2 * SPDX-License-Identifier: BSD-3-Clause
4 * Copyright (c) 1999-2004 Poul-Henning Kamp
5 * Copyright (c) 1999 Michael Smith
6 * Copyright (c) 1989, 1993
7 * The Regents of the University of California. All rights reserved.
8 * (c) UNIX System Laboratories, Inc.
9 * All or some portions of this file are derived from material licensed
10 * to the University of California by American Telephone and Telegraph
11 * Co. or Unix System Laboratories, Inc. and are reproduced herein with
12 * the permission of UNIX System Laboratories, Inc.
14 * Redistribution and use in source and binary forms, with or without
15 * modification, are permitted provided that the following conditions
17 * 1. Redistributions of source code must retain the above copyright
18 * notice, this list of conditions and the following disclaimer.
19 * 2. Redistributions in binary form must reproduce the above copyright
20 * notice, this list of conditions and the following disclaimer in the
21 * documentation and/or other materials provided with the distribution.
22 * 3. Neither the name of the University nor the names of its contributors
23 * may be used to endorse or promote products derived from this software
24 * without specific prior written permission.
26 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
27 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
28 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
29 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
30 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
31 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
32 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
33 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
34 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
35 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
39 #include <sys/cdefs.h>
40 __FBSDID("$FreeBSD$");
42 #include <sys/param.h>
44 #include <sys/eventhandler.h>
45 #include <sys/fcntl.h>
47 #include <sys/kernel.h>
48 #include <sys/libkern.h>
49 #include <sys/malloc.h>
50 #include <sys/mount.h>
51 #include <sys/mutex.h>
52 #include <sys/namei.h>
55 #include <sys/filedesc.h>
56 #include <sys/reboot.h>
58 #include <sys/syscallsubr.h>
59 #include <sys/sysproto.h>
61 #include <sys/sysctl.h>
62 #include <sys/sysent.h>
63 #include <sys/systm.h>
64 #include <sys/vnode.h>
67 #include <geom/geom.h>
69 #include <machine/stdarg.h>
71 #include <security/audit/audit.h>
72 #include <security/mac/mac_framework.h>
74 #define VFS_MOUNTARG_SIZE_MAX (1024 * 64)
76 static int vfs_domount(struct thread *td, const char *fstype, char *fspath,
77 uint64_t fsflags, struct vfsoptlist **optlist);
78 static void free_mntarg(struct mntarg *ma);
80 static int usermount = 0;
81 SYSCTL_INT(_vfs, OID_AUTO, usermount, CTLFLAG_RW, &usermount, 0,
82 "Unprivileged users may mount and unmount file systems");
84 static bool default_autoro = false;
85 SYSCTL_BOOL(_vfs, OID_AUTO, default_autoro, CTLFLAG_RW, &default_autoro, 0,
86 "Retry failed r/w mount as r/o if no explicit ro/rw option is specified");
88 MALLOC_DEFINE(M_MOUNT, "mount", "vfs mount structure");
89 MALLOC_DEFINE(M_STATFS, "statfs", "statfs structure");
90 static uma_zone_t mount_zone;
92 /* List of mounted filesystems. */
93 struct mntlist mountlist = TAILQ_HEAD_INITIALIZER(mountlist);
95 /* For any iteration/modification of mountlist */
