diff options
71 files changed, 18919 insertions, 6664 deletions
diff --git a/include/rpcsvc/bootparam_prot.x b/include/rpcsvc/bootparam_prot.x index a6d4f0e..249ed8a 100644 --- a/include/rpcsvc/bootparam_prot.x +++ b/include/rpcsvc/bootparam_prot.x @@ -44,8 +44,6 @@ %#include <sys/param.h> %#include <sys/syslimits.h> %#include <sys/ucred.h> -%#include <nfs/nfsv2.h> -%#include <nfs/nfs.h> #else %#ifndef lint %/*static char sccsid[] = "from: @(#)bootparam_prot.x 1.2 87/06/24 Copyr 1987 Sun Micro";*/ diff --git a/sbin/mount_nfs/mount_nfs.8 b/sbin/mount_nfs/mount_nfs.8 index 0209b49..8ee10f3 100644 --- a/sbin/mount_nfs/mount_nfs.8 +++ b/sbin/mount_nfs/mount_nfs.8 @@ -39,8 +39,9 @@ .Nd mount nfs file systems .Sh SYNOPSIS .Nm mount_nfs -.Op Fl KMPTbcdiklqs +.Op Fl 3KPTUbcdilqs .Op Fl D Ar deadthresh +.Op Fl I Ar readdirsize .Op Fl L Ar leaseterm .Op Fl R Ar retrycnt .Op Fl a Ar maxreadahead @@ -63,10 +64,14 @@ on to the file system tree at the point .Ar node. This command is normally executed by .Xr mount 8 . -It implements the mount protocol as described in RFC 1094, Appendix A. +It implements the mount protocol as described in RFC 1094, Appendix A and +.%T "NFS: Network File System Version 3 Protocol Specification" , +Appendix I. .Pp The options are: .Bl -tag -width indent +.It Fl 3 +Use the NFS Version 3 protocol (Version 2 is the default). .It Fl D Used with NQNFS to set the .Dq "dead server threshold" @@ -80,23 +85,27 @@ Values may be set in the range of 1 - 9, with 9 referring to an (i.e. never assume cached data still valid). This option is not generally recommended and is really an experimental feature. +.It Fl I +Set the readdir read size to the specified value. The value should normally +be a multiple of DIRBLKSIZ that is <= the read size for the mount. .It Fl K Pass Kerberos authenticators to the server for client-to-server user-credential mapping. -This may only be used over TCP mounts between 4.4BSD clients and servers. +This requires that the kernel be built with the NFSKERB option. +(Refer to the INTERNET-DRAFT titled +.%T "Authentication Mechanisms for ONC RPC" , +for more information.) .It Fl L Used with NQNFS to set the lease term to the specified number of seconds. Only use this argument for mounts with a large round trip delay. Values are normally in the 10-30 second range. -.It Fl M -Assume that other clients are not writing a file concurrently with this client. -This implements a slightly less strict consistency criteria than 4.3BSD-Reno -did, that is more in line with most commercial client implementations. -This is recommended for servers that do not support leasing. .It Fl P Use a reserved socket port number. This is useful for mounting servers that require clients to use a -reserved port number. +reserved port number on the mistaken belief that this makes NFS +more secure. (For the rare case where the client has a trusted root account +but untrusworthy users and the network cables are in secure areas this does +help, but for normal desktop clients this does not apply.) .It Fl R Set the retry count for doing the mount to the specified value. .It Fl T @@ -104,11 +113,15 @@ Use TCP transport instead of UDP. This is recommended for servers that are not on the same LAN cable as the client. (NB: This is NOT supported by most non-BSD servers.) +.It Fl U +Force the mount protocol to use UDP transport, even for TCP NFS mounts. +(Necessary for some old BSD servers.) .It Fl a Set the read-ahead count to the specified value. This may be in the range of 0 - 4, and determines how many blocks will be read ahead when a large file is being read sequentially. -This is recommended for mounts with a large bandwidth * delay product. +Trying a value greater than 1 for this is suggested for +mounts with a large bandwidth * delay product. .It Fl b If an initial attempt to contact the server fails, fork off a child to keep trying the mount in the background. @@ -119,10 +132,12 @@ where the filesystem mount is not critical to multiuser operation. For UDP mount points, do not do a .Xr connect 2 . This must be used for servers that do not reply to requests from the -standard port number. +standard NFS port number 2049. .It Fl d -Do not estimate retransmit timeout dynamically. -This may be useful for UDP mounts that exhibit high retry rates. +Turn off the dynamic retransmit timeout estimator. +This may be useful for UDP mounts that exhibit high retry rates, +since it is possible that the dynamically estimated timeout interval is too +short. .It Fl g Set the maximum size of the group list for the credentials to the specified value. @@ -134,20 +149,15 @@ point. Make the mount interruptible, which implies that file system calls that are delayed due to an unresponsive server will fail with EINTR when a termination signal is posted for the process. -.It Fl k -Used with NQNFS to specify -.Dq get a lease -for the file name being looked up. -This is recommended unless the server is complaining about excessive -lease load. .It Fl l -Used with NQNFS to specify that the \fBReaddir_and_Lookup\fR RPC should +Used with NQNFS and NFSV3 to specify that the \fBReaddirPlus\fR RPC should be used. This option reduces RPC traffic for cases such as .Dq "ls -l" , -but increases the lease load on the server. -This is recommended unless the server is complaining about excessive -lease load. +but tends to flood the attribute and name caches with prefetched entries. +Try this option and see whether performance improves or degrades. Probably +most useful for client to server network interconnects with a large bandwidth +times delay product. .It Fl m Set the Kerberos realm to the string argument. Used with the @@ -167,12 +177,16 @@ Use specified port number for NFS requests. The default is to query the portmapper for the NFS port. .El .It Fl q -Use the leasing extensions to the protocol to maintain cache consistency. -This protocol, referred to as Not Quite Nfs (NQNFS), -is only supported by 4.4BSD servers. +Use the leasing extensions to the NFS Version 3 protocol to maintain cache consistency. +This protocol Version 2, referred to as Not Quite Nfs (NQNFS), +is only supported by this updated release of NFS code. +(It is not backwards compatible with the release of NQNFS that went out on +4.4BSD-Lite. To interoperate with a 4.4BSD-Lite NFS system you will have to +avoid this option until you have had an oppurtunity to upgrade the NFS code +on all your 4.4BSD-Lite based systems.) .It Fl r Set the read data size to the specified value. -It should be a power of 2 greater than or equal to 1024. +It should normally be a power of 2 greater than or equal to 1024. This should be used for UDP mounts when the .Dq "fragments dropped due to timeout" value is getting large while actively using a mount point. @@ -197,6 +211,9 @@ Try increasing the interval if .Xr nfsstat 1 shows high retransmit rates while the file system is active or reducing the value if there is a low retransmit rate but long response delay observed. +(Normally, the -d option should be specified when using this option to manually +tune the timeout +interval.) .It Fl w Set the write data size to the specified value. Ditto the comments w.r.t. the diff --git a/sbin/mount_nfs/mount_nfs.c b/sbin/mount_nfs/mount_nfs.c index 9aa15ce..ddba274 100644 --- a/sbin/mount_nfs/mount_nfs.c +++ b/sbin/mount_nfs/mount_nfs.c @@ -59,16 +59,16 @@ static char sccsid[] = "@(#)mount_nfs.c 8.3 (Berkeley) 3/27/94"; #include <netiso/iso.h> #endif -#ifdef KERBEROS +#ifdef NFSKERB #include <kerberosIV/des.h> #include <kerberosIV/krb.h> #endif #include <nfs/rpcv2.h> -#include <nfs/nfsv2.h> -#define KERNEL +#include <nfs/nfsproto.h> +#define _KERNEL #include <nfs/nfs.h> -#undef KERNEL +#undef _KERNEL #include <nfs/nqnfs.h> #include <arpa/inet.h> @@ -86,14 +86,15 @@ static char sccsid[] = "@(#)mount_nfs.c 8.3 (Berkeley) 3/27/94"; #include "mntopts.h" +#ifdef __FreeBSD__ #define ALTF_BG 0x1 #define ALTF_NOCONN 0x2 #define ALTF_DUMBTIMR 0x4 #define ALTF_INTR 0x8 #define ALTF_KERB 0x10 -#define ALTF_NQLOOKLSE 0x20 -#define ALTF_RDIRALOOK 0x40 -#define ALTF_MYWRITE 0x80 +#define ALTF_NFSV3 0x20 +#define ALTF_RDIRPLUS 0x40 +#define ALTF_MNTUDP 0x80 #define ALTF_RESVPORT 0x100 #define ALTF_SEQPACKET 0x200 #define ALTF_NQNFS 0x400 @@ -109,12 +110,12 @@ struct mntopt mopts[] = { { "conn", 1, ALTF_NOCONN, 1 }, { "dumbtimer", 0, ALTF_DUMBTIMR, 1 }, { "intr", 0, ALTF_INTR, 1 }, -#ifdef KERBEROS +#ifdef NFSKERB { "kerb", 0, ALTF_KERB, 1 }, #endif - { "nqlooklease", 0, ALTF_NQLOOKLSE, 1 }, - { "rdiralook", 0, ALTF_RDIRALOOK, 1 }, - { "mywrite", 0, ALTF_MYWRITE, 1 }, + { "nfsv3", 0, ALTF_NFSV3, 1 }, + { "rdirplus", 0, ALTF_RDIRPLUS, 1 }, + { "mntudp", 0, ALTF_MNTUDP, 1 }, { "resvport", 0, ALTF_RESVPORT, 1 }, #ifdef ISO { "seqpacket", 0, ALTF_SEQPACKET, 1 }, @@ -125,17 +126,27 @@ struct mntopt mopts[] = { { "port=", 0, ALTF_PORT, 1 }, { NULL } }; +#else +struct mntopt mopts[] = { + MOPT_STDOPTS, + MOPT_FORCE, + MOPT_UPDATE, + { NULL } +}; +#endif struct nfs_args nfsdefargs = { (struct sockaddr *)0, sizeof (struct sockaddr_in), SOCK_DGRAM, 0, - (nfsv2fh_t *)0, + (u_char *)0, + 0, 0, NFS_WSIZE, NFS_RSIZE, - NFS_TIMEO, + NFS_READDIRSIZE, + 10, NFS_RETRANS, NFS_MAXGRPS, NFS_DEFRAHEAD, @@ -145,20 +156,34 @@ struct nfs_args nfsdefargs = { }; struct nfhret { - u_long stat; - nfsv2fh_t nfh; + u_long stat; + long vers; + long auth; + long fhsize; + u_char nfh[NFSX_V3FHMAX]; }; #define DEF_RETRY 10000 #define BGRND 1 #define ISBGRND 2 int retrycnt = DEF_RETRY; int opflags = 0; +int nfsproto = IPPROTO_UDP; +int mnttcp_ok = 1; u_short port_no = 0; -#ifdef KERBEROS +#ifdef NFSKERB char inst[INST_SZ]; char realm[REALM_SZ]; -KTEXT_ST kt; +struct { + u_long kind; + KTEXT_ST kt; +} ktick; +struct nfsrpc_nickverf kverf; +struct nfsrpc_fullblock kin, kout; +NFSKERBKEY_T kivec; +CREDENTIALS kcr; +struct timeval ktv; +NFSKERBKEYSCHED_T kerb_keysched; #endif int getnfsargs __P((char *, struct nfs_args *)); @@ -182,14 +207,17 @@ main(argc, argv) int mntflags, altflags, i, nfssvc_flag, num; char *name, *p, *spec; struct vfsconf *vfc; -#ifdef KERBEROS +#ifdef NFSKERB uid_t last_ruid; -#endif -#ifdef KERBEROS last_ruid = -1; (void)strcpy(realm, KRB_REALM); -#endif + if (sizeof (struct nfsrpc_nickverf) != RPCX_NICKVERF || + sizeof (struct nfsrpc_fullblock) != RPCX_FULLBLOCK || + ((char *)&ktick.kt) - ((char *)&ktick) != NFSX_UNSIGNED || + ((char *)ktick.kt.dat) - ((char *)&ktick) != 2 * NFSX_UNSIGNED) + fprintf(stderr, "Yikes! NFSKERB structs not packed!!\n"); +#endif /* NFSKERB */ retrycnt = DEF_RETRY; mntflags = 0; @@ -197,8 +225,11 @@ main(argc, argv) nfsargs = nfsdefargs; nfsargsp = &nfsargs; while ((c = getopt(argc, argv, - "a:bcdD:g:iKklL:Mm:o:PpqR:r:sTt:w:x:")) != EOF) + "3a:bcdD:g:I:iKL:lm:o:PpqR:r:sTt:w:x:U")) != EOF) switch (c) { + case '3': + nfsargsp->flags |= NFSMNT_NFSV3; + break; case 'a': num = strtol(optarg, &p, 10); if (*p || num < 0) @@ -226,21 +257,27 @@ main(argc, argv) num = strtol(optarg, &p, 10); if (*p || num <= 0) errx(1, "illegal -g value -- %s", optarg); +#ifdef __FreeBSD__ set_rpc_maxgrouplist(num); +#endif nfsargsp->maxgrouplist = num; nfsargsp->flags |= NFSMNT_MAXGRPS; break; + case 'I': + num = strtol(optarg, &p, 10); + if (*p || num <= 0) + errx(1, "illegal -I value -- %s", optarg); + nfsargsp->readdirsize = num; + nfsargsp->flags |= NFSMNT_READDIRSIZE; + break; case 'i': nfsargsp->flags |= NFSMNT_INT; break; -#ifdef KERBEROS +#ifdef NFSKERB case 'K': nfsargsp->flags |= NFSMNT_KERB; break; #endif - case 'k': - nfsargsp->flags |= NFSMNT_NQLOOKLEASE; - break; case 'L': num = strtol(optarg, &p, 10); if (*p || num < 2) @@ -249,18 +286,16 @@ main(argc, argv) nfsargsp->flags |= NFSMNT_LEASETERM; break; case 'l': - nfsargsp->flags |= NFSMNT_RDIRALOOK; + nfsargsp->flags |= NFSMNT_RDIRPLUS; break; - case 'M': - nfsargsp->flags |= NFSMNT_MYWRITE; - break; -#ifdef KERBEROS +#ifdef NFSKERB case 'm': (void)strncpy(realm, optarg, REALM_SZ - 1); realm[REALM_SZ - 1] = '\0'; break; #endif case 'o': +#ifdef __FreeBSD__ getmntopts(optarg, mopts, &mntflags, &altflags); if(altflags & ALTF_BG) opflags |= BGRND; @@ -270,16 +305,16 @@ main(argc, argv) nfsargsp->flags |= NFSMNT_DUMBTIMR; if(altflags & ALTF_INTR) nfsargsp->flags |= NFSMNT_INT; -#ifdef KERBEROS +#ifdef NFSKERB if(altflags & ALTF_KERB) nfsargsp->flags |= NFSMNT_KERB; #endif - if(altflags & ALTF_NQLOOKLSE) - nfsargsp->flags |= NFSMNT_NQLOOKLEASE; - if(altflags & ALTF_RDIRALOOK) - nfsargsp->flags |= NFSMNT_RDIRALOOK; - if(altflags & ALTF_MYWRITE) - nfsargsp->flags |= NFSMNT_MYWRITE; + if(altflags & ALTF_NFSV3) + nfsargsp->flags |= NFSMNT_NFSV3; + if(altflags & ALTF_RDIRPLUS) + nfsargsp->flags |= NFSMNT_RDIRPLUS; + if(altflags & ALTF_MNTUDP) + mnttcp_ok = 0; if(altflags & ALTF_RESVPORT) nfsargsp->flags |= NFSMNT_RESVPORT; #ifdef ISO @@ -287,14 +322,19 @@ main(argc, argv) nfsargsp->sotype = SOCK_SEQPACKET; #endif if(altflags & ALTF_NQNFS) - nfsargsp->flags |= NFSMNT_NQNFS; + nfsargsp->flags |= (NFSMNT_NQNFS|NFSMNT_NFSV3); if(altflags & ALTF_SOFT) nfsargsp->flags |= NFSMNT_SOFT; - if(altflags & ALTF_TCP) + if(altflags & ALTF_TCP) { nfsargsp->sotype = SOCK_STREAM; + nfsproto = IPPROTO_TCP; + } if(altflags & ALTF_PORT) port_no = atoi(strstr(optarg, "port=") + 5); altflags = 0; +#else + getmntopts(optarg, mopts, &mntflags); +#endif break; case 'P': nfsargsp->flags |= NFSMNT_RESVPORT; @@ -305,7 +345,7 @@ main(argc, argv) break; #endif case 'q': - nfsargsp->flags |= NFSMNT_NQNFS; + nfsargsp->flags |= (NFSMNT_NQNFS | NFSMNT_NFSV3); break; case 'R': num = strtol(optarg, &p, 10); @@ -325,6 +365,7 @@ main(argc, argv) break; case 'T': nfsargsp->sotype = SOCK_STREAM; + nfsproto = IPPROTO_TCP; break; case 't': num = strtol(optarg, &p, 10); @@ -347,6 +388,9 @@ main(argc, argv) nfsargsp->retrans = num; nfsargsp->flags |= NFSMNT_RETRANS; break; + case 'U': + mnttcp_ok = 0; + break; default: usage(); break; @@ -363,6 +407,7 @@ main(argc, argv) if (!getnfsargs(spec, nfsargsp)) exit(1); +#ifdef __FreeBSD__ vfc = getvfsbyname("nfs"); if(!vfc && vfsisloadable("nfs")) { if(vfsload("nfs")) @@ -372,6 +417,9 @@ main(argc, argv) } if (mount(vfc ? vfc->vfc_index : MOUNT_NFS, name, mntflags, nfsargsp)) +#else + if (mount(MOUNT_NFS, name, mntflags, nfsargsp)) +#endif err(1, "%s", name); if (nfsargsp->flags & (NFSMNT_NQNFS | NFSMNT_KERB)) { if ((opflags & ISBGRND) == 0) { @@ -396,13 +444,18 @@ main(argc, argv) } nfssvc_flag = NFSSVC_MNTD | NFSSVC_GOTAUTH | NFSSVC_AUTHINFAIL; -#ifdef KERBEROS +#ifdef NFSKERB /* * Set up as ncd_authuid for the kerberos call. * Must set ruid to ncd_authuid and reset the * ticket name iff ncd_authuid is not the same * as last time, so that the right ticket file * is found. + * Get the Kerberos credential structure so that + * we have the seesion key and get a ticket for + * this uid. + * For more info see the IETF Draft "Authentication + * in ONC RPC". */ if (ncd.ncd_authuid != last_ruid) { char buf[512]; @@ -411,15 +464,62 @@ main(argc, argv) krb_set_tkt_string(buf); last_ruid = ncd.ncd_authuid; } - if (krb_mk_req(&kt, "rcmd", inst, realm, 0) == - KSUCCESS && - kt.length <= (RPCAUTH_MAXSIZ - 2 * NFSX_UNSIGNED)) { - ncd.ncd_authtype = RPCAUTH_NQNFS; - ncd.ncd_authlen = kt.length; - ncd.ncd_authstr = (char *)kt.dat; - nfssvc_flag = NFSSVC_MNTD | NFSSVC_GOTAUTH; + setreuid(ncd.ncd_authuid, 0); + kret = krb_get_cred(NFS_KERBSRV, inst, realm, &kcr); + if (kret == RET_NOTKT) { + kret = get_ad_tkt(NFS_KERBSRV, inst, realm, + DEFAULT_TKT_LIFE); + if (kret == KSUCCESS) + kret = krb_get_cred(NFS_KERBSRV, inst, realm, + &kcr); } -#endif /* KERBEROS */ + if (kret == KSUCCESS) + kret = krb_mk_req(&ktick.kt, NFS_KERBSRV, inst, + realm, 0); + + /* + * Fill in the AKN_FULLNAME authenticator and verfier. + * Along with the Kerberos ticket, we need to build + * the timestamp verifier and encrypt it in CBC mode. + */ + if (kret == KSUCCESS && + ktick.kt.length <= (RPCAUTH_MAXSIZ-3*NFSX_UNSIGNED) + && gettimeofday(&ktv, (struct timezone *)0) == 0) { + ncd.ncd_authtype = RPCAUTH_KERB4; + ncd.ncd_authstr = (u_char *)&ktick; + ncd.ncd_authlen = nfsm_rndup(ktick.kt.length) + + 3 * NFSX_UNSIGNED; + ncd.ncd_verfstr = (u_char *)&kverf; + ncd.ncd_verflen = sizeof (kverf); + bcopy((caddr_t)kcr.session, (caddr_t)ncd.ncd_key, + sizeof (kcr.session)); + kin.t1 = htonl(ktv.tv_sec); + kin.t2 = htonl(ktv.tv_usec); + kin.w1 = htonl(NFS_KERBTTL); + kin.w2 = htonl(NFS_KERBTTL - 1); + bzero((caddr_t)kivec, sizeof (kivec)); + + /* + * Encrypt kin in CBC mode using the session + * key in kcr. + */ + XXX + + /* + * Finally, fill the timestamp verifier into the + * authenticator and verifier. + */ + ktick.kind = htonl(RPCAKN_FULLNAME); + kverf.kind = htonl(RPCAKN_FULLNAME); + NFS_KERBW1(ktick.kt) = kout.w1; + ktick.kt.length = htonl(ktick.kt.length); + kverf.verf.t1 = kout.t1; + kverf.verf.t2 = kout.t2; + kverf.verf.w2 = kout.w2; + nfssvc_flag = NFSSVC_MNTD | NFSSVC_GOTAUTH; + } + setreuid(0, 0); +#endif /* NFSKERB */ } } exit(0); @@ -440,9 +540,9 @@ getnfsargs(spec, nfsargsp) #endif struct timeval pertry, try; enum clnt_stat clnt_stat; - int so = RPC_ANYSOCK, i; + int so = RPC_ANYSOCK, i, nfsvers, mntvers; char *hostp, *delimp; -#ifdef KERBEROS +#ifdef NFSKERB char *cp; #endif u_short tport; @@ -507,7 +607,7 @@ getnfsargs(spec, nfsargsp) warnx("can't get net id for host"); return (0); } -#ifdef KERBEROS +#ifdef NFSKERB if ((nfsargsp->flags & NFSMNT_KERB)) { if ((hp = gethostbyaddr((char *)&saddr.sin_addr.s_addr, sizeof (u_long), AF_INET)) == (struct hostent *)0) { @@ -520,33 +620,46 @@ getnfsargs(spec, nfsargsp) if (cp = strchr(inst, '.')) *cp = '\0'; } -#endif /* KERBEROS */ +#endif /* NFSKERB */ + if (nfsargsp->flags & NFSMNT_NFSV3) { + nfsvers = 3; + mntvers = 3; + } else { + nfsvers = 2; + mntvers = 1; + } nfhret.stat = EACCES; /* Mark not yet successful */ while (retrycnt > 0) { saddr.sin_family = AF_INET; saddr.sin_port = htons(PMAPPORT); if ((tport = port_no ? port_no : pmap_getport(&saddr, RPCPROG_NFS, - NFS_VER2, nfsargsp->sotype == SOCK_STREAM ? IPPROTO_TCP : - IPPROTO_UDP)) == 0) { + nfsvers, nfsproto)) == 0) { if ((opflags & ISBGRND) == 0) clnt_pcreateerror("NFS Portmap"); } else { saddr.sin_port = 0; pertry.tv_sec = 10; pertry.tv_usec = 0; - if ((clp = (nfsargsp->sotype == SOCK_STREAM ? - clnttcp_create(&saddr, RPCPROG_MNT, RPCMNT_VER1, - &so, 0, 0) : - clntudp_create(&saddr, RPCPROG_MNT, RPCMNT_VER1, - pertry, &so))) == NULL) { + if (mnttcp_ok && nfsargsp->sotype == SOCK_STREAM) + clp = clnttcp_create(&saddr, RPCPROG_MNT, mntvers, + &so, 0, 0); + else + clp = clntudp_create(&saddr, RPCPROG_MNT, mntvers, + pertry, &so); + if (clp == NULL) { if ((opflags & ISBGRND) == 0) clnt_pcreateerror("Cannot MNT RPC"); } else { clp->cl_auth = authunix_create_default(); try.tv_sec = 10; try.tv_usec = 0; + if (nfsargsp->flags & NFSMNT_KERB) + nfhret.auth = RPCAUTH_KERB4; + else + nfhret.auth = RPCAUTH_UNIX; + nfhret.vers = mntvers; clnt_stat = clnt_call(clp, RPCMNT_MOUNT, xdr_dir, spec, xdr_fh, &nfhret, try); if (clnt_stat != RPC_SUCCESS) { @@ -596,7 +709,8 @@ getnfsargs(spec, nfsargsp) nfsargsp->addr = (struct sockaddr *) &saddr; nfsargsp->addrlen = sizeof (saddr); } - nfsargsp->fh = &nfhret.nfh; + nfsargsp->fh = nfhret.nfh; + nfsargsp->fhsize = nfhret.fhsize; nfsargsp->hostname = nam; return (1); } @@ -615,22 +729,53 @@ xdr_dir(xdrsp, dirp) int xdr_fh(xdrsp, np) XDR *xdrsp; - struct nfhret *np; + register struct nfhret *np; { - if (!xdr_u_long(xdrsp, &(np->stat))) + register int i; + long auth, authcnt, authfnd = 0; + + if (!xdr_u_long(xdrsp, &np->stat)) return (0); if (np->stat) return (1); - return (xdr_opaque(xdrsp, (caddr_t)&(np->nfh), NFSX_FH)); + switch (np->vers) { + case 1: + np->fhsize = NFSX_V2FH; + return (xdr_opaque(xdrsp, (caddr_t)np->nfh, NFSX_V2FH)); + case 3: + if (!xdr_long(xdrsp, &np->fhsize)) + return (0); + if (np->fhsize <= 0 || np->fhsize > NFSX_V3FHMAX) + return (0); + if (!xdr_opaque(xdrsp, (caddr_t)np->nfh, np->fhsize)) + return (0); + if (!xdr_long(xdrsp, &authcnt)) + return (0); + for (i = 0; i < authcnt; i++) { + if (!xdr_long(xdrsp, &auth)) + return (0); + if (auth == np->auth) + authfnd++; + } + + /* + * Some servers, such as DEC's OSF/1 return a nil authenticator + * list to indicate RPCAUTH_UNIX. + */ + if (!authfnd && (authcnt > 0 || np->auth != RPCAUTH_UNIX)) + np->stat = EAUTH; + return (1); + }; + return (0); } __dead void usage() { - (void)fprintf(stderr, "usage: mount_nfs %s\n%s\n%s\n%s\n", -"[-bcdiKklMPqsT] [-a maxreadahead] [-D deadthresh]", -"\t[-g maxgroups] [-L leaseterm] [-m realm] [-o options] [-R retrycnt]", -"\t[-r readsize] [-t timeout] [-w writesize] [-x retrans]", -"\trhost:path node"); + (void)fprintf(stderr, "\ +usage: mount_nfs [-3KPTUbcdilqs] [-D deadthresh] [-I readdirsize]\n\ + [-L leaseterm] [-R retrycnt] [-a maxreadahead] [-g maxgroups]\n\ + [-m realm] [-o options] [-r readsize] [-t timeout] [-w writesize]\n\ + [-x retrans] rhost:path node\n"); exit(1); } diff --git a/sbin/mountd/exports.5 b/sbin/mountd/exports.5 index d32527f..fc521f2 100644 --- a/sbin/mountd/exports.5 +++ b/sbin/mountd/exports.5 @@ -49,7 +49,9 @@ file specifies remote mount points for the mount protocol per the .Tn NFS server specification; see -.%T "Network File System Protocol Specification \\*(tNRFC\\*(sP 1094, Appendix A" . +.%T "Network File System Protocol Specification \\*(tNRFC\\*(sP 1094, Appendix A" +and +.%T "NFS: Network File System Version 3 Specification, Appendix I" . .Pp Each line in the file (other than comment lines that begin with a #) @@ -71,7 +73,8 @@ The second is to specify the pathname of the root of the filesystem followed by the .Fl alldirs flag; -this form allows the host(s) to mount any directory within the filesystem. +this form allows the host(s) to mount at any point within the filesystem, +including regular files if the ``-r'' option is used on mountd. The pathnames must not have any symbolic links in them and should not have any "." or ".." components. Mount points for a filesystem may appear on multiple lines each with @@ -140,8 +143,7 @@ The .Fl kerb option specifies that the Kerberos authentication server should be used to authenticate and map client credentials. -(Note that this is NOT Sun NFS compatible and -is supported for TCP transport only.) +This requires that the kernel be built with the NFSKERB option. .Pp The .Fl ro diff --git a/sbin/mountd/mountd.8 b/sbin/mountd/mountd.8 index 78f95b5..7168c17 100644 --- a/sbin/mountd/mountd.8 +++ b/sbin/mountd/mountd.8 @@ -30,7 +30,7 @@ .\" SUCH DAMAGE. .\" .\" From: @(#)mountd.8 8.1 (Berkeley) 6/9/93 -.\" $Id: mountd.8,v 1.2 1994/09/22 22:16:49 wollman Exp $ +.\" $Id: mountd.8,v 1.3 1995/02/22 21:42:48 ats Exp $ .\" .Dd September 22, 1994 .Dt MOUNTD 8 @@ -42,7 +42,7 @@ mount requests .Sh SYNOPSIS .Nm /sbin/mountd -.Op Fl n +.Op Fl nr .Op Ar exportsfile .Sh DESCRIPTION .Xr Mountd @@ -54,7 +54,9 @@ listens for service requests at the port indicated in the .Tn NFS server specification; see .%T "Network File System Protocol Specification" , -RFC1094. +RFC1094, Appendix A and +.%T "NFS: Network File System Version 3 Protocol Specification" , +Appendix I. .Pp Options and operands available for .Nm mountd : @@ -65,6 +67,15 @@ The option allows non-root mount requests to be served. This should only be specified if there are clients such as PC's, that require it. +.It Fl r +The +.Fl r +option allows mount RPCs requests for regular files to be served. +Although this seems to violate the mount protocol specification, some diskless +workstations do mount requests for their swapfiles and expect them to be +regular files. Since a regular file cannot be specified in /etc/exports, the +entire file system that the swapfiles reside in will have to be exported with +the ``-alldirs'' flag. .It Ar exportsfile The .Ar exportsfile diff --git a/sbin/mountd/mountd.c b/sbin/mountd/mountd.c index 4434244..0887c59 100644 --- a/sbin/mountd/mountd.c +++ b/sbin/mountd/mountd.c @@ -43,7 +43,7 @@ static char copyright[] = #ifndef lint /*static char sccsid[] = "From: @(#)mountd.c 8.8 (Berkeley) 2/20/94";*/ static const char rcsid[] = - "$Id: mountd.c,v 1.7.2.1 1995/06/08 04:34:11 davidg Exp $"; + "$Id: mountd.c,v 1.8 1995/06/11 19:30:46 rgrimes Exp $"; #endif /*not lint*/ #include <sys/param.h> @@ -62,7 +62,7 @@ static const char rcsid[] = #include <netiso/iso.h> #endif #include <nfs/rpcv2.h> -#include <nfs/nfsv2.h> +#include <nfs/nfsproto.h> #include <arpa/inet.h> @@ -100,6 +100,8 @@ struct dirlist { }; /* dp_flag bits */ #define DP_DEFSET 0x1 +#define DP_HOSTSET 0x2 +#define DP_KERB 0x4 struct exportlist { struct exportlist *ex_next; @@ -139,22 +141,29 @@ struct grouplist { #define GT_IGNORE 0x5 struct hostlist { + int ht_flag; /* Uses DP_xx bits */ struct grouplist *ht_grp; struct hostlist *ht_next; }; +struct fhreturn { + int fhr_flag; + int fhr_vers; + nfsfh_t fhr_fh; +}; + /* Global defs */ char *add_expdir __P((struct dirlist **, char *, int)); void add_dlist __P((struct dirlist **, struct dirlist *, - struct grouplist *)); + struct grouplist *, int)); void add_mlist __P((char *, char *)); int check_dirpath __P((char *)); int check_options __P((struct dirlist *)); -int chk_host __P((struct dirlist *, u_long, int *)); +int chk_host __P((struct dirlist *, u_long, int *, int *)); void del_mlist __P((char *, char *)); struct dirlist *dirp_search __P((struct dirlist *, char *)); int do_mount __P((struct exportlist *, struct grouplist *, int, - struct ucred *, char *, int, struct statfs *)); + struct ucred *, char *, int, struct statfs *)); int do_opt __P((char **, char **, struct exportlist *, struct grouplist *, int *, int *, struct ucred *)); struct exportlist *ex_search __P((fsid_t *)); @@ -165,6 +174,7 @@ void free_grp __P((struct grouplist *)); void free_host __P((struct hostlist *)); void get_exportlist __P((void)); int get_host __P((char *, struct grouplist *, struct grouplist *)); +int get_num __P((char *)); struct hostlist *get_ht __P((void)); int get_line __P((void)); void get_mountlist __P((void)); @@ -183,7 +193,7 @@ void send_umntall __P((void)); int umntall_each __P((caddr_t, struct sockaddr_in *)); int xdr_dir __P((XDR *, char *)); int xdr_explist __P((XDR *, caddr_t)); -int xdr_fhs __P((XDR *, nfsv2fh_t *)); +int xdr_fhs __P((XDR *, caddr_t)); int xdr_mlist __P((XDR *, caddr_t)); /* C library */ @@ -205,7 +215,8 @@ struct ucred def_anon = { 1, { (gid_t) -2 } }; -int root_only = 1; +int resvport_only = 1; +int dir_only = 1; int opt_flags; /* Bits for above */ #define OP_MAPROOT 0x01 @@ -238,6 +249,7 @@ main(argc, argv) { SVCXPRT *udptransp, *tcptransp; int c; +#ifdef __FreeBSD__ struct vfsconf *vfc; vfc = getvfsbyname("nfs"); @@ -250,17 +262,21 @@ main(argc, argv) if(!vfc) { errx(1, "NFS support is not available in the running kernel"); } +#endif /* __FreeBSD__ */ - while ((c = getopt(argc, argv, "dn")) != EOF) + while ((c = getopt(argc, argv, "dnr")) != EOF) switch (c) { + case 'n': + resvport_only = 0; + break; + case 'r': + dir_only = 0; + break; case 'd': debug = debug ? 0 : 1; break; - case 'n': - root_only = 0; - break; default: - fprintf(stderr, "Usage: mountd [-n] [export_file]\n"); + fprintf(stderr, "Usage: mountd [-r] [-n] [export_file]\n"); exit(1); }; argc -= optind; @@ -300,11 +316,12 @@ main(argc, argv) syslog(LOG_ERR, "Can't create socket"); exit(1); } - pmap_unset(RPCPROG_MNT, RPCMNT_VER1); - if (!svc_register(udptransp, RPCPROG_MNT, RPCMNT_VER1, mntsrv, - IPPROTO_UDP) || - !svc_register(tcptransp, RPCPROG_MNT, RPCMNT_VER1, mntsrv, - IPPROTO_TCP)) { + pmap_unset(RPCPROG_MNT, 1); + pmap_unset(RPCPROG_MNT, 3); + if (!svc_register(udptransp, RPCPROG_MNT, 1, mntsrv, IPPROTO_UDP) || + !svc_register(udptransp, RPCPROG_MNT, 3, mntsrv, IPPROTO_UDP) || + !svc_register(tcptransp, RPCPROG_MNT, 1, mntsrv, IPPROTO_TCP) || + !svc_register(tcptransp, RPCPROG_MNT, 3, mntsrv, IPPROTO_TCP)) { syslog(LOG_ERR, "Can't register mount"); exit(1); } @@ -323,28 +340,21 @@ mntsrv(rqstp, transp) { struct exportlist *ep; struct dirlist *dp; - nfsv2fh_t nfh; + struct fhreturn fhr; struct authunix_parms *ucr; struct stat stb; struct statfs fsb; struct hostent *hp; u_long saddr; + u_short sport; char rpcpath[RPCMNT_PATHLEN+1], dirpath[MAXPATHLEN]; - int bad = ENOENT, omask, defset; - uid_t uid = -2; - - /* Get authorization */ - switch (rqstp->rq_cred.oa_flavor) { - case AUTH_UNIX: - ucr = (struct authunix_parms *)rqstp->rq_clntcred; - uid = ucr->aup_uid; - break; - case AUTH_NULL: - default: - break; - } + int bad = ENOENT, defset, hostset; + sigset_t sighup_mask; + sigemptyset(&sighup_mask); + sigaddset(&sighup_mask, SIGHUP); saddr = transp->xp_raddr.sin_addr.s_addr; + sport = ntohs(transp->xp_raddr.sin_port); hp = (struct hostent *)NULL; switch (rqstp->rq_proc) { case NULLPROC: @@ -352,7 +362,7 @@ mntsrv(rqstp, transp) syslog(LOG_ERR, "Can't send reply"); return; case RPCMNT_MOUNT: - if ((uid != 0 && root_only) || uid == -2) { + if (sport >= IPPORT_RESERVED && resvport_only) { svcerr_weakauth(transp); return; } @@ -363,11 +373,13 @@ mntsrv(rqstp, transp) /* * Get the real pathname and make sure it is a directory - * that exists. + * or a regular file if the -r option was specified + * and it exists. */ if (realpath(rpcpath, dirpath) == 0 || stat(dirpath, &stb) < 0 || - (stb.st_mode & S_IFMT) != S_IFDIR || + (!S_ISDIR(stb.st_mode) && + (dir_only || !S_ISREG(stb.st_mode))) || statfs(dirpath, &fsb) < 0) { chdir("/"); /* Just in case realpath doesn't */ if (debug) @@ -378,26 +390,31 @@ mntsrv(rqstp, transp) } /* Check in the exports list */ - omask = sigblock(sigmask(SIGHUP)); + sigprocmask(SIG_BLOCK, &sighup_mask, NULL); ep = ex_search(&fsb.f_fsid); - defset = 0; - if (ep && (chk_host(ep->ex_defdir, saddr, &defset) || + hostset = defset = 0; + if (ep && (chk_host(ep->ex_defdir, saddr, &defset, &hostset) || ((dp = dirp_search(ep->ex_dirl, dirpath)) && - chk_host(dp, saddr, &defset)) || + chk_host(dp, saddr, &defset, &hostset)) || (defset && scan_tree(ep->ex_defdir, saddr) == 0 && scan_tree(ep->ex_dirl, saddr) == 0))) { + if (hostset & DP_HOSTSET) + fhr.fhr_flag = hostset; + else + fhr.fhr_flag = defset; + fhr.fhr_vers = rqstp->rq_vers; /* Get the file handle */ - bzero((caddr_t)&nfh, sizeof(nfh)); - if (getfh(dirpath, (fhandle_t *)&nfh) < 0) { + bzero((caddr_t)&fhr.fhr_fh, sizeof(nfsfh_t)); + if (getfh(dirpath, (fhandle_t *)&fhr.fhr_fh) < 0) { bad = errno; syslog(LOG_ERR, "Can't get fh for %s", dirpath); if (!svc_sendreply(transp, xdr_long, (caddr_t)&bad)) syslog(LOG_ERR, "Can't send reply"); - sigsetmask(omask); + sigprocmask(SIG_UNBLOCK, &sighup_mask, NULL); return; } - if (!svc_sendreply(transp, xdr_fhs, (caddr_t)&nfh)) + if (!svc_sendreply(transp, xdr_fhs, (caddr_t)&fhr)) syslog(LOG_ERR, "Can't send reply"); if (hp == NULL) hp = gethostbyaddr((caddr_t)&saddr, @@ -414,14 +431,14 @@ mntsrv(rqstp, transp) if (!svc_sendreply(transp, xdr_long, (caddr_t)&bad)) syslog(LOG_ERR, "Can't send reply"); } - sigsetmask(omask); + sigprocmask(SIG_UNBLOCK, &sighup_mask, NULL); return; case RPCMNT_DUMP: if (!svc_sendreply(transp, xdr_mlist, (caddr_t)NULL)) syslog(LOG_ERR, "Can't send reply"); return; case RPCMNT_UMOUNT: - if ((uid != 0 && root_only) || uid == -2) { + if (sport >= IPPORT_RESERVED && resvport_only) { svcerr_weakauth(transp); return; } @@ -437,7 +454,7 @@ mntsrv(rqstp, transp) del_mlist(inet_ntoa(transp->xp_raddr.sin_addr), dirpath); return; case RPCMNT_UMNTALL: - if ((uid != 0 && root_only) || uid == -2) { + if (sport >= IPPORT_RESERVED && resvport_only) { svcerr_weakauth(transp); return; } @@ -470,18 +487,37 @@ xdr_dir(xdrsp, dirp) } /* - * Xdr routine to generate fhstatus + * Xdr routine to generate file handle reply */ int -xdr_fhs(xdrsp, nfh) +xdr_fhs(xdrsp, cp) XDR *xdrsp; - nfsv2fh_t *nfh; + caddr_t cp; { - u_long ok = 0; + register struct fhreturn *fhrp = (struct fhreturn *)cp; + u_long ok = 0, len, auth; if (!xdr_long(xdrsp, &ok)) return (0); - return (xdr_opaque(xdrsp, (caddr_t)nfh, NFSX_FH)); + switch (fhrp->fhr_vers) { + case 1: + return (xdr_opaque(xdrsp, (caddr_t)&fhrp->fhr_fh, NFSX_V2FH)); + case 3: + len = NFSX_V3FH; + if (!xdr_long(xdrsp, &len)) + return (0); + if (!xdr_opaque(xdrsp, (caddr_t)&fhrp->fhr_fh, len)) + return (0); + if (fhrp->fhr_flag & DP_KERB) + auth = RPCAUTH_KERB4; + else + auth = RPCAUTH_UNIX; + len = 1; + if (!xdr_long(xdrsp, &len)) + return (0); + return (xdr_long(xdrsp, &auth)); + }; + return (0); } int @@ -521,9 +557,12 @@ xdr_explist(xdrsp, cp) { struct exportlist *ep; int false = 0; - int omask, putdef; + int putdef; + sigset_t sighup_mask; - omask = sigblock(sigmask(SIGHUP)); + sigemptyset(&sighup_mask); + sigaddset(&sighup_mask, SIGHUP); + sigprocmask(SIG_BLOCK, &sighup_mask, NULL); ep = exphead; while (ep) { putdef = 0; @@ -535,12 +574,12 @@ xdr_explist(xdrsp, cp) goto errout; ep = ep->ex_next; } - sigsetmask(omask); + sigprocmask(SIG_UNBLOCK, &sighup_mask, NULL); if (!xdr_bool(xdrsp, &false)) return (0); return (1); errout: - sigsetmask(omask); + sigprocmask(SIG_UNBLOCK, &sighup_mask, NULL); return (0); } @@ -660,6 +699,18 @@ get_exportlist() struct ufs_args ua; struct iso_args ia; struct mfs_args ma; +#ifdef __NetBSD__ + struct msdosfs_args da; + } targs; + + if (!strcmp(fsp->f_fstypename, MOUNT_MFS) || + !strcmp(fsp->f_fstypename, MOUNT_UFS) || + !strcmp(fsp->f_fstypename, MOUNT_MSDOS) || + !strcmp(fsp->f_fstypename, MOUNT_CD9660)) { + targs.ua.fspec = NULL; + targs.ua.export.ex_flags = MNT_DELEXPORT; + if (mount(fsp->f_fstypename, fsp->f_mntonname, +#else } targs; switch (fsp->f_type) { @@ -670,6 +721,7 @@ get_exportlist() targs.ua.fspec = NULL; targs.ua.export.ex_flags = MNT_DELEXPORT; if (mount(fsp->f_type, fsp->f_mntonname, +#endif fsp->f_flags | MNT_UPDATE, (caddr_t)&targs) < 0) syslog(LOG_ERR, "Can't delete exports for %s", @@ -805,6 +857,7 @@ get_exportlist() if (get_host(hst, grp, tgrp)) { syslog(LOG_ERR, "Bad netgroup %s", cp); getexp_err(ep, tgrp); + endnetgrent(); goto nextline; } } else if (get_host(cp, grp, tgrp)) { @@ -865,12 +918,12 @@ get_exportlist() * Success. Update the data structures. */ if (has_host) { - hang_dirp(dirhead, tgrp, ep, (opt_flags & OP_ALLDIRS)); + hang_dirp(dirhead, tgrp, ep, opt_flags); grp->gr_next = grphead; grphead = tgrp; } else { hang_dirp(dirhead, (struct grouplist *)NULL, ep, - (opt_flags & OP_ALLDIRS)); + opt_flags); free_grp(grp); } dirhead = (struct dirlist *)NULL; @@ -994,24 +1047,28 @@ add_expdir(dpp, cp, len) * and update the entry for host. */ void -hang_dirp(dp, grp, ep, alldirs) +hang_dirp(dp, grp, ep, flags) struct dirlist *dp; struct grouplist *grp; struct exportlist *ep; - int alldirs; + int flags; { struct hostlist *hp; struct dirlist *dp2; - if (alldirs) { + if (flags & OP_ALLDIRS) { if (ep->ex_defdir) free((caddr_t)dp); else ep->ex_defdir = dp; - if (grp == (struct grouplist *)NULL) + if (grp == (struct grouplist *)NULL) { ep->ex_defdir->dp_flag |= DP_DEFSET; - else while (grp) { + if (flags & OP_KERB) + ep->ex_defdir->dp_flag |= DP_KERB; + } else while (grp) { hp = get_ht(); + if (flags & OP_KERB) + hp->ht_flag |= DP_KERB; hp->ht_grp = grp; hp->ht_next = ep->ex_defdir->dp_hosts; ep->ex_defdir->dp_hosts = hp; @@ -1024,7 +1081,7 @@ hang_dirp(dp, grp, ep, alldirs) */ while (dp) { dp2 = dp->dp_left; - add_dlist(&ep->ex_dirl, dp, grp); + add_dlist(&ep->ex_dirl, dp, grp, flags); dp = dp2; } } @@ -1035,10 +1092,11 @@ hang_dirp(dp, grp, ep, alldirs) * for the new directory or adding the new node. */ void -add_dlist(dpp, newdp, grp) +add_dlist(dpp, newdp, grp, flags) struct dirlist **dpp; struct dirlist *newdp; struct grouplist *grp; + int flags; { struct dirlist *dp; struct hostlist *hp; @@ -1048,10 +1106,10 @@ add_dlist(dpp, newdp, grp) if (dp) { cmp = strcmp(dp->dp_dirp, newdp->dp_dirp); if (cmp > 0) { - add_dlist(&dp->dp_left, newdp, grp); + add_dlist(&dp->dp_left, newdp, grp, flags); return; } else if (cmp < 0) { - add_dlist(&dp->dp_right, newdp, grp); + add_dlist(&dp->dp_right, newdp, grp, flags); return; } else free((caddr_t)newdp); @@ -1067,13 +1125,18 @@ add_dlist(dpp, newdp, grp) */ do { hp = get_ht(); + if (flags & OP_KERB) + hp->ht_flag |= DP_KERB; hp->ht_grp = grp; hp->ht_next = dp->dp_hosts; dp->dp_hosts = hp; grp = grp->gr_next; } while (grp); - } else + } else { dp->dp_flag |= DP_DEFSET; + if (flags & OP_KERB) + dp->dp_flag |= DP_KERB; + } } /* @@ -1102,10 +1165,11 @@ dirp_search(dp, dirpath) * Scan for a host match in a directory tree. */ int -chk_host(dp, saddr, defsetp) +chk_host(dp, saddr, defsetp, hostsetp) struct dirlist *dp; u_long saddr; int *defsetp; + int *hostsetp; { struct hostlist *hp; struct grouplist *grp; @@ -1113,7 +1177,7 @@ chk_host(dp, saddr, defsetp) if (dp) { if (dp->dp_flag & DP_DEFSET) - *defsetp = 1; + *defsetp = dp->dp_flag; hp = dp->dp_hosts; while (hp) { grp = hp->ht_grp; @@ -1122,15 +1186,19 @@ chk_host(dp, saddr, defsetp) addrp = (u_long **) grp->gr_ptr.gt_hostent->h_addr_list; while (*addrp) { - if (**addrp == saddr) + if (**addrp == saddr) { + *hostsetp = (hp->ht_flag | DP_HOSTSET); return (1); + } addrp++; } break; case GT_NET: if ((saddr & grp->gr_ptr.gt_net.nt_mask) == - grp->gr_ptr.gt_net.nt_net) + grp->gr_ptr.gt_net.nt_net) { + *hostsetp = (hp->ht_flag | DP_HOSTSET); return (1); + } break; }; hp = hp->ht_next; @@ -1147,12 +1215,12 @@ scan_tree(dp, saddr) struct dirlist *dp; u_long saddr; { - int defset; + int defset, hostset; if (dp) { if (scan_tree(dp->dp_left, saddr)) return (1); - if (chk_host(dp, saddr, &defset)) + if (chk_host(dp, saddr, &defset, &hostset)) return (1); if (scan_tree(dp->dp_right, saddr)) return (1); @@ -1312,7 +1380,7 @@ get_host(cp, grp, tgrp) if (isdigit(*cp)) { saddr = inet_addr(cp); if (saddr == -1) { - syslog(LOG_ERR, "Inet_addr failed"); + syslog(LOG_ERR, "Inet_addr failed for %s", cp); return (1); } if ((hp = gethostbyaddr((caddr_t)&saddr, sizeof (saddr), @@ -1326,7 +1394,7 @@ get_host(cp, grp, tgrp) aptr[1] = (char *)NULL; } } else { - syslog(LOG_ERR, "Gethostbyname failed"); + syslog(LOG_ERR, "Gethostbyname failed for %s", cp); return (1); } } @@ -1424,6 +1492,7 @@ get_ht() if (hp == (struct hostlist *)NULL) out_of_mem(); hp->ht_next = (struct hostlist *)NULL; + hp->ht_flag = 0; return (hp); } @@ -1494,6 +1563,9 @@ do_mount(ep, grp, exflags, anoncrp, dirp, dirplen, fsb) struct ufs_args ua; struct iso_args ia; struct mfs_args ma; +#ifdef __NetBSD__ + struct msdosfs_args da; +#endif } args; u_long net; @@ -1569,7 +1641,11 @@ do_mount(ep, grp, exflags, anoncrp, dirp, dirplen, fsb) * Also, needs to know how to export all types of local * exportable file systems and not just MOUNT_UFS. */ +#ifdef __NetBSD__ + while (mount(fsb->f_fstypename, dirp, +#else while (mount(fsb->f_type, dirp, +#endif fsb->f_flags | MNT_UPDATE, (caddr_t)&args) < 0) { if (cp) *cp-- = savedc; @@ -2033,15 +2109,30 @@ check_dirpath(dirp) while (*cp && ret) { if (*cp == '/') { *cp = '\0'; - if (lstat(dirp, &sb) < 0 || - (sb.st_mode & S_IFMT) != S_IFDIR) + if (lstat(dirp, &sb) < 0 || !S_ISDIR(sb.st_mode)) ret = 0; *cp = '/'; } cp++; } - if (lstat(dirp, &sb) < 0 || - (sb.st_mode & S_IFMT) != S_IFDIR) + if (lstat(dirp, &sb) < 0 || !S_ISDIR(sb.st_mode)) ret = 0; return (ret); } + +/* + * Just translate an ascii string to an integer. + */ +int +get_num(cp) + register char *cp; +{ + register int res = 0; + + while (*cp) { + if (*cp < '0' || *cp > '9') + return (-1); + res = res * 10 + (*cp++ - '0'); + } + return (res); +} diff --git a/sbin/nfsd/nfsd.8 b/sbin/nfsd/nfsd.8 index a54a564..24573d9 100644 --- a/sbin/nfsd/nfsd.8 +++ b/sbin/nfsd/nfsd.8 @@ -98,7 +98,8 @@ listens for service requests at the port indicated in the .Tn NFS server specification; see .%T "Network File System Protocol Specification" , -RFC1094. +RFC1094 and +.%T "NFS: Network File System Version 3 Protocol Specification" . .Pp If .Nm nfsd diff --git a/sbin/nfsd/nfsd.c b/sbin/nfsd/nfsd.c index 9d4d226..8c8731d 100644 --- a/sbin/nfsd/nfsd.c +++ b/sbin/nfsd/nfsd.c @@ -63,10 +63,10 @@ static char sccsid[] = "@(#)nfsd.c 8.7 (Berkeley) 2/22/94"; #include <netiso/iso.h> #endif #include <nfs/rpcv2.h> -#include <nfs/nfsv2.h> +#include <nfs/nfsproto.h> #include <nfs/nfs.h> -#ifdef KERBEROS +#ifdef NFSKERB #include <kerberosIV/des.h> #include <kerberosIV/krb.h> #endif @@ -94,16 +94,23 @@ struct nfsd_srvargs nsd; char **Argv = NULL; /* pointer to argument vector */ char *LastArg = NULL; /* end of argv */ -#ifdef KERBEROS +#ifdef NFSKERB char lnam[ANAME_SZ]; KTEXT_ST kt; -AUTH_DAT auth; +AUTH_DAT kauth; char inst[INST_SZ]; +struct nfsrpc_fullblock kin, kout; +struct nfsrpc_fullverf kverf; +NFSKERBKEY_T kivec; +struct timeval ktv; +NFSKERBKEYSCHED_T kerb_keysched; #endif void nonfs __P((int)); void reapchild __P((int)); +#ifdef __FreeBSD__ void setproctitle __P((char *)); +#endif void usage __P((void)); /* @@ -139,11 +146,13 @@ main(argc, argv, envp) #ifdef ISO struct sockaddr_iso isoaddr, isopeer; #endif + struct timeval ktv; fd_set ready, sockbits; int ch, cltpflag, connect_type_cnt, i, len, maxsock, msgsock; int nfsdcnt, nfssvc_flag, on, reregister, sock, tcpflag, tcpsock; int tp4cnt, tp4flag, tp4sock, tpipcnt, tpipflag, tpipsock, udpflag; char *cp, **cpp; +#ifdef __FreeBSD__ struct vfsconf *vfc; vfc = getvfsbyname("nfs"); @@ -156,6 +165,7 @@ main(argc, argv, envp) if(!vfc) { errx(1, "NFS is not available in the running kernel"); } +#endif /* Save start and extent of argv for setproctitle. */ Argv = argv; @@ -241,10 +251,12 @@ main(argc, argv, envp) if (reregister) { if (udpflag && - !pmap_set(RPCPROG_NFS, NFS_VER2, IPPROTO_UDP, NFS_PORT)) + (!pmap_set(RPCPROG_NFS, 2, IPPROTO_UDP, NFS_PORT) || + !pmap_set(RPCPROG_NFS, 3, IPPROTO_UDP, NFS_PORT))) err(1, "can't register with portmap for UDP."); if (tcpflag && - !pmap_set(RPCPROG_NFS, NFS_VER2, IPPROTO_TCP, NFS_PORT)) + (!pmap_set(RPCPROG_NFS, 2, IPPROTO_TCP, NFS_PORT) || + !pmap_set(RPCPROG_NFS, 3, IPPROTO_TCP, NFS_PORT))) err(1, "can't register with portmap for TCP."); exit(0); } @@ -261,11 +273,17 @@ main(argc, argv, envp) continue; } - setproctitle("nfsd-srv"); + setproctitle("server"); nfssvc_flag = NFSSVC_NFSD; nsd.nsd_nfsd = NULL; -#ifdef KERBEROS - nsd.nsd_authstr = (char *)kt.dat; +#ifdef NFSKERB + if (sizeof (struct nfsrpc_fullverf) != RPCX_FULLVERF || + sizeof (struct nfsrpc_fullblock) != RPCX_FULLBLOCK) + syslog(LOG_ERR, "Yikes NFSKERB structs not packed!"); + nsd.nsd_authstr = (u_char *)&kt; + nsd.nsd_authlen = sizeof (kt); + nsd.nsd_verfstr = (u_char *)&kverf; + nsd.nsd_verflen = sizeof (kverf); #endif while (nfssvc(nfssvc_flag, &nsd) < 0) { if (errno != ENEEDAUTH) { @@ -273,14 +291,27 @@ main(argc, argv, envp) exit(1); } nfssvc_flag = NFSSVC_NFSD | NFSSVC_AUTHINFAIL; -#ifdef KERBEROS - kt.length = nsd.nsd_authlen; - kt.mbz = 0; - (void)strcpy(inst, "*"); - if (krb_rd_req(&kt, "rcmd", - inst, nsd.nsd_haddr, &auth, "") == RD_AP_OK && - krb_kntoln(&auth, lnam) == KSUCCESS && - (pwd = getpwnam(lnam)) != NULL) { +#ifdef NFSKERB + /* + * Get the Kerberos ticket out of the authenticator + * verify it and convert the principal name to a user + * name. The user name is then converted to a set of + * user credentials via the password and group file. + * Finally, decrypt the timestamp and validate it. + * For more info see the IETF Draft "Authentication + * in ONC RPC". + */ + kt.length = ntohl(kt.length); + if (gettimeofday(&ktv, (struct timezone *)0) == 0 && + kt.length > 0 && kt.length <= + (RPCAUTH_MAXSIZ - 3 * NFSX_UNSIGNED)) { + kin.w1 = NFS_KERBW1(kt); + kt.mbz = 0; + (void)strcpy(inst, "*"); + if (krb_rd_req(&kt, NFS_KERBSRV, + inst, nsd.nsd_haddr, &kauth, "") == RD_AP_OK && + krb_kntoln(&kauth, lnam) == KSUCCESS && + (pwd = getpwnam(lnam)) != NULL) { cr = &nsd.nsd_cr; cr->cr_uid = pwd->pw_uid; cr->cr_groups[0] = pwd->pw_gid; @@ -301,9 +332,34 @@ main(argc, argv, envp) break; } endgrent(); - nfssvc_flag = NFSSVC_NFSD | NFSSVC_AUTHIN; + + /* + * Get the timestamp verifier out of the + * authenticator and verifier strings. + */ + kin.t1 = kverf.t1; + kin.t2 = kverf.t2; + kin.w2 = kverf.w2; + bzero((caddr_t)kivec, sizeof (kivec)); + bcopy((caddr_t)kauth.session, + (caddr_t)nsd.nsd_key,sizeof(kauth.session)); + + /* + * Decrypt the timestamp verifier in CBC mode. + */ + XXX + + /* + * Validate the timestamp verifier, to + * check that the session key is ok. + */ + nsd.nsd_timestamp.tv_sec = ntohl(kout.t1); + nsd.nsd_timestamp.tv_usec = ntohl(kout.t2); + nsd.nsd_ttl = ntohl(kout.w1); + if ((nsd.nsd_ttl - 1) == ntohl(kout.w2)) + nfssvc_flag = NFSSVC_NFSD | NFSSVC_AUTHIN; } -#endif /* KERBEROS */ +#endif /* NFSKERB */ } exit(0); } @@ -323,7 +379,8 @@ main(argc, argv, envp) syslog(LOG_ERR, "can't bind udp addr"); exit(1); } - if (!pmap_set(RPCPROG_NFS, NFS_VER2, IPPROTO_UDP, NFS_PORT)) { + if (!pmap_set(RPCPROG_NFS, 2, IPPROTO_UDP, NFS_PORT) || + !pmap_set(RPCPROG_NFS, 3, IPPROTO_UDP, NFS_PORT)) { syslog(LOG_ERR, "can't register with udp portmap"); exit(1); } @@ -403,7 +460,8 @@ main(argc, argv, envp) syslog(LOG_ERR, "listen failed"); exit(1); } - if (!pmap_set(RPCPROG_NFS, NFS_VER2, IPPROTO_TCP, NFS_PORT)) { + if (!pmap_set(RPCPROG_NFS, 2, IPPROTO_TCP, NFS_PORT) || + !pmap_set(RPCPROG_NFS, 3, IPPROTO_TCP, NFS_PORT)) { syslog(LOG_ERR, "can't register tcp with portmap"); exit(1); } @@ -492,7 +550,7 @@ main(argc, argv, envp) if (connect_type_cnt == 0) exit(0); - setproctitle("nfsd-master"); + setproctitle("master"); /* * Loop forever accepting connections and passing the sockets @@ -566,7 +624,7 @@ main(argc, argv, envp) void usage() { - (void)fprintf(stderr, "nfsd %s\n", USAGE); + (void)fprintf(stderr, "usage: nfsd %s\n", USAGE); exit(1); } @@ -582,9 +640,10 @@ reapchild(signo) int signo; { - while (wait3(NULL, WNOHANG, NULL)); + while (wait3(NULL, WNOHANG, NULL) > 0); } +#ifdef __FreeBSD__ void setproctitle(a) char *a; @@ -593,9 +652,10 @@ setproctitle(a) char buf[80]; cp = Argv[0]; - (void)snprintf(buf, sizeof(buf), "%s", a); + (void)snprintf(buf, sizeof(buf), "nfsd-%s", a); (void)strncpy(cp, buf, LastArg - cp); cp += strlen(cp); while (cp < LastArg) *cp++ = '\0'; } +#endif /* __FreeBSD__ */ diff --git a/sbin/nfsiod/nfsiod.c b/sbin/nfsiod/nfsiod.c index d066329..0a6985f 100644 --- a/sbin/nfsiod/nfsiod.c +++ b/sbin/nfsiod/nfsiod.c @@ -50,8 +50,10 @@ static char sccsid[] = "@(#)nfsiod.c 8.3 (Berkeley) 2/22/94"; #include <sys/ucred.h> #include <sys/wait.h> #include <sys/mount.h> +#include <sys/time.h> -#include <nfs/nfsv2.h> +#include <nfs/rpcv2.h> +#include <nfs/nfsproto.h> #include <nfs/nfs.h> #include <err.h> diff --git a/sys/nfs/nfs.h b/sys/nfs/nfs.h index cbf80c6..e1a0f07 100644 --- a/sys/nfs/nfs.h +++ b/sys/nfs/nfs.h @@ -34,7 +34,7 @@ * SUCH DAMAGE. * * @(#)nfs.h 8.1 (Berkeley) 6/10/93 - * $Id: nfs.h,v 1.8 1994/11/02 00:11:00 wollman Exp $ + * $Id: nfs.h,v 1.9 1995/02/14 06:22:18 phk Exp $ */ #ifndef _NFS_NFS_H_ @@ -45,11 +45,12 @@ */ #define NFS_MAXIOVEC 34 -#define NFS_HZ 25 /* Ticks per second for NFS timeouts */ -#define NFS_TIMEO (1*NFS_HZ) /* Default timeout = 1 second */ -#define NFS_MINTIMEO (1*NFS_HZ) /* Min timeout to use */ -#define NFS_MAXTIMEO (60*NFS_HZ) /* Max timeout to backoff to */ -#define NFS_MINIDEMTIMEO (5*NFS_HZ) /* Min timeout for non-idempotent ops*/ +#define NFS_TICKINTVL 5 /* Desired time for a tick (msec) */ +#define NFS_HZ (hz / nfs_ticks) /* Ticks/sec */ +#define NFS_TIMEO (1 * NFS_HZ) /* Default timeout = 1 second */ +#define NFS_MINTIMEO (1 * NFS_HZ) /* Min timeout to use */ +#define NFS_MAXTIMEO (60 * NFS_HZ) /* Max timeout to backoff to */ +#define NFS_MINIDEMTIMEO (5 * NFS_HZ) /* Min timeout for non-idempotent ops*/ #define NFS_MAXREXMIT 100 /* Stop counting after this many */ #define NFS_MAXWINDOW 1024 /* Max number of outstanding requests */ #define NFS_RETRANS 10 /* Num of retrans for soft mounts */ @@ -62,13 +63,81 @@ #endif #define NFS_WSIZE 8192 /* Def. write data size <= 8192 */ #define NFS_RSIZE 8192 /* Def. read data size <= 8192 */ +#define NFS_READDIRSIZE 8192 /* Def. readdir size */ #define NFS_DEFRAHEAD 1 /* Def. read ahead # blocks */ #define NFS_MAXRAHEAD 4 /* Max. read ahead # blocks */ -#define NFS_MAXREADDIR NFS_MAXDATA /* Max. size of directory read */ #define NFS_MAXUIDHASH 64 /* Max. # of hashed uid entries/mp */ -#define NFS_MAXASYNCDAEMON 20 /* Max. number async_daemons runable */ -#define NFS_DIRBLKSIZ 1024 /* Size of an NFS directory block */ +#define NFS_MAXASYNCDAEMON 20 /* Max. number async_daemons runable */ +#define NFS_MAXGATHERDELAY 100 /* Max. write gather delay (msec) */ +#ifndef NFS_GATHERDELAY +#define NFS_GATHERDELAY 10 /* Default write gather delay (msec) */ +#endif +#define NFS_DIRBLKSIZ 4096 /* Must be a multiple of DIRBLKSIZ */ + +/* + * Oddballs + */ #define NMOD(a) ((a) % nfs_asyncdaemons) +#define NFS_CMPFH(n, f, s) \ + ((n)->n_fhsize == (s) && !bcmp((caddr_t)(n)->n_fhp, (caddr_t)(f), (s))) +#define NFS_ISV3(v) (VFSTONFS((v)->v_mount)->nm_flag & NFSMNT_NFSV3) +#define NFS_SRVMAXDATA(n) \ + (((n)->nd_flag & ND_NFSV3) ? (((n)->nd_nam2) ? \ + NFS_MAXDGRAMDATA : NFS_MAXDATA) : NFS_V2MAXDATA) + +/* + * XXX + * sys/buf.h should be editted to change B_APPENDWRITE --> B_NEEDCOMMIT, but + * until then... + * Same goes for sys/malloc.h, which needs M_NFSDIROFF, + * M_NFSRVDESC and M_NFSBIGFH added. + * The VA_EXCLUSIVE flag should be added for va_vaflags and set for an + * exclusive create. + * The B_INVAFTERWRITE flag should be set to whatever is required by the + * buffer cache code to say "Invalidate the block after it is written back". + */ +#ifndef B_NEEDCOMMIT +#define B_NEEDCOMMIT B_APPENDWRITE +#endif +#ifndef M_NFSRVDESC +#define M_NFSRVDESC M_TEMP +#endif +#ifndef M_NFSDIROFF +#define M_NFSDIROFF M_TEMP +#endif +#ifndef M_NFSBIGFH +#define M_NFSBIGFH M_TEMP +#endif +#ifndef VA_EXCLUSIVE +#define VA_EXCLUSIVE 0 +#endif +#ifdef __FreeBSD__ +#define B_INVAFTERWRITE B_NOCACHE +#else +#define B_INVAFTERWRITE B_INVAL +#endif + +/* + * These ifdefs try to handle the differences between the various 4.4BSD-Lite + * based vfs interfaces. + * btw: NetBSD-current does have a VOP_LEASDE(), but I don't know how to + * differentiate between NetBSD-1.0 and NetBSD-current, so.. + * I also don't know about BSDi's 2.0 release. + */ +#if !defined(HAS_VOPLEASE) && !defined(__FreeBSD__) && !defined(__NetBSD__) +#define HAS_VOPLEASE 1 +#endif +#if !defined(HAS_VOPREVOKE) && !defined(__FreeBSD__) && !defined(__NetBSD__) +#define HAS_VOPREVOKE 1 +#endif + +/* + * The IO_METASYNC flag should be implemented for local file systems. + * (Until then, it is nothin at all.) + */ +#ifndef IO_METASYNC +#define IO_METASYNC 0 +#endif /* * Set the attribute timeout based on how recently the file has been modified. @@ -80,6 +149,20 @@ (time.tv_sec - (np)->n_mtime) / 10)) /* + * Expected allocation sizes for major data structures. If the actual size + * of the structure exceeds these sizes, then malloc() will be allocating + * almost twice the memory required. This is used in nfs_init() to warn + * the sysadmin that the size of a structure should be reduced. + * (These sizes are always a power of 2. If the kernel malloc() changes + * to one that does not allocate space in powers of 2 size, then this all + * becomes bunk!) + */ +#define NFS_NODEALLOC 256 +#define NFS_MNTALLOC 512 +#define NFS_SVCALLOC 256 +#define NFS_UIDALLOC 128 + +/* * Structures for the nfssvc(2) syscall. Not that anyone but nfsd and mount_nfs * should ever try and use it. */ @@ -95,7 +178,12 @@ struct nfsd_srvargs { u_long nsd_haddr; /* Ip address of client */ struct ucred nsd_cr; /* Cred. uid maps to */ int nsd_authlen; /* Length of auth string (ret) */ - char *nsd_authstr; /* Auth string (ret) */ + u_char *nsd_authstr; /* Auth string (ret) */ + int nsd_verflen; /* and the verfier */ + u_char *nsd_verfstr; + struct timeval nsd_timestamp; /* timestamp from verifier */ + u_long nsd_ttl; /* credential ttl (sec) */ + NFSKERBKEY_T nsd_key; /* Session key */ }; struct nfsd_cargs { @@ -103,7 +191,10 @@ struct nfsd_cargs { uid_t ncd_authuid; /* Effective uid */ int ncd_authtype; /* Type of authenticator */ int ncd_authlen; /* Length of authenticator string */ - char *ncd_authstr; /* Authenticator string */ + u_char *ncd_authstr; /* Authenticator string */ + int ncd_verflen; /* and the verifier */ + u_char *ncd_verfstr; + NFSKERBKEY_T ncd_key; /* Session key */ }; /* @@ -142,6 +233,7 @@ struct nfsstats { int srvnqnfs_leases; int srvnqnfs_maxleases; int srvnqnfs_getleases; + int srvvop_writes; }; /* @@ -173,7 +265,7 @@ struct nfsstats { * such as SIGALRM will not expect file I/O system calls to be interrupted * by them and break. */ -#ifdef KERNEL +#if defined(KERNEL) || defined(_KERNEL) struct uio; struct buf; struct vattr; struct nameidata; /* XXX */ @@ -224,17 +316,29 @@ TAILQ_HEAD(, nfsreq) nfs_reqq; #define R_MUSTRESEND 0x40 /* Must resend request */ #define R_GETONEREP 0x80 /* Probe for one reply only */ -extern struct nfsstats nfsstats; - /* * A list of nfssvc_sock structures is maintained with all the sockets * that require service by the nfsd. * The nfsuid structs hang off of the nfssvc_sock structs in both lru * and uid hash lists. */ -#define NUIDHASHSIZ 32 +#ifndef NFS_UIDHASHSIZ +#define NFS_UIDHASHSIZ 29 /* Tune the size of nfssvc_sock with this */ +#endif #define NUIDHASH(sock, uid) \ - (&(sock)->ns_uidhashtbl[(uid) & (sock)->ns_uidhash]) + (&(sock)->ns_uidhashtbl[(uid) % NFS_UIDHASHSIZ]) +#ifndef NFS_WDELAYHASHSIZ +#define NFS_WDELAYHASHSIZ 16 /* and with this */ +#endif +#define NWDELAYHASH(sock, f) \ + (&(sock)->ns_wdelayhashtbl[(*((u_long *)(f))) % NFS_WDELAYHASHSIZ]) +#ifndef NFS_MUIDHASHSIZ +#define NFS_MUIDHASHSIZ 67 /* Tune the size of nfsmount with this */ +#endif +#define NMUIDHASH(nmp, uid) \ + (&(nmp)->nm_uidhashtbl[(uid) % NFS_MUIDHASHSIZ]) +#define NFSNOHASH(fhsum) \ + (&nfsnodehashtbl[(fhsum) & nfsnodehash]) /* * Network address hash list element @@ -248,35 +352,41 @@ struct nfsuid { TAILQ_ENTRY(nfsuid) nu_lru; /* LRU chain */ LIST_ENTRY(nfsuid) nu_hash; /* Hash list */ int nu_flag; /* Flags */ - uid_t nu_uid; /* Uid mapped by this entry */ union nethostaddr nu_haddr; /* Host addr. for dgram sockets */ struct ucred nu_cr; /* Cred uid mapped to */ + int nu_expire; /* Expiry time (sec) */ + struct timeval nu_timestamp; /* Kerb. timestamp */ + u_long nu_nickname; /* Nickname on server */ + NFSKERBKEY_T nu_key; /* and session key */ }; #define nu_inetaddr nu_haddr.had_inetaddr #define nu_nam nu_haddr.had_nam /* Bits for nu_flag */ #define NU_INETADDR 0x1 +#define NU_NAM 0x2 +#define NU_NETFAM(u) (((u)->nu_flag & NU_INETADDR) ? AF_INET : AF_ISO) struct nfssvc_sock { TAILQ_ENTRY(nfssvc_sock) ns_chain; /* List of all nfssvc_sock's */ TAILQ_HEAD(, nfsuid) ns_uidlruhead; - LIST_HEAD(, nfsuid) *ns_uidhashtbl; - u_long ns_uidhash; - - int ns_flag; - u_long ns_sref; struct file *ns_fp; struct socket *ns_so; - int ns_solock; struct mbuf *ns_nam; - int ns_cc; struct mbuf *ns_raw; struct mbuf *ns_rawend; - int ns_reclen; struct mbuf *ns_rec; struct mbuf *ns_recend; + struct mbuf *ns_frag; + int ns_flag; + int ns_solock; + int ns_cc; + int ns_reclen; int ns_numuids; + u_long ns_sref; + LIST_HEAD(, nfsrv_descript) ns_tq; /* Write gather lists */ + LIST_HEAD(, nfsuid) ns_uidhashtbl[NFS_UIDHASHSIZ]; + LIST_HEAD(nfsrvw_delayhash, nfsrv_descript) ns_wdelayhashtbl[NFS_WDELAYHASHSIZ]; }; /* Bits for "ns_flag" */ @@ -285,6 +395,7 @@ struct nfssvc_sock { #define SLP_NEEDQ 0x04 #define SLP_DISCONN 0x08 #define SLP_GETSTREAM 0x10 +#define SLP_LASTFRAG 0x20 #define SLP_ALLFLAGS 0xff TAILQ_HEAD(, nfssvc_sock) nfssvc_sockhead; @@ -296,73 +407,124 @@ int nfssvc_sockhead_flag; * One of these structures is allocated for each nfsd. */ struct nfsd { - TAILQ_ENTRY(nfsd) nd_chain; /* List of all nfsd's */ - int nd_flag; /* NFSD_ flags */ - struct nfssvc_sock *nd_slp; /* Current socket */ - struct mbuf *nd_nam; /* Client addr for datagram req. */ - struct mbuf *nd_mrep; /* Req. mbuf list */ - struct mbuf *nd_md; - caddr_t nd_dpos; /* Position in list */ - int nd_procnum; /* RPC procedure number */ - u_long nd_retxid; /* RPC xid */ - int nd_repstat; /* Reply status value */ - struct ucred nd_cr; /* Credentials for req. */ - int nd_nqlflag; /* Leasing flag */ - u_long nd_duration; /* Lease duration */ - int nd_authlen; /* Authenticator len */ - u_char nd_authstr[RPCAUTH_MAXSIZ]; /* Authenticator data */ - struct proc *nd_procp; /* Proc ptr */ + TAILQ_ENTRY(nfsd) nfsd_chain; /* List of all nfsd's */ + int nfsd_flag; /* NFSD_ flags */ + struct nfssvc_sock *nfsd_slp; /* Current socket */ + int nfsd_authlen; /* Authenticator len */ + u_char nfsd_authstr[RPCAUTH_MAXSIZ]; /* Authenticator data */ + int nfsd_verflen; /* and the Verifier */ + u_char nfsd_verfstr[RPCVERF_MAXSIZ]; + struct proc *nfsd_procp; /* Proc ptr */ + struct nfsrv_descript *nfsd_nd; /* Associated nfsrv_descript */ }; -/* Bits for "nd_flag" */ +/* Bits for "nfsd_flag" */ #define NFSD_WAITING 0x01 #define NFSD_REQINPROG 0x02 #define NFSD_NEEDAUTH 0x04 #define NFSD_AUTHFAIL 0x08 +/* + * This structure is used by the server for describing each request. + * Some fields are used only when write request gathering is performed. + */ +struct nfsrv_descript { + u_quad_t nd_time; /* Write deadline (usec) */ + off_t nd_off; /* Start byte offset */ + off_t nd_eoff; /* and end byte offset */ + LIST_ENTRY(nfsrv_descript) nd_hash; /* Hash list */ + LIST_ENTRY(nfsrv_descript) nd_tq; /* and timer list */ + LIST_HEAD(,nfsrv_descript) nd_coalesce; /* coalesced writes */ + struct mbuf *nd_mrep; /* Request mbuf list */ + struct mbuf *nd_md; /* Current dissect mbuf */ + struct mbuf *nd_mreq; /* Reply mbuf list */ + struct mbuf *nd_nam; /* and socket addr */ + struct mbuf *nd_nam2; /* return socket addr */ + caddr_t nd_dpos; /* Current dissect pos */ + int nd_procnum; /* RPC # */ + int nd_stable; /* storage type */ + int nd_flag; /* nd_flag */ + int nd_len; /* Length of this write */ + int nd_repstat; /* Reply status */ + u_long nd_retxid; /* Reply xid */ + u_long nd_duration; /* Lease duration */ + struct timeval nd_starttime; /* Time RPC initiated */ + fhandle_t nd_fh; /* File handle */ + struct ucred nd_cr; /* Credentials */ +}; + +/* Bits for "nd_flag" */ +#define ND_READ LEASE_READ +#define ND_WRITE LEASE_WRITE +#define ND_CHECK 0x04 +#define ND_LEASE (ND_READ | ND_WRITE | ND_CHECK) +#define ND_NFSV3 0x08 +#define ND_NQNFS 0x10 +#define ND_KERBNICK 0x20 +#define ND_KERBFULL 0x40 +#define ND_KERBAUTH (ND_KERBNICK | ND_KERBFULL) + TAILQ_HEAD(, nfsd) nfsd_head; int nfsd_head_flag; #define NFSD_CHECKSLP 0x01 +/* + * These macros compare nfsrv_descript structures. + */ +#define NFSW_CONTIG(o, n) \ + ((o)->nd_eoff >= (n)->nd_off && \ + !bcmp((caddr_t)&(o)->nd_fh, (caddr_t)&(n)->nd_fh, NFSX_V3FH)) + +#define NFSW_SAMECRED(o, n) \ + (((o)->nd_flag & ND_KERBAUTH) == ((n)->nd_flag & ND_KERBAUTH) && \ + !bcmp((caddr_t)&(o)->nd_cr, (caddr_t)&(n)->nd_cr, \ + sizeof (struct ucred))) + int nfs_reply __P((struct nfsreq *)); -int nfs_getreq __P((struct nfsd *,int)); +int nfs_getreq __P((struct nfsrv_descript *,struct nfsd *,int)); int nfs_send __P((struct socket *,struct mbuf *,struct mbuf *,struct nfsreq *)); -int nfs_rephead __P((int,struct nfsd *,int,int,u_quad_t *,struct mbuf **,struct mbuf **,caddr_t *)); +int nfs_rephead __P((int,struct nfsrv_descript *,struct nfssvc_sock *,int,int,u_quad_t *,struct mbuf **,struct mbuf **,caddr_t *)); int nfs_sndlock __P((int *,struct nfsreq *)); int nfs_disct __P((struct mbuf **,caddr_t *,int,int,caddr_t *)); int nfs_vinvalbuf __P((struct vnode *,int,struct ucred *,struct proc *,int)); int nfs_readrpc __P((struct vnode *,struct uio *,struct ucred *)); -int nfs_writerpc __P((struct vnode *,struct uio *,struct ucred *,int)); +int nfs_writerpc __P((struct vnode *,struct uio *,struct ucred *,int *,int *)); int nfs_readdirrpc __P((register struct vnode *,struct uio *,struct ucred *)); +int nfs_setattrrpc __P((struct vnode *,struct vattr *,struct ucred *,struct proc *)); int nfs_asyncio __P((struct buf *,struct ucred *)); int nfs_doio __P((struct buf *,struct ucred *,struct proc *)); int nfs_readlinkrpc __P((struct vnode *,struct uio *,struct ucred *)); int nfs_sigintr __P((struct nfsmount *,struct nfsreq *r,struct proc *)); -int nfs_readdirlookrpc __P((struct vnode *,register struct uio *,struct ucred *)); +int nfs_readdirplusrpc __P((struct vnode *,register struct uio *,struct ucred *)); int nfsm_disct __P((struct mbuf **,caddr_t *,int,int,caddr_t *)); -int nfsrv_fhtovp __P((fhandle_t *,int,struct vnode **,struct ucred *,struct nfssvc_sock *,struct mbuf *,int *)); +void nfsm_srvfattr __P((struct nfsrv_descript *,struct vattr *,struct nfs_fattr *)); +void nfsm_srvwcc __P((struct nfsrv_descript *,int,struct vattr *,int,struct vattr *,struct mbuf **,char **)); +void nfsm_srvpostopattr __P((struct nfsrv_descript *,int,struct vattr *,struct mbuf **,char **)); +int nfsrv_fhtovp __P((fhandle_t *,int,struct vnode **,struct ucred *,struct nfssvc_sock *,struct mbuf *,int *,int)); int nfsrv_access __P((struct vnode *,int,struct ucred *,int,struct proc *)); int netaddr_match __P((int,union nethostaddr *,struct mbuf *)); int nfs_request __P((struct vnode *,struct mbuf *,int,struct proc *,struct ucred *,struct mbuf **,struct mbuf **,caddr_t *)); int nfs_loadattrcache __P((struct vnode **,struct mbuf **,caddr_t *,struct vattr *)); -int nfs_namei __P((struct nameidata *,fhandle_t *,int,struct nfssvc_sock *,struct mbuf *,struct mbuf **,caddr_t *,struct proc *)); +int nfs_namei __P((struct nameidata *,fhandle_t *,int,struct nfssvc_sock *,struct mbuf *,struct mbuf **,caddr_t *,struct vnode **,struct proc *,int)); void nfsm_adj __P((struct mbuf *,int,int)); int nfsm_mbuftouio __P((struct mbuf **,struct uio *,int,caddr_t *)); void nfsrv_initcache __P((void)); int nfs_rcvlock __P((struct nfsreq *)); -int nfs_getauth __P((struct nfsmount *,struct nfsreq *,struct ucred *,int *,char **,int *)); +int nfs_getauth __P((struct nfsmount *,struct nfsreq *,struct ucred *,char **,int *,char *,int *,NFSKERBKEY_T)); +int nfs_getnickauth __P((struct nfsmount *,struct ucred *,char **,int *,char *,int)); +int nfs_savenickauth __P((struct nfsmount *,struct ucred *,int,NFSKERBKEY_T,struct mbuf **,char **,struct mbuf *)); int nfs_msg __P((struct proc *,char *,char *)); int nfs_adv __P((struct mbuf **,caddr_t *,int,int)); int nfsrv_getstream __P((struct nfssvc_sock *,int)); void nfs_nhinit __P((void)); void nfs_timer __P((void*)); -struct nfsnodehashhead * nfs_hash __P((nfsv2fh_t *)); +u_long nfs_hash __P((nfsfh_t *,int)); int nfssvc_iod __P((struct proc *)); int nfssvc_nfsd __P((struct nfsd_srvargs *,caddr_t,struct proc *)); int nfssvc_addsock __P((struct file *,struct mbuf *)); -int nfsrv_dorec __P((struct nfssvc_sock *,struct nfsd *)); -int nfsrv_getcache __P((struct mbuf *,struct nfsd *,struct mbuf **)); -void nfsrv_updatecache __P((struct mbuf *,struct nfsd *,int,struct mbuf *)); +int nfsrv_dorec __P((struct nfssvc_sock *,struct nfsd *,struct nfsrv_descript **)); +int nfsrv_getcache __P((struct nfsrv_descript *,struct nfssvc_sock *,struct mbuf **)); +void nfsrv_updatecache __P((struct nfsrv_descript *,int,struct mbuf *)); int mountnfs __P((struct nfs_args *,struct mount *,struct mbuf *,char *,char *,struct vnode **)); int nfs_connect __P((struct nfsmount *,struct nfsreq *)); int nfs_getattrcache __P((struct vnode *,struct vattr *)); @@ -370,9 +532,20 @@ int nfsm_strtmbuf __P((struct mbuf **,char **,char *,long)); int nfs_bioread __P((struct vnode *,struct uio *,int,struct ucred *)); int nfsm_uiotombuf __P((struct uio *,struct mbuf **,int,caddr_t *)); void nfsrv_init __P((int)); +void nfs_clearcommit __P((struct mount *)); +int nfsrv_errmap __P((struct nfsrv_descript *, int)); +void nfsrvw_coalesce __P((struct nfsrv_descript *,struct nfsrv_descript *)); +void nfsrvw_sort __P((gid_t [],int)); +void nfsrv_setcred __P((struct ucred *,struct ucred *)); +int nfs_flush __P((struct vnode *,struct ucred *,int,struct proc *,int)); +int nfs_writebp __P((struct buf *,int)); int nfsrv_vput __P(( struct vnode * )); int nfsrv_vrele __P(( struct vnode * )); int nfsrv_vmio __P(( struct vnode * )); +int nfsrv_writegather __P((struct nfsrv_descript **, struct nfssvc_sock *, + struct proc *, struct mbuf **)); +int nfs_fsinfo __P((struct nfsmount *, struct vnode *, struct ucred *, + struct proc *p)); #endif /* KERNEL */ diff --git a/sys/nfs/nfs_bio.c b/sys/nfs/nfs_bio.c index f929012..f9ae8b6 100644 --- a/sys/nfs/nfs_bio.c +++ b/sys/nfs/nfs_bio.c @@ -34,7 +34,7 @@ * SUCH DAMAGE. * * @(#)nfs_bio.c 8.5 (Berkeley) 1/4/94 - * $Id: nfs_bio.c,v 1.13 1995/05/21 21:39:21 davidg Exp $ + * $Id: nfs_bio.c,v 1.14 1995/05/30 08:12:35 rgrimes Exp $ */ #include <sys/param.h> @@ -49,16 +49,30 @@ #include <vm/vm.h> -#include <nfs/nfsnode.h> #include <nfs/rpcv2.h> -#include <nfs/nfsv2.h> +#include <nfs/nfsproto.h> #include <nfs/nfs.h> #include <nfs/nfsmount.h> #include <nfs/nqnfs.h> +#include <nfs/nfsnode.h> struct buf *nfs_getcacheblk(); extern struct proc *nfs_iodwant[NFS_MAXASYNCDAEMON]; extern int nfs_numasync; +extern struct nfsstats nfsstats; + +/* + * Ifdefs for FreeBSD-current's merged VM/buffer cache. It is unfortunate + * that this isn't done inside getblk() and brelse() so these calls + * wouldn't need to be here. + */ +#ifdef B_VMIO +#define vnode_pager_uncache(vp) +#else +#define vfs_busy_pages(bp, f) +#define vfs_unbusy_pages(bp) +#define vfs_dirty_pages(bp) +#endif /* * Vnode op for read using bio @@ -72,29 +86,28 @@ nfs_bioread(vp, uio, ioflag, cred) struct ucred *cred; { register struct nfsnode *np = VTONFS(vp); - register int biosize, diff; + register int biosize, diff, i; struct buf *bp = 0, *rabp; struct vattr vattr; struct proc *p; - struct nfsmount *nmp; + struct nfsmount *nmp = VFSTONFS(vp->v_mount); daddr_t lbn, rabn; int bufsize; int nra, error = 0, n = 0, on = 0, not_readin; + nfsquad_t tquad; -#ifdef lint - ioflag = ioflag; -#endif /* lint */ #ifdef DIAGNOSTIC if (uio->uio_rw != UIO_READ) panic("nfs_read mode"); #endif if (uio->uio_resid == 0) return (0); - if (uio->uio_offset < 0 && vp->v_type != VDIR) + if (uio->uio_offset < 0) return (EINVAL); - nmp = VFSTONFS(vp->v_mount); - biosize = NFS_MAXDGRAMDATA; p = uio->uio_procp; + if ((nmp->nm_flag & (NFSMNT_NFSV3 | NFSMNT_GOTFSINFO)) == NFSMNT_NFSV3) + (void)nfs_fsinfo(nmp, vp, cred, p); + biosize = nmp->nm_rsize; /* * For nfs, cache consistency can only be maintained approximately. * Although RFC1094 does not specify the criteria, the following is @@ -107,8 +120,6 @@ nfs_bioread(vp, uio, ioflag, cred) * server, so flush all of the file's data out of the cache. * Then force a getattr rpc to ensure that you have up to date * attributes. - * The mount flag NFSMNT_MYWRITE says "Assume that my writes are - * the ones changing the modify time. * NB: This implies that cache data can be read when up to * NFS_ATTRTIMEO seconds out of date. If you find that you need current * attributes this could be forced by setting n_attrstamp to 0 before @@ -116,14 +127,15 @@ nfs_bioread(vp, uio, ioflag, cred) */ if ((nmp->nm_flag & NFSMNT_NQNFS) == 0 && vp->v_type != VLNK) { if (np->n_flag & NMODIFIED) { - if ((nmp->nm_flag & NFSMNT_MYWRITE) == 0 || - vp->v_type != VREG) { + if (vp->v_type != VREG) { + if (vp->v_type != VDIR) + panic("nfs: bioread, not dir"); + nfs_invaldir(vp); error = nfs_vinvalbuf(vp, V_SAVE, cred, p, 1); if (error) return (error); } np->n_attrstamp = 0; - np->n_direofoffset = 0; error = VOP_GETATTR(vp, &vattr, cred, p); if (error) return (error); @@ -133,7 +145,8 @@ nfs_bioread(vp, uio, ioflag, cred) if (error) return (error); if (np->n_mtime != vattr.va_mtime.ts_sec) { - np->n_direofoffset = 0; + if (vp->v_type == VDIR) + nfs_invaldir(vp); error = nfs_vinvalbuf(vp, V_SAVE, cred, p, 1); if (error) return (error); @@ -147,27 +160,24 @@ nfs_bioread(vp, uio, ioflag, cred) * Get a valid lease. If cached data is stale, flush it. */ if (nmp->nm_flag & NFSMNT_NQNFS) { - if (NQNFS_CKINVALID(vp, np, NQL_READ)) { + if (NQNFS_CKINVALID(vp, np, ND_READ)) { do { - error = nqnfs_getlease(vp, NQL_READ, cred, p); + error = nqnfs_getlease(vp, ND_READ, cred, p); } while (error == NQNFS_EXPIRED); if (error) return (error); if (np->n_lrev != np->n_brev || (np->n_flag & NQNFSNONCACHE) || ((np->n_flag & NMODIFIED) && vp->v_type == VDIR)) { - if (vp->v_type == VDIR) { - np->n_direofoffset = 0; - cache_purge(vp); - } + if (vp->v_type == VDIR) + nfs_invaldir(vp); error = nfs_vinvalbuf(vp, V_SAVE, cred, p, 1); if (error) return (error); np->n_brev = np->n_lrev; } } else if (vp->v_type == VDIR && (np->n_flag & NMODIFIED)) { - np->n_direofoffset = 0; - cache_purge(vp); + nfs_invaldir(vp); error = nfs_vinvalbuf(vp, V_SAVE, cred, p, 1); if (error) return (error); @@ -176,33 +186,27 @@ nfs_bioread(vp, uio, ioflag, cred) if (np->n_flag & NQNFSNONCACHE) { switch (vp->v_type) { case VREG: - error = nfs_readrpc(vp, uio, cred); - break; + return (nfs_readrpc(vp, uio, cred)); case VLNK: - error = nfs_readlinkrpc(vp, uio, cred); - break; + return (nfs_readlinkrpc(vp, uio, cred)); case VDIR: - error = nfs_readdirrpc(vp, uio, cred); break; default: printf(" NQNFSNONCACHE: type %x unexpected\n", vp->v_type); - break; }; - return (error); } switch (vp->v_type) { case VREG: nfsstats.biocache_reads++; lbn = uio->uio_offset / biosize; - on = uio->uio_offset & (biosize-1); + on = uio->uio_offset & (biosize - 1); not_readin = 1; /* * Start the read ahead(s), as required. */ - if (nfs_numasync > 0 && nmp->nm_readahead > 0 && - lbn == vp->v_lastr + 1) { + if (nfs_numasync > 0 && nmp->nm_readahead > 0) { for (nra = 0; nra < nmp->nm_readahead && (lbn + 1 + nra) * biosize < np->n_size; nra++) { rabn = lbn + 1 + nra; @@ -296,32 +300,55 @@ again: break; case VDIR: nfsstats.biocache_readdirs++; - lbn = (daddr_t)uio->uio_offset; + lbn = uio->uio_offset / NFS_DIRBLKSIZ; + on = uio->uio_offset & (NFS_DIRBLKSIZ - 1); bp = nfs_getcacheblk(vp, lbn, NFS_DIRBLKSIZ, p); if (!bp) - return (EINTR); - + return (EINTR); if ((bp->b_flags & B_CACHE) == 0) { - bp->b_flags |= B_READ; - vfs_busy_pages(bp, 0); - error = nfs_doio(bp, cred, p); - if (error) { - bp->b_flags |= B_ERROR; - brelse(bp); - return (error); + bp->b_flags |= B_READ; + vfs_busy_pages(bp, 0); + error = nfs_doio(bp, cred, p); + if (error) { + brelse(bp); + while (error == NFSERR_BAD_COOKIE) { + nfs_invaldir(vp); + error = nfs_vinvalbuf(vp, 0, cred, p, 1); + /* + * Yuck! The directory has been modified on the + * server. The only way to get the block is by + * reading from the beginning to get all the + * offset cookies. + */ + for (i = 0; i <= lbn && !error; i++) { + bp = nfs_getcacheblk(vp, i, NFS_DIRBLKSIZ, p); + if (!bp) + return (EINTR); + if ((bp->b_flags & B_DONE) == 0) { + bp->b_flags |= B_READ; + vfs_busy_pages(bp, 0); + error = nfs_doio(bp, cred, p); + if (error) + brelse(bp); + } + } } + if (error) + return (error); + } } /* * If not eof and read aheads are enabled, start one. * (You need the current block first, so that you have the - * directory offset cookie of the next block. + * directory offset cookie of the next block.) */ - rabn = bp->b_blkno; if (nfs_numasync > 0 && nmp->nm_readahead > 0 && - rabn != 0 && rabn != np->n_direofoffset && - !incore(vp, rabn)) { - rabp = nfs_getcacheblk(vp, rabn, NFS_DIRBLKSIZ, p); + (np->n_direofoffset == 0 || + (lbn + 1) * NFS_DIRBLKSIZ < np->n_direofoffset) && + !(np->n_flag & NQNFSNONCACHE) && + !incore(vp, lbn + 1)) { + rabp = nfs_getcacheblk(vp, lbn + 1, NFS_DIRBLKSIZ, p); if (rabp) { if ((rabp->b_flags & (B_CACHE|B_DELWRI)) == 0) { rabp->b_flags |= (B_READ | B_ASYNC); @@ -336,11 +363,10 @@ again: } } } - on = 0; - n = min(uio->uio_resid, NFS_DIRBLKSIZ - bp->b_resid); + n = min(uio->uio_resid, NFS_DIRBLKSIZ - bp->b_resid - on); break; default: - printf(" nfsbioread: type %x unexpected\n",vp->v_type); + printf(" nfs_bioread: type %x unexpected\n",vp->v_type); break; }; @@ -354,11 +380,11 @@ again: n = 0; break; case VDIR: - uio->uio_offset = bp->b_blkno; + if (np->n_flag & NQNFSNONCACHE) + bp->b_flags |= B_INVAL; break; default: - printf(" nfsbioread: type %x unexpected\n",vp->v_type); - break; + printf(" nfs_bioread: type %x unexpected\n",vp->v_type); } brelse(bp); } while (error == 0 && uio->uio_resid > 0 && n > 0); @@ -386,10 +412,10 @@ nfs_write(ap) int ioflag = ap->a_ioflag; struct buf *bp; struct vattr vattr; - struct nfsmount *nmp; - daddr_t lbn; + struct nfsmount *nmp = VFSTONFS(vp->v_mount); + daddr_t lbn, bn; int bufsize; - int n, on, error = 0; + int n, on, error = 0, iomode, must_commit; #ifdef DIAGNOSTIC if (uio->uio_rw != UIO_WRITE) @@ -403,6 +429,8 @@ nfs_write(ap) np->n_flag &= ~NWRITEERR; return (np->n_error); } + if ((nmp->nm_flag & (NFSMNT_NFSV3 | NFSMNT_GOTFSINFO)) == NFSMNT_NFSV3) + (void)nfs_fsinfo(nmp, vp, cred, p); if (ioflag & (IO_APPEND | IO_SYNC)) { if (np->n_flag & NMODIFIED) { np->n_attrstamp = 0; @@ -418,7 +446,6 @@ nfs_write(ap) uio->uio_offset = np->n_size; } } - nmp = VFSTONFS(vp->v_mount); if (uio->uio_offset < 0) return (EINVAL); if (uio->uio_resid == 0) @@ -437,7 +464,7 @@ nfs_write(ap) * will be the same size within a filesystem. nfs_writerpc will * still use nm_wsize when sizing the rpc's. */ - biosize = NFS_MAXDGRAMDATA; + biosize = nmp->nm_rsize; do { /* @@ -445,12 +472,11 @@ nfs_write(ap) */ /* * Check for a valid write lease. - * If non-cachable, just do the rpc */ if ((nmp->nm_flag & NFSMNT_NQNFS) && - NQNFS_CKINVALID(vp, np, NQL_WRITE)) { + NQNFS_CKINVALID(vp, np, ND_WRITE)) { do { - error = nqnfs_getlease(vp, NQL_WRITE, cred, p); + error = nqnfs_getlease(vp, ND_WRITE, cred, p); } while (error == NQNFS_EXPIRED); if (error) return (error); @@ -462,8 +488,13 @@ nfs_write(ap) np->n_brev = np->n_lrev; } } - if (np->n_flag & NQNFSNONCACHE) - return (nfs_writerpc(vp, uio, cred, ioflag)); + if ((np->n_flag & NQNFSNONCACHE) && uio->uio_iovcnt == 1) { + iomode = NFSV3WRITE_FILESYNC; + error = nfs_writerpc(vp, uio, cred, &iomode, &must_commit); + if (must_commit) + nfs_clearcommit(vp->v_mount); + return (error); + } nfsstats.biocache_writes++; lbn = uio->uio_offset / biosize; on = uio->uio_offset & (biosize-1); @@ -509,9 +540,9 @@ again: * In case getblk() and/or bwrite() delayed us. */ if ((nmp->nm_flag & NFSMNT_NQNFS) && - NQNFS_CKINVALID(vp, np, NQL_WRITE)) { + NQNFS_CKINVALID(vp, np, ND_WRITE)) { do { - error = nqnfs_getlease(vp, NQL_WRITE, cred, p); + error = nqnfs_getlease(vp, ND_WRITE, cred, p); } while (error == NQNFS_EXPIRED); if (error) { brelse(bp); @@ -540,7 +571,6 @@ again: bp->b_dirtyoff = on; bp->b_dirtyend = on + n; } -#ifndef notdef if (bp->b_validend == 0 || bp->b_validend < bp->b_dirtyoff || bp->b_validoff > bp->b_dirtyend) { bp->b_validoff = bp->b_dirtyoff; @@ -549,13 +579,6 @@ again: bp->b_validoff = min(bp->b_validoff, bp->b_dirtyoff); bp->b_validend = max(bp->b_validend, bp->b_dirtyend); } -#else - bp->b_validoff = bp->b_dirtyoff; - bp->b_validend = bp->b_dirtyend; -#endif - if (ioflag & IO_APPEND) - bp->b_flags |= B_APPENDWRITE; - /* * If the lease is non-cachable or IO_SYNC do bwrite(). */ @@ -564,10 +587,16 @@ again: error = VOP_BWRITE(bp); if (error) return (error); + if (np->n_flag & NQNFSNONCACHE) { + error = nfs_vinvalbuf(vp, V_SAVE, cred, p, 1); + if (error) + return (error); + } } else if ((n + on) == biosize && (nmp->nm_flag & NFSMNT_NQNFS) == 0) { bp->b_proc = (struct proc *)0; - bawrite(bp); + bp->b_flags |= B_ASYNC; + (void)nfs_writebp(bp, 0); } else bdwrite(bp); } while (uio->uio_resid > 0 && n > 0); @@ -590,6 +619,7 @@ nfs_getcacheblk(vp, bn, size, p) { register struct buf *bp; struct nfsmount *nmp = VFSTONFS(vp->v_mount); + int biosize = nmp->nm_rsize; if (nmp->nm_flag & NFSMNT_INT) { bp = getblk(vp, bn, size, PCATCH, 0); @@ -602,7 +632,7 @@ nfs_getcacheblk(vp, bn, size, p) bp = getblk(vp, bn, size, 0, 0); if( vp->v_type == VREG) - bp->b_blkno = (bn * NFS_MAXDGRAMDATA) / DEV_BSIZE; + bp->b_blkno = (bn * biosize) / DEV_BSIZE; return (bp); } @@ -689,6 +719,7 @@ nfs_asyncio(bp, cred) bp->b_rcred = cred; } } else { + bp->b_flags |= B_WRITEINPROG; if (bp->b_wcred == NOCRED && cred != NOCRED) { crhold(cred); bp->b_wcred = cred; @@ -700,7 +731,25 @@ nfs_asyncio(bp, cred) wakeup((caddr_t)&nfs_iodwant[i]); return (0); } - return (EIO); + + /* + * If it is a read or a write already marked B_WRITEINPROG or B_NOCACHE + * return EIO so the process will call nfs_doio() and do it + * synchronously. + */ + if (bp->b_flags & (B_READ | B_WRITEINPROG | B_NOCACHE)) + return (EIO); + + /* + * Just turn the async write into a delayed write, instead of + * doing in synchronously. Hopefully, at least one of the nfsiods + * is currently doing a write for this file and will pick up the + * delayed writes before going back to sleep. + */ + bp->b_flags |= B_DELWRI; + reassignbuf(bp, bp->b_vp); + biodone(bp); + return (0); } /* @@ -717,9 +766,10 @@ nfs_doio(bp, cr, p) register struct vnode *vp; struct nfsnode *np; struct nfsmount *nmp; - int error = 0, diff, len; + int error = 0, diff, len, iomode, must_commit = 0; struct uio uio; struct iovec io; + nfsquad_t tquad; vp = bp->b_vp; np = VTONFS(vp); @@ -740,15 +790,18 @@ nfs_doio(bp, cr, p) io.iov_len = uiop->uio_resid = bp->b_bcount; /* mapping was done by vmapbuf() */ io.iov_base = bp->b_data; - uiop->uio_offset = bp->b_blkno * DEV_BSIZE; + uiop->uio_offset = ((off_t)bp->b_blkno) * DEV_BSIZE; if (bp->b_flags & B_READ) { uiop->uio_rw = UIO_READ; nfsstats.read_physios++; error = nfs_readrpc(vp, uiop, cr); } else { + int com; + + iomode = NFSV3WRITE_DATASYNC; uiop->uio_rw = UIO_WRITE; nfsstats.write_physios++; - error = nfs_writerpc(vp, uiop, cr,0); + error = nfs_writerpc(vp, uiop, cr, &iomode, &com); } if (error) { bp->b_flags |= B_ERROR; @@ -760,7 +813,7 @@ nfs_doio(bp, cr, p) uiop->uio_rw = UIO_READ; switch (vp->v_type) { case VREG: - uiop->uio_offset = bp->b_blkno * DEV_BSIZE; + uiop->uio_offset = ((off_t)bp->b_blkno) * DEV_BSIZE; nfsstats.read_bios++; error = nfs_readrpc(vp, uiop, cr); if (!error) { @@ -773,7 +826,7 @@ nfs_doio(bp, cr, p) * Just zero fill the rest of the valid area. */ diff = bp->b_bcount - uiop->uio_resid; - len = np->n_size - (bp->b_blkno * DEV_BSIZE + len = np->n_size - (((u_quad_t)bp->b_blkno) * DEV_BSIZE + diff); if (len > 0) { len = min(len, uiop->uio_resid); @@ -786,31 +839,34 @@ nfs_doio(bp, cr, p) } if (p && (vp->v_flag & VTEXT) && (((nmp->nm_flag & NFSMNT_NQNFS) && - NQNFS_CKINVALID(vp, np, NQL_READ) && + NQNFS_CKINVALID(vp, np, ND_READ) && np->n_lrev != np->n_brev) || (!(nmp->nm_flag & NFSMNT_NQNFS) && np->n_mtime != np->n_vattr.va_mtime.ts_sec))) { uprintf("Process killed due to text file modification\n"); psignal(p, SIGKILL); +#ifdef __NetBSD__ + p->p_holdcnt++; +#else p->p_flag |= P_NOSWAP; +#endif } break; case VLNK: - uiop->uio_offset = 0; + uiop->uio_offset = (off_t)0; nfsstats.readlink_bios++; error = nfs_readlinkrpc(vp, uiop, cr); break; case VDIR: - uiop->uio_offset = bp->b_lblkno; nfsstats.readdir_bios++; - if (VFSTONFS(vp->v_mount)->nm_flag & NFSMNT_NQNFS) - error = nfs_readdirlookrpc(vp, uiop, cr); - else - error = nfs_readdirrpc(vp, uiop, cr); - /* - * Save offset cookie in b_blkno. - */ - bp->b_blkno = uiop->uio_offset; + uiop->uio_offset = ((u_quad_t)bp->b_lblkno) * NFS_DIRBLKSIZ; + if (nmp->nm_flag & NFSMNT_RDIRPLUS) { + error = nfs_readdirplusrpc(vp, uiop, cr); + if (error == NFSERR_NOTSUPP) + nmp->nm_flag &= ~NFSMNT_RDIRPLUS; + } + if ((nmp->nm_flag & NFSMNT_RDIRPLUS) == 0) + error = nfs_readdirrpc(vp, uiop, cr); break; default: printf("nfs_doio: type %x unexpected\n",vp->v_type); @@ -821,32 +877,44 @@ nfs_doio(bp, cr, p) bp->b_error = error; } } else { - if (((bp->b_blkno * DEV_BSIZE) + bp->b_dirtyend) > np->n_size) bp->b_dirtyend = np->n_size - (bp->b_blkno * DEV_BSIZE); if (bp->b_dirtyend > bp->b_dirtyoff) { io.iov_len = uiop->uio_resid = bp->b_dirtyend - - bp->b_dirtyoff; - uiop->uio_offset = (bp->b_blkno * DEV_BSIZE) - + bp->b_dirtyoff; + - bp->b_dirtyoff; + uiop->uio_offset = ((off_t)bp->b_blkno) * DEV_BSIZE + + bp->b_dirtyoff; io.iov_base = (char *)bp->b_data + bp->b_dirtyoff; uiop->uio_rw = UIO_WRITE; nfsstats.write_bios++; - if (bp->b_flags & B_APPENDWRITE) - error = nfs_writerpc(vp, uiop, cr, IO_APPEND); + if ((bp->b_flags & (B_ASYNC | B_NEEDCOMMIT | B_NOCACHE)) == B_ASYNC) + iomode = NFSV3WRITE_UNSTABLE; + else + iomode = NFSV3WRITE_FILESYNC; + bp->b_flags |= B_WRITEINPROG; + error = nfs_writerpc(vp, uiop, cr, &iomode, &must_commit); + if (!error && iomode == NFSV3WRITE_UNSTABLE) + bp->b_flags |= B_NEEDCOMMIT; else - error = nfs_writerpc(vp, uiop, cr, 0); - bp->b_flags &= ~(B_WRITEINPROG | B_APPENDWRITE); + bp->b_flags &= ~B_NEEDCOMMIT; + bp->b_flags &= ~B_WRITEINPROG; - /* - * For an interrupted write, the buffer is still valid and the - * write hasn't been pushed to the server yet, so we can't set - * B_ERROR and report the interruption by setting B_EINTR. For - * the B_ASYNC case, B_EINTR is not relevant, so the rpc attempt - * is essentially a noop. - */ - if (error == EINTR) { + /* + * For an interrupted write, the buffer is still valid + * and the write hasn't been pushed to the server yet, + * so we can't set B_ERROR and report the interruption + * by setting B_EINTR. For the B_ASYNC case, B_EINTR + * is not relevant, so the rpc attempt is essentially + * a noop. For the case of a V3 write rpc not being + * committed to stable storage, the block is still + * dirty and requires either a commit rpc or another + * write rpc with iomode == NFSV3WRITE_FILESYNC before + * the block is reused. This is indicated by setting + * the B_DELWRI and B_NEEDCOMMIT flags. + */ + if (error == EINTR + || (!error && (bp->b_flags & B_NEEDCOMMIT))) { bp->b_flags &= ~(B_INVAL|B_NOCACHE); bp->b_flags |= B_DELWRI; @@ -874,6 +942,8 @@ nfs_doio(bp, cr, p) } } bp->b_resid = uiop->uio_resid; + if (must_commit) + nfs_clearcommit(vp->v_mount); biodone(bp); return (error); } diff --git a/sys/nfs/nfs_common.c b/sys/nfs/nfs_common.c index 1653c8b..350ba46 100644 --- a/sys/nfs/nfs_common.c +++ b/sys/nfs/nfs_common.c @@ -34,7 +34,7 @@ * SUCH DAMAGE. * * @(#)nfs_subs.c 8.3 (Berkeley) 1/4/94 - * $Id: nfs_subs.c,v 1.15 1995/05/30 08:12:43 rgrimes Exp $ + * $Id: nfs_subs.c,v 1.16 1995/06/14 06:23:38 joerg Exp $ */ /* @@ -52,6 +52,7 @@ #include <sys/mbuf.h> #include <sys/socket.h> #include <sys/stat.h> +#include <sys/malloc.h> #ifdef VFS_LKM #include <sys/sysent.h> #include <sys/syscall.h> @@ -60,7 +61,7 @@ #include <vm/vm.h> #include <nfs/rpcv2.h> -#include <nfs/nfsv2.h> +#include <nfs/nfsproto.h> #include <nfs/nfsnode.h> #include <nfs/nfs.h> #include <nfs/xdr_subs.h> @@ -78,31 +79,455 @@ #include <netiso/iso.h> #endif -#define TRUE 1 -#define FALSE 0 - /* * Data items converted to xdr at startup, since they are constant * This is kinda hokey, but may save a little time doing byte swaps */ -u_long nfs_procids[NFS_NPROCS]; u_long nfs_xdrneg1; u_long rpc_call, rpc_vers, rpc_reply, rpc_msgdenied, rpc_autherr, - rpc_mismatch, rpc_auth_unix, rpc_msgaccepted, rpc_rejectedcred, + rpc_mismatch, rpc_auth_unix, rpc_msgaccepted, rpc_auth_kerb; -u_long nfs_vers, nfs_prog, nfs_true, nfs_false; +u_long nfs_prog, nqnfs_prog, nfs_true, nfs_false; /* And other global data */ static u_long nfs_xid = 0; -enum vtype ntov_type[7] = { VNON, VREG, VDIR, VBLK, VCHR, VLNK, VNON }; +enum vtype nv2tov_type[8] = { VNON, VREG, VDIR, VBLK, VCHR, VLNK, VNON, VNON }; +enum vtype nv3tov_type[8]={ VNON, VREG, VDIR, VBLK, VCHR, VLNK, VSOCK, VFIFO }; +int nfs_ticks; + +/* + * Mapping of old NFS Version 2 RPC numbers to generic numbers. + */ +int nfsv3_procid[NFS_NPROCS] = { + NFSPROC_NULL, + NFSPROC_GETATTR, + NFSPROC_SETATTR, + NFSPROC_NOOP, + NFSPROC_LOOKUP, + NFSPROC_READLINK, + NFSPROC_READ, + NFSPROC_NOOP, + NFSPROC_WRITE, + NFSPROC_CREATE, + NFSPROC_REMOVE, + NFSPROC_RENAME, + NFSPROC_LINK, + NFSPROC_SYMLINK, + NFSPROC_MKDIR, + NFSPROC_RMDIR, + NFSPROC_READDIR, + NFSPROC_FSSTAT, + NFSPROC_NOOP, + NFSPROC_NOOP, + NFSPROC_NOOP, + NFSPROC_NOOP, + NFSPROC_NOOP, + NFSPROC_NOOP, + NFSPROC_NOOP, + NFSPROC_NOOP +}; + +/* + * and the reverse mapping from generic to Version 2 procedure numbers + */ +int nfsv2_procid[NFS_NPROCS] = { + NFSV2PROC_NULL, + NFSV2PROC_GETATTR, + NFSV2PROC_SETATTR, + NFSV2PROC_LOOKUP, + NFSV2PROC_NOOP, + NFSV2PROC_READLINK, + NFSV2PROC_READ, + NFSV2PROC_WRITE, + NFSV2PROC_CREATE, + NFSV2PROC_MKDIR, + NFSV2PROC_SYMLINK, + NFSV2PROC_CREATE, + NFSV2PROC_REMOVE, + NFSV2PROC_RMDIR, + NFSV2PROC_RENAME, + NFSV2PROC_LINK, + NFSV2PROC_READDIR, + NFSV2PROC_NOOP, + NFSV2PROC_STATFS, + NFSV2PROC_NOOP, + NFSV2PROC_NOOP, + NFSV2PROC_NOOP, + NFSV2PROC_NOOP, + NFSV2PROC_NOOP, + NFSV2PROC_NOOP, + NFSV2PROC_NOOP, +}; + +/* + * Maps errno values to nfs error numbers. + * Use NFSERR_IO as the catch all for ones not specifically defined in + * RFC 1094. + */ +static u_char nfsrv_v2errmap[ELAST] = { + NFSERR_PERM, NFSERR_NOENT, NFSERR_IO, NFSERR_IO, NFSERR_IO, + NFSERR_NXIO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, + NFSERR_IO, NFSERR_IO, NFSERR_ACCES, NFSERR_IO, NFSERR_IO, + NFSERR_IO, NFSERR_EXIST, NFSERR_IO, NFSERR_NODEV, NFSERR_NOTDIR, + NFSERR_ISDIR, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, + NFSERR_IO, NFSERR_FBIG, NFSERR_NOSPC, NFSERR_IO, NFSERR_ROFS, + NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, + NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, + NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, + NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, + NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, + NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, + NFSERR_IO, NFSERR_IO, NFSERR_NAMETOL, NFSERR_IO, NFSERR_IO, + NFSERR_NOTEMPTY, NFSERR_IO, NFSERR_IO, NFSERR_DQUOT, NFSERR_STALE, + NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, + NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, + NFSERR_IO, +}; + +/* + * Maps errno values to nfs error numbers. + * Although it is not obvious whether or not NFS clients really care if + * a returned error value is in the specified list for the procedure, the + * safest thing to do is filter them appropriately. For Version 2, the + * X/Open XNFS document is the only specification that defines error values + * for each RPC (The RFC simply lists all possible error values for all RPCs), + * so I have decided to not do this for Version 2. + * The first entry is the default error return and the rest are the valid + * errors for that RPC in increasing numeric order. + */ +static short nfsv3err_null[] = { + 0, + 0, +}; + +static short nfsv3err_getattr[] = { + NFSERR_IO, + NFSERR_IO, + NFSERR_STALE, + NFSERR_BADHANDLE, + NFSERR_SERVERFAULT, + 0, +}; + +static short nfsv3err_setattr[] = { + NFSERR_IO, + NFSERR_PERM, + NFSERR_IO, + NFSERR_ACCES, + NFSERR_INVAL, + NFSERR_NOSPC, + NFSERR_ROFS, + NFSERR_DQUOT, + NFSERR_STALE, + NFSERR_BADHANDLE, + NFSERR_NOT_SYNC, + NFSERR_SERVERFAULT, + 0, +}; + +static short nfsv3err_lookup[] = { + NFSERR_IO, + NFSERR_NOENT, + NFSERR_IO, + NFSERR_ACCES, + NFSERR_NOTDIR, + NFSERR_NAMETOL, + NFSERR_STALE, + NFSERR_BADHANDLE, + NFSERR_SERVERFAULT, + 0, +}; + +static short nfsv3err_access[] = { + NFSERR_IO, + NFSERR_IO, + NFSERR_STALE, + NFSERR_BADHANDLE, + NFSERR_SERVERFAULT, + 0, +}; + +static short nfsv3err_readlink[] = { + NFSERR_IO, + NFSERR_IO, + NFSERR_ACCES, + NFSERR_INVAL, + NFSERR_STALE, + NFSERR_BADHANDLE, + NFSERR_NOTSUPP, + NFSERR_SERVERFAULT, + 0, +}; + +static short nfsv3err_read[] = { + NFSERR_IO, + NFSERR_IO, + NFSERR_NXIO, + NFSERR_ACCES, + NFSERR_INVAL, + NFSERR_STALE, + NFSERR_BADHANDLE, + NFSERR_SERVERFAULT, + 0, +}; + +static short nfsv3err_write[] = { + NFSERR_IO, + NFSERR_IO, + NFSERR_ACCES, + NFSERR_INVAL, + NFSERR_FBIG, + NFSERR_NOSPC, + NFSERR_ROFS, + NFSERR_DQUOT, + NFSERR_STALE, + NFSERR_BADHANDLE, + NFSERR_SERVERFAULT, + 0, +}; + +static short nfsv3err_create[] = { + NFSERR_IO, + NFSERR_IO, + NFSERR_ACCES, + NFSERR_EXIST, + NFSERR_NOTDIR, + NFSERR_NOSPC, + NFSERR_ROFS, + NFSERR_NAMETOL, + NFSERR_DQUOT, + NFSERR_STALE, + NFSERR_BADHANDLE, + NFSERR_NOTSUPP, + NFSERR_SERVERFAULT, + 0, +}; + +static short nfsv3err_mkdir[] = { + NFSERR_IO, + NFSERR_IO, + NFSERR_ACCES, + NFSERR_EXIST, + NFSERR_NOTDIR, + NFSERR_NOSPC, + NFSERR_ROFS, + NFSERR_NAMETOL, + NFSERR_DQUOT, + NFSERR_STALE, + NFSERR_BADHANDLE, + NFSERR_NOTSUPP, + NFSERR_SERVERFAULT, + 0, +}; + +static short nfsv3err_symlink[] = { + NFSERR_IO, + NFSERR_IO, + NFSERR_ACCES, + NFSERR_EXIST, + NFSERR_NOTDIR, + NFSERR_NOSPC, + NFSERR_ROFS, + NFSERR_NAMETOL, + NFSERR_DQUOT, + NFSERR_STALE, + NFSERR_BADHANDLE, + NFSERR_NOTSUPP, + NFSERR_SERVERFAULT, + 0, +}; + +static short nfsv3err_mknod[] = { + NFSERR_IO, + NFSERR_IO, + NFSERR_ACCES, + NFSERR_EXIST, + NFSERR_NOTDIR, + NFSERR_NOSPC, + NFSERR_ROFS, + NFSERR_NAMETOL, + NFSERR_DQUOT, + NFSERR_STALE, + NFSERR_BADHANDLE, + NFSERR_NOTSUPP, + NFSERR_SERVERFAULT, + NFSERR_BADTYPE, + 0, +}; + +static short nfsv3err_remove[] = { + NFSERR_IO, + NFSERR_NOENT, + NFSERR_IO, + NFSERR_ACCES, + NFSERR_NOTDIR, + NFSERR_ROFS, + NFSERR_NAMETOL, + NFSERR_STALE, + NFSERR_BADHANDLE, + NFSERR_SERVERFAULT, + 0, +}; + +static short nfsv3err_rmdir[] = { + NFSERR_IO, + NFSERR_NOENT, + NFSERR_IO, + NFSERR_ACCES, + NFSERR_EXIST, + NFSERR_NOTDIR, + NFSERR_INVAL, + NFSERR_ROFS, + NFSERR_NAMETOL, + NFSERR_NOTEMPTY, + NFSERR_STALE, + NFSERR_BADHANDLE, + NFSERR_NOTSUPP, + NFSERR_SERVERFAULT, + 0, +}; + +static short nfsv3err_rename[] = { + NFSERR_IO, + NFSERR_NOENT, + NFSERR_IO, + NFSERR_ACCES, + NFSERR_EXIST, + NFSERR_XDEV, + NFSERR_NOTDIR, + NFSERR_ISDIR, + NFSERR_INVAL, + NFSERR_NOSPC, + NFSERR_ROFS, + NFSERR_MLINK, + NFSERR_NAMETOL, + NFSERR_NOTEMPTY, + NFSERR_DQUOT, + NFSERR_STALE, + NFSERR_BADHANDLE, + NFSERR_NOTSUPP, + NFSERR_SERVERFAULT, + 0, +}; + +static short nfsv3err_link[] = { + NFSERR_IO, + NFSERR_IO, + NFSERR_ACCES, + NFSERR_EXIST, + NFSERR_XDEV, + NFSERR_NOTDIR, + NFSERR_INVAL, + NFSERR_NOSPC, + NFSERR_ROFS, + NFSERR_MLINK, + NFSERR_NAMETOL, + NFSERR_DQUOT, + NFSERR_STALE, + NFSERR_BADHANDLE, + NFSERR_NOTSUPP, + NFSERR_SERVERFAULT, + 0, +}; + +static short nfsv3err_readdir[] = { + NFSERR_IO, + NFSERR_IO, + NFSERR_ACCES, + NFSERR_NOTDIR, + NFSERR_STALE, + NFSERR_BADHANDLE, + NFSERR_BAD_COOKIE, + NFSERR_TOOSMALL, + NFSERR_SERVERFAULT, + 0, +}; + +static short nfsv3err_readdirplus[] = { + NFSERR_IO, + NFSERR_IO, + NFSERR_ACCES, + NFSERR_NOTDIR, + NFSERR_STALE, + NFSERR_BADHANDLE, + NFSERR_BAD_COOKIE, + NFSERR_NOTSUPP, + NFSERR_TOOSMALL, + NFSERR_SERVERFAULT, + 0, +}; + +static short nfsv3err_fsstat[] = { + NFSERR_IO, + NFSERR_IO, + NFSERR_STALE, + NFSERR_BADHANDLE, + NFSERR_SERVERFAULT, + 0, +}; + +static short nfsv3err_fsinfo[] = { + NFSERR_STALE, + NFSERR_STALE, + NFSERR_BADHANDLE, + NFSERR_SERVERFAULT, + 0, +}; + +static short nfsv3err_pathconf[] = { + NFSERR_STALE, + NFSERR_STALE, + NFSERR_BADHANDLE, + NFSERR_SERVERFAULT, + 0, +}; + +static short nfsv3err_commit[] = { + NFSERR_IO, + NFSERR_IO, + NFSERR_STALE, + NFSERR_BADHANDLE, + NFSERR_SERVERFAULT, + 0, +}; + +static short *nfsrv_v3errmap[] = { + nfsv3err_null, + nfsv3err_getattr, + nfsv3err_setattr, + nfsv3err_lookup, + nfsv3err_access, + nfsv3err_readlink, + nfsv3err_read, + nfsv3err_write, + nfsv3err_create, + nfsv3err_mkdir, + nfsv3err_symlink, + nfsv3err_mknod, + nfsv3err_remove, + nfsv3err_rmdir, + nfsv3err_rename, + nfsv3err_link, + nfsv3err_readdir, + nfsv3err_readdirplus, + nfsv3err_fsstat, + nfsv3err_fsinfo, + nfsv3err_pathconf, + nfsv3err_commit, +}; + extern struct proc *nfs_iodwant[NFS_MAXASYNCDAEMON]; -extern int nqnfs_piggy[NFS_NPROCS]; extern struct nfsrtt nfsrtt; extern time_t nqnfsstarttime; -extern u_long nqnfs_prog, nqnfs_vers; extern int nqsrv_clockskew; extern int nqsrv_writeslack; extern int nqsrv_maxlease; +extern struct nfsstats nfsstats; +extern int nqnfs_piggy[NFS_NPROCS]; +extern nfstype nfsv2_type[9]; +extern nfstype nfsv3_type[9]; +extern struct nfsnodehashhead *nfsnodehashtbl; +extern u_long nfsnodehash; #ifdef VFS_LKM struct getfh_args; @@ -167,14 +592,16 @@ nfsm_reqh(vp, procid, hsiz, bposp) * Returns the head of the mbuf list. */ struct mbuf * -nfsm_rpchead(cr, nqnfs, procid, auth_type, auth_len, auth_str, mrest, - mrest_len, mbp, xidp) +nfsm_rpchead(cr, nmflag, procid, auth_type, auth_len, auth_str, verf_len, + verf_str, mrest, mrest_len, mbp, xidp) register struct ucred *cr; - int nqnfs; + int nmflag; int procid; int auth_type; int auth_len; char *auth_str; + int verf_len; + char *verf_str; struct mbuf *mrest; int mrest_len; struct mbuf **mbp; @@ -188,15 +615,13 @@ nfsm_rpchead(cr, nqnfs, procid, auth_type, auth_len, auth_str, mrest, int siz, grpsiz, authsiz; authsiz = nfsm_rndup(auth_len); - if (auth_type == RPCAUTH_NQNFS) - authsiz += 2 * NFSX_UNSIGNED; MGETHDR(mb, M_WAIT, MT_DATA); - if ((authsiz + 10*NFSX_UNSIGNED) >= MINCLSIZE) { + if ((authsiz + 10 * NFSX_UNSIGNED) >= MINCLSIZE) { MCLGET(mb, M_WAIT); - } else if ((authsiz + 10*NFSX_UNSIGNED) < MHLEN) { - MH_ALIGN(mb, authsiz + 10*NFSX_UNSIGNED); + } else if ((authsiz + 10 * NFSX_UNSIGNED) < MHLEN) { + MH_ALIGN(mb, authsiz + 10 * NFSX_UNSIGNED); } else { - MH_ALIGN(mb, 8*NFSX_UNSIGNED); + MH_ALIGN(mb, 8 * NFSX_UNSIGNED); } mb->m_len = 0; mreq = mb; @@ -205,20 +630,26 @@ nfsm_rpchead(cr, nqnfs, procid, auth_type, auth_len, auth_str, mrest, /* * First the RPC header. */ - nfsm_build(tl, u_long *, 8*NFSX_UNSIGNED); + nfsm_build(tl, u_long *, 8 * NFSX_UNSIGNED); if (++nfs_xid == 0) nfs_xid++; *tl++ = *xidp = txdr_unsigned(nfs_xid); *tl++ = rpc_call; *tl++ = rpc_vers; - if (nqnfs) { + if (nmflag & NFSMNT_NQNFS) { *tl++ = txdr_unsigned(NQNFS_PROG); - *tl++ = txdr_unsigned(NQNFS_VER1); + *tl++ = txdr_unsigned(NQNFS_VER3); } else { *tl++ = txdr_unsigned(NFS_PROG); - *tl++ = txdr_unsigned(NFS_VER2); + if (nmflag & NFSMNT_NFSV3) + *tl++ = txdr_unsigned(NFS_VER3); + else + *tl++ = txdr_unsigned(NFS_VER2); } - *tl++ = txdr_unsigned(procid); + if (nmflag & NFSMNT_NFSV3) + *tl++ = txdr_unsigned(procid); + else + *tl++ = txdr_unsigned(nfsv2_procid[procid]); /* * And then the authorization cred. @@ -237,10 +668,7 @@ nfsm_rpchead(cr, nqnfs, procid, auth_type, auth_len, auth_str, mrest, for (i = 1; i <= grpsiz; i++) *tl++ = txdr_unsigned(cr->cr_groups[i]); break; - case RPCAUTH_NQNFS: - nfsm_build(tl, u_long *, 2*NFSX_UNSIGNED); - *tl++ = txdr_unsigned(cr->cr_uid); - *tl = txdr_unsigned(auth_len); + case RPCAUTH_KERB4: siz = auth_len; while (siz > 0) { if (M_TRAILINGSPACE(mb) == 0) { @@ -266,11 +694,43 @@ nfsm_rpchead(cr, nqnfs, procid, auth_type, auth_len, auth_str, mrest, } break; }; - nfsm_build(tl, u_long *, 2*NFSX_UNSIGNED); - *tl++ = txdr_unsigned(RPCAUTH_NULL); - *tl = 0; + + /* + * And the verifier... + */ + nfsm_build(tl, u_long *, 2 * NFSX_UNSIGNED); + if (verf_str) { + *tl++ = txdr_unsigned(RPCAUTH_KERB4); + *tl = txdr_unsigned(verf_len); + siz = verf_len; + while (siz > 0) { + if (M_TRAILINGSPACE(mb) == 0) { + MGET(mb2, M_WAIT, MT_DATA); + if (siz >= MINCLSIZE) + MCLGET(mb2, M_WAIT); + mb->m_next = mb2; + mb = mb2; + mb->m_len = 0; + bpos = mtod(mb, caddr_t); + } + i = min(siz, M_TRAILINGSPACE(mb)); + bcopy(verf_str, bpos, i); + mb->m_len += i; + verf_str += i; + bpos += i; + siz -= i; + } + if ((siz = (nfsm_rndup(verf_len) - verf_len)) > 0) { + for (i = 0; i < siz; i++) + *bpos++ = '\0'; + mb->m_len += siz; + } + } else { + *tl++ = txdr_unsigned(RPCAUTH_NULL); + *tl = 0; + } mb->m_next = mrest; - mreq->m_pkthdr.len = authsiz + 10*NFSX_UNSIGNED + mrest_len; + mreq->m_pkthdr.len = authsiz + 10 * NFSX_UNSIGNED + mrest_len; mreq->m_pkthdr.rcvif = (struct ifnet *)0; *mbp = mb; return (mreq); @@ -603,6 +1063,25 @@ nfs_init() { register int i; + /* + * Check to see if major data structures haven't bloated. + */ + if (sizeof (struct nfsnode) > NFS_NODEALLOC) { + printf("struct nfsnode bloated (> %dbytes)\n", NFS_NODEALLOC); + printf("Try reducing NFS_SMALLFH\n"); + } + if (sizeof (struct nfsmount) > NFS_MNTALLOC) { + printf("struct nfsmount bloated (> %dbytes)\n", NFS_MNTALLOC); + printf("Try reducing NFS_MUIDHASHSIZ\n"); + } + if (sizeof (struct nfssvc_sock) > NFS_SVCALLOC) { + printf("struct nfssvc_sock bloated (> %dbytes)\n",NFS_SVCALLOC); + printf("Try reducing NFS_UIDHASHSIZ\n"); + } + if (sizeof (struct nfsuid) > NFS_UIDALLOC) { + printf("struct nfsuid bloated (> %dbytes)\n",NFS_UIDALLOC); + printf("Try unionizing the nu_nickname and nu_flag fields\n"); + } nfsrtt.pos = 0; rpc_vers = txdr_unsigned(RPC_VER2); rpc_call = txdr_unsigned(RPC_CALL); @@ -611,17 +1090,16 @@ nfs_init() rpc_msgaccepted = txdr_unsigned(RPC_MSGACCEPTED); rpc_mismatch = txdr_unsigned(RPC_MISMATCH); rpc_autherr = txdr_unsigned(RPC_AUTHERR); - rpc_rejectedcred = txdr_unsigned(AUTH_REJECTCRED); rpc_auth_unix = txdr_unsigned(RPCAUTH_UNIX); - rpc_auth_kerb = txdr_unsigned(RPCAUTH_NQNFS); - nfs_vers = txdr_unsigned(NFS_VER2); + rpc_auth_kerb = txdr_unsigned(RPCAUTH_KERB4); nfs_prog = txdr_unsigned(NFS_PROG); + nqnfs_prog = txdr_unsigned(NQNFS_PROG); nfs_true = txdr_unsigned(TRUE); nfs_false = txdr_unsigned(FALSE); nfs_xdrneg1 = txdr_unsigned(-1); - /* Loop thru nfs procids */ - for (i = 0; i < NFS_NPROCS; i++) - nfs_procids[i] = txdr_unsigned(i); + nfs_ticks = (hz * NFS_TICKINTVL + 500) / 1000; + if (nfs_ticks < 1) + nfs_ticks = 1; /* Ensure async daemons disabled */ for (i = 0; i < NFS_MAXASYNCDAEMON; i++) nfs_iodwant[i] = (struct proc *)0; @@ -637,8 +1115,6 @@ nfs_init() nqnfsstarttime = boottime.tv_sec + nqsrv_maxlease + nqsrv_clockskew + nqsrv_writeslack; NQLOADNOVRAM(nqnfsstarttime); - nqnfs_prog = txdr_unsigned(NQNFS_PROG); - nqnfs_vers = txdr_unsigned(NQNFS_VER1); CIRCLEQ_INIT(&nqtimerhead); nqfhhashtbl = hashinit(NQLCHSZ, M_NQLEASE, &nqfhhash); } @@ -649,6 +1125,7 @@ nfs_init() TAILQ_INIT(&nfs_reqq); nfs_timer(0); +#ifdef __FreeBSD__ /* * Set up lease_check and lease_updatetime so that other parts * of the system can call us, if we are loadable. @@ -662,6 +1139,7 @@ nfs_init() sysent[SYS_getfh].sy_narg = 2; sysent[SYS_getfh].sy_call = getfh; #endif +#endif return (0); } @@ -689,57 +1167,65 @@ nfs_loadattrcache(vpp, mdp, dposp, vaper) { register struct vnode *vp = *vpp; register struct vattr *vap; - register struct nfsv2_fattr *fp; + register struct nfs_fattr *fp; register struct nfsnode *np; register struct nfsnodehashhead *nhpp; register long t1; - caddr_t dpos, cp2; - int error = 0, isnq; + caddr_t cp2; + int error = 0, rdev; struct mbuf *md; enum vtype vtyp; u_short vmode; - long rdev; struct timespec mtime; struct vnode *nvp; + quad_t tval; + int v3 = NFS_ISV3(vp); md = *mdp; - dpos = *dposp; - t1 = (mtod(md, caddr_t) + md->m_len) - dpos; - isnq = (VFSTONFS(vp->v_mount)->nm_flag & NFSMNT_NQNFS); - error = nfsm_disct(&md, &dpos, NFSX_FATTR(isnq), t1, &cp2); - if (error) + t1 = (mtod(md, caddr_t) + md->m_len) - *dposp; + if (error = nfsm_disct(mdp, dposp, NFSX_FATTR(v3), t1, &cp2)) return (error); - fp = (struct nfsv2_fattr *)cp2; - vtyp = nfstov_type(fp->fa_type); - vmode = fxdr_unsigned(u_short, fp->fa_mode); - /* - * XXX - * - * The duplicate information returned in fa_type and fa_mode - * is an ambiguity in the NFS version 2 protocol. - * - * VREG should be taken literally as a regular file. If a - * server intents to return some type information differently - * in the upper bits of the mode field (e.g. for sockets, or - * FIFOs), NFSv2 mandates fa_type to be VNON. Anyway, we - * leave the examination of the mode bits even in the VREG - * case to avoid breakage for bogus servers, but we make sure - * that there are actually type bits set in the upper part of - * fa_mode (and failing that, trust the va_type field). - * - * NFSv3 cleared the issue, and requires fa_mode to not - * contain any type information (while also introduing sockets - * and FIFOs for fa_type). - */ - if (vtyp == VNON || (vtyp == VREG && (vmode & S_IFMT) != 0)) - vtyp = IFTOVT(vmode); - if (isnq) { - rdev = fxdr_unsigned(long, fp->fa_nqrdev); - fxdr_nqtime(&fp->fa_nqmtime, &mtime); + fp = (struct nfs_fattr *)cp2; + if (v3) { + vtyp = nfsv3tov_type(fp->fa_type); + vmode = fxdr_unsigned(u_short, fp->fa_mode); + rdev = makedev(fxdr_unsigned(u_char, fp->fa3_rdev.specdata1), + fxdr_unsigned(u_char, fp->fa3_rdev.specdata2)); + fxdr_nfsv3time(&fp->fa3_mtime, &mtime); } else { - rdev = fxdr_unsigned(long, fp->fa_nfsrdev); - fxdr_nfstime(&fp->fa_nfsmtime, &mtime); + vtyp = nfsv2tov_type(fp->fa_type); + vmode = fxdr_unsigned(u_short, fp->fa_mode); + /* + * XXX + * + * The duplicate information returned in fa_type and fa_mode + * is an ambiguity in the NFS version 2 protocol. + * + * VREG should be taken literally as a regular file. If a + * server intents to return some type information differently + * in the upper bits of the mode field (e.g. for sockets, or + * FIFOs), NFSv2 mandates fa_type to be VNON. Anyway, we + * leave the examination of the mode bits even in the VREG + * case to avoid breakage for bogus servers, but we make sure + * that there are actually type bits set in the upper part of + * fa_mode (and failing that, trust the va_type field). + * + * NFSv3 cleared the issue, and requires fa_mode to not + * contain any type information (while also introduing sockets + * and FIFOs for fa_type). + */ + if (vtyp == VNON || (vtyp == VREG && (vmode & S_IFMT) != 0)) + vtyp = IFTOVT(vmode); + rdev = fxdr_unsigned(long, fp->fa2_rdev); + fxdr_nfsv2time(&fp->fa2_mtime, &mtime); + + /* + * Really ugly NFSv2 kludge. + */ + if (vtyp == VCHR && rdev == 0xffffffff) + vtyp = VFIFO; } + /* * If v_type == VNON it is a new node, so fill in the v_type, * n_mtime fields. Check to see if it represents a special @@ -749,10 +1235,15 @@ nfs_loadattrcache(vpp, mdp, dposp, vaper) */ np = VTONFS(vp); if (vp->v_type == VNON) { - if (vtyp == VCHR && rdev == 0xffffffff) - vp->v_type = vtyp = VFIFO; - else - vp->v_type = vtyp; + /* + * If we had a lock and it turns out that the vnode + * is an object which we don't want to lock (e.g. VDIR) + * to avoid nasty hanging problems on a server crash, + * then release it here. + */ + if (vtyp != VREG && VOP_ISLOCKED(vp)) + VOP_UNLOCK(vp); + vp->v_type = vtyp; if (vp->v_type == VFIFO) { vp->v_op = fifo_nfsv2nodeop_p; } @@ -773,7 +1264,7 @@ nfs_loadattrcache(vpp, mdp, dposp, vaper) * Reinitialize aliased node. */ np->n_vnode = nvp; - nhpp = nfs_hash(&np->n_fh); + nhpp = NFSNOHASH(nfs_hash(np->n_fhp, np->n_fhsize)); LIST_INSERT_HEAD(nhpp, np, n_hash); *vpp = vp = nvp; } @@ -783,31 +1274,34 @@ nfs_loadattrcache(vpp, mdp, dposp, vaper) vap = &np->n_vattr; vap->va_type = vtyp; vap->va_mode = (vmode & 07777); - vap->va_nlink = fxdr_unsigned(u_short, fp->fa_nlink); - vap->va_uid = fxdr_unsigned(uid_t, fp->fa_uid); - vap->va_gid = fxdr_unsigned(gid_t, fp->fa_gid); vap->va_rdev = (dev_t)rdev; vap->va_mtime = mtime; vap->va_fsid = vp->v_mount->mnt_stat.f_fsid.val[0]; - if (isnq) { - fxdr_hyper(&fp->fa_nqsize, &vap->va_size); - vap->va_blocksize = fxdr_unsigned(long, fp->fa_nqblocksize); - fxdr_hyper(&fp->fa_nqbytes, &vap->va_bytes); - vap->va_fileid = fxdr_unsigned(long, fp->fa_nqfileid); - fxdr_nqtime(&fp->fa_nqatime, &vap->va_atime); - vap->va_flags = fxdr_unsigned(u_long, fp->fa_nqflags); - fxdr_nqtime(&fp->fa_nqctime, &vap->va_ctime); - vap->va_gen = fxdr_unsigned(u_long, fp->fa_nqgen); - fxdr_hyper(&fp->fa_nqfilerev, &vap->va_filerev); + if (v3) { + vap->va_nlink = fxdr_unsigned(u_short, fp->fa_nlink); + vap->va_uid = fxdr_unsigned(uid_t, fp->fa_uid); + vap->va_gid = fxdr_unsigned(gid_t, fp->fa_gid); + fxdr_hyper(&fp->fa3_size, &vap->va_size); + vap->va_blocksize = NFS_FABLKSIZE; + fxdr_hyper(&fp->fa3_used, &vap->va_bytes); + vap->va_fileid = fxdr_unsigned(int, fp->fa3_fileid.nfsuquad[1]); + fxdr_nfsv3time(&fp->fa3_atime, &vap->va_atime); + fxdr_nfsv3time(&fp->fa3_ctime, &vap->va_ctime); + vap->va_flags = 0; + vap->va_filerev = 0; } else { - vap->va_size = fxdr_unsigned(u_long, fp->fa_nfssize); - vap->va_blocksize = fxdr_unsigned(long, fp->fa_nfsblocksize); - vap->va_bytes = fxdr_unsigned(long, fp->fa_nfsblocks) * NFS_FABLKSIZE; - vap->va_fileid = fxdr_unsigned(long, fp->fa_nfsfileid); - fxdr_nfstime(&fp->fa_nfsatime, &vap->va_atime); + vap->va_nlink = fxdr_unsigned(u_short, fp->fa_nlink); + vap->va_uid = fxdr_unsigned(uid_t, fp->fa_uid); + vap->va_gid = fxdr_unsigned(gid_t, fp->fa_gid); + vap->va_size = fxdr_unsigned(u_long, fp->fa2_size); + vap->va_blocksize = fxdr_unsigned(long, fp->fa2_blocksize); + vap->va_bytes = fxdr_unsigned(long, fp->fa2_blocks) * NFS_FABLKSIZE; + vap->va_fileid = fxdr_unsigned(long, fp->fa2_fileid); + fxdr_nfsv2time(&fp->fa2_atime, &vap->va_atime); vap->va_flags = 0; - fxdr_nfstime(&fp->fa_nfsctime, &vap->va_ctime); - vap->va_gen = 0; + vap->va_ctime.ts_sec = fxdr_unsigned(long, fp->fa2_ctime.nfsv2_sec); + vap->va_ctime.ts_nsec = 0; + vap->va_gen = fxdr_unsigned(u_long, fp->fa2_ctime.nfsv2_usec); vap->va_filerev = 0; } if (vap->va_size != np->n_size) { @@ -824,26 +1318,13 @@ nfs_loadattrcache(vpp, mdp, dposp, vaper) np->n_size = vap->va_size; } np->n_attrstamp = time.tv_sec; - *dposp = dpos; - *mdp = md; if (vaper != NULL) { bcopy((caddr_t)vap, (caddr_t)vaper, sizeof(*vap)); -#ifdef notdef - if ((np->n_flag & NMODIFIED) && np->n_size > vap->va_size) - if (np->n_size > vap->va_size) - vaper->va_size = np->n_size; -#endif if (np->n_flag & NCHG) { - if (np->n_flag & NACC) { - vaper->va_atime.ts_sec = np->n_atim.tv_sec; - vaper->va_atime.ts_nsec = - np->n_atim.tv_usec * 1000; - } - if (np->n_flag & NUPD) { - vaper->va_mtime.ts_sec = np->n_mtim.tv_sec; - vaper->va_mtime.ts_nsec = - np->n_mtim.tv_usec * 1000; - } + if (np->n_flag & NACC) + vaper->va_atime = np->n_atim; + if (np->n_flag & NUPD) + vaper->va_mtime = np->n_mtim; } } return (0); @@ -862,12 +1343,7 @@ nfs_getattrcache(vp, vaper) register struct nfsnode *np = VTONFS(vp); register struct vattr *vap; - if (VFSTONFS(vp->v_mount)->nm_flag & NFSMNT_NQLOOKLEASE) { - if (!NQNFS_CKCACHABLE(vp, NQL_READ) || np->n_attrstamp == 0) { - nfsstats.attrcache_misses++; - return (ENOENT); - } - } else if ((time.tv_sec - np->n_attrstamp) >= NFS_ATTRTIMEO(np)) { + if ((time.tv_sec - np->n_attrstamp) >= NFS_ATTRTIMEO(np)) { nfsstats.attrcache_misses++; return (ENOENT); } @@ -887,23 +1363,11 @@ nfs_getattrcache(vp, vaper) np->n_size = vap->va_size; } bcopy((caddr_t)vap, (caddr_t)vaper, sizeof(struct vattr)); -#ifdef notdef - if ((np->n_flag & NMODIFIED) == 0) { - np->n_size = vaper->va_size; - vnode_pager_setsize(vp, (u_long)np->n_size); - } else if (np->n_size > vaper->va_size) - if (np->n_size > vaper->va_size) - vaper->va_size = np->n_size; -#endif if (np->n_flag & NCHG) { - if (np->n_flag & NACC) { - vaper->va_atime.ts_sec = np->n_atim.tv_sec; - vaper->va_atime.ts_nsec = np->n_atim.tv_usec * 1000; - } - if (np->n_flag & NUPD) { - vaper->va_mtime.ts_sec = np->n_mtim.tv_sec; - vaper->va_mtime.ts_nsec = np->n_mtim.tv_usec * 1000; - } + if (np->n_flag & NACC) + vaper->va_atime = np->n_atim; + if (np->n_flag & NUPD) + vaper->va_mtime = np->n_mtim; } return (0); } @@ -912,7 +1376,7 @@ nfs_getattrcache(vp, vaper) * Set up nameidata for a lookup() call and do it */ int -nfs_namei(ndp, fhp, len, slp, nam, mdp, dposp, p) +nfs_namei(ndp, fhp, len, slp, nam, mdp, dposp, retdirp, p, kerbflag) register struct nameidata *ndp; fhandle_t *fhp; int len; @@ -920,7 +1384,9 @@ nfs_namei(ndp, fhp, len, slp, nam, mdp, dposp, p) struct mbuf *nam; struct mbuf **mdp; caddr_t *dposp; + struct vnode **retdirp; struct proc *p; + int kerbflag; { register int i, rem; register struct mbuf *md; @@ -929,6 +1395,7 @@ nfs_namei(ndp, fhp, len, slp, nam, mdp, dposp, p) int error, rdonly; struct componentname *cnp = &ndp->ni_cnd; + *retdirp = (struct vnode *)0; MALLOC(cnp->cn_pnbuf, char *, len + 1, M_NAMEI, M_WAITOK); /* * Copy the name from the mbuf list to ndp->ni_pnbuf @@ -950,7 +1417,7 @@ nfs_namei(ndp, fhp, len, slp, nam, mdp, dposp, p) rem = md->m_len; } if (*fromcp == '\0' || *fromcp == '/') { - error = EINVAL; + error = EACCES; goto out; } cnp->cn_hash += (unsigned char)*fromcp; @@ -964,26 +1431,24 @@ nfs_namei(ndp, fhp, len, slp, nam, mdp, dposp, p) if (len > 0) { if (rem >= len) *dposp += len; - else { - error = nfs_adv(mdp, dposp, len, rem); - if (error) - goto out; - } + else if (error = nfs_adv(mdp, dposp, len, rem)) + goto out; } ndp->ni_pathlen = tocp - cnp->cn_pnbuf; cnp->cn_nameptr = cnp->cn_pnbuf; /* * Extract and set starting directory. */ - error = nfsrv_fhtovp(fhp, FALSE, &dp, ndp->ni_cnd.cn_cred, slp, - nam, &rdonly); - if (error) + if (error = nfsrv_fhtovp(fhp, FALSE, &dp, ndp->ni_cnd.cn_cred, slp, + nam, &rdonly, kerbflag)) goto out; if (dp->v_type != VDIR) { nfsrv_vrele(dp); error = ENOTDIR; goto out; } + VREF(dp); + *retdirp = dp; ndp->ni_startdir = dp; if (rdonly) cnp->cn_flags |= (NOCROSSMOUNT | RDONLY); @@ -993,8 +1458,7 @@ nfs_namei(ndp, fhp, len, slp, nam, mdp, dposp, p) * And call lookup() to do the real work */ cnp->cn_proc = p; - error = lookup(ndp); - if (error) + if (error = lookup(ndp)) goto out; /* * Check for encountering a symbolic link @@ -1087,6 +1551,109 @@ nfsm_adj(mp, len, nul) } /* + * Make these functions instead of macros, so that the kernel text size + * doesn't get too big... + */ +void +nfsm_srvwcc(nfsd, before_ret, before_vap, after_ret, after_vap, mbp, bposp) + struct nfsrv_descript *nfsd; + int before_ret; + register struct vattr *before_vap; + int after_ret; + struct vattr *after_vap; + struct mbuf **mbp; + char **bposp; +{ + register struct mbuf *mb = *mbp, *mb2; + register char *bpos = *bposp; + register u_long *tl; + + if (before_ret) { + nfsm_build(tl, u_long *, NFSX_UNSIGNED); + *tl = nfs_false; + } else { + nfsm_build(tl, u_long *, 7 * NFSX_UNSIGNED); + *tl++ = nfs_true; + txdr_hyper(&(before_vap->va_size), tl); + tl += 2; + txdr_nfsv3time(&(before_vap->va_mtime), tl); + tl += 2; + txdr_nfsv3time(&(before_vap->va_ctime), tl); + } + *bposp = bpos; + *mbp = mb; + nfsm_srvpostopattr(nfsd, after_ret, after_vap, mbp, bposp); +} + +void +nfsm_srvpostopattr(nfsd, after_ret, after_vap, mbp, bposp) + struct nfsrv_descript *nfsd; + int after_ret; + struct vattr *after_vap; + struct mbuf **mbp; + char **bposp; +{ + register struct mbuf *mb = *mbp, *mb2; + register char *bpos = *bposp; + register u_long *tl; + register struct nfs_fattr *fp; + + if (after_ret) { + nfsm_build(tl, u_long *, NFSX_UNSIGNED); + *tl = nfs_false; + } else { + nfsm_build(tl, u_long *, NFSX_UNSIGNED + NFSX_V3FATTR); + *tl++ = nfs_true; + fp = (struct nfs_fattr *)tl; + nfsm_srvfattr(nfsd, after_vap, fp); + } + *mbp = mb; + *bposp = bpos; +} + +void +nfsm_srvfattr(nfsd, vap, fp) + register struct nfsrv_descript *nfsd; + register struct vattr *vap; + register struct nfs_fattr *fp; +{ + + fp->fa_nlink = txdr_unsigned(vap->va_nlink); + fp->fa_uid = txdr_unsigned(vap->va_uid); + fp->fa_gid = txdr_unsigned(vap->va_gid); + if (nfsd->nd_flag & ND_NFSV3) { + fp->fa_type = vtonfsv3_type(vap->va_type); + fp->fa_mode = vtonfsv3_mode(vap->va_mode); + txdr_hyper(&vap->va_size, &fp->fa3_size); + txdr_hyper(&vap->va_bytes, &fp->fa3_used); + fp->fa3_rdev.specdata1 = txdr_unsigned(major(vap->va_rdev)); + fp->fa3_rdev.specdata2 = txdr_unsigned(minor(vap->va_rdev)); + fp->fa3_fsid.nfsuquad[0] = 0; + fp->fa3_fsid.nfsuquad[1] = txdr_unsigned(vap->va_fsid); + fp->fa3_fileid.nfsuquad[0] = 0; + fp->fa3_fileid.nfsuquad[1] = txdr_unsigned(vap->va_fileid); + txdr_nfsv3time(&vap->va_atime, &fp->fa3_atime); + txdr_nfsv3time(&vap->va_mtime, &fp->fa3_mtime); + txdr_nfsv3time(&vap->va_ctime, &fp->fa3_ctime); + } else { + fp->fa_type = vtonfsv2_type(vap->va_type); + fp->fa_mode = vtonfsv2_mode(vap->va_type, vap->va_mode); + fp->fa2_size = txdr_unsigned(vap->va_size); + fp->fa2_blocksize = txdr_unsigned(vap->va_blocksize); + if (vap->va_type == VFIFO) + fp->fa2_rdev = 0xffffffff; + else + fp->fa2_rdev = txdr_unsigned(vap->va_rdev); + fp->fa2_blocks = txdr_unsigned(vap->va_bytes / NFS_FABLKSIZE); + fp->fa2_fsid = txdr_unsigned(vap->va_fsid); + fp->fa2_fileid = txdr_unsigned(vap->va_fileid); + txdr_nfsv2time(&vap->va_atime, &fp->fa2_atime); + txdr_nfsv2time(&vap->va_mtime, &fp->fa2_mtime); + txdr_nfsv2time(&vap->va_ctime, &fp->fa2_ctime); + } +} + +/* * nfsrv_fhtovp() - convert a fh to a vnode ptr (optionally locked) * - look up fsid in mount list (if not found ret error) * - get vp and export rights by calling VFS_FHTOVP() @@ -1094,7 +1661,7 @@ nfsm_adj(mp, len, nul) * - if not lockflag unlock it with VOP_UNLOCK() */ int -nfsrv_fhtovp(fhp, lockflag, vpp, cred, slp, nam, rdonlyp) +nfsrv_fhtovp(fhp, lockflag, vpp, cred, slp, nam, rdonlyp, kerbflag) fhandle_t *fhp; int lockflag; struct vnode **vpp; @@ -1102,6 +1669,7 @@ nfsrv_fhtovp(fhp, lockflag, vpp, cred, slp, nam, rdonlyp) struct nfssvc_sock *slp; struct mbuf *nam; int *rdonlyp; + int kerbflag; { register struct mount *mp; register struct nfsuid *uidp; @@ -1120,19 +1688,13 @@ nfsrv_fhtovp(fhp, lockflag, vpp, cred, slp, nam, rdonlyp) * Check/setup credentials. */ if (exflags & MNT_EXKERB) { - for (uidp = NUIDHASH(slp, cred->cr_uid)->lh_first; uidp != 0; - uidp = uidp->nu_hash.le_next) { - if (uidp->nu_uid == cred->cr_uid) - break; - } - if (uidp == 0) { + if (!kerbflag) { vput(*vpp); - return (NQNFS_AUTHERR); + return (NFSERR_AUTHERR | AUTH_TOOWEAK); } - cred->cr_uid = uidp->nu_cr.cr_uid; - for (i = 0; i < uidp->nu_cr.cr_ngroups; i++) - cred->cr_groups[i] = uidp->nu_cr.cr_groups[i]; - cred->cr_ngroups = i; + } else if (kerbflag) { + vput(*vpp); + return (NFSERR_AUTHERR | AUTH_TOOWEAK); } else if (cred->cr_uid == 0 || (exflags & MNT_EXPORTANON)) { cred->cr_uid = credanon->cr_uid; for (i = 0; i < credanon->cr_ngroups && i < NGROUPS; i++) @@ -1194,6 +1756,145 @@ netaddr_match(family, haddr, nam) return (0); } +static nfsuint64 nfs_nullcookie = { 0, 0 }; +/* + * This function finds the directory cookie that corresponds to the + * logical byte offset given. + */ +nfsuint64 * +nfs_getcookie(np, off, add) + register struct nfsnode *np; + off_t off; + int add; +{ + register struct nfsdmap *dp, *dp2; + register int pos; + + pos = off / NFS_DIRBLKSIZ; + if (pos == 0) { +#ifdef DIAGNOSTIC + if (add) + panic("nfs getcookie add at 0"); +#endif + return (&nfs_nullcookie); + } + pos--; + dp = np->n_cookies.lh_first; + if (!dp) { + if (add) { + MALLOC(dp, struct nfsdmap *, sizeof (struct nfsdmap), + M_NFSDIROFF, M_WAITOK); + dp->ndm_eocookie = 0; + LIST_INSERT_HEAD(&np->n_cookies, dp, ndm_list); + } else + return ((nfsuint64 *)0); + } + while (pos >= NFSNUMCOOKIES) { + pos -= NFSNUMCOOKIES; + if (dp->ndm_list.le_next) { + if (!add && dp->ndm_eocookie < NFSNUMCOOKIES && + pos >= dp->ndm_eocookie) + return ((nfsuint64 *)0); + dp = dp->ndm_list.le_next; + } else if (add) { + MALLOC(dp2, struct nfsdmap *, sizeof (struct nfsdmap), + M_NFSDIROFF, M_WAITOK); + dp2->ndm_eocookie = 0; + LIST_INSERT_AFTER(dp, dp2, ndm_list); + dp = dp2; + } else + return ((nfsuint64 *)0); + } + if (pos >= dp->ndm_eocookie) { + if (add) + dp->ndm_eocookie = pos + 1; + else + return ((nfsuint64 *)0); + } + return (&dp->ndm_cookies[pos]); +} + +/* + * Invalidate cached directory information, except for the actual directory + * blocks (which are invalidated separately). + * Done mainly to avoid the use of stale offset cookies. + */ +void +nfs_invaldir(vp) + register struct vnode *vp; +{ + register struct nfsnode *np = VTONFS(vp); + +#ifdef DIAGNOSTIC + if (vp->v_type != VDIR) + panic("nfs: invaldir not dir"); +#endif + np->n_direofoffset = 0; + np->n_cookieverf.nfsuquad[0] = 0; + np->n_cookieverf.nfsuquad[1] = 0; + if (np->n_cookies.lh_first) + np->n_cookies.lh_first->ndm_eocookie = 0; +} + +/* + * The write verifier has changed (probably due to a server reboot), so all + * B_NEEDCOMMIT blocks will have to be written again. Since they are on the + * dirty block list as B_DELWRI, all this takes is clearing the B_NEEDCOMMIT + * flag. Once done the new write verifier can be set for the mount point. + */ +void +nfs_clearcommit(mp) + struct mount *mp; +{ + register struct vnode *vp, *nvp; + register struct buf *bp, *nbp; + int s; + + s = splbio(); +loop: + for (vp = mp->mnt_vnodelist.lh_first; vp; vp = nvp) { + if (vp->v_mount != mp) /* Paranoia */ + goto loop; + nvp = vp->v_mntvnodes.le_next; + for (bp = vp->v_dirtyblkhd.lh_first; bp; bp = nbp) { + nbp = bp->b_vnbufs.le_next; + if ((bp->b_flags & (B_BUSY | B_DELWRI | B_NEEDCOMMIT)) + == (B_DELWRI | B_NEEDCOMMIT)) + bp->b_flags &= ~B_NEEDCOMMIT; + } + } + splx(s); +} + +/* + * Map errnos to NFS error numbers. For Version 3 also filter out error + * numbers not specified for the associated procedure. + */ +int +nfsrv_errmap(nd, err) + struct nfsrv_descript *nd; + register int err; +{ + register short *defaulterrp, *errp; + + if (nd->nd_flag & ND_NFSV3) { + if (nd->nd_procnum <= NFSPROC_COMMIT) { + errp = defaulterrp = nfsrv_v3errmap[nd->nd_procnum]; + while (*++errp) { + if (*errp == err) + return (err); + else if (*errp > err) + break; + } + return ((int)*defaulterrp); + } else + return (err & 0xffff); + } + if (err <= ELAST) + return ((int)nfsrv_v2errmap[err - 1]); + return (NFSERR_IO); +} + int nfsrv_vmio( struct vnode *vp) { vm_object_t object; diff --git a/sys/nfs/nfs_common.h b/sys/nfs/nfs_common.h index 030a985..9e89c57 100644 --- a/sys/nfs/nfs_common.h +++ b/sys/nfs/nfs_common.h @@ -34,12 +34,13 @@ * SUCH DAMAGE. * * @(#)nfsm_subs.h 8.1 (Berkeley) 6/16/93 - * $Id: nfsm_subs.h,v 1.5 1994/10/17 17:47:42 phk Exp $ + * $Id: nfsm_subs.h,v 1.6 1995/05/30 08:12:51 rgrimes Exp $ */ #ifndef _NFS_NFSM_SUBS_H_ #define _NFS_NFSM_SUBS_H_ + /* * These macros do strange and peculiar things to mbuf chains for * the assistance of the nfs code. To attempt to use them for any @@ -90,49 +91,153 @@ extern struct mbuf *nfsm_reqh(); mb->m_len += (s); \ bpos += (s); } -#define nfsm_dissect(a,c,s) \ +#define nfsm_dissect(a, c, s) \ { t1 = mtod(md, caddr_t)+md->m_len-dpos; \ if (t1 >= (s)) { \ (a) = (c)(dpos); \ dpos += (s); \ + } else if (t1 = nfsm_disct(&md, &dpos, (s), t1, &cp2)) { \ + error = t1; \ + m_freem(mrep); \ + goto nfsmout; \ } else { \ - error = nfsm_disct(&md, &dpos, (s), t1, &cp2); \ - if (error) { \ - m_freem(mrep); \ + (a) = (c)cp2; \ + } } + +#define nfsm_fhtom(v, v3) \ + { if (v3) { \ + t2 = nfsm_rndup(VTONFS(v)->n_fhsize) + NFSX_UNSIGNED; \ + if (t2 <= M_TRAILINGSPACE(mb)) { \ + nfsm_build(tl, u_long *, t2); \ + *tl++ = txdr_unsigned(VTONFS(v)->n_fhsize); \ + *(tl + ((t2>>2) - 2)) = 0; \ + bcopy((caddr_t)VTONFS(v)->n_fhp,(caddr_t)tl, \ + VTONFS(v)->n_fhsize); \ + } else if (t2 = nfsm_strtmbuf(&mb, &bpos, \ + (caddr_t)VTONFS(v)->n_fhp, VTONFS(v)->n_fhsize)) { \ + error = t2; \ + m_freem(mreq); \ goto nfsmout; \ - } else { \ - (a) = (c)cp2; \ } \ + } else { \ + nfsm_build(cp, caddr_t, NFSX_V2FH); \ + bcopy((caddr_t)VTONFS(v)->n_fhp, cp, NFSX_V2FH); \ } } -#define nfsm_fhtom(v) \ - nfsm_build(cp,caddr_t,NFSX_FH); \ - bcopy((caddr_t)&(VTONFS(v)->n_fh), cp, NFSX_FH) +#define nfsm_srvfhtom(f, v3) \ + { if (v3) { \ + nfsm_build(tl, u_long *, NFSX_UNSIGNED + NFSX_V3FH); \ + *tl++ = txdr_unsigned(NFSX_V3FH); \ + bcopy((caddr_t)(f), (caddr_t)tl, NFSX_V3FH); \ + } else { \ + nfsm_build(cp, caddr_t, NFSX_V2FH); \ + bcopy((caddr_t)(f), cp, NFSX_V2FH); \ + } } -#define nfsm_srvfhtom(f) \ - nfsm_build(cp,caddr_t,NFSX_FH); \ - bcopy((caddr_t)(f), cp, NFSX_FH) +#define nfsm_srvpostop_fh(f) \ + { nfsm_build(tl, u_long *, 2 * NFSX_UNSIGNED + NFSX_V3FH); \ + *tl++ = nfs_true; \ + *tl++ = txdr_unsigned(NFSX_V3FH); \ + bcopy((caddr_t)(f), (caddr_t)tl, NFSX_V3FH); \ + } -#define nfsm_mtofh(d,v) \ - { struct nfsnode *np; nfsv2fh_t *fhp; \ - nfsm_dissect(fhp,nfsv2fh_t *,NFSX_FH); \ - error = nfs_nget((d)->v_mount, fhp, &np); \ - if (error) { \ - m_freem(mrep); \ - goto nfsmout; \ +#define nfsm_mtofh(d, v, v3, f) \ + { struct nfsnode *ttnp; nfsfh_t *ttfhp; int ttfhsize; \ + if (v3) { \ + nfsm_dissect(tl, u_long *, NFSX_UNSIGNED); \ + (f) = fxdr_unsigned(int, *tl); \ + } else \ + (f) = 1; \ + if (f) { \ + nfsm_getfh(ttfhp, ttfhsize, (v3)); \ + if (t1 = nfs_nget((d)->v_mount, ttfhp, ttfhsize, \ + &ttnp)) { \ + error = t1; \ + m_freem(mrep); \ + goto nfsmout; \ + } \ + (v) = NFSTOV(ttnp); \ + } \ + if (v3) { \ + nfsm_dissect(tl, u_long *, NFSX_UNSIGNED); \ + if (f) \ + (f) = fxdr_unsigned(int, *tl); \ + else if (fxdr_unsigned(int, *tl)) \ + nfsm_adv(NFSX_V3FATTR); \ } \ - (v) = NFSTOV(np); \ - nfsm_loadattr(v, (struct vattr *)0); \ + if (f) \ + nfsm_loadattr((v), (struct vattr *)0); \ } -#define nfsm_loadattr(v,a) \ - { struct vnode *tvp = (v); \ - error = nfs_loadattrcache(&tvp, &md, &dpos, (a)); \ - if (error) { \ +#define nfsm_getfh(f, s, v3) \ + { if (v3) { \ + nfsm_dissect(tl, u_long *, NFSX_UNSIGNED); \ + if (((s) = fxdr_unsigned(int, *tl)) <= 0 || \ + (s) > NFSX_V3FHMAX) { \ + m_freem(mrep); \ + error = EBADRPC; \ + goto nfsmout; \ + } \ + } else \ + (s) = NFSX_V2FH; \ + nfsm_dissect((f), nfsfh_t *, nfsm_rndup(s)); } + +#define nfsm_loadattr(v, a) \ + { struct vnode *ttvp = (v); \ + if (t1 = nfs_loadattrcache(&ttvp, &md, &dpos, (a))) { \ + error = t1; \ m_freem(mrep); \ goto nfsmout; \ } \ - (v) = tvp; } + (v) = ttvp; } + +#define nfsm_postop_attr(v, f) \ + { struct vnode *ttvp = (v); \ + nfsm_dissect(tl, u_long *, NFSX_UNSIGNED); \ + if ((f) = fxdr_unsigned(int, *tl)) { \ + if (t1 = nfs_loadattrcache(&ttvp, &md, &dpos, \ + (struct vattr *)0)) { \ + error = t1; \ + (f) = 0; \ + m_freem(mrep); \ + goto nfsmout; \ + } \ + (v) = ttvp; \ + } } + +/* Used as (f) for nfsm_wcc_data() */ +#define NFSV3_WCCRATTR 0 +#define NFSV3_WCCCHK 1 + +#define nfsm_wcc_data(v, f) \ + { int ttattrf, ttretf = 0; \ + nfsm_dissect(tl, u_long *, NFSX_UNSIGNED); \ + if (*tl == nfs_true) { \ + nfsm_dissect(tl, u_long *, 6 * NFSX_UNSIGNED); \ + if (f) \ + ttretf = (VTONFS(v)->n_mtime == \ + fxdr_unsigned(u_long, *(tl + 2))); \ + } \ + nfsm_postop_attr((v), ttattrf); \ + if (f) { \ + (f) = ttretf; \ + } else { \ + (f) = ttattrf; \ + } } + +#define nfsm_v3sattr(s, a, u, g) \ + { (s)->sa_modetrue = nfs_true; \ + (s)->sa_mode = vtonfsv3_mode((a)->va_mode); \ + (s)->sa_uidtrue = nfs_true; \ + (s)->sa_uid = txdr_unsigned(u); \ + (s)->sa_gidtrue = nfs_true; \ + (s)->sa_gid = txdr_unsigned(g); \ + (s)->sa_sizefalse = nfs_false; \ + (s)->sa_atimetype = txdr_unsigned(NFSV3SATTRTIME_TOCLIENT); \ + txdr_nfsv3time(&(a)->va_atime, &(s)->sa_atime); \ + (s)->sa_mtimetype = txdr_unsigned(NFSV3SATTRTIME_TOCLIENT); \ + txdr_nfsv3time(&(a)->va_mtime, &(s)->sa_mtime); \ + } #define nfsm_strsiz(s,m) \ { nfsm_dissect(tl,u_long *,NFSX_UNSIGNED); \ @@ -149,16 +254,27 @@ extern struct mbuf *nfsm_reqh(); nfsm_reply(0); \ } } +#define nfsm_srvnamesiz(s) \ + { nfsm_dissect(tl,u_long *,NFSX_UNSIGNED); \ + if (((s) = fxdr_unsigned(long,*tl)) > NFS_MAXNAMLEN) \ + error = NFSERR_NAMETOL; \ + if ((s) <= 0) \ + error = EBADRPC; \ + if (error) \ + nfsm_reply(0); \ + } + #define nfsm_mtouio(p,s) \ if ((s) > 0 && \ - (error = nfsm_mbuftouio(&md,(p),(s),&dpos))) { \ + (t1 = nfsm_mbuftouio(&md,(p),(s),&dpos))) { \ + error = t1; \ m_freem(mrep); \ goto nfsmout; \ } #define nfsm_uiotom(p,s) \ - error = nfsm_uiotombuf((p),&mb,(s),&bpos); \ - if (error) { \ + if (t1 = nfsm_uiotombuf((p),&mb,(s),&bpos)) { \ + error = t1; \ m_freem(mreq); \ goto nfsmout; \ } @@ -172,10 +288,13 @@ extern struct mbuf *nfsm_reqh(); #define nfsm_rndup(a) (((a)+3)&(~0x3)) #define nfsm_request(v, t, p, c) \ - error = nfs_request((v), mreq, (t), (p), \ - (c), &mrep, &md, &dpos); \ - if (error) \ - goto nfsmout + if (error = nfs_request((v), mreq, (t), (p), \ + (c), &mrep, &md, &dpos)) { \ + if (error & NFSERR_RETERR) \ + error &= ~NFSERR_RETERR; \ + else \ + goto nfsmout; \ + } #define nfsm_strtom(a,s,m) \ if ((s) > (m)) { \ @@ -189,12 +308,10 @@ extern struct mbuf *nfsm_reqh(); *tl++ = txdr_unsigned(s); \ *(tl+((t2>>2)-2)) = 0; \ bcopy((caddr_t)(a), (caddr_t)tl, (s)); \ - } else { \ - error = nfsm_strtmbuf(&mb, &bpos, (a), (s)); \ - if (error) { \ - m_freem(mreq); \ - goto nfsmout; \ - } \ + } else if (t2 = nfsm_strtmbuf(&mb, &bpos, (a), (s))) { \ + error = t2; \ + m_freem(mreq); \ + goto nfsmout; \ } #define nfsm_srvdone \ @@ -204,33 +321,53 @@ extern struct mbuf *nfsm_reqh(); #define nfsm_reply(s) \ { \ nfsd->nd_repstat = error; \ - if (error) \ - (void) nfs_rephead(0, nfsd, error, cache, &frev, \ + if (error && !(nfsd->nd_flag & ND_NFSV3)) \ + (void) nfs_rephead(0, nfsd, slp, error, cache, &frev, \ mrq, &mb, &bpos); \ else \ - (void) nfs_rephead((s), nfsd, error, cache, &frev, \ + (void) nfs_rephead((s), nfsd, slp, error, cache, &frev, \ mrq, &mb, &bpos); \ m_freem(mrep); \ mreq = *mrq; \ - if (error) \ + if (error && (!(nfsd->nd_flag & ND_NFSV3) || \ + error == EBADRPC)) \ return(0); \ } +#define nfsm_writereply(s, v3) \ + { \ + nfsd->nd_repstat = error; \ + if (error && !(v3)) \ + (void) nfs_rephead(0, nfsd, slp, error, cache, &frev, \ + &mreq, &mb, &bpos); \ + else \ + (void) nfs_rephead((s), nfsd, slp, error, cache, &frev, \ + &mreq, &mb, &bpos); \ + } + #define nfsm_adv(s) \ - t1 = mtod(md, caddr_t)+md->m_len-dpos; \ + { t1 = mtod(md, caddr_t)+md->m_len-dpos; \ if (t1 >= (s)) { \ dpos += (s); \ - } else { \ - error = nfs_adv(&md, &dpos, (s), t1); \ - if (error) { \ - m_freem(mrep); \ - goto nfsmout; \ - } \ - } + } else if (t1 = nfs_adv(&md, &dpos, (s), t1)) { \ + error = t1; \ + m_freem(mrep); \ + goto nfsmout; \ + } } #define nfsm_srvmtofh(f) \ - nfsm_dissect(tl, u_long *, NFSX_FH); \ - bcopy((caddr_t)tl, (caddr_t)f, NFSX_FH) + { if (nfsd->nd_flag & ND_NFSV3) { \ + nfsm_dissect(tl, u_long *, NFSX_UNSIGNED); \ + if (fxdr_unsigned(int, *tl) != NFSX_V3FH) { \ + error = EBADRPC; \ + nfsm_reply(0); \ + } \ + } \ + nfsm_dissect(tl, u_long *, NFSX_V3FH); \ + bcopy((caddr_t)tl, (caddr_t)(f), NFSX_V3FH); \ + if ((nfsd->nd_flag & ND_NFSV3) == 0) \ + nfsm_adv(NFSX_V2FH - NFSX_V3FH); \ + } #define nfsm_clget \ if (bp >= be) { \ @@ -246,41 +383,57 @@ extern struct mbuf *nfsm_reqh(); } \ tl = (u_long *)bp -#define nfsm_srvfillattr \ - fp->fa_type = vtonfs_type(vap->va_type); \ - fp->fa_mode = vtonfs_mode(vap->va_type, vap->va_mode); \ - fp->fa_nlink = txdr_unsigned(vap->va_nlink); \ - fp->fa_uid = txdr_unsigned(vap->va_uid); \ - fp->fa_gid = txdr_unsigned(vap->va_gid); \ - if (nfsd->nd_nqlflag == NQL_NOVAL) { \ - fp->fa_nfsblocksize = txdr_unsigned(vap->va_blocksize); \ - if (vap->va_type == VFIFO) \ - fp->fa_nfsrdev = 0xffffffff; \ - else \ - fp->fa_nfsrdev = txdr_unsigned(vap->va_rdev); \ - fp->fa_nfsfsid = txdr_unsigned(vap->va_fsid); \ - fp->fa_nfsfileid = txdr_unsigned(vap->va_fileid); \ - fp->fa_nfssize = txdr_unsigned(vap->va_size); \ - fp->fa_nfsblocks = txdr_unsigned(vap->va_bytes / NFS_FABLKSIZE); \ - txdr_nfstime(&vap->va_atime, &fp->fa_nfsatime); \ - txdr_nfstime(&vap->va_mtime, &fp->fa_nfsmtime); \ - txdr_nfstime(&vap->va_ctime, &fp->fa_nfsctime); \ - } else { \ - fp->fa_nqblocksize = txdr_unsigned(vap->va_blocksize); \ - if (vap->va_type == VFIFO) \ - fp->fa_nqrdev = 0xffffffff; \ - else \ - fp->fa_nqrdev = txdr_unsigned(vap->va_rdev); \ - fp->fa_nqfsid = txdr_unsigned(vap->va_fsid); \ - fp->fa_nqfileid = txdr_unsigned(vap->va_fileid); \ - txdr_hyper(&vap->va_size, &fp->fa_nqsize); \ - txdr_hyper(&vap->va_bytes, &fp->fa_nqbytes); \ - txdr_nqtime(&vap->va_atime, &fp->fa_nqatime); \ - txdr_nqtime(&vap->va_mtime, &fp->fa_nqmtime); \ - txdr_nqtime(&vap->va_ctime, &fp->fa_nqctime); \ - fp->fa_nqflags = txdr_unsigned(vap->va_flags); \ - fp->fa_nqgen = txdr_unsigned(vap->va_gen); \ - txdr_hyper(&vap->va_filerev, &fp->fa_nqfilerev); \ - } +#define nfsm_srvfillattr(a, f) \ + nfsm_srvfattr(nfsd, (a), (f)) + +#define nfsm_srvwcc_data(br, b, ar, a) \ + nfsm_srvwcc(nfsd, (br), (b), (ar), (a), &mb, &bpos) + +#define nfsm_srvpostop_attr(r, a) \ + nfsm_srvpostopattr(nfsd, (r), (a), &mb, &bpos) + +#define nfsm_srvsattr(a) \ + { nfsm_dissect(tl, u_long *, NFSX_UNSIGNED); \ + if (*tl == nfs_true) { \ + nfsm_dissect(tl, u_long *, NFSX_UNSIGNED); \ + (a)->va_mode = nfstov_mode(*tl); \ + } \ + nfsm_dissect(tl, u_long *, NFSX_UNSIGNED); \ + if (*tl == nfs_true) { \ + nfsm_dissect(tl, u_long *, NFSX_UNSIGNED); \ + (a)->va_uid = fxdr_unsigned(uid_t, *tl); \ + } \ + nfsm_dissect(tl, u_long *, NFSX_UNSIGNED); \ + if (*tl == nfs_true) { \ + nfsm_dissect(tl, u_long *, NFSX_UNSIGNED); \ + (a)->va_gid = fxdr_unsigned(gid_t, *tl); \ + } \ + nfsm_dissect(tl, u_long *, NFSX_UNSIGNED); \ + if (*tl == nfs_true) { \ + nfsm_dissect(tl, u_long *, 2 * NFSX_UNSIGNED); \ + fxdr_hyper(tl, &(a)->va_size); \ + } \ + nfsm_dissect(tl, u_long *, NFSX_UNSIGNED); \ + switch (fxdr_unsigned(int, *tl)) { \ + case NFSV3SATTRTIME_TOCLIENT: \ + nfsm_dissect(tl, u_long *, 2 * NFSX_UNSIGNED); \ + fxdr_nfsv3time(tl, &(a)->va_atime); \ + break; \ + case NFSV3SATTRTIME_TOSERVER: \ + (a)->va_atime.ts_sec = time.tv_sec; \ + (a)->va_atime.ts_nsec = time.tv_usec * 1000; \ + break; \ + }; \ + nfsm_dissect(tl, u_long *, NFSX_UNSIGNED); \ + switch (fxdr_unsigned(int, *tl)) { \ + case NFSV3SATTRTIME_TOCLIENT: \ + nfsm_dissect(tl, u_long *, 2 * NFSX_UNSIGNED); \ + fxdr_nfsv3time(tl, &(a)->va_mtime); \ + break; \ + case NFSV3SATTRTIME_TOSERVER: \ + (a)->va_mtime.ts_sec = time.tv_sec; \ + (a)->va_mtime.ts_nsec = time.tv_usec * 1000; \ + break; \ + }; } #endif diff --git a/sys/nfs/nfs_node.c b/sys/nfs/nfs_node.c index e5315f4..c002cea 100644 --- a/sys/nfs/nfs_node.c +++ b/sys/nfs/nfs_node.c @@ -34,7 +34,7 @@ * SUCH DAMAGE. * * @(#)nfs_node.c 8.2 (Berkeley) 12/30/93 - * $Id: nfs_node.c,v 1.7 1994/10/17 17:47:33 phk Exp $ + * $Id: nfs_node.c,v 1.8 1995/03/16 18:15:36 bde Exp $ */ #include <sys/param.h> @@ -47,14 +47,12 @@ #include <sys/malloc.h> #include <nfs/rpcv2.h> -#include <nfs/nfsv2.h> +#include <nfs/nfsproto.h> #include <nfs/nfs.h> #include <nfs/nfsnode.h> #include <nfs/nfsmount.h> #include <nfs/nqnfs.h> -#define NFSNOHASH(fhsum) \ - (&nfsnodehashtbl[(fhsum) & nfsnodehash]) LIST_HEAD(nfsnodehashhead, nfsnode) *nfsnodehashtbl; u_long nfsnodehash; @@ -79,19 +77,20 @@ nfs_nhinit() /* * Compute an entry in the NFS hash table structure */ -struct nfsnodehashhead * -nfs_hash(fhp) - register nfsv2fh_t *fhp; +u_long +nfs_hash(fhp, fhsize) + register nfsfh_t *fhp; + int fhsize; { register u_char *fhpp; register u_long fhsum; - int i; + register int i; fhpp = &fhp->fh_bytes[0]; fhsum = 0; - for (i = 0; i < NFSX_FH; i++) + for (i = 0; i < fhsize; i++) fhsum += *fhpp++; - return (NFSNOHASH(fhsum)); + return (fhsum); } /* @@ -101,9 +100,10 @@ nfs_hash(fhp) * nfsnode structure is returned. */ int -nfs_nget(mntp, fhp, npp) +nfs_nget(mntp, fhp, fhsize, npp) struct mount *mntp; - register nfsv2fh_t *fhp; + register nfsfh_t *fhp; + int fhsize; struct nfsnode **npp; { register struct nfsnode *np; @@ -112,11 +112,11 @@ nfs_nget(mntp, fhp, npp) struct vnode *nvp; int error; - nhpp = nfs_hash(fhp); + nhpp = NFSNOHASH(nfs_hash(fhp, fhsize)); loop: for (np = nhpp->lh_first; np != 0; np = np->n_hash.le_next) { - if (mntp != NFSTOV(np)->v_mount || - bcmp((caddr_t)fhp, (caddr_t)&np->n_fh, NFSX_FH)) + if (mntp != NFSTOV(np)->v_mount || np->n_fhsize != fhsize || + bcmp((caddr_t)fhp, (caddr_t)np->n_fhp, fhsize)) continue; vp = NFSTOV(np); if (vget(vp, 1)) @@ -131,27 +131,26 @@ loop: } vp = nvp; MALLOC(np, struct nfsnode *, sizeof *np, M_NFSNODE, M_WAITOK); + bzero((caddr_t)np, sizeof *np); vp->v_data = np; np->n_vnode = vp; /* * Insert the nfsnode in the hash queue for its new file handle */ - np->n_flag = 0; LIST_INSERT_HEAD(nhpp, np, n_hash); - bcopy((caddr_t)fhp, (caddr_t)&np->n_fh, NFSX_FH); - np->n_attrstamp = 0; - np->n_direofoffset = 0; - np->n_sillyrename = (struct sillyrename *)0; - np->n_size = 0; - np->n_mtime = 0; - np->n_lockf = 0; - if (VFSTONFS(mntp)->nm_flag & NFSMNT_NQNFS) { - np->n_brev = 0; - np->n_lrev = 0; - np->n_expiry = (time_t)0; - np->n_timer.cqe_next = (struct nfsnode *)0; - } + if (fhsize > NFS_SMALLFH) { + MALLOC(np->n_fhp, nfsfh_t *, fhsize, M_NFSBIGFH, M_WAITOK); + } else + np->n_fhp = &np->n_fh; + bcopy((caddr_t)fhp, (caddr_t)np->n_fhp, fhsize); + np->n_fhsize = fhsize; *npp = np; + + /* + * Lock the new nfsnode. + */ + VOP_LOCK(vp); + return (0); } @@ -168,7 +167,10 @@ nfs_inactive(ap) np = VTONFS(ap->a_vp); if (prtactive && ap->a_vp->v_usecount != 0) vprint("nfs_inactive: pushing active", ap->a_vp); - sp = np->n_sillyrename; + if (ap->a_vp->v_type != VDIR) + sp = np->n_sillyrename; + else + sp = (struct sillyrename *)0; np->n_sillyrename = (struct sillyrename *)0; if (sp) { /* @@ -178,9 +180,7 @@ nfs_inactive(ap) nfs_removeit(sp); crfree(sp->s_cred); vrele(sp->s_dvp); -#ifdef SILLYSEPARATE - free((caddr_t)sp, M_NFSREQ); -#endif + FREE((caddr_t)sp, M_NFSREQ); } np->n_flag &= (NMODIFIED | NFLUSHINPROG | NFLUSHWANT | NQNFSEVICTED | NQNFSNONCACHE | NQNFSWRITE); @@ -199,6 +199,7 @@ nfs_reclaim(ap) register struct vnode *vp = ap->a_vp; register struct nfsnode *np = VTONFS(vp); register struct nfsmount *nmp = VFSTONFS(vp->v_mount); + register struct nfsdmap *dp, *dp2; if (prtactive && vp->v_usecount != 0) vprint("nfs_reclaim: pushing active", vp); @@ -211,6 +212,24 @@ nfs_reclaim(ap) if ((nmp->nm_flag & NFSMNT_NQNFS) && np->n_timer.cqe_next != 0) { CIRCLEQ_REMOVE(&nmp->nm_timerhead, np, n_timer); } + + /* + * Free up any directory cookie structures and + * large file handle structures that might be associated with + * this nfs node. + */ + if (vp->v_type == VDIR) { + dp = np->n_cookies.lh_first; + while (dp) { + dp2 = dp; + dp = dp->ndm_list.le_next; + FREE((caddr_t)dp2, M_NFSDIROFF); + } + } + if (np->n_fhsize > NFS_SMALLFH) { + FREE((caddr_t)np->n_fhp, M_NFSBIGFH); + } + cache_purge(vp); FREE(vp->v_data, M_NFSNODE); vp->v_data = (void *)0; @@ -227,6 +246,7 @@ nfs_lock(ap) } */ *ap; { register struct vnode *vp = ap->a_vp; + struct nfsnode *np = VTONFS(vp); /* * Ugh, another place where interruptible mounts will get hung. @@ -239,6 +259,33 @@ nfs_lock(ap) } if (vp->v_tag == VT_NON) return (ENOENT); + +#if 0 + /* + * Only lock regular files. If a server crashed while we were + * holding a directory lock, we could easily end up sleeping + * until the server rebooted while holding a lock on the root. + * Locks are only needed for protecting critical sections in + * VMIO at the moment. + * New vnodes will have type VNON but they should be locked + * since they may become VREG. This is checked in loadattrcache + * and unwanted locks are released there. + */ + if (vp->v_type == VREG || vp->v_type == VNON) { + while (np->n_flag & NLOCKED) { + np->n_flag |= NWANTED; + (void) tsleep((caddr_t) np, PINOD, "nfslck2", 0); + /* + * If the vnode has transmuted into a VDIR while we + * were asleep, then skip the lock. + */ + if (vp->v_type != VREG && vp->v_type != VNON) + return (0); + } + np->n_flag |= NLOCKED; + } +#endif + return (0); } @@ -251,6 +298,20 @@ nfs_unlock(ap) struct vnode *a_vp; } */ *ap; { +#if 0 + struct vnode* vp = ap->a_vp; + struct nfsnode* np = VTONFS(vp); + + if (vp->v_type == VREG || vp->v_type == VNON) { + if (!(np->n_flag & NLOCKED)) + panic("nfs_unlock: nfsnode not locked"); + np->n_flag &= ~NLOCKED; + if (np->n_flag & NWANTED) { + np->n_flag &= ~NWANTED; + wakeup((caddr_t) np); + } + } +#endif return (0); } @@ -264,8 +325,7 @@ nfs_islocked(ap) struct vnode *a_vp; } */ *ap; { - - return (0); + return VTONFS(ap->a_vp)->n_flag & NLOCKED ? 