2 * Copyright (c) 1999-2004 Poul-Henning Kamp
3 * Copyright (c) 1999 Michael Smith
4 * Copyright (c) 1989, 1993
5 * The Regents of the University of California. All rights reserved.
6 * (c) UNIX System Laboratories, Inc.
7 * All or some portions of this file are derived from material licensed
8 * to the University of California by American Telephone and Telegraph
9 * Co. or Unix System Laboratories, Inc. and are reproduced herein with
10 * the permission of UNIX System Laboratories, Inc.
12 * Redistribution and use in source and binary forms, with or without
13 * modification, are permitted provided that the following conditions
15 * 1. Redistributions of source code must retain the above copyright
16 * notice, this list of conditions and the following disclaimer.
17 * 2. Redistributions in binary form must reproduce the above copyright
18 * notice, this list of conditions and the following disclaimer in the
19 * documentation and/or other materials provided with the distribution.
20 * 4. Neither the name of the University nor the names of its contributors
21 * may be used to endorse or promote products derived from this software
22 * without specific prior written permission.
24 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
25 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
26 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
27 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
28 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
29 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
30 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
31 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
32 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
33 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
37 #include <sys/cdefs.h>
38 __FBSDID("$FreeBSD$");
40 #include <sys/param.h>
42 #include <sys/clock.h>
44 #include <sys/kernel.h>
45 #include <sys/libkern.h>
46 #include <sys/malloc.h>
47 #include <sys/mount.h>
48 #include <sys/mutex.h>
49 #include <sys/namei.h>
52 #include <sys/filedesc.h>
53 #include <sys/reboot.h>
54 #include <sys/syscallsubr.h>
55 #include <sys/sysproto.h>
57 #include <sys/sysctl.h>
58 #include <sys/sysent.h>
59 #include <sys/systm.h>
60 #include <sys/vnode.h>
63 #include <geom/geom.h>
65 #include <machine/stdarg.h>
67 #include <security/audit/audit.h>
68 #include <security/mac/mac_framework.h>
70 #include "opt_rootdevname.h"
78 #define ROOTNAME "root_device"
79 #define VFS_MOUNTARG_SIZE_MAX (1024 * 64)
81 static int vfs_domount(struct thread *td, const char *fstype,
82 char *fspath, int fsflags, void *fsdata);
83 static struct mount *vfs_mount_alloc(struct vnode *dvp, struct vfsconf *vfsp,
84 const char *fspath, struct thread *td);
85 static int vfs_mountroot_ask(void);
86 static int vfs_mountroot_try(const char *mountfrom);
87 static int vfs_donmount(struct thread *td, int fsflags,
88 struct uio *fsoptions);
89 static void free_mntarg(struct mntarg *ma);
90 static int vfs_getopt_pos(struct vfsoptlist *opts, const char *name);
92 static int usermount = 0;
93 SYSCTL_INT(_vfs, OID_AUTO, usermount, CTLFLAG_RW, &usermount, 0,
94 "Unprivileged users may mount and unmount file systems");
96 MALLOC_DEFINE(M_MOUNT, "mount", "vfs mount structure");
97 MALLOC_DEFINE(M_VNODE_MARKER, "vnodemarker", "vnode marker");
98 static uma_zone_t mount_zone;
100 /* List of mounted filesystems. */
101 struct mntlist mountlist = TAILQ_HEAD_INITIALIZER(mountlist);
103 /* For any iteration/modification of mountlist */
104 struct mtx mountlist_mtx;
105 MTX_SYSINIT(mountlist, &mountlist_mtx, "mountlist", MTX_DEF);
107 TAILQ_HEAD(vfsoptlist, vfsopt);
109 TAILQ_ENTRY(vfsopt) link;
116 * The vnode of the system's root (/ in the filesystem, without chroot
119 struct vnode *rootvnode;
122 * The root filesystem is detailed in the kernel environment variable
123 * vfs.root.mountfrom, which is expected to be in the general format
126 * vfsname := the name of a VFS known to the kernel and capable
127 * of being mounted as root
128 * path := disk device name or other data used by the filesystem
129 * to locate its physical store
133 * Global opts, taken by all filesystems
135 static const char *global_opts[] = {
149 * The root specifiers we will try if RB_CDROM is specified.
151 static char *cdrom_rootdevnames[] = {
157 /* legacy find-root code */
158 char *rootdevnames[2] = {NULL, NULL};
160 # define ROOTDEVNAME NULL
162 static const char *ctrootdevname = ROOTDEVNAME;
165 * ---------------------------------------------------------------------
166 * Functions for building and sanitizing the mount options
169 /* Remove one mount option. */
171 vfs_freeopt(struct vfsoptlist *opts, struct vfsopt *opt)
174 TAILQ_REMOVE(opts, opt, link);
175 free(opt->name, M_MOUNT);
176 if (opt->value != NULL)
177 free(opt->value, M_MOUNT);
179 else if (opt->len != 0)
180 panic("%s: mount option with NULL value but length != 0",
186 /* Release all resources related to the mount options. */
188 vfs_freeopts(struct vfsoptlist *opts)
192 while (!TAILQ_EMPTY(opts)) {
193 opt = TAILQ_FIRST(opts);
194 vfs_freeopt(opts, opt);
200 vfs_deleteopt(struct vfsoptlist *opts, const char *name)
202 struct vfsopt *opt, *temp;
204 TAILQ_FOREACH_SAFE(opt, opts, link, temp) {
205 if (strcmp(opt->name, name) == 0)
206 vfs_freeopt(opts, opt);
211 * Check if options are equal (with or without the "no" prefix).
214 vfs_equalopts(const char *opt1, const char *opt2)
217 /* "opt" vs. "opt" or "noopt" vs. "noopt" */
218 if (strcmp(opt1, opt2) == 0)
220 /* "noopt" vs. "opt" */
221 if (strncmp(opt1, "no", 2) == 0 && strcmp(opt1 + 2, opt2) == 0)
223 /* "opt" vs. "noopt" */
224 if (strncmp(opt2, "no", 2) == 0 && strcmp(opt1, opt2 + 2) == 0)
230 * If a mount option is specified several times,
231 * (with or without the "no" prefix) only keep
232 * the last occurence of it.
235 vfs_sanitizeopts(struct vfsoptlist *opts)
237 struct vfsopt *opt, *opt2, *tmp;
239 TAILQ_FOREACH_REVERSE(opt, opts, vfsoptlist, link) {
240 opt2 = TAILQ_PREV(opt, vfsoptlist, link);
241 while (opt2 != NULL) {
242 if (vfs_equalopts(opt->name, opt2->name)) {
243 tmp = TAILQ_PREV(opt2, vfsoptlist, link);
244 vfs_freeopt(opts, opt2);
247 opt2 = TAILQ_PREV(opt2, vfsoptlist, link);
254 * Build a linked list of mount options from a struct uio.