96 struct mtx mountlist_mtx;
97 MTX_SYSINIT(mountlist, &mountlist_mtx, "mountlist", MTX_DEF);
99 EVENTHANDLER_LIST_DEFINE(vfs_mounted);
100 EVENTHANDLER_LIST_DEFINE(vfs_unmounted);
103 * Global opts, taken by all filesystems
105 static const char *global_opts[] = {
117 mount_init(void *mem, int size, int flags)
121 mp = (struct mount *)mem;
122 mtx_init(&mp->mnt_mtx, "struct mount mtx", NULL, MTX_DEF);
123 mtx_init(&mp->mnt_listmtx, "struct mount vlist mtx", NULL, MTX_DEF);
124 lockinit(&mp->mnt_explock, PVFS, "explock", 0, 0);
129 mount_fini(void *mem, int size)
133 mp = (struct mount *)mem;
134 lockdestroy(&mp->mnt_explock);
135 mtx_destroy(&mp->mnt_listmtx);
136 mtx_destroy(&mp->mnt_mtx);
140 vfs_mount_init(void *dummy __unused)
143 mount_zone = uma_zcreate("Mountpoints", sizeof(struct mount), NULL,
144 NULL, mount_init, mount_fini, UMA_ALIGN_PTR, UMA_ZONE_NOFREE);
146 SYSINIT(vfs_mount, SI_SUB_VFS, SI_ORDER_ANY, vfs_mount_init, NULL);
149 * ---------------------------------------------------------------------
150 * Functions for building and sanitizing the mount options
153 /* Remove one mount option. */
155 vfs_freeopt(struct vfsoptlist *opts, struct vfsopt *opt)
158 TAILQ_REMOVE(opts, opt, link);
159 free(opt->name, M_MOUNT);
160 if (opt->value != NULL)
161 free(opt->value, M_MOUNT);
165 /* Release all resources related to the mount options. */
167 vfs_freeopts(struct vfsoptlist *opts)
171 while (!TAILQ_EMPTY(opts)) {
172 opt = TAILQ_FIRST(opts);
173 vfs_freeopt(opts, opt);
179 vfs_deleteopt(struct vfsoptlist *opts, const char *name)
181 struct vfsopt *opt, *temp;
185 TAILQ_FOREACH_SAFE(opt, opts, link, temp) {
186 if (strcmp(opt->name, name) == 0)
187 vfs_freeopt(opts, opt);
192 vfs_isopt_ro(const char *opt)
195 if (strcmp(opt, "ro") == 0 || strcmp(opt, "rdonly") == 0 ||
196 strcmp(opt, "norw") == 0)
202 vfs_isopt_rw(const char *opt)
205 if (strcmp(opt, "rw") == 0 || strcmp(opt, "noro") == 0)
211 * Check if options are equal (with or without the "no" prefix).
214 vfs_equalopts(const char *opt1, const char *opt2)
218 /* "opt" vs. "opt" or "noopt" vs. "noopt" */
219 if (strcmp(opt1, opt2) == 0)
221 /* "noopt" vs. "opt" */
222 if (strncmp(opt1, "no", 2) == 0 && strcmp(opt1 + 2, opt2) == 0)
224 /* "opt" vs. "noopt" */
225 if (strncmp(opt2, "no", 2) == 0 && strcmp(opt1, opt2 + 2) == 0)
227 while ((p = strchr(opt1, '.')) != NULL &&
228 !strncmp(opt1, opt2, ++p - opt1)) {
231 /* "foo.noopt" vs. "foo.opt" */
232 if (strncmp(opt1, "no", 2) == 0 && strcmp(opt1 + 2, opt2) == 0)
234 /* "foo.opt" vs. "foo.noopt" */
235 if (strncmp(opt2, "no", 2) == 0 && strcmp(opt1, opt2 + 2) == 0)
238 /* "ro" / "rdonly" / "norw" / "rw" / "noro" */
239 if ((vfs_isopt_ro(opt1) || vfs_isopt_rw(opt1)) &&
240 (vfs_isopt_ro(opt2) || vfs_isopt_rw(opt2)))
246 * If a mount option is specified several times,
247 * (with or without the "no" prefix) only keep
248 * the last occurrence of it.
251 vfs_sanitizeopts(struct vfsoptlist *opts)
253 struct vfsopt *opt, *opt2, *tmp;
255 TAILQ_FOREACH_REVERSE(opt, opts, vfsoptlist, link) {
256 opt2 = TAILQ_PREV(opt, vfsoptlist, link);
257 while (opt2 != NULL) {
258 if (vfs_equalopts(opt->name, opt2->name)) {
259 tmp = TAILQ_PREV(opt2, vfsoptlist, link);
260 vfs_freeopt(opts, opt2);
263 opt2 = TAILQ_PREV(opt2, vfsoptlist, link);
270 * Build a linked list of mount options from a struct uio.