1 : 0; } /* diff --git a/sys/nfs/nfs_nqlease.c b/sys/nfs/nfs_nqlease.c index 81571de..d953699 100644 --- a/sys/nfs/nfs_nqlease.c +++ b/sys/nfs/nfs_nqlease.c @@ -34,7 +34,7 @@ * SUCH DAMAGE. * * @(#)nfs_nqlease.c 8.3 (Berkeley) 1/4/94 - * $Id: nfs_nqlease.c,v 1.10 1995/02/15 03:39:58 davidg Exp $ + * $Id: nfs_nqlease.c,v 1.11 1995/05/30 08:12:36 rgrimes Exp $ */ /* @@ -66,7 +66,7 @@ #include <netinet/in.h> #include <nfs/rpcv2.h> -#include <nfs/nfsv2.h> +#include <nfs/nfsproto.h> #include <nfs/nfs.h> #include <nfs/nfsm_subs.h> #include <nfs/xdr_subs.h> @@ -75,7 +75,6 @@ #include <nfs/nfsmount.h> time_t nqnfsstarttime = (time_t)0; -u_long nqnfs_prog, nqnfs_vers; int nqsrv_clockskew = NQ_CLOCKSKEW; int nqsrv_writeslack = NQ_WRITESLACK; int nqsrv_maxlease = NQ_MAXLEASE; @@ -91,14 +90,13 @@ struct mbuf *nfsm_rpchead(); */ int nqnfs_piggy[NFS_NPROCS] = { 0, - NQL_READ, - NQL_WRITE, 0, - NQL_READ, - NQL_READ, - NQL_READ, + ND_WRITE, + ND_READ, 0, - NQL_WRITE, + ND_READ, + ND_READ, + ND_WRITE, 0, 0, 0, @@ -106,24 +104,30 @@ int nqnfs_piggy[NFS_NPROCS] = { 0, 0, 0, - NQL_READ, 0, - NQL_READ, + ND_READ, + ND_READ, + 0, + 0, + 0, + 0, 0, 0, 0, 0, }; -extern nfstype nfs_type[9]; +extern nfstype nfsv2_type[9]; +extern nfstype nfsv3_type[9]; extern struct nfssvc_sock *nfs_udpsock, *nfs_cltpsock; extern int nfsd_waiting; +extern struct nfsstats nfsstats; #define TRUE 1 #define FALSE 0 /* - * Get or check for a lease for "vp", based on NQL_CHECK flag. + * Get or check for a lease for "vp", based on ND_CHECK flag. * The rules are as follows: * - if a current non-caching lease, reply non-caching * - if a current lease for same host only, extend lease @@ -146,11 +150,12 @@ extern int nfsd_waiting; * queue yet. (Ditto for the splsoftclock() and splx(s) calls) */ int -nqsrv_getlease(vp, duration, flags, nd, nam, cachablep, frev, cred) +nqsrv_getlease(vp, duration, flags, slp, procp, nam, cachablep, frev, cred) struct vnode *vp; u_long *duration; int flags; - struct nfsd *nd; + struct nfssvc_sock *slp; + struct proc *procp; struct mbuf *nam; int *cachablep; u_quad_t *frev; @@ -169,13 +174,13 @@ nqsrv_getlease(vp, duration, flags, nd, nam, cachablep, frev, cred) return (0); if (*duration > nqsrv_maxlease) *duration = nqsrv_maxlease; - error = VOP_GETATTR(vp, &vattr, cred, nd->nd_procp); + error = VOP_GETATTR(vp, &vattr, cred, procp); if (error) return (error); *frev = vattr.va_filerev; s = splsoftclock(); tlp = vp->v_lease; - if ((flags & NQL_CHECK) == 0) + if ((flags & ND_CHECK) == 0) nfsstats.srvnqnfs_getleases++; if (tlp == (struct nqlease *)0) { @@ -205,12 +210,12 @@ nqsrv_getlease(vp, duration, flags, nd, nam, cachablep, frev, cred) if (lp) { if ((lp->lc_flag & LC_NONCACHABLE) || (lp->lc_morehosts == (struct nqm *)0 && - nqsrv_cmpnam(nd->nd_slp, nam, &lp->lc_host))) + nqsrv_cmpnam(slp, nam, &lp->lc_host))) goto doreply; - if ((flags & NQL_READ) && (lp->lc_flag & LC_WRITE) == 0) { - if (flags & NQL_CHECK) + if ((flags & ND_READ) && (lp->lc_flag & LC_WRITE) == 0) { + if (flags & ND_CHECK) goto doreply; - if (nqsrv_cmpnam(nd->nd_slp, nam, &lp->lc_host)) + if (nqsrv_cmpnam(slp, nam, &lp->lc_host)) goto doreply; i = 0; if (lp->lc_morehosts) { @@ -222,7 +227,7 @@ nqsrv_getlease(vp, duration, flags, nd, nam, cachablep, frev, cred) ok = 0; } while (ok && (lph->lph_flag & LC_VALID)) { - if (nqsrv_cmpnam(nd->nd_slp, nam, lph)) + if (nqsrv_cmpnam(slp, nam, lph)) goto doreply; if (++i == LC_MOREHOSTSIZ) { i = 0; @@ -242,12 +247,12 @@ nqsrv_getlease(vp, duration, flags, nd, nam, cachablep, frev, cred) bzero((caddr_t)*lphp, sizeof (struct nqm)); lph = (*lphp)->lpm_hosts; } - nqsrv_addhost(lph, nd->nd_slp, nam); + nqsrv_addhost(lph, slp, nam); nqsrv_unlocklease(lp); } else { lp->lc_flag |= LC_NONCACHABLE; nqsrv_locklease(lp); - nqsrv_send_eviction(vp, lp, nd->nd_slp, nam, cred); + nqsrv_send_eviction(vp, lp, slp, nam, cred); nqsrv_waitfor_expiry(lp); nqsrv_unlocklease(lp); } @@ -255,20 +260,20 @@ doreply: /* * Update the lease and return */ - if ((flags & NQL_CHECK) == 0) + if ((flags & ND_CHECK) == 0) nqsrv_instimeq(lp, *duration); if (lp->lc_flag & LC_NONCACHABLE) *cachablep = 0; else { *cachablep = 1; - if (flags & NQL_WRITE) + if (flags & ND_WRITE) lp->lc_flag |= LC_WRITTEN; } splx(s); return (0); } splx(s); - if (flags & NQL_CHECK) + if (flags & ND_CHECK) return (0); /* @@ -285,9 +290,9 @@ doreply: } MALLOC(lp, struct nqlease *, sizeof (struct nqlease), M_NQLEASE, M_WAITOK); bzero((caddr_t)lp, sizeof (struct nqlease)); - if (flags & NQL_WRITE) + if (flags & ND_WRITE) lp->lc_flag |= (LC_WRITE | LC_WRITTEN); - nqsrv_addhost(&lp->lc_host, nd->nd_slp, nam); + nqsrv_addhost(&lp->lc_host, slp, nam); lp->lc_vp = vp; lp->lc_fsid = fh.fh_fsid; bcopy(fh.fh_fid.fid_data, lp->lc_fiddata, @@ -308,9 +313,12 @@ doreply: /* * Local lease check for server syscalls. * Just set up args and let nqsrv_getlease() do the rest. + * nqnfs_vop_lease_check() is the VOP_LEASE() form of the same routine. + * Ifdef'd code in nfsnode.h renames these routines to whatever a particular + * OS needs. */ void -nfs_lease_check(vp, p, cred, flag) +nqnfs_lease_check(vp, p, cred, flag) struct vnode *vp; struct proc *p; struct ucred *cred; @@ -318,14 +326,31 @@ nfs_lease_check(vp, p, cred, flag) { u_long duration = 0; int cache; - struct nfsd nfsd; u_quad_t frev; - nfsd.nd_slp = NQLOCALSLP; - nfsd.nd_procp = p; - (void) nqsrv_getlease(vp, &duration, NQL_CHECK | flag, &nfsd, - (struct mbuf *)0, &cache, &frev, cred); + (void) nqsrv_getlease(vp, &duration, ND_CHECK | flag, NQLOCALSLP, + p, (struct mbuf *)0, &cache, &frev, cred); +} + +#ifdef HAS_VOPLEASE +int +nqnfs_vop_lease_check(ap) + struct vop_lease_args /* { + struct vnode *a_vp; + struct proc *a_p; + struct ucred *a_cred; + int a_flag; + } */ *ap; +{ + u_long duration = 0; + int cache; + u_quad_t frev; + + (void) nqsrv_getlease(ap->a_vp, &duration, ND_CHECK | ap->a_flag, + NQLOCALSLP, ap->a_p, (struct mbuf *)0, &cache, &frev, ap->a_cred); + return (0); } +#endif /* * Add a host to an nqhost structure for a lease. @@ -486,9 +511,9 @@ nqsrv_send_eviction(vp, lp, slp, nam, cred) else solockp = (int *)0; nfsm_reqhead((struct vnode *)0, NQNFSPROC_EVICTED, - NFSX_FH); - nfsm_build(cp, caddr_t, NFSX_FH); - bzero(cp, NFSX_FH); + NFSX_V3FH); + nfsm_build(cp, caddr_t, NFSX_V3FH); + bzero(cp, NFSX_V3FH); fhp = (fhandle_t *)cp; fhp->fh_fsid = vp->v_mount->mnt_stat.f_fsid; VFS_VPTOFH(vp, &fhp->fh_fid); @@ -502,8 +527,9 @@ nqsrv_send_eviction(vp, lp, slp, nam, cred) printf("mbuf siz=%d\n",siz); panic("Bad nfs svc reply"); } - m = nfsm_rpchead(cred, TRUE, NQNFSPROC_EVICTED, - RPCAUTH_UNIX, 5*NFSX_UNSIGNED, (char *)0, + m = nfsm_rpchead(cred, (NFSMNT_NFSV3 | NFSMNT_NQNFS), + NQNFSPROC_EVICTED, + RPCAUTH_UNIX, 5 * NFSX_UNSIGNED, (char *)0, mreq, siz, &mheadend, &xid); /* * For stream protocols, prepend a Sun RPC @@ -677,18 +703,21 @@ nqnfs_serverd() * do the real work. */ int -nqnfsrv_getlease(nfsd, mrep, md, dpos, cred, nam, mrq) - struct nfsd *nfsd; - struct mbuf *mrep, *md; - caddr_t dpos; - struct ucred *cred; - struct mbuf *nam, **mrq; +nqnfsrv_getlease(nfsd, slp, procp, mrq) + struct nfsrv_descript *nfsd; + struct nfssvc_sock *slp; + struct proc *procp; + struct mbuf **mrq; { - register struct nfsv2_fattr *fp; + struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; + struct mbuf *nam = nfsd->nd_nam; + caddr_t dpos = nfsd->nd_dpos; + struct ucred *cred = &nfsd->nd_cr; + register struct nfs_fattr *fp; struct vattr va; register struct vattr *vap = &va; struct vnode *vp; - nfsv2fh_t nfh; + nfsfh_t nfh; fhandle_t *fhp; register u_long *tl; register long t1; @@ -701,29 +730,29 @@ nqnfsrv_getlease(nfsd, mrep, md, dpos, cred, nam, mrq) fhp = &nfh.fh_generic; nfsm_srvmtofh(fhp); - nfsm_dissect(tl, u_long *, 2*NFSX_UNSIGNED); + nfsm_dissect(tl, u_long *, 2 * NFSX_UNSIGNED); flags = fxdr_unsigned(int, *tl++); nfsd->nd_duration = fxdr_unsigned(int, *tl); - error = nfsrv_fhtovp(fhp, - TRUE, &vp, cred, nfsd->nd_slp, nam, &rdonly); + error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam, &rdonly, + (nfsd->nd_flag & ND_KERBAUTH)); if (error) nfsm_reply(0); - if (rdonly && flags == NQL_WRITE) { + if (rdonly && flags == ND_WRITE) { error = EROFS; nfsrv_vput(vp); nfsm_reply(0); } - (void) nqsrv_getlease(vp, &nfsd->nd_duration, flags, nfsd, + (void) nqsrv_getlease(vp, &nfsd->nd_duration, flags, slp, procp, nam, &cache, &frev, cred); - error = VOP_GETATTR(vp, vap, cred, nfsd->nd_procp); + error = VOP_GETATTR(vp, vap, cred, procp); nfsrv_vput(vp); - nfsm_reply(NFSX_NQFATTR + 4*NFSX_UNSIGNED); - nfsm_build(tl, u_long *, 4*NFSX_UNSIGNED); + nfsm_reply(NFSX_V3FATTR + 4 * NFSX_UNSIGNED); + nfsm_build(tl, u_long *, 4 * NFSX_UNSIGNED); *tl++ = txdr_unsigned(cache); *tl++ = txdr_unsigned(nfsd->nd_duration); txdr_hyper(&frev, tl); - nfsm_build(fp, struct nfsv2_fattr *, NFSX_NQFATTR); - nfsm_srvfillattr; + nfsm_build(fp, struct nfs_fattr *, NFSX_V3FATTR); + nfsm_srvfillattr(vap, fp); nfsm_srvdone; } @@ -732,23 +761,28 @@ nqnfsrv_getlease(nfsd, mrep, md, dpos, cred, nam, mrq) * client. Find the entry and expire it. */ int -nqnfsrv_vacated(nfsd, mrep, md, dpos, cred, nam, mrq) - struct nfsd *nfsd; - struct mbuf *mrep, *md; - caddr_t dpos; - struct ucred *cred; - struct mbuf *nam, **mrq; +nqnfsrv_vacated(nfsd, slp, procp, mrq) + struct nfsrv_descript *nfsd; + struct nfssvc_sock *slp; + struct proc *procp; + struct mbuf **mrq; { + struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; + struct mbuf *nam = nfsd->nd_nam; + caddr_t dpos = nfsd->nd_dpos; + struct ucred *cred = &nfsd->nd_cr; register struct nqlease *lp; register struct nqhost *lph; struct nqlease *tlp = (struct nqlease *)0; - nfsv2fh_t nfh; + nfsfh_t nfh; fhandle_t *fhp; register u_long *tl; register long t1; struct nqm *lphnext; - int error = 0, i, len, ok, gotit = 0; - char *cp2; + struct mbuf *mreq, *mb; + int error = 0, i, len, ok, gotit = 0, cache = 0; + char *cp2, *bpos; + u_quad_t frev; fhp = &nfh.fh_generic; nfsm_srvmtofh(fhp); @@ -774,7 +808,7 @@ nqnfsrv_vacated(nfsd, mrep, md, dpos, cred, nam, mrq) lphnext = lp->lc_morehosts; ok = 1; while (ok && (lph->lph_flag & LC_VALID)) { - if (nqsrv_cmpnam(nfsd->nd_slp, nam, lph)) { + if (nqsrv_cmpnam(slp, nam, lph)) { lph->lph_flag |= LC_VACATED; gotit++; break; @@ -812,7 +846,7 @@ nqnfs_getlease(vp, rwflag, cred, p) { register u_long *tl; register caddr_t cp; - register long t1; + register long t1, t2; register struct nfsnode *np; struct nfsmount *nmp = VFSTONFS(vp->v_mount); caddr_t bpos, dpos, cp2; @@ -823,16 +857,16 @@ nqnfs_getlease(vp, rwflag, cred, p) u_quad_t frev; nfsstats.rpccnt[NQNFSPROC_GETLEASE]++; - mb = mreq = nfsm_reqh(vp, NQNFSPROC_GETLEASE, NFSX_FH+2*NFSX_UNSIGNED, + mb = mreq = nfsm_reqh(vp, NQNFSPROC_GETLEASE, NFSX_V3FH+2*NFSX_UNSIGNED, &bpos); - nfsm_fhtom(vp); - nfsm_build(tl, u_long *, 2*NFSX_UNSIGNED); + nfsm_fhtom(vp, 1); + nfsm_build(tl, u_long *, 2 * NFSX_UNSIGNED); *tl++ = txdr_unsigned(rwflag); *tl = txdr_unsigned(nmp->nm_leaseterm); reqtime = time.tv_sec; nfsm_request(vp, NQNFSPROC_GETLEASE, p, cred); np = VTONFS(vp); - nfsm_dissect(tl, u_long *, 4*NFSX_UNSIGNED); + nfsm_dissect(tl, u_long *, 4 * NFSX_UNSIGNED); cachable = fxdr_unsigned(int, *tl++); reqtime += fxdr_unsigned(int, *tl++); if (reqtime > time.tv_sec) { @@ -856,6 +890,8 @@ nqnfs_vacated(vp, cred) register caddr_t cp; register struct mbuf *m; register int i; + register u_long *tl; + register long t1, t2; caddr_t bpos; u_long xid; int error = 0; @@ -865,16 +901,16 @@ nqnfs_vacated(vp, cred) nmp = VFSTONFS(vp->v_mount); nfsstats.rpccnt[NQNFSPROC_VACATED]++; - nfsm_reqhead(vp, NQNFSPROC_VACATED, NFSX_FH); - nfsm_fhtom(vp); + nfsm_reqhead(vp, NQNFSPROC_VACATED, NFSX_V3FH); + nfsm_fhtom(vp, 1); m = mreq; i = 0; while (m) { i += m->m_len; m = m->m_next; } - m = nfsm_rpchead(cred, TRUE, NQNFSPROC_VACATED, - RPCAUTH_UNIX, 5*NFSX_UNSIGNED, (char *)0, + m = nfsm_rpchead(cred, nmp->nm_flag, NQNFSPROC_VACATED, + RPCAUTH_UNIX, 5 * NFSX_UNSIGNED, (char *)0, mreq, i, &mheadend, &xid); if (nmp->nm_sotype == SOCK_STREAM) { M_PREPEND(m, NFSX_UNSIGNED, M_WAIT); @@ -888,6 +924,7 @@ nqnfs_vacated(vp, cred) (void) nfs_send(nmp->nm_so, nmp->nm_nam, m, &myrep); if (nmp->nm_soflags & PR_CONNREQUIRED) nfs_sndunlock(&nmp->nm_flag); +nfsmout: return (error); } @@ -903,29 +940,37 @@ nqnfs_callback(nmp, mrep, md, dpos) register struct vnode *vp; register u_long *tl; register long t1; - nfsv2fh_t nfh; + nfsfh_t nfh; fhandle_t *fhp; struct nfsnode *np; - struct nfsd nd; - int error; - char *cp2; + struct nfsd tnfsd; + struct nfssvc_sock *slp; + struct nfsrv_descript ndesc; + register struct nfsrv_descript *nfsd = &ndesc; + struct mbuf **mrq = (struct mbuf **)0, *mb, *mreq; + int error = 0, cache = 0; + char *cp2, *bpos; + u_quad_t frev; - nd.nd_mrep = mrep; - nd.nd_md = md; - nd.nd_dpos = dpos; - error = nfs_getreq(&nd, FALSE); +#ifndef nolint + slp = NULL; +#endif + nfsd->nd_mrep = mrep; + nfsd->nd_md = md; + nfsd->nd_dpos = dpos; + error = nfs_getreq(nfsd, &tnfsd, FALSE); if (error) return (error); - md = nd.nd_md; - dpos = nd.nd_dpos; - if (nd.nd_procnum != NQNFSPROC_EVICTED) { + md = nfsd->nd_md; + dpos = nfsd->nd_dpos; + if (nfsd->nd_procnum != NQNFSPROC_EVICTED) { m_freem(mrep); return (EPERM); } fhp = &nfh.fh_generic; nfsm_srvmtofh(fhp); m_freem(mrep); - error = nfs_nget(nmp->nm_mountp, &nfh, &np); + error = nfs_nget(nmp->nm_mountp, (nfsfh_t *)fhp, NFSX_V3FH, &np); if (error) return (error); vp = NFSTOV(np); @@ -937,7 +982,7 @@ nqnfs_callback(nmp, mrep, md, dpos) CIRCLEQ_INSERT_HEAD(&nmp->nm_timerhead, np, n_timer); } } - vrele(vp); + vput(vp); nfsm_srvdone; } @@ -960,35 +1005,39 @@ nqnfs_clientd(nmp, cred, ncd, flag, argp, p) register struct nfsnode *np; struct vnode *vp; struct nfsreq myrep; + struct nfsuid *nuidp, *nnuidp; int error = 0, vpid; /* * First initialize some variables */ - nqnfs_prog = txdr_unsigned(NQNFS_PROG); - nqnfs_vers = txdr_unsigned(NQNFS_VER1); /* * If an authorization string is being passed in, get it. */ if ((flag & NFSSVC_GOTAUTH) && - (nmp->nm_flag & (NFSMNT_WAITAUTH | NFSMNT_DISMNT)) == 0) { - if (nmp->nm_flag & NFSMNT_HASAUTH) - panic("cld kerb"); - if ((flag & NFSSVC_AUTHINFAIL) == 0) { - if (ncd->ncd_authlen <= RPCAUTH_MAXSIZ && - copyin(ncd->ncd_authstr, nmp->nm_authstr, - ncd->ncd_authlen) == 0) { - nmp->nm_authtype = ncd->ncd_authtype; - nmp->nm_authlen = ncd->ncd_authlen; - } else - nmp->nm_flag |= NFSMNT_AUTHERR; + (nmp->nm_flag & (NFSMNT_WAITAUTH | NFSMNT_DISMNT)) == 0) { + if (nmp->nm_flag & NFSMNT_HASAUTH) + panic("cld kerb"); + if ((flag & NFSSVC_AUTHINFAIL) == 0) { + if (ncd->ncd_authlen <= nmp->nm_authlen && + ncd->ncd_verflen <= nmp->nm_verflen && + !copyin(ncd->ncd_authstr,nmp->nm_authstr,ncd->ncd_authlen)&& + !copyin(ncd->ncd_verfstr,nmp->nm_verfstr,ncd->ncd_verflen)){ + nmp->nm_authtype = ncd->ncd_authtype; + nmp->nm_authlen = ncd->ncd_authlen; + nmp->nm_verflen = ncd->ncd_verflen; +#ifdef NFSKERB + nmp->nm_key = ncd->ncd_key; +#endif } else - nmp->nm_flag |= NFSMNT_AUTHERR; - nmp->nm_flag |= NFSMNT_HASAUTH; - wakeup((caddr_t)&nmp->nm_authlen); + nmp->nm_flag |= NFSMNT_AUTHERR; + } else + nmp->nm_flag |= NFSMNT_AUTHERR; + nmp->nm_flag |= NFSMNT_HASAUTH; + wakeup((caddr_t)&nmp->nm_authlen); } else - nmp->nm_flag |= NFSMNT_WAITAUTH; + nmp->nm_flag |= NFSMNT_WAITAUTH; /* * Loop every second updating queue until there is a termination sig. @@ -1016,13 +1065,11 @@ nqnfs_clientd(nmp, cred, ncd, flag, argp, p) while (np != (void *)&nmp->nm_timerhead && (nmp->nm_flag & NFSMNT_DISMINPROG) == 0) { vp = NFSTOV(np); -if (vp->v_mount->mnt_stat.f_fsid.val[1] != MOUNT_NFS) panic("trash2"); vpid = vp->v_id; if (np->n_expiry < time.tv_sec) { if (vget(vp, 1) == 0) { nmp->nm_inprog = vp; if (vpid == vp->v_id) { -if (vp->v_mount->mnt_stat.f_fsid.val[1] != MOUNT_NFS) panic("trash3"); CIRCLEQ_REMOVE(&nmp->nm_timerhead, np, n_timer); np->n_timer.cqe_next = 0; if ((np->n_flag & (NMODIFIED | NQNFSEVICTED)) @@ -1047,9 +1094,8 @@ if (vp->v_mount->mnt_stat.f_fsid.val[1] != MOUNT_NFS) panic("trash3"); == NQNFSWRITE && vp->v_dirtyblkhd.lh_first && vget(vp, 1) == 0) { nmp->nm_inprog = vp; -if (vp->v_mount->mnt_stat.f_fsid.val[1] != MOUNT_NFS) panic("trash4"); if (vpid == vp->v_id && - nqnfs_getlease(vp, NQL_WRITE, cred, p)==0) + nqnfs_getlease(vp, ND_WRITE, cred, p)==0) np->n_brev = np->n_lrev; vrele(vp); nmp->nm_inprog = NULLVP; @@ -1084,6 +1130,16 @@ if (vp->v_mount->mnt_stat.f_fsid.val[1] != MOUNT_NFS) panic("trash4"); (void) dounmount(nmp->nm_mountp, 0, p); } } + + /* + * Finally, we can free up the mount structure. + */ + for (nuidp = nmp->nm_uidlruhead.tqh_first; nuidp != 0; nuidp = nnuidp) { + nnuidp = nuidp->nu_lru.tqe_next; + LIST_REMOVE(nuidp, nu_hash); + TAILQ_REMOVE(&nmp->nm_uidlruhead, nuidp, nu_lru); + free((caddr_t)nuidp, M_NFSUID); + } free((caddr_t)nmp, M_NFSMNT); if (error == EWOULDBLOCK) error = 0; @@ -1095,7 +1151,7 @@ if (vp->v_mount->mnt_stat.f_fsid.val[1] != MOUNT_NFS) panic("trash4"); * Called from the settimeofday() syscall. */ void -nfs_lease_updatetime(deltat) +nqnfs_lease_updatetime(deltat) register int deltat; { register struct nqlease *lp; @@ -1116,8 +1172,17 @@ nfs_lease_updatetime(deltat) * Search the mount list for all nqnfs mounts and do their timer * queues. */ +#if NetBSD >= 1994101 + for (mp = mountlist.cqh_first; mp != (void *)&mountlist; + mp = mp->mnt_list.cqe_next) { +#else for (mp = mountlist.tqh_first; mp != NULL; mp = mp->mnt_list.tqe_next) { +#endif +#ifdef __NetBSD__ + if (!strcmp(&mp->mnt_stat.f_fstypename[0], MOUNT_NFS)) { +#else if (mp->mnt_stat.f_fsid.val[1] == MOUNT_NFS) { +#endif nmp = VFSTONFS(mp); if (nmp->nm_flag & NFSMNT_NQNFS) { for (np = nmp->nm_timerhead.cqh_first; @@ -1174,9 +1239,9 @@ nqnfs_clientlease(nmp, np, rwflag, cachable, expiry, frev) if (np->n_timer.cqe_next != 0) { CIRCLEQ_REMOVE(&nmp->nm_timerhead, np, n_timer); - if (rwflag == NQL_WRITE) + if (rwflag == ND_WRITE) np->n_flag |= NQNFSWRITE; - } else if (rwflag == NQL_READ) + } else if (rwflag == ND_READ) np->n_flag &= ~NQNFSWRITE; else np->n_flag |= NQNFSWRITE; diff --git a/sys/nfs/nfs_serv.c b/sys/nfs/nfs_serv.c index 033dae4..f82db3b 100644 --- a/sys/nfs/nfs_serv.c +++ b/sys/nfs/nfs_serv.c @@ -34,11 +34,11 @@ * SUCH DAMAGE. * * @(#)nfs_serv.c 8.3 (Berkeley) 1/12/94 - * $Id: nfs_serv.c,v 1.14.2.1 1995/06/07 07:25:09 davidg Exp $ + * $Id: nfs_serv.c,v 1.15 1995/06/11 19:31:45 rgrimes Exp $ */ /* - * nfs version 2 server calls to vnode ops + * nfs version 2 and 3 server calls to vnode ops * - these routines generally have 3 phases * 1 - break down and validate rpc request in mbuf list * 2 - do the vnode ops for the request @@ -53,6 +53,8 @@ * returning an error from the server function implies a fatal error * such as a badly constructed rpc request that should be dropped without * a reply. + * For Version 3, nfsm_reply() does not return for the error case, since + * most version 3 rpcs return more than the status for error cases. */ #include <sys/param.h> @@ -62,67 +64,98 @@ #include <sys/namei.h> #include <sys/vnode.h> #include <sys/mount.h> +#include <sys/socket.h> +#include <sys/socketvar.h> #include <sys/mbuf.h> #include <sys/dirent.h> #include <sys/stat.h> +#include <sys/kernel.h> +#include <ufs/ufs/dir.h> #include <vm/vm.h> -#include <nfs/nfsv2.h> +#include <nfs/nfsproto.h> #include <nfs/rpcv2.h> #include <nfs/nfs.h> #include <nfs/xdr_subs.h> #include <nfs/nfsm_subs.h> #include <nfs/nqnfs.h> -/* Defs */ -#define TRUE 1 -#define FALSE 0 - /* Global vars */ -extern u_long nfs_procids[NFS_NPROCS]; extern u_long nfs_xdrneg1; extern u_long nfs_false, nfs_true; -nfstype nfs_type[9] = { NFNON, NFREG, NFDIR, NFBLK, NFCHR, NFLNK, NFNON, +extern enum vtype nv3tov_type[8]; +extern struct nfsstats nfsstats; +nfstype nfsv2_type[9] = { NFNON, NFREG, NFDIR, NFBLK, NFCHR, NFLNK, NFNON, NFCHR, NFNON }; +nfstype nfsv3_type[9] = { NFNON, NFREG, NFDIR, NFBLK, NFCHR, NFLNK, NFSOCK, + NFFIFO, NFNON }; +int nfsrvw_procrastinate = NFS_GATHERDELAY * 1000; /* - * nqnfs access service + * nfs v3 access service */ int -nqnfsrv_access(nfsd, mrep, md, dpos, cred, nam, mrq) - struct nfsd *nfsd; - struct mbuf *mrep, *md; - caddr_t dpos; - struct ucred *cred; - struct mbuf *nam, **mrq; +nfsrv3_access(nfsd, slp, procp, mrq) + struct nfsrv_descript *nfsd; + struct nfssvc_sock *slp; + struct proc *procp; + struct mbuf **mrq; { + struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; + struct mbuf *nam = nfsd->nd_nam; + caddr_t dpos = nfsd->nd_dpos; + struct ucred *cred = &nfsd->nd_cr; struct vnode *vp; - nfsv2fh_t nfh; + nfsfh_t nfh; fhandle_t *fhp; register u_long *tl; register long t1; caddr_t bpos; - int error = 0, rdonly, cache = 0, mode = 0; + int error = 0, rdonly, cache, getret; char *cp2; - struct mbuf *mb, *mreq; + struct mbuf *mb, *mreq, *mb2; + struct vattr vattr, *vap = &vattr; + u_long testmode, nfsmode; u_quad_t frev; +#ifndef nolint + cache = 0; +#endif fhp = &nfh.fh_generic; nfsm_srvmtofh(fhp); - nfsm_dissect(tl, u_long *, 3 * NFSX_UNSIGNED); - error = nfsrv_fhtovp(fhp, TRUE, &vp, cred, nfsd->nd_slp, nam, &rdonly); - if (error) - nfsm_reply(0); - if (*tl++ == nfs_true) - mode |= VREAD; - if (*tl++ == nfs_true) - mode |= VWRITE; - if (*tl == nfs_true) - mode |= VEXEC; - error = nfsrv_access(vp, mode, cred, rdonly, nfsd->nd_procp); + nfsm_dissect(tl, u_long *, NFSX_UNSIGNED); + if (error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam, + &rdonly, (nfsd->nd_flag & ND_KERBAUTH))) { + nfsm_reply(NFSX_UNSIGNED); + nfsm_srvpostop_attr(1, (struct vattr *)0); + return (0); + } + nfsmode = fxdr_unsigned(u_long, *tl); + if ((nfsmode & NFSV3ACCESS_READ) && + nfsrv_access(vp, VREAD, cred, rdonly, procp)) + nfsmode &= ~NFSV3ACCESS_READ; + if (vp->v_type == VDIR) + testmode = (NFSV3ACCESS_MODIFY | NFSV3ACCESS_EXTEND | + NFSV3ACCESS_DELETE); + else + testmode = (NFSV3ACCESS_MODIFY | NFSV3ACCESS_EXTEND); + if ((nfsmode & testmode) && + nfsrv_access(vp, VWRITE, cred, rdonly, procp)) + nfsmode &= ~testmode; + if (vp->v_type == VDIR) + testmode = NFSV3ACCESS_LOOKUP; + else + testmode = NFSV3ACCESS_EXECUTE; + if ((nfsmode & testmode) && + nfsrv_access(vp, VEXEC, cred, rdonly, procp)) + nfsmode &= ~testmode; + getret = VOP_GETATTR(vp, vap, cred, procp); nfsrv_vput(vp); - nfsm_reply(0); + nfsm_reply(NFSX_POSTOPATTR(1) + NFSX_UNSIGNED); + nfsm_srvpostop_attr(getret, vap); + nfsm_build(tl, u_long *, NFSX_UNSIGNED); + *tl = txdr_unsigned(nfsmode); nfsm_srvdone; } @@ -130,18 +163,21 @@ nqnfsrv_access(nfsd, mrep, md, dpos, cred, nam, mrq) * nfs getattr service */ int -nfsrv_getattr(nfsd, mrep, md, dpos, cred, nam, mrq) - struct nfsd *nfsd; - struct mbuf *mrep, *md; - caddr_t dpos; - struct ucred *cred; - struct mbuf *nam, **mrq; +nfsrv_getattr(nfsd, slp, procp, mrq) + struct nfsrv_descript *nfsd; + struct nfssvc_sock *slp; + struct proc *procp; + struct mbuf **mrq; { - register struct nfsv2_fattr *fp; + struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; + struct mbuf *nam = nfsd->nd_nam; + caddr_t dpos = nfsd->nd_dpos; + struct ucred *cred = &nfsd->nd_cr; + register struct nfs_fattr *fp; struct vattr va; register struct vattr *vap = &va; struct vnode *vp; - nfsv2fh_t nfh; + nfsfh_t nfh; fhandle_t *fhp; register u_long *tl; register long t1; @@ -153,15 +189,19 @@ nfsrv_getattr(nfsd, mrep, md, dpos, cred, nam, mrq) fhp = &nfh.fh_generic; nfsm_srvmtofh(fhp); - error = nfsrv_fhtovp(fhp, TRUE, &vp, cred, nfsd->nd_slp, nam, &rdonly); - if (error) + if (error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam, + &rdonly, (nfsd->nd_flag & ND_KERBAUTH))) { nfsm_reply(0); - nqsrv_getl(vp, NQL_READ); - error = VOP_GETATTR(vp, vap, cred, nfsd->nd_procp); + return (0); + } + nqsrv_getl(vp, ND_READ); + error = VOP_GETATTR(vp, vap, cred, procp); nfsrv_vput(vp); - nfsm_reply(NFSX_FATTR(nfsd->nd_nqlflag != NQL_NOVAL)); - nfsm_build(fp, struct nfsv2_fattr *, NFSX_FATTR(nfsd->nd_nqlflag != NQL_NOVAL)); - nfsm_srvfillattr; + nfsm_reply(NFSX_FATTR(nfsd->nd_flag & ND_NFSV3)); + if (error) + return (0); + nfsm_build(fp, struct nfs_fattr *, NFSX_FATTR(nfsd->nd_flag & ND_NFSV3)); + nfsm_srvfillattr(vap, fp); nfsm_srvdone; } @@ -169,68 +209,97 @@ nfsrv_getattr(nfsd, mrep, md, dpos, cred, nam, mrq) * nfs setattr service */ int -nfsrv_setattr(nfsd, mrep, md, dpos, cred, nam, mrq) - struct nfsd *nfsd; - struct mbuf *mrep, *md; - caddr_t dpos; - struct ucred *cred; - struct mbuf *nam, **mrq; +nfsrv_setattr(nfsd, slp, procp, mrq) + struct nfsrv_descript *nfsd; + struct nfssvc_sock *slp; + struct proc *procp; + struct mbuf **mrq; { - struct vattr va; + struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; + struct mbuf *nam = nfsd->nd_nam; + caddr_t dpos = nfsd->nd_dpos; + struct ucred *cred = &nfsd->nd_cr; + struct vattr va, preat; register struct vattr *vap = &va; register struct nfsv2_sattr *sp; - register struct nfsv2_fattr *fp; + register struct nfs_fattr *fp; struct vnode *vp; - nfsv2fh_t nfh; + nfsfh_t nfh; fhandle_t *fhp; register u_long *tl; register long t1; caddr_t bpos; - int error = 0, rdonly, cache; + int error = 0, rdonly, cache, preat_ret = 1, postat_ret = 1; + int v3 = (nfsd->nd_flag & ND_NFSV3), gcheck = 0; char *cp2; struct mbuf *mb, *mb2, *mreq; - u_quad_t frev, frev2; + u_quad_t frev; + struct timespec guard; fhp = &nfh.fh_generic; nfsm_srvmtofh(fhp); - nfsm_dissect(sp, struct nfsv2_sattr *, NFSX_SATTR(nfsd->nd_nqlflag != NQL_NOVAL)); - error = nfsrv_fhtovp(fhp, TRUE, &vp, cred, nfsd->nd_slp, nam, &rdonly); - if (error) - nfsm_reply(0); - nqsrv_getl(vp, NQL_WRITE); VATTR_NULL(vap); - /* - * Nah nah nah nah na nah - * There is a bug in the Sun client that puts 0xffff in the mode - * field of sattr when it should put in 0xffffffff. The u_short - * doesn't sign extend. - * --> check the low order 2 bytes for 0xffff - */ - if ((fxdr_unsigned(int, sp->sa_mode) & 0xffff) != 0xffff) - vap->va_mode = nfstov_mode(sp->sa_mode); - if (sp->sa_uid != nfs_xdrneg1) - vap->va_uid = fxdr_unsigned(uid_t, sp->sa_uid); - if (sp->sa_gid != nfs_xdrneg1) - vap->va_gid = fxdr_unsigned(gid_t, sp->sa_gid); - if (nfsd->nd_nqlflag == NQL_NOVAL) { - if (sp->sa_nfssize != nfs_xdrneg1) - vap->va_size = fxdr_unsigned(u_quad_t, sp->sa_nfssize); - if (sp->sa_nfsatime.nfs_sec != nfs_xdrneg1) { + if (v3) { + nfsm_srvsattr(vap); + nfsm_dissect(tl, u_long *, NFSX_UNSIGNED); + gcheck = fxdr_unsigned(int, *tl); + if (gcheck) { + nfsm_dissect(tl, u_long *, 2 * NFSX_UNSIGNED); + fxdr_nfsv3time(tl, &guard); + } + } else { + nfsm_dissect(sp, struct nfsv2_sattr *, NFSX_V2SATTR); + /* + * Nah nah nah nah na nah + * There is a bug in the Sun client that puts 0xffff in the mode + * field of sattr when it should put in 0xffffffff. The u_short + * doesn't sign extend. + * --> check the low order 2 bytes for 0xffff + */ + if ((fxdr_unsigned(int, sp->sa_mode) & 0xffff) != 0xffff) + vap->va_mode = nfstov_mode(sp->sa_mode); + if (sp->sa_uid != nfs_xdrneg1) + vap->va_uid = fxdr_unsigned(uid_t, sp->sa_uid); + if (sp->sa_gid != nfs_xdrneg1) + vap->va_gid = fxdr_unsigned(gid_t, sp->sa_gid); + if (sp->sa_size != nfs_xdrneg1) + vap->va_size = fxdr_unsigned(u_quad_t, sp->sa_size); + if (sp->sa_atime.nfsv2_sec != nfs_xdrneg1) { #ifdef notyet - fxdr_nfstime(&sp->sa_nfsatime, &vap->va_atime); + fxdr_nfsv2time(&sp->sa_atime, &vap->va_atime); #else vap->va_atime.ts_sec = - fxdr_unsigned(long, sp->sa_nfsatime.nfs_sec); + fxdr_unsigned(long, sp->sa_atime.nfsv2_sec); vap->va_atime.ts_nsec = 0; #endif } - if (sp->sa_nfsmtime.nfs_sec != nfs_xdrneg1) - fxdr_nfstime(&sp->sa_nfsmtime, &vap->va_mtime); - } else { - fxdr_hyper(&sp->sa_nqsize, &vap->va_size); - fxdr_nqtime(&sp->sa_nqatime, &vap->va_atime); - fxdr_nqtime(&sp->sa_nqmtime, &vap->va_mtime); - vap->va_flags = fxdr_unsigned(u_long, sp->sa_nqflags); + if (sp->sa_mtime.nfsv2_sec != nfs_xdrneg1) + fxdr_nfsv2time(&sp->sa_mtime, &vap->va_mtime); + + } + + /* + * Now that we have all the fields, lets do it. + */ + if (error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam, + &rdonly, (nfsd->nd_flag & ND_KERBAUTH))) { + nfsm_reply(2 * NFSX_UNSIGNED); + nfsm_srvwcc_data(preat_ret, &preat, postat_ret, vap); + return (0); + } + nqsrv_getl(vp, ND_WRITE); + if (v3) { + error = preat_ret = VOP_GETATTR(vp, &preat, cred, procp); + if (!error && gcheck && + (preat.va_ctime.ts_sec != guard.ts_sec || + preat.va_ctime.ts_nsec != guard.ts_nsec)) + error = NFSERR_NOT_SYNC; + if (error) { + nfsrv_vput(vp); + nfsm_reply(NFSX_WCCDATA(v3)); + nfsm_srvwcc_data(preat_ret, &preat, postat_ret, vap); + return (0); + } } /* @@ -246,27 +315,23 @@ nfsrv_setattr(nfsd, mrep, md, dpos, cred, nam, mrq) if (vp->v_type == VDIR) { error = EISDIR; goto out; - } else { - error = nfsrv_access(vp, VWRITE, cred, rdonly, - nfsd->nd_procp); - if (error) - goto out; - } - } - error = VOP_SETATTR(vp, vap, cred, nfsd->nd_procp); - if (error) { - nfsrv_vput(vp); - nfsm_reply(0); + } else if (error = nfsrv_access(vp, VWRITE, cred, rdonly, + procp)) + goto out; } - error = VOP_GETATTR(vp, vap, cred, nfsd->nd_procp); + error = VOP_SETATTR(vp, vap, cred, procp); + postat_ret = VOP_GETATTR(vp, vap, cred, procp); + if (!error) + error = postat_ret; out: nfsrv_vput(vp); - nfsm_reply(NFSX_FATTR(nfsd->nd_nqlflag != NQL_NOVAL) + 2*NFSX_UNSIGNED); - nfsm_build(fp, struct nfsv2_fattr *, NFSX_FATTR(nfsd->nd_nqlflag != NQL_NOVAL)); - nfsm_srvfillattr; - if (nfsd->nd_nqlflag != NQL_NOVAL) { - nfsm_build(tl, u_long *, 2*NFSX_UNSIGNED); - txdr_hyper(&frev2, tl); + nfsm_reply(NFSX_WCCORFATTR(v3)); + if (v3) { + nfsm_srvwcc_data(preat_ret, &preat, postat_ret, vap); + return (0); + } else { + nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR); + nfsm_srvfillattr(vap, fp); } nfsm_srvdone; } @@ -275,76 +340,74 @@ out: * nfs lookup rpc */ int -nfsrv_lookup(nfsd, mrep, md, dpos, cred, nam, mrq) - struct nfsd *nfsd; - struct mbuf *mrep, *md; - caddr_t dpos; - struct ucred *cred; - struct mbuf *nam, **mrq; +nfsrv_lookup(nfsd, slp, procp, mrq) + struct nfsrv_descript *nfsd; + struct nfssvc_sock *slp; + struct proc *procp; + struct mbuf **mrq; { - register struct nfsv2_fattr *fp; + struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; + struct mbuf *nam = nfsd->nd_nam; + caddr_t dpos = nfsd->nd_dpos; + struct ucred *cred = &nfsd->nd_cr; + register struct nfs_fattr *fp; struct nameidata nd; - struct vnode *vp; - nfsv2fh_t nfh; + struct vnode *vp, *dirp; + nfsfh_t nfh; fhandle_t *fhp; register caddr_t cp; register u_long *tl; register long t1; caddr_t bpos; - int error = 0, cache, cache2, len; - u_long duration2; + int error = 0, cache, len, dirattr_ret = 1; + int v3 = (nfsd->nd_flag & ND_NFSV3); char *cp2; struct mbuf *mb, *mb2, *mreq; - struct vattr va, *vap = &va; - u_quad_t frev, frev2; + struct vattr va, dirattr, *vap = &va; + u_quad_t frev; fhp = &nfh.fh_generic; - duration2 = 0; - if (nfsd->nd_nqlflag != NQL_NOVAL) { - nfsm_dissect(tl, u_long *, NFSX_UNSIGNED); - duration2 = fxdr_unsigned(int, *tl); - } nfsm_srvmtofh(fhp); - nfsm_srvstrsiz(len, NFS_MAXNAMLEN); + nfsm_srvnamesiz(len); nd.ni_cnd.cn_cred = cred; nd.ni_cnd.cn_nameiop = LOOKUP; nd.ni_cnd.cn_flags = LOCKLEAF | SAVESTART; - error = nfs_namei(&nd, fhp, len, nfsd->nd_slp, nam, &md, &dpos, - nfsd->nd_procp); - if (error) - nfsm_reply(0); - nqsrv_getl(nd.ni_startdir, NQL_READ); + error = nfs_namei(&nd, fhp, len, slp, nam, &md, &dpos, + &dirp, procp, (nfsd->nd_flag & ND_KERBAUTH)); + if (dirp) { + if (v3) + dirattr_ret = VOP_GETATTR(dirp, &dirattr, cred, + procp); + nfsrv_vrele(dirp); + } + if (error) { + nfsm_reply(NFSX_POSTOPATTR(v3)); + nfsm_srvpostop_attr(dirattr_ret, &dirattr); + return (0); + } + nqsrv_getl(nd.ni_startdir, ND_READ); nfsrv_vrele(nd.ni_startdir); FREE(nd.ni_cnd.cn_pnbuf, M_NAMEI); vp = nd.ni_vp; bzero((caddr_t)fhp, sizeof(nfh)); fhp->fh_fsid = vp->v_mount->mnt_stat.f_fsid; error = VFS_VPTOFH(vp, &fhp->fh_fid); + if (!error) + error = VOP_GETATTR(vp, vap, cred, procp); + nfsrv_vput(vp); + nfsm_reply(NFSX_SRVFH(v3) + NFSX_POSTOPORFATTR(v3) + NFSX_POSTOPATTR(v3)); if (error) { - nfsrv_vput(vp); - nfsm_reply(0); + nfsm_srvpostop_attr(dirattr_ret, &dirattr); + return (0); } - if (duration2) - (void) nqsrv_getlease(vp, &duration2, NQL_READ, nfsd, - nam, &cache2, &frev2, cred); - error = VOP_GETATTR(vp, vap, cred, nfsd->nd_procp); - nfsrv_vput(vp); - nfsm_reply(NFSX_FH + NFSX_FATTR(nfsd->nd_nqlflag != NQL_NOVAL) + 5*NFSX_UNSIGNED); - if (nfsd->nd_nqlflag != NQL_NOVAL) { - if (duration2) { - nfsm_build(tl, u_long *, 5*NFSX_UNSIGNED); - *tl++ = txdr_unsigned(NQL_READ); - *tl++ = txdr_unsigned(cache2); - *tl++ = txdr_unsigned(duration2); - txdr_hyper(&frev2, tl); - } else { - nfsm_build(tl, u_long *, NFSX_UNSIGNED); - *tl = 0; - } + nfsm_srvfhtom(fhp, v3); + if (v3) { + nfsm_srvpostop_attr(0, vap); + nfsm_srvpostop_attr(dirattr_ret, &dirattr); + } else { + nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR); + nfsm_srvfillattr(vap, fp); } - nfsm_srvfhtom(fhp); - nfsm_build(fp, struct nfsv2_fattr *, NFSX_FATTR(nfsd->nd_nqlflag != NQL_NOVAL)); - nfsm_srvfillattr; nfsm_srvdone; } @@ -352,28 +415,36 @@ nfsrv_lookup(nfsd, mrep, md, dpos, cred, nam, mrq) * nfs readlink service */ int -nfsrv_readlink(nfsd, mrep, md, dpos, cred, nam, mrq) - struct nfsd *nfsd; - struct mbuf *mrep, *md; - caddr_t dpos; - struct ucred *cred; - struct mbuf *nam, **mrq; +nfsrv_readlink(nfsd, slp, procp, mrq) + struct nfsrv_descript *nfsd; + struct nfssvc_sock *slp; + struct proc *procp; + struct mbuf **mrq; { + struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; + struct mbuf *nam = nfsd->nd_nam; + caddr_t dpos = nfsd->nd_dpos; + struct ucred *cred = &nfsd->nd_cr; struct iovec iv[(NFS_MAXPATHLEN+MLEN-1)/MLEN]; register struct iovec *ivp = iv; register struct mbuf *mp; register u_long *tl; register long t1; caddr_t bpos; - int error = 0, rdonly, cache, i, tlen, len; + int error = 0, rdonly, cache, i, tlen, len, getret; + int v3 = (nfsd->nd_flag & ND_NFSV3); char *cp2; - struct mbuf *mb, *mb2, *mp2 = 0, *mp3 = 0, *mreq; + struct mbuf *mb, *mb2, *mp2, *mp3, *mreq; struct vnode *vp; - nfsv2fh_t nfh; + struct vattr attr; + nfsfh_t nfh; fhandle_t *fhp; struct uio io, *uiop = &io; u_quad_t frev; +#ifndef nolint + mp2 = mp3 = (struct mbuf *)0; +#endif fhp = &nfh.fh_generic; nfsm_srvmtofh(fhp); len = 0; @@ -405,22 +476,33 @@ nfsrv_readlink(nfsd, mrep, md, dpos, cred, nam, mrq) uiop->uio_rw = UIO_READ; uiop->uio_segflg = UIO_SYSSPACE; uiop->uio_procp = (struct proc *)0; - error = nfsrv_fhtovp(fhp, TRUE, &vp, cred, nfsd->nd_slp, nam, &rdonly); - if (error) { + if (error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam, + &rdonly, (nfsd->nd_flag & ND_KERBAUTH))) { m_freem(mp3); - nfsm_reply(0); + nfsm_reply(2 * NFSX_UNSIGNED); + nfsm_srvpostop_attr(1, (struct vattr *)0); + return (0); } if (vp->v_type != VLNK) { - error = EINVAL; + if (v3) + error = EINVAL; + else + error = ENXIO; goto out; } - nqsrv_getl(vp, NQL_READ); + nqsrv_getl(vp, ND_READ); error = VOP_READLINK(vp, uiop, cred); out: + getret = VOP_GETATTR(vp, &attr, cred, procp); nfsrv_vput(vp); if (error) m_freem(mp3); - nfsm_reply(NFSX_UNSIGNED); + nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_UNSIGNED); + if (v3) { + nfsm_srvpostop_attr(getret, &attr); + if (error) + return (0); + } if (uiop->uio_resid > 0) { len -= uiop->uio_resid; tlen = nfsm_rndup(len); @@ -436,26 +518,31 @@ out: * nfs read service */ int -nfsrv_read(nfsd, mrep, md, dpos, cred, nam, mrq) - struct nfsd *nfsd; - struct mbuf *mrep, *md; - caddr_t dpos; - struct ucred *cred; - struct mbuf *nam, **mrq; +nfsrv_read(nfsd, slp, procp, mrq) + struct nfsrv_descript *nfsd; + struct nfssvc_sock *slp; + struct proc *procp; + struct mbuf **mrq; { + struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; + struct mbuf *nam = nfsd->nd_nam; + caddr_t dpos = nfsd->nd_dpos; + struct ucred *cred = &nfsd->nd_cr; register struct iovec *iv; struct iovec *iv2; register struct mbuf *m; - register struct nfsv2_fattr *fp; + register struct nfs_fattr *fp; register u_long *tl; register long t1; + register int i; caddr_t bpos; - int error = 0, rdonly, cache, i, cnt, len, left, siz, tlen; + int error = 0, rdonly, cache, cnt, len, left, siz, tlen, getret; + int v3 = (nfsd->nd_flag & ND_NFSV3), reqlen; char *cp2; struct mbuf *mb, *mb2, *mreq; struct mbuf *m2; struct vnode *vp; - nfsv2fh_t nfh; + nfsfh_t nfh; fhandle_t *fhp; struct uio io, *uiop = &io; struct vattr va, *vap = &va; @@ -464,40 +551,57 @@ nfsrv_read(nfsd, mrep, md, dpos, cred, nam, mrq) fhp = &nfh.fh_generic; nfsm_srvmtofh(fhp); - if (nfsd->nd_nqlflag == NQL_NOVAL) { - nfsm_dissect(tl, u_long *, NFSX_UNSIGNED); - off = (off_t)fxdr_unsigned(u_long, *tl); - } else { + if (v3) { nfsm_dissect(tl, u_long *, 2 * NFSX_UNSIGNED); fxdr_hyper(tl, &off); + } else { + nfsm_dissect(tl, u_long *, NFSX_UNSIGNED); + off = (off_t)fxdr_unsigned(u_long, *tl); + } + nfsm_srvstrsiz(reqlen, NFS_SRVMAXDATA(nfsd)); + if (error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam, + &rdonly, (nfsd->nd_flag & ND_KERBAUTH))) { + nfsm_reply(2 * NFSX_UNSIGNED); + nfsm_srvpostop_attr(1, (struct vattr *)0); + return (0); } - nfsm_srvstrsiz(cnt, NFS_MAXDATA); - error = nfsrv_fhtovp(fhp, TRUE, &vp, cred, nfsd->nd_slp, nam, &rdonly); - if (error) - nfsm_reply(0); if (vp->v_type != VREG) { - error = (vp->v_type == VDIR) ? EISDIR : EACCES; - nfsrv_vput(vp); - nfsm_reply(0); + if (v3) + error = EINVAL; + else + error = (vp->v_type == VDIR) ? EISDIR : EACCES; } - nqsrv_getl(vp, NQL_READ); - if ((error = nfsrv_access(vp, VREAD, cred, rdonly, nfsd->nd_procp)) && - (error = nfsrv_access(vp, VEXEC, cred, rdonly, nfsd->nd_procp))) { - nfsrv_vput(vp); - nfsm_reply(0); + if (!error) { + nqsrv_getl(vp, ND_READ); + if (error = nfsrv_access(vp, VREAD, cred, rdonly, procp)) + error = nfsrv_access(vp, VEXEC, cred, rdonly, procp); } - error = VOP_GETATTR(vp, vap, cred, nfsd->nd_procp); + getret = VOP_GETATTR(vp, vap, cred, procp); + if (!error) + error = getret; if (error) { nfsrv_vput(vp); - nfsm_reply(0); + nfsm_reply(NFSX_POSTOPATTR(v3)); + nfsm_srvpostop_attr(getret, vap); + return (0); } if (off >= vap->va_size) cnt = 0; - else if ((off + cnt) > vap->va_size) + else if ((off + reqlen) > vap->va_size) cnt = nfsm_rndup(vap->va_size - off); - nfsm_reply(NFSX_FATTR(nfsd->nd_nqlflag != NQL_NOVAL)+NFSX_UNSIGNED+nfsm_rndup(cnt)); - nfsm_build(fp, struct nfsv2_fattr *, NFSX_FATTR(nfsd->nd_nqlflag != NQL_NOVAL)); - nfsm_build(tl, u_long *, NFSX_UNSIGNED); + else + cnt = reqlen; + nfsm_reply(NFSX_POSTOPORFATTR(v3) + 3 * NFSX_UNSIGNED+nfsm_rndup(cnt)); + if (v3) { + nfsm_build(tl, u_long *, NFSX_V3FATTR + 4 * NFSX_UNSIGNED); + *tl++ = nfs_true; + fp = (struct nfs_fattr *)tl; + tl += (NFSX_V3FATTR / sizeof (u_long)); + } else { + nfsm_build(tl, u_long *, NFSX_V2FATTR + NFSX_UNSIGNED); + fp = (struct nfs_fattr *)tl; + tl += (NFSX_V2FATTR / sizeof (u_long)); + } len = left = cnt; if (cnt > 0) { /* @@ -505,19 +609,11 @@ nfsrv_read(nfsd, mrep, md, dpos, cred, nam, mrq) */ i = 0; m = m2 = mb; - MALLOC(iv, struct iovec *, - ((NFS_MAXDATA+MLEN-1)/MLEN) * sizeof (struct iovec), - M_TEMP, M_WAITOK); - iv2 = iv; while (left > 0) { siz = min(M_TRAILINGSPACE(m), left); if (siz > 0) { - m->m_len += siz; - iv->iov_base = bpos; - iv->iov_len = siz; - iv++; - i++; left -= siz; + i++; } if (left > 0) { MGET(m, M_WAIT, MT_DATA); @@ -525,10 +621,28 @@ nfsrv_read(nfsd, mrep, md, dpos, cred, nam, mrq) m->m_len = 0; m2->m_next = m; m2 = m; - bpos = mtod(m, caddr_t); } } - uiop->uio_iov = iv2; + MALLOC(iv, struct iovec *, i * sizeof (struct iovec), + M_TEMP, M_WAITOK); + uiop->uio_iov = iv2 = iv; + m = mb; + left = cnt; + i = 0; + while (left > 0) { + if (m == NULL) + panic("nfsrv_read iov"); + siz = min(M_TRAILINGSPACE(m), left); + if (siz > 0) { + iv->iov_base = mtod(m, caddr_t) + m->m_len; + iv->iov_len = siz; + m->m_len += siz; + left -= siz; + iv++; + i++; + } + m = m->m_next; + } uiop->uio_iovcnt = i; uiop->uio_offset = off; uiop->uio_resid = cnt; @@ -537,19 +651,30 @@ nfsrv_read(nfsd, mrep, md, dpos, cred, nam, mrq) error = VOP_READ(vp, uiop, IO_NODELOCKED, cred); off = uiop->uio_offset; FREE((caddr_t)iv2, M_TEMP); - if (error || (error = VOP_GETATTR(vp, vap, cred, nfsd->nd_procp))) { + if (error || (getret = VOP_GETATTR(vp, vap, cred, procp))) { + if (!error) + error = getret; m_freem(mreq); nfsrv_vput(vp); - nfsm_reply(0); + nfsm_reply(NFSX_POSTOPATTR(v3)); + nfsm_srvpostop_attr(getret, vap); + return (0); } } else uiop->uio_resid = 0; nfsrv_vput(vp); - nfsm_srvfillattr; + nfsm_srvfillattr(vap, fp); len -= uiop->uio_resid; tlen = nfsm_rndup(len); if (cnt != tlen || tlen != len) - nfsm_adj(mb, cnt-tlen, tlen-len); + nfsm_adj(mb, cnt - tlen, tlen - len); + if (v3) { + *tl++ = txdr_unsigned(len); + if (len < reqlen) + *tl++ = nfs_true; + else + *tl++ = nfs_false; + } *tl = txdr_unsigned(len); nfsm_srvdone; } @@ -558,132 +683,580 @@ nfsrv_read(nfsd, mrep, md, dpos, cred, nam, mrq) * nfs write service */ int -nfsrv_write(nfsd, mrep, md, dpos, cred, nam, mrq) - struct nfsd *nfsd; - struct mbuf *mrep, *md; - caddr_t dpos; - struct ucred *cred; - struct mbuf *nam, **mrq; +nfsrv_write(nfsd, slp, procp, mrq) + struct nfsrv_descript *nfsd; + struct nfssvc_sock *slp; + struct proc *procp; + struct mbuf **mrq; { + struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; + struct mbuf *nam = nfsd->nd_nam; + caddr_t dpos = nfsd->nd_dpos; + struct ucred *cred = &nfsd->nd_cr; register struct iovec *ivp; + register int i, cnt; register struct mbuf *mp; - register struct nfsv2_fattr *fp; - struct iovec iv[NFS_MAXIOVEC]; - struct vattr va; + register struct nfs_fattr *fp; + struct iovec *iv; + struct vattr va, forat; register struct vattr *vap = &va; register u_long *tl; register long t1; caddr_t bpos; - int error = 0, rdonly, cache, siz, len, xfer; - int ioflags = IO_SYNC | IO_NODELOCKED; + int error = 0, rdonly, cache, siz, len, xfer, forat_ret = 1; + int ioflags, aftat_ret = 1, retlen, zeroing, adjust; + int stable = NFSV3WRITE_FILESYNC; + int v3 = (nfsd->nd_flag & ND_NFSV3); char *cp2; struct mbuf *mb, *mb2, *mreq; struct vnode *vp; - nfsv2fh_t nfh; + nfsfh_t nfh; fhandle_t *fhp; struct uio io, *uiop = &io; off_t off; u_quad_t frev; + if (mrep == NULL) { + *mrq = NULL; + return (0); + } fhp = &nfh.fh_generic; nfsm_srvmtofh(fhp); - nfsm_dissect(tl, u_long *, 4 * NFSX_UNSIGNED); - if (nfsd->nd_nqlflag == NQL_NOVAL) { - off = (off_t)fxdr_unsigned(u_long, *++tl); - tl += 2; - } else { + if (v3) { + nfsm_dissect(tl, u_long *, 5 * NFSX_UNSIGNED); fxdr_hyper(tl, &off); + tl += 3; + stable = fxdr_unsigned(int, *tl++); + } else { + nfsm_dissect(tl, u_long *, 4 * NFSX_UNSIGNED); + off = (off_t)fxdr_unsigned(u_long, *++tl); tl += 2; - if (fxdr_unsigned(u_long, *tl++)) - ioflags |= IO_APPEND; } - len = fxdr_unsigned(long, *tl); - if (len > NFS_MAXDATA || len <= 0) { - error = EBADRPC; - nfsm_reply(0); - } - if (dpos == (mtod(md, caddr_t)+md->m_len)) { - mp = md->m_next; - if (mp == NULL) { - error = EBADRPC; - nfsm_reply(0); + retlen = len = fxdr_unsigned(long, *tl); + cnt = i = 0; + + /* + * For NFS Version 2, it is not obvious what a write of zero length + * should do, but I might as well be consistent with Version 3, + * which is to return ok so long as there are no permission problems. + */ + if (len > 0) { + zeroing = 1; + mp = mrep; + while (mp) { + if (mp == md) { + zeroing = 0; + adjust = dpos - mtod(mp, caddr_t); + mp->m_len -= adjust; + if (mp->m_len > 0 && adjust > 0) + NFSMADV(mp, adjust); } - } else { - mp = md; - siz = dpos-mtod(mp, caddr_t); - mp->m_len -= siz; - NFSMADV(mp, siz); + if (zeroing) + mp->m_len = 0; + else if (mp->m_len > 0) { + i += mp->m_len; + if (i > len) { + mp->m_len -= (i - len); + zeroing = 1; + } + if (mp->m_len > 0) + cnt++; + } + mp = mp->m_next; + } } - error = nfsrv_fhtovp(fhp, TRUE, &vp, cred, nfsd->nd_slp, nam, &rdonly); - if (error) - nfsm_reply(0); + if (len > NFS_MAXDATA || len < 0 || i < len) { + error = EIO; + nfsm_reply(2 * NFSX_UNSIGNED); + nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, vap); + return (0); + } + if (error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam, + &rdonly, (nfsd->nd_flag & ND_KERBAUTH))) { + nfsm_reply(2 * NFSX_UNSIGNED); + nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, vap); + return (0); + } + if (v3) + forat_ret = VOP_GETATTR(vp, &forat, cred, procp); if (vp->v_type != VREG) { - error = (vp->v_type == VDIR) ? EISDIR : EACCES; - nfsrv_vput(vp); - nfsm_reply(0); + if (v3) + error = EINVAL; + else + error = (vp->v_type == VDIR) ? EISDIR : EACCES; + } + if (!error) { + nqsrv_getl(vp, ND_WRITE); + error = nfsrv_access(vp, VWRITE, cred, rdonly, procp); } - nqsrv_getl(vp, NQL_WRITE); - error = nfsrv_access(vp, VWRITE, cred, rdonly, nfsd->nd_procp); if (error) { nfsrv_vput(vp); - nfsm_reply(0); + nfsm_reply(NFSX_WCCDATA(v3)); + nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, vap); + return (0); } - uiop->uio_resid = 0; - uiop->uio_rw = UIO_WRITE; - uiop->uio_segflg = UIO_SYSSPACE; - uiop->uio_procp = (struct proc *)0; + + if (len > 0) { + MALLOC(ivp, struct iovec *, cnt * sizeof (struct iovec), M_TEMP, + M_WAITOK); + uiop->uio_iov = iv = ivp; + uiop->uio_iovcnt = cnt; + mp = mrep; + while (mp) { + if (mp->m_len > 0) { + ivp->iov_base = mtod(mp, caddr_t); + ivp->iov_len = mp->m_len; + ivp++; + } + mp = mp->m_next; + } + + /* + * XXX + * The IO_METASYNC flag indicates that all metadata (and not just + * enough to ensure data integrity) mus be written to stable storage + * synchronously. + * (IO_METASYNC is not yet implemented in 4.4BSD-Lite.) + */ + if (stable == NFSV3WRITE_UNSTABLE) + ioflags = IO_NODELOCKED; + else if (stable == NFSV3WRITE_DATASYNC) + ioflags = (IO_SYNC | IO_NODELOCKED); + else + ioflags = (IO_METASYNC | IO_SYNC | IO_NODELOCKED); + uiop->uio_resid = len; + uiop->uio_rw = UIO_WRITE; + uiop->uio_segflg = UIO_SYSSPACE; + uiop->uio_procp = (struct proc *)0; + uiop->uio_offset = off; + error = VOP_WRITE(vp, uiop, ioflags, cred); + nfsstats.srvvop_writes++; + FREE((caddr_t)iv, M_TEMP); + } + aftat_ret = VOP_GETATTR(vp, vap, cred, procp); + nfsrv_vput(vp); + if (!error) + error = aftat_ret; + nfsm_reply(NFSX_PREOPATTR(v3) + NFSX_POSTOPORFATTR(v3) + + 2 * NFSX_UNSIGNED + NFSX_WRITEVERF(v3)); + if (v3) { + nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, vap); + if (error) + return (0); + nfsm_build(tl, u_long *, 4 * NFSX_UNSIGNED); + *tl++ = txdr_unsigned(retlen); + if (stable == NFSV3WRITE_UNSTABLE) + *tl++ = txdr_unsigned(stable); + else + *tl++ = txdr_unsigned(NFSV3WRITE_FILESYNC); + /* + * Actually, there is no need to txdr these fields, + * but it may make the values more human readable, + * for debugging purposes. + */ + *tl++ = txdr_unsigned(boottime.tv_sec); + *tl = txdr_unsigned(boottime.tv_usec); + } else { + nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR); + nfsm_srvfillattr(vap, fp); + } + nfsm_srvdone; +} + +/* + * NFS write service with write gathering support. Called when + * nfsrvw_procrastinate > 0. + * See: Chet Juszczak, "Improving the Write Performance of an NFS Server", + * in Proc. of the Winter 1994 Usenix Conference, pg. 247-259, San Franscisco, + * Jan. 1994. + */ +int +nfsrv_writegather(ndp, slp, procp, mrq) + struct nfsrv_descript **ndp; + struct nfssvc_sock *slp; + struct proc *procp; + struct mbuf **mrq; +{ + register struct iovec *ivp; + register struct mbuf *mp; + register struct nfsrv_descript *wp, *nfsd, *owp, *swp; + register struct nfs_fattr *fp; + register int i; + struct iovec *iov; + struct nfsrvw_delayhash *wpp; + struct ucred *cred; + struct vattr va, forat; + register u_long *tl; + register long t1; + caddr_t bpos, dpos; + int error = 0, rdonly, cache, len, forat_ret = 1; + int ioflags, aftat_ret = 1, s, adjust, v3, zeroing; + char *cp2; + struct mbuf *mb, *mb2, *mreq, *mrep, *md; + struct vnode *vp; + struct uio io, *uiop = &io; + off_t off; + u_quad_t frev, cur_usec; + +#ifndef nolint + i = 0; + len = 0; +#endif + *mrq = NULL; + if (*ndp) { + nfsd = *ndp; + *ndp = NULL; + mrep = nfsd->nd_mrep; + md = nfsd->nd_md; + dpos = nfsd->nd_dpos; + cred = &nfsd->nd_cr; + v3 = (nfsd->nd_flag & ND_NFSV3); + LIST_INIT(&nfsd->nd_coalesce); + nfsd->nd_mreq = NULL; + nfsd->nd_stable = NFSV3WRITE_FILESYNC; + cur_usec = (u_quad_t)time.tv_sec * 1000000 + (u_quad_t)time.tv_usec; + nfsd->nd_time = cur_usec + nfsrvw_procrastinate; + + /* + * Now, get the write header.. + */ + nfsm_srvmtofh(&nfsd->nd_fh); + if (v3) { + nfsm_dissect(tl, u_long *, 5 * NFSX_UNSIGNED); + fxdr_hyper(tl, &nfsd->nd_off); + tl += 3; + nfsd->nd_stable = fxdr_unsigned(int, *tl++); + } else { + nfsm_dissect(tl, u_long *, 4 * NFSX_UNSIGNED); + nfsd->nd_off = (off_t)fxdr_unsigned(u_long, *++tl); + tl += 2; + } + len = fxdr_unsigned(long, *tl); + nfsd->nd_len = len; + nfsd->nd_eoff = nfsd->nd_off + len; + + /* + * Trim the header out of the mbuf list and trim off any trailing + * junk so that the mbuf list has only the write data. + */ + zeroing = 1; + i = 0; + mp = mrep; + while (mp) { + if (mp == md) { + zeroing = 0; + adjust = dpos - mtod(mp, caddr_t); + mp->m_len -= adjust; + if (mp->m_len > 0 && adjust > 0) + NFSMADV(mp, adjust); + } + if (zeroing) + mp->m_len = 0; + else { + i += mp->m_len; + if (i > len) { + mp->m_len -= (i - len); + zeroing = 1; + } + } + mp = mp->m_next; + } + if (len > NFS_MAXDATA || len < 0 || i < len) { +nfsmout: + m_freem(mrep); + error = EIO; + nfsm_writereply(2 * NFSX_UNSIGNED, v3); + if (v3) + nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va); + nfsd->nd_mreq = mreq; + nfsd->nd_mrep = NULL; + nfsd->nd_time = 0; + } + + /* + * Add this entry to the hash and time queues. + */ + s = splsoftclock(); + owp = NULL; + wp = slp->ns_tq.lh_first; + while (wp && wp->nd_time < nfsd->nd_time) { + owp = wp; + wp = wp->nd_tq.le_next; + } + if (owp) { + LIST_INSERT_AFTER(owp, nfsd, nd_tq); + } else { + LIST_INSERT_HEAD(&slp->ns_tq, nfsd, nd_tq); + } + if (nfsd->nd_mrep) { + wpp = NWDELAYHASH(slp, nfsd->nd_fh.fh_fid.fid_data); + owp = NULL; + wp = wpp->lh_first; + while (wp && + bcmp((caddr_t)&nfsd->nd_fh,(caddr_t)&wp->nd_fh,NFSX_V3FH)) { + owp = wp; + wp = wp->nd_hash.le_next; + } + while (wp && wp->nd_off < nfsd->nd_off && + !bcmp((caddr_t)&nfsd->nd_fh,(caddr_t)&wp->nd_fh,NFSX_V3FH)) { + owp = wp; + wp = wp->nd_hash.le_next; + } + if (owp) { + LIST_INSERT_AFTER(owp, nfsd, nd_hash); + + /* + * Search the hash list for overlapping entries and + * coalesce. + */ + for(; nfsd && NFSW_CONTIG(owp, nfsd); nfsd = wp) { + wp = nfsd->nd_hash.le_next; + if (NFSW_SAMECRED(owp, nfsd)) + nfsrvw_coalesce(owp, nfsd); + } + } else { + LIST_INSERT_HEAD(wpp, nfsd, nd_hash); + } + } + splx(s); + } + /* - * Do up to NFS_MAXIOVEC mbufs of write each iteration of the - * loop until done. + * Now, do VOP_WRITE()s for any one(s) that need to be done now + * and generate the associated reply mbuf list(s). */ - while (len > 0 && uiop->uio_resid == 0) { - ivp = iv; - siz = 0; - uiop->uio_iov = ivp; - uiop->uio_iovcnt = 0; - uiop->uio_offset = off; - while (len > 0 && uiop->uio_iovcnt < NFS_MAXIOVEC && mp != NULL) { - ivp->iov_base = mtod(mp, caddr_t); - if (len < mp->m_len) - ivp->iov_len = xfer = len; +loop1: + cur_usec = (u_quad_t)time.tv_sec * 1000000 + (u_quad_t)time.tv_usec; + s = splsoftclock(); + for (nfsd = slp->ns_tq.lh_first; nfsd; nfsd = owp) { + owp = nfsd->nd_tq.le_next; + if (nfsd->nd_time > cur_usec) + break; + if (nfsd->nd_mreq) + continue; + LIST_REMOVE(nfsd, nd_tq); + LIST_REMOVE(nfsd, nd_hash); + splx(s); + mrep = nfsd->nd_mrep; + nfsd->nd_mrep = NULL; + cred = &nfsd->nd_cr; + v3 = (nfsd->nd_flag & ND_NFSV3); + forat_ret = aftat_ret = 1; + error = nfsrv_fhtovp(&nfsd->nd_fh, 1, &vp, cred, slp, + nfsd->nd_nam, &rdonly, (nfsd->nd_flag & ND_KERBAUTH)); + if (!error) { + if (v3) + forat_ret = VOP_GETATTR(vp, &forat, cred, procp); + if (vp->v_type != VREG) { + if (v3) + error = EINVAL; else - ivp->iov_len = xfer = mp->m_len; -#ifdef notdef - /* Not Yet .. */ - if (M_HASCL(mp) && (((u_long)ivp->iov_base) & CLOFSET) == 0) - ivp->iov_op = NULL; /* what should it be ?? */ - else - ivp->iov_op = NULL; -#endif - uiop->uio_iovcnt++; - ivp++; - len -= xfer; - siz += xfer; + error = (vp->v_type == VDIR) ? EISDIR : EACCES; + } + } else + vp = NULL; + if (!error) { + nqsrv_getl(vp, ND_WRITE); + error = nfsrv_access(vp, VWRITE, cred, rdonly, procp); + } + + if (nfsd->nd_stable == NFSV3WRITE_UNSTABLE) + ioflags = IO_NODELOCKED; + else if (nfsd->nd_stable == NFSV3WRITE_DATASYNC) + ioflags = (IO_SYNC | IO_NODELOCKED); + else + ioflags = (IO_METASYNC | IO_SYNC | IO_NODELOCKED); + uiop->uio_rw = UIO_WRITE; + uiop->uio_segflg = UIO_SYSSPACE; + uiop->uio_procp = (struct proc *)0; + uiop->uio_offset = nfsd->nd_off; + uiop->uio_resid = nfsd->nd_eoff - nfsd->nd_off; + if (uiop->uio_resid > 0) { + mp = mrep; + i = 0; + while (mp) { + if (mp->m_len > 0) + i++; + mp = mp->m_next; + } + uiop->uio_iovcnt = i; + MALLOC(iov, struct iovec *, i * sizeof (struct iovec), + M_TEMP, M_WAITOK); + uiop->uio_iov = ivp = iov; + mp = mrep; + while (mp) { + if (mp->m_len > 0) { + ivp->iov_base = mtod(mp, caddr_t); + ivp->iov_len = mp->m_len; + ivp++; + } mp = mp->m_next; + } + if (!error) { + error = VOP_WRITE(vp, uiop, ioflags, cred); + nfsstats.srvvop_writes++; + } + FREE((caddr_t)iov, M_TEMP); } - if (len > 0 && mp == NULL) { - error = EBADRPC; - nfsrv_vput(vp); - nfsm_reply(0); + m_freem(mrep); + if (vp) { + aftat_ret = VOP_GETATTR(vp, &va, cred, procp); + nfsrv_vput(vp); } - uiop->uio_resid = siz; - error = VOP_WRITE(vp, uiop, ioflags, cred); - if (error) { - nfsrv_vput(vp); - nfsm_reply(0); - } - off = uiop->uio_offset; + + /* + * Loop around generating replies for all write rpcs that have + * now been completed. + */ + swp = nfsd; + do { + if (error) { + nfsm_writereply(NFSX_WCCDATA(v3), v3); + if (v3) { + nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va); + } + } else { + nfsm_writereply(NFSX_PREOPATTR(v3) + + NFSX_POSTOPORFATTR(v3) + 2 * NFSX_UNSIGNED + + NFSX_WRITEVERF(v3), v3); + if (v3) { + nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va); + nfsm_build(tl, u_long *, 4 * NFSX_UNSIGNED); + *tl++ = txdr_unsigned(nfsd->nd_len); + *tl++ = txdr_unsigned(swp->nd_stable); + /* + * Actually, there is no need to txdr these fields, + * but it may make the values more human readable, + * for debugging purposes. + */ + *tl++ = txdr_unsigned(boottime.tv_sec); + *tl = txdr_unsigned(boottime.tv_usec); + } else { + nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR); + nfsm_srvfillattr(&va, fp); + } + } + nfsd->nd_mreq = mreq; + if (nfsd->nd_mrep) + panic("nfsrv_write: nd_mrep not free"); + + /* + * Done. Put it at the head of the timer queue so that + * the final phase can return the reply. + */ + s = splsoftclock(); + if (nfsd != swp) { + nfsd->nd_time = 0; + LIST_INSERT_HEAD(&slp->ns_tq, nfsd, nd_tq); + } + nfsd = swp->nd_coalesce.lh_first; + if (nfsd) { + LIST_REMOVE(nfsd, nd_tq); + } + splx(s); + } while (nfsd); + s = splsoftclock(); + swp->nd_time = 0; + LIST_INSERT_HEAD(&slp->ns_tq, swp, nd_tq); + splx(s); + goto loop1; } - error = VOP_GETATTR(vp, vap, cred, nfsd->nd_procp); - nfsrv_vput(vp); - nfsm_reply(NFSX_FATTR(nfsd->nd_nqlflag != NQL_NOVAL)); - nfsm_build(fp, struct nfsv2_fattr *, NFSX_FATTR(nfsd->nd_nqlflag != NQL_NOVAL)); - nfsm_srvfillattr; - if (nfsd->nd_nqlflag != NQL_NOVAL) { - nfsm_build(tl, u_long *, 2*NFSX_UNSIGNED); - txdr_hyper(&vap->va_filerev, tl); + splx(s); + + /* + * Search for a reply to return. + */ + s = splsoftclock(); + for (nfsd = slp->ns_tq.lh_first; nfsd; nfsd = nfsd->nd_tq.le_next) + if (nfsd->nd_mreq) { + LIST_REMOVE(nfsd, nd_tq); + *mrq = nfsd->nd_mreq; + *ndp = nfsd; + break; + } + splx(s); + return (0); +} + +/* + * Coalesce the write request nfsd into owp. To do this we must: + * - remove nfsd from the queues + * - merge nfsd->nd_mrep into owp->nd_mrep + * - update the nd_eoff and nd_stable for owp + * - put nfsd on owp's nd_coalesce list + * NB: Must be called at splsoftclock(). + */ +void +nfsrvw_coalesce(owp, nfsd) + register struct nfsrv_descript *owp; + register struct nfsrv_descript *nfsd; +{ + register int overlap; + register struct mbuf *mp; + + LIST_REMOVE(nfsd, nd_hash); + LIST_REMOVE(nfsd, nd_tq); + if (owp->nd_eoff < nfsd->nd_eoff) { + overlap = owp->nd_eoff - nfsd->nd_off; + if (overlap < 0) + panic("nfsrv_coalesce: bad off"); + if (overlap > 0) + m_adj(nfsd->nd_mrep, overlap); + mp = owp->nd_mrep; + while (mp->m_next) + mp = mp->m_next; + mp->m_next = nfsd->nd_mrep; + owp->nd_eoff = nfsd->nd_eoff; + } else + m_freem(nfsd->nd_mrep); + nfsd->nd_mrep = NULL; + if (nfsd->nd_stable == NFSV3WRITE_FILESYNC) + owp->nd_stable = NFSV3WRITE_FILESYNC; + else if (nfsd->nd_stable == NFSV3WRITE_DATASYNC && + owp->nd_stable == NFSV3WRITE_UNSTABLE) + owp->nd_stable = NFSV3WRITE_DATASYNC; + LIST_INSERT_HEAD(&owp->nd_coalesce, nfsd, nd_tq); +} + +/* + * Sort the group list in increasing numerical order. + * (Insertion sort by Chris Torek, who was grossed out by the bubble sort + * that used to be here.) + */ +void +nfsrvw_sort(list, num) + register gid_t *list; + register int num; +{ + register int i, j; + gid_t v; + + /* Insertion sort. */ + for (i = 1; i < num; i++) { + v = list[i]; + /* find correct slot for value v, moving others up */ + for (j = i; --j >= 0 && v < list[j];) + list[j + 1] = list[j]; + list[j + 1] = v; } - nfsm_srvdone; +} + +/* + * copy credentials making sure that the result can be compared with bcmp(). + */ +void +nfsrv_setcred(incred, outcred) + register struct ucred *incred, *outcred; +{ + register int i; + + bzero((caddr_t)outcred, sizeof (struct ucred)); + outcred->cr_ref = 1; + outcred->cr_uid = incred->cr_uid; + outcred->cr_ngroups = incred->cr_ngroups; + for (i = 0; i < incred->cr_ngroups; i++) + outcred->cr_groups[i] = incred->cr_groups[i]; + nfsrvw_sort(outcred->cr_groups, outcred->cr_ngroups); } /* @@ -691,15 +1264,18 @@ nfsrv_write(nfsd, mrep, md, dpos, cred, nam, mrq) * now does a truncate to 0 length via. setattr if it already exists */ int -nfsrv_create(nfsd, mrep, md, dpos, cred, nam, mrq) - struct nfsd *nfsd; - struct mbuf *mrep, *md; - caddr_t dpos; - struct ucred *cred; - struct mbuf *nam, **mrq; +nfsrv_create(nfsd, slp, procp, mrq) + struct nfsrv_descript *nfsd; + struct nfssvc_sock *slp; + struct proc *procp; + struct mbuf **mrq; { - register struct nfsv2_fattr *fp; - struct vattr va; + struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; + struct mbuf *nam = nfsd->nd_nam; + caddr_t dpos = nfsd->nd_dpos; + struct ucred *cred = &nfsd->nd_cr; + register struct nfs_fattr *fp; + struct vattr va, dirfor, diraft; register struct vattr *vap = &va; register struct nfsv2_sattr *sp; register u_long *tl; @@ -707,75 +1283,131 @@ nfsrv_create(nfsd, mrep, md, dpos, cred, nam, mrq) register caddr_t cp; register long t1; caddr_t bpos; - int error = 0, rdev, cache, len, tsize; + int error = 0, rdev, cache, len, tsize, dirfor_ret = 1, diraft_ret = 1; + int v3 = (nfsd->nd_flag & ND_NFSV3), how, exclusive_flag = 0; char *cp2; struct mbuf *mb, *mb2, *mreq; - struct vnode *vp; - nfsv2fh_t nfh; + struct vnode *vp, *dirp = (struct vnode *)0; + nfsfh_t nfh; fhandle_t *fhp; - u_quad_t frev; + u_quad_t frev, tempsize; + u_char cverf[NFSX_V3CREATEVERF]; +#ifndef nolint + rdev = 0; +#endif nd.ni_cnd.cn_nameiop = 0; fhp = &nfh.fh_generic; nfsm_srvmtofh(fhp); - nfsm_srvstrsiz(len, NFS_MAXNAMLEN); + nfsm_srvnamesiz(len); nd.ni_cnd.cn_cred = cred; nd.ni_cnd.cn_nameiop = CREATE; nd.ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF | SAVESTART; - error = nfs_namei(&nd, fhp, len, nfsd->nd_slp, nam, &md, &dpos, - nfsd->nd_procp); - if (error) - nfsm_reply(0); + error = nfs_namei(&nd, fhp, len, slp, nam, &md, &dpos, + &dirp, procp, (nfsd->nd_flag & ND_KERBAUTH)); + if (dirp) { + if (v3) + dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred, + procp); + else { + nfsrv_vrele(dirp); + dirp = (struct vnode *)0; + } + } + if (error) { + nfsm_reply(NFSX_WCCDATA(v3)); + nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); + if (dirp) + nfsrv_vrele(dirp); + return (0); + } VATTR_NULL(vap); - nfsm_dissect(sp, struct nfsv2_sattr *, NFSX_SATTR(nfsd->nd_nqlflag != NQL_NOVAL)); + if (v3) { + nfsm_dissect(tl, u_long *, NFSX_UNSIGNED); + how = fxdr_unsigned(int, *tl); + switch (how) { + case NFSV3CREATE_GUARDED: + if (nd.ni_vp) { + error = EEXIST; + break; + } + case NFSV3CREATE_UNCHECKED: + nfsm_srvsattr(vap); + break; + case NFSV3CREATE_EXCLUSIVE: + nfsm_dissect(cp, caddr_t, NFSX_V3CREATEVERF); + bcopy(cp, cverf, NFSX_V3CREATEVERF); + exclusive_flag = 1; + if (nd.ni_vp == NULL) + vap->va_mode = 0; + break; + }; + vap->va_type = VREG; + } else { + nfsm_dissect(sp, struct nfsv2_sattr *, NFSX_V2SATTR); + vap->va_type = IFTOVT(fxdr_unsigned(u_long, sp->sa_mode)); + if (vap->va_type == VNON) + vap->va_type = VREG; + vap->va_mode = nfstov_mode(sp->sa_mode); + switch (vap->va_type) { + case VREG: + tsize = fxdr_unsigned(long, sp->sa_size); + if (tsize != -1) + vap->va_size = (u_quad_t)tsize; + break; + case VCHR: + case VBLK: + case VFIFO: + rdev = fxdr_unsigned(long, sp->sa_size); + break; + }; + } + /* * Iff doesn't exist, create it * otherwise just truncate to 0 length * should I set the mode too ?? */ if (nd.ni_vp == NULL) { - vap->va_type = IFTOVT(fxdr_unsigned(u_long, sp->sa_mode)); - if (vap->va_type == VNON) - vap->va_type = VREG; - vap->va_mode = nfstov_mode(sp->sa_mode); - if (nfsd->nd_nqlflag == NQL_NOVAL) - rdev = fxdr_unsigned(long, sp->sa_nfssize); - else - rdev = fxdr_unsigned(long, sp->sa_nqrdev); if (vap->va_type == VREG || vap->va_type == VSOCK) { nfsrv_vrele(nd.ni_startdir); - nqsrv_getl(nd.ni_dvp, NQL_WRITE); - error=VOP_CREATE(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, vap); - if (error) - nfsm_reply(0); - nfsrv_vmio(nd.ni_vp); - FREE(nd.ni_cnd.cn_pnbuf, M_NAMEI); + nqsrv_getl(nd.ni_dvp, ND_WRITE); + error = VOP_CREATE(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, vap); + if (!error) { + nfsrv_vmio(nd.ni_vp); + FREE(nd.ni_cnd.cn_pnbuf, M_NAMEI); + if (exclusive_flag) { + exclusive_flag = 0; + VATTR_NULL(vap); + bcopy(cverf, (caddr_t)&vap->va_atime, + NFSX_V3CREATEVERF); + error = VOP_SETATTR(nd.ni_vp, vap, cred, + procp); + } + } } else if (vap->va_type == VCHR || vap->va_type == VBLK || vap->va_type == VFIFO) { if (vap->va_type == VCHR && rdev == 0xffffffff) vap->va_type = VFIFO; - if (vap->va_type == VFIFO) { - } else { - error = suser(cred, (u_short *)0); - if (error) { - VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); - nfsrv_vput(nd.ni_dvp); - goto out; - } else - vap->va_rdev = (dev_t)rdev; - } - nqsrv_getl(nd.ni_dvp, NQL_WRITE); - error=VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, vap); - if (error) { + if (error = suser(cred, (u_short *)0)) { + nfsrv_vrele(nd.ni_startdir); + free(nd.ni_cnd.cn_pnbuf, M_NAMEI); + VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); + nfsrv_vput(nd.ni_dvp); + nfsm_reply(0); + return (error); + } else + vap->va_rdev = (dev_t)rdev; + nqsrv_getl(nd.ni_dvp, ND_WRITE); + if (error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, vap)) { nfsrv_vrele(nd.ni_startdir); nfsm_reply(0); } nd.ni_cnd.cn_nameiop = LOOKUP; nd.ni_cnd.cn_flags &= ~(LOCKPARENT | SAVESTART); - nd.ni_cnd.cn_proc = nfsd->nd_procp; - nd.ni_cnd.cn_cred = nfsd->nd_procp->p_ucred; - error = lookup(&nd); - if (error) { + nd.ni_cnd.cn_proc = procp; + nd.ni_cnd.cn_cred = cred; + if (error = lookup(&nd)) { free(nd.ni_cnd.cn_pnbuf, M_NAMEI); nfsm_reply(0); } @@ -789,10 +1421,11 @@ nfsrv_create(nfsd, mrep, md, dpos, cred, nam, mrq) nfsm_reply(0); } } else { + nfsrv_vrele(nd.ni_startdir); + free(nd.ni_cnd.cn_pnbuf, M_NAMEI); VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); nfsrv_vput(nd.ni_dvp); error = ENXIO; - goto out; } vp = nd.ni_vp; } else { @@ -804,46 +1437,56 @@ nfsrv_create(nfsd, mrep, md, dpos, cred, nam, mrq) else nfsrv_vput(nd.ni_dvp); VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); - if (nfsd->nd_nqlflag == NQL_NOVAL) { - tsize = fxdr_unsigned(long, sp->sa_nfssize); - if (tsize != -1) - vap->va_size = (u_quad_t)tsize; - else - vap->va_size = -1; - } else - fxdr_hyper(&sp->sa_nqsize, &vap->va_size); if (vap->va_size != -1) { error = nfsrv_access(vp, VWRITE, cred, - (nd.ni_cnd.cn_flags & RDONLY), nfsd->nd_procp); - if (error) { - nfsrv_vput(vp); - nfsm_reply(0); + (nd.ni_cnd.cn_flags & RDONLY), procp); + if (!error) { + nqsrv_getl(vp, ND_WRITE); + tempsize = vap->va_size; + VATTR_NULL(vap); + vap->va_size = tempsize; + error = VOP_SETATTR(vp, vap, cred, + procp); } - nqsrv_getl(vp, NQL_WRITE); - error = VOP_SETATTR(vp, vap, cred, nfsd->nd_procp); - if (error) { + if (error) nfsrv_vput(vp); - nfsm_reply(0); - } } } - bzero((caddr_t)fhp, sizeof(nfh)); - fhp->fh_fsid = vp->v_mount->mnt_stat.f_fsid; - error = VFS_VPTOFH(vp, &fhp->fh_fid); - if (error) { + if (!error) { + bzero((caddr_t)fhp, sizeof(nfh)); + fhp->fh_fsid = vp->v_mount->mnt_stat.f_fsid; + error = VFS_VPTOFH(vp, &fhp->fh_fid); + if (!error) + error = VOP_GETATTR(vp, vap, cred, procp); nfsrv_vput(vp); - nfsm_reply(0); } - error = VOP_GETATTR(vp, vap, cred, nfsd->nd_procp); - nfsrv_vput(vp); - nfsm_reply(NFSX_FH+NFSX_FATTR(nfsd->nd_nqlflag != NQL_NOVAL)); - nfsm_srvfhtom(fhp); - nfsm_build(fp, struct nfsv2_fattr *, NFSX_FATTR(nfsd->nd_nqlflag != NQL_NOVAL)); - nfsm_srvfillattr; - return (error); + if (v3) { + if (exclusive_flag && !error && + bcmp(cverf, (caddr_t)&vap->va_atime, NFSX_V3CREATEVERF)) + error = EEXIST; + diraft_ret = VOP_GETATTR(dirp, &diraft, cred, procp); + nfsrv_vrele(dirp); + } + nfsm_reply(NFSX_SRVFH(v3) + NFSX_FATTR(v3) + NFSX_WCCDATA(v3)); + if (v3) { + if (!error) { + nfsm_srvpostop_fh(fhp); + nfsm_srvpostop_attr(0, vap); + } + nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); + } else { + nfsm_srvfhtom(fhp, v3); + nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR); + nfsm_srvfillattr(vap, fp); + } + return (0); nfsmout: - if (nd.ni_cnd.cn_nameiop || nd.ni_cnd.cn_flags) + if (dirp) + nfsrv_vrele(dirp); + if (nd.ni_cnd.cn_nameiop) { nfsrv_vrele(nd.ni_startdir); + free((caddr_t)nd.ni_cnd.cn_pnbuf, M_NAMEI); + } VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); if (nd.ni_dvp == nd.ni_vp) nfsrv_vrele(nd.ni_dvp); @@ -852,79 +1495,248 @@ nfsmout: if (nd.ni_vp) nfsrv_vput(nd.ni_vp); return (error); +} + +/* + * nfs v3 mknod service + */ +int +nfsrv_mknod(nfsd, slp, procp, mrq) + struct nfsrv_descript *nfsd; + struct nfssvc_sock *slp; + struct proc *procp; + struct mbuf **mrq; +{ + struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; + struct mbuf *nam = nfsd->nd_nam; + caddr_t dpos = nfsd->nd_dpos; + struct ucred *cred = &nfsd->nd_cr; + register struct nfs_fattr *fp; + struct vattr va, dirfor, diraft; + register struct vattr *vap = &va; + register u_long *tl; + struct nameidata nd; + register caddr_t cp; + register long t1; + caddr_t bpos; + int error = 0, cache, len, tsize, dirfor_ret = 1, diraft_ret = 1; + u_long major, minor; + enum vtype vtyp; + char *cp2; + struct mbuf *mb, *mb2, *mreq; + struct vnode *vp, *dirp = (struct vnode *)0; + nfsfh_t nfh; + fhandle_t *fhp; + u_quad_t frev; + nd.ni_cnd.cn_nameiop = 0; + fhp = &nfh.fh_generic; + nfsm_srvmtofh(fhp); + nfsm_srvnamesiz(len); + nd.ni_cnd.cn_cred = cred; + nd.ni_cnd.cn_nameiop = CREATE; + nd.ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF | SAVESTART; + error = nfs_namei(&nd, fhp, len, slp, nam, &md, &dpos, + &dirp, procp, (nfsd->nd_flag & ND_KERBAUTH)); + if (dirp) + dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred, procp); + if (error) { + nfsm_reply(NFSX_WCCDATA(1)); + nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); + if (dirp) + nfsrv_vrele(dirp); + return (0); + } + nfsm_dissect(tl, u_long *, NFSX_UNSIGNED); + vtyp = nfsv3tov_type(*tl); + if (vtyp != VCHR && vtyp != VBLK && vtyp != VSOCK && vtyp != VFIFO) { + nfsrv_vrele(nd.ni_startdir); + free((caddr_t)nd.ni_cnd.cn_pnbuf, M_NAMEI); + error = NFSERR_BADTYPE; + VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); + nfsrv_vput(nd.ni_dvp); + goto out; + } + VATTR_NULL(vap); + nfsm_srvsattr(vap); + if (vtyp == VCHR || vtyp == VBLK) { + nfsm_dissect(tl, u_long *, 2 * NFSX_UNSIGNED); + major = fxdr_unsigned(u_long, *tl++); + minor = fxdr_unsigned(u_long, *tl); + vap->va_rdev = makedev(major, minor); + } + + /* + * Iff doesn't exist, create it. + */ + if (nd.ni_vp) { + nfsrv_vrele(nd.ni_startdir); + free((caddr_t)nd.ni_cnd.cn_pnbuf, M_NAMEI); + error = EEXIST; + VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); + nfsrv_vput(nd.ni_dvp); + goto out; + } + vap->va_type = vtyp; + if (vtyp == VSOCK) { + nfsrv_vrele(nd.ni_startdir); + nqsrv_getl(nd.ni_dvp, ND_WRITE); + error = VOP_CREATE(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, vap); + if (!error) + FREE(nd.ni_cnd.cn_pnbuf, M_NAMEI); + } else { + if (error = suser(cred, (u_short *)0)) { + nfsrv_vrele(nd.ni_startdir); + free((caddr_t)nd.ni_cnd.cn_pnbuf, M_NAMEI); + VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); + nfsrv_vput(nd.ni_dvp); + goto out; + } + nqsrv_getl(nd.ni_dvp, ND_WRITE); + if (error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, vap)) { + nfsrv_vrele(nd.ni_startdir); + goto out; + } + nd.ni_cnd.cn_nameiop = LOOKUP; + nd.ni_cnd.cn_flags &= ~(LOCKPARENT | SAVESTART); + nd.ni_cnd.cn_proc = procp; + nd.ni_cnd.cn_cred = procp->p_ucred; + error = lookup(&nd); + FREE(nd.ni_cnd.cn_pnbuf, M_NAMEI); + if (error) + goto out; + if (nd.ni_cnd.cn_flags & ISSYMLINK) { + nfsrv_vrele(nd.ni_dvp); + nfsrv_vput(nd.ni_vp); + VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); + error = EINVAL; + } + } out: - nfsrv_vrele(nd.ni_startdir); - free(nd.ni_cnd.cn_pnbuf, M_NAMEI); - nfsm_reply(0); + vp = nd.ni_vp; + if (!error) { + bzero((caddr_t)fhp, sizeof(nfh)); + fhp->fh_fsid = vp->v_mount->mnt_stat.f_fsid; + error = VFS_VPTOFH(vp, &fhp->fh_fid); + if (!error) + error = VOP_GETATTR(vp, vap, cred, procp); + nfsrv_vput(vp); + } + diraft_ret = VOP_GETATTR(dirp, &diraft, cred, procp); + nfsrv_vrele(dirp); + nfsm_reply(NFSX_SRVFH(1) + NFSX_POSTOPATTR(1) + NFSX_WCCDATA(1)); + if (!error) { + nfsm_srvpostop_fh(fhp); + nfsm_srvpostop_attr(0, vap); + } + nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); return (0); +nfsmout: + if (dirp) + nfsrv_vrele(dirp); + if (nd.ni_cnd.cn_nameiop) { + nfsrv_vrele(nd.ni_startdir); + free((caddr_t)nd.ni_cnd.cn_pnbuf, M_NAMEI); + } + VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); + if (nd.ni_dvp == nd.ni_vp) + nfsrv_vrele(nd.ni_dvp); + else + nfsrv_vput(nd.ni_dvp); + if (nd.ni_vp) + nfsrv_vput(nd.ni_vp); + return (error); } /* * nfs remove service */ int -nfsrv_remove(nfsd, mrep, md, dpos, cred, nam, mrq) - struct nfsd *nfsd; - struct mbuf *mrep, *md; - caddr_t dpos; - struct ucred *cred; - struct mbuf *nam, **mrq; +nfsrv_remove(nfsd, slp, procp, mrq) + struct nfsrv_descript *nfsd; + struct nfssvc_sock *slp; + struct proc *procp; + struct mbuf **mrq; { + struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; + struct mbuf *nam = nfsd->nd_nam; + caddr_t dpos = nfsd->nd_dpos; + struct ucred *cred = &nfsd->nd_cr; struct nameidata nd; register u_long *tl; register long t1; caddr_t bpos; - int error = 0, cache, len; + int error = 0, cache, len, dirfor_ret = 1, diraft_ret = 1; + int v3 = (nfsd->nd_flag & ND_NFSV3); char *cp2; - struct mbuf *mb, *mreq; - struct vnode *vp; - nfsv2fh_t nfh; + struct mbuf *mb, *mreq, *mb2; + struct vnode *vp, *dirp; + struct vattr dirfor, diraft; + nfsfh_t nfh; fhandle_t *fhp; u_quad_t frev; +#ifndef nolint + vp = (struct vnode *)0; +#endif fhp = &nfh.fh_generic; nfsm_srvmtofh(fhp); - nfsm_srvstrsiz(len, NFS_MAXNAMLEN); + nfsm_srvnamesiz(len); nd.ni_cnd.cn_cred = cred; nd.ni_cnd.cn_nameiop = DELETE; nd.ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF; - error = nfs_namei(&nd, fhp, len, nfsd->nd_slp, nam, &md, &dpos, - nfsd->nd_procp); - if (error) - nfsm_reply(0); - vp = nd.ni_vp; - if (vp->v_type == VDIR && - (error = suser(cred, (u_short *)0))) - goto out; - /* - * The root of a mounted filesystem cannot be deleted. - */ - if (vp->v_flag & VROOT) { - error = EBUSY; - goto out; + error = nfs_namei(&nd, fhp, len, slp, nam, &md, &dpos, + &dirp, procp, (nfsd->nd_flag & ND_KERBAUTH)); + if (dirp) { + if (v3) + dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred, + procp); + else + nfsrv_vrele(dirp); } - (void) vnode_pager_uncache(vp); -out: if (!error) { - int deallocobj = 0; - nqsrv_getl(nd.ni_dvp, NQL_WRITE); - nqsrv_getl(vp, NQL_WRITE); - - if ((vp->v_flag & VVMIO) && vp->v_vmdata) - deallocobj = 1; - error = VOP_REMOVE(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); - if (error == 0 && deallocobj) - vm_object_deallocate((vm_object_t) vp->v_vmdata); - } else { - VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); - if (nd.ni_dvp == vp) - nfsrv_vrele(nd.ni_dvp); - else - nfsrv_vput(nd.ni_dvp); - nfsrv_vput(vp); + vp = nd.ni_vp; + if (vp->v_type == VDIR && + (error = suser(cred, (u_short *)0))) + goto out; + /* + * The root of a mounted filesystem cannot be deleted. + */ + if (vp->v_flag & VROOT) { + error = EBUSY; + goto out; + } + if (vp->v_flag & VTEXT) + (void) vnode_pager_uncache(vp); +out: + if (!error) { + int deallocobj = 0; + nqsrv_getl(nd.ni_dvp, ND_WRITE); + nqsrv_getl(vp, ND_WRITE); + + if ((vp->v_flag & VVMIO) && vp->v_vmdata) + deallocobj = 1; + error = VOP_REMOVE(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); + if (error == 0 && deallocobj) + vm_object_deallocate((vm_object_t) vp->v_vmdata); + } else { + VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); + if (nd.ni_dvp == vp) + nfsrv_vrele(nd.ni_dvp); + else + nfsrv_vput(nd.ni_dvp); + nfsrv_vput(vp); + } + } + if (dirp && v3) { + diraft_ret = VOP_GETATTR(dirp, &diraft, cred, procp); + nfsrv_vrele(dirp); + } + nfsm_reply(NFSX_WCCDATA(v3)); + if (v3) { + nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); + return (0); } - nfsm_reply(0); nfsm_srvdone; } @@ -932,32 +1744,42 @@ out: * nfs rename service */ int -nfsrv_rename(nfsd, mrep, md, dpos, cred, nam, mrq) - struct nfsd *nfsd; - struct mbuf *mrep, *md; - caddr_t dpos; - struct ucred *cred; - struct mbuf *nam, **mrq; +nfsrv_rename(nfsd, slp, procp, mrq) + struct nfsrv_descript *nfsd; + struct nfssvc_sock *slp; + struct proc *procp; + struct mbuf **mrq; { + struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; + struct mbuf *nam = nfsd->nd_nam; + caddr_t dpos = nfsd->nd_dpos; + struct ucred *cred = &nfsd->nd_cr; register u_long *tl; register long t1; caddr_t bpos; - int error = 0, cache, len, len2; + int error = 0, cache, len, len2, fdirfor_ret = 1, fdiraft_ret = 1; + int tdirfor_ret = 1, tdiraft_ret = 1; + int v3 = (nfsd->nd_flag & ND_NFSV3); char *cp2; - struct mbuf *mb, *mreq; + struct mbuf *mb, *mreq, *mb2; struct nameidata fromnd, tond; - struct vnode *fvp = 0, *tvp, *tdvp; - nfsv2fh_t fnfh, tnfh; + struct vnode *fvp, *tvp, *tdvp, *fdirp = (struct vnode *)0; + struct vnode *tdirp = (struct vnode *)0; + struct vattr fdirfor, fdiraft, tdirfor, tdiraft; + nfsfh_t fnfh, tnfh; fhandle_t *ffhp, *tfhp; u_quad_t frev; uid_t saved_uid; +#ifndef nolint + fvp = (struct vnode *)0; +#endif ffhp = &fnfh.fh_generic; tfhp = &tnfh.fh_generic; fromnd.ni_cnd.cn_nameiop = 0; tond.ni_cnd.cn_nameiop = 0; nfsm_srvmtofh(ffhp); - nfsm_srvstrsiz(len, NFS_MAXNAMLEN); + nfsm_srvnamesiz(len); /* * Remember our original uid so that we can reset cr_uid before * the second nfs_namei() call, in case it is remapped. @@ -966,10 +1788,25 @@ nfsrv_rename(nfsd, mrep, md, dpos, cred, nam, mrq) fromnd.ni_cnd.cn_cred = cred; fromnd.ni_cnd.cn_nameiop = DELETE; fromnd.ni_cnd.cn_flags = WANTPARENT | SAVESTART; - error = nfs_namei(&fromnd, ffhp, len, nfsd->nd_slp, nam, &md, &dpos, - nfsd->nd_procp); - if(error) - nfsm_reply(0); + error = nfs_namei(&fromnd, ffhp, len, slp, nam, &md, + &dpos, &fdirp, procp, (nfsd->nd_flag & ND_KERBAUTH)); + if (fdirp) { + if (v3) + fdirfor_ret = VOP_GETATTR(fdirp, &fdirfor, cred, + procp); + else { + nfsrv_vrele(fdirp); + fdirp = (struct vnode *)0; + } + } + if (error) { + nfsm_reply(2 * NFSX_WCCDATA(v3)); + nfsm_srvwcc_data(fdirfor_ret, &fdirfor, fdiraft_ret, &fdiraft); + nfsm_srvwcc_data(tdirfor_ret, &tdirfor, tdiraft_ret, &tdiraft); + if (fdirp) + nfsrv_vrele(fdirp); + return (0); + } fvp = fromnd.ni_vp; nfsm_srvmtofh(tfhp); nfsm_strsiz(len2, NFS_MAXNAMLEN); @@ -977,8 +1814,17 @@ nfsrv_rename(nfsd, mrep, md, dpos, cred, nam, mrq) tond.ni_cnd.cn_cred = cred; tond.ni_cnd.cn_nameiop = RENAME; tond.ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF | NOCACHE | SAVESTART; - error = nfs_namei(&tond, tfhp, len2, nfsd->nd_slp, nam, &md, - &dpos, nfsd->nd_procp); + error = nfs_namei(&tond, tfhp, len2, slp, nam, &md, + &dpos, &tdirp, procp, (nfsd->nd_flag & ND_KERBAUTH)); + if (tdirp) { + if (v3) + tdirfor_ret = VOP_GETATTR(tdirp, &tdirfor, cred, + procp); + else { + nfsrv_vrele(tdirp); + tdirp = (struct vnode *)0; + } + } if (error) { VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); nfsrv_vrele(fromnd.ni_dvp); @@ -989,27 +1835,45 @@ nfsrv_rename(nfsd, mrep, md, dpos, cred, nam, mrq) tvp = tond.ni_vp; if (tvp != NULL) { if (fvp->v_type == VDIR && tvp->v_type != VDIR) { - error = EISDIR; + if (v3) + error = EEXIST; + else + error = EISDIR; goto out; } else if (fvp->v_type != VDIR && tvp->v_type == VDIR) { - error = ENOTDIR; + if (v3) + error = EEXIST; + else + error = ENOTDIR; goto out; } if (tvp->v_type == VDIR && tvp->v_mountedhere) { - error = EXDEV; + if (v3) + error = EXDEV; + else + error = ENOTEMPTY; goto out; } } if (fvp->v_type == VDIR && fvp->v_mountedhere) { - error = EBUSY; + if (v3) + error = EXDEV; + else + error = ENOTEMPTY; goto out; } if (fvp->v_mount != tdvp->v_mount) { - error = EXDEV; + if (v3) + error = EXDEV; + else + error = ENOTEMPTY; goto out; } if (fvp == tdvp) - error = EINVAL; + if (v3) + error = EINVAL; + else + error = ENOTEMPTY; /* * If source is the same as the destination (that is the * same vnode with the same name in the same directory), @@ -1023,10 +1887,10 @@ nfsrv_rename(nfsd, mrep, md, dpos, cred, nam, mrq) out: if (!error) { int deallocobjfrom = 0, deallocobjto = 0; - nqsrv_getl(fromnd.ni_dvp, NQL_WRITE); - nqsrv_getl(tdvp, NQL_WRITE); + nqsrv_getl(fromnd.ni_dvp, ND_WRITE); + nqsrv_getl(tdvp, ND_WRITE); if (tvp) { - nqsrv_getl(tvp, NQL_WRITE); + nqsrv_getl(tvp, ND_WRITE); if ((tvp->v_flag & VVMIO) && tvp->v_vmdata) deallocobjto = 1; } @@ -1050,21 +1914,39 @@ out: VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); nfsrv_vrele(fromnd.ni_dvp); nfsrv_vrele(fvp); + if (error == -1) + error = 0; } nfsrv_vrele(tond.ni_startdir); FREE(tond.ni_cnd.cn_pnbuf, M_NAMEI); out1: + if (fdirp) { + fdiraft_ret = VOP_GETATTR(fdirp, &fdiraft, cred, procp); + nfsrv_vrele(fdirp); + } + if (tdirp) { + tdiraft_ret = VOP_GETATTR(tdirp, &tdiraft, cred, procp); + nfsrv_vrele(tdirp); + } nfsrv_vrele(fromnd.ni_startdir); FREE(fromnd.ni_cnd.cn_pnbuf, M_NAMEI); - nfsm_reply(0); - return (error); + nfsm_reply(2 * NFSX_WCCDATA(v3)); + if (v3) { + nfsm_srvwcc_data(fdirfor_ret, &fdirfor, fdiraft_ret, &fdiraft); + nfsm_srvwcc_data(tdirfor_ret, &tdirfor, tdiraft_ret, &tdiraft); + } + return (0); nfsmout: - if (tond.ni_cnd.cn_nameiop || tond.ni_cnd.cn_flags) { + if (fdirp) + nfsrv_vrele(fdirp); + if (tdirp) + nfsrv_vrele(tdirp); + if (tond.ni_cnd.cn_nameiop) { nfsrv_vrele(tond.ni_startdir); FREE(tond.ni_cnd.cn_pnbuf, M_NAMEI); } - if (fromnd.ni_cnd.cn_nameiop || fromnd.ni_cnd.cn_flags) { + if (fromnd.ni_cnd.cn_nameiop) { nfsrv_vrele(fromnd.ni_startdir); FREE(fromnd.ni_cnd.cn_pnbuf, M_NAMEI); VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); @@ -1078,22 +1960,27 @@ nfsmout: * nfs link service */ int -nfsrv_link(nfsd, mrep, md, dpos, cred, nam, mrq) - struct nfsd *nfsd; - struct mbuf *mrep, *md; - caddr_t dpos; - struct ucred *cred; - struct mbuf *nam, **mrq; +nfsrv_link(nfsd, slp, procp, mrq) + struct nfsrv_descript *nfsd; + struct nfssvc_sock *slp; + struct proc *procp; + struct mbuf **mrq; { + struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; + struct mbuf *nam = nfsd->nd_nam; + caddr_t dpos = nfsd->nd_dpos; + struct ucred *cred = &nfsd->nd_cr; struct nameidata nd; register u_long *tl; register long t1; caddr_t bpos; - int error = 0, rdonly, cache, len; + int error = 0, rdonly, cache, len, dirfor_ret = 1, diraft_ret = 1; + int getret = 1, v3 = (nfsd->nd_flag & ND_NFSV3); char *cp2; - struct mbuf *mb, *mreq; - struct vnode *vp, *xp; - nfsv2fh_t nfh, dnfh; + struct mbuf *mb, *mreq, *mb2; + struct vnode *vp, *xp, *dirp = (struct vnode *)0; + struct vattr dirfor, diraft, at; + nfsfh_t nfh, dnfh; fhandle_t *fhp, *dfhp; u_quad_t frev; @@ -1101,17 +1988,30 @@ nfsrv_link(nfsd, mrep, md, dpos, cred, nam, mrq) dfhp = &dnfh.fh_generic; nfsm_srvmtofh(fhp); nfsm_srvmtofh(dfhp); - nfsm_srvstrsiz(len, NFS_MAXNAMLEN); - error = nfsrv_fhtovp(fhp, FALSE, &vp, cred, nfsd->nd_slp, nam, &rdonly); - if (error) - nfsm_reply(0); + nfsm_srvnamesiz(len); + if (error = nfsrv_fhtovp(fhp, FALSE, &vp, cred, slp, nam, + &rdonly, (nfsd->nd_flag & ND_KERBAUTH))) { + nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_WCCDATA(v3)); + nfsm_srvpostop_attr(getret, &at); + nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); + return (0); + } if (vp->v_type == VDIR && (error = suser(cred, (u_short *)0))) goto out1; nd.ni_cnd.cn_cred = cred; nd.ni_cnd.cn_nameiop = CREATE; nd.ni_cnd.cn_flags = LOCKPARENT; - error = nfs_namei(&nd, dfhp, len, nfsd->nd_slp, nam, &md, &dpos, - nfsd->nd_procp); + error = nfs_namei(&nd, dfhp, len, slp, nam, &md, &dpos, + &dirp, procp, (nfsd->nd_flag & ND_KERBAUTH)); + if (dirp) { + if (v3) + dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred, + procp); + else { + nfsrv_vrele(dirp); + dirp = (struct vnode *)0; + } + } if (error) goto out1; xp = nd.ni_vp; @@ -1124,9 +2024,13 @@ nfsrv_link(nfsd, mrep, md, dpos, cred, nam, mrq) error = EXDEV; out: if (!error) { - nqsrv_getl(vp, NQL_WRITE); - nqsrv_getl(xp, NQL_WRITE); + nqsrv_getl(vp, ND_WRITE); + nqsrv_getl(xp, ND_WRITE); +#if defined(__FreeBSD__) || defined(__NetBSD__) error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd); +#else + error = VOP_LINK(vp, nd.ni_dvp, &nd.ni_cnd); +#endif } else { VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); if (nd.ni_dvp == nd.ni_vp) @@ -1137,8 +2041,19 @@ out: nfsrv_vrele(nd.ni_vp); } out1: + if (v3) + getret = VOP_GETATTR(vp, &at, cred, procp); + if (dirp) { + diraft_ret = VOP_GETATTR(dirp, &diraft, cred, procp); + nfsrv_vrele(dirp); + } nfsrv_vrele(vp); - nfsm_reply(0); + nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_WCCDATA(v3)); + if (v3) { + nfsm_srvpostop_attr(getret, &at); + nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); + return (0); + } nfsm_srvdone; } @@ -1146,42 +2061,58 @@ out1: * nfs symbolic link service */ int -nfsrv_symlink(nfsd, mrep, md, dpos, cred, nam, mrq) - struct nfsd *nfsd; - struct mbuf *mrep, *md; - caddr_t dpos; - struct ucred *cred; - struct mbuf *nam, **mrq; +nfsrv_symlink(nfsd, slp, procp, mrq) + struct nfsrv_descript *nfsd; + struct nfssvc_sock *slp; + struct proc *procp; + struct mbuf **mrq; { + struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; + struct mbuf *nam = nfsd->nd_nam; + caddr_t dpos = nfsd->nd_dpos; + struct ucred *cred = &nfsd->nd_cr; + struct vattr va, dirfor, diraft; struct vnode *ovp; - struct vattr va; struct nameidata nd; register struct vattr *vap = &va; register u_long *tl; register long t1; struct nfsv2_sattr *sp; - caddr_t bpos; + char *bpos, *cp, *pathcp = (char *)0, *cp2; struct uio io; struct iovec iv; - int error = 0, cache, len, len2; - char *pathcp, *cp2; - struct mbuf *mb, *mreq; - nfsv2fh_t nfh; + int error = 0, cache, len, len2, dirfor_ret = 1, diraft_ret = 1; + int v3 = (nfsd->nd_flag & ND_NFSV3); + struct mbuf *mb, *mreq, *mb2; + struct vnode *dirp = (struct vnode *)0; + nfsfh_t nfh; fhandle_t *fhp; u_quad_t frev; int deallocobj = 0; - pathcp = (char *)0; + nd.ni_cnd.cn_nameiop = 0; fhp = &nfh.fh_generic; nfsm_srvmtofh(fhp); - nfsm_srvstrsiz(len, NFS_MAXNAMLEN); + nfsm_srvnamesiz(len); nd.ni_cnd.cn_cred = cred; nd.ni_cnd.cn_nameiop = CREATE; - nd.ni_cnd.cn_flags = LOCKPARENT; - error = nfs_namei(&nd, fhp, len, nfsd->nd_slp, nam, &md, &dpos, - nfsd->nd_procp); + nd.ni_cnd.cn_flags = LOCKPARENT | SAVESTART; + error = nfs_namei(&nd, fhp, len, slp, nam, &md, &dpos, + &dirp, procp, (nfsd->nd_flag & ND_KERBAUTH)); + if (dirp) { + if (v3) + dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred, + procp); + else { + nfsrv_vrele(dirp); + dirp = (struct vnode *)0; + } + } if (error) goto out; + VATTR_NULL(vap); + if (v3) + nfsm_srvsattr(vap); nfsm_strsiz(len2, NFS_MAXPATHLEN); MALLOC(pathcp, caddr_t, len2 + 1, M_TEMP, M_WAITOK); iv.iov_base = pathcp; @@ -1194,9 +2125,14 @@ nfsrv_symlink(nfsd, mrep, md, dpos, cred, nam, mrq) io.uio_rw = UIO_READ; io.uio_procp = (struct proc *)0; nfsm_mtouio(&io, len2); - nfsm_dissect(sp, struct nfsv2_sattr *, NFSX_SATTR(nfsd->nd_nqlflag != NQL_NOVAL)); + if (!v3) { + nfsm_dissect(sp, struct nfsv2_sattr *, NFSX_V2SATTR); + vap->va_mode = fxdr_unsigned(u_short, sp->sa_mode); + } *(pathcp + len2) = '\0'; if (nd.ni_vp) { + nfsrv_vrele(nd.ni_startdir); + free(nd.ni_cnd.cn_pnbuf, M_NAMEI); VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); if (nd.ni_dvp == nd.ni_vp) nfsrv_vrele(nd.ni_dvp); @@ -1206,20 +2142,58 @@ nfsrv_symlink(nfsd, mrep, md, dpos, cred, nam, mrq) error = EEXIST; goto out; } - VATTR_NULL(vap); - vap->va_mode = fxdr_unsigned(u_short, sp->sa_mode); - nqsrv_getl(nd.ni_dvp, NQL_WRITE); + nqsrv_getl(nd.ni_dvp, ND_WRITE); if ((ovp = nd.ni_vp) && (ovp->v_flag & VVMIO) && ovp->v_vmdata) deallocobj = 1; error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, vap, pathcp); if (error == 0 && deallocobj) vm_object_deallocate( (vm_object_t) ovp->v_vmdata); + if (error) + nfsrv_vrele(nd.ni_startdir); + else { + if (v3) { + nd.ni_cnd.cn_nameiop = LOOKUP; + nd.ni_cnd.cn_flags &= ~(LOCKPARENT | SAVESTART | FOLLOW); + nd.ni_cnd.cn_flags |= (NOFOLLOW | LOCKLEAF); + nd.ni_cnd.cn_proc = procp; + nd.ni_cnd.cn_cred = cred; + error = lookup(&nd); + if (!error) { + bzero((caddr_t)fhp, sizeof(nfh)); + fhp->fh_fsid = nd.ni_vp->v_mount->mnt_stat.f_fsid; + error = VFS_VPTOFH(nd.ni_vp, &fhp->fh_fid); + if (!error) + error = VOP_GETATTR(nd.ni_vp, vap, cred, + procp); + nfsrv_vput(nd.ni_vp); + } + } else + nfsrv_vrele(nd.ni_startdir); + FREE(nd.ni_cnd.cn_pnbuf, M_NAMEI); + } out: if (pathcp) FREE(pathcp, M_TEMP); - nfsm_reply(0); - return (error); + if (dirp) { + diraft_ret = VOP_GETATTR(dirp, &diraft, cred, procp); + nfsrv_vrele(dirp); + } + nfsm_reply(NFSX_SRVFH(v3) + NFSX_POSTOPATTR(v3) + NFSX_WCCDATA(v3)); + if (v3) { + if (!error) { + nfsm_srvpostop_fh(fhp); + nfsm_srvpostop_attr(0, vap); + } + nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); + } + return (0); nfsmout: + if (nd.ni_cnd.cn_nameiop) { + nfsrv_vrele(nd.ni_startdir); + free(nd.ni_cnd.cn_pnbuf, M_NAMEI); + } + if (dirp) + nfsrv_vrele(dirp); VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); if (nd.ni_dvp == nd.ni_vp) nfsrv_vrele(nd.ni_dvp); @@ -1236,43 +2210,65 @@ nfsmout: * nfs mkdir service */ int -nfsrv_mkdir(nfsd, mrep, md, dpos, cred, nam, mrq) - struct nfsd *nfsd; - struct mbuf *mrep, *md; - caddr_t dpos; - struct ucred *cred; - struct mbuf *nam, **mrq; +nfsrv_mkdir(nfsd, slp, procp, mrq) + struct nfsrv_descript *nfsd; + struct nfssvc_sock *slp; + struct proc *procp; + struct mbuf **mrq; { - struct vattr va; + struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; + struct mbuf *nam = nfsd->nd_nam; + caddr_t dpos = nfsd->nd_dpos; + struct ucred *cred = &nfsd->nd_cr; + struct vattr va, dirfor, diraft; register struct vattr *vap = &va; - register struct nfsv2_fattr *fp; + register struct nfs_fattr *fp; struct nameidata nd; register caddr_t cp; register u_long *tl; register long t1; caddr_t bpos; - int error = 0, cache, len; + int error = 0, cache, len, dirfor_ret = 1, diraft_ret = 1; + int v3 = (nfsd->nd_flag & ND_NFSV3); char *cp2; struct mbuf *mb, *mb2, *mreq; - struct vnode *vp; - nfsv2fh_t nfh; + struct vnode *vp, *dirp = (struct vnode *)0; + nfsfh_t nfh; fhandle_t *fhp; u_quad_t frev; fhp = &nfh.fh_generic; nfsm_srvmtofh(fhp); - nfsm_srvstrsiz(len, NFS_MAXNAMLEN); + nfsm_srvnamesiz(len); nd.ni_cnd.cn_cred = cred; nd.ni_cnd.cn_nameiop = CREATE; nd.ni_cnd.cn_flags = LOCKPARENT; - error = nfs_namei(&nd, fhp, len, nfsd->nd_slp, nam, &md, &dpos, - nfsd->nd_procp); - if (error) - nfsm_reply(0); - nfsm_dissect(tl, u_long *, NFSX_UNSIGNED); + error = nfs_namei(&nd, fhp, len, slp, nam, &md, &dpos, + &dirp, procp, (nfsd->nd_flag & ND_KERBAUTH)); + if (dirp) { + if (v3) + dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred, + procp); + else { + nfsrv_vrele(dirp); + dirp = (struct vnode *)0; + } + } + if (error) { + nfsm_reply(NFSX_WCCDATA(v3)); + nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); + if (dirp) + nfsrv_vrele(dirp); + return (0); + } VATTR_NULL(vap); + if (v3) { + nfsm_srvsattr(vap); + } else { + nfsm_dissect(tl, u_long *, NFSX_UNSIGNED); + vap->va_mode = nfstov_mode(*tl++); + } vap->va_type = VDIR; - vap->va_mode = nfstov_mode(*tl++); vp = nd.ni_vp; if (vp != NULL) { VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); @@ -1282,28 +2278,40 @@ nfsrv_mkdir(nfsd, mrep, md, dpos, cred, nam, mrq) nfsrv_vput(nd.ni_dvp); nfsrv_vrele(vp); error = EEXIST; - nfsm_reply(0); + goto out; } - nqsrv_getl(nd.ni_dvp, NQL_WRITE); + nqsrv_getl(nd.ni_dvp, ND_WRITE); error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, vap); - if (error) - nfsm_reply(0); - vp = nd.ni_vp; - bzero((caddr_t)fhp, sizeof(nfh)); - fhp->fh_fsid = vp->v_mount->mnt_stat.f_fsid; - error = VFS_VPTOFH(vp, &fhp->fh_fid); - if (error) { + if (!error) { + vp = nd.ni_vp; + bzero((caddr_t)fhp, sizeof(nfh)); + fhp->fh_fsid = vp->v_mount->mnt_stat.f_fsid; + error = VFS_VPTOFH(vp, &fhp->fh_fid); + if (!error) + error = VOP_GETATTR(vp, vap, cred, procp); nfsrv_vput(vp); - nfsm_reply(0); } - error = VOP_GETATTR(vp, vap, cred, nfsd->nd_procp); - nfsrv_vput(vp); - nfsm_reply(NFSX_FH+NFSX_FATTR(nfsd->nd_nqlflag != NQL_NOVAL)); - nfsm_srvfhtom(fhp); - nfsm_build(fp, struct nfsv2_fattr *, NFSX_FATTR(nfsd->nd_nqlflag != NQL_NOVAL)); - nfsm_srvfillattr; - return (error); +out: + if (dirp) { + diraft_ret = VOP_GETATTR(dirp, &diraft, cred, procp); + nfsrv_vrele(dirp); + } + nfsm_reply(NFSX_SRVFH(v3) + NFSX_POSTOPATTR(v3) + NFSX_WCCDATA(v3)); + if (v3) { + if (!error) { + nfsm_srvpostop_fh(fhp); + nfsm_srvpostop_attr(0, vap); + } + nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); + } else { + nfsm_srvfhtom(fhp, v3); + nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR); + nfsm_srvfillattr(vap, fp); + } + return (0); nfsmout: + if (dirp) + nfsrv_vrele(dirp); VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); if (nd.ni_dvp == nd.ni_vp) nfsrv_vrele(nd.ni_dvp); @@ -1318,35 +2326,54 @@ nfsmout: * nfs rmdir service */ int -nfsrv_rmdir(nfsd, mrep, md, dpos, cred, nam, mrq) - struct nfsd *nfsd; - struct mbuf *mrep, *md; - caddr_t dpos; - struct ucred *cred; - struct mbuf *nam, **mrq; +nfsrv_rmdir(nfsd, slp, procp, mrq) + struct nfsrv_descript *nfsd; + struct nfssvc_sock *slp; + struct proc *procp; + struct mbuf **mrq; { + struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; + struct mbuf *nam = nfsd->nd_nam; + caddr_t dpos = nfsd->nd_dpos; + struct ucred *cred = &nfsd->nd_cr; register u_long *tl; register long t1; caddr_t bpos; - int error = 0, cache, len; + int error = 0, cache, len, dirfor_ret = 1, diraft_ret = 1; + int v3 = (nfsd->nd_flag & ND_NFSV3); char *cp2; - struct mbuf *mb, *mreq; - struct vnode *vp; - nfsv2fh_t nfh; + struct mbuf *mb, *mreq, *mb2; + struct vnode *vp, *dirp = (struct vnode *)0; + struct vattr dirfor, diraft; + nfsfh_t nfh; fhandle_t *fhp; struct nameidata nd; u_quad_t frev; fhp = &nfh.fh_generic; nfsm_srvmtofh(fhp); - nfsm_srvstrsiz(len, NFS_MAXNAMLEN); + nfsm_srvnamesiz(len); nd.ni_cnd.cn_cred = cred; nd.ni_cnd.cn_nameiop = DELETE; nd.ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF; - error = nfs_namei(&nd, fhp, len, nfsd->nd_slp, nam, &md, &dpos, - nfsd->nd_procp); - if (error) - nfsm_reply(0); + error = nfs_namei(&nd, fhp, len, slp, nam, &md, &dpos, + &dirp, procp, (nfsd->nd_flag & ND_KERBAUTH)); + if (dirp) { + if (v3) + dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred, + procp); + else { + nfsrv_vrele(dirp); + dirp = (struct vnode *)0; + } + } + if (error) { + nfsm_reply(NFSX_WCCDATA(v3)); + nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); + if (dirp) + nfsrv_vrele(dirp); + return (0); + } vp = nd.ni_vp; if (vp->v_type != VDIR) { error = ENOTDIR; @@ -1366,8 +2393,8 @@ nfsrv_rmdir(nfsd, mrep, md, dpos, cred, nam, mrq) error = EBUSY; out: if (!error) { - nqsrv_getl(nd.ni_dvp, NQL_WRITE); - nqsrv_getl(vp, NQL_WRITE); + nqsrv_getl(nd.ni_dvp, ND_WRITE); + nqsrv_getl(vp, ND_WRITE); error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); } else { VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); @@ -1377,7 +2404,15 @@ out: nfsrv_vput(nd.ni_dvp); nfsrv_vput(vp); } - nfsm_reply(0); + if (dirp) { + diraft_ret = VOP_GETATTR(dirp, &diraft, cred, procp); + nfsrv_vrele(dirp); + } + nfsm_reply(NFSX_WCCDATA(v3)); + if (v3) { + nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); + return (0); + } nfsm_srvdone; } @@ -1401,7 +2436,7 @@ out: * than requested, but this may not apply to all filesystems. For * example, client NFS does not { although it is never remote mounted * anyhow } - * The alternate call nqnfsrv_readdirlook() does lookups as well. + * The alternate call nfsrv_readdirplus() does lookups as well. * PS: The NFS protocol spec. does not clarify what the "count" byte * argument is a count of.. just name strings and file id's or the * entire reply rpc or ... @@ -1411,21 +2446,25 @@ out: * "entry" structures, but are in the rpc. */ struct flrep { - u_long fl_cachable; - u_long fl_duration; - u_long fl_frev[2]; - nfsv2fh_t fl_nfh; - u_long fl_fattr[NFSX_NQFATTR / sizeof (u_long)]; + nfsuint64 fl_off; + u_long fl_postopok; + u_long fl_fattr[NFSX_V3FATTR / sizeof (u_long)]; + u_long fl_fhok; + u_long fl_fhsize; + u_long fl_nfh[NFSX_V3FH / sizeof (u_long)]; }; int -nfsrv_readdir(nfsd, mrep, md, dpos, cred, nam, mrq) - struct nfsd *nfsd; - struct mbuf *mrep, *md; - caddr_t dpos; - struct ucred *cred; - struct mbuf *nam, **mrq; +nfsrv_readdir(nfsd, slp, procp, mrq) + struct nfsrv_descript *nfsd; + struct nfssvc_sock *slp; + struct proc *procp; + struct mbuf **mrq; { + struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; + struct mbuf *nam = nfsd->nd_nam; + caddr_t dpos = nfsd->nd_dpos; + struct ucred *cred = &nfsd->nd_cr; register char *bp, *be; register struct mbuf *mp; register struct dirent *dp; @@ -1436,38 +2475,63 @@ nfsrv_readdir(nfsd, mrep, md, dpos, cred, nam, mrq) struct mbuf *mb, *mb2, *mreq, *mp2; char *cpos, *cend, *cp2, *rbuf; struct vnode *vp; - nfsv2fh_t nfh; + struct vattr at; + nfsfh_t nfh; fhandle_t *fhp; struct uio io; struct iovec iv; - int len, nlen, rem, xfer, tsiz, i, error = 0; - int siz, cnt, fullsiz, eofflag, rdonly, cache; - u_quad_t frev; - u_long off, toff; - int ncookies; - u_int *cookies; - u_int *cookiep; + int len, nlen, rem, xfer, tsiz, i, error = 0, getret = 1; + int siz, cnt, fullsiz, eofflag, rdonly, cache, ncookies; + int v3 = (nfsd->nd_flag & ND_NFSV3); + u_quad_t frev, off, toff, verf; + u_long *cookies = NULL, *cookiep; fhp = &nfh.fh_generic; nfsm_srvmtofh(fhp); - nfsm_dissect(tl, u_long *, 2*NFSX_UNSIGNED); - off = toff = fxdr_unsigned(u_long, *tl++); + if (v3) { + nfsm_dissect(tl, u_long *, 5 * NFSX_UNSIGNED); + fxdr_hyper(tl, &toff); + tl += 2; + fxdr_hyper(tl, &verf); + tl += 2; + } else { + nfsm_dissect(tl, u_long *, 2 * NFSX_UNSIGNED); + toff = fxdr_unsigned(u_quad_t, *tl++); + } + off = toff; cnt = fxdr_unsigned(int, *tl); - siz = ((cnt+NFS_DIRBLKSIZ-1) & ~(NFS_DIRBLKSIZ-1)); - if (cnt > NFS_MAXREADDIR) - siz = NFS_MAXREADDIR; + siz = ((cnt + DIRBLKSIZ - 1) & ~(DIRBLKSIZ - 1)); + xfer = NFS_SRVMAXDATA(nfsd); + if (siz > xfer) + siz = xfer; fullsiz = siz; - error = nfsrv_fhtovp(fhp, TRUE, &vp, cred, nfsd->nd_slp, nam, &rdonly); - if (error) - nfsm_reply(0); - nqsrv_getl(vp, NQL_READ); - error = nfsrv_access(vp, VEXEC, cred, rdonly, nfsd->nd_procp); + if (error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam, + &rdonly, (nfsd->nd_flag & ND_KERBAUTH))) { + nfsm_reply(NFSX_UNSIGNED); + nfsm_srvpostop_attr(getret, &at); + return (0); + } + nqsrv_getl(vp, ND_READ); + if (v3) { + error = getret = VOP_GETATTR(vp, &at, cred, procp); + if (!error && toff && verf != at.va_filerev) + error = NFSERR_BAD_COOKIE; + } + if (!error) + error = nfsrv_access(vp, VEXEC, cred, rdonly, procp); if (error) { nfsrv_vput(vp); - nfsm_reply(0); + nfsm_reply(NFSX_POSTOPATTR(v3)); + nfsm_srvpostop_attr(getret, &at); + return (0); } VOP_UNLOCK(vp); MALLOC(rbuf, caddr_t, siz, M_TEMP, M_WAITOK); +#ifdef __NetBSD__ + ncookies = siz / (5 * NFSX_UNSIGNED); /*7 for V3, but it's an est. so*/ + MALLOC(cookies, u_long *, ncookies * sizeof (u_long *), M_TEMP, + M_WAITOK); +#endif again: iv.iov_base = rbuf; iv.iov_len = fullsiz; @@ -1478,25 +2542,31 @@ again: io.uio_segflg = UIO_SYSSPACE; io.uio_rw = UIO_READ; io.uio_procp = (struct proc *)0; - cookies = NULL; eofflag = 0; +#ifndef __NetBSD__ + if (cookies) { + free((caddr_t)cookies, M_TEMP); + cookies = NULL; + } error = VOP_READDIR(vp, &io, cred, &eofflag, &ncookies, &cookies); +#else + error = VOP_READDIR(vp, &io, cred, &eofflag, cookies, ncookies); +#endif off = (off_t)io.uio_offset; + if (!cookies && !error) + error = NFSERR_PERM; + if (v3) { + getret = VOP_GETATTR(vp, &at, cred, procp); + if (!error) + error = getret; + } if (error) { nfsrv_vrele(vp); free((caddr_t)rbuf, M_TEMP); - nfsm_reply(0); - } - if (cookies == NULL) { - /* - * If the filesystem doen't support cookies, return eof. - */ - nfsrv_vrele(vp); - nfsm_reply(2*NFSX_UNSIGNED); - nfsm_build(tl, u_long *, 2*NFSX_UNSIGNED); - *tl++ = nfs_false; - *tl = nfs_true; - FREE((caddr_t)rbuf, M_TEMP); + if (cookies) + free((caddr_t)cookies, M_TEMP); + nfsm_reply(NFSX_POSTOPATTR(v3)); + nfsm_srvpostop_attr(getret, &at); return (0); } if (io.uio_resid) { @@ -1508,12 +2578,19 @@ again: */ if (siz == 0) { nfsrv_vrele(vp); - nfsm_reply(2*NFSX_UNSIGNED); - nfsm_build(tl, u_long *, 2*NFSX_UNSIGNED); + nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_COOKIEVERF(v3) + + 2 * NFSX_UNSIGNED); + if (v3) { + nfsm_srvpostop_attr(getret, &at); + nfsm_build(tl, u_long *, 4 * NFSX_UNSIGNED); + txdr_hyper(&at.va_filerev, tl); + tl += 2; + } else + nfsm_build(tl, u_long *, 2 * NFSX_UNSIGNED); *tl++ = nfs_false; *tl = nfs_true; FREE((caddr_t)rbuf, M_TEMP); - FREE(cookies, M_TEMP); + FREE((caddr_t)cookies, M_TEMP); return (0); } } @@ -1526,32 +2603,49 @@ again: cend = rbuf + siz; dp = (struct dirent *)cpos; cookiep = cookies; - while (cpos < cend && ncookies > 0 - && (*cookiep <= toff || dp->d_fileno == 0)) { +#ifdef __FreeBSD__ + /* + * For some reason FreeBSD's ufs_readdir() chooses to back the + * directory offset up to a block boundary, so it is necessary to + * skip over the records that preceed the requested offset. This + * requires the assumption that file offset cookies monotonically + * increase. + */ + while (cpos < cend && ncookies > 0 && + (dp->d_fileno == 0 || ((u_quad_t)(*cookiep)) <= toff)) { +#else + while (dp->d_fileno == 0 && cpos < cend && ncookies > 0) { +#endif cpos += dp->d_reclen; dp = (struct dirent *)cpos; cookiep++; ncookies--; } - if (cpos >= cend) { + if (cpos >= cend || ncookies == 0) { toff = off; siz = fullsiz; - FREE(cookies, M_TEMP); goto again; } - len = 3*NFSX_UNSIGNED; /* paranoia, probably can be 0 */ - nfsm_reply(siz); + len = 3 * NFSX_UNSIGNED; /* paranoia, probably can be 0 */ + nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_COOKIEVERF(v3) + siz); + if (v3) { + nfsm_srvpostop_attr(getret, &at); + nfsm_build(tl, u_long *, 2 * NFSX_UNSIGNED); + txdr_hyper(&at.va_filerev, tl); + } mp = mp2 = mb; bp = bpos; be = bp + M_TRAILINGSPACE(mp); /* Loop through the records and build reply */ - while (cpos < cend) { + while (cpos < cend && ncookies > 0) { if (dp->d_fileno != 0) { nlen = dp->d_namlen; rem = nfsm_rndup(nlen)-nlen; - len += (4*NFSX_UNSIGNED + nlen + rem); + len += (4 * NFSX_UNSIGNED + nlen + rem); + if (v3) + len += 2 * NFSX_UNSIGNED; if (len > cnt) { eofflag = 0; break; @@ -1563,6 +2657,11 @@ again: nfsm_clget; *tl = nfs_true; bp += NFSX_UNSIGNED; + if (v3) { + nfsm_clget; + *tl = 0; + bp += NFSX_UNSIGNED; + } nfsm_clget; *tl = txdr_unsigned(dp->d_fileno); bp += NFSX_UNSIGNED; @@ -1591,12 +2690,18 @@ again: nfsm_clget; /* Finish off the record */ + if (v3) { + *tl = 0; + bp += NFSX_UNSIGNED; + nfsm_clget; + } *tl = txdr_unsigned(*cookiep); bp += NFSX_UNSIGNED; } cpos += dp->d_reclen; dp = (struct dirent *)cpos; cookiep++; + ncookies--; } nfsrv_vrele(vp); nfsm_clget; @@ -1613,19 +2718,22 @@ again: mp->m_len = bp - mtod(mp, caddr_t); } else mp->m_len += bp - bpos; - FREE(rbuf, M_TEMP); - FREE(cookies, M_TEMP); + FREE((caddr_t)rbuf, M_TEMP); + FREE((caddr_t)cookies, M_TEMP); nfsm_srvdone; } int -nqnfsrv_readdirlook(nfsd, mrep, md, dpos, cred, nam, mrq) - struct nfsd *nfsd; - struct mbuf *mrep, *md; - caddr_t dpos; - struct ucred *cred; - struct mbuf *nam, **mrq; +nfsrv_readdirplus(nfsd, slp, procp, mrq) + struct nfsrv_descript *nfsd; + struct nfssvc_sock *slp; + struct proc *procp; + struct mbuf **mrq; { + struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; + struct mbuf *nam = nfsd->nd_nam; + caddr_t dpos = nfsd->nd_dpos; + struct ucred *cred = &nfsd->nd_cr; register char *bp, *be; register struct mbuf *mp; register struct dirent *dp; @@ -1637,42 +2745,58 @@ nqnfsrv_readdirlook(nfsd, mrep, md, dpos, cred, nam, mrq) char *cpos, *cend, *cp2, *rbuf; struct vnode *vp, *nvp; struct flrep fl; - nfsv2fh_t nfh; - fhandle_t *fhp; + nfsfh_t nfh; + fhandle_t *fhp, *nfhp = (fhandle_t *)fl.fl_nfh; struct uio io; struct iovec iv; - struct vattr va, *vap = &va; - struct nfsv2_fattr *fp; - int len, nlen, rem, xfer, tsiz, i, error = 0, cache2; - int siz, cnt, fullsiz, eofflag, rdonly, cache; - u_long duration2; - u_quad_t frev, frev2; - u_long off, toff; - int ncookies; - u_int *cookies; - u_int *cookiep; + struct vattr va, at, *vap = &va; + struct nfs_fattr *fp; + int len, nlen, rem, xfer, tsiz, i, error = 0, getret = 1; + int siz, cnt, fullsiz, eofflag, rdonly, cache, dirlen, ncookies; + u_quad_t frev, off, toff, verf; + u_long *cookies = NULL, *cookiep; fhp = &nfh.fh_generic; nfsm_srvmtofh(fhp); - nfsm_dissect(tl, u_long *, 3*NFSX_UNSIGNED); - toff = off = fxdr_unsigned(u_long, *tl++); - cnt = fxdr_unsigned(int, *tl++); - duration2 = fxdr_unsigned(int, *tl); - siz = ((cnt+NFS_DIRBLKSIZ-1) & ~(NFS_DIRBLKSIZ-1)); - if (cnt > NFS_MAXREADDIR) - siz = NFS_MAXREADDIR; + nfsm_dissect(tl, u_long *, 6 * NFSX_UNSIGNED); + fxdr_hyper(tl, &toff); + tl += 2; + fxdr_hyper(tl, &verf); + tl += 2; + siz = fxdr_unsigned(int, *tl++); + cnt = fxdr_unsigned(int, *tl); + off = toff; + siz = ((siz + DIRBLKSIZ - 1) & ~(DIRBLKSIZ - 1)); + xfer = NFS_SRVMAXDATA(nfsd); + if (siz > xfer) + siz = xfer; fullsiz = siz; - error = nfsrv_fhtovp(fhp, TRUE, &vp, cred, nfsd->nd_slp, nam, &rdonly); - if (error) - nfsm_reply(0); - nqsrv_getl(vp, NQL_READ); - error = nfsrv_access(vp, VEXEC, cred, rdonly, nfsd->nd_procp); + if (error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam, + &rdonly, (nfsd->nd_flag & ND_KERBAUTH))) { + nfsm_reply(NFSX_UNSIGNED); + nfsm_srvpostop_attr(getret, &at); + return (0); + } + error = getret = VOP_GETATTR(vp, &at, cred, procp); + if (!error && toff && verf != at.va_filerev) + error = NFSERR_BAD_COOKIE; + if (!error) { + nqsrv_getl(vp, ND_READ); + error = nfsrv_access(vp, VEXEC, cred, rdonly, procp); + } if (error) { nfsrv_vput(vp); - nfsm_reply(0); + nfsm_reply(NFSX_V3POSTOPATTR); + nfsm_srvpostop_attr(getret, &at); + return (0); } VOP_UNLOCK(vp); MALLOC(rbuf, caddr_t, siz, M_TEMP, M_WAITOK); +#ifdef __NetBSD__ + ncookies = siz / (7 * NFSX_UNSIGNED); + MALLOC(cookies, u_long *, ncookies * sizeof (u_long *), M_TEMP, + M_WAITOK); +#endif again: iv.iov_base = rbuf; iv.iov_len = fullsiz; @@ -1683,25 +2807,29 @@ again: io.uio_segflg = UIO_SYSSPACE; io.uio_rw = UIO_READ; io.uio_procp = (struct proc *)0; - cookies = NULL; eofflag = 0; +#ifndef __NetBSD__ + if (cookies) { + free((caddr_t)cookies, M_TEMP); + cookies = NULL; + } error = VOP_READDIR(vp, &io, cred, &eofflag, &ncookies, &cookies); - off = (u_long)io.uio_offset; +#else + error = VOP_READDIR(vp, &io, cred, &eofflag, cookies, ncookies); +#endif + off = (u_quad_t)io.uio_offset; + getret = VOP_GETATTR(vp, &at, cred, procp); + if (!cookies && !error) + error = NFSERR_PERM; + if (!error) + error = getret; if (error) { nfsrv_vrele(vp); + if (cookies) + free((caddr_t)cookies, M_TEMP); free((caddr_t)rbuf, M_TEMP); - nfsm_reply(0); - } - if (cookies == NULL) { - /* - * If the filesystem doen't support cookies, return eof. - */ - nfsrv_vrele(vp); - nfsm_reply(2*NFSX_UNSIGNED); - nfsm_build(tl, u_long *, 2*NFSX_UNSIGNED); - *tl++ = nfs_false; - *tl = nfs_true; - FREE((caddr_t)rbuf, M_TEMP); + nfsm_reply(NFSX_V3POSTOPATTR); + nfsm_srvpostop_attr(getret, &at); return (0); } if (io.uio_resid) { @@ -1713,12 +2841,16 @@ again: */ if (siz == 0) { nfsrv_vrele(vp); - nfsm_reply(2 * NFSX_UNSIGNED); - nfsm_build(tl, u_long *, 2 * NFSX_UNSIGNED); + nfsm_reply(NFSX_V3POSTOPATTR + NFSX_V3COOKIEVERF + + 2 * NFSX_UNSIGNED); + nfsm_srvpostop_attr(getret, &at); + nfsm_build(tl, u_long *, 4 * NFSX_UNSIGNED); + txdr_hyper(&at.va_filerev, tl); + tl += 2; *tl++ = nfs_false; *tl = nfs_true; + FREE((caddr_t)cookies, M_TEMP); FREE((caddr_t)rbuf, M_TEMP); - FREE(cookies, M_TEMP); return (0); } } @@ -1731,28 +2863,41 @@ again: cend = rbuf + siz; dp = (struct dirent *)cpos; cookiep = cookies; - while (cpos < cend && ncookies > 0 - && (*cookiep <= toff || dp->d_fileno == 0)) { +#ifdef __FreeBSD__ + /* + * For some reason FreeBSD's ufs_readdir() chooses to back the + * directory offset up to a block boundary, so it is necessary to + * skip over the records that preceed the requested offset. This + * requires the assumption that file offset cookies monotonically + * increase. + */ + while (cpos < cend && ncookies > 0 && + (dp->d_fileno == 0 || ((u_quad_t)(*cookiep)) <= toff)) { +#else + while (dp->d_fileno == 0 && cpos < cend && ncookies > 0) { +#endif cpos += dp->d_reclen; dp = (struct dirent *)cpos; cookiep++; ncookies--; } - if (cpos >= cend) { + if (cpos >= cend || ncookies == 0) { toff = off; siz = fullsiz; - FREE(cookies, M_TEMP); goto again; } - len = 3 * NFSX_UNSIGNED; /* paranoia, probably can be 0 */ - nfsm_reply(siz); + dirlen = len = NFSX_V3POSTOPATTR + NFSX_V3COOKIEVERF + 2 * NFSX_UNSIGNED; + nfsm_reply(cnt); + nfsm_srvpostop_attr(getret, &at); + nfsm_build(tl, u_long *, 2 * NFSX_UNSIGNED); + txdr_hyper(&at.va_filerev, tl); mp = mp2 = mb; bp = bpos; be = bp + M_TRAILINGSPACE(mp); /* Loop through the records and build reply */ - while (cpos < cend) { + while (cpos < cend && ncookies > 0) { if (dp->d_fileno != 0) { nlen = dp->d_namlen; rem = nfsm_rndup(nlen)-nlen; @@ -1763,59 +2908,51 @@ again: */ if (VFS_VGET(vp->v_mount, dp->d_fileno, &nvp)) goto invalid; - bzero((caddr_t)&fl.fl_nfh, sizeof (nfsv2fh_t)); - fl.fl_nfh.fh_generic.fh_fsid = + bzero((caddr_t)nfhp, NFSX_V3FH); + nfhp->fh_fsid = nvp->v_mount->mnt_stat.f_fsid; - if (VFS_VPTOFH(nvp, &fl.fl_nfh.fh_generic.fh_fid)) { + if (VFS_VPTOFH(nvp, &nfhp->fh_fid)) { vput(nvp); goto invalid; } - if (duration2) { - (void) nqsrv_getlease(nvp, &duration2, NQL_READ, - nfsd, nam, &cache2, &frev2, cred); - fl.fl_duration = txdr_unsigned(duration2); - fl.fl_cachable = txdr_unsigned(cache2); - txdr_hyper(&frev2, fl.fl_frev); - } else - fl.fl_duration = 0; - if (VOP_GETATTR(nvp, vap, cred, nfsd->nd_procp)) { + if (VOP_GETATTR(nvp, vap, cred, procp)) { vput(nvp); goto invalid; } vput(nvp); - fp = (struct nfsv2_fattr *)&fl.fl_fattr; - nfsm_srvfillattr; - len += (4*NFSX_UNSIGNED + nlen + rem + NFSX_FH - + NFSX_NQFATTR); - if (len > cnt) { + + /* + * If either the dircount or maxcount will be + * exceeded, get out now. Both of these lengths + * are calculated conservatively, including all + * XDR overheads. + */ + len += (7 * NFSX_UNSIGNED + nlen + rem + NFSX_V3FH + + NFSX_V3POSTOPATTR); + dirlen += (6 * NFSX_UNSIGNED + nlen + rem); + if (len > cnt || dirlen > fullsiz) { eofflag = 0; break; } + /* * Build the directory record xdr from * the dirent entry. */ + fp = (struct nfs_fattr *)&fl.fl_fattr; + nfsm_srvfillattr(vap, fp); + fl.fl_fhsize = txdr_unsigned(NFSX_V3FH); + fl.fl_fhok = nfs_true; + fl.fl_postopok = nfs_true; + fl.fl_off.nfsuquad[0] = 0; + fl.fl_off.nfsuquad[1] = txdr_unsigned(*cookiep); + nfsm_clget; *tl = nfs_true; bp += NFSX_UNSIGNED; - - /* - * For readdir_and_lookup copy the stuff out. - */ - xfer = sizeof (struct flrep); - cp = (caddr_t)&fl; - while (xfer > 0) { - nfsm_clget; - if ((bp+xfer) > be) - tsiz = be-bp; - else - tsiz = xfer; - bcopy(cp, bp, tsiz); - bp += tsiz; - xfer -= tsiz; - if (xfer > 0) - cp += tsiz; - } + nfsm_clget; + *tl = 0; + bp += NFSX_UNSIGNED; nfsm_clget; *tl = txdr_unsigned(dp->d_fileno); bp += NFSX_UNSIGNED; @@ -1828,8 +2965,8 @@ again: cp = dp->d_name; while (xfer > 0) { nfsm_clget; - if ((bp+xfer) > be) - tsiz = be-bp; + if ((bp + xfer) > be) + tsiz = be - bp; else tsiz = xfer; bcopy(cp, bp, tsiz); @@ -1841,16 +2978,30 @@ again: /* And null pad to a long boundary */ for (i = 0; i < rem; i++) *bp++ = '\0'; - nfsm_clget; - - /* Finish off the record */ - *tl = txdr_unsigned(*cookiep); - bp += NFSX_UNSIGNED; + + /* + * Now copy the flrep structure out. + */ + xfer = sizeof (struct flrep); + cp = (caddr_t)&fl; + while (xfer > 0) { + nfsm_clget; + if ((bp + xfer) > be) + tsiz = be - bp; + else + tsiz = xfer; + bcopy(cp, bp, tsiz); + bp += tsiz; + xfer -= tsiz; + if (xfer > 0) + cp += tsiz; + } } invalid: cpos += dp->d_reclen; dp = (struct dirent *)cpos; cookiep++; + ncookies--; } nfsrv_vrele(vp); nfsm_clget; @@ -1867,8 +3018,69 @@ invalid: mp->m_len = bp - mtod(mp, caddr_t); } else mp->m_len += bp - bpos; - FREE(rbuf, M_TEMP); - FREE(cookies, M_TEMP); + FREE((caddr_t)cookies, M_TEMP); + FREE((caddr_t)rbuf, M_TEMP); + nfsm_srvdone; +} + +/* + * nfs commit service + */ +int +nfsrv_commit(nfsd, slp, procp, mrq) + struct nfsrv_descript *nfsd; + struct nfssvc_sock *slp; + struct proc *procp; + struct mbuf **mrq; +{ + struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; + struct mbuf *nam = nfsd->nd_nam; + caddr_t dpos = nfsd->nd_dpos; + struct ucred *cred = &nfsd->nd_cr; + struct vattr bfor, aft; + struct vnode *vp; + nfsfh_t nfh; + fhandle_t *fhp; + register u_long *tl; + register long t1; + caddr_t bpos; + int error = 0, rdonly, for_ret = 1, aft_ret = 1, cnt, cache; + char *cp2; + struct mbuf *mb, *mb2, *mreq; + u_quad_t frev, off; + +#ifndef nolint + cache = 0; +#endif + fhp = &nfh.fh_generic; + nfsm_srvmtofh(fhp); + nfsm_dissect(tl, u_long *, 3 * NFSX_UNSIGNED); + + /* + * XXX At this time VOP_FSYNC() does not accept offset and byte + * count parameters, so these arguments are useless (someday maybe). + */ + fxdr_hyper(tl, &off); + tl += 2; + cnt = fxdr_unsigned(int, *tl); + if (error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam, + &rdonly, (nfsd->nd_flag & ND_KERBAUTH))) { + nfsm_reply(2 * NFSX_UNSIGNED); + nfsm_srvwcc_data(for_ret, &bfor, aft_ret, &aft); + return (0); + } + for_ret = VOP_GETATTR(vp, &bfor, cred, procp); + error = VOP_FSYNC(vp, cred, MNT_WAIT, procp); + aft_ret = VOP_GETATTR(vp, &aft, cred, procp); + nfsrv_vput(vp); + nfsm_reply(NFSX_V3WCCDATA + NFSX_V3WRITEVERF); + nfsm_srvwcc_data(for_ret, &bfor, aft_ret, &aft); + if (!error) { + nfsm_build(tl, u_long *, NFSX_V3WRITEVERF); + *tl++ = txdr_unsigned(boottime.tv_sec); + *tl = txdr_unsigned(boottime.tv_usec); + } else + return (0); nfsm_srvdone; } @@ -1876,69 +3088,244 @@ invalid: * nfs statfs service */ int -nfsrv_statfs(nfsd, mrep, md, dpos, cred, nam, mrq) - struct nfsd *nfsd; - struct mbuf *mrep, *md; - caddr_t dpos; - struct ucred *cred; - struct mbuf *nam, **mrq; +nfsrv_statfs(nfsd, slp, procp, mrq) + struct nfsrv_descript *nfsd; + struct nfssvc_sock *slp; + struct proc *procp; + struct mbuf **mrq; { + struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; + struct mbuf *nam = nfsd->nd_nam; + caddr_t dpos = nfsd->nd_dpos; + struct ucred *cred = &nfsd->nd_cr; register struct statfs *sf; - register struct nfsv2_statfs *sfp; + register struct nfs_statfs *sfp; register u_long *tl; register long t1; caddr_t bpos; - int error = 0, rdonly, cache = 0, isnq; + int error = 0, rdonly, cache, getret = 1; + int v3 = (nfsd->nd_flag & ND_NFSV3); char *cp2; struct mbuf *mb, *mb2, *mreq; struct vnode *vp; - nfsv2fh_t nfh; + struct vattr at; + nfsfh_t nfh; fhandle_t *fhp; struct statfs statfs; - u_quad_t frev; + u_quad_t frev, tval; +#ifndef nolint + cache = 0; +#endif fhp = &nfh.fh_generic; - isnq = (nfsd->nd_nqlflag != NQL_NOVAL); nfsm_srvmtofh(fhp); - error = nfsrv_fhtovp(fhp, TRUE, &vp, cred, nfsd->nd_slp, nam, &rdonly); - if (error) - nfsm_reply(0); + if (error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam, + &rdonly, (nfsd->nd_flag & ND_KERBAUTH))) { + nfsm_reply(NFSX_UNSIGNED); + nfsm_srvpostop_attr(getret, &at); + return (0); + } sf = &statfs; - error = VFS_STATFS(vp->v_mount, sf, nfsd->nd_procp); + error = VFS_STATFS(vp->v_mount, sf, procp); + getret = VOP_GETATTR(vp, &at, cred, procp); nfsrv_vput(vp); - nfsm_reply(NFSX_STATFS(isnq)); - nfsm_build(sfp, struct nfsv2_statfs *, NFSX_STATFS(isnq)); - sfp->sf_tsize = txdr_unsigned(NFS_MAXDGRAMDATA); - sfp->sf_bsize = txdr_unsigned(sf->f_bsize); - sfp->sf_blocks = txdr_unsigned(sf->f_blocks); - sfp->sf_bfree = txdr_unsigned(sf->f_bfree); - sfp->sf_bavail = txdr_unsigned(sf->f_bavail); - if (isnq) { - sfp->sf_files = txdr_unsigned(sf->f_files); - sfp->sf_ffree = txdr_unsigned(sf->f_ffree); + nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_STATFS(v3)); + if (v3) + nfsm_srvpostop_attr(getret, &at); + if (error) + return (0); + nfsm_build(sfp, struct nfs_statfs *, NFSX_STATFS(v3)); + if (v3) { + tval = (u_quad_t)sf->f_blocks; + tval *= (u_quad_t)sf->f_bsize; + txdr_hyper(&tval, &sfp->sf_tbytes); + tval = (u_quad_t)sf->f_bfree; + tval *= (u_quad_t)sf->f_bsize; + txdr_hyper(&tval, &sfp->sf_fbytes); + tval = (u_quad_t)sf->f_bavail; + tval *= (u_quad_t)sf->f_bsize; + txdr_hyper(&tval, &sfp->sf_abytes); + sfp->sf_tfiles.nfsuquad[0] = 0; + sfp->sf_tfiles.nfsuquad[1] = txdr_unsigned(sf->f_files); + sfp->sf_ffiles.nfsuquad[0] = 0; + sfp->sf_ffiles.nfsuquad[1] = txdr_unsigned(sf->f_ffree); + sfp->sf_afiles.nfsuquad[0] = 0; + sfp->sf_afiles.nfsuquad[1] = txdr_unsigned(sf->f_ffree); + sfp->sf_invarsec = 0; + } else { + sfp->sf_tsize = txdr_unsigned(NFS_MAXDGRAMDATA); + sfp->sf_bsize = txdr_unsigned(sf->f_bsize); + sfp->sf_blocks = txdr_unsigned(sf->f_blocks); + sfp->sf_bfree = txdr_unsigned(sf->f_bfree); + sfp->sf_bavail = txdr_unsigned(sf->f_bavail); } nfsm_srvdone; } /* + * nfs fsinfo service + */ +int +nfsrv_fsinfo(nfsd, slp, procp, mrq) + struct nfsrv_descript *nfsd; + struct nfssvc_sock *slp; + struct proc *procp; + struct mbuf **mrq; +{ + struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; + struct mbuf *nam = nfsd->nd_nam; + caddr_t dpos = nfsd->nd_dpos; + struct ucred *cred = &nfsd->nd_cr; + register u_long *tl; + register struct nfsv3_fsinfo *sip; + register long t1; + caddr_t bpos; + int error = 0, rdonly, cache, getret = 1, pref; + char *cp2; + struct mbuf *mb, *mb2, *mreq; + struct vnode *vp; + struct vattr at; + nfsfh_t nfh; + fhandle_t *fhp; + u_quad_t frev; + +#ifndef nolint + cache = 0; +#endif + fhp = &nfh.fh_generic; + nfsm_srvmtofh(fhp); + if (error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam, + &rdonly, (nfsd->nd_flag & ND_KERBAUTH))) { + nfsm_reply(NFSX_UNSIGNED); + nfsm_srvpostop_attr(getret, &at); + return (0); + } + getret = VOP_GETATTR(vp, &at, cred, procp); + nfsrv_vput(vp); + nfsm_reply(NFSX_V3POSTOPATTR + NFSX_V3FSINFO); + nfsm_srvpostop_attr(getret, &at); + nfsm_build(sip, struct nfsv3_fsinfo *, NFSX_V3FSINFO); + + /* + * XXX + * There should be file system VFS OP(s) to get this information. + * For now, assume ufs. + */ + if (slp->ns_so->so_type == SOCK_DGRAM) + pref = NFS_MAXDGRAMDATA; + else + pref = NFS_MAXDATA; + sip->fs_rtmax = txdr_unsigned(NFS_MAXDATA); + sip->fs_rtpref = txdr_unsigned(pref); + sip->fs_rtmult = txdr_unsigned(NFS_FABLKSIZE); + sip->fs_wtmax = txdr_unsigned(NFS_MAXDATA); + sip->fs_wtpref = txdr_unsigned(pref); + sip->fs_wtmult = txdr_unsigned(NFS_FABLKSIZE); + sip->fs_dtpref = txdr_unsigned(pref); + sip->fs_maxfilesize.nfsuquad[0] = 0xffffffff; + sip->fs_maxfilesize.nfsuquad[1] = 0xffffffff; + sip->fs_timedelta.nfsv3_sec = 0; + sip->fs_timedelta.nfsv3_nsec = txdr_unsigned(1); + sip->fs_properties = txdr_unsigned(NFSV3FSINFO_LINK | + NFSV3FSINFO_SYMLINK | NFSV3FSINFO_HOMOGENEOUS | + NFSV3FSINFO_CANSETTIME); + nfsm_srvdone; +} + +/* + * nfs pathconf service + */ +int +nfsrv_pathconf(nfsd, slp, procp, mrq) + struct nfsrv_descript *nfsd; + struct nfssvc_sock *slp; + struct proc *procp; + struct mbuf **mrq; +{ + struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; + struct mbuf *nam = nfsd->nd_nam; + caddr_t dpos = nfsd->nd_dpos; + struct ucred *cred = &nfsd->nd_cr; + register u_long *tl; + register struct nfsv3_pathconf *pc; + register long t1; + caddr_t bpos; + int error = 0, rdonly, cache, getret = 1, linkmax, namemax; + int chownres, notrunc; + char *cp2; + struct mbuf *mb, *mb2, *mreq; + struct vnode *vp; + struct vattr at; + nfsfh_t nfh; + fhandle_t *fhp; + u_quad_t frev; + +#ifndef nolint + cache = 0; +#endif + fhp = &nfh.fh_generic; + nfsm_srvmtofh(fhp); + if (error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam, + &rdonly, (nfsd->nd_flag & ND_KERBAUTH))) { + nfsm_reply(NFSX_UNSIGNED); + nfsm_srvpostop_attr(getret, &at); + return (0); + } + error = VOP_PATHCONF(vp, _PC_LINK_MAX, &linkmax); + if (!error) + error = VOP_PATHCONF(vp, _PC_NAME_MAX, &namemax); + if (!error) + error = VOP_PATHCONF(vp, _PC_CHOWN_RESTRICTED, &chownres); + if (!error) + error = VOP_PATHCONF(vp, _PC_NO_TRUNC, ¬runc); + getret = VOP_GETATTR(vp, &at, cred, procp); + nfsrv_vput(vp); + nfsm_reply(NFSX_V3POSTOPATTR + NFSX_V3PATHCONF); + nfsm_srvpostop_attr(getret, &at); + if (error) + return (0); + nfsm_build(pc, struct nfsv3_pathconf *, NFSX_V3PATHCONF); + + pc->pc_linkmax = txdr_unsigned(linkmax); + pc->pc_namemax = txdr_unsigned(namemax); + pc->pc_notrunc = txdr_unsigned(notrunc); + pc->pc_chownrestricted = txdr_unsigned(chownres); + + /* + * These should probably be supported by VOP_PATHCONF(), but + * until msdosfs is exportable (why would you want to?), the + * Unix defaults should be ok. + */ + pc->pc_caseinsensitive = nfs_false; + pc->pc_casepreserving = nfs_true; + nfsm_srvdone; +} + +/* * Null operation, used by clients to ping server */ /* ARGSUSED */ int -nfsrv_null(nfsd, mrep, md, dpos, cred, nam, mrq) - struct nfsd *nfsd; - struct mbuf *mrep, *md; - caddr_t dpos; - struct ucred *cred; - struct mbuf *nam, **mrq; +nfsrv_null(nfsd, slp, procp, mrq) + struct nfsrv_descript *nfsd; + struct nfssvc_sock *slp; + struct proc *procp; + struct mbuf **mrq; { + struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; + struct mbuf *nam = nfsd->nd_nam; + caddr_t dpos = nfsd->nd_dpos; + struct ucred *cred = &nfsd->nd_cr; caddr_t bpos; - int error = VNOVAL, cache = 0; + int error = NFSERR_RETVOID, cache; struct mbuf *mb, *mreq; u_quad_t frev; +#ifndef nolint + cache = 0; +#endif nfsm_reply(0); - return (error); + return (0); } /* @@ -1946,24 +3333,30 @@ nfsrv_null(nfsd, mrep, md, dpos, cred, nam, mrq) */ /* ARGSUSED */ int -nfsrv_noop(nfsd, mrep, md, dpos, cred, nam, mrq) - struct nfsd *nfsd; - struct mbuf *mrep, *md; - caddr_t dpos; - struct ucred *cred; - struct mbuf *nam, **mrq; +nfsrv_noop(nfsd, slp, procp, mrq) + struct nfsrv_descript *nfsd; + struct nfssvc_sock *slp; + struct proc *procp; + struct mbuf **mrq; { + struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; + struct mbuf *nam = nfsd->nd_nam; + caddr_t dpos = nfsd->nd_dpos; + struct ucred *cred = &nfsd->nd_cr; caddr_t bpos; - int error, cache = 0; + int error, cache; struct mbuf *mb, *mreq; u_quad_t frev; +#ifndef nolint + cache = 0; +#endif if (nfsd->nd_repstat) error = nfsd->nd_repstat; else error = EPROCUNAVAIL; nfsm_reply(0); - return (error); + return (0); } /* @@ -1997,8 +3390,6 @@ nfsrv_access(vp, flags, cred, rdonly, p) switch (vp->v_type) { case VREG: case VDIR: case VLNK: return (EROFS); - default: - break; } } /* @@ -2008,8 +3399,7 @@ nfsrv_access(vp, flags, cred, rdonly, p) if (vp->v_flag & VTEXT) return (ETXTBSY); } - error = VOP_GETATTR(vp, &vattr, cred, p); - if (error) + if (error = VOP_GETATTR(vp, &vattr, cred, p)) return (error); if ((error = VOP_ACCESS(vp, flags, cred, p)) && cred->cr_uid != vattr.va_uid) diff --git a/sys/nfs/nfs_socket.c b/sys/nfs/nfs_socket.c index 20253ef..8d15b24 100644 --- a/sys/nfs/nfs_socket.c +++ b/sys/nfs/nfs_socket.c @@ -34,7 +34,7 @@ * SUCH DAMAGE. * * @(#)nfs_socket.c 8.3 (Berkeley) 1/12/94 - * $Id: nfs_socket.c,v 1.6 1995/03/16 18:15:37 bde Exp $ + * $Id: nfs_socket.c,v 1.7 1995/05/30 08:12:40 rgrimes Exp $ */ /* @@ -59,7 +59,7 @@ #include <netinet/tcp.h> #include <nfs/rpcv2.h> -#include <nfs/nfsv2.h> +#include <nfs/nfsproto.h> #include <nfs/nfs.h> #include <nfs/xdr_subs.h> #include <nfs/nfsm_subs.h> @@ -94,36 +94,13 @@ * External data, mostly RPC constants in XDR form */ extern u_long rpc_reply, rpc_msgdenied, rpc_mismatch, rpc_vers, rpc_auth_unix, - rpc_msgaccepted, rpc_call, rpc_autherr, rpc_rejectedcred, + rpc_msgaccepted, rpc_call, rpc_autherr, rpc_auth_kerb; -extern u_long nfs_prog, nfs_vers, nqnfs_prog, nqnfs_vers; +extern u_long nfs_prog, nqnfs_prog; extern time_t nqnfsstarttime; -extern int nonidempotent[NFS_NPROCS]; - -/* - * Maps errno values to nfs error numbers. - * Use NFSERR_IO as the catch all for ones not specifically defined in - * RFC 1094. - */ -static int nfsrv_errmap[ELAST] = { - NFSERR_PERM, NFSERR_NOENT, NFSERR_IO, NFSERR_IO, NFSERR_IO, - NFSERR_NXIO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, - NFSERR_IO, NFSERR_IO, NFSERR_ACCES, NFSERR_IO, NFSERR_IO, - NFSERR_IO, NFSERR_EXIST, NFSERR_IO, NFSERR_NODEV, NFSERR_NOTDIR, - NFSERR_ISDIR, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, - NFSERR_IO, NFSERR_FBIG, NFSERR_NOSPC, NFSERR_IO, NFSERR_ROFS, - NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, - NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, - NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, - NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, - NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, - NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, - NFSERR_IO, NFSERR_IO, NFSERR_NAMETOL, NFSERR_IO, NFSERR_IO, - NFSERR_NOTEMPTY, NFSERR_IO, NFSERR_IO, NFSERR_DQUOT, NFSERR_STALE, - NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, - NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, - NFSERR_IO, -}; +extern struct nfsstats nfsstats; +extern int nfsv3_procid[NFS_NPROCS]; +extern int nfs_ticks; /* * Defines which timer to use for the procnum. @@ -134,7 +111,8 @@ static int nfsrv_errmap[ELAST] = { * 4 - write */ static int proct[NFS_NPROCS] = { - 0, 1, 0, 0, 2, 3, 3, 0, 4, 0, 0, 0, 0, 0, 0, 0, 3, 0, 3, 0, 0, 0, 0, + 0, 1, 0, 2, 1, 3, 3, 4, 0, 0, 0, 0, 0, 0, 0, 0, 3, 3, 0, 0, 0, 0, 0, + 0, 0, 0, }; /* @@ -164,10 +142,12 @@ int nfsrv_null(), nfsrv_getattr(), nfsrv_setattr(), nfsrv_lookup(), + nfsrv3_access(), nfsrv_readlink(), nfsrv_read(), nfsrv_write(), nfsrv_create(), + nfsrv_mknod(), nfsrv_remove(), nfsrv_rename(), nfsrv_link(), @@ -175,37 +155,42 @@ int nfsrv_null(), nfsrv_mkdir(), nfsrv_rmdir(), nfsrv_readdir(), + nfsrv_readdirplus(), nfsrv_statfs(), + nfsrv_fsinfo(), + nfsrv_pathconf(), + nfsrv_commit(), nfsrv_noop(), - nqnfsrv_readdirlook(), nqnfsrv_getlease(), - nqnfsrv_vacated(), - nqnfsrv_access(); + nqnfsrv_vacated(); -int (*nfsrv_procs[NFS_NPROCS])() = { +int (*nfsrv3_procs[NFS_NPROCS])() = { nfsrv_null, nfsrv_getattr, nfsrv_setattr, - nfsrv_noop, nfsrv_lookup, + nfsrv3_access, nfsrv_readlink, nfsrv_read, - nfsrv_noop, nfsrv_write, nfsrv_create, + nfsrv_mkdir, + nfsrv_symlink, + nfsrv_mknod, nfsrv_remove, + nfsrv_rmdir, nfsrv_rename, nfsrv_link, - nfsrv_symlink, - nfsrv_mkdir, - nfsrv_rmdir, nfsrv_readdir, + nfsrv_readdirplus, nfsrv_statfs, - nqnfsrv_readdirlook, + nfsrv_fsinfo, + nfsrv_pathconf, + nfsrv_commit, nqnfsrv_getlease, nqnfsrv_vacated, nfsrv_noop, - nqnfsrv_access, + nfsrv_noop }; /* @@ -874,13 +859,16 @@ nfs_request(vp, mrest, procnum, procp, cred, mrp, mdp, dposp) struct nfsmount *nmp; struct mbuf *md, *mheadend; struct nfsnode *np; + char nickv[RPCX_NICKVERF]; time_t reqtime, waituntil; caddr_t dpos, cp2; int t1, nqlflag, cachable, s, error = 0, mrest_len, auth_len, auth_type; int trylater_delay = NQ_TRYLATERDEL, trylater_cnt = 0, failed_auth = 0; + int verf_len, verf_type; u_long xid; u_quad_t frev; - char *auth_str; + char *auth_str, *verf_str; + NFSKERBKEY_T key; /* save session key */ nmp = VFSTONFS(vp->v_mount); MALLOC(rep, struct nfsreq *, sizeof(struct nfsreq), M_NFSREQ, M_WAITOK); @@ -900,19 +888,21 @@ nfs_request(vp, mrest, procnum, procp, cred, mrp, mdp, dposp) * Get the RPC header with authorization. */ kerbauth: - auth_str = (char *)0; + verf_str = auth_str = (char *)0; if (nmp->nm_flag & NFSMNT_KERB) { - if (failed_auth) { - error = nfs_getauth(nmp, rep, cred, &auth_type, - &auth_str, &auth_len); + verf_str = nickv; + verf_len = sizeof (nickv); + auth_type = RPCAUTH_KERB4; + bzero((caddr_t)key, sizeof (key)); + if (failed_auth || nfs_getnickauth(nmp, cred, &auth_str, + &auth_len, verf_str, verf_len)) { + error = nfs_getauth(nmp, rep, cred, &auth_str, + &auth_len, verf_str, &verf_len, key); if (error) { free((caddr_t)rep, M_NFSREQ); m_freem(mrest); return (error); } - } else { - auth_type = RPCAUTH_UNIX; - auth_len = 5 * NFSX_UNSIGNED; } } else { auth_type = RPCAUTH_UNIX; @@ -922,8 +912,8 @@ kerbauth: nmp->nm_numgrps : (cred->cr_ngroups - 1)) << 2) + 5 * NFSX_UNSIGNED; } - m = nfsm_rpchead(cred, (nmp->nm_flag & NFSMNT_NQNFS), procnum, - auth_type, auth_len, auth_str, mrest, mrest_len, &mheadend, &xid); + m = nfsm_rpchead(cred, nmp->nm_flag, procnum, auth_type, auth_len, + auth_str, verf_len, verf_str, mrest, mrest_len, &mheadend, &xid); if (auth_str) free(auth_str, M_TEMP); @@ -1029,12 +1019,12 @@ tryagain: /* * break down the rpc header and check if ok */ - nfsm_dissect(tl, u_long *, 3*NFSX_UNSIGNED); + nfsm_dissect(tl, u_long *, 3 * NFSX_UNSIGNED); if (*tl++ == rpc_msgdenied) { if (*tl == rpc_mismatch) error = EOPNOTSUPP; else if ((nmp->nm_flag & NFSMNT_KERB) && *tl++ == rpc_autherr) { - if (*tl == rpc_rejectedcred && failed_auth == 0) { + if (!failed_auth) { failed_auth++; mheadend->m_next = (struct mbuf *)0; m_freem(mrep); @@ -1051,22 +1041,25 @@ tryagain: } /* - * skip over the auth_verf, someday we may want to cache auth_short's - * for nfs_reqhead(), but for now just dump it + * Grab any Kerberos verifier, otherwise just throw it away. */ - if (*++tl != 0) { - i = nfsm_rndup(fxdr_unsigned(long, *tl)); - nfsm_adv(i); - } + verf_type = fxdr_unsigned(int, *tl++); + i = fxdr_unsigned(int, *tl); + if ((nmp->nm_flag & NFSMNT_KERB) && verf_type == RPCAUTH_KERB4) { + error = nfs_savenickauth(nmp, cred, i, key, &md, &dpos, mrep); + if (error) + goto nfsmout; + } else if (i > 0) + nfsm_adv(nfsm_rndup(i)); nfsm_dissect(tl, u_long *, NFSX_UNSIGNED); /* 0 == ok */ if (*tl == 0) { nfsm_dissect(tl, u_long *, NFSX_UNSIGNED); if (*tl != 0) { error = fxdr_unsigned(int, *tl); - m_freem(mrep); - if ((nmp->nm_flag & NFSMNT_NQNFS) && - error == NQNFS_TRYLATER) { + if ((nmp->nm_flag & NFSMNT_NFSV3) && + error == NFSERR_TRYLATER) { + m_freem(mrep); error = 0; waituntil = time.tv_sec + trylater_delay; while (time.tv_sec < waituntil) @@ -1084,6 +1077,13 @@ tryagain: */ if (error == ESTALE) cache_purge(vp); + if (nmp->nm_flag & NFSMNT_NFSV3) { + *mrp = mrep; + *mdp = md; + *dposp = dpos; + error |= NFSERR_RETERR; + } else + m_freem(mrep); m_freem(rep->r_mreq); free((caddr_t)rep, M_NFSREQ); return (error); @@ -1115,10 +1115,10 @@ tryagain: return (0); } m_freem(mrep); - m_freem(rep->r_mreq); - free((caddr_t)rep, M_NFSREQ); error = EPROTONOSUPPORT; nfsmout: + m_freem(rep->r_mreq); + free((caddr_t)rep, M_NFSREQ); return (error); } @@ -1127,9 +1127,10 @@ nfsmout: * siz arg. is used to decide if adding a cluster is worthwhile */ int -nfs_rephead(siz, nd, err, cache, frev, mrq, mbp, bposp) +nfs_rephead(siz, nd, slp, err, cache, frev, mrq, mbp, bposp) int siz; - struct nfsd *nd; + struct nfsrv_descript *nd; + struct nfssvc_sock *slp; int err; int cache; u_quad_t *frev; @@ -1154,47 +1155,98 @@ nfs_rephead(siz, nd, err, cache, frev, mrq, mbp, bposp) } else mreq->m_data += max_hdr; tl = mtod(mreq, u_long *); - mreq->m_len = 6*NFSX_UNSIGNED; - bpos = ((caddr_t)tl)+mreq->m_len; + mreq->m_len = 6 * NFSX_UNSIGNED; + bpos = ((caddr_t)tl) + mreq->m_len; *tl++ = txdr_unsigned(nd->nd_retxid); *tl++ = rpc_reply; - if (err == ERPCMISMATCH || err == NQNFS_AUTHERR) { + if (err == ERPCMISMATCH || (err & NFSERR_AUTHERR)) { *tl++ = rpc_msgdenied; - if (err == NQNFS_AUTHERR) { + if (err & NFSERR_AUTHERR) { *tl++ = rpc_autherr; - *tl = rpc_rejectedcred; + *tl = txdr_unsigned(err & ~NFSERR_AUTHERR); mreq->m_len -= NFSX_UNSIGNED; bpos -= NFSX_UNSIGNED; } else { *tl++ = rpc_mismatch; - *tl++ = txdr_unsigned(2); - *tl = txdr_unsigned(2); + *tl++ = txdr_unsigned(RPC_VER2); + *tl = txdr_unsigned(RPC_VER2); } } else { *tl++ = rpc_msgaccepted; - *tl++ = 0; - *tl++ = 0; + + /* + * For Kerberos authentication, we must send the nickname + * verifier back, otherwise just RPCAUTH_NULL. + */ + if (nd->nd_flag & ND_KERBFULL) { + register struct nfsuid *nuidp; + struct timeval ktvin, ktvout; + NFSKERBKEYSCHED_T keys; /* stores key schedule */ + + for (nuidp = NUIDHASH(slp, nd->nd_cr.cr_uid)->lh_first; + nuidp != 0; nuidp = nuidp->nu_hash.le_next) { + if (nuidp->nu_cr.cr_uid == nd->nd_cr.cr_uid && + (!nd->nd_nam2 || netaddr_match(NU_NETFAM(nuidp), + &nuidp->nu_haddr, nd->nd_nam2))) + break; + } + if (nuidp) { + ktvin.tv_sec = + txdr_unsigned(nuidp->nu_timestamp.tv_sec - 1); + ktvin.tv_usec = + txdr_unsigned(nuidp->nu_timestamp.tv_usec); + + /* + * Encrypt the timestamp in ecb mode using the + * session key. + */ +#ifdef NFSKERB + XXX +#endif + + *tl++ = rpc_auth_kerb; + *tl++ = txdr_unsigned(3 * NFSX_UNSIGNED); + *tl = ktvout.tv_sec; + nfsm_build(tl, u_long *, 3 * NFSX_UNSIGNED); + *tl++ = ktvout.tv_usec; + *tl++ = txdr_unsigned(nuidp->nu_cr.cr_uid); + } else { + *tl++ = 0; + *tl++ = 0; + } + } else { + *tl++ = 0; + *tl++ = 0; + } switch (err) { case EPROGUNAVAIL: *tl = txdr_unsigned(RPC_PROGUNAVAIL); break; case EPROGMISMATCH: *tl = txdr_unsigned(RPC_PROGMISMATCH); - nfsm_build(tl, u_long *, 2*NFSX_UNSIGNED); - *tl++ = txdr_unsigned(2); - *tl = txdr_unsigned(2); /* someday 3 */ + nfsm_build(tl, u_long *, 2 * NFSX_UNSIGNED); + if (nd->nd_flag & ND_NQNFS) { + *tl++ = txdr_unsigned(3); + *tl = txdr_unsigned(3); + } else { + *tl++ = txdr_unsigned(2); + *tl = txdr_unsigned(3); + } break; case EPROCUNAVAIL: *tl = txdr_unsigned(RPC_PROCUNAVAIL); break; + case EBADRPC: + *tl = txdr_unsigned(RPC_GARBAGE); + break; default: *tl = 0; - if (err != VNOVAL) { + if (err != NFSERR_RETVOID) { nfsm_build(tl, u_long *, NFSX_UNSIGNED); if (err) - *tl = txdr_unsigned(nfsrv_errmap[err - 1]); + *tl = txdr_unsigned(nfsrv_errmap(nd, err)); else - *tl = 0; + *tl = 0; } break; }; @@ -1203,16 +1255,14 @@ nfs_rephead(siz, nd, err, cache, frev, mrq, mbp, bposp) /* * For nqnfs, piggyback lease as requested. */ - if (nd->nd_nqlflag != NQL_NOVAL && err == 0) { - if (nd->nd_nqlflag) { - nfsm_build(tl, u_long *, 5*NFSX_UNSIGNED); - *tl++ = txdr_unsigned(nd->nd_nqlflag); + if ((nd->nd_flag & ND_NQNFS) && err == 0) { + if (nd->nd_flag & ND_LEASE) { + nfsm_build(tl, u_long *, 5 * NFSX_UNSIGNED); + *tl++ = txdr_unsigned(nd->nd_flag & ND_LEASE); *tl++ = txdr_unsigned(cache); *tl++ = txdr_unsigned(nd->nd_duration); txdr_hyper(frev, tl); } else { - if (nd->nd_nqlflag != 0) - panic("nqreph"); nfsm_build(tl, u_long *, NFSX_UNSIGNED); *tl = 0; } @@ -1220,7 +1270,7 @@ nfs_rephead(siz, nd, err, cache, frev, mrq, mbp, bposp) *mrq = mreq; *mbp = mb; *bposp = bpos; - if (err != 0 && err != VNOVAL) + if (err != 0 && err != NFSERR_RETVOID) nfsstats.srvrpc_errs++; return (0); } @@ -1240,8 +1290,10 @@ nfs_timer(arg) register struct socket *so; register struct nfsmount *nmp; register int timeo; + register struct nfssvc_sock *slp; static long lasttime = 0; int s, error; + u_quad_t cur_usec; s = splnet(); for (rep = nfs_reqq.tqh_first; rep != 0; rep = rep->r_chain.tqe_next) { @@ -1338,8 +1390,19 @@ nfs_timer(arg) lasttime = time.tv_sec; nqnfs_serverd(); } + + /* + * Scan the write gathering queues for writes that need to be + * completed now. + */ + cur_usec = (u_quad_t)time.tv_sec * 1000000 + (u_quad_t)time.tv_usec; + for (slp = nfssvc_sockhead.tqh_first; slp != 0; + slp = slp->ns_chain.tqe_next) { + if (slp->ns_tq.lh_first && slp->ns_tq.lh_first->nd_time<=cur_usec) + nfsrv_wakenfsd(slp); + } splx(s); - timeout(nfs_timer, (void *)0, hz / NFS_HZ); + timeout(nfs_timer, (void *)0, nfs_ticks); } /* @@ -1673,7 +1736,7 @@ nfsrv_getstream(slp, waitflag) register struct nfssvc_sock *slp; int waitflag; { - register struct mbuf *m; + register struct mbuf *m, **mpp; register char *cp1, *cp2; register int len; struct mbuf *om, *m2, *recm = 0; @@ -1707,7 +1770,12 @@ nfsrv_getstream(slp, waitflag) } } slp->ns_cc -= NFSX_UNSIGNED; - slp->ns_reclen = ntohl(recmark) & ~0x80000000; + recmark = ntohl(recmark); + slp->ns_reclen = recmark & ~0x80000000; + if (recmark & 0x80000000) + slp->ns_flag |= SLP_LASTFRAG; + else + slp->ns_flag &= ~SLP_LASTFRAG; if (slp->ns_reclen < NFS_MINPACKET || slp->ns_reclen > NFS_MAXPACKET) { slp->ns_flag &= ~SLP_GETSTREAM; return (EPERM); @@ -1761,12 +1829,23 @@ nfsrv_getstream(slp, waitflag) slp->ns_flag &= ~SLP_GETSTREAM; return (0); } - nfs_realign(recm, 10 * NFSX_UNSIGNED); - if (slp->ns_recend) - slp->ns_recend->m_nextpkt = recm; - else - slp->ns_rec = recm; - slp->ns_recend = recm; + + /* + * Accumulate the fragments into a record. + */ + mpp = &slp->ns_frag; + while (*mpp) + mpp = &((*mpp)->m_next); + *mpp = recm; + if (slp->ns_flag & SLP_LASTFRAG) { + nfs_realign(slp->ns_frag, 10 * NFSX_UNSIGNED); + if (slp->ns_recend) + slp->ns_recend->m_nextpkt = slp->ns_frag; + else + slp->ns_rec = slp->ns_frag; + slp->ns_recend = slp->ns_frag; + slp->ns_frag = (struct mbuf *)0; + } } } @@ -1774,13 +1853,16 @@ nfsrv_getstream(slp, waitflag) * Parse an RPC header. */ int -nfsrv_dorec(slp, nd) +nfsrv_dorec(slp, nfsd, ndp) register struct nfssvc_sock *slp; - register struct nfsd *nd; + struct nfsd *nfsd; + struct nfsrv_descript **ndp; { - register struct mbuf *m; + register struct mbuf *m, *nam; + register struct nfsrv_descript *nd; int error; + *ndp = NULL; if ((slp->ns_flag & SLP_VALID) == 0 || (m = slp->ns_rec) == (struct mbuf *)0) return (ENOBUFS); @@ -1790,19 +1872,24 @@ nfsrv_dorec(slp, nd) else slp->ns_recend = (struct mbuf *)0; if (m->m_type == MT_SONAME) { - nd->nd_nam = m; - nd->nd_md = nd->nd_mrep = m->m_next; - m->m_next = (struct mbuf *)0; - } else { - nd->nd_nam = (struct mbuf *)0; - nd->nd_md = nd->nd_mrep = m; - } - nd->nd_dpos = mtod(nd->nd_md, caddr_t); - error = nfs_getreq(nd, TRUE); + nam = m; + m = m->m_next; + nam->m_next = NULL; + } else + nam = NULL; + MALLOC(nd, struct nfsrv_descript *, sizeof (struct nfsrv_descript), + M_NFSRVDESC, M_WAITOK); + nd->nd_md = nd->nd_mrep = m; + nd->nd_nam2 = nam; + nd->nd_dpos = mtod(m, caddr_t); + error = nfs_getreq(nd, nfsd, TRUE); if (error) { - m_freem(nd->nd_nam); + m_freem(nam); + free((caddr_t)nd, M_NFSRVDESC); return (error); } + *ndp = nd; + nfsd->nfsd_nd = nd; return (0); } @@ -1812,8 +1899,9 @@ nfsrv_dorec(slp, nd) * - fill in the cred struct. */ int -nfs_getreq(nd, has_header) - register struct nfsd *nd; +nfs_getreq(nd, nfsd, has_header) + register struct nfsrv_descript *nd; + struct nfsd *nfsd; int has_header; { register int len, i; @@ -1821,57 +1909,67 @@ nfs_getreq(nd, has_header) register long t1; struct uio uio; struct iovec iov; - caddr_t dpos, cp2; + caddr_t dpos, cp2, cp; u_long nfsvers, auth_type; - int error = 0, nqnfs = 0; + uid_t nickuid; + int error = 0, nqnfs = 0, ticklen; struct mbuf *mrep, *md; + register struct nfsuid *nuidp; + struct timeval tvin, tvout; + NFSKERBKEYSCHED_T keys; /* stores key schedule */ mrep = nd->nd_mrep; md = nd->nd_md; dpos = nd->nd_dpos; if (has_header) { - nfsm_dissect(tl, u_long *, 10*NFSX_UNSIGNED); + nfsm_dissect(tl, u_long *, 10 * NFSX_UNSIGNED); nd->nd_retxid = fxdr_unsigned(u_long, *tl++); if (*tl++ != rpc_call) { m_freem(mrep); return (EBADRPC); } - } else { - nfsm_dissect(tl, u_long *, 8*NFSX_UNSIGNED); - } + } else + nfsm_dissect(tl, u_long *, 8 * NFSX_UNSIGNED); nd->nd_repstat = 0; + nd->nd_flag = 0; if (*tl++ != rpc_vers) { nd->nd_repstat = ERPCMISMATCH; nd->nd_procnum = NFSPROC_NOOP; return (0); } - nfsvers = nfs_vers; if (*tl != nfs_prog) { - if (*tl == nqnfs_prog) { + if (*tl == nqnfs_prog) nqnfs++; - nfsvers = nqnfs_vers; - } else { + else { nd->nd_repstat = EPROGUNAVAIL; nd->nd_procnum = NFSPROC_NOOP; return (0); } } tl++; - if (*tl++ != nfsvers) { + nfsvers = fxdr_unsigned(u_long, *tl++); + if (((nfsvers < NFS_VER2 || nfsvers > NFS_VER3) && !nqnfs) || + (nfsvers != NQNFS_VER3 && nqnfs)) { nd->nd_repstat = EPROGMISMATCH; nd->nd_procnum = NFSPROC_NOOP; return (0); } + if (nqnfs) + nd->nd_flag = (ND_NFSV3 | ND_NQNFS); + else if (nfsvers == NFS_VER3) + nd->nd_flag = ND_NFSV3; nd->nd_procnum = fxdr_unsigned(u_long, *tl++); if (nd->nd_procnum == NFSPROC_NULL) return (0); if (nd->nd_procnum >= NFS_NPROCS || - (!nqnfs && nd->nd_procnum > NFSPROC_STATFS) || - (*tl != rpc_auth_unix && *tl != rpc_auth_kerb)) { + (!nqnfs && nd->nd_procnum >= NQNFSPROC_GETLEASE) || + (!nd->nd_flag && nd->nd_procnum > NFSV2PROC_STATFS)) { nd->nd_repstat = EPROCUNAVAIL; nd->nd_procnum = NFSPROC_NOOP; return (0); } + if ((nd->nd_flag & ND_NFSV3) == 0) + nd->nd_procnum = nfsv3_procid[nd->nd_procnum]; auth_type = *tl++; len = fxdr_unsigned(int, *tl++); if (len < 0 || len > RPCAUTH_MAXSIZ) { @@ -1879,6 +1977,7 @@ nfs_getreq(nd, has_header) return (EBADRPC); } + nd->nd_flag &= ~ND_KERBAUTH; /* * Handle auth_unix or auth_kerb. */ @@ -1889,7 +1988,9 @@ nfs_getreq(nd, has_header) return (EBADRPC); } nfsm_adv(nfsm_rndup(len)); - nfsm_dissect(tl, u_long *, 3*NFSX_UNSIGNED); + nfsm_dissect(tl, u_long *, 3 * NFSX_UNSIGNED); + bzero((caddr_t)&nd->nd_cr, sizeof (struct ucred)); + nd->nd_cr.cr_ref = 1; nd->nd_cr.cr_uid = fxdr_unsigned(uid_t, *tl++); nd->nd_cr.cr_gid = fxdr_unsigned(gid_t, *tl++); len = fxdr_unsigned(int, *tl); @@ -1897,45 +1998,124 @@ nfs_getreq(nd, has_header) m_freem(mrep); return (EBADRPC); } - nfsm_dissect(tl, u_long *, (len + 2)*NFSX_UNSIGNED); + nfsm_dissect(tl, u_long *, (len + 2) * NFSX_UNSIGNED); for (i = 1; i <= len; i++) - if (i < NGROUPS) - nd->nd_cr.cr_groups[i] = fxdr_unsigned(gid_t, *tl++); - else - tl++; + if (i < NGROUPS) + nd->nd_cr.cr_groups[i] = fxdr_unsigned(gid_t, *tl++); + else + tl++; nd->nd_cr.cr_ngroups = (len >= NGROUPS) ? NGROUPS : (len + 1); - } else if (auth_type == rpc_auth_kerb) { - nd->nd_cr.cr_uid = fxdr_unsigned(uid_t, *tl++); - nd->nd_authlen = fxdr_unsigned(int, *tl); - uio.uio_resid = nfsm_rndup(nd->nd_authlen); - if (uio.uio_resid > (len - 2 * NFSX_UNSIGNED)) { + if (nd->nd_cr.cr_ngroups > 1) + nfsrvw_sort(nd->nd_cr.cr_groups, nd->nd_cr.cr_ngroups); + len = fxdr_unsigned(int, *++tl); + if (len < 0 || len > RPCAUTH_MAXSIZ) { m_freem(mrep); return (EBADRPC); } - uio.uio_offset = 0; - uio.uio_iov = &iov; - uio.uio_iovcnt = 1; - uio.uio_segflg = UIO_SYSSPACE; - iov.iov_base = (caddr_t)nd->nd_authstr; - iov.iov_len = RPCAUTH_MAXSIZ; - nfsm_mtouio(&uio, uio.uio_resid); - nfsm_dissect(tl, u_long *, 2 * NFSX_UNSIGNED); - nd->nd_flag |= NFSD_NEEDAUTH; - } + if (len > 0) + nfsm_adv(nfsm_rndup(len)); + } else if (auth_type == rpc_auth_kerb) { + switch (fxdr_unsigned(int, *tl++)) { + case RPCAKN_FULLNAME: + ticklen = fxdr_unsigned(int, *tl); + *((u_long *)nfsd->nfsd_authstr) = *tl; + uio.uio_resid = nfsm_rndup(ticklen) + NFSX_UNSIGNED; + nfsd->nfsd_authlen = uio.uio_resid + NFSX_UNSIGNED; + if (uio.uio_resid > (len - 2 * NFSX_UNSIGNED)) { + m_freem(mrep); + return (EBADRPC); + } + uio.uio_offset = 0; + uio.uio_iov = &iov; + uio.uio_iovcnt = 1; + uio.uio_segflg = UIO_SYSSPACE; + iov.iov_base = (caddr_t)&nfsd->nfsd_authstr[4]; + iov.iov_len = RPCAUTH_MAXSIZ - 4; + nfsm_mtouio(&uio, uio.uio_resid); + nfsm_dissect(tl, u_long *, 2 * NFSX_UNSIGNED); + if (*tl++ != rpc_auth_kerb || + fxdr_unsigned(int, *tl) != 4 * NFSX_UNSIGNED) { + printf("Bad kerb verifier\n"); + nd->nd_repstat = (NFSERR_AUTHERR|AUTH_BADVERF); + nd->nd_procnum = NFSPROC_NOOP; + return (0); + } + nfsm_dissect(cp, caddr_t, 4 * NFSX_UNSIGNED); + tl = (u_long *)cp; + if (fxdr_unsigned(int, *tl) != RPCAKN_FULLNAME) { + printf("Not fullname kerb verifier\n"); + nd->nd_repstat = (NFSERR_AUTHERR|AUTH_BADVERF); + nd->nd_procnum = NFSPROC_NOOP; + return (0); + } + cp += NFSX_UNSIGNED; + bcopy(cp, nfsd->nfsd_verfstr, 3 * NFSX_UNSIGNED); + nfsd->nfsd_verflen = 3 * NFSX_UNSIGNED; + nd->nd_flag |= ND_KERBFULL; + nfsd->nfsd_flag |= NFSD_NEEDAUTH; + break; + case RPCAKN_NICKNAME: + if (len != 2 * NFSX_UNSIGNED) { + printf("Kerb nickname short\n"); + nd->nd_repstat = (NFSERR_AUTHERR|AUTH_BADCRED); + nd->nd_procnum = NFSPROC_NOOP; + return (0); + } + nickuid = fxdr_unsigned(uid_t, *tl); + nfsm_dissect(tl, u_long *, 2 * NFSX_UNSIGNED); + if (*tl++ != rpc_auth_kerb || + fxdr_unsigned(int, *tl) != 3 * NFSX_UNSIGNED) { + printf("Kerb nick verifier bad\n"); + nd->nd_repstat = (NFSERR_AUTHERR|AUTH_BADVERF); + nd->nd_procnum = NFSPROC_NOOP; + return (0); + } + nfsm_dissect(tl, u_long *, 3 * NFSX_UNSIGNED); + tvin.tv_sec = *tl++; + tvin.tv_usec = *tl; + + for (nuidp = NUIDHASH(nfsd->nfsd_slp,nickuid)->lh_first; + nuidp != 0; nuidp = nuidp->nu_hash.le_next) { + if (nuidp->nu_cr.cr_uid == nickuid && + (!nd->nd_nam2 || + netaddr_match(NU_NETFAM(nuidp), + &nuidp->nu_haddr, nd->nd_nam2))) + break; + } + if (!nuidp) { + nd->nd_repstat = + (NFSERR_AUTHERR|AUTH_REJECTCRED); + nd->nd_procnum = NFSPROC_NOOP; + return (0); + } - /* - * Do we have any use for the verifier. - * According to the "Remote Procedure Call Protocol Spec." it - * should be AUTH_NULL, but some clients make it AUTH_UNIX? - * For now, just skip over it - */ - len = fxdr_unsigned(int, *++tl); - if (len < 0 || len > RPCAUTH_MAXSIZ) { - m_freem(mrep); - return (EBADRPC); - } - if (len > 0) { - nfsm_adv(nfsm_rndup(len)); + /* + * Now, decrypt the timestamp using the session key + * and validate it. + */ +#ifdef NFSKERB + XXX +#endif + + tvout.tv_sec = fxdr_unsigned(long, tvout.tv_sec); + tvout.tv_usec = fxdr_unsigned(long, tvout.tv_usec); + if (nuidp->nu_expire < time.tv_sec || + nuidp->nu_timestamp.tv_sec > tvout.tv_sec || + (nuidp->nu_timestamp.tv_sec == tvout.tv_sec && + nuidp->nu_timestamp.tv_usec > tvout.tv_usec)) { + nuidp->nu_expire = 0; + nd->nd_repstat = + (NFSERR_AUTHERR|AUTH_REJECTVERF); + nd->nd_procnum = NFSPROC_NOOP; + return (0); + } + nfsrv_setcred(&nuidp->nu_cr, &nd->nd_cr); + nd->nd_flag |= ND_KERBNICK; + }; + } else { + nd->nd_repstat = (NFSERR_AUTHERR | AUTH_REJECTCRED); + nd->nd_procnum = NFSPROC_NOOP; + return (0); } /* @@ -1943,16 +2123,14 @@ nfs_getreq(nd, has_header) */ if (nqnfs && nd->nd_procnum != NQNFSPROC_EVICTED) { nfsm_dissect(tl, u_long *, NFSX_UNSIGNED); - nd->nd_nqlflag = fxdr_unsigned(int, *tl); - if (nd->nd_nqlflag) { + nd->nd_flag |= fxdr_unsigned(int, *tl); + if (nd->nd_flag & ND_LEASE) { nfsm_dissect(tl, u_long *, NFSX_UNSIGNED); nd->nd_duration = fxdr_unsigned(int, *tl); } else nd->nd_duration = NQ_MINLEASE; - } else { - nd->nd_nqlflag = NQL_NOVAL; + } else nd->nd_duration = NQ_MINLEASE; - } nd->nd_md = md; nd->nd_dpos = dpos; return (0); @@ -1973,13 +2151,13 @@ nfsrv_wakenfsd(slp) if ((slp->ns_flag & SLP_VALID) == 0) return; - for (nd = nfsd_head.tqh_first; nd != 0; nd = nd->nd_chain.tqe_next) { - if (nd->nd_flag & NFSD_WAITING) { - nd->nd_flag &= ~NFSD_WAITING; - if (nd->nd_slp) + for (nd = nfsd_head.tqh_first; nd != 0; nd = nd->nfsd_chain.tqe_next) { + if (nd->nfsd_flag & NFSD_WAITING) { + nd->nfsd_flag &= ~NFSD_WAITING; + if (nd->nfsd_slp) panic("nfsd wakeup"); slp->ns_sref++; - nd->nd_slp = slp; + nd->nfsd_slp = slp; wakeup((caddr_t)nd); return; } diff --git a/sys/nfs/nfs_srvcache.c b/sys/nfs/nfs_srvcache.c index 0f31ae0..acdce0d 100644 --- a/sys/nfs/nfs_srvcache.c +++ b/sys/nfs/nfs_srvcache.c @@ -34,7 +34,7 @@ * SUCH DAMAGE. * * @(#)nfs_srvcache.c 8.1 (Berkeley) 6/10/93 - * $Id: nfs_srvcache.c,v 1.4 1994/10/02 17:27:00 phk Exp $ + * $Id: nfs_srvcache.c,v 1.5 1994/10/17 17:47:36 phk Exp $ */ /* @@ -59,11 +59,13 @@ #endif #include <nfs/nfsm_subs.h> #include <nfs/rpcv2.h> -#include <nfs/nfsv2.h> +#include <nfs/nfsproto.h> #include <nfs/nfs.h> #include <nfs/nfsrvcache.h> #include <nfs/nqnfs.h> +extern struct nfsstats nfsstats; +extern int nfsv2_procid[NFS_NPROCS]; long numnfsrvcache, desirednfsrvcache = NFSRVCACHESIZ; #define NFSRCHASH(xid) \ @@ -89,7 +91,6 @@ int nonidempotent[NFS_NPROCS] = { FALSE, FALSE, FALSE, - FALSE, TRUE, TRUE, TRUE, @@ -98,6 +99,10 @@ int nonidempotent[NFS_NPROCS] = { TRUE, TRUE, TRUE, + TRUE, + FALSE, + FALSE, + FALSE, FALSE, FALSE, FALSE, @@ -108,7 +113,7 @@ int nonidempotent[NFS_NPROCS] = { }; /* True iff the rpc reply is an nfs status ONLY! */ -static int repliesstatus[NFS_NPROCS] = { +static int nfsv2_repstat[NFS_NPROCS] = { FALSE, FALSE, FALSE, @@ -127,11 +132,6 @@ static int repliesstatus[NFS_NPROCS] = { TRUE, FALSE, FALSE, - FALSE, - FALSE, - FALSE, - FALSE, - TRUE, }; /* @@ -160,9 +160,9 @@ nfsrv_initcache() * Update/add new request at end of lru list */ int -nfsrv_getcache(nam, nd, repp) - struct mbuf *nam; - register struct nfsd *nd; +nfsrv_getcache(nd, slp, repp) + register struct nfsrv_descript *nd; + struct nfssvc_sock *slp; struct mbuf **repp; { register struct nfsrvcache *rp; @@ -171,13 +171,17 @@ nfsrv_getcache(nam, nd, repp) caddr_t bpos; int ret; - if (nd->nd_nqlflag != NQL_NOVAL) + /* + * Don't cache recent requests for reliable transport protocols. + * (Maybe we should for the case of a reconnect, but..) + */ + if (!nd->nd_nam2) return (RC_DOIT); loop: for (rp = NFSRCHASH(nd->nd_retxid)->lh_first; rp != 0; rp = rp->rc_hash.le_next) { if (nd->nd_retxid == rp->rc_xid && nd->nd_procnum == rp->rc_proc && - netaddr_match(NETFAMILY(rp), &rp->rc_haddr, nam)) { + netaddr_match(NETFAMILY(rp), &rp->rc_haddr, nd->nd_nam)) { if ((rp->rc_flag & RC_LOCKED) != 0) { rp->rc_flag |= RC_WANTED; (void) tsleep((caddr_t)rp, PZERO-1, "nfsrc", 0); @@ -196,7 +200,7 @@ loop: ret = RC_DROPIT; } else if (rp->rc_flag & RC_REPSTATUS) { nfsstats.srvcache_nonidemdonehits++; - nfs_rephead(0, nd, rp->rc_status, + nfs_rephead(0, nd, slp, rp->rc_status, 0, (u_quad_t *)0, repp, &mb, &bpos); ret = RC_REPLY; } else if (rp->rc_flag & RC_REPMBUF) { @@ -243,7 +247,7 @@ loop: TAILQ_INSERT_TAIL(&nfsrvlruhead, rp, rc_lru); rp->rc_state = RC_INPROG; rp->rc_xid = nd->nd_retxid; - saddr = mtod(nam, struct sockaddr_in *); + saddr = mtod(nd->nd_nam, struct sockaddr_in *); switch (saddr->sin_family) { case AF_INET: rp->rc_flag |= RC_INETADDR; @@ -252,7 +256,7 @@ loop: case AF_ISO: default: rp->rc_flag |= RC_NAM; - rp->rc_nam = m_copym(nam, 0, M_COPYALL, M_WAIT); + rp->rc_nam = m_copym(nd->nd_nam, 0, M_COPYALL, M_WAIT); break; }; rp->rc_proc = nd->nd_procnum; @@ -269,21 +273,20 @@ loop: * Update a request cache entry after the rpc has been done */ void -nfsrv_updatecache(nam, nd, repvalid, repmbuf) - struct mbuf *nam; - register struct nfsd *nd; +nfsrv_updatecache(nd, repvalid, repmbuf) + register struct nfsrv_descript *nd; int repvalid; struct mbuf *repmbuf; { register struct nfsrvcache *rp; - if (nd->nd_nqlflag != NQL_NOVAL) + if (!nd->nd_nam2) return; loop: for (rp = NFSRCHASH(nd->nd_retxid)->lh_first; rp != 0; rp = rp->rc_hash.le_next) { if (nd->nd_retxid == rp->rc_xid && nd->nd_procnum == rp->rc_proc && - netaddr_match(NETFAMILY(rp), &rp->rc_haddr, nam)) { + netaddr_match(NETFAMILY(rp), &rp->rc_haddr, nd->nd_nam)) { if ((rp->rc_flag & RC_LOCKED) != 0) { rp->rc_flag |= RC_WANTED; (void) tsleep((caddr_t)rp, PZERO-1, "nfsrc", 0); @@ -296,7 +299,8 @@ loop: * the reply for non-idempotent rpc's. */ if (repvalid && nonidempotent[nd->nd_procnum]) { - if (repliesstatus[nd->nd_procnum]) { + if ((nd->nd_flag & ND_NFSV3) == 0 && + nfsv2_repstat[nfsv2_procid[nd->nd_procnum]]) { rp->rc_status = nd->nd_repstat; rp->rc_flag |= RC_REPSTATUS; } else { diff --git a/sys/nfs/nfs_subs.c b/sys/nfs/nfs_subs.c index 1653c8b..350ba46 100644 --- a/sys/nfs/nfs_subs.c +++ b/sys/nfs/nfs_subs.c @@ -34,7 +34,7 @@ * SUCH DAMAGE. * * @(#)nfs_subs.c 8.3 (Berkeley) 1/4/94 - * $Id: nfs_subs.c,v 1.15 1995/05/30 08:12:43 rgrimes Exp $ + * $Id: nfs_subs.c,v 1.16 1995/06/14 06:23:38 joerg Exp $ */ /* @@ -52,6 +52,7 @@ #include <sys/mbuf.h> #include <sys/socket.h> #include <sys/stat.h> +#include <sys/malloc.h> #ifdef VFS_LKM #include <sys/sysent.h> #include <sys/syscall.h> @@ -60,7 +61,7 @@ #include <vm/vm.h> #include <nfs/rpcv2.h> -#include <nfs/nfsv2.h> +#include <nfs/nfsproto.h> #include <nfs/nfsnode.h> #include <nfs/nfs.h> #include <nfs/xdr_subs.h> @@ -78,31 +79,455 @@ #include <netiso/iso.h> #endif -#define TRUE 1 -#define FALSE 0 - /* * Data items converted to xdr at startup, since they are constant * This is kinda hokey, but may save a little time doing byte swaps */ -u_long nfs_procids[NFS_NPROCS]; u_long nfs_xdrneg1; u_long rpc_call, rpc_vers, rpc_reply, rpc_msgdenied, rpc_autherr, - rpc_mismatch, rpc_auth_unix, rpc_msgaccepted, rpc_rejectedcred, + rpc_mismatch, rpc_auth_unix, rpc_msgaccepted, rpc_auth_kerb; -u_long nfs_vers, nfs_prog, nfs_true, nfs_false; +u_long nfs_prog, nqnfs_prog, nfs_true, nfs_false; /* And other global data */ static u_long nfs_xid = 0; -enum vtype ntov_type[7] = { VNON, VREG, VDIR, VBLK, VCHR, VLNK, VNON }; +enum vtype nv2tov_type[8] = { VNON, VREG, VDIR, VBLK, VCHR, VLNK, VNON, VNON }; +enum vtype nv3tov_type[8]={ VNON, VREG, VDIR, VBLK, VCHR, VLNK, VSOCK, VFIFO }; +int nfs_ticks; + +/* + * Mapping of old NFS Version 2 RPC numbers to generic numbers. + */ +int nfsv3_procid[NFS_NPROCS] = { + NFSPROC_NULL, + NFSPROC_GETATTR, + NFSPROC_SETATTR, + NFSPROC_NOOP, + NFSPROC_LOOKUP, + NFSPROC_READLINK, + NFSPROC_READ, + NFSPROC_NOOP, + NFSPROC_WRITE, + NFSPROC_CREATE, + NFSPROC_REMOVE, + NFSPROC_RENAME, + NFSPROC_LINK, + NFSPROC_SYMLINK, + NFSPROC_MKDIR, + NFSPROC_RMDIR, + NFSPROC_READDIR, + NFSPROC_FSSTAT, + NFSPROC_NOOP, + NFSPROC_NOOP, + NFSPROC_NOOP, + NFSPROC_NOOP, + NFSPROC_NOOP, + NFSPROC_NOOP, + NFSPROC_NOOP, + NFSPROC_NOOP +}; + +/* + * and the reverse mapping from generic to Version 2 procedure numbers + */ +int nfsv2_procid[NFS_NPROCS] = { + NFSV2PROC_NULL, + NFSV2PROC_GETATTR, + NFSV2PROC_SETATTR, + NFSV2PROC_LOOKUP, + NFSV2PROC_NOOP, + NFSV2PROC_READLINK, + NFSV2PROC_READ, + NFSV2PROC_WRITE, + NFSV2PROC_CREATE, + NFSV2PROC_MKDIR, + NFSV2PROC_SYMLINK, + NFSV2PROC_CREATE, + NFSV2PROC_REMOVE, + NFSV2PROC_RMDIR, + NFSV2PROC_RENAME, + NFSV2PROC_LINK, + NFSV2PROC_READDIR, + NFSV2PROC_NOOP, + NFSV2PROC_STATFS, + NFSV2PROC_NOOP, + NFSV2PROC_NOOP, + NFSV2PROC_NOOP, + NFSV2PROC_NOOP, + NFSV2PROC_NOOP, + NFSV2PROC_NOOP, + NFSV2PROC_NOOP, +}; + +/* + * Maps errno values to nfs error numbers. + * Use NFSERR_IO as the catch all for ones not specifically defined in + * RFC 1094. + */ +static u_char nfsrv_v2errmap[ELAST] = { + NFSERR_PERM, NFSERR_NOENT, NFSERR_IO, NFSERR_IO, NFSERR_IO, + NFSERR_NXIO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, + NFSERR_IO, NFSERR_IO, NFSERR_ACCES, NFSERR_IO, NFSERR_IO, + NFSERR_IO, NFSERR_EXIST, NFSERR_IO, NFSERR_NODEV, NFSERR_NOTDIR, + NFSERR_ISDIR, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, + NFSERR_IO, NFSERR_FBIG, NFSERR_NOSPC, NFSERR_IO, NFSERR_ROFS, + NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, + NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, + NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, + NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, + NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, + NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, + NFSERR_IO, NFSERR_IO, NFSERR_NAMETOL, NFSERR_IO, NFSERR_IO, + NFSERR_NOTEMPTY, NFSERR_IO, NFSERR_IO, NFSERR_DQUOT, NFSERR_STALE, + NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, + NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, + NFSERR_IO, +}; + +/* + * Maps errno values to nfs error numbers. + * Although it is not obvious whether or not NFS clients really care if + * a returned error value is in the specified list for the procedure, the + * safest thing to do is filter them appropriately. For Version 2, the + * X/Open XNFS document is the only specification that defines error values + * for each RPC (The RFC simply lists all possible error values for all RPCs), + * so I have decided to not do this for Version 2. + * The first entry is the default error return and the rest are the valid + * errors for that RPC in increasing numeric order. + */ +static short nfsv3err_null[] = { + 0, + 0, +}; + +static short nfsv3err_getattr[] = { + NFSERR_IO, + NFSERR_IO, + NFSERR_STALE, + NFSERR_BADHANDLE, + NFSERR_SERVERFAULT, + 0, +}; + +static short nfsv3err_setattr[] = { + NFSERR_IO, + NFSERR_PERM, + NFSERR_IO, + NFSERR_ACCES, + NFSERR_INVAL, + NFSERR_NOSPC, + NFSERR_ROFS, + NFSERR_DQUOT, + NFSERR_STALE, + NFSERR_BADHANDLE, + NFSERR_NOT_SYNC, + NFSERR_SERVERFAULT, + 0, +}; + +static short nfsv3err_lookup[] = { + NFSERR_IO, + NFSERR_NOENT, + NFSERR_IO, + NFSERR_ACCES, + NFSERR_NOTDIR, + NFSERR_NAMETOL, + NFSERR_STALE, + NFSERR_BADHANDLE, + NFSERR_SERVERFAULT, + 0, +}; + +static short nfsv3err_access[] = { + NFSERR_IO, + NFSERR_IO, + NFSERR_STALE, + NFSERR_BADHANDLE, + NFSERR_SERVERFAULT, + 0, +}; + +static short nfsv3err_readlink[] = { + NFSERR_IO, + NFSERR_IO, + NFSERR_ACCES, + NFSERR_INVAL, + NFSERR_STALE, + NFSERR_BADHANDLE, + NFSERR_NOTSUPP, + NFSERR_SERVERFAULT, + 0, +}; + +static short nfsv3err_read[] = { + NFSERR_IO, + NFSERR_IO, + NFSERR_NXIO, + NFSERR_ACCES, + NFSERR_INVAL, + NFSERR_STALE, + NFSERR_BADHANDLE, + NFSERR_SERVERFAULT, + 0, +}; + +static short nfsv3err_write[] = { + NFSERR_IO, + NFSERR_IO, + NFSERR_ACCES, + NFSERR_INVAL, + NFSERR_FBIG, + NFSERR_NOSPC, + NFSERR_ROFS, + NFSERR_DQUOT, + NFSERR_STALE, + NFSERR_BADHANDLE, + NFSERR_SERVERFAULT, + 0, +}; + +static short nfsv3err_create[] = { + NFSERR_IO, + NFSERR_IO, + NFSERR_ACCES, + NFSERR_EXIST, + NFSERR_NOTDIR, + NFSERR_NOSPC, + NFSERR_ROFS, + NFSERR_NAMETOL, + NFSERR_DQUOT, + NFSERR_STALE, + NFSERR_BADHANDLE, + NFSERR_NOTSUPP, + NFSERR_SERVERFAULT, + 0, +}; + +static short nfsv3err_mkdir[] = { + NFSERR_IO, + NFSERR_IO, + NFSERR_ACCES, + NFSERR_EXIST, + NFSERR_NOTDIR, + NFSERR_NOSPC, + NFSERR_ROFS, + NFSERR_NAMETOL, + NFSERR_DQUOT, + NFSERR_STALE, + NFSERR_BADHANDLE, + NFSERR_NOTSUPP, + NFSERR_SERVERFAULT, + 0, +}; + +static short nfsv3err_symlink[] = { + NFSERR_IO, + NFSERR_IO, + NFSERR_ACCES, + NFSERR_EXIST, + NFSERR_NOTDIR, + NFSERR_NOSPC, + NFSERR_ROFS, + NFSERR_NAMETOL, + NFSERR_DQUOT, + NFSERR_STALE, + NFSERR_BADHANDLE, + NFSERR_NOTSUPP, + NFSERR_SERVERFAULT, + 0, +}; + +static short nfsv3err_mknod[] = { + NFSERR_IO, + NFSERR_IO, + NFSERR_ACCES, + NFSERR_EXIST, + NFSERR_NOTDIR, + NFSERR_NOSPC, + NFSERR_ROFS, + NFSERR_NAMETOL, + NFSERR_DQUOT, + NFSERR_STALE, + NFSERR_BADHANDLE, + NFSERR_NOTSUPP, + NFSERR_SERVERFAULT, + NFSERR_BADTYPE, + 0, +}; + +static short nfsv3err_remove[] = { + NFSERR_IO, + NFSERR_NOENT, + NFSERR_IO, + NFSERR_ACCES, + NFSERR_NOTDIR, + NFSERR_ROFS, + NFSERR_NAMETOL, + NFSERR_STALE, + NFSERR_BADHANDLE, + NFSERR_SERVERFAULT, + 0, +}; + +static short nfsv3err_rmdir[] = { + NFSERR_IO, + NFSERR_NOENT, + NFSERR_IO, + NFSERR_ACCES, + NFSERR_EXIST, + NFSERR_NOTDIR, + NFSERR_INVAL, + NFSERR_ROFS, + NFSERR_NAMETOL, + NFSERR_NOTEMPTY, + NFSERR_STALE, + NFSERR_BADHANDLE, + NFSERR_NOTSUPP, + NFSERR_SERVERFAULT, + 0, +}; + +static short nfsv3err_rename[] = { + NFSERR_IO, + NFSERR_NOENT, + NFSERR_IO, + NFSERR_ACCES, + NFSERR_EXIST, + NFSERR_XDEV, + NFSERR_NOTDIR, + NFSERR_ISDIR, + NFSERR_INVAL, + NFSERR_NOSPC, + NFSERR_ROFS, + NFSERR_MLINK, + NFSERR_NAMETOL, + NFSERR_NOTEMPTY, + NFSERR_DQUOT, + NFSERR_STALE, + NFSERR_BADHANDLE, + NFSERR_NOTSUPP, + NFSERR_SERVERFAULT, + 0, +}; + +static short nfsv3err_link[] = { + NFSERR_IO, + NFSERR_IO, + NFSERR_ACCES, + NFSERR_EXIST, + NFSERR_XDEV, + NFSERR_NOTDIR, + NFSERR_INVAL, + NFSERR_NOSPC, + NFSERR_ROFS, + NFSERR_MLINK, + NFSERR_NAMETOL, + NFSERR_DQUOT, + NFSERR_STALE, + NFSERR_BADHANDLE, + NFSERR_NOTSUPP, + NFSERR_SERVERFAULT, + 0, +}; + +static short nfsv3err_readdir[] = { + NFSERR_IO, + NFSERR_IO, + NFSERR_ACCES, + NFSERR_NOTDIR, + NFSERR_STALE, + NFSERR_BADHANDLE, + NFSERR_BAD_COOKIE, + NFSERR_TOOSMALL, + NFSERR_SERVERFAULT, + 0, +}; + +static short nfsv3err_readdirplus[] = { + NFSERR_IO, + NFSERR_IO, + NFSERR_ACCES, + NFSERR_NOTDIR, + NFSERR_STALE, + NFSERR_BADHANDLE, + NFSERR_BAD_COOKIE, + NFSERR_NOTSUPP, + NFSERR_TOOSMALL, + NFSERR_SERVERFAULT, + 0, +}; + +static short nfsv3err_fsstat[] = { + NFSERR_IO, + NFSERR_IO, + NFSERR_STALE, + NFSERR_BADHANDLE, + NFSERR_SERVERFAULT, + 0, +}; + +static short nfsv3err_fsinfo[] = { + NFSERR_STALE, + NFSERR_STALE, + NFSERR_BADHANDLE, + NFSERR_SERVERFAULT, + 0, +}; + +static short nfsv3err_pathconf[] = { + NFSERR_STALE, + NFSERR_STALE, + NFSERR_BADHANDLE, + NFSERR_SERVERFAULT, + 0, +}; + +static short nfsv3err_commit[] = { + NFSERR_IO, + NFSERR_IO, + NFSERR_STALE, + NFSERR_BADHANDLE, + NFSERR_SERVERFAULT, + 0, +}; + +static short *nfsrv_v3errmap[] = { + nfsv3err_null, + nfsv3err_getattr, + nfsv3err_setattr, + nfsv3err_lookup, + nfsv3err_access, + nfsv3err_readlink, + nfsv3err_read, + nfsv3err_write, + nfsv3err_create, + nfsv3err_mkdir, + nfsv3err_symlink, + nfsv3err_mknod, + nfsv3err_remove, + nfsv3err_rmdir, + nfsv3err_rename, + nfsv3err_link, + nfsv3err_readdir, + nfsv3err_readdirplus, + nfsv3err_fsstat, + nfsv3err_fsinfo, + nfsv3err_pathconf, + nfsv3err_commit, +}; + extern struct proc *nfs_iodwant[NFS_MAXASYNCDAEMON]; -extern int nqnfs_piggy[NFS_NPROCS]; extern struct nfsrtt nfsrtt; extern time_t nqnfsstarttime; -extern u_long nqnfs_prog, nqnfs_vers; extern int nqsrv_clockskew; extern int nqsrv_writeslack; extern int nqsrv_maxlease; +extern struct nfsstats nfsstats; +extern int nqnfs_piggy[NFS_NPROCS]; +extern nfstype nfsv2_type[9]; +extern nfstype nfsv3_type[9]; +extern struct nfsnodehashhead *nfsnodehashtbl; +extern u_long nfsnodehash; #ifdef VFS_LKM struct getfh_args; @@ -167,14 +592,16 @@ nfsm_reqh(vp, procid, hsiz, bposp) * Returns the head of the mbuf list. */ struct mbuf * -nfsm_rpchead(cr, nqnfs, procid, auth_type, auth_len, auth_str, mrest, - mrest_len, mbp, xidp) +nfsm_rpchead(cr, nmflag, procid, auth_type, auth_len, auth_str, verf_len, + verf_str, mrest, mrest_len, mbp, xidp) register struct ucred *cr; - int nqnfs; + int nmflag; int procid; int auth_type; int auth_len; char *auth_str; + int verf_len; + char *verf_str; struct mbuf *mrest; int mrest_len; struct mbuf **mbp; @@ -188,15 +615,13 @@ nfsm_rpchead(cr, nqnfs, procid, auth_type, auth_len, auth_str, mrest, int siz, grpsiz, authsiz; authsiz = nfsm_rndup(auth_len); - if (auth_type == RPCAUTH_NQNFS) - authsiz += 2 * NFSX_UNSIGNED; MGETHDR(mb, M_WAIT, MT_DATA); - if ((authsiz + 10*NFSX_UNSIGNED) >= MINCLSIZE) { + if ((authsiz + 10 * NFSX_UNSIGNED) >= MINCLSIZE) { MCLGET(mb, M_WAIT); - } else if ((authsiz + 10*NFSX_UNSIGNED) < MHLEN) { - MH_ALIGN(mb, authsiz + 10*NFSX_UNSIGNED); + } else if ((authsiz + 10 * NFSX_UNSIGNED) < MHLEN) { + MH_ALIGN(mb, authsiz + 10 * NFSX_UNSIGNED); } else { - MH_ALIGN(mb, 8*NFSX_UNSIGNED); + MH_ALIGN(mb, 8 * NFSX_UNSIGNED); } mb->m_len = 0; mreq = mb; @@ -205,20 +630,26 @@ nfsm_rpchead(cr, nqnfs, procid, auth_type, auth_len, auth_str, mrest, /* * First the RPC header. */ - nfsm_build(tl, u_long *, 8*NFSX_UNSIGNED); + nfsm_build(tl, u_long *, 8 * NFSX_UNSIGNED); if (++nfs_xid == 0) nfs_xid++; *tl++ = *xidp = txdr_unsigned(nfs_xid); *tl++ = rpc_call; *tl++ = rpc_vers; - if (nqnfs) { + if (nmflag & NFSMNT_NQNFS) { *tl++ = txdr_unsigned(NQNFS_PROG); - *tl++ = txdr_unsigned(NQNFS_VER1); + *tl++ = txdr_unsigned(NQNFS_VER3); } else { *tl++ = txdr_unsigned(NFS_PROG); - *tl++ = txdr_unsigned(NFS_VER2); + if (nmflag & NFSMNT_NFSV3) + *tl++ = txdr_unsigned(NFS_VER3); + else + *tl++ = txdr_unsigned(NFS_VER2); } - *tl++ = txdr_unsigned(procid); + if (nmflag & NFSMNT_NFSV3) + *tl++ = txdr_unsigned(procid); + else + *tl++ = txdr_unsigned(nfsv2_procid[procid]); /* * And then the authorization cred. @@ -237,10 +668,7 @@ nfsm_rpchead(cr, nqnfs, procid, auth_type, auth_len, auth_str, mrest, for (i = 1; i <= grpsiz; i++) *tl++ = txdr_unsigned(cr->cr_groups[i]); break; - case RPCAUTH_NQNFS: - nfsm_build(tl, u_long *, 2*NFSX_UNSIGNED); - *tl++ = txdr_unsigned(cr->cr_uid); - *tl = txdr_unsigned(auth_len); + case RPCAUTH_KERB4: siz = auth_len; while (siz > 0) { if (M_TRAILINGSPACE(mb) == 0) { @@ -266,11 +694,43 @@ nfsm_rpchead(cr, nqnfs, procid, auth_type, auth_len, auth_str, mrest, } break; }; - nfsm_build(tl, u_long *, 2*NFSX_UNSIGNED); - *tl++ = txdr_unsigned(RPCAUTH_NULL); - *tl = 0; + + /* + * And the verifier... + */ + nfsm_build(tl, u_long *, 2 * NFSX_UNSIGNED); + if (verf_str) { + *tl++ = txdr_unsigned(RPCAUTH_KERB4); + *tl = txdr_unsigned(verf_len); + siz = verf_len; + while (siz > 0) { + if (M_TRAILINGSPACE(mb) == 0) { + MGET(mb2, M_WAIT, MT_DATA); + if (siz >= MINCLSIZE) + MCLGET(mb2, M_WAIT); + mb->m_next = mb2; + mb = mb2; + mb->m_len = 0; + bpos = mtod(mb, caddr_t); + } + i = min(siz, M_TRAILINGSPACE(mb)); + bcopy(verf_str, bpos, i); + mb->m_len += i; + verf_str += i; + bpos += i; + siz -= i; + } + if ((siz = (nfsm_rndup(verf_len) - verf_len)) > 0) { + for (i = 0; i < siz; i++) + *bpos++ = '\0'; + mb->m_len += siz; + } + } else { + *tl++ = txdr_unsigned(RPCAUTH_NULL); + *tl = 0; + } mb->m_next = mrest; - mreq->m_pkthdr.len = authsiz + 10*NFSX_UNSIGNED + mrest_len; + mreq->m_pkthdr.len = authsiz + 10 * NFSX_UNSIGNED + mrest_len; mreq->m_pkthdr.rcvif = (struct ifnet *)0; *mbp = mb; return (mreq); @@ -603,6 +1063,25 @@ nfs_init() { register int i; + /* + * Check to see if major data structures haven't bloated. + */ + if (sizeof (struct nfsnode) > NFS_NODEALLOC) { + printf("struct nfsnode bloated (> %dbytes)\n", NFS_NODEALLOC); + printf("Try reducing NFS_SMALLFH\n"); + } + if (sizeof (struct nfsmount) > NFS_MNTALLOC) { + printf("struct nfsmount bloated (> %dbytes)\n", NFS_MNTALLOC); + printf("Try reducing NFS_MUIDHASHSIZ\n"); + } + if (sizeof (struct nfssvc_sock) > NFS_SVCALLOC) { + printf("struct nfssvc_sock bloated (> %dbytes)\n",NFS_SVCALLOC); + printf("Try reducing NFS_UIDHASHSIZ\n"); + } + if (sizeof (struct nfsuid) > NFS_UIDALLOC) { + printf("struct nfsuid bloated (> %dbytes)\n",NFS_UIDALLOC); + printf("Try unionizing the nu_nickname and nu_flag fields\n"); + } nfsrtt.pos = 0; rpc_vers = txdr_unsigned(RPC_VER2); rpc_call = txdr_unsigned(RPC_CALL); @@ -611,17 +1090,16 @@ nfs_init() rpc_msgaccepted = txdr_unsigned(RPC_MSGACCEPTED); rpc_mismatch = txdr_unsigned(RPC_MISMATCH); rpc_autherr = txdr_unsigned(RPC_AUTHERR); - rpc_rejectedcred = txdr_unsigned(AUTH_REJECTCRED); rpc_auth_unix = txdr_unsigned(RPCAUTH_UNIX); - rpc_auth_kerb = txdr_unsigned(RPCAUTH_NQNFS); - nfs_vers = txdr_unsigned(NFS_VER2); + rpc_auth_kerb = txdr_unsigned(RPCAUTH_KERB4); nfs_prog = txdr_unsigned(NFS_PROG); + nqnfs_prog = txdr_unsigned(NQNFS_PROG); nfs_true = txdr_unsigned(TRUE); nfs_false = txdr_unsigned(FALSE); nfs_xdrneg1 = txdr_unsigned(-1); - /* Loop thru nfs procids */ - for (i = 0; i < NFS_NPROCS; i++) - nfs_procids[i] = txdr_unsigned(i); + nfs_ticks = (hz * NFS_TICKINTVL + 500) / 1000; + if (nfs_ticks < 1) + nfs_ticks = 1; /* Ensure async daemons disabled */ for (i = 0; i < NFS_MAXASYNCDAEMON; i++) nfs_iodwant[i] = (struct proc *)0; @@ -637,8 +1115,6 @@ nfs_init() nqnfsstarttime = boottime.tv_sec + nqsrv_maxlease + nqsrv_clockskew + nqsrv_writeslack; NQLOADNOVRAM(nqnfsstarttime); - nqnfs_prog = txdr_unsigned(NQNFS_PROG); - nqnfs_vers = txdr_unsigned(NQNFS_VER1); CIRCLEQ_INIT(&nqtimerhead); nqfhhashtbl = hashinit(NQLCHSZ, M_NQLEASE, &nqfhhash); } @@ -649,6 +1125,7 @@ nfs_init() TAILQ_INIT(&nfs_reqq); nfs_timer(0); +#ifdef __FreeBSD__ /* * Set up lease_check and lease_updatetime so that other parts * of the system can call us, if we are loadable. @@ -662,6 +1139,7 @@ nfs_init() sysent[SYS_getfh].sy_narg = 2; sysent[SYS_getfh].sy_call = getfh; #endif +#endif return (0); } @@ -689,57 +1167,65 @@ nfs_loadattrcache(vpp, mdp, dposp, vaper) { register struct vnode *vp = *vpp; register struct vattr *vap; - register struct nfsv2_fattr *fp; + register struct nfs_fattr *fp; register struct nfsnode *np; register struct nfsnodehashhead *nhpp; register long t1; - caddr_t dpos, cp2; - int error = 0, isnq; + caddr_t cp2; + int error = 0, rdev; struct mbuf *md; enum vtype vtyp; u_short vmode; - long rdev; struct timespec mtime; struct vnode *nvp; + quad_t tval; + int v3 = NFS_ISV3(vp); md = *mdp; - dpos = *dposp; - t1 = (mtod(md, caddr_t) + md->m_len) - dpos; - isnq = (VFSTONFS(vp->v_mount)->nm_flag & NFSMNT_NQNFS); - error = nfsm_disct(&md, &dpos, NFSX_FATTR(isnq), t1, &cp2); - if (error) + t1 = (mtod(md, caddr_t) + md->m_len) - *dposp; + if (error = nfsm_disct(mdp, dposp, NFSX_FATTR(v3), t1, &cp2)) return (error); - fp = (struct nfsv2_fattr *)cp2; - vtyp = nfstov_type(fp->fa_type); - vmode = fxdr_unsigned(u_short, fp->fa_mode); - /* - * XXX - * - * The duplicate information returned in fa_type and fa_mode - * is an ambiguity in the NFS version 2 protocol. - * - * VREG should be taken literally as a regular file. If a - * server intents to return some type information differently - * in the upper bits of the mode field (e.g. for sockets, or - * FIFOs), NFSv2 mandates fa_type to be VNON. Anyway, we - * leave the examination of the mode bits even in the VREG - * case to avoid breakage for bogus servers, but we make sure - * that there are actually type bits set in the upper part of - * fa_mode (and failing that, trust the va_type field). - * - * NFSv3 cleared the issue, and requires fa_mode to not - * contain any type information (while also introduing sockets - * and FIFOs for fa_type). - */ - if (vtyp == VNON || (vtyp == VREG && (vmode & S_IFMT) != 0)) - vtyp = IFTOVT(vmode); - if (isnq) { - rdev = fxdr_unsigned(long, fp->fa_nqrdev); - fxdr_nqtime(&fp->fa_nqmtime, &mtime); + fp = (struct nfs_fattr *)cp2; + if (v3) { + vtyp = nfsv3tov_type(fp->fa_type); + vmode = fxdr_unsigned(u_short, fp->fa_mode); + rdev = makedev(fxdr_unsigned(u_char, fp->fa3_rdev.specdata1), + fxdr_unsigned(u_char, fp->fa3_rdev.specdata2)); + fxdr_nfsv3time(&fp->fa3_mtime, &mtime); } else { - rdev = fxdr_unsigned(long, fp->fa_nfsrdev); - fxdr_nfstime(&fp->fa_nfsmtime, &mtime); + vtyp = nfsv2tov_type(fp->fa_type); + vmode = fxdr_unsigned(u_short, fp->fa_mode); + /* + * XXX + * + * The duplicate information returned in fa_type and fa_mode + * is an ambiguity in the NFS version 2 protocol. + * + * VREG should be taken literally as a regular file. If a + * server intents to return some type information differently + * in the upper bits of the mode field (e.g. for sockets, or + * FIFOs), NFSv2 mandates fa_type to be VNON. Anyway, we + * leave the examination of the mode bits even in the VREG + * case to avoid breakage for bogus servers, but we make sure + * that there are actually type bits set in the upper part of + * fa_mode (and failing that, trust the va_type field). + * + * NFSv3 cleared the issue, and requires fa_mode to not + * contain any type information (while also introduing sockets + * and FIFOs for fa_type). + */ + if (vtyp == VNON || (vtyp == VREG && (vmode & S_IFMT) != 0)) + vtyp = IFTOVT(vmode); + rdev = fxdr_unsigned(long, fp->fa2_rdev); + fxdr_nfsv2time(&fp->fa2_mtime, &mtime); + + /* + * Really ugly NFSv2 kludge. + */ + if (vtyp == VCHR && rdev == 0xffffffff) + vtyp = VFIFO; } + /* * If v_type == VNON it is a new node, so fill in the v_type, * n_mtime fields. Check to see if it represents a special @@ -749,10 +1235,15 @@ nfs_loadattrcache(vpp, mdp, dposp, vaper) */ np = VTONFS(vp); if (vp->v_type == VNON) { - if (vtyp == VCHR && rdev == 0xffffffff) - vp->v_type = vtyp = VFIFO; - else - vp->v_type = vtyp; + /* + * If we had a lock and it turns out that the vnode + * is an object which we don't want to lock (e.g. VDIR) + * to avoid nasty hanging problems on a server crash, + * then release it here. + */ + if (vtyp != VREG && VOP_ISLOCKED(vp)) + VOP_UNLOCK(vp); + vp->v_type = vtyp; if (vp->v_type == VFIFO) { vp->v_op = fifo_nfsv2nodeop_p; } @@ -773,7 +1264,7 @@ nfs_loadattrcache(vpp, mdp, dposp, vaper) * Reinitialize aliased node. */ np->n_vnode = nvp; - nhpp = nfs_hash(&np->n_fh); + nhpp = NFSNOHASH(nfs_hash(np->n_fhp, np->n_fhsize)); LIST_INSERT_HEAD(nhpp, np, n_hash); *vpp = vp = nvp; } @@ -783,31 +1274,34 @@ nfs_loadattrcache(vpp, mdp, dposp, vaper) vap = &np->n_vattr; vap->va_type = vtyp; vap->va_mode = (vmode & 07777); - vap->va_nlink = fxdr_unsigned(u_short, fp->fa_nlink); - vap->va_uid = fxdr_unsigned(uid_t, fp->fa_uid); - vap->va_gid = fxdr_unsigned(gid_t, fp->fa_gid); vap->va_rdev = (dev_t)rdev; vap->va_mtime = mtime; vap->va_fsid = vp->v_mount->mnt_stat.f_fsid.val[0]; - if (isnq) { - fxdr_hyper(&fp->fa_nqsize, &vap->va_size); - vap->va_blocksize = fxdr_unsigned(long, fp->fa_nqblocksize); - fxdr_hyper(&fp->fa_nqbytes, &vap->va_bytes); - vap->va_fileid = fxdr_unsigned(long, fp->fa_nqfileid); - fxdr_nqtime(&fp->fa_nqatime, &vap->va_atime); - vap->va_flags = fxdr_unsigned(u_long, fp->fa_nqflags); - fxdr_nqtime(&fp->fa_nqctime, &vap->va_ctime); - vap->va_gen = fxdr_unsigned(u_long, fp->fa_nqgen); - fxdr_hyper(&fp->fa_nqfilerev, &vap->va_filerev); + if (v3) { + vap->va_nlink = fxdr_unsigned(u_short, fp->fa_nlink); + vap->va_uid = fxdr_unsigned(uid_t, fp->fa_uid); + vap->va_gid = fxdr_unsigned(gid_t, fp->fa_gid); + fxdr_hyper(&fp->fa3_size, &vap->va_size); + vap->va_blocksize = NFS_FABLKSIZE; + fxdr_hyper(&fp->fa3_used, &vap->va_bytes); + vap->va_fileid = fxdr_unsigned(int, fp->fa3_fileid.nfsuquad[1]); + fxdr_nfsv3time(&fp->fa3_atime, &vap->va_atime); + fxdr_nfsv3time(&fp->fa3_ctime, &vap->va_ctime); + vap->va_flags = 0; + vap->va_filerev = 0; } else { - vap->va_size = fxdr_unsigned(u_long, fp->fa_nfssize); - vap->va_blocksize = fxdr_unsigned(long, fp->fa_nfsblocksize); - vap->va_bytes = fxdr_unsigned(long, fp->fa_nfsblocks) * NFS_FABLKSIZE; - vap->va_fileid = fxdr_unsigned(long, fp->fa_nfsfileid); - fxdr_nfstime(&fp->fa_nfsatime, &vap->va_atime); + vap->va_nlink = fxdr_unsigned(u_short, fp->fa_nlink); + vap->va_uid = fxdr_unsigned(uid_t, fp->fa_uid); + vap->va_gid = fxdr_unsigned(gid_t, fp->fa_gid); + vap->va_size = fxdr_unsigned(u_long, fp->fa2_size); + vap->va_blocksize = fxdr_unsigned(long, fp->fa2_blocksize); + vap->va_bytes = fxdr_unsigned(long, fp->fa2_blocks) * NFS_FABLKSIZE; + vap->va_fileid = fxdr_unsigned(long, fp->fa2_fileid); + fxdr_nfsv2time(&fp->fa2_atime, &vap->va_atime); vap->va_flags = 0; - fxdr_nfstime(&fp->fa_nfsctime, &vap->va_ctime); - vap->va_gen = 0; + vap->va_ctime.ts_sec = fxdr_unsigned(long, fp->fa2_ctime.nfsv2_sec); + vap->va_ctime.ts_nsec = 0; + vap->va_gen = fxdr_unsigned(u_long, fp->fa2_ctime.nfsv2_usec); vap->va_filerev = 0; } if (vap->va_size != np->n_size) { @@ -824,26 +1318,13 @@ nfs_loadattrcache(vpp, mdp, dposp, vaper) np->n_size = vap->va_size; } np->n_attrstamp = time.tv_sec; - *dposp = dpos; - *mdp = md; if (vaper != NULL) { bcopy((caddr_t)vap, (caddr_t)vaper, sizeof(*vap)); -#ifdef notdef - if ((np->n_flag & NMODIFIED) && np->n_size > vap->va_size) - if (np->n_size > vap->va_size) - vaper->va_size = np->n_size; -#endif if (np->n_flag & NCHG) { - if (np->n_flag & NACC) { - vaper->va_atime.ts_sec = np->n_atim.tv_sec; - vaper->va_atime.ts_nsec = - np->n_atim.tv_usec * 1000; - } - if (np->n_flag & NUPD) { - vaper->va_mtime.ts_sec = np->n_mtim.tv_sec; - vaper->va_mtime.ts_nsec = - np->n_mtim.tv_usec * 1000; - } + if (np->n_flag & NACC) + vaper->va_atime = np->n_atim; + if (np->n_flag & NUPD) + vaper->va_mtime = np->n_mtim; } } return (0); @@ -862,12 +1343,7 @@ nfs_getattrcache(vp, vaper) register struct nfsnode *np = VTONFS(vp); register struct vattr *vap; - if (VFSTONFS(vp->v_mount)->nm_flag & NFSMNT_NQLOOKLEASE) { - if (!NQNFS_CKCACHABLE(vp, NQL_READ) || np->n_attrstamp == 0) { - nfsstats.attrcache_misses++; - return (ENOENT); - } - } else if ((time.tv_sec - np->n_attrstamp) >= NFS_ATTRTIMEO(np)) { + if ((time.tv_sec - np->n_attrstamp) >= NFS_ATTRTIMEO(np)) { nfsstats.attrcache_misses++; return (ENOENT); } @@ -887,23 +1363,11 @@ nfs_getattrcache(vp, vaper) np->n_size = vap->va_size; } bcopy((caddr_t)vap, (caddr_t)vaper, sizeof(struct vattr)); -#ifdef notdef - if ((np->n_flag & NMODIFIED) == 0) { - np->n_size = vaper->va_size; - vnode_pager_setsize(vp, (u_long)np->n_size); - } else if (np->n_size > vaper->va_size) - if (np->n_size > vaper->va_size) - vaper->va_size = np->n_size; -#endif if (np->n_flag & NCHG) { - if (np->n_flag & NACC) { - vaper->va_atime.ts_sec = np->n_atim.tv_sec; - vaper->va_atime.ts_nsec = np->n_atim.tv_usec * 1000; - } - if (np->n_flag & NUPD) { - vaper->va_mtime.ts_sec = np->n_mtim.tv_sec; - vaper->va_mtime.ts_nsec = np->n_mtim.tv_usec * 1000; - } + if (np->n_flag & NACC) + vaper->va_atime = np->n_atim; + if (np->n_flag & NUPD) + vaper->va_mtime = np->n_mtim; } return (0); } @@ -912,7 +1376,7 @@ nfs_getattrcache(vp, vaper) * Set up nameidata for a lookup() call and do it */ int -nfs_namei(ndp, fhp, len, slp, nam, mdp, dposp, p) +nfs_namei(ndp, fhp, len, slp, nam, mdp, dposp, retdirp, p, kerbflag) register struct nameidata *ndp; fhandle_t *fhp; int len; @@ -920,7 +1384,9 @@ nfs_namei(ndp, fhp, len, slp, nam, mdp, dposp, p) struct mbuf *nam; struct mbuf **mdp; caddr_t *dposp; + struct vnode **retdirp; struct proc *p; + int kerbflag; { register int i, rem; register struct mbuf *md; @@ -929,6 +1395,7 @@ nfs_namei(ndp, fhp, len, slp, nam, mdp, dposp, p) int error, rdonly; struct componentname *cnp = &ndp->ni_cnd; + *retdirp = (struct vnode *)0; MALLOC(cnp->cn_pnbuf, char *, len + 1, M_NAMEI, M_WAITOK); /* * Copy the name from the mbuf list to ndp->ni_pnbuf @@ -950,7 +1417,7 @@ nfs_namei(ndp, fhp, len, slp, nam, mdp, dposp, p) rem = md->m_len; } if (*fromcp == '\0' || *fromcp == '/') { - error = EINVAL; + error = EACCES; goto out; } cnp->cn_hash += (unsigned char)*fromcp; @@ -964,26 +1431,24 @@ nfs_namei(ndp, fhp, len, slp, nam, mdp, dposp, p) if (len > 0) { if (rem >= len) *dposp += len; - else { - error = nfs_adv(mdp, dposp, len, rem); - if (error) - goto out; - } + else if (error = nfs_adv(mdp, dposp, len, rem)) + goto out; } ndp->ni_pathlen = tocp - cnp->cn_pnbuf; cnp->cn_nameptr = cnp->cn_pnbuf; /* * Extract and set starting directory. */ - error = nfsrv_fhtovp(fhp, FALSE, &dp, ndp->ni_cnd.cn_cred, slp, - nam, &rdonly); - if (error) + if (error = nfsrv_fhtovp(fhp, FALSE, &dp, ndp->ni_cnd.cn_cred, slp, + nam, &rdonly, kerbflag)) goto out; if (dp->v_type != VDIR) { nfsrv_vrele(dp); error = ENOTDIR; goto out; } + VREF(dp); + *retdirp = dp; ndp->ni_startdir = dp; if (rdonly) cnp->cn_flags |= (NOCROSSMOUNT | RDONLY); @@ -993,8 +1458,7 @@ nfs_namei(ndp, fhp, len, slp, nam, mdp, dposp, p) * And call lookup() to do the real work */ cnp->cn_proc = p; - error = lookup(ndp); - if (error) + if (error = lookup(ndp)) goto out; /* * Check for encountering a symbolic link @@ -1087,6 +1551,109 @@ nfsm_adj(mp, len, nul) } /* + * Make these functions instead of macros, so that the kernel text size + * doesn't get too big... + */ +void +nfsm_srvwcc(nfsd, before_ret, before_vap, after_ret, after_vap, mbp, bposp) + struct nfsrv_descript *nfsd; + int before_ret; + register struct vattr *before_vap; + int after_ret; + struct vattr *after_vap; + struct mbuf **mbp; + char **bposp; +{ + register struct mbuf *mb = *mbp, *mb2; + register char *bpos = *bposp; + register u_long *tl; + + if (before_ret) { + nfsm_build(tl, u_long *, NFSX_UNSIGNED); + *tl = nfs_false; + } else { + nfsm_build(tl, u_long *, 7 * NFSX_UNSIGNED); + *tl++ = nfs_true; + txdr_hyper(&(before_vap->va_size), tl); + tl += 2; + txdr_nfsv3time(&(before_vap->va_mtime), tl); + tl += 2; + txdr_nfsv3time(&(before_vap->va_ctime), tl); + } + *bposp = bpos; + *mbp = mb; + nfsm_srvpostopattr(nfsd, after_ret, after_vap, mbp, bposp); +} + +void +nfsm_srvpostopattr(nfsd, after_ret, after_vap, mbp, bposp) + struct nfsrv_descript *nfsd; + int after_ret; + struct vattr *after_vap; + struct mbuf **mbp; + char **bposp; +{ + register struct mbuf *mb = *mbp, *mb2; + register char *bpos = *bposp; + register u_long *tl; + register struct nfs_fattr *fp; + + if (after_ret) { + nfsm_build(tl, u_long *, NFSX_UNSIGNED); + *tl = nfs_false; + } else { + nfsm_build(tl, u_long *, NFSX_UNSIGNED + NFSX_V3FATTR); + *tl++ = nfs_true; + fp = (struct nfs_fattr *)tl; + nfsm_srvfattr(nfsd, after_vap, fp); + } + *mbp = mb; + *bposp = bpos; +} + +void +nfsm_srvfattr(nfsd, vap, fp) + register struct nfsrv_descript *nfsd; + register struct vattr *vap; + register struct nfs_fattr *fp; +{ + + fp->fa_nlink = txdr_unsigned(vap->va_nlink); + fp->fa_uid = txdr_unsigned(vap->va_uid); + fp->fa_gid = txdr_unsigned(vap->va_gid); + if (nfsd->nd_flag & ND_NFSV3) { + fp->fa_type = vtonfsv3_type(vap->va_type); + fp->fa_mode = vtonfsv3_mode(vap->va_mode); + txdr_hyper(&vap->va_size, &fp->fa3_size); + txdr_hyper(&vap->va_bytes, &fp->fa3_used); + fp->fa3_rdev.specdata1 = txdr_unsigned(major(vap->va_rdev)); + fp->fa3_rdev.specdata2 = txdr_unsigned(minor(vap->va_rdev)); + fp->fa3_fsid.nfsuquad[0] = 0; + fp->fa3_fsid.nfsuquad[1] = txdr_unsigned(vap->va_fsid); + fp->fa3_fileid.nfsuquad[0] = 0; + fp->fa3_fileid.nfsuquad[1] = txdr_unsigned(vap->va_fileid); + txdr_nfsv3time(&vap->va_atime, &fp->fa3_atime); + txdr_nfsv3time(&vap->va_mtime, &fp->fa3_mtime); + txdr_nfsv3time(&vap->va_ctime, &fp->fa3_ctime); + } else { + fp->fa_type = vtonfsv2_type(vap->va_type); + fp->fa_mode = vtonfsv2_mode(vap->va_type, vap->va_mode); + fp->fa2_size = txdr_unsigned(vap->va_size); + fp->fa2_blocksize = txdr_unsigned(vap->va_blocksize); + if (vap->va_type == VFIFO) + fp->fa2_rdev = 0xffffffff; + else + fp->fa2_rdev = txdr_unsigned(vap->va_rdev); + fp->fa2_blocks = txdr_unsigned(vap->va_bytes / NFS_FABLKSIZE); + fp->fa2_fsid = txdr_unsigned(vap->va_fsid); + fp->fa2_fileid = txdr_unsigned(vap->va_fileid); + txdr_nfsv2time(&vap->va_atime, &fp->fa2_atime); + txdr_nfsv2time(&vap->va_mtime, &fp->fa2_mtime); + txdr_nfsv2time(&vap->va_ctime, &fp->fa2_ctime); + } +} + +/* * nfsrv_fhtovp() - convert a fh to a vnode ptr (optionally locked) * - look up fsid in mount list (if not found ret error) * - get vp and export rights by calling VFS_FHTOVP() @@ -1094,7 +1661,7 @@ nfsm_adj(mp, len, nul) * - if not lockflag unlock it with VOP_UNLOCK() */ int -nfsrv_fhtovp(fhp, lockflag, vpp, cred, slp, nam, rdonlyp) +nfsrv_fhtovp(fhp, lockflag, vpp, cred, slp, nam, rdonlyp, kerbflag) fhandle_t *fhp; int lockflag; struct vnode **vpp; @@ -1102,6 +1669,7 @@ nfsrv_fhtovp(fhp, lockflag, vpp, cred, slp, nam, rdonlyp) struct nfssvc_sock *slp; struct mbuf *nam; int *rdonlyp; + int kerbflag; { register struct mount *mp; register struct nfsuid *uidp; @@ -1120,19 +1688,13 @@ nfsrv_fhtovp(fhp, lockflag, vpp, cred, slp, nam, rdonlyp) * Check/setup credentials. */ if (exflags & MNT_EXKERB) { - for (uidp = NUIDHASH(slp, cred->cr_uid)->lh_first; uidp != 0; - uidp = uidp->nu_hash.le_next) { - if (uidp->nu_uid == cred->cr_uid) - break; - } - if (uidp == 0) { + if (!kerbflag) { vput(*vpp); - return (NQNFS_AUTHERR); + return (NFSERR_AUTHERR | AUTH_TOOWEAK); } - cred->cr_uid = uidp->nu_cr.cr_uid; - for (i = 0; i < uidp->nu_cr.cr_ngroups; i++) - cred->cr_groups[i] = uidp->nu_cr.cr_groups[i]; - cred->cr_ngroups = i; + } else if (kerbflag) { + vput(*vpp); + return (NFSERR_AUTHERR | AUTH_TOOWEAK); } else if (cred->cr_uid == 0 || (exflags & MNT_EXPORTANON)) { cred->cr_uid = credanon->cr_uid; for (i = 0; i < credanon->cr_ngroups && i < NGROUPS; i++) @@ -1194,6 +1756,145 @@ netaddr_match(family, haddr, nam) return (0); } +static nfsuint64 nfs_nullcookie = { 0, 0 }; +/* + * This function finds the directory cookie that corresponds to the + * logical byte offset given. + */ +nfsuint64 * +nfs_getcookie(np, off, add) + register struct nfsnode *np; + off_t off; + int add; +{ + register struct nfsdmap *dp, *dp2; + register int pos; + + pos = off / NFS_DIRBLKSIZ; + if (pos == 0) { +#ifdef DIAGNOSTIC + if (add) + panic("nfs getcookie add at 0"); +#endif + return (&nfs_nullcookie); + } + pos--; + dp = np->n_cookies.lh_first; + if (!dp) { + if (add) { + MALLOC(dp, struct nfsdmap *, sizeof (struct nfsdmap), + M_NFSDIROFF, M_WAITOK); + dp->ndm_eocookie = 0; + LIST_INSERT_HEAD(&np->n_cookies, dp, ndm_list); + } else + return ((nfsuint64 *)0); + } + while (pos >= NFSNUMCOOKIES) { + pos -= NFSNUMCOOKIES; + if (dp->ndm_list.le_next) { + if (!add && dp->ndm_eocookie < NFSNUMCOOKIES && + pos >= dp->ndm_eocookie) + return ((nfsuint64 *)0); + dp = dp->ndm_list.le_next; + } else if (add) { + MALLOC(dp2, struct nfsdmap *, sizeof (struct nfsdmap), + M_NFSDIROFF, M_WAITOK); + dp2->ndm_eocookie = 0; + LIST_INSERT_AFTER(dp, dp2, ndm_list); + dp = dp2; + } else + return ((nfsuint64 *)0); + } + if (pos >= dp->ndm_eocookie) { + if (add) + dp->ndm_eocookie = pos + 1; + else + return ((nfsuint64 *)0); + } + return (&dp->ndm_cookies[pos]); +} + +/* + * Invalidate cached directory information, except for the actual directory + * blocks (which are invalidated separately). + * Done mainly to avoid the use of stale offset cookies. + */ +void +nfs_invaldir(vp) + register struct vnode *vp; +{ + register struct nfsnode *np = VTONFS(vp); + +#ifdef DIAGNOSTIC + if (vp->v_type != VDIR) + panic("nfs: invaldir not dir"); +#endif + np->n_direofoffset = 0; + np->n_cookieverf.nfsuquad[0] = 0; + np->n_cookieverf.nfsuquad[1] = 0; + if (np->n_cookies.lh_first) + np->n_cookies.lh_first->ndm_eocookie = 0; +} + +/* + * The write verifier has changed (probably due to a server reboot), so all + * B_NEEDCOMMIT blocks will have to be written again. Since they are on the + * dirty block list as B_DELWRI, all this takes is clearing the B_NEEDCOMMIT + * flag. Once done the new write verifier can be set for the mount point. + */ +void +nfs_clearcommit(mp) + struct mount *mp; +{ + register struct vnode *vp, *nvp; + register struct buf *bp, *nbp; + int s; + + s = splbio(); +loop: + for (vp = mp->mnt_vnodelist.lh_first; vp; vp = nvp) { + if (vp->v_mount != mp) /* Paranoia */ + goto loop; + nvp = vp->v_mntvnodes.le_next; + for (bp = vp->v_dirtyblkhd.lh_first; bp; bp = nbp) { + nbp = bp->b_vnbufs.le_next; + if ((bp->b_flags & (B_BUSY | B_DELWRI | B_NEEDCOMMIT)) + == (B_DELWRI | B_NEEDCOMMIT)) + bp->b_flags &= ~B_NEEDCOMMIT; + } + } + splx(s); +} + +/* + * Map errnos to NFS error numbers. For Version 3 also filter out error + * numbers not specified for the associated procedure. + */ +int +nfsrv_errmap(nd, err) + struct nfsrv_descript *nd; + register int err; +{ + register short *defaulterrp, *errp; + + if (nd->nd_flag & ND_NFSV3) { + if (nd->nd_procnum <= NFSPROC_COMMIT) { + errp = defaulterrp = nfsrv_v3errmap[nd->nd_procnum]; + while (*++errp) { + if (*errp == err) + return (err); + else if (*errp > err) + break; + } + return ((int)*defaulterrp); + } else + return (err & 0xffff); + } + if (err <= ELAST) + return ((int)nfsrv_v2errmap[err - 1]); + return (NFSERR_IO); +} + int nfsrv_vmio( struct vnode *vp) { vm_object_t object; diff --git a/sys/nfs/nfs_syscalls.c b/sys/nfs/nfs_syscalls.c index f7e159f..35a50f4 100644 --- a/sys/nfs/nfs_syscalls.c +++ b/sys/nfs/nfs_syscalls.c @@ -34,7 +34,7 @@ * SUCH DAMAGE. * * @(#)nfs_syscalls.c 8.3 (Berkeley) 1/4/94 - * $Id: nfs_syscalls.c,v 1.5 1994/10/17 17:47:38 phk Exp $ + * $Id: nfs_syscalls.c,v 1.6 1995/05/30 08:12:45 rgrimes Exp $ */ #include <sys/param.h> @@ -61,9 +61,11 @@ #ifdef ISO #include <netiso/iso.h> #endif +#include <nfs/xdr_subs.h> #include <nfs/rpcv2.h> -#include <nfs/nfsv2.h> +#include <nfs/nfsproto.h> #include <nfs/nfs.h> +#include <nfs/nfsm_subs.h> #include <nfs/nfsrvcache.h> #include <nfs/nfsmount.h> #include <nfs/nfsnode.h> @@ -73,13 +75,14 @@ void nfsrv_zapsock __P((struct nfssvc_sock *)); /* Global defs. */ -extern u_long nfs_prog, nfs_vers; -extern int (*nfsrv_procs[NFS_NPROCS])(); +extern int (*nfsrv3_procs[NFS_NPROCS])(); extern struct proc *nfs_iodwant[NFS_MAXASYNCDAEMON]; extern int nfs_numasync; extern time_t nqnfsstarttime; extern int nqsrv_writeslack; extern int nfsrtton; +extern struct nfsstats nfsstats; +extern int nfsrvw_procrastinate; struct nfssvc_sock *nfs_udpsock, *nfs_cltpsock; int nuidhash_max = NFS_MAXUIDHASH; static int nfs_numnfsd = 0; @@ -225,19 +228,25 @@ nfssvc(p, uap, retval) if (error) return (error); if ((uap->flag & NFSSVC_AUTHIN) && ((nfsd = nsd->nsd_nfsd)) && - (nfsd->nd_slp->ns_flag & SLP_VALID)) { - slp = nfsd->nd_slp; + (nfsd->nfsd_slp->ns_flag & SLP_VALID)) { + slp = nfsd->nfsd_slp; /* * First check to see if another nfsd has already * added this credential. */ - for (nuidp = NUIDHASH(slp, nsd->nsd_uid)->lh_first; + for (nuidp = NUIDHASH(slp,nsd->nsd_cr.cr_uid)->lh_first; nuidp != 0; nuidp = nuidp->nu_hash.le_next) { - if (nuidp->nu_uid == nsd->nsd_uid) + if (nuidp->nu_cr.cr_uid == nsd->nsd_cr.cr_uid && + (!nfsd->nfsd_nd->nd_nam2 || + netaddr_match(NU_NETFAM(nuidp), + &nuidp->nu_haddr, nfsd->nfsd_nd->nd_nam2))) break; } - if (!nuidp) { + if (nuidp) { + nfsrv_setcred(&nuidp->nu_cr,&nfsd->nfsd_nd->nd_cr); + nfsd->nfsd_nd->nd_flag |= ND_KERBFULL; + } else { /* * Nope, so we will. */ @@ -257,22 +266,53 @@ nfssvc(p, uap, retval) LIST_REMOVE(nuidp, nu_hash); TAILQ_REMOVE(&slp->ns_uidlruhead, nuidp, nu_lru); + if (nuidp->nu_flag & NU_NAM) + m_freem(nuidp->nu_nam); } + nuidp->nu_flag = 0; nuidp->nu_cr = nsd->nsd_cr; if (nuidp->nu_cr.cr_ngroups > NGROUPS) - nuidp->nu_cr.cr_ngroups = NGROUPS; + nuidp->nu_cr.cr_ngroups = NGROUPS; nuidp->nu_cr.cr_ref = 1; - nuidp->nu_uid = nsd->nsd_uid; + nuidp->nu_timestamp = nsd->nsd_timestamp; + nuidp->nu_expire = time.tv_sec + nsd->nsd_ttl; + /* + * and save the session key in nu_key. + */ + bcopy(nsd->nsd_key, nuidp->nu_key, + sizeof (nsd->nsd_key)); + if (nfsd->nfsd_nd->nd_nam2) { + struct sockaddr_in *saddr; + + saddr = mtod(nfsd->nfsd_nd->nd_nam2, + struct sockaddr_in *); + switch (saddr->sin_family) { + case AF_INET: + nuidp->nu_flag |= NU_INETADDR; + nuidp->nu_inetaddr = + saddr->sin_addr.s_addr; + break; + case AF_ISO: + default: + nuidp->nu_flag |= NU_NAM; + nuidp->nu_nam = m_copym( + nfsd->nfsd_nd->nd_nam2, 0, + M_COPYALL, M_WAIT); + break; + }; + } TAILQ_INSERT_TAIL(&slp->ns_uidlruhead, nuidp, nu_lru); LIST_INSERT_HEAD(NUIDHASH(slp, nsd->nsd_uid), nuidp, nu_hash); - + nfsrv_setcred(&nuidp->nu_cr, + &nfsd->nfsd_nd->nd_cr); + nfsd->nfsd_nd->nd_flag |= ND_KERBFULL; } } } if ((uap->flag & NFSSVC_AUTHINFAIL) && (nfsd = nsd->nsd_nfsd)) - nfsd->nd_flag |= NFSD_AUTHFAIL; + nfsd->nfsd_flag |= NFSD_AUTHFAIL; error = nfssvc_nfsd(nsd, uap->argp, p); } if (error == EINTR || error == ERESTART) @@ -353,8 +393,6 @@ nfssvc_addsock(fp, mynam) slp = (struct nfssvc_sock *) malloc(sizeof (struct nfssvc_sock), M_NFSSVC, M_WAITOK); bzero((caddr_t)slp, sizeof (struct nfssvc_sock)); - slp->ns_uidhashtbl = - hashinit(NUIDHASHSIZ, M_NFSSVC, &slp->ns_uidhash); TAILQ_INIT(&slp->ns_uidlruhead); TAILQ_INSERT_TAIL(&nfssvc_sockhead, slp, ns_chain); } @@ -381,44 +419,47 @@ nfssvc_nfsd(nsd, argp, p) caddr_t argp; struct proc *p; { - register struct mbuf *m, *nam2; + register struct mbuf *m; register int siz; register struct nfssvc_sock *slp; register struct socket *so; register int *solockp; - struct nfsd *nd = nsd->nsd_nfsd; - struct mbuf *mreq, *nam; - struct timeval starttime; + struct nfsd *nfsd = nsd->nsd_nfsd; + struct nfsrv_descript *nd = NULL; + struct mbuf *mreq; struct nfsuid *uidp; - int error = 0, cacherep, s; - int sotype; + int error = 0, cacherep, s, sotype, writes_todo; + u_quad_t cur_usec; +#ifndef nolint + cacherep = RC_DOIT; + writes_todo = 0; +#endif s = splnet(); - if (nd == (struct nfsd *)0) { - nsd->nsd_nfsd = nd = (struct nfsd *) + if (nfsd == (struct nfsd *)0) { + nsd->nsd_nfsd = nfsd = (struct nfsd *) malloc(sizeof (struct nfsd), M_NFSD, M_WAITOK); - bzero((caddr_t)nd, sizeof (struct nfsd)); - nd->nd_procp = p; - nd->nd_cr.cr_ref = 1; - TAILQ_INSERT_TAIL(&nfsd_head, nd, nd_chain); - nd->nd_nqlflag = NQL_NOVAL; + bzero((caddr_t)nfsd, sizeof (struct nfsd)); + nfsd->nfsd_procp = p; + TAILQ_INSERT_TAIL(&nfsd_head, nfsd, nfsd_chain); nfs_numnfsd++; } /* * Loop getting rpc requests until SIGKILL. */ for (;;) { - if ((nd->nd_flag & NFSD_REQINPROG) == 0) { - while (nd->nd_slp == (struct nfssvc_sock *)0 && + if ((nfsd->nfsd_flag & NFSD_REQINPROG) == 0) { + while (nfsd->nfsd_slp == (struct nfssvc_sock *)0 && (nfsd_head_flag & NFSD_CHECKSLP) == 0) { - nd->nd_flag |= NFSD_WAITING; + nfsd->nfsd_flag |= NFSD_WAITING; nfsd_waiting++; - error = tsleep((caddr_t)nd, PSOCK | PCATCH, "nfsd", 0); + error = tsleep((caddr_t)nfsd, PSOCK | PCATCH, + "nfsd", 0); nfsd_waiting--; if (error) goto done; } - if (nd->nd_slp == (struct nfssvc_sock *)0 && + if (nfsd->nfsd_slp == (struct nfssvc_sock *)0 && (nfsd_head_flag & NFSD_CHECKSLP) != 0) { for (slp = nfssvc_sockhead.tqh_first; slp != 0; slp = slp->ns_chain.tqe_next) { @@ -426,14 +467,14 @@ nfssvc_nfsd(nsd, argp, p) == (SLP_VALID | SLP_DOREC)) { slp->ns_flag &= ~SLP_DOREC; slp->ns_sref++; - nd->nd_slp = slp; + nfsd->nfsd_slp = slp; break; } } if (slp == 0) nfsd_head_flag &= ~NFSD_CHECKSLP; } - if ((slp = nd->nd_slp) == (struct nfssvc_sock *)0) + if ((slp = nfsd->nfsd_slp) == (struct nfssvc_sock *)0) continue; if (slp->ns_flag & SLP_VALID) { if (slp->ns_flag & SLP_DISCONN) @@ -446,86 +487,78 @@ nfssvc_nfsd(nsd, argp, p) M_WAIT); nfs_sndunlock(&slp->ns_solock); } - error = nfsrv_dorec(slp, nd); - nd->nd_flag |= NFSD_REQINPROG; + error = nfsrv_dorec(slp, nfsd, &nd); + cur_usec = (u_quad_t)time.tv_sec * 1000000 + + (u_quad_t)time.tv_usec; + if (error && slp->ns_tq.lh_first && + slp->ns_tq.lh_first->nd_time <= cur_usec) { + error = 0; + cacherep = RC_DOIT; + writes_todo = 1; + } else + writes_todo = 0; + nfsd->nfsd_flag |= NFSD_REQINPROG; } } else { error = 0; - slp = nd->nd_slp; + slp = nfsd->nfsd_slp; } if (error || (slp->ns_flag & SLP_VALID) == 0) { - nd->nd_slp = (struct nfssvc_sock *)0; - nd->nd_flag &= ~NFSD_REQINPROG; + if (nd) { + free((caddr_t)nd, M_NFSRVDESC); + nd = NULL; + } + nfsd->nfsd_slp = (struct nfssvc_sock *)0; + nfsd->nfsd_flag &= ~NFSD_REQINPROG; nfsrv_slpderef(slp); continue; } splx(s); so = slp->ns_so; sotype = so->so_type; - starttime = time; if (so->so_proto->pr_flags & PR_CONNREQUIRED) solockp = &slp->ns_solock; else solockp = (int *)0; - /* - * nam == nam2 for connectionless protocols such as UDP - * nam2 == NULL for connection based protocols to disable - * recent request caching. - */ - nam2 = nd->nd_nam; - if (nam2) { - nam = nam2; - cacherep = RC_CHECKIT; - } else { - nam = slp->ns_nam; - cacherep = RC_DOIT; - } - - /* - * Check to see if authorization is needed. - */ - if (nd->nd_flag & NFSD_NEEDAUTH) { - static int logauth = 0; + if (nd) { + nd->nd_starttime = time; + if (nd->nd_nam2) + nd->nd_nam = nd->nd_nam2; + else + nd->nd_nam = slp->ns_nam; - nd->nd_flag &= ~NFSD_NEEDAUTH; - /* - * Check for a mapping already installed. - */ - for (uidp = NUIDHASH(slp, nd->nd_cr.cr_uid)->lh_first; - uidp != 0; uidp = uidp->nu_hash.le_next) { - if (uidp->nu_uid == nd->nd_cr.cr_uid) - break; - } - if (!uidp) { - nsd->nsd_uid = nd->nd_cr.cr_uid; - if (nam2 && logauth++ == 0) - log(LOG_WARNING, "Kerberized NFS using UDP\n"); - nsd->nsd_haddr = - mtod(nam, struct sockaddr_in *)->sin_addr.s_addr; - nsd->nsd_authlen = nd->nd_authlen; - if (copyout(nd->nd_authstr, nsd->nsd_authstr, - nd->nd_authlen) == 0 && - copyout((caddr_t)nsd, argp, sizeof (*nsd)) == 0) - return (ENEEDAUTH); - cacherep = RC_DROPIT; - } - } - if (cacherep == RC_CHECKIT) - cacherep = nfsrv_getcache(nam2, nd, &mreq); + /* + * Check to see if authorization is needed. + */ + if (nfsd->nfsd_flag & NFSD_NEEDAUTH) { + nfsd->nfsd_flag &= ~NFSD_NEEDAUTH; + nsd->nsd_haddr = mtod(nd->nd_nam, + struct sockaddr_in *)->sin_addr.s_addr; + nsd->nsd_authlen = nfsd->nfsd_authlen; + nsd->nsd_verflen = nfsd->nfsd_verflen; + if (!copyout(nfsd->nfsd_authstr,nsd->nsd_authstr, + nfsd->nfsd_authlen) && + !copyout(nfsd->nfsd_verfstr, nsd->nsd_verfstr, + nfsd->nfsd_verflen) && + !copyout((caddr_t)nsd, argp, sizeof (*nsd))) + return (ENEEDAUTH); + cacherep = RC_DROPIT; + } else + cacherep = nfsrv_getcache(nd, slp, &mreq); - /* - * Check for just starting up for NQNFS and send - * fake "try again later" replies to the NQNFS clients. - */ - if (notstarted && nqnfsstarttime <= time.tv_sec) { + /* + * Check for just starting up for NQNFS and send + * fake "try again later" replies to the NQNFS clients. + */ + if (notstarted && nqnfsstarttime <= time.tv_sec) { if (modify_flag) { nqnfsstarttime = time.tv_sec + nqsrv_writeslack; modify_flag = 0; } else notstarted = 0; - } - if (notstarted) { - if (nd->nd_nqlflag == NQL_NOVAL) + } + if (notstarted) { + if ((nd->nd_flag & ND_NQNFS) == 0) cacherep = RC_DROPIT; else if (nd->nd_procnum != NFSPROC_WRITE) { nd->nd_procnum = NFSPROC_NOOP; @@ -533,36 +566,42 @@ nfssvc_nfsd(nsd, argp, p) cacherep = RC_DOIT; } else modify_flag = 1; - } else if (nd->nd_flag & NFSD_AUTHFAIL) { - nd->nd_flag &= ~NFSD_AUTHFAIL; + } else if (nfsd->nfsd_flag & NFSD_AUTHFAIL) { + nfsd->nfsd_flag &= ~NFSD_AUTHFAIL; nd->nd_procnum = NFSPROC_NOOP; - nd->nd_repstat = NQNFS_AUTHERR; + nd->nd_repstat = (NFSERR_AUTHERR | AUTH_TOOWEAK); cacherep = RC_DOIT; + } } - switch (cacherep) { - case RC_DOIT: - error = (*(nfsrv_procs[nd->nd_procnum]))(nd, - nd->nd_mrep, nd->nd_md, nd->nd_dpos, &nd->nd_cr, - nam, &mreq); - if (nd->nd_cr.cr_ref != 1) { - printf("nfssvc cref=%d\n", nd->nd_cr.cr_ref); - panic("nfssvc cref"); - } + /* + * Loop to get all the write rpc relies that have been + * gathered together. + */ + do { + switch (cacherep) { + case RC_DOIT: + if (writes_todo || (nd->nd_procnum == NFSPROC_WRITE && + nfsrvw_procrastinate > 0 && !notstarted)) + error = nfsrv_writegather(&nd, slp, + nfsd->nfsd_procp, &mreq); + else + error = (*(nfsrv3_procs[nd->nd_procnum]))(nd, + slp, nfsd->nfsd_procp, &mreq); + if (mreq == NULL) + break; if (error) { if (nd->nd_procnum != NQNFSPROC_VACATED) nfsstats.srv_errs++; - if (nam2) { - nfsrv_updatecache(nam2, nd, FALSE, mreq); - m_freem(nam2); - } + nfsrv_updatecache(nd, FALSE, mreq); + if (nd->nd_nam2) + m_freem(nd->nd_nam2); break; } nfsstats.srvrpccnt[nd->nd_procnum]++; - if (nam2) - nfsrv_updatecache(nam2, nd, TRUE, mreq); + nfsrv_updatecache(nd, TRUE, mreq); nd->nd_mrep = (struct mbuf *)0; - case RC_REPLY: + case RC_REPLY: m = mreq; siz = 0; while (m) { @@ -587,15 +626,15 @@ nfssvc_nfsd(nsd, argp, p) if (solockp) (void) nfs_sndlock(solockp, (struct nfsreq *)0); if (slp->ns_flag & SLP_VALID) - error = nfs_send(so, nam2, m, (struct nfsreq *)0); + error = nfs_send(so, nd->nd_nam2, m, NULL); else { error = EPIPE; m_freem(m); } if (nfsrtton) - nfsd_rt(&starttime, sotype, nd, nam, cacherep); - if (nam2) - MFREE(nam2, m); + nfsd_rt(sotype, nd, cacherep); + if (nd->nd_nam2) + MFREE(nd->nd_nam2, m); if (nd->nd_mrep) m_freem(nd->nd_mrep); if (error == EPIPE) @@ -603,29 +642,50 @@ nfssvc_nfsd(nsd, argp, p) if (solockp) nfs_sndunlock(solockp); if (error == EINTR || error == ERESTART) { + free((caddr_t)nd, M_NFSRVDESC); nfsrv_slpderef(slp); s = splnet(); goto done; } break; - case RC_DROPIT: + case RC_DROPIT: if (nfsrtton) - nfsd_rt(&starttime, sotype, nd, nam, cacherep); + nfsd_rt(sotype, nd, cacherep); m_freem(nd->nd_mrep); - m_freem(nam2); + m_freem(nd->nd_nam2); break; - }; + }; + if (nd) { + FREE((caddr_t)nd, M_NFSRVDESC); + nd = NULL; + } + + /* + * Check to see if there are outstanding writes that + * need to be serviced. + */ + cur_usec = (u_quad_t)time.tv_sec * 1000000 + + (u_quad_t)time.tv_usec; + s = splsoftclock(); + if (slp->ns_tq.lh_first && + slp->ns_tq.lh_first->nd_time <= cur_usec) { + cacherep = RC_DOIT; + writes_todo = 1; + } else + writes_todo = 0; + splx(s); + } while (writes_todo); s = splnet(); - if (nfsrv_dorec(slp, nd)) { - nd->nd_flag &= ~NFSD_REQINPROG; - nd->nd_slp = (struct nfssvc_sock *)0; + if (nfsrv_dorec(slp, nfsd, &nd)) { + nfsd->nfsd_flag &= ~NFSD_REQINPROG; + nfsd->nfsd_slp = NULL; nfsrv_slpderef(slp); } } done: - TAILQ_REMOVE(&nfsd_head, nd, nd_chain); + TAILQ_REMOVE(&nfsd_head, nfsd, nfsd_chain); splx(s); - free((caddr_t)nd, M_NFSD); + free((caddr_t)nfsd, M_NFSD); nsd->nsd_nfsd = (struct nfsd *)0; if (--nfs_numnfsd == 0) nfsrv_init(TRUE); /* Reinitialize everything */ @@ -641,9 +701,10 @@ int nfssvc_iod(p) struct proc *p; { - register struct buf *bp; + register struct buf *bp, *nbp; register int i, myiod; - int error = 0; + struct vnode *vp; + int error = 0, s; /* * Assign my position or return error if too many already running @@ -662,24 +723,53 @@ nfssvc_iod(p) * Just loop around doin our stuff until SIGKILL */ for (;;) { - while (nfs_bufq.tqh_first == NULL && error == 0) { - nfs_iodwant[myiod] = p; - error = tsleep((caddr_t)&nfs_iodwant[myiod], - PWAIT | PCATCH, "nfsidl", 0); - } - while ((bp = nfs_bufq.tqh_first) != NULL) { - /* Take one off the front of the list */ - TAILQ_REMOVE(&nfs_bufq, bp, b_freelist); - if (bp->b_flags & B_READ) - (void) nfs_doio(bp, bp->b_rcred, (struct proc *)0); - else - (void) nfs_doio(bp, bp->b_wcred, (struct proc *)0); - } - if (error) { - nfs_asyncdaemon[myiod] = 0; - nfs_numasync--; - return (error); - } + while (nfs_bufq.tqh_first == NULL && error == 0) { + nfs_iodwant[myiod] = p; + error = tsleep((caddr_t)&nfs_iodwant[myiod], + PWAIT | PCATCH, "nfsidl", 0); + } + while ((bp = nfs_bufq.tqh_first) != NULL) { + /* Take one off the front of the list */ + TAILQ_REMOVE(&nfs_bufq, bp, b_freelist); + if (bp->b_flags & B_READ) + (void) nfs_doio(bp, bp->b_rcred, (struct proc *)0); + else do { + /* + * Look for a delayed write for the same vnode, so I can do + * it now. We must grab it before calling nfs_doio() to + * avoid any risk of the vnode getting vclean()'d while + * we are doing the write rpc. + */ + vp = bp->b_vp; + s = splbio(); + for (nbp = vp->v_dirtyblkhd.lh_first; nbp; + nbp = nbp->b_vnbufs.le_next) { + if ((nbp->b_flags & + (B_BUSY|B_DELWRI|B_NEEDCOMMIT|B_NOCACHE))!=B_DELWRI) + continue; + bremfree(nbp); + vfs_busy_pages(nbp, 1); + nbp->b_flags |= (B_BUSY|B_ASYNC); + break; + } + splx(s); + /* + * For the delayed write, do the first part of nfs_bwrite() + * up to, but not including nfs_strategy(). + */ + if (nbp) { + nbp->b_flags &= ~(B_READ|B_DONE|B_ERROR|B_DELWRI); + reassignbuf(nbp, nbp->b_vp); + nbp->b_vp->v_numoutput++; + } + (void) nfs_doio(bp, bp->b_wcred, (struct proc *)0); + } while (bp = nbp); + } + if (error) { + nfs_asyncdaemon[myiod] = 0; + nfs_numasync--; + return (error); + } } } @@ -695,9 +785,11 @@ nfsrv_zapsock(slp) register struct nfssvc_sock *slp; { register struct nfsuid *nuidp, *nnuidp; + register struct nfsrv_descript *nwp, *nnwp; struct socket *so; struct file *fp; struct mbuf *m; + int s; slp->ns_flag &= ~SLP_ALLFLAGS; fp = slp->ns_fp; @@ -716,8 +808,18 @@ nfsrv_zapsock(slp) nnuidp = nuidp->nu_lru.tqe_next; LIST_REMOVE(nuidp, nu_hash); TAILQ_REMOVE(&slp->ns_uidlruhead, nuidp, nu_lru); + if (nuidp->nu_flag & NU_NAM) + m_freem(nuidp->nu_nam); free((caddr_t)nuidp, M_NFSUID); } + s = splsoftclock(); + for (nwp = slp->ns_tq.lh_first; nwp; nwp = nnwp) { + nnwp = nwp->nd_tq.le_next; + LIST_REMOVE(nwp, nd_tq); + free((caddr_t)nwp, M_NFSRVDESC); + } + LIST_INIT(&slp->ns_tq); + splx(s); } } @@ -726,13 +828,15 @@ nfsrv_zapsock(slp) * on this mount point porpous out of the kernel and do it. */ int -nfs_getauth(nmp, rep, cred, auth_type, auth_str, auth_len) +nfs_getauth(nmp, rep, cred, auth_str, auth_len, verf_str, verf_len, key) register struct nfsmount *nmp; struct nfsreq *rep; struct ucred *cred; - int *auth_type; char **auth_str; int *auth_len; + char *verf_str; + int *verf_len; + NFSKERBKEY_T key; /* return session key */ { int error = 0; @@ -748,6 +852,9 @@ nfs_getauth(nmp, rep, cred, auth_type, auth_str, auth_len) } nmp->nm_flag &= ~(NFSMNT_WAITAUTH | NFSMNT_WANTAUTH); nmp->nm_authstr = *auth_str = (char *)malloc(RPCAUTH_MAXSIZ, M_TEMP, M_WAITOK); + nmp->nm_authlen = RPCAUTH_MAXSIZ; + nmp->nm_verfstr = verf_str; + nmp->nm_verflen = *verf_len; nmp->nm_authuid = cred->cr_uid; wakeup((caddr_t)&nmp->nm_authstr); @@ -766,8 +873,9 @@ nfs_getauth(nmp, rep, cred, auth_type, auth_str, auth_len) if (error) free((caddr_t)*auth_str, M_TEMP); else { - *auth_type = nmp->nm_authtype; *auth_len = nmp->nm_authlen; + *verf_len = nmp->nm_verflen; + bcopy((caddr_t)nmp->nm_key, (caddr_t)key, sizeof (key)); } nmp->nm_flag &= ~NFSMNT_HASAUTH; nmp->nm_flag |= NFSMNT_WAITAUTH; @@ -779,6 +887,149 @@ nfs_getauth(nmp, rep, cred, auth_type, auth_str, auth_len) } /* + * Get a nickname authenticator and verifier. + */ +int +nfs_getnickauth(nmp, cred, auth_str, auth_len, verf_str, verf_len) + struct nfsmount *nmp; + struct ucred *cred; + char **auth_str; + int *auth_len; + char *verf_str; + int verf_len; +{ + register struct nfsuid *nuidp; + register u_long *nickp, *verfp; + struct timeval ktvin, ktvout; + NFSKERBKEYSCHED_T keys; /* stores key schedule */ + +#ifdef DIAGNOSTIC + if (verf_len < (4 * NFSX_UNSIGNED)) + panic("nfs_getnickauth verf too small"); +#endif + for (nuidp = NMUIDHASH(nmp, cred->cr_uid)->lh_first; + nuidp != 0; nuidp = nuidp->nu_hash.le_next) { + if (nuidp->nu_cr.cr_uid == cred->cr_uid) + break; + } + if (!nuidp || nuidp->nu_expire < time.tv_sec) + return (EACCES); + + /* + * Move to the end of the lru list (end of lru == most recently used). + */ + TAILQ_REMOVE(&nmp->nm_uidlruhead, nuidp, nu_lru); + TAILQ_INSERT_TAIL(&nmp->nm_uidlruhead, nuidp, nu_lru); + + nickp = (u_long *)malloc(2 * NFSX_UNSIGNED, M_TEMP, M_WAITOK); + *nickp++ = txdr_unsigned(RPCAKN_NICKNAME); + *nickp = txdr_unsigned(nuidp->nu_nickname); + *auth_str = (char *)nickp; + *auth_len = 2 * NFSX_UNSIGNED; + + /* + * Now we must encrypt the verifier and package it up. + */ + verfp = (u_long *)verf_str; + *verfp++ = txdr_unsigned(RPCAKN_NICKNAME); + if (time.tv_sec > nuidp->nu_timestamp.tv_sec || + (time.tv_sec == nuidp->nu_timestamp.tv_sec && + time.tv_usec > nuidp->nu_timestamp.tv_usec)) + nuidp->nu_timestamp = time; + else + nuidp->nu_timestamp.tv_usec++; + ktvin.tv_sec = txdr_unsigned(nuidp->nu_timestamp.tv_sec); + ktvin.tv_usec = txdr_unsigned(nuidp->nu_timestamp.tv_usec); + + /* + * Now encrypt the timestamp verifier in ecb mode using the session + * key. + */ +#ifdef NFSKERB + XXX +#endif + + *verfp++ = ktvout.tv_sec; + *verfp++ = ktvout.tv_usec; + *verfp = 0; + return (0); +} + +/* + * Save the current nickname in a hash list entry on the mount point. + */ +int +nfs_savenickauth(nmp, cred, len, key, mdp, dposp, mrep) + register struct nfsmount *nmp; + struct ucred *cred; + int len; + NFSKERBKEY_T key; + struct mbuf **mdp; + char **dposp; + struct mbuf *mrep; +{ + register struct nfsuid *nuidp; + register u_long *tl; + register long t1; + struct mbuf *md = *mdp; + struct timeval ktvin, ktvout; + u_long nick; + NFSKERBKEYSCHED_T keys; + char *dpos = *dposp, *cp2; + int deltasec, error = 0; + + if (len == (3 * NFSX_UNSIGNED)) { + nfsm_dissect(tl, u_long *, 3 * NFSX_UNSIGNED); + ktvin.tv_sec = *tl++; + ktvin.tv_usec = *tl++; + nick = fxdr_unsigned(u_long, *tl); + + /* + * Decrypt the timestamp in ecb mode. + */ +#ifdef NFSKERB + XXX +#endif + ktvout.tv_sec = fxdr_unsigned(long, ktvout.tv_sec); + ktvout.tv_usec = fxdr_unsigned(long, ktvout.tv_usec); + deltasec = time.tv_sec - ktvout.tv_sec; + if (deltasec < 0) + deltasec = -deltasec; + /* + * If ok, add it to the hash list for the mount point. + */ + if (deltasec <= NFS_KERBCLOCKSKEW) { + if (nmp->nm_numuids < nuidhash_max) { + nmp->nm_numuids++; + nuidp = (struct nfsuid *) + malloc(sizeof (struct nfsuid), M_NFSUID, + M_WAITOK); + } else { + nuidp = nmp->nm_uidlruhead.tqh_first; + LIST_REMOVE(nuidp, nu_hash); + TAILQ_REMOVE(&nmp->nm_uidlruhead, nuidp, + nu_lru); + } + nuidp->nu_flag = 0; + nuidp->nu_cr.cr_uid = cred->cr_uid; + nuidp->nu_expire = time.tv_sec + NFS_KERBTTL; + nuidp->nu_timestamp = ktvout; + nuidp->nu_nickname = nick; + bcopy(key, nuidp->nu_key, sizeof (key)); + TAILQ_INSERT_TAIL(&nmp->nm_uidlruhead, nuidp, + nu_lru); + LIST_INSERT_HEAD(NMUIDHASH(nmp, cred->cr_uid), + nuidp, nu_hash); + } + } else + nfsm_adv(nfsm_rndup(len)); +nfsmout: + *mdp = md; + *dposp = dpos; + return (error); +} + +/* * Derefence a server socket structure. If it has no more references and * is no longer valid, you can throw it away. */ @@ -830,16 +1081,12 @@ nfsrv_init(terminating) nfs_udpsock = (struct nfssvc_sock *) malloc(sizeof (struct nfssvc_sock), M_NFSSVC, M_WAITOK); bzero((caddr_t)nfs_udpsock, sizeof (struct nfssvc_sock)); - nfs_udpsock->ns_uidhashtbl = - hashinit(NUIDHASHSIZ, M_NFSSVC, &nfs_udpsock->ns_uidhash); TAILQ_INIT(&nfs_udpsock->ns_uidlruhead); TAILQ_INSERT_HEAD(&nfssvc_sockhead, nfs_udpsock, ns_chain); nfs_cltpsock = (struct nfssvc_sock *) malloc(sizeof (struct nfssvc_sock), M_NFSSVC, M_WAITOK); bzero((caddr_t)nfs_cltpsock, sizeof (struct nfssvc_sock)); - nfs_cltpsock->ns_uidhashtbl = - hashinit(NUIDHASHSIZ, M_NFSSVC, &nfs_cltpsock->ns_uidhash); TAILQ_INIT(&nfs_cltpsock->ns_uidlruhead); TAILQ_INSERT_TAIL(&nfssvc_sockhead, nfs_cltpsock, ns_chain); } @@ -848,11 +1095,9 @@ nfsrv_init(terminating) * Add entries to the server monitor log. */ static void -nfsd_rt(startp, sotype, nd, nam, cacherep) - struct timeval *startp; +nfsd_rt(sotype, nd, cacherep) int sotype; - register struct nfsd *nd; - struct mbuf *nam; + register struct nfsrv_descript *nd; int cacherep; { register struct drt *rt; @@ -866,15 +1111,17 @@ nfsd_rt(startp, sotype, nd, nam, cacherep) rt->flag = DRT_CACHEDROP; if (sotype == SOCK_STREAM) rt->flag |= DRT_TCP; - if (nd->nd_nqlflag != NQL_NOVAL) + if (nd->nd_flag & ND_NQNFS) rt->flag |= DRT_NQNFS; + else if (nd->nd_flag & ND_NFSV3) + rt->flag |= DRT_NFSV3; rt->proc = nd->nd_procnum; - if (mtod(nam, struct sockaddr *)->sa_family == AF_INET) - rt->ipadr = mtod(nam, struct sockaddr_in *)->sin_addr.s_addr; + if (mtod(nd->nd_nam, struct sockaddr *)->sa_family == AF_INET) + rt->ipadr = mtod(nd->nd_nam, struct sockaddr_in *)->sin_addr.s_addr; else - rt->ipadr = INADDR_ANY; - rt->resptime = ((time.tv_sec - startp->tv_sec) * 1000000) + - (time.tv_usec - startp->tv_usec); + rt->ipadr = INADDR_ANY; + rt->resptime = ((time.tv_sec - nd->nd_starttime.tv_sec) * 1000000) + + (time.tv_usec - nd->nd_starttime.tv_usec); rt->tstamp = time; nfsdrt.pos = (nfsdrt.pos + 1) % NFSRTTLOGSIZ; } diff --git a/sys/nfs/nfs_vfsops.c b/sys/nfs/nfs_vfsops.c index 2daa891..2c4070b 100644 --- a/sys/nfs/nfs_vfsops.c +++ b/sys/nfs/nfs_vfsops.c @@ -34,7 +34,7 @@ * SUCH DAMAGE. * * @(#)nfs_vfsops.c 8.3 (Berkeley) 1/4/94 - * $Id: nfs_vfsops.c,v 1.14.2.1 1995/06/02 11:13:15 davidg Exp $ + * $Id: nfs_vfsops.c,v 1.15 1995/06/11 19:31:46 rgrimes Exp $ */ #include <sys/param.h> @@ -59,10 +59,10 @@ #include <netinet/in.h> #include <nfs/rpcv2.h> -#include <nfs/nfsv2.h> +#include <nfs/nfsproto.h> #include <nfs/nfsnode.h> -#include <nfs/nfsmount.h> #include <nfs/nfs.h> +#include <nfs/nfsmount.h> #include <nfs/xdr_subs.h> #include <nfs/nfsm_subs.h> #include <nfs/nfsdiskless.h> @@ -71,11 +71,15 @@ struct nfsstats nfsstats; static int nfs_sysctl(int *, u_int, void *, size_t *, void *, size_t, struct proc *); +extern int nfs_ticks; /* * nfs vfs operations. */ struct vfsops nfs_vfsops = { +#ifdef __NetBSD__ + MOUNT_NFS, +#endif nfs_mount, nfs_start, nfs_unmount, @@ -87,9 +91,13 @@ struct vfsops nfs_vfsops = { nfs_fhtovp, nfs_vptofh, nfs_init, +#ifdef __FreeBSD__ nfs_sysctl +#endif }; +#ifdef __FreeBSD__ VFS_SET(nfs_vfsops, nfs, MOUNT_NFS, VFCF_NETWORK); +#endif /* * This structure must be filled in by a primary bootstrap or bootstrap @@ -99,16 +107,11 @@ VFS_SET(nfs_vfsops, nfs, MOUNT_NFS, VFCF_NETWORK); struct nfs_diskless nfs_diskless = { 0 }; int nfs_diskless_valid = 0; -extern u_long nfs_procids[NFS_NPROCS]; -extern u_long nfs_prog, nfs_vers; void nfs_disconnect __P((struct nfsmount *)); void nfsargs_ntoh __P((struct nfs_args *)); static struct mount *nfs_mountdiskless __P((char *, char *, int, struct sockaddr_in *, struct nfs_args *, register struct vnode **)); -#define TRUE 1 -#define FALSE 0 - /* * nfs statfs call */ @@ -119,40 +122,65 @@ nfs_statfs(mp, sbp, p) struct proc *p; { register struct vnode *vp; - register struct nfsv2_statfs *sfp; + register struct nfs_statfs *sfp; register caddr_t cp; - register long t1; + register u_long *tl; + register long t1, t2; caddr_t bpos, dpos, cp2; - int error = 0, isnq; + struct nfsmount *nmp = VFSTONFS(mp); + int error = 0, v3 = (nmp->nm_flag & NFSMNT_NFSV3), retattr; struct mbuf *mreq, *mrep, *md, *mb, *mb2; - struct nfsmount *nmp; struct ucred *cred; struct nfsnode *np; + u_quad_t tquad; - nmp = VFSTONFS(mp); - isnq = (nmp->nm_flag & NFSMNT_NQNFS); - error = nfs_nget(mp, &nmp->nm_fh, &np); +#ifndef nolint + sfp = (struct nfs_statfs *)0; +#endif + error = nfs_nget(mp, (nfsfh_t *)nmp->nm_fh, nmp->nm_fhsize, &np); if (error) return (error); vp = NFSTOV(np); - nfsstats.rpccnt[NFSPROC_STATFS]++; cred = crget(); cred->cr_ngroups = 1; - nfsm_reqhead(vp, NFSPROC_STATFS, NFSX_FH); - nfsm_fhtom(vp); - nfsm_request(vp, NFSPROC_STATFS, p, cred); - nfsm_dissect(sfp, struct nfsv2_statfs *, NFSX_STATFS(isnq)); + if (v3 && (nmp->nm_flag & NFSMNT_GOTFSINFO) == 0) + (void)nfs_fsinfo(nmp, vp, cred, p); + nfsstats.rpccnt[NFSPROC_FSSTAT]++; + nfsm_reqhead(vp, NFSPROC_FSSTAT, NFSX_FH(v3)); + nfsm_fhtom(vp, v3); + nfsm_request(vp, NFSPROC_FSSTAT, p, cred); + if (v3) + nfsm_postop_attr(vp, retattr); + if (!error) + nfsm_dissect(sfp, struct nfs_statfs *, NFSX_STATFS(v3)); +#ifdef __NetBSD__ +#ifdef COMPAT_09 + sbp->f_type = 2; +#else + sbp->f_type = 0; +#endif +#else sbp->f_type = MOUNT_NFS; +#endif sbp->f_flags = nmp->nm_flag; - sbp->f_iosize = NFS_MAXDGRAMDATA; - sbp->f_bsize = fxdr_unsigned(long, sfp->sf_bsize); - sbp->f_blocks = fxdr_unsigned(long, sfp->sf_blocks); - sbp->f_bfree = fxdr_unsigned(long, sfp->sf_bfree); - sbp->f_bavail = fxdr_unsigned(long, sfp->sf_bavail); - if (isnq) { - sbp->f_files = fxdr_unsigned(long, sfp->sf_files); - sbp->f_ffree = fxdr_unsigned(long, sfp->sf_ffree); + sbp->f_iosize = min(nmp->nm_rsize, nmp->nm_wsize); + if (v3) { + sbp->f_bsize = NFS_FABLKSIZE; + fxdr_hyper(&sfp->sf_tbytes, &tquad); + sbp->f_blocks = (long)(tquad / ((u_quad_t)NFS_FABLKSIZE)); + fxdr_hyper(&sfp->sf_fbytes, &tquad); + sbp->f_bfree = (long)(tquad / ((u_quad_t)NFS_FABLKSIZE)); + fxdr_hyper(&sfp->sf_abytes, &tquad); + sbp->f_bavail = (long)(tquad / ((u_quad_t)NFS_FABLKSIZE)); + sbp->f_files = (fxdr_unsigned(long, sfp->sf_tfiles.nfsuquad[1]) + & 0x7fffffff); + sbp->f_ffree = (fxdr_unsigned(long, sfp->sf_ffiles.nfsuquad[1]) + & 0x7fffffff); } else { + sbp->f_bsize = fxdr_unsigned(long, sfp->sf_bsize); + sbp->f_blocks = fxdr_unsigned(long, sfp->sf_blocks); + sbp->f_bfree = fxdr_unsigned(long, sfp->sf_bfree); + sbp->f_bavail = fxdr_unsigned(long, sfp->sf_bavail); sbp->f_files = 0; sbp->f_ffree = 0; } @@ -161,12 +189,72 @@ nfs_statfs(mp, sbp, p) bcopy(mp->mnt_stat.f_mntfromname, sbp->f_mntfromname, MNAMELEN); } nfsm_reqdone; - vrele(vp); + vput(vp); crfree(cred); return (error); } /* + * nfs version 3 fsinfo rpc call + */ +int +nfs_fsinfo(nmp, vp, cred, p) + register struct nfsmount *nmp; + register struct vnode *vp; + struct ucred *cred; + struct proc *p; +{ + register struct nfsv3_fsinfo *fsp; + register caddr_t cp; + register long t1, t2; + register u_long *tl, pref, max; + caddr_t bpos, dpos, cp2; + int error = 0, retattr; + struct mbuf *mreq, *mrep, *md, *mb, *mb2; + + nfsstats.rpccnt[NFSPROC_FSINFO]++; + nfsm_reqhead(vp, NFSPROC_FSINFO, NFSX_FH(1)); + nfsm_fhtom(vp, 1); + nfsm_request(vp, NFSPROC_FSINFO, p, cred); + nfsm_postop_attr(vp, retattr); + if (!error) { + nfsm_dissect(fsp, struct nfsv3_fsinfo *, NFSX_V3FSINFO); + pref = fxdr_unsigned(u_long, fsp->fs_wtpref); + if (pref < nmp->nm_wsize) + nmp->nm_wsize = (pref + NFS_FABLKSIZE - 1) & + ~(NFS_FABLKSIZE - 1); + max = fxdr_unsigned(u_long, fsp->fs_wtmax); + if (max < nmp->nm_wsize) { + nmp->nm_wsize = max & ~(NFS_FABLKSIZE - 1); + if (nmp->nm_wsize == 0) + nmp->nm_wsize = max; + } + pref = fxdr_unsigned(u_long, fsp->fs_rtpref); + if (pref < nmp->nm_rsize) + nmp->nm_rsize = (pref + NFS_FABLKSIZE - 1) & + ~(NFS_FABLKSIZE - 1); + max = fxdr_unsigned(u_long, fsp->fs_rtmax); + if (max < nmp->nm_rsize) { + nmp->nm_rsize = max & ~(NFS_FABLKSIZE - 1); + if (nmp->nm_rsize == 0) + nmp->nm_rsize = max; + } + pref = fxdr_unsigned(u_long, fsp->fs_dtpref); + if (pref < nmp->nm_readdirsize) + nmp->nm_readdirsize = (pref + NFS_DIRBLKSIZ - 1) & + ~(NFS_DIRBLKSIZ - 1); + if (max < nmp->nm_readdirsize) { + nmp->nm_readdirsize = max & ~(NFS_DIRBLKSIZ - 1); + if (nmp->nm_readdirsize == 0) + nmp->nm_readdirsize = max; + } + nmp->nm_flag |= NFSMNT_GOTFSINFO; + } + nfsm_reqdone; + return (error); +} + +/* * Mount a remote root fs via. nfs. This depends on the info in the * nfs_diskless structure that has been filled in properly by some primary * bootstrap. @@ -268,7 +356,12 @@ nfs_mountroot() * Create a fake mount point just for the swap vnode so that the * swap file can be on a different server from the rootfs. */ - nd->swap_args.fh = (nfsv2fh_t *)nd->swap_fh; + nd->swap_args.fh = nd->swap_fh; + /* + * If using nfsv3_diskless, replace NFSX_V2FH with + * nd->swap_fhsize. + */ + nd->swap_args.fhsize = NFSX_V2FH; l = ntohl(nd->swap_saddr.sin_addr.s_addr); sprintf(buf,"%ld.%ld.%ld.%ld:%s", (l >> 24) & 0xff, (l >> 16) & 0xff, @@ -293,8 +386,12 @@ nfs_mountroot() /* * Create the rootfs mount point. */ - nd->root_args.fh = (nfsv2fh_t *)nd->root_fh; - l = ntohl(nd->root_saddr.sin_addr.s_addr); + nd->root_args.fh = nd->root_fh; + /* + * If using nfsv3_diskless, replace NFSX_V2FH with nd->root_fhsize. + */ + nd->root_args.fhsize = NFSX_V2FH; + l = ntohl(nd->swap_saddr.sin_addr.s_addr); sprintf(buf,"%ld.%ld.%ld.%ld:%s", (l >> 24) & 0xff, (l >> 16) & 0xff, (l >> 8) & 0xff, (l >> 0) & 0xff,nd->root_hostnam); @@ -304,7 +401,11 @@ nfs_mountroot() if (vfs_lock(mp)) panic("nfs_mountroot: vfs_lock"); +#if NetBSD >= 1994101 + CIRCLEQ_INSERT_TAIL(&mountlist, mp, mnt_list); +#else TAILQ_INSERT_TAIL(&mountlist, mp, mnt_list); +#endif mp->mnt_flag |= MNT_ROOTFS; mp->mnt_vnodecovered = NULLVP; vfs_unlock(mp); @@ -386,12 +487,12 @@ nfs_mount(mp, path, data, ndp, p) struct vnode *vp; char pth[MNAMELEN], hst[MNAMELEN]; u_int len; - nfsv2fh_t nfh; + u_char nfh[NFSX_V3FHMAX]; error = copyin(data, (caddr_t)&args, sizeof (struct nfs_args)); if (error) return (error); - error = copyin((caddr_t)args.fh, (caddr_t)&nfh, sizeof (nfsv2fh_t)); + error = copyin((caddr_t)args.fh, (caddr_t)nfh, args.fhsize); if (error) return (error); error = copyinstr(path, pth, MNAMELEN-1, &len); @@ -406,7 +507,7 @@ nfs_mount(mp, path, data, ndp, p) error = sockargs(&nam, (caddr_t)args.addr, args.addrlen, MT_SONAME); if (error) return (error); - args.fh = &nfh; + args.fh = nfh; error = mountnfs(&args, mp, nam, pth, hst, &vp); return (error); } @@ -424,7 +525,7 @@ mountnfs(argp, mp, nam, pth, hst, vpp) { register struct nfsmount *nmp; struct nfsnode *np; - int error; + int error, maxio; if (mp->mnt_flag & MNT_UPDATE) { nmp = VFSTONFS(mp); @@ -435,16 +536,12 @@ mountnfs(argp, mp, nam, pth, hst, vpp) MALLOC(nmp, struct nfsmount *, sizeof (struct nfsmount), M_NFSMNT, M_WAITOK); bzero((caddr_t)nmp, sizeof (struct nfsmount)); + TAILQ_INIT(&nmp->nm_uidlruhead); mp->mnt_data = (qaddr_t)nmp; } getnewfsid(mp, MOUNT_NFS); nmp->nm_mountp = mp; nmp->nm_flag = argp->flags; - if ((nmp->nm_flag & (NFSMNT_NQNFS | NFSMNT_MYWRITE)) == - (NFSMNT_NQNFS | NFSMNT_MYWRITE)) { - error = EPERM; - goto bad; - } if (nmp->nm_flag & NFSMNT_NQNFS) /* * We have to set mnt_maxsymlink to a non-zero value so @@ -457,14 +554,24 @@ mountnfs(argp, mp, nam, pth, hst, vpp) nmp->nm_retry = NFS_RETRANS; nmp->nm_wsize = NFS_WSIZE; nmp->nm_rsize = NFS_RSIZE; + nmp->nm_readdirsize = NFS_READDIRSIZE; nmp->nm_numgrps = NFS_MAXGRPS; nmp->nm_readahead = NFS_DEFRAHEAD; nmp->nm_leaseterm = NQ_DEFLEASE; nmp->nm_deadthresh = NQ_DEADTHRESH; CIRCLEQ_INIT(&nmp->nm_timerhead); nmp->nm_inprog = NULLVP; - bcopy((caddr_t)argp->fh, (caddr_t)&nmp->nm_fh, sizeof(nfsv2fh_t)); + nmp->nm_fhsize = argp->fhsize; + bcopy((caddr_t)argp->fh, (caddr_t)nmp->nm_fh, argp->fhsize); +#ifdef __NetBSD__ +#ifdef COMPAT_09 + mp->mnt_stat.f_type = 2; +#else + mp->mnt_stat.f_type = 0; +#endif +#else mp->mnt_stat.f_type = MOUNT_NFS; +#endif bcopy(hst, mp->mnt_stat.f_mntfromname, MNAMELEN); bcopy(pth, mp->mnt_stat.f_mntonname, MNAMELEN); nmp->nm_nam = nam; @@ -483,29 +590,48 @@ mountnfs(argp, mp, nam, pth, hst, vpp) nmp->nm_retry = NFS_MAXREXMIT; } + if (argp->flags & NFSMNT_NFSV3) { + if (argp->sotype == SOCK_DGRAM) + maxio = NFS_MAXDGRAMDATA; + else + maxio = NFS_MAXDATA; + } else + maxio = NFS_V2MAXDATA; + if ((argp->flags & NFSMNT_WSIZE) && argp->wsize > 0) { nmp->nm_wsize = argp->wsize; /* Round down to multiple of blocksize */ - nmp->nm_wsize &= ~0x1ff; + nmp->nm_wsize &= ~(NFS_FABLKSIZE - 1); if (nmp->nm_wsize <= 0) - nmp->nm_wsize = 512; - else if (nmp->nm_wsize > NFS_MAXDATA) - nmp->nm_wsize = NFS_MAXDATA; + nmp->nm_wsize = NFS_FABLKSIZE; } + if (nmp->nm_wsize > maxio) + nmp->nm_wsize = maxio; if (nmp->nm_wsize > MAXBSIZE) nmp->nm_wsize = MAXBSIZE; if ((argp->flags & NFSMNT_RSIZE) && argp->rsize > 0) { nmp->nm_rsize = argp->rsize; /* Round down to multiple of blocksize */ - nmp->nm_rsize &= ~0x1ff; + nmp->nm_rsize &= ~(NFS_FABLKSIZE - 1); if (nmp->nm_rsize <= 0) - nmp->nm_rsize = 512; - else if (nmp->nm_rsize > NFS_MAXDATA) - nmp->nm_rsize = NFS_MAXDATA; + nmp->nm_rsize = NFS_FABLKSIZE; } + if (nmp->nm_rsize > maxio) + nmp->nm_rsize = maxio; if (nmp->nm_rsize > MAXBSIZE) nmp->nm_rsize = MAXBSIZE; + + if ((argp->flags & NFSMNT_READDIRSIZE) && argp->readdirsize > 0) { + nmp->nm_readdirsize = argp->readdirsize; + /* Round down to multiple of blocksize */ + nmp->nm_readdirsize &= ~(NFS_DIRBLKSIZ - 1); + if (nmp->nm_readdirsize < NFS_DIRBLKSIZ) + nmp->nm_readdirsize = NFS_DIRBLKSIZ; + } + if (nmp->nm_readdirsize > maxio) + nmp->nm_readdirsize = maxio; + if ((argp->flags & NFSMNT_MAXGRPS) && argp->maxgrouplist >= 0 && argp->maxgrouplist <= NFS_MAXGRPS) nmp->nm_numgrps = argp->maxgrouplist; @@ -536,7 +662,7 @@ mountnfs(argp, mp, nam, pth, hst, vpp) * stuck on a dead server and we are holding a lock on the mount * point. */ - mp->mnt_stat.f_iosize = NFS_MAXDGRAMDATA; + mp->mnt_stat.f_iosize = min(nmp->nm_rsize, nmp->nm_wsize); /* * A reference count is needed on the nfsnode representing the * remote root. If this object is not persistent, then backward @@ -545,11 +671,16 @@ mountnfs(argp, mp, nam, pth, hst, vpp) * this problem, because one can identify root inodes by their * number == ROOTINO (2). */ - error = nfs_nget(mp, &nmp->nm_fh, &np); + error = nfs_nget(mp, (nfsfh_t *)nmp->nm_fh, nmp->nm_fhsize, &np); if (error) goto bad; *vpp = NFSTOV(np); + /* + * Lose the lock but keep the ref. + */ + VOP_UNLOCK(*vpp); + return (0); bad: nfs_disconnect(nmp); @@ -592,7 +723,7 @@ nfs_unmount(mp, mntflags, p) * the remote root. See comment in mountnfs(). The VFS unmount() * has done vput on this vnode, otherwise we would get deadlock! */ - error = nfs_nget(mp, &nmp->nm_fh, &np); + error = nfs_nget(mp, (nfsfh_t *)nmp->nm_fh, nmp->nm_fhsize, &np); if (error) return(error); vp = NFSTOV(np); @@ -622,9 +753,9 @@ nfs_unmount(mp, mntflags, p) nmp->nm_flag |= NFSMNT_DISMNT; /* - * There are two reference counts to get rid of here. + * There are two reference counts and one lock to get rid of here. */ - vrele(vp); + vput(vp); vrele(vp); vgone(vp); nfs_disconnect(nmp); @@ -649,10 +780,11 @@ nfs_root(mp, vpp) int error; nmp = VFSTONFS(mp); - error = nfs_nget(mp, &nmp->nm_fh, &np); + error = nfs_nget(mp, (nfsfh_t *)nmp->nm_fh, nmp->nm_fhsize, &np); if (error) return (error); vp = NFSTOV(np); + VOP_UNLOCK(vp); vp->v_type = VDIR; vp->v_flag = VROOT; *vpp = vp; diff --git a/sys/nfs/nfs_vnops.c b/sys/nfs/nfs_vnops.c index 46484ea..769b0bb 100644 --- a/sys/nfs/nfs_vnops.c +++ b/sys/nfs/nfs_vnops.c @@ -34,17 +34,18 @@ * SUCH DAMAGE. * * @(#)nfs_vnops.c 8.5 (Berkeley) 2/13/94 - * $Id: nfs_vnops.c,v 1.14 1995/03/23 09:43:40 davidg Exp $ + * $Id: nfs_vnops.c,v 1.15 1995/05/30 08:12:49 rgrimes Exp $ */ /* - * vnode op calls for sun nfs version 2 + * vnode op calls for Sun NFS version 2 and 3 */ #include <sys/param.h> -#include <sys/proc.h> #include <sys/kernel.h> #include <sys/systm.h> +#include <sys/resourcevar.h> +#include <sys/proc.h> #include <sys/mount.h> #include <sys/buf.h> #include <sys/malloc.h> @@ -53,7 +54,9 @@ #include <sys/namei.h> #include <sys/vnode.h> #include <sys/dirent.h> +#include <sys/fcntl.h> #include <sys/lockf.h> +#include <ufs/ufs/dir.h> #include <vm/vm.h> @@ -61,7 +64,7 @@ #include <miscfs/fifofs/fifo.h> #include <nfs/rpcv2.h> -#include <nfs/nfsv2.h> +#include <nfs/nfsproto.h> #include <nfs/nfs.h> #include <nfs/nfsnode.h> #include <nfs/nfsmount.h> @@ -69,11 +72,24 @@ #include <nfs/nfsm_subs.h> #include <nfs/nqnfs.h> +#include <net/if.h> +#include <netinet/in.h> +#include <netinet/in_var.h> + /* Defs */ #define TRUE 1 #define FALSE 0 /* + * Ifdef for FreeBSD-current merged buffer cache. It is unfortunate that these + * calls are not in getblk() and brelse() so that they would not be necessary + * here. + */ +#ifndef B_VMIO +#define vfs_busy_pages(bp, f) +#endif + +/* * Global vfs data structures for nfs */ int (**nfsv2_vnodeop_p)(); @@ -89,8 +105,14 @@ struct vnodeopv_entry_desc nfsv2_vnodeop_entries[] = { { &vop_setattr_desc, nfs_setattr }, /* setattr */ { &vop_read_desc, nfs_read }, /* read */ { &vop_write_desc, nfs_write }, /* write */ +#ifdef HAS_VOPLEASE + { &vop_lease_desc, nfs_lease_check }, /* lease */ +#endif { &vop_ioctl_desc, nfs_ioctl }, /* ioctl */ { &vop_select_desc, nfs_select }, /* select */ +#ifdef HAS_VOPREVOKE + { &vop_revoke_desc, nfs_revoke }, /* revoke */ +#endif { &vop_mmap_desc, nfs_mmap }, /* mmap */ { &vop_fsync_desc, nfs_fsync }, /* fsync */ { &vop_seek_desc, nfs_seek }, /* seek */ @@ -119,12 +141,14 @@ struct vnodeopv_entry_desc nfsv2_vnodeop_entries[] = { { &vop_vfree_desc, nfs_vfree }, /* vfree */ { &vop_truncate_desc, nfs_truncate }, /* truncate */ { &vop_update_desc, nfs_update }, /* update */ - { &vop_bwrite_desc, vn_bwrite }, + { &vop_bwrite_desc, nfs_bwrite }, { (struct vnodeop_desc*)NULL, (int(*)())NULL } }; struct vnodeopv_desc nfsv2_vnodeop_opv_desc = { &nfsv2_vnodeop_p, nfsv2_vnodeop_entries }; +#ifdef __FreeBSD__ VNODEOP_SET(nfsv2_vnodeop_opv_desc); +#endif /* * Special device vnode ops @@ -142,8 +166,14 @@ struct vnodeopv_entry_desc spec_nfsv2nodeop_entries[] = { { &vop_setattr_desc, nfs_setattr }, /* setattr */ { &vop_read_desc, nfsspec_read }, /* read */ { &vop_write_desc, nfsspec_write }, /* write */ +#ifdef HAS_VOPLEASE + { &vop_lease_desc, spec_lease_check }, /* lease */ +#endif { &vop_ioctl_desc, spec_ioctl }, /* ioctl */ { &vop_select_desc, spec_select }, /* select */ +#ifdef HAS_VOPREVOKE + { &vop_revoke_desc, spec_revoke }, /* revoke */ +#endif { &vop_mmap_desc, spec_mmap }, /* mmap */ { &vop_fsync_desc, nfs_fsync }, /* fsync */ { &vop_seek_desc, spec_seek }, /* seek */ @@ -177,7 +207,9 @@ struct vnodeopv_entry_desc spec_nfsv2nodeop_entries[] = { }; struct vnodeopv_desc spec_nfsv2nodeop_opv_desc = { &spec_nfsv2nodeop_p, spec_nfsv2nodeop_entries }; +#ifdef __FreeBSD__ VNODEOP_SET(spec_nfsv2nodeop_opv_desc); +#endif int (**fifo_nfsv2nodeop_p)(); struct vnodeopv_entry_desc fifo_nfsv2nodeop_entries[] = { @@ -192,8 +224,14 @@ struct vnodeopv_entry_desc fifo_nfsv2nodeop_entries[] = { { &vop_setattr_desc, nfs_setattr }, /* setattr */ { &vop_read_desc, nfsfifo_read }, /* read */ { &vop_write_desc, nfsfifo_write }, /* write */ +#ifdef HAS_VOPLEASE + { &vop_lease_desc, fifo_lease_check }, /* lease */ +#endif { &vop_ioctl_desc, fifo_ioctl }, /* ioctl */ { &vop_select_desc, fifo_select }, /* select */ +#ifdef HAS_VOPREVOKE + { &vop_revoke_desc, fifo_revoke }, /* revoke */ +#endif { &vop_mmap_desc, fifo_mmap }, /* mmap */ { &vop_fsync_desc, nfs_fsync }, /* fsync */ { &vop_seek_desc, fifo_seek }, /* seek */ @@ -227,15 +265,21 @@ struct vnodeopv_entry_desc fifo_nfsv2nodeop_entries[] = { }; struct vnodeopv_desc fifo_nfsv2nodeop_opv_desc = { &fifo_nfsv2nodeop_p, fifo_nfsv2nodeop_entries }; +#ifdef __FreeBSD__ VNODEOP_SET(fifo_nfsv2nodeop_opv_desc); +#endif void nqnfs_clientlease(); +int nfs_commit(); +int nfs_removerpc(); +int nfs_renamerpc(); /* * Global variables */ -extern u_long nfs_procids[NFS_NPROCS]; -extern u_long nfs_prog, nfs_vers, nfs_true, nfs_false; +extern u_long nfs_true, nfs_false; +extern struct nfsstats nfsstats; +extern nfstype nfsv3_type[9]; struct proc *nfs_iodwant[NFS_MAXASYNCDAEMON]; int nfs_numasync = 0; #define DIRHDSIZ (sizeof (struct dirent) - (MAXNAMLEN + 1)) @@ -261,9 +305,9 @@ nfs_null(vp, cred, procp) /* * nfs access vnode op. - * For nfs, just return ok. File accesses may fail later. - * For nqnfs, use the access rpc to check accessibility. If file modes are - * changed on the server, accesses might still fail later. + * For nfs version 2, just return ok. File accesses may fail later. + * For nfs version 3, use the access rpc to check accessibility. If file modes + * are changed on the server, accesses might still fail later. */ int nfs_access(ap) @@ -277,36 +321,56 @@ nfs_access(ap) register struct vnode *vp = ap->a_vp; register u_long *tl; register caddr_t cp; - caddr_t bpos, dpos; - int error = 0; + register int t1, t2; + caddr_t bpos, dpos, cp2; + int error = 0, attrflag; struct mbuf *mreq, *mrep, *md, *mb, *mb2; + u_long mode, rmode; + int v3 = NFS_ISV3(vp); /* - * For nqnfs, do an access rpc, otherwise you are stuck emulating + * For nfs v3, do an access rpc, otherwise you are stuck emulating * ufs_access() locally using the vattr. This may not be correct, * since the server may apply other access criteria such as * client uid-->server uid mapping that we do not know about, but * this is better than just returning anything that is lying about * in the cache. */ - if (VFSTONFS(vp->v_mount)->nm_flag & NFSMNT_NQNFS) { - nfsstats.rpccnt[NQNFSPROC_ACCESS]++; - nfsm_reqhead(vp, NQNFSPROC_ACCESS, NFSX_FH + 3 * NFSX_UNSIGNED); - nfsm_fhtom(vp); - nfsm_build(tl, u_long *, 3 * NFSX_UNSIGNED); + if (v3) { + nfsstats.rpccnt[NFSPROC_ACCESS]++; + nfsm_reqhead(vp, NFSPROC_ACCESS, NFSX_FH(v3) + NFSX_UNSIGNED); + nfsm_fhtom(vp, v3); + nfsm_build(tl, u_long *, NFSX_UNSIGNED); if (ap->a_mode & VREAD) - *tl++ = nfs_true; - else - *tl++ = nfs_false; - if (ap->a_mode & VWRITE) - *tl++ = nfs_true; - else - *tl++ = nfs_false; - if (ap->a_mode & VEXEC) - *tl = nfs_true; + mode = NFSV3ACCESS_READ; else - *tl = nfs_false; - nfsm_request(vp, NQNFSPROC_ACCESS, ap->a_p, ap->a_cred); + mode = 0; + if (vp->v_type == VDIR) { + if (ap->a_mode & VWRITE) + mode |= (NFSV3ACCESS_MODIFY | NFSV3ACCESS_EXTEND | + NFSV3ACCESS_DELETE); + if (ap->a_mode & VEXEC) + mode |= NFSV3ACCESS_LOOKUP; + } else { + if (ap->a_mode & VWRITE) + mode |= (NFSV3ACCESS_MODIFY | NFSV3ACCESS_EXTEND); + if (ap->a_mode & VEXEC) + mode |= NFSV3ACCESS_EXECUTE; + } + *tl = txdr_unsigned(mode); + nfsm_request(vp, NFSPROC_ACCESS, ap->a_p, ap->a_cred); + nfsm_postop_attr(vp, attrflag); + if (!error) { + nfsm_dissect(tl, u_long *, NFSX_UNSIGNED); + rmode = fxdr_unsigned(u_long, *tl); + /* + * The NFS V3 spec does not clarify whether or not + * the returned access bits can be a superset of + * the ones requested, so... + */ + if ((rmode & mode) != mode) + error = EACCES; + } nfsm_reqdone; return (error); } else @@ -337,14 +401,17 @@ nfs_open(ap) int error; if (vp->v_type != VREG && vp->v_type != VDIR && vp->v_type != VLNK) +{ printf("open eacces vtyp=%d\n",vp->v_type); return (EACCES); +} /* * Get a valid lease. If cached data is stale, flush it. */ if (nmp->nm_flag & NFSMNT_NQNFS) { - if (NQNFS_CKINVALID(vp, np, NQL_READ)) { + if (NQNFS_CKINVALID(vp, np, ND_READ)) { do { - error = nqnfs_getlease(vp, NQL_READ, ap->a_cred, ap->a_p); + error = nqnfs_getlease(vp, ND_READ, ap->a_cred, + ap->a_p); } while (error == NQNFS_EXPIRED); if (error) return (error); @@ -353,6 +420,7 @@ nfs_open(ap) if ((error = nfs_vinvalbuf(vp, V_SAVE, ap->a_cred, ap->a_p, 1)) == EINTR) return (error); + (void) vnode_pager_uncache(vp); np->n_brev = np->n_lrev; } } @@ -361,8 +429,10 @@ nfs_open(ap) if ((error = nfs_vinvalbuf(vp, V_SAVE, ap->a_cred, ap->a_p, 1)) == EINTR) return (error); + /* (void) vnode_pager_uncache(vp); */ np->n_attrstamp = 0; - np->n_direofoffset = 0; + if (vp->v_type == VDIR) + np->n_direofoffset = 0; error = VOP_GETATTR(vp, &vattr, ap->a_cred, ap->a_p); if (error) return (error); @@ -372,10 +442,12 @@ nfs_open(ap) if (error) return (error); if (np->n_mtime != vattr.va_mtime.ts_sec) { - np->n_direofoffset = 0; + if (vp->v_type == VDIR) + np->n_direofoffset = 0; if ((error = nfs_vinvalbuf(vp, V_SAVE, ap->a_cred, ap->a_p, 1)) == EINTR) return (error); + /* (void) vnode_pager_uncache(vp); */ np->n_mtime = vattr.va_mtime.ts_sec; } } @@ -387,7 +459,33 @@ nfs_open(ap) /* * nfs close vnode op - * For reg files, invalidate any buffer cache entries. + * What an NFS client should do upon close after writing is a debatable issue. + * Most NFS clients push delayed writes to the server upon close, basically for + * two reasons: + * 1 - So that any write errors may be reported back to the client process + * doing the close system call. By far the two most likely errors are + * NFSERR_NOSPC and NFSERR_DQUOT to indicate space allocation failure. + * 2 - To put a worst case upper bound on cache inconsistency between + * multiple clients for the file. + * There is also a consistency problem for Version 2 of the protocol w.r.t. + * not being able to tell if other clients are writing a file concurrently, + * since there is no way of knowing if the changed modify time in the reply + * is only due to the write for this client. + * (NFS Version 3 provides weak cache consistency data in the reply that + * should be sufficient to detect and handle this case.) + * + * The current code does the following: + * for NFS Version 2 - play it safe and flush/invalidate all dirty buffers + * for NFS Version 3 - flush dirty buffers to the server but don't invalidate + * or commit them (this satisfies 1 and 2 except for the + * case where the server crashes after this close but + * before the commit RPC, which is felt to be "good + * enough". Changing the last argument to nfs_flush() to + * a 1 would force a commit operation, if it is felt a + * commit is necessary now. + * for NQNFS - do nothing now, since 2 is dealt with via leases and + * 1 should be dealt with via an fsync() system call for + * cases where write errors are important. */ /* ARGSUSED */ int @@ -407,7 +505,10 @@ nfs_close(ap) if (vp->v_type == VREG) { if ((VFSTONFS(vp->v_mount)->nm_flag & NFSMNT_NQNFS) == 0 && (np->n_flag & NMODIFIED)) { - error = nfs_vinvalbuf(vp, V_SAVE, ap->a_cred, ap->a_p, 1); + if (NFS_ISV3(vp)) + error = nfs_flush(vp, ap->a_cred, MNT_WAIT, ap->a_p, 0); + else + error = nfs_vinvalbuf(vp, V_SAVE, ap->a_cred, ap->a_p, 1); np->n_attrstamp = 0; } if (np->n_flag & NWRITEERR) { @@ -433,9 +534,12 @@ nfs_getattr(ap) register struct vnode *vp = ap->a_vp; register struct nfsnode *np = VTONFS(vp); register caddr_t cp; + register u_long *tl; + register int t1, t2; caddr_t bpos, dpos; int error = 0; struct mbuf *mreq, *mrep, *md, *mb, *mb2; + int v3 = NFS_ISV3(vp); /* * Update local times for special files. @@ -448,10 +552,11 @@ nfs_getattr(ap) if (nfs_getattrcache(vp, ap->a_vap) == 0) return (0); nfsstats.rpccnt[NFSPROC_GETATTR]++; - nfsm_reqhead(vp, NFSPROC_GETATTR, NFSX_FH); - nfsm_fhtom(vp); + nfsm_reqhead(vp, NFSPROC_GETATTR, NFSX_FH(v3)); + nfsm_fhtom(vp, v3); nfsm_request(vp, NFSPROC_GETATTR, ap->a_p, ap->a_cred); - nfsm_loadattr(vp, ap->a_vap); + if (!error) + nfsm_loadattr(vp, ap->a_vap); nfsm_reqdone; return (error); } @@ -469,92 +574,166 @@ nfs_setattr(ap) struct proc *a_p; } */ *ap; { - register struct nfsv2_sattr *sp; - register caddr_t cp; - register long t1; - caddr_t bpos, dpos, cp2; - u_long *tl; - int error = 0, isnq; - struct mbuf *mreq, *mrep, *md, *mb, *mb2; register struct vnode *vp = ap->a_vp; register struct nfsnode *np = VTONFS(vp); register struct vattr *vap = ap->a_vap; - u_quad_t frev, tsize = 0; + int error = 0; + u_quad_t tsize; +#ifndef nolint + tsize = (u_quad_t)0; +#endif if (vap->va_size != VNOVAL) { - switch (vp->v_type) { - case VDIR: - return (EISDIR); - case VCHR: - case VBLK: + switch (vp->v_type) { + case VDIR: + return (EISDIR); + case VCHR: + case VBLK: if (vap->va_mtime.ts_sec == VNOVAL && vap->va_atime.ts_sec == VNOVAL && vap->va_mode == (u_short)VNOVAL && - vap->va_uid == VNOVAL && - vap->va_gid == VNOVAL) + vap->va_uid == (uid_t)VNOVAL && + vap->va_gid == (gid_t)VNOVAL) return (0); - vap->va_size = VNOVAL; - break; - default: - if (np->n_flag & NMODIFIED) { - error = nfs_vinvalbuf(vp, - vap->va_size ? V_SAVE : 0, - ap->a_cred, ap->a_p, 1); - if (error) - return (error); - } - tsize = np->n_size; - np->n_size = np->n_vattr.va_size = vap->va_size; - vnode_pager_setsize(vp, (u_long)np->n_size); - } - } else if ((vap->va_mtime.ts_sec != VNOVAL || - vap->va_atime.ts_sec != VNOVAL) && (np->n_flag & NMODIFIED)) { - error = nfs_vinvalbuf(vp, V_SAVE, ap->a_cred, ap->a_p, 1); - if (error == EINTR) - return (error); + vap->va_size = VNOVAL; + break; + default: + if (np->n_flag & NMODIFIED) { + if (vap->va_size == 0) + error = nfs_vinvalbuf(vp, 0, + ap->a_cred, ap->a_p, 1); + else + error = nfs_vinvalbuf(vp, V_SAVE, + ap->a_cred, ap->a_p, 1); + if (error) + return (error); + } + tsize = np->n_size; + np->n_size = np->n_vattr.va_size = vap->va_size; + vnode_pager_setsize(vp, (u_long)np->n_size); + }; + } else if ((vap->va_mtime.ts_sec != VNOVAL || + vap->va_atime.ts_sec != VNOVAL) && (np->n_flag & NMODIFIED) && + vp->v_type == VREG && + (error = nfs_vinvalbuf(vp, V_SAVE, ap->a_cred, + ap->a_p, 1)) == EINTR) + return (error); + error = nfs_setattrrpc(vp, vap, ap->a_cred, ap->a_p); + if (error) { + np->n_size = np->n_vattr.va_size = tsize; + vnode_pager_setsize(vp, (u_long)np->n_size); } + return (error); +} + +/* + * Do an nfs setattr rpc. + */ +int +nfs_setattrrpc(vp, vap, cred, procp) + register struct vnode *vp; + register struct vattr *vap; + struct ucred *cred; + struct proc *procp; +{ + register struct nfsv2_sattr *sp; + register caddr_t cp; + register long t1, t2; + caddr_t bpos, dpos, cp2; + u_long *tl; + int error = 0, wccflag = NFSV3_WCCRATTR; + struct mbuf *mreq, *mrep, *md, *mb, *mb2; + u_quad_t frev; + int v3 = NFS_ISV3(vp); + nfsstats.rpccnt[NFSPROC_SETATTR]++; - isnq = (VFSTONFS(vp->v_mount)->nm_flag & NFSMNT_NQNFS); - nfsm_reqhead(vp, NFSPROC_SETATTR, NFSX_FH+NFSX_SATTR(isnq)); - nfsm_fhtom(vp); - nfsm_build(sp, struct nfsv2_sattr *, NFSX_SATTR(isnq)); - if (vap->va_mode == (u_short)-1) - sp->sa_mode = VNOVAL; - else - sp->sa_mode = vtonfs_mode(vp->v_type, vap->va_mode); - if (vap->va_uid == (uid_t)-1) - sp->sa_uid = VNOVAL; - else - sp->sa_uid = txdr_unsigned(vap->va_uid); - if (vap->va_gid == (gid_t)-1) - sp->sa_gid = VNOVAL; - else - sp->sa_gid = txdr_unsigned(vap->va_gid); - if (isnq) { - txdr_hyper(&vap->va_size, &sp->sa_nqsize); - txdr_nqtime(&vap->va_atime, &sp->sa_nqatime); - txdr_nqtime(&vap->va_mtime, &sp->sa_nqmtime); - sp->sa_nqflags = txdr_unsigned(vap->va_flags); - sp->sa_nqrdev = VNOVAL; + nfsm_reqhead(vp, NFSPROC_SETATTR, NFSX_FH(v3) + NFSX_SATTR(v3)); + nfsm_fhtom(vp, v3); + if (v3) { + if (vap->va_mode != (u_short)VNOVAL) { + nfsm_build(tl, u_long *, 2 * NFSX_UNSIGNED); + *tl++ = nfs_true; + *tl = txdr_unsigned(vap->va_mode); + } else { + nfsm_build(tl, u_long *, NFSX_UNSIGNED); + *tl = nfs_false; + } + if (vap->va_uid != (uid_t)VNOVAL) { + nfsm_build(tl, u_long *, 2 * NFSX_UNSIGNED); + *tl++ = nfs_true; + *tl = txdr_unsigned(vap->va_uid); + } else { + nfsm_build(tl, u_long *, NFSX_UNSIGNED); + *tl = nfs_false; + } + if (vap->va_gid != (gid_t)VNOVAL) { + nfsm_build(tl, u_long *, 2 * NFSX_UNSIGNED); + *tl++ = nfs_true; + *tl = txdr_unsigned(vap->va_gid); + } else { + nfsm_build(tl, u_long *, NFSX_UNSIGNED); + *tl = nfs_false; + } + if (vap->va_size != VNOVAL) { + nfsm_build(tl, u_long *, 3 * NFSX_UNSIGNED); + *tl++ = nfs_true; + txdr_hyper(&vap->va_size, tl); + } else { + nfsm_build(tl, u_long *, NFSX_UNSIGNED); + *tl = nfs_false; + } + if (vap->va_atime.ts_sec != VNOVAL) { + if (vap->va_atime.ts_sec != time.tv_sec) { + nfsm_build(tl, u_long *, 3 * NFSX_UNSIGNED); + *tl++ = txdr_unsigned(NFSV3SATTRTIME_TOCLIENT); + txdr_nfsv3time(&vap->va_atime, tl); + } else { + nfsm_build(tl, u_long *, NFSX_UNSIGNED); + *tl = txdr_unsigned(NFSV3SATTRTIME_TOSERVER); + } + } else { + nfsm_build(tl, u_long *, NFSX_UNSIGNED); + *tl = txdr_unsigned(NFSV3SATTRTIME_DONTCHANGE); + } + if (vap->va_mtime.ts_sec != VNOVAL) { + if (vap->va_mtime.ts_sec != time.tv_sec) { + nfsm_build(tl, u_long *, 3 * NFSX_UNSIGNED); + *tl++ = txdr_unsigned(NFSV3SATTRTIME_TOCLIENT); + txdr_nfsv3time(&vap->va_atime, tl); + } else { + nfsm_build(tl, u_long *, NFSX_UNSIGNED); + *tl = txdr_unsigned(NFSV3SATTRTIME_TOSERVER); + } + } else { + nfsm_build(tl, u_long *, NFSX_UNSIGNED); + *tl = txdr_unsigned(NFSV3SATTRTIME_DONTCHANGE); + } + nfsm_build(tl, u_long *, NFSX_UNSIGNED); + *tl = nfs_false; } else { - sp->sa_nfssize = txdr_unsigned(vap->va_size); - txdr_nfstime(&vap->va_atime, &sp->sa_nfsatime); - txdr_nfstime(&vap->va_mtime, &sp->sa_nfsmtime); - } - nfsm_request(vp, NFSPROC_SETATTR, ap->a_p, ap->a_cred); - nfsm_loadattr(vp, (struct vattr *)0); - if ((VFSTONFS(vp->v_mount)->nm_flag & NFSMNT_NQNFS) && - NQNFS_CKCACHABLE(vp, NQL_WRITE)) { - nfsm_dissect(tl, u_long *, 2*NFSX_UNSIGNED); - fxdr_hyper(tl, &frev); - if (frev > np->n_brev) - np->n_brev = frev; + nfsm_build(sp, struct nfsv2_sattr *, NFSX_V2SATTR); + if (vap->va_mode == (u_short)VNOVAL) + sp->sa_mode = VNOVAL; + else + sp->sa_mode = vtonfsv2_mode(vp->v_type, vap->va_mode); + if (vap->va_uid == (uid_t)VNOVAL) + sp->sa_uid = VNOVAL; + else + sp->sa_uid = txdr_unsigned(vap->va_uid); + if (vap->va_gid == (gid_t)VNOVAL) + sp->sa_gid = VNOVAL; + else + sp->sa_gid = txdr_unsigned(vap->va_gid); + sp->sa_size = txdr_unsigned(vap->va_size); + txdr_nfsv2time(&vap->va_atime, &sp->sa_atime); + txdr_nfsv2time(&vap->va_mtime, &sp->sa_mtime); } + nfsm_request(vp, NFSPROC_SETATTR, procp, cred); + if (v3) { + nfsm_wcc_data(vp, wccflag); + } else + nfsm_loadattr(vp, (struct vattr *)0); nfsm_reqdone; - if (error) { - np->n_size = np->n_vattr.va_size = tsize; - vnode_pager_setsize(vp, (u_long)np->n_size); - } return (error); } @@ -576,23 +755,20 @@ nfs_lookup(ap) register struct vnode *dvp = ap->a_dvp; register struct vnode **vpp = ap->a_vpp; register int flags = cnp->cn_flags; - register struct vnode *vdp; + register struct vnode *newvp; register u_long *tl; register caddr_t cp; register long t1, t2; struct nfsmount *nmp; caddr_t bpos, dpos, cp2; - time_t reqtime = 0; struct mbuf *mreq, *mrep, *md, *mb, *mb2; - struct vnode *newvp; long len; - nfsv2fh_t *fhp; + nfsfh_t *fhp; struct nfsnode *np; - int lockparent, wantparent, error = 0; - int nqlflag = 0, cachable = 0; - u_quad_t frev; + int lockparent, wantparent, error = 0, attrflag, fhsize; + int v3 = NFS_ISV3(dvp); - *vpp = NULL; + *vpp = NULLVP; if (dvp->v_type != VDIR) return (ENOTDIR); lockparent = flags & LOCKPARENT; @@ -603,161 +779,141 @@ nfs_lookup(ap) struct vattr vattr; int vpid; - vdp = *vpp; - vpid = vdp->v_id; + newvp = *vpp; + vpid = newvp->v_id; /* * See the comment starting `Step through' in ufs/ufs_lookup.c * for an explanation of the locking protocol */ - if (dvp == vdp) { - VREF(vdp); + if (dvp == newvp) { + VREF(newvp); error = 0; - } else - error = vget(vdp, 1); + } else if (flags & ISDOTDOT) { + VOP_UNLOCK(dvp); + error = vget(newvp, 1); + if (!error && lockparent && (flags & ISLASTCN)) + error = VOP_LOCK(dvp); + } else { + error = vget(newvp, 1); + if (!lockparent || error || !(flags & ISLASTCN)) + VOP_UNLOCK(dvp); + } if (!error) { - if (vpid == vdp->v_id) { - if (nmp->nm_flag & NFSMNT_NQNFS) { - if ((nmp->nm_flag & NFSMNT_NQLOOKLEASE) == 0) { - nfsstats.lookupcache_hits++; - if (cnp->cn_nameiop != LOOKUP && - (flags & ISLASTCN)) - cnp->cn_flags |= SAVENAME; - return (0); - } else if (NQNFS_CKCACHABLE(dvp, NQL_READ)) { - if (np->n_lrev != np->n_brev || - (np->n_flag & NMODIFIED)) { - np->n_direofoffset = 0; - cache_purge(dvp); - error = nfs_vinvalbuf(dvp, 0, - cnp->cn_cred, cnp->cn_proc, - 1); - if (error == EINTR) - return (error); - np->n_brev = np->n_lrev; - } else { - nfsstats.lookupcache_hits++; - if (cnp->cn_nameiop != LOOKUP && - (flags & ISLASTCN)) - cnp->cn_flags |= SAVENAME; - return (0); - } - } - } else if (!VOP_GETATTR(vdp, &vattr, cnp->cn_cred, cnp->cn_proc) && - vattr.va_ctime.ts_sec == VTONFS(vdp)->n_ctime) { + if (vpid == newvp->v_id) { + if (!VOP_GETATTR(newvp, &vattr, cnp->cn_cred, cnp->cn_proc) + && vattr.va_ctime.ts_sec == VTONFS(newvp)->n_ctime) { nfsstats.lookupcache_hits++; if (cnp->cn_nameiop != LOOKUP && (flags & ISLASTCN)) cnp->cn_flags |= SAVENAME; return (0); } - cache_purge(vdp); + cache_purge(newvp); } - vrele(vdp); + vput(newvp); + if (lockparent && dvp != newvp && (flags & ISLASTCN)) + VOP_UNLOCK(dvp); } + error = VOP_LOCK(dvp); + if (error) + return (error); *vpp = NULLVP; } error = 0; + newvp = NULLVP; nfsstats.lookupcache_misses++; nfsstats.rpccnt[NFSPROC_LOOKUP]++; len = cnp->cn_namelen; - nfsm_reqhead(dvp, NFSPROC_LOOKUP, NFSX_FH+NFSX_UNSIGNED+nfsm_rndup(len)); - - /* - * For nqnfs optionally piggyback a getlease request for the name - * being looked up. - */ - if (nmp->nm_flag & NFSMNT_NQNFS) { - nfsm_build(tl, u_long *, NFSX_UNSIGNED); - if ((nmp->nm_flag & NFSMNT_NQLOOKLEASE) && - ((cnp->cn_flags & MAKEENTRY) && - (cnp->cn_nameiop != DELETE || !(flags & ISLASTCN)))) - *tl = txdr_unsigned(nmp->nm_leaseterm); - else - *tl = 0; - } - nfsm_fhtom(dvp); + nfsm_reqhead(dvp, NFSPROC_LOOKUP, + NFSX_FH(v3) + NFSX_UNSIGNED + nfsm_rndup(len)); + nfsm_fhtom(dvp, v3); nfsm_strtom(cnp->cn_nameptr, len, NFS_MAXNAMLEN); - reqtime = time.tv_sec; nfsm_request(dvp, NFSPROC_LOOKUP, cnp->cn_proc, cnp->cn_cred); -nfsmout: if (error) { - if ((cnp->cn_nameiop == CREATE || cnp->cn_nameiop == RENAME) && - (flags & ISLASTCN) && error == ENOENT) - error = EJUSTRETURN; - if (cnp->cn_nameiop != LOOKUP && (flags & ISLASTCN)) - cnp->cn_flags |= SAVENAME; - return (error); - } - if (nmp->nm_flag & NFSMNT_NQNFS) { - nfsm_dissect(tl, u_long *, NFSX_UNSIGNED); - if (*tl) { - nqlflag = fxdr_unsigned(int, *tl); - nfsm_dissect(tl, u_long *, 4*NFSX_UNSIGNED); - cachable = fxdr_unsigned(int, *tl++); - reqtime += fxdr_unsigned(int, *tl++); - fxdr_hyper(tl, &frev); - } else - nqlflag = 0; + nfsm_postop_attr(dvp, attrflag); + m_freem(mrep); + goto nfsmout; } - nfsm_dissect(fhp, nfsv2fh_t *, NFSX_FH); + nfsm_getfh(fhp, fhsize, v3); /* * Handle RENAME case... */ if (cnp->cn_nameiop == RENAME && wantparent && (flags & ISLASTCN)) { - if (!bcmp(np->n_fh.fh_bytes, (caddr_t)fhp, NFSX_FH)) { + if (NFS_CMPFH(np, fhp, fhsize)) { m_freem(mrep); return (EISDIR); } - error = nfs_nget(dvp->v_mount, fhp, &np); - if (error) { + if (error = nfs_nget(dvp->v_mount, fhp, fhsize, &np)) { m_freem(mrep); return (error); } newvp = NFSTOV(np); - error = nfs_loadattrcache(&newvp, &md, &dpos, (struct vattr*)0); - if (error) { - vrele(newvp); - m_freem(mrep); - return (error); - } + if (v3) { + nfsm_postop_attr(newvp, attrflag); + nfsm_postop_attr(dvp, attrflag); + } else + nfsm_loadattr(newvp, (struct vattr *)0); *vpp = newvp; m_freem(mrep); cnp->cn_flags |= SAVENAME; + if (!lockparent) + VOP_UNLOCK(dvp); return (0); } - if (!bcmp(np->n_fh.fh_bytes, (caddr_t)fhp, NFSX_FH)) { + if (flags & ISDOTDOT) { + VOP_UNLOCK(dvp); + error = nfs_nget(dvp->v_mount, fhp, fhsize, &np); + if (error) { + VOP_LOCK(dvp); + return (error); + } + newvp = NFSTOV(np); + if (lockparent && (flags & ISLASTCN) && + (error = VOP_LOCK(dvp))) { + vput(newvp); + return (error); + } + } else if (NFS_CMPFH(np, fhp, fhsize)) { VREF(dvp); newvp = dvp; } else { - error = nfs_nget(dvp->v_mount, fhp, &np); - if (error) { + if (error = nfs_nget(dvp->v_mount, fhp, fhsize, &np)) { m_freem(mrep); return (error); } + if (!lockparent || !(flags & ISLASTCN)) + VOP_UNLOCK(dvp); newvp = NFSTOV(np); } - error = nfs_loadattrcache(&newvp, &md, &dpos, (struct vattr *)0); - if (error) { - vrele(newvp); - m_freem(mrep); - return (error); - } - m_freem(mrep); - *vpp = newvp; + if (v3) { + nfsm_postop_attr(newvp, attrflag); + nfsm_postop_attr(dvp, attrflag); + } else + nfsm_loadattr(newvp, (struct vattr *)0); if (cnp->cn_nameiop != LOOKUP && (flags & ISLASTCN)) cnp->cn_flags |= SAVENAME; if ((cnp->cn_flags & MAKEENTRY) && (cnp->cn_nameiop != DELETE || !(flags & ISLASTCN))) { - if ((nmp->nm_flag & NFSMNT_NQNFS) == 0) - np->n_ctime = np->n_vattr.va_ctime.ts_sec; - else if (nqlflag && reqtime > time.tv_sec) - nqnfs_clientlease(nmp, np, nqlflag, cachable, reqtime, - frev); - cache_enter(dvp, *vpp, cnp); + np->n_ctime = np->n_vattr.va_ctime.ts_sec; + cache_enter(dvp, newvp, cnp); } - return (0); + *vpp = newvp; + nfsm_reqdone; + if (error) { + if (newvp != NULLVP) + vrele(newvp); + if ((cnp->cn_nameiop == CREATE || cnp->cn_nameiop == RENAME) && + (flags & ISLASTCN) && error == ENOENT) { + if (!lockparent) + VOP_UNLOCK(dvp); + error = EJUSTRETURN; + } + if (cnp->cn_nameiop != LOOKUP && (flags & ISLASTCN)) + cnp->cn_flags |= SAVENAME; + } + return (error); } /* @@ -810,18 +966,22 @@ nfs_readlinkrpc(vp, uiop, cred) { register u_long *tl; register caddr_t cp; - register long t1; + register long t1, t2; caddr_t bpos, dpos, cp2; - int error = 0; + int error = 0, len, attrflag; struct mbuf *mreq, *mrep, *md, *mb, *mb2; - long len; + int v3 = NFS_ISV3(vp); nfsstats.rpccnt[NFSPROC_READLINK]++; - nfsm_reqhead(vp, NFSPROC_READLINK, NFSX_FH); - nfsm_fhtom(vp); + nfsm_reqhead(vp, NFSPROC_READLINK, NFSX_FH(v3)); + nfsm_fhtom(vp, v3); nfsm_request(vp, NFSPROC_READLINK, uiop->uio_procp, cred); - nfsm_strsiz(len, NFS_MAXPATHLEN); - nfsm_mtouio(uiop, len); + if (v3) + nfsm_postop_attr(vp, attrflag); + if (!error) { + nfsm_strsiz(len, NFS_MAXPATHLEN); + nfsm_mtouio(uiop, len); + } nfsm_reqdone; return (error); } @@ -838,25 +998,27 @@ nfs_readrpc(vp, uiop, cred) { register u_long *tl; register caddr_t cp; - register long t1; + register long t1, t2; caddr_t bpos, dpos, cp2; - int error = 0; struct mbuf *mreq, *mrep, *md, *mb, *mb2; struct nfsmount *nmp; - long len, retlen, tsiz; + int error = 0, len, retlen, tsiz, eof, attrflag; + int v3 = NFS_ISV3(vp); +#ifndef nolint + eof = 0; +#endif nmp = VFSTONFS(vp->v_mount); tsiz = uiop->uio_resid; - if (uiop->uio_offset + tsiz > 0xffffffff && - (nmp->nm_flag & NFSMNT_NQNFS) == 0) + if (uiop->uio_offset + tsiz > 0xffffffff && !v3) return (EFBIG); while (tsiz > 0) { nfsstats.rpccnt[NFSPROC_READ]++; len = (tsiz > nmp->nm_rsize) ? nmp->nm_rsize : tsiz; - nfsm_reqhead(vp, NFSPROC_READ, NFSX_FH+NFSX_UNSIGNED*3); - nfsm_fhtom(vp); - nfsm_build(tl, u_long *, NFSX_UNSIGNED*3); - if (nmp->nm_flag & NFSMNT_NQNFS) { + nfsm_reqhead(vp, NFSPROC_READ, NFSX_FH(v3) + NFSX_UNSIGNED * 3); + nfsm_fhtom(vp, v3); + nfsm_build(tl, u_long *, NFSX_UNSIGNED * 3); + if (v3) { txdr_hyper(&uiop->uio_offset, tl); *(tl + 2) = txdr_unsigned(len); } else { @@ -865,14 +1027,25 @@ nfs_readrpc(vp, uiop, cred) *tl = 0; } nfsm_request(vp, NFSPROC_READ, uiop->uio_procp, cred); - nfsm_loadattr(vp, (struct vattr *)0); + if (v3) { + nfsm_postop_attr(vp, attrflag); + if (error) { + m_freem(mrep); + goto nfsmout; + } + nfsm_dissect(tl, u_long *, 2 * NFSX_UNSIGNED); + eof = fxdr_unsigned(int, *(tl + 1)); + } else + nfsm_loadattr(vp, (struct vattr *)0); nfsm_strsiz(retlen, nmp->nm_rsize); nfsm_mtouio(uiop, retlen); m_freem(mrep); - if (retlen < len) + tsiz -= retlen; + if (v3) { + if (eof || retlen == 0) + tsiz = 0; + } else if (retlen < len) tsiz = 0; - else - tsiz -= len; } nfsmout: return (error); @@ -882,150 +1055,227 @@ nfsmout: * nfs write call */ int -nfs_writerpc(vp, uiop, cred, ioflags) +nfs_writerpc(vp, uiop, cred, iomode, must_commit) register struct vnode *vp; - struct uio *uiop; + register struct uio *uiop; struct ucred *cred; - int ioflags; + int *iomode, *must_commit; { register u_long *tl; register caddr_t cp; - register long t1; + register int t1, t2, backup; caddr_t bpos, dpos, cp2; - int error = 0; struct mbuf *mreq, *mrep, *md, *mb, *mb2; - struct nfsmount *nmp; + struct nfsmount *nmp = VFSTONFS(vp->v_mount); struct nfsnode *np = VTONFS(vp); u_quad_t frev; - long len, tsiz; + int error = 0, len, tsiz, wccflag = NFSV3_WCCRATTR, rlen, commit; + int v3 = NFS_ISV3(vp), committed = NFSV3WRITE_FILESYNC; - nmp = VFSTONFS(vp->v_mount); +#ifndef DIAGNOSTIC + if (uiop->uio_iovcnt != 1) + panic("nfs: writerpc iovcnt > 1"); +#endif + *must_commit = 0; tsiz = uiop->uio_resid; - if (uiop->uio_offset + tsiz > 0xffffffff && - (nmp->nm_flag & NFSMNT_NQNFS) == 0) + if (uiop->uio_offset + tsiz > 0xffffffff && !v3) return (EFBIG); while (tsiz > 0) { nfsstats.rpccnt[NFSPROC_WRITE]++; len = (tsiz > nmp->nm_wsize) ? nmp->nm_wsize : tsiz; nfsm_reqhead(vp, NFSPROC_WRITE, - NFSX_FH+NFSX_UNSIGNED*4+nfsm_rndup(len)); - nfsm_fhtom(vp); - nfsm_build(tl, u_long *, NFSX_UNSIGNED * 4); - if (nmp->nm_flag & NFSMNT_NQNFS) { + NFSX_FH(v3) + 5 * NFSX_UNSIGNED + nfsm_rndup(len)); + nfsm_fhtom(vp, v3); + if (v3) { + nfsm_build(tl, u_long *, 5 * NFSX_UNSIGNED); txdr_hyper(&uiop->uio_offset, tl); tl += 2; -#ifdef notyet - if (ioflags & IO_APPEND) - *tl++ = txdr_unsigned(1); - else -#endif - *tl++ = 0; + *tl++ = txdr_unsigned(len); + *tl++ = txdr_unsigned(*iomode); } else { + nfsm_build(tl, u_long *, 4 * NFSX_UNSIGNED); *++tl = txdr_unsigned(uiop->uio_offset); tl += 2; } *tl = txdr_unsigned(len); nfsm_uiotom(uiop, len); nfsm_request(vp, NFSPROC_WRITE, uiop->uio_procp, cred); - nfsm_loadattr(vp, (struct vattr *)0); - if (nmp->nm_flag & NFSMNT_MYWRITE) - VTONFS(vp)->n_mtime = VTONFS(vp)->n_vattr.va_mtime.ts_sec; - else if ((nmp->nm_flag & NFSMNT_NQNFS) && - NQNFS_CKCACHABLE(vp, NQL_WRITE)) { - nfsm_dissect(tl, u_long *, 2*NFSX_UNSIGNED); - fxdr_hyper(tl, &frev); - if (frev > np->n_brev) - np->n_brev = frev; - } + if (v3) { + wccflag = NFSV3_WCCCHK; + nfsm_wcc_data(vp, wccflag); + if (!error) { + nfsm_dissect(tl, u_long *, 2 * NFSX_UNSIGNED + + NFSX_V3WRITEVERF); + rlen = fxdr_unsigned(int, *tl++); + if (rlen == 0) { + error = NFSERR_IO; + break; + } else if (rlen < len) { + backup = len - rlen; + uiop->uio_iov->iov_base -= backup; + uiop->uio_iov->iov_len += backup; + uiop->uio_offset -= backup; + uiop->uio_resid += backup; + len = rlen; + } + commit = fxdr_unsigned(int, *tl++); + + /* + * Return the lowest committment level + * obtained by any of the RPCs. + */ + if (committed == NFSV3WRITE_FILESYNC) + committed = commit; + else if (committed == NFSV3WRITE_DATASYNC && + commit == NFSV3WRITE_UNSTABLE) + committed = commit; + if ((nmp->nm_flag & NFSMNT_HASWRITEVERF) == 0) { + bcopy((caddr_t)tl, (caddr_t)nmp->nm_verf, + NFSX_V3WRITEVERF); + nmp->nm_flag |= NFSMNT_HASWRITEVERF; + } else if (bcmp((caddr_t)tl, + (caddr_t)nmp->nm_verf, NFSX_V3WRITEVERF)) { + *must_commit = 1; + bcopy((caddr_t)tl, (caddr_t)nmp->nm_verf, + NFSX_V3WRITEVERF); + } + } + } else + nfsm_loadattr(vp, (struct vattr *)0); + if (wccflag) + VTONFS(vp)->n_mtime = VTONFS(vp)->n_vattr.va_mtime.ts_sec; m_freem(mrep); tsiz -= len; } nfsmout: + *iomode = committed; if (error) uiop->uio_resid = tsiz; return (error); } /* - * nfs mknod call - * This is a kludge. Use a create rpc but with the IFMT bits of the mode - * set to specify the file type and the size field for rdev. + * nfs mknod rpc + * For NFS v2 this is a kludge. Use a create rpc but with the IFMT bits of the + * mode set to specify the file type and the size field for rdev. */ -/* ARGSUSED */ int -nfs_mknod(ap) - struct vop_mknod_args /* { - struct vnode *a_dvp; - struct vnode **a_vpp; - struct componentname *a_cnp; - struct vattr *a_vap; - } */ *ap; +nfs_mknodrpc(dvp, vpp, cnp, vap) + register struct vnode *dvp; + register struct vnode **vpp; + register struct componentname *cnp; + register struct vattr *vap; { - register struct vnode *dvp = ap->a_dvp; - register struct vattr *vap = ap->a_vap; - register struct componentname *cnp = ap->a_cnp; register struct nfsv2_sattr *sp; + register struct nfsv3_sattr *sp3; register u_long *tl; register caddr_t cp; register long t1, t2; - struct vnode *newvp = 0; + struct vnode *newvp = (struct vnode *)0; + struct nfsnode *np = (struct nfsnode *)0; struct vattr vattr; char *cp2; caddr_t bpos, dpos; - int error = 0, isnq; + int error = 0, wccflag = NFSV3_WCCRATTR, gotvp = 0; struct mbuf *mreq, *mrep, *md, *mb, *mb2; u_long rdev; + int v3 = NFS_ISV3(dvp); - isnq = (VFSTONFS(dvp->v_mount)->nm_flag & NFSMNT_NQNFS); if (vap->va_type == VCHR || vap->va_type == VBLK) rdev = txdr_unsigned(vap->va_rdev); - else if (vap->va_type == VFIFO) + else if (vap->va_type == VFIFO || vap->va_type == VSOCK) rdev = 0xffffffff; else { VOP_ABORTOP(dvp, cnp); vput(dvp); return (EOPNOTSUPP); } - error = VOP_GETATTR(dvp, &vattr, cnp->cn_cred, cnp->cn_proc); - if (error) { + if (error = VOP_GETATTR(dvp, &vattr, cnp->cn_cred, cnp->cn_proc)) { VOP_ABORTOP(dvp, cnp); vput(dvp); return (error); } - newvp = NULLVP; - nfsstats.rpccnt[NFSPROC_CREATE]++; - nfsm_reqhead(dvp, NFSPROC_CREATE, - NFSX_FH+NFSX_UNSIGNED+nfsm_rndup(cnp->cn_namelen)+NFSX_SATTR(isnq)); - nfsm_fhtom(dvp); + nfsstats.rpccnt[NFSPROC_MKNOD]++; + nfsm_reqhead(dvp, NFSPROC_MKNOD, NFSX_FH(v3) + 4 * NFSX_UNSIGNED + + + nfsm_rndup(cnp->cn_namelen) + NFSX_SATTR(v3)); + nfsm_fhtom(dvp, v3); nfsm_strtom(cnp->cn_nameptr, cnp->cn_namelen, NFS_MAXNAMLEN); - nfsm_build(sp, struct nfsv2_sattr *, NFSX_SATTR(isnq)); - sp->sa_mode = vtonfs_mode(vap->va_type, vap->va_mode); - sp->sa_uid = txdr_unsigned(cnp->cn_cred->cr_uid); - sp->sa_gid = txdr_unsigned(vattr.va_gid); - if (isnq) { - sp->sa_nqrdev = rdev; - sp->sa_nqflags = 0; - txdr_nqtime(&vap->va_atime, &sp->sa_nqatime); - txdr_nqtime(&vap->va_mtime, &sp->sa_nqmtime); + if (v3) { + nfsm_build(tl, u_long *, NFSX_UNSIGNED + NFSX_V3SRVSATTR); + *tl++ = vtonfsv3_type(vap->va_type); + sp3 = (struct nfsv3_sattr *)tl; + nfsm_v3sattr(sp3, vap, cnp->cn_cred->cr_uid, vattr.va_gid); + if (vap->va_type == VCHR || vap->va_type == VBLK) { + nfsm_build(tl, u_long *, 2 * NFSX_UNSIGNED); + *tl++ = txdr_unsigned(major(vap->va_rdev)); + *tl = txdr_unsigned(minor(vap->va_rdev)); + } } else { - sp->sa_nfssize = rdev; - txdr_nfstime(&vap->va_atime, &sp->sa_nfsatime); - txdr_nfstime(&vap->va_mtime, &sp->sa_nfsmtime); + nfsm_build(sp, struct nfsv2_sattr *, NFSX_V2SATTR); + sp->sa_mode = vtonfsv2_mode(vap->va_type, vap->va_mode); + sp->sa_uid = txdr_unsigned(cnp->cn_cred->cr_uid); + sp->sa_gid = txdr_unsigned(vattr.va_gid); + sp->sa_size = rdev; + txdr_nfsv2time(&vap->va_atime, &sp->sa_atime); + txdr_nfsv2time(&vap->va_mtime, &sp->sa_mtime); } - nfsm_request(dvp, NFSPROC_CREATE, cnp->cn_proc, cnp->cn_cred); - nfsm_mtofh(dvp, newvp); + nfsm_request(dvp, NFSPROC_MKNOD, cnp->cn_proc, cnp->cn_cred); + if (!error) { + nfsm_mtofh(dvp, newvp, v3, gotvp); + if (!gotvp) { + if (newvp) { + vput(newvp); + newvp = (struct vnode *)0; + } + error = nfs_lookitup(dvp, cnp->cn_nameptr, + cnp->cn_namelen, cnp->cn_cred, cnp->cn_proc, &np); + if (!error) + newvp = NFSTOV(np); + } + } + if (v3) + nfsm_wcc_data(dvp, wccflag); nfsm_reqdone; - if (!error && (cnp->cn_flags & MAKEENTRY)) - cache_enter(dvp, newvp, cnp); + if (error) { + if (newvp) + vput(newvp); + } else { + if (cnp->cn_flags & MAKEENTRY) + cache_enter(dvp, newvp, cnp); + *vpp = newvp; + } FREE(cnp->cn_pnbuf, M_NAMEI); VTONFS(dvp)->n_flag |= NMODIFIED; - VTONFS(dvp)->n_attrstamp = 0; - vrele(dvp); - if (newvp != NULLVP) - vrele(newvp); + if (!wccflag) + VTONFS(dvp)->n_attrstamp = 0; + vput(dvp); + return (error); +} + +/* + * nfs mknod vop + * just call nfs_mknodrpc() to do the work. + */ +/* ARGSUSED */ +int +nfs_mknod(ap) + struct vop_mknod_args /* { + struct vnode *a_dvp; + struct vnode **a_vpp; + struct componentname *a_cnp; + struct vattr *a_vap; + } */ *ap; +{ + struct vnode *newvp; + int error; + + error = nfs_mknodrpc(ap->a_dvp, &newvp, ap->a_cnp, ap->a_vap); + if (!error) + vput(newvp); return (error); } +static u_long create_verf; /* * nfs file create call */ @@ -1042,52 +1292,98 @@ nfs_create(ap) register struct vattr *vap = ap->a_vap; register struct componentname *cnp = ap->a_cnp; register struct nfsv2_sattr *sp; + register struct nfsv3_sattr *sp3; register u_long *tl; register caddr_t cp; register long t1, t2; + struct nfsnode *np = (struct nfsnode *)0; + struct vnode *newvp = (struct vnode *)0; caddr_t bpos, dpos, cp2; - int error = 0, isnq; + int error = 0, wccflag = NFSV3_WCCRATTR, gotvp = 0, fmode = 0; struct mbuf *mreq, *mrep, *md, *mb, *mb2; struct vattr vattr; + int v3 = NFS_ISV3(dvp); - error = VOP_GETATTR(dvp, &vattr, cnp->cn_cred, cnp->cn_proc); - if (error) { + /* + * Oops, not for me.. + */ + if (vap->va_type == VSOCK) + return (nfs_mknodrpc(dvp, ap->a_vpp, cnp, vap)); + + if (error = VOP_GETATTR(dvp, &vattr, cnp->cn_cred, cnp->cn_proc)) { VOP_ABORTOP(dvp, cnp); vput(dvp); return (error); } + if (vap->va_vaflags & VA_EXCLUSIVE) + fmode |= O_EXCL; +again: nfsstats.rpccnt[NFSPROC_CREATE]++; - isnq = (VFSTONFS(dvp->v_mount)->nm_flag & NFSMNT_NQNFS); - nfsm_reqhead(dvp, NFSPROC_CREATE, - NFSX_FH+NFSX_UNSIGNED+nfsm_rndup(cnp->cn_namelen)+NFSX_SATTR(isnq)); - nfsm_fhtom(dvp); + nfsm_reqhead(dvp, NFSPROC_CREATE, NFSX_FH(v3) + 2 * NFSX_UNSIGNED + + nfsm_rndup(cnp->cn_namelen) + NFSX_SATTR(v3)); + nfsm_fhtom(dvp, v3); nfsm_strtom(cnp->cn_nameptr, cnp->cn_namelen, NFS_MAXNAMLEN); - nfsm_build(sp, struct nfsv2_sattr *, NFSX_SATTR(isnq)); - sp->sa_mode = vtonfs_mode(vap->va_type, vap->va_mode); - sp->sa_uid = txdr_unsigned(cnp->cn_cred->cr_uid); - sp->sa_gid = txdr_unsigned(vattr.va_gid); - if (isnq) { - u_quad_t qval = 0; - - txdr_hyper(&qval, &sp->sa_nqsize); - sp->sa_nqflags = 0; - sp->sa_nqrdev = -1; - txdr_nqtime(&vap->va_atime, &sp->sa_nqatime); - txdr_nqtime(&vap->va_mtime, &sp->sa_nqmtime); + if (v3) { + nfsm_build(tl, u_long *, NFSX_UNSIGNED); + if (fmode & O_EXCL) { + *tl = txdr_unsigned(NFSV3CREATE_EXCLUSIVE); + nfsm_build(tl, u_long *, NFSX_V3CREATEVERF); + if (in_ifaddr) + *tl++ = IA_SIN(in_ifaddr)->sin_addr.s_addr; + else + *tl++ = create_verf; + *tl = ++create_verf; + } else { + *tl = txdr_unsigned(NFSV3CREATE_UNCHECKED); + nfsm_build(tl, u_long *, NFSX_V3SRVSATTR); + sp3 = (struct nfsv3_sattr *)tl; + nfsm_v3sattr(sp3, vap, cnp->cn_cred->cr_uid, vattr.va_gid); + } } else { - sp->sa_nfssize = 0; - txdr_nfstime(&vap->va_atime, &sp->sa_nfsatime); - txdr_nfstime(&vap->va_mtime, &sp->sa_nfsmtime); + nfsm_build(sp, struct nfsv2_sattr *, NFSX_V2SATTR); + sp->sa_mode = vtonfsv2_mode(vap->va_type, vap->va_mode); + sp->sa_uid = txdr_unsigned(cnp->cn_cred->cr_uid); + sp->sa_gid = txdr_unsigned(vattr.va_gid); + sp->sa_size = 0; + txdr_nfsv2time(&vap->va_atime, &sp->sa_atime); + txdr_nfsv2time(&vap->va_mtime, &sp->sa_mtime); } nfsm_request(dvp, NFSPROC_CREATE, cnp->cn_proc, cnp->cn_cred); - nfsm_mtofh(dvp, *ap->a_vpp); + if (!error) { + nfsm_mtofh(dvp, newvp, v3, gotvp); + if (!gotvp) { + if (newvp) { + vput(newvp); + newvp = (struct vnode *)0; + } + error = nfs_lookitup(dvp, cnp->cn_nameptr, + cnp->cn_namelen, cnp->cn_cred, cnp->cn_proc, &np); + if (!error) + newvp = NFSTOV(np); + } + } + if (v3) + nfsm_wcc_data(dvp, wccflag); nfsm_reqdone; - if (!error && (cnp->cn_flags & MAKEENTRY)) - cache_enter(dvp, *ap->a_vpp, cnp); + if (error) { + if (v3 && (fmode & O_EXCL) && error == NFSERR_NOTSUPP) { + fmode &= ~O_EXCL; + goto again; + } + if (newvp) + vput(newvp); + } else if (v3 && (fmode & O_EXCL)) + error = nfs_setattrrpc(newvp, vap, cnp->cn_cred, cnp->cn_proc); + if (!error) { + if (cnp->cn_flags & MAKEENTRY) + cache_enter(dvp, newvp, cnp); + *ap->a_vpp = newvp; + } FREE(cnp->cn_pnbuf, M_NAMEI); VTONFS(dvp)->n_flag |= NMODIFIED; - VTONFS(dvp)->n_attrstamp = 0; - vrele(dvp); + if (!wccflag) + VTONFS(dvp)->n_attrstamp = 0; + vput(dvp); return (error); } @@ -1122,22 +1418,17 @@ nfs_remove(ap) int error = 0; struct mbuf *mreq, *mrep, *md, *mb, *mb2; struct vattr vattr; + int v3 = NFS_ISV3(dvp); - if (vp->v_usecount > 1) { - if (!np->n_sillyrename) - error = nfs_sillyrename(dvp, vp, cnp); - else if (VOP_GETATTR(vp, &vattr, cnp->cn_cred, cnp->cn_proc) - == 0 && vattr.va_nlink > 1) - /* - * If we already have a silly name but there are more - * than one links, just proceed with the NFS remove - * request, as the bits will remain available (modulo - * network races). This avoids silently ignoring the - * attempted removal of a non-silly entry. - */ - goto doit; - } else { - doit: +#ifndef DIAGNOSTIC + if ((cnp->cn_flags & HASBUF) == 0) + panic("nfs_remove: no name"); + if (vp->v_usecount < 1) + panic("nfs_remove: bad v_usecount"); +#endif + if (vp->v_usecount == 1 || (np->n_sillyrename && + VOP_GETATTR(vp, &vattr, cnp->cn_cred, cnp->cn_proc) == 0 && + vattr.va_nlink > 1)) { /* * Purge the name cache so that the chance of a lookup for * the name succeeding while the remove is in progress is @@ -1147,23 +1438,14 @@ nfs_remove(ap) */ cache_purge(vp); /* - * Throw away biocache buffers. Mainly to avoid - * unnecessary delayed writes. + * throw away biocache buffers, mainly to avoid + * unnecessary delayed writes later. */ error = nfs_vinvalbuf(vp, 0, cnp->cn_cred, cnp->cn_proc, 1); - if (error == EINTR) - return (error); /* Do the rpc */ - nfsstats.rpccnt[NFSPROC_REMOVE]++; - nfsm_reqhead(dvp, NFSPROC_REMOVE, - NFSX_FH+NFSX_UNSIGNED+nfsm_rndup(cnp->cn_namelen)); - nfsm_fhtom(dvp); - nfsm_strtom(cnp->cn_nameptr, cnp->cn_namelen, NFS_MAXNAMLEN); - nfsm_request(dvp, NFSPROC_REMOVE, cnp->cn_proc, cnp->cn_cred); - nfsm_reqdone; - FREE(cnp->cn_pnbuf, M_NAMEI); - VTONFS(dvp)->n_flag |= NMODIFIED; - VTONFS(dvp)->n_attrstamp = 0; + if (error != EINTR) + error = nfs_removerpc(dvp, cnp->cn_nameptr, + cnp->cn_namelen, cnp->cn_cred, cnp->cn_proc); /* * Kludge City: If the first reply to the remove rpc is lost.. * the reply to the retransmitted request will be ENOENT @@ -1172,10 +1454,15 @@ nfs_remove(ap) */ if (error == ENOENT) error = 0; - } + } else if (!np->n_sillyrename) + error = nfs_sillyrename(dvp, vp, cnp); + FREE(cnp->cn_pnbuf, M_NAMEI); np->n_attrstamp = 0; - vrele(dvp); - vrele(vp); + vput(dvp); + if (vp == dvp) + vrele(vp); + else + vput(vp); return (error); } @@ -1186,22 +1473,42 @@ int nfs_removeit(sp) register struct sillyrename *sp; { + + return (nfs_removerpc(sp->s_dvp, sp->s_name, sp->s_namlen, sp->s_cred, + (struct proc *)0)); +} + +/* + * Nfs remove rpc, called from nfs_remove() and nfs_removeit(). + */ +int +nfs_removerpc(dvp, name, namelen, cred, proc) + register struct vnode *dvp; + char *name; + int namelen; + struct ucred *cred; + struct proc *proc; +{ register u_long *tl; register caddr_t cp; - register long t2; - caddr_t bpos, dpos; - int error = 0; + register long t1, t2; + caddr_t bpos, dpos, cp2; + int error = 0, wccflag = NFSV3_WCCRATTR; struct mbuf *mreq, *mrep, *md, *mb, *mb2; + int v3 = NFS_ISV3(dvp); nfsstats.rpccnt[NFSPROC_REMOVE]++; - nfsm_reqhead(sp->s_dvp, NFSPROC_REMOVE, - NFSX_FH+NFSX_UNSIGNED+nfsm_rndup(sp->s_namlen)); - nfsm_fhtom(sp->s_dvp); - nfsm_strtom(sp->s_name, sp->s_namlen, NFS_MAXNAMLEN); - nfsm_request(sp->s_dvp, NFSPROC_REMOVE, NULL, sp->s_cred); + nfsm_reqhead(dvp, NFSPROC_REMOVE, + NFSX_FH(v3) + NFSX_UNSIGNED + nfsm_rndup(namelen)); + nfsm_fhtom(dvp, v3); + nfsm_strtom(name, namelen, NFS_MAXNAMLEN); + nfsm_request(dvp, NFSPROC_REMOVE, proc, cred); + if (v3) + nfsm_wcc_data(dvp, wccflag); nfsm_reqdone; - VTONFS(sp->s_dvp)->n_flag |= NMODIFIED; - VTONFS(sp->s_dvp)->n_attrstamp = 0; + VTONFS(dvp)->n_flag |= NMODIFIED; + if (!wccflag) + VTONFS(dvp)->n_attrstamp = 0; return (error); } @@ -1225,13 +1532,13 @@ nfs_rename(ap) register struct vnode *tdvp = ap->a_tdvp; register struct componentname *tcnp = ap->a_tcnp; register struct componentname *fcnp = ap->a_fcnp; - register u_long *tl; - register caddr_t cp; - register long t2; - caddr_t bpos, dpos; - int error = 0; - struct mbuf *mreq, *mrep, *md, *mb, *mb2; + int error; +#ifndef DIAGNOSTIC + if ((tcnp->cn_flags & HASBUF) == 0 || + (fcnp->cn_flags & HASBUF) == 0) + panic("nfs_rename: no name"); +#endif /* Check for cross-device rename */ if ((fvp->v_mount != tdvp->v_mount) || (tvp && (fvp->v_mount != tvp->v_mount))) { @@ -1239,21 +1546,20 @@ nfs_rename(ap) goto out; } + /* + * If the tvp exists and is in use, sillyrename it before doing the + * rename of the new file over it. + */ + if (tvp && tvp->v_usecount > 1 && !VTONFS(tvp)->n_sillyrename && + !nfs_sillyrename(tdvp, tvp, tcnp)) { + vrele(tvp); + tvp = NULL; + } + + error = nfs_renamerpc(fdvp, fcnp->cn_nameptr, fcnp->cn_namelen, + tdvp, tcnp->cn_nameptr, tcnp->cn_namelen, tcnp->cn_cred, + tcnp->cn_proc); - nfsstats.rpccnt[NFSPROC_RENAME]++; - nfsm_reqhead(fdvp, NFSPROC_RENAME, - (NFSX_FH+NFSX_UNSIGNED)*2+nfsm_rndup(fcnp->cn_namelen)+ - nfsm_rndup(fcnp->cn_namelen)); /* or fcnp->cn_cred?