diff options
Diffstat (limited to 'sys/miscfs')
35 files changed, 2274 insertions, 1768 deletions
diff --git a/sys/miscfs/deadfs/dead_vnops.c b/sys/miscfs/deadfs/dead_vnops.c index f22425c..6d83c61 100644 --- a/sys/miscfs/deadfs/dead_vnops.c +++ b/sys/miscfs/deadfs/dead_vnops.c @@ -77,11 +77,11 @@ static int dead_select __P((struct vop_select_args *)); #define dead_inactive ((int (*) __P((struct vop_inactive_args *)))nullop) #define dead_reclaim ((int (*) __P((struct vop_reclaim_args *)))nullop) static int dead_lock __P((struct vop_lock_args *)); -#define dead_unlock ((int (*) __P((struct vop_unlock_args *)))nullop) +#define dead_unlock ((int (*) __P((struct vop_unlock_args *)))vop_nounlock) static int dead_bmap __P((struct vop_bmap_args *)); static int dead_strategy __P((struct vop_strategy_args *)); static int dead_print __P((struct vop_print_args *)); -#define dead_islocked ((int (*) __P((struct vop_islocked_args *)))nullop) +#define dead_islocked ((int(*) __P((struct vop_islocked_args *)))vop_noislocked) #define dead_pathconf ((int (*) __P((struct vop_pathconf_args *)))dead_ebadf) #define dead_advlock ((int (*) __P((struct vop_advlock_args *)))dead_ebadf) #define dead_blkatoff ((int (*) __P((struct vop_blkatoff_args *)))dead_badop) @@ -191,11 +191,20 @@ dead_read(ap) if (chkvnlock(ap->a_vp)) panic("dead_read: lock"); +#if 0 + /* Lite2 behaviour */ + /* + * Return EOF for tty devices, EIO for others + */ + if ((ap->a_vp->v_flag & VISTTY) == 0) + return (EIO); +#else /* * Return EOF for character devices, EIO for others */ if (ap->a_vp->v_type != VCHR) return (EIO); +#endif return (0); } @@ -282,12 +291,23 @@ static int dead_lock(ap) struct vop_lock_args /* { struct vnode *a_vp; + int a_flags; + struct proc *a_p; } */ *ap; { + struct vnode *vp = ap->a_vp; - if (!chkvnlock(ap->a_vp)) + /* + * Since we are not using the lock manager, we must clear + * the interlock here. + */ + if (ap->a_flags & LK_INTERLOCK) { + simple_unlock(&vp->v_interlock); + ap->a_flags &= ~LK_INTERLOCK; + } + if (!chkvnlock(vp)) return (0); - return (VCALL(ap->a_vp, VOFFSET(vop_lock), ap)); + return (VCALL(vp, VOFFSET(vop_lock), ap)); } /* diff --git a/sys/miscfs/fdesc/fdesc.h b/sys/miscfs/fdesc/fdesc.h index d754655..b4eff1f 100644 --- a/sys/miscfs/fdesc/fdesc.h +++ b/sys/miscfs/fdesc/fdesc.h @@ -61,8 +61,7 @@ typedef enum { } fdntype; struct fdescnode { - struct fdescnode *fd_forw; /* Hash chain */ - struct fdescnode *fd_back; + LIST_ENTRY(fdescnode) fd_hash; /* Hash list */ struct vnode *fd_vnode; /* Back ptr to vnode */ fdntype fd_type; /* Type of this node */ unsigned fd_fd; /* Fd to be dup'ed */ @@ -74,7 +73,7 @@ struct fdescnode { #define VTOFDESC(vp) ((struct fdescnode *)(vp)->v_data) extern dev_t devctty; -extern int fdesc_init __P((void)); +extern int fdesc_init __P((struct vfsconf *)); extern int fdesc_root __P((struct mount *, struct vnode **)); extern int fdesc_allocvp __P((fdntype, int, struct mount *, struct vnode **)); extern vop_t **fdesc_vnodeop_p; diff --git a/sys/miscfs/fdesc/fdesc_vfsops.c b/sys/miscfs/fdesc/fdesc_vfsops.c index 5f6703b..135052e 100644 --- a/sys/miscfs/fdesc/fdesc_vfsops.c +++ b/sys/miscfs/fdesc/fdesc_vfsops.c @@ -1,5 +1,5 @@ /* - * Copyright (c) 1992, 1993 + * Copyright (c) 1992, 1993, 1995 * The Regents of the University of California. All rights reserved. * * This code is derived from software donated to Berkeley by @@ -108,7 +108,7 @@ fdesc_mount(mp, path, data, ndp, p) /* XXX -- don't mark as local to work around fts() problems */ /*mp->mnt_flag |= MNT_LOCAL;*/ mp->mnt_data = (qaddr_t) fmp; - getnewfsid(mp, MOUNT_FDESC); + vfs_getnewfsid(mp); (void) copyinstr(path, mp->mnt_stat.f_mntonname, MNAMELEN - 1, &size); bzero(mp->mnt_stat.f_mntonname + size, MNAMELEN - size); @@ -137,12 +137,8 @@ fdesc_unmount(mp, mntflags, p) int flags = 0; struct vnode *rootvp = VFSTOFDESC(mp)->f_root; - if (mntflags & MNT_FORCE) { - /* fdesc can never be rootfs so don't check for it */ - if (!doforce) - return (EINVAL); + if (mntflags & MNT_FORCE) flags |= FORCECLOSE; - } /* * Clear out buffer cache. I don't think we @@ -176,6 +172,7 @@ fdesc_root(mp, vpp) struct mount *mp; struct vnode **vpp; { + struct proc *p = curproc; /* XXX */ struct vnode *vp; /* @@ -183,24 +180,12 @@ fdesc_root(mp, vpp) */ vp = VFSTOFDESC(mp)->f_root; VREF(vp); - VOP_LOCK(vp); + vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, p); *vpp = vp; return (0); } static int -fdesc_quotactl(mp, cmd, uid, arg, p) - struct mount *mp; - int cmd; - uid_t uid; - caddr_t arg; - struct proc *p; -{ - - return (EOPNOTSUPP); -} - -static int fdesc_statfs(mp, sbp, p) struct mount *mp; struct statfs *sbp; @@ -233,7 +218,6 @@ fdesc_statfs(mp, sbp, p) if (fdp->fd_nfiles < lim) freefd += (lim - fdp->fd_nfiles); - sbp->f_type = MOUNT_FDESC; sbp->f_flags = 0; sbp->f_bsize = DEV_BSIZE; sbp->f_iosize = DEV_BSIZE; @@ -243,6 +227,7 @@ fdesc_statfs(mp, sbp, p) sbp->f_files = lim + 1; /* Allow for "." */ sbp->f_ffree = freefd; /* See comments above */ if (sbp != &mp->mnt_stat) { + sbp->f_type = mp->mnt_vfc->vfc_typenum; bcopy(&mp->mnt_stat.f_fsid, &sbp->f_fsid, sizeof(sbp->f_fsid)); bcopy(mp->mnt_stat.f_mntonname, sbp->f_mntonname, MNAMELEN); bcopy(mp->mnt_stat.f_mntfromname, sbp->f_mntfromname, MNAMELEN); @@ -261,41 +246,15 @@ fdesc_sync(mp, waitfor, cred, p) return (0); } -/* - * Fdesc flat namespace lookup. - * Currently unsupported. - */ -static int -fdesc_vget(mp, ino, vpp) - struct mount *mp; - ino_t ino; - struct vnode **vpp; -{ - - return (EOPNOTSUPP); -} - -static int -fdesc_fhtovp(mp, fhp, nam, vpp, exflagsp, credanonp) - struct mount *mp; - struct fid *fhp; - struct mbuf *nam; - struct vnode **vpp; - int *exflagsp; - struct ucred **credanonp; -{ - - return (EOPNOTSUPP); -} - -static int -fdesc_vptofh(vp, fhp) - struct vnode *vp; - struct fid *fhp; -{ - - return (EOPNOTSUPP); -} +#define fdesc_fhtovp ((int (*) __P((struct mount *, struct fid *, \ + struct mbuf *, struct vnode **, int *, struct ucred **)))eopnotsupp) +#define fdesc_quotactl ((int (*) __P((struct mount *, int, uid_t, caddr_t, \ + struct proc *)))eopnotsupp) +#define fdesc_sysctl ((int (*) __P((int *, u_int, void *, size_t *, void *, \ + size_t, struct proc *)))eopnotsupp) +#define fdesc_vget ((int (*) __P((struct mount *, ino_t, struct vnode **))) \ + eopnotsupp) +#define fdesc_vptofh ((int (*) __P((struct vnode *, struct fid *)))eopnotsupp) static struct vfsops fdesc_vfsops = { fdesc_mount, diff --git a/sys/miscfs/fdesc/fdesc_vnops.c b/sys/miscfs/fdesc/fdesc_vnops.c index 1fe511b..795dfd8 100644 --- a/sys/miscfs/fdesc/fdesc_vnops.c +++ b/sys/miscfs/fdesc/fdesc_vnops.c @@ -77,22 +77,14 @@ FD_STDIN, FD_STDOUT, FD_STDERR must be a sequence n, n+1, n+2 #endif #define NFDCACHE 4 -#define FD_NHASH(ix) ((ix) & NFDCACHE-1) - -/* - * Cache head - */ -struct fdcache { - struct fdescnode *fc_forw; - struct fdescnode *fc_back; -}; - -static struct fdcache fdcache[NFDCACHE]; +#define FD_NHASH(ix) \ + (&fdhashtbl[(ix) & fdhash]) +LIST_HEAD(fdhashhead, fdescnode) *fdhashtbl; +u_long fdhash; static int fdesc_attr __P((int fd, struct vattr *vap, struct ucred *cred, struct proc *p)); static int fdesc_badop __P((void)); -static int fdesc_enotsupp __P((void)); static int fdesc_getattr __P((struct vop_getattr_args *ap)); static struct fdcache * fdesc_hash __P((int ix)); @@ -115,28 +107,15 @@ static int fdesc_write __P((struct vop_write_args *ap)); * Initialise cache headers */ int -fdesc_init() +fdesc_init(vfsp) + struct vfsconf *vfsp; { - struct fdcache *fc; devctty = makedev(nchrdev, 0); - - for (fc = fdcache; fc < fdcache + NFDCACHE; fc++) - fc->fc_forw = fc->fc_back = (struct fdescnode *) fc; + fdhashtbl = hashinit(NFDCACHE, M_CACHE, &fdhash); return (0); } -/* - * Compute hash list for given target vnode - */ -static struct fdcache * -fdesc_hash(ix) - int ix; -{ - - return (&fdcache[FD_NHASH(ix)]); -} - int fdesc_allocvp(ftype, ix, mp, vpp) fdntype ftype; @@ -144,15 +123,16 @@ fdesc_allocvp(ftype, ix, mp, vpp) struct mount *mp; struct vnode **vpp; { - struct fdcache *fc; + struct proc *p = curproc; /* XXX */ + struct fdhashhead *fc; struct fdescnode *fd; int error = 0; + fc = FD_NHASH(ix); loop: - fc = fdesc_hash(ix); - for (fd = fc->fc_forw; fd != (struct fdescnode *) fc; fd = fd->fd_forw) { + for (fd = fc->lh_first; fd != 0; fd = fd->fd_hash.le_next) { if (fd->fd_ix == ix && fd->fd_vnode->v_mount == mp) { - if (vget(fd->fd_vnode, 0)) + if (vget(fd->fd_vnode, 0, p)) goto loop; *vpp = fd->fd_vnode; return (error); @@ -188,8 +168,7 @@ loop: fd->fd_fd = -1; fd->fd_link = 0; fd->fd_ix = ix; - fc = fdesc_hash(ix); - insque(fd, fc); + LIST_INSERT_HEAD(fc, fd, fd_hash); out:; fdcache_lock &= ~FDL_LOCKED; @@ -216,31 +195,28 @@ fdesc_lookup(ap) { struct vnode **vpp = ap->a_vpp; struct vnode *dvp = ap->a_dvp; - char *pname; - struct proc *p; - int nfiles; - unsigned fd = 0; + struct componentname *cnp = ap->a_cnp; + char *pname = cnp->cn_nameptr; + struct proc *p = cnp->cn_proc; + int nfiles = p->p_fd->fd_nfiles; + unsigned fd; int error; struct vnode *fvp; char *ln; - if (ap->a_cnp->cn_nameiop == DELETE || - ap->a_cnp->cn_nameiop == RENAME) { + if (cnp->cn_nameiop == DELETE || cnp->cn_nameiop == RENAME) { error = EROFS; goto bad; } - pname = ap->a_cnp->cn_nameptr; - if (ap->a_cnp->cn_namelen == 1 && *pname == '.') { + VOP_UNLOCK(dvp, 0, p); + if (cnp->cn_namelen == 1 && *pname == '.') { *vpp = dvp; - VREF(dvp); - VOP_LOCK(dvp); + VREF(dvp); + vn_lock(dvp, LK_SHARED | LK_RETRY, p); return (0); } - p = ap->a_cnp->cn_proc; - nfiles = p->p_fd->fd_nfiles; - switch (VTOFDESC(dvp)->fd_type) { default: case Flink: @@ -250,17 +226,17 @@ fdesc_lookup(ap) goto bad; case Froot: - if (ap->a_cnp->cn_namelen == 2 && bcmp(pname, "fd", 2) == 0) { + if (cnp->cn_namelen == 2 && bcmp(pname, "fd", 2) == 0) { error = fdesc_allocvp(Fdevfd, FD_DEVFD, dvp->v_mount, &fvp); if (error) goto bad; *vpp = fvp; fvp->v_type = VDIR; - VOP_LOCK(fvp); + vn_lock(fvp, LK_SHARED | LK_RETRY, p); return (0); } - if (ap->a_cnp->cn_namelen == 3 && bcmp(pname, "tty", 3) == 0) { + if (cnp->cn_namelen == 3 && bcmp(pname, "tty", 3) == 0) { struct vnode *ttyvp = cttyvp(p); if (ttyvp == NULL) { error = ENXIO; @@ -271,12 +247,12 @@ fdesc_lookup(ap) goto bad; *vpp = fvp; fvp->v_type = VFIFO; - VOP_LOCK(fvp); + vn_lock(fvp, LK_SHARED | LK_RETRY, p); return (0); } ln = 0; - switch (ap->a_cnp->cn_namelen) { + switch (cnp->cn_namelen) { case 5: if (bcmp(pname, "stdin", 5) == 0) { ln = "fd/0"; @@ -302,7 +278,7 @@ fdesc_lookup(ap) VTOFDESC(fvp)->fd_link = ln; *vpp = fvp; fvp->v_type = VLNK; - VOP_LOCK(fvp); + vn_lock(fvp, LK_SHARED | LK_RETRY, p); return (0); } else { error = ENOENT; @@ -312,9 +288,10 @@ fdesc_lookup(ap) /* FALL THROUGH */ case Fdevfd: - if (ap->a_cnp->cn_namelen == 2 && bcmp(pname, "..", 2) == 0) { - error = fdesc_root(dvp->v_mount, vpp); - return (error); + if (cnp->cn_namelen == 2 && bcmp(pname, "..", 2) == 0) { + if (error = fdesc_root(dvp->v_mount, vpp)) + goto bad; + return (0); } fd = 0; @@ -338,11 +315,13 @@ fdesc_lookup(ap) if (error) goto bad; VTOFDESC(fvp)->fd_fd = fd; + vn_lock(fvp, LK_SHARED | LK_RETRY, p); *vpp = fvp; return (0); } bad:; + vn_lock(dvp, LK_SHARED | LK_RETRY, p); *vpp = NULL; return (error); } @@ -402,10 +381,10 @@ fdesc_attr(fd, vap, cred, p) error = VOP_GETATTR((struct vnode *) fp->f_data, vap, cred, p); if (error == 0 && vap->va_type == VDIR) { /* - * don't allow directories to show up because - * that causes loops in the namespace. + * directories can cause loops in the namespace, + * so turn off the 'x' bits to avoid trouble. */ - vap->va_type = VFIFO; + vap->va_mode &= ~((VEXEC)|(VEXEC>>3)|(VEXEC>>6)); } break; @@ -591,6 +570,9 @@ fdesc_readdir(ap) struct vnode *a_vp; struct uio *a_uio; struct ucred *a_cred; + int *a_eofflag; + u_long *a_cookies; + int a_ncookies; } */ *ap; { struct uio *uio = ap->a_uio; @@ -598,6 +580,13 @@ fdesc_readdir(ap) int i; int error; + /* + * We don't allow exporting fdesc mounts, and currently local + * requests do not need cookies. + */ + if (ap->a_ncookies) + panic("fdesc_readdir: not hungry"); + switch (VTOFDESC(ap->a_vp)->fd_type) { case Fctty: return (0); @@ -814,6 +803,7 @@ static int fdesc_inactive(ap) struct vop_inactive_args /* { struct vnode *a_vp; + struct proc *a_p; } */ *ap; { struct vnode *vp = ap->a_vp; @@ -822,6 +812,7 @@ fdesc_inactive(ap) * Clear out the v_type field to avoid * nasty things happening in vgone(). */ + VOP_UNLOCK(vp, 0, ap->a_p); vp->v_type = VNON; return (0); } @@ -833,8 +824,9 @@ fdesc_reclaim(ap) } */ *ap; { struct vnode *vp = ap->a_vp; + struct fdescnode *fd = VTOFDESC(vp); - remque(VTOFDESC(vp)); + LIST_REMOVE(fd, fd_hash); FREE(vp->v_data, M_TEMP); vp->v_data = 0; @@ -907,16 +899,6 @@ fdesc_vfree(ap) } /* - * /dev/fd vnode unsupported operation - */ -static int -fdesc_enotsupp() -{ - - return (EOPNOTSUPP); -} - -/* * /dev/fd "should never get here" operation */ static int @@ -927,38 +909,39 @@ fdesc_badop() /* NOTREACHED */ } -#define fdesc_create ((int (*) __P((struct vop_create_args *)))fdesc_enotsupp) -#define fdesc_mknod ((int (*) __P((struct vop_mknod_args *)))fdesc_enotsupp) +#define fdesc_create ((int (*) __P((struct vop_create_args *)))eopnotsupp) +#define fdesc_mknod ((int (*) __P((struct vop_mknod_args *)))eopnotsupp) #define fdesc_close ((int (*) __P((struct vop_close_args *)))nullop) #define fdesc_access ((int (*) __P((struct vop_access_args *)))nullop) -#define fdesc_mmap ((int (*) __P((struct vop_mmap_args *)))fdesc_enotsupp) +#define fdesc_mmap ((int (*) __P((struct vop_mmap_args *)))eopnotsupp) +#define fdesc_revoke vop_revoke #define fdesc_fsync ((int (*) __P((struct vop_fsync_args *)))nullop) #define fdesc_seek ((int (*) __P((struct vop_seek_args *)))nullop) -#define fdesc_remove ((int (*) __P((struct vop_remove_args *)))fdesc_enotsupp) -#define fdesc_link ((int (*) __P((struct vop_link_args *)))fdesc_enotsupp) -#define fdesc_rename ((int (*) __P((struct vop_rename_args *)))fdesc_enotsupp) -#define fdesc_mkdir ((int (*) __P((struct vop_mkdir_args *)))fdesc_enotsupp) -#define fdesc_rmdir ((int (*) __P((struct vop_rmdir_args *)))fdesc_enotsupp) -#define fdesc_symlink ((int (*) __P((struct vop_symlink_args *)))fdesc_enotsupp) +#define fdesc_remove ((int (*) __P((struct vop_remove_args *)))eopnotsupp) +#define fdesc_link ((int (*) __P((struct vop_link_args *)))eopnotsupp) +#define fdesc_rename ((int (*) __P((struct vop_rename_args *)))eopnotsupp) +#define fdesc_mkdir ((int (*) __P((struct vop_mkdir_args *)))eopnotsupp) +#define fdesc_rmdir ((int (*) __P((struct vop_rmdir_args *)))eopnotsupp) +#define fdesc_symlink ((int (*) __P((struct vop_symlink_args *)))eopnotsupp) #define fdesc_abortop ((int (*) __P((struct vop_abortop_args *)))nullop) -#define fdesc_lock ((int (*) __P((struct vop_lock_args *)))nullop) -#define fdesc_unlock ((int (*) __P((struct vop_unlock_args *)))nullop) +#define fdesc_lock ((int (*) __P((struct vop_lock_args *)))vop_nolock) +#define fdesc_unlock ((int (*) __P((struct vop_unlock_args *)))vop_nounlock) #define fdesc_bmap ((int (*) __P((struct vop_bmap_args *)))fdesc_badop) #define fdesc_strategy ((int (*) __P((struct vop_strategy_args *)))fdesc_badop) -#define fdesc_islocked ((int (*) __P((struct vop_islocked_args *)))nullop) -#define fdesc_advlock ((int (*) __P((struct vop_advlock_args *)))fdesc_enotsupp) +#define fdesc_islocked \ + ((int (*) __P((struct vop_islocked_args *)))vop_noislocked) +#define fdesc_advlock ((int (*) __P((struct vop_advlock_args *)))eopnotsupp) #define fdesc_blkatoff \ - ((int (*) __P((struct vop_blkatoff_args *)))fdesc_enotsupp) -#define fdesc_vget ((int (*) __P((struct vop_vget_args *)))fdesc_enotsupp) + ((int (*) __P((struct vop_blkatoff_args *)))eopnotsupp) #define fdesc_valloc ((int(*) __P(( \ struct vnode *pvp, \ int mode, \ struct ucred *cred, \ - struct vnode **vpp))) fdesc_enotsupp) + struct vnode **vpp))) eopnotsupp) #define fdesc_truncate \ - ((int (*) __P((struct vop_truncate_args *)))fdesc_enotsupp) -#define fdesc_update ((int (*) __P((struct vop_update_args *)))fdesc_enotsupp) -#define fdesc_bwrite ((int (*) __P((struct vop_bwrite_args *)))fdesc_enotsupp) + ((int (*) __P((struct vop_truncate_args *)))eopnotsupp) +#define fdesc_update ((int (*) __P((struct vop_update_args *)))eopnotsupp) +#define fdesc_bwrite ((int (*) __P((struct vop_bwrite_args *)))eopnotsupp) static vop_t **fdesc_vnodeop_p; static struct vnodeopv_entry_desc fdesc_vnodeop_entries[] = { @@ -975,6 +958,7 @@ static struct vnodeopv_entry_desc fdesc_vnodeop_entries[] = { { &vop_write_desc, (vop_t *)fdesc_write }, /* write */ { &vop_ioctl_desc, (vop_t *)fdesc_ioctl }, /* ioctl */ { &vop_select_desc, (vop_t *)fdesc_select }, /* select */ + { &vop_revoke_desc, (vop_t *)fdesc_revoke }, /* revoke */ { &vop_mmap_desc, (vop_t *)fdesc_mmap }, /* mmap */ { &vop_fsync_desc, (vop_t *)fdesc_fsync }, /* fsync */ { &vop_seek_desc, (vop_t *)fdesc_seek }, /* seek */ diff --git a/sys/miscfs/fifofs/fifo.h b/sys/miscfs/fifofs/fifo.h index efeaf85..6562159 100644 --- a/sys/miscfs/fifofs/fifo.h +++ b/sys/miscfs/fifofs/fifo.h @@ -30,7 +30,7 @@ * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * - * @(#)fifo.h 8.2 (Berkeley) 2/2/94 + * @(#)fifo.h 8.6 (Berkeley) 5/21/95 * $FreeBSD$ */ @@ -51,8 +51,10 @@ int fifo_close __P((struct vop_close_args *)); #define fifo_setattr ((int (*) __P((struct vop_setattr_args *)))fifo_ebadf) int fifo_read __P((struct vop_read_args *)); int fifo_write __P((struct vop_write_args *)); +#define fifo_lease_check ((int (*) __P((struct vop_lease_args *)))nullop) int fifo_ioctl __P((struct vop_ioctl_args *)); int fifo_select __P((struct vop_select_args *)); +#define fifo_revoke vop_revoke #define fifo_mmap ((int (*) __P((struct vop_mmap_args *)))fifo_badop) #define fifo_fsync ((int (*) __P((struct vop_fsync_args *)))nullop) #define fifo_seek ((int (*) __P((struct vop_seek_args *)))fifo_badop) @@ -65,11 +67,13 @@ int fifo_select __P((struct vop_select_args *)); #define fifo_readdir ((int (*) __P((struct vop_readdir_args *)))fifo_badop) #define fifo_readlink ((int (*) __P((struct vop_readlink_args *)))fifo_badop) #define fifo_abortop ((int (*) __P((struct vop_abortop_args *)))fifo_badop) -#define fifo_inactive ((int (*) __P((struct vop_inactive_args *)))nullop) +int fifo_inactive __P((struct vop_inactive_args *)); #define fifo_reclaim ((int (*) __P((struct vop_reclaim_args *)))nullop) +#define fifo_lock ((int (*) __P((struct vop_lock_args *)))vop_nolock) +#define fifo_unlock ((int (*) __P((struct vop_unlock_args *)))vop_nounlock) int fifo_bmap __P((struct vop_bmap_args *)); #define fifo_strategy ((int (*) __P((struct vop_strategy_args *)))fifo_badop) -#define fifo_islocked ((int (*) __P((struct vop_islocked_args *)))nullop) +#define fifo_islocked ((int(*) __P((struct vop_islocked_args *)))vop_noislocked) int fifo_pathconf __P((struct vop_pathconf_args *)); int fifo_advlock __P((struct vop_advlock_args *)); #define fifo_blkatoff ((int (*) __P((struct vop_blkatoff_args *)))fifo_badop) diff --git a/sys/miscfs/fifofs/fifo_vnops.c b/sys/miscfs/fifofs/fifo_vnops.c index 240354f..8f915ec 100644 --- a/sys/miscfs/fifofs/fifo_vnops.c +++ b/sys/miscfs/fifofs/fifo_vnops.c @@ -1,5 +1,5 @@ /* - * Copyright (c) 1990, 1993 + * Copyright (c) 1990, 1993, 1995 * The Regents of the University of California. All rights reserved. * * Redistribution and use in source and binary forms, with or without @@ -30,12 +30,12 @@ * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * - * @(#)fifo_vnops.c 8.2 (Berkeley) 1/4/94 + * @(#)fifo_vnops.c 8.10 (Berkeley) 5/27/95 * $FreeBSD$ */ #include <sys/param.h> -#include <sys/kernel.h> +#include <sys/systm.h> #include <sys/proc.h> #include <sys/time.h> #include <sys/namei.h> @@ -44,7 +44,6 @@ #include <sys/socket.h> #include <sys/socketvar.h> #include <sys/stat.h> -#include <sys/systm.h> #include <sys/ioctl.h> #include <sys/file.h> #include <sys/errno.h> @@ -64,8 +63,6 @@ struct fifoinfo { }; static int fifo_ebadf __P((void)); -static int fifo_unlock __P((struct vop_unlock_args *)); -static int fifo_lock __P((struct vop_lock_args *)); static int fifo_print __P((struct vop_print_args *)); vop_t **fifo_vnodeop_p; @@ -81,8 +78,10 @@ static struct vnodeopv_entry_desc fifo_vnodeop_entries[] = { { &vop_setattr_desc, (vop_t *)fifo_setattr }, /* setattr */ { &vop_read_desc, (vop_t *)fifo_read }, /* read */ { &vop_write_desc, (vop_t *)fifo_write }, /* write */ + { &vop_lease_desc, (vop_t *)fifo_lease_check }, /* lease */ { &vop_ioctl_desc, (vop_t *)fifo_ioctl }, /* ioctl */ { &vop_select_desc, (vop_t *)fifo_select }, /* select */ + { &vop_revoke_desc, (vop_t *)fifo_revoke }, /* revoke */ { &vop_mmap_desc, (vop_t *)fifo_mmap }, /* mmap */ { &vop_fsync_desc, (vop_t *)fifo_fsync }, /* fsync */ { &vop_seek_desc, (vop_t *)fifo_seek }, /* seek */ @@ -149,8 +148,9 @@ fifo_open(ap) struct proc *a_p; } */ *ap; { - register struct vnode *vp = ap->a_vp; - register struct fifoinfo *fip; + struct vnode *vp = ap->a_vp; + struct fifoinfo *fip; + struct proc *p = ap->a_p; struct socket *rso, *wso; int error; static char openstr[] = "fifo"; @@ -158,14 +158,14 @@ fifo_open(ap) if ((fip = vp->v_fifoinfo) == NULL) { MALLOC(fip, struct fifoinfo *, sizeof(*fip), M_VNODE, M_WAITOK); vp->v_fifoinfo = fip; - error = socreate(AF_UNIX, &rso, SOCK_STREAM, 0, ap->a_p); + error = socreate(AF_LOCAL, &rso, SOCK_STREAM, 0, ap->a_p); if (error) { free(fip, M_VNODE); vp->v_fifoinfo = NULL; return (error); } fip->fi_readsock = rso; - error = socreate(AF_UNIX, &wso, SOCK_STREAM, 0, ap->a_p); + error = socreate(AF_LOCAL, &wso, SOCK_STREAM, 0, ap->a_p); if (error) { (void)soclose(rso); free(fip, M_VNODE); @@ -185,60 +185,52 @@ fifo_open(ap) wso->so_state |= SS_CANTRCVMORE; rso->so_state |= SS_CANTSENDMORE; } - error = 0; - if ((ap->a_mode & (FREAD|FWRITE)) == (FREAD|FWRITE)) { - if (fip->fi_readers == 0) { + if (ap->a_mode & FREAD) { + fip->fi_readers++; + if (fip->fi_readers == 1) { fip->fi_writesock->so_state &= ~SS_CANTSENDMORE; if (fip->fi_writers > 0) wakeup((caddr_t)&fip->fi_writers); } - if (fip->fi_writers == 0) { + } + if (ap->a_mode & FWRITE) { + fip->fi_writers++; + if (fip->fi_writers == 1) { fip->fi_readsock->so_state &= ~SS_CANTRCVMORE; if (fip->fi_readers > 0) wakeup((caddr_t)&fip->fi_readers); } - fip->fi_readers++; - fip->fi_writers++; } - else if (ap->a_mode & FREAD) { - fip->fi_readers++; - if (fip->fi_readers == 1) { - fip->fi_writesock->so_state &= ~SS_CANTSENDMORE; - if (fip->fi_writers > 0) - wakeup((caddr_t)&fip->fi_writers); + if ((ap->a_mode & FREAD) && (ap->a_mode & O_NONBLOCK) == 0) { + while (fip->fi_writers == 0) { + VOP_UNLOCK(vp, 0, p); + error = tsleep((caddr_t)&fip->fi_readers, + PCATCH | PSOCK, openstr, 0); + vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, p); + if (error) + goto bad; } - if (!