The socket code upcalls into the NFS server using the so_upcall

mechanism so that early processing on mbufs can be performed before
a context switch to the NFS server threads.  Because of this, if
the socket code is running without Giant, the NFS server also needs
to be able to run the upcall code without relying on the presence on
Giant.  This change modifies the NFS server to run using a "giant
code lock" covering operation of the whole subsystem.  Work is in
progress to move to data-based locking as part of the NFSv4 server
changes.

Introduce an NFS server subsystem lock, 'nfsd_mtx', and a set of
macros to operate on the lock:

  NFSD_LOCK_ASSERT()    Assert nfsd_mtx owned by current thread
  NFSD_UNLOCK_ASSERT()  Assert nfsd_mtx not owned by current thread
  NFSD_LOCK_DONTCARE()  Advisory: this function doesn't care
  NFSD_LOCK()           Lock nfsd_mtx
  NFSD_UNLOCK()         Unlock nfsd_mtx

Constify a number of global variables/structures in the NFS server
code, as they are not modified and contain constants only:

  nfsrvv2_procid       nfsrv_nfsv3_procid      nonidempotent
  nfsv2_repstat        nfsv2_type              nfsrv_nfsv3_procid
  nfsrvv2_procid       nfsrv_v2errmap          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
  nfsrv_v3errmap

There are additional structures that should be constified but due
to their being passed into general purpose functions without const
arguments, I have not yet converted.

In general, acquire nfsd_mtx when accessing any of the global NFS
structures, including struct nfssvc_sock, struct nfsd, struct
nfsrv_descript.

Release nfsd_mtx whenever calling into VFS, and acquire Giant for
calls into VFS.  Giant is not required for any part of the
operation of the NFS server with the exception of calls into VFS.
Giant will never by acquired in the upcall code path.  However, it
may operate entirely covered by Giant, or not.  If debug.mpsafenet
is set to 0, the system calls will acquire Giant across all
operations, and the upcall will assert Giant.  As such, by default,
this enables locking and allows us to test assertions, but should not
cause any substantial new amount of code to be run without Giant.
Bugs should manifest in the form of lock assertion failures for now.

This approach is similar (but not identical) to modifications to the
BSD/OS NFS server code snapshot provided by BSDi as part of their
SMPng snapshot.  The strategy is almost the same (single lock over
the NFS server), but differs in the following ways:

- Our NFS client and server code bases don't overlap, which means
  both fewer bugs and easier locking (thanks Peter!).  Also means
  NFSD_*() as opposed to NFS_*().

- We make broad use of assertions, whereas the BSD/OS code does not.

- Made slightly different choices about how to handle macros building
  packets but operating with side effects.

- We acquire Giant only when entering VFS from the NFS server daemon
  threads.

- Serious bugs in BSD/OS implementation corrected -- the snapshot we
  received was clearly a work in progress.

Based on ideas from:	BSDi SMPng Snapshot
Reviewed by:		rick@snowhite.cis.uoguelph.ca
Extensive testing by:	kris

1 files changed, 97 insertions, 36 deletions

diff --git a/sys/nfsserver/nfs_srvsubs.c b/sys/nfsserver/nfs_srvsubs.c
index 9da4b1a..7217150 100644
--- a/sys/nfsserver/nfs_srvsubs.c
+++ b/sys/nfsserver/nfs_srvsubs.c
@@ -85,8 +85,8 @@ u_int32_t nfsrv_rpc_call, nfsrv_rpc_vers, nfsrv_rpc_reply,
 u_int32_t nfsrv_nfs_prog, nfsrv_nfs_true, nfsrv_nfs_false;
 
 /* And other global data */
-static nfstype nfsv2_type[9] = { NFNON, NFREG, NFDIR, NFBLK, NFCHR, NFLNK,
-				 NFNON, NFCHR, NFNON };
+static const nfstype nfsv2_type[9] = { NFNON, NFREG, NFDIR, NFBLK, NFCHR,
+				       NFLNK, NFNON, NFCHR, NFNON };
 #define vtonfsv2_type(a)	txdr_unsigned(nfsv2_type[((int32_t)(a))])
 #define vtonfsv3_mode(m)	txdr_unsigned((m) & ALLPERMS)
 