257 vfs_buildopts(struct uio *auio, struct vfsoptlist **options)
259 struct vfsoptlist *opts;
262 unsigned int i, iovcnt;
263 int error, namelen, optlen;
265 opts = malloc(sizeof(struct vfsoptlist), M_MOUNT, M_WAITOK);
268 iovcnt = auio->uio_iovcnt;
269 for (i = 0; i < iovcnt; i += 2) {
270 opt = malloc(sizeof(struct vfsopt), M_MOUNT, M_WAITOK);
271 namelen = auio->uio_iov[i].iov_len;
272 optlen = auio->uio_iov[i + 1].iov_len;
273 opt->name = malloc(namelen, M_MOUNT, M_WAITOK);
278 * Do this early, so jumps to "bad" will free the current
281 TAILQ_INSERT_TAIL(opts, opt, link);
282 memused += sizeof(struct vfsopt) + optlen + namelen;
285 * Avoid consuming too much memory, and attempts to overflow
288 if (memused > VFS_MOUNTARG_SIZE_MAX ||
289 optlen > VFS_MOUNTARG_SIZE_MAX ||
290 namelen > VFS_MOUNTARG_SIZE_MAX) {
295 if (auio->uio_segflg == UIO_SYSSPACE) {
296 bcopy(auio->uio_iov[i].iov_base, opt->name, namelen);
298 error = copyin(auio->uio_iov[i].iov_base, opt->name,
303 /* Ensure names are null-terminated strings. */
304 if (opt->name[namelen - 1] != '\0') {
310 opt->value = malloc(optlen, M_MOUNT, M_WAITOK);
311 if (auio->uio_segflg == UIO_SYSSPACE) {
312 bcopy(auio->uio_iov[i + 1].iov_base, opt->value,
315 error = copyin(auio->uio_iov[i + 1].iov_base,
322 vfs_sanitizeopts(opts);
331 * Merge the old mount options with the new ones passed
332 * in the MNT_UPDATE case.
335 vfs_mergeopts(struct vfsoptlist *toopts, struct vfsoptlist *opts)
337 struct vfsopt *opt, *opt2, *new;
339 TAILQ_FOREACH(opt, opts, link) {
341 * Check that this option hasn't been redefined
342 * nor cancelled with a "no" mount option.
344 opt2 = TAILQ_FIRST(toopts);
345 while (opt2 != NULL) {
346 if (strcmp(opt2->name, opt->name) == 0)
348 if (strncmp(opt2->name, "no", 2) == 0 &&
349 strcmp(opt2->name + 2, opt->name) == 0) {
350 vfs_freeopt(toopts, opt2);
353 opt2 = TAILQ_NEXT(opt2, link);
355 /* We want this option, duplicate it. */
356 new = malloc(sizeof(struct vfsopt), M_MOUNT, M_WAITOK);
357 new->name = malloc(strlen(opt->name) + 1, M_MOUNT, M_WAITOK);
358 strcpy(new->name, opt->name);
360 new->value = malloc(opt->len, M_MOUNT, M_WAITOK);
361 bcopy(opt->value, new->value, opt->len);
366 TAILQ_INSERT_TAIL(toopts, new, link);
373 * Mount a filesystem.
378 struct nmount_args /* {
390 AUDIT_ARG(fflags, uap->flags);
392 /* Kick out MNT_ROOTFS early as it is legal internally */
393 if (uap->flags & MNT_ROOTFS)
396 iovcnt = uap->iovcnt;
398 * Check that we have an even number of iovec's
399 * and that we have at least two options.
401 if ((iovcnt & 1) || (iovcnt < 4))
404 error = copyinuio(uap->iovp, iovcnt, &auio);
408 for (i = 0; i < iovcnt; i++) {
409 if (iov->iov_len > MMAXOPTIONLEN) {
415 error = vfs_donmount(td, uap->flags, auio);
422 * ---------------------------------------------------------------------
423 * Various utility functions
427 vfs_ref(struct mount *mp)
436 vfs_rel(struct mount *mp)
445 mount_init(void *mem, int size, int flags)
449 mp = (struct mount *)mem;
450 mtx_init(&mp->mnt_mtx, "struct mount mtx", NULL, MTX_DEF);
451 lockinit(&mp->mnt_lock, PVFS, "vfslock", 0, 0);
456 mount_fini(void *mem, int size)
460 mp = (struct mount *)mem;
461 lockdestroy(&mp->mnt_lock);
462 mtx_destroy(&mp->mnt_mtx);
466 * Allocate and initialize the mount point struct.
468 static struct mount *
469 vfs_mount_alloc(struct vnode *vp, struct vfsconf *vfsp,
470 const char *fspath, struct thread *td)
474 mp = uma_zalloc(mount_zone, M_WAITOK);
475 bzero(&mp->mnt_startzero,
476 __rangeof(struct mount, mnt_startzero, mnt_endzero));
477 TAILQ_INIT(&mp->mnt_nvnodelist);
478 mp->mnt_nvnodelistsize = 0;
480 (void) vfs_busy(mp, LK_NOWAIT, 0, td);
481 mp->mnt_op = vfsp->vfc_vfsops;
483 vfsp->vfc_refcount++; /* XXX Unlocked */
484 mp->mnt_stat.f_type = vfsp->vfc_typenum;
486 mp->mnt_flag |= vfsp->vfc_flags & MNT_VISFLAGMASK;
489 strlcpy(mp->mnt_stat.f_fstypename, vfsp->vfc_name, MFSNAMELEN);
490 mp->mnt_vnodecovered = vp;
491 mp->mnt_cred = crdup(td->td_ucred);
492 mp->mnt_stat.f_owner = td->td_ucred->cr_uid;
493 strlcpy(mp->mnt_stat.f_mntonname, fspath, MNAMELEN);
494 mp->mnt_iosize_max = DFLTPHYS;
497 mac_create_mount(td->td_ucred, mp);
499 arc4rand(&mp->mnt_hashseed, sizeof mp->mnt_hashseed, 0);
504 * Destroy the mount struct previously allocated by vfs_mount_alloc().
507 vfs_mount_destroy(struct mount *mp)
512 for (i = 0; mp->mnt_ref && i < 3; i++)
513 msleep(mp, MNT_MTX(mp), PVFS, "mntref", hz);
515 * This will always cause a 3 second delay in rebooting due to
516 * refs on the root mountpoint that never go away. Most of these
517 * are held by init which never exits.