(ap->a_mode & O_NONBLOCK)) - while (fip->fi_writers == 0) { - VOP_UNLOCK(vp); - error = tsleep((caddr_t)&fip->fi_readers, - PCATCH | PSOCK, openstr, 0); - VOP_LOCK(vp); - if (error) - break; - } } - else { - fip->fi_writers++; - if (fip->fi_readers == 0 && (ap->a_mode & O_NONBLOCK)) { - error = ENXIO; - } else { - if (fip->fi_writers == 1) { - fip->fi_readsock->so_state &= ~SS_CANTRCVMORE; - if (fip->fi_readers > 0) - wakeup((caddr_t)&fip->fi_readers); + if (ap->a_mode & FWRITE) { + if (ap->a_mode & O_NONBLOCK) { + if (fip->fi_readers == 0) { + error = ENXIO; + goto bad; } + } else { while (fip->fi_readers == 0) { - VOP_UNLOCK(vp); + VOP_UNLOCK(vp, 0, p); error = tsleep((caddr_t)&fip->fi_writers, PCATCH | PSOCK, openstr, 0); - VOP_LOCK(vp); + vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, p); if (error) - break; + goto bad; } } } - if (error) - VOP_CLOSE(vp, ap->a_mode, ap->a_cred, ap->a_p); + return (0); +bad: + VOP_CLOSE(vp, ap->a_mode, ap->a_cred, p); return (error); } @@ -255,8 +247,9 @@ fifo_read(ap) struct ucred *a_cred; } */ *ap; { - register struct uio *uio = ap->a_uio; - register struct socket *rso = ap->a_vp->v_fifoinfo->fi_readsock; + struct uio *uio = ap->a_uio; + struct socket *rso = ap->a_vp->v_fifoinfo->fi_readsock; + struct proc *p = uio->uio_procp; int error, startresid; #ifdef DIAGNOSTIC @@ -268,10 +261,10 @@ fifo_read(ap) if (ap->a_ioflag & IO_NDELAY) rso->so_state |= SS_NBIO; startresid = uio->uio_resid; - VOP_UNLOCK(ap->a_vp); - error = soreceive(rso, (struct mbuf **)0, uio, - (struct mbuf **)0, (struct mbuf **)0, (int*)0); - VOP_LOCK(ap->a_vp); + VOP_UNLOCK(ap->a_vp, 0, p); + error = soreceive(rso, (struct mbuf **)0, uio, (struct mbuf **)0, + (struct mbuf **)0, (int *)0); + vn_lock(ap->a_vp, LK_EXCLUSIVE | LK_RETRY, p); /* * Clear EOF indication after first such return. */ @@ -296,6 +289,7 @@ fifo_write(ap) } */ *ap; { struct socket *wso = ap->a_vp->v_fifoinfo->fi_writesock; + struct proc *p = ap->a_uio->uio_procp; int error; #ifdef DIAGNOSTIC @@ -304,9 +298,9 @@ fifo_write(ap) #endif if (ap->a_ioflag & IO_NDELAY) wso->so_state |= SS_NBIO; - VOP_UNLOCK(ap->a_vp); + VOP_UNLOCK(ap->a_vp, 0, p); error = sosend(wso, (struct mbuf *)0, ap->a_uio, 0, (struct mbuf *)0, 0); - VOP_LOCK(ap->a_vp); + vn_lock(ap->a_vp, LK_EXCLUSIVE | LK_RETRY, p); if (ap->a_ioflag & IO_NDELAY) wso->so_state &= ~SS_NBIO; return (error); @@ -328,14 +322,23 @@ fifo_ioctl(ap) } */ *ap; { struct file filetmp; + int error; if (ap->a_command == FIONBIO) return (0); - if (ap->a_fflag & FREAD) + if (ap->a_fflag & FREAD) { filetmp.f_data = (caddr_t)ap->a_vp->v_fifoinfo->fi_readsock; - else + error = soo_ioctl(&filetmp, ap->a_command, ap->a_data, ap->a_p); + if (error) + return (error); + } + if (ap->a_fflag & FWRITE) { filetmp.f_data = (caddr_t)ap->a_vp->v_fifoinfo->fi_writesock; - return (soo_ioctl(&filetmp, ap->a_command, ap->a_data, ap->a_p)); + error = soo_ioctl(&filetmp, ap->a_command, ap->a_data, ap->a_p); + if (error) + return (error); + } + return (0); } /* ARGSUSED */ @@ -350,12 +353,33 @@ fifo_select(ap) } */ *ap; { struct file filetmp; + int ready; - if (ap->a_fflags & FREAD) + if (ap->a_fflags & FREAD) { filetmp.f_data = (caddr_t)ap->a_vp->v_fifoinfo->fi_readsock; - else + ready = soo_select(&filetmp, ap->a_which, ap->a_p); + if (ready) + return (ready); + } + if (ap->a_fflags & FWRITE) { filetmp.f_data = (caddr_t)ap->a_vp->v_fifoinfo->fi_writesock; - return (soo_select(&filetmp, ap->a_which, ap->a_p)); + ready = soo_select(&filetmp, ap->a_which, ap->a_p); + if (ready) + return (ready); + } + return (0); +} + +int +fifo_inactive(ap) + struct vop_inactive_args /* { + struct vnode *a_vp; + struct proc *a_p; + } */ *ap; +{ + + VOP_UNLOCK(ap->a_vp, 0, ap->a_p); + return (0); } /* @@ -385,31 +409,6 @@ fifo_bmap(ap) } /* - * At the moment we do not do any locking. - */ -/* ARGSUSED */ -static int -fifo_lock(ap) - struct vop_lock_args /* { - struct vnode *a_vp; - } */ *ap; -{ - - return (0); -} - -/* ARGSUSED */ -static int -fifo_unlock(ap) - struct vop_unlock_args /* { - struct vnode *a_vp; - } */ *ap; -{ - - return (0); -} - -/* * Device close routine */ /* ARGSUSED */ @@ -426,16 +425,16 @@ fifo_close(ap) register struct fifoinfo *fip = vp->v_fifoinfo; int error1, error2; - if (ap->a_fflag & FWRITE) { - fip->fi_writers--; - if (fip->fi_writers == 0) - socantrcvmore(fip->fi_readsock); - } if (ap->a_fflag & FREAD) { fip->fi_readers--; if (fip->fi_readers == 0) socantsendmore(fip->fi_writesock); } + if (ap->a_fflag & FWRITE) { + fip->fi_writers--; + if (fip->fi_writers == 0) + socantrcvmore(fip->fi_readsock); + } if (vp->v_usecount > 1) return (0); error1 = soclose(fip->fi_readsock); diff --git a/sys/miscfs/kernfs/kernfs.h b/sys/miscfs/kernfs/kernfs.h index 69496a4..2795bb3 100644 --- a/sys/miscfs/kernfs/kernfs.h +++ b/sys/miscfs/kernfs/kernfs.h @@ -33,7 +33,7 @@ * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * - * @(#)kernfs.h 8.4 (Berkeley) 1/21/94 + * @(#)kernfs.h 8.6 (Berkeley) 3/29/95 * $FreeBSD$ */ @@ -51,7 +51,18 @@ struct kernfs_node { #define VFSTOKERNFS(mp) ((struct kernfs_mount *)((mp)->mnt_data)) #define VTOKERN(vp) ((struct kernfs_node *)(vp)->v_data) +#define kernfs_fhtovp ((int (*) __P((struct mount *, struct fid *, \ + struct mbuf *, struct vnode **, int *, struct ucred **)))eopnotsupp) +#define kernfs_quotactl ((int (*) __P((struct mount *, int, uid_t, caddr_t, \ + struct proc *)))eopnotsupp) +#define kernfs_sync ((int (*) __P((struct mount *, int, struct ucred *, \ + struct proc *)))nullop) +#define kernfs_sysctl ((int (*) __P((int *, u_int, void *, size_t *, void *, \ + size_t, struct proc *)))eopnotsupp) +#define kernfs_vget ((int (*) __P((struct mount *, ino_t, struct vnode **))) \ + eopnotsupp) +#define kernfs_vptofh ((int (*) __P((struct vnode *, struct fid *)))eopnotsupp) extern vop_t **kernfs_vnodeop_p; extern struct vfsops kernfs_vfsops; -extern struct vnode *rrootvp; +extern dev_t rrootdev; #endif /* KERNEL */ diff --git a/sys/miscfs/kernfs/kernfs_vfsops.c b/sys/miscfs/kernfs/kernfs_vfsops.c index 4c045e7..b774487 100644 --- a/sys/miscfs/kernfs/kernfs_vfsops.c +++ b/sys/miscfs/kernfs/kernfs_vfsops.c @@ -1,5 +1,5 @@ /* - * Copyright (c) 1992, 1993 + * Copyright (c) 1992, 1993, 1995 * The Regents of the University of California. All rights reserved. * * This code is derived from software donated to Berkeley by @@ -33,7 +33,7 @@ * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * - * @(#)kernfs_vfsops.c 8.4 (Berkeley) 1/21/94 + * @(#)kernfs_vfsops.c 8.10 (Berkeley) 5/14/95 * $FreeBSD$ */ @@ -55,88 +55,52 @@ #include <miscfs/specfs/specdev.h> #include <miscfs/kernfs/kernfs.h> -struct vnode *rrootvp; +dev_t rrootdev = NODEV; static int cdevvp __P((dev_t dev, struct vnode **vpp)); -static int kernfs_init __P((void)); +static int kernfs_init __P((struct vfsconf *vfsp)); static int kernfs_mount __P((struct mount *mp, char *path, caddr_t data, struct nameidata *ndp, struct proc *p)); static int kernfs_start __P((struct mount *mp, int flags, struct proc *p)); static int kernfs_unmount __P((struct mount *mp, int mntflags, struct proc *p)); static int kernfs_root __P((struct mount *mp, struct vnode **vpp)); -static int kernfs_quotactl __P((struct mount *mp, int cmd, uid_t uid, - caddr_t arg, struct proc *p)); static int kernfs_statfs __P((struct mount *mp, struct statfs *sbp, struct proc *p)); -static int kernfs_sync __P((struct mount *mp, int waitfor, - struct ucred *cred, struct proc *p)); -static int kernfs_vget __P((struct mount *mp, ino_t ino, - struct vnode **vpp)); -static int kernfs_fhtovp __P((struct mount *mp, struct fid *fhp, - struct mbuf *nam, struct vnode **vpp, - int *exflagsp, struct ucred **credanonp)); -static int kernfs_vptofh __P((struct vnode *vp, struct fid *fhp)); -/* - * Create a vnode for a character device. - */ static int -cdevvp(dev, vpp) - dev_t dev; - struct vnode **vpp; +kernfs_init(vfsp) + struct vfsconf *vfsp; { - register struct vnode *vp; - struct vnode *nvp; - int error; - if (dev == NODEV) - return (0); - error = getnewvnode(VT_NON, (struct mount *)0, spec_vnodeop_p, &nvp); - if (error) { - *vpp = 0; - return (error); - } - vp = nvp; - vp->v_type = VCHR; - nvp = checkalias(vp, dev, (struct mount *)0); - if (nvp) { - vput(vp); - vp = nvp; - } - *vpp = vp; return (0); } -static int -kernfs_init() +void +kernfs_get_rrootdev() { - int cmaj; + static int tried = 0; int bmaj = major(rootdev); - int error = ENXIO; + int cmaj; -#ifdef KERNFS_DIAGNOSTIC - printf("kernfs_init\n"); /* printed during system boot */ -#endif + if (tried) { + /* Already did it once. */ + return; + } + tried = 1; if (!bdevsw[bmaj]) { panic("root dev has no bdevsw"); } + if (rootdev == NODEV) + return; for (cmaj = 0; cmaj < nchrdev; cmaj++) { - if (cdevsw[cmaj] - && (cdevsw[cmaj]->d_open == bdevsw[bmaj]->d_open)) { - dev_t cdev = makedev(cmaj, minor(rootdev)); - error = cdevvp(cdev, &rrootvp); - if (error == 0) - break; - } + rrootdev = makedev(cmaj, minor(rootdev)); + if (chrtoblk(rrootdev) == rootdev) + return; } - - if (error) { - printf("kernfs: no raw boot device\n"); - rrootvp = 0; - } - return (0); + rrootdev = NODEV; + printf("kernfs_get_rrootdev: no raw root device\n"); } /* @@ -182,7 +146,7 @@ kernfs_mount(mp, path, data, ndp, p) fmp->kf_root = rvp; mp->mnt_flag |= MNT_LOCAL; mp->mnt_data = (qaddr_t) fmp; - getnewfsid(mp, MOUNT_KERNFS); + vfs_getnewfsid(mp); (void) copyinstr(path, mp->mnt_stat.f_mntonname, MNAMELEN - 1, &size); bzero(mp->mnt_stat.f_mntonname + size, MNAMELEN - size); @@ -192,6 +156,8 @@ kernfs_mount(mp, path, data, ndp, p) #ifdef KERNFS_DIAGNOSTIC printf("kernfs_mount: at %s\n", mp->mnt_stat.f_mntonname); #endif + + kernfs_get_rrootdev(); return (0); } @@ -218,12 +184,8 @@ kernfs_unmount(mp, mntflags, p) printf("kernfs_unmount(mp = %x)\n", mp); #endif - if (mntflags & MNT_FORCE) { - /* kernfs can never be rootfs so don't check for it */ - if (!doforce) - return (EINVAL); + if (mntflags & MNT_FORCE) flags |= FORCECLOSE; - } /* * Clear out buffer cache. I don't think we @@ -263,6 +225,7 @@ kernfs_root(mp, vpp) struct mount *mp; struct vnode **vpp; { + struct proc *p = curproc; /* XXX */ struct vnode *vp; #ifdef KERNFS_DIAGNOSTIC @@ -274,23 +237,12 @@ kernfs_root(mp, vpp) */ vp = VFSTOKERNFS(mp)->kf_root; VREF(vp); - VOP_LOCK(vp); + vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, p); *vpp = vp; return (0); } static int -kernfs_quotactl(mp, cmd, uid, arg, p) - struct mount *mp; - int cmd; - uid_t uid; - caddr_t arg; - struct proc *p; -{ - return (EOPNOTSUPP); -} - -static int kernfs_statfs(mp, sbp, p) struct mount *mp; struct statfs *sbp; @@ -300,7 +252,6 @@ kernfs_statfs(mp, sbp, p) printf("kernfs_statfs(mp = %x)\n", mp); #endif - sbp->f_type = MOUNT_KERNFS; sbp->f_flags = 0; sbp->f_bsize = DEV_BSIZE; sbp->f_iosize = DEV_BSIZE; @@ -310,6 +261,7 @@ kernfs_statfs(mp, sbp, p) sbp->f_files = 0; sbp->f_ffree = 0; if (sbp != &mp->mnt_stat) { + sbp->f_type = mp->mnt_vfc->vfc_typenum; bcopy(&mp->mnt_stat.f_fsid, &sbp->f_fsid, sizeof(sbp->f_fsid)); bcopy(mp->mnt_stat.f_mntonname, sbp->f_mntonname, MNAMELEN); bcopy(mp->mnt_stat.f_mntfromname, sbp->f_mntfromname, MNAMELEN); @@ -317,53 +269,6 @@ kernfs_statfs(mp, sbp, p) return (0); } -static int -kernfs_sync(mp, waitfor, cred, p) - struct mount *mp; - int waitfor; - struct ucred *cred; - struct proc *p; -{ - - return (0); -} - -/* - * Kernfs flat namespace lookup. - * Currently unsupported. - */ -static int -kernfs_vget(mp, ino, vpp) - struct mount *mp; - ino_t ino; - struct vnode **vpp; -{ - - return (EOPNOTSUPP); -} - - -static int -kernfs_fhtovp(mp, fhp, nam, vpp, exflagsp, credanonp) - struct mount *mp; - struct fid *fhp; - struct mbuf *nam; - struct vnode **vpp; - int *exflagsp; - struct ucred **credanonp; -{ - - return (EOPNOTSUPP); -} - -static int -kernfs_vptofh(vp, fhp) - struct vnode *vp; - struct fid *fhp; -{ - return (EOPNOTSUPP); -} - static struct vfsops kernfs_vfsops = { kernfs_mount, kernfs_start, diff --git a/sys/miscfs/kernfs/kernfs_vnops.c b/sys/miscfs/kernfs/kernfs_vnops.c index afd30e7..108c443 100644 --- a/sys/miscfs/kernfs/kernfs_vnops.c +++ b/sys/miscfs/kernfs/kernfs_vnops.c @@ -33,7 +33,7 @@ * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * - * @(#)kernfs_vnops.c 8.6 (Berkeley) 2/10/94 + * @(#)kernfs_vnops.c 8.15 (Berkeley) 5/21/95 * $FreeBSD$ */ @@ -67,41 +67,44 @@ #define DIR_MODE (S_IRUSR|S_IXUSR|S_IRGRP|S_IXGRP|S_IROTH|S_IXOTH) static struct kern_target { + u_char kt_type; + u_char kt_namlen; char *kt_name; void *kt_data; -#define KTT_NULL 1 -#define KTT_TIME 5 -#define KTT_INT 17 -#define KTT_STRING 31 -#define KTT_HOSTNAME 47 -#define KTT_BOOTFILE 49 -#define KTT_AVENRUN 53 - int kt_tag; - int kt_rw; - int kt_vtype; - struct vnode **kt_vp; +#define KTT_NULL 1 +#define KTT_TIME 5 +#define KTT_INT 17 +#define KTT_STRING 31 +#define KTT_HOSTNAME 47 +#define KTT_BOOTFILE 49 +#define KTT_AVENRUN 53 +#define KTT_DEVICE 71 + u_char kt_tag; + u_char kt_vtype; + mode_t kt_mode; } kern_targets[] = { /* NOTE: The name must be less than UIO_MX-16 chars in length */ - /* name data tag ro/rw type vnodep*/ - { ".", 0, KTT_NULL, VREAD, VDIR, NULL }, - { "..", 0, KTT_NULL, VREAD, VDIR, NULL }, - { "boottime", &boottime.tv_sec, KTT_INT, VREAD, VREG, NULL }, - { "copyright", copyright, KTT_STRING, VREAD, VREG, NULL }, - { "hostname", 0, KTT_HOSTNAME,VREAD|VWRITE,VREG, NULL }, - { "bootfile", 0, KTT_BOOTFILE, VREAD, VREG, NULL }, - { "hz", &hz, KTT_INT, VREAD, VREG, NULL }, - { "loadavg", 0, KTT_AVENRUN, VREAD, VREG, NULL }, - { "pagesize", &cnt.v_page_size, KTT_INT, VREAD, VREG, NULL }, - { "physmem", &physmem, KTT_INT, VREAD, VREG, NULL }, +#define N(s) sizeof(s)-1, s + /* name data tag type ro/rw */ + { DT_DIR, N("."), 0, KTT_NULL, VDIR, DIR_MODE }, + { DT_DIR, N(".."), 0, KTT_NULL, VDIR, DIR_MODE }, + { DT_REG, N("boottime"), &boottime.tv_sec, KTT_INT, VREG, READ_MODE }, + { DT_REG, N("copyright"), copyright, KTT_STRING, VREG, READ_MODE }, + { DT_REG, N("hostname"), 0, KTT_HOSTNAME, VREG, WRITE_MODE }, + { DT_REG, N("bootfile"), 0, KTT_BOOTFILE, VREG, READ_MODE }, + { DT_REG, N("hz"), &hz, KTT_INT, VREG, READ_MODE }, + { DT_REG, N("loadavg"), 0, KTT_AVENRUN, VREG, READ_MODE }, + { DT_REG, N("pagesize"), &cnt.v_page_size, KTT_INT, VREG, READ_MODE }, + { DT_REG, N("physmem"), &physmem, KTT_INT, VREG, READ_MODE }, #if 0 - { "root", 0, KTT_NULL, VREAD, VDIR, &rootdir}, - { "rootdev", 0, KTT_NULL, VREAD, VBLK, &rootvp }, - { "rrootdev", 0, KTT_NULL, VREAD, VCHR, &rrootvp}, + { DT_DIR, N("root"), 0, KTT_NULL, VDIR, DIR_MODE }, + { DT_BLK, N("rootdev"), &rootdev, KTT_DEVICE, VBLK, READ_MODE }, + { DT_CHR, N("rrootdev"), &rrootdev, KTT_DEVICE, VCHR, READ_MODE }, #endif - { "time", 0, KTT_TIME, VREAD, VREG, NULL }, - { "version", version, KTT_STRING, VREAD, VREG, NULL }, + { DT_REG, N("time"), 0, KTT_TIME, VREG, READ_MODE }, + { DT_REG, N("version"), version, KTT_STRING, VREG, READ_MODE }, +#undef N }; - static int nkern_targets = sizeof(kern_targets) / sizeof(kern_targets[0]); static int kernfs_access __P((struct vop_access_args *ap)); @@ -130,6 +133,7 @@ kernfs_xread(kt, buf, len, lenp) int len; int *lenp; { + switch (kt->kt_tag) { case KTT_TIME: { struct timeval tv; @@ -183,14 +187,12 @@ kernfs_xread(kt, buf, len, lenp) case KTT_AVENRUN: sprintf(buf, "%ld %ld %ld %ld\n", - averunnable.ldavg[0], - averunnable.ldavg[1], - averunnable.ldavg[2], - averunnable.fscale); + averunnable.ldavg[0], averunnable.ldavg[1], + averunnable.ldavg[2], averunnable.fscale); break; default: - return (EINVAL); + return (EIO); } *lenp = strlen(buf); @@ -203,15 +205,15 @@ kernfs_xwrite(kt, buf, len) char *buf; int len; { + switch (kt->kt_tag) { - case KTT_HOSTNAME: { + case KTT_HOSTNAME: /* XXX BOGUS !!! no check for the length */ if (buf[len-1] == '\n') --len; bcopy(buf, hostname, len); hostname[len] = '\0'; return (0); - } default: return (EIO); @@ -231,26 +233,32 @@ kernfs_lookup(ap) struct componentname * a_cnp; } */ *ap; { + struct componentname *cnp = ap->a_cnp; struct vnode **vpp = ap->a_vpp; struct vnode *dvp = ap->a_dvp; - struct componentname *cnp = ap->a_cnp; + char *pname = cnp->cn_nameptr; + struct proc *p = cnp->cn_proc; + struct kern_target *kt; struct vnode *fvp; int nameiop = cnp->cn_nameiop; int error, i; - char *pname; #ifdef KERNFS_DIAGNOSTIC printf("kernfs_lookup(%x)\n", ap); printf("kernfs_lookup(dp = %x, vpp = %x, cnp = %x)\n", dvp, vpp, ap->a_cnp); -#endif - pname = cnp->cn_nameptr; -#ifdef KERNFS_DIAGNOSTIC printf("kernfs_lookup(%s)\n", pname); #endif + + *vpp = NULLVP; + + if (cnp->cn_nameiop == DELETE || cnp->cn_nameiop == RENAME) + return (EROFS); + + VOP_UNLOCK(dvp, 0, p); if (cnp->cn_namelen == 1 && *pname == '.') { *vpp = dvp; VREF(dvp); - /*VOP_LOCK(dvp);*/ + vn_lock(dvp, LK_SHARED | LK_RETRY, p); return (0); } @@ -258,91 +266,58 @@ kernfs_lookup(ap) if (cnp->cn_namelen == 4 && bcmp(pname, "root", 4) == 0) { *vpp = rootdir; VREF(rootdir); - VOP_LOCK(rootdir); + vn_lock(rootdir, LK_SHARED | LK_RETRY, p) return (0); } #endif - error = ENOENT; - - for (i = 0; i < nkern_targets; i++) { - struct kern_target *kt = &kern_targets[i]; - if (cnp->cn_namelen == strlen(kt->kt_name) && - bcmp(kt->kt_name, pname, cnp->cn_namelen) == 0) { - error = 0; - break; - } + for (kt = kern_targets, i = 0; i < nkern_targets; kt++, i++) { + if (cnp->cn_namelen == kt->kt_namlen && + bcmp(kt->kt_name, pname, cnp->cn_namelen) == 0) + goto found; } #ifdef KERNFS_DIAGNOSTIC - printf("kernfs_lookup: i = %d, error = %d\n", i, error); + printf("kernfs_lookup: i = %d, failed", i); #endif - /* - * If the name wasn't found, and this is not a LOOKUP - * request, we return EOPNOTSUPP so that the initial namei() - * fails and the higher level routines will not try to call - * our VOP_* functions. - */ - if (error) { - if (nameiop != LOOKUP) - error = EOPNOTSUPP; - goto bad; - } - - /* - * DELETE requests are not supported. - */ - if (nameiop == DELETE) { - error = EOPNOTSUPP; - goto bad; - } + vn_lock(dvp, LK_SHARED | LK_RETRY, p); + return (cnp->cn_nameiop == LOOKUP ? ENOENT : EROFS); - /* - * Allow CREATE requests if the name in question can - * be written to. This allows open(name, O_RDWR | O_CREAT) - * to work. Otherwise CREATE requests are not supported. - */ - if (nameiop == CREATE && (kern_targets[i].kt_rw & VWRITE == 0)) { - error = EOPNOTSUPP; - goto bad; - } - - /* - * Check if this name has already has a vnode associated with it. - */ - if (kern_targets[i].kt_vp) { - if (*kern_targets[i].kt_vp) { - *vpp = *kern_targets[i].kt_vp; - VREF(*vpp); - VOP_LOCK(*vpp); - return (0); +found: + if (kt->kt_tag == KTT_DEVICE) { + dev_t *dp = kt->kt_data; + loop: + if (*dp == NODEV || !vfinddev(*dp, kt->kt_vtype, &fvp)) { + vn_lock(dvp, LK_SHARED | LK_RETRY, p); + return (ENOENT); } - error = ENXIO; - goto bad; + *vpp = fvp; + if (vget(fvp, LK_EXCLUSIVE, p)) + goto loop; + return (0); } #ifdef KERNFS_DIAGNOSTIC printf("kernfs_lookup: allocate new vnode\n"); #endif - error = getnewvnode(VT_KERNFS, dvp->v_mount, kernfs_vnodeop_p, &fvp); - if (error) - goto bad; - MALLOC(fvp->v_data, void *, sizeof(struct kernfs_node), M_TEMP, M_WAITOK); - VTOKERN(fvp)->kf_kt = &kern_targets[i]; - fvp->v_type = VTOKERN(fvp)->kf_kt->kt_vtype; + if (error = getnewvnode(VT_KERNFS, dvp->v_mount, kernfs_vnodeop_p, + &fvp)) { + vn_lock(dvp, LK_SHARED | LK_RETRY, p); + return (error); + } + + MALLOC(fvp->v_data, void *, sizeof(struct kernfs_node), M_TEMP, + M_WAITOK); + VTOKERN(fvp)->kf_kt = kt; + fvp->v_type = kt->kt_vtype; + vn_lock(fvp, LK_SHARED | LK_RETRY, p); *vpp = fvp; + #ifdef KERNFS_DIAGNOSTIC printf("kernfs_lookup: newvp = %x\n", fvp); #endif return (0); - -bad:; - *vpp = NULL; -#ifdef KERNFS_DIAGNOSTIC - printf("kernfs_lookup: error = %d\n", error); -#endif - return (error); } static int @@ -354,22 +329,8 @@ kernfs_open(ap) struct proc *a_p; } */ *ap; { - struct vnode *vp = ap->a_vp; - - /* - * Can always open the root (modulo perms) - */ - if (vp->v_flag & VROOT) - return (0); - -#ifdef KERNFS_DIAGNOSTIC - printf("kernfs_open, mode = %x, file = %s\n", - ap->a_mode, VTOKERN(vp)->kf_kt->kt_name); -#endif - - if ((ap->a_mode & FWRITE) && !(VTOKERN(vp)->kf_kt->kt_rw & VWRITE)) - return (EOPNOTSUPP); + /* Only need to check access permissions. */ return (0); } @@ -382,33 +343,45 @@ kernfs_access(ap) struct proc *a_p; } */ *ap; { - struct vnode *vp = ap->a_vp; - struct ucred *cred = ap->a_cred; - mode_t mode = ap->a_mode; - - if (mode & VEXEC) { - if (vp->v_flag & VROOT) - return (0); - return (EACCES); - } + register struct vnode *vp = ap->a_vp; + register struct ucred *cred = ap->a_cred; + mode_t amode = ap->a_mode; + mode_t fmode = + (vp->v_flag & VROOT) ? DIR_MODE : VTOKERN(vp)->kf_kt->kt_mode; + mode_t mask = 0; + register gid_t *gp; + int i; - if (cred->cr_uid == 0) { - if ((vp->v_flag & VROOT) == 0) { - struct kern_target *kt = VTOKERN(vp)->kf_kt; + /* Some files are simply not modifiable. */ + if ((amode & VWRITE) && (fmode & (S_IWUSR|S_IWGRP|S_IWOTH)) == 0) + return (EPERM); - if ((mode & VWRITE) && !(kt->kt_rw & VWRITE)) - return (EROFS); - } + /* Root can do anything else. */ + if (cred->cr_uid == 0) return (0); - } - if (mode & VWRITE) - return (EACCES); + /* Check for group 0 (wheel) permissions. */ + for (i = 0, gp = cred->cr_groups; i < cred->cr_ngroups; i++, gp++) + if (*gp == 0) { + if (amode & VEXEC) + mask |= S_IXGRP; + if (amode & VREAD) + mask |= S_IRGRP; + if (amode & VWRITE) + mask |= S_IWGRP; + return ((fmode & mask) == mask ? 0 : EACCES); + } - return (0); + /* Otherwise, check everyone else. */ + if (amode & VEXEC) + mask |= S_IXOTH; + if (amode & VREAD) + mask |= S_IROTH; + if (amode & VWRITE) + mask |= S_IWOTH; + return ((fmode & mask) == mask ? 0 : EACCES); } - static int kernfs_getattr(ap) struct vop_getattr_args /* { @@ -420,6 +393,7 @@ kernfs_getattr(ap) { struct vnode *vp = ap->a_vp; struct vattr *vap = ap->a_vap; + struct timeval tv; int error = 0; char strbuf[KSTRING]; @@ -428,7 +402,7 @@ kernfs_getattr(ap) vap->va_uid = 0; vap->va_gid = 0; vap->va_fsid = vp->v_mount->mnt_stat.f_fsid.val[0]; - /* vap->va_qsize = 0; */ + vap->va_size = 0; vap->va_blocksize = DEV_BSIZE; { struct timeval tv; @@ -440,7 +414,6 @@ kernfs_getattr(ap) vap->va_gen = 0; vap->va_flags = 0; vap->va_rdev = 0; - /* vap->va_qbytes = 0; */ vap->va_bytes = 0; if (vp->v_flag & VROOT) { @@ -459,14 +432,13 @@ kernfs_getattr(ap) printf("kernfs_getattr: stat target %s\n", kt->kt_name); #endif vap->va_type = kt->kt_vtype; - vap->va_mode = (kt->kt_rw & VWRITE ? WRITE_MODE : READ_MODE); + vap->va_mode = kt->kt_mode; vap->va_nlink = 1; - vap->va_fileid = 3 + (kt - kern_targets) / sizeof(*kt); + vap->va_fileid = 1 + (kt - kern_targets) / sizeof(*kt); error = kernfs_xread(kt, strbuf, sizeof(strbuf), &nbytes); vap->va_size = nbytes; } - vp->v_type = vap->va_type; #ifdef KERNFS_DIAGNOSTIC printf("kernfs_getattr: return error %d\n", error); #endif @@ -509,7 +481,7 @@ kernfs_read(ap) int error, len; char *cp; - if (vp->v_flag & VROOT) + if (vp->v_type == VDIR) return (EOPNOTSUPP); kt = VTOKERN(vp)->kf_kt; @@ -519,12 +491,11 @@ kernfs_read(ap) #endif len = 0; - error = kernfs_xread(kt, strbuf, sizeof(strbuf), &len); - if (error) + if (error = kernfs_xread(kt, strbuf, sizeof(strbuf), &len)) return (error); - cp = strbuf + off; - len -= off; - return (uiomove(cp, len, uio)); + if (len <= off) + return (0); + return (uiomove(&strbuf[off], len - off, uio)); } static int @@ -542,8 +513,8 @@ kernfs_write(ap) int error, xlen; char strbuf[KSTRING]; - if (vp->v_flag & VROOT) - return (0); + if (vp->v_type == VDIR) + return (EOPNOTSUPP); kt = VTOKERN(vp)->kf_kt; @@ -551,8 +522,7 @@ kernfs_write(ap) return (EINVAL); xlen = min(uio->uio_resid, KSTRING-1); - error = uiomove(strbuf, xlen, uio); - if (error) + if (error = uiomove(strbuf, xlen, uio)) return (error); if (uio->uio_resid != 0) @@ -563,33 +533,52 @@ kernfs_write(ap) return (kernfs_xwrite(kt, strbuf, xlen)); } - static int kernfs_readdir(ap) struct vop_readdir_args /* { struct vnode *a_vp; struct uio *a_uio; struct ucred *a_cred; + int *a_eofflag; + u_long *a_cookies; + int a_ncookies; } */ *ap; { + int error, i; struct uio *uio = ap->a_uio; - int i; - int error; + struct kern_target *kt; + struct dirent d; + + if (ap->a_vp->v_type != VDIR) + return (ENOTDIR); + + /* + * We don't allow exporting kernfs mounts, and currently local + * requests do not need cookies. + */ + if (ap->a_ncookies != NULL) + panic("kernfs_readdir: not hungry"); i = uio->uio_offset / UIO_MX; error = 0; - while (uio->uio_resid > 0 && i < nkern_targets) { - struct dirent d; + for (kt = &kern_targets[i]; + uio->uio_resid >= UIO_MX && i < nkern_targets; kt++, i++) { struct dirent *dp = &d; - struct kern_target *kt = &kern_targets[i]; #ifdef KERNFS_DIAGNOSTIC printf("kernfs_readdir: i = %d\n", i); #endif - bzero((caddr_t) dp, UIO_MX); + if (kt->kt_tag == KTT_DEVICE) { + dev_t *dp = kt->kt_data; + struct vnode *fvp; + + if (*dp == NODEV || !vfinddev(*dp, kt->kt_vtype, &fvp)) + continue; + } - dp->d_namlen = strlen(kt->kt_name); - bcopy(kt->kt_name, dp->d_name, dp->d_namlen+1); + bzero((caddr_t)dp, UIO_MX); + dp->d_namlen = kt->kt_namlen; + bcopy(kt->kt_name, dp->d_name, kt->kt_namlen+1); #ifdef KERNFS_DIAGNOSTIC printf("kernfs_readdir: name = %s, len = %d\n", @@ -600,14 +589,12 @@ kernfs_readdir(ap) */ dp->d_reclen = UIO_MX; dp->d_fileno = i + 3; - dp->d_type = DT_UNKNOWN; /* XXX */ + dp->d_type = kt->kt_type; /* * And ship to userland */ - error = uiomove((caddr_t) dp, UIO_MX, uio); - if (error) + if (error = uiomove((caddr_t)dp, UIO_MX, uio)) break; - i++; } uio->uio_offset = i * UIO_MX; @@ -619,18 +606,20 @@ static int kernfs_inactive(ap) struct vop_inactive_args /* { struct vnode *a_vp; + struct proc *a_p; } */ *ap; { struct vnode *vp = ap->a_vp; +#ifdef KERNFS_DIAGNOSTIC + printf("kernfs_inactive(%x)\n", vp); +#endif /* * Clear out the v_type field to avoid * nasty things happening in vgone(). */ + VOP_UNLOCK(vp, 0, ap->a_p); vp->v_type = VNON; -#ifdef KERNFS_DIAGNOSTIC - printf("kernfs_inactive(%x)\n", vp); -#endif return (0); } @@ -641,6 +630,7 @@ kernfs_reclaim(ap) } */ *ap; { struct vnode *vp = ap->a_vp; + #ifdef KERNFS_DIAGNOSTIC printf("kernfs_reclaim(%x)\n", vp); #endif @@ -717,16 +707,6 @@ kernfs_vfree(ap) } /* - * Kernfs vnode unsupported operation - */ -static int -kernfs_enotsupp() -{ - - return (EOPNOTSUPP); -} - -/* * Kernfs "should never get here" operation */ static int @@ -735,42 +715,42 @@ kernfs_badop() return (EIO); } -#define kernfs_create ((int (*) __P((struct vop_create_args *)))kernfs_enotsupp) -#define kernfs_mknod ((int (*) __P((struct vop_mknod_args *)))kernfs_enotsupp) +#define kernfs_create ((int (*) __P((struct vop_create_args *)))eopnotsupp) +#define kernfs_mknod ((int (*) __P((struct vop_mknod_args *)))eopnotsupp) #define kernfs_close ((int (*) __P((struct vop_close_args *)))nullop) -#define kernfs_ioctl ((int (*) __P((struct vop_ioctl_args *)))kernfs_enotsupp) -#define kernfs_select ((int (*) __P((struct vop_select_args *)))kernfs_enotsupp) -#define kernfs_mmap ((int (*) __P((struct vop_mmap_args *)))kernfs_enotsupp) +#define kernfs_ioctl ((int (*) __P((struct vop_ioctl_args *)))eopnotsupp) +#define kernfs_select ((int (*) __P((struct vop_select_args *)))eopnotsupp) +#define kernfs_revoke vop_revoke +#define kernfs_mmap ((int (*) __P((struct vop_mmap_args *)))eopnotsupp) #define kernfs_fsync ((int (*) __P((struct vop_fsync_args *)))nullop) #define kernfs_seek ((int (*) __P((struct vop_seek_args *)))nullop) -#define kernfs_remove ((int (*) __P((struct vop_remove_args *)))kernfs_enotsupp) -#define kernfs_link ((int (*) __P((struct vop_link_args *)))kernfs_enotsupp) -#define kernfs_rename ((int (*) __P((struct vop_rename_args *)))kernfs_enotsupp) -#define kernfs_mkdir ((int (*) __P((struct vop_mkdir_args *)))kernfs_enotsupp) -#define kernfs_rmdir ((int (*) __P((struct vop_rmdir_args *)))kernfs_enotsupp) -#define kernfs_symlink ((int (*) __P((struct vop_symlink_args *)))kernfs_enotsupp) -#define kernfs_readlink \ - ((int (*) __P((struct vop_readlink_args *)))kernfs_enotsupp) +#define kernfs_remove ((int (*) __P((struct vop_remove_args *)))eopnotsupp) +#define kernfs_link ((int (*) __P((struct vop_link_args *)))eopnotsupp) +#define kernfs_rename ((int (*) __P((struct vop_rename_args *)))eopnotsupp) +#define kernfs_mkdir ((int (*) __P((struct vop_mkdir_args *)))eopnotsupp) +#define kernfs_rmdir ((int (*) __P((struct vop_rmdir_args *)))eopnotsupp) +#define kernfs_symlink ((int (*) __P((struct vop_symlink_args *)))eopnotsupp) +#define kernfs_readlink ((int (*) __P((struct vop_readlink_args *)))eopnotsupp) #define kernfs_abortop ((int (*) __P((struct vop_abortop_args *)))nullop) -#define kernfs_lock ((int (*) __P((struct vop_lock_args *)))nullop) -#define kernfs_unlock ((int (*) __P((struct vop_unlock_args *)))nullop) +#define kernfs_lock ((int (*) __P((struct vop_lock_args *)))vop_nolock) +#define kernfs_unlock ((int (*) __P((struct vop_unlock_args *)))vop_nounlock) #define kernfs_bmap ((int (*) __P((struct vop_bmap_args *)))kernfs_badop) -#define kernfs_strategy ((int (*) __P((struct vop_strategy_args *)))kernfs_badop) -#define kernfs_islocked ((int (*) __P((struct vop_islocked_args *)))nullop) -#define kernfs_advlock ((int (*) __P((struct vop_advlock_args *)))kernfs_enotsupp) -#define kernfs_blkatoff \ - ((int (*) __P((struct vop_blkatoff_args *)))kernfs_enotsupp) +#define kernfs_strategy \ + ((int (*) __P((struct vop_strategy_args *)))kernfs_badop) +#define kernfs_islocked \ + ((int (*) __P((struct vop_islocked_args *)))vop_noislocked) +#define kernfs_advlock ((int (*) __P((struct vop_advlock_args *)))eopnotsupp) +#define kernfs_blkatoff ((int (*) __P((struct vop_blkatoff_args *)))eopnotsupp) #define kernfs_valloc ((int(*) __P(( \ struct vnode *pvp, \ int mode, \ struct ucred *cred, \ - struct vnode **vpp))) kernfs_enotsupp) -#define kernfs_truncate \ - ((int (*) __P((struct vop_truncate_args *)))kernfs_enotsupp) -#define kernfs_update ((int (*) __P((struct vop_update_args *)))kernfs_enotsupp) -#define kernfs_bwrite ((int (*) __P((struct vop_bwrite_args *)))kernfs_enotsupp) + struct vnode **vpp))) eopnotsupp) +#define kernfs_truncate ((int (*) __P((struct vop_truncate_args *)))eopnotsupp) +#define kernfs_update ((int (*) __P((struct vop_update_args *)))eopnotsupp) +#define kernfs_bwrite ((int (*) __P((struct vop_bwrite_args *)))eopnotsupp) -vop_t **kernfs_vnodeop_p; +vop_t **kernfs_vnodeop_p; static struct vnodeopv_entry_desc kernfs_vnodeop_entries[] = { { &vop_default_desc, (vop_t *)vn_default_error }, { &vop_lookup_desc, (vop_t *)kernfs_lookup }, /* lookup */ @@ -785,6 +765,7 @@ static struct vnodeopv_entry_desc kernfs_vnodeop_entries[] = { { &vop_write_desc, (vop_t *)kernfs_write }, /* write */ { &vop_ioctl_desc, (vop_t *)kernfs_ioctl }, /* ioctl */ { &vop_select_desc, (vop_t *)kernfs_select }, /* select */ + { &vop_revoke_desc, (vop_t *)kernfs_revoke }, /* revoke */ { &vop_mmap_desc, (vop_t *)kernfs_mmap }, /* mmap */ { &vop_fsync_desc, (vop_t *)kernfs_fsync }, /* fsync */ { &vop_seek_desc, (vop_t *)kernfs_seek }, /* seek */ diff --git a/sys/miscfs/nullfs/null.h b/sys/miscfs/nullfs/null.h index beadb42..70a81b2 100644 --- a/sys/miscfs/nullfs/null.h +++ b/sys/miscfs/nullfs/null.h @@ -33,7 +33,7 @@ * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * - * @(#)null.h 8.2 (Berkeley) 1/21/94 + * @(#)null.h 8.3 (Berkeley) 8/20/94 * * $FreeBSD$ */ @@ -52,8 +52,7 @@ struct null_mount { * A cache of vnode references */ struct null_node { - struct null_node *null_forw; /* Hash chain */ - struct null_node *null_back; + LIST_ENTRY(null_node) null_hash; /* Hash list */ struct vnode *null_lowervp; /* VREFed once */ struct vnode *null_vnode; /* Back pointer */ }; diff --git a/sys/miscfs/nullfs/null_subr.c b/sys/miscfs/nullfs/null_subr.c index a14a7e4..4418631 100644 --- a/sys/miscfs/nullfs/null_subr.c +++ b/sys/miscfs/nullfs/null_subr.c @@ -33,13 +33,14 @@ * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * - * @(#)null_subr.c 8.4 (Berkeley) 1/21/94 + * @(#)null_subr.c 8.7 (Berkeley) 5/14/95 * * $FreeBSD$ */ #include <sys/param.h> #include <sys/systm.h> +#include <sys/proc.h> #include <sys/time.h> #include <sys/types.h> #include <sys/vnode.h> @@ -48,11 +49,8 @@ #include <sys/malloc.h> #include <miscfs/nullfs/null.h> -extern int nullfs_init __P((void)); - #define LOG2_SIZEVNODE 7 /* log2(sizeof struct vnode) */ #define NNULLNODECACHE 16 -#define NULL_NHASH(vp) ((((u_long)vp)>>LOG2_SIZEVNODE) & (NNULLNODECACHE-1)) /* * Null layer cache: @@ -62,51 +60,32 @@ extern int nullfs_init __P((void)); * alias is removed the lower vnode is vrele'd. */ -/* - * Cache head - */ -struct null_node_cache { - struct null_node *ac_forw; - struct null_node *ac_back; -}; - -static struct null_node_cache null_node_cache[NNULLNODECACHE]; +#define NULL_NHASH(vp) \ + (&null_node_hashtbl[(((u_long)vp)>>LOG2_SIZEVNODE) & null_node_hash]) +LIST_HEAD(null_node_hashhead, null_node) *null_node_hashtbl; +u_long null_node_hash; static int null_node_alloc __P((struct mount *mp, struct vnode *lowervp, struct vnode **vpp)); static struct vnode * null_node_find __P((struct mount *mp, struct vnode *lowervp)); -static struct null_node_cache * - null_node_hash __P((struct vnode *lowervp)); /* * Initialise cache headers */ int -nullfs_init() +nullfs_init(vfsp) + struct vfsconf *vfsp; { - struct null_node_cache *ac; + #ifdef NULLFS_DIAGNOSTIC printf("nullfs_init\n"); /* printed during system boot */ #endif - - for (ac = null_node_cache; ac < null_node_cache + NNULLNODECACHE; ac++) - ac->ac_forw = ac->ac_back = (struct null_node *) ac; + null_node_hashtbl = hashinit(NNULLNODECACHE, M_CACHE, &null_node_hash); return (0); } /* - * Compute hash list for given lower vnode - */ -static struct null_node_cache * -null_node_hash(lowervp) -struct vnode *lowervp; -{ - - return (&null_node_cache[NULL_NHASH(lowervp)]); -} - -/* * Return a VREF'ed alias for lower vnode if already exists, else 0. */ static struct vnode * @@ -114,7 +93,8 @@ null_node_find(mp, lowervp) struct mount *mp; struct vnode *lowervp; { - struct null_node_cache *hd; + struct proc *p = curproc; /* XXX */ + struct null_node_hashhead *hd; struct null_node *a; struct vnode *vp; @@ -124,9 +104,9 @@ null_node_find(mp, lowervp) * the lower vnode. If found, the increment the null_node * reference count (but NOT the lower vnode's VREF counter). */ - hd = null_node_hash(lowervp); + hd = NULL_NHASH(lowervp); loop: - for (a = hd->ac_forw; a != (struct null_node *) hd; a = a->null_forw) { + for (a = hd->lh_first; a != 0; a = a->null_hash.le_next) { if (a->null_lowervp == lowervp && NULLTOV(a)->v_mount == mp) { vp = NULLTOV(a); /* @@ -134,7 +114,7 @@ loop: * stuff, but we don't want to lock * the lower node. */ - if (vget(vp, 0)) { + if (vget(vp, 0, p)) { printf ("null_node_find: vget failed.\n"); goto loop; }; @@ -157,7 +137,7 @@ null_node_alloc(mp, lowervp, vpp) struct vnode *lowervp; struct vnode **vpp; { - struct null_node_cache *hd; + struct null_node_hashhead *hd; struct null_node *xp; struct vnode *othervp, *vp; int error; @@ -194,8 +174,8 @@ null_node_alloc(mp, lowervp, vpp) return 0; }; VREF(lowervp); /* Extra VREF will be vrele'd in null_node_create */ - hd = null_node_hash(lowervp); - insque(xp, hd); + hd = NULL_NHASH(lowervp); + LIST_INSERT_HEAD(hd, xp, null_hash); return 0; } @@ -250,9 +230,8 @@ null_node_create(mp, lowervp, newvpp) #ifdef DIAGNOSTIC if (lowervp->v_usecount < 1) { /* Should never happen... */ - vprint ("null_node_create: alias ",aliasvp); - vprint ("null_node_create: lower ",lowervp); - printf ("null_node_create: lower has 0 usecount.\n"); + vprint ("null_node_create: alias ", aliasvp); + vprint ("null_node_create: lower ", lowervp); panic ("null_node_create: lower has 0 usecount."); }; #endif diff --git a/sys/miscfs/nullfs/null_vfsops.c b/sys/miscfs/nullfs/null_vfsops.c index e2aeeba..339b7c1 100644 --- a/sys/miscfs/nullfs/null_vfsops.c +++ b/sys/miscfs/nullfs/null_vfsops.c @@ -1,5 +1,5 @@ /* - * Copyright (c) 1992, 1993 + * Copyright (c) 1992, 1993, 1995 * The Regents of the University of California. All rights reserved. * * This code is derived from software donated to Berkeley by @@ -46,7 +46,7 @@ #include <sys/param.h> #include <sys/systm.h> -#include <sys/kernel.h> +#include <sys/proc.h> #include <sys/time.h> #include <sys/types.h> #include <sys/vnode.h> @@ -55,7 +55,7 @@ #include <sys/malloc.h> #include <miscfs/nullfs/null.h> -extern int nullfs_init __P((void)); +extern int nullfs_init __P((struct vfsconf *)); static int nullfs_fhtovp __P((struct mount *mp, struct fid *fidp, struct mbuf *nam, struct vnode **vpp, @@ -146,7 +146,7 @@ nullfs_mount(mp, path, data, ndp, p) /* * Unlock the node (either the lower or the alias) */ - VOP_UNLOCK(vp); + VOP_UNLOCK(vp, 0, p); /* * Make sure the node alias worked */ @@ -166,7 +166,7 @@ nullfs_mount(mp, path, data, ndp, p) if (NULLVPTOLOWERVP(nullm_rootvp)->v_mount->mnt_flag & MNT_LOCAL) mp->mnt_flag |= MNT_LOCAL; mp->mnt_data = (qaddr_t) xmp; - getnewfsid(mp, MOUNT_LOFS); + vfs_getnewfsid(mp); (void) copyinstr(path, mp->mnt_stat.f_mntonname, MNAMELEN - 1, &size); bzero(mp->mnt_stat.f_mntonname + size, MNAMELEN - size); @@ -212,12 +212,8 @@ nullfs_unmount(mp, mntflags, p) printf("nullfs_unmount(mp = %x)\n", mp); #endif - if (mntflags & MNT_FORCE) { - /* lofs can never be rootfs so don't check for it */ - if (!doforce) - return (EINVAL); + if (mntflags & MNT_FORCE) flags |= FORCECLOSE; - } /* * Clear out buffer cache. I don't think we @@ -259,6 +255,7 @@ nullfs_root(mp, vpp) struct mount *mp; struct vnode **vpp; { + struct proc *p = curproc; /* XXX */ struct vnode *vp; #ifdef NULLFS_DIAGNOSTIC @@ -273,7 +270,7 @@ nullfs_root(mp, vpp) */ vp = MOUNTTONULLMOUNT(mp)->nullm_rootvp; VREF(vp); - VOP_LOCK(vp); + vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, p); *vpp = vp; return 0; } diff --git a/sys/miscfs/nullfs/null_vnops.c b/sys/miscfs/nullfs/null_vnops.c index ec279bc..c1af96f 100644 --- a/sys/miscfs/nullfs/null_vnops.c +++ b/sys/miscfs/nullfs/null_vnops.c @@ -33,7 +33,13 @@ * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * - * @(#)null_vnops.c 8.1 (Berkeley) 6/10/93 + * @(#)null_vnops.c 8.6 (Berkeley) 5/27/95 + * + * Ancestors: + * @(#)lofs_vnops.c 1.2 (Berkeley) 6/18/92 + * $Id: null_vnops.c,v 1.11.2000.1 1996/09/17 14:32:31 peter Exp $ + * ...and... + * @(#)null_vnodeops.c 1.20 92/07/07 UCLA Ficus project * * $FreeBSD$ */ @@ -88,13 +94,21 @@ * in the arguments and, if a vnode is return by the operation, * stacks a null-node on top of the returned vnode. * - * Although bypass handles most operations, - * vop_getattr, _inactive, _reclaim, and _print are not bypassed. - * Vop_getattr must change the fsid being returned. + * Although bypass handles most operations, vop_getattr, vop_lock, + * vop_unlock, vop_inactive, vop_reclaim, and vop_print are not + * bypassed. Vop_getattr must change the fsid being returned. + * Vop_lock and vop_unlock must handle any locking for the + * current vnode as well as pass the lock request down. * Vop_inactive and vop_reclaim are not bypassed so that - * they can handle freeing null-layer specific data. - * Vop_print is not bypassed to avoid excessive debugging - * information. + * they can handle freeing null-layer specific data. Vop_print + * is not bypassed to avoid excessive debugging information. + * Also, certain vnode operations change the locking state within + * the operation (create, mknod, remove, link, rename, mkdir, rmdir, + * and symlink). Ideally these operations should not change the + * lock state, but should be changed to let the caller of the + * function unlock them. Otherwise all intermediate vnode layers + * (such as union, umapfs, etc) must catch these functions to do + * the necessary locking at their layer. * * * INSTANTIATING VNODE STACKS @@ -178,7 +192,7 @@ static int null_bug_bypass = 0; /* for debugging: enables bypass printf'ing */ SYSCTL_INT(_debug, OID_AUTO, nullfs_bug_bypass, CTLFLAG_RW, &null_bug_bypass, 0, ""); -static int null_bypass __P((struct vop_generic_args *ap)); +int null_bypass __P((struct vop_generic_args *ap)); static int null_bwrite __P((struct vop_bwrite_args *ap)); static int null_getattr __P((struct vop_getattr_args *ap)); static int null_inactive __P((struct vop_inactive_args *ap)); @@ -211,7 +225,7 @@ static int null_strategy __P((struct vop_strategy_args *ap)); * - all mapped vnodes are of our vnode-type (NEEDSWORK: * problems on rmdir'ing mount points and renaming?) */ -static int +int null_bypass(ap) struct vop_generic_args /* { struct vnodeop_desc *a_desc; @@ -254,7 +268,8 @@ null_bypass(ap) * are of our type. Check for and don't map any * that aren't. (We must always map first vp or vclean fails.) */ - if (i && (*this_vp_p)->v_op != null_vnodeop_p) { + if (i && (*this_vp_p == NULL || + (*this_vp_p)->v_op != null_vnodeop_p)) { old_vps[i] = NULL; } else { old_vps[i] = *this_vp_p; @@ -317,6 +332,105 @@ null_bypass(ap) return (error); } +/* + * We have to carry on the locking protocol on the null layer vnodes + * as we progress through the tree. We also have to enforce read-only + * if this layer is mounted read-only. + */ +static int +null_lookup(ap) + struct vop_lookup_args /* { + struct vnode * a_dvp; + struct vnode ** a_vpp; + struct componentname * a_cnp; + } */ *ap; +{ + struct componentname *cnp = ap->a_cnp; + struct proc *p = cnp->cn_proc; + int flags = cnp->cn_flags; + struct vop_lock_args lockargs; + struct vop_unlock_args unlockargs; + struct vnode *dvp, *vp; + int error; + + if ((flags & ISLASTCN) && (ap->a_dvp->v_mount->mnt_flag & MNT_RDONLY) && + (cnp->cn_nameiop == DELETE || cnp->cn_nameiop == RENAME)) + return (EROFS); + error = null_bypass(ap); + if (error == EJUSTRETURN && (flags & ISLASTCN) && + (ap->a_dvp->v_mount->mnt_flag & MNT_RDONLY) && + (cnp->cn_nameiop == CREATE || cnp->cn_nameiop == RENAME)) + error = EROFS; + /* + * We must do the same locking and unlocking at this layer as + * is done in the layers below us. We could figure this out + * based on the error return and the LASTCN, LOCKPARENT, and + * LOCKLEAF flags. However, it is more expidient to just find + * out the state of the lower level vnodes and set ours to the + * same state. + */ + dvp = ap->a_dvp; + vp = *ap->a_vpp; + if (dvp == vp) + return (error); + if (!VOP_ISLOCKED(dvp)) { + unlockargs.a_vp = dvp; + unlockargs.a_flags = 0; + unlockargs.a_p = p; + vop_nounlock(&unlockargs); + } + if (vp != NULL && VOP_ISLOCKED(vp)) { + lockargs.a_vp = vp; + lockargs.a_flags = LK_SHARED; + lockargs.a_p = p; + vop_nolock(&lockargs); + } + return (error); +} + +/* + * Setattr call. Disallow write attempts if the layer is mounted read-only. + */ +int +null_setattr(ap) + struct vop_setattr_args /* { + struct vnodeop_desc *a_desc; + struct vnode *a_vp; + struct vattr *a_vap; + struct ucred *a_cred; + struct proc *a_p; + } */ *ap; +{ + struct vnode *vp = ap->a_vp; + struct vattr *vap = ap->a_vap; + + if ((vap->va_flags != VNOVAL || vap->va_uid != (uid_t)VNOVAL || + vap->va_gid != (gid_t)VNOVAL || vap->va_atime.ts_sec != VNOVAL || + vap->va_mtime.ts_sec != VNOVAL || vap->va_mode != (mode_t)VNOVAL) && + (vp->v_mount->mnt_flag & MNT_RDONLY)) + return (EROFS); + if (vap->va_size != VNOVAL) { + switch (vp->v_type) { + case VDIR: + return (EISDIR); + case VCHR: + case VBLK: + case VSOCK: + case VFIFO: + return (0); + case VREG: + case VLNK: + default: + /* + * Disallow write attempts if the filesystem is + * mounted read-only. + */ + if (vp->v_mount->mnt_flag & MNT_RDONLY) + return (EROFS); + } + } + return (null_bypass(ap)); +} /* * We handle getattr only to change the fsid. @@ -331,19 +445,90 @@ null_getattr(ap) } */ *ap; { int error; - error = null_bypass(ap); - if (error) + + if (error = null_bypass(ap)) return (error); /* Requires that arguments be restored. */ ap->a_vap->va_fsid = ap->a_vp->v_mount->mnt_stat.f_fsid.val[0]; return (0); } - static int +null_access(ap) + struct vop_access_args /* { + struct vnode *a_vp; + int a_mode; + struct ucred *a_cred; + struct proc *a_p; + } */ *ap; +{ + struct vnode *vp = ap->a_vp; + mode_t mode = ap->a_mode; + + /* + * Disallow write attempts on read-only layers; + * unless the file is a socket, fifo, or a block or + * character device resident on the file system. + */ + if (mode & VWRITE) { + switch (vp->v_type) { + case VDIR: + case VLNK: + case VREG: + if (vp->v_mount->mnt_flag & MNT_RDONLY) + return (EROFS); + break; + } + } + return (null_bypass(ap)); +} + +/* + * We need to process our own vnode lock and then clear the + * interlock flag as it applies only to our vnode, not the + * vnodes below us on the stack. + */ +int +null_lock(ap) + struct vop_lock_args /* { + struct vnode *a_vp; + int a_flags; + struct proc *a_p; + } */ *ap; +{ + + vop_nolock(ap); + if ((ap->a_flags & LK_TYPE_MASK) == LK_DRAIN) + return (0); + ap->a_flags &= ~LK_INTERLOCK; + return (null_bypass(ap)); +} + +/* + * We need to process our own vnode unlock and then clear the + * interlock flag as it applies only to our vnode, not the + * vnodes below us on the stack. + */ +int +null_unlock(ap) + struct vop_unlock_args /* { + struct vnode *a_vp; + int a_flags; + struct proc *a_p; + } */ *ap; +{ + struct vnode *vp = ap->a_vp; + + vop_nounlock(ap); + ap->a_flags &= ~LK_INTERLOCK; + return (null_bypass(ap)); +} + +int null_inactive(ap) struct vop_inactive_args /* { struct vnode *a_vp; + struct proc *a_p; } */ *ap; { /* @@ -358,6 +543,7 @@ null_inactive(ap) * like they do in the name lookup cache code. * That's too much work for now. */ + VOP_UNLOCK(ap->a_vp, 0, ap->a_p); return (0); } @@ -365,6 +551,7 @@ static int null_reclaim(ap) struct vop_reclaim_args /* { struct vnode *a_vp; + struct proc *a_p; } */ *ap; { struct vnode *vp = ap->a_vp; @@ -377,14 +564,13 @@ null_reclaim(ap) */ /* After this assignment, this node will not be re-used. */ xp->null_lowervp = NULL; - remque(xp); + LIST_REMOVE(xp, null_hash); FREE(vp->v_data, M_TEMP); vp->v_data = NULL; vrele (lowervp); return (0); } - static int null_print(ap) struct vop_print_args /* { @@ -396,7 +582,6 @@ null_print(ap) return (0); } - /* * XXX - vop_strategy must be hand coded because it has no * vnode in its arguments. @@ -422,7 +607,6 @@ null_strategy(ap) return (error); } - /* * XXX - like vop_strategy, vop_bwrite must be hand coded because it has no * vnode in its arguments. @@ -455,7 +639,12 @@ vop_t **null_vnodeop_p; static struct vnodeopv_entry_desc null_vnodeop_entries[] = { { &vop_default_desc, (vop_t *)null_bypass }, + { &vop_lookup_desc, (vop_t *)null_lookup }, + { &vop_setattr_desc, (vop_t *)null_setattr }, { &vop_getattr_desc, (vop_t *)null_getattr }, + { &vop_access_desc, (vop_t *)null_access }, + { &vop_lock_desc, (vop_t *)null_lock }, + { &vop_unlock_desc, (vop_t *)null_unlock }, { &vop_inactive_desc, (vop_t *)null_inactive }, { &vop_reclaim_desc, (vop_t *)null_reclaim }, { &vop_print_desc, (vop_t *)null_print }, diff --git a/sys/miscfs/portal/portal_vfsops.c b/sys/miscfs/portal/portal_vfsops.c index a980f65..74e1542 100644 --- a/sys/miscfs/portal/portal_vfsops.c +++ b/sys/miscfs/portal/portal_vfsops.c @@ -1,5 +1,5 @@ /* - * Copyright (c) 1992, 1993 + * Copyright (c) 1992, 1993, 1995 * The Regents of the University of California. All rights reserved. * * This code is derived from software donated to Berkeley by @@ -33,7 +33,7 @@ * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * - * @(#)portal_vfsops.c 8.6 (Berkeley) 1/21/94 + * @(#)portal_vfsops.c 8.11 (Berkeley) 5/14/95 * * $FreeBSD$ */ @@ -62,28 +62,18 @@ #include <sys/un.h> #include <miscfs/portal/portal.h> -static int portal_init __P((void)); static int portal_mount __P((struct mount *mp, char *path, caddr_t data, struct nameidata *ndp, struct proc *p)); static int portal_start __P((struct mount *mp, int flags, struct proc *p)); static int portal_unmount __P((struct mount *mp, int mntflags, struct proc *p)); static int portal_root __P((struct mount *mp, struct vnode **vpp)); -static int portal_quotactl __P((struct mount *mp, int cmd, uid_t uid, - caddr_t arg, struct proc *p)); static int portal_statfs __P((struct mount *mp, struct statfs *sbp, struct proc *p)); -static int portal_sync __P((struct mount *mp, int waitfor, - struct ucred *cred, struct proc *p)); -static int portal_vget __P((struct mount *mp, ino_t ino, - struct vnode **vpp)); -static int portal_fhtovp __P((struct mount *mp, struct fid *fhp, - struct mbuf *nam, struct vnode **vpp, - int *exflagsp, struct ucred **credanonp)); -static int portal_vptofh __P((struct vnode *vp, struct fid *fhp)); static int -portal_init() +portal_init(vfsp) + struct vfsconf *vfsp; { return (0); @@ -150,7 +140,7 @@ portal_mount(mp, path, data, ndp, p) mp->mnt_flag |= MNT_LOCAL; mp->mnt_data = (qaddr_t) fmp; - getnewfsid(mp, MOUNT_PORTAL); + vfs_getnewfsid(mp); (void)copyinstr(path, mp->mnt_stat.f_mntonname, MNAMELEN - 1, &size); bzero(mp->mnt_stat.f_mntonname + size, MNAMELEN - size); @@ -186,12 +176,8 @@ portal_unmount(mp, mntflags, p) int error, flags = 0; - if (mntflags & MNT_FORCE) { - /* portal can never be rootfs so don't check for it */ - if (!doforce) - return (EINVAL); + if (mntflags & MNT_FORCE) flags |= FORCECLOSE; - } /* * Clear out buffer cache. I don't think we @@ -241,39 +227,26 @@ portal_root(mp, vpp) struct mount *mp; struct vnode **vpp; { + struct proc *p = curproc; /* XXX */ struct vnode *vp; - /* * Return locked reference to root. */ vp = VFSTOPORTAL(mp)->pm_root; VREF(vp); - VOP_LOCK(vp); + vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, p); *vpp = vp; return (0); } static int -portal_quotactl(mp, cmd, uid, arg, p) - struct mount *mp; - int cmd; - uid_t uid; - caddr_t arg; - struct proc *p; -{ - - return (EOPNOTSUPP); -} - -static int portal_statfs(mp, sbp, p) struct mount *mp; struct statfs *sbp; struct proc *p; { - sbp->f_type = MOUNT_PORTAL; sbp->f_flags = 0; sbp->f_bsize = DEV_BSIZE; sbp->f_iosize = DEV_BSIZE; @@ -283,6 +256,7 @@ portal_statfs(mp, sbp, p) sbp->f_files = 1; /* Allow for "." */ sbp->f_ffree = 0; /* See comments above */ if (sbp != &mp->mnt_stat) { + sbp->f_type = mp->mnt_vfc->vfc_typenum; bcopy(&mp->mnt_stat.f_fsid, &sbp->f_fsid, sizeof(sbp->f_fsid)); bcopy(mp->mnt_stat.f_mntonname, sbp->f_mntonname, MNAMELEN); bcopy(mp->mnt_stat.f_mntfromname, sbp->f_mntfromname, MNAMELEN); @@ -290,48 +264,17 @@ portal_statfs(mp, sbp, p) return (0); } -static int -portal_sync(mp, waitfor, cred, p) - struct mount *mp; - int waitfor; - struct ucred *cred; - struct proc *p; -{ - - return (0); -} - -static int -portal_vget(mp, ino, vpp) - struct mount *mp; - ino_t ino; - struct vnode **vpp; -{ - - return (EOPNOTSUPP); -} - -static int -portal_fhtovp(mp, fhp, nam, vpp, exflagsp, credanonp) - struct mount *mp; - struct fid *fhp; - struct mbuf *nam; - struct vnode **vpp; - int *exflagsp; - struct ucred **credanonp; -{ - - return (EOPNOTSUPP); -} - -static int -portal_vptofh(vp, fhp) - struct vnode *vp; - struct fid *fhp; -{ - - return (EOPNOTSUPP); -} +#define portal_fhtovp ((int (*) __P((struct mount *, struct fid *, \ + struct mbuf *, struct vnode **, int *, struct ucred **)))eopnotsupp) +#define portal_quotactl ((int (*) __P((struct mount *, int, uid_t, caddr_t, \ + struct proc *)))eopnotsupp) +#define portal_sync ((int (*) __P((struct mount *, int, struct ucred *, \ + struct proc *)))nullop) +#define portal_sysctl ((int (*) __P((int *, u_int, void *, size_t *, void *, \ + size_t, struct proc *)))eopnotsupp) +#define portal_vget ((int (*) __P((struct mount *, ino_t, struct vnode **))) \ + eopnotsupp) +#define portal_vptofh ((int (*) __P((struct vnode *, struct fid *)))eopnotsupp) static struct vfsops portal_vfsops = { portal_mount, diff --git a/sys/miscfs/portal/portal_vnops.c b/sys/miscfs/portal/portal_vnops.c index 8135a5e..19b439c 100644 --- a/sys/miscfs/portal/portal_vnops.c +++ b/sys/miscfs/portal/portal_vnops.c @@ -33,7 +33,7 @@ * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * - * @(#)portal_vnops.c 8.8 (Berkeley) 1/21/94 + * @(#)portal_vnops.c 8.14 (Berkeley) 5/21/95 * * $FreeBSD$ */ @@ -112,17 +112,25 @@ portal_lookup(ap) struct componentname * a_cnp; } */ *ap; { - char *pname = ap->a_cnp->cn_nameptr; + struct componentname *cnp = ap->a_cnp; + struct vnode **vpp = ap->a_vpp; + struct vnode *dvp = ap->a_dvp; + char *pname = cnp->cn_nameptr; struct portalnode *pt; int error; struct vnode *fvp = 0; char *path; int size; - if (ap->a_cnp->cn_namelen == 1 && *pname == '.') { - *ap->a_vpp = ap->a_dvp; - VREF(ap->a_dvp); - /*VOP_LOCK(ap->a_dvp);*/ + *vpp = NULLVP; + + if (cnp->cn_nameiop == DELETE || cnp->cn_nameiop == RENAME) + return (EROFS); + + if (cnp->cn_namelen == 1 && *pname == '.') { + *vpp = dvp; + VREF(dvp); + /*VOP_LOCK(dvp);*/ return (0); } @@ -134,7 +142,7 @@ portal_lookup(ap) MALLOC(pt, struct portalnode *, sizeof(struct portalnode), M_TEMP, M_WAITOK); - error = getnewvnode(VT_PORTAL, ap->a_dvp->v_mount, portal_vnodeop_p, &fvp); + error = getnewvnode(VT_PORTAL, dvp->v_mount, portal_vnodeop_p, &fvp); if (error) { FREE(pt, M_TEMP); goto bad; @@ -148,22 +156,20 @@ portal_lookup(ap) */ for (size = 0, path = pname; *path; path++) size++; - ap->a_cnp->cn_consume = size - ap->a_cnp->cn_namelen; + cnp->cn_consume = size - cnp->cn_namelen; pt->pt_arg = malloc(size+1, M_TEMP, M_WAITOK); pt->pt_size = size+1; bcopy(pname, pt->pt_arg, pt->pt_size); pt->pt_fileid = portal_fileid++; - *ap->a_vpp = fvp; + *vpp = fvp; /*VOP_LOCK(fvp);*/ return (0); bad:; - if (fvp) { + if (fvp) vrele(fvp); - } - *ap->a_vpp = NULL; return (error); } @@ -444,6 +450,7 @@ portal_getattr(ap) { struct vnode *vp = ap->a_vp; struct vattr *vap = ap->a_vap; + struct timeval tv; bzero(vap, sizeof(*vap)); vattr_null(vap); @@ -452,8 +459,8 @@ portal_getattr(ap) vap->va_fsid = vp->v_mount->mnt_stat.f_fsid.val[0]; vap->va_size = DEV_BSIZE; vap->va_blocksize = DEV_BSIZE; - microtime((struct timeval *)&vap->va_atime); - TIMEVAL_TO_TIMESPEC((struct timeval *)&vap->va_atime, (struct timespec *)&vap->va_atime); + microtime(&tv); + TIMEVAL_TO_TIMESPEC(&tv, &vap->va_atime); vap->va_mtime = vap->va_atime; vap->va_ctime = vap->va_ctime; vap->va_gen = 0; @@ -509,9 +516,19 @@ portal_readdir(ap) struct vnode *a_vp; struct uio *a_uio; struct ucred *a_cred; + int *a_eofflag; + u_long *a_cookies; + int a_ncookies; } */ *ap; { + /* + * We don't allow exporting portal mounts, and currently local + * requests do not need cookies. + */ + if (ap->a_ncookies) + panic("portal_readdir: not hungry"); + return (0); } @@ -519,9 +536,11 @@ static int portal_inactive(ap) struct vop_inactive_args /* { struct vnode *a_vp; + struct proc *a_p; } */ *ap; { + VOP_UNLOCK(ap->a_vp, 0, ap->a_p); return (0); } @@ -639,6 +658,7 @@ portal_badop() #define portal_ioctl ((int (*) __P((struct vop_ioctl_args *)))portal_enotsupp) #define portal_select ((int (*) __P((struct vop_select_args *)))portal_enotsupp) #define portal_mmap ((int (*) __P((struct vop_mmap_args *)))portal_enotsupp) +#define portal_revoke vop_revoke #define portal_fsync ((int (*) __P((struct vop_fsync_args *)))nullop) #define portal_seek ((int (*) __P((struct vop_seek_args *)))nullop) #define portal_remove ((int (*) __P((struct vop_remove_args *)))portal_enotsupp) @@ -651,12 +671,14 @@ portal_badop() #define portal_readlink \ ((int (*) __P((struct vop_readlink_args *)))portal_enotsupp) #define portal_abortop ((int (*) __P((struct vop_abortop_args *)))nullop) -#define portal_lock ((int (*) __P((struct vop_lock_args *)))nullop) -#define portal_unlock ((int (*) __P((struct vop_unlock_args *)))nullop) +#define portal_lock ((int (*) __P((struct vop_lock_args *)))vop_nolock) +#define portal_unlock ((int (*) __P((struct vop_unlock_args *)))vop_nounlock) #define portal_bmap ((int (*) __P((struct vop_bmap_args *)))portal_badop) #define portal_strategy \ ((int (*) __P((struct vop_strategy_args *)))portal_badop) -#define portal_islocked ((int (*) __P((struct vop_islocked_args *)))nullop) +#define portal_islocked \ + ((int (*) __P((struct vop_islocked_args *)))vop_noislocked) +#define fifo_islocked ((int(*) __P((struct vop_islocked_args *)))vop_noislocked) #define portal_advlock \ ((int (*) __P((struct vop_advlock_args *)))portal_enotsupp) #define portal_blkatoff \ @@ -687,6 +709,7 @@ static struct vnodeopv_entry_desc portal_vnodeop_entries[] = { { &vop_ioctl_desc, (vop_t *)portal_ioctl }, /* ioctl */ { &vop_select_desc, (vop_t *)portal_select }, /* select */ { &vop_mmap_desc, (vop_t *)portal_mmap }, /* mmap */ + { &vop_revoke_desc, (vop_t *)portal_revoke }, /* revoke */ { &vop_fsync_desc, (vop_t *)portal_fsync }, /* fsync */ { &vop_seek_desc, (vop_t *)portal_seek }, /* seek */ { &vop_remove_desc, (vop_t *)portal_remove }, /* remove */ diff --git a/sys/miscfs/procfs/procfs.h b/sys/miscfs/procfs/procfs.h index df01a39..ef0d7eb 100644 --- a/sys/miscfs/procfs/procfs.h +++ b/sys/miscfs/procfs/procfs.h @@ -34,8 +34,9 @@ * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * - * @(#)procfs.h 8.6 (Berkeley) 2/3/94 + * @(#)procfs.h 8.9 (Berkeley) 5/14/95 * + * From: * $FreeBSD$ */ @@ -44,6 +45,7 @@ */ typedef enum { Proot, /* the filesystem root */ + Pcurproc, /* symbolic link for curproc */ Pproc, /* a process-specific sub-directory */ Pfile, /* the executable file */ Pmem, /* the process's memory image */ @@ -97,9 +99,9 @@ struct pfsdent { }; #define UIO_MX sizeof(struct pfsdent) #define PROCFS_FILENO(pid, type) \ - (((type) == Proot) ? \ - 2 : \ - ((((pid)+1) << 3) + ((int) (type)))) + (((type) < Pproc) ? \ + ((type) + 2) : \ + ((((pid)+1) << 4) + ((int) (type)))) /* * Convert between pfsnode vnode @@ -113,33 +115,33 @@ struct vfs_namemap { int nm_val; }; -extern int vfs_getuserstr __P((struct uio *, char *, int *)); -extern vfs_namemap_t *vfs_findname __P((vfs_namemap_t *, char *, int)); +int vfs_getuserstr __P((struct uio *, char *, int *)); +vfs_namemap_t *vfs_findname __P((vfs_namemap_t *, char *, int)); /* <machine/reg.h> */ struct reg; struct fpreg; #define PFIND(pid) ((pid) ? pfind(pid) : &proc0) -extern int procfs_freevp __P((struct vnode *)); -extern int procfs_allocvp __P((struct mount *, struct vnode **, long, pfstype)); -extern struct vnode *procfs_findtextvp __P((struct proc *)); -extern int procfs_sstep __P((struct proc *)); -extern void procfs_fix_sstep __P((struct proc *)); -extern int procfs_read_regs __P((struct proc *, struct reg *)); -extern int procfs_write_regs __P((struct proc *, struct reg *)); -extern int procfs_read_fpregs __P((struct proc *, struct fpreg *)); -extern int procfs_write_fpregs __P((struct proc *, struct fpreg *)); -extern int procfs_donote __P((struct proc *, struct proc *, struct pfsnode *pfsp, struct uio *uio)); -extern int procfs_doregs __P((struct proc *, struct proc *, struct pfsnode *pfsp, struct uio *uio)); -extern int procfs_dofpregs __P((struct proc *, struct proc *, struct pfsnode *pfsp, struct uio *uio)); -extern int procfs_domem __P((struct proc *, struct proc *, struct pfsnode *pfsp, struct uio *uio)); -extern int procfs_doctl __P((struct proc *, struct proc *, struct pfsnode *pfsp, struct uio *uio)); -extern int procfs_dostatus __P((struct proc *, struct proc *, struct pfsnode *pfsp, struct uio *uio)); -extern int procfs_domap __P((struct proc *, struct proc *, struct pfsnode *pfsp, struct uio *uio)); -extern int procfs_dotype __P((struct proc *, struct proc *, struct pfsnode *pfsp, struct uio *uio)); +int procfs_freevp __P((struct vnode *)); +int procfs_allocvp __P((struct mount *, struct vnode **, long, pfstype)); +struct vnode *procfs_findtextvp __P((struct proc *)); +int procfs_sstep __P((struct proc *)); +void procfs_fix_sstep __P((struct proc *)); +int procfs_read_regs __P((struct proc *, struct reg *)); +int procfs_write_regs __P((struct proc *, struct reg *)); +int procfs_read_fpregs __P((struct proc *, struct fpreg *)); +int procfs_write_fpregs __P((struct proc *, struct fpreg *)); +int procfs_donote __P((struct proc *, struct proc *, struct pfsnode *pfsp, struct uio *uio)); +int procfs_doregs __P((struct proc *, struct proc *, struct pfsnode *pfsp, struct uio *uio)); +int procfs_dofpregs __P((struct proc *, struct proc *, struct pfsnode *pfsp, struct uio *uio)); +int procfs_domem __P((struct proc *, struct proc *, struct pfsnode *pfsp, struct uio *uio)); +int procfs_doctl __P((struct proc *, struct proc *, struct pfsnode *pfsp, struct uio *uio)); +int procfs_dostatus __P((struct proc *, struct proc *, struct pfsnode *pfsp, struct uio *uio)); +int procfs_domap __P((struct proc *, struct proc *, struct pfsnode *pfsp, struct uio *uio)); +int procfs_dotype __P((struct proc *, struct proc *, struct pfsnode *pfsp, struct uio *uio)); -/* check to see if the process has the "items" (regs/file) */ +/* functions to check whether or not files should be displayed */ int procfs_validfile __P((struct proc *)); int procfs_validfpregs __P((struct proc *)); int procfs_validregs __P((struct proc *)); diff --git a/sys/miscfs/procfs/procfs_ctl.c b/sys/miscfs/procfs/procfs_ctl.c index 4ddf7a7..68b93dd 100644 --- a/sys/miscfs/procfs/procfs_ctl.c +++ b/sys/miscfs/procfs/procfs_ctl.c @@ -34,8 +34,9 @@ * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * - * @(#)procfs_ctl.c 8.3 (Berkeley) 1/21/94 + * @(#)procfs_ctl.c 8.4 (Berkeley) 6/15/94 * + * From: * $FreeBSD$ */ @@ -51,13 +52,13 @@ #include <sys/resourcevar.h> #include <sys/signal.h> #include <sys/signalvar.h> - -#include <vm/vm.h> -#include <vm/vm_param.h> -#include <vm/vm_extern.h> - +#include <sys/ptrace.h> #include <miscfs/procfs/procfs.h> +#ifndef FIX_SSTEP +#define FIX_SSTEP(p) +#endif + /* * True iff process (p) is in trace wait state * relative to process (curp) @@ -67,13 +68,6 @@ (p)->p_pptr == (curp) && \ ((p)->p_flag & P_TRACED)) -#ifdef notdef -#define FIX_SSTEP(p) { \ - procfs_fix_sstep(p); \ - } \ -} -#endif - #define PROCFS_CTL_ATTACH 1 #define PROCFS_CTL_DETACH 2 #define PROCFS_CTL_STEP 3 @@ -220,8 +214,10 @@ procfs_control(curp, p, op) */ case PROCFS_CTL_STEP: PHOLD(p); - procfs_sstep(p); + error = procfs_sstep(p); PRELE(p); + if (error) + return (error); break; /* diff --git a/sys/miscfs/procfs/procfs_fpregs.c b/sys/miscfs/procfs/procfs_fpregs.c index a3cad5f..841cf76 100644 --- a/sys/miscfs/procfs/procfs_fpregs.c +++ b/sys/miscfs/procfs/procfs_fpregs.c @@ -34,8 +34,9 @@ * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * - * @(#)procfs_fpregs.c 8.1 (Berkeley) 1/27/94 + * @(#)procfs_fpregs.c 8.2 (Berkeley) 6/15/94 * + * From: * $FreeBSD$ */ diff --git a/sys/miscfs/procfs/procfs_map.c b/sys/miscfs/procfs/procfs_map.c index 27c44fa..40c40d8 100644 --- a/sys/miscfs/procfs/procfs_map.c +++ b/sys/miscfs/procfs/procfs_map.c @@ -59,7 +59,7 @@ #include <vm/vm_param.h> #include <vm/vm_prot.h> #include <vm/vm_inherit.h> -#include <vm/lock.h> +#include <sys/lock.h> #include <vm/pmap.h> #include <vm/vm_map.h> #include <vm/vm_page.h> diff --git a/sys/miscfs/procfs/procfs_mem.c b/sys/miscfs/procfs/procfs_mem.c index 4e8fac6..06364e5 100644 --- a/sys/miscfs/procfs/procfs_mem.c +++ b/sys/miscfs/procfs/procfs_mem.c @@ -35,7 +35,7 @@ * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * - * @(#)procfs_mem.c 8.4 (Berkeley) 1/21/94 + * @(#)procfs_mem.c 8.5 (Berkeley) 6/15/94 * * $FreeBSD$ */ @@ -55,7 +55,7 @@ #include <vm/vm.h> #include <vm/vm_param.h> #include <vm/vm_prot.h> -#include <vm/lock.h> +#include <sys/lock.h> #include <vm/pmap.h> #include <vm/vm_map.h> #include <vm/vm_kern.h> @@ -295,14 +295,11 @@ procfs_domem(curp, p, pfs, uio) struct pfsnode *pfs; struct uio *uio; { - int error; if (uio->uio_resid == 0) return (0); - error = procfs_rwmem(p, uio); - - return (error); + return (procfs_rwmem(p, uio)); } /* @@ -320,5 +317,6 @@ struct vnode * procfs_findtextvp(p) struct proc *p; { + return (p->p_textvp); } diff --git a/sys/miscfs/procfs/procfs_regs.c b/sys/miscfs/procfs/procfs_regs.c index 9282a3a..56ef233 100644 --- a/sys/miscfs/procfs/procfs_regs.c +++ b/sys/miscfs/procfs/procfs_regs.c @@ -34,8 +34,9 @@ * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * - * @(#)procfs_regs.c 8.3 (Berkeley) 1/27/94 + * @(#)procfs_regs.c 8.4 (Berkeley) 6/15/94 * + * From: * $FreeBSD$ */ diff --git a/sys/miscfs/procfs/procfs_status.c b/sys/miscfs/procfs/procfs_status.c index 432918d..76974b5 100644 --- a/sys/miscfs/procfs/procfs_status.c +++ b/sys/miscfs/procfs/procfs_status.c @@ -34,8 +34,9 @@ * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * - * @(#)procfs_status.c 8.3 (Berkeley) 2/17/94 + * @(#)procfs_status.c 8.4 (Berkeley) 6/15/94 * + * From: * $FreeBSD$ */ @@ -142,7 +143,7 @@ procfs_dostatus(curp, p, pfs, uio) xlen = ps - psbuf; xlen -= uio->uio_offset; ps = psbuf + uio->uio_offset; - xlen = min(xlen, uio->uio_resid); + xlen = imin(xlen, uio->uio_resid); if (xlen <= 0) error = 0; else diff --git a/sys/miscfs/procfs/procfs_subr.c b/sys/miscfs/procfs/procfs_subr.c index d9b3d2f..5ee78f0 100644 --- a/sys/miscfs/procfs/procfs_subr.c +++ b/sys/miscfs/procfs/procfs_subr.c @@ -34,7 +34,7 @@ * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * - * @(#)procfs_subr.c 8.4 (Berkeley) 1/27/94 + * @(#)procfs_subr.c 8.6 (Berkeley) 5/14/95 * * $FreeBSD$ */ @@ -84,18 +84,21 @@ procfs_allocvp(mp, vpp, pid, pfs_type) long pid; pfstype pfs_type; { - int error; + struct proc *p = curproc; /* XXX */ struct pfsnode *pfs; + struct vnode *vp; struct pfsnode **pp; + int error; loop: for (pfs = pfshead; pfs != 0; pfs = pfs->pfs_next) { + vp = PFSTOV(pfs); if (pfs->pfs_pid == pid && pfs->pfs_type == pfs_type && - PFSTOV(pfs)->v_mount == mp) { - if (vget(pfs->pfs_vnode, 0)) + vp->v_mount == mp) { + if (vget(vp, 0, p)) goto loop; - *vpp = pfs->pfs_vnode; + *vpp = vp; return (0); } } @@ -118,17 +121,18 @@ loop: */ MALLOC(pfs, struct pfsnode *, sizeof(struct pfsnode), M_TEMP, M_WAITOK); - error = getnewvnode(VT_PROCFS, mp, procfs_vnodeop_p, vpp); - if (error) { + if (error = getnewvnode(VT_PROCFS, mp, procfs_vnodeop_p, vpp)) { FREE(pfs, M_TEMP); goto out; } + vp = *vpp; + + vp->v_data = pfs; - (*vpp)->v_data = pfs; pfs->pfs_next = 0; pfs->pfs_pid = (pid_t) pid; pfs->pfs_type = pfs_type; - pfs->pfs_vnode = *vpp; + pfs->pfs_vnode = vp; pfs->pfs_flags = 0; pfs->pfs_lockowner = 0; pfs->pfs_fileno = PROCFS_FILENO(pid, pfs_type); @@ -138,33 +142,41 @@ loop: pfs->pfs_mode = (VREAD|VEXEC) | (VREAD|VEXEC) >> 3 | (VREAD|VEXEC) >> 6; + vp->v_type = VDIR; + vp->v_flag = VROOT; + break; + + case Pcurproc: /* /proc/curproc = lr--r--r-- */ + pfs->pfs_mode = (VREAD) | + (VREAD >> 3) | + (VREAD >> 6); + vp->v_type = VLNK; break; case Pproc: pfs->pfs_mode = (VREAD|VEXEC) | (VREAD|VEXEC) >> 3 | (VREAD|VEXEC) >> 6; + vp->v_type = VDIR; break; case Pfile: - pfs->pfs_mode = (VREAD|VWRITE); - break; - case Pmem: pfs->pfs_mode = (VREAD|VWRITE) | (VREAD) >> 3;; break; case Pregs: - pfs->pfs_mode = (VREAD|VWRITE); - break; - case Pfpregs: pfs->pfs_mode = (VREAD|VWRITE); + vp->v_type = VREG; break; case Pctl: + case Pnote: + case Pnotepg: pfs->pfs_mode = (VWRITE); + vp->v_type = VREG; break; case Ptype: @@ -173,14 +185,7 @@ loop: pfs->pfs_mode = (VREAD) | (VREAD >> 3) | (VREAD >> 6); - break; - - case Pnote: - pfs->pfs_mode = (VWRITE); - break; - - case Pnotepg: - pfs->pfs_mode = (VWRITE); + vp->v_type = VREG; break; default: @@ -316,8 +321,7 @@ vfs_getuserstr(uio, buf, buflenp) return (EMSGSIZE); xlen = uio->uio_resid; - error = uiomove(buf, xlen, uio); - if (error) + if (error = uiomove(buf, xlen, uio)) return (error); /* allow multiple writes without seeks */ @@ -339,6 +343,7 @@ vfs_findname(nm, buf, buflen) char *buf; int buflen; { + for (; nm->nm_name; nm++) if (bcmp(buf, nm->nm_name, buflen+1) == 0) return (nm); diff --git a/sys/miscfs/procfs/procfs_vfsops.c b/sys/miscfs/procfs/procfs_vfsops.c index 433b391..5387126 100644 --- a/sys/miscfs/procfs/procfs_vfsops.c +++ b/sys/miscfs/procfs/procfs_vfsops.c @@ -34,7 +34,7 @@ * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * - * @(#)procfs_vfsops.c 8.4 (Berkeley) 1/21/94 + * @(#)procfs_vfsops.c 8.7 (Berkeley) 5/10/95 * * $FreeBSD$ */ @@ -56,24 +56,14 @@ #include <miscfs/procfs/procfs.h> #include <vm/vm.h> /* for PAGE_SIZE */ -static int procfs_fhtovp __P((struct mount *mp, struct fid *fhp, - struct mbuf *nam, struct vnode **vpp, - int *exflagsp, struct ucred **credanonp)); -static int procfs_init __P((void)); +static int procfs_init __P((struct vfsconf *vfsp)); static int procfs_mount __P((struct mount *mp, char *path, caddr_t data, struct nameidata *ndp, struct proc *p)); -static int procfs_quotactl __P((struct mount *mp, int cmds, uid_t uid, - caddr_t arg, struct proc *p)); static int procfs_start __P((struct mount *mp, int flags, struct proc *p)); static int procfs_statfs __P((struct mount *mp, struct statfs *sbp, struct proc *p)); -static int procfs_sync __P((struct mount *mp, int waitfor, - struct ucred *cred, struct proc *p)); static int procfs_unmount __P((struct mount *mp, int mntflags, struct proc *p)); -static int procfs_vget __P((struct mount *mp, ino_t ino, - struct vnode **vpp)); -static int procfs_vptofh __P((struct vnode *vp, struct fid *fhp)); /* * VFS Operations. @@ -101,7 +91,7 @@ procfs_mount(mp, path, data, ndp, p) mp->mnt_flag |= MNT_LOCAL; mp->mnt_data = 0; - getnewfsid(mp, MOUNT_PROCFS); + vfs_getnewfsid(mp); (void) copyinstr(path, (caddr_t)mp->mnt_stat.f_mntonname, MNAMELEN, &size); bzero(mp->mnt_stat.f_mntonname + size, MNAMELEN - size); @@ -125,12 +115,8 @@ procfs_unmount(mp, mntflags, p) int error; int flags = 0; - if (mntflags & MNT_FORCE) { - /* procfs can never be rootfs so don't check for it */ - if (!doforce) - return (EINVAL); + if (mntflags & MNT_FORCE) flags |= FORCECLOSE; - } error = vflush(mp, 0, flags); if (error) @@ -144,24 +130,10 @@ procfs_root(mp, vpp) struct mount *mp; struct vnode **vpp; { - struct pfsnode *pfs; - struct vnode *vp; - int error; - - error = procfs_allocvp(mp, &vp, (pid_t) 0, Proot); - if (error) - return (error); - - vp->v_type = VDIR; - vp->v_flag = VROOT; - pfs = VTOPFS(vp); - *vpp = vp; - return (0); + return (procfs_allocvp(mp, vpp, 0, Proot)); } -/* - */ /* ARGSUSED */ static int procfs_start(mp, flags, p) @@ -182,7 +154,6 @@ procfs_statfs(mp, sbp, p) struct statfs *sbp; struct proc *p; { - sbp->f_type = MOUNT_PROCFS; sbp->f_bsize = PAGE_SIZE; sbp->f_iosize = PAGE_SIZE; sbp->f_blocks = 1; /* avoid divide by zero in some df's */ @@ -192,6 +163,7 @@ procfs_statfs(mp, sbp, p) sbp->f_ffree = maxproc - nprocs; /* approx */ if (sbp != &mp->mnt_stat) { + sbp->f_type = mp->mnt_vfc->vfc_typenum; bcopy(&mp->mnt_stat.f_fsid, &sbp->f_fsid, sizeof(sbp->f_fsid)); bcopy(mp->mnt_stat.f_mntonname, sbp->f_mntonname, MNAMELEN); bcopy(mp->mnt_stat.f_mntfromname, sbp->f_mntfromname, MNAMELEN); @@ -200,68 +172,25 @@ procfs_statfs(mp, sbp, p) return (0); } - -static int -procfs_quotactl(mp, cmds, uid, arg, p) - struct mount *mp; - int cmds; - uid_t uid; - caddr_t arg; - struct proc *p; -{ - - return (EOPNOTSUPP); -} - static int -procfs_sync(mp, waitfor, cred, p) - struct mount *mp; - int waitfor; - struct ucred *cred; - struct proc *p; +procfs_init(vfsp) + struct vfsconf *vfsp; { return (0); } -static int -procfs_vget(mp, ino, vpp) - struct mount *mp; - ino_t ino; - struct vnode **vpp; -{ - - return (EOPNOTSUPP); -} - -static int -procfs_fhtovp(mp, fhp, nam, vpp, exflagsp, credanonp) - struct mount *mp; - struct fid *fhp; - struct mbuf *nam; - struct vnode **vpp; - int *exflagsp; - struct ucred **credanonp; -{ - - return (EINVAL); -} - -static int -procfs_vptofh(vp, fhp) - struct vnode *vp; - struct fid *fhp; -{ - - return EINVAL; -} - -static int -procfs_init() -{ - - return (0); -} +#define procfs_fhtovp ((int (*) __P((struct mount *, struct fid *, \ + struct mbuf *, struct vnode **, int *, struct ucred **)))einval) +#define procfs_quotactl ((int (*) __P((struct mount *, int, uid_t, caddr_t, \ + struct proc *)))eopnotsupp) +#define procfs_sync ((int (*) __P((struct mount *, int, struct ucred *, \ + struct proc *)))nullop) +#define procfs_sysctl ((int (*) __P((int *, u_int, void *, size_t *, void *, \ + size_t, struct proc *)))eopnotsupp) +#define procfs_vget ((int (*) __P((struct mount *, ino_t, struct vnode **))) \ + eopnotsupp) +#define procfs_vptofh ((int (*) __P((struct vnode *, struct fid *)))einval) static struct vfsops procfs_vfsops = { procfs_mount, diff --git a/sys/miscfs/procfs/procfs_vnops.c b/sys/miscfs/procfs/procfs_vnops.c index f497563..85be3ea 100644 --- a/sys/miscfs/procfs/procfs_vnops.c +++ b/sys/miscfs/procfs/procfs_vnops.c @@ -1,6 +1,6 @@ /* - * Copyright (c) 1993 Jan-Simon Pendry - * Copyright (c) 1993 + * Copyright (c) 1993, 1995 Jan-Simon Pendry + * Copyright (c) 1993, 1995 * The Regents of the University of California. All rights reserved. * * This code is derived from software contributed to Berkeley by @@ -34,7 +34,7 @@ * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * - * @(#)procfs_vnops.c 8.6 (Berkeley) 2/7/94 + * @(#)procfs_vnops.c 8.18 (Berkeley) 5/21/95 * * $FreeBSD$ */ @@ -55,8 +55,9 @@ #include <sys/malloc.h> #include <sys/dirent.h> #include <sys/resourcevar.h> -#include <miscfs/procfs/procfs.h> #include <vm/vm.h> /* for PAGE_SIZE */ +#include <machine/reg.h> +#include <miscfs/procfs/procfs.h> static int procfs_abortop __P((struct vop_abortop_args *)); static int procfs_access __P((struct vop_access_args *)); @@ -78,29 +79,30 @@ static int procfs_setattr __P((struct vop_setattr_args *)); * process-specific sub-directories. It is * used in procfs_lookup and procfs_readdir */ -static struct pfsnames { - u_short d_namlen; - char d_name[PROCFS_NAMELEN]; - pfstype d_pfstype; - int (*d_valid) __P((struct proc *)); -} procent[] = { +struct proc_target { + u_char pt_type; + u_char pt_namlen; + char *pt_name; + pfstype pt_pfstype; + int (*pt_valid) __P((struct proc *p)); +} proc_targets[] = { #define N(s) sizeof(s)-1, s - /* namlen, nam, type validp */ - { N("."), Pproc, NULL }, - { N(".."), Proot, NULL }, - { N("file"), Pfile, procfs_validfile }, - { N("mem"), Pmem, NULL }, - { N("regs"), Pregs, procfs_validregs }, - { N("fpregs"), Pfpregs, procfs_validfpregs }, - { N("ctl"), Pctl, NULL }, - { N("status"), Pstatus, NULL }, - { N("note"), Pnote, NULL }, - { N("notepg"), Pnotepg, NULL }, - { N("map"), Pmap, procfs_validmap }, - { N("etype"), Ptype, procfs_validtype }, + /* name type validp */ + { DT_DIR, N("."), Pproc, NULL }, + { DT_DIR, N(".."), Proot, NULL }, + { DT_REG, N("file"), Pfile, procfs_validfile }, + { DT_REG, N("mem"), Pmem, NULL }, + { DT_REG, N("regs"), Pregs, procfs_validregs }, + { DT_REG, N("fpregs"), Pfpregs, procfs_validfpregs }, + { DT_REG, N("ctl"), Pctl, NULL }, + { DT_REG, N("status"), Pstatus, NULL }, + { DT_REG, N("note"), Pnote, NULL }, + { DT_REG, N("notepg"), Pnotepg, NULL }, + { DT_REG, N("map"), Pmap, procfs_validmap }, + { DT_REG, N("etype"), Ptype, procfs_validtype }, #undef N }; -#define Nprocent (sizeof(procent)/sizeof(procent[0])) +static const int nproc_targets = sizeof(proc_targets) / sizeof(proc_targets[0]); static pid_t atopid __P((const char *, u_int)); @@ -117,7 +119,12 @@ static pid_t atopid __P((const char *, u_int)); */ static int procfs_open(ap) - struct vop_open_args *ap; + struct vop_open_args /* { + struct vnode *a_vp; + int a_mode; + struct ucred *a_cred; + struct proc *a_p; + } */ *ap; { struct pfsnode *pfs = VTOPFS(ap->a_vp); @@ -126,11 +133,10 @@ procfs_open(ap) if (PFIND(pfs->pfs_pid) == 0) return (ENOENT); /* was ESRCH, jsp */ - if (((pfs->pfs_flags & FWRITE) && (ap->a_mode & O_EXCL)) || - ((pfs->pfs_flags & O_EXCL) && (ap->a_mode & FWRITE))) + if ((pfs->pfs_flags & FWRITE) && (ap->a_mode & O_EXCL) || + (pfs->pfs_flags & O_EXCL) && (ap->a_mode & FWRITE)) return (EBUSY); - if (ap->a_mode & FWRITE) pfs->pfs_flags = ap->a_mode & (FWRITE|O_EXCL); @@ -152,7 +158,12 @@ procfs_open(ap) */ static int procfs_close(ap) - struct vop_close_args *ap; + struct vop_close_args /* { + struct vnode *a_vp; + int a_fflag; + struct ucred *a_cred; + struct proc *a_p; + } */ *ap; { struct pfsnode *pfs = VTOPFS(ap->a_vp); @@ -174,14 +185,51 @@ procfs_close(ap) */ static int procfs_ioctl(ap) - struct vop_ioctl_args *ap; + struct vop_ioctl_args /* { + struct vnode *a_vp; + int a_command; + caddr_t a_data; + int a_fflag; + struct ucred *a_cred; + struct proc *a_p; + } */ *ap; { return (ENOTTY); } /* - * _inactive is called when the pfsnode + * do block mapping for pfsnode (vp). + * since we don't use the buffer cache + * for procfs this function should never + * be called. in any case, it's not clear + * what part of the kernel ever makes use + * of this function. for sanity, this is the + * usual no-op bmap, although returning + * (EIO) would be a reasonable alternative. + */ +int +procfs_bmap(ap) + struct vop_bmap_args /* { + struct vnode *a_vp; + daddr_t a_bn; + struct vnode **a_vpp; + daddr_t *a_bnp; + int *a_runp; + } */ *ap; +{ + + if (ap->a_vpp != NULL) + *ap->a_vpp = ap->a_vp; + if (ap->a_bnp != NULL) + *ap->a_bnp = ap->a_bn; + if (ap->a_runp != NULL) + *ap->a_runp = 0; + return (0); +} + +/* + * procfs_inactive is called when the pfsnode * is vrele'd and the reference count goes * to zero. (vp) will be on the vnode free * list, so to get it back vget() must be @@ -194,16 +242,20 @@ procfs_ioctl(ap) * chances are that the process will still be * there and PFIND is not free. * - * (vp) is not locked on entry or exit. + * (vp) is locked on entry, but must be unlocked on exit. */ static int procfs_inactive(ap) - struct vop_inactive_args *ap; + struct vop_inactive_args /* { + struct vnode *a_vp; + } */ *ap; { - struct pfsnode *pfs = VTOPFS(ap->a_vp); + struct vnode *vp = ap->a_vp; + struct pfsnode *pfs = VTOPFS(vp); + VOP_UNLOCK(vp, 0, ap->a_p); if (PFIND(pfs->pfs_pid) == 0) - vgone(ap->a_vp); + vgone(vp); return (0); } @@ -217,12 +269,12 @@ procfs_inactive(ap) */ static int procfs_reclaim(ap) - struct vop_reclaim_args *ap; + struct vop_reclaim_args /* { + struct vnode *a_vp; + } */ *ap; { - int error; - error = procfs_freevp(ap->a_vp); - return (error); + return (procfs_freevp(ap->a_vp)); } /* @@ -269,13 +321,14 @@ procfs_pathconf(ap) */ static int procfs_print(ap) - struct vop_print_args *ap; + struct vop_print_args /* { + struct vnode *a_vp; + } */ *ap; { struct pfsnode *pfs = VTOPFS(ap->a_vp); - printf("tag VT_PROCFS, pid %lu, mode %x, flags %lx\n", - pfs->pfs_pid, - pfs->pfs_mode, pfs->pfs_flags); + printf("tag VT_PROCFS, type %s, pid %d, mode %x, flags %x\n", + pfs->pfs_type, pfs->pfs_pid, pfs->pfs_mode, pfs->pfs_flags); return (0); } @@ -287,7 +340,10 @@ procfs_print(ap) */ static int procfs_abortop(ap) - struct vop_abortop_args *ap; + struct vop_abortop_args /* { + struct vnode *a_dvp; + struct componentname *a_cnp; + } */ *ap; { if ((ap->a_cnp->cn_flags & (HASBUF | SAVESTART)) == HASBUF) @@ -316,7 +372,12 @@ procfs_badop() */ static int procfs_getattr(ap) - struct vop_getattr_args *ap; + struct vop_getattr_args /* { + struct vnode *a_vp; + struct vattr *a_vap; + struct ucred *a_cred; + struct proc *a_p; + } */ *ap; { struct pfsnode *pfs = VTOPFS(ap->a_vp); struct vattr *vap = ap->a_vap; @@ -329,6 +390,7 @@ procfs_getattr(ap) */ switch (pfs->pfs_type) { case Proot: + case Pcurproc: procp = 0; break; @@ -353,6 +415,21 @@ procfs_getattr(ap) vap->va_bytes = vap->va_size = 0; /* + * Make all times be current TOD. + * It would be possible to get the process start + * time from the p_stat structure, but there's + * no "file creation" time stamp anyway, and the + * p_stat structure is not addressible if u. gets + * swapped out for that process. + */ + { + struct timeval tv; + microtime(&tv); + TIMEVAL_TO_TIMESPEC(&tv, &vap->va_ctime); + } + vap->va_atime = vap->va_mtime = vap->va_ctime; + + /* * If the process has exercised some setuid or setgid * privilege, then rip away read/write permission so * that only root can gain access. @@ -376,21 +453,6 @@ procfs_getattr(ap) } /* - * Make all times be current TOD. - * It would be possible to get the process start - * time from the p_stat structure, but there's - * no "file creation" time stamp anyway, and the - * p_stat structure is not addressible if u. gets - * swapped out for that process. - */ - { - struct timeval tv; - microtime(&tv); - TIMEVAL_TO_TIMESPEC(&tv, &vap->va_ctime); - } - vap->va_atime = vap->va_mtime = vap->va_ctime; - - /* * now do the object specific fields * * The size could be set from struct reg, but it's hardly @@ -402,17 +464,30 @@ procfs_getattr(ap) switch (pfs->pfs_type) { case Proot: - vap->va_nlink = nprocs + 3; + /* + * Set nlink to 1 to tell fts(3) we don't actually know. + */ + vap->va_nlink = 1; + vap->va_uid = 0; + vap->va_gid = 0; + vap->va_size = vap->va_bytes = DEV_BSIZE; + break; + + case Pcurproc: { + char buf[16]; /* should be enough */ + vap->va_nlink = 1; vap->va_uid = 0; vap->va_gid = 0; - vap->va_bytes = vap->va_size = DEV_BSIZE; + vap->va_size = vap->va_bytes = + sprintf(buf, "%ld", (long)curproc->p_pid); break; + } case Pproc: - vap->va_nlink = Nprocent; + vap->va_nlink = nproc_targets; vap->va_uid = procp->p_ucred->cr_uid; vap->va_gid = procp->p_ucred->cr_gid; - vap->va_bytes = vap->va_size = DEV_BSIZE; + vap->va_size = vap->va_bytes = DEV_BSIZE; break; case Pfile: @@ -436,7 +511,15 @@ procfs_getattr(ap) case Ptype: case Pmap: case Pregs: + vap->va_bytes = vap->va_size = sizeof(struct reg); + vap->va_nlink = 1; + vap->va_uid = procp->p_ucred->cr_uid; + vap->va_gid = procp->p_ucred->cr_gid; + break; + case Pfpregs: + vap->va_bytes = vap->va_size = sizeof(struct fpreg); + case Pctl: case Pstatus: case Pnote: @@ -455,7 +538,12 @@ procfs_getattr(ap) static int procfs_setattr(ap) - struct vop_setattr_args *ap; + struct vop_setattr_args /* { + struct vnode *a_vp; + struct vattr *a_vap; + struct ucred *a_cred; + struct proc *a_p; + } */ *ap; { /* * just fake out attribute setting @@ -484,7 +572,12 @@ procfs_setattr(ap) */ static int procfs_access(ap) - struct vop_access_args *ap; + struct vop_access_args /* { + struct vnode *a_vp; + int a_mode; + struct ucred *a_cred; + struct proc *a_p; + } */ *ap; { struct vattr *vap; struct vattr vattr; @@ -494,8 +587,9 @@ procfs_access(ap) * If you're the super-user, * you always get access. */ - if (ap->a_cred->cr_uid == (uid_t) 0) + if (ap->a_cred->cr_uid == 0) return (0); + vap = &vattr; error = VOP_GETATTR(ap->a_vp, vap, ap->a_cred, ap->a_p); if (error) @@ -510,7 +604,7 @@ procfs_access(ap) gid_t *gp; int i; - (ap->a_mode) >>= 3; + ap->a_mode >>= 3; gp = ap->a_cred->cr_groups; for (i = 0; i < ap->a_cred->cr_ngroups; i++, gp++) if (vap->va_gid == *gp) @@ -537,18 +631,23 @@ found: */ static int procfs_lookup(ap) - struct vop_lookup_args *ap; + struct vop_lookup_args /* { + struct vnode * a_dvp; + struct vnode ** a_vpp; + struct componentname * a_cnp; + } */ *ap; { struct componentname *cnp = ap->a_cnp; struct vnode **vpp = ap->a_vpp; struct vnode *dvp = ap->a_dvp; char *pname = cnp->cn_nameptr; + struct proc *curp = cnp->cn_proc; int error = 0; + struct proc_target *pt; + struct vnode *fvp; pid_t pid; - struct vnode *nvp; struct pfsnode *pfs; - struct proc *procp; - pfstype pfs_type; + struct proc *p; int i; *vpp = NULL; @@ -559,7 +658,7 @@ procfs_lookup(ap) if (cnp->cn_namelen == 1 && *pname == '.') { *vpp = dvp; VREF(dvp); - /*VOP_LOCK(dvp);*/ + /* vn_lock(dvp, LK_EXCLUSIVE | LK_RETRY, curp); */ return (0); } @@ -570,72 +669,52 @@ procfs_lookup(ap) return (EIO); if (CNEQ(cnp, "curproc", 7)) - pid = cnp->cn_proc->p_pid; - else - pid = atopid(pname, cnp->cn_namelen); - if (pid == NO_PID) - return (ENOENT); + return (procfs_allocvp(dvp->v_mount, vpp, 0, Pcurproc)); - procp = PFIND(pid); - if (procp == 0) - return (ENOENT); - - error = procfs_allocvp(dvp->v_mount, &nvp, pid, Pproc); - if (error) - return (error); + pid = atopid(pname, cnp->cn_namelen); + if (pid == NO_PID) + break; - nvp->v_type = VDIR; - pfs = VTOPFS(nvp); + p = PFIND(pid); + if (p == 0) + break; - *vpp = nvp; - return (0); + return (procfs_allocvp(dvp->v_mount, vpp, pid, Pproc)); case Pproc: - if (cnp->cn_flags & ISDOTDOT) { - error = procfs_root(dvp->v_mount, vpp); - return (error); - } - - procp = PFIND(pfs->pfs_pid); - if (procp == 0) - return (ENOENT); + if (cnp->cn_flags & ISDOTDOT) + return (procfs_root(dvp->v_mount, vpp)); - for (i = 0; i < Nprocent; i++) { - struct pfsnames *dp = &procent[i]; + p = PFIND(pfs->pfs_pid); + if (p == 0) + break; - if (cnp->cn_namelen == dp->d_namlen && - bcmp(pname, dp->d_name, dp->d_namlen) == 0 && - (dp->d_valid == NULL || (*dp->d_valid)(procp))) { - pfs_type = dp->d_pfstype; + for (pt = proc_targets, i = 0; i < nproc_targets; pt++, i++) { + if (cnp->cn_namelen == pt->pt_namlen && + bcmp(pt->pt_name, pname, cnp->cn_namelen) == 0 && + (pt->pt_valid == NULL || (*pt->pt_valid)(p))) goto found; - } } - return (ENOENT); + break; found: - if (pfs_type == Pfile) { - nvp = procfs_findtextvp(procp); - if (nvp) { - VREF(nvp); - VOP_LOCK(nvp); - } else { - error = ENXIO; - } - } else { - error = procfs_allocvp(dvp->v_mount, &nvp, - pfs->pfs_pid, pfs_type); - if (error) - return (error); - - nvp->v_type = VREG; - pfs = VTOPFS(nvp); + if (pt->pt_pfstype == Pfile) { + fvp = procfs_findtextvp(p); + /* We already checked that it exists. */ + VREF(fvp); + vn_lock(fvp, LK_EXCLUSIVE | LK_RETRY, curp); + *vpp = fvp; + return (0); } - *vpp = nvp; - return (error); + + return (procfs_allocvp(dvp->v_mount, vpp, pfs->pfs_pid, + pt->pt_pfstype)); default: return (ENOTDIR); } + + return (cnp->cn_nameiop == LOOKUP ? ENOENT : EROFS); } /* @@ -645,6 +724,7 @@ int procfs_validfile(p) struct proc *p; { + return (procfs_findtextvp(p) != NULLVP); } @@ -662,7 +742,14 @@ procfs_validfile(p) */ static int procfs_readdir(ap) - struct vop_readdir_args *ap; + struct vop_readdir_args /* { + struct vnode *a_vp; + struct uio *a_uio; + struct ucred *a_cred; + int *a_eofflag; + u_long *a_cookies; + int a_ncookies; + } */ *ap; { struct uio *uio = ap->a_uio; struct pfsdent d; @@ -672,6 +759,13 @@ procfs_readdir(ap) int count; int i; + /* + * We don't allow exporting procfs mounts, and currently local + * requests do not need cookies. + */ + if (ap->a_ncookies) + panic("procfs_readdir: not hungry"); + pfs = VTOPFS(ap->a_vp); if (uio->uio_resid < UIO_MX) @@ -693,39 +787,28 @@ procfs_readdir(ap) */ case Pproc: { struct proc *p; + struct proc_target *pt; p = PFIND(pfs->pfs_pid); if (p == NULL) break; - while (uio->uio_resid >= UIO_MX) { - struct pfsnames *dt; - - if (i >= Nprocent) - break; - - dt = &procent[i]; - - /* see if we should show this one. */ - if (dt->d_valid && (*dt->d_valid)(p) == 0) { - i++; + for (pt = &proc_targets[i]; + uio->uio_resid >= UIO_MX && i < nproc_targets; pt++, i++) { + if (pt->pt_valid && (*pt->pt_valid)(p) == 0) continue; - } dp->d_reclen = UIO_MX; - dp->d_fileno = PROCFS_FILENO(pfs->pfs_pid, dt->d_pfstype); - dp->d_type = DT_REG; - dp->d_namlen = dt->d_namlen; - bcopy(dt->d_name, dp->d_name, sizeof(dt->d_name)-1); - error = uiomove((caddr_t) dp, UIO_MX, uio); - if (error) + dp->d_fileno = PROCFS_FILENO(pfs->pfs_pid, pt->pt_pfstype); + dp->d_namlen = pt->pt_namlen; + bcopy(pt->pt_name, dp->d_name, pt->pt_namlen + 1); + dp->d_type = pt->pt_type; + + if (error = uiomove((caddr_t)dp, UIO_MX, uio)) break; - count += UIO_MX; - i++; } break; - } /* @@ -738,63 +821,61 @@ procfs_readdir(ap) */ case Proot: { - int pcnt; #ifdef PROCFS_ZOMBIE int doingzomb = 0; #endif + int pcnt = 0; volatile struct proc *p = allproc.lh_first; -#define PROCFS_XFILES 3 /* number of other entries, like "curproc" */ - pcnt = PROCFS_XFILES; - - while (p && uio->uio_resid >= UIO_MX) { + again: + for (; p && uio->uio_resid >= UIO_MX; i++, pcnt++) { bzero((char *) dp, UIO_MX); - dp->d_type = DT_DIR; dp->d_reclen = UIO_MX; switch (i) { - case 0: + case 0: /* `.' */ + case 1: /* `..' */ dp->d_fileno = PROCFS_FILENO(0, Proot); - dp->d_namlen = sprintf(dp->d_name, "."); - break; - - case 1: - dp->d_fileno = PROCFS_FILENO(0, Proot); - dp->d_namlen = sprintf(dp->d_name, ".."); + dp->d_namlen = i + 1; + bcopy("..", dp->d_name, dp->d_namlen); + dp->d_name[i + 1] = '\0'; + dp->d_type = DT_DIR; break; case 2: - /* ship out entry for "curproc" */ - dp->d_fileno = PROCFS_FILENO(PID_MAX+1, Pproc); - dp->d_namlen = sprintf(dp->d_name, "curproc"); + dp->d_fileno = PROCFS_FILENO(0, Pcurproc); + dp->d_namlen = 7; + bcopy("curproc", dp->d_name, 8); + dp->d_type = DT_LNK; break; default: - if (pcnt >= i) { - dp->d_fileno = PROCFS_FILENO(p->p_pid, Pproc); - dp->d_namlen = sprintf(dp->d_name, "%ld", (long) p->p_pid); + while (pcnt < i) { + pcnt++; + p = p->p_list.le_next; + if (!p) + goto done; } - + dp->d_fileno = PROCFS_FILENO(p->p_pid, Pproc); + dp->d_namlen = sprintf(dp->d_name, "%ld", + (long)p->p_pid); + dp->d_type = DT_REG; p = p->p_list.le_next; - -#ifdef PROCFS_ZOMBIE - if (p == 0 && doingzomb == 0) { - doingzomb = 1; - p = zombproc.lh_first; - } -#endif - - if (pcnt++ < i) - continue; - break; } - error = uiomove((caddr_t) dp, UIO_MX, uio); - if (error) + + if (error = uiomove((caddr_t)dp, UIO_MX, uio)) break; - count += UIO_MX; - i++; } + done: + +#ifdef PROCFS_ZOMBIE + if (p == 0 && doingzomb == 0) { + doingzomb = 1; + p = zombproc.lh_first; + goto again; + } +#endif break; @@ -811,6 +892,25 @@ procfs_readdir(ap) } /* + * readlink reads the link of `curproc' + */ +int +procfs_readlink(ap) + struct vop_readlink_args *ap; +{ + struct uio *uio = ap->a_uio; + char buf[16]; /* should be enough */ + int len; + + if (VTOPFS(ap->a_vp)->pfs_fileno != PROCFS_FILENO(0, Pcurproc)) + return (EINVAL); + + len = sprintf(buf, "%ld", (long)curproc->p_pid); + + return (uiomove((caddr_t)buf, len, ap->a_uio)); +} + +/* * convert decimal ascii to pid_t */ static pid_t @@ -838,6 +938,7 @@ atopid(b, len) #define procfs_write procfs_rw #define procfs_select ((int (*) __P((struct vop_select_args *))) procfs_badop) #define procfs_mmap ((int (*) __P((struct vop_mmap_args *))) procfs_badop) +#define procfs_revoke vop_revoke #define procfs_fsync ((int (*) __P((struct vop_fsync_args *))) procfs_badop) #define procfs_seek ((int (*) __P((struct vop_seek_args *))) procfs_badop) #define procfs_remove ((int (*) __P((struct vop_remove_args *))) procfs_badop) @@ -849,7 +950,6 @@ atopid(b, len) #define procfs_readlink ((int (*) __P((struct vop_readlink_args *))) procfs_badop) #define procfs_lock ((int (*) __P((struct vop_lock_args *))) nullop) #define procfs_unlock ((int (*) __P((struct vop_unlock_args *))) nullop) -#define procfs_bmap ((int (*) __P((struct vop_bmap_args *))) procfs_badop) #define procfs_strategy ((int (*) __P((struct vop_strategy_args *))) procfs_badop) #define procfs_islocked ((int (*) __P((struct vop_islocked_args *))) nullop) #define procfs_advlock ((int (*) __P((struct vop_advlock_args *))) procfs_badop) @@ -859,6 +959,9 @@ atopid(b, len) #define procfs_truncate ((int (*) __P((struct vop_truncate_args *))) procfs_badop) #define procfs_update ((int (*) __P((struct vop_update_args *))) nullop) +/* + * procfs vnode operations. + */ vop_t **procfs_vnodeop_p; static struct vnodeopv_entry_desc procfs_vnodeop_entries[] = { { &vop_default_desc, (vop_t *)vn_default_error }, @@ -875,6 +978,7 @@ static struct vnodeopv_entry_desc procfs_vnodeop_entries[] = { { &vop_ioctl_desc, (vop_t *)procfs_ioctl }, /* ioctl */ { &vop_select_desc, (vop_t *)procfs_select }, /* select */ { &vop_mmap_desc, (vop_t *)procfs_mmap }, /* mmap */ + { &vop_revoke_desc, (vop_t *)procfs_revoke }, /* revoke */ { &vop_fsync_desc, (vop_t *)procfs_fsync }, /* fsync */ { &vop_seek_desc, (vop_t *)procfs_seek }, /* seek */ { &vop_remove_desc, (vop_t *)procfs_remove }, /* remove */ diff --git a/sys/miscfs/specfs/spec_vnops.c b/sys/miscfs/specfs/spec_vnops.c index 049f41e..2e8cc04 100644 --- a/sys/miscfs/specfs/spec_vnops.c +++ b/sys/miscfs/specfs/spec_vnops.c @@ -1,5 +1,5 @@ /* - * Copyright (c) 1989, 1993 + * Copyright (c) 1989, 1993, 1995 * The Regents of the University of California. All rights reserved. * * Redistribution and use in source and binary forms, with or without @@ -30,7 +30,7 @@ * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * - * @(#)spec_vnops.c 8.6 (Berkeley) 4/9/94 + * @(#)spec_vnops.c 8.14 (Berkeley) 5/21/95 * $FreeBSD$ */ @@ -79,8 +79,10 @@ static struct vnodeopv_entry_desc spec_vnodeop_entries[] = { { &vop_setattr_desc, (vop_t *)spec_setattr }, /* setattr */ { &vop_read_desc, (vop_t *)spec_read }, /* read */ { &vop_write_desc, (vop_t *)spec_write }, /* write */ + { &vop_lease_desc, (vop_t *)spec_lease_check }, /* lease */ { &vop_ioctl_desc, (vop_t *)spec_ioctl }, /* ioctl */ { &vop_select_desc, (vop_t *)spec_select }, /* select */ + { &vop_revoke_desc, (vop_t *)spec_revoke }, /* revoke */ { &vop_mmap_desc, (vop_t *)spec_mmap }, /* mmap */ { &vop_fsync_desc, (vop_t *)spec_fsync }, /* fsync */ { &vop_seek_desc, (vop_t *)spec_seek }, /* seek */ @@ -148,9 +150,10 @@ spec_open(ap) struct proc *a_p; } */ *ap; { + struct proc *p = ap->a_p; struct vnode *bvp, *vp = ap->a_vp; dev_t bdev, dev = (dev_t)vp->v_rdev; - register int maj = major(dev); + int maj = major(dev); int error; /* @@ -171,7 +174,9 @@ spec_open(ap) * When running in very secure mode, do not allow * opens for writing of any disk character devices. */ - if (securelevel >= 2 && isdisk(dev, VCHR)) + if (securelevel >= 2 + && cdevsw[maj]->d_bdev + && cdevsw[maj]->d_bdev->d_flags == D_DISK) return (EPERM); /* * When running in secure mode, do not allow opens @@ -189,9 +194,20 @@ spec_open(ap) return (EPERM); } } - VOP_UNLOCK(vp); - error = (*cdevsw[maj]->d_open)(dev, ap->a_mode, S_IFCHR, ap->a_p); - VOP_LOCK(vp); +#if 0 + /* + * Lite2 stuff. We will almost certainly do this + * differently with devfs. The only use of this flag + * is in dead_read to make ttys return EOF instead of + * EIO when they are dead. Pre-lite2 FreeBSD returns + * EOF for all character devices. + */ + if (cdevsw[maj]->d_type == D_TTY) + vp->v_flag |= VISTTY; +#endif + VOP_UNLOCK(vp, 0, p); + error = (*cdevsw[maj]->d_open)(dev, ap->a_mode, S_IFCHR, p); + vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, p); return (error); case VBLK: @@ -204,7 +220,7 @@ spec_open(ap) * opens for writing of any disk block devices. */ if (securelevel >= 2 && ap->a_cred != FSCRED && - (ap->a_mode & FWRITE) && isdisk(dev, VBLK)) + (ap->a_mode & FWRITE) && bdevsw[maj]->d_flags == D_DISK) return (EPERM); /* * Do not allow opens of block devices that are @@ -213,9 +229,7 @@ spec_open(ap) error = vfs_mountedon(vp); if (error) return (error); - return ((*bdevsw[maj]->d_open)(dev, ap->a_mode, S_IFBLK, ap->a_p)); - default: - break; + return ((*bdevsw[maj]->d_open)(dev, ap->a_mode, S_IFBLK, p)); } return (0); } @@ -257,10 +271,10 @@ spec_read(ap) switch (vp->v_type) { case VCHR: - VOP_UNLOCK(vp); + VOP_UNLOCK(vp, 0, p); error = (*cdevsw[major(vp->v_rdev)]->d_read) (vp->v_rdev, uio, ap->a_ioflag); - VOP_LOCK(vp); + vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, p); return (error); case VBLK: @@ -335,10 +349,10 @@ spec_write(ap) switch (vp->v_type) { case VCHR: - VOP_UNLOCK(vp); + VOP_UNLOCK(vp, 0, p); error = (*cdevsw[major(vp->v_rdev)]->d_write) (vp->v_rdev, uio, ap->a_ioflag); - VOP_LOCK(vp); + vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, p); return (error); case VBLK: @@ -408,7 +422,7 @@ spec_ioctl(ap) case VBLK: if (ap->a_command == 0 && (int)ap->a_data == B_TAPE) - if (bdevsw[major(dev)]->d_flags & B_TAPE) + if (bdevsw[major(dev)]->d_flags == D_TAPE) return (0); else return (1); @@ -498,6 +512,18 @@ loop: return (0); } +int +spec_inactive(ap) + struct vop_inactive_args /* { + struct vnode *a_vp; + struct proc *a_p; + } */ *ap; +{ + + VOP_UNLOCK(ap->a_vp, 0, ap->a_p); + return (0); +} + /* * Just call the device strategy routine */ @@ -539,31 +565,6 @@ spec_bmap(ap) } /* - * At the moment we do not do any locking. - */ -/* ARGSUSED */ -int -spec_lock(ap) - struct vop_lock_args /* { - struct vnode *a_vp; - } */ *ap; -{ - - return (0); -} - -/* ARGSUSED */ -int -spec_unlock(ap) - struct vop_unlock_args /* { - struct vnode *a_vp; - } */ *ap; -{ - - return (0); -} - -/* * Device close routine */ /* ARGSUSED */ @@ -577,6 +578,7 @@ spec_close(ap) } */ *ap; { register struct vnode *vp = ap->a_vp; + struct proc *p = ap->a_p; dev_t dev = vp->v_rdev; d_close_t *devclose; int mode, error; @@ -631,8 +633,11 @@ spec_close(ap) (vp->v_flag & VXLOCK) == 0) return (0); - if (vp->v_object) - vnode_pager_uncache(vp); + if (vp->v_object) { + vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, p); + vnode_pager_uncache(vp, p); + VOP_UNLOCK(vp, 0, p); + } devclose = bdevsw[major(dev)]->d_close; mode = S_IFBLK; diff --git a/sys/miscfs/specfs/specdev.h b/sys/miscfs/specfs/specdev.h index 8960040..091d628 100644 --- a/sys/miscfs/specfs/specdev.h +++ b/sys/miscfs/specfs/specdev.h @@ -30,7 +30,7 @@ * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * - * @(#)specdev.h 8.2 (Berkeley) 2/2/94 + * @(#)specdev.h 8.6 (Berkeley) 5/21/95 * $FreeBSD$ */ @@ -91,8 +91,10 @@ int spec_close __P((struct vop_close_args *)); #define spec_setattr ((int (*) __P((struct vop_setattr_args *)))spec_ebadf) int spec_read __P((struct vop_read_args *)); int spec_write __P((struct vop_write_args *)); +#define spec_lease_check ((int (*) __P((struct vop_lease_args *)))nullop) int spec_ioctl __P((struct vop_ioctl_args *)); int spec_select __P((struct vop_select_args *)); +#define spec_revoke vop_revoke #define spec_mmap ((int (*) __P((struct vop_mmap_args *)))spec_badop) int spec_fsync __P((struct vop_fsync_args *)); #define spec_seek ((int (*) __P((struct vop_seek_args *)))spec_badop) @@ -105,14 +107,14 @@ int spec_fsync __P((struct vop_fsync_args *)); #define spec_readdir ((int (*) __P((struct vop_readdir_args *)))spec_badop) #define spec_readlink ((int (*) __P((struct vop_readlink_args *)))spec_badop) #define spec_abortop ((int (*) __P((struct vop_abortop_args *)))spec_badop) -#define spec_inactive ((int (*) __P((struct vop_inactive_args *)))nullop) +int spec_inactive __P((struct vop_inactive_args *)); #define spec_reclaim ((int (*) __P((struct vop_reclaim_args *)))nullop) -int spec_lock __P((struct vop_lock_args *)); -int spec_unlock __P((struct vop_unlock_args *)); +#define spec_lock ((int (*) __P((struct vop_lock_args *)))vop_nolock) +#define spec_unlock ((int (*) __P((struct vop_unlock_args *)))vop_nounlock) int spec_bmap __P((struct vop_bmap_args *)); int spec_strategy __P((struct vop_strategy_args *)); int spec_print __P((struct vop_print_args *)); -#define spec_islocked ((int (*) __P((struct vop_islocked_args *)))nullop) +#define spec_islocked ((int(*) __P((struct vop_islocked_args *)))vop_noislocked) int spec_pathconf __P((struct vop_pathconf_args *)); int spec_advlock __P((struct vop_advlock_args *)); int spec_getpages __P((struct vop_getpages_args *)); diff --git a/sys/miscfs/umapfs/umap.h b/sys/miscfs/umapfs/umap.h index db1efd4..54ae097 100644 --- a/sys/miscfs/umapfs/umap.h +++ b/sys/miscfs/umapfs/umap.h @@ -33,7 +33,7 @@ * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * - * @(#)umap.h 8.3 (Berkeley) 1/21/94 + * @(#)umap.h 8.4 (Berkeley) 8/20/94 * * $FreeBSD$ */ @@ -67,8 +67,7 @@ struct umap_mount { * A cache of vnode references */ struct umap_node { - struct umap_node *umap_forw; /* Hash chain */ - struct umap_node *umap_back; + LIST_ENTRY(umap_node) umap_hash; /* Hash list */ struct vnode *umap_lowervp; /* Aliased vnode - VREFed once */ struct vnode *umap_vnode; /* Back pointer to vnode/umap_node */ }; diff --git a/sys/miscfs/umapfs/umap_subr.c b/sys/miscfs/umapfs/umap_subr.c index 8333943..3b3629d 100644 --- a/sys/miscfs/umapfs/umap_subr.c +++ b/sys/miscfs/umapfs/umap_subr.c @@ -1,5 +1,5 @@ /* - * Copyright (c) 1992, 1993 + * Copyright (c) 1992, 1993, 1995 * The Regents of the University of California. All rights reserved. * * This code is derived from software donated to Berkeley by @@ -33,13 +33,14 @@ * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * - * @(#)umap_subr.c 8.6 (Berkeley) 1/26/94 + * @(#)umap_subr.c 8.9 (Berkeley) 5/14/95 * * $FreeBSD$ */ #include <sys/param.h> #include <sys/systm.h> +#include <sys/proc.h> #include <sys/time.h> #include <sys/types.h> #include <sys/vnode.h> @@ -48,11 +49,8 @@ #include <sys/malloc.h> #include <miscfs/umapfs/umap.h> -extern int umapfs_init __P((void)); - #define LOG2_SIZEVNODE 7 /* log2(sizeof struct vnode) */ #define NUMAPNODECACHE 16 -#define UMAP_NHASH(vp) ((((u_long) vp)>>LOG2_SIZEVNODE) & (NUMAPNODECACHE-1)) /* * Null layer cache: @@ -62,52 +60,33 @@ extern int umapfs_init __P((void)); * alias is removed the target vnode is vrele'd. */ -/* - * Cache head - */ -struct umap_node_cache { - struct umap_node *ac_forw; - struct umap_node *ac_back; -}; - -static struct umap_node_cache umap_node_cache[NUMAPNODECACHE]; +#define UMAP_NHASH(vp) \ + (&umap_node_hashtbl[(((u_long)vp)>>LOG2_SIZEVNODE) & umap_node_hash]) +LIST_HEAD(umap_node_hashhead, umap_node) *umap_node_hashtbl; +u_long umap_node_hash; static u_long umap_findid __P((u_long id, u_long map[][2], int nentries)); static int umap_node_alloc __P((struct mount *mp, struct vnode *lowervp, struct vnode **vpp)); static struct vnode * umap_node_find __P((struct mount *mp, struct vnode *targetvp)); -static struct umap_node_cache * - umap_node_hash __P((struct vnode *targetvp)); /* * Initialise cache headers */ int -umapfs_init() +umapfs_init(vfsp) + struct vfsconf *vfsp; { - struct umap_node_cache *ac; + #ifdef UMAPFS_DIAGNOSTIC printf("umapfs_init\n"); /* printed during system boot */ #endif - - for (ac = umap_node_cache; ac < umap_node_cache + NUMAPNODECACHE; ac++) - ac->ac_forw = ac->ac_back = (struct umap_node *) ac; + umap_node_hashtbl = hashinit(NUMAPNODECACHE, M_CACHE, &umap_node_hash); return (0); } /* - * Compute hash list for given target vnode - */ -static struct umap_node_cache * -umap_node_hash(targetvp) - struct vnode *targetvp; -{ - - return (&umap_node_cache[UMAP_NHASH(targetvp)]); -} - -/* * umap_findid is called by various routines in umap_vnodeops.c to * find a user or group id in a map. */ @@ -163,7 +142,8 @@ umap_node_find(mp, targetvp) struct mount *mp; struct vnode *targetvp; { - struct umap_node_cache *hd; + struct proc *p = curproc; /* XXX */ + struct umap_node_hashhead *hd; struct umap_node *a; struct vnode *vp; @@ -177,10 +157,9 @@ umap_node_find(mp, targetvp) * the target vnode. If found, the increment the umap_node * reference count (but NOT the target vnode's VREF counter). */ - hd = umap_node_hash(targetvp); - - loop: - for (a = hd->ac_forw; a != (struct umap_node *) hd; a = a->umap_forw) { + hd = UMAP_NHASH(targetvp); +loop: + for (a = hd->lh_first; a != 0; a = a->umap_hash.le_next) { if (a->umap_lowervp == targetvp && a->umap_vnode->v_mount == mp) { vp = UMAPTOV(a); @@ -189,7 +168,7 @@ umap_node_find(mp, targetvp) * stuff, but we don't want to lock * the lower node. */ - if (vget(vp, 0)) { + if (vget(vp, 0, p)) { #ifdef UMAPFS_DIAGNOSTIC printf ("umap_node_find: vget failed.