*/ - nfsm_fhtom(fdvp); - nfsm_strtom(fcnp->cn_nameptr, fcnp->cn_namelen, NFS_MAXNAMLEN); - nfsm_fhtom(tdvp); - nfsm_strtom(tcnp->cn_nameptr, tcnp->cn_namelen, NFS_MAXNAMLEN); - nfsm_request(fdvp, NFSPROC_RENAME, tcnp->cn_proc, tcnp->cn_cred); - nfsm_reqdone; - VTONFS(fdvp)->n_flag |= NMODIFIED; - VTONFS(fdvp)->n_attrstamp = 0; - VTONFS(tdvp)->n_flag |= NMODIFIED; - VTONFS(tdvp)->n_attrstamp = 0; if (fvp->v_type == VDIR) { if (tvp != NULL && tvp->v_type == VDIR) cache_purge(tdvp); @@ -1285,26 +1591,52 @@ nfs_renameit(sdvp, scnp, sp) struct componentname *scnp; register struct sillyrename *sp; { + return (nfs_renamerpc(sdvp, scnp->cn_nameptr, scnp->cn_namelen, + sdvp, sp->s_name, sp->s_namlen, scnp->cn_cred, scnp->cn_proc)); +} + +/* + * Do an nfs rename rpc. Called from nfs_rename() and nfs_renameit(). + */ +int +nfs_renamerpc(fdvp, fnameptr, fnamelen, tdvp, tnameptr, tnamelen, cred, proc) + register struct vnode *fdvp; + char *fnameptr; + int fnamelen; + register struct vnode *tdvp; + char *tnameptr; + int tnamelen; + struct ucred *cred; + struct proc *proc; +{ register u_long *tl; register caddr_t cp; - register long t2; - caddr_t bpos, dpos; - int error = 0; + register long t1, t2; + caddr_t bpos, dpos, cp2; + int error = 0, fwccflag = NFSV3_WCCRATTR, twccflag = NFSV3_WCCRATTR; struct mbuf *mreq, *mrep, *md, *mb, *mb2; + int v3 = NFS_ISV3(fdvp); nfsstats.rpccnt[NFSPROC_RENAME]++; - nfsm_reqhead(sdvp, NFSPROC_RENAME, - (NFSX_FH+NFSX_UNSIGNED)*2+nfsm_rndup(scnp->cn_namelen)+ - nfsm_rndup(sp->s_namlen)); - nfsm_fhtom(sdvp); - nfsm_strtom(scnp->cn_nameptr, scnp->cn_namelen, NFS_MAXNAMLEN); - nfsm_fhtom(sdvp); - nfsm_strtom(sp->s_name, sp->s_namlen, NFS_MAXNAMLEN); - nfsm_request(sdvp, NFSPROC_RENAME, scnp->cn_proc, scnp->cn_cred); + nfsm_reqhead(fdvp, NFSPROC_RENAME, + (NFSX_FH(v3) + NFSX_UNSIGNED)*2 + nfsm_rndup(fnamelen) + + nfsm_rndup(tnamelen)); + nfsm_fhtom(fdvp, v3); + nfsm_strtom(fnameptr, fnamelen, NFS_MAXNAMLEN); + nfsm_fhtom(tdvp, v3); + nfsm_strtom(tnameptr, tnamelen, NFS_MAXNAMLEN); + nfsm_request(fdvp, NFSPROC_RENAME, proc, cred); + if (v3) { + nfsm_wcc_data(fdvp, fwccflag); + nfsm_wcc_data(tdvp, twccflag); + } nfsm_reqdone; - FREE(scnp->cn_pnbuf, M_NAMEI); - VTONFS(sdvp)->n_flag |= NMODIFIED; - VTONFS(sdvp)->n_attrstamp = 0; + VTONFS(fdvp)->n_flag |= NMODIFIED; + VTONFS(tdvp)->n_flag |= NMODIFIED; + if (!fwccflag) + VTONFS(fdvp)->n_attrstamp = 0; + if (!twccflag) + VTONFS(tdvp)->n_attrstamp = 0; return (error); } @@ -1319,22 +1651,32 @@ nfs_link(ap) struct componentname *a_cnp; } */ *ap; { +#if defined(__FreeBSD__) || defined(__NetBSD__) + /* + * Since the args are reversed in the VOP_LINK() calls, + * switch them back. Argh! + */ + register struct vnode *vp = ap->a_tdvp; + register struct vnode *tdvp = ap->a_vp; +#else register struct vnode *vp = ap->a_vp; register struct vnode *tdvp = ap->a_tdvp; +#endif register struct componentname *cnp = ap->a_cnp; register u_long *tl; register caddr_t cp; - register long t2; - caddr_t bpos, dpos; - int error = 0; + register long t1, t2; + caddr_t bpos, dpos, cp2; + int error = 0, wccflag = NFSV3_WCCRATTR, attrflag = 0; struct mbuf *mreq, *mrep, *md, *mb, *mb2; + int v3 = NFS_ISV3(vp); if (vp->v_mount != tdvp->v_mount) { /*VOP_ABORTOP(vp, cnp);*/ if (tdvp == vp) - vrele(vp); + vrele(tdvp); else - vput(vp); + vput(tdvp); return (EXDEV); } @@ -1343,21 +1685,27 @@ nfs_link(ap) * doesn't get "out of sync" with the server. * XXX There should be a better way! */ - VOP_FSYNC(tdvp, cnp->cn_cred, MNT_WAIT, cnp->cn_proc); + VOP_FSYNC(vp, cnp->cn_cred, MNT_WAIT, cnp->cn_proc); nfsstats.rpccnt[NFSPROC_LINK]++; - nfsm_reqhead(tdvp, NFSPROC_LINK, - NFSX_FH*2+NFSX_UNSIGNED+nfsm_rndup(cnp->cn_namelen)); - nfsm_fhtom(tdvp); - nfsm_fhtom(vp); + nfsm_reqhead(vp, NFSPROC_LINK, + NFSX_FH(v3)*2 + NFSX_UNSIGNED + nfsm_rndup(cnp->cn_namelen)); + nfsm_fhtom(vp, v3); + nfsm_fhtom(tdvp, v3); nfsm_strtom(cnp->cn_nameptr, cnp->cn_namelen, NFS_MAXNAMLEN); - nfsm_request(tdvp, NFSPROC_LINK, cnp->cn_proc, cnp->cn_cred); + nfsm_request(vp, NFSPROC_LINK, cnp->cn_proc, cnp->cn_cred); + if (v3) { + nfsm_postop_attr(vp, attrflag); + nfsm_wcc_data(tdvp, wccflag); + } nfsm_reqdone; FREE(cnp->cn_pnbuf, M_NAMEI); - VTONFS(tdvp)->n_attrstamp = 0; - VTONFS(vp)->n_flag |= NMODIFIED; - VTONFS(vp)->n_attrstamp = 0; - vrele(vp); + VTONFS(tdvp)->n_flag |= NMODIFIED; + if (!attrflag) + VTONFS(vp)->n_attrstamp = 0; + if (!wccflag) + VTONFS(tdvp)->n_attrstamp = 0; + vput(tdvp); /* * Kludge: Map EEXIST => 0 assuming that it is a reply to a retry. */ @@ -1369,7 +1717,6 @@ nfs_link(ap) /* * nfs symbolic link create call */ -/* start here */ int nfs_symlink(ap) struct vop_symlink_args /* { @@ -1384,43 +1731,51 @@ nfs_symlink(ap) register struct vattr *vap = ap->a_vap; register struct componentname *cnp = ap->a_cnp; register struct nfsv2_sattr *sp; + register struct nfsv3_sattr *sp3; register u_long *tl; register caddr_t cp; - register long t2; - caddr_t bpos, dpos; - int slen, error = 0, isnq; + register long t1, t2; + caddr_t bpos, dpos, cp2; + int slen, error = 0, wccflag = NFSV3_WCCRATTR, gotvp; struct mbuf *mreq, *mrep, *md, *mb, *mb2; + struct vnode *newvp = (struct vnode *)0; + int v3 = NFS_ISV3(dvp); nfsstats.rpccnt[NFSPROC_SYMLINK]++; slen = strlen(ap->a_target); - isnq = (VFSTONFS(dvp->v_mount)->nm_flag & NFSMNT_NQNFS); - nfsm_reqhead(dvp, NFSPROC_SYMLINK, NFSX_FH+2*NFSX_UNSIGNED+ - nfsm_rndup(cnp->cn_namelen)+nfsm_rndup(slen)+NFSX_SATTR(isnq)); - nfsm_fhtom(dvp); + nfsm_reqhead(dvp, NFSPROC_SYMLINK, NFSX_FH(v3) + 2*NFSX_UNSIGNED + + nfsm_rndup(cnp->cn_namelen) + nfsm_rndup(slen) + NFSX_SATTR(v3)); + nfsm_fhtom(dvp, v3); nfsm_strtom(cnp->cn_nameptr, cnp->cn_namelen, NFS_MAXNAMLEN); + if (v3) { + nfsm_build(sp3, struct nfsv3_sattr *, NFSX_V3SRVSATTR); + nfsm_v3sattr(sp3, vap, cnp->cn_cred->cr_uid, + cnp->cn_cred->cr_gid); + } nfsm_strtom(ap->a_target, slen, NFS_MAXPATHLEN); - nfsm_build(sp, struct nfsv2_sattr *, NFSX_SATTR(isnq)); - sp->sa_mode = vtonfs_mode(VLNK, vap->va_mode); - sp->sa_uid = txdr_unsigned(cnp->cn_cred->cr_uid); - sp->sa_gid = txdr_unsigned(cnp->cn_cred->cr_gid); - if (isnq) { - quad_t qval = -1; - - txdr_hyper(&qval, &sp->sa_nqsize); - sp->sa_nqflags = 0; - txdr_nqtime(&vap->va_atime, &sp->sa_nqatime); - txdr_nqtime(&vap->va_mtime, &sp->sa_nqmtime); - } else { - sp->sa_nfssize = -1; - txdr_nfstime(&vap->va_atime, &sp->sa_nfsatime); - txdr_nfstime(&vap->va_mtime, &sp->sa_nfsmtime); + if (!v3) { + nfsm_build(sp, struct nfsv2_sattr *, NFSX_V2SATTR); + sp->sa_mode = vtonfsv2_mode(VLNK, vap->va_mode); + sp->sa_uid = txdr_unsigned(cnp->cn_cred->cr_uid); + sp->sa_gid = txdr_unsigned(cnp->cn_cred->cr_gid); + sp->sa_size = -1; + txdr_nfsv2time(&vap->va_atime, &sp->sa_atime); + txdr_nfsv2time(&vap->va_mtime, &sp->sa_mtime); } nfsm_request(dvp, NFSPROC_SYMLINK, cnp->cn_proc, cnp->cn_cred); + if (v3) { + if (!error) + nfsm_mtofh(dvp, newvp, v3, gotvp); + nfsm_wcc_data(dvp, wccflag); + } nfsm_reqdone; + if (newvp) + vput(newvp); FREE(cnp->cn_pnbuf, M_NAMEI); VTONFS(dvp)->n_flag |= NMODIFIED; - VTONFS(dvp)->n_attrstamp = 0; - vrele(dvp); + if (!wccflag) + VTONFS(dvp)->n_attrstamp = 0; + vput(dvp); /* * Kludge: Map EEXIST => 0 assuming that it is a reply to a retry. */ @@ -1444,76 +1799,78 @@ nfs_mkdir(ap) register struct vnode *dvp = ap->a_dvp; register struct vattr *vap = ap->a_vap; register struct componentname *cnp = ap->a_cnp; - register struct vnode **vpp = ap->a_vpp; register struct nfsv2_sattr *sp; + register struct nfsv3_sattr *sp3; register u_long *tl; register caddr_t cp; register long t1, t2; register int len; + struct nfsnode *np = (struct nfsnode *)0; + struct vnode *newvp = (struct vnode *)0; caddr_t bpos, dpos, cp2; - int error = 0, firsttry = 1, isnq; + nfsfh_t *fhp; + int error = 0, wccflag = NFSV3_WCCRATTR, attrflag; + int fhsize, gotvp = 0; struct mbuf *mreq, *mrep, *md, *mb, *mb2; struct vattr vattr; + int v3 = NFS_ISV3(dvp); - error = VOP_GETATTR(dvp, &vattr, cnp->cn_cred, cnp->cn_proc); - if (error) { + if (error = VOP_GETATTR(dvp, &vattr, cnp->cn_cred, cnp->cn_proc)) { VOP_ABORTOP(dvp, cnp); vput(dvp); return (error); } len = cnp->cn_namelen; - isnq = (VFSTONFS(dvp->v_mount)->nm_flag & NFSMNT_NQNFS); nfsstats.rpccnt[NFSPROC_MKDIR]++; nfsm_reqhead(dvp, NFSPROC_MKDIR, - NFSX_FH+NFSX_UNSIGNED+nfsm_rndup(len)+NFSX_SATTR(isnq)); - nfsm_fhtom(dvp); + NFSX_FH(v3) + NFSX_UNSIGNED + nfsm_rndup(len) + NFSX_SATTR(v3)); + nfsm_fhtom(dvp, v3); nfsm_strtom(cnp->cn_nameptr, len, NFS_MAXNAMLEN); - nfsm_build(sp, struct nfsv2_sattr *, NFSX_SATTR(isnq)); - sp->sa_mode = vtonfs_mode(VDIR, vap->va_mode); - sp->sa_uid = txdr_unsigned(cnp->cn_cred->cr_uid); - sp->sa_gid = txdr_unsigned(vattr.va_gid); - if (isnq) { - quad_t qval = -1; - - txdr_hyper(&qval, &sp->sa_nqsize); - sp->sa_nqflags = 0; - txdr_nqtime(&vap->va_atime, &sp->sa_nqatime); - txdr_nqtime(&vap->va_mtime, &sp->sa_nqmtime); + if (v3) { + nfsm_build(sp3, struct nfsv3_sattr *, NFSX_V3SRVSATTR); + nfsm_v3sattr(sp3, vap, cnp->cn_cred->cr_uid, vattr.va_gid); } else { - sp->sa_nfssize = -1; - txdr_nfstime(&vap->va_atime, &sp->sa_nfsatime); - txdr_nfstime(&vap->va_mtime, &sp->sa_nfsmtime); + nfsm_build(sp, struct nfsv2_sattr *, NFSX_V2SATTR); + sp->sa_mode = vtonfsv2_mode(VDIR, vap->va_mode); + sp->sa_uid = txdr_unsigned(cnp->cn_cred->cr_uid); + sp->sa_gid = txdr_unsigned(vattr.va_gid); + sp->sa_size = -1; + txdr_nfsv2time(&vap->va_atime, &sp->sa_atime); + txdr_nfsv2time(&vap->va_mtime, &sp->sa_mtime); } nfsm_request(dvp, NFSPROC_MKDIR, cnp->cn_proc, cnp->cn_cred); - nfsm_mtofh(dvp, *vpp); + if (!error) + nfsm_mtofh(dvp, newvp, v3, gotvp); + if (v3) + nfsm_wcc_data(dvp, wccflag); nfsm_reqdone; VTONFS(dvp)->n_flag |= NMODIFIED; - VTONFS(dvp)->n_attrstamp = 0; + if (!wccflag) + VTONFS(dvp)->n_attrstamp = 0; /* * Kludge: Map EEXIST => 0 assuming that you have a reply to a retry * if we can succeed in looking up the directory. - * "firsttry" is necessary since the macros may "goto nfsmout" which - * is above the if on errors. (Ugh) */ - if (error == EEXIST && firsttry) { - firsttry = 0; - error = 0; - nfsstats.rpccnt[NFSPROC_LOOKUP]++; - *vpp = NULL; - nfsm_reqhead(dvp, NFSPROC_LOOKUP, - NFSX_FH+NFSX_UNSIGNED+nfsm_rndup(len)); - nfsm_fhtom(dvp); - nfsm_strtom(cnp->cn_nameptr, len, NFS_MAXNAMLEN); - nfsm_request(dvp, NFSPROC_LOOKUP, cnp->cn_proc, cnp->cn_cred); - nfsm_mtofh(dvp, *vpp); - if ((*vpp)->v_type != VDIR) { - vput(*vpp); - error = EEXIST; + if (error == EEXIST || (!error && !gotvp)) { + if (newvp) { + vrele(newvp); + newvp = (struct vnode *)0; + } + error = nfs_lookitup(dvp, cnp->cn_nameptr, len, cnp->cn_cred, + cnp->cn_proc, &np); + if (!error) { + newvp = NFSTOV(np); + if (newvp->v_type != VDIR) + error = EEXIST; } - m_freem(mrep); } + if (error) { + if (newvp) + vrele(newvp); + } else + *ap->a_vpp = newvp; FREE(cnp->cn_pnbuf, M_NAMEI); - vrele(dvp); + vput(dvp); return (error); } @@ -1533,31 +1890,35 @@ nfs_rmdir(ap) register struct componentname *cnp = ap->a_cnp; register u_long *tl; register caddr_t cp; - register long t2; - caddr_t bpos, dpos; - int error = 0; + register long t1, t2; + caddr_t bpos, dpos, cp2; + int error = 0, wccflag = NFSV3_WCCRATTR; struct mbuf *mreq, *mrep, *md, *mb, *mb2; + int v3 = NFS_ISV3(dvp); if (dvp == vp) { - vrele(dvp); + vput(dvp); vrele(dvp); FREE(cnp->cn_pnbuf, M_NAMEI); return (EINVAL); } nfsstats.rpccnt[NFSPROC_RMDIR]++; nfsm_reqhead(dvp, NFSPROC_RMDIR, - NFSX_FH+NFSX_UNSIGNED+nfsm_rndup(cnp->cn_namelen)); - nfsm_fhtom(dvp); + NFSX_FH(v3) + NFSX_UNSIGNED + nfsm_rndup(cnp->cn_namelen)); + nfsm_fhtom(dvp, v3); nfsm_strtom(cnp->cn_nameptr, cnp->cn_namelen, NFS_MAXNAMLEN); nfsm_request(dvp, NFSPROC_RMDIR, cnp->cn_proc, cnp->cn_cred); + if (v3) + nfsm_wcc_data(dvp, wccflag); nfsm_reqdone; FREE(cnp->cn_pnbuf, M_NAMEI); VTONFS(dvp)->n_flag |= NMODIFIED; - VTONFS(dvp)->n_attrstamp = 0; + if (!wccflag) + VTONFS(dvp)->n_attrstamp = 0; cache_purge(dvp); cache_purge(vp); - vrele(vp); - vrele(dvp); + vput(vp); + vput(dvp); /* * Kludge: Map ENOENT => 0 assuming that you have a reply to a retry. */ @@ -1568,9 +1929,6 @@ nfs_rmdir(ap) /* * nfs readdir call - * Although cookie is defined as opaque, I translate it to/from net byte - * order so that it looks more sensible. This appears consistent with the - * Ultrix implementation of NFS. */ int nfs_readdir(ap) @@ -1591,10 +1949,10 @@ nfs_readdir(ap) /* * First, check for hit on the EOF offset cache */ - if (uio->uio_offset != 0 && uio->uio_offset == np->n_direofoffset && + if (np->n_direofoffset > 0 && uio->uio_offset >= np->n_direofoffset && (np->n_flag & NMODIFIED) == 0) { if (VFSTONFS(vp->v_mount)->nm_flag & NFSMNT_NQNFS) { - if (NQNFS_CKCACHABLE(vp, NQL_READ)) { + if (NQNFS_CKCACHABLE(vp, ND_READ)) { nfsstats.direofcache_hits++; return (0); } @@ -1622,90 +1980,148 @@ nfs_readdir(ap) */ int nfs_readdirrpc(vp, uiop, cred) - register struct vnode *vp; - struct uio *uiop; + struct vnode *vp; + register struct uio *uiop; struct ucred *cred; + { - register long len; - register struct dirent *dp = 0; + register int len, left; + register struct dirent *dp; register u_long *tl; register caddr_t cp; - register long t1; - long tlen, lastlen = 0; + register long t1, t2; + register nfsuint64 *cookiep; caddr_t bpos, dpos, cp2; - int error = 0; struct mbuf *mreq, *mrep, *md, *mb, *mb2; - struct mbuf *md2; - caddr_t dpos2; - int siz; - int more_dirs = 1; - u_long off, savoff = 0; - struct dirent *savdp = 0; - struct nfsmount *nmp; - struct nfsnode *np = VTONFS(vp); - long tresid; + nfsuint64 cookie; + struct nfsmount *nmp = VFSTONFS(vp->v_mount); + struct nfsnode *dnp = VTONFS(vp); + nfsfh_t *fhp; + u_quad_t frev, fileno; + int error = 0, tlen, more_dirs = 1, blksiz = 0, bigenough = 1, i; + int cachable, attrflag, fhsize; + int v3 = NFS_ISV3(vp); + +#ifndef nolint + dp = (struct dirent *)0; +#endif +#ifndef DIAGNOSTIC + if (uiop->uio_iovcnt != 1 || (uiop->uio_offset & (NFS_DIRBLKSIZ - 1)) || + (uiop->uio_resid & (NFS_DIRBLKSIZ - 1))) + panic("nfs readdirrpc bad uio"); +#endif - nmp = VFSTONFS(vp->v_mount); - tresid = uiop->uio_resid; /* - * Loop around doing readdir rpc's of size uio_resid or nm_rsize, - * whichever is smaller, truncated to a multiple of NFS_DIRBLKSIZ. + * If there is no cookie, assume end of directory. + */ + cookiep = nfs_getcookie(dnp, uiop->uio_offset, 0); + if (cookiep) + cookie = *cookiep; + else + return (0); + /* + * Loop around doing readdir rpc's of size nm_readdirsize + * truncated to a multiple of DIRBLKSIZ. * The stopping criteria is EOF or buffer full. */ - while (more_dirs && uiop->uio_resid >= NFS_DIRBLKSIZ) { + while (more_dirs && bigenough) { nfsstats.rpccnt[NFSPROC_READDIR]++; - nfsm_reqhead(vp, NFSPROC_READDIR, - NFSX_FH + 2 * NFSX_UNSIGNED); - nfsm_fhtom(vp); - nfsm_build(tl, u_long *, 2 * NFSX_UNSIGNED); - off = (u_long)uiop->uio_offset; - *tl++ = txdr_unsigned(off); - *tl = txdr_unsigned(((uiop->uio_resid > nmp->nm_rsize) ? - nmp->nm_rsize : uiop->uio_resid) & ~(NFS_DIRBLKSIZ-1)); + nfsm_reqhead(vp, NFSPROC_READDIR, NFSX_FH(v3) + + NFSX_READDIR(v3)); + nfsm_fhtom(vp, v3); + if (v3) { + nfsm_build(tl, u_long *, 5 * NFSX_UNSIGNED); + *tl++ = cookie.nfsuquad[0]; + *tl++ = cookie.nfsuquad[1]; + *tl++ = dnp->n_cookieverf.nfsuquad[0]; + *tl++ = dnp->n_cookieverf.nfsuquad[1]; + } else { + nfsm_build(tl, u_long *, 2 * NFSX_UNSIGNED); + *tl++ = cookie.nfsuquad[0]; + } + *tl = txdr_unsigned(nmp->nm_readdirsize); nfsm_request(vp, NFSPROC_READDIR, uiop->uio_procp, cred); - siz = 0; + if (v3) { + nfsm_postop_attr(vp, attrflag); + if (!error) { + nfsm_dissect(tl, u_long *, 2 * NFSX_UNSIGNED); + dnp->n_cookieverf.nfsuquad[0] = *tl++; + dnp->n_cookieverf.nfsuquad[1] = *tl; + } else { + m_freem(mrep); + goto nfsmout; + } + } nfsm_dissect(tl, u_long *, NFSX_UNSIGNED); more_dirs = fxdr_unsigned(int, *tl); - /* Save the position so that we can do nfsm_mtouio() later */ - dpos2 = dpos; - md2 = md; - /* loop thru the dir entries, doctoring them to 4bsd form */ -#ifdef lint - dp = (struct dirent *)0; -#endif /* lint */ - while (more_dirs && siz < uiop->uio_resid) { - savoff = off; /* Hold onto offset and dp */ - savdp = dp; - nfsm_dissect(tl, u_long *, 2 * NFSX_UNSIGNED); - dp = (struct dirent *)tl; - dp->d_fileno = fxdr_unsigned(u_long, *tl++); - len = fxdr_unsigned(int, *tl); + while (more_dirs && bigenough) { + if (v3) { + nfsm_dissect(tl, u_long *, 3 * NFSX_UNSIGNED); + fxdr_hyper(tl, &fileno); + len = fxdr_unsigned(int, *(tl + 2)); + } else { + nfsm_dissect(tl, u_long *, 2 * NFSX_UNSIGNED); + fileno = fxdr_unsigned(u_quad_t, *tl++); + len = fxdr_unsigned(int, *tl); + } if (len <= 0 || len > NFS_MAXNAMLEN) { error = EBADRPC; m_freem(mrep); goto nfsmout; } - dp->d_namlen = (u_char)len; - dp->d_type = DT_UNKNOWN; - nfsm_adv(len); /* Point past name */ tlen = nfsm_rndup(len); - /* - * This should not be necessary, but some servers have - * broken XDR such that these bytes are not null filled. - */ - if (tlen != len) { - *dpos = '\0'; /* Null-terminate */ - nfsm_adv(tlen - len); - len = tlen; + if (tlen == len) + tlen += 4; /* To ensure null termination */ + left = DIRBLKSIZ - blksiz; + if ((tlen + DIRHDSIZ) > left) { + dp->d_reclen += left; + uiop->uio_iov->iov_base += left; + uiop->uio_iov->iov_len -= left; + uiop->uio_offset += left; + uiop->uio_resid -= left; + blksiz = 0; } - nfsm_dissect(tl, u_long *, 2 * NFSX_UNSIGNED); - off = fxdr_unsigned(u_long, *tl); - *tl++ = 0; /* Ensures null termination of name */ + if ((tlen + DIRHDSIZ) > uiop->uio_resid) + bigenough = 0; + if (bigenough) { + dp = (struct dirent *)uiop->uio_iov->iov_base; + dp->d_fileno = (int)fileno; + dp->d_namlen = len; + dp->d_reclen = tlen + DIRHDSIZ; + dp->d_type = DT_UNKNOWN; + blksiz += dp->d_reclen; + if (blksiz == DIRBLKSIZ) + blksiz = 0; + uiop->uio_offset += DIRHDSIZ; + uiop->uio_resid -= DIRHDSIZ; + uiop->uio_iov->iov_base += DIRHDSIZ; + uiop->uio_iov->iov_len -= DIRHDSIZ; + nfsm_mtouio(uiop, len); + cp = uiop->uio_iov->iov_base; + tlen -= len; + *cp = '\0'; /* null terminate */ + uiop->uio_iov->iov_base += tlen; + uiop->uio_iov->iov_len -= tlen; + uiop->uio_offset += tlen; + uiop->uio_resid -= tlen; + } else + nfsm_adv(nfsm_rndup(len)); + if (v3) { + nfsm_dissect(tl, u_long *, 3 * NFSX_UNSIGNED); + } else { + nfsm_dissect(tl, u_long *, 2 * NFSX_UNSIGNED); + } + if (bigenough) { + cookie.nfsuquad[0] = *tl++; + if (v3) + cookie.nfsuquad[1] = *tl++; + } else if (v3) + tl += 2; + else + tl++; more_dirs = fxdr_unsigned(int, *tl); - dp->d_reclen = len + 4 * NFSX_UNSIGNED; - siz += dp->d_reclen; } /* * If at end of rpc data, get the eof boolean @@ -1713,189 +2129,224 @@ nfs_readdirrpc(vp, uiop, cred) if (!more_dirs) { nfsm_dissect(tl, u_long *, NFSX_UNSIGNED); more_dirs = (fxdr_unsigned(int, *tl) == 0); - - /* - * If at EOF, cache directory offset - */ - if (!more_dirs) - np->n_direofoffset = off; - } - /* - * If there is too much to fit in the data buffer, use savoff and - * savdp to trim off the last record. - * --> we are not at eof - */ - if (siz > uiop->uio_resid) { - off = savoff; - siz -= dp->d_reclen; - dp = savdp; - more_dirs = 0; /* Paranoia */ } - if (siz > 0) { - lastlen = dp->d_reclen; - md = md2; - dpos = dpos2; - nfsm_mtouio(uiop, siz); - uiop->uio_offset = (off_t)off; - } else - more_dirs = 0; /* Ugh, never happens, but in case.. */ m_freem(mrep); } /* - * Fill last record, iff any, out to a multiple of NFS_DIRBLKSIZ + * Fill last record, iff any, out to a multiple of DIRBLKSIZ * by increasing d_reclen for the last record. */ - if (uiop->uio_resid < tresid) { - len = uiop->uio_resid & (NFS_DIRBLKSIZ - 1); - if (len > 0) { - dp = (struct dirent *) - (uiop->uio_iov->iov_base - lastlen); - dp->d_reclen += len; - uiop->uio_iov->iov_base += len; - uiop->uio_iov->iov_len -= len; - uiop->uio_resid -= len; - } + if (blksiz > 0) { + left = DIRBLKSIZ - blksiz; + dp->d_reclen += left; + uiop->uio_iov->iov_base += left; + uiop->uio_iov->iov_len -= left; + uiop->uio_offset += left; + uiop->uio_resid -= left; + } + + /* + * We are now either at the end of the directory or have filled the + * block. + */ + if (bigenough) + dnp->n_direofoffset = uiop->uio_offset; + else { + if (uiop->uio_resid > 0) + printf("EEK! readdirrpc resid > 0\n"); + cookiep = nfs_getcookie(dnp, uiop->uio_offset, 1); + *cookiep = cookie; } nfsmout: return (error); } /* - * Nqnfs readdir_and_lookup RPC. Used in place of nfs_readdirrpc(). + * NFS V3 readdir plus RPC. Used in place of nfs_readdirrpc(). */ int -nfs_readdirlookrpc(vp, uiop, cred) +nfs_readdirplusrpc(vp, uiop, cred) struct vnode *vp; register struct uio *uiop; struct ucred *cred; { - register int len; - register struct dirent *dp = 0; + register int len, left; + register struct dirent *dp; register u_long *tl; register caddr_t cp; - register long t1; - caddr_t bpos, dpos, cp2; - struct mbuf *mreq, *mrep, *md, *mb, *mb2; + register long t1, t2; + register struct vnode *newvp; + register nfsuint64 *cookiep; + caddr_t bpos, dpos, cp2, dpossav1, dpossav2; + struct mbuf *mreq, *mrep, *md, *mb, *mb2, *mdsav1, *mdsav2; struct nameidata nami, *ndp = &nami; struct componentname *cnp = &ndp->ni_cnd; - u_long off, endoff = 0, fileno; - time_t reqtime, ltime = 0; - struct nfsmount *nmp; - struct nfsnode *np; - struct vnode *newvp; - nfsv2fh_t *fhp; - u_quad_t frev; - int error = 0, tlen, more_dirs = 1, tresid, doit, bigenough, i; - int cachable = 0; - - if (uiop->uio_iovcnt != 1) - panic("nfs rdirlook"); - nmp = VFSTONFS(vp->v_mount); - tresid = uiop->uio_resid; + nfsuint64 cookie; + struct nfsmount *nmp = VFSTONFS(vp->v_mount); + struct nfsnode *dnp = VTONFS(vp), *np; + nfsfh_t *fhp; + u_quad_t frev, fileno; + int error = 0, tlen, more_dirs = 1, blksiz = 0, doit, bigenough = 1, i; + int cachable, attrflag, fhsize; + +#ifndef nolint + dp = (struct dirent *)0; +#endif +#ifndef DIAGNOSTIC + if (uiop->uio_iovcnt != 1 || (uiop->uio_offset & (DIRBLKSIZ - 1)) || + (uiop->uio_resid & (DIRBLKSIZ - 1))) + panic("nfs readdirplusrpc bad uio"); +#endif ndp->ni_dvp = vp; newvp = NULLVP; + /* - * Loop around doing readdir rpc's of size uio_resid or nm_rsize, - * whichever is smaller, truncated to a multiple of NFS_DIRBLKSIZ. + * If there is no cookie, assume end of directory. + */ + cookiep = nfs_getcookie(dnp, uiop->uio_offset, 0); + if (cookiep) + cookie = *cookiep; + else + return (0); + /* + * Loop around doing readdir rpc's of size nm_readdirsize + * truncated to a multiple of DIRBLKSIZ. * The stopping criteria is EOF or buffer full. */ - while (more_dirs && uiop->uio_resid >= NFS_DIRBLKSIZ) { - nfsstats.rpccnt[NQNFSPROC_READDIRLOOK]++; - nfsm_reqhead(vp, NQNFSPROC_READDIRLOOK, - NFSX_FH + 3 * NFSX_UNSIGNED); - nfsm_fhtom(vp); - nfsm_build(tl, u_long *, 3 * NFSX_UNSIGNED); - off = (u_long)uiop->uio_offset; - *tl++ = txdr_unsigned(off); - *tl++ = txdr_unsigned(((uiop->uio_resid > nmp->nm_rsize) ? - nmp->nm_rsize : uiop->uio_resid) & ~(NFS_DIRBLKSIZ-1)); - if (nmp->nm_flag & NFSMNT_NQLOOKLEASE) - *tl = txdr_unsigned(nmp->nm_leaseterm); - else - *tl = 0; - reqtime = time.tv_sec; - nfsm_request(vp, NQNFSPROC_READDIRLOOK, uiop->uio_procp, cred); - nfsm_dissect(tl, u_long *, NFSX_UNSIGNED); + while (more_dirs && bigenough) { + nfsstats.rpccnt[NFSPROC_READDIRPLUS]++; + nfsm_reqhead(vp, NFSPROC_READDIRPLUS, + NFSX_FH(1) + 6 * NFSX_UNSIGNED); + nfsm_fhtom(vp, 1); + nfsm_build(tl, u_long *, 6 * NFSX_UNSIGNED); + *tl++ = cookie.nfsuquad[0]; + *tl++ = cookie.nfsuquad[1]; + *tl++ = dnp->n_cookieverf.nfsuquad[0]; + *tl++ = dnp->n_cookieverf.nfsuquad[1]; + *tl++ = txdr_unsigned(nmp->nm_readdirsize); + *tl = txdr_unsigned(nmp->nm_rsize); + nfsm_request(vp, NFSPROC_READDIRPLUS, uiop->uio_procp, cred); + nfsm_postop_attr(vp, attrflag); + if (error) { + m_freem(mrep); + goto nfsmout; + } + nfsm_dissect(tl, u_long *, 3 * NFSX_UNSIGNED); + dnp->n_cookieverf.nfsuquad[0] = *tl++; + dnp->n_cookieverf.nfsuquad[1] = *tl++; more_dirs = fxdr_unsigned(int, *tl); /* loop thru the dir entries, doctoring them to 4bsd form */ - bigenough = 1; while (more_dirs && bigenough) { - doit = 1; - nfsm_dissect(tl, u_long *, 4 * NFSX_UNSIGNED); - if (nmp->nm_flag & NFSMNT_NQLOOKLEASE) { - cachable = fxdr_unsigned(int, *tl++); - ltime = reqtime + fxdr_unsigned(int, *tl++); - fxdr_hyper(tl, &frev); - } - nfsm_dissect(fhp, nfsv2fh_t *, NFSX_FH); - if (!bcmp(VTONFS(vp)->n_fh.fh_bytes, (caddr_t)fhp, NFSX_FH)) { - VREF(vp); - newvp = vp; - np = VTONFS(vp); - } else { - error = nfs_nget(vp->v_mount, fhp, &np); - if (error) - doit = 0; - newvp = NFSTOV(np); - } - error = nfs_loadattrcache(&newvp, &md, &dpos, - (struct vattr *)0); - if (error) - doit = 0; - nfsm_dissect(tl, u_long *, 2 * NFSX_UNSIGNED); - fileno = fxdr_unsigned(u_long, *tl++); - len = fxdr_unsigned(int, *tl); + nfsm_dissect(tl, u_long *, 3 * NFSX_UNSIGNED); + fxdr_hyper(tl, &fileno); + len = fxdr_unsigned(int, *(tl + 2)); if (len <= 0 || len > NFS_MAXNAMLEN) { error = EBADRPC; m_freem(mrep); goto nfsmout; } - tlen = (len + 4) & ~0x3; + tlen = nfsm_rndup(len); + if (tlen == len) + tlen += 4; /* To ensure null termination*/ + left = DIRBLKSIZ - blksiz; + if ((tlen + DIRHDSIZ) > left) { + dp->d_reclen += left; + uiop->uio_iov->iov_base += left; + uiop->uio_iov->iov_len -= left; + uiop->uio_offset += left; + uiop->uio_resid -= left; + blksiz = 0; + } if ((tlen + DIRHDSIZ) > uiop->uio_resid) bigenough = 0; - if (bigenough && doit) { + if (bigenough) { dp = (struct dirent *)uiop->uio_iov->iov_base; - dp->d_fileno = fileno; + dp->d_fileno = (int)fileno; dp->d_namlen = len; dp->d_reclen = tlen + DIRHDSIZ; - dp->d_type = - IFTODT(VTTOIF(np->n_vattr.va_type)); + dp->d_type = DT_UNKNOWN; + blksiz += dp->d_reclen; + if (blksiz == DIRBLKSIZ) + blksiz = 0; + uiop->uio_offset += DIRHDSIZ; uiop->uio_resid -= DIRHDSIZ; uiop->uio_iov->iov_base += DIRHDSIZ; uiop->uio_iov->iov_len -= DIRHDSIZ; cnp->cn_nameptr = uiop->uio_iov->iov_base; cnp->cn_namelen = len; - ndp->ni_vp = newvp; nfsm_mtouio(uiop, len); cp = uiop->uio_iov->iov_base; tlen -= len; - for (i = 0; i < tlen; i++) - *cp++ = '\0'; + *cp = '\0'; uiop->uio_iov->iov_base += tlen; uiop->uio_iov->iov_len -= tlen; + uiop->uio_offset += tlen; uiop->uio_resid -= tlen; + } else + nfsm_adv(nfsm_rndup(len)); + nfsm_dissect(tl, u_long *, 3 * NFSX_UNSIGNED); + if (bigenough) { + cookie.nfsuquad[0] = *tl++; + cookie.nfsuquad[1] = *tl++; + } else + tl += 2; + + /* + * Since the attributes are before the file handle + * (sigh), we must skip over the attributes and then + * come back and get them. + */ + attrflag = fxdr_unsigned(int, *tl); + if (attrflag) { + dpossav1 = dpos; + mdsav1 = md; + nfsm_adv(NFSX_V3FATTR); + nfsm_dissect(tl, u_long *, NFSX_UNSIGNED); + doit = fxdr_unsigned(int, *tl); + if (doit) { + nfsm_getfh(fhp, fhsize, 1); + if (NFS_CMPFH(dnp, fhp, fhsize)) { + VREF(vp); + newvp = vp; + np = dnp; + } else { + if (error = nfs_nget(vp->v_mount, fhp, + fhsize, &np)) + doit = 0; + else + newvp = NFSTOV(np); + } + } + if (doit) { + dpossav2 = dpos; + dpos = dpossav1; + mdsav2 = md; + md = mdsav1; + nfsm_loadattr(newvp, (struct vattr *)0); + dpos = dpossav2; + md = mdsav2; + dp->d_type = + IFTODT(VTTOIF(np->n_vattr.va_type)); + ndp->ni_vp = newvp; cnp->cn_hash = 0; - for (cp = cnp->cn_nameptr, i = 1; i <= len; i++, cp++) - cnp->cn_hash += (unsigned char)*cp * i; - if ((nmp->nm_flag & NFSMNT_NQLOOKLEASE) && - ltime > time.tv_sec) - nqnfs_clientlease(nmp, np, NQL_READ, - cachable, ltime, frev); + for (cp = cnp->cn_nameptr, i = 1; i <= len; + i++, cp++) + cnp->cn_hash += (unsigned char)*cp * i; if (cnp->cn_namelen <= NCHNAMLEN) cache_enter(ndp->ni_dvp, ndp->ni_vp, cnp); + } } else { - nfsm_adv(nfsm_rndup(len)); + /* Just skip over the file handle */ + nfsm_dissect(tl, u_long *, NFSX_UNSIGNED); + i = fxdr_unsigned(int, *tl); + nfsm_adv(nfsm_rndup(i)); } if (newvp != NULLVP) { - vrele(newvp); - newvp = NULLVP; + vrele(newvp); + newvp = NULLVP; } - nfsm_dissect(tl, u_long *, 2 * NFSX_UNSIGNED); - if (bigenough) - endoff = off = fxdr_unsigned(u_long, *tl++); - else - endoff = fxdr_unsigned(u_long, *tl++); + nfsm_dissect(tl, u_long *, NFSX_UNSIGNED); more_dirs = fxdr_unsigned(int, *tl); } /* @@ -1904,35 +2355,42 @@ nfs_readdirlookrpc(vp, uiop, cred) if (!more_dirs) { nfsm_dissect(tl, u_long *, NFSX_UNSIGNED); more_dirs = (fxdr_unsigned(int, *tl) == 0); - - /* - * If at EOF, cache directory offset - */ - if (!more_dirs) - VTONFS(vp)->n_direofoffset = endoff; } - if (uiop->uio_resid < tresid) - uiop->uio_offset = (off_t)off; - else - more_dirs = 0; m_freem(mrep); } /* * Fill last record, iff any, out to a multiple of NFS_DIRBLKSIZ * by increasing d_reclen for the last record. */ - if (uiop->uio_resid < tresid) { - len = uiop->uio_resid & (NFS_DIRBLKSIZ - 1); - if (len > 0) { - dp->d_reclen += len; - uiop->uio_iov->iov_base += len; - uiop->uio_iov->iov_len -= len; - uiop->uio_resid -= len; - } + if (blksiz > 0) { + left = DIRBLKSIZ - blksiz; + dp->d_reclen += left; + uiop->uio_iov->iov_base += left; + uiop->uio_iov->iov_len -= left; + uiop->uio_offset += left; + uiop->uio_resid -= left; + } + + /* + * We are now either at the end of the directory or have filled the + * block. + */ + if (bigenough) + dnp->n_direofoffset = uiop->uio_offset; + else { + if (uiop->uio_resid > 0) + printf("EEK! readdirplusrpc resid > 0\n"); + cookiep = nfs_getcookie(dnp, uiop->uio_offset, 1); + *cookiep = cookie; } nfsmout: - if (newvp != NULLVP) - vrele(newvp); + if (newvp != NULLVP) { + if (newvp == vp) + vrele(newvp); + else + vput(newvp); + newvp = NULLVP; + } return (error); } static char hextoasc[] = "0123456789abcdef"; @@ -1950,19 +2408,19 @@ nfs_sillyrename(dvp, vp, cnp) struct vnode *dvp, *vp; struct componentname *cnp; { - register struct nfsnode *np; register struct sillyrename *sp; + struct nfsnode *np; int error; short pid; cache_purge(dvp); np = VTONFS(vp); -#ifdef SILLYSEPARATE +#ifndef DIAGNOSTIC + if (vp->v_type == VDIR) + panic("nfs: sillyrename dir"); +#endif MALLOC(sp, struct sillyrename *, sizeof (struct sillyrename), M_NFSREQ, M_WAITOK); -#else - sp = &np->n_silly; -#endif sp->s_cred = crdup(cnp->cn_cred); sp->s_dvp = dvp; VREF(dvp); @@ -1977,65 +2435,149 @@ nfs_sillyrename(dvp, vp, cnp) sp->s_name[5] = hextoasc[(pid >> 12) & 0xf]; /* Try lookitups until we get one that isn't there */ - while (nfs_lookitup(sp, (nfsv2fh_t *)0, cnp->cn_proc) == 0) { + while (nfs_lookitup(dvp, sp->s_name, sp->s_namlen, sp->s_cred, + cnp->cn_proc, (struct nfsnode **)0) == 0) { sp->s_name[4]++; if (sp->s_name[4] > 'z') { error = EINVAL; goto bad; } } - error = nfs_renameit(dvp, cnp, sp); - if (error) + if (error = nfs_renameit(dvp, cnp, sp)) goto bad; - nfs_lookitup(sp, &np->n_fh, cnp->cn_proc); + error = nfs_lookitup(dvp, sp->s_name, sp->s_namlen, sp->s_cred, + cnp->cn_proc, &np); np->n_sillyrename = sp; return (0); bad: vrele(sp->s_dvp); crfree(sp->s_cred); -#ifdef SILLYSEPARATE free((caddr_t)sp, M_NFSREQ); -#endif return (error); } /* - * Look up a file name for silly rename stuff. - * Just like nfs_lookup() except that it doesn't load returned values - * into the nfsnode table. - * If fhp != NULL it copies the returned file handle out + * Look up a file name and optionally either update the file handle or + * allocate an nfsnode, depending on the value of npp. + * npp == NULL --> just do the lookup + * *npp == NULL --> allocate a new nfsnode and make sure attributes are + * handled too + * *npp != NULL --> update the file handle in the vnode */ int -nfs_lookitup(sp, fhp, procp) - register struct sillyrename *sp; - nfsv2fh_t *fhp; +nfs_lookitup(dvp, name, len, cred, procp, npp) + register struct vnode *dvp; + char *name; + int len; + struct ucred *cred; struct proc *procp; + struct nfsnode **npp; { - register struct vnode *vp = sp->s_dvp; register u_long *tl; register caddr_t cp; register long t1, t2; + struct vnode *newvp = (struct vnode *)0; + struct nfsnode *np, *dnp = VTONFS(dvp); caddr_t bpos, dpos, cp2; - int error = 0, isnq; + int error = 0, fhlen, attrflag; struct mbuf *mreq, *mrep, *md, *mb, *mb2; - long len; + nfsfh_t *nfhp; + int v3 = NFS_ISV3(dvp); - isnq = (VFSTONFS(vp->v_mount)->nm_flag & NFSMNT_NQNFS); nfsstats.rpccnt[NFSPROC_LOOKUP]++; - len = sp->s_namlen; - nfsm_reqhead(vp, NFSPROC_LOOKUP, NFSX_FH+NFSX_UNSIGNED+nfsm_rndup(len)); - if (isnq) { - nfsm_build(tl, u_long *, NFSX_UNSIGNED); - *tl = 0; + nfsm_reqhead(dvp, NFSPROC_LOOKUP, + NFSX_FH(v3) + NFSX_UNSIGNED + nfsm_rndup(len)); + nfsm_fhtom(dvp, v3); + nfsm_strtom(name, len, NFS_MAXNAMLEN); + nfsm_request(dvp, NFSPROC_LOOKUP, procp, cred); + if (npp && !error) { + nfsm_getfh(nfhp, fhlen, v3); + if (*npp) { + np = *npp; + if (np->n_fhsize > NFS_SMALLFH && fhlen <= NFS_SMALLFH) { + free((caddr_t)np->n_fhp, M_NFSBIGFH); + np->n_fhp = &np->n_fh; + } else if (np->n_fhsize <= NFS_SMALLFH && fhlen>NFS_SMALLFH) + np->n_fhp =(nfsfh_t *)malloc(fhlen,M_NFSBIGFH,M_WAITOK); + bcopy((caddr_t)nfhp, (caddr_t)np->n_fhp, fhlen); + np->n_fhsize = fhlen; + newvp = NFSTOV(np); + } else if (NFS_CMPFH(dnp, nfhp, fhlen)) { + VREF(dvp); + newvp = dvp; + } else { + error = nfs_nget(dvp->v_mount, nfhp, fhlen, &np); + if (error) { + m_freem(mrep); + return (error); + } + newvp = NFSTOV(np); + } + if (v3) { + nfsm_postop_attr(newvp, attrflag); + if (!attrflag && *npp == NULL) { + m_freem(mrep); + if (newvp == dvp) + vrele(newvp); + else + vput(newvp); + return (ENOENT); + } + } else + nfsm_loadattr(newvp, (struct vattr *)0); } - nfsm_fhtom(vp); - nfsm_strtom(sp->s_name, len, NFS_MAXNAMLEN); - nfsm_request(vp, NFSPROC_LOOKUP, procp, sp->s_cred); - if (fhp != NULL) { - if (isnq) - nfsm_dissect(tl, u_long *, NFSX_UNSIGNED); - nfsm_dissect(cp, caddr_t, NFSX_FH); - bcopy(cp, (caddr_t)fhp, NFSX_FH); + nfsm_reqdone; + if (npp && *npp == NULL) { + if (error) { + if (newvp) + if (newvp == dvp) + vrele(newvp); + else + vput(newvp); + } else + *npp = np; + } + return (error); +} + +/* + * Nfs Version 3 commit rpc + */ +int +nfs_commit(vp, offset, cnt, cred, procp) + register struct vnode *vp; + u_quad_t offset; + int cnt; + struct ucred *cred; + struct proc *procp; +{ + register caddr_t cp; + register u_long *tl; + register int t1, t2; + register struct nfsmount *nmp = VFSTONFS(vp->v_mount); + caddr_t bpos, dpos, cp2; + int error = 0, wccflag = NFSV3_WCCRATTR; + struct mbuf *mreq, *mrep, *md, *mb, *mb2; + + if ((nmp->nm_flag & NFSMNT_HASWRITEVERF) == 0) + return (0); + nfsstats.rpccnt[NFSPROC_COMMIT]++; + nfsm_reqhead(vp, NFSPROC_COMMIT, NFSX_FH(1)); + nfsm_fhtom(vp, 1); + nfsm_build(tl, u_long *, 3 * NFSX_UNSIGNED); + txdr_hyper(&offset, tl); + tl += 2; + *tl = txdr_unsigned(cnt); + nfsm_request(vp, NFSPROC_COMMIT, procp, cred); + nfsm_wcc_data(vp, wccflag); + if (!error) { + nfsm_dissect(tl, u_long *, NFSX_V3WRITEVERF); + if (bcmp((caddr_t)nmp->nm_verf, (caddr_t)tl, + NFSX_V3WRITEVERF)) { + bcopy((caddr_t)tl, (caddr_t)nmp->nm_verf, + NFSX_V3WRITEVERF); + error = NFSERR_STALEWRITEVERF; + } } nfsm_reqdone; return (error); @@ -2044,9 +2586,7 @@ nfs_lookitup(sp, fhp, procp) /* * Kludge City.. * - make nfs_bmap() essentially a no-op that does no translation - * - do nfs_strategy() by faking physical I/O with nfs_readrpc/nfs_writerpc - * after mapping the physical addresses into Kernel Virtual space in the - * nfsiobuf area. + * - do nfs_strategy() by doing I/O with nfs_readrpc/nfs_writerpc * (Maybe I could use the process's page mapping, but I was concerned that * Kernel Write might not be enabled and also figured copyout() would do * a lot more work than bcopy() and also it currently happens in the @@ -2088,8 +2628,8 @@ nfs_strategy(ap) struct proc *p; int error = 0; - if ((bp->b_flags & (B_PHYS|B_ASYNC)) == (B_PHYS|B_ASYNC)) - panic("nfs physio/async"); + if (bp->b_flags & B_PHYS) + panic("nfs physio"); if (bp->b_flags & B_ASYNC) p = (struct proc *)0; else @@ -2129,9 +2669,7 @@ nfs_mmap(ap) } /* - * Flush all the blocks associated with a vnode. - * Walk through the buffer pool and push any dirty pages - * associated with the vnode. + * fsync vnode op. Just call nfs_flush() with commit == 1. */ /* ARGSUSED */ int @@ -2144,29 +2682,124 @@ nfs_fsync(ap) struct proc * a_p; } */ *ap; { - register struct vnode *vp = ap->a_vp; + + return (nfs_flush(ap->a_vp, ap->a_cred, ap->a_waitfor, ap->a_p, 1)); +} + +/* + * Flush all the blocks associated with a vnode. + * Walk through the buffer pool and push any dirty pages + * associated with the vnode. + */ +int +nfs_flush(vp, cred, waitfor, p, commit) + register struct vnode *vp; + struct ucred *cred; + int waitfor; + struct proc *p; + int commit; +{ register struct nfsnode *np = VTONFS(vp); register struct buf *bp; + register int i; struct buf *nbp; - struct nfsmount *nmp; - int s, error = 0, slptimeo = 0, slpflag = 0; + struct nfsmount *nmp = VFSTONFS(vp->v_mount); + int s, error = 0, slptimeo = 0, slpflag = 0, retv, bvecpos; + int passone = 1; + u_quad_t off = (u_quad_t)-1, endoff = 0, toff; +#ifndef NFS_COMMITBVECSIZ +#define NFS_COMMITBVECSIZ 20 +#endif + struct buf *bvec[NFS_COMMITBVECSIZ]; - nmp = VFSTONFS(vp->v_mount); if (nmp->nm_flag & NFSMNT_INT) slpflag = PCATCH; + if (!commit) + passone = 0; + /* + * A b_flags == (B_DELWRI | B_NEEDCOMMIT) block has been written to the + * server, but nas not been committed to stable storage on the server + * yet. On the first pass, the byte range is worked out and the commit + * rpc is done. On the second pass, nfs_writebp() is called to do the + * job. + */ +again: + bvecpos = 0; + if (NFS_ISV3(vp) && commit) { + s = splbio(); + for (bp = vp->v_dirtyblkhd.lh_first; bp; bp = nbp) { + nbp = bp->b_vnbufs.le_next; + if (bvecpos >= NFS_COMMITBVECSIZ) + break; + if ((bp->b_flags & (B_BUSY | B_DELWRI | B_NEEDCOMMIT)) + != (B_DELWRI | B_NEEDCOMMIT)) + continue; + bremfree(bp); + bp->b_flags |= (B_BUSY | B_WRITEINPROG); + vfs_busy_pages(bp, 1); + /* + * A list of these buffers is kept so that the + * second loop knows which buffers have actually + * been committed. This is necessary, since there + * may be a race between the commit rpc and new + * uncommitted writes on the file. + */ + bvec[bvecpos++] = bp; + toff = ((u_quad_t)bp->b_blkno) * DEV_BSIZE + + bp->b_dirtyoff; + if (toff < off) + off = toff; + toff += (u_quad_t)(bp->b_dirtyend - bp->b_dirtyoff); + if (toff > endoff) + endoff = toff; + } + splx(s); + } + if (bvecpos > 0) { + /* + * Commit data on the server, as required. + */ + retv = nfs_commit(vp, off, (int)(endoff - off), cred, p); + if (retv == NFSERR_STALEWRITEVERF) + nfs_clearcommit(vp->v_mount); + /* + * Now, either mark the blocks I/O done or mark the + * blocks dirty, depending on whether the commit + * succeeded. + */ + for (i = 0; i < bvecpos; i++) { + bp = bvec[i]; + bp->b_flags &= ~(B_NEEDCOMMIT | B_WRITEINPROG); + if (retv) { + brelse(bp); + vfs_unbusy_pages(bp); + } else { + vp->v_numoutput++; + bp->b_flags |= B_ASYNC; + bp->b_flags &= ~(B_READ|B_DONE|B_ERROR|B_DELWRI); + bp->b_dirtyoff = bp->b_dirtyend = 0; + reassignbuf(bp, vp); + biodone(bp); + } + } + } + + /* + * Start/do any write(s) that are required. + */ loop: s = splbio(); for (bp = vp->v_dirtyblkhd.lh_first; bp; bp = nbp) { nbp = bp->b_vnbufs.le_next; if (bp->b_flags & B_BUSY) { - if (ap->a_waitfor != MNT_WAIT) + if (waitfor != MNT_WAIT || passone) continue; bp->b_flags |= B_WANTED; error = tsleep((caddr_t)bp, slpflag | (PRIBIO + 1), "nfsfsync", slptimeo); splx(s); if (error) { - if (nfs_sigintr(nmp, (struct nfsreq *)0, ap->a_p)) + if (nfs_sigintr(nmp, (struct nfsreq *)0, p)) return (EINTR); if (slpflag == PCATCH) { slpflag = 0; @@ -2177,21 +2810,29 @@ loop: } if ((bp->b_flags & B_DELWRI) == 0) panic("nfs_fsync: not dirty"); + if ((passone || !commit) && (bp->b_flags & B_NEEDCOMMIT)) + continue; bremfree(bp); - bp->b_flags |= B_BUSY; + if (passone || !commit) + bp->b_flags |= (B_BUSY|B_ASYNC); + else + bp->b_flags |= (B_BUSY|B_ASYNC|B_WRITEINPROG|B_NEEDCOMMIT); splx(s); - bp->b_flags |= B_ASYNC; VOP_BWRITE(bp); goto loop; } splx(s); - if (ap->a_waitfor == MNT_WAIT) { + if (passone) { + passone = 0; + goto again; + } + if (waitfor == MNT_WAIT) { while (vp->v_numoutput) { vp->v_flag |= VBWAIT; error = tsleep((caddr_t)&vp->v_numoutput, slpflag | (PRIBIO + 1), "nfsfsync", slptimeo); if (error) { - if (nfs_sigintr(nmp, (struct nfsreq *)0, ap->a_p)) + if (nfs_sigintr(nmp, (struct nfsreq *)0, p)) return (EINTR); if (slpflag == PCATCH) { slpflag = 0; @@ -2199,7 +2840,7 @@ loop: } } } - if (vp->v_dirtyblkhd.lh_first) { + if (vp->v_dirtyblkhd.lh_first && commit) { goto loop; } } @@ -2207,14 +2848,15 @@ loop: error = np->n_error; np->n_flag &= ~NWRITEERR; } + np->n_flag &= ~NMODIFIED; return (error); } /* * Return POSIX pathconf information applicable to nfs. * - * Currently the NFS protocol does not support getting such - * information from the remote server. + * The NFS V2 protocol doesn't support this, so just return EINVAL + * for V2. */ /* ARGSUSED */ int @@ -2243,6 +2885,7 @@ nfs_advlock(ap) int a_flags; } */ *ap; { +#ifdef __FreeBSD__ register struct nfsnode *np = VTONFS(ap->a_vp); /* @@ -2251,6 +2894,9 @@ nfs_advlock(ap) * that this is a local lock. */ return (lf_advlock(ap, &(np->n_lockf), np->n_size)); +#else + return (EOPNOTSUPP); +#endif } /* @@ -2363,6 +3009,81 @@ nfs_update(ap) } /* + * Just call nfs_writebp() with the force argument set to 1. + */ +int +nfs_bwrite(ap) + struct vop_bwrite_args /* { + struct vnode *a_bp; + } */ *ap; +{ + + return (nfs_writebp(ap->a_bp, 1)); +} + +/* + * This is a clone of vn_bwrite(), except that B_WRITEINPROG isn't set unless + * the force flag is one and it also handles the B_NEEDCOMMIT flag. + */ +int +nfs_writebp(bp, force) + register struct buf *bp; + int force; +{ + register int oldflags = bp->b_flags, retv = 1; + off_t off; + + if(!(bp->b_flags & B_BUSY)) + panic("bwrite: buffer is not busy???"); + + bp->b_flags &= ~(B_READ|B_DONE|B_ERROR|B_DELWRI); + + if ((oldflags & (B_ASYNC|B_DELWRI)) == (B_ASYNC|B_DELWRI)) { + reassignbuf(bp, bp->b_vp); + } + + bp->b_vp->v_numoutput++; + curproc->p_stats->p_ru.ru_oublock++; + + /* + * If B_NEEDCOMMIT is set, a commit rpc may do the trick. If not + * an actual write will have to be scheduled via. VOP_STRATEGY(). + * If B_WRITEINPROG is already set, then push it with a write anyhow. + */ + if (oldflags & (B_NEEDCOMMIT | B_WRITEINPROG) == B_NEEDCOMMIT) { + off = ((u_quad_t)bp->b_blkno) * DEV_BSIZE + bp->b_dirtyoff; + bp->b_flags |= B_WRITEINPROG; + retv = nfs_commit(bp->b_vp, off, bp->b_dirtyend-bp->b_dirtyoff, + bp->b_wcred, bp->b_proc); + bp->b_flags &= ~B_WRITEINPROG; + if (!retv) { + bp->b_dirtyoff = bp->b_dirtyend = 0; + bp->b_flags &= ~B_NEEDCOMMIT; + biodone(bp); + } else if (retv == NFSERR_STALEWRITEVERF) + nfs_clearcommit(bp->b_vp->v_mount); + } + if (retv) { + if (force) + bp->b_flags |= B_WRITEINPROG; + vfs_busy_pages(bp, 1); + VOP_STRATEGY(bp); + } + + if( (oldflags & B_ASYNC) == 0) { + int rtval = biowait(bp); + + if (oldflags & B_DELWRI) { + reassignbuf(bp, bp->b_vp); + } + brelse(bp); + return (rtval); + } + + return (0); +} + +/* * nfs special file access vnode op. * Essentially just get vattr and then imitate iaccess() since the device is * local to the client. @@ -2409,7 +3130,8 @@ nfsspec_access(ap) found: ; } - return ((vap->va_mode & mode) == mode ? 0 : EACCES); + error = (vap->va_mode & mode) == mode ? 0 : EACCES; + return (error); } /* @@ -2430,7 +3152,8 @@ nfsspec_read(ap) * Set access flag. */ np->n_flag |= NACC; - np->n_atim = time; + np->n_atim.ts_sec = time.tv_sec; + np->n_atim.ts_nsec = time.tv_usec * 1000; return (VOCALL(spec_vnodeop_p, VOFFSET(vop_read), ap)); } @@ -2452,7 +3175,8 @@ nfsspec_write(ap) * Set update flag. */ np->n_flag |= NUPD; - np->n_mtim = time; + np->n_mtim.ts_sec = time.tv_sec; + np->n_mtim.ts_nsec = time.tv_usec * 1000; return (VOCALL(spec_vnodeop_p, VOFFSET(vop_write), ap)); } @@ -2479,16 +3203,10 @@ nfsspec_close(ap) if (vp->v_usecount == 1 && (vp->v_mount->mnt_flag & MNT_RDONLY) == 0) { VATTR_NULL(&vattr); - if (np->n_flag & NACC) { - vattr.va_atime.ts_sec = np->n_atim.tv_sec; - vattr.va_atime.ts_nsec = - np->n_atim.tv_usec * 1000; - } - if (np->n_flag & NUPD) { - vattr.va_mtime.ts_sec = np->n_mtim.tv_sec; - vattr.va_mtime.ts_nsec = - np->n_mtim.tv_usec * 1000; - } + if (np->n_flag & NACC) + vattr.va_atime = np->n_atim; + if (np->n_flag & NUPD) + vattr.va_mtime = np->n_mtim; (void)VOP_SETATTR(vp, &vattr, ap->a_cred, ap->a_p); } } @@ -2513,7 +3231,8 @@ nfsfifo_read(ap) * Set access flag. */ np->n_flag |= NACC; - np->n_atim = time; + np->n_atim.ts_sec = time.tv_sec; + np->n_atim.ts_nsec = time.tv_usec * 1000; return (VOCALL(fifo_vnodeop_p, VOFFSET(vop_read), ap)); } @@ -2535,7 +3254,8 @@ nfsfifo_write(ap) * Set update flag. */ np->n_flag |= NUPD; - np->n_mtim = time; + np->n_mtim.ts_sec = time.tv_sec; + np->n_mtim.ts_nsec = time.tv_usec * 1000; return (VOCALL(fifo_vnodeop_p, VOFFSET(vop_write), ap)); } @@ -2558,24 +3278,22 @@ nfsfifo_close(ap) struct vattr vattr; if (np->n_flag & (NACC | NUPD)) { - if (np->n_flag & NACC) - np->n_atim = time; - if (np->n_flag & NUPD) - np->n_mtim = time; + if (np->n_flag & NACC) { + np->n_atim.ts_sec = time.tv_sec; + np->n_atim.ts_nsec = time.tv_usec * 1000; + } + if (np->n_flag & NUPD) { + np->n_mtim.ts_sec = time.tv_sec; + np->n_mtim.ts_nsec = time.tv_usec * 1000; + } np->n_flag |= NCHG; if (vp->v_usecount == 1 && (vp->v_mount->mnt_flag & MNT_RDONLY) == 0) { VATTR_NULL(&vattr); - if (np->n_flag & NACC) { - vattr.va_atime.ts_sec = np->n_atim.tv_sec; - vattr.va_atime.ts_nsec = - np->n_atim.tv_usec * 1000; - } - if (np->n_flag & NUPD) { - vattr.va_mtime.ts_sec = np->n_mtim.tv_sec; - vattr.va_mtime.ts_nsec = - np->n_mtim.tv_usec * 1000; - } + if (np->n_flag & NACC) + vattr.va_atime = np->n_atim; + if (np->n_flag & NUPD) + vattr.va_mtime = np->n_mtim; (void)VOP_SETATTR(vp, &vattr, ap->a_cred, ap->a_p); } } diff --git a/sys/nfs/nfsdiskless.h b/sys/nfs/nfsdiskless.h index 5df85e0..f7371dc 100644 --- a/sys/nfs/nfsdiskless.h +++ b/sys/nfs/nfsdiskless.h @@ -34,7 +34,7 @@ * SUCH DAMAGE. * * @(#)nfsdiskless.h 8.1 (Berkeley) 6/10/93 - * $Id: nfsdiskless.h,v 1.3 1994/08/05 09:28:49 davidg Exp $ + * $Id: nfsdiskless.h,v 1.4 1994/08/21 06:50:09 paul Exp $ */ #ifndef _NFS_NFSDISKLESS_H_ @@ -52,17 +52,45 @@ * NB: All fields are stored in net byte order to avoid hassles with * client/server byte ordering differences. */ + +/* + * I have defined a new structure that can handle an NFS Version 3 file handle + * but the kernel still expects the old Version 2 one to be provided. The + * changes required in nfs_vfsops.c for using the new are documented there in + * comments. (I felt that breaking network booting code by changing this + * structure would not be prudent at this time, since almost all servers are + * still Version 2 anyhow.) + */ +struct nfsv3_diskless { + struct ifaliasreq myif; /* Default interface */ + struct sockaddr_in mygateway; /* Default gateway */ + struct nfs_args swap_args; /* Mount args for swap file */ + int swap_fhsize; /* Size of file handle */ + u_char swap_fh[NFSX_V3FHMAX]; /* Swap file's file handle */ + struct sockaddr_in swap_saddr; /* Address of swap server */ + char swap_hostnam[MNAMELEN]; /* Host name for mount pt */ + int swap_nblks; /* Size of server swap file */ + struct ucred swap_ucred; /* Swap credentials */ + struct nfs_args root_args; /* Mount args for root fs */ + int root_fhsize; /* Size of root file handle */ + u_char root_fh[NFSX_V3FHMAX]; /* File handle of root dir */ + struct sockaddr_in root_saddr; /* Address of root server */ + char root_hostnam[MNAMELEN]; /* Host name for mount pt */ + long root_time; /* Timestamp of root fs */ + char my_hostnam[MAXHOSTNAMELEN]; /* Client host name */ +}; + struct nfs_diskless { struct ifaliasreq myif; /* Default interface */ struct sockaddr_in mygateway; /* Default gateway */ struct nfs_args swap_args; /* Mount args for swap file */ - u_char swap_fh[NFS_FHSIZE]; /* Swap file's file handle */ + u_char swap_fh[NFSX_V2FH]; /* Swap file's file handle */ struct sockaddr_in swap_saddr; /* Address of swap server */ char swap_hostnam[MNAMELEN]; /* Host name for mount pt */ int swap_nblks; /* Size of server swap file */ struct ucred swap_ucred; /* Swap credentials */ struct nfs_args root_args; /* Mount args for root fs */ - u_char root_fh[NFS_FHSIZE]; /* File handle of root dir */ + u_char root_fh[NFSX_V2FH]; /* File handle of root dir */ struct sockaddr_in root_saddr; /* Address of root server */ char root_hostnam[MNAMELEN]; /* Host name for mount pt */ long root_time; /* Timestamp of root fs */ diff --git a/sys/nfs/nfsm_subs.h b/sys/nfs/nfsm_subs.h index 030a985..9e89c57 100644 --- a/sys/nfs/nfsm_subs.h +++ b/sys/nfs/nfsm_subs.h @@ -34,12 +34,13 @@ * SUCH DAMAGE. * * @(#)nfsm_subs.h 8.1 (Berkeley) 6/16/93 - * $Id: nfsm_subs.h,v 1.5 1994/10/17 17:47:42 phk Exp $ + * $Id: nfsm_subs.h,v 1.6 1995/05/30 08:12:51 rgrimes Exp $ */ #ifndef _NFS_NFSM_SUBS_H_ #define _NFS_NFSM_SUBS_H_ + /* * These macros do strange and peculiar things to mbuf chains for * the assistance of the nfs code. To attempt to use them for any @@ -90,49 +91,153 @@ extern struct mbuf *nfsm_reqh(); mb->m_len += (s); \ bpos += (s); } -#define nfsm_dissect(a,c,s) \ +#define nfsm_dissect(a, c, s) \ { t1 = mtod(md, caddr_t)+md->m_len-dpos; \ if (t1 >= (s)) { \ (a) = (c)(dpos); \ dpos += (s); \ + } else if (t1 = nfsm_disct(&md, &dpos, (s), t1, &cp2)) { \ + error = t1; \ + m_freem(mrep); \ + goto nfsmout; \ } else { \ - error = nfsm_disct(&md, &dpos, (s), t1, &cp2); \ - if (error) { \ - m_freem(mrep); \ + (a) = (c)cp2; \ + } } + +#define nfsm_fhtom(v, v3) \ + { if (v3) { \ + t2 = nfsm_rndup(VTONFS(v)->n_fhsize) + NFSX_UNSIGNED; \ + if (t2 <= M_TRAILINGSPACE(mb)) { \ + nfsm_build(tl, u_long *, t2); \ + *tl++ = txdr_unsigned(VTONFS(v)->n_fhsize); \ + *(tl + ((t2>>2) - 2)) = 0; \ + bcopy((caddr_t)VTONFS(v)->n_fhp,(caddr_t)tl, \ + VTONFS(v)->n_fhsize); \ + } else if (t2 = nfsm_strtmbuf(&mb, &bpos, \ + (caddr_t)VTONFS(v)->n_fhp, VTONFS(v)->n_fhsize)) { \ + error = t2; \ + m_freem(mreq); \ goto nfsmout; \ - } else { \ - (a) = (c)cp2; \ } \ + } else { \ + nfsm_build(cp, caddr_t, NFSX_V2FH); \ + bcopy((caddr_t)VTONFS(v)->n_fhp, cp, NFSX_V2FH); \ } } -#define nfsm_fhtom(v) \ - nfsm_build(cp,caddr_t,NFSX_FH); \ - bcopy((caddr_t)&(VTONFS(v)->n_fh), cp, NFSX_FH) +#define nfsm_srvfhtom(f, v3) \ + { if (v3) { \ + nfsm_build(tl, u_long *, NFSX_UNSIGNED + NFSX_V3FH); \ + *tl++ = txdr_unsigned(NFSX_V3FH); \ + bcopy((caddr_t)(f), (caddr_t)tl, NFSX_V3FH); \ + } else { \ + nfsm_build(cp, caddr_t, NFSX_V2FH); \ + bcopy((caddr_t)(f), cp, NFSX_V2FH); \ + } } -#define nfsm_srvfhtom(f) \ - nfsm_build(cp,caddr_t,NFSX_FH); \ - bcopy((caddr_t)(f), cp, NFSX_FH) +#define nfsm_srvpostop_fh(f) \ + { nfsm_build(tl, u_long *, 2 * NFSX_UNSIGNED + NFSX_V3FH); \ + *tl++ = nfs_true; \ + *tl++ = txdr_unsigned(NFSX_V3FH); \ + bcopy((caddr_t)(f), (caddr_t)tl, NFSX_V3FH); \ + } -#define nfsm_mtofh(d,v) \ - { struct nfsnode *np; nfsv2fh_t *fhp; \ - nfsm_dissect(fhp,nfsv2fh_t *,NFSX_FH); \ - error = nfs_nget((d)->v_mount, fhp, &np); \ - if (error) { \ - m_freem(mrep); \ - goto nfsmout; \ +#define nfsm_mtofh(d, v, v3, f) \ + { struct nfsnode *ttnp; nfsfh_t *ttfhp; int ttfhsize; \ + if (v3) { \ + nfsm_dissect(tl, u_long *, NFSX_UNSIGNED); \ + (f) = fxdr_unsigned(int, *tl); \ + } else \ + (f) = 1; \ + if (f) { \ + nfsm_getfh(ttfhp, ttfhsize, (v3)); \ + if (t1 = nfs_nget((d)->v_mount, ttfhp, ttfhsize, \ + &ttnp)) { \ + error = t1; \ + m_freem(mrep); \ + goto nfsmout; \ + } \ + (v) = NFSTOV(ttnp); \ + } \ + if (v3) { \ + nfsm_dissect(tl, u_long *, NFSX_UNSIGNED); \ + if (f) \ + (f) = fxdr_unsigned(int, *tl); \ + else if (fxdr_unsigned(int, *tl)) \ + nfsm_adv(NFSX_V3FATTR); \ } \ - (v) = NFSTOV(np); \ - nfsm_loadattr(v, (struct vattr *)0); \ + if (f) \ + nfsm_loadattr((v), (struct vattr *)0); \ } -#define nfsm_loadattr(v,a) \ - { struct vnode *tvp = (v); \ - error = nfs_loadattrcache(&tvp, &md, &dpos, (a)); \ - if (error) { \ +#define nfsm_getfh(f, s, v3) \ + { if (v3) { \ + nfsm_dissect(tl, u_long *, NFSX_UNSIGNED); \ + if (((s) = fxdr_unsigned(int, *tl)) <= 0 || \ + (s) > NFSX_V3FHMAX) { \ + m_freem(mrep); \ + error = EBADRPC; \ + goto nfsmout; \ + } \ + } else \ + (s) = NFSX_V2FH; \ + nfsm_dissect((f), nfsfh_t *, nfsm_rndup(s)); } + +#define nfsm_loadattr(v, a) \ + { struct vnode *ttvp = (v); \ + if (t1 = nfs_loadattrcache(&ttvp, &md, &dpos, (a))) { \ + error = t1; \ m_freem(mrep); \ goto nfsmout; \ } \ - (v) = tvp; } + (v) = ttvp; } + +#define nfsm_postop_attr(v, f) \ + { struct vnode *ttvp = (v); \ + nfsm_dissect(tl, u_long *, NFSX_UNSIGNED); \ + if ((f) = fxdr_unsigned(int, *tl)) { \ + if (t1 = nfs_loadattrcache(&ttvp, &md, &dpos, \ + (struct vattr *)0)) { \ + error = t1; \ + (f) = 0; \ + m_freem(mrep); \ + goto nfsmout; \ + } \ + (v) = ttvp; \ + } } + +/* Used as (f) for nfsm_wcc_data() */ +#define NFSV3_WCCRATTR 0 +#define NFSV3_WCCCHK 1 + +#define nfsm_wcc_data(v, f) \ + { int ttattrf, ttretf = 0; \ + nfsm_dissect(tl, u_long *, NFSX_UNSIGNED); \ + if (*tl == nfs_true) { \ + nfsm_dissect(tl, u_long *, 6 * NFSX_UNSIGNED); \ + if (f) \ + ttretf = (VTONFS(v)->n_mtime == \ + fxdr_unsigned(u_long, *(tl + 2))); \ + } \ + nfsm_postop_attr((v), ttattrf); \ + if (f) { \ + (f) = ttretf; \ + } else { \ + (f) = ttattrf; \ + } } + +#define nfsm_v3sattr(s, a, u, g) \ + { (s)->sa_modetrue = nfs_true; \ + (s)->sa_mode = vtonfsv3_mode((a)->va_mode); \ + (s)->sa_uidtrue = nfs_true; \ + (s)->sa_uid = txdr_unsigned(u); \ + (s)->sa_gidtrue = nfs_true; \ + (s)->sa_gid = txdr_unsigned(g); \ + (s)->sa_sizefalse = nfs_false; \ + (s)->sa_atimetype = txdr_unsigned(NFSV3SATTRTIME_TOCLIENT); \ + txdr_nfsv3time(&(a)->va_atime, &(s)->sa_atime); \ + (s)->sa_mtimetype = txdr_unsigned(NFSV3SATTRTIME_TOCLIENT); \ + txdr_nfsv3time(&(a)->va_mtime, &(s)->sa_mtime); \ + } #define nfsm_strsiz(s,m) \ { nfsm_dissect(tl,u_long *,NFSX_UNSIGNED); \ @@ -149,16 +254,27 @@ extern struct mbuf *nfsm_reqh(); nfsm_reply(0); \ } } +#define nfsm_srvnamesiz(s) \ + { nfsm_dissect(tl,u_long *,NFSX_UNSIGNED); \ + if (((s) = fxdr_unsigned(long,*tl)) > NFS_MAXNAMLEN) \ + error = NFSERR_NAMETOL; \ + if ((s) <= 0) \ + error = EBADRPC; \ + if (error) \ + nfsm_reply(0); \ + } + #define nfsm_mtouio(p,s) \ if ((s) > 0 && \ - (error = nfsm_mbuftouio(&md,(p),(s),&dpos))) { \ + (t1 = nfsm_mbuftouio(&md,(p),(s),&dpos))) { \ + error = t1; \ m_freem(mrep); \ goto nfsmout; \ } #define nfsm_uiotom(p,s) \ - error = nfsm_uiotombuf((p),&mb,(s),&bpos); \ - if (error) { \ + if (t1 = nfsm_uiotombuf((p),&mb,(s),&bpos)) { \ + error = t1; \ m_freem(mreq); \ goto nfsmout; \ } @@ -172,10 +288,13 @@ extern struct mbuf *nfsm_reqh(); #define nfsm_rndup(a) (((a)+3)&(~0x3)) #define nfsm_request(v, t, p, c) \ - error = nfs_request((v), mreq, (t), (p), \ - (c), &mrep, &md, &dpos); \ - if (error) \ - goto nfsmout + if (error = nfs_request((v), mreq, (t), (p), \ + (c), &mrep, &md, &dpos)) { \ + if (error & NFSERR_RETERR) \ + error &= ~NFSERR_RETERR; \ + else \ + goto nfsmout; \ + } #define nfsm_strtom(a,s,m) \ if ((s) > (m)) { \ @@ -189,12 +308,10 @@ extern struct mbuf *nfsm_reqh(); *tl++ = txdr_unsigned(s); \ *(tl+((t2>>2)-2)) = 0; \ bcopy((caddr_t)(a), (caddr_t)tl, (s)); \ - } else { \ - error = nfsm_strtmbuf(&mb, &bpos, (a), (s)); \ - if (error) { \ - m_freem(mreq); \ - goto nfsmout; \ - } \ + } else if (t2 = nfsm_strtmbuf(&mb, &bpos, (a), (s))) { \ + error = t2; \ + m_freem(mreq); \ + goto nfsmout; \ } #define nfsm_srvdone \ @@ -204,33 +321,53 @@ extern struct mbuf *nfsm_reqh(); #define nfsm_reply(s) \ { \ nfsd->nd_repstat = error; \ - if (error) \ - (void) nfs_rephead(0, nfsd, error, cache, &frev, \ + if (error && !(nfsd->nd_flag & ND_NFSV3)) \ + (void) nfs_rephead(0, nfsd, slp, error, cache, &frev, \ mrq, &mb, &bpos); \ else \ - (void) nfs_rephead((s), nfsd, error, cache, &frev, \ + (void) nfs_rephead((s), nfsd, slp, error, cache, &frev, \ mrq, &mb, &bpos); \ m_freem(mrep); \ mreq = *mrq; \ - if (error) \ + if (error && (!(nfsd->nd_flag & ND_NFSV3) || \ + error == EBADRPC)) \ return(0); \ } +#define nfsm_writereply(s, v3) \ + { \ + nfsd->nd_repstat = error; \ + if (error && !(v3)) \ + (void) nfs_rephead(0, nfsd, slp, error, cache, &frev, \ + &mreq, &mb, &bpos); \ + else \ + (void) nfs_rephead((s), nfsd, slp, error, cache, &frev, \ + &mreq, &mb, &bpos); \ + } + #define nfsm_adv(s) \ - t1 = mtod(md, caddr_t)+md->m_len-dpos; \ + { t1 = mtod(md, caddr_t)+md->m_len-dpos; \ if (t1 >= (s)) { \ dpos += (s); \ - } else { \ - error = nfs_adv(&md, &dpos, (s), t1); \ - if (error) { \ - m_freem(mrep); \ - goto nfsmout; \ - } \ - } + } else if (t1 = nfs_adv(&md, &dpos, (s), t1)) { \ + error = t1; \ + m_freem(mrep); \ + goto nfsmout; \ + } } #define nfsm_srvmtofh(f) \ - nfsm_dissect(tl, u_long *, NFSX_FH); \ - bcopy((caddr_t)tl, (caddr_t)f, NFSX_FH) + { if (nfsd->nd_flag & ND_NFSV3) { \ + nfsm_dissect(tl, u_long *, NFSX_UNSIGNED); \ + if (fxdr_unsigned(int, *tl) != NFSX_V3FH) { \ + error = EBADRPC; \ + nfsm_reply(0); \ + } \ + } \ + nfsm_dissect(tl, u_long *, NFSX_V3FH); \ + bcopy((caddr_t)tl, (caddr_t)(f), NFSX_V3FH); \ + if ((nfsd->nd_flag & ND_NFSV3) == 0) \ + nfsm_adv(NFSX_V2FH - NFSX_V3FH); \ + } #define nfsm_clget \ if (bp >= be) { \ @@ -246,41 +383,57 @@ extern struct mbuf *nfsm_reqh(); } \ tl = (u_long *)bp -#define nfsm_srvfillattr \ - fp->fa_type = vtonfs_type(vap->va_type); \ - fp->fa_mode = vtonfs_mode(vap->va_type, vap->va_mode); \ - fp->fa_nlink = txdr_unsigned(vap->va_nlink); \ - fp->fa_uid = txdr_unsigned(vap->va_uid); \ - fp->fa_gid = txdr_unsigned(vap->va_gid); \ - if (nfsd->nd_nqlflag == NQL_NOVAL) { \ - fp->fa_nfsblocksize = txdr_unsigned(vap->va_blocksize); \ - if (vap->va_type == VFIFO) \ - fp->fa_nfsrdev = 0xffffffff; \ - else \ - fp->fa_nfsrdev = txdr_unsigned(vap->va_rdev); \ - fp->fa_nfsfsid = txdr_unsigned(vap->va_fsid); \ - fp->fa_nfsfileid = txdr_unsigned(vap->va_fileid); \ - fp->fa_nfssize = txdr_unsigned(vap->va_size); \ - fp->fa_nfsblocks = txdr_unsigned(vap->va_bytes / NFS_FABLKSIZE); \ - txdr_nfstime(&vap->va_atime, &fp->fa_nfsatime); \ - txdr_nfstime(&vap->va_mtime, &fp->fa_nfsmtime); \ - txdr_nfstime(&vap->va_ctime, &fp->fa_nfsctime); \ - } else { \ - fp->fa_nqblocksize = txdr_unsigned(vap->va_blocksize); \ - if (vap->va_type == VFIFO) \ - fp->fa_nqrdev = 0xffffffff; \ - else \ - fp->fa_nqrdev = txdr_unsigned(vap->va_rdev); \ - fp->fa_nqfsid = txdr_unsigned(vap->va_fsid); \ - fp->fa_nqfileid = txdr_unsigned(vap->va_fileid); \ - txdr_hyper(&vap->va_size, &fp->fa_nqsize); \ - txdr_hyper(&vap->va_bytes, &fp->fa_nqbytes); \ - txdr_nqtime(&vap->va_atime, &fp->fa_nqatime); \ - txdr_nqtime(&vap->va_mtime, &fp->fa_nqmtime); \ - txdr_nqtime(&vap->va_ctime, &fp->fa_nqctime); \ - fp->fa_nqflags = txdr_unsigned(vap->va_flags); \ - fp->fa_nqgen = txdr_unsigned(vap->va_gen); \ - txdr_hyper(&vap->va_filerev, &fp->fa_nqfilerev); \ - } +#define nfsm_srvfillattr(a, f) \ + nfsm_srvfattr(nfsd, (a), (f)) + +#define nfsm_srvwcc_data(br, b, ar, a) \ + nfsm_srvwcc(nfsd, (br), (b), (ar), (a), &mb, &bpos) + +#define nfsm_srvpostop_attr(r, a) \ + nfsm_srvpostopattr(nfsd, (r), (a), &mb, &bpos) + +#define nfsm_srvsattr(a) \ + { nfsm_dissect(tl, u_long *, NFSX_UNSIGNED); \ + if (*tl == nfs_true) { \ + nfsm_dissect(tl, u_long *, NFSX_UNSIGNED); \ + (a)->va_mode = nfstov_mode(*tl); \ + } \ + nfsm_dissect(tl, u_long *, NFSX_UNSIGNED); \ + if (*tl == nfs_true) { \ + nfsm_dissect(tl, u_long *, NFSX_UNSIGNED); \ + (a)->va_uid = fxdr_unsigned(uid_t, *tl); \ + } \ + nfsm_dissect(tl, u_long *, NFSX_UNSIGNED); \ + if (*tl == nfs_true) { \ + nfsm_dissect(tl, u_long *, NFSX_UNSIGNED); \ + (a)->va_gid = fxdr_unsigned(gid_t, *tl); \ + } \ + nfsm_dissect(tl, u_long *, NFSX_UNSIGNED); \ + if (*tl == nfs_true) { \ + nfsm_dissect(tl, u_long *, 2 * NFSX_UNSIGNED); \ + fxdr_hyper(tl, &(a)->va_size); \ + } \ + nfsm_dissect(tl, u_long *, NFSX_UNSIGNED); \ + switch (fxdr_unsigned(int, *tl)) { \ + case NFSV3SATTRTIME_TOCLIENT: \ + nfsm_dissect(tl, u_long *, 2 * NFSX_UNSIGNED); \ + fxdr_nfsv3time(tl, &(a)->va_atime); \ + break; \ + case NFSV3SATTRTIME_TOSERVER: \ + (a)->va_atime.ts_sec = time.tv_sec; \ + (a)->va_atime.ts_nsec = time.tv_usec * 1000; \ + break; \ + }; \ + nfsm_dissect(tl, u_long *, NFSX_UNSIGNED); \ + switch (fxdr_unsigned(int, *tl)) { \ + case NFSV3SATTRTIME_TOCLIENT: \ + nfsm_dissect(tl, u_long *, 2 * NFSX_UNSIGNED); \ + fxdr_nfsv3time(tl, &(a)->va_mtime); \ + break; \ + case NFSV3SATTRTIME_TOSERVER: \ + (a)->va_mtime.ts_sec = time.tv_sec; \ + (a)->va_mtime.ts_nsec = time.tv_usec * 1000; \ + break; \ + }; } #endif diff --git a/sys/nfs/nfsmount.h b/sys/nfs/nfsmount.h index 99a1b48..6ac33f7 100644 --- a/sys/nfs/nfsmount.h +++ b/sys/nfs/nfsmount.h @@ -34,7 +34,7 @@ * SUCH DAMAGE. * * @(#)nfsmount.h 8.1 (Berkeley) 6/10/93 - * $Id: nfsmount.h,v 1.3 1994/08/21 06:50:10 paul Exp $ + * $Id: nfsmount.h,v 1.4 1994/10/17 17:47:43 phk Exp $ */ #ifndef _NFS_NFSMOUNT_H_ @@ -49,7 +49,8 @@ struct nfsmount { int nm_flag; /* Flags for soft/hard... */ struct mount *nm_mountp; /* Vfs structure for this filesystem */ int nm_numgrps; /* Max. size of groupslist */ - nfsv2fh_t nm_fh; /* File handle of root dir */ + u_char nm_fh[NFSX_V3FHMAX]; /* File handle of root dir */ + int nm_fhsize; /* Size of root file handle */ struct socket *nm_so; /* Rpc socket */ int nm_sotype; /* Type of socket */ int nm_soproto; /* and protocol */ @@ -65,6 +66,7 @@ struct nfsmount { int nm_deadthresh; /* Threshold of timeouts-->dead server*/ int nm_rsize; /* Max size of read rpc */ int nm_wsize; /* Max size of write rpc */ + int nm_readdirsize; /* Size of a readdir rpc */ int nm_readahead; /* Num. of blocks to readahead */ int nm_leaseterm; /* Term (sec) for NQNFS lease */ CIRCLEQ_HEAD(, nfsnode) nm_timerhead; /* Head of lease timer queue */ @@ -73,9 +75,16 @@ struct nfsmount { int nm_authtype; /* Authenticator type */ int nm_authlen; /* and length */ char *nm_authstr; /* Authenticator string */ + char *nm_verfstr; /* and the verifier */ + int nm_verflen; + u_char nm_verf[NFSX_V3WRITEVERF]; /* V3 write verifier */ + NFSKERBKEY_T nm_key; /* and the session key */ + int nm_numuids; /* Number of nfsuid mappings */ + TAILQ_HEAD(, nfsuid) nm_uidlruhead; /* Lists of nfsuid mappings */ + LIST_HEAD(, nfsuid) nm_uidhashtbl[NFS_MUIDHASHSIZ]; }; -#ifdef KERNEL +#if defined(KERNEL) || defined(_KERNEL) /* * Convert mount ptr to nfsmount ptr. */ diff --git a/sys/nfs/nfsnode.h b/sys/nfs/nfsnode.h index dc9162b..18dbde2 100644 --- a/sys/nfs/nfsnode.h +++ b/sys/nfs/nfsnode.h @@ -34,12 +34,16 @@ * SUCH DAMAGE. * * @(#)nfsnode.h 8.4 (Berkeley) 2/13/94 - * $Id: nfsnode.h,v 1.8 1994/10/17 17:47:44 phk Exp $ + * $Id: nfsnode.h,v 1.9 1995/03/16 18:15:42 bde Exp $ */ #ifndef _NFS_NFSNODE_H_ #define _NFS_NFSNODE_H_ +#ifndef _NFS_NFS_H_ +#include <nfs/nfs.h> +#endif + /* * Silly rename structure that hangs off the nfsnode until the name * can be removed by nfs_inactive() @@ -52,36 +56,73 @@ struct sillyrename { }; /* + * This structure is used to save the logical directory offset to + * NFS cookie mappings. + * The mappings are stored in a list headed + * by n_cookies, as required. + * There is one mapping for each NFS_DIRBLKSIZ bytes of directory information + * stored in increasing logical offset byte order. + */ +#define NFSNUMCOOKIES 31 + +struct nfsdmap { + LIST_ENTRY(nfsdmap) ndm_list; + int ndm_eocookie; + nfsuint64 ndm_cookies[NFSNUMCOOKIES]; +}; + +/* * The nfsnode is the nfs equivalent to ufs's inode. Any similarity * is purely coincidental. * There is a unique nfsnode allocated for each active file, * each current directory, each mounted-on file, text file, and the root. * An nfsnode is 'named' by its file handle. (nget/nfs_node.c) + * If this structure exceeds 256 bytes (it is currently 256 using 4.4BSD-Lite + * type definitions), file handles of > 32 bytes should probably be split out + * into a separate MALLOC()'d data structure. (Reduce the size of nfsfh_t by + * changing the definition in sys/mount.h of NFS_SMALLFH.) + * NB: Hopefully the current order of the fields is such that everything will + * be well aligned and, therefore, tightly packed. */ - struct nfsnode { - LIST_ENTRY(nfsnode) n_hash; /* Hash chain */ - CIRCLEQ_ENTRY(nfsnode) n_timer; /* Nqnfs timer chain */ - nfsv2fh_t n_fh; /* NFS File Handle */ - long n_flag; /* Flag for locking.. */ - struct vnode *n_vnode; /* vnode associated with this node */ - struct vattr n_vattr; /* Vnode attribute cache */ - time_t n_attrstamp; /* Time stamp for cached attributes */ - struct sillyrename *n_sillyrename; /* Ptr to silly rename struct */ - u_quad_t n_size; /* Current size of file */ - struct lockf *n_lockf; /* Locking record of file */ - int n_error; /* Save write error value */ - u_long n_direofoffset; /* Dir. EOF offset cache */ - time_t n_mtime; /* Prev modify time. */ - time_t n_ctime; /* Prev create time. */ - u_quad_t n_brev; /* Modify rev when cached */ - u_quad_t n_lrev; /* Modify rev for lease */ - time_t n_expiry; /* Lease expiry time */ - struct sillyrename n_silly; /* Silly rename struct */ - struct timeval n_atim; /* Special file times */ - struct timeval n_mtim; + LIST_ENTRY(nfsnode) n_hash; /* Hash chain */ + CIRCLEQ_ENTRY(nfsnode) n_timer; /* Nqnfs timer chain */ + u_quad_t n_size; /* Current size of file */ + u_quad_t n_brev; /* Modify rev when cached */ + u_quad_t n_lrev; /* Modify rev for lease */ + struct vattr n_vattr; /* Vnode attribute cache */ + time_t n_attrstamp; /* Attr. cache timestamp */ + time_t n_mtime; /* Prev modify time. */ + time_t n_ctime; /* Prev create time. */ + time_t n_expiry; /* Lease expiry time */ + nfsfh_t *n_fhp; /* NFS File Handle */ + struct vnode *n_vnode; /* associated vnode */ + struct lockf *n_lockf; /* Locking record of file */ + int n_error; /* Save write error value */ + union { + struct timespec nf_atim; /* Special file times */ + nfsuint64 nd_cookieverf; /* Cookie verifier (dir only) */ + } n_un1; + union { + struct timespec nf_mtim; + off_t nd_direof; /* Dir. EOF offset cache */ + } n_un2; + union { + struct sillyrename *nf_silly; /* Ptr to silly rename struct */ + LIST_HEAD(, nfsdmap) nd_cook; /* cookies */ + } n_un3; + short n_fhsize; /* size in bytes, of fh */ + short n_flag; /* Flag for locking.. */ + nfsfh_t n_fh; /* Small File Handle */ }; +#define n_atim n_un1.nf_atim +#define n_mtim n_un2.nf_mtim +#define n_sillyrename n_un3.nf_silly +#define n_cookieverf n_un1.nd_cookieverf +#define n_direofoffset n_un2.nd_direof +#define n_cookies n_un3.nd_cook + /* * Flags for n_flag */ @@ -95,6 +136,8 @@ struct nfsnode { #define NACC 0x0100 /* Special file accessed */ #define NUPD 0x0200 /* Special file updated */ #define NCHG 0x0400 /* Special file times changed */ +#define NLOCKED 0x0800 /* node is locked */ +#define NWANTED 0x0100 /* someone wants to lock */ /* * Convert between nfsnode pointers and vnode pointers @@ -107,7 +150,7 @@ struct nfsnode { */ TAILQ_HEAD(, buf) nfs_bufq; -#ifdef KERNEL +#if defined(KERNEL) || defined(_KERNEL) extern int (**fifo_nfsv2nodeop_p)(); extern int (**nfsv2_vnodeop_p)(); extern int (**spec_nfsv2nodeop_p)(); @@ -128,12 +171,25 @@ int nfs_getattr __P((struct vop_getattr_args *)); int nfs_setattr __P((struct vop_setattr_args *)); int nfs_read __P((struct vop_read_args *)); int nfs_write __P((struct vop_write_args *)); +#ifdef HAS_VOPLEASE +#define nfs_lease_check ((int (*) __P((struct vop_lease_args *)))nullop) +#define nqnfs_vop_lease_check lease_check +#else +#ifdef __FreeBSD__ +#define nqnfs_lease_check nfs_lease_check +#else +#define nqnfs_lease_check lease_check +#endif +#endif int nfsspec_read __P((struct vop_read_args *)); int nfsspec_write __P((struct vop_write_args *)); int nfsfifo_read __P((struct vop_read_args *)); int nfsfifo_write __P((struct vop_write_args *)); #define nfs_ioctl ((int (*) __P((struct vop_ioctl_args *)))enoioctl) #define nfs_select ((int (*) __P((struct vop_select_args *)))seltrue) +#ifdef HAS_VOPREVOKE +#define nfs_revoke vop_revoke +#endif int nfs_mmap __P((struct vop_mmap_args *)); int nfs_fsync __P((struct vop_fsync_args *)); #define nfs_seek ((int (*) __P((struct vop_seek_args *)))nullop) @@ -157,6 +213,7 @@ int nfs_islocked __P((struct vop_islocked_args *)); int nfs_pathconf __P((struct vop_pathconf_args *)); int nfs_advlock __P((struct vop_advlock_args *)); int nfs_blkatoff __P((struct vop_blkatoff_args *)); +int nfs_bwrite __P((struct vop_bwrite_args *)); int nfs_vget __P((struct mount *, ino_t, struct vnode **)); int nfs_valloc __P((struct vop_valloc_args *)); #define nfs_reallocblks \ @@ -164,13 +221,20 @@ int nfs_valloc __P((struct vop_valloc_args *)); int nfs_vfree __P((struct vop_vfree_args *)); int nfs_truncate __P((struct vop_truncate_args *)); int nfs_update __P((struct vop_update_args *)); -int nfs_bwrite __P((struct vop_bwrite_args *)); /* other stuff */ int nfs_removeit __P((struct sillyrename *)); -int nfs_nget __P((struct mount *,nfsv2fh_t *,struct nfsnode **)); -int nfs_lookitup __P((struct sillyrename *,nfsv2fh_t *,struct proc *)); +int nfs_nget __P((struct mount *,nfsfh_t *,int,struct nfsnode **)); +int nfs_lookitup __P((struct vnode *,char *,int,struct ucred *,struct proc *,struct nfsnode **)); int nfs_sillyrename __P((struct vnode *,struct vnode *,struct componentname *)); +nfsuint64 *nfs_getcookie __P((struct nfsnode *, off_t, int)); +void nfs_invaldir __P((struct vnode *)); + +#ifdef __FreeBSD__ +#define nqnfs_lease_updatetime nfs_lease_updatetime +#else +#define nqnfs_lease_updatetime lease_updatetime +#endif #endif /* KERNEL */ diff --git a/sys/nfs/nfsproto.h b/sys/nfs/nfsproto.h new file mode 100644 index 0000000..ac1a090 --- /dev/null +++ b/sys/nfs/nfsproto.h @@ -0,0 +1,441 @@ +/* + * Copyright (c) 1989, 1993 + * The Regents of the University of California. All rights reserved. + * + * This code is derived from software contributed to Berkeley by + * Rick Macklem at The University of Guelph. + * + * Redistribution and use in source and binary forms, with or without + * modification, are permitted provided that the following conditions + * are met: + * 1. Redistributions of source code must retain the above copyright + * notice, this list of conditions and the following disclaimer. + * 2. Redistributions in binary form must reproduce the above copyright + * notice, this list of conditions and the following disclaimer in the + * documentation and/or other materials provided with the distribution. + * 3. All advertising materials mentioning features or use of this software + * must display the following acknowledgement: + * This product includes software developed by the University of + * California, Berkeley and its contributors. + * 4. Neither the name of the University nor the names of its contributors + * may be used to endorse or promote products derived from this software + * without specific prior written permission. + * + * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND + * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE + * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE + * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE + * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL + * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS + * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) + * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT + * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY + * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF + * SUCH DAMAGE. + * + * @(#)nfsproto.h 8.1 (Berkeley) 6/10/93 + * $Id: nfsproto.h,v 1.4 1994/08/21 06:50:13 paul Exp $ + */ + +#ifndef _NFS_NFSPROTO_H_ +#define _NFS_NFSPROTO_H_ + +/* + * nfs definitions as per the Version 2 and 3 specs + */ + +/* + * Constants as defined in the Sun NFS Version 2 and 3 specs. + * "NFS: Network File System Protocol Specification" RFC1094 + * and in the "NFS: Network File System Version 3 Protocol + * Specification" + */ + +#define NFS_PORT 2049 +#define NFS_PROG 100003 +#define NFS_VER2 2 +#define NFS_VER3 3 +#define NFS_V2MAXDATA 8192 +#define NFS_MAXDGRAMDATA 16384 +#define NFS_MAXDATA 32768 +#define NFS_MAXPATHLEN 1024 +#define NFS_MAXNAMLEN 255 +#define NFS_MAXPKTHDR 404 +#define NFS_MAXPACKET (NFS_MAXPKTHDR + NFS_MAXDATA) +#define NFS_MINPACKET 20 +#define NFS_FABLKSIZE 512 /* Size in bytes of a block wrt fa_blocks */ + +/* Stat numbers for rpc returns (version 2 and 3) */ +#define NFS_OK 0 +#define NFSERR_PERM 1 +#define NFSERR_NOENT 2 +#define NFSERR_IO 5 +#define NFSERR_NXIO 6 +#define NFSERR_ACCES 13 +#define NFSERR_EXIST 17 +#define NFSERR_XDEV 18 /* Version 3 only */ +#define NFSERR_NODEV 19 +#define NFSERR_NOTDIR 20 +#define NFSERR_ISDIR 21 +#define NFSERR_INVAL 22 /* Version 3 only */ +#define NFSERR_FBIG 27 +#define NFSERR_NOSPC 28 +#define NFSERR_ROFS 30 +#define NFSERR_MLINK 31 /* Version 3 only */ +#define NFSERR_NAMETOL 63 +#define NFSERR_NOTEMPTY 66 +#define NFSERR_DQUOT 69 +#define NFSERR_STALE 70 +#define NFSERR_REMOTE 71 /* Version 3 only */ +#define NFSERR_WFLUSH 99 /* Version 2 only */ +#define NFSERR_BADHANDLE 10001 /* The rest Version 3 only */ +#define NFSERR_NOT_SYNC 10002 +#define NFSERR_BAD_COOKIE 10003 +#define NFSERR_NOTSUPP 10004 +#define NFSERR_TOOSMALL 10005 +#define NFSERR_SERVERFAULT 10006 +#define NFSERR_BADTYPE 10007 +#define NFSERR_JUKEBOX 10008 +#define NFSERR_TRYLATER NFSERR_JUKEBOX +#define NFSERR_STALEWRITEVERF 30001 /* Fake return for nfs_commit() */ + +#define NFSERR_RETVOID 0x20000000 /* Return void, not error */ +#define NFSERR_AUTHERR 0x40000000 /* Mark an authentication error */ +#define NFSERR_RETERR 0x80000000 /* Mark an error return for V3 */ + +/* Sizes in bytes of various nfs rpc components */ +#define NFSX_UNSIGNED 4 + +/* specific to NFS Version 2 */ +#define NFSX_V2FH 32 +#define NFSX_V2FATTR 68 +#define NFSX_V2SATTR 32 +#define NFSX_V2COOKIE 4 +#define NFSX_V2STATFS 20 + +/* specific to NFS Version 3 */ +#define NFSX_V3FH (sizeof (fhandle_t)) /* size this server uses */ +#define NFSX_V3FHMAX 64 /* max. allowed by protocol */ +#define NFSX_V3FATTR 84 +#define NFSX_V3SATTR 60 /* max. all fields filled in */ +#define NFSX_V3SRVSATTR (sizeof (struct nfsv3_sattr)) +#define NFSX_V3POSTOPATTR (NFSX_V3FATTR + NFSX_UNSIGNED) +#define NFSX_V3WCCDATA (NFSX_V3POSTOPATTR + 8 * NFSX_UNSIGNED) +#define NFSX_V3COOKIEVERF 8 +#define NFSX_V3WRITEVERF 8 +#define NFSX_V3CREATEVERF 8 +#define NFSX_V3STATFS 52 +#define NFSX_V3FSINFO 48 +#define NFSX_V3PATHCONF 24 + +/* variants for both versions */ +#define NFSX_FH(v3) ((v3) ? (NFSX_V3FHMAX + NFSX_UNSIGNED) : \ + NFSX_V2FH) +#define NFSX_SRVFH(v3) ((v3) ? NFSX_V3FH : NFSX_V2FH) +#define NFSX_FATTR(v3) ((v3) ? NFSX_V3FATTR : NFSX_V2FATTR) +#define NFSX_PREOPATTR(v3) ((v3) ? (7 * NFSX_UNSIGNED) : 0) +#define NFSX_POSTOPATTR(v3) ((v3) ? (NFSX_V3FATTR + NFSX_UNSIGNED) : 0) +#define NFSX_POSTOPORFATTR(v3) ((v3) ? (NFSX_V3FATTR + NFSX_UNSIGNED) : \ + NFSX_V2FATTR) +#define NFSX_WCCDATA(v3) ((v3) ? NFSX_V3WCCDATA : 0) +#define NFSX_WCCORFATTR(v3) ((v3) ? NFSX_V3WCCDATA : NFSX_V2FATTR) +#define NFSX_SATTR(v3) ((v3) ? NFSX_V3SATTR : NFSX_V2SATTR) +#define NFSX_COOKIEVERF(v3) ((v3) ? NFSX_V3COOKIEVERF : 0) +#define NFSX_WRITEVERF(v3) ((v3) ? NFSX_V3WRITEVERF : 0) +#define NFSX_READDIR(v3) ((v3) ? (5 * NFSX_UNSIGNED) : \ + (2 * NFSX_UNSIGNED)) +#define NFSX_STATFS(v3) ((v3) ? NFSX_V3STATFS : NFSX_V2STATFS) + +/* nfs rpc procedure numbers (before version mapping) */ +#define NFSPROC_NULL 0 +#define NFSPROC_GETATTR 1 +#define NFSPROC_SETATTR 2 +#define NFSPROC_LOOKUP 3 +#define NFSPROC_ACCESS 4 +#define NFSPROC_READLINK 5 +#define NFSPROC_READ 6 +#define NFSPROC_WRITE 7 +#define NFSPROC_CREATE 8 +#define NFSPROC_MKDIR 9 +#define NFSPROC_SYMLINK 10 +#define NFSPROC_MKNOD 11 +#define NFSPROC_REMOVE 12 +#define NFSPROC_RMDIR 13 +#define NFSPROC_RENAME 14 +#define NFSPROC_LINK 15 +#define NFSPROC_READDIR 16 +#define NFSPROC_READDIRPLUS 17 +#define NFSPROC_FSSTAT 18 +#define NFSPROC_FSINFO 19 +#define NFSPROC_PATHCONF 20 +#define NFSPROC_COMMIT 21 + +/* And leasing (nqnfs) procedure numbers (must be last) */ +#define NQNFSPROC_GETLEASE 22 +#define NQNFSPROC_VACATED 23 +#define NQNFSPROC_EVICTED 24 + +#define NFSPROC_NOOP 25 +#define NFS_NPROCS 26 + +/* Actual Version 2 procedure numbers */ +#define NFSV2PROC_NULL 0 +#define NFSV2PROC_GETATTR 1 +#define NFSV2PROC_SETATTR 2 +#define NFSV2PROC_NOOP 3 +#define NFSV2PROC_ROOT NFSV2PROC_NOOP /* Obsolete */ +#define NFSV2PROC_LOOKUP 4 +#define NFSV2PROC_READLINK 5 +#define NFSV2PROC_READ 6 +#define NFSV2PROC_WRITECACHE NFSV2PROC_NOOP /* Obsolete */ +#define NFSV2PROC_WRITE 8 +#define NFSV2PROC_CREATE 9 +#define NFSV2PROC_REMOVE 10 +#define NFSV2PROC_RENAME 11 +#define NFSV2PROC_LINK 12 +#define NFSV2PROC_SYMLINK 13 +#define NFSV2PROC_MKDIR 14 +#define NFSV2PROC_RMDIR 15 +#define NFSV2PROC_READDIR 16 +#define NFSV2PROC_STATFS 17 + +/* + * Constants used by the Version 3 protocol for various RPCs + */ +#define NFSV3SATTRTIME_DONTCHANGE 0 +#define NFSV3SATTRTIME_TOSERVER 1 +#define NFSV3SATTRTIME_TOCLIENT 2 + +#define NFSV3ACCESS_READ 0x01 +#define NFSV3ACCESS_LOOKUP 0x02 +#define NFSV3ACCESS_MODIFY 0x04 +#define NFSV3ACCESS_EXTEND 0x08 +#define NFSV3ACCESS_DELETE 0x10 +#define NFSV3ACCESS_EXECUTE 0x20 + +#define NFSV3WRITE_UNSTABLE 0 +#define NFSV3WRITE_DATASYNC 1 +#define NFSV3WRITE_FILESYNC 2 + +#define NFSV3CREATE_UNCHECKED 0 +#define NFSV3CREATE_GUARDED 1 +#define NFSV3CREATE_EXCLUSIVE 2 + +#define NFSV3FSINFO_LINK 0x01 +#define NFSV3FSINFO_SYMLINK 0x02 +#define NFSV3FSINFO_HOMOGENEOUS 0x08 +#define NFSV3FSINFO_CANSETTIME 0x10 + +/* Conversion macros */ +#define vtonfsv2_mode(t,m) \ + txdr_unsigned(((t) == VFIFO) ? MAKEIMODE(VCHR, (m)) : \ + MAKEIMODE((t), (m))) +#define vtonfsv3_mode(m) txdr_unsigned((m) & 07777) +#define nfstov_mode(a) (fxdr_unsigned(u_short, (a))&07777) +#define vtonfsv2_type(a) txdr_unsigned(nfsv2_type[((long)(a))]) +#define vtonfsv3_type(a) txdr_unsigned(nfsv3_type[((long)(a))]) +#define nfsv2tov_type(a) nv2tov_type[fxdr_unsigned(u_long,(a))&0x7] +#define nfsv3tov_type(a) nv3tov_type[fxdr_unsigned(u_long,(a))&0x7] + +/* File types */ +typedef enum { NFNON=0, NFREG=1, NFDIR=2, NFBLK=3, NFCHR=4, NFLNK=5, + NFSOCK=6, NFFIFO=7 } nfstype; + +/* Structs for common parts of the rpc's */ +/* + * File Handle (32 bytes for version 2), variable up to 64 for version 3. + * File Handles of up to NFS_SMALLFH in size are stored directly in the + * nfs node, whereas larger ones are malloc'd. (This never happens when + * NFS_SMALLFH is set to 64.) + * NFS_SMALLFH should be in the range of 32 to 64 and be divisible by 4. + */ +#ifndef NFS_SMALLFH +#define NFS_SMALLFH 64 +#endif +union nfsfh { + fhandle_t fh_generic; + u_char fh_bytes[NFS_SMALLFH]; +}; +typedef union nfsfh nfsfh_t; + +struct nfsv2_time { + u_long nfsv2_sec; + u_long nfsv2_usec; +}; +typedef struct nfsv2_time nfstime2; + +struct nfsv3_time { + u_long nfsv3_sec; + u_long nfsv3_nsec; +}; +typedef struct nfsv3_time nfstime3; + +/* + * Quads are defined as arrays of 2 longs to ensure dense packing for the + * protocol and to facilitate xdr conversion. + */ +struct nfs_uquad { + u_long nfsuquad[2]; +}; +typedef struct nfs_uquad nfsuint64; + +/* + * Used to convert between two u_longs and a u_quad_t. + */ +union nfs_quadconvert { + u_long lval[2]; + u_quad_t qval; +}; +typedef union nfs_quadconvert nfsquad_t; + +/* + * NFS Version 3 special file number. + */ +struct nfsv3_spec { + u_long specdata1; + u_long specdata2; +}; +typedef struct nfsv3_spec nfsv3spec; + +/* + * File attributes and setable attributes. These structures cover both + * NFS version 2 and the version 3 protocol. Note that the union is only + * used so that one pointer can refer to both variants. These structures + * go out on the wire and must be densely packed, so no quad data types + * are used. (all fields are longs or u_longs or structures of same) + * NB: You can't do sizeof(struct nfs_fattr), you must use the + * NFSX_FATTR(v3) macro. + */ +struct nfs_fattr { + u_long fa_type; + u_long fa_mode; + u_long fa_nlink; + u_long fa_uid; + u_long fa_gid; + union { + struct { + u_long nfsv2fa_size; + u_long nfsv2fa_blocksize; + u_long nfsv2fa_rdev; + u_long nfsv2fa_blocks; + u_long nfsv2fa_fsid; + u_long nfsv2fa_fileid; + nfstime2 nfsv2fa_atime; + nfstime2 nfsv2fa_mtime; + nfstime2 nfsv2fa_ctime; + } fa_nfsv2; + struct { + nfsuint64 nfsv3fa_size; + nfsuint64 nfsv3fa_used; + nfsv3spec nfsv3fa_rdev; + nfsuint64 nfsv3fa_fsid; + nfsuint64 nfsv3fa_fileid; + nfstime3 nfsv3fa_atime; + nfstime3 nfsv3fa_mtime; + nfstime3 nfsv3fa_ctime; + } fa_nfsv3; + } fa_un; +}; + +/* and some ugly defines for accessing union components */ +#define fa2_size fa_un.fa_nfsv2.nfsv2fa_size +#define fa2_blocksize fa_un.fa_nfsv2.nfsv2fa_blocksize +#define fa2_rdev fa_un.fa_nfsv2.nfsv2fa_rdev +#define fa2_blocks fa_un.fa_nfsv2.nfsv2fa_blocks +#define fa2_fsid fa_un.fa_nfsv2.nfsv2fa_fsid +#define fa2_fileid fa_un.fa_nfsv2.nfsv2fa_fileid +#define fa2_atime fa_un.fa_nfsv2.nfsv2fa_atime +#define fa2_mtime fa_un.fa_nfsv2.nfsv2fa_mtime +#define fa2_ctime fa_un.fa_nfsv2.nfsv2fa_ctime +#define fa3_size fa_un.fa_nfsv3.nfsv3fa_size +#define fa3_used fa_un.fa_nfsv3.nfsv3fa_used +#define fa3_rdev fa_un.fa_nfsv3.nfsv3fa_rdev +#define fa3_fsid fa_un.fa_nfsv3.nfsv3fa_fsid +#define fa3_fileid fa_un.fa_nfsv3.nfsv3fa_fileid +#define fa3_atime fa_un.fa_nfsv3.nfsv3fa_atime +#define fa3_mtime fa_un.fa_nfsv3.nfsv3fa_mtime +#define fa3_ctime fa_un.fa_nfsv3.nfsv3fa_ctime + +struct nfsv2_sattr { + u_long sa_mode; + u_long sa_uid; + u_long sa_gid; + u_long sa_size; + nfstime2 sa_atime; + nfstime2 sa_mtime; +}; + +/* + * NFS Version 3 sattr structure for the new node creation case. + */ +struct nfsv3_sattr { + u_long sa_modetrue; + u_long sa_mode; + u_long sa_uidtrue; + u_long sa_uid; + u_long sa_gidtrue; + u_long sa_gid; + u_long sa_sizefalse; + u_long sa_atimetype; + nfstime3 sa_atime; + u_long sa_mtimetype; + nfstime3 sa_mtime; +}; + +struct nfs_statfs { + union { + struct { + u_long nfsv2sf_tsize; + u_long nfsv2sf_bsize; + u_long nfsv2sf_blocks; + u_long nfsv2sf_bfree; + u_long nfsv2sf_bavail; + } sf_nfsv2; + struct { + nfsuint64 nfsv3sf_tbytes; + nfsuint64 nfsv3sf_fbytes; + nfsuint64 nfsv3sf_abytes; + nfsuint64 nfsv3sf_tfiles; + nfsuint64 nfsv3sf_ffiles; + nfsuint64 nfsv3sf_afiles; + u_long nfsv3sf_invarsec; + } sf_nfsv3; + } sf_un; +}; + +#define sf_tsize sf_un.sf_nfsv2.nfsv2sf_tsize +#define sf_bsize sf_un.sf_nfsv2.nfsv2sf_bsize +#define sf_blocks sf_un.sf_nfsv2.nfsv2sf_blocks +#define sf_bfree sf_un.sf_nfsv2.nfsv2sf_bfree +#define sf_bavail sf_un.sf_nfsv2.nfsv2sf_bavail +#define sf_tbytes sf_un.sf_nfsv3.nfsv3sf_tbytes +#define sf_fbytes sf_un.sf_nfsv3.nfsv3sf_fbytes +#define sf_abytes sf_un.sf_nfsv3.nfsv3sf_abytes +#define sf_tfiles sf_un.sf_nfsv3.nfsv3sf_tfiles +#define sf_ffiles sf_un.sf_nfsv3.nfsv3sf_ffiles +#define sf_afiles sf_un.sf_nfsv3.nfsv3sf_afiles +#define sf_invarsec sf_un.sf_nfsv3.nfsv3sf_invarsec + +struct nfsv3_fsinfo { + u_long fs_rtmax; + u_long fs_rtpref; + u_long fs_rtmult; + u_long fs_wtmax; + u_long fs_wtpref; + u_long fs_wtmult; + u_long fs_dtpref; + nfsuint64 fs_maxfilesize; + nfstime3 fs_timedelta; + u_long fs_properties; +}; + +struct nfsv3_pathconf { + u_long pc_linkmax; + u_long pc_namemax; + u_long pc_notrunc; + u_long pc_chownrestricted; + u_long pc_caseinsensitive; + u_long pc_casepreserving; +}; + +#endif diff --git a/sys/nfs/nfsrtt.h b/sys/nfs/nfsrtt.h index 4463f62..38f520e 100644 --- a/sys/nfs/nfsrtt.h +++ b/sys/nfs/nfsrtt.h @@ -34,7 +34,7 @@ * SUCH DAMAGE. * * @(#)nfsrtt.h 8.1 (Berkeley) 6/10/93 - * $Id: nfsrtt.h,v 1.2 1994/08/02 07:52:24 davidg Exp $ + * $Id: nfsrtt.h,v 1.3 1994/08/21 06:50:12 paul Exp $ */ #ifndef _NFS_NFSRTT_H_ @@ -82,6 +82,7 @@ struct nfsrtt { #define DRT_TCP 0x02 /* Client used TCP transport */ #define DRT_CACHEREPLY 0x04 /* Reply was from recent request cache */ #define DRT_CACHEDROP 0x08 /* Rpc request dropped, due to recent reply */ +#define DRT_NFSV3 0x10 /* Rpc used NFS Version 3 */ /* * Server log structure diff --git a/sys/nfs/nfsrvcache.h b/sys/nfs/nfsrvcache.h index b367b9f..dc10bd7 100644 --- a/sys/nfs/nfsrvcache.h +++ b/sys/nfs/nfsrvcache.h @@ -34,7 +34,7 @@ * SUCH DAMAGE. * * @(#)nfsrvcache.h 8.1 (Berkeley) 6/10/93 - * $Id: nfsrvcache.h,v 1.3 1994/08/21 06:50:13 paul Exp $ + * $Id: nfsrvcache.h,v 1.4 1994/10/17 17:47:44 phk Exp $ */ #ifndef _NFS_NFSRVCACHE_H_ @@ -44,7 +44,7 @@ * Definitions for the server recent request cache */ -#define NFSRVCACHESIZ 256 +#define NFSRVCACHESIZ 64 struct nfsrvcache { TAILQ_ENTRY(nfsrvcache) rc_lru; /* LRU chain */ diff --git a/sys/nfs/nfsv2.h b/sys/nfs/nfsv2.h index 62ba338..1ae1a72 100644 --- a/sys/nfs/nfsv2.h +++ b/sys/nfs/nfsv2.h @@ -34,129 +34,201 @@ * SUCH DAMAGE. * * @(#)nfsv2.h 8.1 (Berkeley) 6/10/93 - * $Id: nfsv2.h,v 1.3 1994/08/18 22:35:37 wollman Exp $ + * $Id: nfsv2.h,v 1.4 1994/08/21 06:50:13 paul Exp $ */ -#ifndef _NFS_NFSV2_H_ -#define _NFS_NFSV2_H_ +#include <nfs/nfsproto.h> + +#if 0 + /* - * nfs definitions as per the version 2 specs + * nfs definitions as per the Version 2 and 3 specs */ /* - * Constants as defined in the Sun NFS Version 2 spec. + * Constants as defined in the Sun NFS Version 2 and 3 specs. * "NFS: Network File System Protocol Specification" RFC1094 + * and in the "NFS: Network File System Version 3 Protocol + * Specification" */ #define NFS_PORT 2049 #define NFS_PROG 100003 #define NFS_VER2 2 +#define NFS_VER3 3 #define NFS_MAXDGRAMDATA 8192 #define NFS_MAXDATA 32768 #define NFS_MAXPATHLEN 1024 #define NFS_MAXNAMLEN 255 -#define NFS_FHSIZE 32 #define NFS_MAXPKTHDR 404 -#define NFS_MAXPACKET (NFS_MAXPKTHDR+NFS_MAXDATA) +#define NFS_MAXPACKET (NFS_MAXPKTHDR + NFS_MAXDATA) #define NFS_MINPACKET 20 #define NFS_FABLKSIZE 512 /* Size in bytes of a block wrt fa_blocks */ -/* Stat numbers for rpc returns */ -#define NFS_OK 0 -#define NFSERR_PERM 1 -#define NFSERR_NOENT 2 -#define NFSERR_IO 5 -#define NFSERR_NXIO 6 -#define NFSERR_ACCES 13 -#define NFSERR_EXIST 17 -#define NFSERR_NODEV 19 -#define NFSERR_NOTDIR 20 -#define NFSERR_ISDIR 21 -#define NFSERR_FBIG 27 -#define NFSERR_NOSPC 28 -#define NFSERR_ROFS 30 -#define NFSERR_NAMETOL 63 -#define NFSERR_NOTEMPTY 66 -#define NFSERR_DQUOT 69 -#define NFSERR_STALE 70 -#define NFSERR_WFLUSH 99 +/* Stat numbers for rpc returns (version 2 and 3) */ +#define NFS_OK 0 +#define NFSERR_PERM 1 +#define NFSERR_NOENT 2 +#define NFSERR_IO 5 +#define NFSERR_NXIO 6 +#define NFSERR_ACCES 13 +#define NFSERR_EXIST 17 +#define NFSERR_XDEV 18 /* Version 3 only */ +#define NFSERR_NODEV 19 +#define NFSERR_NOTDIR 20 +#define NFSERR_ISDIR 21 +#define NFSERR_INVAL 22 /* Version 3 only */ +#define NFSERR_FBIG 27 +#define NFSERR_NOSPC 28 +#define NFSERR_ROFS 30 +#define NFSERR_MLINK 31 /* Version 3 only */ +#define NFSERR_NAMETOL 63 +#define NFSERR_NOTEMPTY 66 +#define NFSERR_DQUOT 69 +#define NFSERR_STALE 70 +#define NFSERR_REMOTE 71 /* Version 3 only */ +#define NFSERR_WFLUSH 99 /* Version 2 only */ +#define NFSERR_BADHANDLE 10001 /* The rest Version 3 only */ +#define NFSERR_NOT_SYNC 10002 +#define NFSERR_BAD_COOKIE 10003 +#define NFSERR_NOTSUPP 10004 +#define NFSERR_TOOSMALL 10005 +#define NFSERR_SERVERFAULT 10006 +#define NFSERR_BADTYPE 10007 +#define NFSERR_JUKEBOX 10008 /* Sizes in bytes of various nfs rpc components */ -#define NFSX_FH 32 #define NFSX_UNSIGNED 4 -#define NFSX_NFSFATTR 68 -#define NFSX_NQFATTR 92 -#define NFSX_NFSSATTR 32 -#define NFSX_NQSATTR 44 -#define NFSX_COOKIE 4 -#define NFSX_NFSSTATFS 20 -#define NFSX_NQSTATFS 28 -#define NFSX_FATTR(isnq) ((isnq) ? NFSX_NQFATTR : NFSX_NFSFATTR) -#define NFSX_SATTR(isnq) ((isnq) ? NFSX_NQSATTR : NFSX_NFSSATTR) -#define NFSX_STATFS(isnq) ((isnq) ? NFSX_NQSTATFS : NFSX_NFSSTATFS) -/* nfs rpc procedure numbers */ +/* specific to NFS Version 2 */ +#define NFSX_V2FH 32 +#define NFSX_V2FATTR 68 +#define NFSX_V2SATTR 32 +#define NFSX_V2COOKIE 4 +#define NFSX_V2STATFS 20 + +/* specific to NFS Version 3 */ +#define NFSX_V3FH 16 /* size this server uses */ +#define NFSX_V3FHMAX 64 /* max. allowed by protocol */ +#define NFSX_V3FATTR 84 +#define NFSX_V3SATTR 60 /* max. all fields filled in */ +#define NFSX_V3COOKIEVERF 8 + +/* variants for both versions */ +#define NFSX_FH(v3) ((v3) ? (NFSX_V3FHMAX + NFSX_UNSIGNED) : \ + NFSX_V2FH) +#define NFSX_SRVFH(v3) ((v3) ? NFSX_V3FH : NFSX_V2FH) +#define NFSX_FATTR(v3) ((v3) ? NFSX_V3FATTR : NFSX_V2FATTR) +#define NFSX_POSTOPATTR(v3) ((v3) ? (NFSX_V3FATTR + NFSX_UNSIGNED) : 0) +#define NFSX_POSTOPORFATTR(v3) ((v3) ? (NFSX_V3FATTR + NFSX_UNSIGNED) : \ + NFSX_V2FATTR) +#define NFSX_WCCDATA(v3) ((v3) ? (NFSX_V3FATTR + 32) : 0) +#define NFSX_WCCORFATTR(v3) ((v3) ? (NFSX_V3FATTR + 32) : NFSX_V2FATTR) +#define NFSX_SATTR(v3) ((v3) ? NFSX_V3SATTR : NFSX_V2SATTR) +#define NFSX_COOKIEVERF(v3) ((v3) ? NFSX_V3COOKIEVERF : 0) +#define NFSX_STATFS(isv3) ((isv3) ? NFSX_NFSV3STATFS : NFSX_NFSSTATFS) + +/* nfs rpc procedure numbers (before version mapping) */ #define NFSPROC_NULL 0 #define NFSPROC_GETATTR 1 #define NFSPROC_SETATTR 2 -#define NFSPROC_NOOP 3 -#define NFSPROC_ROOT NFSPROC_NOOP /* Obsolete */ -#define NFSPROC_LOOKUP 4 +#define NFSPROC_LOOKUP 3 +#define NFSPROC_ACCESS 4 #define NFSPROC_READLINK 5 #define NFSPROC_READ 6 -#define NFSPROC_WRITECACHE NFSPROC_NOOP /* Obsolete */ -#define NFSPROC_WRITE 8 -#define NFSPROC_CREATE 9 -#define NFSPROC_REMOVE 10 -#define NFSPROC_RENAME 11 -#define NFSPROC_LINK 12 -#define NFSPROC_SYMLINK 13 -#define NFSPROC_MKDIR 14 -#define NFSPROC_RMDIR 15 +#define NFSPROC_WRITE 7 +#define NFSPROC_CREATE 8 +#define NFSPROC_MKDIR 9 +#define NFSPROC_SYMLINK 10 +#define NFSPROC_MKNOD 11 +#define NFSPROC_REMOVE 12 +#define NFSPROC_RMDIR 13 +#define NFSPROC_RENAME 14 +#define NFSPROC_LINK 15 #define NFSPROC_READDIR 16 -#define NFSPROC_STATFS 17 +#define NFSPROC_READDIRPLUS 17 +#define NFSPROC_FSSTAT 18 +#define NFSPROC_FSINFO 19 +#define NFSPROC_PATHCONF 20 +#define NFSPROC_COMMIT 21 -/* NQ nfs numbers */ -#define NQNFSPROC_READDIRLOOK 18 -#define NQNFSPROC_GETLEASE 19 -#define NQNFSPROC_VACATED 20 -#define NQNFSPROC_EVICTED 21 -#define NQNFSPROC_ACCESS 22 +/* And leasing (nqnfs) procedure numbers */ +#define NQNFSPROC_GETLEASE 22 +#define NQNFSPROC_VACATED 23 +#define NQNFSPROC_EVICTED 24 + +#define NFS_NPROCS 25 + +/* Actual Version 2 procedure numbers */ +#define NFSV2PROC_NULL 0 +#define NFSV2PROC_GETATTR 1 +#define NFSV2PROC_SETATTR 2 +#define NFSV2PROC_NOOP 3 +#define NFSV2PROC_ROOT NFSV2PROC_NOOP /* Obsolete */ +#define NFSV2PROC_LOOKUP 4 +#define NFSV2PROC_READLINK 5 +#define NFSV2PROC_READ 6 +#define NFSV2PROC_WRITECACHE NFSV2PROC_NOOP /* Obsolete */ +#define NFSV2PROC_WRITE 8 +#define NFSV2PROC_CREATE 9 +#define NFSV2PROC_REMOVE 10 +#define NFSV2PROC_RENAME 11 +#define NFSV2PROC_LINK 12 +#define NFSV2PROC_SYMLINK 13 +#define NFSV2PROC_MKDIR 14 +#define NFSV2PROC_RMDIR 15 +#define NFSV2PROC_READDIR 16 +#define NFSV2PROC_STATFS 17 -#define NFS_NPROCS 23 /* Conversion macros */ -#define vtonfs_mode(t,m) \ +extern int vttoif_tab[]; +#define vtonfsv2_mode(t,m) \ txdr_unsigned(((t) == VFIFO) ? MAKEIMODE(VCHR, (m)) : \ MAKEIMODE((t), (m))) -#define nfstov_mode(a) (fxdr_unsigned(u_short, (a))&07777) -#define vtonfs_type(a) txdr_unsigned(nfs_type[((long)(a))]) -#define nfstov_type(a) ntov_type[fxdr_unsigned(u_long,(a))&0x7] +#define vtonfsv3_mode(m) txdr_unsigned((m) & 07777) +#define nfstov_mode(a) (fxdr_unsigned(u_short, (a))&07777) +#define vtonfsv2_type(a) txdr_unsigned(nfsv2_type[((long)(a))]) +#define vtonfsv3_type(a) txdr_unsigned(nfsv3_type[((long)(a))]) +#define nfsv2tov_type(a) nv2tov_type[fxdr_unsigned(u_long,(a))&0x7] +#define nfsv3tov_type(a) nv3tov_type[fxdr_unsigned(u_long,(a))&0x7] /* File types */ -typedef enum { NFNON=0, NFREG=1, NFDIR=2, NFBLK=3, NFCHR=4, NFLNK=5 } nfstype; +typedef enum { NFNON=0, NFREG=1, NFDIR=2, NFBLK=3, NFCHR=4, NFLNK=5, + NFSOCK=6, NFFIFO=7 } nfstype; /* Structs for common parts of the rpc's */ struct nfsv2_time { - u_long nfs_sec; - u_long nfs_usec; + u_long nfsv2_sec; + u_long nfsv2_usec; +}; +typedef struct nfsv2_time nfstime2; + +struct nfsv3_time { + u_long nfsv3_sec; + u_long nfsv3_nsec; }; +typedef struct nfsv3_time nfstime3; -struct nqnfs_time { - u_long nq_sec; - u_long nq_nsec; +/* + * Quads are defined as arrays of 2 longs to ensure dense packing for the + * protocol and to facilitate xdr conversion. + */ +struct nfs_uquad { + u_long nfsuquad[2]; }; +typedef struct nfs_uquad nfsuint64; /* * File attributes and setable attributes. These structures cover both - * NFS version 2 and the NQNFS protocol. Note that the union is only - * used to that one pointer can refer to both variants. These structures + * NFS version 2 and the version 3 protocol. Note that the union is only + * used so that one pointer can refer to both variants. These structures * go out on the wire and must be densely packed, so no quad data types * are used. (all fields are longs or u_longs or structures of same) - * NB: You can't do sizeof(struct nfsv2_fattr), you must use the - * NFSX_FATTR(isnq) macro. + * NB: You can't do sizeof(struct nfs_fattr), you must use the + * NFSX_FATTR(v3) macro. */ -struct nfsv2_fattr { +struct nfs_fattr { u_long fa_type; u_long fa_mode; u_long fa_nlink; @@ -164,102 +236,85 @@ struct nfsv2_fattr { u_long fa_gid; union { struct { - u_long nfsfa_size; - u_long nfsfa_blocksize; - u_long nfsfa_rdev; - u_long nfsfa_blocks; - u_long nfsfa_fsid; - u_long nfsfa_fileid; - struct nfsv2_time nfsfa_atime; - struct nfsv2_time nfsfa_mtime; - struct nfsv2_time nfsfa_ctime; + u_long nfsv2fa_size; + u_long nfsv2fa_blocksize; + u_long nfsv2fa_rdev; + u_long nfsv2fa_blocks; + u_long nfsv2fa_fsid; + u_long nfsv2fa_fileid; + nfstime2 nfsv2fa_atime; + nfstime2 nfsv2fa_mtime; + nfstime2 nfsv2fa_ctime; } fa_nfsv2; struct { - struct { - u_long nqfa_qsize[2]; - } nqfa_size; - u_long nqfa_blocksize; - u_long nqfa_rdev; - struct { - u_long nqfa_qbytes[2]; - } nqfa_bytes; - u_long nqfa_fsid; - u_long nqfa_fileid; - struct nqnfs_time nqfa_atime; - struct nqnfs_time nqfa_mtime; - struct nqnfs_time nqfa_ctime; - u_long nqfa_flags; - u_long nqfa_gen; - struct { - u_long nqfa_qfilerev[2]; - } nqfa_filerev; - } fa_nqnfs; + nfsuint64 nfsv3fa_size; + nfsuint64 nfsv3fa_used; + nfsuint64 nfsv3fa_rdev; + nfsuint64 nfsv3fa_fsid; + nfsuint64 nfsv3fa_fileid; + nfstime3 nfsv3fa_atime; + nfstime3 nfsv3fa_mtime; + nfstime3 nfsv3fa_ctime; + } fa_nfsv3; } fa_un; }; /* and some ugly defines for accessing union components */ -#define fa_nfssize fa_un.fa_nfsv2.nfsfa_size -#define fa_nfsblocksize fa_un.fa_nfsv2.nfsfa_blocksize -#define fa_nfsrdev fa_un.fa_nfsv2.nfsfa_rdev -#define fa_nfsblocks fa_un.fa_nfsv2.nfsfa_blocks -#define fa_nfsfsid fa_un.fa_nfsv2.nfsfa_fsid -#define fa_nfsfileid fa_un.fa_nfsv2.nfsfa_fileid -#define fa_nfsatime fa_un.fa_nfsv2.nfsfa_atime -#define fa_nfsmtime fa_un.fa_nfsv2.nfsfa_mtime -#define fa_nfsctime fa_un.fa_nfsv2.nfsfa_ctime -#define fa_nqsize fa_un.fa_nqnfs.nqfa_size -#define fa_nqblocksize fa_un.fa_nqnfs.nqfa_blocksize -#define fa_nqrdev fa_un.fa_nqnfs.nqfa_rdev -#define fa_nqbytes fa_un.fa_nqnfs.nqfa_bytes -#define fa_nqfsid fa_un.fa_nqnfs.nqfa_fsid -#define fa_nqfileid fa_un.fa_nqnfs.nqfa_fileid -#define fa_nqatime fa_un.fa_nqnfs.nqfa_atime -#define fa_nqmtime fa_un.fa_nqnfs.nqfa_mtime -#define fa_nqctime fa_un.fa_nqnfs.nqfa_ctime -#define fa_nqflags fa_un.fa_nqnfs.nqfa_flags -#define fa_nqgen fa_un.fa_nqnfs.nqfa_gen -#define fa_nqfilerev fa_un.fa_nqnfs.nqfa_filerev +#define fa2_size fa_un.fa_nfsv2.nfsv2fa_size +#define fa2_blocksize fa_un.fa_nfsv2.nfsv2fa_blocksize +#define fa2_rdev fa_un.fa_nfsv2.nfsv2fa_rdev +#define fa2_blocks fa_un.fa_nfsv2.nfsv2fa_blocks +#define fa2_fsid fa_un.fa_nfsv2.nfsv2fa_fsid +#define fa2_fileid fa_un.fa_nfsv2.nfsv2fa_fileid +#define fa2_atime fa_un.fa_nfsv2.nfsv2fa_atime +#define fa2_mtime fa_un.fa_nfsv2.nfsv2fa_mtime +#define fa2_ctime fa_un.fa_nfsv2.nfsv2fa_ctime +#define fa3_size fa_un.fa_nfsv3.nfsv3fa_size +#define fa3_used fa_un.fa_nfsv3.nfsv3fa_used +#define fa3_rdev fa_un.fa_nfsv3.nfsv3fa_rdev +#define fa3_fsid fa_un.fa_nfsv3.nfsv3fa_fsid +#define fa3_fileid fa_un.fa_nfsv3.nfsv3fa_fileid +#define fa3_atime fa_un.fa_nfsv3.nfsv3fa_atime +#define fa3_mtime fa_un.fa_nfsv3.nfsv3fa_mtime +#define fa3_ctime fa_un.fa_nfsv3.nfsv3fa_ctime struct nfsv2_sattr { - u_long sa_mode; - u_long sa_uid; - u_long sa_gid; - union { - struct { - u_long nfssa_size; - struct nfsv2_time nfssa_atime; - struct nfsv2_time nfssa_mtime; - } sa_nfsv2; - struct { - struct { - u_long nqsa_qsize[2]; - } nqsa_size; - struct nqnfs_time nqsa_atime; - struct nqnfs_time nqsa_mtime; - u_long nqsa_flags; - u_long nqsa_rdev; - } sa_nqnfs; - } sa_un; + u_long sa_mode; + u_long sa_uid; + u_long sa_gid; + u_long sa_size; + nfstime2 sa_atime; + nfstime2 sa_mtime; }; -/* and some ugly defines for accessing the unions */ -#define sa_nfssize sa_un.sa_nfsv2.nfssa_size -#define sa_nfsatime sa_un.sa_nfsv2.nfssa_atime -#define sa_nfsmtime sa_un.sa_nfsv2.nfssa_mtime -#define sa_nqsize sa_un.sa_nqnfs.nqsa_size -#define sa_nqatime sa_un.sa_nqnfs.nqsa_atime -#define sa_nqmtime sa_un.sa_nqnfs.nqsa_mtime -#define sa_nqflags sa_un.sa_nqnfs.nqsa_flags -#define sa_nqrdev sa_un.sa_nqnfs.nqsa_rdev - struct nfsv2_statfs { u_long sf_tsize; u_long sf_bsize; u_long sf_blocks; u_long sf_bfree; u_long sf_bavail; - u_long sf_files; /* Nqnfs only */ - u_long sf_ffree; /* ditto */ }; +struct nfsv3_fsstat { + nfsuint64 sf_tbytes; + nfsuint64 sf_fbytes; + nfsuint64 sf_abytes; + nfsuint64 sf_tfiles; + nfsuint64 sf_ffiles; + nfsuint64 sf_afiles; + u_long sf_invarsec; +}; + +struct nfsv3_fsinfo { + u_long fs_rtmax; + u_long fs_rtpref; + u_long fs_rtmult; + u_long fs_wtmax; + u_long fs_wtpref; + u_long fs_wtmult; + u_long fs_dtpref; + nfsuint64 fs_maxfilesize; + nfstime3 fs_time_delta; + u_long fs_properties; +}; #endif diff --git a/sys/nfs/nqnfs.h b/sys/nfs/nqnfs.h index 0f8f12b..194bc90 100644 --- a/sys/nfs/nqnfs.h +++ b/sys/nfs/nqnfs.h @@ -34,7 +34,7 @@ * SUCH DAMAGE. * * @(#)nqnfs.h 8.1 (Berkeley) 6/10/93 - * $Id: nqnfs.h,v 1.5 1994/10/02 17:27:07 phk Exp $ + * $Id: nqnfs.h,v 1.6 1994/10/17 17:47:45 phk Exp $ */ #ifndef _NFS_NQNFS_H_ @@ -58,7 +58,7 @@ #define NQLCHSZ 256 /* Server hash table size */ #define NQNFS_PROG 300105 /* As assigned by Sun */ -#define NQNFS_VER1 1 +#define NQNFS_VER3 3 #define NQNFS_EVICTSIZ 156 /* Size of eviction request in bytes */ /* @@ -143,14 +143,6 @@ struct nqm { }; /* - * Flag bits for flags argument to nqsrv_getlease. - */ -#define NQL_READ LEASE_READ /* Read Request */ -#define NQL_WRITE LEASE_WRITE /* Write Request */ -#define NQL_CHECK 0x4 /* Check for lease */ -#define NQL_NOVAL 0xffffffff /* Invalid */ - -/* * Special value for slp for local server calls. */ #define NQLOCALSLP ((struct nfssvc_sock *) -1) @@ -160,9 +152,9 @@ struct nqm { */ #define nqsrv_getl(v, l) \ (void) nqsrv_getlease((v), &nfsd->nd_duration, \ - ((nfsd->nd_nqlflag != 0 && nfsd->nd_nqlflag != NQL_NOVAL) ? nfsd->nd_nqlflag : \ - ((l) | NQL_CHECK)), \ - nfsd, nam, &cache, &frev, cred) + ((nfsd->nd_flag & ND_LEASE) ? (nfsd->nd_flag & ND_LEASE) : \ + ((l) | ND_CHECK)), \ + slp, procp, nfsd->nd_nam, &cache, &frev, cred) /* * Client side macros that check for a valid lease. @@ -170,13 +162,13 @@ struct nqm { #define NQNFS_CKINVALID(v, n, f) \ ((time.tv_sec > (n)->n_expiry && \ VFSTONFS((v)->v_mount)->nm_timeouts < VFSTONFS((v)->v_mount)->nm_deadthresh) \ - || ((f) == NQL_WRITE && ((n)->n_flag & NQNFSWRITE) == 0)) + || ((f) == ND_WRITE && ((n)->n_flag & NQNFSWRITE) == 0)) #define NQNFS_CKCACHABLE(v, f) \ ((time.tv_sec <= VTONFS(v)->n_expiry || \ VFSTONFS((v)->v_mount)->nm_timeouts >= VFSTONFS((v)->v_mount)->nm_deadthresh) \ && (VTONFS(v)->n_flag & NQNFSNONCACHE) == 0 && \ - ((f) == NQL_READ || (VTONFS(v)->n_flag & NQNFSWRITE))) + ((f) == ND_READ || (VTONFS(v)->n_flag & NQNFSWRITE))) #define NQNFS_NEEDLEASE(v, p) \ (time.tv_sec > VTONFS(v)->n_expiry ? \ @@ -184,7 +176,7 @@ struct nqm { (((time.tv_sec + NQ_RENEWAL) > VTONFS(v)->n_expiry && \ nqnfs_piggy[p]) ? \ ((VTONFS(v)->n_flag & NQNFSWRITE) ? \ - NQL_WRITE : nqnfs_piggy[p]) : 0)) + ND_WRITE : nqnfs_piggy[p]) : 0)) /* * List head for timer queue. @@ -204,13 +196,12 @@ u_long nqfhhash; */ #define NQNFS_EXPIRED 500 #define NQNFS_TRYLATER 501 -#define NQNFS_AUTHERR 502 -#ifdef KERNEL -void nfs_lease_check __P((struct vnode *, struct proc *, struct ucred *, int)); -void nfs_lease_updatetime __P((int)); +#if defined(KERNEL) || defined(_KERNEL) +void nqnfs_lease_check __P((struct vnode *, struct proc *, struct ucred *, int)); +void nqnfs_lease_updatetime __P((int)); int nqsrv_cmpnam __P((struct nfssvc_sock *,struct mbuf *,struct nqhost *)); -int nqsrv_getlease __P((struct vnode *,u_long *,int,struct nfsd *,struct mbuf *,int *,u_quad_t *,struct ucred *)); +int nqsrv_getlease __P((struct vnode *,u_long *,int,struct nfssvc_sock *,struct proc *,struct mbuf *,int *,u_quad_t *,struct ucred *)); int nqnfs_getlease __P((struct vnode *,int,struct ucred *,struct proc *)); int nqnfs_callback __P((struct nfsmount *,struct mbuf *,struct mbuf *,caddr_t)); int nqnfs_clientd __P((struct nfsmount *,struct ucred *,struct nfsd_cargs *,int,caddr_t,struct proc *)); diff --git a/sys/nfs/rpcv2.h b/sys/nfs/rpcv2.h index 2360cdc..b53d5aa 100644 --- a/sys/nfs/rpcv2.h +++ b/sys/nfs/rpcv2.h @@ -34,7 +34,7 @@ * SUCH DAMAGE. * * @(#)rpcv2.h 8.1 (Berkeley) 6/10/93 - * $Id: rpcv2.h,v 1.2 1994/08/02 07:52:29 davidg Exp $ + * $Id: rpcv2.h,v 1.3 1994/08/21 06:50:15 paul Exp $ */ #ifndef _NFS_RPCV2_H_ @@ -52,10 +52,18 @@ #define RPCAUTH_NULL 0 #define RPCAUTH_UNIX 1 #define RPCAUTH_SHORT 2 +#define RPCAUTH_KERB4 4 #define RPCAUTH_NQNFS 300000 #define RPCAUTH_MAXSIZ 400 +#define RPCVERF_MAXSIZ 12 /* For Kerb, can actually be 400 */ #define RPCAUTH_UNIXGIDS 16 +/* + * Constants associated with authentication flavours. + */ +#define RPCAKN_FULLNAME 0 +#define RPCAKN_NICKNAME 1 + /* Rpc Constants */ #define RPC_CALL 0 #define RPC_REPLY 1 @@ -91,4 +99,43 @@ #define RPCMNT_PATHLEN 1024 #define RPCPROG_NFS 100003 +/* + * Structures used for RPCAUTH_KERB4. + */ +struct nfsrpc_fullverf { + u_long t1; + u_long t2; + u_long w2; +}; + +struct nfsrpc_fullblock { + u_long t1; + u_long t2; + u_long w1; + u_long w2; +}; + +struct nfsrpc_nickverf { + u_long kind; + struct nfsrpc_fullverf verf; +}; + +/* + * and their sizes in bytes.. If sizeof (struct nfsrpc_xx) != these + * constants, well then things will break in mount_nfs and nfsd. + */ +#define RPCX_FULLVERF 12 +#define RPCX_FULLBLOCK 16 +#define RPCX_NICKVERF 16 + +#ifdef NFSKERB +XXX +#else +typedef u_char NFSKERBKEY_T[2]; +typedef u_char NFSKERBKEYSCHED_T[2]; +#endif +#define NFS_KERBSRV "rcmd" /* Kerberos Service for NFS */ +#define NFS_KERBTTL (30 * 60) /* Credential ttl (sec) */ +#define NFS_KERBCLOCKSKEW (5 * 60) /* Clock skew (sec) */ +#define NFS_KERBW1(t) (*((u_long *)(&((t).dat[((t).length + 3) & ~0x3])))) #endif diff --git a/sys/nfs/xdr_subs.h b/sys/nfs/xdr_subs.h index ac00385..e8c3636 100644 --- a/sys/nfs/xdr_subs.h +++ b/sys/nfs/xdr_subs.h @@ -34,7 +34,7 @@ * SUCH DAMAGE. * * @(#)xdr_subs.h 8.1 (Berkeley) 6/10/93 - * $Id: xdr_subs.h,v 1.3 1994/08/21 06:50:16 paul Exp $ + * $Id: xdr_subs.h,v 1.4 1995/01/10 13:11:42 davidg Exp $ */ #ifndef _NFS_XDR_SUBS_H_ @@ -54,28 +54,28 @@ #define fxdr_unsigned(t, v) ((t)ntohl((long)(v))) #define txdr_unsigned(v) (htonl((long)(v))) -#define fxdr_nfstime(f, t) { \ - (t)->ts_sec = ntohl(((struct nfsv2_time *)(f))->nfs_sec); \ - if (((struct nfsv2_time *)(f))->nfs_usec != 0xffffffff) \ - (t)->ts_nsec = 1000 * ntohl(((struct nfsv2_time *)(f))->nfs_usec); \ +#define fxdr_nfsv2time(f, t) { \ + (t)->ts_sec = ntohl(((struct nfsv2_time *)(f))->nfsv2_sec); \ + if (((struct nfsv2_time *)(f))->nfsv2_usec != 0xffffffff) \ + (t)->ts_nsec = 1000 * ntohl(((struct nfsv2_time *)(f))->nfsv2_usec); \ else \ - (t)->ts_nsec = -1; \ + (t)->ts_nsec = 0; \ } -#define txdr_nfstime(f, t) { \ - ((struct nfsv2_time *)(t))->nfs_sec = htonl((f)->ts_sec); \ +#define txdr_nfsv2time(f, t) { \ + ((struct nfsv2_time *)(t))->nfsv2_sec = htonl((f)->ts_sec); \ if ((f)->ts_nsec != -1) \ - ((struct nfsv2_time *)(t))->nfs_usec = htonl((f)->ts_nsec / 1000); \ + ((struct nfsv2_time *)(t))->nfsv2_usec = htonl((f)->ts_nsec / 1000); \ else \ - ((struct nfsv2_time *)(t))->nfs_usec = 0xffffffff; \ + ((struct nfsv2_time *)(t))->nfsv2_usec = 0xffffffff; \ } -#define fxdr_nqtime(f, t) { \ - (t)->ts_sec = ntohl(((struct nqnfs_time *)(f))->nq_sec); \ - (t)->ts_nsec = ntohl(((struct nqnfs_time *)(f))->nq_nsec); \ +#define fxdr_nfsv3time(f, t) { \ + (t)->ts_sec = ntohl(((struct nfsv3_time *)(f))->nfsv3_sec); \ + (t)->ts_nsec = ntohl(((struct nfsv3_time *)(f))->nfsv3_nsec); \ } -#define txdr_nqtime(f, t) { \ - ((struct nqnfs_time *)(t))->nq_sec = htonl((f)->ts_sec); \ - ((struct nqnfs_time *)(t))->nq_nsec = htonl((f)->ts_nsec); \ +#define txdr_nfsv3time(f, t) { \ + ((struct nfsv3_time *)(t))->nfsv3_sec = htonl((f)->ts_sec); \ + ((struct nfsv3_time *)(t))->nfsv3_nsec = htonl((f)->ts_nsec); \ } #define fxdr_hyper(f, t) { \ diff --git a/sys/nfsclient/nfs.h b/sys/nfsclient/nfs.h index cbf80c6..e1a0f07 100644 --- a/sys/nfsclient/nfs.h +++ b/sys/nfsclient/nfs.h @@ -34,7 +34,7 @@ * SUCH DAMAGE. * * @(#)nfs.h 8.1 (Berkeley) 6/10/93 - * $Id: nfs.h,v 1.8 1994/11/02 00:11:00 wollman Exp $ + * $Id: nfs.h,v 1.9 1995/02/14 06:22:18 phk Exp $ */ #ifndef _NFS_NFS_H_ @@ -45,11 +45,12 @@ */ #define NFS_MAXIOVEC 34 -#define NFS_HZ 25 /* Ticks per second for NFS timeouts */ -#define NFS_TIMEO (1*NFS_HZ) /* Default timeout = 1 second */ -#define NFS_MINTIMEO (1*NFS_HZ) /* Min timeout to use */ -#define NFS_MAXTIMEO (60*NFS_HZ) /* Max timeout to backoff to */ -#define NFS_MINIDEMTIMEO (5*NFS_HZ) /* Min timeout for non-idempotent ops*/ +#define NFS_TICKINTVL 5 /* Desired time for a tick (msec) */ +#define NFS_HZ (hz / nfs_ticks) /* Ticks/sec */ +#define NFS_TIMEO (1 * NFS_HZ) /* Default timeout = 1 second */ +#define NFS_MINTIMEO (1 * NFS_HZ) /* Min timeout to use */ +#define NFS_MAXTIMEO (60 * NFS_HZ) /* Max timeout to backoff to */ +#define NFS_MINIDEMTIMEO (5 * NFS_HZ) /* Min timeout for non-idempotent ops*/ #define NFS_MAXREXMIT 100 /* Stop counting after this many */ #define NFS_MAXWINDOW 1024 /* Max number of outstanding requests */ #define NFS_RETRANS 10 /* Num of retrans for soft mounts */ @@ -62,13 +63,81 @@ #endif #define NFS_WSIZE 8192 /* Def. write data size <= 8192 */ #define NFS_RSIZE 8192 /* Def. read data size <= 8192 */ +#define NFS_READDIRSIZE 8192 /* Def. readdir size */ #define NFS_DEFRAHEAD 1 /* Def. read ahead # blocks */ #define NFS_MAXRAHEAD 4 /* Max. read ahead # blocks */ -#define NFS_MAXREADDIR NFS_MAXDATA /* Max. size of directory read */ #define NFS_MAXUIDHASH 64 /* Max. # of hashed uid entries/mp */ -#define NFS_MAXASYNCDAEMON 20 /* Max. number async_daemons runable */ -#define NFS_DIRBLKSIZ 1024 /* Size of an NFS directory block */ +#define NFS_MAXASYNCDAEMON 20 /* Max. number async_daemons runable */ +#define NFS_MAXGATHERDELAY 100 /* Max. write gather delay (msec) */ +#ifndef NFS_GATHERDELAY +#define NFS_GATHERDELAY 10 /* Default write gather delay (msec) */ +#endif +#define NFS_DIRBLKSIZ 4096 /* Must be a multiple of DIRBLKSIZ */ + +/* + * Oddballs + */ #define NMOD(a) ((a) % nfs_asyncdaemons) +#define NFS_CMPFH(n, f, s) \ + ((n)->n_fhsize == (s) && !bcmp((caddr_t)(n)->n_fhp, (caddr_t)(f), (s))) +#define NFS_ISV3(v) (VFSTONFS((v)->v_mount)->nm_flag & NFSMNT_NFSV3) +#define NFS_SRVMAXDATA(n) \ + (((n)->nd_flag & ND_NFSV3) ? (((n)->nd_nam2) ? \ + NFS_MAXDGRAMDATA : NFS_MAXDATA) : NFS_V2MAXDATA) + +/* + * XXX + * sys/buf.h should be editted to change B_APPENDWRITE --> B_NEEDCOMMIT, but + * until then... + * Same goes for sys/malloc.h, which needs M_NFSDIROFF, + * M_NFSRVDESC and M_NFSBIGFH added. + * The VA_EXCLUSIVE flag should be added for va_vaflags and set for an + * exclusive create. + * The B_INVAFTERWRITE flag should be set to whatever is required by the + * buffer cache code to say "Invalidate the block after it is written back". + */ +#ifndef B_NEEDCOMMIT +#define B_NEEDCOMMIT B_APPENDWRITE +#endif +#ifndef M_NFSRVDESC +#define M_NFSRVDESC M_TEMP +#endif +#ifndef M_NFSDIROFF +#define M_NFSDIROFF M_TEMP +#endif +#ifndef M_NFSBIGFH +#define M_NFSBIGFH M_TEMP +#endif +#ifndef VA_EXCLUSIVE +#define VA_EXCLUSIVE 0 +#endif +#ifdef __FreeBSD__ +#define B_INVAFTERWRITE B_NOCACHE +#else +#define B_INVAFTERWRITE B_INVAL +#endif + +/* + * These ifdefs try to handle the differences between the various 4.4BSD-Lite + * based vfs interfaces. + * btw: NetBSD-current does have a VOP_LEASDE(), but I don't know how to + * differentiate between NetBSD-1.0 and NetBSD-current, so.. + * I also don't know about BSDi's 2.0 release. + */ +#if !defined(HAS_VOPLEASE) && !defined(__FreeBSD__) && !defined(__NetBSD__) +#define HAS_VOPLEASE 1 +#endif +#if !defined(HAS_VOPREVOKE) && !defined(__FreeBSD__) && !defined(__NetBSD__) +#define HAS_VOPREVOKE 1 +#endif + +/* + * The IO_METASYNC flag should be implemented for local file systems. + * (Until then, it is nothin at all.) + */ +#ifndef IO_METASYNC +#define IO_METASYNC 0 +#endif /* * Set the attribute timeout based on how recently the file has been modified. @@ -80,6 +149,20 @@ (time.tv_sec - (np)->n_mtime) / 10)) /* + * Expected allocation sizes for major data structures. If the actual size + * of the structure exceeds these sizes, then malloc() will be allocating + * almost twice the memory required. This is used in nfs_init() to warn + * the sysadmin that the size of a structure should be reduced. + * (These sizes are always a power of 2. If the kernel malloc() changes + * to one that does not allocate space in powers of 2 size, then this all + * becomes bunk!) + */ +#define NFS_NODEALLOC 256 +#define NFS_MNTALLOC 512 +#define NFS_SVCALLOC 256 +#define NFS_UIDALLOC 128 + +/* * Structures for the nfssvc(2) syscall. Not that anyone but nfsd and mount_nfs * should ever try and use it. */ @@ -95,7 +178,12 @@ struct nfsd_srvargs { u_long nsd_haddr; /* Ip address of client */ struct ucred nsd_cr; /* Cred. uid maps to */ int nsd_authlen; /* Length of auth string (ret) */ - char *nsd_authstr; /* Auth string (ret) */ + u_char *nsd_authstr; /* Auth string (ret) */ + int nsd_verflen; /* and the verfier */ + u_char *nsd_verfstr; + struct timeval nsd_timestamp; /* timestamp from verifier */ + u_long nsd_ttl; /* credential ttl (sec) */ + NFSKERBKEY_T nsd_key; /* Session key */ }; struct nfsd_cargs { @@ -103,7 +191,10 @@ struct nfsd_cargs { uid_t ncd_authuid; /* Effective uid */ int ncd_authtype; /* Type of authenticator */ int ncd_authlen; /* Length of authenticator string */ - char *ncd_authstr; /* Authenticator string */ + u_char *ncd_authstr; /* Authenticator string */ + int ncd_verflen; /* and the verifier */ + u_char *ncd_verfstr; + NFSKERBKEY_T ncd_key; /* Session key */ }; /* @@ -142,6 +233,7 @@ struct nfsstats { int srvnqnfs_leases; int srvnqnfs_maxleases; int srvnqnfs_getleases; + int srvvop_writes; }; /* @@ -173,7 +265,7 @@ struct nfsstats { * such as SIGALRM will not expect file I/O system calls to be interrupted * by them and break. */ -#ifdef KERNEL +#if defined(KERNEL) || defined(_KERNEL) struct uio; struct buf; struct vattr; struct nameidata; /* XXX */ @@ -224,17 +316,29 @@ TAILQ_HEAD(, nfsreq) nfs_reqq; #define R_MUSTRESEND 0x40 /* Must resend request */ #define R_GETONEREP 0x80 /* Probe for one reply only */ -extern struct nfsstats nfsstats; - /* * A list of nfssvc_sock structures is maintained with all the sockets * that require service by the nfsd. * The nfsuid structs hang off of the nfssvc_sock structs in both lru * and uid hash lists. */ -#define NUIDHASHSIZ 32 +#ifndef NFS_UIDHASHSIZ +#define NFS_UIDHASHSIZ 29 /* Tune the size of nfssvc_sock with this */ +#endif #define NUIDHASH(sock, uid) \ - (&(sock)->ns_uidhashtbl[(uid) & (sock)->ns_uidhash]) + (&(sock)->ns_uidhashtbl[(uid) % NFS_UIDHASHSIZ]) +#ifndef NFS_WDELAYHASHSIZ +#define NFS_WDELAYHASHSIZ 16 /* and with this */ +#endif +#define NWDELAYHASH(sock, f) \ + (&(sock)->ns_wdelayhashtbl[(*((u_long *)(f))) % NFS_WDELAYHASHSIZ]) +#ifndef NFS_MUIDHASHSIZ +#define NFS_MUIDHASHSIZ 67 /* Tune the size of nfsmount with this */ +#endif +#define NMUIDHASH(nmp, uid) \ + (&(nmp)->nm_uidhashtbl[(uid) % NFS_MUIDHASHSIZ]) +#define NFSNOHASH(fhsum) \ + (&nfsnodehashtbl[(fhsum) & nfsnodehash]) /* * Network address hash list element @@ -248,35 +352,41 @@ struct nfsuid { TAILQ_ENTRY(nfsuid) nu_lru; /* LRU chain */ LIST_ENTRY(nfsuid) nu_hash; /* Hash list */ int nu_flag; /* Flags */ - uid_t nu_uid; /* Uid mapped by this entry */ union nethostaddr nu_haddr; /* Host addr. for dgram sockets */ struct ucred nu_cr; /* Cred uid mapped to */ + int nu_expire; /* Expiry time (sec) */ + struct timeval nu_timestamp; /* Kerb. timestamp */ + u_long nu_nickname; /* Nickname on server */ + NFSKERBKEY_T nu_key; /* and session key */ }; #define nu_inetaddr nu_haddr.had_inetaddr #define nu_nam nu_haddr.had_nam /* Bits for nu_flag */ #define NU_INETADDR 0x1 +#define NU_NAM 0x2 +#define NU_NETFAM(u) (((u)->nu_flag & NU_INETADDR) ? AF_INET : AF_ISO) struct nfssvc_sock { TAILQ_ENTRY(nfssvc_sock) ns_chain; /* List of all nfssvc_sock's */ TAILQ_HEAD(, nfsuid) ns_uidlruhead; - LIST_HEAD(, nfsuid) *ns_uidhashtbl; - u_long ns_uidhash; - - int ns_flag; - u_long ns_sref; struct file *ns_fp; struct socket *ns_so; - int ns_solock; struct mbuf *ns_nam; - int ns_cc; struct mbuf *ns_raw; struct mbuf *ns_rawend; - int ns_reclen; struct mbuf *ns_rec; struct mbuf *ns_recend; + struct mbuf *ns_frag; + int ns_flag; + int ns_solock; + int ns_cc; + int ns_reclen; int ns_numuids; + u_long ns_sref; + LIST_HEAD(, nfsrv_descript) ns_tq; /* Write gather lists */ + LIST_HEAD(, nfsuid) ns_uidhashtbl[NFS_UIDHASHSIZ]; + LIST_HEAD(nfsrvw_delayhash, nfsrv_descript) ns_wdelayhashtbl[NFS_WDELAYHASHSIZ]; }; /* Bits for "ns_flag" */ @@ -285,6 +395,7 @@ struct nfssvc_sock { #define SLP_NEEDQ 0x04 #define SLP_DISCONN 0x08 #define SLP_GETSTREAM 0x10 +#define SLP_LASTFRAG 0x20 #define SLP_ALLFLAGS 0xff TAILQ_HEAD(, nfssvc_sock) nfssvc_sockhead; @@ -296,73 +407,124 @@ int nfssvc_sockhead_flag; * One of these structures is allocated for each nfsd. */ struct nfsd { - TAILQ_ENTRY(nfsd) nd_chain; /* List of all nfsd's */ - int nd_flag; /* NFSD_ flags */ - struct nfssvc_sock *nd_slp; /* Current socket */ - struct mbuf *nd_nam; /* Client addr for datagram req. */ - struct mbuf *nd_mrep; /* Req. mbuf list */ - struct mbuf *nd_md; - caddr_t nd_dpos; /* Position in list */ - int nd_procnum; /* RPC procedure number */ - u_long nd_retxid; /* RPC xid */ - int nd_repstat; /* Reply status value */ - struct ucred nd_cr; /* Credentials for req. */ - int nd_nqlflag; /* Leasing flag */ - u_long nd_duration; /* Lease duration */ - int nd_authlen; /* Authenticator len */ - u_char nd_authstr[RPCAUTH_MAXSIZ]; /* Authenticator data */ - struct proc *nd_procp; /* Proc ptr */ + TAILQ_ENTRY(nfsd) nfsd_chain; /* List of all nfsd's */ + int nfsd_flag; /* NFSD_ flags */ + struct nfssvc_sock *nfsd_slp; /* Current socket */ + int nfsd_authlen; /* Authenticator len */ + u_char nfsd_authstr[RPCAUTH_MAXSIZ]; /* Authenticator data */ + int nfsd_verflen; /* and the Verifier */ + u_char nfsd_verfstr[RPCVERF_MAXSIZ]; + struct proc *nfsd_procp; /* Proc ptr */ + struct nfsrv_descript *nfsd_nd; /* Associated nfsrv_descript */ }; -/* Bits for "nd_flag" */ +/* Bits for "nfsd_flag" */ #define NFSD_WAITING 0x01 #define NFSD_REQINPROG 0x02 #define NFSD_NEEDAUTH 0x04 #define NFSD_AUTHFAIL 0x08 +/* + * This structure is used by the server for describing each request. + * Some fields are used only when write request gathering is performed. + */ +struct nfsrv_descript { + u_quad_t nd_time; /* Write deadline (usec) */ + off_t nd_off; /* Start byte offset */ + off_t nd_eoff; /* and end byte offset */ + LIST_ENTRY(nfsrv_descript) nd_hash; /* Hash list */ + LIST_ENTRY(nfsrv_descript) nd_tq; /* and timer list */ + LIST_HEAD(,nfsrv_descript) nd_coalesce; /* coalesced writes */ + struct mbuf *nd_mrep; /* Request mbuf list */ + struct mbuf *nd_md; /* Current dissect mbuf */ + struct mbuf *nd_mreq; /* Reply mbuf list */ + struct mbuf *nd_nam; /* and socket addr */ + struct mbuf *nd_nam2; /* return socket addr */ + caddr_t nd_dpos; /* Current dissect pos */ + int nd_procnum; /* RPC # */ + int nd_stable; /* storage type */ + int nd_flag; /* nd_flag */ + int nd_len; /* Length of this write */ + int nd_repstat; /* Reply status */ + u_long nd_retxid; /* Reply xid */ + u_long nd_duration; /* Lease duration */ + struct timeval nd_starttime; /* Time RPC initiated */ + fhandle_t nd_fh; /* File handle */ + struct ucred nd_cr; /* Credentials */ +}; + +/* Bits for "nd_flag" */ +#define ND_READ LEASE_READ +#define ND_WRITE LEASE_WRITE +#define ND_CHECK 0x04 +#define ND_LEASE (ND_READ | ND_WRITE | ND_CHECK) +#define ND_NFSV3 0x08 +#define ND_NQNFS 0x10 +#define ND_KERBNICK 0x20 +#define ND_KERBFULL 0x40 +#define ND_KERBAUTH (ND_KERBNICK | ND_KERBFULL) + TAILQ_HEAD(, nfsd) nfsd_head; int nfsd_head_flag; #define NFSD_CHECKSLP 0x01 +/* + * These macros compare nfsrv_descript structures. + */ +#define NFSW_CONTIG(o, n) \ + ((o)->nd_eoff >= (n)->nd_off && \ + !bcmp((caddr_t)&(o)->nd_fh, (caddr_t)&(n)->nd_fh, NFSX_V3FH)) + +#define NFSW_SAMECRED(o, n) \ + (((o)->nd_flag & ND_KERBAUTH) == ((n)->nd_flag & ND_KERBAUTH) && \ + !bcmp((caddr_t)&(o)->nd_cr, (caddr_t)&(n)->nd_cr, \ + sizeof (struct ucred))) + int nfs_reply __P((struct nfsreq *)); -int nfs_getreq __P((struct nfsd *,int)); +int nfs_getreq __P((struct nfsrv_descript *,struct nfsd *,int)); int nfs_send __P((struct socket *,struct mbuf *,struct mbuf *,struct nfsreq *)); -int nfs_rephead __P((int,struct nfsd *,int,int,u_quad_t *,struct mbuf **,struct mbuf **,caddr_t *)); +int nfs_rephead __P((int,struct nfsrv_descript *,struct nfssvc_sock *,int,int,u_quad_t *,struct mbuf **,struct mbuf **,caddr_t *)); int nfs_sndlock __P((int *,struct nfsreq *)); int nfs_disct __P((struct mbuf **,caddr_t *,int,int,caddr_t *)); int nfs_vinvalbuf __P((struct vnode *,int,struct ucred *,struct proc *,int)); int nfs_readrpc __P((struct vnode *,struct uio *,struct ucred *)); -int nfs_writerpc __P((struct vnode *,struct uio *,struct ucred *,int)); +int nfs_writerpc __P((struct vnode *,struct uio *,struct ucred *,int *,int *)); int nfs_readdirrpc __P((register struct vnode *,struct uio *,struct ucred *)); +int nfs_setattrrpc __P((struct vnode *,struct vattr *,struct ucred *,struct proc *)); int nfs_asyncio __P((struct buf *,struct ucred *)); int nfs_doio __P((struct buf *,struct ucred *,struct proc *)); int nfs_readlinkrpc __P((struct vnode *,struct uio *,struct ucred *)); int nfs_sigintr __P((struct nfsmount *,struct nfsreq *r,struct proc *)); -int nfs_readdirlookrpc __P((struct vnode *,register struct uio *,struct ucred *)); +int nfs_readdirplusrpc __P((struct vnode *,register struct uio *,struct ucred *)); int nfsm_disct __P((struct mbuf **,caddr_t *,int,int,caddr_t *)); -int nfsrv_fhtovp __P((fhandle_t *,int,struct vnode **,struct ucred *,struct nfssvc_sock *,struct mbuf *,int *)); +void nfsm_srvfattr __P((struct nfsrv_descript *,struct vattr *,struct nfs_fattr *)); +void nfsm_srvwcc __P((struct nfsrv_descript *,int,struct vattr *,int,struct vattr *,struct mbuf **,char **)); +void nfsm_srvpostopattr __P((struct nfsrv_descript *,int,struct vattr *,struct mbuf **,char **)); +int nfsrv_fhtovp __P((fhandle_t *,int,struct vnode **,struct ucred *,struct nfssvc_sock *,struct mbuf *,int *,int)); int nfsrv_access __P((struct vnode *,int,struct ucred *,int,struct proc *)); int netaddr_match __P((int,union nethostaddr *,struct mbuf *)); int nfs_request __P((struct vnode *,struct mbuf *,int,struct proc *,struct ucred *,struct mbuf **,struct mbuf **,caddr_t *)); int nfs_loadattrcache __P((struct vnode **,struct mbuf **,caddr_t *,struct vattr *)); -int nfs_namei __P((struct nameidata *,fhandle_t *,int,struct nfssvc_sock *,struct mbuf *,struct mbuf **,caddr_t *,struct proc *)); +int nfs_namei __P((struct nameidata *,fhandle_t *,int,struct nfssvc_sock *,struct mbuf *,struct mbuf **,caddr_t *,struct vnode **,struct proc *,int)); void nfsm_adj __P((struct mbuf *,int,int)); int nfsm_mbuftouio __P((struct mbuf **,struct uio *,int,caddr_t *)); void nfsrv_initcache __P((void)); int nfs_rcvlock __P((struct nfsreq *)); -int nfs_getauth __P((struct nfsmount *,struct nfsreq *,struct ucred *,int *,char **,int *)); +int nfs_getauth __P((struct nfsmount *,struct nfsreq *,struct ucred *,char **,int *,char *,int *,NFSKERBKEY_T)); +int nfs_getnickauth __P((struct nfsmount *,struct ucred *,char **,int *,char *,int)); +int nfs_savenickauth __P((struct nfsmount *,struct ucred *,int,NFSKERBKEY_T,struct mbuf **,char **,struct mbuf *)); int nfs_msg __P((struct proc *,char *,char *)); int nfs_adv __P((struct mbuf **,caddr_t *,int,int)); int nfsrv_getstream __P((struct nfssvc_sock *,int)); void nfs_nhinit __P((void)); void nfs_timer __P((void*)); -struct nfsnodehashhead * nfs_hash __P((nfsv2fh_t *)); +u_long nfs_hash __P((nfsfh_t *,int)); int nfssvc_iod __P((struct proc *)); int nfssvc_nfsd __P((struct nfsd_srvargs *,caddr_t,struct proc *)); int nfssvc_addsock __P((struct file *,struct mbuf *)); -int nfsrv_dorec __P((struct nfssvc_sock *,struct nfsd *)); -int nfsrv_getcache __P((struct mbuf *,struct nfsd *,struct mbuf **)); -void nfsrv_updatecache __P((struct mbuf *,struct nfsd *,int,struct mbuf *)); +int nfsrv_dorec __P((struct nfssvc_sock *,struct nfsd *,struct nfsrv_descript **)); +int nfsrv_getcache __P((struct nfsrv_descript *,struct nfssvc_sock *,struct mbuf **)); +void nfsrv_updatecache __P((struct nfsrv_descript *,int,struct mbuf *)); int mountnfs __P((struct nfs_args *,struct mount *,struct mbuf *,char *,char *,struct vnode **)); int nfs_connect __P((struct nfsmount *,struct nfsreq *)); int nfs_getattrcache __P((struct vnode *,struct vattr *)); @@ -370,9 +532,20 @@ int nfsm_strtmbuf __P((struct mbuf **,char **,char *,long)); int nfs_bioread __P((struct vnode *,struct uio *,int,struct ucred *)); int nfsm_uiotombuf __P((struct uio *,struct mbuf **,int,caddr_t *)); void nfsrv_init __P((int)); +void nfs_clearcommit __P((struct mount *)); +int nfsrv_errmap __P((struct nfsrv_descript *, int)); +void nfsrvw_coalesce __P((struct nfsrv_descript *,struct nfsrv_descript *)); +void nfsrvw_sort __P((gid_t [],int)); +void nfsrv_setcred __P((struct ucred *,struct ucred *)); +int nfs_flush __P((struct vnode *,struct ucred *,int,struct proc *,int)); +int nfs_writebp __P((struct buf *,int)); int nfsrv_vput __P(( struct vnode * )); int nfsrv_vrele __P(( struct vnode * )); int nfsrv_vmio __P(( struct vnode * )); +int nfsrv_writegather __P((struct nfsrv_descript **, struct nfssvc_sock *, + struct proc *, struct mbuf **)); +int nfs_fsinfo __P((struct nfsmount *, struct vnode *, struct ucred *, + struct proc *p)); #endif /* KERNEL */ diff --git a/sys/nfsclient/nfs_bio.c b/sys/nfsclient/nfs_bio.c index f929012..f9ae8b6 100644 --- a/sys/nfsclient/nfs_bio.c +++ b/sys/nfsclient/nfs_bio.c @@ -34,7 +34,7 @@ * SUCH DAMAGE. * * @(#)nfs_bio.c 8.5 (Berkeley) 1/4/94 - * $Id: nfs_bio.c,v 1.13 1995/05/21 21:39:21 davidg Exp $ + * $Id: nfs_bio.c,v 1.14 1995/05/30 08:12:35 rgrimes Exp $ */ #include <sys/param.h> @@ -49,16 +49,30 @@ #include <vm/vm.h> -#include <nfs/nfsnode.h> #include <nfs/rpcv2.h> -#include <nfs/nfsv2.h> +#include <nfs/nfsproto.h> #include <nfs/nfs.h> #include <nfs/nfsmount.h> #include <nfs/nqnfs.h> +#include <nfs/nfsnode.h> struct buf *nfs_getcacheblk(); extern struct proc *nfs_iodwant[NFS_MAXASYNCDAEMON]; extern int nfs_numasync; +extern struct nfsstats nfsstats; + +/* + * Ifdefs for FreeBSD-current's merged VM/buffer cache. It is unfortunate + * that this isn't done inside getblk() and brelse() so these calls + * wouldn't need to be here. + */ +#ifdef B_VMIO +#define vnode_pager_uncache(vp) +#else +#define vfs_busy_pages(bp, f) +#define vfs_unbusy_pages(bp) +#define vfs_dirty_pages(bp) +#endif /* * Vnode op for read using bio @@ -72,29 +86,28 @@ nfs_bioread(vp, uio, ioflag, cred) struct ucred *cred; { register struct nfsnode *np = VTONFS(vp); - register int biosize, diff; + register int biosize, diff, i; struct buf *bp = 0, *rabp; struct vattr vattr; struct proc *p; - struct nfsmount *nmp; + struct nfsmount *nmp = VFSTONFS(vp->v_mount); daddr_t lbn, rabn; int bufsize; int nra, error = 0, n = 0, on = 0, not_readin; + nfsquad_t tquad; -#ifdef lint - ioflag = ioflag; -#endif /* lint */ #ifdef DIAGNOSTIC if (uio->uio_rw != UIO_READ) panic("nfs_read mode"); #endif if (uio->uio_resid == 0) return (0); - if (uio->uio_offset < 0 && vp->v_type != VDIR) + if (uio->uio_offset < 0) return (EINVAL); - nmp = VFSTONFS(vp->v_mount); - biosize = NFS_MAXDGRAMDATA; p = uio->uio_procp; + if ((nmp->nm_flag & (NFSMNT_NFSV3 | NFSMNT_GOTFSINFO)) == NFSMNT_NFSV3) + (void)nfs_fsinfo(nmp, vp, cred, p); + biosize = nmp->nm_rsize; /* * For nfs, cache consistency can only be maintained approximately. * Although RFC1094 does not specify the criteria, the following is @@ -107,8 +120,6 @@ nfs_bioread(vp, uio, ioflag, cred) * server, so flush all of the file's data out of the cache. * Then force a getattr rpc to ensure that you have up to date * attributes. - * The mount flag NFSMNT_MYWRITE says "Assume that my writes are - * the ones changing the modify time. * NB: This implies that cache data can be read when up to * NFS_ATTRTIMEO seconds out of date. If you find that you need current * attributes this could be forced by setting n_attrstamp to 0 before @@ -116,14 +127,15 @@ nfs_bioread(vp, uio, ioflag, cred) */ if ((nmp->nm_flag & NFSMNT_NQNFS) == 0 && vp->v_type != VLNK) { if (np->n_flag & NMODIFIED) { - if ((nmp->nm_flag & NFSMNT_MYWRITE) == 0 || - vp->v_type != VREG) { + if (vp->v_type != VREG) { + if (vp->v_type != VDIR) + panic("nfs: bioread, not dir"); + nfs_invaldir(vp); error = nfs_vinvalbuf(vp, V_SAVE, cred, p, 1); if (error) return (error); } np->n_attrstamp = 0; - np->n_direofoffset = 0; error = VOP_GETATTR(vp, &vattr, cred, p); if (error) return (error); @@ -133,7 +145,8 @@ nfs_bioread(vp, uio, ioflag, cred) if (error) return (error); if (np->n_mtime != vattr.va_mtime.ts_sec) { - np->n_direofoffset = 0; + if (vp->v_type == VDIR) + nfs_invaldir(vp); error = nfs_vinvalbuf(vp, V_SAVE, cred, p, 1); if (error) return (error); @@ -147,27 +160,24 @@ nfs_bioread(vp, uio, ioflag, cred) * Get a valid lease. If cached data is stale, flush it. */ if (nmp->nm_flag & NFSMNT_NQNFS) { - if (NQNFS_CKINVALID(vp, np, NQL_READ)) { + if (NQNFS_CKINVALID(vp, np, ND_READ)) { do { - error = nqnfs_getlease(vp, NQL_READ, cred, p); + error = nqnfs_getlease(vp, ND_READ, cred, p); } while (error == NQNFS_EXPIRED); if (error) return (error); if (np->n_lrev != np->n_brev || (np->n_flag & NQNFSNONCACHE) || ((np->n_flag & NMODIFIED) && vp->v_type == VDIR)) { - if (vp->v_type == VDIR) { - np->n_direofoffset = 0; - cache_purge(vp); - } + if (vp->v_type == VDIR) + nfs_invaldir(vp); error = nfs_vinvalbuf(vp, V_SAVE, cred, p, 1); if (error) return (error); np->n_brev = np->n_lrev; } } else if (vp->v_type == VDIR && (np->n_flag & NMODIFIED)) { - np->n_direofoffset = 0; - cache_purge(vp); + nfs_invaldir(vp); error = nfs_vinvalbuf(vp, V_SAVE, cred, p, 1); if (error) return (error); @@ -176,33 +186,27 @@ nfs_bioread(vp, uio, ioflag, cred) if (np->n_flag & NQNFSNONCACHE) { switch (vp->v_type) { case VREG: - error = nfs_readrpc(vp, uio, cred); - break; + return (nfs_readrpc(vp, uio, cred)); case VLNK: - error = nfs_readlinkrpc(vp, uio, cred); - break; + return (nfs_readlinkrpc(vp, uio, cred)); case VDIR: - error = nfs_readdirrpc(vp, uio, cred); break; default: printf(" NQNFSNONCACHE: type %x unexpected\n", vp->v_type); - break; }; - return (error); } switch (vp->v_type) { case VREG: nfsstats.biocache_reads++; lbn = uio->uio_offset / biosize; - on = uio->uio_offset & (biosize-1); + on = uio->uio_offset & (biosize - 1); not_readin = 1; /* * Start the read ahead(s), as required. */ - if (nfs_numasync > 0 && nmp->nm_readahead > 0 && - lbn == vp->v_lastr + 1) { + if (nfs_numasync > 0 && nmp->nm_readahead > 0) { for (nra = 0; nra < nmp->nm_readahead && (lbn + 1 + nra) * biosize < np->n_size; nra++) { rabn = lbn + 1 + nra; @@ -296,32 +300,55 @@ again: break; case VDIR: nfsstats.biocache_readdirs++; - lbn = (daddr_t)uio->uio_offset; + lbn = uio->uio_offset / NFS_DIRBLKSIZ; + on = uio->uio_offset & (NFS_DIRBLKSIZ - 1); bp = nfs_getcacheblk(vp, lbn, NFS_DIRBLKSIZ, p); if (!bp) - return (EINTR); - + return (EINTR); if ((bp->b_flags & B_CACHE) == 0) { - bp->b_flags |= B_READ; - vfs_busy_pages(bp, 0); - error = nfs_doio(bp, cred, p); - if (error) { - bp->b_flags |= B_ERROR; - brelse(bp); - return (error); + bp->b_flags |= B_READ; + vfs_busy_pages(bp, 0); + error = nfs_doio(bp, cred, p); + if (error) { + brelse(bp); + while (error == NFSERR_BAD_COOKIE) { + nfs_invaldir(vp); + error = nfs_vinvalbuf(vp, 0, cred, p, 1); + /* + * Yuck! The directory has been modified on the + * server. The only way to get the block is by + * reading from the beginning to get all the + * offset cookies. + */ + for (i = 0; i <= lbn && !error; i++) { + bp = nfs_getcacheblk(vp, i, NFS_DIRBLKSIZ, p); + if (!bp) + return (EINTR); + if ((bp->b_flags & B_DONE) == 0) { + bp->b_flags |= B_READ; + vfs_busy_pages(bp, 0); + error = nfs_doio(bp, cred, p); + if (error) + brelse(bp); + } + } } + if (error) + return (error); + } } /* * If not eof and read aheads are enabled, start one. * (You need the current block first, so that you have the - * directory offset cookie of the next block. + * directory offset cookie of the next block.) */ - rabn = bp->b_blkno; if (nfs_numasync > 0 && nmp->nm_readahead > 0 && - rabn != 0 && rabn != np->n_direofoffset && - !incore(vp, rabn)) { - rabp = nfs_getcacheblk(vp, rabn, NFS_DIRBLKSIZ, p); + (np->n_direofoffset == 0 || + (lbn + 1) * NFS_DIRBLKSIZ < np->n_direofoffset) && + !(np->n_flag & NQNFSNONCACHE) && + !incore(vp, lbn + 1)) { + rabp = nfs_getcacheblk(vp, lbn + 1, NFS_DIRBLKSIZ, p); if (rabp) { if ((rabp->b_flags & (B_CACHE|B_DELWRI)) == 0) { rabp->b_flags |= (B_READ | B_ASYNC); @@ -336,11 +363,10 @@ again: } } } - on = 0; - n = min(uio->uio_resid, NFS_DIRBLKSIZ - bp->b_resid); + n = min(uio->uio_resid, NFS_DIRBLKSIZ - bp->b_resid - on); break; default: - printf(" nfsbioread: type %x unexpected\n",vp->v_type); + printf(" nfs_bioread: type %x unexpected\n",vp->v_type); break; }; @@ -354,11 +380,11 @@ again: n = 0; break; case VDIR: - uio->uio_offset = bp->b_blkno; + if (np->n_flag & NQNFSNONCACHE) + bp->b_flags |= B_INVAL; break; default: - printf(" nfsbioread: type %x unexpected\n",vp->v_type); - break; + printf(" nfs_bioread: type %x unexpected\n",vp->v_type); } brelse(bp); } while (error == 0 && uio->uio_resid > 0 && n > 0); @@ -386,10 +412,10 @@ nfs_write(ap) int ioflag = ap->a_ioflag; struct buf *bp; struct vattr vattr; - struct nfsmount *nmp; - daddr_t lbn; + struct nfsmount *nmp = VFSTONFS(vp->v_mount); + daddr_t lbn, bn; int bufsize; - int n, on, error = 0; + int n, on, error = 0, iomode, must_commit; #ifdef DIAGNOSTIC if (uio->uio_rw != UIO_WRITE) @@ -403,6 +429,8 @@ nfs_write(ap) np->n_flag &= ~NWRITEERR; return (np->n_error); } + if ((nmp->nm_flag & (NFSMNT_NFSV3 | NFSMNT_GOTFSINFO)) == NFSMNT_NFSV3) + (void)nfs_fsinfo(nmp, vp, cred, p); if (ioflag & (IO_APPEND | IO_SYNC)) { if (np->n_flag & NMODIFIED) { np->n_attrstamp = 0; @@ -418,7 +446,6 @@ nfs_write(ap) uio->uio_offset = np->n_size; } } - nmp = VFSTONFS(vp->v_mount); if (uio->uio_offset < 0) return (EINVAL); if (uio->uio_resid == 0) @@ -437,7 +464,7 @@ nfs_write(ap) * will be the same size within a filesystem. nfs_writerpc will * still use nm_wsize when sizing the rpc's. */ - biosize = NFS_MAXDGRAMDATA; + biosize = nmp->nm_rsize; do { /* @@ -445,12 +472,11 @@ nfs_write(ap) */ /* * Check for a valid write lease. - * If non-cachable, just do the rpc */ if ((nmp->nm_flag & NFSMNT_NQNFS) && - NQNFS_CKINVALID(vp, np, NQL_WRITE)) { + NQNFS_CKINVALID(vp, np, ND_WRITE)) { do { - error = nqnfs_getlease(vp, NQL_WRITE, cred, p); + error = nqnfs_getlease(vp, ND_WRITE, cred, p); } while (error == NQNFS_EXPIRED); if (error) return (error); @@ -462,8 +488,13 @@ nfs_write(ap) np->n_brev = np->n_lrev; } } - if (np->n_flag & NQNFSNONCACHE) - return (nfs_writerpc(vp, uio, cred, ioflag)); + if ((np->n_flag & NQNFSNONCACHE) && uio->uio_iovcnt == 1) { + iomode = NFSV3WRITE_FILESYNC; + error = nfs_writerpc(vp, uio, cred, &iomode, &must_commit); + if (must_commit) + nfs_clearcommit(vp->v_mount); + return (error); + } nfsstats.biocache_writes++; lbn = uio->uio_offset / biosize; on = uio->uio_offset & (biosize-1); @@ -509,9 +540,9 @@ again: * In case getblk() and/or bwrite() delayed us. */ if ((nmp->nm_flag & NFSMNT_NQNFS) && - NQNFS_CKINVALID(vp, np, NQL_WRITE)) { + NQNFS_CKINVALID(vp, np, ND_WRITE)) { do { - error = nqnfs_getlease(vp, NQL_WRITE, cred, p); + error = nqnfs_getlease(vp, ND_WRITE, cred, p); } while (error == NQNFS_EXPIRED); if (error) { brelse(bp); @@ -540,7 +571,6 @@ again: bp->b_dirtyoff = on; bp->b_dirtyend = on + n; } -#ifndef notdef if (bp->b_validend == 0 || bp->b_validend < bp->b_dirtyoff || bp->b_validoff > bp->b_dirtyend) { bp->b_validoff = bp->b_dirtyoff; @@ -549,13 +579,6 @@ again: bp->b_validoff = min(bp->b_validoff, bp->b_dirtyoff); bp->b_validend = max(bp->b_validend, bp->b_dirtyend); } -#else - bp->b_validoff = bp->b_dirtyoff; - bp->b_validend = bp->b_dirtyend; -#endif - if (ioflag & IO_APPEND) - bp->b_flags |= B_APPENDWRITE; - /* * If the lease is non-cachable or IO_SYNC do bwrite(). */ @@ -564,10 +587,16 @@ again: error = VOP_BWRITE(bp); if (error) return (error); + if (np->n_flag & NQNFSNONCACHE) { + error = nfs_vinvalbuf(vp, V_SAVE, cred, p, 1); + if (error) + return (error); + } } else if ((n + on) == biosize && (nmp->nm_flag & NFSMNT_NQNFS) == 0) { bp->b_proc = (struct proc *)0; - bawrite(bp); + bp->b_flags |= B_ASYNC; + (void)nfs_writebp(bp, 0); } else bdwrite(bp); } while (uio->uio_resid > 0 && n > 0); @@ -590,6 +619,7 @@ nfs_getcacheblk(vp, bn, size, p) { register struct buf *bp; struct nfsmount *nmp = VFSTONFS(vp->v_mount); + int biosize = nmp->nm_rsize; if (nmp->nm_flag & NFSMNT_INT) { bp = getblk(vp, bn, size, PCATCH, 0); @@ -602,7 +632,7 @@ nfs_getcacheblk(vp, bn, size, p) bp = getblk(vp, bn, size, 0, 0); if( vp->v_type == VREG) - bp->b_blkno = (bn * NFS_MAXDGRAMDATA) / DEV_BSIZE; + bp->b_blkno = (bn * biosize) / DEV_BSIZE; return (bp); } @@ -689,6 +719,7 @@ nfs_asyncio(bp, cred) bp->b_rcred = cred; } } else { + bp->b_flags |= B_WRITEINPROG; if (bp->b_wcred == NOCRED && cred != NOCRED) { crhold(cred); bp->b_wcred = cred; @@ -700,7 +731,25 @@ nfs_asyncio(bp, cred) wakeup((caddr_t)&nfs_iodwant[i]); return (0); } - return (EIO); + + /* + * If it is a read or a write already marked B_WRITEINPROG or B_NOCACHE + * return EIO so the process will call nfs_doio() and do it + * synchronously. + */ + if (bp->b_flags & (B_READ | B_WRITEINPROG | B_NOCACHE)) + return (EIO); + + /* + * Just turn the async write into a delayed write, instead of + * doing in synchronously. Hopefully, at least one of the nfsiods + * is currently doing a write for this file and will pick up the + * delayed writes before going back to sleep. + */ + bp->b_flags |= B_DELWRI; + reassignbuf(bp, bp->b_vp); + biodone(bp); + return (0); } /* @@ -717,9 +766,10 @@ nfs_doio(bp, cr, p) register struct vnode *vp; struct nfsnode *np; struct nfsmount *nmp; - int error = 0, diff, len; + int error = 0, diff, len, iomode, must_commit = 0; struct uio uio; struct iovec io; + nfsquad_t tquad; vp = bp->b_vp; np = VTONFS(vp); @@ -740,15 +790,18 @@ nfs_doio(bp, cr, p) io.iov_len = uiop->uio_resid = bp->b_bcount; /* mapping was done by vmapbuf() */ io.iov_base = bp->b_data; - uiop->uio_offset = bp->b_blkno * DEV_BSIZE; + uiop->uio_offset = ((off_t)bp->b_blkno) * DEV_BSIZE; if (bp->b_flags & B_READ) { uiop->uio_rw = UIO_READ; nfsstats.read_physios++; error = nfs_readrpc(vp, uiop, cr); } else { + int com; + + iomode = NFSV3WRITE_DATASYNC; uiop->uio_rw = UIO_WRITE; nfsstats.write_physios++; - error = nfs_writerpc(vp, uiop, cr,0); + error = nfs_writerpc(vp, uiop, cr, &iomode, &com); } if (error) { bp->b_flags |= B_ERROR; @@ -760,7 +813,7 @@ nfs_doio(bp, cr, p) uiop->uio_rw = UIO_READ; switch (vp->v_type) { case VREG: - uiop->uio_offset = bp->b_blkno * DEV_BSIZE; + uiop->uio_offset = ((off_t)bp->b_blkno) * DEV_BSIZE; nfsstats.read_bios++; error = nfs_readrpc(vp, uiop, cr); if (!error) { @@ -773,7 +826,7 @@ nfs_doio(bp, cr, p) * Just zero fill the rest of the valid area. */ diff = bp->b_bcount - uiop->uio_resid; - len = np->n_size - (bp->b_blkno * DEV_BSIZE + len = np->n_size - (((u_quad_t)bp->b_blkno) * DEV_BSIZE + diff); if (len > 0) { len = min(len, uiop->uio_resid); @@ -786,31 +839,34 @@ nfs_doio(bp, cr, p) } if (p && (vp->v_flag & VTEXT) && (((nmp->nm_flag & NFSMNT_NQNFS) && - NQNFS_CKINVALID(vp, np, NQL_READ) && + NQNFS_CKINVALID(vp, np, ND_READ) && np->n_lrev != np->n_brev) || (!