519 if (i == 3 && (!rebooting || bootverbose))
520 printf("Mount point %s had %d dangling refs\n",
521 mp->mnt_stat.f_mntonname, mp->mnt_ref);
522 if (mp->mnt_holdcnt != 0) {
523 printf("Waiting for mount point to be unheld\n");
524 while (mp->mnt_holdcnt != 0) {
525 mp->mnt_holdcntwaiters++;
526 msleep(&mp->mnt_holdcnt, MNT_MTX(mp),
527 PZERO, "mntdestroy", 0);
528 mp->mnt_holdcntwaiters--;
530 printf("mount point unheld\n");
532 if (mp->mnt_writeopcount > 0) {
533 printf("Waiting for mount point write ops\n");
534 while (mp->mnt_writeopcount > 0) {
535 mp->mnt_kern_flag |= MNTK_SUSPEND;
536 msleep(&mp->mnt_writeopcount,
538 PZERO, "mntdestroy2", 0);
540 printf("mount point write ops completed\n");
542 if (mp->mnt_secondary_writes > 0) {
543 printf("Waiting for mount point secondary write ops\n");
544 while (mp->mnt_secondary_writes > 0) {
545 mp->mnt_kern_flag |= MNTK_SUSPEND;
546 msleep(&mp->mnt_secondary_writes,
548 PZERO, "mntdestroy3", 0);
550 printf("mount point secondary write ops completed\n");
553 mp->mnt_vfc->vfc_refcount--;
554 if (!TAILQ_EMPTY(&mp->mnt_nvnodelist)) {
557 TAILQ_FOREACH(vp, &mp->mnt_nvnodelist, v_nmntvnodes)
559 panic("unmount: dangling vnode");
562 if (mp->mnt_kern_flag & MNTK_MWAIT)
564 if (mp->mnt_writeopcount != 0)
565 panic("vfs_mount_destroy: nonzero writeopcount");
566 if (mp->mnt_secondary_writes != 0)
567 panic("vfs_mount_destroy: nonzero secondary_writes");
568 if (mp->mnt_nvnodelistsize != 0)
569 panic("vfs_mount_destroy: nonzero nvnodelistsize");
570 mp->mnt_writeopcount = -1000;
571 mp->mnt_nvnodelistsize = -1000;
572 mp->mnt_secondary_writes = -1000;
575 mac_destroy_mount(mp);
577 if (mp->mnt_opt != NULL)
578 vfs_freeopts(mp->mnt_opt);
579 crfree(mp->mnt_cred);
580 uma_zfree(mount_zone, mp);
584 vfs_donmount(struct thread *td, int fsflags, struct uio *fsoptions)
586 struct vfsoptlist *optlist;
587 struct vfsopt *opt, *noro_opt;
588 char *fstype, *fspath, *errmsg;
589 int error, fstypelen, fspathlen, errmsg_len, errmsg_pos;
590 int has_rw, has_noro;
598 error = vfs_buildopts(fsoptions, &optlist);
602 if (vfs_getopt(optlist, "errmsg", (void **)&errmsg, &errmsg_len) == 0)
603 errmsg_pos = vfs_getopt_pos(optlist, "errmsg");
606 * We need these two options before the others,
607 * and they are mandatory for any filesystem.
608 * Ensure they are NUL terminated as well.
611 error = vfs_getopt(optlist, "fstype", (void **)&fstype, &fstypelen);
612 if (error || fstype[fstypelen - 1] != '\0') {
615 strncpy(errmsg, "Invalid fstype", errmsg_len);
619 error = vfs_getopt(optlist, "fspath", (void **)&fspath, &fspathlen);
620 if (error || fspath[fspathlen - 1] != '\0') {
623 strncpy(errmsg, "Invalid fspath", errmsg_len);
628 * We need to see if we have the "update" option
629 * before we call vfs_domount(), since vfs_domount() has special
630 * logic based on MNT_UPDATE. This is very important
631 * when we want to update the root filesystem.
633 TAILQ_FOREACH(opt, optlist, link) {
634 if (strcmp(opt->name, "update") == 0)
635 fsflags |= MNT_UPDATE;
636 else if (strcmp(opt->name, "async") == 0)
637 fsflags |= MNT_ASYNC;
638 else if (strcmp(opt->name, "force") == 0)
639 fsflags |= MNT_FORCE;
640 else if (strcmp(opt->name, "multilabel") == 0)
641 fsflags |= MNT_MULTILABEL;
642 else if (strcmp(opt->name, "noasync") == 0)
643 fsflags &= ~MNT_ASYNC;
644 else if (strcmp(opt->name, "noatime") == 0)
645 fsflags |= MNT_NOATIME;
646 else if (strcmp(opt->name, "noclusterr") == 0)
647 fsflags |= MNT_NOCLUSTERR;
648 else if (strcmp(opt->name, "noclusterw") == 0)
649 fsflags |= MNT_NOCLUSTERW;
650 else if (strcmp(opt->name, "noexec") == 0)
651 fsflags |= MNT_NOEXEC;
652 else if (strcmp(opt->name, "nosuid") == 0)
653 fsflags |= MNT_NOSUID;
654 else if (strcmp(opt->name, "nosymfollow") == 0)
655 fsflags |= MNT_NOSYMFOLLOW;
656 else if (strcmp(opt->name, "noro") == 0) {
657 fsflags &= ~MNT_RDONLY;
660 else if (strcmp(opt->name, "rw") == 0) {
661 fsflags &= ~MNT_RDONLY;
664 else if (strcmp(opt->name, "ro") == 0 ||
665 strcmp(opt->name, "rdonly") == 0)
666 fsflags |= MNT_RDONLY;
667 else if (strcmp(opt->name, "snapshot") == 0)
668 fsflags |= MNT_SNAPSHOT;
669 else if (strcmp(opt->name, "suiddir") == 0)
670 fsflags |= MNT_SUIDDIR;
671 else if (strcmp(opt->name, "sync") == 0)
672 fsflags |= MNT_SYNCHRONOUS;
673 else if (strcmp(opt->name, "union") == 0)
674 fsflags |= MNT_UNION;
678 * If "rw" was specified as a mount option, and we
679 * are trying to update a mount-point from "ro" to "rw",
680 * we need a mount option "noro", since in vfs_mergeopts(),
681 * "noro" will cancel "ro", but "rw" will not do anything.
683 if (has_rw && !has_noro) {
684 noro_opt = malloc(sizeof(struct vfsopt), M_MOUNT, M_WAITOK);
685 noro_opt->name = strdup("noro", M_MOUNT);
686 noro_opt->value = NULL;
688 TAILQ_INSERT_TAIL(optlist, noro_opt, link);
692 * Be ultra-paranoid about making sure the type and fspath
693 * variables will fit in our mp buffers, including the
696 if (fstypelen >= MFSNAMELEN - 1 || fspathlen >= MNAMELEN - 1) {
697 error = ENAMETOOLONG;
702 error = vfs_domount(td, fstype, fspath, fsflags, optlist);
705 /* copyout the errmsg */
706 if (errmsg_pos != -1 && ((2 * errmsg_pos + 1) < fsoptions->uio_iovcnt)
707 && errmsg_len > 0 && errmsg != NULL) {
708 if (fsoptions->uio_segflg == UIO_SYSSPACE) {
710 fsoptions->uio_iov[2 * errmsg_pos + 1].iov_base,
711 fsoptions->uio_iov[2 * errmsg_pos + 1].iov_len);
714 fsoptions->uio_iov[2 * errmsg_pos + 1].iov_base,
715 fsoptions->uio_iov[2 * errmsg_pos + 1].iov_len);
720 vfs_freeopts(optlist);
727 #ifndef _SYS_SYSPROTO_H_
739 struct mount_args /* {
747 struct vfsconf *vfsp = NULL;
748 struct mntarg *ma = NULL;
751 AUDIT_ARG(fflags, uap->flags);
753 /* Kick out MNT_ROOTFS early as it is legal internally */
754 uap->flags &= ~MNT_ROOTFS;
756 fstype = malloc(MFSNAMELEN, M_TEMP, M_WAITOK);
757 error = copyinstr(uap->type, fstype, MFSNAMELEN, NULL);
759 free(fstype, M_TEMP);
763 AUDIT_ARG(text, fstype);
765 vfsp = vfs_byname_kld(fstype, td, &error);
766 free(fstype, M_TEMP);
771 if (vfsp->vfc_vfsops->vfs_cmount == NULL) {
776 ma = mount_argsu(ma, "fstype", uap->type, MNAMELEN);
777 ma = mount_argsu(ma, "fspath", uap->path, MNAMELEN);
778 ma = mount_argb(ma, uap->flags & MNT_RDONLY, "noro");
779 ma = mount_argb(ma, !(uap->flags & MNT_NOSUID), "nosuid");
780 ma = mount_argb(ma, !(uap->flags & MNT_NOEXEC), "noexec");
782 error = vfsp->vfc_vfsops->vfs_cmount(ma, uap->data, uap->flags, td);
789 * vfs_domount(): actually attempt a filesystem mount.