273 vfs_buildopts(struct uio *auio, struct vfsoptlist **options)
275 struct vfsoptlist *opts;
277 size_t memused, namelen, optlen;
278 unsigned int i, iovcnt;
281 opts = malloc(sizeof(struct vfsoptlist), M_MOUNT, M_WAITOK);
284 iovcnt = auio->uio_iovcnt;
285 for (i = 0; i < iovcnt; i += 2) {
286 namelen = auio->uio_iov[i].iov_len;
287 optlen = auio->uio_iov[i + 1].iov_len;
288 memused += sizeof(struct vfsopt) + optlen + namelen;
290 * Avoid consuming too much memory, and attempts to overflow
293 if (memused > VFS_MOUNTARG_SIZE_MAX ||
294 optlen > VFS_MOUNTARG_SIZE_MAX ||
295 namelen > VFS_MOUNTARG_SIZE_MAX) {
300 opt = malloc(sizeof(struct vfsopt), M_MOUNT, M_WAITOK);
301 opt->name = malloc(namelen, M_MOUNT, M_WAITOK);
308 * Do this early, so jumps to "bad" will free the current
311 TAILQ_INSERT_TAIL(opts, opt, link);
313 if (auio->uio_segflg == UIO_SYSSPACE) {
314 bcopy(auio->uio_iov[i].iov_base, opt->name, namelen);
316 error = copyin(auio->uio_iov[i].iov_base, opt->name,
321 /* Ensure names are null-terminated strings. */
322 if (namelen == 0 || opt->name[namelen - 1] != '\0') {
328 opt->value = malloc(optlen, M_MOUNT, M_WAITOK);
329 if (auio->uio_segflg == UIO_SYSSPACE) {
330 bcopy(auio->uio_iov[i + 1].iov_base, opt->value,
333 error = copyin(auio->uio_iov[i + 1].iov_base,
340 vfs_sanitizeopts(opts);
349 * Merge the old mount options with the new ones passed
350 * in the MNT_UPDATE case.
352 * XXX: This function will keep a "nofoo" option in the new
353 * options. E.g, if the option's canonical name is "foo",
354 * "nofoo" ends up in the mount point's active options.
357 vfs_mergeopts(struct vfsoptlist *toopts, struct vfsoptlist *oldopts)
359 struct vfsopt *opt, *new;
361 TAILQ_FOREACH(opt, oldopts, link) {
362 new = malloc(sizeof(struct vfsopt), M_MOUNT, M_WAITOK);
363 new->name = strdup(opt->name, M_MOUNT);
365 new->value = malloc(opt->len, M_MOUNT, M_WAITOK);
366 bcopy(opt->value, new->value, opt->len);
370 new->seen = opt->seen;
371 TAILQ_INSERT_HEAD(toopts, new, link);
373 vfs_sanitizeopts(toopts);
377 * Mount a filesystem.
379 #ifndef _SYS_SYSPROTO_H_
387 sys_nmount(struct thread *td, struct nmount_args *uap)
395 * Mount flags are now 64-bits. On 32-bit archtectures only
396 * 32-bits are passed in, but from here on everything handles
397 * 64-bit flags correctly.
401 AUDIT_ARG_FFLAGS(flags);
402 CTR4(KTR_VFS, "%s: iovp %p with iovcnt %d and flags %d", __func__,
403 uap->iovp, uap->iovcnt, flags);
406 * Filter out MNT_ROOTFS. We do not want clients of nmount() in
407 * userspace to set this flag, but we must filter it out if we want
408 * MNT_UPDATE on the root file system to work.
409 * MNT_ROOTFS should only be set by the kernel when mounting its
412 flags &= ~MNT_ROOTFS;
414 iovcnt = uap->iovcnt;
416 * Check that we have an even number of iovec's
417 * and that we have at least two options.
419 if ((iovcnt & 1) || (iovcnt < 4)) {
420 CTR2(KTR_VFS, "%s: failed for invalid iovcnt %d", __func__,
425 error = copyinuio(uap->iovp, iovcnt, &auio);
427 CTR2(KTR_VFS, "%s: failed for invalid uio op with %d errno",
431 error = vfs_donmount(td, flags, auio);
438 * ---------------------------------------------------------------------
439 * Various utility functions
443 vfs_ref(struct mount *mp)
446 CTR2(KTR_VFS, "%s: mp %p", __func__, mp);
453 vfs_rel(struct mount *mp)
456 CTR2(KTR_VFS, "%s: mp %p", __func__, mp);
463 * Allocate and initialize the mount point struct.