@@ -100,10 +100,12 @@ int nfsd_head_flag;
 static int nfs_prev_nfssvc_sy_narg;
 static sy_call_t *nfs_prev_nfssvc_sy_call;
 
+struct mtx nfsd_mtx;
+
 /*
  * Mapping of old NFS Version 2 RPC numbers to generic numbers.
  */
-int nfsrv_nfsv3_procid[NFS_NPROCS] = {
+const int nfsrv_nfsv3_procid[NFS_NPROCS] = {
 	NFSPROC_NULL,
 	NFSPROC_GETATTR,
 	NFSPROC_SETATTR,
@@ -132,7 +134,7 @@ int nfsrv_nfsv3_procid[NFS_NPROCS] = {
 /*
  * and the reverse mapping from generic to Version 2 procedure numbers
  */
-int nfsrvv2_procid[NFS_NPROCS] = {
+const int nfsrvv2_procid[NFS_NPROCS] = {
 	NFSV2PROC_NULL,
 	NFSV2PROC_GETATTR,
 	NFSV2PROC_SETATTR,
@@ -163,7 +165,7 @@ int nfsrvv2_procid[NFS_NPROCS] = {
  * Use 0 (which gets converted to NFSERR_IO) as the catch all for ones not
  * specifically defined in RFC 1094.
  */
-static u_char nfsrv_v2errmap[ELAST] = {
+static const u_char nfsrv_v2errmap[ELAST] = {
   NFSERR_PERM,	NFSERR_NOENT,	0,		0,		0,	
   NFSERR_NXIO,	0,		0,		0,		0,	
   0,		0,		NFSERR_ACCES,	0,		0,	
@@ -192,12 +194,12 @@ static u_char nfsrv_v2errmap[ELAST] = {
  * 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[] = {
+static const short nfsv3err_null[] = {
 	0,
 	0,
 };
 
-static short nfsv3err_getattr[] = {
+static const short nfsv3err_getattr[] = {
 	NFSERR_IO,
 	NFSERR_IO,
 	NFSERR_STALE,
@@ -206,7 +208,7 @@ static short nfsv3err_getattr[] = {
 	0,
 };
 
-static short nfsv3err_setattr[] = {
+static const short nfsv3err_setattr[] = {
 	NFSERR_IO,
 	NFSERR_PERM,
 	NFSERR_IO,
@@ -222,7 +224,7 @@ static short nfsv3err_setattr[] = {
 	0,
 };
 
-static short nfsv3err_lookup[] = {
+static const short nfsv3err_lookup[] = {
 	NFSERR_IO,
 	NFSERR_NOENT,
 	NFSERR_IO,
@@ -235,7 +237,7 @@ static short nfsv3err_lookup[] = {
 	0,
 };
 
-static short nfsv3err_access[] = {
+static const short nfsv3err_access[] = {
 	NFSERR_IO,
 	NFSERR_IO,
 	NFSERR_STALE,
@@ -244,7 +246,7 @@ static short nfsv3err_access[] = {
 	0,
 };
 
-static short nfsv3err_readlink[] = {
+static const short nfsv3err_readlink[] = {
 	NFSERR_IO,
 	NFSERR_IO,
 	NFSERR_ACCES,
@@ -256,7 +258,7 @@ static short nfsv3err_readlink[] = {
 	0,
 };
 
-static short nfsv3err_read[] = {
+static const short nfsv3err_read[] = {
 	NFSERR_IO,
 	NFSERR_IO,
 	NFSERR_NXIO,
@@ -268,7 +270,7 @@ static short nfsv3err_read[] = {
 	0,
 };
 