793 struct thread *td, /* Calling thread. */
794 const char *fstype, /* Filesystem type. */
795 char *fspath, /* Mount path. */
796 int fsflags, /* Flags common to all filesystems. */
797 void *fsdata /* Options local to the filesystem. */
802 struct vfsconf *vfsp;
803 struct export_args export;
808 mtx_assert(&Giant, MA_OWNED);
810 * Be ultra-paranoid about making sure the type and fspath
811 * variables will fit in our mp buffers, including the
814 if (strlen(fstype) >= MFSNAMELEN || strlen(fspath) >= MNAMELEN)
815 return (ENAMETOOLONG);
817 if (jailed(td->td_ucred) || usermount == 0) {
818 if ((error = priv_check(td, PRIV_VFS_MOUNT)) != 0)
823 * Do not allow NFS export or MNT_SUIDDIR by unprivileged users.
825 if (fsflags & MNT_EXPORTED) {
826 error = priv_check(td, PRIV_VFS_MOUNT_EXPORTED);
830 if (fsflags & MNT_SUIDDIR) {
831 error = priv_check(td, PRIV_VFS_MOUNT_SUIDDIR);
836 * Silently enforce MNT_NOSUID and MNT_USER for unprivileged users.
838 if ((fsflags & (MNT_NOSUID | MNT_USER)) != (MNT_NOSUID | MNT_USER)) {
839 if (priv_check(td, PRIV_VFS_MOUNT_NONUSER) != 0)
840 fsflags |= MNT_NOSUID | MNT_USER;
843 /* Load KLDs before we lock the covered vnode to avoid reversals. */
845 if ((fsflags & MNT_UPDATE) == 0) {
846 /* Don't try to load KLDs if we're mounting the root. */
847 if (fsflags & MNT_ROOTFS)
848 vfsp = vfs_byname(fstype);
850 vfsp = vfs_byname_kld(fstype, td, &error);
855 * Get vnode to be covered
857 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1, UIO_SYSSPACE,
859 if ((error = namei(&nd)) != 0)
861 NDFREE(&nd, NDF_ONLY_PNBUF);
863 if (fsflags & MNT_UPDATE) {
864 if ((vp->v_vflag & VV_ROOT) == 0) {
872 * We only allow the filesystem to be reloaded if it
873 * is currently mounted read-only.
875 if ((fsflags & MNT_RELOAD) &&
876 ((mp->mnt_flag & MNT_RDONLY) == 0)) {
879 return (EOPNOTSUPP); /* Needs translation */
883 * Only privileged root, or (if MNT_USER is set) the user that
884 * did the original mount is permitted to update it.
886 error = vfs_suser(mp, td);
891 if (vfs_busy(mp, LK_NOWAIT, 0, td)) {
896 if ((vp->v_iflag & VI_MOUNT) != 0 ||
897 vp->v_mountedhere != NULL) {
903 vp->v_iflag |= VI_MOUNT;
906 mp->mnt_flag |= fsflags &
907 (MNT_RELOAD | MNT_FORCE | MNT_UPDATE | MNT_SNAPSHOT | MNT_ROOTFS);
909 VOP_UNLOCK(vp, 0, td);
910 mp->mnt_optnew = fsdata;
911 vfs_mergeopts(mp->mnt_optnew, mp->mnt_opt);
914 * If the user is not root, ensure that they own the directory
915 * onto which we are attempting to mount.
917 error = VOP_GETATTR(vp, &va, td->td_ucred, td);
922 if (va.va_uid != td->td_ucred->cr_uid) {
923 error = priv_check_cred(td->td_ucred, PRIV_VFS_ADMIN,
930 error = vinvalbuf(vp, V_SAVE, td, 0, 0);
935 if (vp->v_type != VDIR) {
940 if ((vp->v_iflag & VI_MOUNT) != 0 ||
941 vp->v_mountedhere != NULL) {
946 vp->v_iflag |= VI_MOUNT;
950 * Allocate and initialize the filesystem.
952 mp = vfs_mount_alloc(vp, vfsp, fspath, td);
953 VOP_UNLOCK(vp, 0, td);
955 /* XXXMAC: pass to vfs_mount_alloc? */
956 mp->mnt_optnew = fsdata;
960 * Set the mount level flags.
963 mp->mnt_flag = (mp->mnt_flag & ~MNT_UPDATEMASK) |
964 (fsflags & (MNT_UPDATEMASK | MNT_FORCE | MNT_ROOTFS |
966 if ((mp->mnt_flag & MNT_ASYNC) == 0)
967 mp->mnt_kern_flag &= ~MNTK_ASYNC;
970 * Mount the filesystem.
971 * XXX The final recipients of VFS_MOUNT just overwrite the ndp they
972 * get. No freeing of cn_pnbuf.
974 error = VFS_MOUNT(mp, td);
977 * Process the export option only if we are
978 * updating mount options.
980 if (!error && (fsflags & MNT_UPDATE)) {
981 if (vfs_copyopt(mp->mnt_optnew, "export", &export,
982 sizeof(export)) == 0)
983 error = vfs_export(mp, &export);
987 if (mp->mnt_opt != NULL)
988 vfs_freeopts(mp->mnt_opt);
989 mp->mnt_opt = mp->mnt_optnew;
990 (void)VFS_STATFS(mp, &mp->mnt_stat, td);
993 * Prevent external consumers of mount options from reading
996 mp->mnt_optnew = NULL;
997 if (mp->mnt_flag & MNT_UPDATE) {
1000 mp->mnt_flag = (mp->mnt_flag & MNT_QUOTA) |
1001 (flag & ~MNT_QUOTA);
1003 mp->mnt_flag &= ~(MNT_UPDATE | MNT_RELOAD |
1004 MNT_FORCE | MNT_SNAPSHOT);
1005 if ((mp->mnt_flag & MNT_ASYNC) != 0 && mp->mnt_noasync == 0)
1006 mp->mnt_kern_flag |= MNTK_ASYNC;
1008 mp->mnt_kern_flag &= ~MNTK_ASYNC;
1010 if ((mp->mnt_flag & MNT_RDONLY) == 0) {
1011 if (mp->mnt_syncer == NULL)
1012 error = vfs_allocate_syncvnode(mp);
1014 if (mp->mnt_syncer != NULL)
1015 vrele(mp->mnt_syncer);
1016 mp->mnt_syncer = NULL;
1020 vp->v_iflag &= ~VI_MOUNT;
1026 if ((mp->mnt_flag & MNT_ASYNC) != 0 && mp->mnt_noasync == 0)
1027 mp->mnt_kern_flag |= MNTK_ASYNC;
1029 mp->mnt_kern_flag &= ~MNTK_ASYNC;
1031 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td);
1033 * Put the new filesystem on the mount list after root.