\n"); #endif @@ -217,7 +196,7 @@ umap_node_alloc(mp, lowervp, vpp) struct vnode *lowervp; struct vnode **vpp; { - struct umap_node_cache *hd; + struct umap_node_hashhead *hd; struct umap_node *xp; struct vnode *othervp, *vp; int error; @@ -257,8 +236,8 @@ umap_node_alloc(mp, lowervp, vpp) return (0); } VREF(lowervp); /* Extra VREF will be vrele'd in umap_node_create */ - hd = umap_node_hash(lowervp); - insque(xp, hd); + hd = UMAP_NHASH(lowervp); + LIST_INSERT_HEAD(hd, xp, umap_hash); return (0); } diff --git a/sys/miscfs/umapfs/umap_vfsops.c b/sys/miscfs/umapfs/umap_vfsops.c index 517c291..897e2ce 100644 --- a/sys/miscfs/umapfs/umap_vfsops.c +++ b/sys/miscfs/umapfs/umap_vfsops.c @@ -1,5 +1,5 @@ /* - * Copyright (c) 1992, 1993 + * Copyright (c) 1992, 1993, 1995 * The Regents of the University of California. All rights reserved. * * This code is derived from software donated to Berkeley by @@ -33,7 +33,7 @@ * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * - * @(#)umap_vfsops.c 8.3 (Berkeley) 1/21/94 + * @(#)umap_vfsops.c 8.8 (Berkeley) 5/14/95 * * $FreeBSD$ */ @@ -46,6 +46,7 @@ #include <sys/param.h> #include <sys/systm.h> #include <sys/kernel.h> +#include <sys/proc.h> #include <sys/time.h> #include <sys/types.h> #include <sys/vnode.h> @@ -54,7 +55,7 @@ #include <sys/malloc.h> #include <miscfs/umapfs/umap.h> -extern int umapfs_init __P((void)); +extern int umapfs_init __P((struct vfsconf *)); static int umapfs_fhtovp __P((struct mount *mp, struct fid *fidp, struct mbuf *nam, struct vnode **vpp, @@ -187,7 +188,7 @@ umapfs_mount(mp, path, data, ndp, p) /* * Unlock the node (either the lower or the alias) */ - VOP_UNLOCK(vp); + VOP_UNLOCK(vp, 0, p); /* * Make sure the node alias worked */ @@ -207,7 +208,7 @@ umapfs_mount(mp, path, data, ndp, p) if (UMAPVPTOLOWERVP(umapm_rootvp)->v_mount->mnt_flag & MNT_LOCAL) mp->mnt_flag |= MNT_LOCAL; mp->mnt_data = (qaddr_t) amp; - getnewfsid(mp, MOUNT_LOFS); + vfs_getnewfsid(mp); (void) copyinstr(path, mp->mnt_stat.f_mntonname, MNAMELEN - 1, &size); bzero(mp->mnt_stat.f_mntonname + size, MNAMELEN - size); @@ -253,12 +254,8 @@ umapfs_unmount(mp, mntflags, p) printf("umapfs_unmount(mp = %x)\n", mp); #endif - if (mntflags & MNT_FORCE) { - /* lofs can never be rootfs so don't check for it */ - if (!doforce) - return (EINVAL); + if (mntflags & MNT_FORCE) flags |= FORCECLOSE; - } /* * Clear out buffer cache. I don't think we @@ -300,6 +297,7 @@ umapfs_root(mp, vpp) struct mount *mp; struct vnode **vpp; { + struct proc *p = curproc; /* XXX */ struct vnode *vp; #ifdef UMAPFS_DIAGNOSTIC @@ -314,7 +312,7 @@ umapfs_root(mp, vpp) */ vp = MOUNTTOUMAPMOUNT(mp)->umapm_rootvp; VREF(vp); - VOP_LOCK(vp); + vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, p); *vpp = vp; return (0); } @@ -429,4 +427,3 @@ static struct vfsops umap_vfsops = { }; VFS_SET(umap_vfsops, umap, MOUNT_UMAP, VFCF_LOOPBACK); - diff --git a/sys/miscfs/umapfs/umap_vnops.c b/sys/miscfs/umapfs/umap_vnops.c index 7c14cda..17064f4 100644 --- a/sys/miscfs/umapfs/umap_vnops.c +++ b/sys/miscfs/umapfs/umap_vnops.c @@ -33,7 +33,7 @@ * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * - * @(#)umap_vnops.c 8.3 (Berkeley) 1/5/94 + * @(#)umap_vnops.c 8.6 (Berkeley) 5/22/95 * $FreeBSD$ */ @@ -336,10 +336,52 @@ umap_getattr(ap) return (0); } +/* + * We need to process our own vnode lock and then clear the + * interlock flag as it applies only to our vnode, not the + * vnodes below us on the stack. + */ +int +umap_lock(ap) + struct vop_lock_args /* { + struct vnode *a_vp; + int a_flags; + struct proc *a_p; + } */ *ap; +{ + + vop_nolock(ap); + if ((ap->a_flags & LK_TYPE_MASK) == LK_DRAIN) + return (0); + ap->a_flags &= ~LK_INTERLOCK; + return (null_bypass(ap)); +} + +/* + * We need to process our own vnode unlock and then clear the + * interlock flag as it applies only to our vnode, not the + * vnodes below us on the stack. + */ +int +umap_unlock(ap) + struct vop_unlock_args /* { + struct vnode *a_vp; + int a_flags; + struct proc *a_p; + } */ *ap; +{ + struct vnode *vp = ap->a_vp; + + vop_nounlock(ap); + ap->a_flags &= ~LK_INTERLOCK; + return (null_bypass(ap)); +} + static int umap_inactive(ap) struct vop_inactive_args /* { struct vnode *a_vp; + struct proc *a_p; } */ *ap; { /* @@ -349,6 +391,7 @@ umap_inactive(ap) * cache and reusable. * */ + VOP_UNLOCK(ap->a_vp, 0, ap->a_p); return (0); } @@ -364,7 +407,7 @@ umap_reclaim(ap) /* After this assignment, this node will not be re-used. */ xp->umap_lowervp = NULL; - remque(xp); + LIST_REMOVE(xp, umap_hash); FREE(vp->v_data, M_TEMP); vp->v_data = NULL; vrele(lowervp); @@ -487,6 +530,8 @@ static struct vnodeopv_entry_desc umap_vnodeop_entries[] = { { &vop_default_desc, (vop_t *)umap_bypass }, { &vop_getattr_desc, (vop_t *)umap_getattr }, + { &vop_lock_desc, (vop_t *)umap_lock }, + { &vop_unlock_desc, (vop_t *)umap_unlock }, { &vop_inactive_desc, (vop_t *)umap_inactive }, { &vop_reclaim_desc, (vop_t *)umap_reclaim }, { &vop_print_desc, (vop_t *)umap_print }, diff --git a/sys/miscfs/union/union.h b/sys/miscfs/union/union.h index 9c740d3..c956921 100644 --- a/sys/miscfs/union/union.h +++ b/sys/miscfs/union/union.h @@ -34,7 +34,7 @@ * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * - * @(#)union.h 8.2 (Berkeley) 2/17/94 + * @(#)union.h 8.9 (Berkeley) 12/10/94 * $FreeBSD$ */ @@ -75,10 +75,14 @@ struct union_node { struct vnode *un_uppervp; /* overlaying object */ struct vnode *un_lowervp; /* underlying object */ struct vnode *un_dirvp; /* Parent dir of uppervp */ + struct vnode *un_pvp; /* Parent vnode */ char *un_path; /* saved component name */ int un_hash; /* saved un_path hash value */ int un_openl; /* # of opens on lowervp */ - int un_flags; + unsigned int un_flags; + struct vnode **un_dircache; /* cached union stack */ + off_t un_uppersz; /* size of upper object */ + off_t un_lowersz; /* size of lower object */ #ifdef DIAGNOSTIC pid_t un_pid; #endif @@ -88,16 +92,23 @@ struct union_node { #define UN_LOCKED 0x02 #define UN_ULOCK 0x04 /* Upper node is locked */ #define UN_KLOCK 0x08 /* Keep upper node locked on vput */ +#define UN_CACHED 0x10 /* In union cache */ extern int union_allocvp __P((struct vnode **, struct mount *, struct vnode *, struct vnode *, struct componentname *, struct vnode *, - struct vnode *)); + struct vnode *, int)); extern int union_freevp __P((struct vnode *)); -extern int union_copyfile __P((struct proc *, struct ucred *, - struct vnode *, struct vnode *)); +extern int union_copyfile __P((struct vnode *, struct vnode *, + struct ucred *, struct proc *)); +extern int union_copyup __P((struct union_node *, int, struct ucred *, + struct proc *)); +extern int union_dowhiteout __P((struct union_node *, struct ucred *, + struct proc *)); extern int union_mkshadow __P((struct union_mount *, struct vnode *, struct componentname *, struct vnode **)); +extern int union_mkwhiteout __P((struct union_mount *, struct vnode *, + struct componentname *, char *)); extern int union_vn_create __P((struct vnode **, struct union_node *, struct proc *)); extern int union_vn_close __P((struct vnode *, int, struct ucred *, @@ -108,6 +119,7 @@ extern void union_removed_upper __P((struct union_node *un)); extern struct vnode *union_lowervp __P((struct vnode *)); extern void union_newlower __P((struct union_node *, struct vnode *)); extern void union_newupper __P((struct union_node *, struct vnode *)); +extern void union_newsize __P((struct vnode *, off_t, off_t)); #define MOUNTTOUNIONMOUNT(mp) ((struct union_mount *)((mp)->mnt_data)) #define VTOUNION(vp) ((struct union_node *)(vp)->v_data) diff --git a/sys/miscfs/union/union_subr.c b/sys/miscfs/union/union_subr.c index 6549f7d..63e25e6 100644 --- a/sys/miscfs/union/union_subr.c +++ b/sys/miscfs/union/union_subr.c @@ -34,7 +34,7 @@ * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * - * @(#)union_subr.c 8.4 (Berkeley) 2/17/94 + * @(#)union_subr.c 8.20 (Berkeley) 5/20/95 * $FreeBSD$ */ @@ -48,6 +48,9 @@ #include <sys/file.h> #include <sys/filedesc.h> #include <sys/queue.h> +#include <sys/mount.h> +#include <sys/stat.h> +#include <vm/vm.h> /* for vnode_pager_setsize */ #include <miscfs/union/union.h> #include <sys/proc.h> @@ -118,31 +121,38 @@ union_updatevp(un, uppervp, lowervp) { int ohash = UNION_HASH(un->un_uppervp, un->un_lowervp); int nhash = UNION_HASH(uppervp, lowervp); + int docache = (lowervp != NULLVP || uppervp != NULLVP); + int lhash, hhash, uhash; - if (ohash != nhash) { - /* - * Ensure locking is ordered from lower to higher - * to avoid deadlocks. - */ - if (nhash < ohash) { - int t = ohash; - ohash = nhash; - nhash = t; - } + /* + * Ensure locking is ordered from lower to higher + * to avoid deadlocks. + */ + if (nhash < ohash) { + lhash = nhash; + uhash = ohash; + } else { + lhash = ohash; + uhash = nhash; + } - while (union_list_lock(ohash)) + if (lhash != uhash) + while (union_list_lock(lhash)) continue; - while (union_list_lock(nhash)) - continue; + while (union_list_lock(uhash)) + continue; - LIST_REMOVE(un, un_cache); - union_list_unlock(ohash); - } else { - while (union_list_lock(nhash)) - continue; + if (ohash != nhash || !docache) { + if (un->un_flags & UN_CACHED) { + un->un_flags &= ~UN_CACHED; + LIST_REMOVE(un, un_cache); + } } + if (ohash != nhash) + union_list_unlock(ohash); + if (un->un_lowervp != lowervp) { if (un->un_lowervp) { vrele(un->un_lowervp); @@ -156,6 +166,7 @@ union_updatevp(un, uppervp, lowervp) } } un->un_lowervp = lowervp; + un->un_lowersz = VNOVAL; } if (un->un_uppervp != uppervp) { @@ -163,10 +174,13 @@ union_updatevp(un, uppervp, lowervp) vrele(un->un_uppervp); un->un_uppervp = uppervp; + un->un_uppersz = VNOVAL; } - if (ohash != nhash) + if (docache && (ohash != nhash)) { LIST_INSERT_HEAD(&unhead[nhash], un, un_cache); + un->un_flags |= UN_CACHED; + } union_list_unlock(nhash); } @@ -190,6 +204,47 @@ union_newupper(un, uppervp) } /* + * Keep track of size changes in the underlying vnodes. + * If the size changes, then callback to the vm layer + * giving priority to the upper layer size. + */ +void +union_newsize(vp, uppersz, lowersz) + struct vnode *vp; + off_t uppersz, lowersz; +{ + struct union_node *un; + off_t sz; + + /* only interested in regular files */ + if (vp->v_type != VREG) + return; + + un = VTOUNION(vp); + sz = VNOVAL; + + if ((uppersz != VNOVAL) && (un->un_uppersz != uppersz)) { + un->un_uppersz = uppersz; + if (sz == VNOVAL) + sz = un->un_uppersz; + } + + if ((lowersz != VNOVAL) && (un->un_lowersz != lowersz)) { + un->un_lowersz = lowersz; + if (sz == VNOVAL) + sz = un->un_lowersz; + } + + if (sz != VNOVAL) { +#ifdef UNION_DIAGNOSTIC + printf("union: %s size now %ld\n", + uppersz != VNOVAL ? "upper" : "lower", (long) sz); +#endif + vnode_pager_setsize(vp, sz); + } +} + +/* * allocate a union_node/vnode pair. the vnode is * referenced and locked. the new vnode is returned * via (vpp). (mp) is the mountpoint of the union filesystem, @@ -221,19 +276,22 @@ union_newupper(un, uppervp) * the vnode free list. */ int -union_allocvp(vpp, mp, undvp, dvp, cnp, uppervp, lowervp) +union_allocvp(vpp, mp, undvp, dvp, cnp, uppervp, lowervp, docache) struct vnode **vpp; struct mount *mp; - struct vnode *undvp; + struct vnode *undvp; /* parent union vnode */ struct vnode *dvp; /* may be null */ struct componentname *cnp; /* may be null */ struct vnode *uppervp; /* may be null */ struct vnode *lowervp; /* may be null */ + int docache; { int error; struct union_node *un = 0; struct vnode *xlowervp = NULLVP; - int hash = 0; + struct union_mount *um = MOUNTTOUNIONMOUNT(mp); + int hash; + int vflag; int try; if (uppervp == NULLVP && lowervp == NULLVP) @@ -244,8 +302,22 @@ union_allocvp(vpp, mp, undvp, dvp, cnp, uppervp, lowervp) lowervp = NULLVP; } + /* detect the root vnode (and aliases) */ + vflag = 0; + if ((uppervp == um->um_uppervp) && + ((lowervp == NULLVP) || lowervp == um->um_lowervp)) { + if (lowervp == NULLVP) { + lowervp = um->um_lowervp; + if (lowervp != NULLVP) + VREF(lowervp); + } + vflag = VROOT; + } + loop: - for (try = 0; try < 3; try++) { + if (!docache) { + un = 0; + } else for (try = 0; try < 3; try++) { switch (try) { case 0: if (lowervp == NULLVP) @@ -276,7 +348,8 @@ loop: (un->un_uppervp == uppervp || un->un_uppervp == NULLVP) && (UNIONTOV(un)->v_mount == mp)) { - if (vget(UNIONTOV(un), 0)) { + if (vget(UNIONTOV(un), 0, + cnp ? cnp->cn_proc : NULL)) { union_list_unlock(hash); goto loop; } @@ -359,8 +432,7 @@ loop: */ if (lowervp != un->un_lowervp) { union_newlower(un, lowervp); - if (cnp && (lowervp != NULLVP) && - (lowervp->v_type == VREG)) { + if (cnp && (lowervp != NULLVP)) { un->un_hash = cnp->cn_hash; un->un_path = malloc(cnp->cn_namelen+1, M_TEMP, M_WAITOK); @@ -377,14 +449,16 @@ loop: return (0); } - /* - * otherwise lock the vp list while we call getnewvnode - * since that can block. - */ - hash = UNION_HASH(uppervp, lowervp); + if (docache) { + /* + * otherwise lock the vp list while we call getnewvnode + * since that can block. + */ + hash = UNION_HASH(uppervp, lowervp); - if (union_list_lock(hash)) - goto loop; + if (union_list_lock(hash)) + goto loop; + } error = getnewvnode(VT_UNION, mp, union_vnodeop_p, vpp); if (error) { @@ -403,6 +477,7 @@ loop: MALLOC((*vpp)->v_data, void *, sizeof(struct union_node), M_TEMP, M_WAITOK); + (*vpp)->v_flag |= vflag; if (uppervp) (*vpp)->v_type = uppervp->v_type; else @@ -410,7 +485,13 @@ loop: un = VTOUNION(*vpp); un->un_vnode = *vpp; un->un_uppervp = uppervp; + un->un_uppersz = VNOVAL; un->un_lowervp = lowervp; + un->un_lowersz = VNOVAL; + un->un_pvp = undvp; + if (undvp != NULLVP) + VREF(undvp); + un->un_dircache = 0; un->un_openl = 0; un->un_flags = UN_LOCKED; if (un->un_uppervp) @@ -421,7 +502,7 @@ loop: else un->un_pid = -1; #endif - if (cnp && (lowervp != NULLVP) && (lowervp->v_type == VREG)) { + if (cnp && (lowervp != NULLVP)) { un->un_hash = cnp->cn_hash; un->un_path = malloc(cnp->cn_namelen+1, M_TEMP, M_WAITOK); bcopy(cnp->cn_nameptr, un->un_path, cnp->cn_namelen); @@ -434,13 +515,17 @@ loop: un->un_dirvp = 0; } - LIST_INSERT_HEAD(&unhead[hash], un, un_cache); + if (docache) { + LIST_INSERT_HEAD(&unhead[hash], un, un_cache); + un->un_flags |= UN_CACHED; + } if (xlowervp) vrele(xlowervp); out: - union_list_unlock(hash); + if (docache) + union_list_unlock(hash); return (error); } @@ -451,13 +536,18 @@ union_freevp(vp) { struct union_node *un = VTOUNION(vp); - LIST_REMOVE(un, un_cache); + if (un->un_flags & UN_CACHED) { + un->un_flags &= ~UN_CACHED; + LIST_REMOVE(un, un_cache); + } - if (un->un_uppervp) + if (un->un_pvp != NULLVP) + vrele(un->un_pvp); + if (un->un_uppervp != NULLVP) vrele(un->un_uppervp); - if (un->un_lowervp) + if (un->un_lowervp != NULLVP) vrele(un->un_lowervp); - if (un->un_dirvp) + if (un->un_dirvp != NULLVP) vrele(un->un_dirvp); if (un->un_path) free(un->un_path, M_TEMP); @@ -474,11 +564,11 @@ union_freevp(vp) * and (tvp) are locked on entry and exit. */ int -union_copyfile(p, cred, fvp, tvp) - struct proc *p; - struct ucred *cred; +union_copyfile(fvp, tvp, cred, p) struct vnode *fvp; struct vnode *tvp; + struct ucred *cred; + struct proc *p; { char *buf; struct uio uio; @@ -497,12 +587,12 @@ union_copyfile(p, cred, fvp, tvp) uio.uio_segflg = UIO_SYSSPACE; uio.uio_offset = 0; - VOP_UNLOCK(fvp); /* XXX */ - LEASE_CHECK(fvp, p, cred, LEASE_READ); - VOP_LOCK(fvp); /* XXX */ - VOP_UNLOCK(tvp); /* XXX */ - LEASE_CHECK(tvp, p, cred, LEASE_WRITE); - VOP_LOCK(tvp); /* XXX */ + VOP_UNLOCK(fvp, 0, p); /* XXX */ + VOP_LEASE(fvp, p, cred, LEASE_READ); + vn_lock(fvp, LK_EXCLUSIVE | LK_RETRY, p); /* XXX */ + VOP_UNLOCK(tvp, 0, p); /* XXX */ + VOP_LEASE(tvp, p, cred, LEASE_WRITE); + vn_lock(tvp, LK_EXCLUSIVE | LK_RETRY, p); /* XXX */ buf = malloc(MAXBSIZE, M_TEMP, M_WAITOK); @@ -542,6 +632,123 @@ union_copyfile(p, cred, fvp, tvp) } /* + * (un) is assumed to be locked on entry and remains + * locked on exit. + */ +int +union_copyup(un, docopy, cred, p) + struct union_node *un; + int docopy; + struct ucred *cred; + struct proc *p; +{ + int error; + struct vnode *lvp, *uvp; + + error = union_vn_create(&uvp, un, p); + if (error) + return (error); + + /* at this point, uppervp is locked */ + union_newupper(un, uvp); + un->un_flags |= UN_ULOCK; + + lvp = un->un_lowervp; + + if (docopy) { + /* + * XX - should not ignore errors + * from VOP_CLOSE + */ + vn_lock(lvp, LK_EXCLUSIVE | LK_RETRY, p); + error = VOP_OPEN(lvp, FREAD, cred, p); + if (error == 0) { + error = union_copyfile(lvp, uvp, cred, p); + VOP_UNLOCK(lvp, 0, p); + (void) VOP_CLOSE(lvp, FREAD, cred, p); + } +#ifdef UNION_DIAGNOSTIC + if (error == 0) + uprintf("union: copied up %s\n", un->un_path); +#endif + + } + un->un_flags &= ~UN_ULOCK; + VOP_UNLOCK(uvp, 0, p); + union_vn_close(uvp, FWRITE, cred, p); + vn_lock(uvp, LK_EXCLUSIVE | LK_RETRY, p); + un->un_flags |= UN_ULOCK; + + /* + * Subsequent IOs will go to the top layer, so + * call close on the lower vnode and open on the + * upper vnode to ensure that the filesystem keeps + * its references counts right. This doesn't do + * the right thing with (cred) and (FREAD) though. + * Ignoring error returns is not right, either. + */ + if (error == 0) { + int i; + + for (i = 0; i < un->un_openl; i++) { + (void) VOP_CLOSE(lvp, FREAD, cred, p); + (void) VOP_OPEN(uvp, FREAD, cred, p); + } + un->un_openl = 0; + } + + return (error); + +} + +static int +union_relookup(um, dvp, vpp, cnp, cn, path, pathlen) + struct union_mount *um; + struct vnode *dvp; + struct vnode **vpp; + struct componentname *cnp; + struct componentname *cn; + char *path; + int pathlen; +{ + int error; + + /* + * A new componentname structure must be faked up because + * there is no way to know where the upper level cnp came + * from or what it is being used for. This must duplicate + * some of the work done by NDINIT, some of the work done + * by namei, some of the work done by lookup and some of + * the work done by VOP_LOOKUP when given a CREATE flag. + * Conclusion: Horrible. + * + * The pathname buffer will be FREEed by VOP_MKDIR. + */ + cn->cn_namelen = pathlen; + cn->cn_pnbuf = malloc(cn->cn_namelen+1, M_NAMEI, M_WAITOK); + bcopy(path, cn->cn_pnbuf, cn->cn_namelen); + cn->cn_pnbuf[cn->cn_namelen] = '\0'; + + cn->cn_nameiop = CREATE; + cn->cn_flags = (LOCKPARENT|HASBUF|SAVENAME|SAVESTART|ISLASTCN); + cn->cn_proc = cnp->cn_proc; + if (um->um_op == UNMNT_ABOVE) + cn->cn_cred = cnp->cn_cred; + else + cn->cn_cred = um->um_cred; + cn->cn_nameptr = cn->cn_pnbuf; + cn->cn_hash = cnp->cn_hash; + cn->cn_consume = cnp->cn_consume; + + VREF(dvp); + error = relookup(dvp, vpp, cn); + if (!error) + vrele(dvp); + + return (error); +} + +/* * Create a shadow directory in the upper layer. * The new vnode is returned locked. * @@ -565,6 +772,19 @@ union_mkshadow(um, dvp, cnp, vpp) struct proc *p = cnp->cn_proc; struct componentname cn; + error = union_relookup(um, dvp, vpp, cnp, &cn, + cnp->cn_nameptr, cnp->cn_namelen); + if (error) + return (error); + + if (*vpp) { + VOP_ABORTOP(dvp, &cn); + VOP_UNLOCK(dvp, 0, p); + vrele(*vpp); + *vpp = NULLVP; + return (EEXIST); + } + /* * policy: when creating the shadow directory in the * upper layer, create it owned by the user who did @@ -573,55 +793,62 @@ union_mkshadow(um, dvp, cnp, vpp) * mkdir syscall). (jsp, kb) */ - /* - * A new componentname structure must be faked up because - * there is no way to know where the upper level cnp came - * from or what it is being used for. This must duplicate - * some of the work done by NDINIT, some of the work done - * by namei, some of the work done by lookup and some of - * the work done by VOP_LOOKUP when given a CREATE flag. - * Conclusion: Horrible. - * - * The pathname buffer will be FREEed by VOP_MKDIR. - */ - cn.cn_pnbuf = malloc(cnp->cn_namelen+1, M_NAMEI, M_WAITOK); - bcopy(cnp->cn_nameptr, cn.cn_pnbuf, cnp->cn_namelen); - cn.cn_pnbuf[cnp->cn_namelen] = '\0'; + VATTR_NULL(&va); + va.va_type = VDIR; + va.va_mode = um->um_cmode; - cn.cn_nameiop = CREATE; - cn.cn_flags = (LOCKPARENT|HASBUF|SAVENAME|SAVESTART|ISLASTCN); - cn.cn_proc = cnp->cn_proc; - if (um->um_op == UNMNT_ABOVE) - cn.cn_cred = cnp->cn_cred; - else - cn.cn_cred = um->um_cred; - cn.cn_nameptr = cn.cn_pnbuf; - cn.cn_namelen = cnp->cn_namelen; - cn.cn_hash = cnp->cn_hash; - cn.cn_consume = cnp->cn_consume; + /* VOP_LEASE: dvp is locked */ + VOP_LEASE(dvp, p, cn.cn_cred, LEASE_WRITE); - VREF(dvp); - error = relookup(dvp, vpp, &cn); - if (error) + error = VOP_MKDIR(dvp, vpp, &cn, &va); + return (error); +} + +/* + * Create a whiteout entry in the upper layer. + * + * (um) points to the union mount structure for access to the + * the mounting process's credentials. + * (dvp) is the directory in which to create the whiteout. + * it is locked on entry and exit. + * (cnp) is the componentname to be created. + */ +int +union_mkwhiteout(um, dvp, cnp, path) + struct union_mount *um; + struct vnode *dvp; + struct componentname *cnp; + char *path; +{ + int error; + struct vattr va; + struct proc *p = cnp->cn_proc; + struct vnode *wvp; + struct componentname cn; + + VOP_UNLOCK(dvp, 0, p); + error = union_relookup(um, dvp, &wvp, cnp, &cn, path, strlen(path)); + if (error) { + vn_lock(dvp, LK_EXCLUSIVE | LK_RETRY, p); return (error); - vrele(dvp); + } - if (*vpp) { + if (wvp) { VOP_ABORTOP(dvp, &cn); - VOP_UNLOCK(dvp); - vrele(*vpp); - *vpp = NULLVP; + vrele(dvp); + vrele(wvp); return (EEXIST); } - VATTR_NULL(&va); - va.va_type = VDIR; - va.va_mode = um->um_cmode; + /* VOP_LEASE: dvp is locked */ + VOP_LEASE(dvp, p, p->p_ucred, LEASE_WRITE); + + error = VOP_WHITEOUT(dvp, &cn, CREATE); + if (error) + VOP_ABORTOP(dvp, &cn); - /* LEASE_CHECK: dvp is locked */ - LEASE_CHECK(dvp, p, p->p_ucred, LEASE_WRITE); + vrele(dvp); - error = VOP_MKDIR(dvp, vpp, &cn, &va); return (error); } @@ -699,9 +926,8 @@ union_vn_create(vpp, un, p) VATTR_NULL(vap); vap->va_type = VREG; vap->va_mode = cmode; - LEASE_CHECK(un->un_dirvp, p, cred, LEASE_WRITE); - error = VOP_CREATE(un->un_dirvp, &vp, &cn, vap); - if (error) + VOP_LEASE(un->un_dirvp, p, cred, LEASE_WRITE); + if (error = VOP_CREATE(un->un_dirvp, &vp, &cn, vap)) return (error); error = VOP_OPEN(vp, fmode, cred, p); @@ -722,6 +948,7 @@ union_vn_close(vp, fmode, cred, p) struct ucred *cred; struct proc *p; { + if (fmode & FWRITE) --vp->v_writecount; return (VOP_CLOSE(vp, fmode, cred, p)); @@ -731,24 +958,137 @@ void union_removed_upper(un) struct union_node *un; { + struct proc *p = curproc; /* XXX */ + + union_newupper(un, NULLVP); + if (un->un_flags & UN_CACHED) { + un->un_flags &= ~UN_CACHED; + LIST_REMOVE(un, un_cache); + } + if (un->un_flags & UN_ULOCK) { un->un_flags &= ~UN_ULOCK; - VOP_UNLOCK(un->un_uppervp); + VOP_UNLOCK(un->un_uppervp, 0, p); } - - union_newupper(un, NULLVP); } +#if 0 struct vnode * union_lowervp(vp) struct vnode *vp; { struct union_node *un = VTOUNION(vp); - if (un->un_lowervp && (vp->v_type == un->un_lowervp->v_type)) { - if (vget(un->un_lowervp, 0)) - return (NULLVP); + if ((un->un_lowervp != NULLVP) && + (vp->v_type == un->un_lowervp->v_type)) { + if (vget(un->un_lowervp, 0) == 0) + return (un->un_lowervp); } - return (un->un_lowervp); + return (NULLVP); +} +#endif + +/* + * determine whether a whiteout is needed + * during a remove/rmdir operation. + */ +int +union_dowhiteout(un, cred, p) + struct union_node *un; + struct ucred *cred; + struct proc *p; +{ + struct vattr va; + + if (un->un_lowervp != NULLVP) + return (1); + + if (VOP_GETATTR(un->un_uppervp, &va, cred, p) == 0 && + (va.va_flags & OPAQUE)) + return (1); + + return (0); +} + +static void +union_dircache_r(vp, vppp, cntp) + struct vnode *vp; + struct vnode ***vppp; + int *cntp; +{ + struct union_node *un; + + if (vp->v_op != union_vnodeop_p) { + if (vppp) { + VREF(vp); + *(*vppp)++ = vp; + if (--(*cntp) == 0) + panic("union: dircache table too small"); + } else { + (*cntp)++; + } + + return; + } + + un = VTOUNION(vp); + if (un->un_uppervp != NULLVP) + union_dircache_r(un->un_uppervp, vppp, cntp); + if (un->un_lowervp != NULLVP) + union_dircache_r(un->un_lowervp, vppp, cntp); +} + +struct vnode * +union_dircache(vp, p) + struct vnode *vp; + struct proc *p; +{ + int cnt; + struct vnode *nvp; + struct vnode **vpp; + struct vnode **dircache; + struct union_node *un; + int error; + + vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, p); + dircache = VTOUNION(vp)->un_dircache; + + nvp = NULLVP; + + if (dircache == 0) { + cnt = 0; + union_dircache_r(vp, 0, &cnt); + cnt++; + dircache = (struct vnode **) + malloc(cnt * sizeof(struct vnode *), + M_TEMP, M_WAITOK); + vpp = dircache; + union_dircache_r(vp, &vpp, &cnt); + *vpp = NULLVP; + vpp = dircache + 1; + } else { + vpp = dircache; + do { + if (*vpp++ == VTOUNION(vp)->un_uppervp) + break; + } while (*vpp != NULLVP); + } + + if (*vpp == NULLVP) + goto out; + + vn_lock(*vpp, LK_EXCLUSIVE | LK_RETRY, p); + VREF(*vpp); + error = union_allocvp(&nvp, vp->v_mount, NULLVP, NULLVP, 0, *vpp, NULLVP, 0); + if (error) + goto out; + + VTOUNION(vp)->un_dircache = 0; + un = VTOUNION(nvp); + un->un_dircache = dircache; + +out: + VOP_UNLOCK(vp, 0, p); + return (nvp); } diff --git a/sys/miscfs/union/union_vfsops.c b/sys/miscfs/union/union_vfsops.c index a1a6a0d..0295961 100644 --- a/sys/miscfs/union/union_vfsops.c +++ b/sys/miscfs/union/union_vfsops.c @@ -1,6 +1,6 @@ /* - * Copyright (c) 1994 The Regents of the University of California. - * Copyright (c) 1994 Jan-Simon Pendry. + * Copyright (c) 1994, 1995 The Regents of the University of California. + * Copyright (c) 1994, 1995 Jan-Simon Pendry. * All rights reserved. * * This code is derived from software donated to Berkeley by @@ -34,7 +34,7 @@ * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * - * @(#)union_vfsops.c 8.7 (Berkeley) 3/5/94 + * @(#)union_vfsops.c 8.20 (Berkeley) 5/20/95 * $FreeBSD$ */ @@ -56,7 +56,7 @@ #include <sys/queue.h> #include <miscfs/union/union.h> -extern int union_init __P((void)); +extern int union_init __P((struct vfsconf *)); extern int union_fhtovp __P((struct mount *mp, struct fid *fidp, struct mbuf *nam, struct vnode **vpp, @@ -92,7 +92,7 @@ union_mount(mp, path, data, ndp, p) struct union_args args; struct vnode *lowerrootvp = NULLVP; struct vnode *upperrootvp = NULLVP; - struct union_mount *um; + struct union_mount *um = 0; struct ucred *cred = 0; struct ucred *scred; struct vattr va; @@ -118,34 +118,6 @@ union_mount(mp, path, data, ndp, p) } /* - * Take a copy of the process's credentials. This isn't - * quite right since the euid will always be zero and we - * want to get the "real" users credentials. So fix up - * the uid field after taking the copy. - */ - cred = crdup(p->p_ucred); - cred->cr_uid = p->p_cred->p_ruid; - - /* - * Ensure the *real* user has write permission on the - * mounted-on directory. This allows the mount_union - * command to be made setuid root so allowing anyone - * to do union mounts onto any directory on which they - * have write permission and which they also own. - */ - error = VOP_GETATTR(mp->mnt_vnodecovered, &va, cred, p); - if (error) - goto bad; - if ((va.va_uid != cred->cr_uid) && - (cred->cr_uid != 0)) { - error = EACCES; - goto bad; - } - error = VOP_ACCESS(mp->mnt_vnodecovered, VWRITE, cred, p); - if (error) - goto bad; - - /* * Get argument */ error = copyin(data, (caddr_t)&args, sizeof(struct union_args)); @@ -156,18 +128,10 @@ union_mount(mp, path, data, ndp, p) VREF(lowerrootvp); /* - * Find upper node. Use the real process credentials, - * not the effective ones since this will have come - * through a setuid process (mount_union). All this - * messing around with permissions is entirely bogus - * and should be removed by allowing any user straight - * past the mount system call. + * Find upper node. */ - scred = p->p_ucred; - p->p_ucred = cred; NDINIT(ndp, LOOKUP, FOLLOW|WANTPARENT, UIO_USERSPACE, args.target, p); - p->p_ucred = scred; error = namei(ndp); if (error) @@ -218,7 +182,18 @@ union_mount(mp, path, data, ndp, p) goto bad; } - um->um_cred = cred; + /* + * Unless the mount is readonly, ensure that the top layer + * supports whiteout operations + */ + if ((mp->mnt_flag & MNT_RDONLY) == 0) { + error = VOP_WHITEOUT(um->um_uppervp, (struct componentname *) 0, LOOKUP); + if (error) + goto bad; + } + + um->um_cred = p->p_ucred; + crhold(um->um_cred); um->um_cmode = UN_DIRMODE &~ p->p_fd->fd_cmask; /* @@ -246,24 +221,18 @@ union_mount(mp, path, data, ndp, p) */ mp->mnt_flag |= (um->um_uppervp->v_mount->mnt_flag & MNT_RDONLY); - /* - * This is a user mount. Privilege check for unmount - * will be done in union_unmount. - */ - mp->mnt_flag |= MNT_USER; - mp->mnt_data = (qaddr_t) um; - getnewfsid(mp, MOUNT_UNION); + vfs_getnewfsid(mp); (void) copyinstr(path, mp->mnt_stat.f_mntonname, MNAMELEN - 1, &size); bzero(mp->mnt_stat.f_mntonname + size, MNAMELEN - size); switch (um->um_op) { case UNMNT_ABOVE: - cp = "<above>"; + cp = "<above>:"; break; case UNMNT_BELOW: - cp = "<below>"; + cp = "<below>:"; break; case UNMNT_REPLACE: cp = ""; @@ -287,6 +256,8 @@ union_mount(mp, path, data, ndp, p) return (0); bad: + if (um) + free(um, M_UFSMNT); if (cred) crfree(cred); if (upperrootvp) @@ -323,40 +294,55 @@ union_unmount(mp, mntflags, p) struct union_mount *um = MOUNTTOUNIONMOUNT(mp); struct vnode *um_rootvp; int error; + int freeing; int flags = 0; #ifdef UNION_DIAGNOSTIC printf("union_unmount(mp = %x)\n", mp); #endif - /* only the mounter, or superuser can unmount */ - if ((p->p_cred->p_ruid != um->um_cred->cr_uid) && - (error = suser(p->p_ucred, &p->p_acflag))) + if (mntflags & MNT_FORCE) + flags |= FORCECLOSE; + + if (error = union_root(mp, &um_rootvp)) return (error); - if (mntflags & MNT_FORCE) { - /* union can never be rootfs so don't check for it */ - if (!doforce) - return (EINVAL); - flags |= FORCECLOSE; + /* + * Keep flushing vnodes from the mount list. + * This is needed because of the un_pvp held + * reference to the parent vnode. + * If more vnodes have been freed on a given pass, + * the try again. The loop will iterate at most + * (d) times, where (d) is the maximum tree depth + * in the filesystem. + */ + for (freeing = 0; vflush(mp, um_rootvp, flags) != 0;) { + struct vnode *vp; + int n; + + /* count #vnodes held on mount list */ + for (n = 0, vp = mp->mnt_vnodelist.lh_first; + vp != NULLVP; + vp = vp->v_mntvnodes.le_next) + n++; + + /* if this is unchanged then stop */ + if (n == freeing) + break; + + /* otherwise try once more time */ + freeing = n; } - error = union_root(mp, &um_rootvp); - if (error) - return (error); + /* At this point the root vnode should have a single reference */ if (um_rootvp->v_usecount > 1) { vput(um_rootvp); return (EBUSY); } - error = vflush(mp, um_rootvp, flags); - if (error) { - vput(um_rootvp); - return (error); - } #ifdef UNION_DIAGNOSTIC - vprint("alias root of lower", um_rootvp); -#endif + vprint("union root", um_rootvp); +#endif /* * Discard references to upper and lower target vnodes. */ @@ -385,16 +371,11 @@ union_root(mp, vpp) struct mount *mp; struct vnode **vpp; { + struct proc *p = curproc; /* XXX */ struct union_mount *um = MOUNTTOUNIONMOUNT(mp); int error; int loselock; -#ifdef UNION_DIAGNOSTIC - printf("union_root(mp = %x, lvp = %x, uvp = %x)\n", mp, - um->um_lowervp, - um->um_uppervp); -#endif - /* * Return locked reference to root. */ @@ -403,7 +384,7 @@ union_root(mp, vpp) VOP_ISLOCKED(um->um_uppervp)) { loselock = 1; } else { - VOP_LOCK(um->um_uppervp); + vn_lock(um->um_uppervp, LK_EXCLUSIVE | LK_RETRY, p); loselock = 0; } if (um->um_lowervp) @@ -413,16 +394,17 @@ union_root(mp, vpp) (struct vnode *) 0, (struct componentname *) 0, um->um_uppervp, - um->um_lowervp); + um->um_lowervp, + 1); if (error) { - if (!loselock) - VOP_UNLOCK(um->um_uppervp); - vrele(um->um_uppervp); + if (loselock) + vrele(um->um_uppervp); + else + vput(um->um_uppervp); if (um->um_lowervp) vrele(um->um_lowervp); } else { - (*vpp)->v_flag |= VROOT; if (loselock) VTOUNION(*vpp)->un_flags &= ~UN_ULOCK; } @@ -431,18 +413,6 @@ union_root(mp, vpp) } int -union_quotactl(mp, cmd, uid, arg, p) - struct mount *mp; - int cmd; - uid_t uid; - caddr_t arg; - struct proc *p; -{ - - return (EOPNOTSUPP); -} - -int union_statfs(mp, sbp, p) struct mount *mp; struct statfs *sbp; @@ -485,7 +455,6 @@ union_statfs(mp, sbp, p) if (error) return (error); - sbp->f_type = MOUNT_UNION; sbp->f_flags = mstat.f_flags; sbp->f_bsize = mstat.f_bsize; sbp->f_iosize = mstat.f_iosize; @@ -496,18 +465,23 @@ union_statfs(mp, sbp, p) * kind of sense. none of this makes sense though. */ - if (mstat.f_bsize != lbsize) { + if (mstat.f_bsize != lbsize) sbp->f_blocks = sbp->f_blocks * lbsize / mstat.f_bsize; - sbp->f_bfree = sbp->f_bfree * lbsize / mstat.f_bsize; - sbp->f_bavail = sbp->f_bavail * lbsize / mstat.f_bsize; - } + + /* + * The "total" fields count total resources in all layers, + * the "free" fields count only those resources which are + * free in the upper layer (since only the upper layer + * is writeable). + */ sbp->f_blocks += mstat.f_blocks; - sbp->f_bfree += mstat.f_bfree; - sbp->f_bavail += mstat.f_bavail; + sbp->f_bfree = mstat.f_bfree; + sbp->f_bavail = mstat.f_bavail; sbp->f_files += mstat.f_files; - sbp->f_ffree += mstat.f_ffree; + sbp->f_ffree = mstat.f_ffree; if (sbp != &mp->mnt_stat) { + sbp->f_type = mp->mnt_vfc->vfc_typenum; bcopy(&mp->mnt_stat.f_fsid, &sbp->f_fsid, sizeof(sbp->f_fsid)); bcopy(mp->mnt_stat.f_mntonname, sbp->f_mntonname, MNAMELEN); bcopy(mp->mnt_stat.f_mntfromname, sbp->f_mntfromname, MNAMELEN); @@ -515,51 +489,21 @@ union_statfs(mp, sbp, p) return (0); } -int -union_sync(mp, waitfor, cred, p) - struct mount *mp; - int waitfor; - struct ucred *cred; - struct proc *p; -{ - - /* - * XXX - Assumes no data cached at union layer. - */ - return (0); -} - -int -union_vget(mp, ino, vpp) - struct mount *mp; - ino_t ino; - struct vnode **vpp; -{ - - return (EOPNOTSUPP); -} - -int -union_fhtovp(mp, fidp, nam, vpp, exflagsp, credanonp) - struct mount *mp; - struct fid *fidp; - struct mbuf *nam; - struct vnode **vpp; - int *exflagsp; - struct ucred **credanonp; -{ - - return (EOPNOTSUPP); -} - -int -union_vptofh(vp, fhp) - struct vnode *vp; - struct fid *fhp; -{ - - return (EOPNOTSUPP); -} +/* + * XXX - Assumes no data cached at union layer. + */ +#define union_sync ((int (*) __P((struct mount *, int, struct ucred *, \ + struct proc *)))nullop) + +#define union_fhtovp ((int (*) __P((struct mount *, struct fid *, \ + struct mbuf *, struct vnode **, int *, struct ucred **)))eopnotsupp) +#define union_quotactl ((int (*) __P((struct mount *, int, uid_t, caddr_t, \ + struct proc *)))eopnotsupp) +#define union_sysctl ((int (*) __P((int *, u_int, void *, size_t *, void *, \ + size_t, struct proc *)))eopnotsupp) +#define union_vget ((int (*) __P((struct mount *, ino_t, struct vnode **))) \ + eopnotsupp) +#define union_vptofh ((int (*) __P((struct vnode *, struct fid *)))eopnotsupp) struct vfsops union_vfsops = { union_mount, diff --git a/sys/miscfs/union/union_vnops.c b/sys/miscfs/union/union_vnops.c index 5940243..4a66dc0 100644 --- a/sys/miscfs/union/union_vnops.c +++ b/sys/miscfs/union/union_vnops.c @@ -1,7 +1,7 @@ /* - * Copyright (c) 1992, 1993, 1994 The Regents of the University of California. - * Copyright (c) 1992, 1993, 1994 Jan-Simon Pendry. - * All rights reserved. + * Copyright (c) 1992, 1993, 1994, 1995 Jan-Simon Pendry. + * Copyright (c) 1992, 1993, 1994, 1995 + * The Regents of the University of California. All rights reserved. * * This code is derived from software contributed to Berkeley by * Jan-Simon Pendry. @@ -34,7 +34,7 @@ * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * - * @(#)union_vnops.c 8.6 (Berkeley) 2/17/94 + * @(#)union_vnops.c 8.32 (Berkeley) 6/23/95 * $FreeBSD$ */ @@ -43,7 +43,7 @@ #include <sys/proc.h> #include <sys/file.h> #include <sys/time.h> -#include <sys/kernel.h> +#include <sys/stat.h> #include <sys/types.h> #include <sys/vnode.h> #include <sys/mount.h> @@ -51,16 +51,12 @@ #include <sys/malloc.h> #include <sys/buf.h> #include <sys/queue.h> +#include <sys/lock.h> #include <miscfs/union/union.h> -/* FIXUP throws the lock on the uppervp vnode if the union_node is already - * locked and the uppervp vnode is not. Before, this was thrown regardless - * of the state of the union_node which resulted in locked vnodes which - * were never unlocked (since the union would never be unlocked). - */ -#define FIXUP(un) { \ - if (((un)->un_flags & (UN_LOCKED|UN_ULOCK)) == UN_LOCKED) { \ - union_fixup(un); \ +#define FIXUP(un, p) { \ + if (((un)->un_flags & UN_ULOCK) == 0) { \ + union_fixup(un, p); \ } \ } @@ -70,7 +66,7 @@ extern int union_advlock __P((struct vop_advlock_args *ap)); extern int union_bmap __P((struct vop_bmap_args *ap)); extern int union_close __P((struct vop_close_args *ap)); extern int union_create __P((struct vop_create_args *ap)); -static void union_fixup __P((struct union_node *un)); +static void union_fixup __P((struct union_node *un, struct proc *p)); extern int union_fsync __P((struct vop_fsync_args *ap)); extern int union_getattr __P((struct vop_getattr_args *ap)); extern int union_inactive __P((struct vop_inactive_args *ap)); @@ -79,7 +75,7 @@ extern int union_islocked __P((struct vop_islocked_args *ap)); extern int union_link __P((struct vop_link_args *ap)); extern int union_lock __P((struct vop_lock_args *ap)); extern int union_lookup __P((struct vop_lookup_args *ap)); -static int union_lookup1 __P((struct vnode *udvp, struct vnode *dvp, +static int union_lookup1 __P((struct vnode *udvp, struct vnode **dvpp, struct vnode **vpp, struct componentname *cnp)); extern int union_mkdir __P((struct vop_mkdir_args *ap)); @@ -100,29 +96,34 @@ extern int union_select __P((struct vop_select_args *ap)); extern int union_setattr __P((struct vop_setattr_args *ap)); extern int union_strategy __P((struct vop_strategy_args *ap)); extern int union_symlink __P((struct vop_symlink_args *ap)); -extern int union_unlock __P((struct vop_lock_args *ap)); +extern int union_unlock __P((struct vop_unlock_args *ap)); extern int union_write __P((struct vop_read_args *ap)); static void -union_fixup(un) +union_fixup(un, p) struct union_node *un; + struct proc *p; { - VOP_LOCK(un->un_uppervp); + vn_lock(un->un_uppervp, LK_EXCLUSIVE | LK_RETRY, p); un->un_flags |= UN_ULOCK; } static int -union_lookup1(udvp, dvp, vpp, cnp) +union_lookup1(udvp, dvpp, vpp, cnp) struct vnode *udvp; - struct vnode *dvp; + struct vnode **dvpp; struct vnode **vpp; struct componentname *cnp; { int error; + struct proc *p = cnp->cn_proc; struct vnode *tdvp; + struct vnode *dvp; struct mount *mp; + dvp = *dvpp; + /* * If stepping up the directory tree, check for going * back across the mount point, in which case do what @@ -130,21 +131,18 @@ union_lookup1(udvp, dvp, vpp, cnp) * hierarchy. */ if (cnp->cn_flags & ISDOTDOT) { - for (;;) { + while ((dvp != udvp) && (dvp->v_flag & VROOT)) { /* * Don't do the NOCROSSMOUNT check * at this level. By definition, * union fs deals with namespaces, not * filesystems. */ - if ((dvp->v_flag & VROOT) == 0) - break; - tdvp = dvp; - dvp = dvp->v_mount->mnt_vnodecovered; + *dvpp = dvp = dvp->v_mount->mnt_vnodecovered; vput(tdvp); VREF(dvp); - VOP_LOCK(dvp); + vn_lock(dvp, LK_EXCLUSIVE | LK_RETRY, p); } } @@ -158,7 +156,7 @@ union_lookup1(udvp, dvp, vpp, cnp) * here to allow it to be unlocked again (phew) in union_lookup. */ if (dvp != tdvp && !(cnp->cn_flags & ISLASTCN)) - VOP_LOCK(dvp); + vn_lock(dvp, LK_EXCLUSIVE | LK_RETRY, p); dvp = tdvp; @@ -170,13 +168,11 @@ union_lookup1(udvp, dvp, vpp, cnp) while (dvp != udvp && (dvp->v_type == VDIR) && (mp = dvp->v_mountedhere)) { - if (mp->mnt_flag & MNT_MLOCK) { - mp->mnt_flag |= MNT_MWAIT; - (void) tsleep((caddr_t) mp, PVFS, "unlkup", 0); + if (vfs_busy(mp, 0, 0, p)) continue; - } error = VFS_ROOT(mp, &tdvp); + vfs_unbusy(mp, p); if (error) { vput(dvp); return (error); @@ -206,9 +202,28 @@ union_lookup(ap) struct vnode *dvp = ap->a_dvp; struct union_node *dun = VTOUNION(dvp); struct componentname *cnp = ap->a_cnp; + struct proc *p = cnp->cn_proc; int lockparent = cnp->cn_flags & LOCKPARENT; struct union_mount *um = MOUNTTOUNIONMOUNT(dvp->v_mount); - struct ucred *saved_cred = 0; + struct ucred *saved_cred; + int iswhiteout; + struct vattr va; + +#ifdef notyet + if (cnp->cn_namelen == 3 && + cnp->cn_nameptr[2] == '.' && + cnp->cn_nameptr[1] == '.' && + cnp->cn_nameptr[0] == '.') { + dvp = *ap->a_vpp = LOWERVP(ap->a_dvp); + if (dvp == NULLVP) + return (ENOENT); + VREF(dvp); + vn_lock(dvp, LK_EXCLUSIVE | LK_RETRY, p); + if (!lockparent || !(cnp->cn_flags & ISLASTCN)) + VOP_UNLOCK(ap->a_dvp, 0, p); + return (0); + } +#endif cnp->cn_flags |= LOCKPARENT; @@ -216,6 +231,7 @@ union_lookup(ap) lowerdvp = dun->un_lowervp; uppervp = NULLVP; lowervp = NULLVP; + iswhiteout = 0; /* * do the lookup in the upper level. @@ -223,9 +239,9 @@ union_lookup(ap) * then assume that something special is going * on and just return that vnode. */ - if (upperdvp) { - FIXUP(dun); - uerror = union_lookup1(um->um_uppervp, upperdvp, + if (upperdvp != NULLVP) { + FIXUP(dun, p); + uerror = union_lookup1(um->um_uppervp, &upperdvp, &uppervp, cnp); /*if (uppervp == upperdvp) dun->un_flags |= UN_KLOCK;*/ @@ -236,6 +252,16 @@ union_lookup(ap) cnp->cn_flags &= ~LOCKPARENT; return (uerror); } + if (uerror == ENOENT || uerror == EJUSTRETURN) { + if (cnp->cn_flags & ISWHITEOUT) { + iswhiteout = 1; + } else if (lowerdvp != NULLVP) { + lerror = VOP_GETATTR(upperdvp, &va, + cnp->cn_cred, cnp->cn_proc); + if (lerror == 0 && (va.va_flags & OPAQUE)) + iswhiteout = 1; + } + } } else { uerror = ENOENT; } @@ -247,10 +273,10 @@ union_lookup(ap) * back from the upper layer and return the lower vnode * instead. */ - if (lowerdvp) { + if (lowerdvp != NULLVP && !iswhiteout) { int nameiop; - VOP_LOCK(lowerdvp); + vn_lock(lowerdvp, LK_EXCLUSIVE | LK_RETRY, p); /* * Only do a LOOKUP on the bottom node, since @@ -262,17 +288,17 @@ union_lookup(ap) saved_cred = cnp->cn_cred; cnp->cn_cred = um->um_cred; } - lerror = union_lookup1(um->um_lowervp, lowerdvp, + lerror = union_lookup1(um->um_lowervp, &lowerdvp, &lowervp, cnp); if (um->um_op == UNMNT_BELOW) cnp->cn_cred = saved_cred; cnp->cn_nameiop = nameiop; if (lowervp != lowerdvp) - VOP_UNLOCK(lowerdvp); + VOP_UNLOCK(lowerdvp, 0, p); if (cnp->cn_consume != 0) { - if (uppervp) { + if (uppervp != NULLVP) { if (uppervp == upperdvp) vrele(uppervp); else @@ -286,6 +312,14 @@ union_lookup(ap) } } else { lerror = ENOENT; + if ((cnp->cn_flags & ISDOTDOT) && dun->un_pvp != NULLVP) { + lowervp = LOWERVP(dun->un_pvp); + if (lowervp != NULLVP) { + VREF(lowervp); + vn_lock(lowervp, LK_EXCLUSIVE | LK_RETRY, p); + lerror = 0; + } + } } if (!lockparent) @@ -326,13 +360,13 @@ union_lookup(ap) if (uerror != 0 /* && (lerror == 0) */ ) { if (lowervp->v_type == VDIR) { /* case 2b. */ dun->un_flags &= ~UN_ULOCK; - VOP_UNLOCK(upperdvp); + VOP_UNLOCK(upperdvp, 0, p); uerror = union_mkshadow(um, upperdvp, cnp, &uppervp); - VOP_LOCK(upperdvp); + vn_lock(upperdvp, LK_EXCLUSIVE | LK_RETRY, p); dun->un_flags |= UN_ULOCK; if (uerror) { - if (lowervp) { + if (lowervp != NULLVP) { vput(lowervp); lowervp = NULLVP; } @@ -341,21 +375,21 @@ union_lookup(ap) } } - if (lowervp) - VOP_UNLOCK(lowervp); + if (lowervp != NULLVP) + VOP_UNLOCK(lowervp, 0, p); error = union_allocvp(ap->a_vpp, dvp->v_mount, dvp, upperdvp, cnp, - uppervp, lowervp); + uppervp, lowervp, 1); if (error) { - if (uppervp) + if (uppervp != NULLVP) vput(uppervp); - if (lowervp) + if (lowervp != NULLVP) vrele(lowervp); } else { if (*ap->a_vpp != dvp) if (!lockparent || !