-static short nfsv3err_write[] = {
+static const short nfsv3err_write[] = {
 	NFSERR_IO,
 	NFSERR_IO,
 	NFSERR_ACCES,
@@ -283,7 +285,7 @@ static short nfsv3err_write[] = {
 	0,
 };
 
-static short nfsv3err_create[] = {
+static const short nfsv3err_create[] = {
 	NFSERR_IO,
 	NFSERR_IO,
 	NFSERR_ACCES,
@@ -300,7 +302,7 @@ static short nfsv3err_create[] = {
 	0,
 };
 
-static short nfsv3err_mkdir[] = {
+static const short nfsv3err_mkdir[] = {
 	NFSERR_IO,
 	NFSERR_IO,
 	NFSERR_ACCES,
@@ -317,7 +319,7 @@ static short nfsv3err_mkdir[] = {
 	0,
 };
 
-static short nfsv3err_symlink[] = {
+static const short nfsv3err_symlink[] = {
 	NFSERR_IO,
 	NFSERR_IO,
 	NFSERR_ACCES,
@@ -334,7 +336,7 @@ static short nfsv3err_symlink[] = {
 	0,
 };
 
-static short nfsv3err_mknod[] = {
+static const short nfsv3err_mknod[] = {
 	NFSERR_IO,
 	NFSERR_IO,
 	NFSERR_ACCES,
@@ -352,7 +354,7 @@ static short nfsv3err_mknod[] = {
 	0,
 };
 
-static short nfsv3err_remove[] = {
+static const short nfsv3err_remove[] = {
 	NFSERR_IO,
 	NFSERR_NOENT,
 	NFSERR_IO,
@@ -366,7 +368,7 @@ static short nfsv3err_remove[] = {
 	0,
 };
 
-static short nfsv3err_rmdir[] = {
+static const short nfsv3err_rmdir[] = {
 	NFSERR_IO,
 	NFSERR_NOENT,
 	NFSERR_IO,
@@ -384,7 +386,7 @@ static short nfsv3err_rmdir[] = {
 	0,
 };
 
-static short nfsv3err_rename[] = {
+static const short nfsv3err_rename[] = {
 	NFSERR_IO,
 	NFSERR_NOENT,
 	NFSERR_IO,
@@ -407,7 +409,7 @@ static short nfsv3err_rename[] = {
 	0,
 };
 
-static short nfsv3err_link[] = {
+static const short nfsv3err_link[] = {
 	NFSERR_IO,
 	NFSERR_IO,
 	NFSERR_ACCES,
@@ -427,7 +429,7 @@ static short nfsv3err_link[] = {
 	0,
 };
 
-static short nfsv3err_readdir[] = {
+static const short nfsv3err_readdir[] = {
 	NFSERR_IO,
 	NFSERR_IO,
 	NFSERR_ACCES,
@@ -440,7 +442,7 @@ static short nfsv3err_readdir[] = {
 	0,
 };
 
-static short nfsv3err_readdirplus[] = {
+static const short nfsv3err_readdirplus[] = {
 	NFSERR_IO,
 	NFSERR_IO,
 	NFSERR_ACCES,
@@ -454,7 +456,7 @@ static short nfsv3err_readdirplus[] = {
 	0,
 };
 
-static short nfsv3err_fsstat[] = {
+static const short nfsv3err_fsstat[] = {
 	NFSERR_IO,
 	NFSERR_IO,
 	NFSERR_STALE,
@@ -463,7 +465,7 @@ static short nfsv3err_fsstat[] = {
 	0,
 };
 
-static short nfsv3err_fsinfo[] = {
+static const short nfsv3err_fsinfo[] = {
 	NFSERR_STALE,
 	NFSERR_STALE,
 	NFSERR_BADHANDLE,
@@ -471,7 +473,7 @@ static short nfsv3err_fsinfo[] = {
 	0,
 };
 
-static short nfsv3err_pathconf[] = {
+static const short nfsv3err_pathconf[] = {
 	NFSERR_STALE,
 	NFSERR_STALE,
 	NFSERR_BADHANDLE,
@@ -479,7 +481,7 @@ static short nfsv3err_pathconf[] = {
 	0,
 };
 