1037 struct vnode *newdp;
1040 vp->v_iflag &= ~VI_MOUNT;
1042 vp->v_mountedhere = mp;
1043 mtx_lock(&mountlist_mtx);
1044 TAILQ_INSERT_TAIL(&mountlist, mp, mnt_list);
1045 mtx_unlock(&mountlist_mtx);
1046 vfs_event_signal(NULL, VQ_MOUNT, 0);
1047 if (VFS_ROOT(mp, LK_EXCLUSIVE, &newdp, td))
1048 panic("mount: lost mount");
1049 mountcheckdirs(vp, newdp);
1051 VOP_UNLOCK(vp, 0, td);
1052 if ((mp->mnt_flag & MNT_RDONLY) == 0)
1053 error = vfs_allocate_syncvnode(mp);
1059 vp->v_iflag &= ~VI_MOUNT;
1062 vfs_mount_destroy(mp);
1069 * Unmount a filesystem.
1071 * Note: unmount takes a path to the vnode mounted on as argument, not
1072 * special file (as before).
1074 #ifndef _SYS_SYSPROTO_H_
1075 struct unmount_args {
1084 register struct unmount_args /* {
1091 int error, id0, id1;
1093 if (jailed(td->td_ucred) || usermount == 0) {
1094 error = priv_check(td, PRIV_VFS_UNMOUNT);
1099 pathbuf = malloc(MNAMELEN, M_TEMP, M_WAITOK);
1100 error = copyinstr(uap->path, pathbuf, MNAMELEN, NULL);
1102 free(pathbuf, M_TEMP);
1105 AUDIT_ARG(upath, td, pathbuf, ARG_UPATH1);
1107 if (uap->flags & MNT_BYFSID) {
1108 /* Decode the filesystem ID. */
1109 if (sscanf(pathbuf, "FSID:%d:%d", &id0, &id1) != 2) {
1111 free(pathbuf, M_TEMP);
1115 mtx_lock(&mountlist_mtx);
1116 TAILQ_FOREACH_REVERSE(mp, &mountlist, mntlist, mnt_list) {
1117 if (mp->mnt_stat.f_fsid.val[0] == id0 &&
1118 mp->mnt_stat.f_fsid.val[1] == id1)
1121 mtx_unlock(&mountlist_mtx);
1123 mtx_lock(&mountlist_mtx);
1124 TAILQ_FOREACH_REVERSE(mp, &mountlist, mntlist, mnt_list) {
1125 if (strcmp(mp->mnt_stat.f_mntonname, pathbuf) == 0)
1128 mtx_unlock(&mountlist_mtx);
1130 free(pathbuf, M_TEMP);
1133 * Previously we returned ENOENT for a nonexistent path and
1134 * EINVAL for a non-mountpoint. We cannot tell these apart
1135 * now, so in the !MNT_BYFSID case return the more likely
1136 * EINVAL for compatibility.
1139 return ((uap->flags & MNT_BYFSID) ? ENOENT : EINVAL);
1143 * Don't allow unmounting the root filesystem.
1145 if (mp->mnt_flag & MNT_ROOTFS) {
1149 error = dounmount(mp, uap->flags, td);
1155 * Do the actual filesystem unmount.
1158 dounmount(mp, flags, td)
1163 struct vnode *coveredvp, *fsrootvp;
1168 mtx_assert(&Giant, MA_OWNED);
1170 if ((coveredvp = mp->mnt_vnodecovered) != NULL) {
1171 mnt_gen_r = mp->mnt_gen;
1174 error = vn_lock(coveredvp, LK_EXCLUSIVE | LK_INTERLOCK, td);
1177 * Check for mp being unmounted while waiting for the
1178 * covered vnode lock.
1182 if (coveredvp->v_mountedhere != mp ||
1183 coveredvp->v_mountedhere->mnt_gen != mnt_gen_r) {
1184 VOP_UNLOCK(coveredvp, 0, td);
1189 * Only privileged root, or (if MNT_USER is set) the user that did the
1190 * original mount is permitted to unmount this filesystem.
1192 error = vfs_suser(mp, td);
1195 VOP_UNLOCK(coveredvp, 0, td);
1200 if (mp->mnt_kern_flag & MNTK_UNMOUNT) {
1203 VOP_UNLOCK(coveredvp, 0, td);
1206 mp->mnt_kern_flag |= MNTK_UNMOUNT;
1207 /* Allow filesystems to detect that a forced unmount is in progress. */
1208 if (flags & MNT_FORCE)
1209 mp->mnt_kern_flag |= MNTK_UNMOUNTF;
1210 error = lockmgr(&mp->mnt_lock, LK_DRAIN | LK_INTERLOCK |
1211 ((flags & MNT_FORCE) ? 0 : LK_NOWAIT), MNT_MTX(mp), td);
1214 mp->mnt_kern_flag &= ~(MNTK_UNMOUNT | MNTK_UNMOUNTF);
1215 if (mp->mnt_kern_flag & MNTK_MWAIT)
1219 VOP_UNLOCK(coveredvp, 0, td);
1222 vn_start_write(NULL, &mp, V_WAIT);
1224 if (mp->mnt_flag & MNT_EXPUBLIC)
1225 vfs_setpublicfs(NULL, NULL, NULL);
1227 vfs_msync(mp, MNT_WAIT);
1229 async_flag = mp->mnt_flag & MNT_ASYNC;
1230 mp->mnt_flag &= ~MNT_ASYNC;
1231 mp->mnt_kern_flag &= ~MNTK_ASYNC;
1233 cache_purgevfs(mp); /* remove cache entries for this file sys */
1234 if (mp->mnt_syncer != NULL)
1235 vrele(mp->mnt_syncer);
1237 * For forced unmounts, move process cdir/rdir refs on the fs root
1238 * vnode to the covered vnode. For non-forced unmounts we want
1239 * such references to cause an EBUSY error.