(nmp->nm_flag & NFSMNT_NQNFS) && np->n_mtime != np->n_vattr.va_mtime.ts_sec))) { uprintf("Process killed due to text file modification\n"); psignal(p, SIGKILL); +#ifdef __NetBSD__ + p->p_holdcnt++; +#else p->p_flag |= P_NOSWAP; +#endif } break; case VLNK: - uiop->uio_offset = 0; + uiop->uio_offset = (off_t)0; nfsstats.readlink_bios++; error = nfs_readlinkrpc(vp, uiop, cr); break; case VDIR: - uiop->uio_offset = bp->b_lblkno; nfsstats.readdir_bios++; - if (VFSTONFS(vp->v_mount)->nm_flag & NFSMNT_NQNFS) - error = nfs_readdirlookrpc(vp, uiop, cr); - else - error = nfs_readdirrpc(vp, uiop, cr); - /* - * Save offset cookie in b_blkno. - */ - bp->b_blkno = uiop->uio_offset; + uiop->uio_offset = ((u_quad_t)bp->b_lblkno) * NFS_DIRBLKSIZ; + if (nmp->nm_flag & NFSMNT_RDIRPLUS) { + error = nfs_readdirplusrpc(vp, uiop, cr); + if (error == NFSERR_NOTSUPP) + nmp->nm_flag &= ~NFSMNT_RDIRPLUS; + } + if ((nmp->nm_flag & NFSMNT_RDIRPLUS) == 0) + error = nfs_readdirrpc(vp, uiop, cr); break; default: printf("nfs_doio: type %x unexpected\n",vp->v_type); @@ -821,32 +877,44 @@ nfs_doio(bp, cr, p) bp->b_error = error; } } else { - if (((bp->b_blkno * DEV_BSIZE) + bp->b_dirtyend) > np->n_size) bp->b_dirtyend = np->n_size - (bp->b_blkno * DEV_BSIZE); if (bp->b_dirtyend > bp->b_dirtyoff) { io.iov_len = uiop->uio_resid = bp->b_dirtyend - - bp->b_dirtyoff; - uiop->uio_offset = (bp->b_blkno * DEV_BSIZE) - + bp->b_dirtyoff; + - bp->b_dirtyoff; + uiop->uio_offset = ((off_t)bp->b_blkno) * DEV_BSIZE + + bp->b_dirtyoff; io.iov_base = (char *)bp->b_data + bp->b_dirtyoff; uiop->uio_rw = UIO_WRITE; nfsstats.write_bios++; - if (bp->b_flags & B_APPENDWRITE) - error = nfs_writerpc(vp, uiop, cr, IO_APPEND); + if ((bp->b_flags & (B_ASYNC | B_NEEDCOMMIT | B_NOCACHE)) == B_ASYNC) + iomode = NFSV3WRITE_UNSTABLE; + else + iomode = NFSV3WRITE_FILESYNC; + bp->b_flags |= B_WRITEINPROG; + error = nfs_writerpc(vp, uiop, cr, &iomode, &must_commit); + if (!error && iomode == NFSV3WRITE_UNSTABLE) + bp->b_flags |= B_NEEDCOMMIT; else - error = nfs_writerpc(vp, uiop, cr, 0); - bp->b_flags &= ~(B_WRITEINPROG | B_APPENDWRITE); + bp->b_flags &= ~B_NEEDCOMMIT; + bp->b_flags &= ~B_WRITEINPROG; - /* - * For an interrupted write, the buffer is still valid and the - * write hasn't been pushed to the server yet, so we can't set - * B_ERROR and report the interruption by setting B_EINTR. For - * the B_ASYNC case, B_EINTR is not relevant, so the rpc attempt - * is essentially a noop. - */ - if (error == EINTR) { + /* + * For an interrupted write, the buffer is still valid + * and the write hasn't been pushed to the server yet, + * so we can't set B_ERROR and report the interruption + * by setting B_EINTR. For the B_ASYNC case, B_EINTR + * is not relevant, so the rpc attempt is essentially + * a noop. For the case of a V3 write rpc not being + * committed to stable storage, the block is still + * dirty and requires either a commit rpc or another + * write rpc with iomode == NFSV3WRITE_FILESYNC before + * the block is reused. This is indicated by setting + * the B_DELWRI and B_NEEDCOMMIT flags. + */ + if (error == EINTR + || (!error && (bp->b_flags & B_NEEDCOMMIT))) { bp->b_flags &= ~(B_INVAL|B_NOCACHE); bp->b_flags |= B_DELWRI; @@ -874,6 +942,8 @@ nfs_doio(bp, cr, p) } } bp->b_resid = uiop->uio_resid; + if (must_commit) + nfs_clearcommit(vp->v_mount); biodone(bp); return (error); } diff --git a/sys/nfsclient/nfs_nfsiod.c b/sys/nfsclient/nfs_nfsiod.c index f7e159f..35a50f4 100644 --- a/sys/nfsclient/nfs_nfsiod.c +++ b/sys/nfsclient/nfs_nfsiod.c @@ -34,7 +34,7 @@ * SUCH DAMAGE. * * @(#)nfs_syscalls.c 8.3 (Berkeley) 1/4/94 - * $Id: nfs_syscalls.c,v 1.5 1994/10/17 17:47:38 phk Exp $ + * $Id: nfs_syscalls.c,v 1.6 1995/05/30 08:12:45 rgrimes Exp $ */ #include <sys/param.h> @@ -61,9 +61,11 @@ #ifdef ISO #include <netiso/iso.h> #endif +#include <nfs/xdr_subs.h> #include <nfs/rpcv2.h> -#include <nfs/nfsv2.h> +#include <nfs/nfsproto.h> #include <nfs/nfs.h> +#include <nfs/nfsm_subs.h> #include <nfs/nfsrvcache.h> #include <nfs/nfsmount.h> #include <nfs/nfsnode.h> @@ -73,13 +75,14 @@ void nfsrv_zapsock __P((struct nfssvc_sock *)); /* Global defs. */ -extern u_long nfs_prog, nfs_vers; -extern int (*nfsrv_procs[NFS_NPROCS])(); +extern int (*nfsrv3_procs[NFS_NPROCS])(); extern struct proc *nfs_iodwant[NFS_MAXASYNCDAEMON]; extern int nfs_numasync; extern time_t nqnfsstarttime; extern int nqsrv_writeslack; extern int nfsrtton; +extern struct nfsstats nfsstats; +extern int nfsrvw_procrastinate; struct nfssvc_sock *nfs_udpsock, *nfs_cltpsock; int nuidhash_max = NFS_MAXUIDHASH; static int nfs_numnfsd = 0; @@ -225,19 +228,25 @@ nfssvc(p, uap, retval) if (error) return (error); if ((uap->flag & NFSSVC_AUTHIN) && ((nfsd = nsd->nsd_nfsd)) && - (nfsd->nd_slp->ns_flag & SLP_VALID)) { - slp = nfsd->nd_slp; + (nfsd->nfsd_slp->ns_flag & SLP_VALID)) { + slp = nfsd->nfsd_slp; /* * First check to see if another nfsd has already * added this credential. */ - for (nuidp = NUIDHASH(slp, nsd->nsd_uid)->lh_first; + for (nuidp = NUIDHASH(slp,nsd->nsd_cr.cr_uid)->lh_first; nuidp != 0; nuidp = nuidp->nu_hash.le_next) { - if (nuidp->nu_uid == nsd->nsd_uid) + if (nuidp->nu_cr.cr_uid == nsd->nsd_cr.cr_uid && + (!nfsd->nfsd_nd->nd_nam2 || + netaddr_match(NU_NETFAM(nuidp), + &nuidp->nu_haddr, nfsd->nfsd_nd->nd_nam2))) break; } - if (!nuidp) { + if (nuidp) { + nfsrv_setcred(&nuidp->nu_cr,&nfsd->nfsd_nd->nd_cr); + nfsd->nfsd_nd->nd_flag |= ND_KERBFULL; + } else { /* * Nope, so we will. */ @@ -257,22 +266,53 @@ nfssvc(p, uap, retval) LIST_REMOVE(nuidp, nu_hash); TAILQ_REMOVE(&slp->ns_uidlruhead, nuidp, nu_lru); + if (nuidp->nu_flag & NU_NAM) + m_freem(nuidp->nu_nam); } + nuidp->nu_flag = 0; nuidp->nu_cr = nsd->nsd_cr; if (nuidp->nu_cr.cr_ngroups > NGROUPS) - nuidp->nu_cr.cr_ngroups = NGROUPS; + nuidp->nu_cr.cr_ngroups = NGROUPS; nuidp->nu_cr.cr_ref = 1; - nuidp->nu_uid = nsd->nsd_uid; + nuidp->nu_timestamp = nsd->nsd_timestamp; + nuidp->nu_expire = time.tv_sec + nsd->nsd_ttl; + /* + * and save the session key in nu_key. + */ + bcopy(nsd->nsd_key, nuidp->nu_key, + sizeof (nsd->nsd_key)); + if (nfsd->nfsd_nd->nd_nam2) { + struct sockaddr_in *saddr; + + saddr = mtod(nfsd->nfsd_nd->nd_nam2, + struct sockaddr_in *); + switch (saddr->sin_family) { + case AF_INET: + nuidp->nu_flag |= NU_INETADDR; + nuidp->nu_inetaddr = + saddr->sin_addr.s_addr; + break; + case AF_ISO: + default: + nuidp->nu_flag |= NU_NAM; + nuidp->nu_nam = m_copym( + nfsd->nfsd_nd->nd_nam2, 0, + M_COPYALL, M_WAIT); + break; + }; + } TAILQ_INSERT_TAIL(&slp->ns_uidlruhead, nuidp, nu_lru); LIST_INSERT_HEAD(NUIDHASH(slp, nsd->nsd_uid), nuidp, nu_hash); - + nfsrv_setcred(&nuidp->nu_cr, + &nfsd->nfsd_nd->nd_cr); + nfsd->nfsd_nd->nd_flag |= ND_KERBFULL; } } } if ((uap->flag & NFSSVC_AUTHINFAIL) && (nfsd = nsd->nsd_nfsd)) - nfsd->nd_flag |= NFSD_AUTHFAIL; + nfsd->nfsd_flag |= NFSD_AUTHFAIL; error = nfssvc_nfsd(nsd, uap->argp, p); } if (error == EINTR || error == ERESTART) @@ -353,8 +393,6 @@ nfssvc_addsock(fp, mynam) slp = (struct nfssvc_sock *) malloc(sizeof (struct nfssvc_sock), M_NFSSVC, M_WAITOK); bzero((caddr_t)slp, sizeof (struct nfssvc_sock)); - slp->ns_uidhashtbl = - hashinit(NUIDHASHSIZ, M_NFSSVC, &slp->ns_uidhash); TAILQ_INIT(&slp->ns_uidlruhead); TAILQ_INSERT_TAIL(&nfssvc_sockhead, slp, ns_chain); } @@ -381,44 +419,47 @@ nfssvc_nfsd(nsd, argp, p) caddr_t argp; struct proc *p; { - register struct mbuf *m, *nam2; + register struct mbuf *m; register int siz; register struct nfssvc_sock *slp; register struct socket *so; register int *solockp; - struct nfsd *nd = nsd->nsd_nfsd; - struct mbuf *mreq, *nam; - struct timeval starttime; + struct nfsd *nfsd = nsd->nsd_nfsd; + struct nfsrv_descript *nd = NULL; + struct mbuf *mreq; struct nfsuid *uidp; - int error = 0, cacherep, s; - int sotype; + int error = 0, cacherep, s, sotype, writes_todo; + u_quad_t cur_usec; +#ifndef nolint + cacherep = RC_DOIT; + writes_todo = 0; +#endif s = splnet(); - if (nd == (struct nfsd *)0) { - nsd->nsd_nfsd = nd = (struct nfsd *) + if (nfsd == (struct nfsd *)0) { + nsd->nsd_nfsd = nfsd = (struct nfsd *) malloc(sizeof (struct nfsd), M_NFSD, M_WAITOK); - bzero((caddr_t)nd, sizeof (struct nfsd)); - nd->nd_procp = p; - nd->nd_cr.cr_ref = 1; - TAILQ_INSERT_TAIL(&nfsd_head, nd, nd_chain); - nd->nd_nqlflag = NQL_NOVAL; + bzero((caddr_t)nfsd, sizeof (struct nfsd)); + nfsd->nfsd_procp = p; + TAILQ_INSERT_TAIL(&nfsd_head, nfsd, nfsd_chain); nfs_numnfsd++; } /* * Loop getting rpc requests until SIGKILL. */ for (;;) { - if ((nd->nd_flag & NFSD_REQINPROG) == 0) { - while (nd->nd_slp == (struct nfssvc_sock *)0 && + if ((nfsd->nfsd_flag & NFSD_REQINPROG) == 0) { + while (nfsd->nfsd_slp == (struct nfssvc_sock *)0 && (nfsd_head_flag & NFSD_CHECKSLP) == 0) { - nd->nd_flag |= NFSD_WAITING; + nfsd->nfsd_flag |= NFSD_WAITING; nfsd_waiting++; - error = tsleep((caddr_t)nd, PSOCK | PCATCH, "nfsd", 0); + error = tsleep((caddr_t)nfsd, PSOCK | PCATCH, + "nfsd", 0); nfsd_waiting--; if (error) goto done; } - if (nd->nd_slp == (struct nfssvc_sock *)0 && + if (nfsd->nfsd_slp == (struct nfssvc_sock *)0 && (nfsd_head_flag & NFSD_CHECKSLP) != 0) { for (slp = nfssvc_sockhead.tqh_first; slp != 0; slp = slp->ns_chain.tqe_next) { @@ -426,14 +467,14 @@ nfssvc_nfsd(nsd, argp, p) == (SLP_VALID | SLP_DOREC)) { slp->ns_flag &= ~SLP_DOREC; slp->ns_sref++; - nd->nd_slp = slp; + nfsd->nfsd_slp = slp; break; } } if (slp == 0) nfsd_head_flag &= ~NFSD_CHECKSLP; } - if ((slp = nd->nd_slp) == (struct nfssvc_sock *)0) + if ((slp = nfsd->nfsd_slp) == (struct nfssvc_sock *)0) continue; if (slp->ns_flag & SLP_VALID) { if (slp->ns_flag & SLP_DISCONN) @@ -446,86 +487,78 @@ nfssvc_nfsd(nsd, argp, p) M_WAIT); nfs_sndunlock(&slp->ns_solock); } - error = nfsrv_dorec(slp, nd); - nd->nd_flag |= NFSD_REQINPROG; + error = nfsrv_dorec(slp, nfsd, &nd); + cur_usec = (u_quad_t)time.tv_sec * 1000000 + + (u_quad_t)time.tv_usec; + if (error && slp->ns_tq.lh_first && + slp->ns_tq.lh_first->nd_time <= cur_usec) { + error = 0; + cacherep = RC_DOIT; + writes_todo = 1; + } else + writes_todo = 0; + nfsd->nfsd_flag |= NFSD_REQINPROG; } } else { error = 0; - slp = nd->nd_slp; + slp = nfsd->nfsd_slp; } if (error || (slp->ns_flag & SLP_VALID) == 0) { - nd->nd_slp = (struct nfssvc_sock *)0; - nd->nd_flag &= ~NFSD_REQINPROG; + if (nd) { + free((caddr_t)nd, M_NFSRVDESC); + nd = NULL; + } + nfsd->nfsd_slp = (struct nfssvc_sock *)0; + nfsd->nfsd_flag &= ~NFSD_REQINPROG; nfsrv_slpderef(slp); continue; } splx(s); so = slp->ns_so; sotype = so->so_type; - starttime = time; if (so->so_proto->pr_flags & PR_CONNREQUIRED) solockp = &slp->ns_solock; else solockp = (int *)0; - /* - * nam == nam2 for connectionless protocols such as UDP - * nam2 == NULL for connection based protocols to disable - * recent request caching. - */ - nam2 = nd->nd_nam; - if (nam2) { - nam = nam2; - cacherep = RC_CHECKIT; - } else { - nam = slp->ns_nam; - cacherep = RC_DOIT; - } - - /* - * Check to see if authorization is needed. - */ - if (nd->nd_flag & NFSD_NEEDAUTH) { - static int logauth = 0; + if (nd) { + nd->nd_starttime = time; + if (nd->nd_nam2) + nd->nd_nam = nd->nd_nam2; + else + nd->nd_nam = slp->ns_nam; - nd->nd_flag &= ~NFSD_NEEDAUTH; - /* - * Check for a mapping already installed. - */ - for (uidp = NUIDHASH(slp, nd->nd_cr.cr_uid)->lh_first; - uidp != 0; uidp = uidp->nu_hash.le_next) { - if (uidp->nu_uid == nd->nd_cr.cr_uid) - break; - } - if (!uidp) { - nsd->nsd_uid = nd->nd_cr.cr_uid; - if (nam2 && logauth++ == 0) - log(LOG_WARNING, "Kerberized NFS using UDP\n"); - nsd->nsd_haddr = - mtod(nam, struct sockaddr_in *)->sin_addr.s_addr; - nsd->nsd_authlen = nd->nd_authlen; - if (copyout(nd->nd_authstr, nsd->nsd_authstr, - nd->nd_authlen) == 0 && - copyout((caddr_t)nsd, argp, sizeof (*nsd)) == 0) - return (ENEEDAUTH); - cacherep = RC_DROPIT; - } - } - if (cacherep == RC_CHECKIT) - cacherep = nfsrv_getcache(nam2, nd, &mreq); + /* + * Check to see if authorization is needed. + */ + if (nfsd->nfsd_flag & NFSD_NEEDAUTH) { + nfsd->nfsd_flag &= ~NFSD_NEEDAUTH; + nsd->nsd_haddr = mtod(nd->nd_nam, + struct sockaddr_in *)->sin_addr.s_addr; + nsd->nsd_authlen = nfsd->nfsd_authlen; + nsd->nsd_verflen = nfsd->nfsd_verflen; + if (!copyout(nfsd->nfsd_authstr,nsd->nsd_authstr, + nfsd->nfsd_authlen) && + !copyout(nfsd->nfsd_verfstr, nsd->nsd_verfstr, + nfsd->nfsd_verflen) && + !copyout((caddr_t)nsd, argp, sizeof (*nsd))) + return (ENEEDAUTH); + cacherep = RC_DROPIT; + } else + cacherep = nfsrv_getcache(nd, slp, &mreq); - /* - * Check for just starting up for NQNFS and send - * fake "try again later" replies to the NQNFS clients. - */ - if (notstarted && nqnfsstarttime <= time.tv_sec) { + /* + * Check for just starting up for NQNFS and send + * fake "try again later" replies to the NQNFS clients. + */ + if (notstarted && nqnfsstarttime <= time.tv_sec) { if (modify_flag) { nqnfsstarttime = time.tv_sec + nqsrv_writeslack; modify_flag = 0; } else notstarted = 0; - } - if (notstarted) { - if (nd->nd_nqlflag == NQL_NOVAL) + } + if (notstarted) { + if ((nd->nd_flag & ND_NQNFS) == 0) cacherep = RC_DROPIT; else if (nd->nd_procnum != NFSPROC_WRITE) { nd->nd_procnum = NFSPROC_NOOP; @@ -533,36 +566,42 @@ nfssvc_nfsd(nsd, argp, p) cacherep = RC_DOIT; } else modify_flag = 1; - } else if (nd->nd_flag & NFSD_AUTHFAIL) { - nd->nd_flag &= ~NFSD_AUTHFAIL; + } else if (nfsd->nfsd_flag & NFSD_AUTHFAIL) { + nfsd->nfsd_flag &= ~NFSD_AUTHFAIL; nd->nd_procnum = NFSPROC_NOOP; - nd->nd_repstat = NQNFS_AUTHERR; + nd->nd_repstat = (NFSERR_AUTHERR | AUTH_TOOWEAK); cacherep = RC_DOIT; + } } - switch (cacherep) { - case RC_DOIT: - error = (*(nfsrv_procs[nd->nd_procnum]))(nd, - nd->nd_mrep, nd->nd_md, nd->nd_dpos, &nd->nd_cr, - nam, &mreq); - if (nd->nd_cr.cr_ref != 1) { - printf("nfssvc cref=%d\n", nd->nd_cr.cr_ref); - panic("nfssvc cref"); - } + /* + * Loop to get all the write rpc relies that have been + * gathered together. + */ + do { + switch (cacherep) { + case RC_DOIT: + if (writes_todo || (nd->nd_procnum == NFSPROC_WRITE && + nfsrvw_procrastinate > 0 && !notstarted)) + error = nfsrv_writegather(&nd, slp, + nfsd->nfsd_procp, &mreq); + else + error = (*(nfsrv3_procs[nd->nd_procnum]))(nd, + slp, nfsd->nfsd_procp, &mreq); + if (mreq == NULL) + break; if (error) { if (nd->nd_procnum != NQNFSPROC_VACATED) nfsstats.srv_errs++; - if (nam2) { - nfsrv_updatecache(nam2, nd, FALSE, mreq); - m_freem(nam2); - } + nfsrv_updatecache(nd, FALSE, mreq); + if (nd->nd_nam2) + m_freem(nd->nd_nam2); break; } nfsstats.srvrpccnt[nd->nd_procnum]++; - if (nam2) - nfsrv_updatecache(nam2, nd, TRUE, mreq); + nfsrv_updatecache(nd, TRUE, mreq); nd->nd_mrep = (struct mbuf *)0; - case RC_REPLY: + case RC_REPLY: m = mreq; siz = 0; while (m) { @@ -587,15 +626,15 @@ nfssvc_nfsd(nsd, argp, p) if (solockp) (void) nfs_sndlock(solockp, (struct nfsreq *)0); if (slp->ns_flag & SLP_VALID) - error = nfs_send(so, nam2, m, (struct nfsreq *)0); + error = nfs_send(so, nd->nd_nam2, m, NULL); else { error = EPIPE; m_freem(m); } if (nfsrtton) - nfsd_rt(&starttime, sotype, nd, nam, cacherep); - if (nam2) - MFREE(nam2, m); + nfsd_rt(sotype, nd, cacherep); + if (nd->nd_nam2) + MFREE(nd->nd_nam2, m); if (nd->nd_mrep) m_freem(nd->nd_mrep); if (error == EPIPE) @@ -603,29 +642,50 @@ nfssvc_nfsd(nsd, argp, p) if (solockp) nfs_sndunlock(solockp); if (error == EINTR || error == ERESTART) { + free((caddr_t)nd, M_NFSRVDESC); nfsrv_slpderef(slp); s = splnet(); goto done; } break; - case RC_DROPIT: + case RC_DROPIT: if (nfsrtton) - nfsd_rt(&starttime, sotype, nd, nam, cacherep); + nfsd_rt(sotype, nd, cacherep); m_freem(nd->nd_mrep); - m_freem(nam2); + m_freem(nd->nd_nam2); break; - }; + }; + if (nd) { + FREE((caddr_t)nd, M_NFSRVDESC); + nd = NULL; + } + + /* + * Check to see if there are outstanding writes that + * need to be serviced. + */ + cur_usec = (u_quad_t)time.tv_sec * 1000000 + + (u_quad_t)time.tv_usec; + s = splsoftclock(); + if (slp->ns_tq.lh_first && + slp->ns_tq.lh_first->nd_time <= cur_usec) { + cacherep = RC_DOIT; + writes_todo = 1; + } else + writes_todo = 0; + splx(s); + } while (writes_todo); s = splnet(); - if (nfsrv_dorec(slp, nd)) { - nd->nd_flag &= ~NFSD_REQINPROG; - nd->nd_slp = (struct nfssvc_sock *)0; + if (nfsrv_dorec(slp, nfsd, &nd)) { + nfsd->nfsd_flag &= ~NFSD_REQINPROG; + nfsd->nfsd_slp = NULL; nfsrv_slpderef(slp); } } done: - TAILQ_REMOVE(&nfsd_head, nd, nd_chain); + TAILQ_REMOVE(&nfsd_head, nfsd, nfsd_chain); splx(s); - free((caddr_t)nd, M_NFSD); + free((caddr_t)nfsd, M_NFSD); nsd->nsd_nfsd = (struct nfsd *)0; if (--nfs_numnfsd == 0) nfsrv_init(TRUE); /* Reinitialize everything */ @@ -641,9 +701,10 @@ int nfssvc_iod(p) struct proc *p; { - register struct buf *bp; + register struct buf *bp, *nbp; register int i, myiod; - int error = 0; + struct vnode *vp; + int error = 0, s; /* * Assign my position or return error if too many already running @@ -662,24 +723,53 @@ nfssvc_iod(p) * Just loop around doin our stuff until SIGKILL */ for (;;) { - while (nfs_bufq.tqh_first == NULL && error == 0) { - nfs_iodwant[myiod] = p; - error = tsleep((caddr_t)&nfs_iodwant[myiod], - PWAIT | PCATCH, "nfsidl", 0); - } - while ((bp = nfs_bufq.tqh_first) != NULL) { - /* Take one off the front of the list */ - TAILQ_REMOVE(&nfs_bufq, bp, b_freelist); - if (bp->b_flags & B_READ) - (void) nfs_doio(bp, bp->b_rcred, (struct proc *)0); - else - (void) nfs_doio(bp, bp->b_wcred, (struct proc *)0); - } - if (error) { - nfs_asyncdaemon[myiod] = 0; - nfs_numasync--; - return (error); - } + while (nfs_bufq.tqh_first == NULL && error == 0) { + nfs_iodwant[myiod] = p; + error = tsleep((caddr_t)&nfs_iodwant[myiod], + PWAIT | PCATCH, "nfsidl", 0); + } + while ((bp = nfs_bufq.tqh_first) != NULL) { + /* Take one off the front of the list */ + TAILQ_REMOVE(&nfs_bufq, bp, b_freelist); + if (bp->b_flags & B_READ) + (void) nfs_doio(bp, bp->b_rcred, (struct proc *)0); + else do { + /* + * Look for a delayed write for the same vnode, so I can do + * it now. We must grab it before calling nfs_doio() to + * avoid any risk of the vnode getting vclean()'d while + * we are doing the write rpc. + */ + vp = bp->b_vp; + s = splbio(); + for (nbp = vp->v_dirtyblkhd.lh_first; nbp; + nbp = nbp->b_vnbufs.le_next) { + if ((nbp->b_flags & + (B_BUSY|B_DELWRI|B_NEEDCOMMIT|B_NOCACHE))!=B_DELWRI) + continue; + bremfree(nbp); + vfs_busy_pages(nbp, 1); + nbp->b_flags |= (B_BUSY|B_ASYNC); + break; + } + splx(s); + /* + * For the delayed write, do the first part of nfs_bwrite() + * up to, but not including nfs_strategy(). + */ + if (nbp) { + nbp->b_flags &= ~(B_READ|B_DONE|B_ERROR|B_DELWRI); + reassignbuf(nbp, nbp->b_vp); + nbp->b_vp->v_numoutput++; + } + (void) nfs_doio(bp, bp->b_wcred, (struct proc *)0); + } while (bp = nbp); + } + if (error) { + nfs_asyncdaemon[myiod] = 0; + nfs_numasync--; + return (error); + } } } @@ -695,9 +785,11 @@ nfsrv_zapsock(slp) register struct nfssvc_sock *slp; { register struct nfsuid *nuidp, *nnuidp; + register struct nfsrv_descript *nwp, *nnwp; struct socket *so; struct file *fp; struct mbuf *m; + int s; slp->ns_flag &= ~SLP_ALLFLAGS; fp = slp->ns_fp; @@ -716,8 +808,18 @@ nfsrv_zapsock(slp) nnuidp = nuidp->nu_lru.tqe_next; LIST_REMOVE(nuidp, nu_hash); TAILQ_REMOVE(&slp->ns_uidlruhead, nuidp, nu_lru); + if (nuidp->nu_flag & NU_NAM) + m_freem(nuidp->nu_nam); free((caddr_t)nuidp, M_NFSUID); } + s = splsoftclock(); + for (nwp = slp->ns_tq.lh_first; nwp; nwp = nnwp) { + nnwp = nwp->nd_tq.le_next; + LIST_REMOVE(nwp, nd_tq); + free((caddr_t)nwp, M_NFSRVDESC); + } + LIST_INIT(&slp->ns_tq); + splx(s); } } @@ -726,13 +828,15 @@ nfsrv_zapsock(slp) * on this mount point porpous out of the kernel and do it. */ int -nfs_getauth(nmp, rep, cred, auth_type, auth_str, auth_len) +nfs_getauth(nmp, rep, cred, auth_str, auth_len, verf_str, verf_len, key) register struct nfsmount *nmp; struct nfsreq *rep; struct ucred *cred; - int *auth_type; char **auth_str; int *auth_len; + char *verf_str; + int *verf_len; + NFSKERBKEY_T key; /* return session key */ { int error = 0; @@ -748,6 +852,9 @@ nfs_getauth(nmp, rep, cred, auth_type, auth_str, auth_len) } nmp->nm_flag &= ~(NFSMNT_WAITAUTH | NFSMNT_WANTAUTH); nmp->nm_authstr = *auth_str = (char *)malloc(RPCAUTH_MAXSIZ, M_TEMP, M_WAITOK); + nmp->nm_authlen = RPCAUTH_MAXSIZ; + nmp->nm_verfstr = verf_str; + nmp->nm_verflen = *verf_len; nmp->nm_authuid = cred->cr_uid; wakeup((caddr_t)&nmp->nm_authstr); @@ -766,8 +873,9 @@ nfs_getauth(nmp, rep, cred, auth_type, auth_str, auth_len) if (error) free((caddr_t)*auth_str, M_TEMP); else { - *auth_type = nmp->nm_authtype; *auth_len = nmp->nm_authlen; + *verf_len = nmp->nm_verflen; + bcopy((caddr_t)nmp->nm_key, (caddr_t)key, sizeof (key)); } nmp->nm_flag &= ~NFSMNT_HASAUTH; nmp->nm_flag |= NFSMNT_WAITAUTH; @@ -779,6 +887,149 @@ nfs_getauth(nmp, rep, cred, auth_type, auth_str, auth_len) } /* + * Get a nickname authenticator and verifier. + */ +int +nfs_getnickauth(nmp, cred, auth_str, auth_len, verf_str, verf_len) + struct nfsmount *nmp; + struct ucred *cred; + char **auth_str; + int *auth_len; + char *verf_str; + int verf_len; +{ + register struct nfsuid *nuidp; + register u_long *nickp, *verfp; + struct timeval ktvin, ktvout; + NFSKERBKEYSCHED_T keys; /* stores key schedule */ + +#ifdef DIAGNOSTIC + if (verf_len < (4 * NFSX_UNSIGNED)) + panic("nfs_getnickauth verf too small"); +#endif + for (nuidp = NMUIDHASH(nmp, cred->cr_uid)->lh_first; + nuidp != 0; nuidp = nuidp->nu_hash.le_next) { + if (nuidp->nu_cr.cr_uid == cred->cr_uid) + break; + } + if (!nuidp || nuidp->nu_expire < time.tv_sec) + return (EACCES); + + /* + * Move to the end of the lru list (end of lru == most recently used). + */ + TAILQ_REMOVE(&nmp->nm_uidlruhead, nuidp, nu_lru); + TAILQ_INSERT_TAIL(&nmp->nm_uidlruhead, nuidp, nu_lru); + + nickp = (u_long *)malloc(2 * NFSX_UNSIGNED, M_TEMP, M_WAITOK); + *nickp++ = txdr_unsigned(RPCAKN_NICKNAME); + *nickp = txdr_unsigned(nuidp->nu_nickname); + *auth_str = (char *)nickp; + *auth_len = 2 * NFSX_UNSIGNED; + + /* + * Now we must encrypt the verifier and package it up. + */ + verfp = (u_long *)verf_str; + *verfp++ = txdr_unsigned(RPCAKN_NICKNAME); + if (time.tv_sec > nuidp->nu_timestamp.tv_sec || + (time.tv_sec == nuidp->nu_timestamp.tv_sec && + time.tv_usec > nuidp->nu_timestamp.tv_usec)) + nuidp->nu_timestamp = time; + else + nuidp->nu_timestamp.tv_usec++; + ktvin.tv_sec = txdr_unsigned(nuidp->nu_timestamp.tv_sec); + ktvin.tv_usec = txdr_unsigned(nuidp->nu_timestamp.tv_usec); + + /* + * Now encrypt the timestamp verifier in ecb mode using the session + * key. + */ +#ifdef NFSKERB + XXX +#endif + + *verfp++ = ktvout.tv_sec; + *verfp++ = ktvout.tv_usec; + *verfp = 0; + return (0); +} + +/* + * Save the current nickname in a hash list entry on the mount point. + */ +int +nfs_savenickauth(nmp, cred, len, key, mdp, dposp, mrep) + register struct nfsmount *nmp; + struct ucred *cred; + int len; + NFSKERBKEY_T key; + struct mbuf **mdp; + char **dposp; + struct mbuf *mrep; +{ + register struct nfsuid *nuidp; + register u_long *tl; + register long t1; + struct mbuf *md = *mdp; + struct timeval ktvin, ktvout; + u_long nick; + NFSKERBKEYSCHED_T keys; + char *dpos = *dposp, *cp2; + int deltasec, error = 0; + + if (len == (3 * NFSX_UNSIGNED)) { + nfsm_dissect(tl, u_long *, 3 * NFSX_UNSIGNED); + ktvin.tv_sec = *tl++; + ktvin.tv_usec = *tl++; + nick = fxdr_unsigned(u_long, *tl); + + /* + * Decrypt the timestamp in ecb mode. + */ +#ifdef NFSKERB + XXX +#endif + ktvout.tv_sec = fxdr_unsigned(long, ktvout.tv_sec); + ktvout.tv_usec = fxdr_unsigned(long, ktvout.tv_usec); + deltasec = time.tv_sec - ktvout.tv_sec; + if (deltasec < 0) + deltasec = -deltasec; + /* + * If ok, add it to the hash list for the mount point. + */ + if (deltasec <= NFS_KERBCLOCKSKEW) { + if (nmp->nm_numuids < nuidhash_max) { + nmp->nm_numuids++; + nuidp = (struct nfsuid *) + malloc(sizeof (struct nfsuid), M_NFSUID, + M_WAITOK); + } else { + nuidp = nmp->nm_uidlruhead.tqh_first; + LIST_REMOVE(nuidp, nu_hash); + TAILQ_REMOVE(&nmp->nm_uidlruhead, nuidp, + nu_lru); + } + nuidp->nu_flag = 0; + nuidp->nu_cr.cr_uid = cred->cr_uid; + nuidp->nu_expire = time.tv_sec + NFS_KERBTTL; + nuidp->nu_timestamp = ktvout; + nuidp->nu_nickname = nick; + bcopy(key, nuidp->nu_key, sizeof (key)); + TAILQ_INSERT_TAIL(&nmp->nm_uidlruhead, nuidp, + nu_lru); + LIST_INSERT_HEAD(NMUIDHASH(nmp, cred->cr_uid), + nuidp, nu_hash); + } + } else + nfsm_adv(nfsm_rndup(len)); +nfsmout: + *mdp = md; + *dposp = dpos; + return (error); +} + +/* * Derefence a server socket structure. If it has no more references and * is no longer valid, you can throw it away. */ @@ -830,16 +1081,12 @@ nfsrv_init(terminating) nfs_udpsock = (struct nfssvc_sock *) malloc(sizeof (struct nfssvc_sock), M_NFSSVC, M_WAITOK); bzero((caddr_t)nfs_udpsock, sizeof (struct nfssvc_sock)); - nfs_udpsock->ns_uidhashtbl = - hashinit(NUIDHASHSIZ, M_NFSSVC, &nfs_udpsock->ns_uidhash); TAILQ_INIT(&nfs_udpsock->ns_uidlruhead); TAILQ_INSERT_HEAD(&nfssvc_sockhead, nfs_udpsock, ns_chain); nfs_cltpsock = (struct nfssvc_sock *) malloc(sizeof (struct nfssvc_sock), M_NFSSVC, M_WAITOK); bzero((caddr_t)nfs_cltpsock, sizeof (struct nfssvc_sock)); - nfs_cltpsock->ns_uidhashtbl = - hashinit(NUIDHASHSIZ, M_NFSSVC, &nfs_cltpsock->ns_uidhash); TAILQ_INIT(&nfs_cltpsock->ns_uidlruhead); TAILQ_INSERT_TAIL(&nfssvc_sockhead, nfs_cltpsock, ns_chain); } @@ -848,11 +1095,9 @@ nfsrv_init(terminating) * Add entries to the server monitor log. */ static void -nfsd_rt(startp, sotype, nd, nam, cacherep) - struct timeval *startp; +nfsd_rt(sotype, nd, cacherep) int sotype; - register struct nfsd *nd; - struct mbuf *nam; + register struct nfsrv_descript *nd; int cacherep; { register struct drt *rt; @@ -866,15 +1111,17 @@ nfsd_rt(startp, sotype, nd, nam, cacherep) rt->flag = DRT_CACHEDROP; if (sotype == SOCK_STREAM) rt->flag |= DRT_TCP; - if (nd->nd_nqlflag != NQL_NOVAL) + if (nd->nd_flag & ND_NQNFS) rt->flag |= DRT_NQNFS; + else if (nd->nd_flag & ND_NFSV3) + rt->flag |= DRT_NFSV3; rt->proc = nd->nd_procnum; - if (mtod(nam, struct sockaddr *)->sa_family == AF_INET) - rt->ipadr = mtod(nam, struct sockaddr_in *)->sin_addr.s_addr; + if (mtod(nd->nd_nam, struct sockaddr *)->sa_family == AF_INET) + rt->ipadr = mtod(nd->nd_nam, struct sockaddr_in *)->sin_addr.s_addr; else - rt->ipadr = INADDR_ANY; - rt->resptime = ((time.tv_sec - startp->tv_sec) * 1000000) + - (time.tv_usec - startp->tv_usec); + rt->ipadr = INADDR_ANY; + rt->resptime = ((time.tv_sec - nd->nd_starttime.tv_sec) * 1000000) + + (time.tv_usec - nd->nd_starttime.tv_usec); rt->tstamp = time; nfsdrt.pos = (nfsdrt.pos + 1) % NFSRTTLOGSIZ; } diff --git a/sys/nfsclient/nfs_node.c b/sys/nfsclient/nfs_node.c index e5315f4..c002cea 100644 --- a/sys/nfsclient/nfs_node.c +++ b/sys/nfsclient/nfs_node.c @@ -34,7 +34,7 @@ * SUCH DAMAGE. * * @(#)nfs_node.c 8.2 (Berkeley) 12/30/93 - * $Id: nfs_node.c,v 1.7 1994/10/17 17:47:33 phk Exp $ + * $Id: nfs_node.c,v 1.8 1995/03/16 18:15:36 bde Exp $ */ #include <sys/param.h> @@ -47,14 +47,12 @@ #include <sys/malloc.h> #include <nfs/rpcv2.h> -#include <nfs/nfsv2.h> +#include <nfs/nfsproto.h> #include <nfs/nfs.h> #include <nfs/nfsnode.h> #include <nfs/nfsmount.h> #include <nfs/nqnfs.h> -#define NFSNOHASH(fhsum) \ - (&nfsnodehashtbl[(fhsum) & nfsnodehash]) LIST_HEAD(nfsnodehashhead, nfsnode) *nfsnodehashtbl; u_long nfsnodehash; @@ -79,19 +77,20 @@ nfs_nhinit() /* * Compute an entry in the NFS hash table structure */ -struct nfsnodehashhead * -nfs_hash(fhp) - register nfsv2fh_t *fhp; +u_long +nfs_hash(fhp, fhsize) + register nfsfh_t *fhp; + int fhsize; { register u_char *fhpp; register u_long fhsum; - int i; + register int i; fhpp = &fhp->fh_bytes[0]; fhsum = 0; - for (i = 0; i < NFSX_FH; i++) + for (i = 0; i < fhsize; i++) fhsum += *fhpp++; - return (NFSNOHASH(fhsum)); + return (fhsum); } /* @@ -101,9 +100,10 @@ nfs_hash(fhp) * nfsnode structure is returned. */ int -nfs_nget(mntp, fhp, npp) +nfs_nget(mntp, fhp, fhsize, npp) struct mount *mntp; - register nfsv2fh_t *fhp; + register nfsfh_t *fhp; + int fhsize; struct nfsnode **npp; { register struct nfsnode *np; @@ -112,11 +112,11 @@ nfs_nget(mntp, fhp, npp) struct vnode *nvp; int error; - nhpp = nfs_hash(fhp); + nhpp = NFSNOHASH(nfs_hash(fhp, fhsize)); loop: for (np = nhpp->lh_first; np != 0; np = np->n_hash.le_next) { - if (mntp != NFSTOV(np)->v_mount || - bcmp((caddr_t)fhp, (caddr_t)&np->n_fh, NFSX_FH)) + if (mntp != NFSTOV(np)->v_mount || np->n_fhsize != fhsize || + bcmp((caddr_t)fhp, (caddr_t)np->n_fhp, fhsize)) continue; vp = NFSTOV(np); if (vget(vp, 1)) @@ -131,27 +131,26 @@ loop: } vp = nvp; MALLOC(np, struct nfsnode *, sizeof *np, M_NFSNODE, M_WAITOK); + bzero((caddr_t)np, sizeof *np); vp->v_data = np; np->n_vnode = vp; /* * Insert the nfsnode in the hash queue for its new file handle */ - np->n_flag = 0; LIST_INSERT_HEAD(nhpp, np, n_hash); - bcopy((caddr_t)fhp, (caddr_t)&np->n_fh, NFSX_FH); - np->n_attrstamp = 0; - np->n_direofoffset = 0; - np->n_sillyrename = (struct sillyrename *)0; - np->n_size = 0; - np->n_mtime = 0; - np->n_lockf = 0; - if (VFSTONFS(mntp)->nm_flag & NFSMNT_NQNFS) { - np->n_brev = 0; - np->n_lrev = 0; - np->n_expiry = (time_t)0; - np->n_timer.cqe_next = (struct nfsnode *)0; - } + if (fhsize > NFS_SMALLFH) { + MALLOC(np->n_fhp, nfsfh_t *, fhsize, M_NFSBIGFH, M_WAITOK); + } else + np->n_fhp = &np->n_fh; + bcopy((caddr_t)fhp, (caddr_t)np->n_fhp, fhsize); + np->n_fhsize = fhsize; *npp = np; + + /* + * Lock the new nfsnode. + */ + VOP_LOCK(vp); + return (0); } @@ -168,7 +167,10 @@ nfs_inactive(ap) np = VTONFS(ap->a_vp); if (prtactive && ap->a_vp->v_usecount != 0) vprint("nfs_inactive: pushing active", ap->a_vp); - sp = np->n_sillyrename; + if (ap->a_vp->v_type != VDIR) + sp = np->n_sillyrename; + else + sp = (struct sillyrename *)0; np->n_sillyrename = (struct sillyrename *)0; if (sp) { /* @@ -178,9 +180,7 @@ nfs_inactive(ap) nfs_removeit(sp); crfree(sp->s_cred); vrele(sp->s_dvp); -#ifdef SILLYSEPARATE - free((caddr_t)sp, M_NFSREQ); -#endif + FREE((caddr_t)sp, M_NFSREQ); } np->n_flag &= (NMODIFIED | NFLUSHINPROG | NFLUSHWANT | NQNFSEVICTED | NQNFSNONCACHE | NQNFSWRITE); @@ -199,6 +199,7 @@ nfs_reclaim(ap) register struct vnode *vp = ap->a_vp; register struct nfsnode *np = VTONFS(vp); register struct nfsmount *nmp = VFSTONFS(vp->v_mount); + register struct nfsdmap *dp, *dp2; if (prtactive && vp->v_usecount != 0) vprint("nfs_reclaim: pushing active", vp); @@ -211,6 +212,24 @@ nfs_reclaim(ap) if ((nmp->nm_flag & NFSMNT_NQNFS) && np->n_timer.cqe_next != 0) { CIRCLEQ_REMOVE(&nmp->nm_timerhead, np, n_timer); } + + /* + * Free up any directory cookie structures and + * large file handle structures that might be associated with + * this nfs node. + */ + if (vp->v_type == VDIR) { + dp = np->n_cookies.lh_first; + while (dp) { + dp2 = dp; + dp = dp->ndm_list.le_next; + FREE((caddr_t)dp2, M_NFSDIROFF); + } + } + if (np->n_fhsize > NFS_SMALLFH) { + FREE((caddr_t)np->n_fhp, M_NFSBIGFH); + } + cache_purge(vp); FREE(vp->v_data, M_NFSNODE); vp->v_data = (void *)0; @@ -227,6 +246,7 @@ nfs_lock(ap) } */ *ap; { register struct vnode *vp = ap->a_vp; + struct nfsnode *np = VTONFS(vp); /* * Ugh, another place where interruptible mounts will get hung. @@ -239,6 +259,33 @@ nfs_lock(ap) } if (vp->v_tag == VT_NON) return (ENOENT); + +#if 0 + /* + * Only lock regular files. If a server crashed while we were + * holding a directory lock, we could easily end up sleeping + * until the server rebooted while holding a lock on the root. + * Locks are only needed for protecting critical sections in + * VMIO at the moment. + * New vnodes will have type VNON but they should be locked + * since they may become VREG. This is checked in loadattrcache + * and unwanted locks are released there. + */ + if (vp->v_type == VREG || vp->v_type == VNON) { + while (np->n_flag & NLOCKED) { + np->n_flag |= NWANTED; + (void) tsleep((caddr_t) np, PINOD, "nfslck2", 0); + /* + * If the vnode has transmuted into a VDIR while we + * were asleep, then skip the lock. + */ + if (vp->v_type != VREG && vp->v_type != VNON) + return (0); + } + np->n_flag |= NLOCKED; + } +#endif + return (0); } @@ -251,6 +298,20 @@ nfs_unlock(ap) struct vnode *a_vp; } */ *ap; { +#if 0 + struct vnode* vp = ap->a_vp; + struct nfsnode* np = VTONFS(vp); + + if (vp->v_type == VREG || vp->v_type == VNON) { + if (!(np->n_flag & NLOCKED)) + panic("nfs_unlock: nfsnode not locked"); + np->n_flag &= ~NLOCKED; + if (np->n_flag & NWANTED) { + np->n_flag &= ~NWANTED; + wakeup((caddr_t) np); + } + } +#endif return (0); } @@ -264,8 +325,7 @@ nfs_islocked(ap) struct vnode *a_vp; } */ *ap; { - - return (0); + return VTONFS(ap->a_vp)->n_flag & NLOCKED ? 1 : 0; } /* diff --git a/sys/nfsclient/nfs_socket.c b/sys/nfsclient/nfs_socket.c index 20253ef..8d15b24 100644 --- a/sys/nfsclient/nfs_socket.c +++ b/sys/nfsclient/nfs_socket.c @@ -34,7 +34,7 @@ * SUCH DAMAGE. * * @(#)nfs_socket.c 8.3 (Berkeley) 1/12/94 - * $Id: nfs_socket.c,v 1.6 1995/03/16 18:15:37 bde Exp $ + * $Id: nfs_socket.c,v 1.7 1995/05/30 08:12:40 rgrimes Exp $ */ /* @@ -59,7 +59,7 @@ #include <netinet/tcp.h> #include <nfs/rpcv2.h> -#include <nfs/nfsv2.h> +#include <nfs/nfsproto.h> #include <nfs/nfs.h> #include <nfs/xdr_subs.h> #include <nfs/nfsm_subs.h> @@ -94,36 +94,13 @@ * External data, mostly RPC constants in XDR form */ extern u_long rpc_reply, rpc_msgdenied, rpc_mismatch, rpc_vers, rpc_auth_unix, - rpc_msgaccepted, rpc_call, rpc_autherr, rpc_rejectedcred, + rpc_msgaccepted, rpc_call, rpc_autherr, rpc_auth_kerb; -extern u_long nfs_prog, nfs_vers, nqnfs_prog, nqnfs_vers; +extern u_long nfs_prog, nqnfs_prog; extern time_t nqnfsstarttime; -extern int nonidempotent[NFS_NPROCS]; - -/* - * Maps errno values to nfs error numbers. - * Use NFSERR_IO as the catch all for ones not specifically defined in - * RFC 1094. - */ -static int nfsrv_errmap[ELAST] = { - NFSERR_PERM, NFSERR_NOENT, NFSERR_IO, NFSERR_IO, NFSERR_IO, - NFSERR_NXIO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, - NFSERR_IO, NFSERR_IO, NFSERR_ACCES, NFSERR_IO, NFSERR_IO, - NFSERR_IO, NFSERR_EXIST, NFSERR_IO, NFSERR_NODEV, NFSERR_NOTDIR, - NFSERR_ISDIR, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, - NFSERR_IO, NFSERR_FBIG, NFSERR_NOSPC, NFSERR_IO, NFSERR_ROFS, - NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, - NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, - NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, - NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, - NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, - NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, - NFSERR_IO, NFSERR_IO, NFSERR_NAMETOL, NFSERR_IO, NFSERR_IO, - NFSERR_NOTEMPTY, NFSERR_IO, NFSERR_IO, NFSERR_DQUOT, NFSERR_STALE, - NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, - NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, - NFSERR_IO, -}; +extern struct nfsstats nfsstats; +extern int nfsv3_procid[NFS_NPROCS]; +extern int nfs_ticks; /* * Defines which timer to use for the procnum. @@ -134,7 +111,8 @@ static int nfsrv_errmap[ELAST] = { * 4 - write */ static int proct[NFS_NPROCS] = { - 0, 1, 0, 0, 2, 3, 3, 0, 4, 0, 0, 0, 0, 0, 0, 0, 3, 0, 3, 0, 0, 0, 0, + 0, 1, 0, 2, 1, 3, 3, 4, 0, 0, 0, 0, 0, 0, 0, 0, 3, 3, 0, 0, 0, 0, 0, + 0, 0, 0, }; /* @@ -164,10 +142,12 @@ int nfsrv_null(), nfsrv_getattr(), nfsrv_setattr(), nfsrv_lookup(), + nfsrv3_access(), nfsrv_readlink(), nfsrv_read(), nfsrv_write(), nfsrv_create(), + nfsrv_mknod(), nfsrv_remove(), nfsrv_rename(), nfsrv_link(), @@ -175,37 +155,42 @@ int nfsrv_null(), nfsrv_mkdir(), nfsrv_rmdir(), nfsrv_readdir(), + nfsrv_readdirplus(), nfsrv_statfs(), + nfsrv_fsinfo(), + nfsrv_pathconf(), + nfsrv_commit(), nfsrv_noop(), - nqnfsrv_readdirlook(), nqnfsrv_getlease(), - nqnfsrv_vacated(), - nqnfsrv_access(); + nqnfsrv_vacated(); -int (*nfsrv_procs[NFS_NPROCS])() = { +int (*nfsrv3_procs[NFS_NPROCS])() = { nfsrv_null, nfsrv_getattr, nfsrv_setattr, - nfsrv_noop, nfsrv_lookup, + nfsrv3_access, nfsrv_readlink, nfsrv_read, - nfsrv_noop, nfsrv_write, nfsrv_create, + nfsrv_mkdir, + nfsrv_symlink, + nfsrv_mknod, nfsrv_remove, + nfsrv_rmdir, nfsrv_rename, nfsrv_link, - nfsrv_symlink, - nfsrv_mkdir, - nfsrv_rmdir, nfsrv_readdir, + nfsrv_readdirplus, nfsrv_statfs, - nqnfsrv_readdirlook, + nfsrv_fsinfo, + nfsrv_pathconf, + nfsrv_commit, nqnfsrv_getlease, nqnfsrv_vacated, nfsrv_noop, - nqnfsrv_access, + nfsrv_noop }; /* @@ -874,13 +859,16 @@ nfs_request(vp, mrest, procnum, procp, cred, mrp, mdp, dposp) struct nfsmount *nmp; struct mbuf *md, *mheadend; struct nfsnode *np; + char nickv[RPCX_NICKVERF]; time_t reqtime, waituntil; caddr_t dpos, cp2; int t1, nqlflag, cachable, s, error = 0, mrest_len, auth_len, auth_type; int trylater_delay = NQ_TRYLATERDEL, trylater_cnt = 0, failed_auth = 0; + int verf_len, verf_type; u_long xid; u_quad_t frev; - char *auth_str; + char *auth_str, *verf_str; + NFSKERBKEY_T key; /* save session key */ nmp = VFSTONFS(vp->v_mount); MALLOC(rep, struct nfsreq *, sizeof(struct nfsreq), M_NFSREQ, M_WAITOK); @@ -900,19 +888,21 @@ nfs_request(vp, mrest, procnum, procp, cred, mrp, mdp, dposp) * Get the RPC header with authorization. */ kerbauth: - auth_str = (char *)0; + verf_str = auth_str = (char *)0; if (nmp->nm_flag & NFSMNT_KERB) { - if (failed_auth) { - error = nfs_getauth(nmp, rep, cred, &auth_type, - &auth_str, &auth_len); + verf_str = nickv; + verf_len = sizeof (nickv); + auth_type = RPCAUTH_KERB4; + bzero((caddr_t)key, sizeof (key)); + if (failed_auth || nfs_getnickauth(nmp, cred, &auth_str, + &auth_len, verf_str, verf_len)) { + error = nfs_getauth(nmp, rep, cred, &auth_str, + &auth_len, verf_str, &verf_len, key); if (error) { free((caddr_t)rep, M_NFSREQ); m_freem(mrest); return (error); } - } else { - auth_type = RPCAUTH_UNIX; - auth_len = 5 * NFSX_UNSIGNED; } } else { auth_type = RPCAUTH_UNIX; @@ -922,8 +912,8 @@ kerbauth: nmp->nm_numgrps : (cred->cr_ngroups - 1)) << 2) + 5 * NFSX_UNSIGNED; } - m = nfsm_rpchead(cred, (nmp->nm_flag & NFSMNT_NQNFS), procnum, - auth_type, auth_len, auth_str, mrest, mrest_len, &mheadend, &xid); + m = nfsm_rpchead(cred, nmp->nm_flag, procnum, auth_type, auth_len, + auth_str, verf_len, verf_str, mrest, mrest_len, &mheadend, &xid); if (auth_str) free(auth_str, M_TEMP); @@ -1029,12 +1019,12 @@ tryagain: /* * break down the rpc header and check if ok */ - nfsm_dissect(tl, u_long *, 3*NFSX_UNSIGNED); + nfsm_dissect(tl, u_long *, 3 * NFSX_UNSIGNED); if (*tl++ == rpc_msgdenied) { if (*tl == rpc_mismatch) error = EOPNOTSUPP; else if ((nmp->nm_flag & NFSMNT_KERB) && *tl++ == rpc_autherr) { - if (*tl == rpc_rejectedcred && failed_auth == 0) { + if (!failed_auth) { failed_auth++; mheadend->m_next = (struct mbuf *)0; m_freem(mrep); @@ -1051,22 +1041,25 @@ tryagain: } /* - * skip over the auth_verf, someday we may want to cache auth_short's - * for nfs_reqhead(), but for now just dump it + * Grab any Kerberos verifier, otherwise just throw it away. */ - if (*++tl != 0) { - i = nfsm_rndup(fxdr_unsigned(long, *tl)); - nfsm_adv(i); - } + verf_type = fxdr_unsigned(int, *tl++); + i = fxdr_unsigned(int, *tl); + if ((nmp->nm_flag & NFSMNT_KERB) && verf_type == RPCAUTH_KERB4) { + error = nfs_savenickauth(nmp, cred, i, key, &md, &dpos, mrep); + if (error) + goto nfsmout; + } else if (i > 0) + nfsm_adv(nfsm_rndup(i)); nfsm_dissect(tl, u_long *, NFSX_UNSIGNED); /* 0 == ok */ if (*tl == 0) { nfsm_dissect(tl, u_long *, NFSX_UNSIGNED); if (*tl != 0) { error = fxdr_unsigned(int, *tl); - m_freem(mrep); - if ((nmp->nm_flag & NFSMNT_NQNFS) && - error == NQNFS_TRYLATER) { + if ((nmp->nm_flag & NFSMNT_NFSV3) && + error == NFSERR_TRYLATER) { + m_freem(mrep); error = 0; waituntil = time.tv_sec + trylater_delay; while (time.tv_sec < waituntil) @@ -1084,6 +1077,13 @@ tryagain: */ if (error == ESTALE) cache_purge(vp); + if (nmp->nm_flag & NFSMNT_NFSV3) { + *mrp = mrep; + *mdp = md; + *dposp = dpos; + error |= NFSERR_RETERR; + } else + m_freem(mrep); m_freem(rep->r_mreq); free((caddr_t)rep, M_NFSREQ); return (error); @@ -1115,10 +1115,10 @@ tryagain: return (0); } m_freem(mrep); - m_freem(rep->r_mreq); - free((caddr_t)rep, M_NFSREQ); error = EPROTONOSUPPORT; nfsmout: + m_freem(rep->r_mreq); + free((caddr_t)rep, M_NFSREQ); return (error); } @@ -1127,9 +1127,10 @@ nfsmout: * siz arg. is used to decide if adding a cluster is worthwhile */ int -nfs_rephead(siz, nd, err, cache, frev, mrq, mbp, bposp) +nfs_rephead(siz, nd, slp, err, cache, frev, mrq, mbp, bposp) int siz; - struct nfsd *nd; + struct nfsrv_descript *nd; + struct nfssvc_sock *slp; int err; int cache; u_quad_t *frev; @@ -1154,47 +1155,98 @@ nfs_rephead(siz, nd, err, cache, frev, mrq, mbp, bposp) } else mreq->m_data += max_hdr; tl = mtod(mreq, u_long *); - mreq->m_len = 6*NFSX_UNSIGNED; - bpos = ((caddr_t)tl)+mreq->m_len; + mreq->m_len = 6 * NFSX_UNSIGNED; + bpos = ((caddr_t)tl) + mreq->m_len; *tl++ = txdr_unsigned(nd->nd_retxid); *tl++ = rpc_reply; - if (err == ERPCMISMATCH || err == NQNFS_AUTHERR) { + if (err == ERPCMISMATCH || (err & NFSERR_AUTHERR)) { *tl++ = rpc_msgdenied; - if (err == NQNFS_AUTHERR) { + if (err & NFSERR_AUTHERR) { *tl++ = rpc_autherr; - *tl = rpc_rejectedcred; + *tl = txdr_unsigned(err & ~NFSERR_AUTHERR); mreq->m_len -= NFSX_UNSIGNED; bpos -= NFSX_UNSIGNED; } else { *tl++ = rpc_mismatch; - *tl++ = txdr_unsigned(2); - *tl = txdr_unsigned(2); + *tl++ = txdr_unsigned(RPC_VER2); + *tl = txdr_unsigned(RPC_VER2); } } else { *tl++ = rpc_msgaccepted; - *tl++ = 0; - *tl++ = 0; + + /* + * For Kerberos authentication, we must send the nickname + * verifier back, otherwise just RPCAUTH_NULL. + */ + if (nd->nd_flag & ND_KERBFULL) { + register struct nfsuid *nuidp; + struct timeval ktvin, ktvout; + NFSKERBKEYSCHED_T keys; /* stores key schedule */ + + for (nuidp = NUIDHASH(slp, nd->nd_cr.cr_uid)->lh_first; + nuidp != 0; nuidp = nuidp->nu_hash.le_next) { + if (nuidp->nu_cr.cr_uid == nd->nd_cr.cr_uid && + (!nd->nd_nam2 || netaddr_match(NU_NETFAM(nuidp), + &nuidp->nu_haddr, nd->nd_nam2))) + break; + } + if (nuidp) { + ktvin.tv_sec = + txdr_unsigned(nuidp->nu_timestamp.tv_sec - 1); + ktvin.tv_usec = + txdr_unsigned(nuidp->nu_timestamp.tv_usec); + + /* + * Encrypt the timestamp in ecb mode using the + * session key. + */ +#ifdef NFSKERB + XXX +#endif + + *tl++ = rpc_auth_kerb; + *tl++ = txdr_unsigned(3 * NFSX_UNSIGNED); + *tl = ktvout.tv_sec; + nfsm_build(tl, u_long *, 3 * NFSX_UNSIGNED); + *tl++ = ktvout.tv_usec; + *tl++ = txdr_unsigned(nuidp->nu_cr.cr_uid); + } else { + *tl++ = 0; + *tl++ = 0; + } + } else { + *tl++ = 0; + *tl++ = 0; + } switch (err) { case EPROGUNAVAIL: *tl = txdr_unsigned(RPC_PROGUNAVAIL); break; case EPROGMISMATCH: *tl = txdr_unsigned(RPC_PROGMISMATCH); - nfsm_build(tl, u_long *, 2*NFSX_UNSIGNED); - *tl++ = txdr_unsigned(2); - *tl = txdr_unsigned(2); /* someday 3 */ + nfsm_build(tl, u_long *, 2 * NFSX_UNSIGNED); + if (nd->nd_flag & ND_NQNFS) { + *tl++ = txdr_unsigned(3); + *tl = txdr_unsigned(3); + } else { + *tl++ = txdr_unsigned(2); + *tl = txdr_unsigned(3); + } break; case EPROCUNAVAIL: *tl = txdr_unsigned(RPC_PROCUNAVAIL); break; + case EBADRPC: + *tl = txdr_unsigned(RPC_GARBAGE); + break; default: *tl = 0; - if (err != VNOVAL) { + if (err != NFSERR_RETVOID) { nfsm_build(tl, u_long *, NFSX_UNSIGNED); if (err) - *tl = txdr_unsigned(nfsrv_errmap[err - 1]); + *tl = txdr_unsigned(nfsrv_errmap(nd, err)); else - *tl = 0; + *tl = 0; } break; }; @@ -1203,16 +1255,14 @@ nfs_rephead(siz, nd, err, cache, frev, mrq, mbp, bposp) /* * For nqnfs, piggyback lease as requested. */ - if (nd->nd_nqlflag != NQL_NOVAL && err == 0) { - if (nd->nd_nqlflag) { - nfsm_build(tl, u_long *, 5*NFSX_UNSIGNED); - *tl++ = txdr_unsigned(nd->nd_nqlflag); + if ((nd->nd_flag & ND_NQNFS) && err == 0) { + if (nd->nd_flag & ND_LEASE) { + nfsm_build(tl, u_long *, 5 * NFSX_UNSIGNED); + *tl++ = txdr_unsigned(nd->nd_flag & ND_LEASE); *tl++ = txdr_unsigned(cache); *tl++ = txdr_unsigned(nd->nd_duration); txdr_hyper(frev, tl); } else { - if (nd->nd_nqlflag != 0) - panic("nqreph"); nfsm_build(tl, u_long *, NFSX_UNSIGNED); *tl = 0; } @@ -1220,7 +1270,7 @@ nfs_rephead(siz, nd, err, cache, frev, mrq, mbp, bposp) *mrq = mreq; *mbp = mb; *bposp = bpos; - if (err != 0 && err != VNOVAL) + if (err != 0 && err != NFSERR_RETVOID) nfsstats.srvrpc_errs++; return (0); } @@ -1240,8 +1290,10 @@ nfs_timer(arg) register struct socket *so; register struct nfsmount *nmp; register int timeo; + register struct nfssvc_sock *slp; static long lasttime = 0; int s, error; + u_quad_t cur_usec; s = splnet(); for (rep = nfs_reqq.tqh_first; rep != 0; rep = rep->r_chain.tqe_next) { @@ -1338,8 +1390,19 @@ nfs_timer(arg) lasttime = time.tv_sec; nqnfs_serverd(); } + + /* + * Scan the write gathering queues for writes that need to be + * completed now. + */ + cur_usec = (u_quad_t)time.tv_sec * 1000000 + (u_quad_t)time.tv_usec; + for (slp = nfssvc_sockhead.tqh_first; slp != 0; + slp = slp->ns_chain.tqe_next) { + if (slp->ns_tq.lh_first && slp->ns_tq.lh_first->nd_time<=cur_usec) + nfsrv_wakenfsd(slp); + } splx(s); - timeout(nfs_timer, (void *)0, hz / NFS_HZ); + timeout(nfs_timer, (void *)0, nfs_ticks); } /* @@ -1673,7 +1736,7 @@ nfsrv_getstream(slp, waitflag) register struct nfssvc_sock *slp; int waitflag; { - register struct mbuf *m; + register struct mbuf *m, **mpp; register char *cp1, *cp2; register int len; struct mbuf *om, *m2, *recm = 0; @@ -1707,7 +1770,12 @@ nfsrv_getstream(slp, waitflag) } } slp->ns_cc -= NFSX_UNSIGNED; - slp->ns_reclen = ntohl(recmark) & ~0x80000000; + recmark = ntohl(recmark); + slp->ns_reclen = recmark & ~0x80000000; + if (recmark & 0x80000000) + slp->ns_flag |= SLP_LASTFRAG; + else + slp->ns_flag &= ~SLP_LASTFRAG; if (slp->ns_reclen < NFS_MINPACKET || slp->ns_reclen > NFS_MAXPACKET) { slp->ns_flag &= ~SLP_GETSTREAM; return (EPERM); @@ -1761,12 +1829,23 @@ nfsrv_getstream(slp, waitflag) slp->ns_flag &= ~SLP_GETSTREAM; return (0); } - nfs_realign(recm, 10 * NFSX_UNSIGNED); - if (slp->ns_recend) - slp->ns_recend->m_nextpkt = recm; - else - slp->ns_rec = recm; - slp->ns_recend = recm; + + /* + * Accumulate the fragments into a record. + */ + mpp = &slp->ns_frag; + while (*mpp) + mpp = &((*mpp)->m_next); + *mpp = recm; + if (slp->ns_flag & SLP_LASTFRAG) { + nfs_realign(slp->ns_frag, 10 * NFSX_UNSIGNED); + if (slp->ns_recend) + slp->ns_recend->m_nextpkt = slp->ns_frag; + else + slp->ns_rec = slp->ns_frag; + slp->ns_recend = slp->ns_frag; + slp->ns_frag = (struct mbuf *)0; + } } } @@ -1774,13 +1853,16 @@ nfsrv_getstream(slp, waitflag) * Parse an RPC header. */ int -nfsrv_dorec(slp, nd) +nfsrv_dorec(slp, nfsd, ndp) register struct nfssvc_sock *slp; - register struct nfsd *nd; + struct nfsd *nfsd; + struct nfsrv_descript **ndp; { - register struct mbuf *m; + register struct mbuf *m, *nam; + register struct nfsrv_descript *nd; int error; + *ndp = NULL; if ((slp->ns_flag & SLP_VALID) == 0 || (m = slp->ns_rec) == (struct mbuf *)0) return (ENOBUFS); @@ -1790,19 +1872,24 @@ nfsrv_dorec(slp, nd) else slp->ns_recend = (struct mbuf *)0; if (m->m_type == MT_SONAME) { - nd->nd_nam = m; - nd->nd_md = nd->nd_mrep = m->m_next; - m->m_next = (struct mbuf *)0; - } else { - nd->nd_nam = (struct mbuf *)0; - nd->nd_md = nd->nd_mrep = m; - } - nd->nd_dpos = mtod(nd->nd_md, caddr_t); - error = nfs_getreq(nd, TRUE); + nam = m; + m = m->m_next; + nam->m_next = NULL; + } else + nam = NULL; + MALLOC(nd, struct nfsrv_descript *, sizeof (struct nfsrv_descript), + M_NFSRVDESC, M_WAITOK); + nd->nd_md = nd->nd_mrep = m; + nd->nd_nam2 = nam; + nd->nd_dpos = mtod(m, caddr_t); + error = nfs_getreq(nd, nfsd, TRUE); if (error) { - m_freem(nd->nd_nam); + m_freem(nam); + free((caddr_t)nd, M_NFSRVDESC); return (error); } + *ndp = nd; + nfsd->nfsd_nd = nd; return (0); } @@ -1812,8 +1899,9 @@ nfsrv_dorec(slp, nd) * - fill in the cred struct. */ int -nfs_getreq(nd, has_header) - register struct nfsd *nd; +nfs_getreq(nd, nfsd, has_header) + register struct nfsrv_descript *nd; + struct nfsd *nfsd; int has_header; { register int len, i; @@ -1821,57 +1909,67 @@ nfs_getreq(nd, has_header) register long t1; struct uio uio; struct iovec iov; - caddr_t dpos, cp2; + caddr_t dpos, cp2, cp; u_long nfsvers, auth_type; - int error = 0, nqnfs = 0; + uid_t nickuid; + int error = 0, nqnfs = 0, ticklen; struct mbuf *mrep, *md; + register struct nfsuid *nuidp; + struct timeval tvin, tvout; + NFSKERBKEYSCHED_T keys; /* stores key schedule */ mrep = nd->nd_mrep; md = nd->nd_md; dpos = nd->nd_dpos; if (has_header) { - nfsm_dissect(tl, u_long *, 10*NFSX_UNSIGNED); + nfsm_dissect(tl, u_long *, 10 * NFSX_UNSIGNED); nd->nd_retxid = fxdr_unsigned(u_long, *tl++); if (*tl++ != rpc_call) { m_freem(mrep); return (EBADRPC); } - } else { - nfsm_dissect(tl, u_long *, 8*NFSX_UNSIGNED); - } + } else + nfsm_dissect(tl, u_long *, 8 * NFSX_UNSIGNED); nd->nd_repstat = 0; + nd->nd_flag = 0; if (*tl++ != rpc_vers) { nd->nd_repstat = ERPCMISMATCH; nd->nd_procnum = NFSPROC_NOOP; return (0); } - nfsvers = nfs_vers; if (*tl != nfs_prog) { - if (*tl == nqnfs_prog) { + if (*tl == nqnfs_prog) nqnfs++; - nfsvers = nqnfs_vers; - } else { + else { nd->nd_repstat = EPROGUNAVAIL; nd->nd_procnum = NFSPROC_NOOP; return (0); } } tl++; - if (*tl++ != nfsvers) { + nfsvers = fxdr_unsigned(u_long, *tl++); + if (((nfsvers < NFS_VER2 || nfsvers > NFS_VER3) && !nqnfs) || + (nfsvers != NQNFS_VER3 && nqnfs)) { nd->nd_repstat = EPROGMISMATCH; nd->nd_procnum = NFSPROC_NOOP; return (0); } + if (nqnfs) + nd->nd_flag = (ND_NFSV3 | ND_NQNFS); + else if (nfsvers == NFS_VER3) + nd->nd_flag = ND_NFSV3; nd->nd_procnum = fxdr_unsigned(u_long, *tl++); if (nd->nd_procnum == NFSPROC_NULL) return (0); if (nd->nd_procnum >= NFS_NPROCS || - (!nqnfs && nd->nd_procnum > NFSPROC_STATFS) || - (*tl != rpc_auth_unix && *tl != rpc_auth_kerb)) { + (!nqnfs && nd->nd_procnum >= NQNFSPROC_GETLEASE) || + (!nd->nd_flag && nd->nd_procnum > NFSV2PROC_STATFS)) { nd->nd_repstat = EPROCUNAVAIL; nd->nd_procnum = NFSPROC_NOOP; return (0); } + if ((nd->nd_flag & ND_NFSV3) == 0) + nd->nd_procnum = nfsv3_procid[nd->nd_procnum]; auth_type = *tl++; len = fxdr_unsigned(int, *tl++); if (len < 0 || len > RPCAUTH_MAXSIZ) { @@ -1879,6 +1977,7 @@ nfs_getreq(nd, has_header) return (EBADRPC); } + nd->nd_flag &= ~ND_KERBAUTH; /* * Handle auth_unix or auth_kerb. */ @@ -1889,7 +1988,9 @@ nfs_getreq(nd, has_header) return (EBADRPC); } nfsm_adv(nfsm_rndup(len)); - nfsm_dissect(tl, u_long *, 3*NFSX_UNSIGNED); + nfsm_dissect(tl, u_long *, 3 * NFSX_UNSIGNED); + bzero((caddr_t)&nd->nd_cr, sizeof (struct ucred)); + nd->nd_cr.cr_ref = 1; nd->nd_cr.cr_uid = fxdr_unsigned(uid_t, *tl++); nd->nd_cr.cr_gid = fxdr_unsigned(gid_t, *tl++); len = fxdr_unsigned(int, *tl); @@ -1897,45 +1998,124 @@ nfs_getreq(nd, has_header) m_freem(mrep); return (EBADRPC); } - nfsm_dissect(tl, u_long *, (len + 2)*NFSX_UNSIGNED); + nfsm_dissect(tl, u_long *, (len + 2) * NFSX_UNSIGNED); for (i = 1; i <= len; i++) - if (i < NGROUPS) - nd->nd_cr.cr_groups[i] = fxdr_unsigned(gid_t, *tl++); - else - tl++; + if (i < NGROUPS) + nd->nd_cr.cr_groups[i] = fxdr_unsigned(gid_t, *tl++); + else + tl++; nd->nd_cr.cr_ngroups = (len >= NGROUPS) ? NGROUPS : (len + 1); - } else if (auth_type == rpc_auth_kerb) { - nd->nd_cr.cr_uid = fxdr_unsigned(uid_t, *tl++); - nd->nd_authlen = fxdr_unsigned(int, *tl); - uio.uio_resid = nfsm_rndup(nd->nd_authlen); - if (uio.uio_resid > (len - 2 * NFSX_UNSIGNED)) { + if (nd->nd_cr.cr_ngroups > 1) + nfsrvw_sort(nd->nd_cr.cr_groups, nd->nd_cr.cr_ngroups); + len = fxdr_unsigned(int, *++tl); + if (len < 0 || len > RPCAUTH_MAXSIZ) { m_freem(mrep); return (EBADRPC); } - uio.uio_offset = 0; - uio.uio_iov = &iov; - uio.uio_iovcnt = 1; - uio.uio_segflg = UIO_SYSSPACE; - iov.iov_base = (caddr_t)nd->nd_authstr; - iov.iov_len = RPCAUTH_MAXSIZ; - nfsm_mtouio(&uio, uio.uio_resid); - nfsm_dissect(tl, u_long *, 2 * NFSX_UNSIGNED); - nd->nd_flag |= NFSD_NEEDAUTH; - } + if (len > 0) + nfsm_adv(nfsm_rndup(len)); + } else if (auth_type == rpc_auth_kerb) { + switch (fxdr_unsigned(int, *tl++)) { + case RPCAKN_FULLNAME: + ticklen = fxdr_unsigned(int, *tl); + *((u_long *)nfsd->nfsd_authstr) = *tl; + uio.uio_resid = nfsm_rndup(ticklen) + NFSX_UNSIGNED; + nfsd->nfsd_authlen = uio.uio_resid + NFSX_UNSIGNED; + if (uio.uio_resid > (len - 2 * NFSX_UNSIGNED)) { + m_freem(mrep); + return (EBADRPC); + } + uio.uio_offset = 0; + uio.uio_iov = &iov; + uio.uio_iovcnt = 1; + uio.uio_segflg = UIO_SYSSPACE; + iov.iov_base = (caddr_t)&nfsd->nfsd_authstr[4]; + iov.iov_len = RPCAUTH_MAXSIZ - 4; + nfsm_mtouio(&uio, uio.uio_resid); + nfsm_dissect(tl, u_long *, 2 * NFSX_UNSIGNED); + if (*tl++ != rpc_auth_kerb || + fxdr_unsigned(int, *tl) != 4 * NFSX_UNSIGNED) { + printf("Bad kerb verifier\n"); + nd->nd_repstat = (NFSERR_AUTHERR|AUTH_BADVERF); + nd->nd_procnum = NFSPROC_NOOP; + return (0); + } + nfsm_dissect(cp, caddr_t, 4 * NFSX_UNSIGNED); + tl = (u_long *)cp; + if (fxdr_unsigned(int, *tl) != RPCAKN_FULLNAME) { + printf("Not fullname kerb verifier\n"); + nd->nd_repstat = (NFSERR_AUTHERR|AUTH_BADVERF); + nd->nd_procnum = NFSPROC_NOOP; + return (0); + } + cp += NFSX_UNSIGNED; + bcopy(cp, nfsd->nfsd_verfstr, 3 * NFSX_UNSIGNED); + nfsd->nfsd_verflen = 3 * NFSX_UNSIGNED; + nd->nd_flag |= ND_KERBFULL; + nfsd->nfsd_flag |= NFSD_NEEDAUTH; + break; + case RPCAKN_NICKNAME: + if (len != 2 * NFSX_UNSIGNED) { + printf("Kerb nickname short\n"); + nd->nd_repstat = (NFSERR_AUTHERR|AUTH_BADCRED); + nd->nd_procnum = NFSPROC_NOOP; + return (0); + } + nickuid = fxdr_unsigned(uid_t, *tl); + nfsm_dissect(tl, u_long *, 2 * NFSX_UNSIGNED); + if (*tl++ != rpc_auth_kerb || + fxdr_unsigned(int, *tl) != 3 * NFSX_UNSIGNED) { + printf("Kerb nick verifier bad\n"); + nd->nd_repstat = (NFSERR_AUTHERR|AUTH_BADVERF); + nd->nd_procnum = NFSPROC_NOOP; + return (0); + } + nfsm_dissect(tl, u_long *, 3 * NFSX_UNSIGNED); + tvin.tv_sec = *tl++; + tvin.tv_usec = *tl; + + for (nuidp = NUIDHASH(nfsd->nfsd_slp,nickuid)->lh_first; + nuidp != 0; nuidp = nuidp->nu_hash.le_next) { + if (nuidp->nu_cr.cr_uid == nickuid && + (!nd->nd_nam2 || + netaddr_match(NU_NETFAM(nuidp), + &nuidp->nu_haddr, nd->nd_nam2))) + break; + } + if (!nuidp) { + nd->nd_repstat = + (NFSERR_AUTHERR|AUTH_REJECTCRED); + nd->nd_procnum = NFSPROC_NOOP; + return (0); + } - /* - * Do we have any use for the verifier. - * According to the "Remote Procedure Call Protocol Spec." it - * should be AUTH_NULL, but some clients make it AUTH_UNIX? - * For now, just skip over it - */ - len = fxdr_unsigned(int, *++tl); - if (len < 0 || len > RPCAUTH_MAXSIZ) { - m_freem(mrep); - return (EBADRPC); - } - if (len > 0) { - nfsm_adv(nfsm_rndup(len)); + /* + * Now, decrypt the timestamp using the session key + * and validate it. + */ +#ifdef NFSKERB + XXX +#endif + + tvout.tv_sec = fxdr_unsigned(long, tvout.tv_sec); + tvout.tv_usec = fxdr_unsigned(long, tvout.tv_usec); + if (nuidp->nu_expire < time.tv_sec || + nuidp->nu_timestamp.tv_sec > tvout.tv_sec || + (nuidp->nu_timestamp.tv_sec == tvout.tv_sec && + nuidp->nu_timestamp.tv_usec > tvout.tv_usec)) { + nuidp->nu_expire = 0; + nd->nd_repstat = + (NFSERR_AUTHERR|AUTH_REJECTVERF); + nd->nd_procnum = NFSPROC_NOOP; + return (0); + } + nfsrv_setcred(&nuidp->nu_cr, &nd->nd_cr); + nd->nd_flag |= ND_KERBNICK; + }; + } else { + nd->nd_repstat = (NFSERR_AUTHERR | AUTH_REJECTCRED); + nd->nd_procnum = NFSPROC_NOOP; + return (0); } /* @@ -1943,16 +2123,14 @@ nfs_getreq(nd, has_header) */ if (nqnfs && nd->nd_procnum != NQNFSPROC_EVICTED) { nfsm_dissect(tl, u_long *, NFSX_UNSIGNED); - nd->nd_nqlflag = fxdr_unsigned(int, *tl); - if (nd->nd_nqlflag) { + nd->nd_flag |= fxdr_unsigned(int, *tl); + if (nd->nd_flag & ND_LEASE) { nfsm_dissect(tl, u_long *, NFSX_UNSIGNED); nd->nd_duration = fxdr_unsigned(int, *tl); } else nd->nd_duration = NQ_MINLEASE; - } else { - nd->nd_nqlflag = NQL_NOVAL; + } else nd->nd_duration = NQ_MINLEASE; - } nd->nd_md = md; nd->nd_dpos = dpos; return (0); @@ -1973,13 +2151,13 @@ nfsrv_wakenfsd(slp) if ((slp->ns_flag & SLP_VALID) == 0) return; - for (nd = nfsd_head.tqh_first; nd != 0; nd = nd->nd_chain.tqe_next) { - if (nd->nd_flag & NFSD_WAITING) { - nd->nd_flag &= ~NFSD_WAITING; - if (nd->nd_slp) + for (nd = nfsd_head.tqh_first; nd != 0; nd = nd->nfsd_chain.tqe_next) { + if (nd->nfsd_flag & NFSD_WAITING) { + nd->nfsd_flag &= ~NFSD_WAITING; + if (nd->nfsd_slp) panic("nfsd wakeup"); slp->ns_sref++; - nd->nd_slp = slp; + nd->nfsd_slp = slp; wakeup((caddr_t)nd); return; } diff --git a/sys/nfsclient/nfs_subs.c b/sys/nfsclient/nfs_subs.c index 1653c8b..350ba46 100644 --- a/sys/nfsclient/nfs_subs.c +++ b/sys/nfsclient/nfs_subs.c @@ -34,7 +34,7 @@ * SUCH DAMAGE. * * @(#)nfs_subs.c 8.3 (Berkeley) 1/4/94 - * $Id: nfs_subs.c,v 1.15 1995/05/30 08:12:43 rgrimes Exp $ + * $Id: nfs_subs.c,v 1.16 1995/06/14 06:23:38 joerg Exp $ */ /* @@ -52,6 +52,7 @@ #include <sys/mbuf.h> #include <sys/socket.h> #include <sys/stat.h> +#include <sys/malloc.h> #ifdef VFS_LKM #include <sys/sysent.h> #include <sys/syscall.h> @@ -60,7 +61,7 @@ #include <vm/vm.h> #include <nfs/rpcv2.h> -#include <nfs/nfsv2.h> +#include <nfs/nfsproto.h> #include <nfs/nfsnode.h> #include <nfs/nfs.h> #include <nfs/xdr_subs.h> @@ -78,31 +79,455 @@ #include <netiso/iso.h> #endif -#define TRUE 1 -#define FALSE 0 - /* * Data items converted to xdr at startup, since they are constant * This is kinda hokey, but may save a little time doing byte swaps */ -u_long nfs_procids[NFS_NPROCS]; u_long nfs_xdrneg1; u_long rpc_call, rpc_vers, rpc_reply, rpc_msgdenied, rpc_autherr, - rpc_mismatch, rpc_auth_unix, rpc_msgaccepted, rpc_rejectedcred, + rpc_mismatch, rpc_auth_unix, rpc_msgaccepted, rpc_auth_kerb; -u_long nfs_vers, nfs_prog, nfs_true, nfs_false; +u_long nfs_prog, nqnfs_prog, nfs_true, nfs_false; /* And other global data */ static u_long nfs_xid = 0; -enum vtype ntov_type[7] = { VNON, VREG, VDIR, VBLK, VCHR, VLNK, VNON }; +enum vtype nv2tov_type[8] = { VNON, VREG, VDIR, VBLK, VCHR, VLNK, VNON, VNON }; +enum vtype nv3tov_type[8]={ VNON, VREG, VDIR, VBLK, VCHR, VLNK, VSOCK, VFIFO }; +int nfs_ticks; + +/* + * Mapping of old NFS Version 2 RPC numbers to generic numbers. + */ +int nfsv3_procid[NFS_NPROCS] = { + NFSPROC_NULL, + NFSPROC_GETATTR, + NFSPROC_SETATTR, + NFSPROC_NOOP, + NFSPROC_LOOKUP, + NFSPROC_READLINK, + NFSPROC_READ, + NFSPROC_NOOP, + NFSPROC_WRITE, + NFSPROC_CREATE, + NFSPROC_REMOVE, + NFSPROC_RENAME, + NFSPROC_LINK, + NFSPROC_SYMLINK, + NFSPROC_MKDIR, + NFSPROC_RMDIR, + NFSPROC_READDIR, + NFSPROC_FSSTAT, + NFSPROC_NOOP, + NFSPROC_NOOP, + NFSPROC_NOOP, + NFSPROC_NOOP, + NFSPROC_NOOP, + NFSPROC_NOOP, + NFSPROC_NOOP, + NFSPROC_NOOP +}; + +/* + * and the reverse mapping from generic to Version 2 procedure numbers + */ +int nfsv2_procid[NFS_NPROCS] = { + NFSV2PROC_NULL, + NFSV2PROC_GETATTR, + NFSV2PROC_SETATTR, + NFSV2PROC_LOOKUP, + NFSV2PROC_NOOP, + NFSV2PROC_READLINK, + NFSV2PROC_READ, + NFSV2PROC_WRITE, + NFSV2PROC_CREATE, + NFSV2PROC_MKDIR, + NFSV2PROC_SYMLINK, + NFSV2PROC_CREATE, + NFSV2PROC_REMOVE, + NFSV2PROC_RMDIR, + NFSV2PROC_RENAME, + NFSV2PROC_LINK, + NFSV2PROC_READDIR, + NFSV2PROC_NOOP, + NFSV2PROC_STATFS, + NFSV2PROC_NOOP, + NFSV2PROC_NOOP, + NFSV2PROC_NOOP, + NFSV2PROC_NOOP, + NFSV2PROC_NOOP, + NFSV2PROC_NOOP, + NFSV2PROC_NOOP, +}; + +/* + * Maps errno values to nfs error numbers. + * Use NFSERR_IO as the catch all for ones not specifically defined in + * RFC 1094. + */ +static u_char nfsrv_v2errmap[ELAST] = { + NFSERR_PERM, NFSERR_NOENT, NFSERR_IO, NFSERR_IO, NFSERR_IO, + NFSERR_NXIO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, + NFSERR_IO, NFSERR_IO, NFSERR_ACCES, NFSERR_IO, NFSERR_IO, + NFSERR_IO, NFSERR_EXIST, NFSERR_IO, NFSERR_NODEV, NFSERR_NOTDIR, + NFSERR_ISDIR, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, + NFSERR_IO, NFSERR_FBIG, NFSERR_NOSPC, NFSERR_IO, NFSERR_ROFS, + NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, + NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, + NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, + NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, + NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, + NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, + NFSERR_IO, NFSERR_IO, NFSERR_NAMETOL, NFSERR_IO, NFSERR_IO, + NFSERR_NOTEMPTY, NFSERR_IO, NFSERR_IO, NFSERR_DQUOT, NFSERR_STALE, + NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, + NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, + NFSERR_IO, +}; + +/* + * Maps errno values to nfs error numbers. + * Although it is not obvious whether or not NFS clients really care if + * a returned error value is in the specified list for the procedure, the + * safest thing to do is filter them appropriately. For Version 2, the + * X/Open XNFS document is the only specification that defines error values + * for each RPC (The RFC simply lists all possible error values for all RPCs), + * so I have decided to not do this for Version 2. + * The first entry is the default error return and the rest are the valid + * errors for that RPC in increasing numeric order. + */ +static short nfsv3err_null[] = { + 0, + 0, +}; + +static short nfsv3err_getattr[] = { + NFSERR_IO, + NFSERR_IO, + NFSERR_STALE, + NFSERR_BADHANDLE, + NFSERR_SERVERFAULT, + 0, +}; + +static short nfsv3err_setattr[] = { + NFSERR_IO, + NFSERR_PERM, + NFSERR_IO, + NFSERR_ACCES, + NFSERR_INVAL, + NFSERR_NOSPC, + NFSERR_ROFS, + NFSERR_DQUOT, + NFSERR_STALE, + NFSERR_BADHANDLE, + NFSERR_NOT_SYNC, + NFSERR_SERVERFAULT, + 0, +}; + +static short nfsv3err_lookup[] = { + NFSERR_IO, + NFSERR_NOENT, + NFSERR_IO, + NFSERR_ACCES, + NFSERR_NOTDIR, + NFSERR_NAMETOL, + NFSERR_STALE, + NFSERR_BADHANDLE, + NFSERR_SERVERFAULT, + 0, +}; + +static short nfsv3err_access[] = { + NFSERR_IO, + NFSERR_IO, + NFSERR_STALE, + NFSERR_BADHANDLE, + NFSERR_SERVERFAULT, + 0, +}; + +static short nfsv3err_readlink[] = { + NFSERR_IO, + NFSERR_IO, + NFSERR_ACCES, + NFSERR_INVAL, + NFSERR_STALE, + NFSERR_BADHANDLE, + NFSERR_NOTSUPP, + NFSERR_SERVERFAULT, + 0, +}; + +static short nfsv3err_read[] = { + NFSERR_IO, + NFSERR_IO, + NFSERR_NXIO, + NFSERR_ACCES, + NFSERR_INVAL, + NFSERR_STALE, + NFSERR_BADHANDLE, + NFSERR_SERVERFAULT, + 0, +}; + +static short nfsv3err_write[] = { + NFSERR_IO, + NFSERR_IO, + NFSERR_ACCES, + NFSERR_INVAL, + NFSERR_FBIG, + NFSERR_NOSPC, + NFSERR_ROFS, + NFSERR_DQUOT, + NFSERR_STALE, + NFSERR_BADHANDLE, + NFSERR_SERVERFAULT, + 0, +}; + +static short nfsv3err_create[] = { + NFSERR_IO, + NFSERR_IO, + NFSERR_ACCES, + NFSERR_EXIST, + NFSERR_NOTDIR, + NFSERR_NOSPC, + NFSERR_ROFS, + NFSERR_NAMETOL, + NFSERR_DQUOT, + NFSERR_STALE, + NFSERR_BADHANDLE, + NFSERR_NOTSUPP, + NFSERR_SERVERFAULT, + 0, +}; + +static short nfsv3err_mkdir[] = { + NFSERR_IO, + NFSERR_IO, + NFSERR_ACCES, + NFSERR_EXIST, + NFSERR_NOTDIR, + NFSERR_NOSPC, + NFSERR_ROFS, + NFSERR_NAMETOL, + NFSERR_DQUOT, + NFSERR_STALE, + NFSERR_BADHANDLE, + NFSERR_NOTSUPP, + NFSERR_SERVERFAULT, + 0, +}; + +static short nfsv3err_symlink[] = { + NFSERR_IO, + NFSERR_IO, + NFSERR_ACCES, + NFSERR_EXIST, + NFSERR_NOTDIR, + NFSERR_NOSPC, + NFSERR_ROFS, + NFSERR_NAMETOL, + NFSERR_DQUOT, + NFSERR_STALE, + NFSERR_BADHANDLE, + NFSERR_NOTSUPP, + NFSERR_SERVERFAULT, + 0, +}; + +static short nfsv3err_mknod[] = { + NFSERR_IO, + NFSERR_IO, + NFSERR_ACCES, + NFSERR_EXIST, + NFSERR_NOTDIR, + NFSERR_NOSPC, + NFSERR_ROFS, + NFSERR_NAMETOL, + NFSERR_DQUOT, + NFSERR_STALE, + NFSERR_BADHANDLE, + NFSERR_NOTSUPP, + NFSERR_SERVERFAULT, + NFSERR_BADTYPE, + 0, +}; + +static short nfsv3err_remove[] = { + NFSERR_IO, + NFSERR_NOENT, + NFSERR_IO, + NFSERR_ACCES, + NFSERR_NOTDIR, + NFSERR_ROFS, + NFSERR_NAMETOL, + NFSERR_STALE, + NFSERR_BADHANDLE, + NFSERR_SERVERFAULT, + 0, +}; + +static short nfsv3err_rmdir[] = { + NFSERR_IO, + NFSERR_NOENT, + NFSERR_IO, + NFSERR_ACCES, + NFSERR_EXIST, + NFSERR_NOTDIR, + NFSERR_INVAL, + NFSERR_ROFS, + NFSERR_NAMETOL, + NFSERR_NOTEMPTY, + NFSERR_STALE, + NFSERR_BADHANDLE, + NFSERR_NOTSUPP, + NFSERR_SERVERFAULT, + 0, +}; + +static short nfsv3err_rename[] = { + NFSERR_IO, + NFSERR_NOENT, + NFSERR_IO, + NFSERR_ACCES, + NFSERR_EXIST, + NFSERR_XDEV, + NFSERR_NOTDIR, + NFSERR_ISDIR, + NFSERR_INVAL, + NFSERR_NOSPC, + NFSERR_ROFS, + NFSERR_MLINK, + NFSERR_NAMETOL, + NFSERR_NOTEMPTY, + NFSERR_DQUOT, + NFSERR_STALE, + NFSERR_BADHANDLE, + NFSERR_NOTSUPP, + NFSERR_SERVERFAULT, + 0, +}; + +static short nfsv3err_link[] = { + NFSERR_IO, + NFSERR_IO, + NFSERR_ACCES, + NFSERR_EXIST, + NFSERR_XDEV, + NFSERR_NOTDIR, + NFSERR_INVAL, + NFSERR_NOSPC, + NFSERR_ROFS, + NFSERR_MLINK, + NFSERR_NAMETOL, + NFSERR_DQUOT, + NFSERR_STALE, + NFSERR_BADHANDLE, + NFSERR_NOTSUPP, + NFSERR_SERVERFAULT, + 0, +}; + +static short nfsv3err_readdir[] = { + NFSERR_IO, + NFSERR_IO, + NFSERR_ACCES, + NFSERR_NOTDIR, + NFSERR_STALE, + NFSERR_BADHANDLE, + NFSERR_BAD_COOKIE, + NFSERR_TOOSMALL, + NFSERR_SERVERFAULT, + 0, +}; + +static short nfsv3err_readdirplus[] = { + NFSERR_IO, + NFSERR_IO, + NFSERR_ACCES, + NFSERR_NOTDIR, + NFSERR_STALE, + NFSERR_BADHANDLE, + NFSERR_BAD_COOKIE, + NFSERR_NOTSUPP, + NFSERR_TOOSMALL, + NFSERR_SERVERFAULT, + 0, +}; + +static short nfsv3err_fsstat[] = { + NFSERR_IO, + NFSERR_IO, + NFSERR_STALE, + NFSERR_BADHANDLE, + NFSERR_SERVERFAULT, + 0, +}; + +static short nfsv3err_fsinfo[] = { + NFSERR_STALE, + NFSERR_STALE, + NFSERR_BADHANDLE, + NFSERR_SERVERFAULT, + 0, +}; + +static short nfsv3err_pathconf[] = { + NFSERR_STALE, + NFSERR_STALE, + NFSERR_BADHANDLE, + NFSERR_SERVERFAULT, + 0, +}; + +static short nfsv3err_commit[] = { + NFSERR_IO, + NFSERR_IO, + NFSERR_STALE, + NFSERR_BADHANDLE, + NFSERR_SERVERFAULT, + 0, +}; + +static short *nfsrv_v3errmap[] = { + nfsv3err_null, + nfsv3err_getattr, + nfsv3err_setattr, + nfsv3err_lookup, + nfsv3err_access, + nfsv3err_readlink, + nfsv3err_read, + nfsv3err_write, + nfsv3err_create, + nfsv3err_mkdir, + nfsv3err_symlink, + nfsv3err_mknod, + nfsv3err_remove, + nfsv3err_rmdir, + nfsv3err_rename, + nfsv3err_link, + nfsv3err_readdir, + nfsv3err_readdirplus, + nfsv3err_fsstat, + nfsv3err_fsinfo, + nfsv3err_pathconf, + nfsv3err_commit, +}; + extern struct proc *nfs_iodwant[NFS_MAXASYNCDAEMON]; -extern int nqnfs_piggy[NFS_NPROCS]; extern struct nfsrtt nfsrtt; extern time_t nqnfsstarttime; -extern u_long nqnfs_prog, nqnfs_vers; extern int nqsrv_clockskew; extern int nqsrv_writeslack; extern int nqsrv_maxlease; +extern struct nfsstats nfsstats; +extern int nqnfs_piggy[NFS_NPROCS]; +extern nfstype nfsv2_type[9]; +extern nfstype nfsv3_type[9]; +extern struct nfsnodehashhead *nfsnodehashtbl; +extern u_long nfsnodehash; #ifdef VFS_LKM struct getfh_args; @@ -167,14 +592,16 @@ nfsm_reqh(vp, procid, hsiz, bposp) * Returns the head of the mbuf list. */ struct mbuf * -nfsm_rpchead(cr, nqnfs, procid, auth_type, auth_len, auth_str, mrest, - mrest_len, mbp, xidp) +nfsm_rpchead(cr, nmflag, procid, auth_type, auth_len, auth_str, verf_len, + verf_str, mrest, mrest_len, mbp, xidp) register struct ucred *cr; - int nqnfs; + int nmflag; int procid; int auth_type; int auth_len; char *auth_str; + int verf_len; + char *verf_str; struct mbuf *mrest; int mrest_len; struct mbuf **mbp; @@ -188,15 +615,13 @@ nfsm_rpchead(cr, nqnfs, procid, auth_type, auth_len, auth_str, mrest, int siz, grpsiz, authsiz; authsiz = nfsm_rndup(auth_len); - if (auth_type == RPCAUTH_NQNFS) - authsiz += 2 * NFSX_UNSIGNED; MGETHDR(mb, M_WAIT, MT_DATA); - if ((authsiz + 10*NFSX_UNSIGNED) >= MINCLSIZE) { + if ((authsiz + 10 * NFSX_UNSIGNED) >= MINCLSIZE) { MCLGET(mb, M_WAIT); - } else if ((authsiz + 10*NFSX_UNSIGNED) < MHLEN) { - MH_ALIGN(mb, authsiz + 10*NFSX_UNSIGNED); + } else if ((authsiz + 10 * NFSX_UNSIGNED) < MHLEN) { + MH_ALIGN(mb, authsiz + 10 * NFSX_UNSIGNED); } else { - MH_ALIGN(mb, 8*NFSX_UNSIGNED); + MH_ALIGN(mb, 8 * NFSX_UNSIGNED); } mb->m_len = 0; mreq = mb; @@ -205,20 +630,26 @@ nfsm_rpchead(cr, nqnfs, procid, auth_type, auth_len, auth_str, mrest, /* * First the RPC header. */ - nfsm_build(tl, u_long *, 8*NFSX_UNSIGNED); + nfsm_build(tl, u_long *, 8 * NFSX_UNSIGNED); if (++nfs_xid == 0) nfs_xid++; *tl++ = *xidp = txdr_unsigned(nfs_xid); *tl++ = rpc_call; *tl++ = rpc_vers; - if (nqnfs) { + if (nmflag & NFSMNT_NQNFS) { *tl++ = txdr_unsigned(NQNFS_PROG); - *tl++ = txdr_unsigned(NQNFS_VER1); + *tl++ = txdr_unsigned(NQNFS_VER3); } else { *tl++ = txdr_unsigned(NFS_PROG); - *tl++ = txdr_unsigned(NFS_VER2); + if (nmflag & NFSMNT_NFSV3) + *tl++ = txdr_unsigned(NFS_VER3); + else + *tl++ = txdr_unsigned(NFS_VER2); } - *tl++ = txdr_unsigned(procid); + if (nmflag & NFSMNT_NFSV3) + *tl++ = txdr_unsigned(procid); + else + *tl++ = txdr_unsigned(nfsv2_procid[procid]); /* * And then the authorization cred. @@ -237,10 +668,7 @@ nfsm_rpchead(cr, nqnfs, procid, auth_type, auth_len, auth_str, mrest, for (i = 1; i <= grpsiz; i++) *tl++ = txdr_unsigned(cr->cr_groups[i]); break; - case RPCAUTH_NQNFS: - nfsm_build(tl, u_long *, 2*NFSX_UNSIGNED); - *tl++ = txdr_unsigned(cr->cr_uid); - *tl = txdr_unsigned(auth_len); + case RPCAUTH_KERB4: siz = auth_len; while (siz > 0) { if (M_TRAILINGSPACE(mb) == 0) { @@ -266,11 +694,43 @@ nfsm_rpchead(cr, nqnfs, procid, auth_type, auth_len, auth_str, mrest, } break; }; - nfsm_build(tl, u_long *, 2*NFSX_UNSIGNED); - *tl++ = txdr_unsigned(RPCAUTH_NULL); - *tl = 0; + + /* + * And the verifier... + */ + nfsm_build(tl, u_long *, 2 * NFSX_UNSIGNED); + if (verf_str) { + *tl++ = txdr_unsigned(RPCAUTH_KERB4); + *tl = txdr_unsigned(verf_len); + siz = verf_len; + while (siz > 0) { + if (M_TRAILINGSPACE(mb) == 0) { + MGET(mb2, M_WAIT, MT_DATA); + if (siz >= MINCLSIZE) + MCLGET(mb2, M_WAIT); + mb->m_next = mb2; + mb = mb2; + mb->m_len = 0; + bpos = mtod(mb, caddr_t); + } + i = min(siz, M_TRAILINGSPACE(mb)); + bcopy(verf_str, bpos, i); + mb->m_len += i; + verf_str += i; + bpos += i; + siz -= i; + } + if ((siz = (nfsm_rndup(verf_len) - verf_len)) > 0) { + for (i = 0; i < siz; i++) + *bpos++ = '\0'; + mb->m_len += siz; + } + } else { + *tl++ = txdr_unsigned(RPCAUTH_NULL); + *tl = 0; + } mb->m_next = mrest; - mreq->m_pkthdr.len = authsiz + 10*NFSX_UNSIGNED + mrest_len; + mreq->m_pkthdr.len = authsiz + 10 * NFSX_UNSIGNED + mrest_len; mreq->m_pkthdr.rcvif = (struct ifnet *)0; *mbp = mb; return (mreq); @@ -603,6 +1063,25 @@ nfs_init() { register int i; + /* + * Check to see if major data structures haven't bloated. + */ + if (sizeof (struct nfsnode) > NFS_NODEALLOC) { + printf("struct nfsnode bloated (> %dbytes)\n", NFS_NODEALLOC); + printf("Try reducing NFS_SMALLFH\n"); + } + if (sizeof (struct nfsmount) > NFS_MNTALLOC) { + printf("struct nfsmount bloated (> %dbytes)\n", NFS_MNTALLOC); + printf("Try reducing NFS_MUIDHASHSIZ\n"); + } + if (sizeof (struct nfssvc_sock) > NFS_SVCALLOC) { + printf("struct nfssvc_sock bloated (> %dbytes)\n",NFS_SVCALLOC); + printf("Try reducing NFS_UIDHASHSIZ\n"); + } + if (sizeof (struct nfsuid) > NFS_UIDALLOC) { + printf("struct nfsuid bloated (> %dbytes)\n",NFS_UIDALLOC); + printf("Try unionizing the nu_nickname and nu_flag fields\n"); + } nfsrtt.pos = 0; rpc_vers = txdr_unsigned(RPC_VER2); rpc_call = txdr_unsigned(RPC_CALL); @@ -611,17 +1090,16 @@ nfs_init() rpc_msgaccepted = txdr_unsigned(RPC_MSGACCEPTED); rpc_mismatch = txdr_unsigned(RPC_MISMATCH); rpc_autherr = txdr_unsigned(RPC_AUTHERR); - rpc_rejectedcred = txdr_unsigned(AUTH_REJECTCRED); rpc_auth_unix = txdr_unsigned(RPCAUTH_UNIX); - rpc_auth_kerb = txdr_unsigned(RPCAUTH_NQNFS); - nfs_vers = txdr_unsigned(NFS_VER2); + rpc_auth_kerb = txdr_unsigned(RPCAUTH_KERB4); nfs_prog = txdr_unsigned(NFS_PROG); + nqnfs_prog = txdr_unsigned(NQNFS_PROG); nfs_true = txdr_unsigned(TRUE); nfs_false = txdr_unsigned(FALSE); nfs_xdrneg1 = txdr_unsigned(-1); - /* Loop thru nfs procids */ - for (i = 0; i < NFS_NPROCS; i++) - nfs_procids[i] = txdr_unsigned(i); + nfs_ticks = (hz * NFS_TICKINTVL + 500) / 1000; + if (nfs_ticks < 1) + nfs_ticks = 1; /* Ensure async daemons disabled */ for (i = 0; i < NFS_MAXASYNCDAEMON; i++) nfs_iodwant[i] = (struct proc *)0; @@ -637,8 +1115,6 @@ nfs_init() nqnfsstarttime = boottime.tv_sec + nqsrv_maxlease + nqsrv_clockskew + nqsrv_writeslack; NQLOADNOVRAM(nqnfsstarttime); - nqnfs_prog = txdr_unsigned(NQNFS_PROG); - nqnfs_vers = txdr_unsigned(NQNFS_VER1); CIRCLEQ_INIT(&nqtimerhead); nqfhhashtbl = hashinit(NQLCHSZ, M_NQLEASE, &nqfhhash); } @@ -649,6 +1125,7 @@ nfs_init() TAILQ_INIT(&nfs_reqq); nfs_timer(0); +#ifdef __FreeBSD__ /* * Set up lease_check and lease_updatetime so that other parts * of the system can call us, if we are loadable. @@ -662,6 +1139,7 @@ nfs_init() sysent[SYS_getfh].sy_narg = 2; sysent[SYS_getfh].sy_call = getfh; #endif +#endif return (0); } @@ -689,57 +1167,65 @@ nfs_loadattrcache(vpp, mdp, dposp, vaper) { register struct vnode *vp = *vpp; register struct vattr *vap; - register struct nfsv2_fattr *fp; + register struct nfs_fattr *fp; register struct nfsnode *np; register struct nfsnodehashhead *nhpp; register long t1; - caddr_t dpos, cp2; - int error = 0, isnq; + caddr_t cp2; + int error = 0, rdev; struct mbuf *md; enum vtype vtyp; u_short vmode; - long rdev; struct timespec mtime; struct vnode *nvp; + quad_t tval; + int v3 = NFS_ISV3(vp); md = *mdp; - dpos = *dposp; - t1 = (mtod(md, caddr_t) + md->m_len) - dpos; - isnq = (VFSTONFS(vp->v_mount)->nm_flag & NFSMNT_NQNFS); - error = nfsm_disct(&md, &dpos, NFSX_FATTR(isnq), t1, &cp2); - if (error) + t1 = (mtod(md, caddr_t) + md->m_len) - *dposp; + if (error = nfsm_disct(mdp, dposp, NFSX_FATTR(v3), t1, &cp2)) return (error); - fp = (struct nfsv2_fattr *)cp2; - vtyp = nfstov_type(fp->fa_type); - vmode = fxdr_unsigned(u_short, fp->fa_mode); - /* - * XXX - * - * The duplicate information returned in fa_type and fa_mode - * is an ambiguity in the NFS version 2 protocol. - * - * VREG should be taken literally as a regular file. If a - * server intents to return some type information differently - * in the upper bits of the mode field (e.g. for sockets, or - * FIFOs), NFSv2 mandates fa_type to be VNON. Anyway, we - * leave the examination of the mode bits even in the VREG - * case to avoid breakage for bogus servers, but we make sure - * that there are actually type bits set in the upper part of - * fa_mode (and failing that, trust the va_type field). - * - * NFSv3 cleared the issue, and requires fa_mode to not - * contain any type information (while also introduing sockets - * and FIFOs for fa_type). - */ - if (vtyp == VNON || (vtyp == VREG && (vmode & S_IFMT) != 0)) - vtyp = IFTOVT(vmode); - if (isnq) { - rdev = fxdr_unsigned(long, fp->fa_nqrdev); - fxdr_nqtime(&fp->fa_nqmtime, &mtime); + fp = (struct nfs_fattr *)cp2; + if (v3) { + vtyp = nfsv3tov_type(fp->fa_type); + vmode = fxdr_unsigned(u_short, fp->fa_mode); + rdev = makedev(fxdr_unsigned(u_char, fp->fa3_rdev.specdata1), + fxdr_unsigned(u_char, fp->fa3_rdev.specdata2)); + fxdr_nfsv3time(&fp->fa3_mtime, &mtime); } else { - rdev = fxdr_unsigned(long, fp->fa_nfsrdev); - fxdr_nfstime(&fp->fa_nfsmtime, &mtime); + vtyp = nfsv2tov_type(fp->fa_type); + vmode = fxdr_unsigned(u_short, fp->fa_mode); + /* + * XXX + * + * The duplicate information returned in fa_type and fa_mode + * is an ambiguity in the NFS version 2 protocol. + * + * VREG should be taken literally as a regular file. If a + * server intents to return some type information differently + * in the upper bits of the mode field (e.g. for sockets, or + * FIFOs), NFSv2 mandates fa_type to be VNON. Anyway, we + * leave the examination of the mode bits even in the VREG + * case to avoid breakage for bogus servers, but we make sure + * that there are actually type bits set in the upper part of + * fa_mode (and failing that, trust the va_type field). + * + * NFSv3 cleared the issue, and requires fa_mode to not + * contain any type information (while also introduing sockets + * and FIFOs for fa_type). + */ + if (vtyp == VNON || (vtyp == VREG && (vmode & S_IFMT) != 0)) + vtyp = IFTOVT(vmode); + rdev = fxdr_unsigned(long, fp->fa2_rdev); + fxdr_nfsv2time(&fp->fa2_mtime, &mtime); + + /* + * Really ugly NFSv2 kludge. + */ + if (vtyp == VCHR && rdev == 0xffffffff) + vtyp = VFIFO; } + /* * If v_type == VNON it is a new node, so fill in the v_type, * n_mtime fields. Check to see if it represents a special @@ -749,10 +1235,15 @@ nfs_loadattrcache(vpp, mdp, dposp, vaper) */ np = VTONFS(vp); if (vp->v_type == VNON) { - if (vtyp == VCHR && rdev == 0xffffffff) - vp->v_type = vtyp = VFIFO; - else - vp->v_type = vtyp; + /* + * If we had a lock and it turns out that the vnode + * is an object which we don't want to lock (e.g. VDIR) + * to avoid nasty hanging problems on a server crash, + * then release it here. + */ + if (vtyp != VREG && VOP_ISLOCKED(vp)) + VOP_UNLOCK(vp); + vp->v_type = vtyp; if (vp->v_type == VFIFO) { vp->v_op = fifo_nfsv2nodeop_p; } @@ -773,7 +1264,7 @@ nfs_loadattrcache(vpp, mdp, dposp, vaper) * Reinitialize aliased node. */ np->n_vnode = nvp; - nhpp = nfs_hash(&np->n_fh); + nhpp = NFSNOHASH(nfs_hash(np->n_fhp, np->n_fhsize)); LIST_INSERT_HEAD(nhpp, np, n_hash); *vpp = vp = nvp; } @@ -783,31 +1274,34 @@ nfs_loadattrcache(vpp, mdp, dposp, vaper) vap = &np->n_vattr; vap->va_type = vtyp; vap->va_mode = (vmode & 07777); - vap->va_nlink = fxdr_unsigned(u_short, fp->fa_nlink); - vap->va_uid = fxdr_unsigned(uid_t, fp->fa_uid); - vap->va_gid = fxdr_unsigned(gid_t, fp->fa_gid); vap->va_rdev = (dev_t)rdev; vap->va_mtime = mtime; vap->va_fsid = vp->v_mount->mnt_stat.f_fsid.val[0]; - if (isnq) { - fxdr_hyper(&fp->fa_nqsize, &vap->va_size); - vap->va_blocksize = fxdr_unsigned(long, fp->fa_nqblocksize); - fxdr_hyper(&fp->fa_nqbytes, &vap->va_bytes); - vap->va_fileid = fxdr_unsigned(long, fp->fa_nqfileid); - fxdr_nqtime(&fp->fa_nqatime, &vap->va_atime); - vap->va_flags = fxdr_unsigned(u_long, fp->fa_nqflags); - fxdr_nqtime(&fp->fa_nqctime, &vap->va_ctime); - vap->va_gen = fxdr_unsigned(u_long, fp->fa_nqgen); - fxdr_hyper(&fp->fa_nqfilerev, &vap->va_filerev); + if (v3) { + vap->va_nlink = fxdr_unsigned(u_short, fp->fa_nlink); + vap->va_uid = fxdr_unsigned(uid_t, fp->fa_uid); + vap->va_gid = fxdr_unsigned(gid_t, fp->fa_gid); + fxdr_hyper(&fp->fa3_size, &vap->va_size); + vap->va_blocksize = NFS_FABLKSIZE; + fxdr_hyper(&fp->fa3_used, &vap->va_bytes); + vap->va_fileid = fxdr_unsigned(int, fp->fa3_fileid.nfsuquad[1]); + fxdr_nfsv3time(&fp->fa3_atime, &vap->va_atime); + fxdr_nfsv3time(&fp->fa3_ctime, &vap->va_ctime); + vap->va_flags = 0; + vap->va_filerev = 0; } else { - vap->va_size = fxdr_unsigned(u_long, fp->fa_nfssize); - vap->va_blocksize = fxdr_unsigned(long, fp->fa_nfsblocksize); - vap->va_bytes = fxdr_unsigned(long, fp->fa_nfsblocks) * NFS_FABLKSIZE; - vap->va_fileid = fxdr_unsigned(long, fp->fa_nfsfileid); - fxdr_nfstime(&fp->fa_nfsatime, &vap->va_atime); + vap->va_nlink = fxdr_unsigned(u_short, fp->fa_nlink); + vap->va_uid = fxdr_unsigned(uid_t, fp->fa_uid); + vap->va_gid = fxdr_unsigned(gid_t, fp->fa_gid); + vap->va_size = fxdr_unsigned(u_long, fp->fa2_size); + vap->va_blocksize = fxdr_unsigned(long, fp->fa2_blocksize); + vap->va_bytes = fxdr_unsigned(long, fp->fa2_blocks) * NFS_FABLKSIZE; + vap->va_fileid = fxdr_unsigned(long, fp->fa2_fileid); + fxdr_nfsv2time(&fp->fa2_atime, &vap->va_atime); vap->va_flags = 0; - fxdr_nfstime(&fp->fa_nfsctime, &vap->va_ctime); - vap->va_gen = 0; + vap->va_ctime.ts_sec = fxdr_unsigned(long, fp->fa2_ctime.nfsv2_sec); + vap->va_ctime.ts_nsec = 0; + vap->va_gen = fxdr_unsigned(u_long, fp->fa2_ctime.nfsv2_usec); vap->va_filerev = 0; } if (vap->va_size != np->n_size) { @@ -824,26 +1318,13 @@ nfs_loadattrcache(vpp, mdp, dposp, vaper) np->n_size = vap->va_size; } np->n_attrstamp = time.tv_sec; - *dposp = dpos; - *mdp = md; if (vaper != NULL) { bcopy((caddr_t)vap, (caddr_t)vaper, sizeof(*vap)); -#ifdef notdef - if ((np->n_flag & NMODIFIED) && np->n_size > vap->va_size) - if (np->n_size > vap->va_size) - vaper->va_size = np->n_size; -#endif if (np->n_flag & NCHG) { - if (np->n_flag & NACC) { - vaper->va_atime.ts_sec = np->n_atim.tv_sec; - vaper->va_atime.ts_nsec = - np->n_atim.tv_usec * 1000; - } - if (np->n_flag & NUPD) { - vaper->va_mtime.ts_sec = np->n_mtim.tv_sec; - vaper->va_mtime.ts_nsec = - np->n_mtim.tv_usec * 1000; - } + if (np->n_flag & NACC) + vaper->va_atime = np->n_atim; + if (np->n_flag & NUPD) + vaper->va_mtime = np->n_mtim; } } return (0); @@ -862,12 +1343,7 @@ nfs_getattrcache(vp, vaper) register struct nfsnode *np = VTONFS(vp); register struct vattr *vap; - if (VFSTONFS(vp->v_mount)->nm_flag & NFSMNT_NQLOOKLEASE) { - if (!NQNFS_CKCACHABLE(vp, NQL_READ) || np->n_attrstamp == 0) { - nfsstats.attrcache_misses++; - return (ENOENT); - } - } else if ((time.tv_sec - np->n_attrstamp) >= NFS_ATTRTIMEO(np)) { + if ((time.tv_sec - np->n_attrstamp) >= NFS_ATTRTIMEO(np)) { nfsstats.attrcache_misses++; return (ENOENT); } @@ -887,23 +1363,11 @@ nfs_getattrcache(vp, vaper) np->n_size = vap->va_size; } bcopy((caddr_t)vap, (caddr_t)vaper, sizeof(struct vattr)); -#ifdef notdef - if ((np->n_flag & NMODIFIED) == 0) { - np->n_size = vaper->va_size; - vnode_pager_setsize(vp, (u_long)np->n_size); - } else if (np->n_size > vaper->va_size) - if (np->n_size > vaper->va_size) - vaper->va_size = np->n_size; -#endif if (np->n_flag & NCHG) { - if (np->n_flag & NACC) { - vaper->va_atime.ts_sec = np->n_atim.tv_sec; - vaper->va_atime.ts_nsec = np->n_atim.tv_usec * 1000; - } - if (np->n_flag & NUPD) { - vaper->va_mtime.ts_sec = np->n_mtim.tv_sec; - vaper->va_mtime.ts_nsec = np->n_mtim.tv_usec * 1000; - } + if (np->n_flag & NACC) + vaper->va_atime = np->n_atim; + if (np->n_flag & NUPD) + vaper->va_mtime = np->n_mtim; } return (0); } @@ -912,7 +1376,7 @@ nfs_getattrcache(vp, vaper) * Set up nameidata for a lookup() call and do it */ int -nfs_namei(ndp, fhp, len, slp, nam, mdp, dposp, p) +nfs_namei(ndp, fhp, len, slp, nam, mdp, dposp, retdirp, p, kerbflag) register struct nameidata *ndp; fhandle_t *fhp; int len; @@ -920,7 +1384,9 @@ nfs_namei(ndp, fhp, len, slp, nam, mdp, dposp, p) struct mbuf *nam; struct mbuf **mdp; caddr_t *dposp; + struct vnode **retdirp; struct proc *p; + int kerbflag; { register int i, rem; register struct mbuf *md; @@ -929,6 +1395,7 @@ nfs_namei(ndp, fhp, len, slp, nam, mdp, dposp, p) int error, rdonly; struct componentname *cnp = &ndp->ni_cnd; + *retdirp = (struct vnode *)0; MALLOC(cnp->cn_pnbuf, char *, len + 1, M_NAMEI, M_WAITOK); /* * Copy the name from the mbuf list to ndp->ni_pnbuf @@ -950,7 +1417,7 @@ nfs_namei(ndp, fhp, len, slp, nam, mdp, dposp, p) rem = md->m_len; } if (*fromcp == '\0' || *fromcp == '/') { - error = EINVAL; + error = EACCES; goto out; } cnp->cn_hash += (unsigned char)*fromcp; @@ -964,26 +1431,24 @@ nfs_namei(ndp, fhp, len, slp, nam, mdp, dposp, p) if (len > 0) { if (rem >= len) *dposp += len; - else { - error = nfs_adv(mdp, dposp, len, rem); - if (error) - goto out; - } + else if (error = nfs_adv(mdp, dposp, len, rem)) + goto out; } ndp->ni_pathlen = tocp - cnp->cn_pnbuf; cnp->cn_nameptr = cnp->cn_pnbuf; /* * Extract and set starting directory. */ - error = nfsrv_fhtovp(fhp, FALSE, &dp, ndp->ni_cnd.cn_cred, slp, - nam, &rdonly); - if (error) + if (error = nfsrv_fhtovp(fhp, FALSE, &dp, ndp->ni_cnd.cn_cred, slp, + nam, &rdonly, kerbflag)) goto out; if (dp->v_type != VDIR) { nfsrv_vrele(dp); error = ENOTDIR; goto out; } + VREF(dp); + *retdirp = dp; ndp->ni_startdir = dp; if (rdonly) cnp->cn_flags |= (NOCROSSMOUNT | RDONLY); @@ -993,8 +1458,7 @@ nfs_namei(ndp, fhp, len, slp, nam, mdp, dposp, p) * And call lookup() to do the real work */ cnp->cn_proc = p; - error = lookup(ndp); - if (error) + if (error = lookup(ndp)) goto out; /* * Check for encountering a symbolic link @@ -1087,6 +1551,109 @@ nfsm_adj(mp, len, nul) } /* + * Make these functions instead of macros, so that the kernel text size + * doesn't get too big... + */ +void +nfsm_srvwcc(nfsd, before_ret, before_vap, after_ret, after_vap, mbp, bposp) + struct nfsrv_descript *nfsd; + int before_ret; + register struct vattr *before_vap; + int after_ret; + struct vattr *after_vap; + struct mbuf **mbp; + char **bposp; +{ + register struct mbuf *mb = *mbp, *mb2; + register char *bpos = *bposp; + register u_long *tl; + + if (before_ret) { + nfsm_build(tl, u_long *, NFSX_UNSIGNED); + *tl = nfs_false; + } else { + nfsm_build(tl, u_long *, 7 * NFSX_UNSIGNED); + *tl++ = nfs_true; + txdr_hyper(&(before_vap->va_size), tl); + tl += 2; + txdr_nfsv3time(&(before_vap->va_mtime), tl); + tl += 2; + txdr_nfsv3time(&(before_vap->va_ctime), tl); + } + *bposp = bpos; + *mbp = mb; + nfsm_srvpostopattr(nfsd, after_ret, after_vap, mbp, bposp); +} + +void +nfsm_srvpostopattr(nfsd, after_ret, after_vap, mbp, bposp) + struct nfsrv_descript *nfsd; + int after_ret; + struct vattr *after_vap; + struct mbuf **mbp; + char **bposp; +{ + register struct mbuf *mb = *mbp, *mb2; + register char *bpos = *bposp; + register u_long *tl; + register struct nfs_fattr *fp; + + if (after_ret) { + nfsm_build(tl, u_long *, NFSX_UNSIGNED); + *tl = nfs_false; + } else { + nfsm_build(tl, u_long *, NFSX_UNSIGNED + NFSX_V3FATTR); + *tl++ = nfs_true; + fp = (struct nfs_fattr *)tl; + nfsm_srvfattr(nfsd, after_vap, fp); + } + *mbp = mb; + *bposp = bpos; +} + +void +nfsm_srvfattr(nfsd, vap, fp) + register struct nfsrv_descript *nfsd; + register struct vattr *vap; + register struct nfs_fattr *fp; +{ + + fp->fa_nlink = txdr_unsigned(vap->va_nlink); + fp->fa_uid = txdr_unsigned(vap->va_uid); + fp->fa_gid = txdr_unsigned(vap->va_gid); + if (nfsd->nd_flag & ND_NFSV3) { + fp->fa_type = vtonfsv3_type(vap->va_type); + fp->fa_mode = vtonfsv3_mode(vap->va_mode); + txdr_hyper(&vap->va_size, &fp->fa3_size); + txdr_hyper(&vap->va_bytes, &fp->fa3_used); + fp->fa3_rdev.specdata1 = txdr_unsigned(major(vap->va_rdev)); + fp->fa3_rdev.specdata2 = txdr_unsigned(minor(vap->va_rdev)); + fp->fa3_fsid.nfsuquad[0] = 0; + fp->fa3_fsid.nfsuquad[1] = txdr_unsigned(vap->va_fsid); + fp->fa3_fileid.nfsuquad[0] = 0; + fp->fa3_fileid.nfsuquad[1] = txdr_unsigned(vap->va_fileid); + txdr_nfsv3time(&vap->va_atime, &fp->fa3_atime); + txdr_nfsv3time(&vap->va_mtime, &fp->fa3_mtime); + txdr_nfsv3time(&vap->va_ctime, &fp->fa3_ctime); + } else { + fp->fa_type = vtonfsv2_type(vap->va_type); + fp->fa_mode = vtonfsv2_mode(vap->va_type, vap->va_mode); + fp->fa2_size = txdr_unsigned(vap->va_size); + fp->fa2_blocksize = txdr_unsigned(vap->va_blocksize); + if (vap->va_type == VFIFO) + fp->fa2_rdev = 0xffffffff; + else + fp->fa2_rdev = txdr_unsigned(vap->va_rdev); + fp->fa2_blocks = txdr_unsigned(vap->va_bytes / NFS_FABLKSIZE); + fp->fa2_fsid = txdr_unsigned(vap->va_fsid); + fp->fa2_fileid = txdr_unsigned(vap->va_fileid); + txdr_nfsv2time(&vap->va_atime, &fp->fa2_atime); + txdr_nfsv2time(&vap->va_mtime, &fp->fa2_mtime); + txdr_nfsv2time(&vap->va_ctime, &fp->fa2_ctime); + } +} + +/* * nfsrv_fhtovp() - convert a fh to a vnode ptr (optionally locked) * - look up fsid in mount list (if not found ret error) * - get vp and export rights by calling VFS_FHTOVP() @@ -1094,7 +1661,7 @@ nfsm_adj(mp, len, nul) * - if not lockflag unlock it with VOP_UNLOCK() */ int -nfsrv_fhtovp(fhp, lockflag, vpp, cred, slp, nam, rdonlyp) +nfsrv_fhtovp(fhp, lockflag, vpp, cred, slp, nam, rdonlyp, kerbflag) fhandle_t *fhp; int lockflag; struct vnode **vpp; @@ -1102,6 +1669,7 @@ nfsrv_fhtovp(fhp, lockflag, vpp, cred, slp, nam, rdonlyp) struct nfssvc_sock *slp; struct mbuf *nam; int *rdonlyp; + int kerbflag; { register struct mount *mp; register struct nfsuid *uidp; @@ -1120,19 +1688,13 @@ nfsrv_fhtovp(fhp, lockflag, vpp, cred, slp, nam, rdonlyp) * Check/setup credentials. */ if (exflags & MNT_EXKERB) { - for (uidp = NUIDHASH(slp, cred->cr_uid)->lh_first; uidp != 0; - uidp = uidp->nu_hash.le_next) { - if (uidp->nu_uid == cred->cr_uid) - break; - } - if (uidp == 0) { + if (!kerbflag) { vput(*vpp); - return (NQNFS_AUTHERR); + return (NFSERR_AUTHERR | AUTH_TOOWEAK); } - cred->cr_uid = uidp->nu_cr.cr_uid; - for (i = 0; i < uidp->nu_cr.cr_ngroups; i++) - cred->cr_groups[i] = uidp->nu_cr.cr_groups[i]; - cred->cr_ngroups = i; + } else if (kerbflag) { + vput(*vpp); + return (NFSERR_AUTHERR | AUTH_TOOWEAK); } else if (cred->cr_uid == 0 || (exflags & MNT_EXPORTANON)) { cred->cr_uid = credanon->cr_uid; for (i = 0; i < credanon->cr_ngroups && i < NGROUPS; i++) @@ -1194,6 +1756,145 @@ netaddr_match(family, haddr, nam) return (0); } +static nfsuint64 nfs_nullcookie = { 0, 0 }; +/* + * This function finds the directory cookie that corresponds to the + * logical byte offset given. + */ +nfsuint64 * +nfs_getcookie(np, off, add) + register struct nfsnode *np; + off_t off; + int add; +{ + register struct nfsdmap *dp, *dp2; + register int pos; + + pos = off / NFS_DIRBLKSIZ; + if (pos == 0) { +#ifdef DIAGNOSTIC + if (add) + panic("nfs getcookie add at 0"); +#endif + return (&nfs_nullcookie); + } + pos--; + dp = np->n_cookies.lh_first; + if (!dp) { + if (add) { + MALLOC(dp, struct nfsdmap *, sizeof (struct nfsdmap), + M_NFSDIROFF, M_WAITOK); + dp->ndm_eocookie = 0; + LIST_INSERT_HEAD(&np->n_cookies, dp, ndm_list); + } else + return ((nfsuint64 *)0); + } + while (pos >= NFSNUMCOOKIES) { + pos -= NFSNUMCOOKIES; + if (dp->ndm_list.le_next) { + if (!add && dp->ndm_eocookie < NFSNUMCOOKIES && + pos >= dp->ndm_eocookie) + return ((nfsuint64 *)0); + dp = dp->ndm_list.le_next; + } else if (add) { + MALLOC(dp2, struct nfsdmap *, sizeof (struct nfsdmap), + M_NFSDIROFF, M_WAITOK); + dp2->ndm_eocookie = 0; + LIST_INSERT_AFTER(dp, dp2, ndm_list); + dp = dp2; + } else + return ((nfsuint64 *)0); + } + if (pos >= dp->ndm_eocookie) { + if (add) + dp->ndm_eocookie = pos + 1; + else + return ((nfsuint64 *)0); + } + return (&dp->ndm_cookies[pos]); +} + +/* + * Invalidate cached directory information, except for the actual directory + * blocks (which are invalidated separately). + * Done mainly to avoid the use of stale offset cookies. + */ +void +nfs_invaldir(vp) + register struct vnode *vp; +{ + register struct nfsnode *np = VTONFS(vp); + +#ifdef DIAGNOSTIC + if (vp->v_type != VDIR) + panic("nfs: invaldir not dir"); +#endif + np->n_direofoffset = 0; + np->n_cookieverf.nfsuquad[0] = 0; + np->n_cookieverf.nfsuquad[1] = 0; + if (np->n_cookies.lh_first) + np->n_cookies.lh_first->ndm_eocookie = 0; +} + +/* + * The write verifier has changed (probably due to a server reboot), so all + * B_NEEDCOMMIT blocks will have to be written again. Since they are on the + * dirty block list as B_DELWRI, all this takes is clearing the B_NEEDCOMMIT + * flag. Once done the new write verifier can be set for the mount point. + */ +void +nfs_clearcommit(mp) + struct mount *mp; +{ + register struct vnode *vp, *nvp; + register struct buf *bp, *nbp; + int s; + + s = splbio(); +loop: + for (vp = mp->mnt_vnodelist.lh_first; vp; vp = nvp) { + if (vp->v_mount != mp) /* Paranoia */ + goto loop; + nvp = vp->v_mntvnodes.le_next; + for (bp = vp->v_dirtyblkhd.lh_first; bp; bp = nbp) { + nbp = bp->b_vnbufs.le_next; + if ((bp->b_flags & (B_BUSY | B_DELWRI | B_NEEDCOMMIT)) + == (B_DELWRI | B_NEEDCOMMIT)) + bp->b_flags &= ~B_NEEDCOMMIT; + } + } + splx(s); +} + +/* + * Map errnos to NFS error numbers. For Version 3 also filter out error + * numbers not specified for the associated procedure. + */ +int +nfsrv_errmap(nd, err) + struct nfsrv_descript *nd; + register int err; +{ + register short *defaulterrp, *errp; + + if (nd->nd_flag & ND_NFSV3) { + if (nd->nd_procnum <= NFSPROC_COMMIT) { + errp = defaulterrp = nfsrv_v3errmap[nd->nd_procnum]; + while (*++errp) { + if (*errp == err) + return (err); + else if (*errp > err) + break; + } + return ((int)*defaulterrp); + } else + return (err & 0xffff); + } + if (err <= ELAST) + return ((int)nfsrv_v2errmap[err - 1]); + return (NFSERR_IO); +} + int nfsrv_vmio( struct vnode *vp) { vm_object_t object; diff --git a/sys/nfsclient/nfs_vfsops.c b/sys/nfsclient/nfs_vfsops.c index 2daa891..2c4070b 100644 --- a/sys/nfsclient/nfs_vfsops.c +++ b/sys/nfsclient/nfs_vfsops.c @@ -34,7 +34,7 @@ * SUCH DAMAGE. * * @(#)nfs_vfsops.c 8.3 (Berkeley) 1/4/94 - * $Id: nfs_vfsops.c,v 1.14.2.1 1995/06/02 11:13:15 davidg Exp $ + * $Id: nfs_vfsops.c,v 1.15 1995/06/11 19:31:46 rgrimes Exp $ */ #include <sys/param.h> @@ -59,10 +59,10 @@ #include <netinet/in.h> #include <nfs/rpcv2.h> -#include <nfs/nfsv2.h> +#include <nfs/nfsproto.h> #include <nfs/nfsnode.h> -#include <nfs/nfsmount.h> #include <nfs/nfs.h> +#include <nfs/nfsmount.h> #include <nfs/xdr_subs.h> #include <nfs/nfsm_subs.h> #include <nfs/nfsdiskless.h> @@ -71,11 +71,15 @@ struct nfsstats nfsstats; static int nfs_sysctl(int *, u_int, void *, size_t *, void *, size_t, struct proc *); +extern int nfs_ticks; /* * nfs vfs operations. */ struct vfsops nfs_vfsops = { +#ifdef __NetBSD__ + MOUNT_NFS, +#endif nfs_mount, nfs_start, nfs_unmount, @@ -87,9 +91,13 @@ struct vfsops nfs_vfsops = { nfs_fhtovp, nfs_vptofh, nfs_init, +#ifdef __FreeBSD__ nfs_sysctl +#endif }; +#ifdef __FreeBSD__ VFS_SET(nfs_vfsops, nfs, MOUNT_NFS, VFCF_NETWORK); +#endif /* * This structure must be filled in by a primary bootstrap or bootstrap @@ -99,16 +107,11 @@ VFS_SET(nfs_vfsops, nfs, MOUNT_NFS, VFCF_NETWORK); struct nfs_diskless nfs_diskless = { 0 }; int nfs_diskless_valid = 0; -extern u_long nfs_procids[NFS_NPROCS]; -extern u_long nfs_prog, nfs_vers; void nfs_disconnect __P((struct nfsmount *)); void nfsargs_ntoh __P((struct nfs_args *)); static struct mount *nfs_mountdiskless __P((char *, char *, int, struct sockaddr_in *, struct nfs_args *, register struct vnode **)); -#define TRUE 1 -#define FALSE 0 - /* * nfs statfs call */ @@ -119,40 +122,65 @@ nfs_statfs(mp, sbp, p) struct proc *p; { register struct vnode *vp; - register struct nfsv2_statfs *sfp; + register struct nfs_statfs *sfp; register caddr_t cp; - register long t1; + register u_long *tl; + register long t1, t2; caddr_t bpos, dpos, cp2; - int error = 0, isnq; + struct nfsmount *nmp = VFSTONFS(mp); + int error = 0, v3 = (nmp->nm_flag & NFSMNT_NFSV3), retattr; struct mbuf *mreq, *mrep, *md, *mb, *mb2; - struct nfsmount *nmp; struct ucred *cred; struct nfsnode *np; + u_quad_t tquad; - nmp = VFSTONFS(mp); - isnq = (nmp->nm_flag & NFSMNT_NQNFS); - error = nfs_nget(mp, &nmp->nm_fh, &np); +#ifndef nolint + sfp = (struct nfs_statfs *)0; +#endif + error = nfs_nget(mp, (nfsfh_t *)nmp->nm_fh, nmp->nm_fhsize, &np); if (error) return (error); vp = NFSTOV(np); - nfsstats.rpccnt[NFSPROC_STATFS]++; cred = crget(); cred->cr_ngroups = 1; - nfsm_reqhead(vp, NFSPROC_STATFS, NFSX_FH); - nfsm_fhtom(vp); - nfsm_request(vp, NFSPROC_STATFS, p, cred); - nfsm_dissect(sfp, struct nfsv2_statfs *, NFSX_STATFS(isnq)); + if (v3 && (nmp->nm_flag & NFSMNT_GOTFSINFO) == 0) + (void)nfs_fsinfo(nmp, vp, cred, p); + nfsstats.rpccnt[NFSPROC_FSSTAT]++; + nfsm_reqhead(vp, NFSPROC_FSSTAT, NFSX_FH(v3)); + nfsm_fhtom(vp, v3); + nfsm_request(vp, NFSPROC_FSSTAT, p, cred); + if (v3) + nfsm_postop_attr(vp, retattr); + if (!error) + nfsm_dissect(sfp, struct nfs_statfs *, NFSX_STATFS(v3)); +#ifdef __NetBSD__ +#ifdef COMPAT_09 + sbp->f_type = 2; +#else + sbp->f_type = 0; +#endif +#else sbp->f_type = MOUNT_NFS; +#endif sbp->f_flags = nmp->nm_flag; - sbp->f_iosize = NFS_MAXDGRAMDATA; - sbp->f_bsize = fxdr_unsigned(long, sfp->sf_bsize); - sbp->f_blocks = fxdr_unsigned(long, sfp->sf_blocks); - sbp->f_bfree = fxdr_unsigned(long, sfp->sf_bfree); - sbp->f_bavail = fxdr_unsigned(long, sfp->sf_bavail); - if (isnq) { - sbp->f_files = fxdr_unsigned(long, sfp->sf_files); - sbp->f_ffree = fxdr_unsigned(long, sfp->sf_ffree); + sbp->f_iosize = min(nmp->nm_rsize, nmp->nm_wsize); + if (v3) { + sbp->f_bsize = NFS_FABLKSIZE; + fxdr_hyper(&sfp->sf_tbytes, &tquad); + sbp->f_blocks = (long)(tquad / ((u_quad_t)NFS_FABLKSIZE)); + fxdr_hyper(&sfp->sf_fbytes, &tquad); + sbp->f_bfree = (long)(tquad / ((u_quad_t)NFS_FABLKSIZE)); + fxdr_hyper(&sfp->sf_abytes, &tquad); + sbp->f_bavail = (long)(tquad / ((u_quad_t)NFS_FABLKSIZE)); + sbp->f_files = (fxdr_unsigned(long, sfp->sf_tfiles.nfsuquad[1]) + & 0x7fffffff); + sbp->f_ffree = (fxdr_unsigned(long, sfp->sf_ffiles.nfsuquad[1]) + & 0x7fffffff); } else { + sbp->f_bsize = fxdr_unsigned(long, sfp->sf_bsize); + sbp->f_blocks = fxdr_unsigned(long, sfp->sf_blocks); + sbp->f_bfree = fxdr_unsigned(long, sfp->sf_bfree); + sbp->f_bavail = fxdr_unsigned(long, sfp->sf_bavail); sbp->f_files = 0; sbp->f_ffree = 0; } @@ -161,12 +189,72 @@ nfs_statfs(mp, sbp, p) bcopy(mp->mnt_stat.f_mntfromname, sbp->f_mntfromname, MNAMELEN); } nfsm_reqdone; - vrele(vp); + vput(vp); crfree(cred); return (error); } /* + * nfs version 3 fsinfo rpc call + */ +int +nfs_fsinfo(nmp, vp, cred, p) + register struct nfsmount *nmp; + register struct vnode *vp; + struct ucred *cred; + struct proc *p; +{ + register struct nfsv3_fsinfo *fsp; + register caddr_t cp; + register long t1, t2; + register u_long *tl, pref, max; + caddr_t bpos, dpos, cp2; + int error = 0, retattr; + struct mbuf *mreq, *mrep, *md, *mb, *mb2; + + nfsstats.rpccnt[NFSPROC_FSINFO]++; + nfsm_reqhead(vp, NFSPROC_FSINFO, NFSX_FH(1)); + nfsm_fhtom(vp, 1); + nfsm_request(vp, NFSPROC_FSINFO, p, cred); + nfsm_postop_attr(vp, retattr); + if (!error) { + nfsm_dissect(fsp, struct nfsv3_fsinfo *, NFSX_V3FSINFO); + pref = fxdr_unsigned(u_long, fsp->fs_wtpref); + if (pref < nmp->nm_wsize) + nmp->nm_wsize = (pref + NFS_FABLKSIZE - 1) & + ~(NFS_FABLKSIZE - 1); + max = fxdr_unsigned(u_long, fsp->fs_wtmax); + if (max < nmp->nm_wsize) { + nmp->nm_wsize = max & ~(NFS_FABLKSIZE - 1); + if (nmp->nm_wsize == 0) + nmp->nm_wsize = max; + } + pref = fxdr_unsigned(u_long, fsp->fs_rtpref); + if (pref < nmp->nm_rsize) + nmp->nm_rsize = (pref + NFS_FABLKSIZE - 1) & + ~(NFS_FABLKSIZE - 1); + max = fxdr_unsigned(u_long, fsp->fs_rtmax); + if (max < nmp->nm_rsize) { + nmp->nm_rsize = max & ~(NFS_FABLKSIZE - 1); + if (nmp->nm_rsize == 0) + nmp->nm_rsize = max; + } + pref = fxdr_unsigned(u_long, fsp->fs_dtpref); + if (pref < nmp->nm_readdirsize) + nmp->nm_readdirsize = (pref + NFS_DIRBLKSIZ - 1) & + ~(NFS_DIRBLKSIZ - 1); + if (max < nmp->nm_readdirsize) { + nmp->nm_readdirsize = max & ~(NFS_DIRBLKSIZ - 1); + if (nmp->nm_readdirsize == 0) + nmp->nm_readdirsize = max; + } + nmp->nm_flag |= NFSMNT_GOTFSINFO; + } + nfsm_reqdone; + return (error); +} + +/* * Mount a remote root fs via. nfs. This depends on the info in the * nfs_diskless structure that has been filled in properly by some primary * bootstrap. @@ -268,7 +356,12 @@ nfs_mountroot() * Create a fake mount point just for the swap vnode so that the * swap file can be on a different server from the rootfs. */ - nd->swap_args.fh = (nfsv2fh_t *)nd->swap_fh; + nd->swap_args.fh = nd->swap_fh; + /* + * If using nfsv3_diskless, replace NFSX_V2FH with + * nd->swap_fhsize. + */ + nd->swap_args.fhsize = NFSX_V2FH; l = ntohl(nd->swap_saddr.sin_addr.s_addr); sprintf(buf,"%ld.%ld.%ld.%ld:%s", (l >> 24) & 0xff, (l >> 16) & 0xff, @@ -293,8 +386,12 @@ nfs_mountroot() /* * Create the rootfs mount point. */ - nd->root_args.fh = (nfsv2fh_t *)nd->root_fh; - l = ntohl(nd->root_saddr.sin_addr.s_addr); + nd->root_args.fh = nd->root_fh; + /* + * If using nfsv3_diskless, replace NFSX_V2FH with nd->root_fhsize. + */ + nd->root_args.fhsize = NFSX_V2FH; + l = ntohl(nd->swap_saddr.sin_addr.s_addr); sprintf(buf,"%ld.%ld.%ld.%ld:%s", (l >> 24) & 0xff, (l >> 16) & 0xff, (l >> 8) & 0xff, (l >> 0) & 0xff,nd->root_hostnam); @@ -304,7 +401,11 @@ nfs_mountroot() if (vfs_lock(mp)) panic("nfs_mountroot: vfs_lock"); +#if NetBSD >= 1994101 + CIRCLEQ_INSERT_TAIL(&mountlist, mp, mnt_list); +#else TAILQ_INSERT_TAIL(&mountlist, mp, mnt_list); +#endif mp->mnt_flag |= MNT_ROOTFS; mp->mnt_vnodecovered = NULLVP; vfs_unlock(mp); @@ -386,12 +487,12 @@ nfs_mount(mp, path, data, ndp, p) struct vnode *vp; char pth[MNAMELEN], hst[MNAMELEN]; u_int len; - nfsv2fh_t nfh; + u_char nfh[NFSX_V3FHMAX]; error = copyin(data, (caddr_t)&args, sizeof (struct nfs_args)); if (error) return (error); - error = copyin((caddr_t)args.fh, (caddr_t)&nfh, sizeof (nfsv2fh_t)); + error = copyin((caddr_t)args.fh, (caddr_t)nfh, args.fhsize); if (error) return (error); error = copyinstr(path, pth, MNAMELEN-1, &len); @@ -406,7 +507,7 @@ nfs_mount(mp, path, data, ndp, p) error = sockargs(&nam, (caddr_t)args.addr, args.addrlen, MT_SONAME); if (error) return (error); - args.fh = &nfh; + args.fh = nfh; error = mountnfs(&args, mp, nam, pth, hst, &vp); return (error); } @@ -424,7 +525,7 @@ mountnfs(argp, mp, nam, pth, hst, vpp) { register struct nfsmount *nmp; struct nfsnode *np; - int error; + int error, maxio; if (mp->mnt_flag & MNT_UPDATE) { nmp = VFSTONFS(mp); @@ -435,16 +536,12 @@ mountnfs(argp, mp, nam, pth, hst, vpp) MALLOC(nmp, struct nfsmount *, sizeof (struct nfsmount), M_NFSMNT, M_WAITOK); bzero((caddr_t)nmp, sizeof (struct nfsmount)); + TAILQ_INIT(&nmp->nm_uidlruhead); mp->mnt_data = (qaddr_t)nmp; } getnewfsid(mp, MOUNT_NFS); nmp->nm_mountp = mp; nmp->nm_flag = argp->flags; - if ((nmp->nm_flag & (NFSMNT_NQNFS | NFSMNT_MYWRITE)) == - (NFSMNT_NQNFS | NFSMNT_MYWRITE)) { - error = EPERM; - goto bad; - } if (nmp->nm_flag & NFSMNT_NQNFS) /* * We have to set mnt_maxsymlink to a non-zero value so @@ -457,14 +554,24 @@ mountnfs(argp, mp, nam, pth, hst, vpp) nmp->nm_retry = NFS_RETRANS; nmp->nm_wsize = NFS_WSIZE; nmp->nm_rsize = NFS_RSIZE; + nmp->nm_readdirsize = NFS_READDIRSIZE; nmp->nm_numgrps = NFS_MAXGRPS; nmp->nm_readahead = NFS_DEFRAHEAD; nmp->nm_leaseterm = NQ_DEFLEASE; nmp->nm_deadthresh = NQ_DEADTHRESH; CIRCLEQ_INIT(&nmp->nm_timerhead); nmp->nm_inprog = NULLVP; - bcopy((caddr_t)argp->fh, (caddr_t)&nmp->nm_fh, sizeof(nfsv2fh_t)); + nmp->nm_fhsize = argp->fhsize; + bcopy((caddr_t)argp->fh, (caddr_t)nmp->nm_fh, argp->fhsize); +#ifdef __NetBSD__ +#ifdef COMPAT_09 + mp->mnt_stat.f_type = 2; +#else + mp->mnt_stat.f_type = 0; +#endif +#else mp->mnt_stat.f_type = MOUNT_NFS; +#endif bcopy(hst, mp->mnt_stat.f_mntfromname, MNAMELEN); bcopy(pth, mp->mnt_stat.f_mntonname, MNAMELEN); nmp->nm_nam = nam; @@ -483,29 +590,48 @@ mountnfs(argp, mp, nam, pth, hst, vpp) nmp->nm_retry = NFS_MAXREXMIT; } + if (argp->flags & NFSMNT_NFSV3) { + if (argp->sotype == SOCK_DGRAM) + maxio = NFS_MAXDGRAMDATA; + else + maxio = NFS_MAXDATA; + } else + maxio = NFS_V2MAXDATA; + if ((argp->flags & NFSMNT_WSIZE) && argp->wsize > 0) { nmp->nm_wsize = argp->wsize; /* Round down to multiple of blocksize */ - nmp->nm_wsize &= ~0x1ff; + nmp->nm_wsize &= ~(NFS_FABLKSIZE - 1); if (nmp->nm_wsize <= 0) - nmp->nm_wsize = 512; - else if (nmp->nm_wsize > NFS_MAXDATA) - nmp->nm_wsize = NFS_MAXDATA; + nmp->nm_wsize = NFS_FABLKSIZE; } + if (nmp->nm_wsize > maxio) + nmp->nm_wsize = maxio; if (nmp->nm_wsize > MAXBSIZE) nmp->nm_wsize = MAXBSIZE; if ((argp->flags & NFSMNT_RSIZE) && argp->rsize > 0) { nmp->nm_rsize = argp->rsize; /* Round down to multiple of blocksize */ - nmp->nm_rsize &= ~0x1ff; + nmp->nm_rsize &= ~(NFS_FABLKSIZE - 1); if (nmp->nm_rsize <= 0) - nmp->nm_rsize = 512; - else if (nmp->nm_rsize > NFS_MAXDATA) - nmp->nm_rsize = NFS_MAXDATA; + nmp->nm_rsize = NFS_FABLKSIZE; } + if (nmp->nm_rsize > maxio) + nmp->nm_rsize = maxio; if (nmp->nm_rsize > MAXBSIZE) nmp->nm_rsize = MAXBSIZE; + + if ((argp->flags & NFSMNT_READDIRSIZE) && argp->readdirsize > 0) { + nmp->nm_readdirsize = argp->readdirsize; + /* Round down to multiple of blocksize */ + nmp->nm_readdirsize &= ~(NFS_DIRBLKSIZ - 1); + if (nmp->nm_readdirsize < NFS_DIRBLKSIZ) + nmp->nm_readdirsize = NFS_DIRBLKSIZ; + } + if (nmp->nm_readdirsize > maxio) + nmp->nm_readdirsize = maxio; + if ((argp->flags & NFSMNT_MAXGRPS) && argp->maxgrouplist >= 0 && argp->maxgrouplist <= NFS_MAXGRPS) nmp->nm_numgrps = argp->maxgrouplist; @@ -536,7 +662,7 @@ mountnfs(argp, mp, nam, pth, hst, vpp) * stuck on a dead server and we are holding a lock on the mount * point. */ - mp->mnt_stat.f_iosize = NFS_MAXDGRAMDATA; + mp->mnt_stat.f_iosize = min(nmp->nm_rsize, nmp->nm_wsize); /* * A reference count is needed on the nfsnode representing the * remote root. If this object is not persistent, then backward @@ -545,11 +671,16 @@ mountnfs(argp, mp, nam, pth, hst, vpp) * this problem, because one can identify root inodes by their * number == ROOTINO (2). */ - error = nfs_nget(mp, &nmp->nm_fh, &np); + error = nfs_nget(mp, (nfsfh_t *)nmp->nm_fh, nmp->nm_fhsize, &np); if (error) goto bad; *vpp = NFSTOV(np); + /* + * Lose the lock but keep the ref. + */ + VOP_UNLOCK(*vpp); + return (0); bad: nfs_disconnect(nmp); @@ -592,7 +723,7 @@ nfs_unmount(mp, mntflags, p) * the remote root. See comment in mountnfs(). The VFS unmount() * has done vput on this vnode, otherwise we would get deadlock! */ - error = nfs_nget(mp, &nmp->nm_fh, &np); + error = nfs_nget(mp, (nfsfh_t *)nmp->nm_fh, nmp->nm_fhsize, &np); if (error) return(error); vp = NFSTOV(np); @@ -622,9 +753,9 @@ nfs_unmount(mp, mntflags, p) nmp->nm_flag |= NFSMNT_DISMNT; /* - * There are two reference counts to get rid of here. + * There are two reference counts and one lock to get rid of here. */ - vrele(vp); + vput(vp); vrele(vp); vgone(vp); nfs_disconnect(nmp); @@ -649,10 +780,11 @@ nfs_root(mp, vpp) int error; nmp = VFSTONFS(mp); - error = nfs_nget(mp, &nmp->nm_fh, &np); + error = nfs_nget(mp, (nfsfh_t *)nmp->nm_fh, nmp->nm_fhsize, &np); if (error) return (error); vp = NFSTOV(np); + VOP_UNLOCK(vp); vp->v_type = VDIR; vp->v_flag = VROOT; *vpp = vp; diff --git a/sys/nfsclient/nfs_vnops.c b/sys/nfsclient/nfs_vnops.c index 46484ea..769b0bb 100644 --- a/sys/nfsclient/nfs_vnops.c +++ b/sys/nfsclient/nfs_vnops.c @@ -34,17 +34,18 @@ * SUCH DAMAGE. * * @(#)nfs_vnops.c 8.5 (Berkeley) 2/13/94 - * $Id: nfs_vnops.c,v 1.14 1995/03/23 09:43:40 davidg Exp $ + * $Id: nfs_vnops.c,v 1.15 1995/05/30 08:12:49 rgrimes Exp $ */ /* - * vnode op calls for sun nfs version 2 + * vnode op calls for Sun NFS version 2 and 3 */ #include <sys/param.h> -#include <sys/proc.h> #include <sys/kernel.h> #include <sys/systm.h> +#include <sys/resourcevar.h> +#include <sys/proc.h> #include <sys/mount.h> #include <sys/buf.h> #include <sys/malloc.h> @@ -53,7 +54,9 @@ #include <sys/namei.h> #include <sys/vnode.h> #include <sys/dirent.h> +#include <sys/fcntl.h> #include <sys/lockf.h> +#include <ufs/ufs/dir.h> #include <vm/vm.h> @@ -61,7 +64,7 @@ #include <miscfs/fifofs/fifo.h> #include <nfs/rpcv2.h> -#include <nfs/nfsv2.h> +#include <nfs/nfsproto.h> #include <nfs/nfs.h> #include <nfs/nfsnode.h> #include <nfs/nfsmount.h> @@ -69,11 +72,24 @@ #include <nfs/nfsm_subs.h> #include <nfs/nqnfs.h> +#include <net/if.h> +#include <netinet/in.h> +#include <netinet/in_var.h> + /* Defs */ #define TRUE 1 #define FALSE 0 /* + * Ifdef for FreeBSD-current merged buffer cache. It is unfortunate that these + * calls are not in getblk() and brelse() so that they would not be necessary + * here. + */ +#ifndef B_VMIO +#define vfs_busy_pages(bp, f) +#endif + +/* * Global vfs data structures for nfs */ int (**nfsv2_vnodeop_p)(); @@ -89,8 +105,14 @@ struct vnodeopv_entry_desc nfsv2_vnodeop_entries[] = { { &vop_setattr_desc, nfs_setattr }, /* setattr */ { &vop_read_desc, nfs_read }, /* read */ { &vop_write_desc, nfs_write }, /* write */ +#ifdef HAS_VOPLEASE + { &vop_lease_desc, nfs_lease_check }, /* lease */ +#endif { &vop_ioctl_desc, nfs_ioctl }, /* ioctl */ { &vop_select_desc, nfs_select }, /* select */ +#ifdef HAS_VOPREVOKE + { &vop_revoke_desc, nfs_revoke }, /* revoke */ +#endif { &vop_mmap_desc, nfs_mmap }, /* mmap */ { &vop_fsync_desc, nfs_fsync }, /* fsync */ { &vop_seek_desc, nfs_seek }, /* seek */ @@ -119,12 +141,14 @@ struct vnodeopv_entry_desc nfsv2_vnodeop_entries[] = { { &vop_vfree_desc, nfs_vfree }, /* vfree */ { &vop_truncate_desc, nfs_truncate }, /* truncate */ { &vop_update_desc, nfs_update }, /* update */ - { &vop_bwrite_desc, vn_bwrite }, + { &vop_bwrite_desc, nfs_bwrite }, { (struct vnodeop_desc*)NULL, (int(*)())NULL } }; struct vnodeopv_desc nfsv2_vnodeop_opv_desc = { &nfsv2_vnodeop_p, nfsv2_vnodeop_entries }; +#ifdef __FreeBSD__ VNODEOP_SET(nfsv2_vnodeop_opv_desc); +#endif /* * Special device vnode ops @@ -142,8 +166,14 @@ struct vnodeopv_entry_desc spec_nfsv2nodeop_entries[] = { { &vop_setattr_desc, nfs_setattr }, /* setattr */ { &vop_read_desc, nfsspec_read }, /* read */ { &vop_write_desc, nfsspec_write }, /* write */ +#ifdef HAS_VOPLEASE + { &vop_lease_desc, spec_lease_check }, /* lease */ +#endif { &vop_ioctl_desc, spec_ioctl }, /* ioctl */ { &vop_select_desc, spec_select }, /* select */ +#ifdef HAS_VOPREVOKE + { &vop_revoke_desc, spec_revoke }, /* revoke */ +#endif { &vop_mmap_desc, spec_mmap }, /* mmap */ { &vop_fsync_desc, nfs_fsync }, /* fsync */ { &vop_seek_desc, spec_seek }, /* seek */ @@ -177,7 +207,9 @@ struct vnodeopv_entry_desc spec_nfsv2nodeop_entries[] = { }; struct vnodeopv_desc spec_nfsv2nodeop_opv_desc = { &spec_nfsv2nodeop_p, spec_nfsv2nodeop_entries }; +#ifdef __FreeBSD__ VNODEOP_SET(spec_nfsv2nodeop_opv_desc); +#endif int (**fifo_nfsv2nodeop_p)(); struct vnodeopv_entry_desc fifo_nfsv2nodeop_entries[] = { @@ -192,8 +224,14 @@ struct vnodeopv_entry_desc fifo_nfsv2nodeop_entries[] = { { &vop_setattr_desc, nfs_setattr }, /* setattr */ { &vop_read_desc, nfsfifo_read }, /* read */ { &vop_write_desc, nfsfifo_write }, /* write */ +#ifdef HAS_VOPLEASE + { &vop_lease_desc, fifo_lease_check }, /* lease */ +#endif { &vop_ioctl_desc, fifo_ioctl }, /* ioctl */ { &vop_select_desc, fifo_select }, /* select */ +#ifdef HAS_VOPREVOKE + { &vop_revoke_desc, fifo_revoke }, /* revoke */ +#endif { &vop_mmap_desc, fifo_mmap }, /* mmap */ { &vop_fsync_desc, nfs_fsync }, /* fsync */ { &vop_seek_desc, fifo_seek }, /* seek */ @@ -227,15 +265,21 @@ struct vnodeopv_entry_desc fifo_nfsv2nodeop_entries[] = { }; struct vnodeopv_desc fifo_nfsv2nodeop_opv_desc = { &fifo_nfsv2nodeop_p, fifo_nfsv2nodeop_entries }; +#ifdef __FreeBSD__ VNODEOP_SET(fifo_nfsv2nodeop_opv_desc); +#endif void nqnfs_clientlease(); +int nfs_commit(); +int nfs_removerpc(); +int nfs_renamerpc(); /* * Global variables */ -extern u_long nfs_procids[NFS_NPROCS]; -extern u_long nfs_prog, nfs_vers, nfs_true, nfs_false; +extern u_long nfs_true, nfs_false; +extern struct nfsstats nfsstats; +extern nfstype nfsv3_type[9]; struct proc *nfs_iodwant[NFS_MAXASYNCDAEMON]; int nfs_numasync = 0; #define DIRHDSIZ (sizeof (struct dirent) - (MAXNAMLEN + 1)) @@ -261,9 +305,9 @@ nfs_null(vp, cred, procp) /* * nfs access vnode op. - * For nfs, just return ok. File accesses may fail later. - * For nqnfs, use the access rpc to check accessibility. If file modes are - * changed on the server, accesses might still fail later. + * For nfs version 2, just return ok. File accesses may fail later. + * For nfs version 3, use the access rpc to check accessibility. If file modes + * are changed on the server, accesses might still fail later. */ int nfs_access(ap) @@ -277,36 +321,56 @@ nfs_access(ap) register struct vnode *vp = ap->a_vp; register u_long *tl; register caddr_t cp; - caddr_t bpos, dpos; - int error = 0; + register int t1, t2; + caddr_t bpos, dpos, cp2; + int error = 0, attrflag; struct mbuf *mreq, *mrep, *md, *mb, *mb2; + u_long mode, rmode; + int v3 = NFS_ISV3(vp); /* - * For nqnfs, do an access rpc, otherwise you are stuck emulating + * For nfs v3, do an access rpc, otherwise you are stuck emulating * ufs_access() locally using the vattr. This may not be correct, * since the server may apply other access criteria such as * client uid-->server uid mapping that we do not know about, but * this is better than just returning anything that is lying about * in the cache. */ - if (VFSTONFS(vp->v_mount)->nm_flag & NFSMNT_NQNFS) { - nfsstats.rpccnt[NQNFSPROC_ACCESS]++; - nfsm_reqhead(vp, NQNFSPROC_ACCESS, NFSX_FH + 3 * NFSX_UNSIGNED); - nfsm_fhtom(vp); - nfsm_build(tl, u_long *, 3 * NFSX_UNSIGNED); + if (v3) { + nfsstats.rpccnt[NFSPROC_ACCESS]++; + nfsm_reqhead(vp, NFSPROC_ACCESS, NFSX_FH(v3) + NFSX_UNSIGNED); + nfsm_fhtom(vp, v3); + nfsm_build(tl, u_long *, NFSX_UNSIGNED); if (ap->a_mode & VREAD) - *tl++ = nfs_true; - else - *tl++ = nfs_false; - if (ap->a_mode & VWRITE) - *tl++ = nfs_true; - else - *tl++ = nfs_false; - if (ap->a_mode & VEXEC) - *tl = nfs_true; + mode = NFSV3ACCESS_READ; else - *tl = nfs_false; - nfsm_request(vp, NQNFSPROC_ACCESS, ap->a_p, ap->a_cred); + mode = 0; + if (vp->v_type == VDIR) { + if (ap->a_mode & VWRITE) + mode |= (NFSV3ACCESS_MODIFY | NFSV3ACCESS_EXTEND | + NFSV3ACCESS_DELETE); + if (ap->a_mode & VEXEC) + mode |= NFSV3ACCESS_LOOKUP; + } else { + if (ap->a_mode & VWRITE) + mode |= (NFSV3ACCESS_MODIFY | NFSV3ACCESS_EXTEND); + if (ap->a_mode & VEXEC) + mode |= NFSV3ACCESS_EXECUTE; + } + *tl = txdr_unsigned(mode); + nfsm_request(vp, NFSPROC_ACCESS, ap->a_p, ap->a_cred); + nfsm_postop_attr(vp, attrflag); + if (!error) { + nfsm_dissect(tl, u_long *, NFSX_UNSIGNED); + rmode = fxdr_unsigned(u_long, *tl); + /* + * The NFS V3 spec does not clarify whether or not + * the returned access bits can be a superset of + * the ones requested, so... + */ + if ((rmode & mode) != mode) + error = EACCES; + } nfsm_reqdone; return (error); } else @@ -337,14 +401,17 @@ nfs_open(ap) int error; if (vp->v_type != VREG && vp->v_type != VDIR && vp->v_type != VLNK) +{ printf("open eacces vtyp=%d\n",vp->v_type); return (EACCES); +} /* * Get a valid lease. If cached data is stale, flush it. */ if (nmp->nm_flag & NFSMNT_NQNFS) { - if (NQNFS_CKINVALID(vp, np, NQL_READ)) { + if (NQNFS_CKINVALID(vp, np, ND_READ)) { do { - error = nqnfs_getlease(vp, NQL_READ, ap->a_cred, ap->a_p); + error = nqnfs_getlease(vp, ND_READ, ap->a_cred, + ap->a_p); } while (error == NQNFS_EXPIRED); if (error) return (error); @@ -353,6 +420,7 @@ nfs_open(ap) if ((error = nfs_vinvalbuf(vp, V_SAVE, ap->a_cred, ap->a_p, 1)) == EINTR) return (error); + (void) vnode_pager_uncache(vp); np->n_brev = np->n_lrev; } } @@ -361,8 +429,10 @@ nfs_open(ap) if ((error = nfs_vinvalbuf(vp, V_SAVE, ap->a_cred, ap->a_p, 1)) == EINTR) return (error); + /* (void) vnode_pager_uncache(vp); */ np->n_attrstamp = 0; - np->n_direofoffset = 0; + if (vp->v_type == VDIR) + np->n_direofoffset = 0; error = VOP_GETATTR(vp, &vattr, ap->a_cred, ap->a_p); if (error) return (error); @@ -372,10 +442,12 @@ nfs_open(ap) if (error) return (error); if (np->n_mtime != vattr.va_mtime.ts_sec) { - np->n_direofoffset = 0; + if (vp->v_type == VDIR) + np->n_direofoffset = 0; if ((error = nfs_vinvalbuf(vp, V_SAVE, ap->a_cred, ap->a_p, 1)) == EINTR) return (error); + /* (void) vnode_pager_uncache(vp); */ np->n_mtime = vattr.va_mtime.ts_sec; } } @@ -387,7 +459,33 @@ nfs_open(ap) /* * nfs close vnode op - * For reg files, invalidate any buffer cache entries. + * What an NFS client should do upon close after writing is a debatable issue. + * Most NFS clients push delayed writes to the server upon close, basically for + * two reasons: + * 1 - So that any write errors may be reported back to the client process + * doing the close system call. By far the two most likely errors are + * NFSERR_NOSPC and NFSERR_DQUOT to indicate space allocation failure. + * 2 - To put a worst case upper bound on cache inconsistency between + * multiple clients for the file. + * There is also a consistency problem for Version 2 of the protocol w.r.t. + * not being able to tell if other clients are writing a file concurrently, + * since there is no way of knowing if the changed modify time in the reply + * is only due to the write for this client. + * (NFS Version 3 provides weak cache consistency data in the reply that + * should be sufficient to detect and handle this case.) + * + * The current code does the following: + * for NFS Version 2 - play it safe and flush/invalidate all dirty buffers + * for NFS Version 3 - flush dirty buffers to the server but don't invalidate + * or commit them (this satisfies 1 and 2 except for the + * case where the server crashes after this close but + * before the commit RPC, which is felt to be "good + * enough". Changing the last argument to nfs_flush() to + * a 1 would force a commit operation, if it is felt a + * commit is necessary now. + * for NQNFS - do nothing now, since 2 is dealt with via leases and + * 1 should be dealt with via an fsync() system call for + * cases where write errors are important. */ /* ARGSUSED */ int @@ -407,7 +505,10 @@ nfs_close(ap) if (vp->v_type == VREG) { if ((VFSTONFS(vp->v_mount)->nm_flag & NFSMNT_NQNFS) == 0 && (np->n_flag & NMODIFIED)) { - error = nfs_vinvalbuf(vp, V_SAVE, ap->a_cred, ap->a_p, 1); + if (NFS_ISV3(vp)) + error = nfs_flush(vp, ap->a_cred, MNT_WAIT, ap->a_p, 0); + else + error = nfs_vinvalbuf(vp, V_SAVE, ap->a_cred, ap->a_p, 1); np->n_attrstamp = 0; } if (np->n_flag & NWRITEERR) { @@ -433,9 +534,12 @@ nfs_getattr(ap) register struct vnode *vp = ap->a_vp; register struct nfsnode *np = VTONFS(vp); register caddr_t cp; + register u_long *tl; + register int t1, t2; caddr_t bpos, dpos; int error = 0; struct mbuf *mreq, *mrep, *md, *mb, *mb2; + int v3 = NFS_ISV3(vp); /* * Update local times for special files. @@ -448,10 +552,11 @@ nfs_getattr(ap) if (nfs_getattrcache(vp, ap->a_vap) == 0) return (0); nfsstats.rpccnt[NFSPROC_GETATTR]++; - nfsm_reqhead(vp, NFSPROC_GETATTR, NFSX_FH); - nfsm_fhtom(vp); + nfsm_reqhead(vp, NFSPROC_GETATTR, NFSX_FH(v3)); + nfsm_fhtom(vp, v3); nfsm_request(vp, NFSPROC_GETATTR, ap->a_p, ap->a_cred); - nfsm_loadattr(vp, ap->a_vap); + if (!error) + nfsm_loadattr(vp, ap->a_vap); nfsm_reqdone; return (error); } @@ -469,92 +574,166 @@ nfs_setattr(ap) struct proc *a_p; } */ *ap; { - register struct nfsv2_sattr *sp; - register caddr_t cp; - register long t1; - caddr_t bpos, dpos, cp2; - u_long *tl; - int error = 0, isnq; - struct mbuf *mreq, *mrep, *md, *mb, *mb2; register struct vnode *vp = ap->a_vp; register struct nfsnode *np = VTONFS(vp); register struct vattr *vap = ap->a_vap; - u_quad_t frev, tsize = 0; + int error = 0; + u_quad_t tsize; +#ifndef nolint + tsize = (u_quad_t)0; +#endif if (vap->va_size != VNOVAL) { - switch (vp->v_type) { - case VDIR: - return (EISDIR); - case VCHR: - case VBLK: + switch (vp->v_type) { + case VDIR: + return (EISDIR); + case VCHR: + case VBLK: if (vap->va_mtime.ts_sec == VNOVAL && vap->va_atime.ts_sec == VNOVAL && vap->va_mode == (u_short)VNOVAL && - vap->va_uid == VNOVAL && - vap->va_gid == VNOVAL) + vap->va_uid == (uid_t)VNOVAL && + vap->va_gid == (gid_t)VNOVAL) return (0); - vap->va_size = VNOVAL; - break; - default: - if (np->n_flag & NMODIFIED) { - error = nfs_vinvalbuf(vp, - vap->va_size ? V_SAVE : 0, - ap->a_cred, ap->a_p, 1); - if (error) - return (error); - } - tsize = np->n_size; - np->n_size = np->n_vattr.va_size = vap->va_size; - vnode_pager_setsize(vp, (u_long)np->n_size); - } - } else if ((vap->va_mtime.ts_sec != VNOVAL || - vap->va_atime.ts_sec != VNOVAL) && (np->n_flag & NMODIFIED)) { - error = nfs_vinvalbuf(vp, V_SAVE, ap->a_cred, ap->a_p, 1); - if (error == EINTR) - return (error); + vap->va_size = VNOVAL; + break; + default: + if (np->n_flag & NMODIFIED) { + if (vap->va_size == 0) + error = nfs_vinvalbuf(vp, 0, + ap->a_cred, ap->a_p, 1); + else + error = nfs_vinvalbuf(vp, V_SAVE, + ap->a_cred, ap->a_p, 1); + if (error) + return (error); + } + tsize = np->n_size; + np->n_size = np->n_vattr.va_size = vap->va_size; + vnode_pager_setsize(vp, (u_long)np->n_size); + }; + } else if ((vap->va_mtime.ts_sec != VNOVAL || + vap->va_atime.ts_sec != VNOVAL) && (np->n_flag & NMODIFIED) && + vp->v_type == VREG && + (error = nfs_vinvalbuf(vp, V_SAVE, ap->a_cred, + ap->a_p, 1)) == EINTR) + return (error); + error = nfs_setattrrpc(vp, vap, ap->a_cred, ap->a_p); + if (error) { + np->n_size = np->n_vattr.va_size = tsize; + vnode_pager_setsize(vp, (u_long)np->n_size); } + return (error); +} + +/* + * Do an nfs setattr rpc. + */ +int +nfs_setattrrpc(vp, vap, cred, procp) + register struct vnode *vp; + register struct vattr *vap; + struct ucred *cred; + struct proc *procp; +{ + register struct nfsv2_sattr *sp; + register caddr_t cp; + register long t1, t2; + caddr_t bpos, dpos, cp2; + u_long *tl; + int error = 0, wccflag = NFSV3_WCCRATTR; + struct mbuf *mreq, *mrep, *md, *mb, *mb2; + u_quad_t frev; + int v3 = NFS_ISV3(vp); + nfsstats.rpccnt[NFSPROC_SETATTR]++; - isnq = (VFSTONFS(vp->v_mount)->nm_flag & NFSMNT_NQNFS); - nfsm_reqhead(vp, NFSPROC_SETATTR, NFSX_FH+NFSX_SATTR(isnq)); - nfsm_fhtom(vp); - nfsm_build(sp, struct nfsv2_sattr *, NFSX_SATTR(isnq)); - if (vap->va_mode == (u_short)-1) - sp->sa_mode = VNOVAL; - else - sp->sa_mode = vtonfs_mode(vp->v_type, vap->va_mode); - if (vap->va_uid == (uid_t)-1) - sp->sa_uid = VNOVAL; - else - sp->sa_uid = txdr_unsigned(vap->va_uid); - if (vap->va_gid == (gid_t)-1) - sp->sa_gid = VNOVAL; - else - sp->sa_gid = txdr_unsigned(vap->va_gid); - if (isnq) { - txdr_hyper(&vap->va_size, &sp->sa_nqsize); - txdr_nqtime(&vap->va_atime, &sp->sa_nqatime); - txdr_nqtime(&vap->va_mtime, &sp->sa_nqmtime); - sp->sa_nqflags = txdr_unsigned(vap->va_flags); - sp->sa_nqrdev = VNOVAL; + nfsm_reqhead(vp, NFSPROC_SETATTR, NFSX_FH(v3) + NFSX_SATTR(v3)); + nfsm_fhtom(vp, v3); + if (v3) { + if (vap->va_mode != (u_short)VNOVAL) { + nfsm_build(tl, u_long *, 2 * NFSX_UNSIGNED); + *tl++ = nfs_true; + *tl = txdr_unsigned(vap->va_mode); + } else { + nfsm_build(tl, u_long *, NFSX_UNSIGNED); + *tl = nfs_false; + } + if (vap->va_uid != (uid_t)VNOVAL) { + nfsm_build(tl, u_long *, 2 * NFSX_UNSIGNED); + *tl++ = nfs_true; + *tl = txdr_unsigned(vap->va_uid); + } else { + nfsm_build(tl, u_long *, NFSX_UNSIGNED); + *tl = nfs_false; + } + if (vap->va_gid != (gid_t)VNOVAL) { + nfsm_build(tl, u_long *, 2 * NFSX_UNSIGNED); + *tl++ = nfs_true; + *tl = txdr_unsigned(vap->va_gid); + } else { + nfsm_build(tl, u_long *, NFSX_UNSIGNED); + *tl = nfs_false; + } + if (vap->va_size != VNOVAL) { + nfsm_build(tl, u_long *, 3 * NFSX_UNSIGNED); + *tl++ = nfs_true; + txdr_hyper(&vap->va_size, tl); + } else { + nfsm_build(tl, u_long *, NFSX_UNSIGNED); + *tl = nfs_false; + } + if (vap->va_atime.ts_sec != VNOVAL) { + if (vap->va_atime.ts_sec != time.tv_sec) { + nfsm_build(tl, u_long *, 3 * NFSX_UNSIGNED); + *tl++ = txdr_unsigned(NFSV3SATTRTIME_TOCLIENT); + txdr_nfsv3time(&vap->va_atime, tl); + } else { + nfsm_build(tl, u_long *, NFSX_UNSIGNED); + *tl = txdr_unsigned(NFSV3SATTRTIME_TOSERVER); + } + } else { + nfsm_build(tl, u_long *, NFSX_UNSIGNED); + *tl = txdr_unsigned(NFSV3SATTRTIME_DONTCHANGE); + } + if (vap->va_mtime.ts_sec != VNOVAL) { + if (vap->va_mtime.ts_sec != time.tv_sec) { + nfsm_build(tl, u_long *, 3 * NFSX_UNSIGNED); + *tl++ = txdr_unsigned(NFSV3SATTRTIME_TOCLIENT); + txdr_nfsv3time(&vap->va_atime, tl); + } else { + nfsm_build(tl, u_long *, NFSX_UNSIGNED); + *tl = txdr_unsigned(NFSV3SATTRTIME_TOSERVER); + } + } else { + nfsm_build(tl, u_long *, NFSX_UNSIGNED); + *tl = txdr_unsigned(NFSV3SATTRTIME_DONTCHANGE); + } + nfsm_build(tl, u_long *, NFSX_UNSIGNED); + *tl = nfs_false; } else { - sp->sa_nfssize = txdr_unsigned(vap->va_size); - txdr_nfstime(&vap->va_atime, &sp->sa_nfsatime); - txdr_nfstime(&vap->va_mtime, &sp->sa_nfsmtime); - } - nfsm_request(vp, NFSPROC_SETATTR, ap->a_p, ap->a_cred); - nfsm_loadattr(vp, (struct vattr *)0); - if ((VFSTONFS(vp->v_mount)->nm_flag & NFSMNT_NQNFS) && - NQNFS_CKCACHABLE(vp, NQL_WRITE)) { - nfsm_dissect(tl, u_long *, 2*NFSX_UNSIGNED); - fxdr_hyper(tl, &frev); - if (frev > np->n_brev) - np->n_brev = frev; + nfsm_build(sp, struct nfsv2_sattr *, NFSX_V2SATTR); + if (vap->va_mode == (u_short)VNOVAL) + sp->sa_mode = VNOVAL; + else + sp->sa_mode = vtonfsv2_mode(vp->v_type, vap->va_mode); + if (vap->va_uid == (uid_t)VNOVAL) + sp->sa_uid = VNOVAL; + else + sp->sa_uid = txdr_unsigned(vap->va_uid); + if (vap->va_gid == (gid_t)VNOVAL) + sp->sa_gid = VNOVAL; + else + sp->sa_gid = txdr_unsigned(vap->va_gid); + sp->sa_size = txdr_unsigned(vap->va_size); + txdr_nfsv2time(&vap->va_atime, &sp->sa_atime); + txdr_nfsv2time(&vap->va_mtime, &sp->sa_mtime); } + nfsm_request(vp, NFSPROC_SETATTR, procp, cred); + if (v3) { + nfsm_wcc_data(vp, wccflag); + } else + nfsm_loadattr(vp, (struct vattr *)0); nfsm_reqdone; - if (error) { - np->n_size = np->n_vattr.va_size = tsize; - vnode_pager_setsize(vp, (u_long)np->n_size); - } return (error); } @@ -576,23 +755,20 @@ nfs_lookup(ap) register struct vnode *dvp = ap->a_dvp; register struct vnode **vpp = ap->a_vpp; register int flags = cnp->cn_flags; - register struct vnode *vdp; + register struct vnode *newvp; register u_long *tl; register caddr_t cp; register long t1, t2; struct nfsmount *nmp; caddr_t bpos, dpos, cp2; - time_t reqtime = 0; struct mbuf *mreq, *mrep, *md, *mb, *mb2; - struct vnode *newvp; long len; - nfsv2fh_t *fhp; + nfsfh_t *fhp; struct nfsnode *np; - int lockparent, wantparent, error = 0; - int nqlflag = 0, cachable = 0; - u_quad_t frev; + int lockparent, wantparent, error = 0, attrflag, fhsize; + int v3 = NFS_ISV3(dvp); - *vpp = NULL; + *vpp = NULLVP; if (dvp->v_type != VDIR) return (ENOTDIR); lockparent = flags & LOCKPARENT; @@ -603,161 +779,141 @@ nfs_lookup(ap) struct vattr vattr; int vpid; - vdp = *vpp; - vpid = vdp->v_id; + newvp = *vpp; + vpid = newvp->v_id; /* * See the comment starting `Step through' in ufs/ufs_lookup.c * for an explanation of the locking protocol */ - if (dvp == vdp) { - VREF(vdp); + if (dvp == newvp) { + VREF(newvp); error = 0; - } else - error = vget(vdp, 1); + } else if (flags & ISDOTDOT) { + VOP_UNLOCK(dvp); + error = vget(newvp, 1); + if (!error && lockparent && (flags & ISLASTCN)) + error = VOP_LOCK(dvp); + } else { + error = vget(newvp, 1); + if (!lockparent || error || !(flags & ISLASTCN)) + VOP_UNLOCK(dvp); + } if (!error) { - if (vpid == vdp->v_id) { - if (nmp->nm_flag & NFSMNT_NQNFS) { - if ((nmp->nm_flag & NFSMNT_NQLOOKLEASE) == 0) { - nfsstats.lookupcache_hits++; - if (cnp->cn_nameiop != LOOKUP && - (flags & ISLASTCN)) - cnp->cn_flags |= SAVENAME; - return (0); - } else if (NQNFS_CKCACHABLE(dvp, NQL_READ)) { - if (np->n_lrev != np->n_brev || - (np->n_flag & NMODIFIED)) { - np->n_direofoffset = 0; - cache_purge(dvp); - error = nfs_vinvalbuf(dvp, 0, - cnp->cn_cred, cnp->cn_proc, - 1); - if (error == EINTR) - return (error); - np->n_brev = np->n_lrev; - } else { - nfsstats.lookupcache_hits++; - if (cnp->cn_nameiop != LOOKUP && - (flags & ISLASTCN)) - cnp->cn_flags |= SAVENAME; - return (0); - } - } - } else if (!VOP_GETATTR(vdp, &vattr, cnp->cn_cred, cnp->cn_proc) && - vattr.va_ctime.ts_sec == VTONFS(vdp)->n_ctime) { + if (vpid == newvp->v_id) { + if (!VOP_GETATTR(newvp, &vattr, cnp->cn_cred, cnp->cn_proc) + && vattr.va_ctime.ts_sec == VTONFS(newvp)->n_ctime) { nfsstats.lookupcache_hits++; if (cnp->cn_nameiop != LOOKUP && (flags & ISLASTCN)) cnp->cn_flags |= SAVENAME; return (0); } - cache_purge(vdp); + cache_purge(newvp); } - vrele(vdp); + vput(newvp); + if (lockparent && dvp != newvp && (flags & ISLASTCN)) + VOP_UNLOCK(dvp); } + error = VOP_LOCK(dvp); + if (error) + return (error); *vpp = NULLVP; } error = 0; + newvp = NULLVP; nfsstats.lookupcache_misses++; nfsstats.rpccnt[NFSPROC_LOOKUP]++; len = cnp->cn_namelen; - nfsm_reqhead(dvp, NFSPROC_LOOKUP, NFSX_FH+NFSX_UNSIGNED+nfsm_rndup(len)); - - /* - * For nqnfs optionally piggyback a getlease request for the name - * being looked up. - */ - if (nmp->nm_flag & NFSMNT_NQNFS) { - nfsm_build(tl, u_long *, NFSX_UNSIGNED); - if ((nmp->nm_flag & NFSMNT_NQLOOKLEASE) && - ((cnp->cn_flags & MAKEENTRY) && - (cnp->cn_nameiop != DELETE || !(flags & ISLASTCN)))) - *tl = txdr_unsigned(nmp->nm_leaseterm); - else - *tl = 0; - } - nfsm_fhtom(dvp); + nfsm_reqhead(dvp, NFSPROC_LOOKUP, + NFSX_FH(v3) + NFSX_UNSIGNED + nfsm_rndup(len)); + nfsm_fhtom(dvp, v3); nfsm_strtom(cnp->cn_nameptr, len, NFS_MAXNAMLEN); - reqtime = time.tv_sec; nfsm_request(dvp, NFSPROC_LOOKUP, cnp->cn_proc, cnp->cn_cred); -nfsmout: if (error) { - if ((cnp->cn_nameiop == CREATE || cnp->cn_nameiop == RENAME) && - (flags & ISLASTCN) && error == ENOENT) - error = EJUSTRETURN; - if (cnp->cn_nameiop != LOOKUP && (flags & ISLASTCN)) - cnp->cn_flags |= SAVENAME; - return (error); - } - if (nmp->nm_flag & NFSMNT_NQNFS) { - nfsm_dissect(tl, u_long *, NFSX_UNSIGNED); - if (*tl) { - nqlflag = fxdr_unsigned(int, *tl); - nfsm_dissect(tl, u_long *, 4*NFSX_UNSIGNED); - cachable = fxdr_unsigned(int, *tl++); - reqtime += fxdr_unsigned(int, *tl++); - fxdr_hyper(tl, &frev); - } else - nqlflag = 0; + nfsm_postop_attr(dvp, attrflag); + m_freem(mrep); + goto nfsmout; } - nfsm_dissect(fhp, nfsv2fh_t *, NFSX_FH); + nfsm_getfh(fhp, fhsize, v3); /* * Handle RENAME case... */ if (cnp->cn_nameiop == RENAME && wantparent && (flags & ISLASTCN)) { - if (!bcmp(np->n_fh.fh_bytes, (caddr_t)fhp, NFSX_FH)) { + if (NFS_CMPFH(np, fhp, fhsize)) { m_freem(mrep); return (EISDIR); } - error = nfs_nget(dvp->v_mount, fhp, &np); - if (error) { + if (error = nfs_nget(dvp->v_mount, fhp, fhsize, &np)) { m_freem(mrep); return (error); } newvp = NFSTOV(np); - error = nfs_loadattrcache(&newvp, &md, &dpos, (struct vattr*)0); - if (error) { - vrele(newvp); - m_freem(mrep); - return (error); - } + if (v3) { + nfsm_postop_attr(newvp, attrflag); + nfsm_postop_attr(dvp, attrflag); + } else + nfsm_loadattr(newvp, (struct vattr *)0); *vpp = newvp; m_freem(mrep); cnp->cn_flags |= SAVENAME; + if (!lockparent) + VOP_UNLOCK(dvp); return (0); } - if (!bcmp(np->n_fh.fh_bytes, (caddr_t)fhp, NFSX_FH)) { + if (flags & ISDOTDOT) { + VOP_UNLOCK(dvp); + error = nfs_nget(dvp->v_mount, fhp, fhsize, &np); + if (error) { + VOP_LOCK(dvp); + return (error); + } + newvp = NFSTOV(np); + if (lockparent && (flags & ISLASTCN) && + (error = VOP_LOCK(dvp))) { + vput(newvp); + return (error); + } + } else if (NFS_CMPFH(np, fhp, fhsize)) { VREF(dvp); newvp = dvp; } else { - error = nfs_nget(dvp->v_mount, fhp, &np); - if (error) { + if (error = nfs_nget(dvp->v_mount, fhp, fhsize, &np)) { m_freem(mrep); return (error); } + if (!lockparent || !(flags & ISLASTCN)) + VOP_UNLOCK(dvp); newvp = NFSTOV(np); } - error = nfs_loadattrcache(&newvp, &md, &dpos, (struct vattr *)0); - if (error) { - vrele(newvp); - m_freem(mrep); - return (error); - } - m_freem(mrep); - *vpp = newvp; + if (v3) { + nfsm_postop_attr(newvp, attrflag); + nfsm_postop_attr(dvp, attrflag); + } else + nfsm_loadattr(newvp, (struct vattr *)0); if (cnp->cn_nameiop != LOOKUP && (flags & ISLASTCN)) cnp->cn_flags |= SAVENAME; if ((cnp->cn_flags & MAKEENTRY) && (cnp->cn_nameiop != DELETE || !(flags & ISLASTCN))) { - if ((nmp->nm_flag & NFSMNT_NQNFS) == 0) - np->n_ctime = np->n_vattr.va_ctime.ts_sec; - else if (nqlflag && reqtime > time.tv_sec) - nqnfs_clientlease(nmp, np, nqlflag, cachable, reqtime, - frev); - cache_enter(dvp, *vpp, cnp); + np->n_ctime = np->n_vattr.va_ctime.ts_sec; + cache_enter(dvp, newvp, cnp); } - return (0); + *vpp = newvp; + nfsm_reqdone; + if (error) { + if (newvp != NULLVP) + vrele(newvp); + if ((cnp->cn_nameiop == CREATE || cnp->cn_nameiop == RENAME) && + (flags & ISLASTCN) && error == ENOENT) { + if (!lockparent) + VOP_UNLOCK(dvp); + error = EJUSTRETURN; + } + if (cnp->cn_nameiop != LOOKUP && (flags & ISLASTCN)) + cnp->cn_flags |= SAVENAME; + } + return (error); } /* @@ -810,18 +966,22 @@ nfs_readlinkrpc(vp, uiop, cred) { register u_long *tl; register caddr_t cp; - register long t1; + register long t1, t2; caddr_t bpos, dpos, cp2; - int error = 0; + int error = 0, len, attrflag; struct mbuf *mreq, *mrep, *md, *mb, *mb2; - long len; + int v3 = NFS_ISV3(vp); nfsstats.rpccnt[NFSPROC_READLINK]++; - nfsm_reqhead(vp, NFSPROC_READLINK, NFSX_FH); - nfsm_fhtom(vp); + nfsm_reqhead(vp, NFSPROC_READLINK, NFSX_FH(v3)); + nfsm_fhtom(vp, v3); nfsm_request(vp, NFSPROC_READLINK, uiop->uio_procp, cred); - nfsm_strsiz(len, NFS_MAXPATHLEN); - nfsm_mtouio(uiop, len); + if (v3) + nfsm_postop_attr(vp, attrflag); + if (!error) { + nfsm_strsiz(len, NFS_MAXPATHLEN); + nfsm_mtouio(uiop, len); + } nfsm_reqdone; return (error); } @@ -838,25 +998,27 @@ nfs_readrpc(vp, uiop, cred) { register u_long *tl; register caddr_t cp; - register long t1; + register long t1, t2; caddr_t bpos, dpos, cp2; - int error = 0; struct mbuf *mreq, *mrep, *md, *mb, *mb2; struct nfsmount *nmp; - long len, retlen, tsiz; + int error = 0, len, retlen, tsiz, eof, attrflag; + int v3 = NFS_ISV3(vp); +#ifndef nolint + eof = 0; +#endif nmp = VFSTONFS(vp->v_mount); tsiz = uiop->uio_resid; - if (uiop->uio_offset + tsiz > 0xffffffff && - (nmp->nm_flag & NFSMNT_NQNFS) == 0) + if (uiop->uio_offset + tsiz > 0xffffffff && !v3) return (EFBIG); while (tsiz > 0) { nfsstats.rpccnt[NFSPROC_READ]++; len = (tsiz > nmp->nm_rsize) ? nmp->nm_rsize : tsiz; - nfsm_reqhead(vp, NFSPROC_READ, NFSX_FH+NFSX_UNSIGNED*3); - nfsm_fhtom(vp); - nfsm_build(tl, u_long *, NFSX_UNSIGNED*3); - if (nmp->nm_flag & NFSMNT_NQNFS) { + nfsm_reqhead(vp, NFSPROC_READ, NFSX_FH(v3) + NFSX_UNSIGNED * 3); + nfsm_fhtom(vp, v3); + nfsm_build(tl, u_long *, NFSX_UNSIGNED * 3); + if (v3) { txdr_hyper(&uiop->uio_offset, tl); *(tl + 2) = txdr_unsigned(len); } else { @@ -865,14 +1027,25 @@ nfs_readrpc(vp, uiop, cred) *tl = 0; } nfsm_request(vp, NFSPROC_READ, uiop->uio_procp, cred); - nfsm_loadattr(vp, (struct vattr *)0); + if (v3) { + nfsm_postop_attr(vp, attrflag); + if (error) { + m_freem(mrep); + goto nfsmout; + } + nfsm_dissect(tl, u_long *, 2 * NFSX_UNSIGNED); + eof = fxdr_unsigned(int, *(tl + 1)); + } else + nfsm_loadattr(vp, (struct vattr *)0); nfsm_strsiz(retlen, nmp->nm_rsize); nfsm_mtouio(uiop, retlen); m_freem(mrep); - if (retlen < len) + tsiz -= retlen; + if (v3) { + if (eof || retlen == 0) + tsiz = 0; + } else if (retlen < len) tsiz = 0; - else - tsiz -= len; } nfsmout: return (error); @@ -882,150 +1055,227 @@ nfsmout: * nfs write call */ int -nfs_writerpc(vp, uiop, cred, ioflags) +nfs_writerpc(vp, uiop, cred, iomode, must_commit) register struct vnode *vp; - struct uio *uiop; + register struct uio *uiop; struct ucred *cred; - int ioflags; + int *iomode, *must_commit; { register u_long *tl; register caddr_t cp; - register long t1; + register int t1, t2, backup; caddr_t bpos, dpos, cp2; - int error = 0; struct mbuf *mreq, *mrep, *md, *mb, *mb2; - struct nfsmount *nmp; + struct nfsmount *nmp = VFSTONFS(vp->v_mount); struct nfsnode *np = VTONFS(vp); u_quad_t frev; - long len, tsiz; + int error = 0, len, tsiz, wccflag = NFSV3_WCCRATTR, rlen, commit; + int v3 = NFS_ISV3(vp), committed = NFSV3WRITE_FILESYNC; - nmp = VFSTONFS(vp->v_mount); +#ifndef DIAGNOSTIC + if (uiop->uio_iovcnt != 1) + panic("nfs: writerpc iovcnt > 1"); +#endif + *must_commit = 0; tsiz = uiop->uio_resid; - if (uiop->uio_offset + tsiz > 0xffffffff && - (nmp->nm_flag & NFSMNT_NQNFS) == 0) + if (uiop->uio_offset + tsiz > 0xffffffff && !v3) return (EFBIG); while (tsiz > 0) { nfsstats.rpccnt[NFSPROC_WRITE]++; len = (tsiz > nmp->nm_wsize) ? nmp->nm_wsize : tsiz; nfsm_reqhead(vp, NFSPROC_WRITE, - NFSX_FH+NFSX_UNSIGNED*4+nfsm_rndup(len)); - nfsm_fhtom(vp); - nfsm_build(tl, u_long *, NFSX_UNSIGNED * 4); - if (nmp->nm_flag & NFSMNT_NQNFS) { + NFSX_FH(v3) + 5 * NFSX_UNSIGNED + nfsm_rndup(len)); + nfsm_fhtom(vp, v3); + if (v3) { + nfsm_build(tl, u_long *, 5 * NFSX_UNSIGNED); txdr_hyper(&uiop->uio_offset, tl); tl += 2; -#ifdef notyet - if (ioflags & IO_APPEND) - *tl++ = txdr_unsigned(1); - else -#endif - *tl++ = 0; + *tl++ = txdr_unsigned(len); + *tl++ = txdr_unsigned(*iomode); } else { + nfsm_build(tl, u_long *, 4 * NFSX_UNSIGNED); *++tl = txdr_unsigned(uiop->uio_offset); tl += 2; } *tl = txdr_unsigned(len); nfsm_uiotom(uiop, len); nfsm_request(vp, NFSPROC_WRITE, uiop->uio_procp, cred); - nfsm_loadattr(vp, (struct vattr *)0); - if (nmp->nm_flag & NFSMNT_MYWRITE) - VTONFS(vp)->n_mtime = VTONFS(vp)->n_vattr.va_mtime.ts_sec; - else if ((nmp->nm_flag & NFSMNT_NQNFS) && - NQNFS_CKCACHABLE(vp, NQL_WRITE)) { - nfsm_dissect(tl, u_long *, 2*NFSX_UNSIGNED); - fxdr_hyper(tl, &frev); - if (frev > np->n_brev) - np->n_brev = frev; - } + if (v3) { + wccflag = NFSV3_WCCCHK; + nfsm_wcc_data(vp, wccflag); + if (!error) { + nfsm_dissect(tl, u_long *, 2 * NFSX_UNSIGNED + + NFSX_V3WRITEVERF); + rlen = fxdr_unsigned(int, *tl++); + if (rlen == 0) { + error = NFSERR_IO; + break; + } else if (rlen < len) { + backup = len - rlen; + uiop->uio_iov->iov_base -= backup; + uiop->uio_iov->iov_len += backup; + uiop->uio_offset -= backup; + uiop->uio_resid += backup; + len = rlen; + } + commit = fxdr_unsigned(int, *tl++); + + /* + * Return the lowest committment level + * obtained by any of the RPCs. + */ + if (committed == NFSV3WRITE_FILESYNC) + committed = commit; + else if (committed == NFSV3WRITE_DATASYNC && + commit == NFSV3WRITE_UNSTABLE) + committed = commit; + if ((nmp->nm_flag & NFSMNT_HASWRITEVERF) == 0) { + bcopy((caddr_t)tl, (caddr_t)nmp->nm_verf, + NFSX_V3WRITEVERF); + nmp->nm_flag |= NFSMNT_HASWRITEVERF; + } else if (bcmp((caddr_t)tl, + (caddr_t)nmp->nm_verf, NFSX_V3WRITEVERF)) { + *must_commit = 1; + bcopy((caddr_t)tl, (caddr_t)nmp->nm_verf, + NFSX_V3WRITEVERF); + } + } + } else + nfsm_loadattr(vp, (struct vattr *)0); + if (wccflag) + VTONFS(vp)->n_mtime = VTONFS(vp)->n_vattr.va_mtime.ts_sec; m_freem(mrep); tsiz -= len; } nfsmout: + *iomode = committed; if (error) uiop->uio_resid = tsiz; return (error); } /* - * nfs mknod call - * This is a kludge. Use a create rpc but with the IFMT bits of the mode - * set to specify the file type and the size field for rdev. + * nfs mknod rpc + * For NFS v2 this is a kludge. Use a create rpc but with the IFMT bits of the + * mode set to specify the file type and the size field for rdev. */ -/* ARGSUSED */ int -nfs_mknod(ap) - struct vop_mknod_args /* { - struct vnode *a_dvp; - struct vnode **a_vpp; - struct componentname *a_cnp; - struct vattr *a_vap; - } */ *ap; +nfs_mknodrpc(dvp, vpp, cnp, vap) + register struct vnode *dvp; + register struct vnode **vpp; + register struct componentname *cnp; + register struct vattr *vap; { - register struct vnode *dvp = ap->a_dvp; - register struct vattr *vap = ap->a_vap; - register struct componentname *cnp = ap->a_cnp; register struct nfsv2_sattr *sp; + register struct nfsv3_sattr *sp3; register u_long *tl; register caddr_t cp; register long t1, t2; - struct vnode *newvp = 0; + struct vnode *newvp = (struct vnode *)0; + struct nfsnode *np = (struct nfsnode *)0; struct vattr vattr; char *cp2; caddr_t bpos, dpos; - int error = 0, isnq; + int error = 0, wccflag = NFSV3_WCCRATTR, gotvp = 0; struct mbuf *mreq, *mrep, *md, *mb, *mb2; u_long rdev; + int v3 = NFS_ISV3(dvp); - isnq = (VFSTONFS(dvp->v_mount)->nm_flag & NFSMNT_NQNFS); if (vap->va_type == VCHR || vap->va_type == VBLK) rdev = txdr_unsigned(vap->va_rdev); - else if (vap->va_type == VFIFO) + else if (vap->va_type == VFIFO || vap->va_type == VSOCK) rdev = 0xffffffff; else { VOP_ABORTOP(dvp, cnp); vput(dvp); return (EOPNOTSUPP); } - error = VOP_GETATTR(dvp, &vattr, cnp->cn_cred, cnp->cn_proc); - if (error) { + if (error = VOP_GETATTR(dvp, &vattr, cnp->cn_cred, cnp->cn_proc)) { VOP_ABORTOP(dvp, cnp); vput(dvp); return (error); } - newvp = NULLVP; - nfsstats.rpccnt[NFSPROC_CREATE]++; - nfsm_reqhead(dvp, NFSPROC_CREATE, - NFSX_FH+NFSX_UNSIGNED+nfsm_rndup(cnp->cn_namelen)+NFSX_SATTR(isnq)); - nfsm_fhtom(dvp); + nfsstats.rpccnt[NFSPROC_MKNOD]++; + nfsm_reqhead(dvp, NFSPROC_MKNOD, NFSX_FH(v3) + 4 * NFSX_UNSIGNED + + + nfsm_rndup(cnp->cn_namelen) + NFSX_SATTR(v3)); + nfsm_fhtom(dvp, v3); nfsm_strtom(cnp->cn_nameptr, cnp->cn_namelen, NFS_MAXNAMLEN); - nfsm_build(sp, struct nfsv2_sattr *, NFSX_SATTR(isnq)); - sp->sa_mode = vtonfs_mode(vap->va_type, vap->va_mode); - sp->sa_uid = txdr_unsigned(cnp->cn_cred->cr_uid); - sp->sa_gid = txdr_unsigned(vattr.va_gid); - if (isnq) { - sp->sa_nqrdev = rdev; - sp->sa_nqflags = 0; - txdr_nqtime(&vap->va_atime, &sp->sa_nqatime); - txdr_nqtime(&vap->va_mtime, &sp->sa_nqmtime); + if (v3) { + nfsm_build(tl, u_long *, NFSX_UNSIGNED + NFSX_V3SRVSATTR); + *tl++ = vtonfsv3_type(vap->va_type); + sp3 = (struct nfsv3_sattr *)tl; + nfsm_v3sattr(sp3, vap, cnp->cn_cred->cr_uid, vattr.va_gid); + if (vap->va_type == VCHR || vap->va_type == VBLK) { + nfsm_build(tl, u_long *, 2 * NFSX_UNSIGNED); + *tl++ = txdr_unsigned(major(vap->va_rdev)); + *tl = txdr_unsigned(minor(vap->va_rdev)); + } } else { - sp->sa_nfssize = rdev; - txdr_nfstime(&vap->va_atime, &sp->sa_nfsatime); - txdr_nfstime(&vap->va_mtime, &sp->sa_nfsmtime); + nfsm_build(sp, struct nfsv2_sattr *, NFSX_V2SATTR); + sp->sa_mode = vtonfsv2_mode(vap->va_type, vap->va_mode); + sp->sa_uid = txdr_unsigned(cnp->cn_cred->cr_uid); + sp->sa_gid = txdr_unsigned(vattr.va_gid); + sp->sa_size = rdev; + txdr_nfsv2time(&vap->va_atime, &sp->sa_atime); + txdr_nfsv2time(&vap->va_mtime, &sp->sa_mtime); } - nfsm_request(dvp, NFSPROC_CREATE, cnp->cn_proc, cnp->cn_cred); - nfsm_mtofh(dvp, newvp); + nfsm_request(dvp, NFSPROC_MKNOD, cnp->cn_proc, cnp->cn_cred); + if (!error) { + nfsm_mtofh(dvp, newvp, v3, gotvp); + if (!gotvp) { + if (newvp) { + vput(newvp); + newvp = (struct vnode *)0; + } + error = nfs_lookitup(dvp, cnp->cn_nameptr, + cnp->cn_namelen, cnp->cn_cred, cnp->cn_proc, &np); + if (!error) + newvp = NFSTOV(np); + } + } + if (v3) + nfsm_wcc_data(dvp, wccflag); nfsm_reqdone; - if (!error && (cnp->cn_flags & MAKEENTRY)) - cache_enter(dvp, newvp, cnp); + if (error) { + if (newvp) + vput(newvp); + } else { + if (cnp->cn_flags & MAKEENTRY) + cache_enter(dvp, newvp, cnp); + *vpp = newvp; + } FREE(cnp->cn_pnbuf, M_NAMEI); VTONFS(dvp)->n_flag |= NMODIFIED; - VTONFS(dvp)->n_attrstamp = 0; - vrele(dvp); - if (newvp != NULLVP) - vrele(newvp); + if (!wccflag) + VTONFS(dvp)->n_attrstamp = 0; + vput(dvp); + return (error); +} + +/* + * nfs mknod vop + * just call nfs_mknodrpc() to do the work. + */ +/* ARGSUSED */ +int +nfs_mknod(ap) + struct vop_mknod_args /* { + struct vnode *a_dvp; + struct vnode **a_vpp; + struct componentname *a_cnp; + struct vattr *a_vap; + } */ *ap; +{ + struct vnode *newvp; + int error; + + error = nfs_mknodrpc(ap->a_dvp, &newvp, ap->a_cnp, ap->a_vap); + if (!error) + vput(newvp); return (error); } +static u_long create_verf; /* * nfs file create call */ @@ -1042,52 +1292,98 @@ nfs_create(ap) register struct vattr *vap = ap->a_vap; register struct componentname *cnp = ap->a_cnp; register struct nfsv2_sattr *sp; + register struct nfsv3_sattr *sp3; register u_long *tl; register caddr_t cp; register long t1, t2; + struct nfsnode *np = (struct nfsnode *)0; + struct vnode *newvp = (struct vnode *)0; caddr_t bpos, dpos, cp2; - int error = 0, isnq; + int error = 0, wccflag = NFSV3_WCCRATTR, gotvp = 0, fmode = 0; struct mbuf *mreq, *mrep, *md, *mb, *mb2; struct vattr vattr; + int v3 = NFS_ISV3(dvp); - error = VOP_GETATTR(dvp, &vattr, cnp->cn_cred, cnp->cn_proc); - if (error) { + /* + * Oops, not for me.. + */ + if (vap->va_type == VSOCK) + return (nfs_mknodrpc(dvp, ap->a_vpp, cnp, vap)); + + if (error = VOP_GETATTR(dvp, &vattr, cnp->cn_cred, cnp->cn_proc)) { VOP_ABORTOP(dvp, cnp); vput(dvp); return (error); } + if (vap->va_vaflags & VA_EXCLUSIVE) + fmode |= O_EXCL; +again: nfsstats.rpccnt[NFSPROC_CREATE]++; - isnq = (VFSTONFS(dvp->v_mount)->nm_flag & NFSMNT_NQNFS); - nfsm_reqhead(dvp, NFSPROC_CREATE, - NFSX_FH+NFSX_UNSIGNED+nfsm_rndup(cnp->cn_namelen)+NFSX_SATTR(isnq)); - nfsm_fhtom(dvp); + nfsm_reqhead(dvp, NFSPROC_CREATE, NFSX_FH(v3) + 2 * NFSX_UNSIGNED + + nfsm_rndup(cnp->cn_namelen) + NFSX_SATTR(v3)); + nfsm_fhtom(dvp, v3); nfsm_strtom(cnp->cn_nameptr, cnp->cn_namelen, NFS_MAXNAMLEN); - nfsm_build(sp, struct nfsv2_sattr *, NFSX_SATTR(isnq)); - sp->sa_mode = vtonfs_mode(vap->va_type, vap->va_mode); - sp->sa_uid = txdr_unsigned(cnp->cn_cred->cr_uid); - sp->sa_gid = txdr_unsigned(vattr.va_gid); - if (isnq) { - u_quad_t qval = 0; - - txdr_hyper(&qval, &sp->sa_nqsize); - sp->sa_nqflags = 0; - sp->sa_nqrdev = -1; - txdr_nqtime(&vap->va_atime, &sp->sa_nqatime); - txdr_nqtime(&vap->va_mtime, &sp->sa_nqmtime); + if (v3) { + nfsm_build(tl, u_long *, NFSX_UNSIGNED); + if (fmode & O_EXCL) { + *tl = txdr_unsigned(NFSV3CREATE_EXCLUSIVE); + nfsm_build(tl, u_long *, NFSX_V3CREATEVERF); + if (in_ifaddr) + *tl++ = IA_SIN(in_ifaddr)->sin_addr.s_addr; + else + *tl++ = create_verf; + *tl = ++create_verf; + } else { + *tl = txdr_unsigned(NFSV3CREATE_UNCHECKED); + nfsm_build(tl, u_long *, NFSX_V3SRVSATTR); + sp3 = (struct nfsv3_sattr *)tl; + nfsm_v3sattr(sp3, vap, cnp->cn_cred->cr_uid, vattr.va_gid); + } } else { - sp->sa_nfssize = 0; - txdr_nfstime(&vap->va_atime, &sp->sa_nfsatime); - txdr_nfstime(&vap->va_mtime, &sp->sa_nfsmtime); + nfsm_build(sp, struct nfsv2_sattr *, NFSX_V2SATTR); + sp->sa_mode = vtonfsv2_mode(vap->va_type, vap->va_mode); + sp->sa_uid = txdr_unsigned(cnp->cn_cred->cr_uid); + sp->sa_gid = txdr_unsigned(vattr.va_gid); + sp->sa_size = 0; + txdr_nfsv2time(&vap->va_atime, &sp->sa_atime); + txdr_nfsv2time(&vap->va_mtime, &sp->sa_mtime); } nfsm_request(dvp, NFSPROC_CREATE, cnp->cn_proc, cnp->cn_cred); - nfsm_mtofh(dvp, *ap->a_vpp); + if (!error) { + nfsm_mtofh(dvp, newvp, v3, gotvp); + if (!gotvp) { + if (newvp) { + vput(newvp); + newvp = (struct vnode *)0; + } + error = nfs_lookitup(dvp, cnp->cn_nameptr, + cnp->cn_namelen, cnp->cn_cred, cnp->cn_proc, &np); + if (!error) + newvp = NFSTOV(np); + } + } + if (v3) + nfsm_wcc_data(dvp, wccflag); nfsm_reqdone; - if (!error && (cnp->cn_flags & MAKEENTRY)) - cache_enter(dvp, *ap->a_vpp, cnp); + if (error) { + if (v3 && (fmode & O_EXCL) && error == NFSERR_NOTSUPP) { + fmode &= ~O_EXCL; + goto again; + } + if (newvp) + vput(newvp); + } else if (v3 && (fmode & O_EXCL)) + error = nfs_setattrrpc(newvp, vap, cnp->cn_cred, cnp->cn_proc); + if (!error) { + if (cnp->cn_flags & MAKEENTRY) + cache_enter(dvp, newvp, cnp); + *ap->a_vpp = newvp; + } FREE(cnp->cn_pnbuf, M_NAMEI); VTONFS(dvp)->n_flag |= NMODIFIED; - VTONFS(dvp)->n_attrstamp = 0; - vrele(dvp); + if (!wccflag) + VTONFS(dvp)->n_attrstamp = 0; + vput(dvp); return (error); } @@ -1122,22 +1418,17 @@ nfs_remove(ap) int error = 0; struct mbuf *mreq, *mrep, *md, *mb, *mb2; struct vattr vattr; + int v3 = NFS_ISV3(dvp); - if (vp->v_usecount > 1) { - if (!np->n_sillyrename) - error = nfs_sillyrename(dvp, vp, cnp); - else if (VOP_GETATTR(vp, &vattr, cnp->cn_cred, cnp->cn_proc) - == 0 && vattr.va_nlink > 1) - /* - * If we already have a silly name but there are more - * than one links, just proceed with the NFS remove - * request, as the bits will remain available (modulo - * network races). This avoids silently ignoring the - * attempted removal of a non-silly entry. - */ - goto doit; - } else { - doit: +#ifndef DIAGNOSTIC + if ((cnp->cn_flags & HASBUF) == 0) + panic("nfs_remove: no name"); + if (vp->v_usecount < 1) + panic("nfs_remove: bad v_usecount"); +#endif + if (vp->v_usecount == 1 || (np->n_sillyrename && + VOP_GETATTR(vp, &vattr, cnp->cn_cred, cnp->cn_proc) == 0 && + vattr.va_nlink > 1)) { /* * Purge the name cache so that the chance of a lookup for * the name succeeding while the remove is in progress is @@ -1147,23 +1438,14 @@ nfs_remove(ap) */ cache_purge(vp); /* - * Throw away biocache buffers. Mainly to avoid - * unnecessary delayed writes. + * throw away biocache buffers, mainly to avoid + * unnecessary delayed writes later. */ error = nfs_vinvalbuf(vp, 0, cnp->cn_cred, cnp->cn_proc, 1); - if (error == EINTR) - return (error); /* Do the rpc */ - nfsstats.rpccnt[NFSPROC_REMOVE]++; - nfsm_reqhead(dvp, NFSPROC_REMOVE, - NFSX_FH+NFSX_UNSIGNED+nfsm_rndup(cnp->cn_namelen)); - nfsm_fhtom(dvp); - nfsm_strtom(cnp->cn_nameptr, cnp->cn_namelen, NFS_MAXNAMLEN); - nfsm_request(dvp, NFSPROC_REMOVE, cnp->cn_proc, cnp->cn_cred); - nfsm_reqdone; - FREE(cnp->cn_pnbuf, M_NAMEI); - VTONFS(dvp)->n_flag |= NMODIFIED; - VTONFS(dvp)->n_attrstamp = 0; + if (error != EINTR) + error = nfs_removerpc(dvp, cnp->cn_nameptr, + cnp->cn_namelen, cnp->cn_cred, cnp->cn_proc); /* * Kludge City: If the first reply to the remove rpc is lost.. * the reply to the retransmitted request will be ENOENT @@ -1172,10 +1454,15 @@ nfs_remove(ap) */ if (error == ENOENT) error = 0; - } + } else if (!np->n_sillyrename) + error = nfs_sillyrename(dvp, vp, cnp); + FREE(cnp->cn_pnbuf, M_NAMEI); np->n_attrstamp = 0; - vrele(dvp); - vrele(vp); + vput(dvp); + if (vp == dvp) + vrele(vp); + else + vput(vp); return (error); } @@ -1186,22 +1473,42 @@ int nfs_removeit(sp) register struct sillyrename *sp; { + + return (nfs_removerpc(sp->s_dvp, sp->s_name, sp->s_namlen, sp->s_cred, + (struct proc *)0)); +} + +/* + * Nfs remove rpc, called from nfs_remove() and nfs_removeit(). + */ +int +nfs_removerpc(dvp, name, namelen, cred, proc) + register struct vnode *dvp; + char *name; + int namelen; + struct ucred *cred; + struct proc *proc; +{ register u_long *tl; register caddr_t cp; - register long t2; - caddr_t bpos, dpos; - int error = 0; + register long t1, t2; + caddr_t bpos, dpos, cp2; + int error = 0, wccflag = NFSV3_WCCRATTR; struct mbuf *mreq, *mrep, *md, *mb, *mb2; + int v3 = NFS_ISV3(dvp); nfsstats.rpccnt[NFSPROC_REMOVE]++; - nfsm_reqhead(sp->s_dvp, NFSPROC_REMOVE, - NFSX_FH+NFSX_UNSIGNED+nfsm_rndup(sp->s_namlen)); - nfsm_fhtom(sp->s_dvp); - nfsm_strtom(sp->s_name, sp->s_namlen, NFS_MAXNAMLEN); - nfsm_request(sp->s_dvp, NFSPROC_REMOVE, NULL, sp->s_cred); + nfsm_reqhead(dvp, NFSPROC_REMOVE, + NFSX_FH(v3) + NFSX_UNSIGNED + nfsm_rndup(namelen)); + nfsm_fhtom(dvp, v3); + nfsm_strtom(name, namelen, NFS_MAXNAMLEN); + nfsm_request(dvp, NFSPROC_REMOVE, proc, cred); + if (v3) + nfsm_wcc_data(dvp, wccflag); nfsm_reqdone; - VTONFS(sp->s_dvp)->n_flag |= NMODIFIED; - VTONFS(sp->s_dvp)->n_attrstamp = 0; + VTONFS(dvp)->n_flag |= NMODIFIED; + if (!wccflag) + VTONFS(dvp)->n_attrstamp = 0; return (error); } @@ -1225,13 +1532,13 @@ nfs_rename(ap) register struct vnode *tdvp = ap->a_tdvp; register struct componentname *tcnp = ap->a_tcnp; register struct componentname *fcnp = ap->a_fcnp; - register u_long *tl; - register caddr_t cp; - register long t2; - caddr_t bpos, dpos; - int error = 0; - struct mbuf *mreq, *mrep, *md, *mb, *mb2; + int error; +#ifndef DIAGNOSTIC + if ((tcnp->cn_flags & HASBUF) == 0 || + (fcnp->cn_flags & HASBUF) == 0) + panic("nfs_rename: no name"); +#endif /* Check for cross-device rename */ if ((fvp->v_mount != tdvp->v_mount) || (tvp && (fvp->v_mount != tvp->v_mount))) { @@ -1239,21 +1546,20 @@ nfs_rename(ap) goto out; } + /* + * If the tvp exists and is in use, sillyrename it before doing the + * rename of the new file over it. + */ + if (tvp && tvp->v_usecount > 1 && !VTONFS(tvp)->n_sillyrename && + !nfs_sillyrename(tdvp, tvp, tcnp)) { + vrele(tvp); + tvp = NULL; + } + + error = nfs_renamerpc(fdvp, fcnp->cn_nameptr, fcnp->cn_namelen, + tdvp, tcnp->cn_nameptr, tcnp->cn_namelen, tcnp->cn_cred, + tcnp->cn_proc); - nfsstats.rpccnt[NFSPROC_RENAME]++; - nfsm_reqhead(fdvp, NFSPROC_RENAME, - (NFSX_FH+NFSX_UNSIGNED)*2+nfsm_rndup(fcnp->cn_namelen)+ - nfsm_rndup(fcnp->cn_namelen)); /* or fcnp->cn_cred?