466 vfs_mount_alloc(struct vnode *vp, struct vfsconf *vfsp, const char *fspath,
471 mp = uma_zalloc(mount_zone, M_WAITOK);
472 bzero(&mp->mnt_startzero,
473 __rangeof(struct mount, mnt_startzero, mnt_endzero));
474 TAILQ_INIT(&mp->mnt_nvnodelist);
475 mp->mnt_nvnodelistsize = 0;
476 TAILQ_INIT(&mp->mnt_activevnodelist);
477 mp->mnt_activevnodelistsize = 0;
478 TAILQ_INIT(&mp->mnt_tmpfreevnodelist);
479 mp->mnt_tmpfreevnodelistsize = 0;
481 (void) vfs_busy(mp, MBF_NOWAIT);
482 atomic_add_acq_int(&vfsp->vfc_refcount, 1);
483 mp->mnt_op = vfsp->vfc_vfsops;
485 mp->mnt_stat.f_type = vfsp->vfc_typenum;
487 strlcpy(mp->mnt_stat.f_fstypename, vfsp->vfc_name, MFSNAMELEN);
488 mp->mnt_vnodecovered = vp;
489 mp->mnt_cred = crdup(cred);
490 mp->mnt_stat.f_owner = cred->cr_uid;
491 strlcpy(mp->mnt_stat.f_mntonname, fspath, MNAMELEN);
492 mp->mnt_iosize_max = DFLTPHYS;
495 mac_mount_create(cred, mp);
497 arc4rand(&mp->mnt_hashseed, sizeof mp->mnt_hashseed, 0);
498 TAILQ_INIT(&mp->mnt_uppers);
503 * Destroy the mount struct previously allocated by vfs_mount_alloc().
506 vfs_mount_destroy(struct mount *mp)
510 mp->mnt_kern_flag |= MNTK_REFEXPIRE;
511 if (mp->mnt_kern_flag & MNTK_MWAIT) {
512 mp->mnt_kern_flag &= ~MNTK_MWAIT;
516 msleep(mp, MNT_MTX(mp), PVFS, "mntref", 0);
517 KASSERT(mp->mnt_ref == 0,
518 ("%s: invalid refcount in the drain path @ %s:%d", __func__,
519 __FILE__, __LINE__));
520 if (mp->mnt_writeopcount != 0)
521 panic("vfs_mount_destroy: nonzero writeopcount");
522 if (mp->mnt_secondary_writes != 0)
523 panic("vfs_mount_destroy: nonzero secondary_writes");
524 atomic_subtract_rel_int(&mp->mnt_vfc->vfc_refcount, 1);
525 if (!TAILQ_EMPTY(&mp->mnt_nvnodelist)) {
528 TAILQ_FOREACH(vp, &mp->mnt_nvnodelist, v_nmntvnodes)
529 vn_printf(vp, "dangling vnode ");
530 panic("unmount: dangling vnode");
532 KASSERT(TAILQ_EMPTY(&mp->mnt_uppers), ("mnt_uppers"));
533 if (mp->mnt_nvnodelistsize != 0)
534 panic("vfs_mount_destroy: nonzero nvnodelistsize");
535 if (mp->mnt_activevnodelistsize != 0)
536 panic("vfs_mount_destroy: nonzero activevnodelistsize");
537 if (mp->mnt_tmpfreevnodelistsize != 0)
538 panic("vfs_mount_destroy: nonzero tmpfreevnodelistsize");
539 if (mp->mnt_lockref != 0)
540 panic("vfs_mount_destroy: nonzero lock refcount");
542 if (mp->mnt_vnodecovered != NULL)
543 vrele(mp->mnt_vnodecovered);
545 mac_mount_destroy(mp);
547 if (mp->mnt_opt != NULL)
548 vfs_freeopts(mp->mnt_opt);
549 crfree(mp->mnt_cred);
550 uma_zfree(mount_zone, mp);
554 vfs_should_downgrade_to_ro_mount(uint64_t fsflags, int error)
556 /* This is an upgrade of an exisiting mount. */
557 if ((fsflags & MNT_UPDATE) != 0)
559 /* This is already an R/O mount. */
560 if ((fsflags & MNT_RDONLY) != 0)
564 case ENODEV: /* generic, geom, ... */
565 case EACCES: /* cam/scsi, ... */
566 case EROFS: /* md, mmcsd, ... */
568 * These errors can be returned by the storage layer to signal
569 * that the media is read-only. No harm in the R/O mount
570 * attempt if the error was returned for some other reason.