(cnp->cn_flags & ISLASTCN)) - VOP_UNLOCK(dvp); + VOP_UNLOCK(dvp, 0, p); } return (error); @@ -372,28 +406,26 @@ union_create(ap) { struct union_node *un = VTOUNION(ap->a_dvp); struct vnode *dvp = un->un_uppervp; + struct componentname *cnp = ap->a_cnp; + struct proc *p = cnp->cn_proc; - if (dvp) { + if (dvp != NULLVP) { int error; struct vnode *vp; + struct mount *mp; - FIXUP(un); + FIXUP(un, p); VREF(dvp); un->un_flags |= UN_KLOCK; + mp = ap->a_dvp->v_mount; vput(ap->a_dvp); - error = VOP_CREATE(dvp, &vp, ap->a_cnp, ap->a_vap); + error = VOP_CREATE(dvp, &vp, cnp, ap->a_vap); if (error) return (error); - error = union_allocvp( - ap->a_vpp, - ap->a_dvp->v_mount, - ap->a_dvp, - NULLVP, - ap->a_cnp, - vp, - NULLVP); + error = union_allocvp(ap->a_vpp, mp, NULLVP, NULLVP, cnp, vp, + NULLVP, 1); if (error) vput(vp); return (error); @@ -404,6 +436,25 @@ union_create(ap) } int +union_whiteout(ap) + struct vop_whiteout_args /* { + struct vnode *a_dvp; + struct componentname *a_cnp; + int a_flags; + } */ *ap; +{ + struct union_node *un = VTOUNION(ap->a_dvp); + struct componentname *cnp = ap->a_cnp; + struct proc *p = cnp->cn_proc; + + if (un->un_uppervp == NULLVP) + return (EOPNOTSUPP); + + FIXUP(un, p); + return (VOP_WHITEOUT(un->un_uppervp, cnp, ap->a_flags)); +} + +int union_mknod(ap) struct vop_mknod_args /* { struct vnode *a_dvp; @@ -414,29 +465,27 @@ union_mknod(ap) { struct union_node *un = VTOUNION(ap->a_dvp); struct vnode *dvp = un->un_uppervp; + struct componentname *cnp = ap->a_cnp; + struct proc *p = cnp->cn_proc; - if (dvp) { + if (dvp != NULLVP) { int error; struct vnode *vp; + struct mount *mp; - FIXUP(un); + FIXUP(un, p); VREF(dvp); un->un_flags |= UN_KLOCK; + mp = ap->a_dvp->v_mount; vput(ap->a_dvp); - error = VOP_MKNOD(dvp, &vp, ap->a_cnp, ap->a_vap); + error = VOP_MKNOD(dvp, &vp, cnp, ap->a_vap); if (error) return (error); - if (vp) { - error = union_allocvp( - ap->a_vpp, - ap->a_dvp->v_mount, - ap->a_dvp, - NULLVP, - ap->a_cnp, - vp, - NULLVP); + if (vp != NULLVP) { + error = union_allocvp(ap->a_vpp, mp, NULLVP, NULLVP, + cnp, vp, NULLVP, 1); if (error) vput(vp); } @@ -476,77 +525,7 @@ union_open(ap) */ tvp = un->un_lowervp; if ((ap->a_mode & FWRITE) && (tvp->v_type == VREG)) { - struct vnode *vp; - int i; - - /* - * Open the named file in the upper layer. Note that - * the file may have come into existence *since* the - * lookup was done, since the upper layer may really - * be a loopback mount of some other filesystem... - * so open the file with exclusive create and barf if - * it already exists. - * XXX - perhaps should re-lookup the node (once more - * with feeling) and simply open that. Who knows. - */ - error = union_vn_create(&vp, un, p); - if (error) - return (error); - - /* at this point, uppervp is locked */ - union_newupper(un, vp); - un->un_flags |= UN_ULOCK; - - /* - * Now, if the file is being opened with truncation, - * then the (new) upper vnode is ready to fly, - * otherwise the data from the lower vnode must be - * copied to the upper layer first. This only works - * for regular files (check is made above). - */ - if ((mode & O_TRUNC) == 0) { - /* - * XXX - should not ignore errors - * from VOP_CLOSE - */ - VOP_LOCK(tvp); - error = VOP_OPEN(tvp, FREAD, cred, p); - if (error == 0) { - error = union_copyfile(p, cred, - tvp, un->un_uppervp); - VOP_UNLOCK(tvp); - (void) VOP_CLOSE(tvp, FREAD, cred, p); - } else { - VOP_UNLOCK(tvp); - } - -#ifdef UNION_DIAGNOSTIC - if (!error) - uprintf("union: copied up %s\n", - un->un_path); -#endif - } - - un->un_flags &= ~UN_ULOCK; - VOP_UNLOCK(un->un_uppervp); - union_vn_close(un->un_uppervp, FWRITE, cred, p); - VOP_LOCK(un->un_uppervp); - un->un_flags |= UN_ULOCK; - - /* - * Subsequent IOs will go to the top layer, so - * call close on the lower vnode and open on the - * upper vnode to ensure that the filesystem keeps - * its references counts right. This doesn't do - * the right thing with (cred) and (FREAD) though. - * Ignoring error returns is not righ, either. - */ - for (i = 0; i < un->un_openl; i++) { - (void) VOP_CLOSE(tvp, FREAD, cred, p); - (void) VOP_OPEN(un->un_uppervp, FREAD, cred, p); - } - un->un_openl = 0; - + error = union_copyup(un, (mode&O_TRUNC) == 0, cred, p); if (error == 0) error = VOP_OPEN(un->un_uppervp, mode, cred, p); return (error); @@ -556,14 +535,14 @@ union_open(ap) * Just open the lower vnode */ un->un_openl++; - VOP_LOCK(tvp); + vn_lock(tvp, LK_EXCLUSIVE | LK_RETRY, p); error = VOP_OPEN(tvp, mode, cred, p); - VOP_UNLOCK(tvp); + VOP_UNLOCK(tvp, 0, p); return (error); } - FIXUP(un); + FIXUP(un, p); error = VOP_OPEN(tvp, mode, cred, p); @@ -582,9 +561,7 @@ union_close(ap) struct union_node *un = VTOUNION(ap->a_vp); struct vnode *vp; - if (un->un_uppervp) { - vp = un->un_uppervp; - } else { + if ((vp = un->un_uppervp) == NULLVP) { #ifdef UNION_DIAGNOSTIC if (un->un_openl <= 0) panic("union: un_openl cnt"); @@ -593,7 +570,8 @@ union_close(ap) vp = un->un_lowervp; } - return (VOP_CLOSE(vp, ap->a_fflag, ap->a_cred, ap->a_p)); + ap->a_vp = vp; + return (VCALL(vp, VOFFSET(vop_close), ap)); } /* @@ -615,27 +593,29 @@ union_access(ap) } */ *ap; { struct union_node *un = VTOUNION(ap->a_vp); + struct proc *p = ap->a_p; int error = EACCES; struct vnode *vp; - vp = un->un_uppervp; - if (vp) { - FIXUP(un); - return (VOP_ACCESS(vp, ap->a_mode, ap->a_cred, ap->a_p)); + if ((vp = un->un_uppervp) != NULLVP) { + FIXUP(un, p); + ap->a_vp = vp; + return (VCALL(vp, VOFFSET(vop_access), ap)); } - vp = un->un_lowervp; - if (vp) { - VOP_LOCK(vp); - error = VOP_ACCESS(vp, ap->a_mode, ap->a_cred, ap->a_p); + if ((vp = un->un_lowervp) != NULLVP) { + vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, p); + ap->a_vp = vp; + error = VCALL(vp, VOFFSET(vop_access), ap); if (error == 0) { struct union_mount *um = MOUNTTOUNIONMOUNT(vp->v_mount); - if (um->um_op == UNMNT_BELOW) - error = VOP_ACCESS(vp, ap->a_mode, - um->um_cred, ap->a_p); + if (um->um_op == UNMNT_BELOW) { + ap->a_cred = um->um_cred; + error = VCALL(vp, VOFFSET(vop_access), ap); + } } - VOP_UNLOCK(vp); + VOP_UNLOCK(vp, 0, p); if (error) return (error); } @@ -644,7 +624,8 @@ union_access(ap) } /* - * We handle getattr only to change the fsid. + * We handle getattr only to change the fsid and + * track object sizes */ int union_getattr(ap) @@ -658,6 +639,7 @@ union_getattr(ap) int error; struct union_node *un = VTOUNION(ap->a_vp); struct vnode *vp = un->un_uppervp; + struct proc *p = ap->a_p; struct vattr *vap; struct vattr va; @@ -675,10 +657,21 @@ union_getattr(ap) vp = un->un_uppervp; if (vp != NULLVP) { - FIXUP(un); + /* + * It's not clear whether VOP_GETATTR is to be + * called with the vnode locked or not. stat() calls + * it with (vp) locked, and fstat calls it with + * (vp) unlocked. + * In the mean time, compensate here by checking + * the union_node's lock flag. + */ + if (un->un_flags & UN_LOCKED) + FIXUP(un, p); + error = VOP_GETATTR(vp, vap, ap->a_cred, ap->a_p); if (error) return (error); + union_newsize(ap->a_vp, vap->va_size, VNOVAL); } if (vp == NULLVP) { @@ -691,17 +684,16 @@ union_getattr(ap) } if (vp != NULLVP) { - VOP_LOCK(vp); error = VOP_GETATTR(vp, vap, ap->a_cred, ap->a_p); - VOP_UNLOCK(vp); if (error) return (error); + union_newsize(ap->a_vp, VNOVAL, vap->va_size); } if ((vap != ap->a_vap) && (vap->va_type == VDIR)) ap->a_vap->va_nlink += vap->va_nlink; - vap->va_fsid = ap->a_vp->v_mount->mnt_stat.f_fsid.val[0]; + ap->a_vap->va_fsid = ap->a_vp->v_mount->mnt_stat.f_fsid.val[0]; return (0); } @@ -715,6 +707,7 @@ union_setattr(ap) } */ *ap; { struct union_node *un = VTOUNION(ap->a_vp); + struct proc *p = ap->a_p; int error; /* @@ -724,21 +717,11 @@ union_setattr(ap) */ if ((un->un_uppervp == NULLVP) && /* assert(un->un_lowervp != NULLVP) */ - (un->un_lowervp->v_type == VREG) && - (ap->a_vap->va_size == 0)) { - struct vnode *vp; - - error = union_vn_create(&vp, un, ap->a_p); + (un->un_lowervp->v_type == VREG)) { + error = union_copyup(un, (ap->a_vap->va_size != 0), + ap->a_cred, ap->a_p); if (error) return (error); - - /* at this point, uppervp is locked */ - union_newupper(un, vp); - - VOP_UNLOCK(vp); - union_vn_close(un->un_uppervp, FWRITE, ap->a_cred, ap->a_p); - VOP_LOCK(vp); - un->un_flags |= UN_ULOCK; } /* @@ -746,9 +729,11 @@ union_setattr(ap) * otherwise return read-only filesystem error. */ if (un->un_uppervp != NULLVP) { - FIXUP(un); + FIXUP(un, p); error = VOP_SETATTR(un->un_uppervp, ap->a_vap, ap->a_cred, ap->a_p); + if ((error == 0) && (ap->a_vap->va_size != VNOVAL)) + union_newsize(ap->a_vp, ap->a_vap->va_size, VNOVAL); } else { error = EROFS; } @@ -766,16 +751,36 @@ union_read(ap) } */ *ap; { int error; + struct proc *p = ap->a_uio->uio_procp; struct vnode *vp = OTHERVP(ap->a_vp); int dolock = (vp == LOWERVP(ap->a_vp)); if (dolock) - VOP_LOCK(vp); + vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, p); else - FIXUP(VTOUNION(ap->a_vp)); + FIXUP(VTOUNION(ap->a_vp), p); error = VOP_READ(vp, ap->a_uio, ap->a_ioflag, ap->a_cred); if (dolock) - VOP_UNLOCK(vp); + VOP_UNLOCK(vp, 0, p); + + /* + * XXX + * perhaps the size of the underlying object has changed under + * our feet. take advantage of the offset information present + * in the uio structure. + */ + if (error == 0) { + struct union_node *un = VTOUNION(ap->a_vp); + off_t cur = ap->a_uio->uio_offset; + + if (vp == un->un_uppervp) { + if (cur > un->un_uppersz) + union_newsize(ap->a_vp, cur, VNOVAL); + } else { + if (cur > un->un_lowersz) + union_newsize(ap->a_vp, VNOVAL, cur); + } + } return (error); } @@ -790,21 +795,47 @@ union_write(ap) } */ *ap; { int error; - struct vnode *vp = OTHERVP(ap->a_vp); - int dolock = (vp == LOWERVP(ap->a_vp)); + struct vnode *vp; + struct union_node *un = VTOUNION(ap->a_vp); + struct proc *p = ap->a_uio->uio_procp; - if (dolock) - VOP_LOCK(vp); - else - FIXUP(VTOUNION(ap->a_vp)); + vp = UPPERVP(ap->a_vp); + if (vp == NULLVP) + panic("union: missing upper layer in write"); + + FIXUP(un, p); error = VOP_WRITE(vp, ap->a_uio, ap->a_ioflag, ap->a_cred); - if (dolock) - VOP_UNLOCK(vp); + + /* + * the size of the underlying object may be changed by the + * write. + */ + if (error == 0) { + off_t cur = ap->a_uio->uio_offset; + + if (cur > un->un_uppersz) + union_newsize(ap->a_vp, cur, VNOVAL); + } return (error); } int +union_lease(ap) + struct vop_lease_args /* { + struct vnode *a_vp; + struct proc *a_p; + struct ucred *a_cred; + int a_flag; + } */ *ap; +{ + register struct vnode *ovp = OTHERVP(ap->a_vp); + + ap->a_vp = ovp; + return (VCALL(ovp, VOFFSET(vop_lease), ap)); +} + +int union_ioctl(ap) struct vop_ioctl_args /* { struct vnode *a_vp; @@ -815,9 +846,10 @@ union_ioctl(ap) struct proc *a_p; } */ *ap; { + register struct vnode *ovp = OTHERVP(ap->a_vp); - return (VOP_IOCTL(OTHERVP(ap->a_vp), ap->a_command, ap->a_data, - ap->a_fflag, ap->a_cred, ap->a_p)); + ap->a_vp = ovp; + return (VCALL(ovp, VOFFSET(vop_ioctl), ap)); } int @@ -830,9 +862,28 @@ union_select(ap) struct proc *a_p; } */ *ap; { + register struct vnode *ovp = OTHERVP(ap->a_vp); - return (VOP_SELECT(OTHERVP(ap->a_vp), ap->a_which, ap->a_fflags, - ap->a_cred, ap->a_p)); + ap->a_vp = ovp; + return (VCALL(ovp, VOFFSET(vop_select), ap)); +} + +int +union_revoke(ap) + struct vop_revoke_args /* { + struct vnode *a_vp; + int a_flags; + struct proc *a_p; + } */ *ap; +{ + struct vnode *vp = ap->a_vp; + + if (UPPERVP(vp)) + VOP_REVOKE(UPPERVP(vp), ap->a_flags); + if (LOWERVP(vp)) + VOP_REVOKE(LOWERVP(vp), ap->a_flags); + vgone(vp); + return (0); } int @@ -844,9 +895,10 @@ union_mmap(ap) struct proc *a_p; } */ *ap; { + register struct vnode *ovp = OTHERVP(ap->a_vp); - return (VOP_MMAP(OTHERVP(ap->a_vp), ap->a_fflags, - ap->a_cred, ap->a_p)); + ap->a_vp = ovp; + return (VCALL(ovp, VOFFSET(vop_mmap), ap)); } int @@ -859,19 +911,19 @@ union_fsync(ap) } */ *ap; { int error = 0; + struct proc *p = ap->a_p; struct vnode *targetvp = OTHERVP(ap->a_vp); - if (targetvp) { + if (targetvp != NULLVP) { int dolock = (targetvp == LOWERVP(ap->a_vp)); if (dolock) - VOP_LOCK(targetvp); + vn_lock(targetvp, LK_EXCLUSIVE | LK_RETRY, p); else - FIXUP(VTOUNION(ap->a_vp)); - error = VOP_FSYNC(targetvp, ap->a_cred, - ap->a_waitfor, ap->a_p); + FIXUP(VTOUNION(ap->a_vp), p); + error = VOP_FSYNC(targetvp, ap->a_cred, ap->a_waitfor, p); if (dolock) - VOP_UNLOCK(targetvp); + VOP_UNLOCK(targetvp, 0, p); } return (error); @@ -886,8 +938,10 @@ union_seek(ap) struct ucred *a_cred; } */ *ap; { + register struct vnode *ovp = OTHERVP(ap->a_vp); - return (VOP_SEEK(OTHERVP(ap->a_vp), ap->a_oldoff, ap->a_newoff, ap->a_cred)); + ap->a_vp = ovp; + return (VCALL(ovp, VOFFSET(vop_seek), ap)); } int @@ -901,34 +955,37 @@ union_remove(ap) int error; struct union_node *dun = VTOUNION(ap->a_dvp); struct union_node *un = VTOUNION(ap->a_vp); + struct componentname *cnp = ap->a_cnp; + struct proc *p = cnp->cn_proc; - if (dun->un_uppervp && un->un_uppervp) { + if (dun->un_uppervp == NULLVP) + panic("union remove: null upper vnode"); + + if (un->un_uppervp != NULLVP) { struct vnode *dvp = dun->un_uppervp; struct vnode *vp = un->un_uppervp; - FIXUP(dun); + FIXUP(dun, p); VREF(dvp); dun->un_flags |= UN_KLOCK; vput(ap->a_dvp); - FIXUP(un); + FIXUP(un, p); VREF(vp); un->un_flags |= UN_KLOCK; vput(ap->a_vp); - error = VOP_REMOVE(dvp, vp, ap->a_cnp); + if (union_dowhiteout(un, cnp->cn_cred, cnp->cn_proc)) + cnp->cn_flags |= DOWHITEOUT; + error = VOP_REMOVE(dvp, vp, cnp); if (!error) union_removed_upper(un); - - /* - * XXX: should create a whiteout here - */ } else { - /* - * XXX: should create a whiteout here - */ + FIXUP(dun, p); + error = union_mkwhiteout( + MOUNTTOUNIONMOUNT(UNIONTOV(dun)->v_mount), + dun->un_uppervp, ap->a_cnp, un->un_path); vput(ap->a_dvp); vput(ap->a_vp); - error = EROFS; } return (error); @@ -942,34 +999,51 @@ union_link(ap) struct componentname *a_cnp; } */ *ap; { - int error; - struct union_node *dun = VTOUNION(ap->a_vp); - struct union_node *un = VTOUNION(ap->a_tdvp); - - if (dun->un_uppervp && un->un_uppervp) { - struct vnode *dvp = dun->un_uppervp; - struct vnode *vp = un->un_uppervp; + int error = 0; + struct componentname *cnp = ap->a_cnp; + struct proc *p = cnp->cn_proc; + struct union_node *un; + struct vnode *vp; + struct vnode *tdvp; - FIXUP(dun); - VREF(dvp); - dun->un_flags |= UN_KLOCK; - vput(ap->a_vp); - FIXUP(un); - VREF(vp); - vrele(ap->a_tdvp); + un = VTOUNION(ap->a_tdvp); - error = VOP_LINK(dvp, vp, ap->a_cnp); + if (ap->a_tdvp->v_op != ap->a_vp->v_op) { + vp = ap->a_vp; } else { - /* - * XXX: need to copy to upper layer - * and do the link there. - */ - vput(ap->a_vp); - vrele(ap->a_tdvp); + struct union_node *tun = VTOUNION(ap->a_vp); + if (tun->un_uppervp == NULLVP) { + vn_lock(ap->a_vp, LK_EXCLUSIVE | LK_RETRY, p); + if (un->un_uppervp == tun->un_dirvp) { + un->un_flags &= ~UN_ULOCK; + VOP_UNLOCK(un->un_uppervp, 0, p); + } + error = union_copyup(tun, 1, cnp->cn_cred, p); + if (un->un_uppervp == tun->un_dirvp) { + vn_lock(un->un_uppervp, + LK_EXCLUSIVE | LK_RETRY, p); + un->un_flags |= UN_ULOCK; + } + VOP_UNLOCK(ap->a_vp, 0, p); + } + vp = tun->un_uppervp; + } + + tdvp = un->un_uppervp; + if (tdvp == NULLVP) error = EROFS; + + if (error) { + vput(ap->a_tdvp); + return (error); } - return (error); + FIXUP(un, p); + VREF(tdvp); + un->un_flags |= UN_KLOCK; + vput(ap->a_tdvp); + + return (VOP_LINK(vp, tdvp, cnp)); } int @@ -993,11 +1067,16 @@ union_rename(ap) if (fdvp->v_op == union_vnodeop_p) { /* always true */ struct union_node *un = VTOUNION(fdvp); if (un->un_uppervp == NULLVP) { - error = EROFS; + /* + * this should never happen in normal + * operation but might if there was + * a problem creating the top-level shadow + * directory. + */ + error = EXDEV; goto bad; } - FIXUP(un); fdvp = un->un_uppervp; VREF(fdvp); vrele(ap->a_fdvp); @@ -1006,11 +1085,14 @@ union_rename(ap) if (fvp->v_op == union_vnodeop_p) { /* always true */ struct union_node *un = VTOUNION(fvp); if (un->un_uppervp == NULLVP) { - error = EROFS; + /* XXX: should do a copyup */ + error = EXDEV; goto bad; } - FIXUP(un); + if (un->un_lowervp != NULLVP) + ap->a_fcnp->cn_flags |= DOWHITEOUT; + fvp = un->un_uppervp; VREF(fvp); vrele(ap->a_fvp); @@ -1019,7 +1101,13 @@ union_rename(ap) if (tdvp->v_op == union_vnodeop_p) { struct union_node *un = VTOUNION(tdvp); if (un->un_uppervp == NULLVP) { - error = EROFS; + /* + * this should never happen in normal + * operation but might if there was + * a problem creating the top-level shadow + * directory. + */ + error = EXDEV; goto bad; } @@ -1029,16 +1117,14 @@ union_rename(ap) vput(ap->a_tdvp); } - if (tvp && tvp->v_op == union_vnodeop_p) { + if (tvp != NULLVP && tvp->v_op == union_vnodeop_p) { struct union_node *un = VTOUNION(tvp); - if (un->un_uppervp == NULLVP) { - error = EROFS; - goto bad; - } tvp = un->un_uppervp; - VREF(tvp); - un->un_flags |= UN_KLOCK; + if (tvp != NULLVP) { + VREF(tvp); + un->un_flags |= UN_KLOCK; + } vput(ap->a_tvp); } @@ -1048,7 +1134,7 @@ bad: vrele(fdvp); vrele(fvp); vput(tdvp); - if (tvp) + if (tvp != NULLVP) vput(tvp); return (error); @@ -1065,27 +1151,26 @@ union_mkdir(ap) { struct union_node *un = VTOUNION(ap->a_dvp); struct vnode *dvp = un->un_uppervp; + struct componentname *cnp = ap->a_cnp; + struct proc *p = cnp->cn_proc; - if (dvp) { + if (dvp != NULLVP) { int error; struct vnode *vp; - FIXUP(un); + FIXUP(un, p); VREF(dvp); un->un_flags |= UN_KLOCK; - vput(ap->a_dvp); - error = VOP_MKDIR(dvp, &vp, ap->a_cnp, ap->a_vap); - if (error) + VOP_UNLOCK(ap->a_dvp, 0, p); + error = VOP_MKDIR(dvp, &vp, cnp, ap->a_vap); + if (error) { + vrele(ap->a_dvp); return (error); + } - error = union_allocvp( - ap->a_vpp, - ap->a_dvp->v_mount, - ap->a_dvp, - NULLVP, - ap->a_cnp, - vp, - NULLVP); + error = union_allocvp(ap->a_vpp, ap->a_dvp->v_mount, ap->a_dvp, + NULLVP, cnp, vp, NULLVP, 1); + vrele(ap->a_dvp); if (error) vput(vp); return (error); @@ -1106,34 +1191,37 @@ union_rmdir(ap) int error; struct union_node *dun = VTOUNION(ap->a_dvp); struct union_node *un = VTOUNION(ap->a_vp); + struct componentname *cnp = ap->a_cnp; + struct proc *p = cnp->cn_proc; - if (dun->un_uppervp && un->un_uppervp) { + if (dun->un_uppervp == NULLVP) + panic("union rmdir: null upper vnode"); + + if (un->un_uppervp != NULLVP) { struct vnode *dvp = dun->un_uppervp; struct vnode *vp = un->un_uppervp; - FIXUP(dun); + FIXUP(dun, p); VREF(dvp); dun->un_flags |= UN_KLOCK; vput(ap->a_dvp); - FIXUP(un); + FIXUP(un, p); VREF(vp); un->un_flags |= UN_KLOCK; vput(ap->a_vp); + if (union_dowhiteout(un, cnp->cn_cred, cnp->cn_proc)) + cnp->cn_flags |= DOWHITEOUT; error = VOP_RMDIR(dvp, vp, ap->a_cnp); if (!error) union_removed_upper(un); - - /* - * XXX: should create a whiteout here - */ } else { - /* - * XXX: should create a whiteout here - */ + FIXUP(dun, p); + error = union_mkwhiteout( + MOUNTTOUNIONMOUNT(UNIONTOV(dun)->v_mount), + dun->un_uppervp, ap->a_cnp, un->un_path); vput(ap->a_dvp); vput(ap->a_vp); - error = EROFS; } return (error); @@ -1151,17 +1239,18 @@ union_symlink(ap) { struct union_node *un = VTOUNION(ap->a_dvp); struct vnode *dvp = un->un_uppervp; + struct componentname *cnp = ap->a_cnp; + struct proc *p = cnp->cn_proc; - if (dvp) { + if (dvp != NULLVP) { int error; struct vnode *vp; - FIXUP(un); + FIXUP(un, p); VREF(dvp); un->un_flags |= UN_KLOCK; vput(ap->a_dvp); - error = VOP_SYMLINK(dvp, &vp, ap->a_cnp, - ap->a_vap, ap->a_target); + error = VOP_SYMLINK(dvp, &vp, cnp, ap->a_vap, ap->a_target); *ap->a_vpp = NULLVP; return (error); } @@ -1183,17 +1272,21 @@ union_readdir(ap) struct vnode *a_vp; struct uio *a_uio; struct ucred *a_cred; + int *a_eofflag; + u_long *a_cookies; + int a_ncookies; } */ *ap; { - int error = 0; struct union_node *un = VTOUNION(ap->a_vp); + struct vnode *uvp = un->un_uppervp; + struct proc *p = ap->a_uio->uio_procp; - if (un->un_uppervp) { - FIXUP(un); - error = VOP_READDIR(un->un_uppervp, ap->a_uio, ap->a_cred, NULL, NULL, NULL); - } + if (uvp == NULLVP) + return (0); - return (error); + FIXUP(un, p); + ap->a_vp = uvp; + return (VCALL(uvp, VOFFSET(vop_readdir), ap)); } int @@ -1205,16 +1298,19 @@ union_readlink(ap) } */ *ap; { int error; + struct uio *uio = ap->a_uio; + struct proc *p = uio->uio_procp; struct vnode *vp = OTHERVP(ap->a_vp); int dolock = (vp == LOWERVP(ap->a_vp)); if (dolock) - VOP_LOCK(vp); + vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, p); else - FIXUP(VTOUNION(ap->a_vp)); - error = VOP_READLINK(vp, ap->a_uio, ap->a_cred); + FIXUP(VTOUNION(ap->a_vp), p); + ap->a_vp = vp; + error = VCALL(vp, VOFFSET(vop_readlink), ap); if (dolock) - VOP_UNLOCK(vp); + VOP_UNLOCK(vp, 0, p); return (error); } @@ -1227,6 +1323,8 @@ union_abortop(ap) } */ *ap; { int error; + struct componentname *cnp = ap->a_cnp; + struct proc *p = cnp->cn_proc; struct vnode *vp = OTHERVP(ap->a_dvp); struct union_node *un = VTOUNION(ap->a_dvp); int islocked = un->un_flags & UN_LOCKED; @@ -1234,13 +1332,14 @@ union_abortop(ap) if (islocked) { if (dolock) - VOP_LOCK(vp); + vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, p); else - FIXUP(VTOUNION(ap->a_dvp)); + FIXUP(VTOUNION(ap->a_dvp), p); } - error = VOP_ABORTOP(vp, ap->a_cnp); + ap->a_dvp = vp; + error = VCALL(vp, VOFFSET(vop_abortop), ap); if (islocked && dolock) - VOP_UNLOCK(vp); + VOP_UNLOCK(vp, 0, p); return (error); } @@ -1249,8 +1348,13 @@ int union_inactive(ap) struct vop_inactive_args /* { struct vnode *a_vp; + struct proc *a_p; } */ *ap; { + struct vnode *vp = ap->a_vp; + struct proc *p = ap->a_p; + struct union_node *un = VTOUNION(vp); + struct vnode **vpp; /* * Do nothing (and _don't_ bypass). @@ -1265,12 +1369,17 @@ union_inactive(ap) * That's too much work for now. */ -#ifdef UNION_DIAGNOSTIC - struct union_node *un = VTOUNION(ap->a_vp); + if (un->un_dircache != 0) { + for (vpp = un->un_dircache; *vpp != NULLVP; vpp++) + vrele(*vpp); + free(un->un_dircache, M_TEMP); + un->un_dircache = 0; + } - if (un->un_flags & UN_LOCKED) - panic("union: inactivating locked node"); -#endif + VOP_UNLOCK(vp, 0, p); + + if ((un->un_flags & UN_CACHED) == 0) + vgone(vp); return (0); } @@ -1292,24 +1401,39 @@ union_lock(ap) struct vop_lock_args *ap; { struct vnode *vp = ap->a_vp; + struct proc *p = ap->a_p; + int flags = ap->a_flags; struct union_node *un; + int error; -start: - while (vp->v_flag & VXLOCK) { - vp->v_flag |= VXWANT; - (void) tsleep((caddr_t)vp, PINOD, "unnlk1", 0); - } + vop_nolock(ap); + /* + * Need to do real lockmgr-style locking here. + * in the mean time, draining won't work quite right, + * which could lead to a few race conditions. + * the following test was here, but is not quite right, we + * still need to take the lock: + if ((flags & LK_TYPE_MASK) == LK_DRAIN) + return (0); + */ + flags &= ~LK_INTERLOCK; +start: un = VTOUNION(vp); - if (un->un_uppervp) { - if ((un->un_flags & UN_ULOCK) == 0) { + if (un->un_uppervp != NULLVP) { + if (((un->un_flags & UN_ULOCK) == 0) && + (vp->v_usecount != 0)) { + error = vn_lock(un->un_uppervp, flags, p); + if (error) + return (error); un->un_flags |= UN_ULOCK; - VOP_LOCK(un->un_uppervp); } #ifdef DIAGNOSTIC - if (un->un_flags & UN_KLOCK) - panic("union: dangling upper lock"); + if (un->un_flags & UN_KLOCK) { + vprint("union: dangling klock", vp); + panic("union: dangling upper lock (%lx)", vp); + } #endif } @@ -1320,7 +1444,7 @@ start: panic("union: locking against myself"); #endif un->un_flags |= UN_WANT; - (void) tsleep((caddr_t) &un->un_flags, PINOD, "unnlk2", 0); + tsleep((caddr_t)&un->un_flags, PINOD, "unionlk2", 0); goto start; } @@ -1335,11 +1459,27 @@ start: return (0); } +/* + * When operations want to vput() a union node yet retain a lock on + * the upper vnode (say, to do some further operations like link(), + * mkdir(), ...), they set UN_KLOCK on the union node, then call + * vput() which calls VOP_UNLOCK() and comes here. union_unlock() + * unlocks the union node (leaving the upper vnode alone), clears the + * KLOCK flag, and then returns to vput(). The caller then does whatever + * is left to do with the upper vnode, and ensures that it gets unlocked. + * + * If UN_KLOCK isn't set, then the upper vnode is unlocked here. + */ int union_unlock(ap) - struct vop_lock_args *ap; + struct vop_unlock_args /* { + struct vnode *a_vp; + int a_flags; + struct proc *a_p; + } */ *ap; { struct union_node *un = VTOUNION(ap->a_vp); + struct proc *p = ap->a_p; #ifdef DIAGNOSTIC if ((un->un_flags & UN_LOCKED) == 0) @@ -1352,7 +1492,7 @@ union_unlock(ap) un->un_flags &= ~UN_LOCKED; if ((un->un_flags & (UN_ULOCK|UN_KLOCK)) == UN_ULOCK) - VOP_UNLOCK(un->un_uppervp); + VOP_UNLOCK(un->un_uppervp, 0, p); un->un_flags &= ~(UN_ULOCK|UN_KLOCK); @@ -1364,6 +1504,7 @@ union_unlock(ap) #ifdef DIAGNOSTIC un->un_pid = 0; #endif + vop_nounlock(ap); return (0); } @@ -1380,16 +1521,18 @@ union_bmap(ap) } */ *ap; { int error; + struct proc *p = curproc; /* XXX */ struct vnode *vp = OTHERVP(ap->a_vp); int dolock = (vp == LOWERVP(ap->a_vp)); if (dolock) - VOP_LOCK(vp); + vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, p); else - FIXUP(VTOUNION(ap->a_vp)); - error = VOP_BMAP(vp, ap->a_bn, ap->a_vpp, ap->a_bnp, ap->a_runp, ap->a_runb); + FIXUP(VTOUNION(ap->a_vp), p); + ap->a_vp = vp; + error = VCALL(vp, VOFFSET(vop_bmap), ap); if (dolock) - VOP_UNLOCK(vp); + VOP_UNLOCK(vp, 0, p); return (error); } @@ -1404,6 +1547,11 @@ union_print(ap) printf("\ttag VT_UNION, vp=%p, uppervp=%p, lowervp=%p\n", vp, UPPERVP(vp), LOWERVP(vp)); + if (UPPERVP(vp) != NULLVP) + vprint("union: upper", UPPERVP(vp)); + if (LOWERVP(vp) != NULLVP) + vprint("union: lower", LOWERVP(vp)); + return (0); } @@ -1426,16 +1574,18 @@ union_pathconf(ap) } */ *ap; { int error; + struct proc *p = curproc; /* XXX */ struct vnode *vp = OTHERVP(ap->a_vp); int dolock = (vp == LOWERVP(ap->a_vp)); if (dolock) - VOP_LOCK(vp); + vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, p); else - FIXUP(VTOUNION(ap->a_vp)); - error = VOP_PATHCONF(vp, ap->a_name, ap->a_retval); + FIXUP(VTOUNION(ap->a_vp), p); + ap->a_vp = vp; + error = VCALL(vp, VOFFSET(vop_pathconf), ap); if (dolock) - VOP_UNLOCK(vp); + VOP_UNLOCK(vp, 0, p); return (error); } @@ -1450,9 +1600,10 @@ union_advlock(ap) int a_flags; } */ *ap; { + register struct vnode *ovp = OTHERVP(ap->a_vp); - return (VOP_ADVLOCK(OTHERVP(ap->a_vp), ap->a_id, ap->a_op, - ap->a_fl, ap->a_flags)); + ap->a_vp = ovp; + return (VCALL(ovp, VOFFSET(vop_advlock), ap)); } @@ -1496,6 +1647,7 @@ struct vnodeopv_entry_desc union_vnodeop_entries[] = { { &vop_default_desc, (vop_t *)vn_default_error }, { &vop_lookup_desc, (vop_t *)union_lookup }, /* lookup */ { &vop_create_desc, (vop_t *)union_create }, /* create */ + { &vop_whiteout_desc, (vop_t *)union_whiteout }, /* whiteout */ { &vop_mknod_desc, (vop_t *)union_mknod }, /* mknod */ { &vop_open_desc, (vop_t *)union_open }, /* open */ { &vop_close_desc, (vop_t *)union_close }, /* close */ @@ -1504,8 +1656,10 @@ struct vnodeopv_entry_desc union_vnodeop_entries[] = { { &vop_setattr_desc, (vop_t *)union_setattr }, /* setattr */ { &vop_read_desc, (vop_t *)union_read }, /* read */ { &vop_write_desc, (vop_t *)union_write }, /* write */ + { &vop_lease_desc, (vop_t *)union_lease }, /* lease */ { &vop_ioctl_desc, (vop_t *)union_ioctl }, /* ioctl */ { &vop_select_desc, (vop_t *)union_select }, /* select */ + { &vop_revoke_desc, (vop_t *)union_revoke }, /* revoke */ { &vop_mmap_desc, (vop_t *)union_mmap }, /* mmap */ { &vop_fsync_desc, (vop_t *)union_fsync }, /* fsync */ { &vop_seek_desc, (vop_t *)union_seek }, /* seek */ |