1241 if ((flags & MNT_FORCE) &&
1242 VFS_ROOT(mp, LK_EXCLUSIVE, &fsrootvp, td) == 0) {
1243 if (mp->mnt_vnodecovered != NULL)
1244 mountcheckdirs(fsrootvp, mp->mnt_vnodecovered);
1245 if (fsrootvp == rootvnode) {
1251 if (((mp->mnt_flag & MNT_RDONLY) ||
1252 (error = VFS_SYNC(mp, MNT_WAIT, td)) == 0) ||
1253 (flags & MNT_FORCE)) {
1254 error = VFS_UNMOUNT(mp, flags, td);
1256 vn_finished_write(mp);
1258 /* Undo cdir/rdir and rootvnode changes made above. */
1259 if ((flags & MNT_FORCE) &&
1260 VFS_ROOT(mp, LK_EXCLUSIVE, &fsrootvp, td) == 0) {
1261 if (mp->mnt_vnodecovered != NULL)
1262 mountcheckdirs(mp->mnt_vnodecovered, fsrootvp);
1263 if (rootvnode == NULL) {
1264 rootvnode = fsrootvp;
1269 if ((mp->mnt_flag & MNT_RDONLY) == 0 && mp->mnt_syncer == NULL)
1270 (void) vfs_allocate_syncvnode(mp);
1272 mp->mnt_kern_flag &= ~(MNTK_UNMOUNT | MNTK_UNMOUNTF);
1273 mp->mnt_flag |= async_flag;
1274 if ((mp->mnt_flag & MNT_ASYNC) != 0 && mp->mnt_noasync == 0)
1275 mp->mnt_kern_flag |= MNTK_ASYNC;
1276 lockmgr(&mp->mnt_lock, LK_RELEASE, NULL, td);
1277 if (mp->mnt_kern_flag & MNTK_MWAIT)
1281 VOP_UNLOCK(coveredvp, 0, td);
1284 mtx_lock(&mountlist_mtx);
1285 TAILQ_REMOVE(&mountlist, mp, mnt_list);
1286 mtx_unlock(&mountlist_mtx);
1287 if (coveredvp != NULL) {
1288 coveredvp->v_mountedhere = NULL;
1291 vfs_event_signal(NULL, VQ_UNMOUNT, 0);
1292 lockmgr(&mp->mnt_lock, LK_RELEASE, NULL, td);
1293 vfs_mount_destroy(mp);
1298 * ---------------------------------------------------------------------
1299 * Mounting of root filesystem
1303 struct root_hold_token {
1305 LIST_ENTRY(root_hold_token) list;
1308 static LIST_HEAD(, root_hold_token) root_holds =
1309 LIST_HEAD_INITIALIZER(&root_holds);
1311 struct root_hold_token *
1312 root_mount_hold(const char *identifier)
1314 struct root_hold_token *h;
1316 h = malloc(sizeof *h, M_DEVBUF, M_ZERO | M_WAITOK);
1317 h->who = identifier;
1318 mtx_lock(&mountlist_mtx);
1319 LIST_INSERT_HEAD(&root_holds, h, list);
1320 mtx_unlock(&mountlist_mtx);
1325 root_mount_rel(struct root_hold_token *h)
1328 mtx_lock(&mountlist_mtx);
1329 LIST_REMOVE(h, list);
1330 wakeup(&root_holds);
1331 mtx_unlock(&mountlist_mtx);
1336 root_mount_wait(void)
1338 struct root_hold_token *h;
1344 mtx_lock(&mountlist_mtx);
1345 if (LIST_EMPTY(&root_holds)) {
1346 mtx_unlock(&mountlist_mtx);
1349 printf("Root mount waiting for:");
1350 LIST_FOREACH(h, &root_holds, list)
1351 printf(" %s", h->who);
1353 msleep(&root_holds, &mountlist_mtx, PZERO | PDROP, "roothold",
1359 set_rootvnode(struct thread *td)
1363 if (VFS_ROOT(TAILQ_FIRST(&mountlist), LK_EXCLUSIVE, &rootvnode, td))
1364 panic("Cannot find root vnode");
1367 FILEDESC_LOCK(p->p_fd);
1369 if (p->p_fd->fd_cdir != NULL)
1370 vrele(p->p_fd->fd_cdir);
1371 p->p_fd->fd_cdir = rootvnode;
1374 if (p->p_fd->fd_rdir != NULL)
1375 vrele(p->p_fd->fd_rdir);
1376 p->p_fd->fd_rdir = rootvnode;
1379 FILEDESC_UNLOCK(p->p_fd);
1381 VOP_UNLOCK(rootvnode, 0, td);
1385 * Mount /devfs as our root filesystem, but do not put it on the mountlist
1386 * yet. Create a /dev -> / symlink so that absolute pathnames will lookup.
1392 struct thread *td = curthread;
1393 struct vfsoptlist *opts;
1394 struct vfsconf *vfsp;
1395 struct mount *mp = NULL;
1398 vfsp = vfs_byname("devfs");
1399 KASSERT(vfsp != NULL, ("Could not find devfs by name"));
1403 mp = vfs_mount_alloc(NULLVP, vfsp, "/dev", td);
1405 error = VFS_MOUNT(mp, td);
1406 KASSERT(error == 0, ("VFS_MOUNT(devfs) failed %d", error));
1410 opts = malloc(sizeof(struct vfsoptlist), M_MOUNT, M_WAITOK);
1414 mtx_lock(&mountlist_mtx);
1415 TAILQ_INSERT_HEAD(&mountlist, mp, mnt_list);
1416 mtx_unlock(&mountlist_mtx);
1420 error = kern_symlink(td, "/", "dev", UIO_SYSSPACE);
1422 printf("kern_symlink /dev -> / returns %d\n", error);
1426 * Surgically move our devfs to be mounted on /dev.
1430 devfs_fixup(struct thread *td)
1432 struct nameidata nd;
1434 struct vnode *vp, *dvp;
1437 /* Remove our devfs mount from the mountlist and purge the cache */
1438 mtx_lock(&mountlist_mtx);
1439 mp = TAILQ_FIRST(&mountlist);
1440 TAILQ_REMOVE(&mountlist, mp, mnt_list);
1441 mtx_unlock(&mountlist_mtx);
1444 VFS_ROOT(mp, LK_EXCLUSIVE, &dvp, td);
1446 dvp->v_iflag &= ~VI_MOUNT;
1447 dvp->v_mountedhere = NULL;
1450 /* Set up the real rootvnode, and purge the cache */
1451 TAILQ_FIRST(&mountlist)->mnt_vnodecovered = NULL;
1453 cache_purgevfs(rootvnode->v_mount);
1455 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_SYSSPACE, "/dev", td);
1458 printf("Lookup of /dev for devfs, error: %d\n", error);
1461 NDFREE(&nd, NDF_ONLY_PNBUF);
1463 if (vp->v_type != VDIR) {
1466 error = vinvalbuf(vp, V_SAVE, td, 0, 0);
1471 mp->mnt_vnodecovered = vp;
1472 vp->v_mountedhere = mp;
1473 mtx_lock(&mountlist_mtx);
1474 TAILQ_INSERT_TAIL(&mountlist, mp, mnt_list);
1475 mtx_unlock(&mountlist_mtx);
1476 VOP_UNLOCK(vp, 0, td);
1480 /* Unlink the no longer needed /dev/dev -> / symlink */
1481 kern_unlink(td, "/dev/dev", UIO_SYSSPACE);
1485 * Report errors during filesystem mounting.
1488 vfs_mount_error(struct mount *mp, const char *fmt, ...)
1490 struct vfsoptlist *moptlist = mp->mnt_optnew;
1495 error = vfs_getopt(moptlist, "errmsg", (void **)&errmsg, &len);
1496 if (error || errmsg == NULL || len <= 0)
1500 vsnprintf(errmsg, (size_t)len, fmt, ap);
1505 * Find and mount the root filesystem
1511 int error, i, asked = 0;
1515 mount_zone = uma_zcreate("Mountpoints", sizeof(struct mount),
1516 NULL, NULL, mount_init, mount_fini,
1517 UMA_ALIGN_PTR, UMA_ZONE_NOFREE);
1521 * We are booted with instructions to prompt for the root filesystem.
1523 if (boothowto & RB_ASKNAME) {
1524 if (!vfs_mountroot_ask())
1530 * The root filesystem information is compiled in, and we are
1531 * booted with instructions to use it.
1533 if (ctrootdevname != NULL && (boothowto & RB_DFLTROOT)) {
1534 if (!vfs_mountroot_try(ctrootdevname))
1536 ctrootdevname = NULL;
1540 * We've been given the generic "use CDROM as root" flag. This is
1541 * necessary because one media may be used in many different
1542 * devices, so we need to search for them.