*/ - nfsm_fhtom(fdvp); - nfsm_strtom(fcnp->cn_nameptr, fcnp->cn_namelen, NFS_MAXNAMLEN); - nfsm_fhtom(tdvp); - nfsm_strtom(tcnp->cn_nameptr, tcnp->cn_namelen, NFS_MAXNAMLEN); - nfsm_request(fdvp, NFSPROC_RENAME, tcnp->cn_proc, tcnp->cn_cred); - nfsm_reqdone; - VTONFS(fdvp)->n_flag |= NMODIFIED; - VTONFS(fdvp)->n_attrstamp = 0; - VTONFS(tdvp)->n_flag |= NMODIFIED; - VTONFS(tdvp)->n_attrstamp = 0; if (fvp->v_type == VDIR) { if (tvp != NULL && tvp->v_type == VDIR) cache_purge(tdvp); @@ -1285,26 +1591,52 @@ nfs_renameit(sdvp, scnp, sp) struct componentname *scnp; register struct sillyrename *sp; { + return (nfs_renamerpc(sdvp, scnp->cn_nameptr, scnp->cn_namelen, + sdvp, sp->s_name, sp->s_namlen, scnp->cn_cred, scnp->cn_proc)); +} + +/* + * Do an nfs rename rpc. Called from nfs_rename() and nfs_renameit(). + */ +int +nfs_renamerpc(fdvp, fnameptr, fnamelen, tdvp, tnameptr, tnamelen, cred, proc) + register struct vnode *fdvp; + char *fnameptr; + int fnamelen; + register struct vnode *tdvp; + char *tnameptr; + int tnamelen; + struct ucred *cred; + struct proc *proc; +{ register u_long *tl; register caddr_t cp; - register long t2; - caddr_t bpos, dpos; - int error = 0; + register long t1, t2; + caddr_t bpos, dpos, cp2; + int error = 0, fwccflag = NFSV3_WCCRATTR, twccflag = NFSV3_WCCRATTR; struct mbuf *mreq, *mrep, *md, *mb, *mb2; + int v3 = NFS_ISV3(fdvp); nfsstats.rpccnt[NFSPROC_RENAME]++; - nfsm_reqhead(sdvp, NFSPROC_RENAME, - (NFSX_FH+NFSX_UNSIGNED)*2+nfsm_rndup(scnp->cn_namelen)+ - nfsm_rndup(sp->s_namlen)); - nfsm_fhtom(sdvp); - nfsm_strtom(scnp->cn_nameptr, scnp->cn_namelen, NFS_MAXNAMLEN); - nfsm_fhtom(sdvp); - nfsm_strtom(sp->s_name, sp->s_namlen, NFS_MAXNAMLEN); - nfsm_request(sdvp, NFSPROC_RENAME, scnp->cn_proc, scnp->cn_cred); + nfsm_reqhead(fdvp, NFSPROC_RENAME, + (NFSX_FH(v3) + NFSX_UNSIGNED)*2 + nfsm_rndup(fnamelen) + + nfsm_rndup(tnamelen)); + nfsm_fhtom(fdvp, v3); + nfsm_strtom(fnameptr, fnamelen, NFS_MAXNAMLEN); + nfsm_fhtom(tdvp, v3); + nfsm_strtom(tnameptr, tnamelen, NFS_MAXNAMLEN); + nfsm_request(fdvp, NFSPROC_RENAME, proc, cred); + if (v3) { + nfsm_wcc_data(fdvp, fwccflag); + nfsm_wcc_data(tdvp, twccflag); + } nfsm_reqdone; - FREE(scnp->cn_pnbuf, M_NAMEI); - VTONFS(sdvp)->n_flag |= NMODIFIED; - VTONFS(sdvp)->n_attrstamp = 0; + VTONFS(fdvp)->n_flag |= NMODIFIED; + VTONFS(tdvp)->n_flag |= NMODIFIED; + if (!fwccflag) + VTONFS(fdvp)->n_attrstamp = 0; + if (!twccflag) + VTONFS(tdvp)->n_attrstamp = 0; return (error); } @@ -1319,22 +1651,32 @@ nfs_link(ap) struct componentname *a_cnp; } */ *ap; { +#if defined(__FreeBSD__) || defined(__NetBSD__) + /* + * Since the args are reversed in the VOP_LINK() calls, + * switch them back. Argh! + */ + register struct vnode *vp = ap->a_tdvp; + register struct vnode *tdvp = ap->a_vp; +#else register struct vnode *vp = ap->a_vp; register struct vnode *tdvp = ap->a_tdvp; +#endif register struct componentname *cnp = ap->a_cnp; register u_long *tl; register caddr_t cp; - register long t2; - caddr_t bpos, dpos; - int error = 0; + register long t1, t2; + caddr_t bpos, dpos, cp2; + int error = 0, wccflag = NFSV3_WCCRATTR, attrflag = 0; struct mbuf *mreq, *mrep, *md, *mb, *mb2; + int v3 = NFS_ISV3(vp); if (vp->v_mount != tdvp->v_mount) { /*VOP_ABORTOP(vp, cnp);*/ if (tdvp == vp) - vrele(vp); + vrele(tdvp); else - vput(vp); + vput(tdvp); return (EXDEV); } @@ -1343,21 +1685,27 @@ nfs_link(ap) * doesn't get "out of sync" with the server. * XXX There should be a better way! */ - VOP_FSYNC(tdvp, cnp->cn_cred, MNT_WAIT, cnp->cn_proc); + VOP_FSYNC(vp, cnp->cn_cred, MNT_WAIT, cnp->cn_proc); nfsstats.rpccnt[NFSPROC_LINK]++; - nfsm_reqhead(tdvp, NFSPROC_LINK, - NFSX_FH*2+NFSX_UNSIGNED+nfsm_rndup(cnp->cn_namelen)); - nfsm_fhtom(tdvp); - nfsm_fhtom(vp); + nfsm_reqhead(vp, NFSPROC_LINK, + NFSX_FH(v3)*2 + NFSX_UNSIGNED + nfsm_rndup(cnp->cn_namelen)); + nfsm_fhtom(vp, v3); + nfsm_fhtom(tdvp, v3); nfsm_strtom(cnp->cn_nameptr, cnp->cn_namelen, NFS_MAXNAMLEN); - nfsm_request(tdvp, NFSPROC_LINK, cnp->cn_proc, cnp->cn_cred); + nfsm_request(vp, NFSPROC_LINK, cnp->cn_proc, cnp->cn_cred); + if (v3) { + nfsm_postop_attr(vp, attrflag); + nfsm_wcc_data(tdvp, wccflag); + } nfsm_reqdone; FREE(cnp->cn_pnbuf, M_NAMEI); - VTONFS(tdvp)->n_attrstamp = 0; - VTONFS(vp)->n_flag |= NMODIFIED; - VTONFS(vp)->n_attrstamp = 0; - vrele(vp); + VTONFS(tdvp)->n_flag |= NMODIFIED; + if (!attrflag) + VTONFS(vp)->n_attrstamp = 0; + if (!wccflag) + VTONFS(tdvp)->n_attrstamp = 0; + vput(tdvp); /* * Kludge: Map EEXIST => 0 assuming that it is a reply to a retry. */ @@ -1369,7 +1717,6 @@ nfs_link(ap) /* * nfs symbolic link create call */ -/* start here */ int nfs_symlink(ap) struct vop_symlink_args /* { @@ -1384,43 +1731,51 @@ nfs_symlink(ap) register struct vattr *vap = ap->a_vap; register struct componentname *cnp = ap->a_cnp; register struct nfsv2_sattr *sp; + register struct nfsv3_sattr *sp3; register u_long *tl; register caddr_t cp; - register long t2; - caddr_t bpos, dpos; - int slen, error = 0, isnq; + register long t1, t2; + caddr_t bpos, dpos, cp2; + int slen, error = 0, wccflag = NFSV3_WCCRATTR, gotvp; struct mbuf *mreq, *mrep, *md, *mb, *mb2; + struct vnode *newvp = (struct vnode *)0; + int v3 = NFS_ISV3(dvp); nfsstats.rpccnt[NFSPROC_SYMLINK]++; slen = strlen(ap->a_target); - isnq = (VFSTONFS(dvp->v_mount)->nm_flag & NFSMNT_NQNFS); - nfsm_reqhead(dvp, NFSPROC_SYMLINK, NFSX_FH+2*NFSX_UNSIGNED+ - nfsm_rndup(cnp->cn_namelen)+nfsm_rndup(slen)+NFSX_SATTR(isnq)); - nfsm_fhtom(dvp); + nfsm_reqhead(dvp, NFSPROC_SYMLINK, NFSX_FH(v3) + 2*NFSX_UNSIGNED + + nfsm_rndup(cnp->cn_namelen) + nfsm_rndup(slen) + NFSX_SATTR(v3)); + nfsm_fhtom(dvp, v3); nfsm_strtom(cnp->cn_nameptr, cnp->cn_namelen, NFS_MAXNAMLEN); + if (v3) { + nfsm_build(sp3, struct nfsv3_sattr *, NFSX_V3SRVSATTR); + nfsm_v3sattr(sp3, vap, cnp->cn_cred->cr_uid, + cnp->cn_cred->cr_gid); + } nfsm_strtom(ap->a_target, slen, NFS_MAXPATHLEN); - nfsm_build(sp, struct nfsv2_sattr *, NFSX_SATTR(isnq)); - sp->sa_mode = vtonfs_mode(VLNK, vap->va_mode); - sp->sa_uid = txdr_unsigned(cnp->cn_cred->cr_uid); - sp->sa_gid = txdr_unsigned(cnp->cn_cred->cr_gid); - if (isnq) { - quad_t qval = -1; - - txdr_hyper(&qval, &sp->sa_nqsize); - sp->sa_nqflags = 0; - txdr_nqtime(&vap->va_atime, &sp->sa_nqatime); - txdr_nqtime(&vap->va_mtime, &sp->sa_nqmtime); - } else { - sp->sa_nfssize = -1; - txdr_nfstime(&vap->va_atime, &sp->sa_nfsatime); - txdr_nfstime(&vap->va_mtime, &sp->sa_nfsmtime); + if (!v3) { + nfsm_build(sp, struct nfsv2_sattr *, NFSX_V2SATTR); + sp->sa_mode = vtonfsv2_mode(VLNK, vap->va_mode); + sp->sa_uid = txdr_unsigned(cnp->cn_cred->cr_uid); + sp->sa_gid = txdr_unsigned(cnp->cn_cred->cr_gid); + sp->sa_size = -1; + txdr_nfsv2time(&vap->va_atime, &sp->sa_atime); + txdr_nfsv2time(&vap->va_mtime, &sp->sa_mtime); } nfsm_request(dvp, NFSPROC_SYMLINK, cnp->cn_proc, cnp->cn_cred); + if (v3) { + if (!error) + nfsm_mtofh(dvp, newvp, v3, gotvp); + nfsm_wcc_data(dvp, wccflag); + } nfsm_reqdone; + if (newvp) + vput(newvp); FREE(cnp->cn_pnbuf, M_NAMEI); VTONFS(dvp)->n_flag |= NMODIFIED; - VTONFS(dvp)->n_attrstamp = 0; - vrele(dvp); + if (!wccflag) + VTONFS(dvp)->n_attrstamp = 0; + vput(dvp); /* * Kludge: Map EEXIST => 0 assuming that it is a reply to a retry. */ @@ -1444,76 +1799,78 @@ nfs_mkdir(ap) register struct vnode *dvp = ap->a_dvp; register struct vattr *vap = ap->a_vap; register struct componentname *cnp = ap->a_cnp; - register struct vnode **vpp = ap->a_vpp; register struct nfsv2_sattr *sp; + register struct nfsv3_sattr *sp3; register u_long *tl; register caddr_t cp; register long t1, t2; register int len; + struct nfsnode *np = (struct nfsnode *)0; + struct vnode *newvp = (struct vnode *)0; caddr_t bpos, dpos, cp2; - int error = 0, firsttry = 1, isnq; + nfsfh_t *fhp; + int error = 0, wccflag = NFSV3_WCCRATTR, attrflag; + int fhsize, gotvp = 0; struct mbuf *mreq, *mrep, *md, *mb, *mb2; struct vattr vattr; + int v3 = NFS_ISV3(dvp); - error = VOP_GETATTR(dvp, &vattr, cnp->cn_cred, cnp->cn_proc); - if (error) { + if (error = VOP_GETATTR(dvp, &vattr, cnp->cn_cred, cnp->cn_proc)) { VOP_ABORTOP(dvp, cnp); vput(dvp); return (error); } len = cnp->cn_namelen; - isnq = (VFSTONFS(dvp->v_mount)->nm_flag & NFSMNT_NQNFS); nfsstats.rpccnt[NFSPROC_MKDIR]++; nfsm_reqhead(dvp, NFSPROC_MKDIR, - NFSX_FH+NFSX_UNSIGNED+nfsm_rndup(len)+NFSX_SATTR(isnq)); - nfsm_fhtom(dvp); + NFSX_FH(v3) + NFSX_UNSIGNED + nfsm_rndup(len) + NFSX_SATTR(v3)); + nfsm_fhtom(dvp, v3); nfsm_strtom(cnp->cn_nameptr, len, NFS_MAXNAMLEN); - nfsm_build(sp, struct nfsv2_sattr *, NFSX_SATTR(isnq)); - sp->sa_mode = vtonfs_mode(VDIR, vap->va_mode); - sp->sa_uid = txdr_unsigned(cnp->cn_cred->cr_uid); - sp->sa_gid = txdr_unsigned(vattr.va_gid); - if (isnq) { - quad_t qval = -1; - - txdr_hyper(&qval, &sp->sa_nqsize); - sp->sa_nqflags = 0; - txdr_nqtime(&vap->va_atime, &sp->sa_nqatime); - txdr_nqtime(&vap->va_mtime, &sp->sa_nqmtime); + if (v3) { + nfsm_build(sp3, struct nfsv3_sattr *, NFSX_V3SRVSATTR); + nfsm_v3sattr(sp3, vap, cnp->cn_cred->cr_uid, vattr.va_gid); } else { - sp->sa_nfssize = -1; - txdr_nfstime(&vap->va_atime, &sp->sa_nfsatime); - txdr_nfstime(&vap->va_mtime, &sp->sa_nfsmtime); + nfsm_build(sp, struct nfsv2_sattr *, NFSX_V2SATTR); + sp->sa_mode = vtonfsv2_mode(VDIR, vap->va_mode); + sp->sa_uid = txdr_unsigned(cnp->cn_cred->cr_uid); + sp->sa_gid = txdr_unsigned(vattr.va_gid); + sp->sa_size = -1; + txdr_nfsv2time(&vap->va_atime, &sp->sa_atime); + txdr_nfsv2time(&vap->va_mtime, &sp->sa_mtime); } nfsm_request(dvp, NFSPROC_MKDIR, cnp->cn_proc, cnp->cn_cred); - nfsm_mtofh(dvp, *vpp); + if (!error) + nfsm_mtofh(dvp, newvp, v3, gotvp); + if (v3) + nfsm_wcc_data(dvp, wccflag); nfsm_reqdone; VTONFS(dvp)->n_flag |= NMODIFIED; - VTONFS(dvp)->n_attrstamp = 0; + if (!wccflag) + VTONFS(dvp)->n_attrstamp = 0; /* * Kludge: Map EEXIST => 0 assuming that you have a reply to a retry * if we can succeed in looking up the directory. - * "firsttry" is necessary since the macros may "goto nfsmout" which - * is above the if on errors. (Ugh) */ - if (error == EEXIST && firsttry) { - firsttry = 0; - error = 0; - nfsstats.rpccnt[NFSPROC_LOOKUP]++; - *vpp = NULL; - nfsm_reqhead(dvp, NFSPROC_LOOKUP, - NFSX_FH+NFSX_UNSIGNED+nfsm_rndup(len)); - nfsm_fhtom(dvp); - nfsm_strtom(cnp->cn_nameptr, len, NFS_MAXNAMLEN); - nfsm_request(dvp, NFSPROC_LOOKUP, cnp->cn_proc, cnp->cn_cred); - nfsm_mtofh(dvp, *vpp); - if ((*vpp)->v_type != VDIR) { - vput(*vpp); - error = EEXIST; + if (error == EEXIST || (!error && !gotvp)) { + if (newvp) { + vrele(newvp); + newvp = (struct vnode *)0; + } + error = nfs_lookitup(dvp, cnp->cn_nameptr, len, cnp->cn_cred, + cnp->cn_proc, &np); + if (!error) { + newvp = NFSTOV(np); + if (newvp->v_type != VDIR) + error = EEXIST; } - m_freem(mrep); } + if (error) { + if (newvp) + vrele(newvp); + } else + *ap->a_vpp = newvp; FREE(cnp->cn_pnbuf, M_NAMEI); - vrele(dvp); + vput(dvp); return (error); } @@ -1533,31 +1890,35 @@ nfs_rmdir(ap) register struct componentname *cnp = ap->a_cnp; register u_long *tl; register caddr_t cp; - register long t2; - caddr_t bpos, dpos; - int error = 0; + register long t1, t2; + caddr_t bpos, dpos, cp2; + int error = 0, wccflag = NFSV3_WCCRATTR; struct mbuf *mreq, *mrep, *md, *mb, *mb2; + int v3 = NFS_ISV3(dvp); if (dvp == vp) { - vrele(dvp); + vput(dvp); vrele(dvp); FREE(cnp->cn_pnbuf, M_NAMEI); return (EINVAL); } nfsstats.rpccnt[NFSPROC_RMDIR]++; nfsm_reqhead(dvp, NFSPROC_RMDIR, - NFSX_FH+NFSX_UNSIGNED+nfsm_rndup(cnp->cn_namelen)); - nfsm_fhtom(dvp); + NFSX_FH(v3) + NFSX_UNSIGNED + nfsm_rndup(cnp->cn_namelen)); + nfsm_fhtom(dvp, v3); nfsm_strtom(cnp->cn_nameptr, cnp->cn_namelen, NFS_MAXNAMLEN); nfsm_request(dvp, NFSPROC_RMDIR, cnp->cn_proc, cnp->cn_cred); + if (v3) + nfsm_wcc_data(dvp, wccflag); nfsm_reqdone; FREE(cnp->cn_pnbuf, M_NAMEI); VTONFS(dvp)->n_flag |= NMODIFIED; - VTONFS(dvp)->n_attrstamp = 0; + if (!wccflag) + VTONFS(dvp)->n_attrstamp = 0; cache_purge(dvp); cache_purge(vp); - vrele(vp); - vrele(dvp); + vput(vp); + vput(dvp); /* * Kludge: Map ENOENT => 0 assuming that you have a reply to a retry. */ @@ -1568,9 +1929,6 @@ nfs_rmdir(ap) /* * nfs readdir call - * Although cookie is defined as opaque, I translate it to/from net byte - * order so that it looks more sensible. This appears consistent with the - * Ultrix implementation of NFS. */ int nfs_readdir(ap) @@ -1591,10 +1949,10 @@ nfs_readdir(ap) /* * First, check for hit on the EOF offset cache */ - if (uio->uio_offset != 0 && uio->uio_offset == np->n_direofoffset && + if (np->n_direofoffset > 0 && uio->uio_offset >= np->n_direofoffset && (np->n_flag & NMODIFIED) == 0) { if (VFSTONFS(vp->v_mount)->nm_flag & NFSMNT_NQNFS) { - if (NQNFS_CKCACHABLE(vp, NQL_READ)) { + if (NQNFS_CKCACHABLE(vp, ND_READ)) { nfsstats.direofcache_hits++; return (0); } @@ -1622,90 +1980,148 @@ nfs_readdir(ap) */ int nfs_readdirrpc(vp, uiop, cred) - register struct vnode *vp; - struct uio *uiop; + struct vnode *vp; + register struct uio *uiop; struct ucred *cred; + { - register long len; - register struct dirent *dp = 0; + register int len, left; + register struct dirent *dp; register u_long *tl; register caddr_t cp; - register long t1; - long tlen, lastlen = 0; + register long t1, t2; + register nfsuint64 *cookiep; caddr_t bpos, dpos, cp2; - int error = 0; struct mbuf *mreq, *mrep, *md, *mb, *mb2; - struct mbuf *md2; - caddr_t dpos2; - int siz; - int more_dirs = 1; - u_long off, savoff = 0; - struct dirent *savdp = 0; - struct nfsmount *nmp; - struct nfsnode *np = VTONFS(vp); - long tresid; + nfsuint64 cookie; + struct nfsmount *nmp = VFSTONFS(vp->v_mount); + struct nfsnode *dnp = VTONFS(vp); + nfsfh_t *fhp; + u_quad_t frev, fileno; + int error = 0, tlen, more_dirs = 1, blksiz = 0, bigenough = 1, i; + int cachable, attrflag, fhsize; + int v3 = NFS_ISV3(vp); + +#ifndef nolint + dp = (struct dirent *)0; +#endif +#ifndef DIAGNOSTIC + if (uiop->uio_iovcnt != 1 || (uiop->uio_offset & (NFS_DIRBLKSIZ - 1)) || + (uiop->uio_resid & (NFS_DIRBLKSIZ - 1))) + panic("nfs readdirrpc bad uio"); +#endif - nmp = VFSTONFS(vp->v_mount); - tresid = uiop->uio_resid; /* - * Loop around doing readdir rpc's of size uio_resid or nm_rsize, - * whichever is smaller, truncated to a multiple of NFS_DIRBLKSIZ. + * If there is no cookie, assume end of directory. + */ + cookiep = nfs_getcookie(dnp, uiop->uio_offset, 0); + if (cookiep) + cookie = *cookiep; + else + return (0); + /* + * Loop around doing readdir rpc's of size nm_readdirsize + * truncated to a multiple of DIRBLKSIZ. * The stopping criteria is EOF or buffer full. */ - while (more_dirs && uiop->uio_resid >= NFS_DIRBLKSIZ) { + while (more_dirs && bigenough) { nfsstats.rpccnt[NFSPROC_READDIR]++; - nfsm_reqhead(vp, NFSPROC_READDIR, - NFSX_FH + 2 * NFSX_UNSIGNED); - nfsm_fhtom(vp); - nfsm_build(tl, u_long *, 2 * NFSX_UNSIGNED); - off = (u_long)uiop->uio_offset; - *tl++ = txdr_unsigned(off); - *tl = txdr_unsigned(((uiop->uio_resid > nmp->nm_rsize) ? - nmp->nm_rsize : uiop->uio_resid) & ~(NFS_DIRBLKSIZ-1)); + nfsm_reqhead(vp, NFSPROC_READDIR, NFSX_FH(v3) + + NFSX_READDIR(v3)); + nfsm_fhtom(vp, v3); + if (v3) { + nfsm_build(tl, u_long *, 5 * NFSX_UNSIGNED); + *tl++ = cookie.nfsuquad[0]; + *tl++ = cookie.nfsuquad[1]; + *tl++ = dnp->n_cookieverf.nfsuquad[0]; + *tl++ = dnp->n_cookieverf.nfsuquad[1]; + } else { + nfsm_build(tl, u_long *, 2 * NFSX_UNSIGNED); + *tl++ = cookie.nfsuquad[0]; + } + *tl = txdr_unsigned(nmp->nm_readdirsize); nfsm_request(vp, NFSPROC_READDIR, uiop->uio_procp, cred); - siz = 0; + if (v3) { + nfsm_postop_attr(vp, attrflag); + if (!error) { + nfsm_dissect(tl, u_long *, 2 * NFSX_UNSIGNED); + dnp->n_cookieverf.nfsuquad[0] = *tl++; + dnp->n_cookieverf.nfsuquad[1] = *tl; + } else { + m_freem(mrep); + goto nfsmout; + } + } nfsm_dissect(tl, u_long *, NFSX_UNSIGNED); more_dirs = fxdr_unsigned(int, *tl); - /* Save the position so that we can do nfsm_mtouio() later */ - dpos2 = dpos; - md2 = md; - /* loop thru the dir entries, doctoring them to 4bsd form */ -#ifdef lint - dp = (struct dirent *)0; -#endif /* lint */ - while (more_dirs && siz < uiop->uio_resid) { - savoff = off; /* Hold onto offset and dp */ - savdp = dp; - nfsm_dissect(tl, u_long *, 2 * NFSX_UNSIGNED); - dp = (struct dirent *)tl; - dp->d_fileno = fxdr_unsigned(u_long, *tl++); - len = fxdr_unsigned(int, *tl); + while (more_dirs && bigenough) { + if (v3) { + nfsm_dissect(tl, u_long *, 3 * NFSX_UNSIGNED); + fxdr_hyper(tl, &fileno); + len = fxdr_unsigned(int, *(tl + 2)); + } else { + nfsm_dissect(tl, u_long *, 2 * NFSX_UNSIGNED); + fileno = fxdr_unsigned(u_quad_t, *tl++); + len = fxdr_unsigned(int, *tl); + } if (len <= 0 || len > NFS_MAXNAMLEN) { error = EBADRPC; m_freem(mrep); goto nfsmout; } - dp->d_namlen = (u_char)len; - dp->d_type = DT_UNKNOWN; - nfsm_adv(len); /* Point past name */ tlen = nfsm_rndup(len); - /* - * This should not be necessary, but some servers have - * broken XDR such that these bytes are not null filled. - */ - if (tlen != len) { - *dpos = '\0'; /* Null-terminate */ - nfsm_adv(tlen - len); - len = tlen; + if (tlen == len) + tlen += 4; /* To ensure null termination */ + left = DIRBLKSIZ - blksiz; + if ((tlen + DIRHDSIZ) > left) { + dp->d_reclen += left; + uiop->uio_iov->iov_base += left; + uiop->uio_iov->iov_len -= left; + uiop->uio_offset += left; + uiop->uio_resid -= left; + blksiz = 0; } - nfsm_dissect(tl, u_long *, 2 * NFSX_UNSIGNED); - off = fxdr_unsigned(u_long, *tl); - *tl++ = 0; /* Ensures null termination of name */ + if ((tlen + DIRHDSIZ) > uiop->uio_resid) + bigenough = 0; + if (bigenough) { + dp = (struct dirent *)uiop->uio_iov->iov_base; + dp->d_fileno = (int)fileno; + dp->d_namlen = len; + dp->d_reclen = tlen + DIRHDSIZ; + dp->d_type = DT_UNKNOWN; + blksiz += dp->d_reclen; + if (blksiz == DIRBLKSIZ) + blksiz = 0; + uiop->uio_offset += DIRHDSIZ; + uiop->uio_resid -= DIRHDSIZ; + uiop->uio_iov->iov_base += DIRHDSIZ; + uiop->uio_iov->iov_len -= DIRHDSIZ; + nfsm_mtouio(uiop, len); + cp = uiop->uio_iov->iov_base; + tlen -= len; + *cp = '\0'; /* null terminate */ + uiop->uio_iov->iov_base += tlen; + uiop->uio_iov->iov_len -= tlen; + uiop->uio_offset += tlen; + uiop->uio_resid -= tlen; + } else + nfsm_adv(nfsm_rndup(len)); + if (v3) { + nfsm_dissect(tl, u_long *, 3 * NFSX_UNSIGNED); + } else { + nfsm_dissect(tl, u_long *, 2 * NFSX_UNSIGNED); + } + if (bigenough) { + cookie.nfsuquad[0] = *tl++; + if (v3) + cookie.nfsuquad[1] = *tl++; + } else if (v3) + tl += 2; + else + tl++; more_dirs = fxdr_unsigned(int, *tl); - dp->d_reclen = len + 4 * NFSX_UNSIGNED; - siz += dp->d_reclen; } /* * If at end of rpc data, get the eof boolean @@ -1713,189 +2129,224 @@ nfs_readdirrpc(vp, uiop, cred) if (!more_dirs) { nfsm_dissect(tl, u_long *, NFSX_UNSIGNED); more_dirs = (fxdr_unsigned(int, *tl) == 0); - - /* - * If at EOF, cache directory offset - */ - if (!more_dirs) - np->n_direofoffset = off; - } - /* - * If there is too much to fit in the data buffer, use savoff and - * savdp to trim off the last record. - * --> we are not at eof - */ - if (siz > uiop->uio_resid) { - off = savoff; - siz -= dp->d_reclen; - dp = savdp; - more_dirs = 0; /* Paranoia */ } - if (siz > 0) { - lastlen = dp->d_reclen; - md = md2; - dpos = dpos2; - nfsm_mtouio(uiop, siz); - uiop->uio_offset = (off_t)off; - } else - more_dirs = 0; /* Ugh, never happens, but in case.. */ m_freem(mrep); } /* - * Fill last record, iff any, out to a multiple of NFS_DIRBLKSIZ + * Fill last record, iff any, out to a multiple of DIRBLKSIZ * by increasing d_reclen for the last record. */ - if (uiop->uio_resid < tresid) { - len = uiop->uio_resid & (NFS_DIRBLKSIZ - 1); - if (len > 0) { - dp = (struct dirent *) - (uiop->uio_iov->iov_base - lastlen); - dp->d_reclen += len; - uiop->uio_iov->iov_base += len; - uiop->uio_iov->iov_len -= len; - uiop->uio_resid -= len; - } + if (blksiz > 0) { + left = DIRBLKSIZ - blksiz; + dp->d_reclen += left; + uiop->uio_iov->iov_base += left; + uiop->uio_iov->iov_len -= left; + uiop->uio_offset += left; + uiop->uio_resid -= left; + } + + /* + * We are now either at the end of the directory or have filled the + * block. + */ + if (bigenough) + dnp->n_direofoffset = uiop->uio_offset; + else { + if (uiop->uio_resid > 0) + printf("EEK! readdirrpc resid > 0\n"); + cookiep = nfs_getcookie(dnp, uiop->uio_offset, 1); + *cookiep = cookie; } nfsmout: return (error); } /* - * Nqnfs readdir_and_lookup RPC. Used in place of nfs_readdirrpc(). + * NFS V3 readdir plus RPC. Used in place of nfs_readdirrpc(). */ int -nfs_readdirlookrpc(vp, uiop, cred) +nfs_readdirplusrpc(vp, uiop, cred) struct vnode *vp; register struct uio *uiop; struct ucred *cred; { - register int len; - register struct dirent *dp = 0; + register int len, left; + register struct dirent *dp; register u_long *tl; register caddr_t cp; - register long t1; - caddr_t bpos, dpos, cp2; - struct mbuf *mreq, *mrep, *md, *mb, *mb2; + register long t1, t2; + register struct vnode *newvp; + register nfsuint64 *cookiep; + caddr_t bpos, dpos, cp2, dpossav1, dpossav2; + struct mbuf *mreq, *mrep, *md, *mb, *mb2, *mdsav1, *mdsav2; struct nameidata nami, *ndp = &nami; struct componentname *cnp = &ndp->ni_cnd; - u_long off, endoff = 0, fileno; - time_t reqtime, ltime = 0; - struct nfsmount *nmp; - struct nfsnode *np; - struct vnode *newvp; - nfsv2fh_t *fhp; - u_quad_t frev; - int error = 0, tlen, more_dirs = 1, tresid, doit, bigenough, i; - int cachable = 0; - - if (uiop->uio_iovcnt != 1) - panic("nfs rdirlook"); - nmp = VFSTONFS(vp->v_mount); - tresid = uiop->uio_resid; + nfsuint64 cookie; + struct nfsmount *nmp = VFSTONFS(vp->v_mount); + struct nfsnode *dnp = VTONFS(vp), *np; + nfsfh_t *fhp; + u_quad_t frev, fileno; + int error = 0, tlen, more_dirs = 1, blksiz = 0, doit, bigenough = 1, i; + int cachable, attrflag, fhsize; + +#ifndef nolint + dp = (struct dirent *)0; +#endif +#ifndef DIAGNOSTIC + if (uiop->uio_iovcnt != 1 || (uiop->uio_offset & (DIRBLKSIZ - 1)) || + (uiop->uio_resid & (DIRBLKSIZ - 1))) + panic("nfs readdirplusrpc bad uio"); +#endif ndp->ni_dvp = vp; newvp = NULLVP; + /* - * Loop around doing readdir rpc's of size uio_resid or nm_rsize, - * whichever is smaller, truncated to a multiple of NFS_DIRBLKSIZ. + * If there is no cookie, assume end of directory. + */ + cookiep = nfs_getcookie(dnp, uiop->uio_offset, 0); + if (cookiep) + cookie = *cookiep; + else + return (0); + /* + * Loop around doing readdir rpc's of size nm_readdirsize + * truncated to a multiple of DIRBLKSIZ. * The stopping criteria is EOF or buffer full. */ - while (more_dirs && uiop->uio_resid >= NFS_DIRBLKSIZ) { - nfsstats.rpccnt[NQNFSPROC_READDIRLOOK]++; - nfsm_reqhead(vp, NQNFSPROC_READDIRLOOK, - NFSX_FH + 3 * NFSX_UNSIGNED); - nfsm_fhtom(vp); - nfsm_build(tl, u_long *, 3 * NFSX_UNSIGNED); - off = (u_long)uiop->uio_offset; - *tl++ = txdr_unsigned(off); - *tl++ = txdr_unsigned(((uiop->uio_resid > nmp->nm_rsize) ? - nmp->nm_rsize : uiop->uio_resid) & ~(NFS_DIRBLKSIZ-1)); - if (nmp->nm_flag & NFSMNT_NQLOOKLEASE) - *tl = txdr_unsigned(nmp->nm_leaseterm); - else - *tl = 0; - reqtime = time.tv_sec; - nfsm_request(vp, NQNFSPROC_READDIRLOOK, uiop->uio_procp, cred); - nfsm_dissect(tl, u_long *, NFSX_UNSIGNED); + while (more_dirs && bigenough) { + nfsstats.rpccnt[NFSPROC_READDIRPLUS]++; + nfsm_reqhead(vp, NFSPROC_READDIRPLUS, + NFSX_FH(1) + 6 * NFSX_UNSIGNED); + nfsm_fhtom(vp, 1); + nfsm_build(tl, u_long *, 6 * NFSX_UNSIGNED); + *tl++ = cookie.nfsuquad[0]; + *tl++ = cookie.nfsuquad[1]; + *tl++ = dnp->n_cookieverf.nfsuquad[0]; + *tl++ = dnp->n_cookieverf.nfsuquad[1]; + *tl++ = txdr_unsigned(nmp->nm_readdirsize); + *tl = txdr_unsigned(nmp->nm_rsize); + nfsm_request(vp, NFSPROC_READDIRPLUS, uiop->uio_procp, cred); + nfsm_postop_attr(vp, attrflag); + if (error) { + m_freem(mrep); + goto nfsmout; + } + nfsm_dissect(tl, u_long *, 3 * NFSX_UNSIGNED); + dnp->n_cookieverf.nfsuquad[0] = *tl++; + dnp->n_cookieverf.nfsuquad[1] = *tl++; more_dirs = fxdr_unsigned(int, *tl); /* loop thru the dir entries, doctoring them to 4bsd form */ - bigenough = 1; while (more_dirs && bigenough) { - doit = 1; - nfsm_dissect(tl, u_long *, 4 * NFSX_UNSIGNED); - if (nmp->nm_flag & NFSMNT_NQLOOKLEASE) { - cachable = fxdr_unsigned(int, *tl++); - ltime = reqtime + fxdr_unsigned(int, *tl++); - fxdr_hyper(tl, &frev); - } - nfsm_dissect(fhp, nfsv2fh_t *, NFSX_FH); - if (!bcmp(VTONFS(vp)->n_fh.fh_bytes, (caddr_t)fhp, NFSX_FH)) { - VREF(vp); - newvp = vp; - np = VTONFS(vp); - } else { - error = nfs_nget(vp->v_mount, fhp, &np); - if (error) - doit = 0; - newvp = NFSTOV(np); - } - error = nfs_loadattrcache(&newvp, &md, &dpos, - (struct vattr *)0); - if (error) - doit = 0; - nfsm_dissect(tl, u_long *, 2 * NFSX_UNSIGNED); - fileno = fxdr_unsigned(u_long, *tl++); - len = fxdr_unsigned(int, *tl); + nfsm_dissect(tl, u_long *, 3 * NFSX_UNSIGNED); + fxdr_hyper(tl, &fileno); + len = fxdr_unsigned(int, *(tl + 2)); if (len <= 0 || len > NFS_MAXNAMLEN) { error = EBADRPC; m_freem(mrep); goto nfsmout; } - tlen = (len + 4) & ~0x3; + tlen = nfsm_rndup(len); + if (tlen == len) + tlen += 4; /* To ensure null termination*/ + left = DIRBLKSIZ - blksiz; + if ((tlen + DIRHDSIZ) > left) { + dp->d_reclen += left; + uiop->uio_iov->iov_base += left; + uiop->uio_iov->iov_len -= left; + uiop->uio_offset += left; + uiop->uio_resid -= left; + blksiz = 0; + } if ((tlen + DIRHDSIZ) > uiop->uio_resid) bigenough = 0; - if (bigenough && doit) { + if (bigenough) { dp = (struct dirent *)uiop->uio_iov->iov_base; - dp->d_fileno = fileno; + dp->d_fileno = (int)fileno; dp->d_namlen = len; dp->d_reclen = tlen + DIRHDSIZ; - dp->d_type = - IFTODT(VTTOIF(np->n_vattr.va_type)); + dp->d_type = DT_UNKNOWN; + blksiz += dp->d_reclen; + if (blksiz == DIRBLKSIZ) + blksiz = 0; + uiop->uio_offset += DIRHDSIZ; uiop->uio_resid -= DIRHDSIZ; uiop->uio_iov->iov_base += DIRHDSIZ; uiop->uio_iov->iov_len -= DIRHDSIZ; cnp->cn_nameptr = uiop->uio_iov->iov_base; cnp->cn_namelen = len; - ndp->ni_vp = newvp; nfsm_mtouio(uiop, len); cp = uiop->uio_iov->iov_base; tlen -= len; - for (i = 0; i < tlen; i++) - *cp++ = '\0'; + *cp = '\0'; uiop->uio_iov->iov_base += tlen; uiop->uio_iov->iov_len -= tlen; + uiop->uio_offset += tlen; uiop->uio_resid -= tlen; + } else + nfsm_adv(nfsm_rndup(len)); + nfsm_dissect(tl, u_long *, 3 * NFSX_UNSIGNED); + if (bigenough) { + cookie.nfsuquad[0] = *tl++; + cookie.nfsuquad[1] = *tl++; + } else + tl += 2; + + /* + * Since the attributes are before the file handle + * (sigh), we must skip over the attributes and then + * come back and get them. + */ + attrflag = fxdr_unsigned(int, *tl); + if (attrflag) { + dpossav1 = dpos; + mdsav1 = md; + nfsm_adv(NFSX_V3FATTR); + nfsm_dissect(tl, u_long *, NFSX_UNSIGNED); + doit = fxdr_unsigned(int, *tl); + if (doit) { + nfsm_getfh(fhp, fhsize, 1); + if (NFS_CMPFH(dnp, fhp, fhsize)) { + VREF(vp); + newvp = vp; + np = dnp; + } else { + if (error = nfs_nget(vp->v_mount, fhp, + fhsize, &np)) + doit = 0; + else + newvp = NFSTOV(np); + } + } + if (doit) { + dpossav2 = dpos; + dpos = dpossav1; + mdsav2 = md; + md = mdsav1; + nfsm_loadattr(newvp, (struct vattr *)0); + dpos = dpossav2; + md = mdsav2; + dp->d_type = + IFTODT(VTTOIF(np->n_vattr.va_type)); + ndp->ni_vp = newvp; cnp->cn_hash = 0; - for (cp = cnp->cn_nameptr, i = 1; i <= len; i++, cp++) - cnp->cn_hash += (unsigned char)*cp * i; - if ((nmp->nm_flag & NFSMNT_NQLOOKLEASE) && - ltime > time.tv_sec) - nqnfs_clientlease(nmp, np, NQL_READ, - cachable, ltime, frev); + for (cp = cnp->cn_nameptr, i = 1; i <= len; + i++, cp++) + cnp->cn_hash += (unsigned char)*cp * i; if (cnp->cn_namelen <= NCHNAMLEN) cache_enter(ndp->ni_dvp, ndp->ni_vp, cnp); + } } else { - nfsm_adv(nfsm_rndup(len)); + /* Just skip over the file handle */ + nfsm_dissect(tl, u_long *, NFSX_UNSIGNED); + i = fxdr_unsigned(int, *tl); + nfsm_adv(nfsm_rndup(i)); } if (newvp != NULLVP) { - vrele(newvp); - newvp = NULLVP; + vrele(newvp); + newvp = NULLVP; } - nfsm_dissect(tl, u_long *, 2 * NFSX_UNSIGNED); - if (bigenough) - endoff = off = fxdr_unsigned(u_long, *tl++); - else - endoff = fxdr_unsigned(u_long, *tl++); + nfsm_dissect(tl, u_long *, NFSX_UNSIGNED); more_dirs = fxdr_unsigned(int, *tl); } /* @@ -1904,35 +2355,42 @@ nfs_readdirlookrpc(vp, uiop, cred) if (!more_dirs) { nfsm_dissect(tl, u_long *, NFSX_UNSIGNED); more_dirs = (fxdr_unsigned(int, *tl) == 0); - - /* - * If at EOF, cache directory offset - */ - if (!more_dirs) - VTONFS(vp)->n_direofoffset = endoff; } - if (uiop->uio_resid < tresid) - uiop->uio_offset = (off_t)off; - else - more_dirs = 0; m_freem(mrep); } /* * Fill last record, iff any, out to a multiple of NFS_DIRBLKSIZ * by increasing d_reclen for the last record. */ - if (uiop->uio_resid < tresid) { - len = uiop->uio_resid & (NFS_DIRBLKSIZ - 1); - if (len > 0) { - dp->d_reclen += len; - uiop->uio_iov->iov_base += len; - uiop->uio_iov->iov_len -= len; - uiop->uio_resid -= len; - } + if (blksiz > 0) { + left = DIRBLKSIZ - blksiz; + dp->d_reclen += left; + uiop->uio_iov->iov_base += left; + uiop->uio_iov->iov_len -= left; + uiop->uio_offset += left; + uiop->uio_resid -= left; + } + + /* + * We are now either at the end of the directory or have filled the + * block. + */ + if (bigenough) + dnp->n_direofoffset = uiop->uio_offset; + else { + if (uiop->uio_resid > 0) + printf("EEK! readdirplusrpc resid > 0\n"); + cookiep = nfs_getcookie(dnp, uiop->uio_offset, 1); + *cookiep = cookie; } nfsmout: - if (newvp != NULLVP) - vrele(newvp); + if (newvp != NULLVP) { + if (newvp == vp) + vrele(newvp); + else + vput(newvp); + newvp = NULLVP; + } return (error); } static char hextoasc[] = "0123456789abcdef"; @@ -1950,19 +2408,19 @@ nfs_sillyrename(dvp, vp, cnp) struct vnode *dvp, *vp; struct componentname *cnp; { - register struct nfsnode *np; register struct sillyrename *sp; + struct nfsnode *np; int error; short pid; cache_purge(dvp); np = VTONFS(vp); -#ifdef SILLYSEPARATE +#ifndef DIAGNOSTIC + if (vp->v_type == VDIR) + panic("nfs: sillyrename dir"); +#endif MALLOC(sp, struct sillyrename *, sizeof (struct sillyrename), M_NFSREQ, M_WAITOK); -#else - sp = &np->n_silly; -#endif sp->s_cred = crdup(cnp->cn_cred); sp->s_dvp = dvp; VREF(dvp); @@ -1977,65 +2435,149 @@ nfs_sillyrename(dvp, vp, cnp) sp->s_name[5] = hextoasc[(pid >> 12) & 0xf]; /* Try lookitups until we get one that isn't there */ - while (nfs_lookitup(sp, (nfsv2fh_t *)0, cnp->cn_proc) == 0) { + while (nfs_lookitup(dvp, sp->s_name, sp->s_namlen, sp->s_cred, + cnp->cn_proc, (struct nfsnode **)0) == 0) { sp->s_name[4]++; if (sp->s_name[4] > 'z') { error = EINVAL; goto bad; } } - error = nfs_renameit(dvp, cnp, sp); - if (error) + if (error = nfs_renameit(dvp, cnp, sp)) goto bad; - nfs_lookitup(sp, &np->n_fh, cnp->cn_proc); + error = nfs_lookitup(dvp, sp->s_name, sp->s_namlen, sp->s_cred, + cnp->cn_proc, &np); np->n_sillyrename = sp; return (0); bad: vrele(sp->s_dvp); crfree(sp->s_cred); -#ifdef SILLYSEPARATE free((caddr_t)sp, M_NFSREQ); -#endif return (error); } /* - * Look up a file name for silly rename stuff. - * Just like nfs_lookup() except that it doesn't load returned values - * into the nfsnode table. - * If fhp != NULL it copies the returned file handle out + * Look up a file name and optionally either update the file handle or + * allocate an nfsnode, depending on the value of npp. + * npp == NULL --> just do the lookup + * *npp == NULL --> allocate a new nfsnode and make sure attributes are + * handled too + * *npp != NULL --> update the file handle in the vnode */ int -nfs_lookitup(sp, fhp, procp) - register struct sillyrename *sp; - nfsv2fh_t *fhp; +nfs_lookitup(dvp, name, len, cred, procp, npp) + register struct vnode *dvp; + char *name; + int len; + struct ucred *cred; struct proc *procp; + struct nfsnode **npp; { - register struct vnode *vp = sp->s_dvp; register u_long *tl; register caddr_t cp; register long t1, t2; + struct vnode *newvp = (struct vnode *)0; + struct nfsnode *np, *dnp = VTONFS(dvp); caddr_t bpos, dpos, cp2; - int error = 0, isnq; + int error = 0, fhlen, attrflag; struct mbuf *mreq, *mrep, *md, *mb, *mb2; - long len; + nfsfh_t *nfhp; + int v3 = NFS_ISV3(dvp); - isnq = (VFSTONFS(vp->v_mount)->nm_flag & NFSMNT_NQNFS); nfsstats.rpccnt[NFSPROC_LOOKUP]++; - len = sp->s_namlen; - nfsm_reqhead(vp, NFSPROC_LOOKUP, NFSX_FH+NFSX_UNSIGNED+nfsm_rndup(len)); - if (isnq) { - nfsm_build(tl, u_long *, NFSX_UNSIGNED); - *tl = 0; + nfsm_reqhead(dvp, NFSPROC_LOOKUP, + NFSX_FH(v3) + NFSX_UNSIGNED + nfsm_rndup(len)); + nfsm_fhtom(dvp, v3); + nfsm_strtom(name, len, NFS_MAXNAMLEN); + nfsm_request(dvp, NFSPROC_LOOKUP, procp, cred); + if (npp && !error) { + nfsm_getfh(nfhp, fhlen, v3); + if (*npp) { + np = *npp; + if (np->n_fhsize > NFS_SMALLFH && fhlen <= NFS_SMALLFH) { + free((caddr_t)np->n_fhp, M_NFSBIGFH); + np->n_fhp = &np->n_fh; + } else if (np->n_fhsize <= NFS_SMALLFH && fhlen>NFS_SMALLFH) + np->n_fhp =(nfsfh_t *)malloc(fhlen,M_NFSBIGFH,M_WAITOK); + bcopy((caddr_t)nfhp, (caddr_t)np->n_fhp, fhlen); + np->n_fhsize = fhlen; + newvp = NFSTOV(np); + } else if (NFS_CMPFH(dnp, nfhp, fhlen)) { + VREF(dvp); + newvp = dvp; + } else { + error = nfs_nget(dvp->v_mount, nfhp, fhlen, &np); + if (error) { + m_freem(mrep); + return (error); + } + newvp = NFSTOV(np); + } + if (v3) { + nfsm_postop_attr(newvp, attrflag); + if (!attrflag && *npp == NULL) { + m_freem(mrep); + if (newvp == dvp) + vrele(newvp); + else + vput(newvp); + return (ENOENT); + } + } else + nfsm_loadattr(newvp, (struct vattr *)0); } - nfsm_fhtom(vp); - nfsm_strtom(sp->s_name, len, NFS_MAXNAMLEN); - nfsm_request(vp, NFSPROC_LOOKUP, procp, sp->s_cred); - if (fhp != NULL) { - if (isnq) - nfsm_dissect(tl, u_long *, NFSX_UNSIGNED); - nfsm_dissect(cp, caddr_t, NFSX_FH); - bcopy(cp, (caddr_t)fhp, NFSX_FH); + nfsm_reqdone; + if (npp && *npp == NULL) { + if (error) { + if (newvp) + if (newvp == dvp) + vrele(newvp); + else + vput(newvp); + } else + *npp = np; + } + return (error); +} + +/* + * Nfs Version 3 commit rpc + */ +int +nfs_commit(vp, offset, cnt, cred, procp) + register struct vnode *vp; + u_quad_t offset; + int cnt; + struct ucred *cred; + struct proc *procp; +{ + register caddr_t cp; + register u_long *tl; + register int t1, t2; + register struct nfsmount *nmp = VFSTONFS(vp->v_mount); + caddr_t bpos, dpos, cp2; + int error = 0, wccflag = NFSV3_WCCRATTR; + struct mbuf *mreq, *mrep, *md, *mb, *mb2; + + if ((nmp->nm_flag & NFSMNT_HASWRITEVERF) == 0) + return (0); + nfsstats.rpccnt[NFSPROC_COMMIT]++; + nfsm_reqhead(vp, NFSPROC_COMMIT, NFSX_FH(1)); + nfsm_fhtom(vp, 1); + nfsm_build(tl, u_long *, 3 * NFSX_UNSIGNED); + txdr_hyper(&offset, tl); + tl += 2; + *tl = txdr_unsigned(cnt); + nfsm_request(vp, NFSPROC_COMMIT, procp, cred); + nfsm_wcc_data(vp, wccflag); + if (!error) { + nfsm_dissect(tl, u_long *, NFSX_V3WRITEVERF); + if (bcmp((caddr_t)nmp->nm_verf, (caddr_t)tl, + NFSX_V3WRITEVERF)) { + bcopy((caddr_t)tl, (caddr_t)nmp->nm_verf, + NFSX_V3WRITEVERF); + error = NFSERR_STALEWRITEVERF; + } } nfsm_reqdone; return (error); @@ -2044,9 +2586,7 @@ nfs_lookitup(sp, fhp, procp) /* * Kludge City.. * - make nfs_bmap() essentially a no-op that does no translation - * - do nfs_strategy() by faking physical I/O with nfs_readrpc/nfs_writerpc - * after mapping the physical addresses into Kernel Virtual space in the - * nfsiobuf area. + * - do nfs_strategy() by doing I/O with nfs_readrpc/nfs_writerpc * (Maybe I could use the process's page mapping, but I was concerned that * Kernel Write might not be enabled and also figured copyout() would do * a lot more work than bcopy() and also it currently happens in the @@ -2088,8 +2628,8 @@ nfs_strategy(ap) struct proc *p; int error = 0; - if ((bp->b_flags & (B_PHYS|B_ASYNC)) == (B_PHYS|B_ASYNC)) - panic("nfs physio/async"); + if (bp->b_flags & B_PHYS) + panic("nfs physio"); if (bp->b_flags & B_ASYNC) p = (struct proc *)0; else @@ -2129,9 +2669,7 @@ nfs_mmap(ap) } /* - * Flush all the blocks associated with a vnode. - * Walk through the buffer pool and push any dirty pages - * associated with the vnode. + * fsync vnode op. Just call nfs_flush() with commit == 1. */ /* ARGSUSED */ int @@ -2144,29 +2682,124 @@ nfs_fsync(ap) struct proc * a_p; } */ *ap; { - register struct vnode *vp = ap->a_vp; + + return (nfs_flush(ap->a_vp, ap->a_cred, ap->a_waitfor, ap->a_p, 1)); +} + +/* + * Flush all the blocks associated with a vnode. + * Walk through the buffer pool and push any dirty pages + * associated with the vnode. + */ +int +nfs_flush(vp, cred, waitfor, p, commit) + register struct vnode *vp; + struct ucred *cred; + int waitfor; + struct proc *p; + int commit; +{ register struct nfsnode *np = VTONFS(vp); register struct buf *bp; + register int i; struct buf *nbp; - struct nfsmount *nmp; - int s, error = 0, slptimeo = 0, slpflag = 0; + struct nfsmount *nmp = VFSTONFS(vp->v_mount); + int s, error = 0, slptimeo = 0, slpflag = 0, retv, bvecpos; + int passone = 1; + u_quad_t off = (u_quad_t)-1, endoff = 0, toff; +#ifndef NFS_COMMITBVECSIZ +#define NFS_COMMITBVECSIZ 20 +#endif + struct buf *bvec[NFS_COMMITBVECSIZ]; - nmp = VFSTONFS(vp->v_mount); if (nmp->nm_flag & NFSMNT_INT) slpflag = PCATCH; + if (!commit) + passone = 0; + /* + * A b_flags == (B_DELWRI | B_NEEDCOMMIT) block has been written to the + * server, but nas not been committed to stable storage on the server + * yet. On the first pass, the byte range is worked out and the commit + * rpc is done. On the second pass, nfs_writebp() is called to do the + * job. + */ +again: + bvecpos = 0; + if (NFS_ISV3(vp) && commit) { + s = splbio(); + for (bp = vp->v_dirtyblkhd.lh_first; bp; bp = nbp) { + nbp = bp->b_vnbufs.le_next; + if (bvecpos >= NFS_COMMITBVECSIZ) + break; + if ((bp->b_flags & (B_BUSY | B_DELWRI | B_NEEDCOMMIT)) + != (B_DELWRI | B_NEEDCOMMIT)) + continue; + bremfree(bp); + bp->b_flags |= (B_BUSY | B_WRITEINPROG); + vfs_busy_pages(bp, 1); + /* + * A list of these buffers is kept so that the + * second loop knows which buffers have actually + * been committed. This is necessary, since there + * may be a race between the commit rpc and new + * uncommitted writes on the file. + */ + bvec[bvecpos++] = bp; + toff = ((u_quad_t)bp->b_blkno) * DEV_BSIZE + + bp->b_dirtyoff; + if (toff < off) + off = toff; + toff += (u_quad_t)(bp->b_dirtyend - bp->b_dirtyoff); + if (toff > endoff) + endoff = toff; + } + splx(s); + } + if (bvecpos > 0) { + /* + * Commit data on the server, as required. + */ + retv = nfs_commit(vp, off, (int)(endoff - off), cred, p); + if (retv == NFSERR_STALEWRITEVERF) + nfs_clearcommit(vp->v_mount); + /* + * Now, either mark the blocks I/O done or mark the + * blocks dirty, depending on whether the commit + * succeeded. + */ + for (i = 0; i < bvecpos; i++) { + bp = bvec[i]; + bp->b_flags &= ~(B_NEEDCOMMIT | B_WRITEINPROG); + if (retv) { + brelse(bp); + vfs_unbusy_pages(bp); + } else { + vp->v_numoutput++; + bp->b_flags |= B_ASYNC; + bp->b_flags &= ~(B_READ|B_DONE|B_ERROR|B_DELWRI); + bp->b_dirtyoff = bp->b_dirtyend = 0; + reassignbuf(bp, vp); + biodone(bp); + } + } + } + + /* + * Start/do any write(s) that are required. + */ loop: s = splbio(); for (bp = vp->v_dirtyblkhd.lh_first; bp; bp = nbp) { nbp = bp->b_vnbufs.le_next; if (bp->b_flags & B_BUSY) { - if (ap->a_waitfor != MNT_WAIT) + if (waitfor != MNT_WAIT || passone) continue; bp->b_flags |= B_WANTED; error = tsleep((caddr_t)bp, slpflag | (PRIBIO + 1), "nfsfsync", slptimeo); splx(s); if (error) { - if (nfs_sigintr(nmp, (struct nfsreq *)0, ap->a_p)) + if (nfs_sigintr(nmp, (struct nfsreq *)0, p)) return (EINTR); if (slpflag == PCATCH) { slpflag = 0; @@ -2177,21 +2810,29 @@ loop: } if ((bp->b_flags & B_DELWRI) == 0) panic("nfs_fsync: not dirty"); + if ((passone || !commit) && (bp->b_flags & B_NEEDCOMMIT)) + continue; bremfree(bp); - bp->b_flags |= B_BUSY; + if (passone || !commit) + bp->b_flags |= (B_BUSY|B_ASYNC); + else + bp->b_flags |= (B_BUSY|B_ASYNC|B_WRITEINPROG|B_NEEDCOMMIT); splx(s); - bp->b_flags |= B_ASYNC; VOP_BWRITE(bp); goto loop; } splx(s); - if (ap->a_waitfor == MNT_WAIT) { + if (passone) { + passone = 0; + goto again; + } + if (waitfor == MNT_WAIT) { while (vp->v_numoutput) { vp->v_flag |= VBWAIT; error = tsleep((caddr_t)&vp->v_numoutput, slpflag | (PRIBIO + 1), "nfsfsync", slptimeo); if (error) { - if (nfs_sigintr(nmp, (struct nfsreq *)0, ap->a_p)) + if (nfs_sigintr(nmp, (struct nfsreq *)0, p)) return (EINTR); if (slpflag == PCATCH) { slpflag = 0; @@ -2199,7 +2840,7 @@ loop: } } } - if (vp->v_dirtyblkhd.lh_first) { + if (vp->v_dirtyblkhd.lh_first && commit) { goto loop; } } @@ -2207,14 +2848,15 @@ loop: error = np->n_error; np->n_flag &= ~NWRITEERR; } + np->n_flag &= ~NMODIFIED; return (error); } /* * Return POSIX pathconf information applicable to nfs. * - * Currently the NFS protocol does not support getting such - * information from the remote server. + * The NFS V2 protocol doesn't support this, so just return EINVAL + * for V2. */ /* ARGSUSED */ int @@ -2243,6 +2885,7 @@ nfs_advlock(ap) int a_flags; } */ *ap; { +#ifdef __FreeBSD__ register struct nfsnode *np = VTONFS(ap->a_vp); /* @@ -2251,6 +2894,9 @@ nfs_advlock(ap) * that this is a local lock. */ return (lf_advlock(ap, &(np->n_lockf), np->n_size)); +#else + return (EOPNOTSUPP); +#endif } /* @@ -2363,6 +3009,81 @@ nfs_update(ap) } /* + * Just call nfs_writebp() with the force argument set to 1. + */ +int +nfs_bwrite(ap) + struct vop_bwrite_args /* { + struct vnode *a_bp; + } */ *ap; +{ + + return (nfs_writebp(ap->a_bp, 1)); +} + +/* + * This is a clone of vn_bwrite(), except that B_WRITEINPROG isn't set unless + * the force flag is one and it also handles the B_NEEDCOMMIT flag. + */ +int +nfs_writebp(bp, force) + register struct buf *bp; + int force; +{ + register int oldflags = bp->b_flags, retv = 1; + off_t off; + + if(!(bp->b_flags & B_BUSY)) + panic("bwrite: buffer is not busy???"); + + bp->b_flags &= ~(B_READ|B_DONE|B_ERROR|B_DELWRI); + + if ((oldflags & (B_ASYNC|B_DELWRI)) == (B_ASYNC|B_DELWRI)) { + reassignbuf(bp, bp->b_vp); + } + + bp->b_vp->v_numoutput++; + curproc->p_stats->p_ru.ru_oublock++; + + /* + * If B_NEEDCOMMIT is set, a commit rpc may do the trick. If not + * an actual write will have to be scheduled via. VOP_STRATEGY(). + * If B_WRITEINPROG is already set, then push it with a write anyhow. + */ + if (oldflags & (B_NEEDCOMMIT | B_WRITEINPROG) == B_NEEDCOMMIT) { + off = ((u_quad_t)bp->b_blkno) * DEV_BSIZE + bp->b_dirtyoff; + bp->b_flags |= B_WRITEINPROG; + retv = nfs_commit(bp->b_vp, off, bp->b_dirtyend-bp->b_dirtyoff, + bp->b_wcred, bp->b_proc); + bp->b_flags &= ~B_WRITEINPROG; + if (!retv) { + bp->b_dirtyoff = bp->b_dirtyend = 0; + bp->b_flags &= ~B_NEEDCOMMIT; + biodone(bp); + } else if (retv == NFSERR_STALEWRITEVERF) + nfs_clearcommit(bp->b_vp->v_mount); + } + if (retv) { + if (force) + bp->b_flags |= B_WRITEINPROG; + vfs_busy_pages(bp, 1); + VOP_STRATEGY(bp); + } + + if( (oldflags & B_ASYNC) == 0) { + int rtval = biowait(bp); + + if (oldflags & B_DELWRI) { + reassignbuf(bp, bp->b_vp); + } + brelse(bp); + return (rtval); + } + + return (0); +} + +/* * nfs special file access vnode op. * Essentially just get vattr and then imitate iaccess() since the device is * local to the client. @@ -2409,7 +3130,8 @@ nfsspec_access(ap) found: ; } - return ((vap->va_mode & mode) == mode ? 0 : EACCES); + error = (vap->va_mode & mode) == mode ? 0 : EACCES; + return (error); } /* @@ -2430,7 +3152,8 @@ nfsspec_read(ap) * Set access flag. */ np->n_flag |= NACC; - np->n_atim = time; + np->n_atim.ts_sec = time.tv_sec; + np->n_atim.ts_nsec = time.tv_usec * 1000; return (VOCALL(spec_vnodeop_p, VOFFSET(vop_read), ap)); } @@ -2452,7 +3175,8 @@ nfsspec_write(ap) * Set update flag. */ np->n_flag |= NUPD; - np->n_mtim = time; + np->n_mtim.ts_sec = time.tv_sec; + np->n_mtim.ts_nsec = time.tv_usec * 1000; return (VOCALL(spec_vnodeop_p, VOFFSET(vop_write), ap)); } @@ -2479,16 +3203,10 @@ nfsspec_close(ap) if (vp->v_usecount == 1 && (vp->v_mount->mnt_flag & MNT_RDONLY) == 0) { VATTR_NULL(&vattr); - if (np->n_flag & NACC) { - vattr.va_atime.ts_sec = np->n_atim.tv_sec; - vattr.va_atime.ts_nsec = - np->n_atim.tv_usec * 1000; - } - if (np->n_flag & NUPD) { - vattr.va_mtime.ts_sec = np->n_mtim.tv_sec; - vattr.va_mtime.ts_nsec = - np->n_mtim.tv_usec * 1000; - } + if (np->n_flag & NACC) + vattr.va_atime = np->n_atim; + if (np->n_flag & NUPD) + vattr.va_mtime = np->n_mtim; (void)VOP_SETATTR(vp, &vattr, ap->a_cred, ap->a_p); } } @@ -2513,7 +3231,8 @@ nfsfifo_read(ap) * Set access flag. */ np->n_flag |= NACC; - np->n_atim = time; + np->n_atim.ts_sec = time.tv_sec; + np->n_atim.ts_nsec = time.tv_usec * 1000; return (VOCALL(fifo_vnodeop_p, VOFFSET(vop_read), ap)); } @@ -2535,7 +3254,8 @@ nfsfifo_write(ap) * Set update flag. */ np->n_flag |= NUPD; - np->n_mtim = time; + np->n_mtim.ts_sec = time.tv_sec; + np->n_mtim.ts_nsec = time.tv_usec * 1000; return (VOCALL(fifo_vnodeop_p, VOFFSET(vop_write), ap)); } @@ -2558,24 +3278,22 @@ nfsfifo_close(ap) struct vattr vattr; if (np->n_flag & (NACC | NUPD)) { - if (np->n_flag & NACC) - np->n_atim = time; - if (np->n_flag & NUPD) - np->n_mtim = time; + if (np->n_flag & NACC) { + np->n_atim.ts_sec = time.tv_sec; + np->n_atim.ts_nsec = time.tv_usec * 1000; + } + if (np->n_flag & NUPD) { + np->n_mtim.ts_sec = time.tv_sec; + np->n_mtim.ts_nsec = time.tv_usec * 1000; + } np->n_flag |= NCHG; if (vp->v_usecount == 1 && (vp->v_mount->mnt_flag & MNT_RDONLY) == 0) { VATTR_NULL(&vattr); - if (np->n_flag & NACC) { - vattr.va_atime.ts_sec = np->n_atim.tv_sec; - vattr.va_atime.ts_nsec = - np->n_atim.tv_usec * 1000; - } - if (np->n_flag & NUPD) { - vattr.va_mtime.ts_sec = np->n_mtim.tv_sec; - vattr.va_mtime.ts_nsec = - np->n_mtim.tv_usec * 1000; - } + if (np->n_flag & NACC) + vattr.va_atime = np->n_atim; + if (np->n_flag & NUPD) + vattr.va_mtime = np->n_mtim; (void)VOP_SETATTR(vp, &vattr, ap->a_cred, ap->a_p); } } diff --git a/sys/nfsclient/nfsargs.h b/sys/nfsclient/nfsargs.h index cbf80c6..e1a0f07 100644 --- a/sys/nfsclient/nfsargs.h +++ b/sys/nfsclient/nfsargs.h @@ -34,7 +34,7 @@ * SUCH DAMAGE. * * @(#)nfs.h 8.1 (Berkeley) 6/10/93 - * $Id: nfs.h,v 1.8 1994/11/02 00:11:00 wollman Exp $ + * $Id: nfs.h,v 1.9 1995/02/14 06:22:18 phk Exp $ */ #ifndef _NFS_NFS_H_ @@ -45,11 +45,12 @@ */ #define NFS_MAXIOVEC 34 -#define NFS_HZ 25 /* Ticks per second for NFS timeouts */ -#define NFS_TIMEO (1*NFS_HZ) /* Default timeout = 1 second */ -#define NFS_MINTIMEO (1*NFS_HZ) /* Min timeout to use */ -#define NFS_MAXTIMEO (60*NFS_HZ) /* Max timeout to backoff to */ -#define NFS_MINIDEMTIMEO (5*NFS_HZ) /* Min timeout for non-idempotent ops*/ +#define NFS_TICKINTVL 5 /* Desired time for a tick (msec) */ +#define NFS_HZ (hz / nfs_ticks) /* Ticks/sec */ +#define NFS_TIMEO (1 * NFS_HZ) /* Default timeout = 1 second */ +#define NFS_MINTIMEO (1 * NFS_HZ) /* Min timeout to use */ +#define NFS_MAXTIMEO (60 * NFS_HZ) /* Max timeout to backoff to */ +#define NFS_MINIDEMTIMEO (5 * NFS_HZ) /* Min timeout for non-idempotent ops*/ #define NFS_MAXREXMIT 100 /* Stop counting after this many */ #define NFS_MAXWINDOW 1024 /* Max number of outstanding requests */ #define NFS_RETRANS 10 /* Num of retrans for soft mounts */ @@ -62,13 +63,81 @@ #endif #define NFS_WSIZE 8192 /* Def. write data size <= 8192 */ #define NFS_RSIZE 8192 /* Def. read data size <= 8192 */ +#define NFS_READDIRSIZE 8192 /* Def. readdir size */ #define NFS_DEFRAHEAD 1 /* Def. read ahead # blocks */ #define NFS_MAXRAHEAD 4 /* Max. read ahead # blocks */ -#define NFS_MAXREADDIR NFS_MAXDATA /* Max. size of directory read */ #define NFS_MAXUIDHASH 64 /* Max. # of hashed uid entries/mp */ -#define NFS_MAXASYNCDAEMON 20 /* Max. number async_daemons runable */ -#define NFS_DIRBLKSIZ 1024 /* Size of an NFS directory block */ +#define NFS_MAXASYNCDAEMON 20 /* Max. number async_daemons runable */ +#define NFS_MAXGATHERDELAY 100 /* Max. write gather delay (msec) */ +#ifndef NFS_GATHERDELAY +#define NFS_GATHERDELAY 10 /* Default write gather delay (msec) */ +#endif +#define NFS_DIRBLKSIZ 4096 /* Must be a multiple of DIRBLKSIZ */ + +/* + * Oddballs + */ #define NMOD(a) ((a) % nfs_asyncdaemons) +#define NFS_CMPFH(n, f, s) \ + ((n)->n_fhsize == (s) && !bcmp((caddr_t)(n)->n_fhp, (caddr_t)(f), (s))) +#define NFS_ISV3(v) (VFSTONFS((v)->v_mount)->nm_flag & NFSMNT_NFSV3) +#define NFS_SRVMAXDATA(n) \ + (((n)->nd_flag & ND_NFSV3) ? (((n)->nd_nam2) ? \ + NFS_MAXDGRAMDATA : NFS_MAXDATA) : NFS_V2MAXDATA) + +/* + * XXX + * sys/buf.h should be editted to change B_APPENDWRITE --> B_NEEDCOMMIT, but + * until then... + * Same goes for sys/malloc.h, which needs M_NFSDIROFF, + * M_NFSRVDESC and M_NFSBIGFH added. + * The VA_EXCLUSIVE flag should be added for va_vaflags and set for an + * exclusive create. + * The B_INVAFTERWRITE flag should be set to whatever is required by the + * buffer cache code to say "Invalidate the block after it is written back". + */ +#ifndef B_NEEDCOMMIT +#define B_NEEDCOMMIT B_APPENDWRITE +#endif +#ifndef M_NFSRVDESC +#define M_NFSRVDESC M_TEMP +#endif +#ifndef M_NFSDIROFF +#define M_NFSDIROFF M_TEMP +#endif +#ifndef M_NFSBIGFH +#define M_NFSBIGFH M_TEMP +#endif +#ifndef VA_EXCLUSIVE +#define VA_EXCLUSIVE 0 +#endif +#ifdef __FreeBSD__ +#define B_INVAFTERWRITE B_NOCACHE +#else +#define B_INVAFTERWRITE B_INVAL +#endif + +/* + * These ifdefs try to handle the differences between the various 4.4BSD-Lite + * based vfs interfaces. + * btw: NetBSD-current does have a VOP_LEASDE(), but I don't know how to + * differentiate between NetBSD-1.0 and NetBSD-current, so.. + * I also don't know about BSDi's 2.0 release. + */ +#if !defined(HAS_VOPLEASE) && !defined(__FreeBSD__) && !defined(__NetBSD__) +#define HAS_VOPLEASE 1 +#endif +#if !defined(HAS_VOPREVOKE) && !defined(__FreeBSD__) && !defined(__NetBSD__) +#define HAS_VOPREVOKE 1 +#endif + +/* + * The IO_METASYNC flag should be implemented for local file systems. + * (Until then, it is nothin at all.) + */ +#ifndef IO_METASYNC +#define IO_METASYNC 0 +#endif /* * Set the attribute timeout based on how recently the file has been modified. @@ -80,6 +149,20 @@ (time.tv_sec - (np)->n_mtime) / 10)) /* + * Expected allocation sizes for major data structures. If the actual size + * of the structure exceeds these sizes, then malloc() will be allocating + * almost twice the memory required. This is used in nfs_init() to warn + * the sysadmin that the size of a structure should be reduced. + * (These sizes are always a power of 2. If the kernel malloc() changes + * to one that does not allocate space in powers of 2 size, then this all + * becomes bunk!) + */ +#define NFS_NODEALLOC 256 +#define NFS_MNTALLOC 512 +#define NFS_SVCALLOC 256 +#define NFS_UIDALLOC 128 + +/* * Structures for the nfssvc(2) syscall. Not that anyone but nfsd and mount_nfs * should ever try and use it. */ @@ -95,7 +178,12 @@ struct nfsd_srvargs { u_long nsd_haddr; /* Ip address of client */ struct ucred nsd_cr; /* Cred. uid maps to */ int nsd_authlen; /* Length of auth string (ret) */ - char *nsd_authstr; /* Auth string (ret) */ + u_char *nsd_authstr; /* Auth string (ret) */ + int nsd_verflen; /* and the verfier */ + u_char *nsd_verfstr; + struct timeval nsd_timestamp; /* timestamp from verifier */ + u_long nsd_ttl; /* credential ttl (sec) */ + NFSKERBKEY_T nsd_key; /* Session key */ }; struct nfsd_cargs { @@ -103,7 +191,10 @@ struct nfsd_cargs { uid_t ncd_authuid; /* Effective uid */ int ncd_authtype; /* Type of authenticator */ int ncd_authlen; /* Length of authenticator string */ - char *ncd_authstr; /* Authenticator string */ + u_char *ncd_authstr; /* Authenticator string */ + int ncd_verflen; /* and the verifier */ + u_char *ncd_verfstr; + NFSKERBKEY_T ncd_key; /* Session key */ }; /* @@ -142,6 +233,7 @@ struct nfsstats { int srvnqnfs_leases; int srvnqnfs_maxleases; int srvnqnfs_getleases; + int srvvop_writes; }; /* @@ -173,7 +265,7 @@ struct nfsstats { * such as SIGALRM will not expect file I/O system calls to be interrupted * by them and break. */ -#ifdef KERNEL +#if defined(KERNEL) || defined(_KERNEL) struct uio; struct buf; struct vattr; struct nameidata; /* XXX */ @@ -224,17 +316,29 @@ TAILQ_HEAD(, nfsreq) nfs_reqq; #define R_MUSTRESEND 0x40 /* Must resend request */ #define R_GETONEREP 0x80 /* Probe for one reply only */ -extern struct nfsstats nfsstats; - /* * A list of nfssvc_sock structures is maintained with all the sockets * that require service by the nfsd. * The nfsuid structs hang off of the nfssvc_sock structs in both lru * and uid hash lists. */ -#define NUIDHASHSIZ 32 +#ifndef NFS_UIDHASHSIZ +#define NFS_UIDHASHSIZ 29 /* Tune the size of nfssvc_sock with this */ +#endif #define NUIDHASH(sock, uid) \ - (&(sock)->ns_uidhashtbl[(uid) & (sock)->ns_uidhash]) + (&(sock)->ns_uidhashtbl[(uid) % NFS_UIDHASHSIZ]) +#ifndef NFS_WDELAYHASHSIZ +#define NFS_WDELAYHASHSIZ 16 /* and with this */ +#endif +#define NWDELAYHASH(sock, f) \ + (&(sock)->ns_wdelayhashtbl[(*((u_long *)(f))) % NFS_WDELAYHASHSIZ]) +#ifndef NFS_MUIDHASHSIZ +#define NFS_MUIDHASHSIZ 67 /* Tune the size of nfsmount with this */ +#endif +#define NMUIDHASH(nmp, uid) \ + (&(nmp)->nm_uidhashtbl[(uid) % NFS_MUIDHASHSIZ]) +#define NFSNOHASH(fhsum) \ + (&nfsnodehashtbl[(fhsum) & nfsnodehash]) /* * Network address hash list element @@ -248,35 +352,41 @@ struct nfsuid { TAILQ_ENTRY(nfsuid) nu_lru; /* LRU chain */ LIST_ENTRY(nfsuid) nu_hash; /* Hash list */ int nu_flag; /* Flags */ - uid_t nu_uid; /* Uid mapped by this entry */ union nethostaddr nu_haddr; /* Host addr. for dgram sockets */ struct ucred nu_cr; /* Cred uid mapped to */ + int nu_expire; /* Expiry time (sec) */ + struct timeval nu_timestamp; /* Kerb. timestamp */ + u_long nu_nickname; /* Nickname on server */ + NFSKERBKEY_T nu_key; /* and session key */ }; #define nu_inetaddr nu_haddr.had_inetaddr #define nu_nam nu_haddr.had_nam /* Bits for nu_flag */ #define NU_INETADDR 0x1 +#define NU_NAM 0x2 +#define NU_NETFAM(u) (((u)->nu_flag & NU_INETADDR) ? AF_INET : AF_ISO) struct nfssvc_sock { TAILQ_ENTRY(nfssvc_sock) ns_chain; /* List of all nfssvc_sock's */ TAILQ_HEAD(, nfsuid) ns_uidlruhead; - LIST_HEAD(, nfsuid) *ns_uidhashtbl; - u_long ns_uidhash; - - int ns_flag; - u_long ns_sref; struct file *ns_fp; struct socket *ns_so; - int ns_solock; struct mbuf *ns_nam; - int ns_cc; struct mbuf *ns_raw; struct mbuf *ns_rawend; - int ns_reclen; struct mbuf *ns_rec; struct mbuf *ns_recend; + struct mbuf *ns_frag; + int ns_flag; + int ns_solock; + int ns_cc; + int ns_reclen; int ns_numuids; + u_long ns_sref; + LIST_HEAD(, nfsrv_descript) ns_tq; /* Write gather lists */ + LIST_HEAD(, nfsuid) ns_uidhashtbl[NFS_UIDHASHSIZ]; + LIST_HEAD(nfsrvw_delayhash, nfsrv_descript) ns_wdelayhashtbl[NFS_WDELAYHASHSIZ]; }; /* Bits for "ns_flag" */ @@ -285,6 +395,7 @@ struct nfssvc_sock { #define SLP_NEEDQ 0x04 #define SLP_DISCONN 0x08 #define SLP_GETSTREAM 0x10 +#define SLP_LASTFRAG 0x20 #define SLP_ALLFLAGS 0xff TAILQ_HEAD(, nfssvc_sock) nfssvc_sockhead; @@ -296,73 +407,124 @@ int nfssvc_sockhead_flag; * One of these structures is allocated for each nfsd. */ struct nfsd { - TAILQ_ENTRY(nfsd) nd_chain; /* List of all nfsd's */ - int nd_flag; /* NFSD_ flags */ - struct nfssvc_sock *nd_slp; /* Current socket */ - struct mbuf *nd_nam; /* Client addr for datagram req. */ - struct mbuf *nd_mrep; /* Req. mbuf list */ - struct mbuf *nd_md; - caddr_t nd_dpos; /* Position in list */ - int nd_procnum; /* RPC procedure number */ - u_long nd_retxid; /* RPC xid */ - int nd_repstat; /* Reply status value */ - struct ucred nd_cr; /* Credentials for req. */ - int nd_nqlflag; /* Leasing flag */ - u_long nd_duration; /* Lease duration */ - int nd_authlen; /* Authenticator len */ - u_char nd_authstr[RPCAUTH_MAXSIZ]; /* Authenticator data */ - struct proc *nd_procp; /* Proc ptr */ + TAILQ_ENTRY(nfsd) nfsd_chain; /* List of all nfsd's */ + int nfsd_flag; /* NFSD_ flags */ + struct nfssvc_sock *nfsd_slp; /* Current socket */ + int nfsd_authlen; /* Authenticator len */ + u_char nfsd_authstr[RPCAUTH_MAXSIZ]; /* Authenticator data */ + int nfsd_verflen; /* and the Verifier */ + u_char nfsd_verfstr[RPCVERF_MAXSIZ]; + struct proc *nfsd_procp; /* Proc ptr */ + struct nfsrv_descript *nfsd_nd; /* Associated nfsrv_descript */ }; -/* Bits for "nd_flag" */ +/* Bits for "nfsd_flag" */ #define NFSD_WAITING 0x01 #define NFSD_REQINPROG 0x02 #define NFSD_NEEDAUTH 0x04 #define NFSD_AUTHFAIL 0x08 +/* + * This structure is used by the server for describing each request. + * Some fields are used only when write request gathering is performed. + */ +struct nfsrv_descript { + u_quad_t nd_time; /* Write deadline (usec) */ + off_t nd_off; /* Start byte offset */ + off_t nd_eoff; /* and end byte offset */ + LIST_ENTRY(nfsrv_descript) nd_hash; /* Hash list */ + LIST_ENTRY(nfsrv_descript) nd_tq; /* and timer list */ + LIST_HEAD(,nfsrv_descript) nd_coalesce; /* coalesced writes */ + struct mbuf *nd_mrep; /* Request mbuf list */ + struct mbuf *nd_md; /* Current dissect mbuf */ + struct mbuf *nd_mreq; /* Reply mbuf list */ + struct mbuf *nd_nam; /* and socket addr */ + struct mbuf *nd_nam2; /* return socket addr */ + caddr_t nd_dpos; /* Current dissect pos */ + int nd_procnum; /* RPC # */ + int nd_stable; /* storage type */ + int nd_flag; /* nd_flag */ + int nd_len; /* Length of this write */ + int nd_repstat; /* Reply status */ + u_long nd_retxid; /* Reply xid */ + u_long nd_duration; /* Lease duration */ + struct timeval nd_starttime; /* Time RPC initiated */ + fhandle_t nd_fh; /* File handle */ + struct ucred nd_cr; /* Credentials */ +}; + +/* Bits for "nd_flag" */ +#define ND_READ LEASE_READ +#define ND_WRITE LEASE_WRITE +#define ND_CHECK 0x04 +#define ND_LEASE (ND_READ | ND_WRITE | ND_CHECK) +#define ND_NFSV3 0x08 +#define ND_NQNFS 0x10 +#define ND_KERBNICK 0x20 +#define ND_KERBFULL 0x40 +#define ND_KERBAUTH (ND_KERBNICK | ND_KERBFULL) + TAILQ_HEAD(, nfsd) nfsd_head; int nfsd_head_flag; #define NFSD_CHECKSLP 0x01 +/* + * These macros compare nfsrv_descript structures. + */ +#define NFSW_CONTIG(o, n) \ + ((o)->nd_eoff >= (n)->nd_off && \ + !bcmp((caddr_t)&(o)->nd_fh, (caddr_t)&(n)->nd_fh, NFSX_V3FH)) + +#define NFSW_SAMECRED(o, n) \ + (((o)->nd_flag & ND_KERBAUTH) == ((n)->nd_flag & ND_KERBAUTH) && \ + !bcmp((caddr_t)&(o)->nd_cr, (caddr_t)&(n)->nd_cr, \ + sizeof (struct ucred))) + int nfs_reply __P((struct nfsreq *)); -int nfs_getreq __P((struct nfsd *,int)); +int nfs_getreq __P((struct nfsrv_descript *,struct nfsd *,int)); int nfs_send __P((struct socket *,struct mbuf *,struct mbuf *,struct nfsreq *)); -int nfs_rephead __P((int,struct nfsd *,int,int,u_quad_t *,struct mbuf **,struct mbuf **,caddr_t *)); +int nfs_rephead __P((int,struct nfsrv_descript *,struct nfssvc_sock *,int,int,u_quad_t *,struct mbuf **,struct mbuf **,caddr_t *)); int nfs_sndlock __P((int *,struct nfsreq *)); int nfs_disct __P((struct mbuf **,caddr_t *,int,int,caddr_t *)); int nfs_vinvalbuf __P((struct vnode *,int,struct ucred *,struct proc *,int)); int nfs_readrpc __P((struct vnode *,struct uio *,struct ucred *)); -int nfs_writerpc __P((struct vnode *,struct uio *,struct ucred *,int)); +int nfs_writerpc __P((struct vnode *,struct uio *,struct ucred *,int *,int *)); int nfs_readdirrpc __P((register struct vnode *,struct uio *,struct ucred *)); +int nfs_setattrrpc __P((struct vnode *,struct vattr *,struct ucred *,struct proc *)); int nfs_asyncio __P((struct buf *,struct ucred *)); int nfs_doio __P((struct buf *,struct ucred *,struct proc *)); int nfs_readlinkrpc __P((struct vnode *,struct uio *,struct ucred *)); int nfs_sigintr __P((struct nfsmount *,struct nfsreq *r,struct proc *)); -int nfs_readdirlookrpc __P((struct vnode *,register struct uio *,struct ucred *)); +int nfs_readdirplusrpc __P((struct vnode *,register struct uio *,struct ucred *)); int nfsm_disct __P((struct mbuf **,caddr_t *,int,int,caddr_t *)); -int nfsrv_fhtovp __P((fhandle_t *,int,struct vnode **,struct ucred *,struct nfssvc_sock *,struct mbuf *,int *)); +void nfsm_srvfattr __P((struct nfsrv_descript *,struct vattr *,struct nfs_fattr *)); +void nfsm_srvwcc __P((struct nfsrv_descript *,int,struct vattr *,int,struct vattr *,struct mbuf **,char **)); +void nfsm_srvpostopattr __P((struct nfsrv_descript *,int,struct vattr *,struct mbuf **,char **)); +int nfsrv_fhtovp __P((fhandle_t *,int,struct vnode **,struct ucred *,struct nfssvc_sock *,struct mbuf *,int *,int)); int nfsrv_access __P((struct vnode *,int,struct ucred *,int,struct proc *)); int netaddr_match __P((int,union nethostaddr *,struct mbuf *)); int nfs_request __P((struct vnode *,struct mbuf *,int,struct proc *,struct ucred *,struct mbuf **,struct mbuf **,caddr_t *)); int nfs_loadattrcache __P((struct vnode **,struct mbuf **,caddr_t *,struct vattr *)); -int nfs_namei __P((struct nameidata *,fhandle_t *,int,struct nfssvc_sock *,struct mbuf *,struct mbuf **,caddr_t *,struct proc *)); +int nfs_namei __P((struct nameidata *,fhandle_t *,int,struct nfssvc_sock *,struct mbuf *,struct mbuf **,caddr_t *,struct vnode **,struct proc *,int)); void nfsm_adj __P((struct mbuf *,int,int)); int nfsm_mbuftouio __P((struct mbuf **,struct uio *,int,caddr_t *)); void nfsrv_initcache __P((void)); int nfs_rcvlock __P((struct nfsreq *)); -int nfs_getauth __P((struct nfsmount *,struct nfsreq *,struct ucred *,int *,char **,int *)); +int nfs_getauth __P((struct nfsmount *,struct nfsreq *,struct ucred *,char **,int *,char *,int *,NFSKERBKEY_T)); +int nfs_getnickauth __P((struct nfsmount *,struct ucred *,char **,int *,char *,int)); +int nfs_savenickauth __P((struct nfsmount *,struct ucred *,int,NFSKERBKEY_T,struct mbuf **,char **,struct mbuf *)); int nfs_msg __P((struct proc *,char *,char *)); int nfs_adv __P((struct mbuf **,caddr_t *,int,int)); int nfsrv_getstream __P((struct nfssvc_sock *,int)); void nfs_nhinit __P((void)); void nfs_timer __P((void*)); -struct nfsnodehashhead * nfs_hash __P((nfsv2fh_t *)); +u_long nfs_hash __P((nfsfh_t *,int)); int nfssvc_iod __P((struct proc *)); int nfssvc_nfsd __P((struct nfsd_srvargs *,caddr_t,struct proc *)); int nfssvc_addsock __P((struct file *,struct mbuf *)); -int nfsrv_dorec __P((struct nfssvc_sock *,struct nfsd *)); -int nfsrv_getcache __P((struct mbuf *,struct nfsd *,struct mbuf **)); -void nfsrv_updatecache __P((struct mbuf *,struct nfsd *,int,struct mbuf *)); +int nfsrv_dorec __P((struct nfssvc_sock *,struct nfsd *,struct nfsrv_descript **)); +int nfsrv_getcache __P((struct nfsrv_descript *,struct nfssvc_sock *,struct mbuf **)); +void nfsrv_updatecache __P((struct nfsrv_descript *,int,struct mbuf *)); int mountnfs __P((struct nfs_args *,struct mount *,struct mbuf *,char *,char *,struct vnode **)); int nfs_connect __P((struct nfsmount *,struct nfsreq *)); int nfs_getattrcache __P((struct vnode *,struct vattr *)); @@ -370,9 +532,20 @@ int nfsm_strtmbuf __P((struct mbuf **,char **,char *,long)); int nfs_bioread __P((struct vnode *,struct uio *,int,struct ucred *)); int nfsm_uiotombuf __P((struct uio *,struct mbuf **,int,caddr_t *)); void nfsrv_init __P((int)); +void nfs_clearcommit __P((struct mount *)); +int nfsrv_errmap __P((struct nfsrv_descript *, int)); +void nfsrvw_coalesce __P((struct nfsrv_descript *,struct nfsrv_descript *)); +void nfsrvw_sort __P((gid_t [],int)); +void nfsrv_setcred __P((struct ucred *,struct ucred *)); +int nfs_flush __P((struct vnode *,struct ucred *,int,struct proc *,int)); +int nfs_writebp __P((struct buf *,int)); int nfsrv_vput __P(( struct vnode * )); int nfsrv_vrele __P(( struct vnode * )); int nfsrv_vmio __P(( struct vnode * )); +int nfsrv_writegather __P((struct nfsrv_descript **, struct nfssvc_sock *, + struct proc *, struct mbuf **)); +int nfs_fsinfo __P((struct nfsmount *, struct vnode *, struct ucred *, + struct proc *p)); #endif /* KERNEL */ diff --git a/sys/nfsclient/nfsdiskless.h b/sys/nfsclient/nfsdiskless.h index 5df85e0..f7371dc 100644 --- a/sys/nfsclient/nfsdiskless.h +++ b/sys/nfsclient/nfsdiskless.h @@ -34,7 +34,7 @@ * SUCH DAMAGE. * * @(#)nfsdiskless.h 8.1 (Berkeley) 6/10/93 - * $Id: nfsdiskless.h,v 1.3 1994/08/05 09:28:49 davidg Exp $ + * $Id: nfsdiskless.h,v 1.4 1994/08/21 06:50:09 paul Exp $ */ #ifndef _NFS_NFSDISKLESS_H_ @@ -52,17 +52,45 @@ * NB: All fields are stored in net byte order to avoid hassles with * client/server byte ordering differences. */ + +/* + * I have defined a new structure that can handle an NFS Version 3 file handle + * but the kernel still expects the old Version 2 one to be provided. The + * changes required in nfs_vfsops.c for using the new are documented there in + * comments. (I felt that breaking network booting code by changing this + * structure would not be prudent at this time, since almost all servers are + * still Version 2 anyhow.) + */ +struct nfsv3_diskless { + struct ifaliasreq myif; /* Default interface */ + struct sockaddr_in mygateway; /* Default gateway */ + struct nfs_args swap_args; /* Mount args for swap file */ + int swap_fhsize; /* Size of file handle */ + u_char swap_fh[NFSX_V3FHMAX]; /* Swap file's file handle */ + struct sockaddr_in swap_saddr; /* Address of swap server */ + char swap_hostnam[MNAMELEN]; /* Host name for mount pt */ + int swap_nblks; /* Size of server swap file */ + struct ucred swap_ucred; /* Swap credentials */ + struct nfs_args root_args; /* Mount args for root fs */ + int root_fhsize; /* Size of root file handle */ + u_char root_fh[NFSX_V3FHMAX]; /* File handle of root dir */ + struct sockaddr_in root_saddr; /* Address of root server */ + char root_hostnam[MNAMELEN]; /* Host name for mount pt */ + long root_time; /* Timestamp of root fs */ + char my_hostnam[MAXHOSTNAMELEN]; /* Client host name */ +}; + struct nfs_diskless { struct ifaliasreq myif; /* Default interface */ struct sockaddr_in mygateway; /* Default gateway */ struct nfs_args swap_args; /* Mount args for swap file */ - u_char swap_fh[NFS_FHSIZE]; /* Swap file's file handle */ + u_char swap_fh[NFSX_V2FH]; /* Swap file's file handle */ struct sockaddr_in swap_saddr; /* Address of swap server */ char swap_hostnam[MNAMELEN]; /* Host name for mount pt */ int swap_nblks; /* Size of server swap file */ struct ucred swap_ucred; /* Swap credentials */ struct nfs_args root_args; /* Mount args for root fs */ - u_char root_fh[NFS_FHSIZE]; /* File handle of root dir */ + u_char root_fh[NFSX_V2FH]; /* File handle of root dir */ struct sockaddr_in root_saddr; /* Address of root server */ char root_hostnam[MNAMELEN]; /* Host name for mount pt */ long root_time; /* Timestamp of root fs */ diff --git a/sys/nfsclient/nfsm_subs.h b/sys/nfsclient/nfsm_subs.h index 030a985..9e89c57 100644 --- a/sys/nfsclient/nfsm_subs.h +++ b/sys/nfsclient/nfsm_subs.h @@ -34,12 +34,13 @@ * SUCH DAMAGE. * * @(#)nfsm_subs.h 8.1 (Berkeley) 6/16/93 - * $Id: nfsm_subs.h,v 1.5 1994/10/17 17:47:42 phk Exp $ + * $Id: nfsm_subs.h,v 1.6 1995/05/30 08:12:51 rgrimes Exp $ */ #ifndef _NFS_NFSM_SUBS_H_ #define _NFS_NFSM_SUBS_H_ + /* * These macros do strange and peculiar things to mbuf chains for * the assistance of the nfs code. To attempt to use them for any @@ -90,49 +91,153 @@ extern struct mbuf *nfsm_reqh(); mb->m_len += (s); \ bpos += (s); } -#define nfsm_dissect(a,c,s) \ +#define nfsm_dissect(a, c, s) \ { t1 = mtod(md, caddr_t)+md->m_len-dpos; \ if (t1 >= (s)) { \ (a) = (c)(dpos); \ dpos += (s); \ + } else if (t1 = nfsm_disct(&md, &dpos, (s), t1, &cp2)) { \ + error = t1; \ + m_freem(mrep); \ + goto nfsmout; \ } else { \ - error = nfsm_disct(&md, &dpos, (s), t1, &cp2); \ - if (error) { \ - m_freem(mrep); \ + (a) = (c)cp2; \ + } } + +#define nfsm_fhtom(v, v3) \ + { if (v3) { \ + t2 = nfsm_rndup(VTONFS(v)->n_fhsize) + NFSX_UNSIGNED; \ + if (t2 <= M_TRAILINGSPACE(mb)) { \ + nfsm_build(tl, u_long *, t2); \ + *tl++ = txdr_unsigned(VTONFS(v)->n_fhsize); \ + *(tl + ((t2>>2) - 2)) = 0; \ + bcopy((caddr_t)VTONFS(v)->n_fhp,(caddr_t)tl, \ + VTONFS(v)->n_fhsize); \ + } else if (t2 = nfsm_strtmbuf(&mb, &bpos, \ + (caddr_t)VTONFS(v)->n_fhp, VTONFS(v)->n_fhsize)) { \ + error = t2; \ + m_freem(mreq); \ goto nfsmout; \ - } else { \ - (a) = (c)cp2; \ } \ + } else { \ + nfsm_build(cp, caddr_t, NFSX_V2FH); \ + bcopy((caddr_t)VTONFS(v)->n_fhp, cp, NFSX_V2FH); \ } } -#define nfsm_fhtom(v) \ - nfsm_build(cp,caddr_t,NFSX_FH); \ - bcopy((caddr_t)&(VTONFS(v)->n_fh), cp, NFSX_FH) +#define nfsm_srvfhtom(f, v3) \ + { if (v3) { \ + nfsm_build(tl, u_long *, NFSX_UNSIGNED + NFSX_V3FH); \ + *tl++ = txdr_unsigned(NFSX_V3FH); \ + bcopy((caddr_t)(f), (caddr_t)tl, NFSX_V3FH); \ + } else { \ + nfsm_build(cp, caddr_t, NFSX_V2FH); \ + bcopy((caddr_t)(f), cp, NFSX_V2FH); \ + } } -#define nfsm_srvfhtom(f) \ - nfsm_build(cp,caddr_t,NFSX_FH); \ - bcopy((caddr_t)(f), cp, NFSX_FH) +#define nfsm_srvpostop_fh(f) \ + { nfsm_build(tl, u_long *, 2 * NFSX_UNSIGNED + NFSX_V3FH); \ + *tl++ = nfs_true; \ + *tl++ = txdr_unsigned(NFSX_V3FH); \ + bcopy((caddr_t)(f), (caddr_t)tl, NFSX_V3FH); \ + } -#define nfsm_mtofh(d,v) \ - { struct nfsnode *np; nfsv2fh_t *fhp; \ - nfsm_dissect(fhp,nfsv2fh_t *,NFSX_FH); \ - error = nfs_nget((d)->v_mount, fhp, &np); \ - if (error) { \ - m_freem(mrep); \ - goto nfsmout; \ +#define nfsm_mtofh(d, v, v3, f) \ + { struct nfsnode *ttnp; nfsfh_t *ttfhp; int ttfhsize; \ + if (v3) { \ + nfsm_dissect(tl, u_long *, NFSX_UNSIGNED); \ + (f) = fxdr_unsigned(int, *tl); \ + } else \ + (f) = 1; \ + if (f) { \ + nfsm_getfh(ttfhp, ttfhsize, (v3)); \ + if (t1 = nfs_nget((d)->v_mount, ttfhp, ttfhsize, \ + &ttnp)) { \ + error = t1; \ + m_freem(mrep); \ + goto nfsmout; \ + } \ + (v) = NFSTOV(ttnp); \ + } \ + if (v3) { \ + nfsm_dissect(tl, u_long *, NFSX_UNSIGNED); \ + if (f) \ + (f) = fxdr_unsigned(int, *tl); \ + else if (fxdr_unsigned(int, *tl)) \ + nfsm_adv(NFSX_V3FATTR); \ } \ - (v) = NFSTOV(np); \ - nfsm_loadattr(v, (struct vattr *)0); \ + if (f) \ + nfsm_loadattr((v), (struct vattr *)0); \ } -#define nfsm_loadattr(v,a) \ - { struct vnode *tvp = (v); \ - error = nfs_loadattrcache(&tvp, &md, &dpos, (a)); \ - if (error) { \ +#define nfsm_getfh(f, s, v3) \ + { if (v3) { \ + nfsm_dissect(tl, u_long *, NFSX_UNSIGNED); \ + if (((s) = fxdr_unsigned(int, *tl)) <= 0 || \ + (s) > NFSX_V3FHMAX) { \ + m_freem(mrep); \ + error = EBADRPC; \ + goto nfsmout; \ + } \ + } else \ + (s) = NFSX_V2FH; \ + nfsm_dissect((f), nfsfh_t *, nfsm_rndup(s)); } + +#define nfsm_loadattr(v, a) \ + { struct vnode *ttvp = (v); \ + if (t1 = nfs_loadattrcache(&ttvp, &md, &dpos, (a))) { \ + error = t1; \ m_freem(mrep); \ goto nfsmout; \ } \ - (v) = tvp; } + (v) = ttvp; } + +#define nfsm_postop_attr(v, f) \ + { struct vnode *ttvp = (v); \ + nfsm_dissect(tl, u_long *, NFSX_UNSIGNED); \ + if ((f) = fxdr_unsigned(int, *tl)) { \ + if (t1 = nfs_loadattrcache(&ttvp, &md, &dpos, \ + (struct vattr *)0)) { \ + error = t1; \ + (f) = 0; \ + m_freem(mrep); \ + goto nfsmout; \ + } \ + (v) = ttvp; \ + } } + +/* Used as (f) for nfsm_wcc_data() */ +#define NFSV3_WCCRATTR 0 +#define NFSV3_WCCCHK 1 + +#define nfsm_wcc_data(v, f) \ + { int ttattrf, ttretf = 0; \ + nfsm_dissect(tl, u_long *, NFSX_UNSIGNED); \ + if (*tl == nfs_true) { \ + nfsm_dissect(tl, u_long *, 6 * NFSX_UNSIGNED); \ + if (f) \ + ttretf = (VTONFS(v)->n_mtime == \ + fxdr_unsigned(u_long, *(tl + 2))); \ + } \ + nfsm_postop_attr((v), ttattrf); \ + if (f) { \ + (f) = ttretf; \ + } else { \ + (f) = ttattrf; \ + } } + +#define nfsm_v3sattr(s, a, u, g) \ + { (s)->sa_modetrue = nfs_true; \ + (s)->sa_mode = vtonfsv3_mode((a)->va_mode); \ + (s)->sa_uidtrue = nfs_true; \ + (s)->sa_uid = txdr_unsigned(u); \ + (s)->sa_gidtrue = nfs_true; \ + (s)->sa_gid = txdr_unsigned(g); \ + (s)->sa_sizefalse = nfs_false; \ + (s)->sa_atimetype = txdr_unsigned(NFSV3SATTRTIME_TOCLIENT); \ + txdr_nfsv3time(&(a)->va_atime, &(s)->sa_atime); \ + (s)->sa_mtimetype = txdr_unsigned(NFSV3SATTRTIME_TOCLIENT); \ + txdr_nfsv3time(&(a)->va_mtime, &(s)->sa_mtime); \ + } #define nfsm_strsiz(s,m) \ { nfsm_dissect(tl,u_long *,NFSX_UNSIGNED); \ @@ -149,16 +254,27 @@ extern struct mbuf *nfsm_reqh(); nfsm_reply(0); \ } } +#define nfsm_srvnamesiz(s) \ + { nfsm_dissect(tl,u_long *,NFSX_UNSIGNED); \ + if (((s) = fxdr_unsigned(long,*tl)) > NFS_MAXNAMLEN) \ + error = NFSERR_NAMETOL; \ + if ((s) <= 0) \ + error = EBADRPC; \ + if (error) \ + nfsm_reply(0); \ + } + #define nfsm_mtouio(p,s) \ if ((s) > 0 && \ - (error = nfsm_mbuftouio(&md,(p),(s),&dpos))) { \ + (t1 = nfsm_mbuftouio(&md,(p),(s),&dpos))) { \ + error = t1; \ m_freem(mrep); \ goto nfsmout; \ } #define nfsm_uiotom(p,s) \ - error = nfsm_uiotombuf((p),&mb,(s),&bpos); \ - if (error) { \ + if (t1 = nfsm_uiotombuf((p),&mb,(s),&bpos)) { \ + error = t1; \ m_freem(mreq); \ goto nfsmout; \ } @@ -172,10 +288,13 @@ extern struct mbuf *nfsm_reqh(); #define nfsm_rndup(a) (((a)+3)&(~0x3)) #define nfsm_request(v, t, p, c) \ - error = nfs_request((v), mreq, (t), (p), \ - (c), &mrep, &md, &dpos); \ - if (error) \ - goto nfsmout + if (error = nfs_request((v), mreq, (t), (p), \ + (c), &mrep, &md, &dpos)) { \ + if (error & NFSERR_RETERR) \ + error &= ~NFSERR_RETERR; \ + else \ + goto nfsmout; \ + } #define nfsm_strtom(a,s,m) \ if ((s) > (m)) { \ @@ -189,12 +308,10 @@ extern struct mbuf *nfsm_reqh(); *tl++ = txdr_unsigned(s); \ *(tl+((t2>>2)-2)) = 0; \ bcopy((caddr_t)(a), (caddr_t)tl, (s)); \ - } else { \ - error = nfsm_strtmbuf(&mb, &bpos, (a), (s)); \ - if (error) { \ - m_freem(mreq); \ - goto nfsmout; \ - } \ + } else if (t2 = nfsm_strtmbuf(&mb, &bpos, (a), (s))) { \ + error = t2; \ + m_freem(mreq); \ + goto nfsmout; \ } #define nfsm_srvdone \ @@ -204,33 +321,53 @@ extern struct mbuf *nfsm_reqh(); #define nfsm_reply(s) \ { \ nfsd->nd_repstat = error; \ - if (error) \ - (void) nfs_rephead(0, nfsd, error, cache, &frev, \ + if (error && !(nfsd->nd_flag & ND_NFSV3)) \ + (void) nfs_rephead(0, nfsd, slp, error, cache, &frev, \ mrq, &mb, &bpos); \ else \ - (void) nfs_rephead((s), nfsd, error, cache, &frev, \ + (void) nfs_rephead((s), nfsd, slp, error, cache, &frev, \ mrq, &mb, &bpos); \ m_freem(mrep); \ mreq = *mrq; \ - if (error) \ + if (error && (!(nfsd->nd_flag & ND_NFSV3) || \ + error == EBADRPC)) \ return(0); \ } +#define nfsm_writereply(s, v3) \ + { \ + nfsd->nd_repstat = error; \ + if (error && !(v3)) \ + (void) nfs_rephead(0, nfsd, slp, error, cache, &frev, \ + &mreq, &mb, &bpos); \ + else \ + (void) nfs_rephead((s), nfsd, slp, error, cache, &frev, \ + &mreq, &mb, &bpos); \ + } + #define nfsm_adv(s) \ - t1 = mtod(md, caddr_t)+md->m_len-dpos; \ + { t1 = mtod(md, caddr_t)+md->m_len-dpos; \ if (t1 >= (s)) { \ dpos += (s); \ - } else { \ - error = nfs_adv(&md, &dpos, (s), t1); \ - if (error) { \ - m_freem(mrep); \ - goto nfsmout; \ - } \ - } + } else if (t1 = nfs_adv(&md, &dpos, (s), t1)) { \ + error = t1; \ + m_freem(mrep); \ + goto nfsmout; \ + } } #define nfsm_srvmtofh(f) \ - nfsm_dissect(tl, u_long *, NFSX_FH); \ - bcopy((caddr_t)tl, (caddr_t)f, NFSX_FH) + { if (nfsd->nd_flag & ND_NFSV3) { \ + nfsm_dissect(tl, u_long *, NFSX_UNSIGNED); \ + if (fxdr_unsigned(int, *tl) != NFSX_V3FH) { \ + error = EBADRPC; \ + nfsm_reply(0); \ + } \ + } \ + nfsm_dissect(tl, u_long *, NFSX_V3FH); \ + bcopy((caddr_t)tl, (caddr_t)(f), NFSX_V3FH); \ + if ((nfsd->nd_flag & ND_NFSV3) == 0) \ + nfsm_adv(NFSX_V2FH - NFSX_V3FH); \ + } #define nfsm_clget \ if (bp >= be) { \ @@ -246,41 +383,57 @@ extern struct mbuf *nfsm_reqh(); } \ tl = (u_long *)bp -#define nfsm_srvfillattr \ - fp->fa_type = vtonfs_type(vap->va_type); \ - fp->fa_mode = vtonfs_mode(vap->va_type, vap->va_mode); \ - fp->fa_nlink = txdr_unsigned(vap->va_nlink); \ - fp->fa_uid = txdr_unsigned(vap->va_uid); \ - fp->fa_gid = txdr_unsigned(vap->va_gid); \ - if (nfsd->nd_nqlflag == NQL_NOVAL) { \ - fp->fa_nfsblocksize = txdr_unsigned(vap->va_blocksize); \ - if (vap->va_type == VFIFO) \ - fp->fa_nfsrdev = 0xffffffff; \ - else \ - fp->fa_nfsrdev = txdr_unsigned(vap->va_rdev); \ - fp->fa_nfsfsid = txdr_unsigned(vap->va_fsid); \ - fp->fa_nfsfileid = txdr_unsigned(vap->va_fileid); \ - fp->fa_nfssize = txdr_unsigned(vap->va_size); \ - fp->fa_nfsblocks = txdr_unsigned(vap->va_bytes / NFS_FABLKSIZE); \ - txdr_nfstime(&vap->va_atime, &fp->fa_nfsatime); \ - txdr_nfstime(&vap->va_mtime, &fp->fa_nfsmtime); \ - txdr_nfstime(&vap->va_ctime, &fp->fa_nfsctime); \ - } else { \ - fp->fa_nqblocksize = txdr_unsigned(vap->va_blocksize); \ - if (vap->va_type == VFIFO) \ - fp->fa_nqrdev = 0xffffffff; \ - else \ - fp->fa_nqrdev = txdr_unsigned(vap->va_rdev); \ - fp->fa_nqfsid = txdr_unsigned(vap->va_fsid); \ - fp->fa_nqfileid = txdr_unsigned(vap->va_fileid); \ - txdr_hyper(&vap->va_size, &fp->fa_nqsize); \ - txdr_hyper(&vap->va_bytes, &fp->fa_nqbytes); \ - txdr_nqtime(&vap->va_atime, &fp->fa_nqatime); \ - txdr_nqtime(&vap->va_mtime, &fp->fa_nqmtime); \ - txdr_nqtime(&vap->va_ctime, &fp->fa_nqctime); \ - fp->fa_nqflags = txdr_unsigned(vap->va_flags); \ - fp->fa_nqgen = txdr_unsigned(vap->va_gen); \ - txdr_hyper(&vap->va_filerev, &fp->fa_nqfilerev); \ - } +#define nfsm_srvfillattr(a, f) \ + nfsm_srvfattr(nfsd, (a), (f)) + +#define nfsm_srvwcc_data(br, b, ar, a) \ + nfsm_srvwcc(nfsd, (br), (b), (ar), (a), &mb, &bpos) + +#define nfsm_srvpostop_attr(r, a) \ + nfsm_srvpostopattr(nfsd, (r), (a), &mb, &bpos) + +#define nfsm_srvsattr(a) \ + { nfsm_dissect(tl, u_long *, NFSX_UNSIGNED); \ + if (*tl == nfs_true) { \ + nfsm_dissect(tl, u_long *, NFSX_UNSIGNED); \ + (a)->va_mode = nfstov_mode(*tl); \ + } \ + nfsm_dissect(tl, u_long *, NFSX_UNSIGNED); \ + if (*tl == nfs_true) { \ + nfsm_dissect(tl, u_long *, NFSX_UNSIGNED); \ + (a)->va_uid = fxdr_unsigned(uid_t, *tl); \ + } \ + nfsm_dissect(tl, u_long *, NFSX_UNSIGNED); \ + if (*tl == nfs_true) { \ + nfsm_dissect(tl, u_long *, NFSX_UNSIGNED); \ + (a)->va_gid = fxdr_unsigned(gid_t, *tl); \ + } \ + nfsm_dissect(tl, u_long *, NFSX_UNSIGNED); \ + if (*tl == nfs_true) { \ + nfsm_dissect(tl, u_long *, 2 * NFSX_UNSIGNED); \ + fxdr_hyper(tl, &(a)->va_size); \ + } \ + nfsm_dissect(tl, u_long *, NFSX_UNSIGNED); \ + switch (fxdr_unsigned(int, *tl)) { \ + case NFSV3SATTRTIME_TOCLIENT: \ + nfsm_dissect(tl, u_long *, 2 * NFSX_UNSIGNED); \ + fxdr_nfsv3time(tl, &(a)->va_atime); \ + break; \ + case NFSV3SATTRTIME_TOSERVER: \ + (a)->va_atime.ts_sec = time.tv_sec; \ + (a)->va_atime.ts_nsec = time.tv_usec * 1000; \ + break; \ + }; \ + nfsm_dissect(tl, u_long *, NFSX_UNSIGNED); \ + switch (fxdr_unsigned(int, *tl)) { \ + case NFSV3SATTRTIME_TOCLIENT: \ + nfsm_dissect(tl, u_long *, 2 * NFSX_UNSIGNED); \ + fxdr_nfsv3time(tl, &(a)->va_mtime); \ + break; \ + case NFSV3SATTRTIME_TOSERVER: \ + (a)->va_mtime.ts_sec = time.tv_sec; \ + (a)->va_mtime.ts_nsec = time.tv_usec * 1000; \ + break; \ + }; } #endif diff --git a/sys/nfsclient/nfsmount.h b/sys/nfsclient/nfsmount.h index 99a1b48..6ac33f7 100644 --- a/sys/nfsclient/nfsmount.h +++ b/sys/nfsclient/nfsmount.h @@ -34,7 +34,7 @@ * SUCH DAMAGE. * * @(#)nfsmount.h 8.1 (Berkeley) 6/10/93 - * $Id: nfsmount.h,v 1.3 1994/08/21 06:50:10 paul Exp $ + * $Id: nfsmount.h,v 1.4 1994/10/17 17:47:43 phk Exp $ */ #ifndef _NFS_NFSMOUNT_H_ @@ -49,7 +49,8 @@ struct nfsmount { int nm_flag; /* Flags for soft/hard... */ struct mount *nm_mountp; /* Vfs structure for this filesystem */ int nm_numgrps; /* Max. size of groupslist */ - nfsv2fh_t nm_fh; /* File handle of root dir */ + u_char nm_fh[NFSX_V3FHMAX]; /* File handle of root dir */ + int nm_fhsize; /* Size of root file handle */ struct socket *nm_so; /* Rpc socket */ int nm_sotype; /* Type of socket */ int nm_soproto; /* and protocol */ @@ -65,6 +66,7 @@ struct nfsmount { int nm_deadthresh; /* Threshold of timeouts-->dead server*/ int nm_rsize; /* Max size of read rpc */ int nm_wsize; /* Max size of write rpc */ + int nm_readdirsize; /* Size of a readdir rpc */ int nm_readahead; /* Num. of blocks to readahead */ int nm_leaseterm; /* Term (sec) for NQNFS lease */ CIRCLEQ_HEAD(, nfsnode) nm_timerhead; /* Head of lease timer queue */ @@ -73,9 +75,16 @@ struct nfsmount { int nm_authtype; /* Authenticator type */ int nm_authlen; /* and length */ char *nm_authstr; /* Authenticator string */ + char *nm_verfstr; /* and the verifier */ + int nm_verflen; + u_char nm_verf[NFSX_V3WRITEVERF]; /* V3 write verifier */ + NFSKERBKEY_T nm_key; /* and the session key */ + int nm_numuids; /* Number of nfsuid mappings */ + TAILQ_HEAD(, nfsuid) nm_uidlruhead; /* Lists of nfsuid mappings */ + LIST_HEAD(, nfsuid) nm_uidhashtbl[NFS_MUIDHASHSIZ]; }; -#ifdef KERNEL +#if defined(KERNEL) || defined(_KERNEL) /* * Convert mount ptr to nfsmount ptr. */ diff --git a/sys/nfsclient/nfsnode.h b/sys/nfsclient/nfsnode.h index dc9162b..18dbde2 100644 --- a/sys/nfsclient/nfsnode.h +++ b/sys/nfsclient/nfsnode.h @@ -34,12 +34,16 @@ * SUCH DAMAGE. * * @(#)nfsnode.h 8.4 (Berkeley) 2/13/94 - * $Id: nfsnode.h,v 1.8 1994/10/17 17:47:44 phk Exp $ + * $Id: nfsnode.h,v 1.9 1995/03/16 18:15:42 bde Exp $ */ #ifndef _NFS_NFSNODE_H_ #define _NFS_NFSNODE_H_ +#ifndef _NFS_NFS_H_ +#include <nfs/nfs.h> +#endif + /* * Silly rename structure that hangs off the nfsnode until the name * can be removed by nfs_inactive() @@ -52,36 +56,73 @@ struct sillyrename { }; /* + * This structure is used to save the logical directory offset to + * NFS cookie mappings. + * The mappings are stored in a list headed + * by n_cookies, as required. + * There is one mapping for each NFS_DIRBLKSIZ bytes of directory information + * stored in increasing logical offset byte order. + */ +#define NFSNUMCOOKIES 31 + +struct nfsdmap { + LIST_ENTRY(nfsdmap) ndm_list; + int ndm_eocookie; + nfsuint64 ndm_cookies[NFSNUMCOOKIES]; +}; + +/* * The nfsnode is the nfs equivalent to ufs's inode. Any similarity * is purely coincidental. * There is a unique nfsnode allocated for each active file, * each current directory, each mounted-on file, text file, and the root. * An nfsnode is 'named' by its file handle. (nget/nfs_node.c) + * If this structure exceeds 256 bytes (it is currently 256 using 4.4BSD-Lite + * type definitions), file handles of > 32 bytes should probably be split out + * into a separate MALLOC()'d data structure. (Reduce the size of nfsfh_t by + * changing the definition in sys/mount.h of NFS_SMALLFH.) + * NB: Hopefully the current order of the fields is such that everything will + * be well aligned and, therefore, tightly packed. */ - struct nfsnode { - LIST_ENTRY(nfsnode) n_hash; /* Hash chain */ - CIRCLEQ_ENTRY(nfsnode) n_timer; /* Nqnfs timer chain */ - nfsv2fh_t n_fh; /* NFS File Handle */ - long n_flag; /* Flag for locking.. */ - struct vnode *n_vnode; /* vnode associated with this node */ - struct vattr n_vattr; /* Vnode attribute cache */ - time_t n_attrstamp; /* Time stamp for cached attributes */ - struct sillyrename *n_sillyrename; /* Ptr to silly rename struct */ - u_quad_t n_size; /* Current size of file */ - struct lockf *n_lockf; /* Locking record of file */ - int n_error; /* Save write error value */ - u_long n_direofoffset; /* Dir. EOF offset cache */ - time_t n_mtime; /* Prev modify time. */ - time_t n_ctime; /* Prev create time. */ - u_quad_t n_brev; /* Modify rev when cached */ - u_quad_t n_lrev; /* Modify rev for lease */ - time_t n_expiry; /* Lease expiry time */ - struct sillyrename n_silly; /* Silly rename struct */ - struct timeval n_atim; /* Special file times */ - struct timeval n_mtim; + LIST_ENTRY(nfsnode) n_hash; /* Hash chain */ + CIRCLEQ_ENTRY(nfsnode) n_timer; /* Nqnfs timer chain */ + u_quad_t n_size; /* Current size of file */ + u_quad_t n_brev; /* Modify rev when cached */ + u_quad_t n_lrev; /* Modify rev for lease */ + struct vattr n_vattr; /* Vnode attribute cache */ + time_t n_attrstamp; /* Attr. cache timestamp */ + time_t n_mtime; /* Prev modify time. */ + time_t n_ctime; /* Prev create time. */ + time_t n_expiry; /* Lease expiry time */ + nfsfh_t *n_fhp; /* NFS File Handle */ + struct vnode *n_vnode; /* associated vnode */ + struct lockf *n_lockf; /* Locking record of file */ + int n_error; /* Save write error value */ + union { + struct timespec nf_atim; /* Special file times */ + nfsuint64 nd_cookieverf; /* Cookie verifier (dir only) */ + } n_un1; + union { + struct timespec nf_mtim; + off_t nd_direof; /* Dir. EOF offset cache */ + } n_un2; + union { + struct sillyrename *nf_silly; /* Ptr to silly rename struct */ + LIST_HEAD(, nfsdmap) nd_cook; /* cookies */ + } n_un3; + short n_fhsize; /* size in bytes, of fh */ + short n_flag; /* Flag for locking.. */ + nfsfh_t n_fh; /* Small File Handle */ }; +#define n_atim n_un1.nf_atim +#define n_mtim n_un2.nf_mtim +#define n_sillyrename n_un3.nf_silly +#define n_cookieverf n_un1.nd_cookieverf +#define n_direofoffset n_un2.nd_direof +#define n_cookies n_un3.nd_cook + /* * Flags for n_flag */ @@ -95,6 +136,8 @@ struct nfsnode { #define NACC 0x0100 /* Special file accessed */ #define NUPD 0x0200 /* Special file updated */ #define NCHG 0x0400 /* Special file times changed */ +#define NLOCKED 0x0800 /* node is locked */ +#define NWANTED 0x0100 /* someone wants to lock */ /* * Convert between nfsnode pointers and vnode pointers @@ -107,7 +150,7 @@ struct nfsnode { */ TAILQ_HEAD(, buf) nfs_bufq; -#ifdef KERNEL +#if defined(KERNEL) || defined(_KERNEL) extern int (**fifo_nfsv2nodeop_p)(); extern int (**nfsv2_vnodeop_p)(); extern int (**spec_nfsv2nodeop_p)(); @@ -128,12 +171,25 @@ int nfs_getattr __P((struct vop_getattr_args *)); int nfs_setattr __P((struct vop_setattr_args *)); int nfs_read __P((struct vop_read_args *)); int nfs_write __P((struct vop_write_args *)); +#ifdef HAS_VOPLEASE +#define nfs_lease_check ((int (*) __P((struct vop_lease_args *)))nullop) +#define nqnfs_vop_lease_check lease_check +#else +#ifdef __FreeBSD__ +#define nqnfs_lease_check nfs_lease_check +#else +#define nqnfs_lease_check lease_check +#endif +#endif int nfsspec_read __P((struct vop_read_args *)); int nfsspec_write __P((struct vop_write_args *)); int nfsfifo_read __P((struct vop_read_args *)); int nfsfifo_write __P((struct vop_write_args *)); #define nfs_ioctl ((int (*) __P((struct vop_ioctl_args *)))enoioctl) #define nfs_select ((int (*) __P((struct vop_select_args *)))seltrue) +#ifdef HAS_VOPREVOKE +#define nfs_revoke vop_revoke +#endif int nfs_mmap __P((struct vop_mmap_args *)); int nfs_fsync __P((struct vop_fsync_args *)); #define nfs_seek ((int (*) __P((struct vop_seek_args *)))nullop) @@ -157,6 +213,7 @@ int nfs_islocked __P((struct vop_islocked_args *)); int nfs_pathconf __P((struct vop_pathconf_args *)); int nfs_advlock __P((struct vop_advlock_args *)); int nfs_blkatoff __P((struct vop_blkatoff_args *)); +int nfs_bwrite __P((struct vop_bwrite_args *)); int nfs_vget __P((struct mount *, ino_t, struct vnode **)); int nfs_valloc __P((struct vop_valloc_args *)); #define nfs_reallocblks \ @@ -164,13 +221,20 @@ int nfs_valloc __P((struct vop_valloc_args *)); int nfs_vfree __P((struct vop_vfree_args *)); int nfs_truncate __P((struct vop_truncate_args *)); int nfs_update __P((struct vop_update_args *)); -int nfs_bwrite __P((struct vop_bwrite_args *)); /* other stuff */ int nfs_removeit __P((struct sillyrename *)); -int nfs_nget __P((struct mount *,nfsv2fh_t *,struct nfsnode **)); -int nfs_lookitup __P((struct sillyrename *,nfsv2fh_t *,struct proc *)); +int nfs_nget __P((struct mount *,nfsfh_t *,int,struct nfsnode **)); +int nfs_lookitup __P((struct vnode *,char *,int,struct ucred *,struct proc *,struct nfsnode **)); int nfs_sillyrename __P((struct vnode *,struct vnode *,struct componentname *)); +nfsuint64 *nfs_getcookie __P((struct nfsnode *, off_t, int)); +void nfs_invaldir __P((struct vnode *)); + +#ifdef __FreeBSD__ +#define nqnfs_lease_updatetime nfs_lease_updatetime +#else +#define nqnfs_lease_updatetime lease_updatetime +#endif #endif /* KERNEL */ diff --git a/sys/nfsclient/nfsstats.h b/sys/nfsclient/nfsstats.h index cbf80c6..e1a0f07 100644 --- a/sys/nfsclient/nfsstats.h +++ b/sys/nfsclient/nfsstats.h @@ -34,7 +34,7 @@ * SUCH DAMAGE. * * @(#)nfs.h 8.1 (Berkeley) 6/10/93 - * $Id: nfs.h,v 1.8 1994/11/02 00:11:00 wollman Exp $ + * $Id: nfs.h,v 1.9 1995/02/14 06:22:18 phk Exp $ */ #ifndef _NFS_NFS_H_ @@ -45,11 +45,12 @@ */ #define NFS_MAXIOVEC 34 -#define NFS_HZ 25 /* Ticks per second for NFS timeouts */ -#define NFS_TIMEO (1*NFS_HZ) /* Default timeout = 1 second */ -#define NFS_MINTIMEO (1*NFS_HZ) /* Min timeout to use */ -#define NFS_MAXTIMEO (60*NFS_HZ) /* Max timeout to backoff to */ -#define NFS_MINIDEMTIMEO (5*NFS_HZ) /* Min timeout for non-idempotent ops*/ +#define NFS_TICKINTVL 5 /* Desired time for a tick (msec) */ +#define NFS_HZ (hz / nfs_ticks) /* Ticks/sec */ +#define NFS_TIMEO (1 * NFS_HZ) /* Default timeout = 1 second */ +#define NFS_MINTIMEO (1 * NFS_HZ) /* Min timeout to use */ +#define NFS_MAXTIMEO (60 * NFS_HZ) /* Max timeout to backoff to */ +#define NFS_MINIDEMTIMEO (5 * NFS_HZ) /* Min timeout for non-idempotent ops*/ #define NFS_MAXREXMIT 100 /* Stop counting after this many */ #define NFS_MAXWINDOW 1024 /* Max number of outstanding requests */ #define NFS_RETRANS 10 /* Num of retrans for soft mounts */ @@ -62,13 +63,81 @@ #endif #define NFS_WSIZE 8192 /* Def. write data size <= 8192 */ #define NFS_RSIZE 8192 /* Def. read data size <= 8192 */ +#define NFS_READDIRSIZE 8192 /* Def. readdir size */ #define NFS_DEFRAHEAD 1 /* Def. read ahead # blocks */ #define NFS_MAXRAHEAD 4 /* Max. read ahead # blocks */ -#define NFS_MAXREADDIR NFS_MAXDATA /* Max. size of directory read */ #define NFS_MAXUIDHASH 64 /* Max. # of hashed uid entries/mp */ -#define NFS_MAXASYNCDAEMON 20 /* Max. number async_daemons runable */ -#define NFS_DIRBLKSIZ 1024 /* Size of an NFS directory block */ +#define NFS_MAXASYNCDAEMON 20 /* Max. number async_daemons runable */ +#define NFS_MAXGATHERDELAY 100 /* Max. write gather delay (msec) */ +#ifndef NFS_GATHERDELAY +#define NFS_GATHERDELAY 10 /* Default write gather delay (msec) */ +#endif +#define NFS_DIRBLKSIZ 4096 /* Must be a multiple of DIRBLKSIZ */ + +/* + * Oddballs + */ #define NMOD(a) ((a) % nfs_asyncdaemons) +#define NFS_CMPFH(n, f, s) \ + ((n)->n_fhsize == (s) && !bcmp((caddr_t)(n)->n_fhp, (caddr_t)(f), (s))) +#define NFS_ISV3(v) (VFSTONFS((v)->v_mount)->nm_flag & NFSMNT_NFSV3) +#define NFS_SRVMAXDATA(n) \ + (((n)->nd_flag & ND_NFSV3) ? (((n)->nd_nam2) ? \ + NFS_MAXDGRAMDATA : NFS_MAXDATA) : NFS_V2MAXDATA) + +/* + * XXX + * sys/buf.h should be editted to change B_APPENDWRITE --> B_NEEDCOMMIT, but + * until then... + * Same goes for sys/malloc.h, which needs M_NFSDIROFF, + * M_NFSRVDESC and M_NFSBIGFH added. + * The VA_EXCLUSIVE flag should be added for va_vaflags and set for an + * exclusive create. + * The B_INVAFTERWRITE flag should be set to whatever is required by the + * buffer cache code to say "Invalidate the block after it is written back". + */ +#ifndef B_NEEDCOMMIT +#define B_NEEDCOMMIT B_APPENDWRITE +#endif +#ifndef M_NFSRVDESC +#define M_NFSRVDESC M_TEMP +#endif +#ifndef M_NFSDIROFF +#define M_NFSDIROFF M_TEMP +#endif +#ifndef M_NFSBIGFH +#define M_NFSBIGFH M_TEMP +#endif +#ifndef VA_EXCLUSIVE +#define VA_EXCLUSIVE 0 +#endif +#ifdef __FreeBSD__ +#define B_INVAFTERWRITE B_NOCACHE +#else +#define B_INVAFTERWRITE B_INVAL +#endif + +/* + * These ifdefs try to handle the differences between the various 4.4BSD-Lite + * based vfs interfaces. + * btw: NetBSD-current does have a VOP_LEASDE(), but I don't know how to + * differentiate between NetBSD-1.0 and NetBSD-current, so.. + * I also don't know about BSDi's 2.0 release. + */ +#if !defined(HAS_VOPLEASE) && !defined(__FreeBSD__) && !defined(__NetBSD__) +#define HAS_VOPLEASE 1 +#endif +#if !defined(HAS_VOPREVOKE) && !defined(__FreeBSD__) && !defined(__NetBSD__) +#define HAS_VOPREVOKE 1 +#endif + +/* + * The IO_METASYNC flag should be implemented for local file systems. + * (Until then, it is nothin at all.) + */ +#ifndef IO_METASYNC +#define IO_METASYNC 0 +#endif /* * Set the attribute timeout based on how recently the file has been modified. @@ -80,6 +149,20 @@ (time.tv_sec - (np)->n_mtime) / 10)) /* + * Expected allocation sizes for major data structures. If the actual size + * of the structure exceeds these sizes, then malloc() will be allocating + * almost twice the memory required. This is used in nfs_init() to warn + * the sysadmin that the size of a structure should be reduced. + * (These sizes are always a power of 2. If the kernel malloc() changes + * to one that does not allocate space in powers of 2 size, then this all + * becomes bunk!) + */ +#define NFS_NODEALLOC 256 +#define NFS_MNTALLOC 512 +#define NFS_SVCALLOC 256 +#define NFS_UIDALLOC 128 + +/* * Structures for the nfssvc(2) syscall. Not that anyone but nfsd and mount_nfs * should ever try and use it. */ @@ -95,7 +178,12 @@ struct nfsd_srvargs { u_long nsd_haddr; /* Ip address of client */ struct ucred nsd_cr; /* Cred. uid maps to */ int nsd_authlen; /* Length of auth string (ret) */ - char *nsd_authstr; /* Auth string (ret) */ + u_char *nsd_authstr; /* Auth string (ret) */ + int nsd_verflen; /* and the verfier */ + u_char *nsd_verfstr; + struct timeval nsd_timestamp; /* timestamp from verifier */ + u_long nsd_ttl; /* credential ttl (sec) */ + NFSKERBKEY_T nsd_key; /* Session key */ }; struct nfsd_cargs { @@ -103,7 +191,10 @@ struct nfsd_cargs { uid_t ncd_authuid; /* Effective uid */ int ncd_authtype; /* Type of authenticator */ int ncd_authlen; /* Length of authenticator string */ - char *ncd_authstr; /* Authenticator string */ + u_char *ncd_authstr; /* Authenticator string */ + int ncd_verflen; /* and the verifier */ + u_char *ncd_verfstr; + NFSKERBKEY_T ncd_key; /* Session key */ }; /* @@ -142,6 +233,7 @@ struct nfsstats { int srvnqnfs_leases; int srvnqnfs_maxleases; int srvnqnfs_getleases; + int srvvop_writes; }; /* @@ -173,7 +265,7 @@ struct nfsstats { * such as SIGALRM will not expect file I/O system calls to be interrupted * by them and break. */ -#ifdef KERNEL +#if defined(KERNEL) || defined(_KERNEL) struct uio; struct buf; struct vattr; struct nameidata; /* XXX */ @@ -224,17 +316,29 @@ TAILQ_HEAD(, nfsreq) nfs_reqq; #define R_MUSTRESEND 0x40 /* Must resend request */ #define R_GETONEREP 0x80 /* Probe for one reply only */ -extern struct nfsstats nfsstats; - /* * A list of nfssvc_sock structures is maintained with all the sockets * that require service by the nfsd. * The nfsuid structs hang off of the nfssvc_sock structs in both lru * and uid hash lists. */ -#define NUIDHASHSIZ 32 +#ifndef NFS_UIDHASHSIZ +#define NFS_UIDHASHSIZ 29 /* Tune the size of nfssvc_sock with this */ +#endif #define NUIDHASH(sock, uid) \ - (&(sock)->ns_uidhashtbl[(uid) & (sock)->ns_uidhash]) + (&(sock)->ns_uidhashtbl[(uid) % NFS_UIDHASHSIZ]) +#ifndef NFS_WDELAYHASHSIZ +#define NFS_WDELAYHASHSIZ 16 /* and with this */ +#endif +#define NWDELAYHASH(sock, f) \ + (&(sock)->ns_wdelayhashtbl[(*((u_long *)(f))) % NFS_WDELAYHASHSIZ]) +#ifndef NFS_MUIDHASHSIZ +#define NFS_MUIDHASHSIZ 67 /* Tune the size of nfsmount with this */ +#endif +#define NMUIDHASH(nmp, uid) \ + (&(nmp)->nm_uidhashtbl[(uid) % NFS_MUIDHASHSIZ]) +#define NFSNOHASH(fhsum) \ + (&nfsnodehashtbl[(fhsum) & nfsnodehash]) /* * Network address hash list element @@ -248,35 +352,41 @@ struct nfsuid { TAILQ_ENTRY(nfsuid) nu_lru; /* LRU chain */ LIST_ENTRY(nfsuid) nu_hash; /* Hash list */ int nu_flag; /* Flags */ - uid_t nu_uid; /* Uid mapped by this entry */ union nethostaddr nu_haddr; /* Host addr. for dgram sockets */ struct ucred nu_cr; /* Cred uid mapped to */ + int nu_expire; /* Expiry time (sec) */ + struct timeval nu_timestamp; /* Kerb. timestamp */ + u_long nu_nickname; /* Nickname on server */ + NFSKERBKEY_T nu_key; /* and session key */ }; #define nu_inetaddr nu_haddr.had_inetaddr #define nu_nam nu_haddr.had_nam /* Bits for nu_flag */ #define NU_INETADDR 0x1 +#define NU_NAM 0x2 +#define NU_NETFAM(u) (((u)->nu_flag & NU_INETADDR) ? AF_INET : AF_ISO) struct nfssvc_sock { TAILQ_ENTRY(nfssvc_sock) ns_chain; /* List of all nfssvc_sock's */ TAILQ_HEAD(, nfsuid) ns_uidlruhead; - LIST_HEAD(, nfsuid) *ns_uidhashtbl; - u_long ns_uidhash; - - int ns_flag; - u_long ns_sref; struct file *ns_fp; struct socket *ns_so; - int ns_solock; struct mbuf *ns_nam; - int ns_cc; struct mbuf *ns_raw; struct mbuf *ns_rawend; - int ns_reclen; struct mbuf *ns_rec; struct mbuf *ns_recend; + struct mbuf *ns_frag; + int ns_flag; + int ns_solock; + int ns_cc; + int ns_reclen; int ns_numuids; + u_long ns_sref; + LIST_HEAD(, nfsrv_descript) ns_tq; /* Write gather lists */ + LIST_HEAD(, nfsuid) ns_uidhashtbl[NFS_UIDHASHSIZ]; + LIST_HEAD(nfsrvw_delayhash, nfsrv_descript) ns_wdelayhashtbl[NFS_WDELAYHASHSIZ]; }; /* Bits for "ns_flag" */ @@ -285,6 +395,7 @@ struct nfssvc_sock { #define SLP_NEEDQ 0x04 #define SLP_DISCONN 0x08 #define SLP_GETSTREAM 0x10 +#define SLP_LASTFRAG 0x20 #define SLP_ALLFLAGS 0xff TAILQ_HEAD(, nfssvc_sock) nfssvc_sockhead; @@ -296,73 +407,124 @@ int nfssvc_sockhead_flag; * One of these structures is allocated for each nfsd. */ struct nfsd { - TAILQ_ENTRY(nfsd) nd_chain; /* List of all nfsd's */ - int nd_flag; /* NFSD_ flags */ - struct nfssvc_sock *nd_slp; /* Current socket */ - struct mbuf *nd_nam; /* Client addr for datagram req. */ - struct mbuf *nd_mrep; /* Req. mbuf list */ - struct mbuf *nd_md; - caddr_t nd_dpos; /* Position in list */ - int nd_procnum; /* RPC procedure number */ - u_long nd_retxid; /* RPC xid */ - int nd_repstat; /* Reply status value */ - struct ucred nd_cr; /* Credentials for req. */ - int nd_nqlflag; /* Leasing flag */ - u_long nd_duration; /* Lease duration */ - int nd_authlen; /* Authenticator len */ - u_char nd_authstr[RPCAUTH_MAXSIZ]; /* Authenticator data */ - struct proc *nd_procp; /* Proc ptr */ + TAILQ_ENTRY(nfsd) nfsd_chain; /* List of all nfsd's */ + int nfsd_flag; /* NFSD_ flags */ + struct nfssvc_sock *nfsd_slp; /* Current socket */ + int nfsd_authlen; /* Authenticator len */ + u_char nfsd_authstr[RPCAUTH_MAXSIZ]; /* Authenticator data */ + int nfsd_verflen; /* and the Verifier */ + u_char nfsd_verfstr[RPCVERF_MAXSIZ]; + struct proc *nfsd_procp; /* Proc ptr */ + struct nfsrv_descript *nfsd_nd; /* Associated nfsrv_descript */ }; -/* Bits for "nd_flag" */ +/* Bits for "nfsd_flag" */ #define NFSD_WAITING 0x01 #define NFSD_REQINPROG 0x02 #define NFSD_NEEDAUTH 0x04 #define NFSD_AUTHFAIL 0x08 +/* + * This structure is used by the server for describing each request. + * Some fields are used only when write request gathering is performed. + */ +struct nfsrv_descript { + u_quad_t nd_time; /* Write deadline (usec) */ + off_t nd_off; /* Start byte offset */ + off_t nd_eoff; /* and end byte offset */ + LIST_ENTRY(nfsrv_descript) nd_hash; /* Hash list */ + LIST_ENTRY(nfsrv_descript) nd_tq; /* and timer list */ + LIST_HEAD(,nfsrv_descript) nd_coalesce; /* coalesced writes */ + struct mbuf *nd_mrep; /* Request mbuf list */ + struct mbuf *nd_md; /* Current dissect mbuf */ + struct mbuf *nd_mreq; /* Reply mbuf list */ + struct mbuf *nd_nam; /* and socket addr */ + struct mbuf *nd_nam2; /* return socket addr */ + caddr_t nd_dpos; /* Current dissect pos */ + int nd_procnum; /* RPC # */ + int nd_stable; /* storage type */ + int nd_flag; /* nd_flag */ + int nd_len; /* Length of this write */ + int nd_repstat; /* Reply status */ + u_long nd_retxid; /* Reply xid */ + u_long nd_duration; /* Lease duration */ + struct timeval nd_starttime; /* Time RPC initiated */ + fhandle_t nd_fh; /* File handle */ + struct ucred nd_cr; /* Credentials */ +}; + +/* Bits for "nd_flag" */ +#define ND_READ LEASE_READ +#define ND_WRITE LEASE_WRITE +#define ND_CHECK 0x04 +#define ND_LEASE (ND_READ | ND_WRITE | ND_CHECK) +#define ND_NFSV3 0x08 +#define ND_NQNFS 0x10 +#define ND_KERBNICK 0x20 +#define ND_KERBFULL 0x40 +#define ND_KERBAUTH (ND_KERBNICK | ND_KERBFULL) + TAILQ_HEAD(, nfsd) nfsd_head; int nfsd_head_flag; #define NFSD_CHECKSLP 0x01 +/* + * These macros compare nfsrv_descript structures. + */ +#define NFSW_CONTIG(o, n) \ + ((o)->nd_eoff >= (n)->nd_off && \ + !bcmp((caddr_t)&(o)->nd_fh, (caddr_t)&(n)->nd_fh, NFSX_V3FH)) + +#define NFSW_SAMECRED(o, n) \ + (((o)->nd_flag & ND_KERBAUTH) == ((n)->nd_flag & ND_KERBAUTH) && \ + !bcmp((caddr_t)&(o)->nd_cr, (caddr_t)&(n)->nd_cr, \ + sizeof (struct ucred))) + int nfs_reply __P((struct nfsreq *)); -int nfs_getreq __P((struct nfsd *,int)); +int nfs_getreq __P((struct nfsrv_descript *,struct nfsd *,int)); int nfs_send __P((struct socket *,struct mbuf *,struct mbuf *,struct nfsreq *)); -int nfs_rephead __P((int,struct nfsd *,int,int,u_quad_t *,struct mbuf **,struct mbuf **,caddr_t *)); +int nfs_rephead __P((int,struct nfsrv_descript *,struct nfssvc_sock *,int,int,u_quad_t *,struct mbuf **,struct mbuf **,caddr_t *)); int nfs_sndlock __P((int *,struct nfsreq *)); int nfs_disct __P((struct mbuf **,caddr_t *,int,int,caddr_t *)); int nfs_vinvalbuf __P((struct vnode *,int,struct ucred *,struct proc *,int)); int nfs_readrpc __P((struct vnode *,struct uio *,struct ucred *)); -int nfs_writerpc __P((struct vnode *,struct uio *,struct ucred *,int)); +int nfs_writerpc __P((struct vnode *,struct uio *,struct ucred *,int *,int *)); int nfs_readdirrpc __P((register struct vnode *,struct uio *,struct ucred *)); +int nfs_setattrrpc __P((struct vnode *,struct vattr *,struct ucred *,struct proc *)); int nfs_asyncio __P((struct buf *,struct ucred *)); int nfs_doio __P((struct buf *,struct ucred *,struct proc *)); int nfs_readlinkrpc __P((struct vnode *,struct uio *,struct ucred *)); int nfs_sigintr __P((struct nfsmount *,struct nfsreq *r,struct proc *)); -int nfs_readdirlookrpc __P((struct vnode *,register struct uio *,struct ucred *)); +int nfs_readdirplusrpc __P((struct vnode *,register struct uio *,struct ucred *)); int nfsm_disct __P((struct mbuf **,caddr_t *,int,int,caddr_t *)); -int nfsrv_fhtovp __P((fhandle_t *,int,struct vnode **,struct ucred *,struct nfssvc_sock *,struct mbuf *,int *)); +void nfsm_srvfattr __P((struct nfsrv_descript *,struct vattr *,struct nfs_fattr *)); +void nfsm_srvwcc __P((struct nfsrv_descript *,int,struct vattr *,int,struct vattr *,struct mbuf **,char **)); +void nfsm_srvpostopattr __P((struct nfsrv_descript *,int,struct vattr *,struct mbuf **,char **)); +int nfsrv_fhtovp __P((fhandle_t *,int,struct vnode **,struct ucred *,struct nfssvc_sock *,struct mbuf *,int *,int)); int nfsrv_access __P((struct vnode *,int,struct ucred *,int,struct proc *)); int netaddr_match __P((int,union nethostaddr *,struct mbuf *)); int nfs_request __P((struct vnode *,struct mbuf *,int,struct proc *,struct ucred *,struct mbuf **,struct mbuf **,caddr_t *)); int nfs_loadattrcache __P((struct vnode **,struct mbuf **,caddr_t *,struct vattr *)); -int nfs_namei __P((struct nameidata *,fhandle_t *,int,struct nfssvc_sock *,struct mbuf *,struct mbuf **,caddr_t *,struct proc *)); +int nfs_namei __P((struct nameidata *,fhandle_t *,int,struct nfssvc_sock *,struct mbuf *,struct mbuf **,caddr_t *,struct vnode **,struct proc *,int)); void nfsm_adj __P((struct mbuf *,int,int)); int nfsm_mbuftouio __P((struct mbuf **,struct uio *,int,caddr_t *)); void nfsrv_initcache __P((void)); int nfs_rcvlock __P((struct nfsreq *)); -int nfs_getauth __P((struct nfsmount *,struct nfsreq *,struct ucred *,int *,char **,int *)); +int nfs_getauth __P((struct nfsmount *,struct nfsreq *,struct ucred *,char **,int *,char *,int *,NFSKERBKEY_T)); +int nfs_getnickauth __P((struct nfsmount *,struct ucred *,char **,int *,char *,int)); +int nfs_savenickauth __P((struct nfsmount *,struct ucred *,int,NFSKERBKEY_T,struct mbuf **,char **,struct mbuf *)); int nfs_msg __P((struct proc *,char *,char *)); int nfs_adv __P((struct mbuf **,caddr_t *,int,int)); int nfsrv_getstream __P((struct nfssvc_sock *,int)); void nfs_nhinit __P((void)); void nfs_timer __P((void*)); -struct nfsnodehashhead * nfs_hash __P((nfsv2fh_t *)); +u_long nfs_hash __P((nfsfh_t *,int)); int nfssvc_iod __P((struct proc *)); int nfssvc_nfsd __P((struct nfsd_srvargs *,caddr_t,struct proc *)); int nfssvc_addsock __P((struct file *,struct mbuf *)); -int nfsrv_dorec __P((struct nfssvc_sock *,struct nfsd *)); -int nfsrv_getcache __P((struct mbuf *,struct nfsd *,struct mbuf **)); -void nfsrv_updatecache __P((struct mbuf *,struct nfsd *,int,struct mbuf *)); +int nfsrv_dorec __P((struct nfssvc_sock *,struct nfsd *,struct nfsrv_descript **)); +int nfsrv_getcache __P((struct nfsrv_descript *,struct nfssvc_sock *,struct mbuf **)); +void nfsrv_updatecache __P((struct nfsrv_descript *,int,struct mbuf *)); int mountnfs __P((struct nfs_args *,struct mount *,struct mbuf *,char *,char *,struct vnode **)); int nfs_connect __P((struct nfsmount *,struct nfsreq *)); int nfs_getattrcache __P((struct vnode *,struct vattr *)); @@ -370,9 +532,20 @@ int nfsm_strtmbuf __P((struct mbuf **,char **,char *,long)); int nfs_bioread __P((struct vnode *,struct uio *,int,struct ucred *)); int nfsm_uiotombuf __P((struct uio *,struct mbuf **,int,caddr_t *)); void nfsrv_init __P((int)); +void nfs_clearcommit __P((struct mount *)); +int nfsrv_errmap __P((struct nfsrv_descript *, int)); +void nfsrvw_coalesce __P((struct nfsrv_descript *,struct nfsrv_descript *)); +void nfsrvw_sort __P((gid_t [],int)); +void nfsrv_setcred __P((struct ucred *,struct ucred *)); +int nfs_flush __P((struct vnode *,struct ucred *,int,struct proc *,int)); +int nfs_writebp __P((struct buf *,int)); int nfsrv_vput __P(( struct vnode * )); int nfsrv_vrele __P(( struct vnode * )); int nfsrv_vmio __P(( struct vnode * )); +int nfsrv_writegather __P((struct nfsrv_descript **, struct nfssvc_sock *, + struct proc *, struct mbuf **)); +int nfs_fsinfo __P((struct nfsmount *, struct vnode *, struct ucred *, + struct proc *p)); #endif /* KERNEL */ diff --git a/sys/nfsserver/nfs.h b/sys/nfsserver/nfs.h index cbf80c6..e1a0f07 100644 --- a/sys/nfsserver/nfs.h +++ b/sys/nfsserver/nfs.h @@ -34,7 +34,7 @@ * SUCH DAMAGE. * * @(#)nfs.h 8.1 (Berkeley) 6/10/93 - * $Id: nfs.h,v 1.8 1994/11/02 00:11:00 wollman Exp $ + * $Id: nfs.h,v 1.9 1995/02/14 06:22:18 phk Exp $ */ #ifndef _NFS_NFS_H_ @@ -45,11 +45,12 @@ */ #define NFS_MAXIOVEC 34 -#define NFS_HZ 25 /* Ticks per second for NFS timeouts */ -#define NFS_TIMEO (1*NFS_HZ) /* Default timeout = 1 second */ -#define NFS_MINTIMEO (1*NFS_HZ) /* Min timeout to use */ -#define NFS_MAXTIMEO (60*NFS_HZ) /* Max timeout to backoff to */ -#define NFS_MINIDEMTIMEO (5*NFS_HZ) /* Min timeout for non-idempotent ops*/ +#define NFS_TICKINTVL 5 /* Desired time for a tick (msec) */ +#define NFS_HZ (hz / nfs_ticks) /* Ticks/sec */ +#define NFS_TIMEO (1 * NFS_HZ) /* Default timeout = 1 second */ +#define NFS_MINTIMEO (1 * NFS_HZ) /* Min timeout to use */ +#define NFS_MAXTIMEO (60 * NFS_HZ) /* Max timeout to backoff to */ +#define NFS_MINIDEMTIMEO (5 * NFS_HZ) /* Min timeout for non-idempotent ops*/ #define NFS_MAXREXMIT 100 /* Stop counting after this many */ #define NFS_MAXWINDOW 1024 /* Max number of outstanding requests */ #define NFS_RETRANS 10 /* Num of retrans for soft mounts */ @@ -62,13 +63,81 @@ #endif #define NFS_WSIZE 8192 /* Def. write data size <= 8192 */ #define NFS_RSIZE 8192 /* Def. read data size <= 8192 */ +#define NFS_READDIRSIZE 8192 /* Def. readdir size */ #define NFS_DEFRAHEAD 1 /* Def. read ahead # blocks */ #define NFS_MAXRAHEAD 4 /* Max. read ahead # blocks */ -#define NFS_MAXREADDIR NFS_MAXDATA /* Max. size of directory read */ #define NFS_MAXUIDHASH 64 /* Max. # of hashed uid entries/mp */ -#define NFS_MAXASYNCDAEMON 20 /* Max. number async_daemons runable */ -#define NFS_DIRBLKSIZ 1024 /* Size of an NFS directory block */ +#define NFS_MAXASYNCDAEMON 20 /* Max. number async_daemons runable */ +#define NFS_MAXGATHERDELAY 100 /* Max. write gather delay (msec) */ +#ifndef NFS_GATHERDELAY +#define NFS_GATHERDELAY 10 /* Default write gather delay (msec) */ +#endif +#define NFS_DIRBLKSIZ 4096 /* Must be a multiple of DIRBLKSIZ */ + +/* + * Oddballs + */ #define NMOD(a) ((a) % nfs_asyncdaemons) +#define NFS_CMPFH(n, f, s) \ + ((n)->n_fhsize == (s) && !bcmp((caddr_t)(n)->n_fhp, (caddr_t)(f), (s))) +#define NFS_ISV3(v) (VFSTONFS((v)->v_mount)->nm_flag & NFSMNT_NFSV3) +#define NFS_SRVMAXDATA(n) \ + (((n)->nd_flag & ND_NFSV3) ? (((n)->nd_nam2) ? \ + NFS_MAXDGRAMDATA : NFS_MAXDATA) : NFS_V2MAXDATA) + +/* + * XXX + * sys/buf.h should be editted to change B_APPENDWRITE --> B_NEEDCOMMIT, but + * until then... + * Same goes for sys/malloc.h, which needs M_NFSDIROFF, + * M_NFSRVDESC and M_NFSBIGFH added. + * The VA_EXCLUSIVE flag should be added for va_vaflags and set for an + * exclusive create. + * The B_INVAFTERWRITE flag should be set to whatever is required by the + * buffer cache code to say "Invalidate the block after it is written back". + */ +#ifndef B_NEEDCOMMIT +#define B_NEEDCOMMIT B_APPENDWRITE +#endif +#ifndef M_NFSRVDESC +#define M_NFSRVDESC M_TEMP +#endif +#ifndef M_NFSDIROFF +#define M_NFSDIROFF M_TEMP +#endif +#ifndef M_NFSBIGFH +#define M_NFSBIGFH M_TEMP +#endif +#ifndef VA_EXCLUSIVE +#define VA_EXCLUSIVE 0 +#endif +#ifdef __FreeBSD__ +#define B_INVAFTERWRITE B_NOCACHE +#else +#define B_INVAFTERWRITE B_INVAL +#endif + +/* + * These ifdefs try to handle the differences between the various 4.4BSD-Lite + * based vfs interfaces. + * btw: NetBSD-current does have a VOP_LEASDE(), but I don't know how to + * differentiate between NetBSD-1.0 and NetBSD-current, so.. + * I also don't know about BSDi's 2.0 release. + */ +#if !defined(HAS_VOPLEASE) && !defined(__FreeBSD__) && !defined(__NetBSD__) +#define HAS_VOPLEASE 1 +#endif +#if !defined(HAS_VOPREVOKE) && !defined(__FreeBSD__) && !defined(__NetBSD__) +#define HAS_VOPREVOKE 1 +#endif + +/* + * The IO_METASYNC flag should be implemented for local file systems. + * (Until then, it is nothin at all.) + */ +#ifndef IO_METASYNC +#define IO_METASYNC 0 +#endif /* * Set the attribute timeout based on how recently the file has been modified. @@ -80,6 +149,20 @@ (time.tv_sec - (np)->n_mtime) / 10)) /* + * Expected allocation sizes for major data structures. If the actual size + * of the structure exceeds these sizes, then malloc() will be allocating + * almost twice the memory required. This is used in nfs_init() to warn + * the sysadmin that the size of a structure should be reduced. + * (These sizes are always a power of 2. If the kernel malloc() changes + * to one that does not allocate space in powers of 2 size, then this all + * becomes bunk!) + */ +#define NFS_NODEALLOC 256 +#define NFS_MNTALLOC 512 +#define NFS_SVCALLOC 256 +#define NFS_UIDALLOC 128 + +/* * Structures for the nfssvc(2) syscall. Not that anyone but nfsd and mount_nfs * should ever try and use it. */ @@ -95,7 +178,12 @@ struct nfsd_srvargs { u_long nsd_haddr; /* Ip address of client */ struct ucred nsd_cr; /* Cred. uid maps to */ int nsd_authlen; /* Length of auth string (ret) */ - char *nsd_authstr; /* Auth string (ret) */ + u_char *nsd_authstr; /* Auth string (ret) */ + int nsd_verflen; /* and the verfier */ + u_char *nsd_verfstr; + struct timeval nsd_timestamp; /* timestamp from verifier */ + u_long nsd_ttl; /* credential ttl (sec) */ + NFSKERBKEY_T nsd_key; /* Session key */ }; struct nfsd_cargs { @@ -103,7 +191,10 @@ struct nfsd_cargs { uid_t ncd_authuid; /* Effective uid */ int ncd_authtype; /* Type of authenticator */ int ncd_authlen; /* Length of authenticator string */ - char *ncd_authstr; /* Authenticator string */ + u_char *ncd_authstr; /* Authenticator string */ + int ncd_verflen; /* and the verifier */ + u_char *ncd_verfstr; + NFSKERBKEY_T ncd_key; /* Session key */ }; /* @@ -142,6 +233,7 @@ struct nfsstats { int srvnqnfs_leases; int srvnqnfs_maxleases; int srvnqnfs_getleases; + int srvvop_writes; }; /* @@ -173,7 +265,7 @@ struct nfsstats { * such as SIGALRM will not expect file I/O system calls to be interrupted * by them and break. */ -#ifdef KERNEL +#if defined(KERNEL) || defined(_KERNEL) struct uio; struct buf; struct vattr; struct nameidata; /* XXX */ @@ -224,17 +316,29 @@ TAILQ_HEAD(, nfsreq) nfs_reqq; #define R_MUSTRESEND 0x40 /* Must resend request */ #define R_GETONEREP 0x80 /* Probe for one reply only */ -extern struct nfsstats nfsstats; - /* * A list of nfssvc_sock structures is maintained with all the sockets * that require service by the nfsd. * The nfsuid structs hang off of the nfssvc_sock structs in both lru * and uid hash lists. */ -#define NUIDHASHSIZ 32 +#ifndef NFS_UIDHASHSIZ +#define NFS_UIDHASHSIZ 29 /* Tune the size of nfssvc_sock with this */ +#endif #define NUIDHASH(sock, uid) \ - (&(sock)->ns_uidhashtbl[(uid) & (sock)->ns_uidhash]) + (&(sock)->ns_uidhashtbl[(uid) % NFS_UIDHASHSIZ]) +#ifndef NFS_WDELAYHASHSIZ +#define NFS_WDELAYHASHSIZ 16 /* and with this */ +#endif +#define NWDELAYHASH(sock, f) \ + (&(sock)->ns_wdelayhashtbl[(*((u_long *)(f))) % NFS_WDELAYHASHSIZ]) +#ifndef NFS_MUIDHASHSIZ +#define NFS_MUIDHASHSIZ 67 /* Tune the size of nfsmount with this */ +#endif +#define NMUIDHASH(nmp, uid) \ + (&(nmp)->nm_uidhashtbl[(uid) % NFS_MUIDHASHSIZ]) +#define NFSNOHASH(fhsum) \ + (&nfsnodehashtbl[(fhsum) & nfsnodehash]) /* * Network address hash list element @@ -248,35 +352,41 @@ struct nfsuid { TAILQ_ENTRY(nfsuid) nu_lru; /* LRU chain */ LIST_ENTRY(nfsuid) nu_hash; /* Hash list */ int nu_flag; /* Flags */ - uid_t nu_uid; /* Uid mapped by this entry */ union nethostaddr nu_haddr; /* Host addr. for dgram sockets */ struct ucred nu_cr; /* Cred uid mapped to */ + int nu_expire; /* Expiry time (sec) */ + struct timeval nu_timestamp; /* Kerb. timestamp */ + u_long nu_nickname; /* Nickname on server */ + NFSKERBKEY_T nu_key; /* and session key */ }; #define nu_inetaddr nu_haddr.had_inetaddr #define nu_nam nu_haddr.had_nam /* Bits for nu_flag */ #define NU_INETADDR 0x1 +#define NU_NAM 0x2 +#define NU_NETFAM(u) (((u)->nu_flag & NU_INETADDR) ? AF_INET : AF_ISO) struct nfssvc_sock { TAILQ_ENTRY(nfssvc_sock) ns_chain; /* List of all nfssvc_sock's */ TAILQ_HEAD(, nfsuid) ns_uidlruhead; - LIST_HEAD(, nfsuid) *ns_uidhashtbl; - u_long ns_uidhash; - - int ns_flag; - u_long ns_sref; struct file *ns_fp; struct socket *ns_so; - int ns_solock; struct mbuf *ns_nam; - int ns_cc; struct mbuf *ns_raw; struct mbuf *ns_rawend; - int ns_reclen; struct mbuf *ns_rec; struct mbuf *ns_recend; + struct mbuf *ns_frag; + int ns_flag; + int ns_solock; + int ns_cc; + int ns_reclen; int ns_numuids; + u_long ns_sref; + LIST_HEAD(, nfsrv_descript) ns_tq; /* Write gather lists */ + LIST_HEAD(, nfsuid) ns_uidhashtbl[NFS_UIDHASHSIZ]; + LIST_HEAD(nfsrvw_delayhash, nfsrv_descript) ns_wdelayhashtbl[NFS_WDELAYHASHSIZ]; }; /* Bits for "ns_flag" */ @@ -285,6 +395,7 @@ struct nfssvc_sock { #define SLP_NEEDQ 0x04 #define SLP_DISCONN 0x08 #define SLP_GETSTREAM 0x10 +#define SLP_LASTFRAG 0x20 #define SLP_ALLFLAGS 0xff TAILQ_HEAD(, nfssvc_sock) nfssvc_sockhead; @@ -296,73 +407,124 @@ int nfssvc_sockhead_flag; * One of these structures is allocated for each nfsd. */ struct nfsd { - TAILQ_ENTRY(nfsd) nd_chain; /* List of all nfsd's */ - int nd_flag; /* NFSD_ flags */ - struct nfssvc_sock *nd_slp; /* Current socket */ - struct mbuf *nd_nam; /* Client addr for datagram req. */ - struct mbuf *nd_mrep; /* Req. mbuf list */ - struct mbuf *nd_md; - caddr_t nd_dpos; /* Position in list */ - int nd_procnum; /* RPC procedure number */ - u_long nd_retxid; /* RPC xid */ - int nd_repstat; /* Reply status value */ - struct ucred nd_cr; /* Credentials for req. */ - int nd_nqlflag; /* Leasing flag */ - u_long nd_duration; /* Lease duration */ - int nd_authlen; /* Authenticator len */ - u_char nd_authstr[RPCAUTH_MAXSIZ]; /* Authenticator data */ - struct proc *nd_procp; /* Proc ptr */ + TAILQ_ENTRY(nfsd) nfsd_chain; /* List of all nfsd's */ + int nfsd_flag; /* NFSD_ flags */ + struct nfssvc_sock *nfsd_slp; /* Current socket */ + int nfsd_authlen; /* Authenticator len */ + u_char nfsd_authstr[RPCAUTH_MAXSIZ]; /* Authenticator data */ + int nfsd_verflen; /* and the Verifier */ + u_char nfsd_verfstr[RPCVERF_MAXSIZ]; + struct proc *nfsd_procp; /* Proc ptr */ + struct nfsrv_descript *nfsd_nd; /* Associated nfsrv_descript */ }; -/* Bits for "nd_flag" */ +/* Bits for "nfsd_flag" */ #define NFSD_WAITING 0x01 #define NFSD_REQINPROG 0x02 #define NFSD_NEEDAUTH 0x04 #define NFSD_AUTHFAIL 0x08 +/* + * This structure is used by the server for describing each request. + * Some fields are used only when write request gathering is performed. + */ +struct nfsrv_descript { + u_quad_t nd_time; /* Write deadline (usec) */ + off_t nd_off; /* Start byte offset */ + off_t nd_eoff; /* and end byte offset */ + LIST_ENTRY(nfsrv_descript) nd_hash; /* Hash list */ + LIST_ENTRY(nfsrv_descript) nd_tq; /* and timer list */ + LIST_HEAD(,nfsrv_descript) nd_coalesce; /* coalesced writes */ + struct mbuf *nd_mrep; /* Request mbuf list */ + struct mbuf *nd_md; /* Current dissect mbuf */ + struct mbuf *nd_mreq; /* Reply mbuf list */ + struct mbuf *nd_nam; /* and socket addr */ + struct mbuf *nd_nam2; /* return socket addr */ + caddr_t nd_dpos; /* Current dissect pos */ + int nd_procnum; /* RPC # */ + int nd_stable; /* storage type */ + int nd_flag; /* nd_flag */ + int nd_len; /* Length of this write */ + int nd_repstat; /* Reply status */ + u_long nd_retxid; /* Reply xid */ + u_long nd_duration; /* Lease duration */ + struct timeval nd_starttime; /* Time RPC initiated */ + fhandle_t nd_fh; /* File handle */ + struct ucred nd_cr; /* Credentials */ +}; + +/* Bits for "nd_flag" */ +#define ND_READ LEASE_READ +#define ND_WRITE LEASE_WRITE +#define ND_CHECK 0x04 +#define ND_LEASE (ND_READ | ND_WRITE | ND_CHECK) +#define ND_NFSV3 0x08 +#define ND_NQNFS 0x10 +#define ND_KERBNICK 0x20 +#define ND_KERBFULL 0x40 +#define ND_KERBAUTH (ND_KERBNICK | ND_KERBFULL) + TAILQ_HEAD(, nfsd) nfsd_head; int nfsd_head_flag; #define NFSD_CHECKSLP 0x01 +/* + * These macros compare nfsrv_descript structures. + */ +#define NFSW_CONTIG(o, n) \ + ((o)->nd_eoff >= (n)->nd_off && \ + !bcmp((caddr_t)&(o)->nd_fh, (caddr_t)&(n)->nd_fh, NFSX_V3FH)) + +#define NFSW_SAMECRED(o, n) \ + (((o)->nd_flag & ND_KERBAUTH) == ((n)->nd_flag & ND_KERBAUTH) && \ + !bcmp((caddr_t)&(o)->nd_cr, (caddr_t)&(n)->nd_cr, \ + sizeof (struct ucred))) + int nfs_reply __P((struct nfsreq *)); -int nfs_getreq __P((struct nfsd *,int)); +int nfs_getreq __P((struct nfsrv_descript *,struct nfsd *,int)); int nfs_send __P((struct socket *,struct mbuf *,struct mbuf *,struct nfsreq *)); -int nfs_rephead __P((int,struct nfsd *,int,int,u_quad_t *,struct mbuf **,struct mbuf **,caddr_t *)); +int nfs_rephead __P((int,struct nfsrv_descript *,struct nfssvc_sock *,int,int,u_quad_t *,struct mbuf **,struct mbuf **,caddr_t *)); int nfs_sndlock __P((int *,struct nfsreq *)); int nfs_disct __P((struct mbuf **,caddr_t *,int,int,caddr_t *)); int nfs_vinvalbuf __P((struct vnode *,int,struct ucred *,struct proc *,int)); int nfs_readrpc __P((struct vnode *,struct uio *,struct ucred *)); -int nfs_writerpc __P((struct vnode *,struct uio *,struct ucred *,int)); +int nfs_writerpc __P((struct vnode *,struct uio *,struct ucred *,int *,int *)); int nfs_readdirrpc __P((register struct vnode *,struct uio *,struct ucred *)); +int nfs_setattrrpc __P((struct vnode *,struct vattr *,struct ucred *,struct proc *)); int nfs_asyncio __P((struct buf *,struct ucred *)); int nfs_doio __P((struct buf *,struct ucred *,struct proc *)); int nfs_readlinkrpc __P((struct vnode *,struct uio *,struct ucred *)); int nfs_sigintr __P((struct nfsmount *,struct nfsreq *r,struct proc *)); -int nfs_readdirlookrpc __P((struct vnode *,register struct uio *,struct ucred *)); +int nfs_readdirplusrpc __P((struct vnode *,register struct uio *,struct ucred *)); int nfsm_disct __P((struct mbuf **,caddr_t *,int,int,caddr_t *)); -int nfsrv_fhtovp __P((fhandle_t *,int,struct vnode **,struct ucred *,struct nfssvc_sock *,struct mbuf *,int *)); +void nfsm_srvfattr __P((struct nfsrv_descript *,struct vattr *,struct nfs_fattr *)); +void nfsm_srvwcc __P((struct nfsrv_descript *,int,struct vattr *,int,struct vattr *,struct mbuf **,char **)); +void nfsm_srvpostopattr __P((struct nfsrv_descript *,int,struct vattr *,struct mbuf **,char **)); +int nfsrv_fhtovp __P((fhandle_t *,int,struct vnode **,struct ucred *,struct nfssvc_sock *,struct mbuf *,int *,int)); int nfsrv_access __P((struct vnode *,int,struct ucred *,int,struct proc *)); int netaddr_match __P((int,union nethostaddr *,struct mbuf *)); int nfs_request __P((struct vnode *,struct mbuf *,int,struct proc *,struct ucred *,struct mbuf **,struct mbuf **,caddr_t *)); int nfs_loadattrcache __P((struct vnode **,struct mbuf **,caddr_t *,struct vattr *)); -int nfs_namei __P((struct nameidata *,fhandle_t *,int,struct nfssvc_sock *,struct mbuf *,struct mbuf **,caddr_t *,struct proc *)); +int nfs_namei __P((struct nameidata *,fhandle_t *,int,struct nfssvc_sock *,struct mbuf *,struct mbuf **,caddr_t *,struct vnode **,struct proc *,int)); void nfsm_adj __P((struct mbuf *,int,int)); int nfsm_mbuftouio __P((struct mbuf **,struct uio *,int,caddr_t *)); void nfsrv_initcache __P((void)); int nfs_rcvlock __P((struct nfsreq *)); -int nfs_getauth __P((struct nfsmount *,struct nfsreq *,struct ucred *,int *,char **,int *)); +int nfs_getauth __P((struct nfsmount *,struct nfsreq *,struct ucred *,char **,int *,char *,int *,NFSKERBKEY_T)); +int nfs_getnickauth __P((struct nfsmount *,struct ucred *,char **,int *,char *,int)); +int nfs_savenickauth __P((struct nfsmount *,struct ucred *,int,NFSKERBKEY_T,struct mbuf **,char **,struct mbuf *)); int nfs_msg __P((struct proc *,char *,char *)); int nfs_adv __P((struct mbuf **,caddr_t *,int,int)); int nfsrv_getstream __P((struct nfssvc_sock *,int)); void nfs_nhinit __P((void)); void nfs_timer __P((void*)); -struct nfsnodehashhead * nfs_hash __P((nfsv2fh_t *)); +u_long nfs_hash __P((nfsfh_t *,int)); int nfssvc_iod __P((struct proc *)); int nfssvc_nfsd __P((struct nfsd_srvargs *,caddr_t,struct proc *)); int nfssvc_addsock __P((struct file *,struct mbuf *)); -int nfsrv_dorec __P((struct nfssvc_sock *,struct nfsd *)); -int nfsrv_getcache __P((struct mbuf *,struct nfsd *,struct mbuf **)); -void nfsrv_updatecache __P((struct mbuf *,struct nfsd *,int,struct mbuf *)); +int nfsrv_dorec __P((struct nfssvc_sock *,struct nfsd *,struct nfsrv_descript **)); +int nfsrv_getcache __P((struct nfsrv_descript *,struct nfssvc_sock *,struct mbuf **)); +void nfsrv_updatecache __P((struct nfsrv_descript *,int,struct mbuf *)); int mountnfs __P((struct nfs_args *,struct mount *,struct mbuf *,char *,char *,struct vnode **)); int nfs_connect __P((struct nfsmount *,struct nfsreq *)); int nfs_getattrcache __P((struct vnode *,struct vattr *)); @@ -370,9 +532,20 @@ int nfsm_strtmbuf __P((struct mbuf **,char **,char *,long)); int nfs_bioread __P((struct vnode *,struct uio *,int,struct ucred *)); int nfsm_uiotombuf __P((struct uio *,struct mbuf **,int,caddr_t *)); void nfsrv_init __P((int)); +void nfs_clearcommit __P((struct mount *)); +int nfsrv_errmap __P((struct nfsrv_descript *, int)); +void nfsrvw_coalesce __P((struct nfsrv_descript *,struct nfsrv_descript *)); +void nfsrvw_sort __P((gid_t [],int)); +void nfsrv_setcred __P((struct ucred *,struct ucred *)); +int nfs_flush __P((struct vnode *,struct ucred *,int,struct proc *,int)); +int nfs_writebp __P((struct buf *,int)); int nfsrv_vput __P(( struct vnode * )); int nfsrv_vrele __P(( struct vnode * )); int nfsrv_vmio __P(( struct vnode * )); +int nfsrv_writegather __P((struct nfsrv_descript **, struct nfssvc_sock *, + struct proc *, struct mbuf **)); +int nfs_fsinfo __P((struct nfsmount *, struct vnode *, struct ucred *, + struct proc *p)); #endif /* KERNEL */ diff --git a/sys/nfsserver/nfs_serv.c b/sys/nfsserver/nfs_serv.c index 033dae4..f82db3b 100644 --- a/sys/nfsserver/nfs_serv.c +++ b/sys/nfsserver/nfs_serv.c @@ -34,11 +34,11 @@ * SUCH DAMAGE. * * @(#)nfs_serv.c 8.3 (Berkeley) 1/12/94 - * $Id: nfs_serv.c,v 1.14.2.1 1995/06/07 07:25:09 davidg Exp $ + * $Id: nfs_serv.c,v 1.15 1995/06/11 19:31:45 rgrimes Exp $ */ /* - * nfs version 2 server calls to vnode ops + * nfs version 2 and 3 server calls to vnode ops * - these routines generally have 3 phases * 1 - break down and validate rpc request in mbuf list * 2 - do the vnode ops for the request @@ -53,6 +53,8 @@ * returning an error from the server function implies a fatal error * such as a badly constructed rpc request that should be dropped without * a reply. + * For Version 3, nfsm_reply() does not return for the error case, since + * most version 3 rpcs return more than the status for error cases. */ #include <sys/param.h> @@ -62,67 +64,98 @@ #include <sys/namei.h> #include <sys/vnode.h> #include <sys/mount.h> +#include <sys/socket.h> +#include <sys/socketvar.h> #include <sys/mbuf.h> #include <sys/dirent.h> #include <sys/stat.h> +#include <sys/kernel.h> +#include <ufs/ufs/dir.h> #include <vm/vm.h> -#include <nfs/nfsv2.h> +#include <nfs/nfsproto.h> #include <nfs/rpcv2.h> #include <nfs/nfs.h> #include <nfs/xdr_subs.h> #include <nfs/nfsm_subs.h> #include <nfs/nqnfs.h> -/* Defs */ -#define TRUE 1 -#define FALSE 0 - /* Global vars */ -extern u_long nfs_procids[NFS_NPROCS]; extern u_long nfs_xdrneg1; extern u_long nfs_false, nfs_true; -nfstype nfs_type[9] = { NFNON, NFREG, NFDIR, NFBLK, NFCHR, NFLNK, NFNON, +extern enum vtype nv3tov_type[8]; +extern struct nfsstats nfsstats; +nfstype nfsv2_type[9] = { NFNON, NFREG, NFDIR, NFBLK, NFCHR, NFLNK, NFNON, NFCHR, NFNON }; +nfstype nfsv3_type[9] = { NFNON, NFREG, NFDIR, NFBLK, NFCHR, NFLNK, NFSOCK, + NFFIFO, NFNON }; +int nfsrvw_procrastinate = NFS_GATHERDELAY * 1000; /* - * nqnfs access service + * nfs v3 access service */ int -nqnfsrv_access(nfsd, mrep, md, dpos, cred, nam, mrq) - struct nfsd *nfsd; - struct mbuf *mrep, *md; - caddr_t dpos; - struct ucred *cred; - struct mbuf *nam, **mrq; +nfsrv3_access(nfsd, slp, procp, mrq) + struct nfsrv_descript *nfsd; + struct nfssvc_sock *slp; + struct proc *procp; + struct mbuf **mrq; { + struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; + struct mbuf *nam = nfsd->nd_nam; + caddr_t dpos = nfsd->nd_dpos; + struct ucred *cred = &nfsd->nd_cr; struct vnode *vp; - nfsv2fh_t nfh; + nfsfh_t nfh; fhandle_t *fhp; register u_long *tl; register long t1; caddr_t bpos; - int error = 0, rdonly, cache = 0, mode = 0; + int error = 0, rdonly, cache, getret; char *cp2; - struct mbuf *mb, *mreq; + struct mbuf *mb, *mreq, *mb2; + struct vattr vattr, *vap = &vattr; + u_long testmode, nfsmode; u_quad_t frev; +#ifndef nolint + cache = 0; +#endif fhp = &nfh.fh_generic; nfsm_srvmtofh(fhp); - nfsm_dissect(tl, u_long *, 3 * NFSX_UNSIGNED); - error = nfsrv_fhtovp(fhp, TRUE, &vp, cred, nfsd->nd_slp, nam, &rdonly); - if (error) - nfsm_reply(0); - if (*tl++ == nfs_true) - mode |= VREAD; - if (*tl++ == nfs_true) - mode |= VWRITE; - if (*tl == nfs_true) - mode |= VEXEC; - error = nfsrv_access(vp, mode, cred, rdonly, nfsd->nd_procp); + nfsm_dissect(tl, u_long *, NFSX_UNSIGNED); + if (error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam, + &rdonly, (nfsd->nd_flag & ND_KERBAUTH))) { + nfsm_reply(NFSX_UNSIGNED); + nfsm_srvpostop_attr(1, (struct vattr *)0); + return (0); + } + nfsmode = fxdr_unsigned(u_long, *tl); + if ((nfsmode & NFSV3ACCESS_READ) && + nfsrv_access(vp, VREAD, cred, rdonly, procp)) + nfsmode &= ~NFSV3ACCESS_READ; + if (vp->v_type == VDIR) + testmode = (NFSV3ACCESS_MODIFY | NFSV3ACCESS_EXTEND | + NFSV3ACCESS_DELETE); + else + testmode = (NFSV3ACCESS_MODIFY | NFSV3ACCESS_EXTEND); + if ((nfsmode & testmode) && + nfsrv_access(vp, VWRITE, cred, rdonly, procp)) + nfsmode &= ~testmode; + if (vp->v_type == VDIR) + testmode = NFSV3ACCESS_LOOKUP; + else + testmode = NFSV3ACCESS_EXECUTE; + if ((nfsmode & testmode) && + nfsrv_access(vp, VEXEC, cred, rdonly, procp)) + nfsmode &= ~testmode; + getret = VOP_GETATTR(vp, vap, cred, procp); nfsrv_vput(vp); - nfsm_reply(0); + nfsm_reply(NFSX_POSTOPATTR(1) + NFSX_UNSIGNED); + nfsm_srvpostop_attr(getret, vap); + nfsm_build(tl, u_long *, NFSX_UNSIGNED); + *tl = txdr_unsigned(nfsmode); nfsm_srvdone; } @@ -130,18 +163,21 @@ nqnfsrv_access(nfsd, mrep, md, dpos, cred, nam, mrq) * nfs getattr service */ int -nfsrv_getattr(nfsd, mrep, md, dpos, cred, nam, mrq) - struct nfsd *nfsd; - struct mbuf *mrep, *md; - caddr_t dpos; - struct ucred *cred; - struct mbuf *nam, **mrq; +nfsrv_getattr(nfsd, slp, procp, mrq) + struct nfsrv_descript *nfsd; + struct nfssvc_sock *slp; + struct proc *procp; + struct mbuf **mrq; { - register struct nfsv2_fattr *fp; + struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; + struct mbuf *nam = nfsd->nd_nam; + caddr_t dpos = nfsd->nd_dpos; + struct ucred *cred = &nfsd->nd_cr; + register struct nfs_fattr *fp; struct vattr va; register struct vattr *vap = &va; struct vnode *vp; - nfsv2fh_t nfh; + nfsfh_t nfh; fhandle_t *fhp; register u_long *tl; register long t1; @@ -153,15 +189,19 @@ nfsrv_getattr(nfsd, mrep, md, dpos, cred, nam, mrq) fhp = &nfh.fh_generic; nfsm_srvmtofh(fhp); - error = nfsrv_fhtovp(fhp, TRUE, &vp, cred, nfsd->nd_slp, nam, &rdonly); - if (error) + if (error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam, + &rdonly, (nfsd->nd_flag & ND_KERBAUTH))) { nfsm_reply(0); - nqsrv_getl(vp, NQL_READ); - error = VOP_GETATTR(vp, vap, cred, nfsd->nd_procp); + return (0); + } + nqsrv_getl(vp, ND_READ); + error = VOP_GETATTR(vp, vap, cred, procp); nfsrv_vput(vp); - nfsm_reply(NFSX_FATTR(nfsd->nd_nqlflag != NQL_NOVAL)); - nfsm_build(fp, struct nfsv2_fattr *, NFSX_FATTR(nfsd->nd_nqlflag != NQL_NOVAL)); - nfsm_srvfillattr; + nfsm_reply(NFSX_FATTR(nfsd->nd_flag & ND_NFSV3)); + if (error) + return (0); + nfsm_build(fp, struct nfs_fattr *, NFSX_FATTR(nfsd->nd_flag & ND_NFSV3)); + nfsm_srvfillattr(vap, fp); nfsm_srvdone; } @@ -169,68 +209,97 @@ nfsrv_getattr(nfsd, mrep, md, dpos, cred, nam, mrq) * nfs setattr service */ int -nfsrv_setattr(nfsd, mrep, md, dpos, cred, nam, mrq) - struct nfsd *nfsd; - struct mbuf *mrep, *md; - caddr_t dpos; - struct ucred *cred; - struct mbuf *nam, **mrq; +nfsrv_setattr(nfsd, slp, procp, mrq) + struct nfsrv_descript *nfsd; + struct nfssvc_sock *slp; + struct proc *procp; + struct mbuf **mrq; { - struct vattr va; + struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; + struct mbuf *nam = nfsd->nd_nam; + caddr_t dpos = nfsd->nd_dpos; + struct ucred *cred = &nfsd->nd_cr; + struct vattr va, preat; register struct vattr *vap = &va; register struct nfsv2_sattr *sp; - register struct nfsv2_fattr *fp; + register struct nfs_fattr *fp; struct vnode *vp; - nfsv2fh_t nfh; + nfsfh_t nfh; fhandle_t *fhp; register u_long *tl; register long t1; caddr_t bpos; - int error = 0, rdonly, cache; + int error = 0, rdonly, cache, preat_ret = 1, postat_ret = 1; + int v3 = (nfsd->nd_flag & ND_NFSV3), gcheck = 0; char *cp2; struct mbuf *mb, *mb2, *mreq; - u_quad_t frev, frev2; + u_quad_t frev; + struct timespec guard; fhp = &nfh.fh_generic; nfsm_srvmtofh(fhp); - nfsm_dissect(sp, struct nfsv2_sattr *, NFSX_SATTR(nfsd->nd_nqlflag != NQL_NOVAL)); - error = nfsrv_fhtovp(fhp, TRUE, &vp, cred, nfsd->nd_slp, nam, &rdonly); - if (error) - nfsm_reply(0); - nqsrv_getl(vp, NQL_WRITE); VATTR_NULL(vap); - /* - * Nah nah nah nah na nah - * There is a bug in the Sun client that puts 0xffff in the mode - * field of sattr when it should put in 0xffffffff. The u_short - * doesn't sign extend. - * --> check the low order 2 bytes for 0xffff - */ - if ((fxdr_unsigned(int, sp->sa_mode) & 0xffff) != 0xffff) - vap->va_mode = nfstov_mode(sp->sa_mode); - if (sp->sa_uid != nfs_xdrneg1) - vap->va_uid = fxdr_unsigned(uid_t, sp->sa_uid); - if (sp->sa_gid != nfs_xdrneg1) - vap->va_gid = fxdr_unsigned(gid_t, sp->sa_gid); - if (nfsd->nd_nqlflag == NQL_NOVAL) { - if (sp->sa_nfssize != nfs_xdrneg1) - vap->va_size = fxdr_unsigned(u_quad_t, sp->sa_nfssize); - if (sp->sa_nfsatime.nfs_sec != nfs_xdrneg1) { + if (v3) { + nfsm_srvsattr(vap); + nfsm_dissect(tl, u_long *, NFSX_UNSIGNED); + gcheck = fxdr_unsigned(int, *tl); + if (gcheck) { + nfsm_dissect(tl, u_long *, 2 * NFSX_UNSIGNED); + fxdr_nfsv3time(tl, &guard); + } + } else { + nfsm_dissect(sp, struct nfsv2_sattr *, NFSX_V2SATTR); + /* + * Nah nah nah nah na nah + * There is a bug in the Sun client that puts 0xffff in the mode + * field of sattr when it should put in 0xffffffff. The u_short + * doesn't sign extend. + * --> check the low order 2 bytes for 0xffff + */ + if ((fxdr_unsigned(int, sp->sa_mode) & 0xffff) != 0xffff) + vap->va_mode = nfstov_mode(sp->sa_mode); + if (sp->sa_uid != nfs_xdrneg1) + vap->va_uid = fxdr_unsigned(uid_t, sp->sa_uid); + if (sp->sa_gid != nfs_xdrneg1) + vap->va_gid = fxdr_unsigned(gid_t, sp->sa_gid); + if (sp->sa_size != nfs_xdrneg1) + vap->va_size = fxdr_unsigned(u_quad_t, sp->sa_size); + if (sp->sa_atime.nfsv2_sec != nfs_xdrneg1) { #ifdef notyet - fxdr_nfstime(&sp->sa_nfsatime, &vap->va_atime); + fxdr_nfsv2time(&sp->sa_atime, &vap->va_atime); #else vap->va_atime.ts_sec = - fxdr_unsigned(long, sp->sa_nfsatime.nfs_sec); + fxdr_unsigned(long, sp->sa_atime.nfsv2_sec); vap->va_atime.ts_nsec = 0; #endif } - if (sp->sa_nfsmtime.nfs_sec != nfs_xdrneg1) - fxdr_nfstime(&sp->sa_nfsmtime, &vap->va_mtime); - } else { - fxdr_hyper(&sp->sa_nqsize, &vap->va_size); - fxdr_nqtime(&sp->sa_nqatime, &vap->va_atime); - fxdr_nqtime(&sp->sa_nqmtime, &vap->va_mtime); - vap->va_flags = fxdr_unsigned(u_long, sp->sa_nqflags); + if (sp->sa_mtime.nfsv2_sec != nfs_xdrneg1) + fxdr_nfsv2time(&sp->sa_mtime, &vap->va_mtime); + + } + + /* + * Now that we have all the fields, lets do it. + */ + if (error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam, + &rdonly, (nfsd->nd_flag & ND_KERBAUTH))) { + nfsm_reply(2 * NFSX_UNSIGNED); + nfsm_srvwcc_data(preat_ret, &preat, postat_ret, vap); + return (0); + } + nqsrv_getl(vp, ND_WRITE); + if (v3) { + error = preat_ret = VOP_GETATTR(vp, &preat, cred, procp); + if (!error && gcheck && + (preat.va_ctime.ts_sec != guard.ts_sec || + preat.va_ctime.ts_nsec != guard.ts_nsec)) + error = NFSERR_NOT_SYNC; + if (error) { + nfsrv_vput(vp); + nfsm_reply(NFSX_WCCDATA(v3)); + nfsm_srvwcc_data(preat_ret, &preat, postat_ret, vap); + return (0); + } } /* @@ -246,27 +315,23 @@ nfsrv_setattr(nfsd, mrep, md, dpos, cred, nam, mrq) if (vp->v_type == VDIR) { error = EISDIR; goto out; - } else { - error = nfsrv_access(vp, VWRITE, cred, rdonly, - nfsd->nd_procp); - if (error) - goto out; - } - } - error = VOP_SETATTR(vp, vap, cred, nfsd->nd_procp); - if (error) { - nfsrv_vput(vp); - nfsm_reply(0); + } else if (error = nfsrv_access(vp, VWRITE, cred, rdonly, + procp)) + goto out; } - error = VOP_GETATTR(vp, vap, cred, nfsd->nd_procp); + error = VOP_SETATTR(vp, vap, cred, procp); + postat_ret = VOP_GETATTR(vp, vap, cred, procp); + if (!error) + error = postat_ret; out: nfsrv_vput(vp); - nfsm_reply(NFSX_FATTR(nfsd->nd_nqlflag != NQL_NOVAL) + 2*NFSX_UNSIGNED); - nfsm_build(fp, struct nfsv2_fattr *, NFSX_FATTR(nfsd->nd_nqlflag != NQL_NOVAL)); - nfsm_srvfillattr; - if (nfsd->nd_nqlflag != NQL_NOVAL) { - nfsm_build(tl, u_long *, 2*NFSX_UNSIGNED); - txdr_hyper(&frev2, tl); + nfsm_reply(NFSX_WCCORFATTR(v3)); + if (v3) { + nfsm_srvwcc_data(preat_ret, &preat, postat_ret, vap); + return (0); + } else { + nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR); + nfsm_srvfillattr(vap, fp); } nfsm_srvdone; } @@ -275,76 +340,74 @@ out: * nfs lookup rpc */ int -nfsrv_lookup(nfsd, mrep, md, dpos, cred, nam, mrq) - struct nfsd *nfsd; - struct mbuf *mrep, *md; - caddr_t dpos; - struct ucred *cred; - struct mbuf *nam, **mrq; +nfsrv_lookup(nfsd, slp, procp, mrq) + struct nfsrv_descript *nfsd; + struct nfssvc_sock *slp; + struct proc *procp; + struct mbuf **mrq; { - register struct nfsv2_fattr *fp; + struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; + struct mbuf *nam = nfsd->nd_nam; + caddr_t dpos = nfsd->nd_dpos; + struct ucred *cred = &nfsd->nd_cr; + register struct nfs_fattr *fp; struct nameidata nd; - struct vnode *vp; - nfsv2fh_t nfh; + struct vnode *vp, *dirp; + nfsfh_t nfh; fhandle_t *fhp; register caddr_t cp; register u_long *tl; register long t1; caddr_t bpos; - int error = 0, cache, cache2, len; - u_long duration2; + int error = 0, cache, len, dirattr_ret = 1; + int v3 = (nfsd->nd_flag & ND_NFSV3); char *cp2; struct mbuf *mb, *mb2, *mreq; - struct vattr va, *vap = &va; - u_quad_t frev, frev2; + struct vattr va, dirattr, *vap = &va; + u_quad_t frev; fhp = &nfh.fh_generic; - duration2 = 0; - if (nfsd->nd_nqlflag != NQL_NOVAL) { - nfsm_dissect(tl, u_long *, NFSX_UNSIGNED); - duration2 = fxdr_unsigned(int, *tl); - } nfsm_srvmtofh(fhp); - nfsm_srvstrsiz(len, NFS_MAXNAMLEN); + nfsm_srvnamesiz(len); nd.ni_cnd.cn_cred = cred; nd.ni_cnd.cn_nameiop = LOOKUP; nd.ni_cnd.cn_flags = LOCKLEAF | SAVESTART; - error = nfs_namei(&nd, fhp, len, nfsd->nd_slp, nam, &md, &dpos, - nfsd->nd_procp); - if (error) - nfsm_reply(0); - nqsrv_getl(nd.ni_startdir, NQL_READ); + error = nfs_namei(&nd, fhp, len, slp, nam, &md, &dpos, + &dirp, procp, (nfsd->nd_flag & ND_KERBAUTH)); + if (dirp) { + if (v3) + dirattr_ret = VOP_GETATTR(dirp, &dirattr, cred, + procp); + nfsrv_vrele(dirp); + } + if (error) { + nfsm_reply(NFSX_POSTOPATTR(v3)); + nfsm_srvpostop_attr(dirattr_ret, &dirattr); + return (0); + } + nqsrv_getl(nd.ni_startdir, ND_READ); nfsrv_vrele(nd.ni_startdir); FREE(nd.ni_cnd.cn_pnbuf, M_NAMEI); vp = nd.ni_vp; bzero((caddr_t)fhp, sizeof(nfh)); fhp->fh_fsid = vp->v_mount->mnt_stat.f_fsid; error = VFS_VPTOFH(vp, &fhp->fh_fid); + if (!error) + error = VOP_GETATTR(vp, vap, cred, procp); + nfsrv_vput(vp); + nfsm_reply(NFSX_SRVFH(v3) + NFSX_POSTOPORFATTR(v3) + NFSX_POSTOPATTR(v3)); if (error) { - nfsrv_vput(vp); - nfsm_reply(0); + nfsm_srvpostop_attr(dirattr_ret, &dirattr); + return (0); } - if (duration2) - (void) nqsrv_getlease(vp, &duration2, NQL_READ, nfsd, - nam, &cache2, &frev2, cred); - error = VOP_GETATTR(vp, vap, cred, nfsd->nd_procp); - nfsrv_vput(vp); - nfsm_reply(NFSX_FH + NFSX_FATTR(nfsd->nd_nqlflag != NQL_NOVAL) + 5*NFSX_UNSIGNED); - if (nfsd->nd_nqlflag != NQL_NOVAL) { - if (duration2) { - nfsm_build(tl, u_long *, 5*NFSX_UNSIGNED); - *tl++ = txdr_unsigned(NQL_READ); - *tl++ = txdr_unsigned(cache2); - *tl++ = txdr_unsigned(duration2); - txdr_hyper(&frev2, tl); - } else { - nfsm_build(tl, u_long *, NFSX_UNSIGNED); - *tl = 0; - } + nfsm_srvfhtom(fhp, v3); + if (v3) { + nfsm_srvpostop_attr(0, vap); + nfsm_srvpostop_attr(dirattr_ret, &dirattr); + } else { + nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR); + nfsm_srvfillattr(vap, fp); } - nfsm_srvfhtom(fhp); - nfsm_build(fp, struct nfsv2_fattr *, NFSX_FATTR(nfsd->nd_nqlflag != NQL_NOVAL)); - nfsm_srvfillattr; nfsm_srvdone; } @@ -352,28 +415,36 @@ nfsrv_lookup(nfsd, mrep, md, dpos, cred, nam, mrq) * nfs readlink service */ int -nfsrv_readlink(nfsd, mrep, md, dpos, cred, nam, mrq) - struct nfsd *nfsd; - struct mbuf *mrep, *md; - caddr_t dpos; - struct ucred *cred; - struct mbuf *nam, **mrq; +nfsrv_readlink(nfsd, slp, procp, mrq) + struct nfsrv_descript *nfsd; + struct nfssvc_sock *slp; + struct proc *procp; + struct mbuf **mrq; { + struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; + struct mbuf *nam = nfsd->nd_nam; + caddr_t dpos = nfsd->nd_dpos; + struct ucred *cred = &nfsd->nd_cr; struct iovec iv[(NFS_MAXPATHLEN+MLEN-1)/MLEN]; register struct iovec *ivp = iv; register struct mbuf *mp; register u_long *tl; register long t1; caddr_t bpos; - int error = 0, rdonly, cache, i, tlen, len; + int error = 0, rdonly, cache, i, tlen, len, getret; + int v3 = (nfsd->nd_flag & ND_NFSV3); char *cp2; - struct mbuf *mb, *mb2, *mp2 = 0, *mp3 = 0, *mreq; + struct mbuf *mb, *mb2, *mp2, *mp3, *mreq; struct vnode *vp; - nfsv2fh_t nfh; + struct vattr attr; + nfsfh_t nfh; fhandle_t *fhp; struct uio io, *uiop = &io; u_quad_t frev; +#ifndef nolint + mp2 = mp3 = (struct mbuf *)0; +#endif fhp = &nfh.fh_generic; nfsm_srvmtofh(fhp); len = 0; @@ -405,22 +476,33 @@ nfsrv_readlink(nfsd, mrep, md, dpos, cred, nam, mrq) uiop->uio_rw = UIO_READ; uiop->uio_segflg = UIO_SYSSPACE; uiop->uio_procp = (struct proc *)0; - error = nfsrv_fhtovp(fhp, TRUE, &vp, cred, nfsd->nd_slp, nam, &rdonly); - if (error) { + if (error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam, + &rdonly, (nfsd->nd_flag & ND_KERBAUTH))) { m_freem(mp3); - nfsm_reply(0); + nfsm_reply(2 * NFSX_UNSIGNED); + nfsm_srvpostop_attr(1, (struct vattr *)0); + return (0); } if (vp->v_type != VLNK) { - error = EINVAL; + if (v3) + error = EINVAL; + else + error = ENXIO; goto out; } - nqsrv_getl(vp, NQL_READ); + nqsrv_getl(vp, ND_READ); error = VOP_READLINK(vp, uiop, cred); out: + getret = VOP_GETATTR(vp, &attr, cred, procp); nfsrv_vput(vp); if (error) m_freem(mp3); - nfsm_reply(NFSX_UNSIGNED); + nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_UNSIGNED); + if (v3) { + nfsm_srvpostop_attr(getret, &attr); + if (error) + return (0); + } if (uiop->uio_resid > 0) { len -= uiop->uio_resid; tlen = nfsm_rndup(len); @@ -436,26 +518,31 @@ out: * nfs read service */ int -nfsrv_read(nfsd, mrep, md, dpos, cred, nam, mrq) - struct nfsd *nfsd; - struct mbuf *mrep, *md; - caddr_t dpos; - struct ucred *cred; - struct mbuf *nam, **mrq; +nfsrv_read(nfsd, slp, procp, mrq) + struct nfsrv_descript *nfsd; + struct nfssvc_sock *slp; + struct proc *procp; + struct mbuf **mrq; { + struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; + struct mbuf *nam = nfsd->nd_nam; + caddr_t dpos = nfsd->nd_dpos; + struct ucred *cred = &nfsd->nd_cr; register struct iovec *iv; struct iovec *iv2; register struct mbuf *m; - register struct nfsv2_fattr *fp; + register struct nfs_fattr *fp; register u_long *tl; register long t1; + register int i; caddr_t bpos; - int error = 0, rdonly, cache, i, cnt, len, left, siz, tlen; + int error = 0, rdonly, cache, cnt, len, left, siz, tlen, getret; + int v3 = (nfsd->nd_flag & ND_NFSV3), reqlen; char *cp2; struct mbuf *mb, *mb2, *mreq; struct mbuf *m2; struct vnode *vp; - nfsv2fh_t nfh; + nfsfh_t nfh; fhandle_t *fhp; struct uio io, *uiop = &io; struct vattr va, *vap = &va; @@ -464,40 +551,57 @@ nfsrv_read(nfsd, mrep, md, dpos, cred, nam, mrq) fhp = &nfh.fh_generic; nfsm_srvmtofh(fhp); - if (nfsd->nd_nqlflag == NQL_NOVAL) { - nfsm_dissect(tl, u_long *, NFSX_UNSIGNED); - off = (off_t)fxdr_unsigned(u_long, *tl); - } else { + if (v3) { nfsm_dissect(tl, u_long *, 2 * NFSX_UNSIGNED); fxdr_hyper(tl, &off); + } else { + nfsm_dissect(tl, u_long *, NFSX_UNSIGNED); + off = (off_t)fxdr_unsigned(u_long, *tl); + } + nfsm_srvstrsiz(reqlen, NFS_SRVMAXDATA(nfsd)); + if (error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam, + &rdonly, (nfsd->nd_flag & ND_KERBAUTH))) { + nfsm_reply(2 * NFSX_UNSIGNED); + nfsm_srvpostop_attr(1, (struct vattr *)0); + return (0); } - nfsm_srvstrsiz(cnt, NFS_MAXDATA); - error = nfsrv_fhtovp(fhp, TRUE, &vp, cred, nfsd->nd_slp, nam, &rdonly); - if (error) - nfsm_reply(0); if (vp->v_type != VREG) { - error = (vp->v_type == VDIR) ? EISDIR : EACCES; - nfsrv_vput(vp); - nfsm_reply(0); + if (v3) + error = EINVAL; + else + error = (vp->v_type == VDIR) ? EISDIR : EACCES; } - nqsrv_getl(vp, NQL_READ); - if ((error = nfsrv_access(vp, VREAD, cred, rdonly, nfsd->nd_procp)) && - (error = nfsrv_access(vp, VEXEC, cred, rdonly, nfsd->nd_procp))) { - nfsrv_vput(vp); - nfsm_reply(0); + if (!error) { + nqsrv_getl(vp, ND_READ); + if (error = nfsrv_access(vp, VREAD, cred, rdonly, procp)) + error = nfsrv_access(vp, VEXEC, cred, rdonly, procp); } - error = VOP_GETATTR(vp, vap, cred, nfsd->nd_procp); + getret = VOP_GETATTR(vp, vap, cred, procp); + if (!error) + error = getret; if (error) { nfsrv_vput(vp); - nfsm_reply(0); + nfsm_reply(NFSX_POSTOPATTR(v3)); + nfsm_srvpostop_attr(getret, vap); + return (0); } if (off >= vap->va_size) cnt = 0; - else if ((off + cnt) > vap->va_size) + else if ((off + reqlen) > vap->va_size) cnt = nfsm_rndup(vap->va_size - off); - nfsm_reply(NFSX_FATTR(nfsd->nd_nqlflag != NQL_NOVAL)+NFSX_UNSIGNED+nfsm_rndup(cnt)); - nfsm_build(fp, struct nfsv2_fattr *, NFSX_FATTR(nfsd->nd_nqlflag != NQL_NOVAL)); - nfsm_build(tl, u_long *, NFSX_UNSIGNED); + else + cnt = reqlen; + nfsm_reply(NFSX_POSTOPORFATTR(v3) + 3 * NFSX_UNSIGNED+nfsm_rndup(cnt)); + if (v3) { + nfsm_build(tl, u_long *, NFSX_V3FATTR + 4 * NFSX_UNSIGNED); + *tl++ = nfs_true; + fp = (struct nfs_fattr *)tl; + tl += (NFSX_V3FATTR / sizeof (u_long)); + } else { + nfsm_build(tl, u_long *, NFSX_V2FATTR + NFSX_UNSIGNED); + fp = (struct nfs_fattr *)tl; + tl += (NFSX_V2FATTR / sizeof (u_long)); + } len = left = cnt; if (cnt > 0) { /* @@ -505,19 +609,11 @@ nfsrv_read(nfsd, mrep, md, dpos, cred, nam, mrq) */ i = 0; m = m2 = mb; - MALLOC(iv, struct iovec *, - ((NFS_MAXDATA+MLEN-1)/MLEN) * sizeof (struct iovec), - M_TEMP, M_WAITOK); - iv2 = iv; while (left > 0) { siz = min(M_TRAILINGSPACE(m), left); if (siz > 0) { - m->m_len += siz; - iv->iov_base = bpos; - iv->iov_len = siz; - iv++; - i++; left -= siz; + i++; } if (left > 0) { MGET(m, M_WAIT, MT_DATA); @@ -525,10 +621,28 @@ nfsrv_read(nfsd, mrep, md, dpos, cred, nam, mrq) m->m_len = 0; m2->m_next = m; m2 = m; - bpos = mtod(m, caddr_t); } } - uiop->uio_iov = iv2; + MALLOC(iv, struct iovec *, i * sizeof (struct iovec), + M_TEMP, M_WAITOK); + uiop->uio_iov = iv2 = iv; + m = mb; + left = cnt; + i = 0; + while (left > 0) { + if (m == NULL) + panic("nfsrv_read iov"); + siz = min(M_TRAILINGSPACE(m), left); + if (siz > 0) { + iv->iov_base = mtod(m, caddr_t) + m->m_len; + iv->iov_len = siz; + m->m_len += siz; + left -= siz; + iv++; + i++; + } + m = m->m_next; + } uiop->uio_iovcnt = i; uiop->uio_offset = off; uiop->uio_resid = cnt; @@ -537,19 +651,30 @@ nfsrv_read(nfsd, mrep, md, dpos, cred, nam, mrq) error = VOP_READ(vp, uiop, IO_NODELOCKED, cred); off = uiop->uio_offset; FREE((caddr_t)iv2, M_TEMP); - if (error || (error = VOP_GETATTR(vp, vap, cred, nfsd->nd_procp))) { + if (error || (getret = VOP_GETATTR(vp, vap, cred, procp))) { + if (!error) + error = getret; m_freem(mreq); nfsrv_vput(vp); - nfsm_reply(0); + nfsm_reply(NFSX_POSTOPATTR(v3)); + nfsm_srvpostop_attr(getret, vap); + return (0); } } else uiop->uio_resid = 0; nfsrv_vput(vp); - nfsm_srvfillattr; + nfsm_srvfillattr(vap, fp); len -= uiop->uio_resid; tlen = nfsm_rndup(len); if (cnt != tlen || tlen != len) - nfsm_adj(mb, cnt-tlen, tlen-len); + nfsm_adj(mb, cnt - tlen, tlen - len); + if (v3) { + *tl++ = txdr_unsigned(len); + if (len < reqlen) + *tl++ = nfs_true; + else + *tl++ = nfs_false; + } *tl = txdr_unsigned(len); nfsm_srvdone; } @@ -558,132 +683,580 @@ nfsrv_read(nfsd, mrep, md, dpos, cred, nam, mrq) * nfs write service */ int -nfsrv_write(nfsd, mrep, md, dpos, cred, nam, mrq) - struct nfsd *nfsd; - struct mbuf *mrep, *md; - caddr_t dpos; - struct ucred *cred; - struct mbuf *nam, **mrq; +nfsrv_write(nfsd, slp, procp, mrq) + struct nfsrv_descript *nfsd; + struct nfssvc_sock *slp; + struct proc *procp; + struct mbuf **mrq; { + struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; + struct mbuf *nam = nfsd->nd_nam; + caddr_t dpos = nfsd->nd_dpos; + struct ucred *cred = &nfsd->nd_cr; register struct iovec *ivp; + register int i, cnt; register struct mbuf *mp; - register struct nfsv2_fattr *fp; - struct iovec iv[NFS_MAXIOVEC]; - struct vattr va; + register struct nfs_fattr *fp; + struct iovec *iv; + struct vattr va, forat; register struct vattr *vap = &va; register u_long *tl; register long t1; caddr_t bpos; - int error = 0, rdonly, cache, siz, len, xfer; - int ioflags = IO_SYNC | IO_NODELOCKED; + int error = 0, rdonly, cache, siz, len, xfer, forat_ret = 1; + int ioflags, aftat_ret = 1, retlen, zeroing, adjust; + int stable = NFSV3WRITE_FILESYNC; + int v3 = (nfsd->nd_flag & ND_NFSV3); char *cp2; struct mbuf *mb, *mb2, *mreq; struct vnode *vp; - nfsv2fh_t nfh; + nfsfh_t nfh; fhandle_t *fhp; struct uio io, *uiop = &io; off_t off; u_quad_t frev; + if (mrep == NULL) { + *mrq = NULL; + return (0); + } fhp = &nfh.fh_generic; nfsm_srvmtofh(fhp); - nfsm_dissect(tl, u_long *, 4 * NFSX_UNSIGNED); - if (nfsd->nd_nqlflag == NQL_NOVAL) { - off = (off_t)fxdr_unsigned(u_long, *++tl); - tl += 2; - } else { + if (v3) { + nfsm_dissect(tl, u_long *, 5 * NFSX_UNSIGNED); fxdr_hyper(tl, &off); + tl += 3; + stable = fxdr_unsigned(int, *tl++); + } else { + nfsm_dissect(tl, u_long *, 4 * NFSX_UNSIGNED); + off = (off_t)fxdr_unsigned(u_long, *++tl); tl += 2; - if (fxdr_unsigned(u_long, *tl++)) - ioflags |= IO_APPEND; } - len = fxdr_unsigned(long, *tl); - if (len > NFS_MAXDATA || len <= 0) { - error = EBADRPC; - nfsm_reply(0); - } - if (dpos == (mtod(md, caddr_t)+md->m_len)) { - mp = md->m_next; - if (mp == NULL) { - error = EBADRPC; - nfsm_reply(0); + retlen = len = fxdr_unsigned(long, *tl); + cnt = i = 0; + + /* + * For NFS Version 2, it is not obvious what a write of zero length + * should do, but I might as well be consistent with Version 3, + * which is to return ok so long as there are no permission problems. + */ + if (len > 0) { + zeroing = 1; + mp = mrep; + while (mp) { + if (mp == md) { + zeroing = 0; + adjust = dpos - mtod(mp, caddr_t); + mp->m_len -= adjust; + if (mp->m_len > 0 && adjust > 0) + NFSMADV(mp, adjust); } - } else { - mp = md; - siz = dpos-mtod(mp, caddr_t); - mp->m_len -= siz; - NFSMADV(mp, siz); + if (zeroing) + mp->m_len = 0; + else if (mp->m_len > 0) { + i += mp->m_len; + if (i > len) { + mp->m_len -= (i - len); + zeroing = 1; + } + if (mp->m_len > 0) + cnt++; + } + mp = mp->m_next; + } } - error = nfsrv_fhtovp(fhp, TRUE, &vp, cred, nfsd->nd_slp, nam, &rdonly); - if (error) - nfsm_reply(0); + if (len > NFS_MAXDATA || len < 0 || i < len) { + error = EIO; + nfsm_reply(2 * NFSX_UNSIGNED); + nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, vap); + return (0); + } + if (error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam, + &rdonly, (nfsd->nd_flag & ND_KERBAUTH))) { + nfsm_reply(2 * NFSX_UNSIGNED); + nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, vap); + return (0); + } + if (v3) + forat_ret = VOP_GETATTR(vp, &forat, cred, procp); if (vp->v_type != VREG) { - error = (vp->v_type == VDIR) ? EISDIR : EACCES; - nfsrv_vput(vp); - nfsm_reply(0); + if (v3) + error = EINVAL; + else + error = (vp->v_type == VDIR) ? EISDIR : EACCES; + } + if (!error) { + nqsrv_getl(vp, ND_WRITE); + error = nfsrv_access(vp, VWRITE, cred, rdonly, procp); } - nqsrv_getl(vp, NQL_WRITE); - error = nfsrv_access(vp, VWRITE, cred, rdonly, nfsd->nd_procp); if (error) { nfsrv_vput(vp); - nfsm_reply(0); + nfsm_reply(NFSX_WCCDATA(v3)); + nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, vap); + return (0); } - uiop->uio_resid = 0; - uiop->uio_rw = UIO_WRITE; - uiop->uio_segflg = UIO_SYSSPACE; - uiop->uio_procp = (struct proc *)0; + + if (len > 0) { + MALLOC(ivp, struct iovec *, cnt * sizeof (struct iovec), M_TEMP, + M_WAITOK); + uiop->uio_iov = iv = ivp; + uiop->uio_iovcnt = cnt; + mp = mrep; + while (mp) { + if (mp->m_len > 0) { + ivp->iov_base = mtod(mp, caddr_t); + ivp->iov_len = mp->m_len; + ivp++; + } + mp = mp->m_next; + } + + /* + * XXX + * The IO_METASYNC flag indicates that all metadata (and not just + * enough to ensure data integrity) mus be written to stable storage + * synchronously. + * (IO_METASYNC is not yet implemented in 4.4BSD-Lite.) + */ + if (stable == NFSV3WRITE_UNSTABLE) + ioflags = IO_NODELOCKED; + else if (stable == NFSV3WRITE_DATASYNC) + ioflags = (IO_SYNC | IO_NODELOCKED); + else + ioflags = (IO_METASYNC | IO_SYNC | IO_NODELOCKED); + uiop->uio_resid = len; + uiop->uio_rw = UIO_WRITE; + uiop->uio_segflg = UIO_SYSSPACE; + uiop->uio_procp = (struct proc *)0; + uiop->uio_offset = off; + error = VOP_WRITE(vp, uiop, ioflags, cred); + nfsstats.srvvop_writes++; + FREE((caddr_t)iv, M_TEMP); + } + aftat_ret = VOP_GETATTR(vp, vap, cred, procp); + nfsrv_vput(vp); + if (!error) + error = aftat_ret; + nfsm_reply(NFSX_PREOPATTR(v3) + NFSX_POSTOPORFATTR(v3) + + 2 * NFSX_UNSIGNED + NFSX_WRITEVERF(v3)); + if (v3) { + nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, vap); + if (error) + return (0); + nfsm_build(tl, u_long *, 4 * NFSX_UNSIGNED); + *tl++ = txdr_unsigned(retlen); + if (stable == NFSV3WRITE_UNSTABLE) + *tl++ = txdr_unsigned(stable); + else + *tl++ = txdr_unsigned(NFSV3WRITE_FILESYNC); + /* + * Actually, there is no need to txdr these fields, + * but it may make the values more human readable, + * for debugging purposes. + */ + *tl++ = txdr_unsigned(boottime.tv_sec); + *tl = txdr_unsigned(boottime.tv_usec); + } else { + nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR); + nfsm_srvfillattr(vap, fp); + } + nfsm_srvdone; +} + +/* + * NFS write service with write gathering support. Called when + * nfsrvw_procrastinate > 0. + * See: Chet Juszczak, "Improving the Write Performance of an NFS Server", + * in Proc. of the Winter 1994 Usenix Conference, pg. 247-259, San Franscisco, + * Jan. 1994. + */ +int +nfsrv_writegather(ndp, slp, procp, mrq) + struct nfsrv_descript **ndp; + struct nfssvc_sock *slp; + struct proc *procp; + struct mbuf **mrq; +{ + register struct iovec *ivp; + register struct mbuf *mp; + register struct nfsrv_descript *wp, *nfsd, *owp, *swp; + register struct nfs_fattr *fp; + register int i; + struct iovec *iov; + struct nfsrvw_delayhash *wpp; + struct ucred *cred; + struct vattr va, forat; + register u_long *tl; + register long t1; + caddr_t bpos, dpos; + int error = 0, rdonly, cache, len, forat_ret = 1; + int ioflags, aftat_ret = 1, s, adjust, v3, zeroing; + char *cp2; + struct mbuf *mb, *mb2, *mreq, *mrep, *md; + struct vnode *vp; + struct uio io, *uiop = &io; + off_t off; + u_quad_t frev, cur_usec; + +#ifndef nolint + i = 0; + len = 0; +#endif + *mrq = NULL; + if (*ndp) { + nfsd = *ndp; + *ndp = NULL; + mrep = nfsd->nd_mrep; + md = nfsd->nd_md; + dpos = nfsd->nd_dpos; + cred = &nfsd->nd_cr; + v3 = (nfsd->nd_flag & ND_NFSV3); + LIST_INIT(&nfsd->nd_coalesce); + nfsd->nd_mreq = NULL; + nfsd->nd_stable = NFSV3WRITE_FILESYNC; + cur_usec = (u_quad_t)time.tv_sec * 1000000 + (u_quad_t)time.tv_usec; + nfsd->nd_time = cur_usec + nfsrvw_procrastinate; + + /* + * Now, get the write header.. + */ + nfsm_srvmtofh(&nfsd->nd_fh); + if (v3) { + nfsm_dissect(tl, u_long *, 5 * NFSX_UNSIGNED); + fxdr_hyper(tl, &nfsd->nd_off); + tl += 3; + nfsd->nd_stable = fxdr_unsigned(int, *tl++); + } else { + nfsm_dissect(tl, u_long *, 4 * NFSX_UNSIGNED); + nfsd->nd_off = (off_t)fxdr_unsigned(u_long, *++tl); + tl += 2; + } + len = fxdr_unsigned(long, *tl); + nfsd->nd_len = len; + nfsd->nd_eoff = nfsd->nd_off + len; + + /* + * Trim the header out of the mbuf list and trim off any trailing + * junk so that the mbuf list has only the write data. + */ + zeroing = 1; + i = 0; + mp = mrep; + while (mp) { + if (mp == md) { + zeroing = 0; + adjust = dpos - mtod(mp, caddr_t); + mp->m_len -= adjust; + if (mp->m_len > 0 && adjust > 0) + NFSMADV(mp, adjust); + } + if (zeroing) + mp->m_len = 0; + else { + i += mp->m_len; + if (i > len) { + mp->m_len -= (i - len); + zeroing = 1; + } + } + mp = mp->m_next; + } + if (len > NFS_MAXDATA || len < 0 || i < len) { +nfsmout: + m_freem(mrep); + error = EIO; + nfsm_writereply(2 * NFSX_UNSIGNED, v3); + if (v3) + nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va); + nfsd->nd_mreq = mreq; + nfsd->nd_mrep = NULL; + nfsd->nd_time = 0; + } + + /* + * Add this entry to the hash and time queues. + */ + s = splsoftclock(); + owp = NULL; + wp = slp->ns_tq.lh_first; + while (wp && wp->nd_time < nfsd->nd_time) { + owp = wp; + wp = wp->nd_tq.le_next; + } + if (owp) { + LIST_INSERT_AFTER(owp, nfsd, nd_tq); + } else { + LIST_INSERT_HEAD(&slp->ns_tq, nfsd, nd_tq); + } + if (nfsd->nd_mrep) { + wpp = NWDELAYHASH(slp, nfsd->nd_fh.fh_fid.fid_data); + owp = NULL; + wp = wpp->lh_first; + while (wp && + bcmp((caddr_t)&nfsd->nd_fh,(caddr_t)&wp->nd_fh,NFSX_V3FH)) { + owp = wp; + wp = wp->nd_hash.le_next; + } + while (wp && wp->nd_off < nfsd->nd_off && + !bcmp((caddr_t)&nfsd->nd_fh,(caddr_t)&wp->nd_fh,NFSX_V3FH)) { + owp = wp; + wp = wp->nd_hash.le_next; + } + if (owp) { + LIST_INSERT_AFTER(owp, nfsd, nd_hash); + + /* + * Search the hash list for overlapping entries and + * coalesce. + */ + for(; nfsd && NFSW_CONTIG(owp, nfsd); nfsd = wp) { + wp = nfsd->nd_hash.le_next; + if (NFSW_SAMECRED(owp, nfsd)) + nfsrvw_coalesce(owp, nfsd); + } + } else { + LIST_INSERT_HEAD(wpp, nfsd, nd_hash); + } + } + splx(s); + } + /* - * Do up to NFS_MAXIOVEC mbufs of write each iteration of the - * loop until done. + * Now, do VOP_WRITE()s for any one(s) that need to be done now + * and generate the associated reply mbuf list(s). */ - while (len > 0 && uiop->uio_resid == 0) { - ivp = iv; - siz = 0; - uiop->uio_iov = ivp; - uiop->uio_iovcnt = 0; - uiop->uio_offset = off; - while (len > 0 && uiop->uio_iovcnt < NFS_MAXIOVEC && mp != NULL) { - ivp->iov_base = mtod(mp, caddr_t); - if (len < mp->m_len) - ivp->iov_len = xfer = len; +loop1: + cur_usec = (u_quad_t)time.tv_sec * 1000000 + (u_quad_t)time.tv_usec; + s = splsoftclock(); + for (nfsd = slp->ns_tq.lh_first; nfsd; nfsd = owp) { + owp = nfsd->nd_tq.le_next; + if (nfsd->nd_time > cur_usec) + break; + if (nfsd->nd_mreq) + continue; + LIST_REMOVE(nfsd, nd_tq); + LIST_REMOVE(nfsd, nd_hash); + splx(s); + mrep = nfsd->nd_mrep; + nfsd->nd_mrep = NULL; + cred = &nfsd->nd_cr; + v3 = (nfsd->nd_flag & ND_NFSV3); + forat_ret = aftat_ret = 1; + error = nfsrv_fhtovp(&nfsd->nd_fh, 1, &vp, cred, slp, + nfsd->nd_nam, &rdonly, (nfsd->nd_flag & ND_KERBAUTH)); + if (!error) { + if (v3) + forat_ret = VOP_GETATTR(vp, &forat, cred, procp); + if (vp->v_type != VREG) { + if (v3) + error = EINVAL; else - ivp->iov_len = xfer = mp->m_len; -#ifdef notdef - /* Not Yet .. */ - if (M_HASCL(mp) && (((u_long)ivp->iov_base) & CLOFSET) == 0) - ivp->iov_op = NULL; /* what should it be ?? */ - else - ivp->iov_op = NULL; -#endif - uiop->uio_iovcnt++; - ivp++; - len -= xfer; - siz += xfer; + error = (vp->v_type == VDIR) ? EISDIR : EACCES; + } + } else + vp = NULL; + if (!error) { + nqsrv_getl(vp, ND_WRITE); + error = nfsrv_access(vp, VWRITE, cred, rdonly, procp); + } + + if (nfsd->nd_stable == NFSV3WRITE_UNSTABLE) + ioflags = IO_NODELOCKED; + else if (nfsd->nd_stable == NFSV3WRITE_DATASYNC) + ioflags = (IO_SYNC | IO_NODELOCKED); + else + ioflags = (IO_METASYNC | IO_SYNC | IO_NODELOCKED); + uiop->uio_rw = UIO_WRITE; + uiop->uio_segflg = UIO_SYSSPACE; + uiop->uio_procp = (struct proc *)0; + uiop->uio_offset = nfsd->nd_off; + uiop->uio_resid = nfsd->nd_eoff - nfsd->nd_off; + if (uiop->uio_resid > 0) { + mp = mrep; + i = 0; + while (mp) { + if (mp->m_len > 0) + i++; + mp = mp->m_next; + } + uiop->uio_iovcnt = i; + MALLOC(iov, struct iovec *, i * sizeof (struct iovec), + M_TEMP, M_WAITOK); + uiop->uio_iov = ivp = iov; + mp = mrep; + while (mp) { + if (mp->m_len > 0) { + ivp->iov_base = mtod(mp, caddr_t); + ivp->iov_len = mp->m_len; + ivp++; + } mp = mp->m_next; + } + if (!error) { + error = VOP_WRITE(vp, uiop, ioflags, cred); + nfsstats.srvvop_writes++; + } + FREE((caddr_t)iov, M_TEMP); } - if (len > 0 && mp == NULL) { - error = EBADRPC; - nfsrv_vput(vp); - nfsm_reply(0); + m_freem(mrep); + if (vp) { + aftat_ret = VOP_GETATTR(vp, &va, cred, procp); + nfsrv_vput(vp); } - uiop->uio_resid = siz; - error = VOP_WRITE(vp, uiop, ioflags, cred); - if (error) { - nfsrv_vput(vp); - nfsm_reply(0); - } - off = uiop->uio_offset; + + /* + * Loop around generating replies for all write rpcs that have + * now been completed. + */ + swp = nfsd; + do { + if (error) { + nfsm_writereply(NFSX_WCCDATA(v3), v3); + if (v3) { + nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va); + } + } else { + nfsm_writereply(NFSX_PREOPATTR(v3) + + NFSX_POSTOPORFATTR(v3) + 2 * NFSX_UNSIGNED + + NFSX_WRITEVERF(v3), v3); + if (v3) { + nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va); + nfsm_build(tl, u_long *, 4 * NFSX_UNSIGNED); + *tl++ = txdr_unsigned(nfsd->nd_len); + *tl++ = txdr_unsigned(swp->nd_stable); + /* + * Actually, there is no need to txdr these fields, + * but it may make the values more human readable, + * for debugging purposes. + */ + *tl++ = txdr_unsigned(boottime.tv_sec); + *tl = txdr_unsigned(boottime.tv_usec); + } else { + nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR); + nfsm_srvfillattr(&va, fp); + } + } + nfsd->nd_mreq = mreq; + if (nfsd->nd_mrep) + panic("nfsrv_write: nd_mrep not free"); + + /* + * Done. Put it at the head of the timer queue so that + * the final phase can return the reply. + */ + s = splsoftclock(); + if (nfsd != swp) { + nfsd->nd_time = 0; + LIST_INSERT_HEAD(&slp->ns_tq, nfsd, nd_tq); + } + nfsd = swp->nd_coalesce.lh_first; + if (nfsd) { + LIST_REMOVE(nfsd, nd_tq); + } + splx(s); + } while (nfsd); + s = splsoftclock(); + swp->nd_time = 0; + LIST_INSERT_HEAD(&slp->ns_tq, swp, nd_tq); + splx(s); + goto loop1; } - error = VOP_GETATTR(vp, vap, cred, nfsd->nd_procp); - nfsrv_vput(vp); - nfsm_reply(NFSX_FATTR(nfsd->nd_nqlflag != NQL_NOVAL)); - nfsm_build(fp, struct nfsv2_fattr *, NFSX_FATTR(nfsd->nd_nqlflag != NQL_NOVAL)); - nfsm_srvfillattr; - if (nfsd->nd_nqlflag != NQL_NOVAL) { - nfsm_build(tl, u_long *, 2*NFSX_UNSIGNED); - txdr_hyper(&vap->va_filerev, tl); + splx(s); + + /* + * Search for a reply to return. + */ + s = splsoftclock(); + for (nfsd = slp->ns_tq.lh_first; nfsd; nfsd = nfsd->nd_tq.le_next) + if (nfsd->nd_mreq) { + LIST_REMOVE(nfsd, nd_tq); + *mrq = nfsd->nd_mreq; + *ndp = nfsd; + break; + } + splx(s); + return (0); +} + +/* + * Coalesce the write request nfsd into owp. To do this we must: + * - remove nfsd from the queues + * - merge nfsd->nd_mrep into owp->nd_mrep + * - update the nd_eoff and nd_stable for owp + * - put nfsd on owp's nd_coalesce list + * NB: Must be called at splsoftclock(). + */ +void +nfsrvw_coalesce(owp, nfsd) + register struct nfsrv_descript *owp; + register struct nfsrv_descript *nfsd; +{ + register int overlap; + register struct mbuf *mp; + + LIST_REMOVE(nfsd, nd_hash); + LIST_REMOVE(nfsd, nd_tq); + if (owp->nd_eoff < nfsd->nd_eoff) { + overlap = owp->nd_eoff - nfsd->nd_off; + if (overlap < 0) + panic("nfsrv_coalesce: bad off"); + if (overlap > 0) + m_adj(nfsd->nd_mrep, overlap); + mp = owp->nd_mrep; + while (mp->m_next) + mp = mp->m_next; + mp->m_next = nfsd->nd_mrep; + owp->nd_eoff = nfsd->nd_eoff; + } else + m_freem(nfsd->nd_mrep); + nfsd->nd_mrep = NULL; + if (nfsd->nd_stable == NFSV3WRITE_FILESYNC) + owp->nd_stable = NFSV3WRITE_FILESYNC; + else if (nfsd->nd_stable == NFSV3WRITE_DATASYNC && + owp->nd_stable == NFSV3WRITE_UNSTABLE) + owp->nd_stable = NFSV3WRITE_DATASYNC; + LIST_INSERT_HEAD(&owp->nd_coalesce, nfsd, nd_tq); +} + +/* + * Sort the group list in increasing numerical order. + * (Insertion sort by Chris Torek, who was grossed out by the bubble sort + * that used to be here.) + */ +void +nfsrvw_sort(list, num) + register gid_t *list; + register int num; +{ + register int i, j; + gid_t v; + + /* Insertion sort. */ + for (i = 1; i < num; i++) { + v = list[i]; + /* find correct slot for value v, moving others up */ + for (j = i; --j >= 0 && v < list[j];) + list[j + 1] = list[j]; + list[j + 1] = v; } - nfsm_srvdone; +} + +/* + * copy credentials making sure that the result can be compared with bcmp(). + */ +void +nfsrv_setcred(incred, outcred) + register struct ucred *incred, *outcred; +{ + register int i; + + bzero((caddr_t)outcred, sizeof (struct ucred)); + outcred->cr_ref = 1; + outcred->cr_uid = incred->cr_uid; + outcred->cr_ngroups = incred->cr_ngroups; + for (i = 0; i < incred->cr_ngroups; i++) + outcred->cr_groups[i] = incred->cr_groups[i]; + nfsrvw_sort(outcred->cr_groups, outcred->cr_ngroups); } /* @@ -691,15 +1264,18 @@ nfsrv_write(nfsd, mrep, md, dpos, cred, nam, mrq) * now does a truncate to 0 length via. setattr if it already exists */ int -nfsrv_create(nfsd, mrep, md, dpos, cred, nam, mrq) - struct nfsd *nfsd; - struct mbuf *mrep, *md; - caddr_t dpos; - struct ucred *cred; - struct mbuf *nam, **mrq; +nfsrv_create(nfsd, slp, procp, mrq) + struct nfsrv_descript *nfsd; + struct nfssvc_sock *slp; + struct proc *procp; + struct mbuf **mrq; { - register struct nfsv2_fattr *fp; - struct vattr va; + struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; + struct mbuf *nam = nfsd->nd_nam; + caddr_t dpos = nfsd->nd_dpos; + struct ucred *cred = &nfsd->nd_cr; + register struct nfs_fattr *fp; + struct vattr va, dirfor, diraft; register struct vattr *vap = &va; register struct nfsv2_sattr *sp; register u_long *tl; @@ -707,75 +1283,131 @@ nfsrv_create(nfsd, mrep, md, dpos, cred, nam, mrq) register caddr_t cp; register long t1; caddr_t bpos; - int error = 0, rdev, cache, len, tsize; + int error = 0, rdev, cache, len, tsize, dirfor_ret = 1, diraft_ret = 1; + int v3 = (nfsd->nd_flag & ND_NFSV3), how, exclusive_flag = 0; char *cp2; struct mbuf *mb, *mb2, *mreq; - struct vnode *vp; - nfsv2fh_t nfh; + struct vnode *vp, *dirp = (struct vnode *)0; + nfsfh_t nfh; fhandle_t *fhp; - u_quad_t frev; + u_quad_t frev, tempsize; + u_char cverf[NFSX_V3CREATEVERF]; +#ifndef nolint + rdev = 0; +#endif nd.ni_cnd.cn_nameiop = 0; fhp = &nfh.fh_generic; nfsm_srvmtofh(fhp); - nfsm_srvstrsiz(len, NFS_MAXNAMLEN); + nfsm_srvnamesiz(len); nd.ni_cnd.cn_cred = cred; nd.ni_cnd.cn_nameiop = CREATE; nd.ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF | SAVESTART; - error = nfs_namei(&nd, fhp, len, nfsd->nd_slp, nam, &md, &dpos, - nfsd->nd_procp); - if (error) - nfsm_reply(0); + error = nfs_namei(&nd, fhp, len, slp, nam, &md, &dpos, + &dirp, procp, (nfsd->nd_flag & ND_KERBAUTH)); + if (dirp) { + if (v3) + dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred, + procp); + else { + nfsrv_vrele(dirp); + dirp = (struct vnode *)0; + } + } + if (error) { + nfsm_reply(NFSX_WCCDATA(v3)); + nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); + if (dirp) + nfsrv_vrele(dirp); + return (0); + } VATTR_NULL(vap); - nfsm_dissect(sp, struct nfsv2_sattr *, NFSX_SATTR(nfsd->nd_nqlflag != NQL_NOVAL)); + if (v3) { + nfsm_dissect(tl, u_long *, NFSX_UNSIGNED); + how = fxdr_unsigned(int, *tl); + switch (how) { + case NFSV3CREATE_GUARDED: + if (nd.ni_vp) { + error = EEXIST; + break; + } + case NFSV3CREATE_UNCHECKED: + nfsm_srvsattr(vap); + break; + case NFSV3CREATE_EXCLUSIVE: + nfsm_dissect(cp, caddr_t, NFSX_V3CREATEVERF); + bcopy(cp, cverf, NFSX_V3CREATEVERF); + exclusive_flag = 1; + if (nd.ni_vp == NULL) + vap->va_mode = 0; + break; + }; + vap->va_type = VREG; + } else { + nfsm_dissect(sp, struct nfsv2_sattr *, NFSX_V2SATTR); + vap->va_type = IFTOVT(fxdr_unsigned(u_long, sp->sa_mode)); + if (vap->va_type == VNON) + vap->va_type = VREG; + vap->va_mode = nfstov_mode(sp->sa_mode); + switch (vap->va_type) { + case VREG: + tsize = fxdr_unsigned(long, sp->sa_size); + if (tsize != -1) + vap->va_size = (u_quad_t)tsize; + break; + case VCHR: + case VBLK: + case VFIFO: + rdev = fxdr_unsigned(long, sp->sa_size); + break; + }; + } + /* * Iff doesn't exist, create it * otherwise just truncate to 0 length * should I set the mode too ?? */ if (nd.ni_vp == NULL) { - vap->va_type = IFTOVT(fxdr_unsigned(u_long, sp->sa_mode)); - if (vap->va_type == VNON) - vap->va_type = VREG; - vap->va_mode = nfstov_mode(sp->sa_mode); - if (nfsd->nd_nqlflag == NQL_NOVAL) - rdev = fxdr_unsigned(long, sp->sa_nfssize); - else - rdev = fxdr_unsigned(long, sp->sa_nqrdev); if (vap->va_type == VREG || vap->va_type == VSOCK) { nfsrv_vrele(nd.ni_startdir); - nqsrv_getl(nd.ni_dvp, NQL_WRITE); - error=VOP_CREATE(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, vap); - if (error) - nfsm_reply(0); - nfsrv_vmio(nd.ni_vp); - FREE(nd.ni_cnd.cn_pnbuf, M_NAMEI); + nqsrv_getl(nd.ni_dvp, ND_WRITE); + error = VOP_CREATE(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, vap); + if (!error) { + nfsrv_vmio(nd.ni_vp); + FREE(nd.ni_cnd.cn_pnbuf, M_NAMEI); + if (exclusive_flag) { + exclusive_flag = 0; + VATTR_NULL(vap); + bcopy(cverf, (caddr_t)&vap->va_atime, + NFSX_V3CREATEVERF); + error = VOP_SETATTR(nd.ni_vp, vap, cred, + procp); + } + } } else if (vap->va_type == VCHR || vap->va_type == VBLK || vap->va_type == VFIFO) { if (vap->va_type == VCHR && rdev == 0xffffffff) vap->va_type = VFIFO; - if (vap->va_type == VFIFO) { - } else { - error = suser(cred, (u_short *)0); - if (error) { - VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); - nfsrv_vput(nd.ni_dvp); - goto out; - } else - vap->va_rdev = (dev_t)rdev; - } - nqsrv_getl(nd.ni_dvp, NQL_WRITE); - error=VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, vap); - if (error) { + if (error = suser(cred, (u_short *)0)) { + nfsrv_vrele(nd.ni_startdir); + free(nd.ni_cnd.cn_pnbuf, M_NAMEI); + VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); + nfsrv_vput(nd.ni_dvp); + nfsm_reply(0); + return (error); + } else + vap->va_rdev = (dev_t)rdev; + nqsrv_getl(nd.ni_dvp, ND_WRITE); + if (error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, vap)) { nfsrv_vrele(nd.ni_startdir); nfsm_reply(0); } nd.ni_cnd.cn_nameiop = LOOKUP; nd.ni_cnd.cn_flags &= ~(LOCKPARENT | SAVESTART); - nd.ni_cnd.cn_proc = nfsd->nd_procp; - nd.ni_cnd.cn_cred = nfsd->nd_procp->p_ucred; - error = lookup(&nd); - if (error) { + nd.ni_cnd.cn_proc = procp; + nd.ni_cnd.cn_cred = cred; + if (error = lookup(&nd)) { free(nd.ni_cnd.cn_pnbuf, M_NAMEI); nfsm_reply(0); } @@ -789,10 +1421,11 @@ nfsrv_create(nfsd, mrep, md, dpos, cred, nam, mrq) nfsm_reply(0); } } else { + nfsrv_vrele(nd.ni_startdir); + free(nd.ni_cnd.cn_pnbuf, M_NAMEI); VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); nfsrv_vput(nd.ni_dvp); error = ENXIO; - goto out; } vp = nd.ni_vp; } else { @@ -804,46 +1437,56 @@ nfsrv_create(nfsd, mrep, md, dpos, cred, nam, mrq) else nfsrv_vput(nd.ni_dvp); VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); - if (nfsd->nd_nqlflag == NQL_NOVAL) { - tsize = fxdr_unsigned(long, sp->sa_nfssize); - if (tsize != -1) - vap->va_size = (u_quad_t)tsize; - else - vap->va_size = -1; - } else - fxdr_hyper(&sp->sa_nqsize, &vap->va_size); if (vap->va_size != -1) { error = nfsrv_access(vp, VWRITE, cred, - (nd.ni_cnd.cn_flags & RDONLY), nfsd->nd_procp); - if (error) { - nfsrv_vput(vp); - nfsm_reply(0); + (nd.ni_cnd.cn_flags & RDONLY), procp); + if (!error) { + nqsrv_getl(vp, ND_WRITE); + tempsize = vap->va_size; + VATTR_NULL(vap); + vap->va_size = tempsize; + error = VOP_SETATTR(vp, vap, cred, + procp); } - nqsrv_getl(vp, NQL_WRITE); - error = VOP_SETATTR(vp, vap, cred, nfsd->nd_procp); - if (error) { + if (error) nfsrv_vput(vp); - nfsm_reply(0); - } } } - bzero((caddr_t)fhp, sizeof(nfh)); - fhp->fh_fsid = vp->v_mount->mnt_stat.f_fsid; - error = VFS_VPTOFH(vp, &fhp->fh_fid); - if (error) { + if (!error) { + bzero((caddr_t)fhp, sizeof(nfh)); + fhp->fh_fsid = vp->v_mount->mnt_stat.f_fsid; + error = VFS_VPTOFH(vp, &fhp->fh_fid); + if (!error) + error = VOP_GETATTR(vp, vap, cred, procp); nfsrv_vput(vp); - nfsm_reply(0); } - error = VOP_GETATTR(vp, vap, cred, nfsd->nd_procp); - nfsrv_vput(vp); - nfsm_reply(NFSX_FH+NFSX_FATTR(nfsd->nd_nqlflag != NQL_NOVAL)); - nfsm_srvfhtom(fhp); - nfsm_build(fp, struct nfsv2_fattr *, NFSX_FATTR(nfsd->nd_nqlflag != NQL_NOVAL)); - nfsm_srvfillattr; - return (error); + if (v3) { + if (exclusive_flag && !error && + bcmp(cverf, (caddr_t)&vap->va_atime, NFSX_V3CREATEVERF)) + error = EEXIST; + diraft_ret = VOP_GETATTR(dirp, &diraft, cred, procp); + nfsrv_vrele(dirp); + } + nfsm_reply(NFSX_SRVFH(v3) + NFSX_FATTR(v3) + NFSX_WCCDATA(v3)); + if (v3) { + if (!error) { + nfsm_srvpostop_fh(fhp); + nfsm_srvpostop_attr(0, vap); + } + nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); + } else { + nfsm_srvfhtom(fhp, v3); + nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR); + nfsm_srvfillattr(vap, fp); + } + return (0); nfsmout: - if (nd.ni_cnd.cn_nameiop || nd.ni_cnd.cn_flags) + if (dirp) + nfsrv_vrele(dirp); + if (nd.ni_cnd.cn_nameiop) { nfsrv_vrele(nd.ni_startdir); + free((caddr_t)nd.ni_cnd.cn_pnbuf, M_NAMEI); + } VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); if (nd.ni_dvp == nd.ni_vp) nfsrv_vrele(nd.ni_dvp); @@ -852,79 +1495,248 @@ nfsmout: if (nd.ni_vp) nfsrv_vput(nd.ni_vp); return (error); +} + +/* + * nfs v3 mknod service + */ +int +nfsrv_mknod(nfsd, slp, procp, mrq) + struct nfsrv_descript *nfsd; + struct nfssvc_sock *slp; + struct proc *procp; + struct mbuf **mrq; +{ + struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; + struct mbuf *nam = nfsd->nd_nam; + caddr_t dpos = nfsd->nd_dpos; + struct ucred *cred = &nfsd->nd_cr; + register struct nfs_fattr *fp; + struct vattr va, dirfor, diraft; + register struct vattr *vap = &va; + register u_long *tl; + struct nameidata nd; + register caddr_t cp; + register long t1; + caddr_t bpos; + int error = 0, cache, len, tsize, dirfor_ret = 1, diraft_ret = 1; + u_long major, minor; + enum vtype vtyp; + char *cp2; + struct mbuf *mb, *mb2, *mreq; + struct vnode *vp, *dirp = (struct vnode *)0; + nfsfh_t nfh; + fhandle_t *fhp; + u_quad_t frev; + nd.ni_cnd.cn_nameiop = 0; + fhp = &nfh.fh_generic; + nfsm_srvmtofh(fhp); + nfsm_srvnamesiz(len); + nd.ni_cnd.cn_cred = cred; + nd.ni_cnd.cn_nameiop = CREATE; + nd.ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF | SAVESTART; + error = nfs_namei(&nd, fhp, len, slp, nam, &md, &dpos, + &dirp, procp, (nfsd->nd_flag & ND_KERBAUTH)); + if (dirp) + dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred, procp); + if (error) { + nfsm_reply(NFSX_WCCDATA(1)); + nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); + if (dirp) + nfsrv_vrele(dirp); + return (0); + } + nfsm_dissect(tl, u_long *, NFSX_UNSIGNED); + vtyp = nfsv3tov_type(*tl); + if (vtyp != VCHR && vtyp != VBLK && vtyp != VSOCK && vtyp != VFIFO) { + nfsrv_vrele(nd.ni_startdir); + free((caddr_t)nd.ni_cnd.cn_pnbuf, M_NAMEI); + error = NFSERR_BADTYPE; + VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); + nfsrv_vput(nd.ni_dvp); + goto out; + } + VATTR_NULL(vap); + nfsm_srvsattr(vap); + if (vtyp == VCHR || vtyp == VBLK) { + nfsm_dissect(tl, u_long *, 2 * NFSX_UNSIGNED); + major = fxdr_unsigned(u_long, *tl++); + minor = fxdr_unsigned(u_long, *tl); + vap->va_rdev = makedev(major, minor); + } + + /* + * Iff doesn't exist, create it. + */ + if (nd.ni_vp) { + nfsrv_vrele(nd.ni_startdir); + free((caddr_t)nd.ni_cnd.cn_pnbuf, M_NAMEI); + error = EEXIST; + VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); + nfsrv_vput(nd.ni_dvp); + goto out; + } + vap->va_type = vtyp; + if (vtyp == VSOCK) { + nfsrv_vrele(nd.ni_startdir); + nqsrv_getl(nd.ni_dvp, ND_WRITE); + error = VOP_CREATE(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, vap); + if (!error) + FREE(nd.ni_cnd.cn_pnbuf, M_NAMEI); + } else { + if (error = suser(cred, (u_short *)0)) { + nfsrv_vrele(nd.ni_startdir); + free((caddr_t)nd.ni_cnd.cn_pnbuf, M_NAMEI); + VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); + nfsrv_vput(nd.ni_dvp); + goto out; + } + nqsrv_getl(nd.ni_dvp, ND_WRITE); + if (error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, vap)) { + nfsrv_vrele(nd.ni_startdir); + goto out; + } + nd.ni_cnd.cn_nameiop = LOOKUP; + nd.ni_cnd.cn_flags &= ~(LOCKPARENT | SAVESTART); + nd.ni_cnd.cn_proc = procp; + nd.ni_cnd.cn_cred = procp->p_ucred; + error = lookup(&nd); + FREE(nd.ni_cnd.cn_pnbuf, M_NAMEI); + if (error) + goto out; + if (nd.ni_cnd.cn_flags & ISSYMLINK) { + nfsrv_vrele(nd.ni_dvp); + nfsrv_vput(nd.ni_vp); + VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); + error = EINVAL; + } + } out: - nfsrv_vrele(nd.ni_startdir); - free(nd.ni_cnd.cn_pnbuf, M_NAMEI); - nfsm_reply(0); + vp = nd.ni_vp; + if (!error) { + bzero((caddr_t)fhp, sizeof(nfh)); + fhp->fh_fsid = vp->v_mount->mnt_stat.f_fsid; + error = VFS_VPTOFH(vp, &fhp->fh_fid); + if (!error) + error = VOP_GETATTR(vp, vap, cred, procp); + nfsrv_vput(vp); + } + diraft_ret = VOP_GETATTR(dirp, &diraft, cred, procp); + nfsrv_vrele(dirp); + nfsm_reply(NFSX_SRVFH(1) + NFSX_POSTOPATTR(1) + NFSX_WCCDATA(1)); + if (!error) { + nfsm_srvpostop_fh(fhp); + nfsm_srvpostop_attr(0, vap); + } + nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); return (0); +nfsmout: + if (dirp) + nfsrv_vrele(dirp); + if (nd.ni_cnd.cn_nameiop) { + nfsrv_vrele(nd.ni_startdir); + free((caddr_t)nd.ni_cnd.cn_pnbuf, M_NAMEI); + } + VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); + if (nd.ni_dvp == nd.ni_vp) + nfsrv_vrele(nd.ni_dvp); + else + nfsrv_vput(nd.ni_dvp); + if (nd.ni_vp) + nfsrv_vput(nd.ni_vp); + return (error); } /* * nfs remove service */ int -nfsrv_remove(nfsd, mrep, md, dpos, cred, nam, mrq) - struct nfsd *nfsd; - struct mbuf *mrep, *md; - caddr_t dpos; - struct ucred *cred; - struct mbuf *nam, **mrq; +nfsrv_remove(nfsd, slp, procp, mrq) + struct nfsrv_descript *nfsd; + struct nfssvc_sock *slp; + struct proc *procp; + struct mbuf **mrq; { + struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; + struct mbuf *nam = nfsd->nd_nam; + caddr_t dpos = nfsd->nd_dpos; + struct ucred *cred = &nfsd->nd_cr; struct nameidata nd; register u_long *tl; register long t1; caddr_t bpos; - int error = 0, cache, len; + int error = 0, cache, len, dirfor_ret = 1, diraft_ret = 1; + int v3 = (nfsd->nd_flag & ND_NFSV3); char *cp2; - struct mbuf *mb, *mreq; - struct vnode *vp; - nfsv2fh_t nfh; + struct mbuf *mb, *mreq, *mb2; + struct vnode *vp, *dirp; + struct vattr dirfor, diraft; + nfsfh_t nfh; fhandle_t *fhp; u_quad_t frev; +#ifndef nolint + vp = (struct vnode *)0; +#endif fhp = &nfh.fh_generic; nfsm_srvmtofh(fhp); - nfsm_srvstrsiz(len, NFS_MAXNAMLEN); + nfsm_srvnamesiz(len); nd.ni_cnd.cn_cred = cred; nd.ni_cnd.cn_nameiop = DELETE; nd.ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF; - error = nfs_namei(&nd, fhp, len, nfsd->nd_slp, nam, &md, &dpos, - nfsd->nd_procp); - if (error) - nfsm_reply(0); - vp = nd.ni_vp; - if (vp->v_type == VDIR && - (error = suser(cred, (u_short *)0))) - goto out; - /* - * The root of a mounted filesystem cannot be deleted. - */ - if (vp->v_flag & VROOT) { - error = EBUSY; - goto out; + error = nfs_namei(&nd, fhp, len, slp, nam, &md, &dpos, + &dirp, procp, (nfsd->nd_flag & ND_KERBAUTH)); + if (dirp) { + if (v3) + dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred, + procp); + else + nfsrv_vrele(dirp); } - (void) vnode_pager_uncache(vp); -out: if (!error) { - int deallocobj = 0; - nqsrv_getl(nd.ni_dvp, NQL_WRITE); - nqsrv_getl(vp, NQL_WRITE); - - if ((vp->v_flag & VVMIO) && vp->v_vmdata) - deallocobj = 1; - error = VOP_REMOVE(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); - if (error == 0 && deallocobj) - vm_object_deallocate((vm_object_t) vp->v_vmdata); - } else { - VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); - if (nd.ni_dvp == vp) - nfsrv_vrele(nd.ni_dvp); - else - nfsrv_vput(nd.ni_dvp); - nfsrv_vput(vp); + vp = nd.ni_vp; + if (vp->v_type == VDIR && + (error = suser(cred, (u_short *)0))) + goto out; + /* + * The root of a mounted filesystem cannot be deleted. + */ + if (vp->v_flag & VROOT) { + error = EBUSY; + goto out; + } + if (vp->v_flag & VTEXT) + (void) vnode_pager_uncache(vp); +out: + if (!error) { + int deallocobj = 0; + nqsrv_getl(nd.ni_dvp, ND_WRITE); + nqsrv_getl(vp, ND_WRITE); + + if ((vp->v_flag & VVMIO) && vp->v_vmdata) + deallocobj = 1; + error = VOP_REMOVE(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); + if (error == 0 && deallocobj) + vm_object_deallocate((vm_object_t) vp->v_vmdata); + } else { + VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); + if (nd.ni_dvp == vp) + nfsrv_vrele(nd.ni_dvp); + else + nfsrv_vput(nd.ni_dvp); + nfsrv_vput(vp); + } + } + if (dirp && v3) { + diraft_ret = VOP_GETATTR(dirp, &diraft, cred, procp); + nfsrv_vrele(dirp); + } + nfsm_reply(NFSX_WCCDATA(v3)); + if (v3) { + nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); + return (0); } - nfsm_reply(0); nfsm_srvdone; } @@ -932,32 +1744,42 @@ out: * nfs rename service */ int -nfsrv_rename(nfsd, mrep, md, dpos, cred, nam, mrq) - struct nfsd *nfsd; - struct mbuf *mrep, *md; - caddr_t dpos; - struct ucred *cred; - struct mbuf *nam, **mrq; +nfsrv_rename(nfsd, slp, procp, mrq) + struct nfsrv_descript *nfsd; + struct nfssvc_sock *slp; + struct proc *procp; + struct mbuf **mrq; { + struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; + struct mbuf *nam = nfsd->nd_nam; + caddr_t dpos = nfsd->nd_dpos; + struct ucred *cred = &nfsd->nd_cr; register u_long *tl; register long t1; caddr_t bpos; - int error = 0, cache, len, len2; + int error = 0, cache, len, len2, fdirfor_ret = 1, fdiraft_ret = 1; + int tdirfor_ret = 1, tdiraft_ret = 1; + int v3 = (nfsd->nd_flag & ND_NFSV3); char *cp2; - struct mbuf *mb, *mreq; + struct mbuf *mb, *mreq, *mb2; struct nameidata fromnd, tond; - struct vnode *fvp = 0, *tvp, *tdvp; - nfsv2fh_t fnfh, tnfh; + struct vnode *fvp, *tvp, *tdvp, *fdirp = (struct vnode *)0; + struct vnode *tdirp = (struct vnode *)0; + struct vattr fdirfor, fdiraft, tdirfor, tdiraft; + nfsfh_t fnfh, tnfh; fhandle_t *ffhp, *tfhp; u_quad_t frev; uid_t saved_uid; +#ifndef nolint + fvp = (struct vnode *)0; +#endif ffhp = &fnfh.fh_generic; tfhp = &tnfh.fh_generic; fromnd.ni_cnd.cn_nameiop = 0; tond.ni_cnd.cn_nameiop = 0; nfsm_srvmtofh(ffhp); - nfsm_srvstrsiz(len, NFS_MAXNAMLEN); + nfsm_srvnamesiz(len); /* * Remember our original uid so that we can reset cr_uid before * the second nfs_namei() call, in case it is remapped. @@ -966,10 +1788,25 @@ nfsrv_rename(nfsd, mrep, md, dpos, cred, nam, mrq) fromnd.ni_cnd.cn_cred = cred; fromnd.ni_cnd.cn_nameiop = DELETE; fromnd.ni_cnd.cn_flags = WANTPARENT | SAVESTART; - error = nfs_namei(&fromnd, ffhp, len, nfsd->nd_slp, nam, &md, &dpos, - nfsd->nd_procp); - if(error) - nfsm_reply(0); + error = nfs_namei(&fromnd, ffhp, len, slp, nam, &md, + &dpos, &fdirp, procp, (nfsd->nd_flag & ND_KERBAUTH)); + if (fdirp) { + if (v3) + fdirfor_ret = VOP_GETATTR(fdirp, &fdirfor, cred, + procp); + else { + nfsrv_vrele(fdirp); + fdirp = (struct vnode *)0; + } + } + if (error) { + nfsm_reply(2 * NFSX_WCCDATA(v3)); + nfsm_srvwcc_data(fdirfor_ret, &fdirfor, fdiraft_ret, &fdiraft); + nfsm_srvwcc_data(tdirfor_ret, &tdirfor, tdiraft_ret, &tdiraft); + if (fdirp) + nfsrv_vrele(fdirp); + return (0); + } fvp = fromnd.ni_vp; nfsm_srvmtofh(tfhp); nfsm_strsiz(len2, NFS_MAXNAMLEN); @@ -977,8 +1814,17 @@ nfsrv_rename(nfsd, mrep, md, dpos, cred, nam, mrq) tond.ni_cnd.cn_cred = cred; tond.ni_cnd.cn_nameiop = RENAME; tond.ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF | NOCACHE | SAVESTART; - error = nfs_namei(&tond, tfhp, len2, nfsd->nd_slp, nam, &md, - &dpos, nfsd->nd_procp); + error = nfs_namei(&tond, tfhp, len2, slp, nam, &md, + &dpos, &tdirp, procp, (nfsd->nd_flag & ND_KERBAUTH)); + if (tdirp) { + if (v3) + tdirfor_ret = VOP_GETATTR(tdirp, &tdirfor, cred, + procp); + else { + nfsrv_vrele(tdirp); + tdirp = (struct vnode *)0; + } + } if (error) { VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); nfsrv_vrele(fromnd.ni_dvp); @@ -989,27 +1835,45 @@ nfsrv_rename(nfsd, mrep, md, dpos, cred, nam, mrq) tvp = tond.ni_vp; if (tvp != NULL) { if (fvp->v_type == VDIR && tvp->v_type != VDIR) { - error = EISDIR; + if (v3) + error = EEXIST; + else + error = EISDIR; goto out; } else if (fvp->v_type != VDIR && tvp->v_type == VDIR) { - error = ENOTDIR; + if (v3) + error = EEXIST; + else + error = ENOTDIR; goto out; } if (tvp->v_type == VDIR && tvp->v_mountedhere) { - error = EXDEV; + if (v3) + error = EXDEV; + else + error = ENOTEMPTY; goto out; } } if (fvp->v_type == VDIR && fvp->v_mountedhere) { - error = EBUSY; + if (v3) + error = EXDEV; + else + error = ENOTEMPTY; goto out; } if (fvp->v_mount != tdvp->v_mount) { - error = EXDEV; + if (v3) + error = EXDEV; + else + error = ENOTEMPTY; goto out; } if (fvp == tdvp) - error = EINVAL; + if (v3) + error = EINVAL; + else + error = ENOTEMPTY; /* * If source is the same as the destination (that is the * same vnode with the same name in the same directory), @@ -1023,10 +1887,10 @@ nfsrv_rename(nfsd, mrep, md, dpos, cred, nam, mrq) out: if (!error) { int deallocobjfrom = 0, deallocobjto = 0; - nqsrv_getl(fromnd.ni_dvp, NQL_WRITE); - nqsrv_getl(tdvp, NQL_WRITE); + nqsrv_getl(fromnd.ni_dvp, ND_WRITE); + nqsrv_getl(tdvp, ND_WRITE); if (tvp) { - nqsrv_getl(tvp, NQL_WRITE); + nqsrv_getl(tvp, ND_WRITE); if ((tvp->v_flag & VVMIO) && tvp->v_vmdata) deallocobjto = 1; } @@ -1050,21 +1914,39 @@ out: VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); nfsrv_vrele(fromnd.ni_dvp); nfsrv_vrele(fvp); + if (error == -1) + error = 0; } nfsrv_vrele(tond.ni_startdir); FREE(tond.ni_cnd.cn_pnbuf, M_NAMEI); out1: + if (fdirp) { + fdiraft_ret = VOP_GETATTR(fdirp, &fdiraft, cred, procp); + nfsrv_vrele(fdirp); + } + if (tdirp) { + tdiraft_ret = VOP_GETATTR(tdirp, &tdiraft, cred, procp); + nfsrv_vrele(tdirp); + } nfsrv_vrele(fromnd.ni_startdir); FREE(fromnd.ni_cnd.cn_pnbuf, M_NAMEI); - nfsm_reply(0); - return (error); + nfsm_reply(2 * NFSX_WCCDATA(v3)); + if (v3) { + nfsm_srvwcc_data(fdirfor_ret, &fdirfor, fdiraft_ret, &fdiraft); + nfsm_srvwcc_data(tdirfor_ret, &tdirfor, tdiraft_ret, &tdiraft); + } + return (0); nfsmout: - if (tond.ni_cnd.cn_nameiop || tond.ni_cnd.cn_flags) { + if (fdirp) + nfsrv_vrele(fdirp); + if (tdirp) + nfsrv_vrele(tdirp); + if (tond.ni_cnd.cn_nameiop) { nfsrv_vrele(tond.ni_startdir); FREE(tond.ni_cnd.cn_pnbuf, M_NAMEI); } - if (fromnd.ni_cnd.cn_nameiop || fromnd.ni_cnd.cn_flags) { + if (fromnd.ni_cnd.cn_nameiop) { nfsrv_vrele(fromnd.ni_startdir); FREE(fromnd.ni_cnd.cn_pnbuf, M_NAMEI); VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd); @@ -1078,22 +1960,27 @@ nfsmout: * nfs link service */ int -nfsrv_link(nfsd, mrep, md, dpos, cred, nam, mrq) - struct nfsd *nfsd; - struct mbuf *mrep, *md; - caddr_t dpos; - struct ucred *cred; - struct mbuf *nam, **mrq; +nfsrv_link(nfsd, slp, procp, mrq) + struct nfsrv_descript *nfsd; + struct nfssvc_sock *slp; + struct proc *procp; + struct mbuf **mrq; { + struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; + struct mbuf *nam = nfsd->nd_nam; + caddr_t dpos = nfsd->nd_dpos; + struct ucred *cred = &nfsd->nd_cr; struct nameidata nd; register u_long *tl; register long t1; caddr_t bpos; - int error = 0, rdonly, cache, len; + int error = 0, rdonly, cache, len, dirfor_ret = 1, diraft_ret = 1; + int getret = 1, v3 = (nfsd->nd_flag & ND_NFSV3); char *cp2; - struct mbuf *mb, *mreq; - struct vnode *vp, *xp; - nfsv2fh_t nfh, dnfh; + struct mbuf *mb, *mreq, *mb2; + struct vnode *vp, *xp, *dirp = (struct vnode *)0; + struct vattr dirfor, diraft, at; + nfsfh_t nfh, dnfh; fhandle_t *fhp, *dfhp; u_quad_t frev; @@ -1101,17 +1988,30 @@ nfsrv_link(nfsd, mrep, md, dpos, cred, nam, mrq) dfhp = &dnfh.fh_generic; nfsm_srvmtofh(fhp); nfsm_srvmtofh(dfhp); - nfsm_srvstrsiz(len, NFS_MAXNAMLEN); - error = nfsrv_fhtovp(fhp, FALSE, &vp, cred, nfsd->nd_slp, nam, &rdonly); - if (error) - nfsm_reply(0); + nfsm_srvnamesiz(len); + if (error = nfsrv_fhtovp(fhp, FALSE, &vp, cred, slp, nam, + &rdonly, (nfsd->nd_flag & ND_KERBAUTH))) { + nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_WCCDATA(v3)); + nfsm_srvpostop_attr(getret, &at); + nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); + return (0); + } if (vp->v_type == VDIR && (error = suser(cred, (u_short *)0))) goto out1; nd.ni_cnd.cn_cred = cred; nd.ni_cnd.cn_nameiop = CREATE; nd.ni_cnd.cn_flags = LOCKPARENT; - error = nfs_namei(&nd, dfhp, len, nfsd->nd_slp, nam, &md, &dpos, - nfsd->nd_procp); + error = nfs_namei(&nd, dfhp, len, slp, nam, &md, &dpos, + &dirp, procp, (nfsd->nd_flag & ND_KERBAUTH)); + if (dirp) { + if (v3) + dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred, + procp); + else { + nfsrv_vrele(dirp); + dirp = (struct vnode *)0; + } + } if (error) goto out1; xp = nd.ni_vp; @@ -1124,9 +2024,13 @@ nfsrv_link(nfsd, mrep, md, dpos, cred, nam, mrq) error = EXDEV; out: if (!error) { - nqsrv_getl(vp, NQL_WRITE); - nqsrv_getl(xp, NQL_WRITE); + nqsrv_getl(vp, ND_WRITE); + nqsrv_getl(xp, ND_WRITE); +#if defined(__FreeBSD__) || defined(__NetBSD__) error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd); +#else + error = VOP_LINK(vp, nd.ni_dvp, &nd.ni_cnd); +#endif } else { VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); if (nd.ni_dvp == nd.ni_vp) @@ -1137,8 +2041,19 @@ out: nfsrv_vrele(nd.ni_vp); } out1: + if (v3) + getret = VOP_GETATTR(vp, &at, cred, procp); + if (dirp) { + diraft_ret = VOP_GETATTR(dirp, &diraft, cred, procp); + nfsrv_vrele(dirp); + } nfsrv_vrele(vp); - nfsm_reply(0); + nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_WCCDATA(v3)); + if (v3) { + nfsm_srvpostop_attr(getret, &at); + nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); + return (0); + } nfsm_srvdone; } @@ -1146,42 +2061,58 @@ out1: * nfs symbolic link service */ int -nfsrv_symlink(nfsd, mrep, md, dpos, cred, nam, mrq) - struct nfsd *nfsd; - struct mbuf *mrep, *md; - caddr_t dpos; - struct ucred *cred; - struct mbuf *nam, **mrq; +nfsrv_symlink(nfsd, slp, procp, mrq) + struct nfsrv_descript *nfsd; + struct nfssvc_sock *slp; + struct proc *procp; + struct mbuf **mrq; { + struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; + struct mbuf *nam = nfsd->nd_nam; + caddr_t dpos = nfsd->nd_dpos; + struct ucred *cred = &nfsd->nd_cr; + struct vattr va, dirfor, diraft; struct vnode *ovp; - struct vattr va; struct nameidata nd; register struct vattr *vap = &va; register u_long *tl; register long t1; struct nfsv2_sattr *sp; - caddr_t bpos; + char *bpos, *cp, *pathcp = (char *)0, *cp2; struct uio io; struct iovec iv; - int error = 0, cache, len, len2; - char *pathcp, *cp2; - struct mbuf *mb, *mreq; - nfsv2fh_t nfh; + int error = 0, cache, len, len2, dirfor_ret = 1, diraft_ret = 1; + int v3 = (nfsd->nd_flag & ND_NFSV3); + struct mbuf *mb, *mreq, *mb2; + struct vnode *dirp = (struct vnode *)0; + nfsfh_t nfh; fhandle_t *fhp; u_quad_t frev; int deallocobj = 0; - pathcp = (char *)0; + nd.ni_cnd.cn_nameiop = 0; fhp = &nfh.fh_generic; nfsm_srvmtofh(fhp); - nfsm_srvstrsiz(len, NFS_MAXNAMLEN); + nfsm_srvnamesiz(len); nd.ni_cnd.cn_cred = cred; nd.ni_cnd.cn_nameiop = CREATE; - nd.ni_cnd.cn_flags = LOCKPARENT; - error = nfs_namei(&nd, fhp, len, nfsd->nd_slp, nam, &md, &dpos, - nfsd->nd_procp); + nd.ni_cnd.cn_flags = LOCKPARENT | SAVESTART; + error = nfs_namei(&nd, fhp, len, slp, nam, &md, &dpos, + &dirp, procp, (nfsd->nd_flag & ND_KERBAUTH)); + if (dirp) { + if (v3) + dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred, + procp); + else { + nfsrv_vrele(dirp); + dirp = (struct vnode *)0; + } + } if (error) goto out; + VATTR_NULL(vap); + if (v3) + nfsm_srvsattr(vap); nfsm_strsiz(len2, NFS_MAXPATHLEN); MALLOC(pathcp, caddr_t, len2 + 1, M_TEMP, M_WAITOK); iv.iov_base = pathcp; @@ -1194,9 +2125,14 @@ nfsrv_symlink(nfsd, mrep, md, dpos, cred, nam, mrq) io.uio_rw = UIO_READ; io.uio_procp = (struct proc *)0; nfsm_mtouio(&io, len2); - nfsm_dissect(sp, struct nfsv2_sattr *, NFSX_SATTR(nfsd->nd_nqlflag != NQL_NOVAL)); + if (!v3) { + nfsm_dissect(sp, struct nfsv2_sattr *, NFSX_V2SATTR); + vap->va_mode = fxdr_unsigned(u_short, sp->sa_mode); + } *(pathcp + len2) = '\0'; if (nd.ni_vp) { + nfsrv_vrele(nd.ni_startdir); + free(nd.ni_cnd.cn_pnbuf, M_NAMEI); VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); if (nd.ni_dvp == nd.ni_vp) nfsrv_vrele(nd.ni_dvp); @@ -1206,20 +2142,58 @@ nfsrv_symlink(nfsd, mrep, md, dpos, cred, nam, mrq) error = EEXIST; goto out; } - VATTR_NULL(vap); - vap->va_mode = fxdr_unsigned(u_short, sp->sa_mode); - nqsrv_getl(nd.ni_dvp, NQL_WRITE); + nqsrv_getl(nd.ni_dvp, ND_WRITE); if ((ovp = nd.ni_vp) && (ovp->v_flag & VVMIO) && ovp->v_vmdata) deallocobj = 1; error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, vap, pathcp); if (error == 0 && deallocobj) vm_object_deallocate( (vm_object_t) ovp->v_vmdata); + if (error) + nfsrv_vrele(nd.ni_startdir); + else { + if (v3) { + nd.ni_cnd.cn_nameiop = LOOKUP; + nd.ni_cnd.cn_flags &= ~(LOCKPARENT | SAVESTART | FOLLOW); + nd.ni_cnd.cn_flags |= (NOFOLLOW | LOCKLEAF); + nd.ni_cnd.cn_proc = procp; + nd.ni_cnd.cn_cred = cred; + error = lookup(&nd); + if (!error) { + bzero((caddr_t)fhp, sizeof(nfh)); + fhp->fh_fsid = nd.ni_vp->v_mount->mnt_stat.f_fsid; + error = VFS_VPTOFH(nd.ni_vp, &fhp->fh_fid); + if (!error) + error = VOP_GETATTR(nd.ni_vp, vap, cred, + procp); + nfsrv_vput(nd.ni_vp); + } + } else + nfsrv_vrele(nd.ni_startdir); + FREE(nd.ni_cnd.cn_pnbuf, M_NAMEI); + } out: if (pathcp) FREE(pathcp, M_TEMP); - nfsm_reply(0); - return (error); + if (dirp) { + diraft_ret = VOP_GETATTR(dirp, &diraft, cred, procp); + nfsrv_vrele(dirp); + } + nfsm_reply(NFSX_SRVFH(v3) + NFSX_POSTOPATTR(v3) + NFSX_WCCDATA(v3)); + if (v3) { + if (!error) { + nfsm_srvpostop_fh(fhp); + nfsm_srvpostop_attr(0, vap); + } + nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); + } + return (0); nfsmout: + if (nd.ni_cnd.cn_nameiop) { + nfsrv_vrele(nd.ni_startdir); + free(nd.ni_cnd.cn_pnbuf, M_NAMEI); + } + if (dirp) + nfsrv_vrele(dirp); VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); if (nd.ni_dvp == nd.ni_vp) nfsrv_vrele(nd.ni_dvp); @@ -1236,43 +2210,65 @@ nfsmout: * nfs mkdir service */ int -nfsrv_mkdir(nfsd, mrep, md, dpos, cred, nam, mrq) - struct nfsd *nfsd; - struct mbuf *mrep, *md; - caddr_t dpos; - struct ucred *cred; - struct mbuf *nam, **mrq; +nfsrv_mkdir(nfsd, slp, procp, mrq) + struct nfsrv_descript *nfsd; + struct nfssvc_sock *slp; + struct proc *procp; + struct mbuf **mrq; { - struct vattr va; + struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; + struct mbuf *nam = nfsd->nd_nam; + caddr_t dpos = nfsd->nd_dpos; + struct ucred *cred = &nfsd->nd_cr; + struct vattr va, dirfor, diraft; register struct vattr *vap = &va; - register struct nfsv2_fattr *fp; + register struct nfs_fattr *fp; struct nameidata nd; register caddr_t cp; register u_long *tl; register long t1; caddr_t bpos; - int error = 0, cache, len; + int error = 0, cache, len, dirfor_ret = 1, diraft_ret = 1; + int v3 = (nfsd->nd_flag & ND_NFSV3); char *cp2; struct mbuf *mb, *mb2, *mreq; - struct vnode *vp; - nfsv2fh_t nfh; + struct vnode *vp, *dirp = (struct vnode *)0; + nfsfh_t nfh; fhandle_t *fhp; u_quad_t frev; fhp = &nfh.fh_generic; nfsm_srvmtofh(fhp); - nfsm_srvstrsiz(len, NFS_MAXNAMLEN); + nfsm_srvnamesiz(len); nd.ni_cnd.cn_cred = cred; nd.ni_cnd.cn_nameiop = CREATE; nd.ni_cnd.cn_flags = LOCKPARENT; - error = nfs_namei(&nd, fhp, len, nfsd->nd_slp, nam, &md, &dpos, - nfsd->nd_procp); - if (error) - nfsm_reply(0); - nfsm_dissect(tl, u_long *, NFSX_UNSIGNED); + error = nfs_namei(&nd, fhp, len, slp, nam, &md, &dpos, + &dirp, procp, (nfsd->nd_flag & ND_KERBAUTH)); + if (dirp) { + if (v3) + dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred, + procp); + else { + nfsrv_vrele(dirp); + dirp = (struct vnode *)0; + } + } + if (error) { + nfsm_reply(NFSX_WCCDATA(v3)); + nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); + if (dirp) + nfsrv_vrele(dirp); + return (0); + } VATTR_NULL(vap); + if (v3) { + nfsm_srvsattr(vap); + } else { + nfsm_dissect(tl, u_long *, NFSX_UNSIGNED); + vap->va_mode = nfstov_mode(*tl++); + } vap->va_type = VDIR; - vap->va_mode = nfstov_mode(*tl++); vp = nd.ni_vp; if (vp != NULL) { VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); @@ -1282,28 +2278,40 @@ nfsrv_mkdir(nfsd, mrep, md, dpos, cred, nam, mrq) nfsrv_vput(nd.ni_dvp); nfsrv_vrele(vp); error = EEXIST; - nfsm_reply(0); + goto out; } - nqsrv_getl(nd.ni_dvp, NQL_WRITE); + nqsrv_getl(nd.ni_dvp, ND_WRITE); error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, vap); - if (error) - nfsm_reply(0); - vp = nd.ni_vp; - bzero((caddr_t)fhp, sizeof(nfh)); - fhp->fh_fsid = vp->v_mount->mnt_stat.f_fsid; - error = VFS_VPTOFH(vp, &fhp->fh_fid); - if (error) { + if (!error) { + vp = nd.ni_vp; + bzero((caddr_t)fhp, sizeof(nfh)); + fhp->fh_fsid = vp->v_mount->mnt_stat.f_fsid; + error = VFS_VPTOFH(vp, &fhp->fh_fid); + if (!error) + error = VOP_GETATTR(vp, vap, cred, procp); nfsrv_vput(vp); - nfsm_reply(0); } - error = VOP_GETATTR(vp, vap, cred, nfsd->nd_procp); - nfsrv_vput(vp); - nfsm_reply(NFSX_FH+NFSX_FATTR(nfsd->nd_nqlflag != NQL_NOVAL)); - nfsm_srvfhtom(fhp); - nfsm_build(fp, struct nfsv2_fattr *, NFSX_FATTR(nfsd->nd_nqlflag != NQL_NOVAL)); - nfsm_srvfillattr; - return (error); +out: + if (dirp) { + diraft_ret = VOP_GETATTR(dirp, &diraft, cred, procp); + nfsrv_vrele(dirp); + } + nfsm_reply(NFSX_SRVFH(v3) + NFSX_POSTOPATTR(v3) + NFSX_WCCDATA(v3)); + if (v3) { + if (!error) { + nfsm_srvpostop_fh(fhp); + nfsm_srvpostop_attr(0, vap); + } + nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); + } else { + nfsm_srvfhtom(fhp, v3); + nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR); + nfsm_srvfillattr(vap, fp); + } + return (0); nfsmout: + if (dirp) + nfsrv_vrele(dirp); VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); if (nd.ni_dvp == nd.ni_vp) nfsrv_vrele(nd.ni_dvp); @@ -1318,35 +2326,54 @@ nfsmout: * nfs rmdir service */ int -nfsrv_rmdir(nfsd, mrep, md, dpos, cred, nam, mrq) - struct nfsd *nfsd; - struct mbuf *mrep, *md; - caddr_t dpos; - struct ucred *cred; - struct mbuf *nam, **mrq; +nfsrv_rmdir(nfsd, slp, procp, mrq) + struct nfsrv_descript *nfsd; + struct nfssvc_sock *slp; + struct proc *procp; + struct mbuf **mrq; { + struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; + struct mbuf *nam = nfsd->nd_nam; + caddr_t dpos = nfsd->nd_dpos; + struct ucred *cred = &nfsd->nd_cr; register u_long *tl; register long t1; caddr_t bpos; - int error = 0, cache, len; + int error = 0, cache, len, dirfor_ret = 1, diraft_ret = 1; + int v3 = (nfsd->nd_flag & ND_NFSV3); char *cp2; - struct mbuf *mb, *mreq; - struct vnode *vp; - nfsv2fh_t nfh; + struct mbuf *mb, *mreq, *mb2; + struct vnode *vp, *dirp = (struct vnode *)0; + struct vattr dirfor, diraft; + nfsfh_t nfh; fhandle_t *fhp; struct nameidata nd; u_quad_t frev; fhp = &nfh.fh_generic; nfsm_srvmtofh(fhp); - nfsm_srvstrsiz(len, NFS_MAXNAMLEN); + nfsm_srvnamesiz(len); nd.ni_cnd.cn_cred = cred; nd.ni_cnd.cn_nameiop = DELETE; nd.ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF; - error = nfs_namei(&nd, fhp, len, nfsd->nd_slp, nam, &md, &dpos, - nfsd->nd_procp); - if (error) - nfsm_reply(0); + error = nfs_namei(&nd, fhp, len, slp, nam, &md, &dpos, + &dirp, procp, (nfsd->nd_flag & ND_KERBAUTH)); + if (dirp) { + if (v3) + dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred, + procp); + else { + nfsrv_vrele(dirp); + dirp = (struct vnode *)0; + } + } + if (error) { + nfsm_reply(NFSX_WCCDATA(v3)); + nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); + if (dirp) + nfsrv_vrele(dirp); + return (0); + } vp = nd.ni_vp; if (vp->v_type != VDIR) { error = ENOTDIR; @@ -1366,8 +2393,8 @@ nfsrv_rmdir(nfsd, mrep, md, dpos, cred, nam, mrq) error = EBUSY; out: if (!error) { - nqsrv_getl(nd.ni_dvp, NQL_WRITE); - nqsrv_getl(vp, NQL_WRITE); + nqsrv_getl(nd.ni_dvp, ND_WRITE); + nqsrv_getl(vp, ND_WRITE); error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd); } else { VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd); @@ -1377,7 +2404,15 @@ out: nfsrv_vput(nd.ni_dvp); nfsrv_vput(vp); } - nfsm_reply(0); + if (dirp) { + diraft_ret = VOP_GETATTR(dirp, &diraft, cred, procp); + nfsrv_vrele(dirp); + } + nfsm_reply(NFSX_WCCDATA(v3)); + if (v3) { + nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft); + return (0); + } nfsm_srvdone; } @@ -1401,7 +2436,7 @@ out: * than requested, but this may not apply to all filesystems. For * example, client NFS does not { although it is never remote mounted * anyhow } - * The alternate call nqnfsrv_readdirlook() does lookups as well. + * The alternate call nfsrv_readdirplus() does lookups as well. * PS: The NFS protocol spec. does not clarify what the "count" byte * argument is a count of.. just name strings and file id's or the * entire reply rpc or ... @@ -1411,21 +2446,25 @@ out: * "entry" structures, but are in the rpc. */ struct flrep { - u_long fl_cachable; - u_long fl_duration; - u_long fl_frev[2]; - nfsv2fh_t fl_nfh; - u_long fl_fattr[NFSX_NQFATTR / sizeof (u_long)]; + nfsuint64 fl_off; + u_long fl_postopok; + u_long fl_fattr[NFSX_V3FATTR / sizeof (u_long)]; + u_long fl_fhok; + u_long fl_fhsize; + u_long fl_nfh[NFSX_V3FH / sizeof (u_long)]; }; int -nfsrv_readdir(nfsd, mrep, md, dpos, cred, nam, mrq) - struct nfsd *nfsd; - struct mbuf *mrep, *md; - caddr_t dpos; - struct ucred *cred; - struct mbuf *nam, **mrq; +nfsrv_readdir(nfsd, slp, procp, mrq) + struct nfsrv_descript *nfsd; + struct nfssvc_sock *slp; + struct proc *procp; + struct mbuf **mrq; { + struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; + struct mbuf *nam = nfsd->nd_nam; + caddr_t dpos = nfsd->nd_dpos; + struct ucred *cred = &nfsd->nd_cr; register char *bp, *be; register struct mbuf *mp; register struct dirent *dp; @@ -1436,38 +2475,63 @@ nfsrv_readdir(nfsd, mrep, md, dpos, cred, nam, mrq) struct mbuf *mb, *mb2, *mreq, *mp2; char *cpos, *cend, *cp2, *rbuf; struct vnode *vp; - nfsv2fh_t nfh; + struct vattr at; + nfsfh_t nfh; fhandle_t *fhp; struct uio io; struct iovec iv; - int len, nlen, rem, xfer, tsiz, i, error = 0; - int siz, cnt, fullsiz, eofflag, rdonly, cache; - u_quad_t frev; - u_long off, toff; - int ncookies; - u_int *cookies; - u_int *cookiep; + int len, nlen, rem, xfer, tsiz, i, error = 0, getret = 1; + int siz, cnt, fullsiz, eofflag, rdonly, cache, ncookies; + int v3 = (nfsd->nd_flag & ND_NFSV3); + u_quad_t frev, off, toff, verf; + u_long *cookies = NULL, *cookiep; fhp = &nfh.fh_generic; nfsm_srvmtofh(fhp); - nfsm_dissect(tl, u_long *, 2*NFSX_UNSIGNED); - off = toff = fxdr_unsigned(u_long, *tl++); + if (v3) { + nfsm_dissect(tl, u_long *, 5 * NFSX_UNSIGNED); + fxdr_hyper(tl, &toff); + tl += 2; + fxdr_hyper(tl, &verf); + tl += 2; + } else { + nfsm_dissect(tl, u_long *, 2 * NFSX_UNSIGNED); + toff = fxdr_unsigned(u_quad_t, *tl++); + } + off = toff; cnt = fxdr_unsigned(int, *tl); - siz = ((cnt+NFS_DIRBLKSIZ-1) & ~(NFS_DIRBLKSIZ-1)); - if (cnt > NFS_MAXREADDIR) - siz = NFS_MAXREADDIR; + siz = ((cnt + DIRBLKSIZ - 1) & ~(DIRBLKSIZ - 1)); + xfer = NFS_SRVMAXDATA(nfsd); + if (siz > xfer) + siz = xfer; fullsiz = siz; - error = nfsrv_fhtovp(fhp, TRUE, &vp, cred, nfsd->nd_slp, nam, &rdonly); - if (error) - nfsm_reply(0); - nqsrv_getl(vp, NQL_READ); - error = nfsrv_access(vp, VEXEC, cred, rdonly, nfsd->nd_procp); + if (error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam, + &rdonly, (nfsd->nd_flag & ND_KERBAUTH))) { + nfsm_reply(NFSX_UNSIGNED); + nfsm_srvpostop_attr(getret, &at); + return (0); + } + nqsrv_getl(vp, ND_READ); + if (v3) { + error = getret = VOP_GETATTR(vp, &at, cred, procp); + if (!error && toff && verf != at.va_filerev) + error = NFSERR_BAD_COOKIE; + } + if (!error) + error = nfsrv_access(vp, VEXEC, cred, rdonly, procp); if (error) { nfsrv_vput(vp); - nfsm_reply(0); + nfsm_reply(NFSX_POSTOPATTR(v3)); + nfsm_srvpostop_attr(getret, &at); + return (0); } VOP_UNLOCK(vp); MALLOC(rbuf, caddr_t, siz, M_TEMP, M_WAITOK); +#ifdef __NetBSD__ + ncookies = siz / (5 * NFSX_UNSIGNED); /*7 for V3, but it's an est. so*/ + MALLOC(cookies, u_long *, ncookies * sizeof (u_long *), M_TEMP, + M_WAITOK); +#endif again: iv.iov_base = rbuf; iv.iov_len = fullsiz; @@ -1478,25 +2542,31 @@ again: io.uio_segflg = UIO_SYSSPACE; io.uio_rw = UIO_READ; io.uio_procp = (struct proc *)0; - cookies = NULL; eofflag = 0; +#ifndef __NetBSD__ + if (cookies) { + free((caddr_t)cookies, M_TEMP); + cookies = NULL; + } error = VOP_READDIR(vp, &io, cred, &eofflag, &ncookies, &cookies); +#else + error = VOP_READDIR(vp, &io, cred, &eofflag, cookies, ncookies); +#endif off = (off_t)io.uio_offset; + if (!cookies && !error) + error = NFSERR_PERM; + if (v3) { + getret = VOP_GETATTR(vp, &at, cred, procp); + if (!error) + error = getret; + } if (error) { nfsrv_vrele(vp); free((caddr_t)rbuf, M_TEMP); - nfsm_reply(0); - } - if (cookies == NULL) { - /* - * If the filesystem doen't support cookies, return eof. - */ - nfsrv_vrele(vp); - nfsm_reply(2*NFSX_UNSIGNED); - nfsm_build(tl, u_long *, 2*NFSX_UNSIGNED); - *tl++ = nfs_false; - *tl = nfs_true; - FREE((caddr_t)rbuf, M_TEMP); + if (cookies) + free((caddr_t)cookies, M_TEMP); + nfsm_reply(NFSX_POSTOPATTR(v3)); + nfsm_srvpostop_attr(getret, &at); return (0); } if (io.uio_resid) { @@ -1508,12 +2578,19 @@ again: */ if (siz == 0) { nfsrv_vrele(vp); - nfsm_reply(2*NFSX_UNSIGNED); - nfsm_build(tl, u_long *, 2*NFSX_UNSIGNED); + nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_COOKIEVERF(v3) + + 2 * NFSX_UNSIGNED); + if (v3) { + nfsm_srvpostop_attr(getret, &at); + nfsm_build(tl, u_long *, 4 * NFSX_UNSIGNED); + txdr_hyper(&at.va_filerev, tl); + tl += 2; + } else + nfsm_build(tl, u_long *, 2 * NFSX_UNSIGNED); *tl++ = nfs_false; *tl = nfs_true; FREE((caddr_t)rbuf, M_TEMP); - FREE(cookies, M_TEMP); + FREE((caddr_t)cookies, M_TEMP); return (0); } } @@ -1526,32 +2603,49 @@ again: cend = rbuf + siz; dp = (struct dirent *)cpos; cookiep = cookies; - while (cpos < cend && ncookies > 0 - && (*cookiep <= toff || dp->d_fileno == 0)) { +#ifdef __FreeBSD__ + /* + * For some reason FreeBSD's ufs_readdir() chooses to back the + * directory offset up to a block boundary, so it is necessary to + * skip over the records that preceed the requested offset. This + * requires the assumption that file offset cookies monotonically + * increase. + */ + while (cpos < cend && ncookies > 0 && + (dp->d_fileno == 0 || ((u_quad_t)(*cookiep)) <= toff)) { +#else + while (dp->d_fileno == 0 && cpos < cend && ncookies > 0) { +#endif cpos += dp->d_reclen; dp = (struct dirent *)cpos; cookiep++; ncookies--; } - if (cpos >= cend) { + if (cpos >= cend || ncookies == 0) { toff = off; siz = fullsiz; - FREE(cookies, M_TEMP); goto again; } - len = 3*NFSX_UNSIGNED; /* paranoia, probably can be 0 */ - nfsm_reply(siz); + len = 3 * NFSX_UNSIGNED; /* paranoia, probably can be 0 */ + nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_COOKIEVERF(v3) + siz); + if (v3) { + nfsm_srvpostop_attr(getret, &at); + nfsm_build(tl, u_long *, 2 * NFSX_UNSIGNED); + txdr_hyper(&at.va_filerev, tl); + } mp = mp2 = mb; bp = bpos; be = bp + M_TRAILINGSPACE(mp); /* Loop through the records and build reply */ - while (cpos < cend) { + while (cpos < cend && ncookies > 0) { if (dp->d_fileno != 0) { nlen = dp->d_namlen; rem = nfsm_rndup(nlen)-nlen; - len += (4*NFSX_UNSIGNED + nlen + rem); + len += (4 * NFSX_UNSIGNED + nlen + rem); + if (v3) + len += 2 * NFSX_UNSIGNED; if (len > cnt) { eofflag = 0; break; @@ -1563,6 +2657,11 @@ again: nfsm_clget; *tl = nfs_true; bp += NFSX_UNSIGNED; + if (v3) { + nfsm_clget; + *tl = 0; + bp += NFSX_UNSIGNED; + } nfsm_clget; *tl = txdr_unsigned(dp->d_fileno); bp += NFSX_UNSIGNED; @@ -1591,12 +2690,18 @@ again: nfsm_clget; /* Finish off the record */ + if (v3) { + *tl = 0; + bp += NFSX_UNSIGNED; + nfsm_clget; + } *tl = txdr_unsigned(*cookiep); bp += NFSX_UNSIGNED; } cpos += dp->d_reclen; dp = (struct dirent *)cpos; cookiep++; + ncookies--; } nfsrv_vrele(vp); nfsm_clget; @@ -1613,19 +2718,22 @@ again: mp->m_len = bp - mtod(mp, caddr_t); } else mp->m_len += bp - bpos; - FREE(rbuf, M_TEMP); - FREE(cookies, M_TEMP); + FREE((caddr_t)rbuf, M_TEMP); + FREE((caddr_t)cookies, M_TEMP); nfsm_srvdone; } int -nqnfsrv_readdirlook(nfsd, mrep, md, dpos, cred, nam, mrq) - struct nfsd *nfsd; - struct mbuf *mrep, *md; - caddr_t dpos; - struct ucred *cred; - struct mbuf *nam, **mrq; +nfsrv_readdirplus(nfsd, slp, procp, mrq) + struct nfsrv_descript *nfsd; + struct nfssvc_sock *slp; + struct proc *procp; + struct mbuf **mrq; { + struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; + struct mbuf *nam = nfsd->nd_nam; + caddr_t dpos = nfsd->nd_dpos; + struct ucred *cred = &nfsd->nd_cr; register char *bp, *be; register struct mbuf *mp; register struct dirent *dp; @@ -1637,42 +2745,58 @@ nqnfsrv_readdirlook(nfsd, mrep, md, dpos, cred, nam, mrq) char *cpos, *cend, *cp2, *rbuf; struct vnode *vp, *nvp; struct flrep fl; - nfsv2fh_t nfh; - fhandle_t *fhp; + nfsfh_t nfh; + fhandle_t *fhp, *nfhp = (fhandle_t *)fl.fl_nfh; struct uio io; struct iovec iv; - struct vattr va, *vap = &va; - struct nfsv2_fattr *fp; - int len, nlen, rem, xfer, tsiz, i, error = 0, cache2; - int siz, cnt, fullsiz, eofflag, rdonly, cache; - u_long duration2; - u_quad_t frev, frev2; - u_long off, toff; - int ncookies; - u_int *cookies; - u_int *cookiep; + struct vattr va, at, *vap = &va; + struct nfs_fattr *fp; + int len, nlen, rem, xfer, tsiz, i, error = 0, getret = 1; + int siz, cnt, fullsiz, eofflag, rdonly, cache, dirlen, ncookies; + u_quad_t frev, off, toff, verf; + u_long *cookies = NULL, *cookiep; fhp = &nfh.fh_generic; nfsm_srvmtofh(fhp); - nfsm_dissect(tl, u_long *, 3*NFSX_UNSIGNED); - toff = off = fxdr_unsigned(u_long, *tl++); - cnt = fxdr_unsigned(int, *tl++); - duration2 = fxdr_unsigned(int, *tl); - siz = ((cnt+NFS_DIRBLKSIZ-1) & ~(NFS_DIRBLKSIZ-1)); - if (cnt > NFS_MAXREADDIR) - siz = NFS_MAXREADDIR; + nfsm_dissect(tl, u_long *, 6 * NFSX_UNSIGNED); + fxdr_hyper(tl, &toff); + tl += 2; + fxdr_hyper(tl, &verf); + tl += 2; + siz = fxdr_unsigned(int, *tl++); + cnt = fxdr_unsigned(int, *tl); + off = toff; + siz = ((siz + DIRBLKSIZ - 1) & ~(DIRBLKSIZ - 1)); + xfer = NFS_SRVMAXDATA(nfsd); + if (siz > xfer) + siz = xfer; fullsiz = siz; - error = nfsrv_fhtovp(fhp, TRUE, &vp, cred, nfsd->nd_slp, nam, &rdonly); - if (error) - nfsm_reply(0); - nqsrv_getl(vp, NQL_READ); - error = nfsrv_access(vp, VEXEC, cred, rdonly, nfsd->nd_procp); + if (error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam, + &rdonly, (nfsd->nd_flag & ND_KERBAUTH))) { + nfsm_reply(NFSX_UNSIGNED); + nfsm_srvpostop_attr(getret, &at); + return (0); + } + error = getret = VOP_GETATTR(vp, &at, cred, procp); + if (!error && toff && verf != at.va_filerev) + error = NFSERR_BAD_COOKIE; + if (!error) { + nqsrv_getl(vp, ND_READ); + error = nfsrv_access(vp, VEXEC, cred, rdonly, procp); + } if (error) { nfsrv_vput(vp); - nfsm_reply(0); + nfsm_reply(NFSX_V3POSTOPATTR); + nfsm_srvpostop_attr(getret, &at); + return (0); } VOP_UNLOCK(vp); MALLOC(rbuf, caddr_t, siz, M_TEMP, M_WAITOK); +#ifdef __NetBSD__ + ncookies = siz / (7 * NFSX_UNSIGNED); + MALLOC(cookies, u_long *, ncookies * sizeof (u_long *), M_TEMP, + M_WAITOK); +#endif again: iv.iov_base = rbuf; iv.iov_len = fullsiz; @@ -1683,25 +2807,29 @@ again: io.uio_segflg = UIO_SYSSPACE; io.uio_rw = UIO_READ; io.uio_procp = (struct proc *)0; - cookies = NULL; eofflag = 0; +#ifndef __NetBSD__ + if (cookies) { + free((caddr_t)cookies, M_TEMP); + cookies = NULL; + } error = VOP_READDIR(vp, &io, cred, &eofflag, &ncookies, &cookies); - off = (u_long)io.uio_offset; +#else + error = VOP_READDIR(vp, &io, cred, &eofflag, cookies, ncookies); +#endif + off = (u_quad_t)io.uio_offset; + getret = VOP_GETATTR(vp, &at, cred, procp); + if (!cookies && !error) + error = NFSERR_PERM; + if (!error) + error = getret; if (error) { nfsrv_vrele(vp); + if (cookies) + free((caddr_t)cookies, M_TEMP); free((caddr_t)rbuf, M_TEMP); - nfsm_reply(0); - } - if (cookies == NULL) { - /* - * If the filesystem doen't support cookies, return eof. - */ - nfsrv_vrele(vp); - nfsm_reply(2*NFSX_UNSIGNED); - nfsm_build(tl, u_long *, 2*NFSX_UNSIGNED); - *tl++ = nfs_false; - *tl = nfs_true; - FREE((caddr_t)rbuf, M_TEMP); + nfsm_reply(NFSX_V3POSTOPATTR); + nfsm_srvpostop_attr(getret, &at); return (0); } if (io.uio_resid) { @@ -1713,12 +2841,16 @@ again: */ if (siz == 0) { nfsrv_vrele(vp); - nfsm_reply(2 * NFSX_UNSIGNED); - nfsm_build(tl, u_long *, 2 * NFSX_UNSIGNED); + nfsm_reply(NFSX_V3POSTOPATTR + NFSX_V3COOKIEVERF + + 2 * NFSX_UNSIGNED); + nfsm_srvpostop_attr(getret, &at); + nfsm_build(tl, u_long *, 4 * NFSX_UNSIGNED); + txdr_hyper(&at.va_filerev, tl); + tl += 2; *tl++ = nfs_false; *tl = nfs_true; + FREE((caddr_t)cookies, M_TEMP); FREE((caddr_t)rbuf, M_TEMP); - FREE(cookies, M_TEMP); return (0); } } @@ -1731,28 +2863,41 @@ again: cend = rbuf + siz; dp = (struct dirent *)cpos; cookiep = cookies; - while (cpos < cend && ncookies > 0 - && (*cookiep <= toff || dp->d_fileno == 0)) { +#ifdef __FreeBSD__ + /* + * For some reason FreeBSD's ufs_readdir() chooses to back the + * directory offset up to a block boundary, so it is necessary to + * skip over the records that preceed the requested offset. This + * requires the assumption that file offset cookies monotonically + * increase. + */ + while (cpos < cend && ncookies > 0 && + (dp->d_fileno == 0 || ((u_quad_t)(*cookiep)) <= toff)) { +#else + while (dp->d_fileno == 0 && cpos < cend && ncookies > 0) { +#endif cpos += dp->d_reclen; dp = (struct dirent *)cpos; cookiep++; ncookies--; } - if (cpos >= cend) { + if (cpos >= cend || ncookies == 0) { toff = off; siz = fullsiz; - FREE(cookies, M_TEMP); goto again; } - len = 3 * NFSX_UNSIGNED; /* paranoia, probably can be 0 */ - nfsm_reply(siz); + dirlen = len = NFSX_V3POSTOPATTR + NFSX_V3COOKIEVERF + 2 * NFSX_UNSIGNED; + nfsm_reply(cnt); + nfsm_srvpostop_attr(getret, &at); + nfsm_build(tl, u_long *, 2 * NFSX_UNSIGNED); + txdr_hyper(&at.va_filerev, tl); mp = mp2 = mb; bp = bpos; be = bp + M_TRAILINGSPACE(mp); /* Loop through the records and build reply */ - while (cpos < cend) { + while (cpos < cend && ncookies > 0) { if (dp->d_fileno != 0) { nlen = dp->d_namlen; rem = nfsm_rndup(nlen)-nlen; @@ -1763,59 +2908,51 @@ again: */ if (VFS_VGET(vp->v_mount, dp->d_fileno, &nvp)) goto invalid; - bzero((caddr_t)&fl.fl_nfh, sizeof (nfsv2fh_t)); - fl.fl_nfh.fh_generic.fh_fsid = + bzero((caddr_t)nfhp, NFSX_V3FH); + nfhp->fh_fsid = nvp->v_mount->mnt_stat.f_fsid; - if (VFS_VPTOFH(nvp, &fl.fl_nfh.fh_generic.fh_fid)) { + if (VFS_VPTOFH(nvp, &nfhp->fh_fid)) { vput(nvp); goto invalid; } - if (duration2) { - (void) nqsrv_getlease(nvp, &duration2, NQL_READ, - nfsd, nam, &cache2, &frev2, cred); - fl.fl_duration = txdr_unsigned(duration2); - fl.fl_cachable = txdr_unsigned(cache2); - txdr_hyper(&frev2, fl.fl_frev); - } else - fl.fl_duration = 0; - if (VOP_GETATTR(nvp, vap, cred, nfsd->nd_procp)) { + if (VOP_GETATTR(nvp, vap, cred, procp)) { vput(nvp); goto invalid; } vput(nvp); - fp = (struct nfsv2_fattr *)&fl.fl_fattr; - nfsm_srvfillattr; - len += (4*NFSX_UNSIGNED + nlen + rem + NFSX_FH - + NFSX_NQFATTR); - if (len > cnt) { + + /* + * If either the dircount or maxcount will be + * exceeded, get out now. Both of these lengths + * are calculated conservatively, including all + * XDR overheads. + */ + len += (7 * NFSX_UNSIGNED + nlen + rem + NFSX_V3FH + + NFSX_V3POSTOPATTR); + dirlen += (6 * NFSX_UNSIGNED + nlen + rem); + if (len > cnt || dirlen > fullsiz) { eofflag = 0; break; } + /* * Build the directory record xdr from * the dirent entry. */ + fp = (struct nfs_fattr *)&fl.fl_fattr; + nfsm_srvfillattr(vap, fp); + fl.fl_fhsize = txdr_unsigned(NFSX_V3FH); + fl.fl_fhok = nfs_true; + fl.fl_postopok = nfs_true; + fl.fl_off.nfsuquad[0] = 0; + fl.fl_off.nfsuquad[1] = txdr_unsigned(*cookiep); + nfsm_clget; *tl = nfs_true; bp += NFSX_UNSIGNED; - - /* - * For readdir_and_lookup copy the stuff out. - */ - xfer = sizeof (struct flrep); - cp = (caddr_t)&fl; - while (xfer > 0) { - nfsm_clget; - if ((bp+xfer) > be) - tsiz = be-bp; - else - tsiz = xfer; - bcopy(cp, bp, tsiz); - bp += tsiz; - xfer -= tsiz; - if (xfer > 0) - cp += tsiz; - } + nfsm_clget; + *tl = 0; + bp += NFSX_UNSIGNED; nfsm_clget; *tl = txdr_unsigned(dp->d_fileno); bp += NFSX_UNSIGNED; @@ -1828,8 +2965,8 @@ again: cp = dp->d_name; while (xfer > 0) { nfsm_clget; - if ((bp+xfer) > be) - tsiz = be-bp; + if ((bp + xfer) > be) + tsiz = be - bp; else tsiz = xfer; bcopy(cp, bp, tsiz); @@ -1841,16 +2978,30 @@ again: /* And null pad to a long boundary */ for (i = 0; i < rem; i++) *bp++ = '\0'; - nfsm_clget; - - /* Finish off the record */ - *tl = txdr_unsigned(*cookiep); - bp += NFSX_UNSIGNED; + + /* + * Now copy the flrep structure out. + */ + xfer = sizeof (struct flrep); + cp = (caddr_t)&fl; + while (xfer > 0) { + nfsm_clget; + if ((bp + xfer) > be) + tsiz = be - bp; + else + tsiz = xfer; + bcopy(cp, bp, tsiz); + bp += tsiz; + xfer -= tsiz; + if (xfer > 0) + cp += tsiz; + } } invalid: cpos += dp->d_reclen; dp = (struct dirent *)cpos; cookiep++; + ncookies--; } nfsrv_vrele(vp); nfsm_clget; @@ -1867,8 +3018,69 @@ invalid: mp->m_len = bp - mtod(mp, caddr_t); } else mp->m_len += bp - bpos; - FREE(rbuf, M_TEMP); - FREE(cookies, M_TEMP); + FREE((caddr_t)cookies, M_TEMP); + FREE((caddr_t)rbuf, M_TEMP); + nfsm_srvdone; +} + +/* + * nfs commit service + */ +int +nfsrv_commit(nfsd, slp, procp, mrq) + struct nfsrv_descript *nfsd; + struct nfssvc_sock *slp; + struct proc *procp; + struct mbuf **mrq; +{ + struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; + struct mbuf *nam = nfsd->nd_nam; + caddr_t dpos = nfsd->nd_dpos; + struct ucred *cred = &nfsd->nd_cr; + struct vattr bfor, aft; + struct vnode *vp; + nfsfh_t nfh; + fhandle_t *fhp; + register u_long *tl; + register long t1; + caddr_t bpos; + int error = 0, rdonly, for_ret = 1, aft_ret = 1, cnt, cache; + char *cp2; + struct mbuf *mb, *mb2, *mreq; + u_quad_t frev, off; + +#ifndef nolint + cache = 0; +#endif + fhp = &nfh.fh_generic; + nfsm_srvmtofh(fhp); + nfsm_dissect(tl, u_long *, 3 * NFSX_UNSIGNED); + + /* + * XXX At this time VOP_FSYNC() does not accept offset and byte + * count parameters, so these arguments are useless (someday maybe). + */ + fxdr_hyper(tl, &off); + tl += 2; + cnt = fxdr_unsigned(int, *tl); + if (error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam, + &rdonly, (nfsd->nd_flag & ND_KERBAUTH))) { + nfsm_reply(2 * NFSX_UNSIGNED); + nfsm_srvwcc_data(for_ret, &bfor, aft_ret, &aft); + return (0); + } + for_ret = VOP_GETATTR(vp, &bfor, cred, procp); + error = VOP_FSYNC(vp, cred, MNT_WAIT, procp); + aft_ret = VOP_GETATTR(vp, &aft, cred, procp); + nfsrv_vput(vp); + nfsm_reply(NFSX_V3WCCDATA + NFSX_V3WRITEVERF); + nfsm_srvwcc_data(for_ret, &bfor, aft_ret, &aft); + if (!error) { + nfsm_build(tl, u_long *, NFSX_V3WRITEVERF); + *tl++ = txdr_unsigned(boottime.tv_sec); + *tl = txdr_unsigned(boottime.tv_usec); + } else + return (0); nfsm_srvdone; } @@ -1876,69 +3088,244 @@ invalid: * nfs statfs service */ int -nfsrv_statfs(nfsd, mrep, md, dpos, cred, nam, mrq) - struct nfsd *nfsd; - struct mbuf *mrep, *md; - caddr_t dpos; - struct ucred *cred; - struct mbuf *nam, **mrq; +nfsrv_statfs(nfsd, slp, procp, mrq) + struct nfsrv_descript *nfsd; + struct nfssvc_sock *slp; + struct proc *procp; + struct mbuf **mrq; { + struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; + struct mbuf *nam = nfsd->nd_nam; + caddr_t dpos = nfsd->nd_dpos; + struct ucred *cred = &nfsd->nd_cr; register struct statfs *sf; - register struct nfsv2_statfs *sfp; + register struct nfs_statfs *sfp; register u_long *tl; register long t1; caddr_t bpos; - int error = 0, rdonly, cache = 0, isnq; + int error = 0, rdonly, cache, getret = 1; + int v3 = (nfsd->nd_flag & ND_NFSV3); char *cp2; struct mbuf *mb, *mb2, *mreq; struct vnode *vp; - nfsv2fh_t nfh; + struct vattr at; + nfsfh_t nfh; fhandle_t *fhp; struct statfs statfs; - u_quad_t frev; + u_quad_t frev, tval; +#ifndef nolint + cache = 0; +#endif fhp = &nfh.fh_generic; - isnq = (nfsd->nd_nqlflag != NQL_NOVAL); nfsm_srvmtofh(fhp); - error = nfsrv_fhtovp(fhp, TRUE, &vp, cred, nfsd->nd_slp, nam, &rdonly); - if (error) - nfsm_reply(0); + if (error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam, + &rdonly, (nfsd->nd_flag & ND_KERBAUTH))) { + nfsm_reply(NFSX_UNSIGNED); + nfsm_srvpostop_attr(getret, &at); + return (0); + } sf = &statfs; - error = VFS_STATFS(vp->v_mount, sf, nfsd->nd_procp); + error = VFS_STATFS(vp->v_mount, sf, procp); + getret = VOP_GETATTR(vp, &at, cred, procp); nfsrv_vput(vp); - nfsm_reply(NFSX_STATFS(isnq)); - nfsm_build(sfp, struct nfsv2_statfs *, NFSX_STATFS(isnq)); - sfp->sf_tsize = txdr_unsigned(NFS_MAXDGRAMDATA); - sfp->sf_bsize = txdr_unsigned(sf->f_bsize); - sfp->sf_blocks = txdr_unsigned(sf->f_blocks); - sfp->sf_bfree = txdr_unsigned(sf->f_bfree); - sfp->sf_bavail = txdr_unsigned(sf->f_bavail); - if (isnq) { - sfp->sf_files = txdr_unsigned(sf->f_files); - sfp->sf_ffree = txdr_unsigned(sf->f_ffree); + nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_STATFS(v3)); + if (v3) + nfsm_srvpostop_attr(getret, &at); + if (error) + return (0); + nfsm_build(sfp, struct nfs_statfs *, NFSX_STATFS(v3)); + if (v3) { + tval = (u_quad_t)sf->f_blocks; + tval *= (u_quad_t)sf->f_bsize; + txdr_hyper(&tval, &sfp->sf_tbytes); + tval = (u_quad_t)sf->f_bfree; + tval *= (u_quad_t)sf->f_bsize; + txdr_hyper(&tval, &sfp->sf_fbytes); + tval = (u_quad_t)sf->f_bavail; + tval *= (u_quad_t)sf->f_bsize; + txdr_hyper(&tval, &sfp->sf_abytes); + sfp->sf_tfiles.nfsuquad[0] = 0; + sfp->sf_tfiles.nfsuquad[1] = txdr_unsigned(sf->f_files); + sfp->sf_ffiles.nfsuquad[0] = 0; + sfp->sf_ffiles.nfsuquad[1] = txdr_unsigned(sf->f_ffree); + sfp->sf_afiles.nfsuquad[0] = 0; + sfp->sf_afiles.nfsuquad[1] = txdr_unsigned(sf->f_ffree); + sfp->sf_invarsec = 0; + } else { + sfp->sf_tsize = txdr_unsigned(NFS_MAXDGRAMDATA); + sfp->sf_bsize = txdr_unsigned(sf->f_bsize); + sfp->sf_blocks = txdr_unsigned(sf->f_blocks); + sfp->sf_bfree = txdr_unsigned(sf->f_bfree); + sfp->sf_bavail = txdr_unsigned(sf->f_bavail); } nfsm_srvdone; } /* + * nfs fsinfo service + */ +int +nfsrv_fsinfo(nfsd, slp, procp, mrq) + struct nfsrv_descript *nfsd; + struct nfssvc_sock *slp; + struct proc *procp; + struct mbuf **mrq; +{ + struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; + struct mbuf *nam = nfsd->nd_nam; + caddr_t dpos = nfsd->nd_dpos; + struct ucred *cred = &nfsd->nd_cr; + register u_long *tl; + register struct nfsv3_fsinfo *sip; + register long t1; + caddr_t bpos; + int error = 0, rdonly, cache, getret = 1, pref; + char *cp2; + struct mbuf *mb, *mb2, *mreq; + struct vnode *vp; + struct vattr at; + nfsfh_t nfh; + fhandle_t *fhp; + u_quad_t frev; + +#ifndef nolint + cache = 0; +#endif + fhp = &nfh.fh_generic; + nfsm_srvmtofh(fhp); + if (error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam, + &rdonly, (nfsd->nd_flag & ND_KERBAUTH))) { + nfsm_reply(NFSX_UNSIGNED); + nfsm_srvpostop_attr(getret, &at); + return (0); + } + getret = VOP_GETATTR(vp, &at, cred, procp); + nfsrv_vput(vp); + nfsm_reply(NFSX_V3POSTOPATTR + NFSX_V3FSINFO); + nfsm_srvpostop_attr(getret, &at); + nfsm_build(sip, struct nfsv3_fsinfo *, NFSX_V3FSINFO); + + /* + * XXX + * There should be file system VFS OP(s) to get this information. + * For now, assume ufs. + */ + if (slp->ns_so->so_type == SOCK_DGRAM) + pref = NFS_MAXDGRAMDATA; + else + pref = NFS_MAXDATA; + sip->fs_rtmax = txdr_unsigned(NFS_MAXDATA); + sip->fs_rtpref = txdr_unsigned(pref); + sip->fs_rtmult = txdr_unsigned(NFS_FABLKSIZE); + sip->fs_wtmax = txdr_unsigned(NFS_MAXDATA); + sip->fs_wtpref = txdr_unsigned(pref); + sip->fs_wtmult = txdr_unsigned(NFS_FABLKSIZE); + sip->fs_dtpref = txdr_unsigned(pref); + sip->fs_maxfilesize.nfsuquad[0] = 0xffffffff; + sip->fs_maxfilesize.nfsuquad[1] = 0xffffffff; + sip->fs_timedelta.nfsv3_sec = 0; + sip->fs_timedelta.nfsv3_nsec = txdr_unsigned(1); + sip->fs_properties = txdr_unsigned(NFSV3FSINFO_LINK | + NFSV3FSINFO_SYMLINK | NFSV3FSINFO_HOMOGENEOUS | + NFSV3FSINFO_CANSETTIME); + nfsm_srvdone; +} + +/* + * nfs pathconf service + */ +int +nfsrv_pathconf(nfsd, slp, procp, mrq) + struct nfsrv_descript *nfsd; + struct nfssvc_sock *slp; + struct proc *procp; + struct mbuf **mrq; +{ + struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; + struct mbuf *nam = nfsd->nd_nam; + caddr_t dpos = nfsd->nd_dpos; + struct ucred *cred = &nfsd->nd_cr; + register u_long *tl; + register struct nfsv3_pathconf *pc; + register long t1; + caddr_t bpos; + int error = 0, rdonly, cache, getret = 1, linkmax, namemax; + int chownres, notrunc; + char *cp2; + struct mbuf *mb, *mb2, *mreq; + struct vnode *vp; + struct vattr at; + nfsfh_t nfh; + fhandle_t *fhp; + u_quad_t frev; + +#ifndef nolint + cache = 0; +#endif + fhp = &nfh.fh_generic; + nfsm_srvmtofh(fhp); + if (error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam, + &rdonly, (nfsd->nd_flag & ND_KERBAUTH))) { + nfsm_reply(NFSX_UNSIGNED); + nfsm_srvpostop_attr(getret, &at); + return (0); + } + error = VOP_PATHCONF(vp, _PC_LINK_MAX, &linkmax); + if (!error) + error = VOP_PATHCONF(vp, _PC_NAME_MAX, &namemax); + if (!error) + error = VOP_PATHCONF(vp, _PC_CHOWN_RESTRICTED, &chownres); + if (!error) + error = VOP_PATHCONF(vp, _PC_NO_TRUNC, ¬runc); + getret = VOP_GETATTR(vp, &at, cred, procp); + nfsrv_vput(vp); + nfsm_reply(NFSX_V3POSTOPATTR + NFSX_V3PATHCONF); + nfsm_srvpostop_attr(getret, &at); + if (error) + return (0); + nfsm_build(pc, struct nfsv3_pathconf *, NFSX_V3PATHCONF); + + pc->pc_linkmax = txdr_unsigned(linkmax); + pc->pc_namemax = txdr_unsigned(namemax); + pc->pc_notrunc = txdr_unsigned(notrunc); + pc->pc_chownrestricted = txdr_unsigned(chownres); + + /* + * These should probably be supported by VOP_PATHCONF(), but + * until msdosfs is exportable (why would you want to?), the + * Unix defaults should be ok. + */ + pc->pc_caseinsensitive = nfs_false; + pc->pc_casepreserving = nfs_true; + nfsm_srvdone; +} + +/* * Null operation, used by clients to ping server */ /* ARGSUSED */ int -nfsrv_null(nfsd, mrep, md, dpos, cred, nam, mrq) - struct nfsd *nfsd; - struct mbuf *mrep, *md; - caddr_t dpos; - struct ucred *cred; - struct mbuf *nam, **mrq; +nfsrv_null(nfsd, slp, procp, mrq) + struct nfsrv_descript *nfsd; + struct nfssvc_sock *slp; + struct proc *procp; + struct mbuf **mrq; { + struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; + struct mbuf *nam = nfsd->nd_nam; + caddr_t dpos = nfsd->nd_dpos; + struct ucred *cred = &nfsd->nd_cr; caddr_t bpos; - int error = VNOVAL, cache = 0; + int error = NFSERR_RETVOID, cache; struct mbuf *mb, *mreq; u_quad_t frev; +#ifndef nolint + cache = 0; +#endif nfsm_reply(0); - return (error); + return (0); } /* @@ -1946,24 +3333,30 @@ nfsrv_null(nfsd, mrep, md, dpos, cred, nam, mrq) */ /* ARGSUSED */ int -nfsrv_noop(nfsd, mrep, md, dpos, cred, nam, mrq) - struct nfsd *nfsd; - struct mbuf *mrep, *md; - caddr_t dpos; - struct ucred *cred; - struct mbuf *nam, **mrq; +nfsrv_noop(nfsd, slp, procp, mrq) + struct nfsrv_descript *nfsd; + struct nfssvc_sock *slp; + struct proc *procp; + struct mbuf **mrq; { + struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md; + struct mbuf *nam = nfsd->nd_nam; + caddr_t dpos = nfsd->nd_dpos; + struct ucred *cred = &nfsd->nd_cr; caddr_t bpos; - int error, cache = 0; + int error, cache; struct mbuf *mb, *mreq; u_quad_t frev; +#ifndef nolint + cache = 0; +#endif if (nfsd->nd_repstat) error = nfsd->nd_repstat; else error = EPROCUNAVAIL; nfsm_reply(0); - return (error); + return (0); } /* @@ -1997,8 +3390,6 @@ nfsrv_access(vp, flags, cred, rdonly, p) switch (vp->v_type) { case VREG: case VDIR: case VLNK: return (EROFS); - default: - break; } } /* @@ -2008,8 +3399,7 @@ nfsrv_access(vp, flags, cred, rdonly, p) if (vp->v_flag & VTEXT) return (ETXTBSY); } - error = VOP_GETATTR(vp, &vattr, cred, p); - if (error) + if (error = VOP_GETATTR(vp, &vattr, cred, p)) return (error); if ((error = VOP_ACCESS(vp, flags, cred, p)) && cred->cr_uid != vattr.va_uid) diff --git a/sys/nfsserver/nfs_srvcache.c b/sys/nfsserver/nfs_srvcache.c index 0f31ae0..acdce0d 100644 --- a/sys/nfsserver/nfs_srvcache.c +++ b/sys/nfsserver/nfs_srvcache.c @@ -34,7 +34,7 @@ * SUCH DAMAGE. * * @(#)nfs_srvcache.c 8.1 (Berkeley) 6/10/93 - * $Id: nfs_srvcache.c,v 1.4 1994/10/02 17:27:00 phk Exp $ + * $Id: nfs_srvcache.c,v 1.5 1994/10/17 17:47:36 phk Exp $ */ /* @@ -59,11 +59,13 @@ #endif #include <nfs/nfsm_subs.h> #include <nfs/rpcv2.h> -#include <nfs/nfsv2.h> +#include <nfs/nfsproto.h> #include <nfs/nfs.h> #include <nfs/nfsrvcache.h> #include <nfs/nqnfs.h> +extern struct nfsstats nfsstats; +extern int nfsv2_procid[NFS_NPROCS]; long numnfsrvcache, desirednfsrvcache = NFSRVCACHESIZ; #define NFSRCHASH(xid) \ @@ -89,7 +91,6 @@ int nonidempotent[NFS_NPROCS] = { FALSE, FALSE, FALSE, - FALSE, TRUE, TRUE, TRUE, @@ -98,6 +99,10 @@ int nonidempotent[NFS_NPROCS] = { TRUE, TRUE, TRUE, + TRUE, + FALSE, + FALSE, + FALSE, FALSE, FALSE, FALSE, @@ -108,7 +113,7 @@ int nonidempotent[NFS_NPROCS] = { }; /* True iff the rpc reply is an nfs status ONLY! */ -static int repliesstatus[NFS_NPROCS] = { +static int nfsv2_repstat[NFS_NPROCS] = { FALSE, FALSE, FALSE, @@ -127,11 +132,6 @@ static int repliesstatus[NFS_NPROCS] = { TRUE, FALSE, FALSE, - FALSE, - FALSE, - FALSE, - FALSE, - TRUE, }; /* @@ -160,9 +160,9 @@ nfsrv_initcache() * Update/add new request at end of lru list */ int -nfsrv_getcache(nam, nd, repp) - struct mbuf *nam; - register struct nfsd *nd; +nfsrv_getcache(nd, slp, repp) + register struct nfsrv_descript *nd; + struct nfssvc_sock *slp; struct mbuf **repp; { register struct nfsrvcache *rp; @@ -171,13 +171,17 @@ nfsrv_getcache(nam, nd, repp) caddr_t bpos; int ret; - if (nd->nd_nqlflag != NQL_NOVAL) + /* + * Don't cache recent requests for reliable transport protocols. + * (Maybe we should for the case of a reconnect, but..) + */ + if (!nd->nd_nam2) return (RC_DOIT); loop: for (rp = NFSRCHASH(nd->nd_retxid)->lh_first; rp != 0; rp = rp->rc_hash.le_next) { if (nd->nd_retxid == rp->rc_xid && nd->nd_procnum == rp->rc_proc && - netaddr_match(NETFAMILY(rp), &rp->rc_haddr, nam)) { + netaddr_match(NETFAMILY(rp), &rp->rc_haddr, nd->nd_nam)) { if ((rp->rc_flag & RC_LOCKED) != 0) { rp->rc_flag |= RC_WANTED; (void) tsleep((caddr_t)rp, PZERO-1, "nfsrc", 0); @@ -196,7 +200,7 @@ loop: ret = RC_DROPIT; } else if (rp->rc_flag & RC_REPSTATUS) { nfsstats.srvcache_nonidemdonehits++; - nfs_rephead(0, nd, rp->rc_status, + nfs_rephead(0, nd, slp, rp->rc_status, 0, (u_quad_t *)0, repp, &mb, &bpos); ret = RC_REPLY; } else if (rp->rc_flag & RC_REPMBUF) { @@ -243,7 +247,7 @@ loop: TAILQ_INSERT_TAIL(&nfsrvlruhead, rp, rc_lru); rp->rc_state = RC_INPROG; rp->rc_xid = nd->nd_retxid; - saddr = mtod(nam, struct sockaddr_in *); + saddr = mtod(nd->nd_nam, struct sockaddr_in *); switch (saddr->sin_family) { case AF_INET: rp->rc_flag |= RC_INETADDR; @@ -252,7 +256,7 @@ loop: case AF_ISO: default: rp->rc_flag |= RC_NAM; - rp->rc_nam = m_copym(nam, 0, M_COPYALL, M_WAIT); + rp->rc_nam = m_copym(nd->nd_nam, 0, M_COPYALL, M_WAIT); break; }; rp->rc_proc = nd->nd_procnum; @@ -269,21 +273,20 @@ loop: * Update a request cache entry after the rpc has been done */ void -nfsrv_updatecache(nam, nd, repvalid, repmbuf) - struct mbuf *nam; - register struct nfsd *nd; +nfsrv_updatecache(nd, repvalid, repmbuf) + register struct nfsrv_descript *nd; int repvalid; struct mbuf *repmbuf; { register struct nfsrvcache *rp; - if (nd->nd_nqlflag != NQL_NOVAL) + if (!nd->nd_nam2) return; loop: for (rp = NFSRCHASH(nd->nd_retxid)->lh_first; rp != 0; rp = rp->rc_hash.le_next) { if (nd->nd_retxid == rp->rc_xid && nd->nd_procnum == rp->rc_proc && - netaddr_match(NETFAMILY(rp), &rp->rc_haddr, nam)) { + netaddr_match(NETFAMILY(rp), &rp->rc_haddr, nd->nd_nam)) { if ((rp->rc_flag & RC_LOCKED) != 0) { rp->rc_flag |= RC_WANTED; (void) tsleep((caddr_t)rp, PZERO-1, "nfsrc", 0); @@ -296,7 +299,8 @@ loop: * the reply for non-idempotent rpc's. */ if (repvalid && nonidempotent[nd->nd_procnum]) { - if (repliesstatus[nd->nd_procnum]) { + if ((nd->nd_flag & ND_NFSV3) == 0 && + nfsv2_repstat[nfsv2_procid[nd->nd_procnum]]) { rp->rc_status = nd->nd_repstat; rp->rc_flag |= RC_REPSTATUS; } else { diff --git a/sys/nfsserver/nfs_srvsock.c b/sys/nfsserver/nfs_srvsock.c index 20253ef..8d15b24 100644 --- a/sys/nfsserver/nfs_srvsock.c +++ b/sys/nfsserver/nfs_srvsock.c @@ -34,7 +34,7 @@ * SUCH DAMAGE. * * @(#)nfs_socket.c 8.3 (Berkeley) 1/12/94 - * $Id: nfs_socket.c,v 1.6 1995/03/16 18:15:37 bde Exp $ + * $Id: nfs_socket.c,v 1.7 1995/05/30 08:12:40 rgrimes Exp $ */ /* @@ -59,7 +59,7 @@ #include <netinet/tcp.h> #include <nfs/rpcv2.h> -#include <nfs/nfsv2.h> +#include <nfs/nfsproto.h> #include <nfs/nfs.h> #include <nfs/xdr_subs.h> #include <nfs/nfsm_subs.h> @@ -94,36 +94,13 @@ * External data, mostly RPC constants in XDR form */ extern u_long rpc_reply, rpc_msgdenied, rpc_mismatch, rpc_vers, rpc_auth_unix, - rpc_msgaccepted, rpc_call, rpc_autherr, rpc_rejectedcred, + rpc_msgaccepted, rpc_call, rpc_autherr, rpc_auth_kerb; -extern u_long nfs_prog, nfs_vers, nqnfs_prog, nqnfs_vers; +extern u_long nfs_prog, nqnfs_prog; extern time_t nqnfsstarttime; -extern int nonidempotent[NFS_NPROCS]; - -/* - * Maps errno values to nfs error numbers. - * Use NFSERR_IO as the catch all for ones not specifically defined in - * RFC 1094. - */ -static int nfsrv_errmap[ELAST] = { - NFSERR_PERM, NFSERR_NOENT, NFSERR_IO, NFSERR_IO, NFSERR_IO, - NFSERR_NXIO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, - NFSERR_IO, NFSERR_IO, NFSERR_ACCES, NFSERR_IO, NFSERR_IO, - NFSERR_IO, NFSERR_EXIST, NFSERR_IO, NFSERR_NODEV, NFSERR_NOTDIR, - NFSERR_ISDIR, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, - NFSERR_IO, NFSERR_FBIG, NFSERR_NOSPC, NFSERR_IO, NFSERR_ROFS, - NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, - NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, - NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, - NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, - NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, - NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, - NFSERR_IO, NFSERR_IO, NFSERR_NAMETOL, NFSERR_IO, NFSERR_IO, - NFSERR_NOTEMPTY, NFSERR_IO, NFSERR_IO, NFSERR_DQUOT, NFSERR_STALE, - NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, - NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, - NFSERR_IO, -}; +extern struct nfsstats nfsstats; +extern int nfsv3_procid[NFS_NPROCS]; +extern int nfs_ticks; /* * Defines which timer to use for the procnum. @@ -134,7 +111,8 @@ static int nfsrv_errmap[ELAST] = { * 4 - write */ static int proct[NFS_NPROCS] = { - 0, 1, 0, 0, 2, 3, 3, 0, 4, 0, 0, 0, 0, 0, 0, 0, 3, 0, 3, 0, 0, 0, 0, + 0, 1, 0, 2, 1, 3, 3, 4, 0, 0, 0, 0, 0, 0, 0, 0, 3, 3, 0, 0, 0, 0, 0, + 0, 0, 0, }; /* @@ -164,10 +142,12 @@ int nfsrv_null(), nfsrv_getattr(), nfsrv_setattr(), nfsrv_lookup(), + nfsrv3_access(), nfsrv_readlink(), nfsrv_read(), nfsrv_write(), nfsrv_create(), + nfsrv_mknod(), nfsrv_remove(), nfsrv_rename(), nfsrv_link(), @@ -175,37 +155,42 @@ int nfsrv_null(), nfsrv_mkdir(), nfsrv_rmdir(), nfsrv_readdir(), + nfsrv_readdirplus(), nfsrv_statfs(), + nfsrv_fsinfo(), + nfsrv_pathconf(), + nfsrv_commit(), nfsrv_noop(), - nqnfsrv_readdirlook(), nqnfsrv_getlease(), - nqnfsrv_vacated(), - nqnfsrv_access(); + nqnfsrv_vacated(); -int (*nfsrv_procs[NFS_NPROCS])() = { +int (*nfsrv3_procs[NFS_NPROCS])() = { nfsrv_null, nfsrv_getattr, nfsrv_setattr, - nfsrv_noop, nfsrv_lookup, + nfsrv3_access, nfsrv_readlink, nfsrv_read, - nfsrv_noop, nfsrv_write, nfsrv_create, + nfsrv_mkdir, + nfsrv_symlink, + nfsrv_mknod, nfsrv_remove, + nfsrv_rmdir, nfsrv_rename, nfsrv_link, - nfsrv_symlink, - nfsrv_mkdir, - nfsrv_rmdir, nfsrv_readdir, + nfsrv_readdirplus, nfsrv_statfs, - nqnfsrv_readdirlook, + nfsrv_fsinfo, + nfsrv_pathconf, + nfsrv_commit, nqnfsrv_getlease, nqnfsrv_vacated, nfsrv_noop, - nqnfsrv_access, + nfsrv_noop }; /* @@ -874,13 +859,16 @@ nfs_request(vp, mrest, procnum, procp, cred, mrp, mdp, dposp) struct nfsmount *nmp; struct mbuf *md, *mheadend; struct nfsnode *np; + char nickv[RPCX_NICKVERF]; time_t reqtime, waituntil; caddr_t dpos, cp2; int t1, nqlflag, cachable, s, error = 0, mrest_len, auth_len, auth_type; int trylater_delay = NQ_TRYLATERDEL, trylater_cnt = 0, failed_auth = 0; + int verf_len, verf_type; u_long xid; u_quad_t frev; - char *auth_str; + char *auth_str, *verf_str; + NFSKERBKEY_T key; /* save session key */ nmp = VFSTONFS(vp->v_mount); MALLOC(rep, struct nfsreq *, sizeof(struct nfsreq), M_NFSREQ, M_WAITOK); @@ -900,19 +888,21 @@ nfs_request(vp, mrest, procnum, procp, cred, mrp, mdp, dposp) * Get the RPC header with authorization. */ kerbauth: - auth_str = (char *)0; + verf_str = auth_str = (char *)0; if (nmp->nm_flag & NFSMNT_KERB) { - if (failed_auth) { - error = nfs_getauth(nmp, rep, cred, &auth_type, - &auth_str, &auth_len); + verf_str = nickv; + verf_len = sizeof (nickv); + auth_type = RPCAUTH_KERB4; + bzero((caddr_t)key, sizeof (key)); + if (failed_auth || nfs_getnickauth(nmp, cred, &auth_str, + &auth_len, verf_str, verf_len)) { + error = nfs_getauth(nmp, rep, cred, &auth_str, + &auth_len, verf_str, &verf_len, key); if (error) { free((caddr_t)rep, M_NFSREQ); m_freem(mrest); return (error); } - } else { - auth_type = RPCAUTH_UNIX; - auth_len = 5 * NFSX_UNSIGNED; } } else { auth_type = RPCAUTH_UNIX; @@ -922,8 +912,8 @@ kerbauth: nmp->nm_numgrps : (cred->cr_ngroups - 1)) << 2) + 5 * NFSX_UNSIGNED; } - m = nfsm_rpchead(cred, (nmp->nm_flag & NFSMNT_NQNFS), procnum, - auth_type, auth_len, auth_str, mrest, mrest_len, &mheadend, &xid); + m = nfsm_rpchead(cred, nmp->nm_flag, procnum, auth_type, auth_len, + auth_str, verf_len, verf_str, mrest, mrest_len, &mheadend, &xid); if (auth_str) free(auth_str, M_TEMP); @@ -1029,12 +1019,12 @@ tryagain: /* * break down the rpc header and check if ok */ - nfsm_dissect(tl, u_long *, 3*NFSX_UNSIGNED); + nfsm_dissect(tl, u_long *, 3 * NFSX_UNSIGNED); if (*tl++ == rpc_msgdenied) { if (*tl == rpc_mismatch) error = EOPNOTSUPP; else if ((nmp->nm_flag & NFSMNT_KERB) && *tl++ == rpc_autherr) { - if (*tl == rpc_rejectedcred && failed_auth == 0) { + if (!failed_auth) { failed_auth++; mheadend->m_next = (struct mbuf *)0; m_freem(mrep); @@ -1051,22 +1041,25 @@ tryagain: } /* - * skip over the auth_verf, someday we may want to cache auth_short's - * for nfs_reqhead(), but for now just dump it + * Grab any Kerberos verifier, otherwise just throw it away. */ - if (*++tl != 0) { - i = nfsm_rndup(fxdr_unsigned(long, *tl)); - nfsm_adv(i); - } + verf_type = fxdr_unsigned(int, *tl++); + i = fxdr_unsigned(int, *tl); + if ((nmp->nm_flag & NFSMNT_KERB) && verf_type == RPCAUTH_KERB4) { + error = nfs_savenickauth(nmp, cred, i, key, &md, &dpos, mrep); + if (error) + goto nfsmout; + } else if (i > 0) + nfsm_adv(nfsm_rndup(i)); nfsm_dissect(tl, u_long *, NFSX_UNSIGNED); /* 0 == ok */ if (*tl == 0) { nfsm_dissect(tl, u_long *, NFSX_UNSIGNED); if (*tl != 0) { error = fxdr_unsigned(int, *tl); - m_freem(mrep); - if ((nmp->nm_flag & NFSMNT_NQNFS) && - error == NQNFS_TRYLATER) { + if ((nmp->nm_flag & NFSMNT_NFSV3) && + error == NFSERR_TRYLATER) { + m_freem(mrep); error = 0; waituntil = time.tv_sec + trylater_delay; while (time.tv_sec < waituntil) @@ -1084,6 +1077,13 @@ tryagain: */ if (error == ESTALE) cache_purge(vp); + if (nmp->nm_flag & NFSMNT_NFSV3) { + *mrp = mrep; + *mdp = md; + *dposp = dpos; + error |= NFSERR_RETERR; + } else + m_freem(mrep); m_freem(rep->r_mreq); free((caddr_t)rep, M_NFSREQ); return (error); @@ -1115,10 +1115,10 @@ tryagain: return (0); } m_freem(mrep); - m_freem(rep->r_mreq); - free((caddr_t)rep, M_NFSREQ); error = EPROTONOSUPPORT; nfsmout: + m_freem(rep->r_mreq); + free((caddr_t)rep, M_NFSREQ); return (error); } @@ -1127,9 +1127,10 @@ nfsmout: * siz arg. is used to decide if adding a cluster is worthwhile */ int -nfs_rephead(siz, nd, err, cache, frev, mrq, mbp, bposp) +nfs_rephead(siz, nd, slp, err, cache, frev, mrq, mbp, bposp) int siz; - struct nfsd *nd; + struct nfsrv_descript *nd; + struct nfssvc_sock *slp; int err; int cache; u_quad_t *frev; @@ -1154,47 +1155,98 @@ nfs_rephead(siz, nd, err, cache, frev, mrq, mbp, bposp) } else mreq->m_data += max_hdr; tl = mtod(mreq, u_long *); - mreq->m_len = 6*NFSX_UNSIGNED; - bpos = ((caddr_t)tl)+mreq->m_len; + mreq->m_len = 6 * NFSX_UNSIGNED; + bpos = ((caddr_t)tl) + mreq->m_len; *tl++ = txdr_unsigned(nd->nd_retxid); *tl++ = rpc_reply; - if (err == ERPCMISMATCH || err == NQNFS_AUTHERR) { + if (err == ERPCMISMATCH || (err & NFSERR_AUTHERR)) { *tl++ = rpc_msgdenied; - if (err == NQNFS_AUTHERR) { + if (err & NFSERR_AUTHERR) { *tl++ = rpc_autherr; - *tl = rpc_rejectedcred; + *tl = txdr_unsigned(err & ~NFSERR_AUTHERR); mreq->m_len -= NFSX_UNSIGNED; bpos -= NFSX_UNSIGNED; } else { *tl++ = rpc_mismatch; - *tl++ = txdr_unsigned(2); - *tl = txdr_unsigned(2); + *tl++ = txdr_unsigned(RPC_VER2); + *tl = txdr_unsigned(RPC_VER2); } } else { *tl++ = rpc_msgaccepted; - *tl++ = 0; - *tl++ = 0; + + /* + * For Kerberos authentication, we must send the nickname + * verifier back, otherwise just RPCAUTH_NULL. + */ + if (nd->nd_flag & ND_KERBFULL) { + register struct nfsuid *nuidp; + struct timeval ktvin, ktvout; + NFSKERBKEYSCHED_T keys; /* stores key schedule */ + + for (nuidp = NUIDHASH(slp, nd->nd_cr.cr_uid)->lh_first; + nuidp != 0; nuidp = nuidp->nu_hash.le_next) { + if (nuidp->nu_cr.cr_uid == nd->nd_cr.cr_uid && + (!nd->nd_nam2 || netaddr_match(NU_NETFAM(nuidp), + &nuidp->nu_haddr, nd->nd_nam2))) + break; + } + if (nuidp) { + ktvin.tv_sec = + txdr_unsigned(nuidp->nu_timestamp.tv_sec - 1); + ktvin.tv_usec = + txdr_unsigned(nuidp->nu_timestamp.tv_usec); + + /* + * Encrypt the timestamp in ecb mode using the + * session key. + */ +#ifdef NFSKERB + XXX +#endif + + *tl++ = rpc_auth_kerb; + *tl++ = txdr_unsigned(3 * NFSX_UNSIGNED); + *tl = ktvout.tv_sec; + nfsm_build(tl, u_long *, 3 * NFSX_UNSIGNED); + *tl++ = ktvout.tv_usec; + *tl++ = txdr_unsigned(nuidp->nu_cr.cr_uid); + } else { + *tl++ = 0; + *tl++ = 0; + } + } else { + *tl++ = 0; + *tl++ = 0; + } switch (err) { case EPROGUNAVAIL: *tl = txdr_unsigned(RPC_PROGUNAVAIL); break; case EPROGMISMATCH: *tl = txdr_unsigned(RPC_PROGMISMATCH); - nfsm_build(tl, u_long *, 2*NFSX_UNSIGNED); - *tl++ = txdr_unsigned(2); - *tl = txdr_unsigned(2); /* someday 3 */ + nfsm_build(tl, u_long *, 2 * NFSX_UNSIGNED); + if (nd->nd_flag & ND_NQNFS) { + *tl++ = txdr_unsigned(3); + *tl = txdr_unsigned(3); + } else { + *tl++ = txdr_unsigned(2); + *tl = txdr_unsigned(3); + } break; case EPROCUNAVAIL: *tl = txdr_unsigned(RPC_PROCUNAVAIL); break; + case EBADRPC: + *tl = txdr_unsigned(RPC_GARBAGE); + break; default: *tl = 0; - if (err != VNOVAL) { + if (err != NFSERR_RETVOID) { nfsm_build(tl, u_long *, NFSX_UNSIGNED); if (err) - *tl = txdr_unsigned(nfsrv_errmap[err - 1]); + *tl = txdr_unsigned(nfsrv_errmap(nd, err)); else - *tl = 0; + *tl = 0; } break; }; @@ -1203,16 +1255,14 @@ nfs_rephead(siz, nd, err, cache, frev, mrq, mbp, bposp) /* * For nqnfs, piggyback lease as requested. */ - if (nd->nd_nqlflag != NQL_NOVAL && err == 0) { - if (nd->nd_nqlflag) { - nfsm_build(tl, u_long *, 5*NFSX_UNSIGNED); - *tl++ = txdr_unsigned(nd->nd_nqlflag); + if ((nd->nd_flag & ND_NQNFS) && err == 0) { + if (nd->nd_flag & ND_LEASE) { + nfsm_build(tl, u_long *, 5 * NFSX_UNSIGNED); + *tl++ = txdr_unsigned(nd->nd_flag & ND_LEASE); *tl++ = txdr_unsigned(cache); *tl++ = txdr_unsigned(nd->nd_duration); txdr_hyper(frev, tl); } else { - if (nd->nd_nqlflag != 0) - panic("nqreph"); nfsm_build(tl, u_long *, NFSX_UNSIGNED); *tl = 0; } @@ -1220,7 +1270,7 @@ nfs_rephead(siz, nd, err, cache, frev, mrq, mbp, bposp) *mrq = mreq; *mbp = mb; *bposp = bpos; - if (err != 0 && err != VNOVAL) + if (err != 0 && err != NFSERR_RETVOID) nfsstats.srvrpc_errs++; return (0); } @@ -1240,8 +1290,10 @@ nfs_timer(arg) register struct socket *so; register struct nfsmount *nmp; register int timeo; + register struct nfssvc_sock *slp; static long lasttime = 0; int s, error; + u_quad_t cur_usec; s = splnet(); for (rep = nfs_reqq.tqh_first; rep != 0; rep = rep->r_chain.tqe_next) { @@ -1338,8 +1390,19 @@ nfs_timer(arg) lasttime = time.tv_sec; nqnfs_serverd(); } + + /* + * Scan the write gathering queues for writes that need to be + * completed now. + */ + cur_usec = (u_quad_t)time.tv_sec * 1000000 + (u_quad_t)time.tv_usec; + for (slp = nfssvc_sockhead.tqh_first; slp != 0; + slp = slp->ns_chain.tqe_next) { + if (slp->ns_tq.lh_first && slp->ns_tq.lh_first->nd_time<=cur_usec) + nfsrv_wakenfsd(slp); + } splx(s); - timeout(nfs_timer, (void *)0, hz / NFS_HZ); + timeout(nfs_timer, (void *)0, nfs_ticks); } /* @@ -1673,7 +1736,7 @@ nfsrv_getstream(slp, waitflag) register struct nfssvc_sock *slp; int waitflag; { - register struct mbuf *m; + register struct mbuf *m, **mpp; register char *cp1, *cp2; register int len; struct mbuf *om, *m2, *recm = 0; @@ -1707,7 +1770,12 @@ nfsrv_getstream(slp, waitflag) } } slp->ns_cc -= NFSX_UNSIGNED; - slp->ns_reclen = ntohl(recmark) & ~0x80000000; + recmark = ntohl(recmark); + slp->ns_reclen = recmark & ~0x80000000; + if (recmark & 0x80000000) + slp->ns_flag |= SLP_LASTFRAG; + else + slp->ns_flag &= ~SLP_LASTFRAG; if (slp->ns_reclen < NFS_MINPACKET || slp->ns_reclen > NFS_MAXPACKET) { slp->ns_flag &= ~SLP_GETSTREAM; return (EPERM); @@ -1761,12 +1829,23 @@ nfsrv_getstream(slp, waitflag) slp->ns_flag &= ~SLP_GETSTREAM; return (0); } - nfs_realign(recm, 10 * NFSX_UNSIGNED); - if (slp->ns_recend) - slp->ns_recend->m_nextpkt = recm; - else - slp->ns_rec = recm; - slp->ns_recend = recm; + + /* + * Accumulate the fragments into a record. + */ + mpp = &slp->ns_frag; + while (*mpp) + mpp = &((*mpp)->m_next); + *mpp = recm; + if (slp->ns_flag & SLP_LASTFRAG) { + nfs_realign(slp->ns_frag, 10 * NFSX_UNSIGNED); + if (slp->ns_recend) + slp->ns_recend->m_nextpkt = slp->ns_frag; + else + slp->ns_rec = slp->ns_frag; + slp->ns_recend = slp->ns_frag; + slp->ns_frag = (struct mbuf *)0; + } } } @@ -1774,13 +1853,16 @@ nfsrv_getstream(slp, waitflag) * Parse an RPC header. */ int -nfsrv_dorec(slp, nd) +nfsrv_dorec(slp, nfsd, ndp) register struct nfssvc_sock *slp; - register struct nfsd *nd; + struct nfsd *nfsd; + struct nfsrv_descript **ndp; { - register struct mbuf *m; + register struct mbuf *m, *nam; + register struct nfsrv_descript *nd; int error; + *ndp = NULL; if ((slp->ns_flag & SLP_VALID) == 0 || (m = slp->ns_rec) == (struct mbuf *)0) return (ENOBUFS); @@ -1790,19 +1872,24 @@ nfsrv_dorec(slp, nd) else slp->ns_recend = (struct mbuf *)0; if (m->m_type == MT_SONAME) { - nd->nd_nam = m; - nd->nd_md = nd->nd_mrep = m->m_next; - m->m_next = (struct mbuf *)0; - } else { - nd->nd_nam = (struct mbuf *)0; - nd->nd_md = nd->nd_mrep = m; - } - nd->nd_dpos = mtod(nd->nd_md, caddr_t); - error = nfs_getreq(nd, TRUE); + nam = m; + m = m->m_next; + nam->m_next = NULL; + } else + nam = NULL; + MALLOC(nd, struct nfsrv_descript *, sizeof (struct nfsrv_descript), + M_NFSRVDESC, M_WAITOK); + nd->nd_md = nd->nd_mrep = m; + nd->nd_nam2 = nam; + nd->nd_dpos = mtod(m, caddr_t); + error = nfs_getreq(nd, nfsd, TRUE); if (error) { - m_freem(nd->nd_nam); + m_freem(nam); + free((caddr_t)nd, M_NFSRVDESC); return (error); } + *ndp = nd; + nfsd->nfsd_nd = nd; return (0); } @@ -1812,8 +1899,9 @@ nfsrv_dorec(slp, nd) * - fill in the cred struct. */ int -nfs_getreq(nd, has_header) - register struct nfsd *nd; +nfs_getreq(nd, nfsd, has_header) + register struct nfsrv_descript *nd; + struct nfsd *nfsd; int has_header; { register int len, i; @@ -1821,57 +1909,67 @@ nfs_getreq(nd, has_header) register long t1; struct uio uio; struct iovec iov; - caddr_t dpos, cp2; + caddr_t dpos, cp2, cp; u_long nfsvers, auth_type; - int error = 0, nqnfs = 0; + uid_t nickuid; + int error = 0, nqnfs = 0, ticklen; struct mbuf *mrep, *md; + register struct nfsuid *nuidp; + struct timeval tvin, tvout; + NFSKERBKEYSCHED_T keys; /* stores key schedule */ mrep = nd->nd_mrep; md = nd->nd_md; dpos = nd->nd_dpos; if (has_header) { - nfsm_dissect(tl, u_long *, 10*NFSX_UNSIGNED); + nfsm_dissect(tl, u_long *, 10 * NFSX_UNSIGNED); nd->nd_retxid = fxdr_unsigned(u_long, *tl++); if (*tl++ != rpc_call) { m_freem(mrep); return (EBADRPC); } - } else { - nfsm_dissect(tl, u_long *, 8*NFSX_UNSIGNED); - } + } else + nfsm_dissect(tl, u_long *, 8 * NFSX_UNSIGNED); nd->nd_repstat = 0; + nd->nd_flag = 0; if (*tl++ != rpc_vers) { nd->nd_repstat = ERPCMISMATCH; nd->nd_procnum = NFSPROC_NOOP; return (0); } - nfsvers = nfs_vers; if (*tl != nfs_prog) { - if (*tl == nqnfs_prog) { + if (*tl == nqnfs_prog) nqnfs++; - nfsvers = nqnfs_vers; - } else { + else { nd->nd_repstat = EPROGUNAVAIL; nd->nd_procnum = NFSPROC_NOOP; return (0); } } tl++; - if (*tl++ != nfsvers) { + nfsvers = fxdr_unsigned(u_long, *tl++); + if (((nfsvers < NFS_VER2 || nfsvers > NFS_VER3) && !nqnfs) || + (nfsvers != NQNFS_VER3 && nqnfs)) { nd->nd_repstat = EPROGMISMATCH; nd->nd_procnum = NFSPROC_NOOP; return (0); } + if (nqnfs) + nd->nd_flag = (ND_NFSV3 | ND_NQNFS); + else if (nfsvers == NFS_VER3) + nd->nd_flag = ND_NFSV3; nd->nd_procnum = fxdr_unsigned(u_long, *tl++); if (nd->nd_procnum == NFSPROC_NULL) return (0); if (nd->nd_procnum >= NFS_NPROCS || - (!nqnfs && nd->nd_procnum > NFSPROC_STATFS) || - (*tl != rpc_auth_unix && *tl != rpc_auth_kerb)) { + (!nqnfs && nd->nd_procnum >= NQNFSPROC_GETLEASE) || + (!nd->nd_flag && nd->nd_procnum > NFSV2PROC_STATFS)) { nd->nd_repstat = EPROCUNAVAIL; nd->nd_procnum = NFSPROC_NOOP; return (0); } + if ((nd->nd_flag & ND_NFSV3) == 0) + nd->nd_procnum = nfsv3_procid[nd->nd_procnum]; auth_type = *tl++; len = fxdr_unsigned(int, *tl++); if (len < 0 || len > RPCAUTH_MAXSIZ) { @@ -1879,6 +1977,7 @@ nfs_getreq(nd, has_header) return (EBADRPC); } + nd->nd_flag &= ~ND_KERBAUTH; /* * Handle auth_unix or auth_kerb. */ @@ -1889,7 +1988,9 @@ nfs_getreq(nd, has_header) return (EBADRPC); } nfsm_adv(nfsm_rndup(len)); - nfsm_dissect(tl, u_long *, 3*NFSX_UNSIGNED); + nfsm_dissect(tl, u_long *, 3 * NFSX_UNSIGNED); + bzero((caddr_t)&nd->nd_cr, sizeof (struct ucred)); + nd->nd_cr.cr_ref = 1; nd->nd_cr.cr_uid = fxdr_unsigned(uid_t, *tl++); nd->nd_cr.cr_gid = fxdr_unsigned(gid_t, *tl++); len = fxdr_unsigned(int, *tl); @@ -1897,45 +1998,124 @@ nfs_getreq(nd, has_header) m_freem(mrep); return (EBADRPC); } - nfsm_dissect(tl, u_long *, (len + 2)*NFSX_UNSIGNED); + nfsm_dissect(tl, u_long *, (len + 2) * NFSX_UNSIGNED); for (i = 1; i <= len; i++) - if (i < NGROUPS) - nd->nd_cr.cr_groups[i] = fxdr_unsigned(gid_t, *tl++); - else - tl++; + if (i < NGROUPS) + nd->nd_cr.cr_groups[i] = fxdr_unsigned(gid_t, *tl++); + else + tl++; nd->nd_cr.cr_ngroups = (len >= NGROUPS) ? NGROUPS : (len + 1); - } else if (auth_type == rpc_auth_kerb) { - nd->nd_cr.cr_uid = fxdr_unsigned(uid_t, *tl++); - nd->nd_authlen = fxdr_unsigned(int, *tl); - uio.uio_resid = nfsm_rndup(nd->nd_authlen); - if (uio.uio_resid > (len - 2 * NFSX_UNSIGNED)) { + if (nd->nd_cr.cr_ngroups > 1) + nfsrvw_sort(nd->nd_cr.cr_groups, nd->nd_cr.cr_ngroups); + len = fxdr_unsigned(int, *++tl); + if (len < 0 || len > RPCAUTH_MAXSIZ) { m_freem(mrep); return (EBADRPC); } - uio.uio_offset = 0; - uio.uio_iov = &iov; - uio.uio_iovcnt = 1; - uio.uio_segflg = UIO_SYSSPACE; - iov.iov_base = (caddr_t)nd->nd_authstr; - iov.iov_len = RPCAUTH_MAXSIZ; - nfsm_mtouio(&uio, uio.uio_resid); - nfsm_dissect(tl, u_long *, 2 * NFSX_UNSIGNED); - nd->nd_flag |= NFSD_NEEDAUTH; - } + if (len > 0) + nfsm_adv(nfsm_rndup(len)); + } else if (auth_type == rpc_auth_kerb) { + switch (fxdr_unsigned(int, *tl++)) { + case RPCAKN_FULLNAME: + ticklen = fxdr_unsigned(int, *tl); + *((u_long *)nfsd->nfsd_authstr) = *tl; + uio.uio_resid = nfsm_rndup(ticklen) + NFSX_UNSIGNED; + nfsd->nfsd_authlen = uio.uio_resid + NFSX_UNSIGNED; + if (uio.uio_resid > (len - 2 * NFSX_UNSIGNED)) { + m_freem(mrep); + return (EBADRPC); + } + uio.uio_offset = 0; + uio.uio_iov = &iov; + uio.uio_iovcnt = 1; + uio.uio_segflg = UIO_SYSSPACE; + iov.iov_base = (caddr_t)&nfsd->nfsd_authstr[4]; + iov.iov_len = RPCAUTH_MAXSIZ - 4; + nfsm_mtouio(&uio, uio.uio_resid); + nfsm_dissect(tl, u_long *, 2 * NFSX_UNSIGNED); + if (*tl++ != rpc_auth_kerb || + fxdr_unsigned(int, *tl) != 4 * NFSX_UNSIGNED) { + printf("Bad kerb verifier\n"); + nd->nd_repstat = (NFSERR_AUTHERR|AUTH_BADVERF); + nd->nd_procnum = NFSPROC_NOOP; + return (0); + } + nfsm_dissect(cp, caddr_t, 4 * NFSX_UNSIGNED); + tl = (u_long *)cp; + if (fxdr_unsigned(int, *tl) != RPCAKN_FULLNAME) { + printf("Not fullname kerb verifier\n"); + nd->nd_repstat = (NFSERR_AUTHERR|AUTH_BADVERF); + nd->nd_procnum = NFSPROC_NOOP; + return (0); + } + cp += NFSX_UNSIGNED; + bcopy(cp, nfsd->nfsd_verfstr, 3 * NFSX_UNSIGNED); + nfsd->nfsd_verflen = 3 * NFSX_UNSIGNED; + nd->nd_flag |= ND_KERBFULL; + nfsd->nfsd_flag |= NFSD_NEEDAUTH; + break; + case RPCAKN_NICKNAME: + if (len != 2 * NFSX_UNSIGNED) { + printf("Kerb nickname short\n"); + nd->nd_repstat = (NFSERR_AUTHERR|AUTH_BADCRED); + nd->nd_procnum = NFSPROC_NOOP; + return (0); + } + nickuid = fxdr_unsigned(uid_t, *tl); + nfsm_dissect(tl, u_long *, 2 * NFSX_UNSIGNED); + if (*tl++ != rpc_auth_kerb || + fxdr_unsigned(int, *tl) != 3 * NFSX_UNSIGNED) { + printf("Kerb nick verifier bad\n"); + nd->nd_repstat = (NFSERR_AUTHERR|AUTH_BADVERF); + nd->nd_procnum = NFSPROC_NOOP; + return (0); + } + nfsm_dissect(tl, u_long *, 3 * NFSX_UNSIGNED); + tvin.tv_sec = *tl++; + tvin.tv_usec = *tl; + + for (nuidp = NUIDHASH(nfsd->nfsd_slp,nickuid)->lh_first; + nuidp != 0; nuidp = nuidp->nu_hash.le_next) { + if (nuidp->nu_cr.cr_uid == nickuid && + (!nd->nd_nam2 || + netaddr_match(NU_NETFAM(nuidp), + &nuidp->nu_haddr, nd->nd_nam2))) + break; + } + if (!nuidp) { + nd->nd_repstat = + (NFSERR_AUTHERR|AUTH_REJECTCRED); + nd->nd_procnum = NFSPROC_NOOP; + return (0); + } - /* - * Do we have any use for the verifier. - * According to the "Remote Procedure Call Protocol Spec." it - * should be AUTH_NULL, but some clients make it AUTH_UNIX? - * For now, just skip over it - */ - len = fxdr_unsigned(int, *++tl); - if (len < 0 || len > RPCAUTH_MAXSIZ) { - m_freem(mrep); - return (EBADRPC); - } - if (len > 0) { - nfsm_adv(nfsm_rndup(len)); + /* + * Now, decrypt the timestamp using the session key + * and validate it. + */ +#ifdef NFSKERB + XXX +#endif + + tvout.tv_sec = fxdr_unsigned(long, tvout.tv_sec); + tvout.tv_usec = fxdr_unsigned(long, tvout.tv_usec); + if (nuidp->nu_expire < time.tv_sec || + nuidp->nu_timestamp.tv_sec > tvout.tv_sec || + (nuidp->nu_timestamp.tv_sec == tvout.tv_sec && + nuidp->nu_timestamp.tv_usec > tvout.tv_usec)) { + nuidp->nu_expire = 0; + nd->nd_repstat = + (NFSERR_AUTHERR|AUTH_REJECTVERF); + nd->nd_procnum = NFSPROC_NOOP; + return (0); + } + nfsrv_setcred(&nuidp->nu_cr, &nd->nd_cr); + nd->nd_flag |= ND_KERBNICK; + }; + } else { + nd->nd_repstat = (NFSERR_AUTHERR | AUTH_REJECTCRED); + nd->nd_procnum = NFSPROC_NOOP; + return (0); } /* @@ -1943,16 +2123,14 @@ nfs_getreq(nd, has_header) */ if (nqnfs && nd->nd_procnum != NQNFSPROC_EVICTED) { nfsm_dissect(tl, u_long *, NFSX_UNSIGNED); - nd->nd_nqlflag = fxdr_unsigned(int, *tl); - if (nd->nd_nqlflag) { + nd->nd_flag |= fxdr_unsigned(int, *tl); + if (nd->nd_flag & ND_LEASE) { nfsm_dissect(tl, u_long *, NFSX_UNSIGNED); nd->nd_duration = fxdr_unsigned(int, *tl); } else nd->nd_duration = NQ_MINLEASE; - } else { - nd->nd_nqlflag = NQL_NOVAL; + } else nd->nd_duration = NQ_MINLEASE; - } nd->nd_md = md; nd->nd_dpos = dpos; return (0); @@ -1973,13 +2151,13 @@ nfsrv_wakenfsd(slp) if ((slp->ns_flag & SLP_VALID) == 0) return; - for (nd = nfsd_head.tqh_first; nd != 0; nd = nd->nd_chain.tqe_next) { - if (nd->nd_flag & NFSD_WAITING) { - nd->nd_flag &= ~NFSD_WAITING; - if (nd->nd_slp) + for (nd = nfsd_head.tqh_first; nd != 0; nd = nd->nfsd_chain.tqe_next) { + if (nd->nfsd_flag & NFSD_WAITING) { + nd->nfsd_flag &= ~NFSD_WAITING; + if (nd->nfsd_slp) panic("nfsd wakeup"); slp->ns_sref++; - nd->nd_slp = slp; + nd->nfsd_slp = slp; wakeup((caddr_t)nd); return; } diff --git a/sys/nfsserver/nfs_srvsubs.c b/sys/nfsserver/nfs_srvsubs.c index 1653c8b..350ba46 100644 --- a/sys/nfsserver/nfs_srvsubs.c +++ b/sys/nfsserver/nfs_srvsubs.c @@ -34,7 +34,7 @@ * SUCH DAMAGE. * * @(#)nfs_subs.c 8.3 (Berkeley) 1/4/94 - * $Id: nfs_subs.c,v 1.15 1995/05/30 08:12:43 rgrimes Exp $ + * $Id: nfs_subs.c,v 1.16 1995/06/14 06:23:38 joerg Exp $ */ /* @@ -52,6 +52,7 @@ #include <sys/mbuf.h> #include <sys/socket.h> #include <sys/stat.h> +#include <sys/malloc.h> #ifdef VFS_LKM #include <sys/sysent.h> #include <sys/syscall.h> @@ -60,7 +61,7 @@ #include <vm/vm.h> #include <nfs/rpcv2.h> -#include <nfs/nfsv2.h> +#include <nfs/nfsproto.h> #include <nfs/nfsnode.h> #include <nfs/nfs.h> #include <nfs/xdr_subs.h> @@ -78,31 +79,455 @@ #include <netiso/iso.h> #endif -#define TRUE 1 -#define FALSE 0 - /* * Data items converted to xdr at startup, since they are constant * This is kinda hokey, but may save a little time doing byte swaps */ -u_long nfs_procids[NFS_NPROCS]; u_long nfs_xdrneg1; u_long rpc_call, rpc_vers, rpc_reply, rpc_msgdenied, rpc_autherr, - rpc_mismatch, rpc_auth_unix, rpc_msgaccepted, rpc_rejectedcred, + rpc_mismatch, rpc_auth_unix, rpc_msgaccepted, rpc_auth_kerb; -u_long nfs_vers, nfs_prog, nfs_true, nfs_false; +u_long nfs_prog, nqnfs_prog, nfs_true, nfs_false; /* And other global data */ static u_long nfs_xid = 0; -enum vtype ntov_type[7] = { VNON, VREG, VDIR, VBLK, VCHR, VLNK, VNON }; +enum vtype nv2tov_type[8] = { VNON, VREG, VDIR, VBLK, VCHR, VLNK, VNON, VNON }; +enum vtype nv3tov_type[8]={ VNON, VREG, VDIR, VBLK, VCHR, VLNK, VSOCK, VFIFO }; +int nfs_ticks; + +/* + * Mapping of old NFS Version 2 RPC numbers to generic numbers. + */ +int nfsv3_procid[NFS_NPROCS] = { + NFSPROC_NULL, + NFSPROC_GETATTR, + NFSPROC_SETATTR, + NFSPROC_NOOP, + NFSPROC_LOOKUP, + NFSPROC_READLINK, + NFSPROC_READ, + NFSPROC_NOOP, + NFSPROC_WRITE, + NFSPROC_CREATE, + NFSPROC_REMOVE, + NFSPROC_RENAME, + NFSPROC_LINK, + NFSPROC_SYMLINK, + NFSPROC_MKDIR, + NFSPROC_RMDIR, + NFSPROC_READDIR, + NFSPROC_FSSTAT, + NFSPROC_NOOP, + NFSPROC_NOOP, + NFSPROC_NOOP, + NFSPROC_NOOP, + NFSPROC_NOOP, + NFSPROC_NOOP, + NFSPROC_NOOP, + NFSPROC_NOOP +}; + +/* + * and the reverse mapping from generic to Version 2 procedure numbers + */ +int nfsv2_procid[NFS_NPROCS] = { + NFSV2PROC_NULL, + NFSV2PROC_GETATTR, + NFSV2PROC_SETATTR, + NFSV2PROC_LOOKUP, + NFSV2PROC_NOOP, + NFSV2PROC_READLINK, + NFSV2PROC_READ, + NFSV2PROC_WRITE, + NFSV2PROC_CREATE, + NFSV2PROC_MKDIR, + NFSV2PROC_SYMLINK, + NFSV2PROC_CREATE, + NFSV2PROC_REMOVE, + NFSV2PROC_RMDIR, + NFSV2PROC_RENAME, + NFSV2PROC_LINK, + NFSV2PROC_READDIR, + NFSV2PROC_NOOP, + NFSV2PROC_STATFS, + NFSV2PROC_NOOP, + NFSV2PROC_NOOP, + NFSV2PROC_NOOP, + NFSV2PROC_NOOP, + NFSV2PROC_NOOP, + NFSV2PROC_NOOP, + NFSV2PROC_NOOP, +}; + +/* + * Maps errno values to nfs error numbers. + * Use NFSERR_IO as the catch all for ones not specifically defined in + * RFC 1094. + */ +static u_char nfsrv_v2errmap[ELAST] = { + NFSERR_PERM, NFSERR_NOENT, NFSERR_IO, NFSERR_IO, NFSERR_IO, + NFSERR_NXIO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, + NFSERR_IO, NFSERR_IO, NFSERR_ACCES, NFSERR_IO, NFSERR_IO, + NFSERR_IO, NFSERR_EXIST, NFSERR_IO, NFSERR_NODEV, NFSERR_NOTDIR, + NFSERR_ISDIR, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, + NFSERR_IO, NFSERR_FBIG, NFSERR_NOSPC, NFSERR_IO, NFSERR_ROFS, + NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, + NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, + NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, + NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, + NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, + NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, + NFSERR_IO, NFSERR_IO, NFSERR_NAMETOL, NFSERR_IO, NFSERR_IO, + NFSERR_NOTEMPTY, NFSERR_IO, NFSERR_IO, NFSERR_DQUOT, NFSERR_STALE, + NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, + NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, + NFSERR_IO, +}; + +/* + * Maps errno values to nfs error numbers. + * Although it is not obvious whether or not NFS clients really care if + * a returned error value is in the specified list for the procedure, the + * safest thing to do is filter them appropriately. For Version 2, the + * X/Open XNFS document is the only specification that defines error values + * for each RPC (The RFC simply lists all possible error values for all RPCs), + * so I have decided to not do this for Version 2. + * The first entry is the default error return and the rest are the valid + * errors for that RPC in increasing numeric order. + */ +static short nfsv3err_null[] = { + 0, + 0, +}; + +static short nfsv3err_getattr[] = { + NFSERR_IO, + NFSERR_IO, + NFSERR_STALE, + NFSERR_BADHANDLE, + NFSERR_SERVERFAULT, + 0, +}; + +static short nfsv3err_setattr[] = { + NFSERR_IO, + NFSERR_PERM, + NFSERR_IO, + NFSERR_ACCES, + NFSERR_INVAL, + NFSERR_NOSPC, + NFSERR_ROFS, + NFSERR_DQUOT, + NFSERR_STALE, + NFSERR_BADHANDLE, + NFSERR_NOT_SYNC, + NFSERR_SERVERFAULT, + 0, +}; + +static short nfsv3err_lookup[] = { + NFSERR_IO, + NFSERR_NOENT, + NFSERR_IO, + NFSERR_ACCES, + NFSERR_NOTDIR, + NFSERR_NAMETOL, + NFSERR_STALE, + NFSERR_BADHANDLE, + NFSERR_SERVERFAULT, + 0, +}; + +static short nfsv3err_access[] = { + NFSERR_IO, + NFSERR_IO, + NFSERR_STALE, + NFSERR_BADHANDLE, + NFSERR_SERVERFAULT, + 0, +}; + +static short nfsv3err_readlink[] = { + NFSERR_IO, + NFSERR_IO, + NFSERR_ACCES, + NFSERR_INVAL, + NFSERR_STALE, + NFSERR_BADHANDLE, + NFSERR_NOTSUPP, + NFSERR_SERVERFAULT, + 0, +}; + +static short nfsv3err_read[] = { + NFSERR_IO, + NFSERR_IO, + NFSERR_NXIO, + NFSERR_ACCES, + NFSERR_INVAL, + NFSERR_STALE, + NFSERR_BADHANDLE, + NFSERR_SERVERFAULT, + 0, +}; + +static short nfsv3err_write[] = { + NFSERR_IO, + NFSERR_IO, + NFSERR_ACCES, + NFSERR_INVAL, + NFSERR_FBIG, + NFSERR_NOSPC, + NFSERR_ROFS, + NFSERR_DQUOT, + NFSERR_STALE, + NFSERR_BADHANDLE, + NFSERR_SERVERFAULT, + 0, +}; + +static short nfsv3err_create[] = { + NFSERR_IO, + NFSERR_IO, + NFSERR_ACCES, + NFSERR_EXIST, + NFSERR_NOTDIR, + NFSERR_NOSPC, + NFSERR_ROFS, + NFSERR_NAMETOL, + NFSERR_DQUOT, + NFSERR_STALE, + NFSERR_BADHANDLE, + NFSERR_NOTSUPP, + NFSERR_SERVERFAULT, + 0, +}; + +static short nfsv3err_mkdir[] = { + NFSERR_IO, + NFSERR_IO, + NFSERR_ACCES, + NFSERR_EXIST, + NFSERR_NOTDIR, + NFSERR_NOSPC, + NFSERR_ROFS, + NFSERR_NAMETOL, + NFSERR_DQUOT, + NFSERR_STALE, + NFSERR_BADHANDLE, + NFSERR_NOTSUPP, + NFSERR_SERVERFAULT, + 0, +}; + +static short nfsv3err_symlink[] = { + NFSERR_IO, + NFSERR_IO, + NFSERR_ACCES, + NFSERR_EXIST, + NFSERR_NOTDIR, + NFSERR_NOSPC, + NFSERR_ROFS, + NFSERR_NAMETOL, + NFSERR_DQUOT, + NFSERR_STALE, + NFSERR_BADHANDLE, + NFSERR_NOTSUPP, + NFSERR_SERVERFAULT, + 0, +}; + +static short nfsv3err_mknod[] = { + NFSERR_IO, + NFSERR_IO, + NFSERR_ACCES, + NFSERR_EXIST, + NFSERR_NOTDIR, + NFSERR_NOSPC, + NFSERR_ROFS, + NFSERR_NAMETOL, + NFSERR_DQUOT, + NFSERR_STALE, + NFSERR_BADHANDLE, + NFSERR_NOTSUPP, + NFSERR_SERVERFAULT, + NFSERR_BADTYPE, + 0, +}; + +static short