579 vfs_donmount(struct thread *td, uint64_t fsflags, struct uio *fsoptions)
581 struct vfsoptlist *optlist;
582 struct vfsopt *opt, *tmp_opt;
583 char *fstype, *fspath, *errmsg;
584 int error, fstypelen, fspathlen, errmsg_len, errmsg_pos;
587 errmsg = fspath = NULL;
588 errmsg_len = fspathlen = 0;
590 autoro = default_autoro;
592 error = vfs_buildopts(fsoptions, &optlist);
596 if (vfs_getopt(optlist, "errmsg", (void **)&errmsg, &errmsg_len) == 0)
597 errmsg_pos = vfs_getopt_pos(optlist, "errmsg");
600 * We need these two options before the others,
601 * and they are mandatory for any filesystem.
602 * Ensure they are NUL terminated as well.
605 error = vfs_getopt(optlist, "fstype", (void **)&fstype, &fstypelen);
606 if (error || fstype[fstypelen - 1] != '\0') {
609 strncpy(errmsg, "Invalid fstype", errmsg_len);
613 error = vfs_getopt(optlist, "fspath", (void **)&fspath, &fspathlen);
614 if (error || fspath[fspathlen - 1] != '\0') {
617 strncpy(errmsg, "Invalid fspath", errmsg_len);
622 * We need to see if we have the "update" option
623 * before we call vfs_domount(), since vfs_domount() has special
624 * logic based on MNT_UPDATE. This is very important
625 * when we want to update the root filesystem.
627 TAILQ_FOREACH_SAFE(opt, optlist, link, tmp_opt) {
628 if (strcmp(opt->name, "update") == 0) {
629 fsflags |= MNT_UPDATE;
630 vfs_freeopt(optlist, opt);
632 else if (strcmp(opt->name, "async") == 0)
633 fsflags |= MNT_ASYNC;
634 else if (strcmp(opt->name, "force") == 0) {
635 fsflags |= MNT_FORCE;
636 vfs_freeopt(optlist, opt);
638 else if (strcmp(opt->name, "reload") == 0) {
639 fsflags |= MNT_RELOAD;
640 vfs_freeopt(optlist, opt);
642 else if (strcmp(opt->name, "multilabel") == 0)
643 fsflags |= MNT_MULTILABEL;
644 else if (strcmp(opt->name, "noasync") == 0)
645 fsflags &= ~MNT_ASYNC;
646 else if (strcmp(opt->name, "noatime") == 0)
647 fsflags |= MNT_NOATIME;
648 else if (strcmp(opt->name, "atime") == 0) {
649 free(opt->name, M_MOUNT);
650 opt->name = strdup("nonoatime", M_MOUNT);
652 else if (strcmp(opt->name, "noclusterr") == 0)
653 fsflags |= MNT_NOCLUSTERR;
654 else if (strcmp(opt->name, "clusterr") == 0) {
655 free(opt->name, M_MOUNT);
656 opt->name = strdup("nonoclusterr", M_MOUNT);
658 else if (strcmp(opt->name, "noclusterw") == 0)
659 fsflags |= MNT_NOCLUSTERW;
660 else if (strcmp(opt->name, "clusterw") == 0) {
661 free(opt->name, M_MOUNT);
662 opt->name = strdup("nonoclusterw", M_MOUNT);
664 else if (strcmp(opt->name, "noexec") == 0)
665 fsflags |= MNT_NOEXEC;
666 else if (strcmp(opt->name, "exec") == 0) {
667 free(opt->name, M_MOUNT);
668 opt->name = strdup("nonoexec", M_MOUNT);
670 else if (strcmp(opt->name, "nosuid") == 0)
671 fsflags |= MNT_NOSUID;
672 else if (strcmp(opt->name, "suid") == 0) {
673 free(opt->name, M_MOUNT);