1544 if (boothowto & RB_CDROM) {
1545 for (i = 0; cdrom_rootdevnames[i] != NULL; i++) {
1546 if (!vfs_mountroot_try(cdrom_rootdevnames[i]))
1552 * Try to use the value read by the loader from /etc/fstab, or
1553 * supplied via some other means. This is the preferred
1556 cp = getenv("vfs.root.mountfrom");
1558 error = vfs_mountroot_try(cp);
1565 * Try values that may have been computed by code during boot
1567 if (!vfs_mountroot_try(rootdevnames[0]))
1569 if (!vfs_mountroot_try(rootdevnames[1]))
1573 * If we (still) have a compiled-in default, try it.
1575 if (ctrootdevname != NULL)
1576 if (!vfs_mountroot_try(ctrootdevname))
1579 * Everything so far has failed, prompt on the console if we haven't
1580 * already tried that.
1583 if (!vfs_mountroot_ask())
1586 panic("Root mount failed, startup aborted.");
1590 * Mount (mountfrom) as the root filesystem.
1593 vfs_mountroot_try(const char *mountfrom)
1596 char *vfsname, *path;
1606 if (mountfrom == NULL)
1607 return (error); /* don't complain */
1608 printf("Trying to mount root from %s\n", mountfrom);
1610 /* parse vfs name and path */
1611 vfsname = malloc(MFSNAMELEN, M_MOUNT, M_WAITOK);
1612 path = malloc(MNAMELEN, M_MOUNT, M_WAITOK);
1613 vfsname[0] = path[0] = 0;
1614 sprintf(patt, "%%%d[a-z0-9]:%%%ds", MFSNAMELEN, MNAMELEN);
1615 if (sscanf(mountfrom, patt, vfsname, path) < 1)
1618 if (path[0] == '\0')
1619 strcpy(path, ROOTNAME);
1621 error = kernel_vmount(
1622 MNT_RDONLY | MNT_ROOTFS,
1629 * We mount devfs prior to mounting the / FS, so the first
1630 * entry will typically be devfs.
1632 mp = TAILQ_FIRST(&mountlist);
1633 KASSERT(mp != NULL, ("%s: mountlist is empty", __func__));
1636 * Iterate over all currently mounted file systems and use
1637 * the time stamp found to check and/or initialize the RTC.
1638 * Typically devfs has no time stamp and the only other FS
1639 * is the actual / FS.
1640 * Call inittodr() only once and pass it the largest of the
1641 * timestamps we encounter.
1645 if (mp->mnt_time > timebase)
1646 timebase = mp->mnt_time;
1647 mp = TAILQ_NEXT(mp, mnt_list);
1648 } while (mp != NULL);
1651 devfs_fixup(curthread);
1654 free(path, M_MOUNT);
1655 free(vfsname, M_MOUNT);
1660 * ---------------------------------------------------------------------
1661 * Interactive root filesystem selection code.
1665 vfs_mountroot_ask(void)
1670 printf("\nManual root filesystem specification:\n");
1671 printf(" <fstype>:<device> Mount <device> using filesystem <fstype>\n");
1672 #if defined(__amd64__) || defined(__i386__) || defined(__ia64__)
1673 printf(" eg. ufs:da0s1a\n");
1675 printf(" eg. ufs:/dev/da0a\n");
1677 printf(" ? List valid disk boot devices\n");
1678 printf(" <empty line> Abort manual input\n");
1679 printf("\nmountroot> ");
1680 gets(name, sizeof(name), 1);
1681 if (name[0] == '\0')
1683 if (name[0] == '?') {
1684 printf("\nList of GEOM managed disk devices:\n ");
1688 if (!vfs_mountroot_try(name))
1694 * ---------------------------------------------------------------------
1695 * Functions for querying mount options/arguments from filesystems.
1699 * Check that no unknown options are given
1702 vfs_filteropt(struct vfsoptlist *opts, const char **legal)
1707 TAILQ_FOREACH(opt, opts, link) {
1709 if (p[0] == 'n' && p[1] == 'o')
1711 for(t = global_opts; *t != NULL; t++)
1716 for(t = legal; *t != NULL; t++)
1721 printf("mount option <%s> is unknown\n", p);
1728 * Get a mount option by its name.
1730 * Return 0 if the option was found, ENOENT otherwise.
1731 * If len is non-NULL it will be filled with the length
1732 * of the option. If buf is non-NULL, it will be filled
1733 * with the address of the option.
1736 vfs_getopt(opts, name, buf, len)
1737 struct vfsoptlist *opts;
1744 KASSERT(opts != NULL, ("vfs_getopt: caller passed 'opts' as NULL"));
1746 TAILQ_FOREACH(opt, opts, link) {
1747 if (strcmp(name, opt->name) == 0) {
1759 vfs_getopt_pos(struct vfsoptlist *opts, const char *name)
1768 TAILQ_FOREACH(opt, opts, link) {
1769 if (strcmp(name, opt->name) == 0)
1777 vfs_getopts(struct vfsoptlist *opts, const char *name, int *error)
1782 TAILQ_FOREACH(opt, opts, link) {
1783 if (strcmp(name, opt->name) != 0)
1785 if (((char *)opt->value)[opt->len - 1] != '\0') {
1789 return (opt->value);
1796 vfs_flagopt(struct vfsoptlist *opts, const char *name, u_int *w, u_int val)
1800 TAILQ_FOREACH(opt, opts, link) {
1801 if (strcmp(name, opt->name) == 0) {
1813 vfs_scanopt(struct vfsoptlist *opts, const char *name, const char *fmt, ...)
1819 KASSERT(opts != NULL, ("vfs_getopt: caller passed 'opts' as NULL"));
1821 TAILQ_FOREACH(opt, opts, link) {
1822 if (strcmp(name, opt->name) != 0)
1824 if (((char *)opt->value)[opt->len - 1] != '\0')
1827 ret = vsscanf(opt->value, fmt, ap);
1835 * Find and copy a mount option.
1837 * The size of the buffer has to be specified
1838 * in len, if it is not the same length as the
1839 * mount option, EINVAL is returned.
1840 * Returns ENOENT if the option is not found.