-static short nfsv3err_commit[] = {
+static const short nfsv3err_commit[] = {
 	NFSERR_IO,
 	NFSERR_IO,
 	NFSERR_STALE,
@@ -488,7 +490,7 @@ static short nfsv3err_commit[] = {
 	0,
 };
 
-static short *nfsrv_v3errmap[] = {
+static const short *nfsrv_v3errmap[] = {
 	nfsv3err_null,
 	nfsv3err_getattr,
 	nfsv3err_setattr,
@@ -520,8 +522,11 @@ static int
 nfsrv_modevent(module_t mod, int type, void *data)
 {
 
+	NET_LOCK_GIANT();
+
 	switch (type) {
 	case MOD_LOAD:
+		mtx_init(&nfsd_mtx, "nfsd_mtx", NULL, MTX_DEF);
 		nfsrv_rpc_vers = txdr_unsigned(RPC_VER2);
 		nfsrv_rpc_call = txdr_unsigned(RPC_CALL);
 		nfsrv_rpc_reply = txdr_unsigned(RPC_REPLY);
@@ -538,10 +543,11 @@ nfsrv_modevent(module_t mod, int type, void *data)
 		if (nfsrv_ticks < 1)
 			nfsrv_ticks = 1;
 
-		nfsrv_init(0);		/* Init server data structures */
 		nfsrv_initcache();	/* Init the server request cache */
-
+		NFSD_LOCK();
+		nfsrv_init(0);		/* Init server data structures */
 		callout_init(&nfsrv_callout, 0);
+		NFSD_UNLOCK();
 		nfsrv_timer(0);
 
 		nfs_prev_nfssvc_sy_narg = sysent[SYS_nfssvc].sy_narg;
@@ -557,8 +563,10 @@ nfsrv_modevent(module_t mod, int type, void *data)
 		callout_stop(&nfsrv_callout);
 		sysent[SYS_nfssvc].sy_narg = nfs_prev_nfssvc_sy_narg;
 		sysent[SYS_nfssvc].sy_call = nfs_prev_nfssvc_sy_call;
+		mtx_destroy(&nfsd_mtx);
 		break;
 	}
+	NET_UNLOCK_GIANT();
 	return 0;
 }
 static moduledata_t nfsserver_mod = {
@@ -603,6 +611,10 @@ nfs_namei(struct nameidata *ndp, fhandle_t *fhp, int len,
 	struct componentname *cnp = &ndp->ni_cnd;
 	int lockleaf = (cnp->cn_flags & LOCKLEAF) != 0;
 
+	NFSD_LOCK_ASSERT();
+	NFSD_UNLOCK();
+	mtx_lock(&Giant);	/* VFS */
+
 	*retdirp = NULL;
 	cnp->cn_flags |= NOMACCHECK;
 	cnp->cn_pnbuf = uma_zalloc(namei_zone, M_WAITOK);
@@ -646,8 +658,12 @@ nfs_namei(struct nameidata *ndp, fhandle_t *fhp, int len,
 	/*
 	 * Extract and set starting directory.
 	 */
+	mtx_unlock(&Giant);	/* VFS */
+	NFSD_LOCK();
 	error = nfsrv_fhtovp(fhp, FALSE, &dp, ndp->ni_cnd.cn_cred, slp,
 	    nam, &rdonly, pubflag);
+	NFSD_UNLOCK();
+	mtx_lock(&Giant);	/* VFS */
 	if (error)
 		goto out;
 	if (dp->v_type != VDIR) {
@@ -868,6 +884,8 @@ out:
 	} else if ((ndp->ni_cnd.cn_flags & (WANTPARENT|LOCKPARENT)) == 0) {
 		ndp->ni_dvp = NULL;
 	}
+	mtx_unlock(&Giant);	/* VFS */
+	NFSD_LOCK();
 	return (error);
 }
 