674 opt->name = strdup("nonosuid", M_MOUNT);
676 else if (strcmp(opt->name, "nosymfollow") == 0)
677 fsflags |= MNT_NOSYMFOLLOW;
678 else if (strcmp(opt->name, "symfollow") == 0) {
679 free(opt->name, M_MOUNT);
680 opt->name = strdup("nonosymfollow", M_MOUNT);
682 else if (strcmp(opt->name, "noro") == 0) {
683 fsflags &= ~MNT_RDONLY;
686 else if (strcmp(opt->name, "rw") == 0) {
687 fsflags &= ~MNT_RDONLY;
690 else if (strcmp(opt->name, "ro") == 0) {
691 fsflags |= MNT_RDONLY;
694 else if (strcmp(opt->name, "rdonly") == 0) {
695 free(opt->name, M_MOUNT);
696 opt->name = strdup("ro", M_MOUNT);
697 fsflags |= MNT_RDONLY;
700 else if (strcmp(opt->name, "autoro") == 0) {
701 vfs_freeopt(optlist, opt);
704 else if (strcmp(opt->name, "suiddir") == 0)
705 fsflags |= MNT_SUIDDIR;
706 else if (strcmp(opt->name, "sync") == 0)
707 fsflags |= MNT_SYNCHRONOUS;
708 else if (strcmp(opt->name, "union") == 0)
709 fsflags |= MNT_UNION;
710 else if (strcmp(opt->name, "automounted") == 0) {
711 fsflags |= MNT_AUTOMOUNTED;
712 vfs_freeopt(optlist, opt);
717 * Be ultra-paranoid about making sure the type and fspath
718 * variables will fit in our mp buffers, including the
721 if (fstypelen > MFSNAMELEN || fspathlen > MNAMELEN) {
722 error = ENAMETOOLONG;
726 error = vfs_domount(td, fstype, fspath, fsflags, &optlist);
729 * See if we can mount in the read-only mode if the error code suggests
730 * that it could be possible and the mount options allow for that.
731 * Never try it if "[no]{ro|rw}" has been explicitly requested and not
732 * overridden by "autoro".
734 if (autoro && vfs_should_downgrade_to_ro_mount(fsflags, error)) {
735 printf("%s: R/W mount failed, possibly R/O media,"
736 " trying R/O mount\n", __func__);
737 fsflags |= MNT_RDONLY;
738 error = vfs_domount(td, fstype, fspath, fsflags, &optlist);
741 /* copyout the errmsg */
742 if (errmsg_pos != -1 && ((2 * errmsg_pos + 1) < fsoptions->uio_iovcnt)
743 && errmsg_len > 0 && errmsg != NULL) {
744 if (fsoptions->uio_segflg == UIO_SYSSPACE) {
746 fsoptions->uio_iov[2 * errmsg_pos + 1].iov_base,
747 fsoptions->uio_iov[2 * errmsg_pos + 1].iov_len);
750 fsoptions->uio_iov[2 * errmsg_pos + 1].iov_base,
751 fsoptions->uio_iov[2 * errmsg_pos + 1].iov_len);
756 vfs_freeopts(optlist);
763 #ifndef _SYS_SYSPROTO_H_
773 sys_mount(struct thread *td, struct mount_args *uap)
776 struct vfsconf *vfsp = NULL;
777 struct mntarg *ma = NULL;
782 * Mount flags are now 64-bits. On 32-bit architectures only
783 * 32-bits are passed in, but from here on everything handles
784 * 64-bit flags correctly.
788 AUDIT_ARG_FFLAGS(flags);
791 * Filter out MNT_ROOTFS. We do not want clients of mount() in
792 * userspace to set this flag, but we must filter it out if we want
793 * MNT_UPDATE on the root file system to work.