nfsv3err_remove[] = { + NFSERR_IO, + NFSERR_NOENT, + NFSERR_IO, + NFSERR_ACCES, + NFSERR_NOTDIR, + NFSERR_ROFS, + NFSERR_NAMETOL, + NFSERR_STALE, + NFSERR_BADHANDLE, + NFSERR_SERVERFAULT, + 0, +}; + +static short nfsv3err_rmdir[] = { + NFSERR_IO, + NFSERR_NOENT, + NFSERR_IO, + NFSERR_ACCES, + NFSERR_EXIST, + NFSERR_NOTDIR, + NFSERR_INVAL, + NFSERR_ROFS, + NFSERR_NAMETOL, + NFSERR_NOTEMPTY, + NFSERR_STALE, + NFSERR_BADHANDLE, + NFSERR_NOTSUPP, + NFSERR_SERVERFAULT, + 0, +}; + +static short nfsv3err_rename[] = { + NFSERR_IO, + NFSERR_NOENT, + NFSERR_IO, + NFSERR_ACCES, + NFSERR_EXIST, + NFSERR_XDEV, + NFSERR_NOTDIR, + NFSERR_ISDIR, + NFSERR_INVAL, + NFSERR_NOSPC, + NFSERR_ROFS, + NFSERR_MLINK, + NFSERR_NAMETOL, + NFSERR_NOTEMPTY, + NFSERR_DQUOT, + NFSERR_STALE, + NFSERR_BADHANDLE, + NFSERR_NOTSUPP, + NFSERR_SERVERFAULT, + 0, +}; + +static short nfsv3err_link[] = { + NFSERR_IO, + NFSERR_IO, + NFSERR_ACCES, + NFSERR_EXIST, + NFSERR_XDEV, + NFSERR_NOTDIR, + NFSERR_INVAL, + NFSERR_NOSPC, + NFSERR_ROFS, + NFSERR_MLINK, + NFSERR_NAMETOL, + NFSERR_DQUOT, + NFSERR_STALE, + NFSERR_BADHANDLE, + NFSERR_NOTSUPP, + NFSERR_SERVERFAULT, + 0, +}; + +static short nfsv3err_readdir[] = { + NFSERR_IO, + NFSERR_IO, + NFSERR_ACCES, + NFSERR_NOTDIR, + NFSERR_STALE, + NFSERR_BADHANDLE, + NFSERR_BAD_COOKIE, + NFSERR_TOOSMALL, + NFSERR_SERVERFAULT, + 0, +}; + +static short nfsv3err_readdirplus[] = { + NFSERR_IO, + NFSERR_IO, + NFSERR_ACCES, + NFSERR_NOTDIR, + NFSERR_STALE, + NFSERR_BADHANDLE, + NFSERR_BAD_COOKIE, + NFSERR_NOTSUPP, + NFSERR_TOOSMALL, + NFSERR_SERVERFAULT, + 0, +}; + +static short nfsv3err_fsstat[] = { + NFSERR_IO, + NFSERR_IO, + NFSERR_STALE, + NFSERR_BADHANDLE, + NFSERR_SERVERFAULT, + 0, +}; + +static short nfsv3err_fsinfo[] = { + NFSERR_STALE, + NFSERR_STALE, + NFSERR_BADHANDLE, + NFSERR_SERVERFAULT, + 0, +}; + +static short nfsv3err_pathconf[] = { + NFSERR_STALE, + NFSERR_STALE, + NFSERR_BADHANDLE, + NFSERR_SERVERFAULT, + 0, +}; + +static short nfsv3err_commit[] = { + NFSERR_IO, + NFSERR_IO, + NFSERR_STALE, + NFSERR_BADHANDLE, + NFSERR_SERVERFAULT, + 0, +}; + +static short *nfsrv_v3errmap[] = { + nfsv3err_null, + nfsv3err_getattr, + nfsv3err_setattr, + nfsv3err_lookup, + nfsv3err_access, + nfsv3err_readlink, + nfsv3err_read, + nfsv3err_write, + nfsv3err_create, + nfsv3err_mkdir, + nfsv3err_symlink, + nfsv3err_mknod, + nfsv3err_remove, + nfsv3err_rmdir, + nfsv3err_rename, + nfsv3err_link, + nfsv3err_readdir, + nfsv3err_readdirplus, + nfsv3err_fsstat, + nfsv3err_fsinfo, + nfsv3err_pathconf, + nfsv3err_commit, +}; + extern struct proc *nfs_iodwant[NFS_MAXASYNCDAEMON]; -extern int nqnfs_piggy[NFS_NPROCS]; extern struct nfsrtt nfsrtt; extern time_t nqnfsstarttime; -extern u_long nqnfs_prog, nqnfs_vers; extern int nqsrv_clockskew; extern int nqsrv_writeslack; extern int nqsrv_maxlease; +extern struct nfsstats nfsstats; +extern int nqnfs_piggy[NFS_NPROCS]; +extern nfstype nfsv2_type[9]; +extern nfstype nfsv3_type[9]; +extern struct nfsnodehashhead *nfsnodehashtbl; +extern u_long nfsnodehash; #ifdef VFS_LKM struct getfh_args; @@ -167,14 +592,16 @@ nfsm_reqh(vp, procid, hsiz, bposp) * Returns the head of the mbuf list. */ struct mbuf * -nfsm_rpchead(cr, nqnfs, procid, auth_type, auth_len, auth_str, mrest, - mrest_len, mbp, xidp) +nfsm_rpchead(cr, nmflag, procid, auth_type, auth_len, auth_str, verf_len, + verf_str, mrest, mrest_len, mbp, xidp) register struct ucred *cr; - int nqnfs; + int nmflag; int procid; int auth_type; int auth_len; char *auth_str; + int verf_len; + char *verf_str; struct mbuf *mrest; int mrest_len; struct mbuf **mbp; @@ -188,15 +615,13 @@ nfsm_rpchead(cr, nqnfs, procid, auth_type, auth_len, auth_str, mrest, int siz, grpsiz, authsiz; authsiz = nfsm_rndup(auth_len); - if (auth_type == RPCAUTH_NQNFS) - authsiz += 2 * NFSX_UNSIGNED; MGETHDR(mb, M_WAIT, MT_DATA); - if ((authsiz + 10*NFSX_UNSIGNED) >= MINCLSIZE) { + if ((authsiz + 10 * NFSX_UNSIGNED) >= MINCLSIZE) { MCLGET(mb, M_WAIT); - } else if ((authsiz + 10*NFSX_UNSIGNED) < MHLEN) { - MH_ALIGN(mb, authsiz + 10*NFSX_UNSIGNED); + } else if ((authsiz + 10 * NFSX_UNSIGNED) < MHLEN) { + MH_ALIGN(mb, authsiz + 10 * NFSX_UNSIGNED); } else { - MH_ALIGN(mb, 8*NFSX_UNSIGNED); + MH_ALIGN(mb, 8 * NFSX_UNSIGNED); } mb->m_len = 0; mreq = mb; @@ -205,20 +630,26 @@ nfsm_rpchead(cr, nqnfs, procid, auth_type, auth_len, auth_str, mrest, /* * First the RPC header. */ - nfsm_build(tl, u_long *, 8*NFSX_UNSIGNED); + nfsm_build(tl, u_long *, 8 * NFSX_UNSIGNED); if (++nfs_xid == 0) nfs_xid++; *tl++ = *xidp = txdr_unsigned(nfs_xid); *tl++ = rpc_call; *tl++ = rpc_vers; - if (nqnfs) { + if (nmflag & NFSMNT_NQNFS) { *tl++ = txdr_unsigned(NQNFS_PROG); - *tl++ = txdr_unsigned(NQNFS_VER1); + *tl++ = txdr_unsigned(NQNFS_VER3); } else { *tl++ = txdr_unsigned(NFS_PROG); - *tl++ = txdr_unsigned(NFS_VER2); + if (nmflag & NFSMNT_NFSV3) + *tl++ = txdr_unsigned(NFS_VER3); + else + *tl++ = txdr_unsigned(NFS_VER2); } - *tl++ = txdr_unsigned(procid); + if (nmflag & NFSMNT_NFSV3) + *tl++ = txdr_unsigned(procid); + else + *tl++ = txdr_unsigned(nfsv2_procid[procid]); /* * And then the authorization cred. @@ -237,10 +668,7 @@ nfsm_rpchead(cr, nqnfs, procid, auth_type, auth_len, auth_str, mrest, for (i = 1; i <= grpsiz; i++) *tl++ = txdr_unsigned(cr->cr_groups[i]); break; - case RPCAUTH_NQNFS: - nfsm_build(tl, u_long *, 2*NFSX_UNSIGNED); - *tl++ = txdr_unsigned(cr->cr_uid); - *tl = txdr_unsigned(auth_len); + case RPCAUTH_KERB4: siz = auth_len; while (siz > 0) { if (M_TRAILINGSPACE(mb) == 0) { @@ -266,11 +694,43 @@ nfsm_rpchead(cr, nqnfs, procid, auth_type, auth_len, auth_str, mrest, } break; }; - nfsm_build(tl, u_long *, 2*NFSX_UNSIGNED); - *tl++ = txdr_unsigned(RPCAUTH_NULL); - *tl = 0; + + /* + * And the verifier... + */ + nfsm_build(tl, u_long *, 2 * NFSX_UNSIGNED); + if (verf_str) { + *tl++ = txdr_unsigned(RPCAUTH_KERB4); + *tl = txdr_unsigned(verf_len); + siz = verf_len; + while (siz > 0) { + if (M_TRAILINGSPACE(mb) == 0) { + MGET(mb2, M_WAIT, MT_DATA); + if (siz >= MINCLSIZE) + MCLGET(mb2, M_WAIT); + mb->m_next = mb2; + mb = mb2; + mb->m_len = 0; + bpos = mtod(mb, caddr_t); + } + i = min(siz, M_TRAILINGSPACE(mb)); + bcopy(verf_str, bpos, i); + mb->m_len += i; + verf_str += i; + bpos += i; + siz -= i; + } + if ((siz = (nfsm_rndup(verf_len) - verf_len)) > 0) { + for (i = 0; i < siz; i++) + *bpos++ = '\0'; + mb->m_len += siz; + } + } else { + *tl++ = txdr_unsigned(RPCAUTH_NULL); + *tl = 0; + } mb->m_next = mrest; - mreq->m_pkthdr.len = authsiz + 10*NFSX_UNSIGNED + mrest_len; + mreq->m_pkthdr.len = authsiz + 10 * NFSX_UNSIGNED + mrest_len; mreq->m_pkthdr.rcvif = (struct ifnet *)0; *mbp = mb; return (mreq); @@ -603,6 +1063,25 @@ nfs_init() { register int i; + /* + * Check to see if major data structures haven't bloated. + */ + if (sizeof (struct nfsnode) > NFS_NODEALLOC) { + printf("struct nfsnode bloated (> %dbytes)\n", NFS_NODEALLOC); + printf("Try reducing NFS_SMALLFH\n"); + } + if (sizeof (struct nfsmount) > NFS_MNTALLOC) { + printf("struct nfsmount bloated (> %dbytes)\n", NFS_MNTALLOC); + printf("Try reducing NFS_MUIDHASHSIZ\n"); + } + if (sizeof (struct nfssvc_sock) > NFS_SVCALLOC) { + printf("struct nfssvc_sock bloated (> %dbytes)\n",NFS_SVCALLOC); + printf("Try reducing NFS_UIDHASHSIZ\n"); + } + if (sizeof (struct nfsuid) > NFS_UIDALLOC) { + printf("struct nfsuid bloated (> %dbytes)\n",NFS_UIDALLOC); + printf("Try unionizing the nu_nickname and nu_flag fields\n"); + } nfsrtt.pos = 0; rpc_vers = txdr_unsigned(RPC_VER2); rpc_call = txdr_unsigned(RPC_CALL); @@ -611,17 +1090,16 @@ nfs_init() rpc_msgaccepted = txdr_unsigned(RPC_MSGACCEPTED); rpc_mismatch = txdr_unsigned(RPC_MISMATCH); rpc_autherr = txdr_unsigned(RPC_AUTHERR); - rpc_rejectedcred = txdr_unsigned(AUTH_REJECTCRED); rpc_auth_unix = txdr_unsigned(RPCAUTH_UNIX); - rpc_auth_kerb = txdr_unsigned(RPCAUTH_NQNFS); - nfs_vers = txdr_unsigned(NFS_VER2); + rpc_auth_kerb = txdr_unsigned(RPCAUTH_KERB4); nfs_prog = txdr_unsigned(NFS_PROG); + nqnfs_prog = txdr_unsigned(NQNFS_PROG); nfs_true = txdr_unsigned(TRUE); nfs_false = txdr_unsigned(FALSE); nfs_xdrneg1 = txdr_unsigned(-1); - /* Loop thru nfs procids */ - for (i = 0; i < NFS_NPROCS; i++) - nfs_procids[i] = txdr_unsigned(i); + nfs_ticks = (hz * NFS_TICKINTVL + 500) / 1000; + if (nfs_ticks < 1) + nfs_ticks = 1; /* Ensure async daemons disabled */ for (i = 0; i < NFS_MAXASYNCDAEMON; i++) nfs_iodwant[i] = (struct proc *)0; @@ -637,8 +1115,6 @@ nfs_init() nqnfsstarttime = boottime.tv_sec + nqsrv_maxlease + nqsrv_clockskew + nqsrv_writeslack; NQLOADNOVRAM(nqnfsstarttime); - nqnfs_prog = txdr_unsigned(NQNFS_PROG); - nqnfs_vers = txdr_unsigned(NQNFS_VER1); CIRCLEQ_INIT(&nqtimerhead); nqfhhashtbl = hashinit(NQLCHSZ, M_NQLEASE, &nqfhhash); } @@ -649,6 +1125,7 @@ nfs_init() TAILQ_INIT(&nfs_reqq); nfs_timer(0); +#ifdef __FreeBSD__ /* * Set up lease_check and lease_updatetime so that other parts * of the system can call us, if we are loadable. @@ -662,6 +1139,7 @@ nfs_init() sysent[SYS_getfh].sy_narg = 2; sysent[SYS_getfh].sy_call = getfh; #endif +#endif return (0); } @@ -689,57 +1167,65 @@ nfs_loadattrcache(vpp, mdp, dposp, vaper) { register struct vnode *vp = *vpp; register struct vattr *vap; - register struct nfsv2_fattr *fp; + register struct nfs_fattr *fp; register struct nfsnode *np; register struct nfsnodehashhead *nhpp; register long t1; - caddr_t dpos, cp2; - int error = 0, isnq; + caddr_t cp2; + int error = 0, rdev; struct mbuf *md; enum vtype vtyp; u_short vmode; - long rdev; struct timespec mtime; struct vnode *nvp; + quad_t tval; + int v3 = NFS_ISV3(vp); md = *mdp; - dpos = *dposp; - t1 = (mtod(md, caddr_t) + md->m_len) - dpos; - isnq = (VFSTONFS(vp->v_mount)->nm_flag & NFSMNT_NQNFS); - error = nfsm_disct(&md, &dpos, NFSX_FATTR(isnq), t1, &cp2); - if (error) + t1 = (mtod(md, caddr_t) + md->m_len) - *dposp; + if (error = nfsm_disct(mdp, dposp, NFSX_FATTR(v3), t1, &cp2)) return (error); - fp = (struct nfsv2_fattr *)cp2; - vtyp = nfstov_type(fp->fa_type); - vmode = fxdr_unsigned(u_short, fp->fa_mode); - /* - * XXX - * - * The duplicate information returned in fa_type and fa_mode - * is an ambiguity in the NFS version 2 protocol. - * - * VREG should be taken literally as a regular file. If a - * server intents to return some type information differently - * in the upper bits of the mode field (e.g. for sockets, or - * FIFOs), NFSv2 mandates fa_type to be VNON. Anyway, we - * leave the examination of the mode bits even in the VREG - * case to avoid breakage for bogus servers, but we make sure - * that there are actually type bits set in the upper part of - * fa_mode (and failing that, trust the va_type field). - * - * NFSv3 cleared the issue, and requires fa_mode to not - * contain any type information (while also introduing sockets - * and FIFOs for fa_type). - */ - if (vtyp == VNON || (vtyp == VREG && (vmode & S_IFMT) != 0)) - vtyp = IFTOVT(vmode); - if (isnq) { - rdev = fxdr_unsigned(long, fp->fa_nqrdev); - fxdr_nqtime(&fp->fa_nqmtime, &mtime); + fp = (struct nfs_fattr *)cp2; + if (v3) { + vtyp = nfsv3tov_type(fp->fa_type); + vmode = fxdr_unsigned(u_short, fp->fa_mode); + rdev = makedev(fxdr_unsigned(u_char, fp->fa3_rdev.specdata1), + fxdr_unsigned(u_char, fp->fa3_rdev.specdata2)); + fxdr_nfsv3time(&fp->fa3_mtime, &mtime); } else { - rdev = fxdr_unsigned(long, fp->fa_nfsrdev); - fxdr_nfstime(&fp->fa_nfsmtime, &mtime); + vtyp = nfsv2tov_type(fp->fa_type); + vmode = fxdr_unsigned(u_short, fp->fa_mode); + /* + * XXX + * + * The duplicate information returned in fa_type and fa_mode + * is an ambiguity in the NFS version 2 protocol. + * + * VREG should be taken literally as a regular file. If a + * server intents to return some type information differently + * in the upper bits of the mode field (e.g. for sockets, or + * FIFOs), NFSv2 mandates fa_type to be VNON. Anyway, we + * leave the examination of the mode bits even in the VREG + * case to avoid breakage for bogus servers, but we make sure + * that there are actually type bits set in the upper part of + * fa_mode (and failing that, trust the va_type field). + * + * NFSv3 cleared the issue, and requires fa_mode to not + * contain any type information (while also introduing sockets + * and FIFOs for fa_type). + */ + if (vtyp == VNON || (vtyp == VREG && (vmode & S_IFMT) != 0)) + vtyp = IFTOVT(vmode); + rdev = fxdr_unsigned(long, fp->fa2_rdev); + fxdr_nfsv2time(&fp->fa2_mtime, &mtime); + + /* + * Really ugly NFSv2 kludge. + */ + if (vtyp == VCHR && rdev == 0xffffffff) + vtyp = VFIFO; } + /* * If v_type == VNON it is a new node, so fill in the v_type, * n_mtime fields. Check to see if it represents a special @@ -749,10 +1235,15 @@ nfs_loadattrcache(vpp, mdp, dposp, vaper) */ np = VTONFS(vp); if (vp->v_type == VNON) { - if (vtyp == VCHR && rdev == 0xffffffff) - vp->v_type = vtyp = VFIFO; - else - vp->v_type = vtyp; + /* + * If we had a lock and it turns out that the vnode + * is an object which we don't want to lock (e.g. VDIR) + * to avoid nasty hanging problems on a server crash, + * then release it here. + */ + if (vtyp != VREG && VOP_ISLOCKED(vp)) + VOP_UNLOCK(vp); + vp->v_type = vtyp; if (vp->v_type == VFIFO) { vp->v_op = fifo_nfsv2nodeop_p; } @@ -773,7 +1264,7 @@ nfs_loadattrcache(vpp, mdp, dposp, vaper) * Reinitialize aliased node. */ np->n_vnode = nvp; - nhpp = nfs_hash(&np->n_fh); + nhpp = NFSNOHASH(nfs_hash(np->n_fhp, np->n_fhsize)); LIST_INSERT_HEAD(nhpp, np, n_hash); *vpp = vp = nvp; } @@ -783,31 +1274,34 @@ nfs_loadattrcache(vpp, mdp, dposp, vaper) vap = &np->n_vattr; vap->va_type = vtyp; vap->va_mode = (vmode & 07777); - vap->va_nlink = fxdr_unsigned(u_short, fp->fa_nlink); - vap->va_uid = fxdr_unsigned(uid_t, fp->fa_uid); - vap->va_gid = fxdr_unsigned(gid_t, fp->fa_gid); vap->va_rdev = (dev_t)rdev; vap->va_mtime = mtime; vap->va_fsid = vp->v_mount->mnt_stat.f_fsid.val[0]; - if (isnq) { - fxdr_hyper(&fp->fa_nqsize, &vap->va_size); - vap->va_blocksize = fxdr_unsigned(long, fp->fa_nqblocksize); - fxdr_hyper(&fp->fa_nqbytes, &vap->va_bytes); - vap->va_fileid = fxdr_unsigned(long, fp->fa_nqfileid); - fxdr_nqtime(&fp->fa_nqatime, &vap->va_atime); - vap->va_flags = fxdr_unsigned(u_long, fp->fa_nqflags); - fxdr_nqtime(&fp->fa_nqctime, &vap->va_ctime); - vap->va_gen = fxdr_unsigned(u_long, fp->fa_nqgen); - fxdr_hyper(&fp->fa_nqfilerev, &vap->va_filerev); + if (v3) { + vap->va_nlink = fxdr_unsigned(u_short, fp->fa_nlink); + vap->va_uid = fxdr_unsigned(uid_t, fp->fa_uid); + vap->va_gid = fxdr_unsigned(gid_t, fp->fa_gid); + fxdr_hyper(&fp->fa3_size, &vap->va_size); + vap->va_blocksize = NFS_FABLKSIZE; + fxdr_hyper(&fp->fa3_used, &vap->va_bytes); + vap->va_fileid = fxdr_unsigned(int, fp->fa3_fileid.nfsuquad[1]); + fxdr_nfsv3time(&fp->fa3_atime, &vap->va_atime); + fxdr_nfsv3time(&fp->fa3_ctime, &vap->va_ctime); + vap->va_flags = 0; + vap->va_filerev = 0; } else { - vap->va_size = fxdr_unsigned(u_long, fp->fa_nfssize); - vap->va_blocksize = fxdr_unsigned(long, fp->fa_nfsblocksize); - vap->va_bytes = fxdr_unsigned(long, fp->fa_nfsblocks) * NFS_FABLKSIZE; - vap->va_fileid = fxdr_unsigned(long, fp->fa_nfsfileid); - fxdr_nfstime(&fp->fa_nfsatime, &vap->va_atime); + vap->va_nlink = fxdr_unsigned(u_short, fp->fa_nlink); + vap->va_uid = fxdr_unsigned(uid_t, fp->fa_uid); + vap->va_gid = fxdr_unsigned(gid_t, fp->fa_gid); + vap->va_size = fxdr_unsigned(u_long, fp->fa2_size); + vap->va_blocksize = fxdr_unsigned(long, fp->fa2_blocksize); + vap->va_bytes = fxdr_unsigned(long, fp->fa2_blocks) * NFS_FABLKSIZE; + vap->va_fileid = fxdr_unsigned(long, fp->fa2_fileid); + fxdr_nfsv2time(&fp->fa2_atime, &vap->va_atime); vap->va_flags = 0; - fxdr_nfstime(&fp->fa_nfsctime, &vap->va_ctime); - vap->va_gen = 0; + vap->va_ctime.ts_sec = fxdr_unsigned(long, fp->fa2_ctime.nfsv2_sec); + vap->va_ctime.ts_nsec = 0; + vap->va_gen = fxdr_unsigned(u_long, fp->fa2_ctime.nfsv2_usec); vap->va_filerev = 0; } if (vap->va_size != np->n_size) { @@ -824,26 +1318,13 @@ nfs_loadattrcache(vpp, mdp, dposp, vaper) np->n_size = vap->va_size; } np->n_attrstamp = time.tv_sec; - *dposp = dpos; - *mdp = md; if (vaper != NULL) { bcopy((caddr_t)vap, (caddr_t)vaper, sizeof(*vap)); -#ifdef notdef - if ((np->n_flag & NMODIFIED) && np->n_size > vap->va_size) - if (np->n_size > vap->va_size) - vaper->va_size = np->n_size; -#endif if (np->n_flag & NCHG) { - if (np->n_flag & NACC) { - vaper->va_atime.ts_sec = np->n_atim.tv_sec; - vaper->va_atime.ts_nsec = - np->n_atim.tv_usec * 1000; - } - if (np->n_flag & NUPD) { - vaper->va_mtime.ts_sec = np->n_mtim.tv_sec; - vaper->va_mtime.ts_nsec = - np->n_mtim.tv_usec * 1000; - } + if (np->n_flag & NACC) + vaper->va_atime = np->n_atim; + if (np->n_flag & NUPD) + vaper->va_mtime = np->n_mtim; } } return (0); @@ -862,12 +1343,7 @@ nfs_getattrcache(vp, vaper) register struct nfsnode *np = VTONFS(vp); register struct vattr *vap; - if (VFSTONFS(vp->v_mount)->nm_flag & NFSMNT_NQLOOKLEASE) { - if (!NQNFS_CKCACHABLE(vp, NQL_READ) || np->n_attrstamp == 0) { - nfsstats.attrcache_misses++; - return (ENOENT); - } - } else if ((time.tv_sec - np->n_attrstamp) >= NFS_ATTRTIMEO(np)) { + if ((time.tv_sec - np->n_attrstamp) >= NFS_ATTRTIMEO(np)) { nfsstats.attrcache_misses++; return (ENOENT); } @@ -887,23 +1363,11 @@ nfs_getattrcache(vp, vaper) np->n_size = vap->va_size; } bcopy((caddr_t)vap, (caddr_t)vaper, sizeof(struct vattr)); -#ifdef notdef - if ((np->n_flag & NMODIFIED) == 0) { - np->n_size = vaper->va_size; - vnode_pager_setsize(vp, (u_long)np->n_size); - } else if (np->n_size > vaper->va_size) - if (np->n_size > vaper->va_size) - vaper->va_size = np->n_size; -#endif if (np->n_flag & NCHG) { - if (np->n_flag & NACC) { - vaper->va_atime.ts_sec = np->n_atim.tv_sec; - vaper->va_atime.ts_nsec = np->n_atim.tv_usec * 1000; - } - if (np->n_flag & NUPD) { - vaper->va_mtime.ts_sec = np->n_mtim.tv_sec; - vaper->va_mtime.ts_nsec = np->n_mtim.tv_usec * 1000; - } + if (np->n_flag & NACC) + vaper->va_atime = np->n_atim; + if (np->n_flag & NUPD) + vaper->va_mtime = np->n_mtim; } return (0); } @@ -912,7 +1376,7 @@ nfs_getattrcache(vp, vaper) * Set up nameidata for a lookup() call and do it */ int -nfs_namei(ndp, fhp, len, slp, nam, mdp, dposp, p) +nfs_namei(ndp, fhp, len, slp, nam, mdp, dposp, retdirp, p, kerbflag) register struct nameidata *ndp; fhandle_t *fhp; int len; @@ -920,7 +1384,9 @@ nfs_namei(ndp, fhp, len, slp, nam, mdp, dposp, p) struct mbuf *nam; struct mbuf **mdp; caddr_t *dposp; + struct vnode **retdirp; struct proc *p; + int kerbflag; { register int i, rem; register struct mbuf *md; @@ -929,6 +1395,7 @@ nfs_namei(ndp, fhp, len, slp, nam, mdp, dposp, p) int error, rdonly; struct componentname *cnp = &ndp->ni_cnd; + *retdirp = (struct vnode *)0; MALLOC(cnp->cn_pnbuf, char *, len + 1, M_NAMEI, M_WAITOK); /* * Copy the name from the mbuf list to ndp->ni_pnbuf @@ -950,7 +1417,7 @@ nfs_namei(ndp, fhp, len, slp, nam, mdp, dposp, p) rem = md->m_len; } if (*fromcp == '\0' || *fromcp == '/') { - error = EINVAL; + error = EACCES; goto out; } cnp->cn_hash += (unsigned char)*fromcp; @@ -964,26 +1431,24 @@ nfs_namei(ndp, fhp, len, slp, nam, mdp, dposp, p) if (len > 0) { if (rem >= len) *dposp += len; - else { - error = nfs_adv(mdp, dposp, len, rem); - if (error) - goto out; - } + else if (error = nfs_adv(mdp, dposp, len, rem)) + goto out; } ndp->ni_pathlen = tocp - cnp->cn_pnbuf; cnp->cn_nameptr = cnp->cn_pnbuf; /* * Extract and set starting directory. */ - error = nfsrv_fhtovp(fhp, FALSE, &dp, ndp->ni_cnd.cn_cred, slp, - nam, &rdonly); - if (error) + if (error = nfsrv_fhtovp(fhp, FALSE, &dp, ndp->ni_cnd.cn_cred, slp, + nam, &rdonly, kerbflag)) goto out; if (dp->v_type != VDIR) { nfsrv_vrele(dp); error = ENOTDIR; goto out; } + VREF(dp); + *retdirp = dp; ndp->ni_startdir = dp; if (rdonly) cnp->cn_flags |= (NOCROSSMOUNT | RDONLY); @@ -993,8 +1458,7 @@ nfs_namei(ndp, fhp, len, slp, nam, mdp, dposp, p) * And call lookup() to do the real work */ cnp->cn_proc = p; - error = lookup(ndp); - if (error) + if (error = lookup(ndp)) goto out; /* * Check for encountering a symbolic link @@ -1087,6 +1551,109 @@ nfsm_adj(mp, len, nul) } /* + * Make these functions instead of macros, so that the kernel text size + * doesn't get too big... + */ +void +nfsm_srvwcc(nfsd, before_ret, before_vap, after_ret, after_vap, mbp, bposp) + struct nfsrv_descript *nfsd; + int before_ret; + register struct vattr *before_vap; + int after_ret; + struct vattr *after_vap; + struct mbuf **mbp; + char **bposp; +{ + register struct mbuf *mb = *mbp, *mb2; + register char *bpos = *bposp; + register u_long *tl; + + if (before_ret) { + nfsm_build(tl, u_long *, NFSX_UNSIGNED); + *tl = nfs_false; + } else { + nfsm_build(tl, u_long *, 7 * NFSX_UNSIGNED); + *tl++ = nfs_true; + txdr_hyper(&(before_vap->va_size), tl); + tl += 2; + txdr_nfsv3time(&(before_vap->va_mtime), tl); + tl += 2; + txdr_nfsv3time(&(before_vap->va_ctime), tl); + } + *bposp = bpos; + *mbp = mb; + nfsm_srvpostopattr(nfsd, after_ret, after_vap, mbp, bposp); +} + +void +nfsm_srvpostopattr(nfsd, after_ret, after_vap, mbp, bposp) + struct nfsrv_descript *nfsd; + int after_ret; + struct vattr *after_vap; + struct mbuf **mbp; + char **bposp; +{ + register struct mbuf *mb = *mbp, *mb2; + register char *bpos = *bposp; + register u_long *tl; + register struct nfs_fattr *fp; + + if (after_ret) { + nfsm_build(tl, u_long *, NFSX_UNSIGNED); + *tl = nfs_false; + } else { + nfsm_build(tl, u_long *, NFSX_UNSIGNED + NFSX_V3FATTR); + *tl++ = nfs_true; + fp = (struct nfs_fattr *)tl; + nfsm_srvfattr(nfsd, after_vap, fp); + } + *mbp = mb; + *bposp = bpos; +} + +void +nfsm_srvfattr(nfsd, vap, fp) + register struct nfsrv_descript *nfsd; + register struct vattr *vap; + register struct nfs_fattr *fp; +{ + + fp->fa_nlink = txdr_unsigned(vap->va_nlink); + fp->fa_uid = txdr_unsigned(vap->va_uid); + fp->fa_gid = txdr_unsigned(vap->va_gid); + if (nfsd->nd_flag & ND_NFSV3) { + fp->fa_type = vtonfsv3_type(vap->va_type); + fp->fa_mode = vtonfsv3_mode(vap->va_mode); + txdr_hyper(&vap->va_size, &fp->fa3_size); + txdr_hyper(&vap->va_bytes, &fp->fa3_used); + fp->fa3_rdev.specdata1 = txdr_unsigned(major(vap->va_rdev)); + fp->fa3_rdev.specdata2 = txdr_unsigned(minor(vap->va_rdev)); + fp->fa3_fsid.nfsuquad[0] = 0; + fp->fa3_fsid.nfsuquad[1] = txdr_unsigned(vap->va_fsid); + fp->fa3_fileid.nfsuquad[0] = 0; + fp->fa3_fileid.nfsuquad[1] = txdr_unsigned(vap->va_fileid); + txdr_nfsv3time(&vap->va_atime, &fp->fa3_atime); + txdr_nfsv3time(&vap->va_mtime, &fp->fa3_mtime); + txdr_nfsv3time(&vap->va_ctime, &fp->fa3_ctime); + } else { + fp->fa_type = vtonfsv2_type(vap->va_type); + fp->fa_mode = vtonfsv2_mode(vap->va_type, vap->va_mode); + fp->fa2_size = txdr_unsigned(vap->va_size); + fp->fa2_blocksize = txdr_unsigned(vap->va_blocksize); + if (vap->va_type == VFIFO) + fp->fa2_rdev = 0xffffffff; + else + fp->fa2_rdev = txdr_unsigned(vap->va_rdev); + fp->fa2_blocks = txdr_unsigned(vap->va_bytes / NFS_FABLKSIZE); + fp->fa2_fsid = txdr_unsigned(vap->va_fsid); + fp->fa2_fileid = txdr_unsigned(vap->va_fileid); + txdr_nfsv2time(&vap->va_atime, &fp->fa2_atime); + txdr_nfsv2time(&vap->va_mtime, &fp->fa2_mtime); + txdr_nfsv2time(&vap->va_ctime, &fp->fa2_ctime); + } +} + +/* * nfsrv_fhtovp() - convert a fh to a vnode ptr (optionally locked) * - look up fsid in mount list (if not found ret error) * - get vp and export rights by calling VFS_FHTOVP() @@ -1094,7 +1661,7 @@ nfsm_adj(mp, len, nul) * - if not lockflag unlock it with VOP_UNLOCK() */ int -nfsrv_fhtovp(fhp, lockflag, vpp, cred, slp, nam, rdonlyp) +nfsrv_fhtovp(fhp, lockflag, vpp, cred, slp, nam, rdonlyp, kerbflag) fhandle_t *fhp; int lockflag; struct vnode **vpp; @@ -1102,6 +1669,7 @@ nfsrv_fhtovp(fhp, lockflag, vpp, cred, slp, nam, rdonlyp) struct nfssvc_sock *slp; struct mbuf *nam; int *rdonlyp; + int kerbflag; { register struct mount *mp; register struct nfsuid *uidp; @@ -1120,19 +1688,13 @@ nfsrv_fhtovp(fhp, lockflag, vpp, cred, slp, nam, rdonlyp) * Check/setup credentials. */ if (exflags & MNT_EXKERB) { - for (uidp = NUIDHASH(slp, cred->cr_uid)->lh_first; uidp != 0; - uidp = uidp->nu_hash.le_next) { - if (uidp->nu_uid == cred->cr_uid) - break; - } - if (uidp == 0) { + if (!kerbflag) { vput(*vpp); - return (NQNFS_AUTHERR); + return (NFSERR_AUTHERR | AUTH_TOOWEAK); } - cred->cr_uid = uidp->nu_cr.cr_uid; - for (i = 0; i < uidp->nu_cr.cr_ngroups; i++) - cred->cr_groups[i] = uidp->nu_cr.cr_groups[i]; - cred->cr_ngroups = i; + } else if (kerbflag) { + vput(*vpp); + return (NFSERR_AUTHERR | AUTH_TOOWEAK); } else if (cred->cr_uid == 0 || (exflags & MNT_EXPORTANON)) { cred->cr_uid = credanon->cr_uid; for (i = 0; i < credanon->cr_ngroups && i < NGROUPS; i++) @@ -1194,6 +1756,145 @@ netaddr_match(family, haddr, nam) return (0); } +static nfsuint64 nfs_nullcookie = { 0, 0 }; +/* + * This function finds the directory cookie that corresponds to the + * logical byte offset given. + */ +nfsuint64 * +nfs_getcookie(np, off, add) + register struct nfsnode *np; + off_t off; + int add; +{ + register struct nfsdmap *dp, *dp2; + register int pos; + + pos = off / NFS_DIRBLKSIZ; + if (pos == 0) { +#ifdef DIAGNOSTIC + if (add) + panic("nfs getcookie add at 0"); +#endif + return (&nfs_nullcookie); + } + pos--; + dp = np->n_cookies.lh_first; + if (!dp) { + if (add) { + MALLOC(dp, struct nfsdmap *, sizeof (struct nfsdmap), + M_NFSDIROFF, M_WAITOK); + dp->ndm_eocookie = 0; + LIST_INSERT_HEAD(&np->n_cookies, dp, ndm_list); + } else + return ((nfsuint64 *)0); + } + while (pos >= NFSNUMCOOKIES) { + pos -= NFSNUMCOOKIES; + if (dp->ndm_list.le_next) { + if (!add && dp->ndm_eocookie < NFSNUMCOOKIES && + pos >= dp->ndm_eocookie) + return ((nfsuint64 *)0); + dp = dp->ndm_list.le_next; + } else if (add) { + MALLOC(dp2, struct nfsdmap *, sizeof (struct nfsdmap), + M_NFSDIROFF, M_WAITOK); + dp2->ndm_eocookie = 0; + LIST_INSERT_AFTER(dp, dp2, ndm_list); + dp = dp2; + } else + return ((nfsuint64 *)0); + } + if (pos >= dp->ndm_eocookie) { + if (add) + dp->ndm_eocookie = pos + 1; + else + return ((nfsuint64 *)0); + } + return (&dp->ndm_cookies[pos]); +} + +/* + * Invalidate cached directory information, except for the actual directory + * blocks (which are invalidated separately). + * Done mainly to avoid the use of stale offset cookies. + */ +void +nfs_invaldir(vp) + register struct vnode *vp; +{ + register struct nfsnode *np = VTONFS(vp); + +#ifdef DIAGNOSTIC + if (vp->v_type != VDIR) + panic("nfs: invaldir not dir"); +#endif + np->n_direofoffset = 0; + np->n_cookieverf.nfsuquad[0] = 0; + np->n_cookieverf.nfsuquad[1] = 0; + if (np->n_cookies.lh_first) + np->n_cookies.lh_first->ndm_eocookie = 0; +} + +/* + * The write verifier has changed (probably due to a server reboot), so all + * B_NEEDCOMMIT blocks will have to be written again. Since they are on the + * dirty block list as B_DELWRI, all this takes is clearing the B_NEEDCOMMIT + * flag. Once done the new write verifier can be set for the mount point. + */ +void +nfs_clearcommit(mp) + struct mount *mp; +{ + register struct vnode *vp, *nvp; + register struct buf *bp, *nbp; + int s; + + s = splbio(); +loop: + for (vp = mp->mnt_vnodelist.lh_first; vp; vp = nvp) { + if (vp->v_mount != mp) /* Paranoia */ + goto loop; + nvp = vp->v_mntvnodes.le_next; + for (bp = vp->v_dirtyblkhd.lh_first; bp; bp = nbp) { + nbp = bp->b_vnbufs.le_next; + if ((bp->b_flags & (B_BUSY | B_DELWRI | B_NEEDCOMMIT)) + == (B_DELWRI | B_NEEDCOMMIT)) + bp->b_flags &= ~B_NEEDCOMMIT; + } + } + splx(s); +} + +/* + * Map errnos to NFS error numbers. For Version 3 also filter out error + * numbers not specified for the associated procedure. + */ +int +nfsrv_errmap(nd, err) + struct nfsrv_descript *nd; + register int err; +{ + register short *defaulterrp, *errp; + + if (nd->nd_flag & ND_NFSV3) { + if (nd->nd_procnum <= NFSPROC_COMMIT) { + errp = defaulterrp = nfsrv_v3errmap[nd->nd_procnum]; + while (*++errp) { + if (*errp == err) + return (err); + else if (*errp > err) + break; + } + return ((int)*defaulterrp); + } else + return (err & 0xffff); + } + if (err <= ELAST) + return ((int)nfsrv_v2errmap[err - 1]); + return (NFSERR_IO); +} + int nfsrv_vmio( struct vnode *vp) { vm_object_t object; diff --git a/sys/nfsserver/nfs_syscalls.c b/sys/nfsserver/nfs_syscalls.c index f7e159f..35a50f4 100644 --- a/sys/nfsserver/nfs_syscalls.c +++ b/sys/nfsserver/nfs_syscalls.c @@ -34,7 +34,7 @@ * SUCH DAMAGE. * * @(#)nfs_syscalls.c 8.3 (Berkeley) 1/4/94 - * $Id: nfs_syscalls.c,v 1.5 1994/10/17 17:47:38 phk Exp $ + * $Id: nfs_syscalls.c,v 1.6 1995/05/30 08:12:45 rgrimes Exp $ */ #include <sys/param.h> @@ -61,9 +61,11 @@ #ifdef ISO #include <netiso/iso.h> #endif +#include <nfs/xdr_subs.h> #include <nfs/rpcv2.h> -#include <nfs/nfsv2.h> +#include <nfs/nfsproto.h> #include <nfs/nfs.h> +#include <nfs/nfsm_subs.h> #include <nfs/nfsrvcache.h> #include <nfs/nfsmount.h> #include <nfs/nfsnode.h> @@ -73,13 +75,14 @@ void nfsrv_zapsock __P((struct nfssvc_sock *)); /* Global defs. */ -extern u_long nfs_prog, nfs_vers; -extern int (*nfsrv_procs[NFS_NPROCS])(); +extern int (*nfsrv3_procs[NFS_NPROCS])(); extern struct proc *nfs_iodwant[NFS_MAXASYNCDAEMON]; extern int nfs_numasync; extern time_t nqnfsstarttime; extern int nqsrv_writeslack; extern int nfsrtton; +extern struct nfsstats nfsstats; +extern int nfsrvw_procrastinate; struct nfssvc_sock *nfs_udpsock, *nfs_cltpsock; int nuidhash_max = NFS_MAXUIDHASH; static int nfs_numnfsd = 0; @@ -225,19 +228,25 @@ nfssvc(p, uap, retval) if (error) return (error); if ((uap->flag & NFSSVC_AUTHIN) && ((nfsd = nsd->nsd_nfsd)) && - (nfsd->nd_slp->ns_flag & SLP_VALID)) { - slp = nfsd->nd_slp; + (nfsd->nfsd_slp->ns_flag & SLP_VALID)) { + slp = nfsd->nfsd_slp; /* * First check to see if another nfsd has already * added this credential. */ - for (nuidp = NUIDHASH(slp, nsd->nsd_uid)->lh_first; + for (nuidp = NUIDHASH(slp,nsd->nsd_cr.cr_uid)->lh_first; nuidp != 0; nuidp = nuidp->nu_hash.le_next) { - if (nuidp->nu_uid == nsd->nsd_uid) + if (nuidp->nu_cr.cr_uid == nsd->nsd_cr.cr_uid && + (!nfsd->nfsd_nd->nd_nam2 || + netaddr_match(NU_NETFAM(nuidp), + &nuidp->nu_haddr, nfsd->nfsd_nd->nd_nam2))) break; } - if (!nuidp) { + if (nuidp) { + nfsrv_setcred(&nuidp->nu_cr,&nfsd->nfsd_nd->nd_cr); + nfsd->nfsd_nd->nd_flag |= ND_KERBFULL; + } else { /* * Nope, so we will. */ @@ -257,22 +266,53 @@ nfssvc(p, uap, retval) LIST_REMOVE(nuidp, nu_hash); TAILQ_REMOVE(&slp->ns_uidlruhead, nuidp, nu_lru); + if (nuidp->nu_flag & NU_NAM) + m_freem(nuidp->nu_nam); } + nuidp->nu_flag = 0; nuidp->nu_cr = nsd->nsd_cr; if (nuidp->nu_cr.cr_ngroups > NGROUPS) - nuidp->nu_cr.cr_ngroups = NGROUPS; + nuidp->nu_cr.cr_ngroups = NGROUPS; nuidp->nu_cr.cr_ref = 1; - nuidp->nu_uid = nsd->nsd_uid; + nuidp->nu_timestamp = nsd->nsd_timestamp; + nuidp->nu_expire = time.tv_sec + nsd->nsd_ttl; + /* + * and save the session key in nu_key. + */ + bcopy(nsd->nsd_key, nuidp->nu_key, + sizeof (nsd->nsd_key)); + if (nfsd->nfsd_nd->nd_nam2) { + struct sockaddr_in *saddr; + + saddr = mtod(nfsd->nfsd_nd->nd_nam2, + struct sockaddr_in *); + switch (saddr->sin_family) { + case AF_INET: + nuidp->nu_flag |= NU_INETADDR; + nuidp->nu_inetaddr = + saddr->sin_addr.s_addr; + break; + case AF_ISO: + default: + nuidp->nu_flag |= NU_NAM; + nuidp->nu_nam = m_copym( + nfsd->nfsd_nd->nd_nam2, 0, + M_COPYALL, M_WAIT); + break; + }; + } TAILQ_INSERT_TAIL(&slp->ns_uidlruhead, nuidp, nu_lru); LIST_INSERT_HEAD(NUIDHASH(slp, nsd->nsd_uid), nuidp, nu_hash); - + nfsrv_setcred(&nuidp->nu_cr, + &nfsd->nfsd_nd->nd_cr); + nfsd->nfsd_nd->nd_flag |= ND_KERBFULL; } } } if ((uap->flag & NFSSVC_AUTHINFAIL) && (nfsd = nsd->nsd_nfsd)) - nfsd->nd_flag |= NFSD_AUTHFAIL; + nfsd->nfsd_flag |= NFSD_AUTHFAIL; error = nfssvc_nfsd(nsd, uap->argp, p); } if (error == EINTR || error == ERESTART) @@ -353,8 +393,6 @@ nfssvc_addsock(fp, mynam) slp = (struct nfssvc_sock *) malloc(sizeof (struct nfssvc_sock), M_NFSSVC, M_WAITOK); bzero((caddr_t)slp, sizeof (struct nfssvc_sock)); - slp->ns_uidhashtbl = - hashinit(NUIDHASHSIZ, M_NFSSVC, &slp->ns_uidhash); TAILQ_INIT(&slp->ns_uidlruhead); TAILQ_INSERT_TAIL(&nfssvc_sockhead, slp, ns_chain); } @@ -381,44 +419,47 @@ nfssvc_nfsd(nsd, argp, p) caddr_t argp; struct proc *p; { - register struct mbuf *m, *nam2; + register struct mbuf *m; register int siz; register struct nfssvc_sock *slp; register struct socket *so; register int *solockp; - struct nfsd *nd = nsd->nsd_nfsd; - struct mbuf *mreq, *nam; - struct timeval starttime; + struct nfsd *nfsd = nsd->nsd_nfsd; + struct nfsrv_descript *nd = NULL; + struct mbuf *mreq; struct nfsuid *uidp; - int error = 0, cacherep, s; - int sotype; + int error = 0, cacherep, s, sotype, writes_todo; + u_quad_t cur_usec; +#ifndef nolint + cacherep = RC_DOIT; + writes_todo = 0; +#endif s = splnet(); - if (nd == (struct nfsd *)0) { - nsd->nsd_nfsd = nd = (struct nfsd *) + if (nfsd == (struct nfsd *)0) { + nsd->nsd_nfsd = nfsd = (struct nfsd *) malloc(sizeof (struct nfsd), M_NFSD, M_WAITOK); - bzero((caddr_t)nd, sizeof (struct nfsd)); - nd->nd_procp = p; - nd->nd_cr.cr_ref = 1; - TAILQ_INSERT_TAIL(&nfsd_head, nd, nd_chain); - nd->nd_nqlflag = NQL_NOVAL; + bzero((caddr_t)nfsd, sizeof (struct nfsd)); + nfsd->nfsd_procp = p; + TAILQ_INSERT_TAIL(&nfsd_head, nfsd, nfsd_chain); nfs_numnfsd++; } /* * Loop getting rpc requests until SIGKILL. */ for (;;) { - if ((nd->nd_flag & NFSD_REQINPROG) == 0) { - while (nd->nd_slp == (struct nfssvc_sock *)0 && + if ((nfsd->nfsd_flag & NFSD_REQINPROG) == 0) { + while (nfsd->nfsd_slp == (struct nfssvc_sock *)0 && (nfsd_head_flag & NFSD_CHECKSLP) == 0) { - nd->nd_flag |= NFSD_WAITING; + nfsd->nfsd_flag |= NFSD_WAITING; nfsd_waiting++; - error = tsleep((caddr_t)nd, PSOCK | PCATCH, "nfsd", 0); + error = tsleep((caddr_t)nfsd, PSOCK | PCATCH, + "nfsd", 0); nfsd_waiting--; if (error) goto done; } - if (nd->nd_slp == (struct nfssvc_sock *)0 && + if (nfsd->nfsd_slp == (struct nfssvc_sock *)0 && (nfsd_head_flag & NFSD_CHECKSLP) != 0) { for (slp = nfssvc_sockhead.tqh_first; slp != 0; slp = slp->ns_chain.tqe_next) { @@ -426,14 +467,14 @@ nfssvc_nfsd(nsd, argp, p) == (SLP_VALID | SLP_DOREC)) { slp->ns_flag &= ~SLP_DOREC; slp->ns_sref++; - nd->nd_slp = slp; + nfsd->nfsd_slp = slp; break; } } if (slp == 0) nfsd_head_flag &= ~NFSD_CHECKSLP; } - if ((slp = nd->nd_slp) == (struct nfssvc_sock *)0) + if ((slp = nfsd->nfsd_slp) == (struct nfssvc_sock *)0) continue; if (slp->ns_flag & SLP_VALID) { if (slp->ns_flag & SLP_DISCONN) @@ -446,86 +487,78 @@ nfssvc_nfsd(nsd, argp, p) M_WAIT); nfs_sndunlock(&slp->ns_solock); } - error = nfsrv_dorec(slp, nd); - nd->nd_flag |= NFSD_REQINPROG; + error = nfsrv_dorec(slp, nfsd, &nd); + cur_usec = (u_quad_t)time.tv_sec * 1000000 + + (u_quad_t)time.tv_usec; + if (error && slp->ns_tq.lh_first && + slp->ns_tq.lh_first->nd_time <= cur_usec) { + error = 0; + cacherep = RC_DOIT; + writes_todo = 1; + } else + writes_todo = 0; + nfsd->nfsd_flag |= NFSD_REQINPROG; } } else { error = 0; - slp = nd->nd_slp; + slp = nfsd->nfsd_slp; } if (error || (slp->ns_flag & SLP_VALID) == 0) { - nd->nd_slp = (struct nfssvc_sock *)0; - nd->nd_flag &= ~NFSD_REQINPROG; + if (nd) { + free((caddr_t)nd, M_NFSRVDESC); + nd = NULL; + } + nfsd->nfsd_slp = (struct nfssvc_sock *)0; + nfsd->nfsd_flag &= ~NFSD_REQINPROG; nfsrv_slpderef(slp); continue; } splx(s); so = slp->ns_so; sotype = so->so_type; - starttime = time; if (so->so_proto->pr_flags & PR_CONNREQUIRED) solockp = &slp->ns_solock; else solockp = (int *)0; - /* - * nam == nam2 for connectionless protocols such as UDP - * nam2 == NULL for connection based protocols to disable - * recent request caching. - */ - nam2 = nd->nd_nam; - if (nam2) { - nam = nam2; - cacherep = RC_CHECKIT; - } else { - nam = slp->ns_nam; - cacherep = RC_DOIT; - } - - /* - * Check to see if authorization is needed. - */ - if (nd->nd_flag & NFSD_NEEDAUTH) { - static int logauth = 0; + if (nd) { + nd->nd_starttime = time; + if (nd->nd_nam2) + nd->nd_nam = nd->nd_nam2; + else + nd->nd_nam = slp->ns_nam; - nd->nd_flag &= ~NFSD_NEEDAUTH; - /* - * Check for a mapping already installed. - */ - for (uidp = NUIDHASH(slp, nd->nd_cr.cr_uid)->lh_first; - uidp != 0; uidp = uidp->nu_hash.le_next) { - if (uidp->nu_uid == nd->nd_cr.cr_uid) - break; - } - if (!uidp) { - nsd->nsd_uid = nd->nd_cr.cr_uid; - if (nam2 && logauth++ == 0) - log(LOG_WARNING, "Kerberized NFS using UDP\n"); - nsd->nsd_haddr = - mtod(nam, struct sockaddr_in *)->sin_addr.s_addr; - nsd->nsd_authlen = nd->nd_authlen; - if (copyout(nd->nd_authstr, nsd->nsd_authstr, - nd->nd_authlen) == 0 && - copyout((caddr_t)nsd, argp, sizeof (*nsd)) == 0) - return (ENEEDAUTH); - cacherep = RC_DROPIT; - } - } - if (cacherep == RC_CHECKIT) - cacherep = nfsrv_getcache(nam2, nd, &mreq); + /* + * Check to see if authorization is needed. + */ + if (nfsd->nfsd_flag & NFSD_NEEDAUTH) { + nfsd->nfsd_flag &= ~NFSD_NEEDAUTH; + nsd->nsd_haddr = mtod(nd->nd_nam, + struct sockaddr_in *)->sin_addr.s_addr; + nsd->nsd_authlen = nfsd->nfsd_authlen; + nsd->nsd_verflen = nfsd->nfsd_verflen; + if (!copyout(nfsd->nfsd_authstr,nsd->nsd_authstr, + nfsd->nfsd_authlen) && + !copyout(nfsd->nfsd_verfstr, nsd->nsd_verfstr, + nfsd->nfsd_verflen) && + !copyout((caddr_t)nsd, argp, sizeof (*nsd))) + return (ENEEDAUTH); + cacherep = RC_DROPIT; + } else + cacherep = nfsrv_getcache(nd, slp, &mreq); - /* - * Check for just starting up for NQNFS and send - * fake "try again later" replies to the NQNFS clients. - */ - if (notstarted && nqnfsstarttime <= time.tv_sec) { + /* + * Check for just starting up for NQNFS and send + * fake "try again later" replies to the NQNFS clients. + */ + if (notstarted && nqnfsstarttime <= time.tv_sec) { if (modify_flag) { nqnfsstarttime = time.tv_sec + nqsrv_writeslack; modify_flag = 0; } else notstarted = 0; - } - if (notstarted) { - if (nd->nd_nqlflag == NQL_NOVAL) + } + if (notstarted) { + if ((nd->nd_flag & ND_NQNFS) == 0) cacherep = RC_DROPIT; else if (nd->nd_procnum != NFSPROC_WRITE) { nd->nd_procnum = NFSPROC_NOOP; @@ -533,36 +566,42 @@ nfssvc_nfsd(nsd, argp, p) cacherep = RC_DOIT; } else modify_flag = 1; - } else if (nd->nd_flag & NFSD_AUTHFAIL) { - nd->nd_flag &= ~NFSD_AUTHFAIL; + } else if (nfsd->nfsd_flag & NFSD_AUTHFAIL) { + nfsd->nfsd_flag &= ~NFSD_AUTHFAIL; nd->nd_procnum = NFSPROC_NOOP; - nd->nd_repstat = NQNFS_AUTHERR; + nd->nd_repstat = (NFSERR_AUTHERR | AUTH_TOOWEAK); cacherep = RC_DOIT; + } } - switch (cacherep) { - case RC_DOIT: - error = (*(nfsrv_procs[nd->nd_procnum]))(nd, - nd->nd_mrep, nd->nd_md, nd->nd_dpos, &nd->nd_cr, - nam, &mreq); - if (nd->nd_cr.cr_ref != 1) { - printf("nfssvc cref=%d\n", nd->nd_cr.cr_ref); - panic("nfssvc cref"); - } + /* + * Loop to get all the write rpc relies that have been + * gathered together. + */ + do { + switch (cacherep) { + case RC_DOIT: + if (writes_todo || (nd->nd_procnum == NFSPROC_WRITE && + nfsrvw_procrastinate > 0 && !notstarted)) + error = nfsrv_writegather(&nd, slp, + nfsd->nfsd_procp, &mreq); + else + error = (*(nfsrv3_procs[nd->nd_procnum]))(nd, + slp, nfsd->nfsd_procp, &mreq); + if (mreq == NULL) + break; if (error) { if (nd->nd_procnum != NQNFSPROC_VACATED) nfsstats.srv_errs++; - if (nam2) { - nfsrv_updatecache(nam2, nd, FALSE, mreq); - m_freem(nam2); - } + nfsrv_updatecache(nd, FALSE, mreq); + if (nd->nd_nam2) + m_freem(nd->nd_nam2); break; } nfsstats.srvrpccnt[nd->nd_procnum]++; - if (nam2) - nfsrv_updatecache(nam2, nd, TRUE, mreq); + nfsrv_updatecache(nd, TRUE, mreq); nd->nd_mrep = (struct mbuf *)0; - case RC_REPLY: + case RC_REPLY: m = mreq; siz = 0; while (m) { @@ -587,15 +626,15 @@ nfssvc_nfsd(nsd, argp, p) if (solockp) (void) nfs_sndlock(solockp, (struct nfsreq *)0); if (slp->ns_flag & SLP_VALID) - error = nfs_send(so, nam2, m, (struct nfsreq *)0); + error = nfs_send(so, nd->nd_nam2, m, NULL); else { error = EPIPE; m_freem(m); } if (nfsrtton) - nfsd_rt(&starttime, sotype, nd, nam, cacherep); - if (nam2) - MFREE(nam2, m); + nfsd_rt(sotype, nd, cacherep); + if (nd->nd_nam2) + MFREE(nd->nd_nam2, m); if (nd->nd_mrep) m_freem(nd->nd_mrep); if (error == EPIPE) @@ -603,29 +642,50 @@ nfssvc_nfsd(nsd, argp, p) if (solockp) nfs_sndunlock(solockp); if (error == EINTR || error == ERESTART) { + free((caddr_t)nd, M_NFSRVDESC); nfsrv_slpderef(slp); s = splnet(); goto done; } break; - case RC_DROPIT: + case RC_DROPIT: if (nfsrtton) - nfsd_rt(&starttime, sotype, nd, nam, cacherep); + nfsd_rt(sotype, nd, cacherep); m_freem(nd->nd_mrep); - m_freem(nam2); + m_freem(nd->nd_nam2); break; - }; + }; + if (nd) { + FREE((caddr_t)nd, M_NFSRVDESC); + nd = NULL; + } + + /* + * Check to see if there are outstanding writes that + * need to be serviced. + */ + cur_usec = (u_quad_t)time.tv_sec * 1000000 + + (u_quad_t)time.tv_usec; + s = splsoftclock(); + if (slp->ns_tq.lh_first && + slp->ns_tq.lh_first->nd_time <= cur_usec) { + cacherep = RC_DOIT; + writes_todo = 1; + } else + writes_todo = 0; + splx(s); + } while (writes_todo); s = splnet(); - if (nfsrv_dorec(slp, nd)) { - nd->nd_flag &= ~NFSD_REQINPROG; - nd->nd_slp = (struct nfssvc_sock *)0; + if (nfsrv_dorec(slp, nfsd, &nd)) { + nfsd->nfsd_flag &= ~NFSD_REQINPROG; + nfsd->nfsd_slp = NULL; nfsrv_slpderef(slp); } } done: - TAILQ_REMOVE(&nfsd_head, nd, nd_chain); + TAILQ_REMOVE(&nfsd_head, nfsd, nfsd_chain); splx(s); - free((caddr_t)nd, M_NFSD); + free((caddr_t)nfsd, M_NFSD); nsd->nsd_nfsd = (struct nfsd *)0; if (--nfs_numnfsd == 0) nfsrv_init(TRUE); /* Reinitialize everything */ @@ -641,9 +701,10 @@ int nfssvc_iod(p) struct proc *p; { - register struct buf *bp; + register struct buf *bp, *nbp; register int i, myiod; - int error = 0; + struct vnode *vp; + int error = 0, s; /* * Assign my position or return error if too many already running @@ -662,24 +723,53 @@ nfssvc_iod(p) * Just loop around doin our stuff until SIGKILL */ for (;;) { - while (nfs_bufq.tqh_first == NULL && error == 0) { - nfs_iodwant[myiod] = p; - error = tsleep((caddr_t)&nfs_iodwant[myiod], - PWAIT | PCATCH, "nfsidl", 0); - } - while ((bp = nfs_bufq.tqh_first) != NULL) { - /* Take one off the front of the list */ - TAILQ_REMOVE(&nfs_bufq, bp, b_freelist); - if (bp->b_flags & B_READ) - (void) nfs_doio(bp, bp->b_rcred, (struct proc *)0); - else - (void) nfs_doio(bp, bp->b_wcred, (struct proc *)0); - } - if (error) { - nfs_asyncdaemon[myiod] = 0; - nfs_numasync--; - return (error); - } + while (nfs_bufq.tqh_first == NULL && error == 0) { + nfs_iodwant[myiod] = p; + error = tsleep((caddr_t)&nfs_iodwant[myiod], + PWAIT | PCATCH, "nfsidl", 0); + } + while ((bp = nfs_bufq.tqh_first) != NULL) { + /* Take one off the front of the list */ + TAILQ_REMOVE(&nfs_bufq, bp, b_freelist); + if (bp->b_flags & B_READ) + (void) nfs_doio(bp, bp->b_rcred, (struct proc *)0); + else do { + /* + * Look for a delayed write for the same vnode, so I can do + * it now. We must grab it before calling nfs_doio() to + * avoid any risk of the vnode getting vclean()'d while + * we are doing the write rpc. + */ + vp = bp->b_vp; + s = splbio(); + for (nbp = vp->v_dirtyblkhd.lh_first; nbp; + nbp = nbp->b_vnbufs.le_next) { + if ((nbp->b_flags & + (B_BUSY|B_DELWRI|B_NEEDCOMMIT|B_NOCACHE))!=B_DELWRI) + continue; + bremfree(nbp); + vfs_busy_pages(nbp, 1); + nbp->b_flags |= (B_BUSY|B_ASYNC); + break; + } + splx(s); + /* + * For the delayed write, do the first part of nfs_bwrite() + * up to, but not including nfs_strategy(). + */ + if (nbp) { + nbp->b_flags &= ~(B_READ|B_DONE|B_ERROR|B_DELWRI); + reassignbuf(nbp, nbp->b_vp); + nbp->b_vp->v_numoutput++; + } + (void) nfs_doio(bp, bp->b_wcred, (struct proc *)0); + } while (bp = nbp); + } + if (error) { + nfs_asyncdaemon[myiod] = 0; + nfs_numasync--; + return (error); + } } } @@ -695,9 +785,11 @@ nfsrv_zapsock(slp) register struct nfssvc_sock *slp; { register struct nfsuid *nuidp, *nnuidp; + register struct nfsrv_descript *nwp, *nnwp; struct socket *so; struct file *fp; struct mbuf *m; + int s; slp->ns_flag &= ~SLP_ALLFLAGS; fp = slp->ns_fp; @@ -716,8 +808,18 @@ nfsrv_zapsock(slp) nnuidp = nuidp->nu_lru.tqe_next; LIST_REMOVE(nuidp, nu_hash); TAILQ_REMOVE(&slp->ns_uidlruhead, nuidp, nu_lru); + if (nuidp->nu_flag & NU_NAM) + m_freem(nuidp->nu_nam); free((caddr_t)nuidp, M_NFSUID); } + s = splsoftclock(); + for (nwp = slp->ns_tq.lh_first; nwp; nwp = nnwp) { + nnwp = nwp->nd_tq.le_next; + LIST_REMOVE(nwp, nd_tq); + free((caddr_t)nwp, M_NFSRVDESC); + } + LIST_INIT(&slp->ns_tq); + splx(s); } } @@ -726,13 +828,15 @@ nfsrv_zapsock(slp) * on this mount point porpous out of the kernel and do it. */ int -nfs_getauth(nmp, rep, cred, auth_type, auth_str, auth_len) +nfs_getauth(nmp, rep, cred, auth_str, auth_len, verf_str, verf_len, key) register struct nfsmount *nmp; struct nfsreq *rep; struct ucred *cred; - int *auth_type; char **auth_str; int *auth_len; + char *verf_str; + int *verf_len; + NFSKERBKEY_T key; /* return session key */ { int error = 0; @@ -748,6 +852,9 @@ nfs_getauth(nmp, rep, cred, auth_type, auth_str, auth_len) } nmp->nm_flag &= ~(NFSMNT_WAITAUTH | NFSMNT_WANTAUTH); nmp->nm_authstr = *auth_str = (char *)malloc(RPCAUTH_MAXSIZ, M_TEMP, M_WAITOK); + nmp->nm_authlen = RPCAUTH_MAXSIZ; + nmp->nm_verfstr = verf_str; + nmp->nm_verflen = *verf_len; nmp->nm_authuid = cred->cr_uid; wakeup((caddr_t)&nmp->nm_authstr); @@ -766,8 +873,9 @@ nfs_getauth(nmp, rep, cred, auth_type, auth_str, auth_len) if (error) free((caddr_t)*auth_str, M_TEMP); else { - *auth_type = nmp->nm_authtype; *auth_len = nmp->nm_authlen; + *verf_len = nmp->nm_verflen; + bcopy((caddr_t)nmp->nm_key, (caddr_t)key, sizeof (key)); } nmp->nm_flag &= ~NFSMNT_HASAUTH; nmp->nm_flag |= NFSMNT_WAITAUTH; @@ -779,6 +887,149 @@ nfs_getauth(nmp, rep, cred, auth_type, auth_str, auth_len) } /* + * Get a nickname authenticator and verifier. + */ +int +nfs_getnickauth(nmp, cred, auth_str, auth_len, verf_str, verf_len) + struct nfsmount *nmp; + struct ucred *cred; + char **auth_str; + int *auth_len; + char *verf_str; + int verf_len; +{ + register struct nfsuid *nuidp; + register u_long *nickp, *verfp; + struct timeval ktvin, ktvout; + NFSKERBKEYSCHED_T keys; /* stores key schedule */ + +#ifdef DIAGNOSTIC + if (verf_len < (4 * NFSX_UNSIGNED)) + panic("nfs_getnickauth verf too small"); +#endif + for (nuidp = NMUIDHASH(nmp, cred->cr_uid)->lh_first; + nuidp != 0; nuidp = nuidp->nu_hash.le_next) { + if (nuidp->nu_cr.cr_uid == cred->cr_uid) + break; + } + if (!nuidp || nuidp->nu_expire < time.tv_sec) + return (EACCES); + + /* + * Move to the end of the lru list (end of lru == most recently used). + */ + TAILQ_REMOVE(&nmp->nm_uidlruhead, nuidp, nu_lru); + TAILQ_INSERT_TAIL(&nmp->nm_uidlruhead, nuidp, nu_lru); + + nickp = (u_long *)malloc(2 * NFSX_UNSIGNED, M_TEMP, M_WAITOK); + *nickp++ = txdr_unsigned(RPCAKN_NICKNAME); + *nickp = txdr_unsigned(nuidp->nu_nickname); + *auth_str = (char *)nickp; + *auth_len = 2 * NFSX_UNSIGNED; + + /* + * Now we must encrypt the verifier and package it up. + */ + verfp = (u_long *)verf_str; + *verfp++ = txdr_unsigned(RPCAKN_NICKNAME); + if (time.tv_sec > nuidp->nu_timestamp.tv_sec || + (time.tv_sec == nuidp->nu_timestamp.tv_sec && + time.tv_usec > nuidp->nu_timestamp.tv_usec)) + nuidp->nu_timestamp = time; + else + nuidp->nu_timestamp.tv_usec++; + ktvin.tv_sec = txdr_unsigned(nuidp->nu_timestamp.tv_sec); + ktvin.tv_usec = txdr_unsigned(nuidp->nu_timestamp.tv_usec); + + /* + * Now encrypt the timestamp verifier in ecb mode using the session + * key. + */ +#ifdef NFSKERB + XXX +#endif + + *verfp++ = ktvout.tv_sec; + *verfp++ = ktvout.tv_usec; + *verfp = 0; + return (0); +} + +/* + * Save the current nickname in a hash list entry on the mount point. + */ +int +nfs_savenickauth(nmp, cred, len, key, mdp, dposp, mrep) + register struct nfsmount *nmp; + struct ucred *cred; + int len; + NFSKERBKEY_T key; + struct mbuf **mdp; + char **dposp; + struct mbuf *mrep; +{ + register struct nfsuid *nuidp; + register u_long *tl; + register long t1; + struct mbuf *md = *mdp; + struct timeval ktvin, ktvout; + u_long nick; + NFSKERBKEYSCHED_T keys; + char *dpos = *dposp, *cp2; + int deltasec, error = 0; + + if (len == (3 * NFSX_UNSIGNED)) { + nfsm_dissect(tl, u_long *, 3 * NFSX_UNSIGNED); + ktvin.tv_sec = *tl++; + ktvin.tv_usec = *tl++; + nick = fxdr_unsigned(u_long, *tl); + + /* + * Decrypt the timestamp in ecb mode. + */ +#ifdef NFSKERB + XXX +#endif + ktvout.tv_sec = fxdr_unsigned(long, ktvout.tv_sec); + ktvout.tv_usec = fxdr_unsigned(long, ktvout.tv_usec); + deltasec = time.tv_sec - ktvout.tv_sec; + if (deltasec < 0) + deltasec = -deltasec; + /* + * If ok, add it to the hash list for the mount point. + */ + if (deltasec <= NFS_KERBCLOCKSKEW) { + if (nmp->nm_numuids < nuidhash_max) { + nmp->nm_numuids++; + nuidp = (struct nfsuid *) + malloc(sizeof (struct nfsuid), M_NFSUID, + M_WAITOK); + } else { + nuidp = nmp->nm_uidlruhead.tqh_first; + LIST_REMOVE(nuidp, nu_hash); + TAILQ_REMOVE(&nmp->nm_uidlruhead, nuidp, + nu_lru); + } + nuidp->nu_flag = 0; + nuidp->nu_cr.cr_uid = cred->cr_uid; + nuidp->nu_expire = time.tv_sec + NFS_KERBTTL; + nuidp->nu_timestamp = ktvout; + nuidp->nu_nickname = nick; + bcopy(key, nuidp->nu_key, sizeof (key)); + TAILQ_INSERT_TAIL(&nmp->nm_uidlruhead, nuidp, + nu_lru); + LIST_INSERT_HEAD(NMUIDHASH(nmp, cred->cr_uid), + nuidp, nu_hash); + } + } else + nfsm_adv(nfsm_rndup(len)); +nfsmout: + *mdp = md; + *dposp = dpos; + return (error); +} + +/* * Derefence a server socket structure. If it has no more references and * is no longer valid, you can throw it away. */ @@ -830,16 +1081,12 @@ nfsrv_init(terminating) nfs_udpsock = (struct nfssvc_sock *) malloc(sizeof (struct nfssvc_sock), M_NFSSVC, M_WAITOK); bzero((caddr_t)nfs_udpsock, sizeof (struct nfssvc_sock)); - nfs_udpsock->ns_uidhashtbl = - hashinit(NUIDHASHSIZ, M_NFSSVC, &nfs_udpsock->ns_uidhash); TAILQ_INIT(&nfs_udpsock->ns_uidlruhead); TAILQ_INSERT_HEAD(&nfssvc_sockhead, nfs_udpsock, ns_chain); nfs_cltpsock = (struct nfssvc_sock *) malloc(sizeof (struct nfssvc_sock), M_NFSSVC, M_WAITOK); bzero((caddr_t)nfs_cltpsock, sizeof (struct nfssvc_sock)); - nfs_cltpsock->ns_uidhashtbl = - hashinit(NUIDHASHSIZ, M_NFSSVC, &nfs_cltpsock->ns_uidhash); TAILQ_INIT(&nfs_cltpsock->ns_uidlruhead); TAILQ_INSERT_TAIL(&nfssvc_sockhead, nfs_cltpsock, ns_chain); } @@ -848,11 +1095,9 @@ nfsrv_init(terminating) * Add entries to the server monitor log. */ static void -nfsd_rt(startp, sotype, nd, nam, cacherep) - struct timeval *startp; +nfsd_rt(sotype, nd, cacherep) int sotype; - register struct nfsd *nd; - struct mbuf *nam; + register struct nfsrv_descript *nd; int cacherep; { register struct drt *rt; @@ -866,15 +1111,17 @@ nfsd_rt(startp, sotype, nd, nam, cacherep) rt->flag = DRT_CACHEDROP; if (sotype == SOCK_STREAM) rt->flag |= DRT_TCP; - if (nd->nd_nqlflag != NQL_NOVAL) + if (nd->nd_flag & ND_NQNFS) rt->flag |= DRT_NQNFS; + else if (nd->nd_flag & ND_NFSV3) + rt->flag |= DRT_NFSV3; rt->proc = nd->nd_procnum; - if (mtod(nam, struct sockaddr *)->sa_family == AF_INET) - rt->ipadr = mtod(nam, struct sockaddr_in *)->sin_addr.s_addr; + if (mtod(nd->nd_nam, struct sockaddr *)->sa_family == AF_INET) + rt->ipadr = mtod(nd->nd_nam, struct sockaddr_in *)->sin_addr.s_addr; else - rt->ipadr = INADDR_ANY; - rt->resptime = ((time.tv_sec - startp->tv_sec) * 1000000) + - (time.tv_usec - startp->tv_usec); + rt->ipadr = INADDR_ANY; + rt->resptime = ((time.tv_sec - nd->nd_starttime.tv_sec) * 1000000) + + (time.tv_usec - nd->nd_starttime.tv_usec); rt->tstamp = time; nfsdrt.pos = (nfsdrt.pos + 1) % NFSRTTLOGSIZ; } diff --git a/sys/nfsserver/nfsm_subs.h b/sys/nfsserver/nfsm_subs.h index 030a985..9e89c57 100644 --- a/sys/nfsserver/nfsm_subs.h +++ b/sys/nfsserver/nfsm_subs.h @@ -34,12 +34,13 @@ * SUCH DAMAGE. * * @(#)nfsm_subs.h 8.1 (Berkeley) 6/16/93 - * $Id: nfsm_subs.h,v 1.5 1994/10/17 17:47:42 phk Exp $ + * $Id: nfsm_subs.h,v 1.6 1995/05/30 08:12:51 rgrimes Exp $ */ #ifndef _NFS_NFSM_SUBS_H_ #define _NFS_NFSM_SUBS_H_ + /* * These macros do strange and peculiar things to mbuf chains for * the assistance of the nfs code. To attempt to use them for any @@ -90,49 +91,153 @@ extern struct mbuf *nfsm_reqh(); mb->m_len += (s); \ bpos += (s); } -#define nfsm_dissect(a,c,s) \ +#define nfsm_dissect(a, c, s) \ { t1 = mtod(md, caddr_t)+md->m_len-dpos; \ if (t1 >= (s)) { \ (a) = (c)(dpos); \ dpos += (s); \ + } else if (t1 = nfsm_disct(&md, &dpos, (s), t1, &cp2)) { \ + error = t1; \ + m_freem(mrep); \ + goto nfsmout; \ } else { \ - error = nfsm_disct(&md, &dpos, (s), t1, &cp2); \ - if (error) { \ - m_freem(mrep); \ + (a) = (c)cp2; \ + } } + +#define nfsm_fhtom(v, v3) \ + { if (v3) { \ + t2 = nfsm_rndup(VTONFS(v)->n_fhsize) + NFSX_UNSIGNED; \ + if (t2 <= M_TRAILINGSPACE(mb)) { \ + nfsm_build(tl, u_long *, t2); \ + *tl++ = txdr_unsigned(VTONFS(v)->n_fhsize); \ + *(tl + ((t2>>2) - 2)) = 0; \ + bcopy((caddr_t)VTONFS(v)->n_fhp,(caddr_t)tl, \ + VTONFS(v)->n_fhsize); \ + } else if (t2 = nfsm_strtmbuf(&mb, &bpos, \ + (caddr_t)VTONFS(v)->n_fhp, VTONFS(v)->n_fhsize)) { \ + error = t2; \ + m_freem(mreq); \ goto nfsmout; \ - } else { \ - (a) = (c)cp2; \ } \ + } else { \ + nfsm_build(cp, caddr_t, NFSX_V2FH); \ + bcopy((caddr_t)VTONFS(v)->n_fhp, cp, NFSX_V2FH); \ } } -#define nfsm_fhtom(v) \ - nfsm_build(cp,caddr_t,NFSX_FH); \ - bcopy((caddr_t)&(VTONFS(v)->n_fh), cp, NFSX_FH) +#define nfsm_srvfhtom(f, v3) \ + { if (v3) { \ + nfsm_build(tl, u_long *, NFSX_UNSIGNED + NFSX_V3FH); \ + *tl++ = txdr_unsigned(NFSX_V3FH); \ + bcopy((caddr_t)(f), (caddr_t)tl, NFSX_V3FH); \ + } else { \ + nfsm_build(cp, caddr_t, NFSX_V2FH); \ + bcopy((caddr_t)(f), cp, NFSX_V2FH); \ + } } -#define nfsm_srvfhtom(f) \ - nfsm_build(cp,caddr_t,NFSX_FH); \ - bcopy((caddr_t)(f), cp, NFSX_FH) +#define nfsm_srvpostop_fh(f) \ + { nfsm_build(tl, u_long *, 2 * NFSX_UNSIGNED + NFSX_V3FH); \ + *tl++ = nfs_true; \ + *tl++ = txdr_unsigned(NFSX_V3FH); \ + bcopy((caddr_t)(f), (caddr_t)tl, NFSX_V3FH); \ + } -#define nfsm_mtofh(d,v) \ - { struct nfsnode *np; nfsv2fh_t *fhp; \ - nfsm_dissect(fhp,nfsv2fh_t *,NFSX_FH); \ - error = nfs_nget((d)->v_mount, fhp, &np); \ - if (error) { \ - m_freem(mrep); \ - goto nfsmout; \ +#define nfsm_mtofh(d, v, v3, f) \ + { struct nfsnode *ttnp; nfsfh_t *ttfhp; int ttfhsize; \ + if (v3) { \ + nfsm_dissect(tl, u_long *, NFSX_UNSIGNED); \ + (f) = fxdr_unsigned(int, *tl); \ + } else \ + (f) = 1; \ + if (f) { \ + nfsm_getfh(ttfhp, ttfhsize, (v3)); \ + if (t1 = nfs_nget((d)->v_mount, ttfhp, ttfhsize, \ + &ttnp)) { \ + error = t1; \ + m_freem(mrep); \ + goto nfsmout; \ + } \ + (v) = NFSTOV(ttnp); \ + } \ + if (v3) { \ + nfsm_dissect(tl, u_long *, NFSX_UNSIGNED); \ + if (f) \ + (f) = fxdr_unsigned(int, *tl); \ + else if (fxdr_unsigned(int, *tl)) \ + nfsm_adv(NFSX_V3FATTR); \ } \ - (v) = NFSTOV(np); \ - nfsm_loadattr(v, (struct vattr *)0); \ + if (f) \ + nfsm_loadattr((v), (struct vattr *)0); \ } -#define nfsm_loadattr(v,a) \ - { struct vnode *tvp = (v); \ - error = nfs_loadattrcache(&tvp, &md, &dpos, (a)); \ - if (error) { \ +#define nfsm_getfh(f, s, v3) \ + { if (v3) { \ + nfsm_dissect(tl, u_long *, NFSX_UNSIGNED); \ + if (((s) = fxdr_unsigned(int, *tl)) <= 0 || \ + (s) > NFSX_V3FHMAX) { \ + m_freem(mrep); \ + error = EBADRPC; \ + goto nfsmout; \ + } \ + } else \ + (s) = NFSX_V2FH; \ + nfsm_dissect((f), nfsfh_t *, nfsm_rndup(s)); } + +#define nfsm_loadattr(v, a) \ + { struct vnode *ttvp = (v); \ + if (t1 = nfs_loadattrcache(&ttvp, &md, &dpos, (a))) { \ + error = t1; \ m_freem(mrep); \ goto nfsmout; \ } \ - (v) = tvp; } + (v) = ttvp; } + +#define nfsm_postop_attr(v, f) \ + { struct vnode *ttvp = (v); \ + nfsm_dissect(tl, u_long *, NFSX_UNSIGNED); \ + if ((f) = fxdr_unsigned(int, *tl)) { \ + if (t1 = nfs_loadattrcache(&ttvp, &md, &dpos, \ + (struct vattr *)0)) { \ + error = t1; \ + (f) = 0; \ + m_freem(mrep); \ + goto nfsmout; \ + } \ + (v) = ttvp; \ + } } + +/* Used as (f) for nfsm_wcc_data() */ +#define NFSV3_WCCRATTR 0 +#define NFSV3_WCCCHK 1 + +#define nfsm_wcc_data(v, f) \ + { int ttattrf, ttretf = 0; \ + nfsm_dissect(tl, u_long *, NFSX_UNSIGNED); \ + if (*tl == nfs_true) { \ + nfsm_dissect(tl, u_long *, 6 * NFSX_UNSIGNED); \ + if (f) \ + ttretf = (VTONFS(v)->n_mtime == \ + fxdr_unsigned(u_long, *(tl + 2))); \ + } \ + nfsm_postop_attr((v), ttattrf); \ + if (f) { \ + (f) = ttretf; \ + } else { \ + (f) = ttattrf; \ + } } + +#define nfsm_v3sattr(s, a, u, g) \ + { (s)->sa_modetrue = nfs_true; \ + (s)->sa_mode = vtonfsv3_mode((a)->va_mode); \ + (s)->sa_uidtrue = nfs_true; \ + (s)->sa_uid = txdr_unsigned(u); \ + (s)->sa_gidtrue = nfs_true; \ + (s)->sa_gid = txdr_unsigned(g); \ + (s)->sa_sizefalse = nfs_false; \ + (s)->sa_atimetype = txdr_unsigned(NFSV3SATTRTIME_TOCLIENT); \ + txdr_nfsv3time(&(a)->va_atime, &(s)->sa_atime); \ + (s)->sa_mtimetype = txdr_unsigned(NFSV3SATTRTIME_TOCLIENT); \ + txdr_nfsv3time(&(a)->va_mtime, &(s)->sa_mtime); \ + } #define nfsm_strsiz(s,m) \ { nfsm_dissect(tl,u_long *,NFSX_UNSIGNED); \ @@ -149,16 +254,27 @@ extern struct mbuf *nfsm_reqh(); nfsm_reply(0); \ } } +#define nfsm_srvnamesiz(s) \ + { nfsm_dissect(tl,u_long *,NFSX_UNSIGNED); \ + if (((s) = fxdr_unsigned(long,*tl)) > NFS_MAXNAMLEN) \ + error = NFSERR_NAMETOL; \ + if ((s) <= 0) \ + error = EBADRPC; \ + if (error) \ + nfsm_reply(0); \ + } + #define nfsm_mtouio(p,s) \ if ((s) > 0 && \ - (error = nfsm_mbuftouio(&md,(p),(s),&dpos))) { \ + (t1 = nfsm_mbuftouio(&md,(p),(s),&dpos))) { \ + error = t1; \ m_freem(mrep); \ goto nfsmout; \ } #define nfsm_uiotom(p,s) \ - error = nfsm_uiotombuf((p),&mb,(s),&bpos); \ - if (error) { \ + if (t1 = nfsm_uiotombuf((p),&mb,(s),&bpos)) { \ + error = t1; \ m_freem(mreq); \ goto nfsmout; \ } @@ -172,10 +288,13 @@ extern struct mbuf *nfsm_reqh(); #define nfsm_rndup(a) (((a)+3)&(~0x3)) #define nfsm_request(v, t, p, c) \ - error = nfs_request((v), mreq, (t), (p), \ - (c), &mrep, &md, &dpos); \ - if (error) \ - goto nfsmout + if (error = nfs_request((v), mreq, (t), (p), \ + (c), &mrep, &md, &dpos)) { \ + if (error & NFSERR_RETERR) \ + error &= ~NFSERR_RETERR; \ + else \ + goto nfsmout; \ + } #define nfsm_strtom(a,s,m) \ if ((s) > (m)) { \ @@ -189,12 +308,10 @@ extern struct mbuf *nfsm_reqh(); *tl++ = txdr_unsigned(s); \ *(tl+((t2>>2)-2)) = 0; \ bcopy((caddr_t)(a), (caddr_t)tl, (s)); \ - } else { \ - error = nfsm_strtmbuf(&mb, &bpos, (a), (s)); \ - if (error) { \ - m_freem(mreq); \ - goto nfsmout; \ - } \ + } else if (t2 = nfsm_strtmbuf(&mb, &bpos, (a), (s))) { \ + error = t2; \ + m_freem(mreq); \ + goto nfsmout; \ } #define nfsm_srvdone \ @@ -204,33 +321,53 @@ extern struct mbuf *nfsm_reqh(); #define nfsm_reply(s) \ { \ nfsd->nd_repstat = error; \ - if (error) \ - (void) nfs_rephead(0, nfsd, error, cache, &frev, \ + if (error && !(nfsd->nd_flag & ND_NFSV3)) \ + (void) nfs_rephead(0, nfsd, slp, error, cache, &frev, \ mrq, &mb, &bpos); \ else \ - (void) nfs_rephead((s), nfsd, error, cache, &frev, \ + (void) nfs_rephead((s), nfsd, slp, error, cache, &frev, \ mrq, &mb, &bpos); \ m_freem(mrep); \ mreq = *mrq; \ - if (error) \ + if (error && (!(nfsd->nd_flag & ND_NFSV3) || \ + error == EBADRPC)) \ return(0); \ } +#define nfsm_writereply(s, v3) \ + { \ + nfsd->nd_repstat = error; \ + if (error && !(v3)) \ + (void) nfs_rephead(0, nfsd, slp, error, cache, &frev, \ + &mreq, &mb, &bpos); \ + else \ + (void) nfs_rephead((s), nfsd, slp, error, cache, &frev, \ + &mreq, &mb, &bpos); \ + } + #define nfsm_adv(s) \ - t1 = mtod(md, caddr_t)+md->m_len-dpos; \ + { t1 = mtod(md, caddr_t)+md->m_len-dpos; \ if (t1 >= (s)) { \ dpos += (s); \ - } else { \ - error = nfs_adv(&md, &dpos, (s), t1); \ - if (error) { \ - m_freem(mrep); \ - goto nfsmout; \ - } \ - } + } else if (t1 = nfs_adv(&md, &dpos, (s), t1)) { \ + error = t1; \ + m_freem(mrep); \ + goto nfsmout; \ + } } #define nfsm_srvmtofh(f) \ - nfsm_dissect(tl, u_long *, NFSX_FH); \ - bcopy((caddr_t)tl, (caddr_t)f, NFSX_FH) + { if (nfsd->nd_flag & ND_NFSV3) { \ + nfsm_dissect(tl, u_long *, NFSX_UNSIGNED); \ + if (fxdr_unsigned(int, *tl) != NFSX_V3FH) { \ + error = EBADRPC; \ + nfsm_reply(0); \ + } \ + } \ + nfsm_dissect(tl, u_long *, NFSX_V3FH); \ + bcopy((caddr_t)tl, (caddr_t)(f), NFSX_V3FH); \ + if ((nfsd->nd_flag & ND_NFSV3) == 0) \ + nfsm_adv(NFSX_V2FH - NFSX_V3FH); \ + } #define nfsm_clget \ if (bp >= be) { \ @@ -246,41 +383,57 @@ extern struct mbuf *nfsm_reqh(); } \ tl = (u_long *)bp -#define nfsm_srvfillattr \ - fp->fa_type = vtonfs_type(vap->va_type); \ - fp->fa_mode = vtonfs_mode(vap->va_type, vap->va_mode); \ - fp->fa_nlink = txdr_unsigned(vap->va_nlink); \ - fp->fa_uid = txdr_unsigned(vap->va_uid); \ - fp->fa_gid = txdr_unsigned(vap->va_gid); \ - if (nfsd->nd_nqlflag == NQL_NOVAL) { \ - fp->fa_nfsblocksize = txdr_unsigned(vap->va_blocksize); \ - if (vap->va_type == VFIFO) \ - fp->fa_nfsrdev = 0xffffffff; \ - else \ - fp->fa_nfsrdev = txdr_unsigned(vap->va_rdev); \ - fp->fa_nfsfsid = txdr_unsigned(vap->va_fsid); \ - fp->fa_nfsfileid = txdr_unsigned(vap->va_fileid); \ - fp->fa_nfssize = txdr_unsigned(vap->va_size); \ - fp->fa_nfsblocks = txdr_unsigned(vap->va_bytes / NFS_FABLKSIZE); \ - txdr_nfstime(&vap->va_atime, &fp->fa_nfsatime); \ - txdr_nfstime(&vap->va_mtime, &fp->fa_nfsmtime); \ - txdr_nfstime(&vap->va_ctime, &fp->fa_nfsctime); \ - } else { \ - fp->fa_nqblocksize = txdr_unsigned(vap->va_blocksize); \ - if (vap->va_type == VFIFO) \ - fp->fa_nqrdev = 0xffffffff; \ - else \ - fp->fa_nqrdev = txdr_unsigned(vap->va_rdev); \ - fp->fa_nqfsid = txdr_unsigned(vap->va_fsid); \ - fp->fa_nqfileid = txdr_unsigned(vap->va_fileid); \ - txdr_hyper(&vap->va_size, &fp->fa_nqsize); \ - txdr_hyper(&vap->va_bytes, &fp->fa_nqbytes); \ - txdr_nqtime(&vap->va_atime, &fp->fa_nqatime); \ - txdr_nqtime(&vap->va_mtime, &fp->fa_nqmtime); \ - txdr_nqtime(&vap->va_ctime, &fp->fa_nqctime); \ - fp->fa_nqflags = txdr_unsigned(vap->va_flags); \ - fp->fa_nqgen = txdr_unsigned(vap->va_gen); \ - txdr_hyper(&vap->va_filerev, &fp->fa_nqfilerev); \ - } +#define nfsm_srvfillattr(a, f) \ + nfsm_srvfattr(nfsd, (a), (f)) + +#define nfsm_srvwcc_data(br, b, ar, a) \ + nfsm_srvwcc(nfsd, (br), (b), (ar), (a), &mb, &bpos) + +#define nfsm_srvpostop_attr(r, a) \ + nfsm_srvpostopattr(nfsd, (r), (a), &mb, &bpos) + +#define nfsm_srvsattr(a) \ + { nfsm_dissect(tl, u_long *, NFSX_UNSIGNED); \ + if (*tl == nfs_true) { \ + nfsm_dissect(tl, u_long *, NFSX_UNSIGNED); \ + (a)->va_mode = nfstov_mode(*tl); \ + } \ + nfsm_dissect(tl, u_long *, NFSX_UNSIGNED); \ + if (*tl == nfs_true) { \ + nfsm_dissect(tl, u_long *, NFSX_UNSIGNED); \ + (a)->va_uid = fxdr_unsigned(uid_t, *tl); \ + } \ + nfsm_dissect(tl, u_long *, NFSX_UNSIGNED); \ + if (*tl == nfs_true) { \ + nfsm_dissect(tl, u_long *, NFSX_UNSIGNED); \ + (a)->va_gid = fxdr_unsigned(gid_t, *tl); \ + } \ + nfsm_dissect(tl, u_long *, NFSX_UNSIGNED); \ + if (*tl == nfs_true) { \ + nfsm_dissect(tl, u_long *, 2 * NFSX_UNSIGNED); \ + fxdr_hyper(tl, &(a)->va_size); \ + } \ + nfsm_dissect(tl, u_long *, NFSX_UNSIGNED); \ + switch (fxdr_unsigned(int, *tl)) { \ + case NFSV3SATTRTIME_TOCLIENT: \ + nfsm_dissect(tl, u_long *, 2 * NFSX_UNSIGNED); \ + fxdr_nfsv3time(tl, &(a)->va_atime); \ + break; \ + case NFSV3SATTRTIME_TOSERVER: \ + (a)->va_atime.ts_sec = time.tv_sec; \ + (a)->va_atime.ts_nsec = time.tv_usec * 1000; \ + break; \ + }; \ + nfsm_dissect(tl, u_long *, NFSX_UNSIGNED); \ + switch (fxdr_unsigned(int, *tl)) { \ + case NFSV3SATTRTIME_TOCLIENT: \ + nfsm_dissect(tl, u_long *, 2 * NFSX_UNSIGNED); \ + fxdr_nfsv3time(tl, &(a)->va_mtime); \ + break; \ + case NFSV3SATTRTIME_TOSERVER: \ + (a)->va_mtime.ts_sec = time.tv_sec; \ + (a)->va_mtime.ts_nsec = time.tv_usec * 1000; \ + break; \ + }; } #endif diff --git a/sys/nfsserver/nfsproto.h b/sys/nfsserver/nfsproto.h new file mode 100644 index 0000000..ac1a090 --- /dev/null +++ b/sys/nfsserver/nfsproto.h @@ -0,0 +1,441 @@ +/* + * Copyright (c) 1989, 1993 + * The Regents of the University of California. All rights reserved. + * + * This code is derived from software contributed to Berkeley by + * Rick Macklem at The University of Guelph. + * + * Redistribution and use in source and binary forms, with or without + * modification, are permitted provided that the following conditions + * are met: + * 1. Redistributions of source code must retain the above copyright + * notice, this list of conditions and the following disclaimer. + * 2. Redistributions in binary form must reproduce the above copyright + * notice, this list of conditions and the following disclaimer in the + * documentation and/or other materials provided with the distribution. + * 3. All advertising materials mentioning features or use of this software + * must display the following acknowledgement: + * This product includes software developed by the University of + * California, Berkeley and its contributors. + * 4. Neither the name of the University nor the names of its contributors + * may be used to endorse or promote products derived from this software + * without specific prior written permission. + * + * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND + * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE + * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE + * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE + * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL + * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS + * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) + * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT + * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY + * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF + * SUCH DAMAGE. + * + * @(#)nfsproto.h 8.1 (Berkeley) 6/10/93 + * $Id: nfsproto.h,v 1.4 1994/08/21 06:50:13 paul Exp $ + */ + +#ifndef _NFS_NFSPROTO_H_ +#define _NFS_NFSPROTO_H_ + +/* + * nfs definitions as per the Version 2 and 3 specs + */ + +/* + * Constants as defined in the Sun NFS Version 2 and 3 specs. + * "NFS: Network File System Protocol Specification" RFC1094 + * and in the "NFS: Network File System Version 3 Protocol + * Specification" + */ + +#define NFS_PORT 2049 +#define NFS_PROG 100003 +#define NFS_VER2 2 +#define NFS_VER3 3 +#define NFS_V2MAXDATA 8192 +#define NFS_MAXDGRAMDATA 16384 +#define NFS_MAXDATA 32768 +#define NFS_MAXPATHLEN 1024 +#define NFS_MAXNAMLEN 255 +#define NFS_MAXPKTHDR 404 +#define NFS_MAXPACKET (NFS_MAXPKTHDR + NFS_MAXDATA) +#define NFS_MINPACKET 20 +#define NFS_FABLKSIZE 512 /* Size in bytes of a block wrt fa_blocks */ + +/* Stat numbers for rpc returns (version 2 and 3) */ +#define NFS_OK 0 +#define NFSERR_PERM 1 +#define NFSERR_NOENT 2 +#define NFSERR_IO 5 +#define NFSERR_NXIO 6 +#define NFSERR_ACCES 13 +#define NFSERR_EXIST 17 +#define NFSERR_XDEV 18 /* Version 3 only */ +#define NFSERR_NODEV 19 +#define NFSERR_NOTDIR 20 +#define NFSERR_ISDIR 21 +#define NFSERR_INVAL 22 /* Version 3 only */ +#define NFSERR_FBIG 27 +#define NFSERR_NOSPC 28 +#define NFSERR_ROFS 30 +#define NFSERR_MLINK 31 /* Version 3 only */ +#define NFSERR_NAMETOL 63 +#define NFSERR_NOTEMPTY 66 +#define NFSERR_DQUOT 69 +#define NFSERR_STALE 70 +#define NFSERR_REMOTE 71 /* Version 3 only */ +#define NFSERR_WFLUSH 99 /* Version 2 only */ +#define NFSERR_BADHANDLE 10001 /* The rest Version 3 only */ +#define NFSERR_NOT_SYNC 10002 +#define NFSERR_BAD_COOKIE 10003 +#define NFSERR_NOTSUPP 10004 +#define NFSERR_TOOSMALL 10005 +#define NFSERR_SERVERFAULT 10006 +#define NFSERR_BADTYPE 10007 +#define NFSERR_JUKEBOX 10008 +#define NFSERR_TRYLATER NFSERR_JUKEBOX +#define NFSERR_STALEWRITEVERF 30001 /* Fake return for nfs_commit() */ + +#define NFSERR_RETVOID 0x20000000 /* Return void, not error */ +#define NFSERR_AUTHERR 0x40000000 /* Mark an authentication error */ +#define NFSERR_RETERR 0x80000000 /* Mark an error return for V3 */ + +/* Sizes in bytes of various nfs rpc components */ +#define NFSX_UNSIGNED 4 + +/* specific to NFS Version 2 */ +#define NFSX_V2FH 32 +#define NFSX_V2FATTR 68 +#define NFSX_V2SATTR 32 +#define NFSX_V2COOKIE 4 +#define NFSX_V2STATFS 20 + +/* specific to NFS Version 3 */ +#define NFSX_V3FH (sizeof (fhandle_t)) /* size this server uses */ +#define NFSX_V3FHMAX 64 /* max. allowed by protocol */ +#define NFSX_V3FATTR 84 +#define NFSX_V3SATTR 60 /* max. all fields filled in */ +#define NFSX_V3SRVSATTR (sizeof (struct nfsv3_sattr)) +#define NFSX_V3POSTOPATTR (NFSX_V3FATTR + NFSX_UNSIGNED) +#define NFSX_V3WCCDATA (NFSX_V3POSTOPATTR + 8 * NFSX_UNSIGNED) +#define NFSX_V3COOKIEVERF 8 +#define NFSX_V3WRITEVERF 8 +#define NFSX_V3CREATEVERF 8 +#define NFSX_V3STATFS 52 +#define NFSX_V3FSINFO 48 +#define NFSX_V3PATHCONF 24 + +/* variants for both versions */ +#define NFSX_FH(v3) ((v3) ? (NFSX_V3FHMAX + NFSX_UNSIGNED) : \ + NFSX_V2FH) +#define NFSX_SRVFH(v3) ((v3) ? NFSX_V3FH : NFSX_V2FH) +#define NFSX_FATTR(v3) ((v3) ? NFSX_V3FATTR : NFSX_V2FATTR) +#define NFSX_PREOPATTR(v3) ((v3) ? (7 * NFSX_UNSIGNED) : 0) +#define NFSX_POSTOPATTR(v3) ((v3) ? (NFSX_V3FATTR + NFSX_UNSIGNED) : 0) +#define NFSX_POSTOPORFATTR(v3) ((v3) ? (NFSX_V3FATTR + NFSX_UNSIGNED) : \ + NFSX_V2FATTR) +#define NFSX_WCCDATA(v3) ((v3) ? NFSX_V3WCCDATA : 0) +#define NFSX_WCCORFATTR(v3) ((v3) ? NFSX_V3WCCDATA : NFSX_V2FATTR) +#define NFSX_SATTR(v3) ((v3) ? NFSX_V3SATTR : NFSX_V2SATTR) +#define NFSX_COOKIEVERF(v3) ((v3) ? NFSX_V3COOKIEVERF : 0) +#define NFSX_WRITEVERF(v3) ((v3) ? NFSX_V3WRITEVERF : 0) +#define NFSX_READDIR(v3) ((v3) ? (5 * NFSX_UNSIGNED) : \ + (2 * NFSX_UNSIGNED)) +#define NFSX_STATFS(v3) ((v3) ? NFSX_V3STATFS : NFSX_V2STATFS) + +/* nfs rpc procedure numbers (before version mapping) */ +#define NFSPROC_NULL 0 +#define NFSPROC_GETATTR 1 +#define NFSPROC_SETATTR 2 +#define NFSPROC_LOOKUP 3 +#define NFSPROC_ACCESS 4 +#define NFSPROC_READLINK 5 +#define NFSPROC_READ 6 +#define NFSPROC_WRITE 7 +#define NFSPROC_CREATE 8 +#define NFSPROC_MKDIR 9 +#define NFSPROC_SYMLINK 10 +#define NFSPROC_MKNOD 11 +#define NFSPROC_REMOVE 12 +#define NFSPROC_RMDIR 13 +#define NFSPROC_RENAME 14 +#define NFSPROC_LINK 15 +#define NFSPROC_READDIR 16 +#define NFSPROC_READDIRPLUS 17 +#define NFSPROC_FSSTAT 18 +#define NFSPROC_FSINFO 19 +#define NFSPROC_PATHCONF 20 +#define NFSPROC_COMMIT 21 + +/* And leasing (nqnfs) procedure numbers (must be last) */ +#define NQNFSPROC_GETLEASE 22 +#define NQNFSPROC_VACATED 23 +#define NQNFSPROC_EVICTED 24 + +#define NFSPROC_NOOP 25 +#define NFS_NPROCS 26 + +/* Actual Version 2 procedure numbers */ +#define NFSV2PROC_NULL 0 +#define NFSV2PROC_GETATTR 1 +#define NFSV2PROC_SETATTR 2 +#define NFSV2PROC_NOOP 3 +#define NFSV2PROC_ROOT NFSV2PROC_NOOP /* Obsolete */ +#define NFSV2PROC_LOOKUP 4 +#define NFSV2PROC_READLINK 5 +#define NFSV2PROC_READ 6 +#define NFSV2PROC_WRITECACHE NFSV2PROC_NOOP /* Obsolete */ +#define NFSV2PROC_WRITE 8 +#define NFSV2PROC_CREATE 9 +#define NFSV2PROC_REMOVE 10 +#define NFSV2PROC_RENAME 11 +#define NFSV2PROC_LINK 12 +#define NFSV2PROC_SYMLINK 13 +#define NFSV2PROC_MKDIR 14 +#define NFSV2PROC_RMDIR 15 +#define NFSV2PROC_READDIR 16 +#define NFSV2PROC_STATFS 17 + +/* + * Constants used by the Version 3 protocol for various RPCs + */ +#define NFSV3SATTRTIME_DONTCHANGE 0 +#define NFSV3SATTRTIME_TOSERVER 1 +#define NFSV3SATTRTIME_TOCLIENT 2 + +#define NFSV3ACCESS_READ 0x01 +#define NFSV3ACCESS_LOOKUP 0x02 +#define NFSV3ACCESS_MODIFY 0x04 +#define NFSV3ACCESS_EXTEND 0x08 +#define NFSV3ACCESS_DELETE 0x10 +#define NFSV3ACCESS_EXECUTE 0x20 + +#define NFSV3WRITE_UNSTABLE 0 +#define NFSV3WRITE_DATASYNC 1 +#define NFSV3WRITE_FILESYNC 2 + +#define NFSV3CREATE_UNCHECKED 0 +#define NFSV3CREATE_GUARDED 1 +#define NFSV3CREATE_EXCLUSIVE 2 + +#define NFSV3FSINFO_LINK 0x01 +#define NFSV3FSINFO_SYMLINK 0x02 +#define NFSV3FSINFO_HOMOGENEOUS 0x08 +#define NFSV3FSINFO_CANSETTIME 0x10 + +/* Conversion macros */ +#define vtonfsv2_mode(t,m) \ + txdr_unsigned(((t) == VFIFO) ? MAKEIMODE(VCHR, (m)) : \ + MAKEIMODE((t), (m))) +#define vtonfsv3_mode(m) txdr_unsigned((m) & 07777) +#define nfstov_mode(a) (fxdr_unsigned(u_short, (a))&07777) +#define vtonfsv2_type(a) txdr_unsigned(nfsv2_type[((long)(a))]) +#define vtonfsv3_type(a) txdr_unsigned(nfsv3_type[((long)(a))]) +#define nfsv2tov_type(a) nv2tov_type[fxdr_unsigned(u_long,(a))&0x7] +#define nfsv3tov_type(a) nv3tov_type[fxdr_unsigned(u_long,(a))&0x7] + +/* File types */ +typedef enum { NFNON=0, NFREG=1, NFDIR=2, NFBLK=3, NFCHR=4, NFLNK=5, + NFSOCK=6, NFFIFO=7 } nfstype; + +/* Structs for common parts of the rpc's */ +/* + * File Handle (32 bytes for version 2), variable up to 64 for version 3. + * File Handles of up to NFS_SMALLFH in size are stored directly in the + * nfs node, whereas larger ones are malloc'd. (This never happens when + * NFS_SMALLFH is set to 64.) + * NFS_SMALLFH should be in the range of 32 to 64 and be divisible by 4. + */ +#ifndef NFS_SMALLFH +#define NFS_SMALLFH 64 +#endif +union nfsfh { + fhandle_t fh_generic; + u_char fh_bytes[NFS_SMALLFH]; +}; +typedef union nfsfh nfsfh_t; + +struct nfsv2_time { + u_long nfsv2_sec; + u_long nfsv2_usec; +}; +typedef struct nfsv2_time nfstime2; + +struct nfsv3_time { + u_long nfsv3_sec; + u_long nfsv3_nsec; +}; +typedef struct nfsv3_time nfstime3; + +/* + * Quads are defined as arrays of 2 longs to ensure dense packing for the + * protocol and to facilitate xdr conversion. + */ +struct nfs_uquad { + u_long nfsuquad[2]; +}; +typedef struct nfs_uquad nfsuint64; + +/* + * Used to convert between two u_longs and a u_quad_t. + */ +union nfs_quadconvert { + u_long lval[2]; + u_quad_t qval; +}; +typedef union nfs_quadconvert nfsquad_t; + +/* + * NFS Version 3 special file number. + */ +struct nfsv3_spec { + u_long specdata1; + u_long specdata2; +}; +typedef struct nfsv3_spec nfsv3spec; + +/* + * File attributes and setable attributes. These structures cover both + * NFS version 2 and the version 3 protocol. Note that the union is only + * used so that one pointer can refer to both variants. These structures + * go out on the wire and must be densely packed, so no quad data types + * are used. (all fields are longs or u_longs or structures of same) + * NB: You can't do sizeof(struct nfs_fattr), you must use the + * NFSX_FATTR(v3) macro. + */ +struct nfs_fattr { + u_long fa_type; + u_long fa_mode; + u_long fa_nlink; + u_long fa_uid; + u_long fa_gid; + union { + struct { + u_long nfsv2fa_size; + u_long nfsv2fa_blocksize; + u_long nfsv2fa_rdev; + u_long nfsv2fa_blocks; + u_long nfsv2fa_fsid; + u_long nfsv2fa_fileid; + nfstime2 nfsv2fa_atime; + nfstime2 nfsv2fa_mtime; + nfstime2 nfsv2fa_ctime; + } fa_nfsv2; + struct { + nfsuint64 nfsv3fa_size; + nfsuint64 nfsv3fa_used; + nfsv3spec nfsv3fa_rdev; + nfsuint64 nfsv3fa_fsid; + nfsuint64 nfsv3fa_fileid; + nfstime3 nfsv3fa_atime; + nfstime3 nfsv3fa_mtime; + nfstime3 nfsv3fa_ctime; + } fa_nfsv3; + } fa_un; +}; + +/* and some ugly defines for accessing union components */ +#define fa2_size fa_un.fa_nfsv2.nfsv2fa_size +#define fa2_blocksize fa_un.fa_nfsv2.nfsv2fa_blocksize +#define fa2_rdev fa_un.fa_nfsv2.nfsv2fa_rdev +#define fa2_blocks fa_un.fa_nfsv2.nfsv2fa_blocks +#define fa2_fsid fa_un.fa_nfsv2.nfsv2fa_fsid +#define fa2_fileid fa_un.fa_nfsv2.nfsv2fa_fileid +#define fa2_atime fa_un.fa_nfsv2.nfsv2fa_atime +#define fa2_mtime fa_un.fa_nfsv2.nfsv2fa_mtime +#define fa2_ctime fa_un.fa_nfsv2.nfsv2fa_ctime +#define fa3_size fa_un.fa_nfsv3.nfsv3fa_size +#define fa3_used fa_un.fa_nfsv3.nfsv3fa_used +#define fa3_rdev fa_un.fa_nfsv3.nfsv3fa_rdev +#define fa3_fsid fa_un.fa_nfsv3.nfsv3fa_fsid +#define fa3_fileid fa_un.fa_nfsv3.nfsv3fa_fileid +#define fa3_atime fa_un.fa_nfsv3.nfsv3fa_atime +#define fa3_mtime fa_un.fa_nfsv3.nfsv3fa_mtime +#define fa3_ctime fa_un.fa_nfsv3.nfsv3fa_ctime + +struct nfsv2_sattr { + u_long sa_mode; + u_long sa_uid; + u_long sa_gid; + u_long sa_size; + nfstime2 sa_atime; + nfstime2 sa_mtime; +}; + +/* + * NFS Version 3 sattr structure for the new node creation case. + */ +struct nfsv3_sattr { + u_long sa_modetrue; + u_long sa_mode; + u_long sa_uidtrue; + u_long sa_uid; + u_long sa_gidtrue; + u_long sa_gid; + u_long sa_sizefalse; + u_long sa_atimetype; + nfstime3 sa_atime; + u_long sa_mtimetype; + nfstime3 sa_mtime; +}; + +struct nfs_statfs { + union { + struct { + u_long nfsv2sf_tsize; + u_long nfsv2sf_bsize; + u_long nfsv2sf_blocks; + u_long nfsv2sf_bfree; + u_long nfsv2sf_bavail; + } sf_nfsv2; + struct { + nfsuint64 nfsv3sf_tbytes; + nfsuint64 nfsv3sf_fbytes; + nfsuint64 nfsv3sf_abytes; + nfsuint64 nfsv3sf_tfiles; + nfsuint64 nfsv3sf_ffiles; + nfsuint64 nfsv3sf_afiles; + u_long nfsv3sf_invarsec; + } sf_nfsv3; + } sf_un; +}; + +#define sf_tsize sf_un.sf_nfsv2.nfsv2sf_tsize +#define sf_bsize sf_un.sf_nfsv2.nfsv2sf_bsize +#define sf_blocks sf_un.sf_nfsv2.nfsv2sf_blocks +#define sf_bfree sf_un.sf_nfsv2.nfsv2sf_bfree +#define sf_bavail sf_un.sf_nfsv2.nfsv2sf_bavail +#define sf_tbytes sf_un.sf_nfsv3.nfsv3sf_tbytes +#define sf_fbytes sf_un.sf_nfsv3.nfsv3sf_fbytes +#define sf_abytes sf_un.sf_nfsv3.nfsv3sf_abytes +#define sf_tfiles sf_un.sf_nfsv3.nfsv3sf_tfiles +#define sf_ffiles sf_un.sf_nfsv3.nfsv3sf_ffiles +#define sf_afiles sf_un.sf_nfsv3.nfsv3sf_afiles +#define sf_invarsec sf_un.sf_nfsv3.nfsv3sf_invarsec + +struct nfsv3_fsinfo { + u_long fs_rtmax; + u_long fs_rtpref; + u_long fs_rtmult; + u_long fs_wtmax; + u_long fs_wtpref; + u_long fs_wtmult; + u_long fs_dtpref; + nfsuint64 fs_maxfilesize; + nfstime3 fs_timedelta; + u_long fs_properties; +}; + +struct nfsv3_pathconf { + u_long pc_linkmax; + u_long pc_namemax; + u_long pc_notrunc; + u_long pc_chownrestricted; + u_long pc_caseinsensitive; + u_long pc_casepreserving; +}; + +#endif diff --git a/sys/nfsserver/nfsrvcache.h b/sys/nfsserver/nfsrvcache.h index b367b9f..dc10bd7 100644 --- a/sys/nfsserver/nfsrvcache.h +++ b/sys/nfsserver/nfsrvcache.h @@ -34,7 +34,7 @@ * SUCH DAMAGE. * * @(#)nfsrvcache.h 8.1 (Berkeley) 6/10/93 - * $Id: nfsrvcache.h,v 1.3 1994/08/21 06:50:13 paul Exp $ + * $Id: nfsrvcache.h,v 1.4 1994/10/17 17:47:44 phk Exp $ */ #ifndef _NFS_NFSRVCACHE_H_ @@ -44,7 +44,7 @@ * Definitions for the server recent request cache */ -#define NFSRVCACHESIZ 256 +#define NFSRVCACHESIZ 64 struct nfsrvcache { TAILQ_ENTRY(nfsrvcache) rc_lru; /* LRU chain */ diff --git a/sys/nfsserver/nfsrvstats.h b/sys/nfsserver/nfsrvstats.h index cbf80c6..e1a0f07 100644 --- a/sys/nfsserver/nfsrvstats.h +++ b/sys/nfsserver/nfsrvstats.h @@ -34,7 +34,7 @@ * SUCH DAMAGE. * * @(#)nfs.h 8.1 (Berkeley) 6/10/93 - * $Id: nfs.h,v 1.8 1994/11/02 00:11:00 wollman Exp $ + * $Id: nfs.h,v 1.9 1995/02/14 06:22:18 phk Exp $ */ #ifndef _NFS_NFS_H_ @@ -45,11 +45,12 @@ */ #define NFS_MAXIOVEC 34 -#define NFS_HZ 25 /* Ticks per second for NFS timeouts */ -#define NFS_TIMEO (1*NFS_HZ) /* Default timeout = 1 second */ -#define NFS_MINTIMEO (1*NFS_HZ) /* Min timeout to use */ -#define NFS_MAXTIMEO (60*NFS_HZ) /* Max timeout to backoff to */ -#define NFS_MINIDEMTIMEO (5*NFS_HZ) /* Min timeout for non-idempotent ops*/ +#define NFS_TICKINTVL 5 /* Desired time for a tick (msec) */ +#define NFS_HZ (hz / nfs_ticks) /* Ticks/sec */ +#define NFS_TIMEO (1 * NFS_HZ) /* Default timeout = 1 second */ +#define NFS_MINTIMEO (1 * NFS_HZ) /* Min timeout to use */ +#define NFS_MAXTIMEO (60 * NFS_HZ) /* Max timeout to backoff to */ +#define NFS_MINIDEMTIMEO (5 * NFS_HZ) /* Min timeout for non-idempotent ops*/ #define NFS_MAXREXMIT 100 /* Stop counting after this many */ #define NFS_MAXWINDOW 1024 /* Max number of outstanding requests */ #define NFS_RETRANS 10 /* Num of retrans for soft mounts */ @@ -62,13 +63,81 @@ #endif #define NFS_WSIZE 8192 /* Def. write data size <= 8192 */ #define NFS_RSIZE 8192 /* Def. read data size <= 8192 */ +#define NFS_READDIRSIZE 8192 /* Def. readdir size */ #define NFS_DEFRAHEAD 1 /* Def. read ahead # blocks */ #define NFS_MAXRAHEAD 4 /* Max. read ahead # blocks */ -#define NFS_MAXREADDIR NFS_MAXDATA /* Max. size of directory read */ #define NFS_MAXUIDHASH 64 /* Max. # of hashed uid entries/mp */ -#define NFS_MAXASYNCDAEMON 20 /* Max. number async_daemons runable */ -#define NFS_DIRBLKSIZ 1024 /* Size of an NFS directory block */ +#define NFS_MAXASYNCDAEMON 20 /* Max. number async_daemons runable */ +#define NFS_MAXGATHERDELAY 100 /* Max. write gather delay (msec) */ +#ifndef NFS_GATHERDELAY +#define NFS_GATHERDELAY 10 /* Default write gather delay (msec) */ +#endif +#define NFS_DIRBLKSIZ 4096 /* Must be a multiple of DIRBLKSIZ */ + +/* + * Oddballs + */ #define NMOD(a) ((a) % nfs_asyncdaemons) +#define NFS_CMPFH(n, f, s) \ + ((n)->n_fhsize == (s) && !bcmp((caddr_t)(n)->n_fhp, (caddr_t)(f), (s))) +#define NFS_ISV3(v) (VFSTONFS((v)->v_mount)->nm_flag & NFSMNT_NFSV3) +#define NFS_SRVMAXDATA(n) \ + (((n)->nd_flag & ND_NFSV3) ? (((n)->nd_nam2) ? \ + NFS_MAXDGRAMDATA : NFS_MAXDATA) : NFS_V2MAXDATA) + +/* + * XXX + * sys/buf.h should be editted to change B_APPENDWRITE --> B_NEEDCOMMIT, but + * until then... + * Same goes for sys/malloc.h, which needs M_NFSDIROFF, + * M_NFSRVDESC and M_NFSBIGFH added. + * The VA_EXCLUSIVE flag should be added for va_vaflags and set for an + * exclusive create. + * The B_INVAFTERWRITE flag should be set to whatever is required by the + * buffer cache code to say "Invalidate the block after it is written back". + */ +#ifndef B_NEEDCOMMIT +#define B_NEEDCOMMIT B_APPENDWRITE +#endif +#ifndef M_NFSRVDESC +#define M_NFSRVDESC M_TEMP +#endif +#ifndef M_NFSDIROFF +#define M_NFSDIROFF M_TEMP +#endif +#ifndef M_NFSBIGFH +#define M_NFSBIGFH M_TEMP +#endif +#ifndef VA_EXCLUSIVE +#define VA_EXCLUSIVE 0 +#endif +#ifdef __FreeBSD__ +#define B_INVAFTERWRITE B_NOCACHE +#else +#define B_INVAFTERWRITE B_INVAL +#endif + +/* + * These ifdefs try to handle the differences between the various 4.4BSD-Lite + * based vfs interfaces. + * btw: NetBSD-current does have a VOP_LEASDE(), but I don't know how to + * differentiate between NetBSD-1.0 and NetBSD-current, so.. + * I also don't know about BSDi's 2.0 release. + */ +#if !defined(HAS_VOPLEASE) && !defined(__FreeBSD__) && !defined(__NetBSD__) +#define HAS_VOPLEASE 1 +#endif +#if !defined(HAS_VOPREVOKE) && !defined(__FreeBSD__) && !defined(__NetBSD__) +#define HAS_VOPREVOKE 1 +#endif + +/* + * The IO_METASYNC flag should be implemented for local file systems. + * (Until then, it is nothin at all.) + */ +#ifndef IO_METASYNC +#define IO_METASYNC 0 +#endif /* * Set the attribute timeout based on how recently the file has been modified. @@ -80,6 +149,20 @@ (time.tv_sec - (np)->n_mtime) / 10)) /* + * Expected allocation sizes for major data structures. If the actual size + * of the structure exceeds these sizes, then malloc() will be allocating + * almost twice the memory required. This is used in nfs_init() to warn + * the sysadmin that the size of a structure should be reduced. + * (These sizes are always a power of 2. If the kernel malloc() changes + * to one that does not allocate space in powers of 2 size, then this all + * becomes bunk!) + */ +#define NFS_NODEALLOC 256 +#define NFS_MNTALLOC 512 +#define NFS_SVCALLOC 256 +#define NFS_UIDALLOC 128 + +/* * Structures for the nfssvc(2) syscall. Not that anyone but nfsd and mount_nfs * should ever try and use it. */ @@ -95,7 +178,12 @@ struct nfsd_srvargs { u_long nsd_haddr; /* Ip address of client */ struct ucred nsd_cr; /* Cred. uid maps to */ int nsd_authlen; /* Length of auth string (ret) */ - char *nsd_authstr; /* Auth string (ret) */ + u_char *nsd_authstr; /* Auth string (ret) */ + int nsd_verflen; /* and the verfier */ + u_char *nsd_verfstr; + struct timeval nsd_timestamp; /* timestamp from verifier */ + u_long nsd_ttl; /* credential ttl (sec) */ + NFSKERBKEY_T nsd_key; /* Session key */ }; struct nfsd_cargs { @@ -103,7 +191,10 @@ struct nfsd_cargs { uid_t ncd_authuid; /* Effective uid */ int ncd_authtype; /* Type of authenticator */ int ncd_authlen; /* Length of authenticator string */ - char *ncd_authstr; /* Authenticator string */ + u_char *ncd_authstr; /* Authenticator string */ + int ncd_verflen; /* and the verifier */ + u_char *ncd_verfstr; + NFSKERBKEY_T ncd_key; /* Session key */ }; /* @@ -142,6 +233,7 @@ struct nfsstats { int srvnqnfs_leases; int srvnqnfs_maxleases; int srvnqnfs_getleases; + int srvvop_writes; }; /* @@ -173,7 +265,7 @@ struct nfsstats { * such as SIGALRM will not expect file I/O system calls to be interrupted * by them and break. */ -#ifdef KERNEL +#if defined(KERNEL) || defined(_KERNEL) struct uio; struct buf; struct vattr; struct nameidata; /* XXX */ @@ -224,17 +316,29 @@ TAILQ_HEAD(, nfsreq) nfs_reqq; #define R_MUSTRESEND 0x40 /* Must resend request */ #define R_GETONEREP 0x80 /* Probe for one reply only */ -extern struct nfsstats nfsstats; - /* * A list of nfssvc_sock structures is maintained with all the sockets * that require service by the nfsd. * The nfsuid structs hang off of the nfssvc_sock structs in both lru * and uid hash lists. */ -#define NUIDHASHSIZ 32 +#ifndef NFS_UIDHASHSIZ +#define NFS_UIDHASHSIZ 29 /* Tune the size of nfssvc_sock with this */ +#endif #define NUIDHASH(sock, uid) \ - (&(sock)->ns_uidhashtbl[(uid) & (sock)->ns_uidhash]) + (&(sock)->ns_uidhashtbl[(uid) % NFS_UIDHASHSIZ]) +#ifndef NFS_WDELAYHASHSIZ +#define NFS_WDELAYHASHSIZ 16 /* and with this */ +#endif +#define NWDELAYHASH(sock, f) \ + (&(sock)->ns_wdelayhashtbl[(*((u_long *)(f))) % NFS_WDELAYHASHSIZ]) +#ifndef NFS_MUIDHASHSIZ +#define NFS_MUIDHASHSIZ 67 /* Tune the size of nfsmount with this */ +#endif +#define NMUIDHASH(nmp, uid) \ + (&(nmp)->nm_uidhashtbl[(uid) % NFS_MUIDHASHSIZ]) +#define NFSNOHASH(fhsum) \ + (&nfsnodehashtbl[(fhsum) & nfsnodehash]) /* * Network address hash list element @@ -248,35 +352,41 @@ struct nfsuid { TAILQ_ENTRY(nfsuid) nu_lru; /* LRU chain */ LIST_ENTRY(nfsuid) nu_hash; /* Hash list */ int nu_flag; /* Flags */ - uid_t nu_uid; /* Uid mapped by this entry */ union nethostaddr nu_haddr; /* Host addr. for dgram sockets */ struct ucred nu_cr; /* Cred uid mapped to */ + int nu_expire; /* Expiry time (sec) */ + struct timeval nu_timestamp; /* Kerb. timestamp */ + u_long nu_nickname; /* Nickname on server */ + NFSKERBKEY_T nu_key; /* and session key */ }; #define nu_inetaddr nu_haddr.had_inetaddr #define nu_nam nu_haddr.had_nam /* Bits for nu_flag */ #define NU_INETADDR 0x1 +#define NU_NAM 0x2 +#define NU_NETFAM(u) (((u)->nu_flag & NU_INETADDR) ? AF_INET : AF_ISO) struct nfssvc_sock { TAILQ_ENTRY(nfssvc_sock) ns_chain; /* List of all nfssvc_sock's */ TAILQ_HEAD(, nfsuid) ns_uidlruhead; - LIST_HEAD(, nfsuid) *ns_uidhashtbl; - u_long ns_uidhash; - - int ns_flag; - u_long ns_sref; struct file *ns_fp; struct socket *ns_so; - int ns_solock; struct mbuf *ns_nam; - int ns_cc; struct mbuf *ns_raw; struct mbuf *ns_rawend; - int ns_reclen; struct mbuf *ns_rec; struct mbuf *ns_recend; + struct mbuf *ns_frag; + int ns_flag; + int ns_solock; + int ns_cc; + int ns_reclen; int ns_numuids; + u_long ns_sref; + LIST_HEAD(, nfsrv_descript) ns_tq; /* Write gather lists */ + LIST_HEAD(, nfsuid) ns_uidhashtbl[NFS_UIDHASHSIZ]; + LIST_HEAD(nfsrvw_delayhash, nfsrv_descript) ns_wdelayhashtbl[NFS_WDELAYHASHSIZ]; }; /* Bits for "ns_flag" */ @@ -285,6 +395,7 @@ struct nfssvc_sock { #define SLP_NEEDQ 0x04 #define SLP_DISCONN 0x08 #define SLP_GETSTREAM 0x10 +#define SLP_LASTFRAG 0x20 #define SLP_ALLFLAGS 0xff TAILQ_HEAD(, nfssvc_sock) nfssvc_sockhead; @@ -296,73 +407,124 @@ int nfssvc_sockhead_flag; * One of these structures is allocated for each nfsd. */ struct nfsd { - TAILQ_ENTRY(nfsd) nd_chain; /* List of all nfsd's */ - int nd_flag; /* NFSD_ flags */ - struct nfssvc_sock *nd_slp; /* Current socket */ - struct mbuf *nd_nam; /* Client addr for datagram req. */ - struct mbuf *nd_mrep; /* Req. mbuf list */ - struct mbuf *nd_md; - caddr_t nd_dpos; /* Position in list */ - int nd_procnum; /* RPC procedure number */ - u_long nd_retxid; /* RPC xid */ - int nd_repstat; /* Reply status value */ - struct ucred nd_cr; /* Credentials for req. */ - int nd_nqlflag; /* Leasing flag */ - u_long nd_duration; /* Lease duration */ - int nd_authlen; /* Authenticator len */ - u_char nd_authstr[RPCAUTH_MAXSIZ]; /* Authenticator data */ - struct proc *nd_procp; /* Proc ptr */ + TAILQ_ENTRY(nfsd) nfsd_chain; /* List of all nfsd's */ + int nfsd_flag; /* NFSD_ flags */ + struct nfssvc_sock *nfsd_slp; /* Current socket */ + int nfsd_authlen; /* Authenticator len */ + u_char nfsd_authstr[RPCAUTH_MAXSIZ]; /* Authenticator data */ + int nfsd_verflen; /* and the Verifier */ + u_char nfsd_verfstr[RPCVERF_MAXSIZ]; + struct proc *nfsd_procp; /* Proc ptr */ + struct nfsrv_descript *nfsd_nd; /* Associated nfsrv_descript */ }; -/* Bits for "nd_flag" */ +/* Bits for "nfsd_flag" */ #define NFSD_WAITING 0x01 #define NFSD_REQINPROG 0x02 #define NFSD_NEEDAUTH 0x04 #define NFSD_AUTHFAIL 0x08 +/* + * This structure is used by the server for describing each request. + * Some fields are used only when write request gathering is performed. + */ +struct nfsrv_descript { + u_quad_t nd_time; /* Write deadline (usec) */ + off_t nd_off; /* Start byte offset */ + off_t nd_eoff; /* and end byte offset */ + LIST_ENTRY(nfsrv_descript) nd_hash; /* Hash list */ + LIST_ENTRY(nfsrv_descript) nd_tq; /* and timer list */ + LIST_HEAD(,nfsrv_descript) nd_coalesce; /* coalesced writes */ + struct mbuf *nd_mrep; /* Request mbuf list */ + struct mbuf *nd_md; /* Current dissect mbuf */ + struct mbuf *nd_mreq; /* Reply mbuf list */ + struct mbuf *nd_nam; /* and socket addr */ + struct mbuf *nd_nam2; /* return socket addr */ + caddr_t nd_dpos; /* Current dissect pos */ + int nd_procnum; /* RPC # */ + int nd_stable; /* storage type */ + int nd_flag; /* nd_flag */ + int nd_len; /* Length of this write */ + int nd_repstat; /* Reply status */ + u_long nd_retxid; /* Reply xid */ + u_long nd_duration; /* Lease duration */ + struct timeval nd_starttime; /* Time RPC initiated */ + fhandle_t nd_fh; /* File handle */ + struct ucred nd_cr; /* Credentials */ +}; + +/* Bits for "nd_flag" */ +#define ND_READ LEASE_READ +#define ND_WRITE LEASE_WRITE +#define ND_CHECK 0x04 +#define ND_LEASE (ND_READ | ND_WRITE | ND_CHECK) +#define ND_NFSV3 0x08 +#define ND_NQNFS 0x10 +#define ND_KERBNICK 0x20 +#define ND_KERBFULL 0x40 +#define ND_KERBAUTH (ND_KERBNICK | ND_KERBFULL) + TAILQ_HEAD(, nfsd) nfsd_head; int nfsd_head_flag; #define NFSD_CHECKSLP 0x01 +/* + * These macros compare nfsrv_descript structures. + */ +#define NFSW_CONTIG(o, n) \ + ((o)->nd_eoff >= (n)->nd_off && \ + !bcmp((caddr_t)&(o)->nd_fh, (caddr_t)&(n)->nd_fh, NFSX_V3FH)) + +#define NFSW_SAMECRED(o, n) \ + (((o)->nd_flag & ND_KERBAUTH) == ((n)->nd_flag & ND_KERBAUTH) && \ + !bcmp((caddr_t)&(o)->nd_cr, (caddr_t)&(n)->nd_cr, \ + sizeof (struct ucred))) + int nfs_reply __P((struct nfsreq *)); -int nfs_getreq __P((struct nfsd *,int)); +int nfs_getreq __P((struct nfsrv_descript *,struct nfsd *,int)); int nfs_send __P((struct socket *,struct mbuf *,struct mbuf *,struct nfsreq *)); -int nfs_rephead __P((int,struct nfsd *,int,int,u_quad_t *,struct mbuf **,struct mbuf **,caddr_t *)); +int nfs_rephead __P((int,struct nfsrv_descript *,struct nfssvc_sock *,int,int,u_quad_t *,struct mbuf **,struct mbuf **,caddr_t *)); int nfs_sndlock __P((int *,struct nfsreq *)); int nfs_disct __P((struct mbuf **,caddr_t *,int,int,caddr_t *)); int nfs_vinvalbuf __P((struct vnode *,int,struct ucred *,struct proc *,int)); int nfs_readrpc __P((struct vnode *,struct uio *,struct ucred *)); -int nfs_writerpc __P((struct vnode *,struct uio *,struct ucred *,int)); +int nfs_writerpc __P((struct vnode *,struct uio *,struct ucred *,int *,int *)); int nfs_readdirrpc __P((register struct vnode *,struct uio *,struct ucred *)); +int nfs_setattrrpc __P((struct vnode *,struct vattr *,struct ucred *,struct proc *)); int nfs_asyncio __P((struct buf *,struct ucred *)); int nfs_doio __P((struct buf *,struct ucred *,struct proc *)); int nfs_readlinkrpc __P((struct vnode *,struct uio *,struct ucred *)); int nfs_sigintr __P((struct nfsmount *,struct nfsreq *r,struct proc *)); -int nfs_readdirlookrpc __P((struct vnode *,register struct uio *,struct ucred *)); +int nfs_readdirplusrpc __P((struct vnode *,register struct uio *,struct ucred *)); int nfsm_disct __P((struct mbuf **,caddr_t *,int,int,caddr_t *)); -int nfsrv_fhtovp __P((fhandle_t *,int,struct vnode **,struct ucred *,struct nfssvc_sock *,struct mbuf *,int *)); +void nfsm_srvfattr __P((struct nfsrv_descript *,struct vattr *,struct nfs_fattr *)); +void nfsm_srvwcc __P((struct nfsrv_descript *,int,struct vattr *,int,struct vattr *,struct mbuf **,char **)); +void nfsm_srvpostopattr __P((struct nfsrv_descript *,int,struct vattr *,struct mbuf **,char **)); +int nfsrv_fhtovp __P((fhandle_t *,int,struct vnode **,struct ucred *,struct nfssvc_sock *,struct mbuf *,int *,int)); int nfsrv_access __P((struct vnode *,int,struct ucred *,int,struct proc *)); int netaddr_match __P((int,union nethostaddr *,struct mbuf *)); int nfs_request __P((struct vnode *,struct mbuf *,int,struct proc *,struct ucred *,struct mbuf **,struct mbuf **,caddr_t *)); int nfs_loadattrcache __P((struct vnode **,struct mbuf **,caddr_t *,struct vattr *)); -int nfs_namei __P((struct nameidata *,fhandle_t *,int,struct nfssvc_sock *,struct mbuf *,struct mbuf **,caddr_t *,struct proc *)); +int nfs_namei __P((struct nameidata *,fhandle_t *,int,struct nfssvc_sock *,struct mbuf *,struct mbuf **,caddr_t *,struct vnode **,struct proc *,int)); void nfsm_adj __P((struct mbuf *,int,int)); int nfsm_mbuftouio __P((struct mbuf **,struct uio *,int,caddr_t *)); void nfsrv_initcache __P((void)); int nfs_rcvlock __P((struct nfsreq *)); -int nfs_getauth __P((struct nfsmount *,struct nfsreq *,struct ucred *,int *,char **,int *)); +int nfs_getauth __P((struct nfsmount *,struct nfsreq *,struct ucred *,char **,int *,char *,int *,NFSKERBKEY_T)); +int nfs_getnickauth __P((struct nfsmount *,struct ucred *,char **,int *,char *,int)); +int nfs_savenickauth __P((struct nfsmount *,struct ucred *,int,NFSKERBKEY_T,struct mbuf **,char **,struct mbuf *)); int nfs_msg __P((struct proc *,char *,char *)); int nfs_adv __P((struct mbuf **,caddr_t *,int,int)); int nfsrv_getstream __P((struct nfssvc_sock *,int)); void nfs_nhinit __P((void)); void nfs_timer __P((void*)); -struct nfsnodehashhead * nfs_hash __P((nfsv2fh_t *)); +u_long nfs_hash __P((nfsfh_t *,int)); int nfssvc_iod __P((struct proc *)); int nfssvc_nfsd __P((struct nfsd_srvargs *,caddr_t,struct proc *)); int nfssvc_addsock __P((struct file *,struct mbuf *)); -int nfsrv_dorec __P((struct nfssvc_sock *,struct nfsd *)); -int nfsrv_getcache __P((struct mbuf *,struct nfsd *,struct mbuf **)); -void nfsrv_updatecache __P((struct mbuf *,struct nfsd *,int,struct mbuf *)); +int nfsrv_dorec __P((struct nfssvc_sock *,struct nfsd *,struct nfsrv_descript **)); +int nfsrv_getcache __P((struct nfsrv_descript *,struct nfssvc_sock *,struct mbuf **)); +void nfsrv_updatecache __P((struct nfsrv_descript *,int,struct mbuf *)); int mountnfs __P((struct nfs_args *,struct mount *,struct mbuf *,char *,char *,struct vnode **)); int nfs_connect __P((struct nfsmount *,struct nfsreq *)); int nfs_getattrcache __P((struct vnode *,struct vattr *)); @@ -370,9 +532,20 @@ int nfsm_strtmbuf __P((struct mbuf **,char **,char *,long)); int nfs_bioread __P((struct vnode *,struct uio *,int,struct ucred *)); int nfsm_uiotombuf __P((struct uio *,struct mbuf **,int,caddr_t *)); void nfsrv_init __P((int)); +void nfs_clearcommit __P((struct mount *)); +int nfsrv_errmap __P((struct nfsrv_descript *, int)); +void nfsrvw_coalesce __P((struct nfsrv_descript *,struct nfsrv_descript *)); +void nfsrvw_sort __P((gid_t [],int)); +void nfsrv_setcred __P((struct ucred *,struct ucred *)); +int nfs_flush __P((struct vnode *,struct ucred *,int,struct proc *,int)); +int nfs_writebp __P((struct buf *,int)); int nfsrv_vput __P(( struct vnode * )); int nfsrv_vrele __P(( struct vnode * )); int nfsrv_vmio __P(( struct vnode * )); +int nfsrv_writegather __P((struct nfsrv_descript **, struct nfssvc_sock *, + struct proc *, struct mbuf **)); +int nfs_fsinfo __P((struct nfsmount *, struct vnode *, struct ucred *, + struct proc *p)); #endif /* KERNEL */ diff --git a/sys/sys/mount.h b/sys/sys/mount.h index 1e6b19a..0580e98 100644 --- a/sys/sys/mount.h +++ b/sys/sys/mount.h @@ -31,7 +31,7 @@ * SUCH DAMAGE. * * @(#)mount.h 8.13 (Berkeley) 3/27/94 - * $Id: mount.h,v 1.18 1995/05/21 21:39:24 davidg Exp $ + * $Id: mount.h,v 1.19 1995/05/30 08:14:28 rgrimes Exp $ */ #ifndef _SYS_MOUNT_H_ @@ -403,15 +403,6 @@ struct iso_args { #ifdef NFS /* - * File Handle (32 bytes for version 2), variable up to 1024 for version 3 - */ -union nfsv2fh { - fhandle_t fh_generic; - u_char fh_bytes[32]; -}; -typedef union nfsv2fh nfsv2fh_t; - -/* * Arguments to mount NFS */ struct nfs_args { @@ -419,10 +410,12 @@ struct nfs_args { int addrlen; /* length of address */ int sotype; /* Socket type */ int proto; /* and Protocol */ - nfsv2fh_t *fh; /* File handle to be mounted */ + u_char *fh; /* File handle to be mounted */ + int fhsize; /* Size, in bytes, of fh */ int flags; /* flags */ int wsize; /* write size in bytes */ int rsize; /* read size in bytes */ + int readdirsize; /* readdir size in bytes */ int timeo; /* initial timeout in .1 secs */ int retrans; /* times to retry send */ int maxgrouplist; /* Max. size of group list */ @@ -432,7 +425,6 @@ struct nfs_args { char *hostname; /* server's name */ }; - /* * NFS mount option flags */ @@ -445,16 +437,19 @@ struct nfs_args { #define NFSMNT_INT 0x00000040 /* allow interrupts on hard mount */ #define NFSMNT_NOCONN 0x00000080 /* Don't Connect the socket */ #define NFSMNT_NQNFS 0x00000100 /* Use Nqnfs protocol */ -#define NFSMNT_MYWRITE 0x00000200 /* Assume writes were mine */ +#define NFSMNT_NFSV3 0x00000200 /* Use NFS Version 3 protocol */ #define NFSMNT_KERB 0x00000400 /* Use Kerberos authentication */ #define NFSMNT_DUMBTIMR 0x00000800 /* Don't estimate rtt dynamically */ -#define NFSMNT_RDIRALOOK 0x00001000 /* Do lookup with readdir (nqnfs) */ -#define NFSMNT_LEASETERM 0x00002000 /* set lease term (nqnfs) */ -#define NFSMNT_READAHEAD 0x00004000 /* set read ahead */ -#define NFSMNT_DEADTHRESH 0x00008000 /* set dead server retry thresh */ -#define NFSMNT_NQLOOKLEASE 0x00010000 /* Get lease for lookup */ -#define NFSMNT_RESVPORT 0x00020000 /* Allocate a reserved port */ -#define NFSMNT_INTERNAL 0xffe00000 /* Bits set internally */ +#define NFSMNT_LEASETERM 0x00001000 /* set lease term (nqnfs) */ +#define NFSMNT_READAHEAD 0x00002000 /* set read ahead */ +#define NFSMNT_DEADTHRESH 0x00004000 /* set dead server retry thresh */ +#define NFSMNT_RESVPORT 0x00008000 /* Allocate a reserved port */ +#define NFSMNT_RDIRPLUS 0x00010000 /* Use Readdirplus for V3 */ +#define NFSMNT_READDIRSIZE 0x00020000 /* Set readdir size */ +#define NFSMNT_INTERNAL 0xfffc0000 /* Bits set internally */ +#define NFSMNT_HASWRITEVERF 0x00040000 /* Has write verifier for V3 */ +#define NFSMNT_GOTPATHCONF 0x00080000 /* Got the V3 pathconf info */ +#define NFSMNT_GOTFSINFO 0x00100000 /* Got the V3 fsinfo */ #define NFSMNT_MNTD 0x00200000 /* Mnt server for mnt point */ #define NFSMNT_DISMINPROG 0x00400000 /* Dismount in progress */ #define NFSMNT_DISMNT 0x00800000 /* Dismounted */ diff --git a/usr.bin/fstat/fstat.c b/usr.bin/fstat/fstat.c index 304f3fc..96116f9 100644 --- a/usr.bin/fstat/fstat.c +++ b/usr.bin/fstat/fstat.c @@ -62,7 +62,7 @@ static char sccsid[] = "@(#)fstat.c 8.1 (Berkeley) 6/6/93"; #undef KERNEL #define NFS #include <sys/mount.h> -#include <nfs/nfsv2.h> +#include <nfs/nfsproto.h> #include <nfs/rpcv2.h> #include <nfs/nfs.h> #include <nfs/nfsnode.h> diff --git a/usr.bin/nfsstat/nfsstat.c b/usr.bin/nfsstat/nfsstat.c index 169e8d1..0f54f12 100644 --- a/usr.bin/nfsstat/nfsstat.c +++ b/usr.bin/nfsstat/nfsstat.c @@ -43,13 +43,14 @@ static char copyright[] = #ifndef lint /*static char sccsid[] = "From: @(#)nfsstat.c 8.1 (Berkeley) 6/6/93";*/ static const char rcsid[] = - "$Id$"; + "$Id: nfsstat.c,v 1.2 1994/10/23 23:25:43 wollman Exp $"; #endif /* not lint */ #include <sys/param.h> #include <sys/mount.h> #include <sys/sysctl.h> -#include <nfs/nfsv2.h> +#include <nfs/rpcv2.h> +#include <nfs/nfsproto.h> #include <nfs/nfs.h> #include <signal.h> #include <fcntl.h> @@ -73,7 +74,10 @@ kvm_t *kd; static int deadkernel = 0; -void intpr(void), printhdr(void), sidewaysintpr(u_int), usage(void); +void intpr __P((void)); +void printhdr __P((void)); +void sidewaysintpr __P((u_int)); +void usage __P((void)); main(argc, argv) int argc; @@ -146,7 +150,8 @@ main(argc, argv) * for dead ones. */ void -readstats(struct nfsstats *stp) +readstats(stp) + struct nfsstats *stp; { if(deadkernel) { if(kvm_read(kd, (u_long)nl[N_NFSSTAT].n_value, stp, @@ -171,7 +176,7 @@ readstats(struct nfsstats *stp) * Print a description of the nfs stats. */ void -intpr(void) +intpr() { struct nfsstats nfsstats; @@ -193,7 +198,7 @@ intpr(void) nfsstats.rpccnt[NFSPROC_REMOVE]); printf("%9.9s %9.9s %9.9s %9.9s %9.9s %9.9s %9.9s %9.9s\n", "Rename", "Link", "Symlink", "Mkdir", "Rmdir", - "Readdir", "Statfs", "RdirLook"); + "Readdir", "RdirPlus", "Access"); printf("%9d %9d %9d %9d %9d %9d %9d %9d\n", nfsstats.rpccnt[NFSPROC_RENAME], nfsstats.rpccnt[NFSPROC_LINK], @@ -201,11 +206,17 @@ intpr(void) nfsstats.rpccnt[NFSPROC_MKDIR], nfsstats.rpccnt[NFSPROC_RMDIR], nfsstats.rpccnt[NFSPROC_READDIR], - nfsstats.rpccnt[NFSPROC_STATFS], - nfsstats.rpccnt[NQNFSPROC_READDIRLOOK]); - printf("%9.9s %9.9s %9.9s\n", + nfsstats.rpccnt[NFSPROC_READDIRPLUS], + nfsstats.rpccnt[NFSPROC_ACCESS]); + printf("%9.9s %9.9s %9.9s %9.9s %9.9s %9.9s %9.9s %9.9s\n", + "Mknod", "Fsstat", "Fsinfo", "PathConf", "Commit", "GLease", "Vacate", "Evict"); - printf("%9d %9d %9d\n", + printf("%9d %9d %9d %9d %9d %9d %9d %9d\n", + nfsstats.rpccnt[NFSPROC_MKNOD], + nfsstats.rpccnt[NFSPROC_FSSTAT], + nfsstats.rpccnt[NFSPROC_FSINFO], + nfsstats.rpccnt[NFSPROC_PATHCONF], + nfsstats.rpccnt[NFSPROC_COMMIT], nfsstats.rpccnt[NQNFSPROC_GETLEASE], nfsstats.rpccnt[NQNFSPROC_VACATED], nfsstats.rpccnt[NQNFSPROC_EVICTED]); @@ -256,7 +267,7 @@ intpr(void) nfsstats.srvrpccnt[NFSPROC_REMOVE]); printf("%9.9s %9.9s %9.9s %9.9s %9.9s %9.9s %9.9s %9.9s\n", "Rename", "Link", "Symlink", "Mkdir", "Rmdir", - "Readdir", "Statfs", "RdirLook"); + "Readdir", "RdirPlus", "Access"); printf("%9d %9d %9d %9d %9d %9d %9d %9d\n", nfsstats.srvrpccnt[NFSPROC_RENAME], nfsstats.srvrpccnt[NFSPROC_LINK], @@ -264,11 +275,17 @@ intpr(void) nfsstats.srvrpccnt[NFSPROC_MKDIR], nfsstats.srvrpccnt[NFSPROC_RMDIR], nfsstats.srvrpccnt[NFSPROC_READDIR], - nfsstats.srvrpccnt[NFSPROC_STATFS], - nfsstats.srvrpccnt[NQNFSPROC_READDIRLOOK]); - printf("%9.9s %9.9s %9.9s\n", + nfsstats.srvrpccnt[NFSPROC_READDIRPLUS], + nfsstats.srvrpccnt[NFSPROC_ACCESS]); + printf("%9.9s %9.9s %9.9s %9.9s %9.9s %9.9s %9.9s %9.9s\n", + "Mknod", "Fsstat", "Fsinfo", "PathConf", "Commit", "GLease", "Vacate", "Evict"); - printf("%9d %9d %9d\n", + printf("%9d %9d %9d %9d %9d %9d %9d %9d\n", + nfsstats.srvrpccnt[NFSPROC_MKNOD], + nfsstats.srvrpccnt[NFSPROC_FSSTAT], + nfsstats.srvrpccnt[NFSPROC_FSINFO], + nfsstats.srvrpccnt[NFSPROC_PATHCONF], + nfsstats.srvrpccnt[NFSPROC_COMMIT], nfsstats.srvrpccnt[NQNFSPROC_GETLEASE], nfsstats.srvrpccnt[NQNFSPROC_VACATED], nfsstats.srvrpccnt[NQNFSPROC_EVICTED]); @@ -291,6 +308,13 @@ intpr(void) nfsstats.srvnqnfs_leases, nfsstats.srvnqnfs_maxleases, nfsstats.srvnqnfs_getleases); + printf("Server Write Gathering:\n"); + printf("%9.9s %9.9s %9.9s\n", + "WriteOps", "WriteRPC", "Opsaved"); + printf("%9d %9d %9d\n", + nfsstats.srvvop_writes, + nfsstats.srvrpccnt[NFSPROC_WRITE], + nfsstats.srvrpccnt[NFSPROC_WRITE] - nfsstats.srvvop_writes); } u_char signalled; /* set if alarm goes off "early" */ @@ -321,23 +345,25 @@ sidewaysintpr(interval) } readstats(&nfsstats); printf("Client: %8d %8d %8d %8d %8d %8d %8d %8d\n", - nfsstats.rpccnt[1]-lastst.rpccnt[1], - nfsstats.rpccnt[4]-lastst.rpccnt[4], - nfsstats.rpccnt[5]-lastst.rpccnt[5], - nfsstats.rpccnt[6]-lastst.rpccnt[6], - nfsstats.rpccnt[8]-lastst.rpccnt[8], - nfsstats.rpccnt[11]-lastst.rpccnt[11], - nfsstats.rpccnt[12]-lastst.rpccnt[12], - nfsstats.rpccnt[16]-lastst.rpccnt[16]); + nfsstats.rpccnt[NFSPROC_GETATTR]-lastst.rpccnt[NFSPROC_GETATTR], + nfsstats.rpccnt[NFSPROC_LOOKUP]-lastst.rpccnt[NFSPROC_LOOKUP], + nfsstats.rpccnt[NFSPROC_READLINK]-lastst.rpccnt[NFSPROC_READLINK], + nfsstats.rpccnt[NFSPROC_READ]-lastst.rpccnt[NFSPROC_READ], + nfsstats.rpccnt[NFSPROC_WRITE]-lastst.rpccnt[NFSPROC_WRITE], + nfsstats.rpccnt[NFSPROC_RENAME]-lastst.rpccnt[NFSPROC_RENAME], + nfsstats.rpccnt[NFSPROC_ACCESS]-lastst.rpccnt[NFSPROC_ACCESS], + (nfsstats.rpccnt[NFSPROC_READDIR]-lastst.rpccnt[NFSPROC_READDIR]) + +(nfsstats.rpccnt[NFSPROC_READDIRPLUS]-lastst.rpccnt[NFSPROC_READDIRPLUS])); printf("Server: %8d %8d %8d %8d %8d %8d %8d %8d\n", - nfsstats.srvrpccnt[1]-lastst.srvrpccnt[1], - nfsstats.srvrpccnt[4]-lastst.srvrpccnt[4], - nfsstats.srvrpccnt[5]-lastst.srvrpccnt[5], - nfsstats.srvrpccnt[6]-lastst.srvrpccnt[6], - nfsstats.srvrpccnt[8]-lastst.srvrpccnt[8], - nfsstats.srvrpccnt[11]-lastst.srvrpccnt[11], - nfsstats.srvrpccnt[12]-lastst.srvrpccnt[12], - nfsstats.srvrpccnt[16]-lastst.srvrpccnt[16]); + nfsstats.srvrpccnt[NFSPROC_GETATTR]-lastst.srvrpccnt[NFSPROC_GETATTR], + nfsstats.srvrpccnt[NFSPROC_LOOKUP]-lastst.srvrpccnt[NFSPROC_LOOKUP], + nfsstats.srvrpccnt[NFSPROC_READLINK]-lastst.srvrpccnt[NFSPROC_READLINK], + nfsstats.srvrpccnt[NFSPROC_READ]-lastst.srvrpccnt[NFSPROC_READ], + nfsstats.srvrpccnt[NFSPROC_WRITE]-lastst.srvrpccnt[NFSPROC_WRITE], + nfsstats.srvrpccnt[NFSPROC_RENAME]-lastst.srvrpccnt[NFSPROC_RENAME], + nfsstats.srvrpccnt[NFSPROC_ACCESS]-lastst.srvrpccnt[NFSPROC_ACCESS], + (nfsstats.srvrpccnt[NFSPROC_READDIR]-lastst.srvrpccnt[NFSPROC_READDIR]) + +(nfsstats.srvrpccnt[NFSPROC_READDIRPLUS]-lastst.srvrpccnt[NFSPROC_READDIRPLUS])); lastst = nfsstats; fflush(stdout); oldmask = sigblock(sigmask(SIGALRM)); @@ -355,7 +381,7 @@ printhdr() { printf(" %8.8s %8.8s %8.8s %8.8s %8.8s %8.8s %8.8s %8.8s\n", "Getattr", "Lookup", "Readlink", "Read", "Write", "Rename", - "Link", "Readdir"); + "Access", "Readdir"); fflush(stdout); } diff --git a/usr.bin/showmount/showmount.8 b/usr.bin/showmount/showmount.8 index fe84e1e..05c89a9 100644 --- a/usr.bin/showmount/showmount.8 +++ b/usr.bin/showmount/showmount.8 @@ -42,7 +42,7 @@ .Nd show remote nfs mounts on host .Sh SYNOPSIS .Nm showmount -.Op Fl ade +.Op Fl ade3 .Op Ar host .Sh DESCRIPTION .Nm Showmount @@ -56,7 +56,10 @@ file systems mounted on the host. See .%T "NFS: Network File System Protocol Specification" , RFC 1094, -Appendix A , +Appendix A, +and +.%T "NFS: Network File System Version 3 Protocol Specification" , +Appendix I, for a detailed description of the protocol. .Bl -tag -width Ds .It Fl a @@ -70,6 +73,8 @@ List directory paths of mount points instead of hosts Show the .Ar host Ns 's exports list +.It Fl 3 +Use mount protocol Version 3, compatible with NFS Version 3. .El .Sh SEE ALSO .Xr mount 1 , diff --git a/usr.bin/showmount/showmount.c b/usr.bin/showmount/showmount.c index e5a3752..8cb342d 100644 --- a/usr.bin/showmount/showmount.c +++ b/usr.bin/showmount/showmount.c @@ -90,6 +90,7 @@ int xdr_mntdump(), xdr_exports(); * This command queries the NFS mount daemon for it's mount list and/or * it's exports list and prints them out. * See "NFS: Network File System Protocol Specification, RFC1094, Appendix A" + * and the "Network File System Protocol XXX.." * for detailed information on the protocol. */ main(argc, argv) @@ -101,12 +102,12 @@ main(argc, argv) register struct grouplist *grp; extern char *optarg; extern int optind; - register int rpcs = 0; + register int rpcs = 0, mntvers = 1; char ch; char *host; int estat; - while ((ch = getopt(argc, argv, "ade")) != EOF) + while ((ch = getopt(argc, argv, "ade3")) != EOF) switch((char)ch) { case 'a': if (type == 0) { @@ -125,6 +126,9 @@ main(argc, argv) case 'e': rpcs |= DOEXPORTS; break; + case '3': + mntvers = 3; + break; case '?': default: usage(); @@ -141,7 +145,7 @@ main(argc, argv) rpcs = DODUMP; if (rpcs & DODUMP) - if ((estat = callrpc(host, RPCPROG_MNT, RPCMNT_VER1, + if ((estat = callrpc(host, RPCPROG_MNT, mntvers, RPCMNT_DUMP, xdr_void, (char *)0, xdr_mntdump, (char *)&mntdump)) != 0) { clnt_perrno(estat); @@ -149,7 +153,7 @@ main(argc, argv) exit(1); } if (rpcs & DOEXPORTS) - if ((estat = callrpc(host, RPCPROG_MNT, RPCMNT_VER1, + if ((estat = callrpc(host, RPCPROG_MNT, mntvers, RPCMNT_EXPORT, xdr_void, (char *)0, xdr_exports, (char *)&exports)) != 0) { clnt_perrno(estat); diff --git a/usr.sbin/amd/amd/afs_ops.c b/usr.sbin/amd/amd/afs_ops.c index 73f39a8..d10a262 100644 --- a/usr.sbin/amd/amd/afs_ops.c +++ b/usr.sbin/amd/amd/afs_ops.c @@ -37,7 +37,7 @@ * * @(#)afs_ops.c 8.1 (Berkeley) 6/6/93 * - * $Id: afs_ops.c,v 1.1.1.1 1994/05/26 05:22:03 rgrimes Exp $ + * $Id: afs_ops.c,v 1.2 1995/05/30 03:45:43 rgrimes Exp $ * */ @@ -144,6 +144,9 @@ char *opts; } NFS_FH_DREF(nfs_args.fh, (NFS_FH_TYPE) fhp); +#ifdef NFSv3 + nfs_args.fhsize = FHSIZE; +#endif /* * Create sockaddr to point to the local machine. 127.0.0.1 diff --git a/usr.sbin/amd/amd/nfs_ops.c b/usr.sbin/amd/amd/nfs_ops.c index 1d92974..eed5b71 100644 --- a/usr.sbin/amd/amd/nfs_ops.c +++ b/usr.sbin/amd/amd/nfs_ops.c @@ -35,7 +35,7 @@ * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * - * $Id: nfs_ops.c,v 1.2 1995/02/13 01:56:22 wpaul Exp $ + * $Id: nfs_ops.c,v 1.3 1995/05/30 03:45:52 rgrimes Exp $ */ #ifndef lint @@ -529,6 +529,9 @@ mntfs *mf; * set mount args */ NFS_FH_DREF(nfs_args.fh, (NFS_FH_TYPE) fhp->fhstatus_u.fhs_fhandle); +#ifdef NFSv3 + nfs_args.fhsize = FHSIZE; +#endif #ifdef ULTRIX_HACK nfs_args.optstr = mnt.mnt_opts; diff --git a/usr.sbin/amd/config/os-bsd44.h b/usr.sbin/amd/config/os-bsd44.h index fa41cfe..863ebfa 100644 --- a/usr.sbin/amd/config/os-bsd44.h +++ b/usr.sbin/amd/config/os-bsd44.h @@ -37,7 +37,7 @@ * * @(#)os-bsd44.h 8.1 (Berkeley) 6/6/93 * - * $Id: os-bsd44.h,v 1.1.1.1 1994/05/26 05:22:07 rgrimes Exp $ + * $Id: os-bsd44.h,v 1.2 1995/01/20 20:58:54 wollman Exp $ * * 4.4 BSD definitions for Amd (automounter) */ @@ -61,6 +61,7 @@ */ #define NFS_44 #define HAS_TCP_NFS +#define NFSv3 /* * Does this OS have NDBM support? @@ -170,7 +171,11 @@ struct mntent { * Type of a file handle */ #undef NFS_FH_TYPE +#ifdef NFSv3 +#define NFS_FH_TYPE fhandle_t * +#else #define NFS_FH_TYPE nfsv2fh_t * +#endif /* * How to get a mount list diff --git a/usr.sbin/mountd/exports.5 b/usr.sbin/mountd/exports.5 index d32527f..fc521f2 100644 --- a/usr.sbin/mountd/exports.5 +++ b/usr.sbin/mountd/exports.5 @@ -49,7 +49,9 @@ file specifies remote mount points for the mount protocol per the .Tn NFS server specification; see -.%T "Network File System Protocol Specification \\*(tNRFC\\*(sP 1094, Appendix A" . +.%T "Network File System Protocol Specification \\*(tNRFC\\*(sP 1094, Appendix A" +and +.%T "NFS: Network File System Version 3 Specification, Appendix I" . .Pp Each line in the file (other than comment lines that begin with a #) @@ -71,7 +73,8 @@ The second is to specify the pathname of the root of the filesystem followed by the .Fl alldirs flag; -this form allows the host(s) to mount any directory within the filesystem. +this form allows the host(s) to mount at any point within the filesystem, +including regular files if the ``-r'' option is used on mountd. The pathnames must not have any symbolic links in them and should not have any "." or ".." components. Mount points for a filesystem may appear on multiple lines each with @@ -140,8 +143,7 @@ The .Fl kerb option specifies that the Kerberos authentication server should be used to authenticate and map client credentials. -(Note that this is NOT Sun NFS compatible and -is supported for TCP transport only.) +This requires that the kernel be built with the NFSKERB option. .Pp The .Fl ro diff --git a/usr.sbin/mountd/mountd.8 b/usr.sbin/mountd/mountd.8 index 78f95b5..7168c17 100644 --- a/usr.sbin/mountd/mountd.8 +++ b/usr.sbin/mountd/mountd.8 @@ -30,7 +30,7 @@ .\" SUCH DAMAGE. .\" .\" From: @(#)mountd.8 8.1 (Berkeley) 6/9/93 -.\" $Id: mountd.8,v 1.2 1994/09/22 22:16:49 wollman Exp $ +.\" $Id: mountd.8,v 1.3 1995/02/22 21:42:48 ats Exp $ .\" .Dd September 22, 1994 .Dt MOUNTD 8 @@ -42,7 +42,7 @@ mount requests .Sh SYNOPSIS .Nm /sbin/mountd -.Op Fl n +.Op Fl nr .Op Ar exportsfile .Sh DESCRIPTION .Xr Mountd @@ -54,7 +54,9 @@ listens for service requests at the port indicated in the .Tn NFS server specification; see .%T "Network File System Protocol Specification" , -RFC1094. +RFC1094, Appendix A and +.%T "NFS: Network File System Version 3 Protocol Specification" , +Appendix I. .Pp Options and operands available for .Nm mountd : @@ -65,6 +67,15 @@ The option allows non-root mount requests to be served. This should only be specified if there are clients such as PC's, that require it. +.It Fl r +The +.Fl r +option allows mount RPCs requests for regular files to be served. +Although this seems to violate the mount protocol specification, some diskless +workstations do mount requests for their swapfiles and expect them to be +regular files. Since a regular file cannot be specified in /etc/exports, the +entire file system that the swapfiles reside in will have to be exported with +the ``-alldirs'' flag. .It Ar exportsfile The .Ar exportsfile diff --git a/usr.sbin/mountd/mountd.c b/usr.sbin/mountd/mountd.c index 4434244..0887c59 100644 --- a/usr.sbin/mountd/mountd.c +++ b/usr.sbin/mountd/mountd.c @@ -43,7 +43,7 @@ static char copyright[] = #ifndef lint /*static char sccsid[] = "From: @(#)mountd.c 8.8 (Berkeley) 2/20/94";*/ static const char rcsid[] = - "$Id: mountd.c,v 1.7.2.1 1995/06/08 04:34:11 davidg Exp $"; + "$Id: mountd.c,v 1.8 1995/06/11 19:30:46 rgrimes Exp $"; #endif /*not lint*/ #include <sys/param.h> @@ -62,7 +62,7 @@ static const char rcsid[] = #include <netiso/iso.h> #endif #include <nfs/rpcv2.h> -#include <nfs/nfsv2.h> +#include <nfs/nfsproto.h> #include <arpa/inet.h> @@ -100,6 +100,8 @@ struct dirlist { }; /* dp_flag bits */ #define DP_DEFSET 0x1 +#define DP_HOSTSET 0x2 +#define DP_KERB 0x4 struct exportlist { struct exportlist *ex_next; @@ -139,22 +141,29 @@ struct grouplist { #define GT_IGNORE 0x5 struct hostlist { + int ht_flag; /* Uses DP_xx bits */ struct grouplist *ht_grp; struct hostlist *ht_next; }; +struct fhreturn { + int fhr_flag; + int fhr_vers; + nfsfh_t fhr_fh; +}; + /* Global defs */ char *add_expdir __P((struct dirlist **, char *, int)); void add_dlist __P((struct dirlist **, struct dirlist *, - struct grouplist *)); + struct grouplist *, int)); void add_mlist __P((char *, char *)); int check_dirpath __P((char *)); int check_options __P((struct dirlist *)); -int chk_host __P((struct dirlist *, u_long, int *)); +int chk_host __P((struct dirlist *, u_long, int *, int *)); void del_mlist __P((char *, char *)); struct dirlist *dirp_search __P((struct dirlist *, char *)); int do_mount __P((struct exportlist *, struct grouplist *, int, - struct ucred *, char *, int, struct statfs *)); + struct ucred *, char *, int, struct statfs *)); int do_opt __P((char **, char **, struct exportlist *, struct grouplist *, int *, int *, struct ucred *)); struct exportlist *ex_search __P((fsid_t *)); @@ -165,6 +174,7 @@ void free_grp __P((struct grouplist *)); void free_host __P((struct hostlist *)); void get_exportlist __P((void)); int get_host __P((char *, struct grouplist *, struct grouplist *)); +int get_num __P((char *)); struct hostlist *get_ht __P((void)); int get_line __P((void)); void get_mountlist __P((void)); @@ -183,7 +193,7 @@ void send_umntall __P((void)); int umntall_each __P((caddr_t, struct sockaddr_in *)); int xdr_dir __P((XDR *, char *)); int xdr_explist __P((XDR *, caddr_t)); -int xdr_fhs __P((XDR *, nfsv2fh_t *)); +int xdr_fhs __P((XDR *, caddr_t)); int xdr_mlist __P((XDR *, caddr_t)); /* C library */ @@ -205,7 +215,8 @@ struct ucred def_anon = { 1, { (gid_t) -2 } }; -int root_only = 1; +int resvport_only = 1; +int dir_only = 1; int opt_flags; /* Bits for above */ #define OP_MAPROOT 0x01 @@ -238,6 +249,7 @@ main(argc, argv) { SVCXPRT *udptransp, *tcptransp; int c; +#ifdef __FreeBSD__ struct vfsconf *vfc; vfc = getvfsbyname("nfs"); @@ -250,17 +262,21 @@ main(argc, argv) if(!vfc) { errx(1, "NFS support is not available in the running kernel"); } +#endif /* __FreeBSD__ */ - while ((c = getopt(argc, argv, "dn")) != EOF) + while ((c = getopt(argc, argv, "dnr")) != EOF) switch (c) { + case 'n': + resvport_only = 0; + break; + case 'r': + dir_only = 0; + break; case 'd': debug = debug ? 0 : 1; break; - case 'n': - root_only = 0; - break; default: - fprintf(stderr, "Usage: mountd [-n] [export_file]\n"); + fprintf(stderr, "Usage: mountd [-r] [-n] [export_file]\n"); exit(1); }; argc -= optind; @@ -300,11 +316,12 @@ main(argc, argv) syslog(LOG_ERR, "Can't create socket"); exit(1); } - pmap_unset(RPCPROG_MNT, RPCMNT_VER1); - if (!svc_register(udptransp, RPCPROG_MNT, RPCMNT_VER1, mntsrv, - IPPROTO_UDP) || - !svc_register(tcptransp, RPCPROG_MNT, RPCMNT_VER1, mntsrv, - IPPROTO_TCP)) { + pmap_unset(RPCPROG_MNT, 1); + pmap_unset(RPCPROG_MNT, 3); + if (!svc_register(udptransp, RPCPROG_MNT, 1, mntsrv, IPPROTO_UDP) || + !svc_register(udptransp, RPCPROG_MNT, 3, mntsrv, IPPROTO_UDP) || + !svc_register(tcptransp, RPCPROG_MNT, 1, mntsrv, IPPROTO_TCP) || + !svc_register(tcptransp, RPCPROG_MNT, 3, mntsrv, IPPROTO_TCP)) { syslog(LOG_ERR, "Can't register mount"); exit(1); } @@ -323,28 +340,21 @@ mntsrv(rqstp, transp) { struct exportlist *ep; struct dirlist *dp; - nfsv2fh_t nfh; + struct fhreturn fhr; struct authunix_parms *ucr; struct stat stb; struct statfs fsb; struct hostent *hp; u_long saddr; + u_short sport; char rpcpath[RPCMNT_PATHLEN+1], dirpath[MAXPATHLEN]; - int bad = ENOENT, omask, defset; - uid_t uid = -2; - - /* Get authorization */ - switch (rqstp->rq_cred.oa_flavor) { - case AUTH_UNIX: - ucr = (struct authunix_parms *)rqstp->rq_clntcred; - uid = ucr->aup_uid; - break; - case AUTH_NULL: - default: - break; - } + int bad = ENOENT, defset, hostset; + sigset_t sighup_mask; + sigemptyset(&sighup_mask); + sigaddset(&sighup_mask, SIGHUP); saddr = transp->xp_raddr.sin_addr.s_addr; + sport = ntohs(transp->xp_raddr.sin_port); hp = (struct hostent *)NULL; switch (rqstp->rq_proc) { case NULLPROC: @@ -352,7 +362,7 @@ mntsrv(rqstp, transp) syslog(LOG_ERR, "Can't send reply"); return; case RPCMNT_MOUNT: - if ((uid != 0 && root_only) || uid == -2) { + if (sport >= IPPORT_RESERVED && resvport_only) { svcerr_weakauth(transp); return; } @@ -363,11 +373,13 @@ mntsrv(rqstp, transp) /* * Get the real pathname and make sure it is a directory - * that exists. + * or a regular file if the -r option was specified + * and it exists. */ if (realpath(rpcpath, dirpath) == 0 || stat(dirpath, &stb) < 0 || - (stb.st_mode & S_IFMT) != S_IFDIR || + (!S_ISDIR(stb.st_mode) && + (dir_only || !S_ISREG(stb.st_mode))) || statfs(dirpath, &fsb) < 0) { chdir("/"); /* Just in case realpath doesn't */ if (debug) @@ -378,26 +390,31 @@ mntsrv(rqstp, transp) } /* Check in the exports list */ - omask = sigblock(sigmask(SIGHUP)); + sigprocmask(SIG_BLOCK, &sighup_mask, NULL); ep = ex_search(&fsb.f_fsid); - defset = 0; - if (ep && (chk_host(ep->ex_defdir, saddr, &defset) || + hostset = defset = 0; + if (ep && (chk_host(ep->ex_defdir, saddr, &defset, &hostset) || ((dp = dirp_search(ep->ex_dirl, dirpath)) && - chk_host(dp, saddr, &defset)) || + chk_host(dp, saddr, &defset, &hostset)) || (defset && scan_tree(ep->ex_defdir, saddr) == 0 && scan_tree(ep->ex_dirl, saddr) == 0))) { + if (hostset & DP_HOSTSET) + fhr.fhr_flag = hostset; + else + fhr.fhr_flag = defset; + fhr.fhr_vers = rqstp->rq_vers; /* Get the file handle */ - bzero((caddr_t)&nfh, sizeof(nfh)); - if (getfh(dirpath, (fhandle_t *)&nfh) < 0) { + bzero((caddr_t)&fhr.fhr_fh, sizeof(nfsfh_t)); + if (getfh(dirpath, (fhandle_t *)&fhr.fhr_fh) < 0) { bad = errno; syslog(LOG_ERR, "Can't get fh for %s", dirpath); if (!svc_sendreply(transp, xdr_long, (caddr_t)&bad)) syslog(LOG_ERR, "Can't send reply"); - sigsetmask(omask); + sigprocmask(SIG_UNBLOCK, &sighup_mask, NULL); return; } - if (!svc_sendreply(transp, xdr_fhs, (caddr_t)&nfh)) + if (!svc_sendreply(transp, xdr_fhs, (caddr_t)&fhr)) syslog(LOG_ERR, "Can't send reply"); if (hp == NULL) hp = gethostbyaddr((caddr_t)&saddr, @@ -414,14 +431,14 @@ mntsrv(rqstp, transp) if (!svc_sendreply(transp, xdr_long, (caddr_t)&bad)) syslog(LOG_ERR, "Can't send reply"); } - sigsetmask(omask); + sigprocmask(SIG_UNBLOCK, &sighup_mask, NULL); return; case RPCMNT_DUMP: if (!svc_sendreply(transp, xdr_mlist, (caddr_t)NULL)) syslog(LOG_ERR, "Can't send reply"); return; case RPCMNT_UMOUNT: - if ((uid != 0 && root_only) || uid == -2) { + if (sport >= IPPORT_RESERVED && resvport_only) { svcerr_weakauth(transp); return; } @@ -437,7 +454,7 @@ mntsrv(rqstp, transp) del_mlist(inet_ntoa(transp->xp_raddr.sin_addr), dirpath); return; case RPCMNT_UMNTALL: - if ((uid != 0 && root_only) || uid == -2) { + if (sport >= IPPORT_RESERVED && resvport_only) { svcerr_weakauth(transp); return; } @@ -470,18 +487,37 @@ xdr_dir(xdrsp, dirp) } /* - * Xdr routine to generate fhstatus + * Xdr routine to generate file handle reply */ int -xdr_fhs(xdrsp, nfh) +xdr_fhs(xdrsp, cp) XDR *xdrsp; - nfsv2fh_t *nfh; + caddr_t cp; { - u_long ok = 0; + register struct fhreturn *fhrp = (struct fhreturn *)cp; + u_long ok = 0, len, auth; if (!xdr_long(xdrsp, &ok)) return (0); - return (xdr_opaque(xdrsp, (caddr_t)nfh, NFSX_FH)); + switch (fhrp->fhr_vers) { + case 1: + return (xdr_opaque(xdrsp, (caddr_t)&fhrp->fhr_fh, NFSX_V2FH)); + case 3: + len = NFSX_V3FH; + if (!xdr_long(xdrsp, &len)) + return (0); + if (!xdr_opaque(xdrsp, (caddr_t)&fhrp->fhr_fh, len)) + return (0); + if (fhrp->fhr_flag & DP_KERB) + auth = RPCAUTH_KERB4; + else + auth = RPCAUTH_UNIX; + len = 1; + if (!xdr_long(xdrsp, &len)) + return (0); + return (xdr_long(xdrsp, &auth)); + }; + return (0); } int @@ -521,9 +557,12 @@ xdr_explist(xdrsp, cp) { struct exportlist *ep; int false = 0; - int omask, putdef; + int putdef; + sigset_t sighup_mask; - omask = sigblock(sigmask(SIGHUP)); + sigemptyset(&sighup_mask); + sigaddset(&sighup_mask, SIGHUP); + sigprocmask(SIG_BLOCK, &sighup_mask, NULL); ep = exphead; while (ep) { putdef = 0; @@ -535,12 +574,12 @@ xdr_explist(xdrsp, cp) goto errout; ep = ep->ex_next; } - sigsetmask(omask); + sigprocmask(SIG_UNBLOCK, &sighup_mask, NULL); if (!xdr_bool(xdrsp, &false)) return (0); return (1); errout: - sigsetmask(omask); + sigprocmask(SIG_UNBLOCK, &sighup_mask, NULL); return (0); } @@ -660,6 +699,18 @@ get_exportlist() struct ufs_args ua; struct iso_args ia; struct mfs_args ma; +#ifdef __NetBSD__ + struct msdosfs_args da; + } targs; + + if (!strcmp(fsp->f_fstypename, MOUNT_MFS) || + !strcmp(fsp->f_fstypename, MOUNT_UFS) || + !strcmp(fsp->f_fstypename, MOUNT_MSDOS) || + !strcmp(fsp->f_fstypename, MOUNT_CD9660)) { + targs.ua.fspec = NULL; + targs.ua.export.ex_flags = MNT_DELEXPORT; + if (mount(fsp->f_fstypename, fsp->f_mntonname, +#else } targs; switch (fsp->f_type) { @@ -670,6 +721,7 @@ get_exportlist() targs.ua.fspec = NULL; targs.ua.export.ex_flags = MNT_DELEXPORT; if (mount(fsp->f_type, fsp->f_mntonname, +#endif fsp->f_flags | MNT_UPDATE, (caddr_t)&targs) < 0) syslog(LOG_ERR, "Can't delete exports for %s", @@ -805,6 +857,7 @@ get_exportlist() if (get_host(hst, grp, tgrp)) { syslog(LOG_ERR, "Bad netgroup %s", cp); getexp_err(ep, tgrp); + endnetgrent(); goto nextline; } } else if (get_host(cp, grp, tgrp)) { @@ -865,12 +918,12 @@ get_exportlist() * Success. Update the data structures. */ if (has_host) { - hang_dirp(dirhead, tgrp, ep, (opt_flags & OP_ALLDIRS)); + hang_dirp(dirhead, tgrp, ep, opt_flags); grp->gr_next = grphead; grphead = tgrp; } else { hang_dirp(dirhead, (struct grouplist *)NULL, ep, - (opt_flags & OP_ALLDIRS)); + opt_flags); free_grp(grp); } dirhead = (struct dirlist *)NULL; @@ -994,24 +1047,28 @@ add_expdir(dpp, cp, len) * and update the entry for host. */ void -hang_dirp(dp, grp, ep, alldirs) +hang_dirp(dp, grp, ep, flags) struct dirlist *dp; struct grouplist *grp; struct exportlist *ep; - int alldirs; + int flags; { struct hostlist *hp; struct dirlist *dp2; - if (alldirs) { + if (flags & OP_ALLDIRS) { if (ep->ex_defdir) free((caddr_t)dp); else ep->ex_defdir = dp; - if (grp == (struct grouplist *)NULL) + if (grp == (struct grouplist *)NULL) { ep->ex_defdir->dp_flag |= DP_DEFSET; - else while (grp) { + if (flags & OP_KERB) + ep->ex_defdir->dp_flag |= DP_KERB; + } else while (grp) { hp = get_ht(); + if (flags & OP_KERB) + hp->ht_flag |= DP_KERB; hp->ht_grp = grp; hp->ht_next = ep->ex_defdir->dp_hosts; ep->ex_defdir->dp_hosts = hp; @@ -1024,7 +1081,7 @@ hang_dirp(dp, grp, ep, alldirs) */ while (dp) { dp2 = dp->dp_left; - add_dlist(&ep->ex_dirl, dp, grp); + add_dlist(&ep->ex_dirl, dp, grp, flags); dp = dp2; } } @@ -1035,10 +1092,11 @@ hang_dirp(dp, grp, ep, alldirs) * for the new directory or adding the new node. */ void -add_dlist(dpp, newdp, grp) +add_dlist(dpp, newdp, grp, flags) struct dirlist **dpp; struct dirlist *newdp; struct grouplist *grp; + int flags; { struct dirlist *dp; struct hostlist *hp; @@ -1048,10 +1106,10 @@ add_dlist(dpp, newdp, grp) if (dp) { cmp = strcmp(dp->dp_dirp, newdp->dp_dirp); if (cmp > 0) { - add_dlist(&dp->dp_left, newdp, grp); + add_dlist(&dp->dp_left, newdp, grp, flags); return; } else if (cmp < 0) { - add_dlist(&dp->dp_right, newdp, grp); + add_dlist(&dp->dp_right, newdp, grp, flags); return; } else free((caddr_t)newdp); @@ -1067,13 +1125,18 @@ add_dlist(dpp, newdp, grp) */ do { hp = get_ht(); + if (flags & OP_KERB) + hp->ht_flag |= DP_KERB; hp->ht_grp = grp; hp->ht_next = dp->dp_hosts; dp->dp_hosts = hp; grp = grp->gr_next; } while (grp); - } else + } else { dp->dp_flag |= DP_DEFSET; + if (flags & OP_KERB) + dp->dp_flag |= DP_KERB; + } } /* @@ -1102,10 +1165,11 @@ dirp_search(dp, dirpath) * Scan for a host match in a directory tree. */ int -chk_host(dp, saddr, defsetp) +chk_host(dp, saddr, defsetp, hostsetp) struct dirlist *dp; u_long saddr; int *defsetp; + int *hostsetp; { struct hostlist *hp; struct grouplist *grp; @@ -1113,7 +1177,7 @@ chk_host(dp, saddr, defsetp) if (dp) { if (dp->dp_flag & DP_DEFSET) - *defsetp = 1; + *defsetp = dp->dp_flag; hp = dp->dp_hosts; while (hp) { grp = hp->ht_grp; @@ -1122,15 +1186,19 @@ chk_host(dp, saddr, defsetp) addrp = (u_long **) grp->gr_ptr.gt_hostent->h_addr_list; while (*addrp) { - if (**addrp == saddr) + if (**addrp == saddr) { + *hostsetp = (hp->ht_flag | DP_HOSTSET); return (1); + } addrp++; } break; case GT_NET: if ((saddr & grp->gr_ptr.gt_net.nt_mask) == - grp->gr_ptr.gt_net.nt_net) + grp->gr_ptr.gt_net.nt_net) { + *hostsetp = (hp->ht_flag | DP_HOSTSET); return (1); + } break; }; hp = hp->ht_next; @@ -1147,12 +1215,12 @@ scan_tree(dp, saddr) struct dirlist *dp; u_long saddr; { - int defset; + int defset, hostset; if (dp) { if (scan_tree(dp->dp_left, saddr)) return (1); - if (chk_host(dp, saddr, &defset)) + if (chk_host(dp, saddr, &defset, &hostset)) return (1); if (scan_tree(dp->dp_right, saddr)) return (1); @@ -1312,7 +1380,7 @@ get_host(cp, grp, tgrp) if (isdigit(*cp)) { saddr = inet_addr(cp); if (saddr == -1) { - syslog(LOG_ERR, "Inet_addr failed"); + syslog(LOG_ERR, "Inet_addr failed for %s", cp); return (1); } if ((hp = gethostbyaddr((caddr_t)&saddr, sizeof (saddr), @@ -1326,7 +1394,7 @@ get_host(cp, grp, tgrp) aptr[1] = (char *)NULL; } } else { - syslog(LOG_ERR, "Gethostbyname failed"); + syslog(LOG_ERR, "Gethostbyname failed for %s", cp); return (1); } } @@ -1424,6 +1492,7 @@ get_ht() if (hp == (struct hostlist *)NULL) out_of_mem(); hp->ht_next = (struct hostlist *)NULL; + hp->ht_flag = 0; return (hp); } @@ -1494,6 +1563,9 @@ do_mount(ep, grp, exflags, anoncrp, dirp, dirplen, fsb) struct ufs_args ua; struct iso_args ia; struct mfs_args ma; +#ifdef __NetBSD__ + struct msdosfs_args da; +#endif } args; u_long net; @@ -1569,7 +1641,11 @@ do_mount(ep, grp, exflags, anoncrp, dirp, dirplen, fsb) * Also, needs to know how to export all types of local * exportable file systems and not just MOUNT_UFS. */ +#ifdef __NetBSD__ + while (mount(fsb->f_fstypename, dirp, +#else while (mount(fsb->f_type, dirp, +#endif fsb->f_flags | MNT_UPDATE, (caddr_t)&args) < 0) { if (cp) *cp-- = savedc; @@ -2033,15 +2109,30 @@ check_dirpath(dirp) while (*cp && ret) { if (*cp == '/') { *cp = '\0'; - if (lstat(dirp, &sb) < 0 || - (sb.st_mode & S_IFMT) != S_IFDIR) + if (lstat(dirp, &sb) < 0 || !S_ISDIR(sb.st_mode)) ret = 0; *cp = '/'; } cp++; } - if (lstat(dirp, &sb) < 0 || - (sb.st_mode & S_IFMT) != S_IFDIR) + if (lstat(dirp, &sb) < 0 || !S_ISDIR(sb.st_mode)) ret = 0; return (ret); } + +/* + * Just translate an ascii string to an integer. + */ +int +get_num(cp) + register char *cp; +{ + register int res = 0; + + while (*cp) { + if (*cp < '0' || *cp > '9') + return (-1); + res = res * 10 + (*cp++ - '0'); + } + return (res); +} diff --git a/usr.sbin/nfsd/nfsd.8 b/usr.sbin/nfsd/nfsd.8 index a54a564..24573d9 100644 --- a/usr.sbin/nfsd/nfsd.8 +++ b/usr.sbin/nfsd/nfsd.8 @@ -98,7 +98,8 @@ listens for service requests at the port indicated in the .Tn NFS server specification; see .%T "Network File System Protocol Specification" , -RFC1094. +RFC1094 and +.%T "NFS: Network File System Version 3 Protocol Specification" . .Pp If .Nm nfsd diff --git a/usr.sbin/nfsd/nfsd.c b/usr.sbin/nfsd/nfsd.c index 9d4d226..8c8731d 100644 --- a/usr.sbin/nfsd/nfsd.c +++ b/usr.sbin/nfsd/nfsd.c @@ -63,10 +63,10 @@ static char sccsid[] = "@(#)nfsd.c 8.7 (Berkeley) 2/22/94"; #include <netiso/iso.h> #endif #include <nfs/rpcv2.h> -#include <nfs/nfsv2.h> +#include <nfs/nfsproto.h> #include <nfs/nfs.h> -#ifdef KERBEROS +#ifdef NFSKERB #include <kerberosIV/des.h> #include <kerberosIV/krb.h> #endif @@ -94,16 +94,23 @@ struct nfsd_srvargs nsd; char **Argv = NULL; /* pointer to argument vector */ char *LastArg = NULL; /* end of argv */ -#ifdef KERBEROS +#ifdef NFSKERB char lnam[ANAME_SZ]; KTEXT_ST kt; -AUTH_DAT auth; +AUTH_DAT kauth; char inst[INST_SZ]; +struct nfsrpc_fullblock kin, kout; +struct nfsrpc_fullverf kverf; +NFSKERBKEY_T kivec; +struct timeval ktv; +NFSKERBKEYSCHED_T kerb_keysched; #endif void nonfs __P((int)); void reapchild __P((int)); +#ifdef __FreeBSD__ void setproctitle __P((char *)); +#endif void usage __P((void)); /* @@ -139,11 +146,13 @@ main(argc, argv, envp) #ifdef ISO struct sockaddr_iso isoaddr, isopeer; #endif + struct timeval ktv; fd_set ready, sockbits; int ch, cltpflag, connect_type_cnt, i, len, maxsock, msgsock; int nfsdcnt, nfssvc_flag, on, reregister, sock, tcpflag, tcpsock; int tp4cnt, tp4flag, tp4sock, tpipcnt, tpipflag, tpipsock, udpflag; char *cp, **cpp; +#ifdef __FreeBSD__ struct vfsconf *vfc; vfc = getvfsbyname("nfs"); @@ -156,6 +165,7 @@ main(argc, argv, envp) if(!vfc) { errx(1, "NFS is not available in the running kernel"); } +#endif /* Save start and extent of argv for setproctitle. */ Argv = argv; @@ -241,10 +251,12 @@ main(argc, argv, envp) if (reregister) { if (udpflag && - !pmap_set(RPCPROG_NFS, NFS_VER2, IPPROTO_UDP, NFS_PORT)) + (!pmap_set(RPCPROG_NFS, 2, IPPROTO_UDP, NFS_PORT) || + !pmap_set(RPCPROG_NFS, 3, IPPROTO_UDP, NFS_PORT))) err(1, "can't register with portmap for UDP."); if (tcpflag && - !pmap_set(RPCPROG_NFS, NFS_VER2, IPPROTO_TCP, NFS_PORT)) + (!pmap_set(RPCPROG_NFS, 2, IPPROTO_TCP, NFS_PORT) || + !pmap_set(RPCPROG_NFS, 3, IPPROTO_TCP, NFS_PORT))) err(1, "can't register with portmap for TCP."); exit(0); } @@ -261,11 +273,17 @@ main(argc, argv, envp) continue; } - setproctitle("nfsd-srv"); + setproctitle("server"); nfssvc_flag = NFSSVC_NFSD; nsd.nsd_nfsd = NULL; -#ifdef KERBEROS - nsd.nsd_authstr = (char *)kt.dat; +#ifdef NFSKERB + if (sizeof (struct nfsrpc_fullverf) != RPCX_FULLVERF || + sizeof (struct nfsrpc_fullblock) != RPCX_FULLBLOCK) + syslog(LOG_ERR, "Yikes NFSKERB structs not packed!"); + nsd.nsd_authstr = (u_char *)&kt; + nsd.nsd_authlen = sizeof (kt); + nsd.nsd_verfstr = (u_char *)&kverf; + nsd.nsd_verflen = sizeof (kverf); #endif while (nfssvc(nfssvc_flag, &nsd) < 0) { if (errno != ENEEDAUTH) { @@ -273,14 +291,27 @@ main(argc, argv, envp) exit(1); } nfssvc_flag = NFSSVC_NFSD | NFSSVC_AUTHINFAIL; -#ifdef KERBEROS - kt.length = nsd.nsd_authlen; - kt.mbz = 0; - (void)strcpy(inst, "*"); - if (krb_rd_req(&kt, "rcmd", - inst, nsd.nsd_haddr, &auth, "") == RD_AP_OK && - krb_kntoln(&auth, lnam) == KSUCCESS && - (pwd = getpwnam(lnam)) != NULL) { +#ifdef NFSKERB + /* + * Get the Kerberos ticket out of the authenticator + * verify it and convert the principal name to a user + * name. The user name is then converted to a set of + * user credentials via the password and group file. + * Finally, decrypt the timestamp and validate it. + * For more info see the IETF Draft "Authentication + * in ONC RPC". + */ + kt.length = ntohl(kt.length); + if (gettimeofday(&ktv, (struct timezone *)0) == 0 && + kt.length > 0 && kt.length <= + (RPCAUTH_MAXSIZ - 3 * NFSX_UNSIGNED)) { + kin.w1 = NFS_KERBW1(kt); + kt.mbz = 0; + (void)strcpy(inst, "*"); + if (krb_rd_req(&kt, NFS_KERBSRV, + inst, nsd.nsd_haddr, &kauth, "") == RD_AP_OK && + krb_kntoln(&kauth, lnam) == KSUCCESS && + (pwd = getpwnam(lnam)) != NULL) { cr = &nsd.nsd_cr; cr->cr_uid = pwd->pw_uid; cr->cr_groups[0] = pwd->pw_gid; @@ -301,9 +332,34 @@ main(argc, argv, envp) break; } endgrent(); - nfssvc_flag = NFSSVC_NFSD | NFSSVC_AUTHIN; + + /* + * Get the timestamp verifier out of the + * authenticator and verifier strings. + */ + kin.t1 = kverf.t1; + kin.t2 = kverf.t2; + kin.w2 = kverf.w2; + bzero((caddr_t)kivec, sizeof (kivec)); + bcopy((caddr_t)kauth.session, + (caddr_t)nsd.nsd_key,sizeof(kauth.session)); + + /* + * Decrypt the timestamp verifier in CBC mode. + */ + XXX + + /* + * Validate the timestamp verifier, to + * check that the session key is ok. + */ + nsd.nsd_timestamp.tv_sec = ntohl(kout.t1); + nsd.nsd_timestamp.tv_usec = ntohl(kout.t2); + nsd.nsd_ttl = ntohl(kout.w1); + if ((nsd.nsd_ttl - 1) == ntohl(kout.w2)) + nfssvc_flag = NFSSVC_NFSD | NFSSVC_AUTHIN; } -#endif /* KERBEROS */ +#endif /* NFSKERB */ } exit(0); } @@ -323,7 +379,8 @@ main(argc, argv, envp) syslog(LOG_ERR, "can't bind udp addr"); exit(1); } - if (!pmap_set(RPCPROG_NFS, NFS_VER2, IPPROTO_UDP, NFS_PORT)) { + if (!pmap_set(RPCPROG_NFS, 2, IPPROTO_UDP, NFS_PORT) || + !pmap_set(RPCPROG_NFS, 3, IPPROTO_UDP, NFS_PORT)) { syslog(LOG_ERR, "can't register with udp portmap"); exit(1); } @@ -403,7 +460,8 @@ main(argc, argv, envp) syslog(LOG_ERR, "listen failed"); exit(1); } - if (!pmap_set(RPCPROG_NFS, NFS_VER2, IPPROTO_TCP, NFS_PORT)) { + if (!pmap_set(RPCPROG_NFS, 2, IPPROTO_TCP, NFS_PORT) || + !pmap_set(RPCPROG_NFS, 3, IPPROTO_TCP, NFS_PORT)) { syslog(LOG_ERR, "can't register tcp with portmap"); exit(1); } @@ -492,7 +550,7 @@ main(argc, argv, envp) if (connect_type_cnt == 0) exit(0); - setproctitle("nfsd-master"); + setproctitle("master"); /* * Loop forever accepting connections and passing the sockets @@ -566,7 +624,7 @@ main(argc, argv, envp) void usage() { - (void)fprintf(stderr, "nfsd %s\n", USAGE); + (void)fprintf(stderr, "usage: nfsd %s\n", USAGE); exit(1); } @@ -582,9 +640,10 @@ reapchild(signo) int signo; { - while (wait3(NULL, WNOHANG, NULL)); + while (wait3(NULL, WNOHANG, NULL) > 0); } +#ifdef __FreeBSD__ void setproctitle(a) char *a; @@ -593,9 +652,10 @@ setproctitle(a) char buf[80]; cp = Argv[0]; - (void)snprintf(buf, sizeof(buf), "%s", a); + (void)snprintf(buf, sizeof(buf), "nfsd-%s", a); (void)strncpy(cp, buf, LastArg - cp); cp += strlen(cp); while (cp < LastArg) *cp++ = '\0'; } +#endif /* __FreeBSD__ */ diff --git a/usr.sbin/pstat/pstat.c b/usr.sbin/pstat/pstat.c index 736c96dd..3693557 100644 --- a/usr.sbin/pstat/pstat.c +++ b/usr.sbin/pstat/pstat.c @@ -54,6 +54,8 @@ static char sccsid[] = "@(#)pstat.c 8.9 (Berkeley) 2/16/94"; #undef NFS #undef KERNEL #include <sys/stat.h> +#include <nfs/rpcv2.h> +#include <nfs/nfsproto.h> #include <nfs/nfsnode.h> #include <sys/ioctl.h> #include <sys/ioctl_compat.h> /* XXX NTTYDISC is too well hidden */ |