@@ -882,6 +900,8 @@ nfsm_adj(struct mbuf *mp, int len, int nul)
 	int count, i;
 	char *cp;
 
+	NFSD_LOCK_DONTCARE();
+
 	/*
 	 * Trim from tail.  Scan the mbuf chain,
 	 * calculating its length and finding the last mbuf.
@@ -1043,6 +1063,8 @@ nfsrv_fhtovp(fhandle_t *fhp, int lockflag, struct vnode **vpp,
 	struct sockaddr_int *saddr;
 #endif
 
+	NFSD_LOCK_ASSERT();
+
 	*vpp = NULL;
 
 	if (nfs_ispublicfh(fhp)) {
@@ -1054,12 +1076,14 @@ nfsrv_fhtovp(fhandle_t *fhp, int lockflag, struct vnode **vpp,
 	mp = vfs_getvfs(&fhp->fh_fsid);
 	if (!mp)
 		return (ESTALE);
+	NFSD_UNLOCK();
+	mtx_lock(&Giant);	/* VFS */
 	error = VFS_CHECKEXP(mp, nam, &exflags, &credanon);
 	if (error)
-		return (error);
+		goto out;
 	error = VFS_FHTOVP(mp, &fhp->fh_fid, vpp);
 	if (error)
-		return (error);
+		goto out;
 #ifdef MNT_EXNORESPORT
 	if (!(exflags & (MNT_EXNORESPORT|MNT_EXPUBLIC))) {
 		saddr = (struct sockaddr_in *)nam;
@@ -1069,7 +1093,7 @@ nfsrv_fhtovp(fhandle_t *fhp, int lockflag, struct vnode **vpp,
 		    ntohs(saddr->sin_port) >= IPPORT_RESERVED) {
 			vput(*vpp);
 			*vpp = NULL;
-			return (NFSERR_AUTHERR | AUTH_TOOWEAK);
+			error = NFSERR_AUTHERR | AUTH_TOOWEAK;
 		}
 	}
 #endif
@@ -1091,7 +1115,10 @@ nfsrv_fhtovp(fhandle_t *fhp, int lockflag, struct vnode **vpp,
 
 	if (!lockflag)
 		VOP_UNLOCK(*vpp, 0, td);
-	return (0);
+out:
+	mtx_unlock(&Giant);	/* VFS */
+	NFSD_LOCK();
+	return (error);
 }
 
 
@@ -1106,6 +1133,8 @@ nfs_ispublicfh(fhandle_t *fhp)
 	char *cp = (char *)fhp;
 	int i;
 
+	NFSD_LOCK_DONTCARE();
+
 	for (i = 0; i < NFSX_V3FH; i++)
 		if (*cp++ != 0)
 			return (FALSE);
@@ -1124,6 +1153,8 @@ netaddr_match(int family, union nethostaddr *haddr, struct sockaddr *nam)
 {
 	struct sockaddr_in *inetaddr;
 
+	NFSD_LOCK_DONTCARE();
+
 	switch (family) {
 	case AF_INET:
 		inetaddr = (struct sockaddr_in *)nam;
@@ -1159,9 +1190,11 @@ netaddr_match(int family, union nethostaddr *haddr, struct sockaddr *nam)
 int
 nfsrv_errmap(struct nfsrv_descript *nd, int err)
 {
-	short *defaulterrp, *errp;
+	const short *defaulterrp, *errp;
 	int e;
 
+	NFSD_LOCK_DONTCARE();
+
 	if (nd->nd_flag & ND_NFSV3) {
 	    if (nd->nd_procnum <= NFSPROC_COMMIT) {
 		errp = defaulterrp = nfsrv_v3errmap[nd->nd_procnum];
@@ -1187,6 +1220,9 @@ int
 nfsrv_object_create(struct vnode *vp)
 {
 
+	GIANT_REQUIRED;
+	NFSD_UNLOCK_ASSERT();
+
 	if (vp == NULL || vp->v_type != VREG)
 		return (1);
 	return (vfs_object_create(vp, curthread, curthread->td_ucred));
@@ -1203,6 +1239,8 @@ nfsrvw_sort(gid_t *list, int num)
 	int i, j;
 	gid_t v;
 