1843 vfs_copyopt(opts, name, dest, len)
1844 struct vfsoptlist *opts;
1851 KASSERT(opts != NULL, ("vfs_copyopt: caller passed 'opts' as NULL"));
1853 TAILQ_FOREACH(opt, opts, link) {
1854 if (strcmp(name, opt->name) == 0) {
1855 if (len != opt->len)
1857 bcopy(opt->value, dest, opt->len);
1865 * This is a helper function for filesystems to traverse their
1866 * vnodes. See MNT_VNODE_FOREACH() in sys/mount.h
1870 __mnt_vnode_next(struct vnode **mvp, struct mount *mp)
1874 mtx_assert(MNT_MTX(mp), MA_OWNED);
1876 KASSERT((*mvp)->v_mount == mp, ("marker vnode mount list mismatch"));
1877 vp = TAILQ_NEXT(*mvp, v_nmntvnodes);
1878 while (vp != NULL && vp->v_type == VMARKER)
1879 vp = TAILQ_NEXT(vp, v_nmntvnodes);
1881 /* Check if we are done */
1883 __mnt_vnode_markerfree(mvp, mp);
1886 TAILQ_REMOVE(&mp->mnt_nvnodelist, *mvp, v_nmntvnodes);
1887 TAILQ_INSERT_AFTER(&mp->mnt_nvnodelist, vp, *mvp, v_nmntvnodes);
1892 __mnt_vnode_first(struct vnode **mvp, struct mount *mp)
1896 mtx_assert(MNT_MTX(mp), MA_OWNED);
1898 vp = TAILQ_FIRST(&mp->mnt_nvnodelist);
1899 while (vp != NULL && vp->v_type == VMARKER)
1900 vp = TAILQ_NEXT(vp, v_nmntvnodes);
1902 /* Check if we are done */
1909 *mvp = (struct vnode *) malloc(sizeof(struct vnode),
1913 (*mvp)->v_type = VMARKER;
1915 vp = TAILQ_FIRST(&mp->mnt_nvnodelist);
1916 while (vp != NULL && vp->v_type == VMARKER)
1917 vp = TAILQ_NEXT(vp, v_nmntvnodes);
1919 /* Check if we are done */
1922 free(*mvp, M_VNODE_MARKER);
1926 if (mp->mnt_holdcnt == 0 && mp->mnt_holdcntwaiters != 0)
1927 wakeup(&mp->mnt_holdcnt);
1930 mp->mnt_markercnt++;
1931 (*mvp)->v_mount = mp;
1932 TAILQ_INSERT_AFTER(&mp->mnt_nvnodelist, vp, *mvp, v_nmntvnodes);
1938 __mnt_vnode_markerfree(struct vnode **mvp, struct mount *mp)
1944 mtx_assert(MNT_MTX(mp), MA_OWNED);
1946 KASSERT((*mvp)->v_mount == mp, ("marker vnode mount list mismatch"));
1947 TAILQ_REMOVE(&mp->mnt_nvnodelist, *mvp, v_nmntvnodes);
1949 free(*mvp, M_VNODE_MARKER);
1953 mp->mnt_markercnt--;
1955 if (mp->mnt_holdcnt == 0 && mp->mnt_holdcntwaiters != 0)
1956 wakeup(&mp->mnt_holdcnt);
1961 __vfs_statfs(struct mount *mp, struct statfs *sbp, struct thread *td)
1965 error = mp->mnt_op->vfs_statfs(mp, &mp->mnt_stat, td);
1966 if (sbp != &mp->mnt_stat)
1967 *sbp = mp->mnt_stat;
1972 vfs_mountedfrom(struct mount *mp, const char *from)
1975 bzero(mp->mnt_stat.f_mntfromname, sizeof mp->mnt_stat.f_mntfromname);
1976 strlcpy(mp->mnt_stat.f_mntfromname, from,
1977 sizeof mp->mnt_stat.f_mntfromname);
1981 * ---------------------------------------------------------------------
1982 * This is the api for building mount args and mounting filesystems from
1983 * inside the kernel.
1985 * The API works by accumulation of individual args. First error is
1988 * XXX: should be documented in new manpage kernel_mount(9)
1991 /* A memory allocation which must be freed when we are done */
1993 SLIST_ENTRY(mntaarg) next;
1996 /* The header for the mount arguments */
2001 SLIST_HEAD(, mntaarg) list;
2005 * Add a boolean argument.
2007 * flag is the boolean value.
2008 * name must start with "no".
2011 mount_argb(struct mntarg *ma, int flag, const char *name)
2014 KASSERT(name[0] == 'n' && name[1] == 'o',
2015 ("mount_argb(...,%s): name must start with 'no'", name));
2017 return (mount_arg(ma, name + (flag ? 2 : 0), NULL, 0));
2021 * Add an argument printf style
2024 mount_argf(struct mntarg *ma, const char *name, const char *fmt, ...)
2027 struct mntaarg *maa;
2032 ma = malloc(sizeof *ma, M_MOUNT, M_WAITOK | M_ZERO);
2033 SLIST_INIT(&ma->list);
2038 ma->v = realloc(ma->v, sizeof *ma->v * (ma->len + 2),
2040 ma->v[ma->len].iov_base = (void *)(uintptr_t)name;
2041 ma->v[ma->len].iov_len = strlen(name) + 1;
2044 sb = sbuf_new(NULL, NULL, 0, SBUF_AUTOEXTEND);
2046 sbuf_vprintf(sb, fmt, ap);
2049 len = sbuf_len(sb) + 1;
2050 maa = malloc(sizeof *maa + len, M_MOUNT, M_WAITOK | M_ZERO);
2051 SLIST_INSERT_HEAD(&ma->list, maa, next);
2052 bcopy(sbuf_data(sb), maa + 1, len);
2055 ma->v[ma->len].iov_base = maa + 1;
2056 ma->v[ma->len].iov_len = len;
2063 * Add an argument which is a userland string.
2066 mount_argsu(struct mntarg *ma, const char *name, const void *val, int len)
2068 struct mntaarg *maa;
2074 ma = malloc(sizeof *ma, M_MOUNT, M_WAITOK | M_ZERO);
2075 SLIST_INIT(&ma->list);
2079 maa = malloc(sizeof *maa + len, M_MOUNT, M_WAITOK | M_ZERO);
2080 SLIST_INSERT_HEAD(&ma->list, maa, next);
2081 tbuf = (void *)(maa + 1);
2082 ma->error = copyinstr(val, tbuf, len, NULL);
2083 return (mount_arg(ma, name, tbuf, -1));
2089 * If length is -1, use printf.
2092 mount_arg(struct mntarg *ma, const char *name, const void *val, int len)
2096 ma = malloc(sizeof *ma, M_MOUNT, M_WAITOK | M_ZERO);
2097 SLIST_INIT(&ma->list);
2102 ma->v = realloc(ma->v, sizeof *ma->v * (ma->len + 2),
2104 ma->v[ma->len].iov_base = (void *)(uintptr_t)name;
2105 ma->v[ma->len].iov_len = strlen(name) + 1;
2108 ma->v[ma->len].iov_base = (void *)(uintptr_t)val;
2110 ma->v[ma->len].iov_len = strlen(val) + 1;
2112 ma->v[ma->len].iov_len = len;
2118 * Free a mntarg structure
2121 free_mntarg(struct mntarg *ma)
2123 struct mntaarg *maa;
2125 while (!SLIST_EMPTY(&ma->list)) {
2126 maa = SLIST_FIRST(&ma->list);
2127 SLIST_REMOVE_HEAD(&ma->list, next);
2130 free(ma->v, M_MOUNT);
2135 * Mount a filesystem
2138 kernel_mount(struct mntarg *ma, int flags)
2143 KASSERT(ma != NULL, ("kernel_mount NULL ma"));
2144 KASSERT(ma->v != NULL, ("kernel_mount NULL ma->v"));
2145 KASSERT(!(ma->len & 1), ("kernel_mount odd ma->len (%d)", ma->len));
2147 auio.uio_iov = ma->v;
2148 auio.uio_iovcnt = ma->len;
2149 auio.uio_segflg = UIO_SYSSPACE;
2153 error = vfs_donmount(curthread, flags, &auio);
2159 * A printflike function to mount a filesystem.
2162 kernel_vmount(int flags, ...)
2164 struct mntarg *ma = NULL;
2170 va_start(ap, flags);
2172 cp = va_arg(ap, const char *);
2175 vp = va_arg(ap, const void *);
2176 ma = mount_arg(ma, cp, vp, -1);
2180 error = kernel_mount(ma, flags);