+	NFSD_LOCK_DONTCARE();
+
 	/* Insertion sort. */
 	for (i = 1; i < num; i++) {
 		v = list[i];
@@ -1221,6 +1259,8 @@ nfsrv_setcred(struct ucred *incred, struct ucred *outcred)
 {
 	int i;
 
+	NFSD_LOCK_DONTCARE();
+
 	bzero((caddr_t)outcred, sizeof (struct ucred));
 	outcred->cr_ref = 1;
 	outcred->cr_uid = incred->cr_uid;
@@ -1239,6 +1279,8 @@ nfsm_srvfhtom_xx(fhandle_t *f, int v3, struct mbuf **mb, caddr_t *bpos)
 {
 	u_int32_t *tl;
 
+	NFSD_LOCK_DONTCARE();
+
 	if (v3) {
 		tl = nfsm_build_xx(NFSX_UNSIGNED + NFSX_V3FH, mb, bpos);
 		*tl++ = txdr_unsigned(NFSX_V3FH);
@@ -1265,6 +1307,8 @@ nfsm_srvstrsiz_xx(int *s, int m, struct mbuf **md, caddr_t *dpos)
 {
 	u_int32_t *tl;
 
+	NFSD_LOCK_DONTCARE();
+
 	tl = nfsm_dissect_xx(NFSX_UNSIGNED, md, dpos);
 	if (tl == NULL)
 		return EBADRPC;
@@ -1279,6 +1323,8 @@ nfsm_srvnamesiz_xx(int *s, int m, struct mbuf **md, caddr_t *dpos)
 {
 	u_int32_t *tl;
 
+	NFSD_LOCK_DONTCARE();
+
 	tl = nfsm_dissect_xx(NFSX_UNSIGNED, md, dpos);
 	if (tl == NULL)
 		return EBADRPC;
@@ -1292,15 +1338,26 @@ nfsm_srvnamesiz_xx(int *s, int m, struct mbuf **md, caddr_t *dpos)
 
 void
 nfsm_clget_xx(u_int32_t **tl, struct mbuf *mb, struct mbuf **mp,
-    char **bp, char **be, caddr_t bpos)
+    char **bp, char **be, caddr_t bpos, int droplock)
 {
 	struct mbuf *nmp;
 
+	NFSD_LOCK_DONTCARE();
+
+	if (droplock)
+		NFSD_LOCK_ASSERT();
+	else
+		NFSD_UNLOCK_ASSERT();
+
 	if (*bp >= *be) {
 		if (*mp == mb)
 			(*mp)->m_len += *bp - bpos;
+		if (droplock)
+			NFSD_UNLOCK();
 		MGET(nmp, M_TRYWAIT, MT_DATA);
 		MCLGET(nmp, M_TRYWAIT);
+		if (droplock)
+			NFSD_LOCK();
 		nmp->m_len = NFSMSIZ(nmp);
 		(*mp)->m_next = nmp;
 		*mp = nmp;
@@ -1317,6 +1374,8 @@ nfsm_srvmtofh_xx(fhandle_t *f, struct nfsrv_descript *nfsd, struct mbuf **md,
 	u_int32_t *tl;
 	int fhlen;
 
+	NFSD_LOCK_DONTCARE();
+
 	if (nfsd->nd_flag & ND_NFSV3) {
 		tl = nfsm_dissect_xx(NFSX_UNSIGNED, md, dpos);
 		if (tl == NULL)
@@ -1343,6 +1402,8 @@ nfsm_srvsattr_xx(struct vattr *a, struct mbuf **md, caddr_t *dpos)
 {
 	u_int32_t *tl;
 
+	NFSD_LOCK_DONTCARE();
+
 	tl = nfsm_dissect_xx(NFSX_UNSIGNED, md, dpos);
 	if (tl == NULL)
 		return EBADRPC;