794 * MNT_ROOTFS should only be set by the kernel when mounting its
797 flags &= ~MNT_ROOTFS;
799 fstype = malloc(MFSNAMELEN, M_TEMP, M_WAITOK);
800 error = copyinstr(uap->type, fstype, MFSNAMELEN, NULL);
802 free(fstype, M_TEMP);
806 AUDIT_ARG_TEXT(fstype);
807 vfsp = vfs_byname_kld(fstype, td, &error);
808 free(fstype, M_TEMP);
811 if (((vfsp->vfc_flags & VFCF_SBDRY) != 0 &&
812 vfsp->vfc_vfsops_sd->vfs_cmount == NULL) ||
813 ((vfsp->vfc_flags & VFCF_SBDRY) == 0 &&
814 vfsp->vfc_vfsops->vfs_cmount == NULL))
817 ma = mount_argsu(ma, "fstype", uap->type, MFSNAMELEN);
818 ma = mount_argsu(ma, "fspath", uap->path, MNAMELEN);
819 ma = mount_argb(ma, flags & MNT_RDONLY, "noro");
820 ma = mount_argb(ma, !(flags & MNT_NOSUID), "nosuid");
821 ma = mount_argb(ma, !(flags & MNT_NOEXEC), "noexec");
823 if ((vfsp->vfc_flags & VFCF_SBDRY) != 0)
824 return (vfsp->vfc_vfsops_sd->vfs_cmount(ma, uap->data, flags));
825 return (vfsp->vfc_vfsops->vfs_cmount(ma, uap->data, flags));
829 * vfs_domount_first(): first file system mount (not update)
833 struct thread *td, /* Calling thread. */
834 struct vfsconf *vfsp, /* File system type. */
835 char *fspath, /* Mount path. */
836 struct vnode *vp, /* Vnode to be covered. */
837 uint64_t fsflags, /* Flags common to all filesystems. */
838 struct vfsoptlist **optlist /* Options local to the filesystem. */
846 ASSERT_VOP_ELOCKED(vp, __func__);
847 KASSERT((fsflags & MNT_UPDATE) == 0, ("MNT_UPDATE shouldn't be here"));
850 * If the jail of the calling thread lacks permission for this type of
851 * file system, deny immediately.
853 if (jailed(td->td_ucred) && !prison_allow(td->td_ucred,
854 vfsp->vfc_prison_flag)) {
860 * If the user is not root, ensure that they own the directory
861 * onto which we are attempting to mount.
863 error = VOP_GETATTR(vp, &va, td->td_ucred);
864 if (error == 0 && va.va_uid != td->td_ucred->cr_uid)
865 error = priv_check_cred(td->td_ucred, PRIV_VFS_ADMIN);
867 error = vinvalbuf(vp, V_SAVE, 0, 0);
868 if (error == 0 && vp->v_type != VDIR)
872 if ((vp->v_iflag & VI_MOUNT) == 0 && vp->v_mountedhere == NULL)
873 vp->v_iflag |= VI_MOUNT;
884 /* Allocate and initialize the filesystem. */
885 mp = vfs_mount_alloc(vp, vfsp, fspath, td->td_ucred);
886 /* XXXMAC: pass to vfs_mount_alloc? */
887 mp->mnt_optnew = *optlist;
888 /* Set the mount level flags. */
889 mp->mnt_flag = (fsflags & (MNT_UPDATEMASK | MNT_ROOTFS | MNT_RDONLY));
892 * Mount the filesystem.
893 * XXX The final recipients of VFS_MOUNT just overwrite the ndp they
894 * get. No freeing of cn_pnbuf.
897 if ((error = VFS_MOUNT(mp)) != 0 ||
898 (error1 = VFS_STATFS(mp, &mp->mnt_stat)) != 0 ||
899 (error1 = VFS_ROOT(mp, LK_EXCLUSIVE, &newdp)) != 0) {
902 if ((error1 = VFS_UNMOUNT(mp, 0)) != 0)
903 printf("VFS_UNMOUNT returned %d\n", error1);
906 mp->mnt_vnodecovered = NULL;
907 vfs_mount_destroy(mp);
909 vp->v_iflag &= ~VI_MOUNT;
914 VOP_UNLOCK(newdp, 0);
916 if (mp->mnt_opt != NULL)
917 vfs_freeopts(mp->mnt_opt);
918 mp->mnt_opt = mp->mnt_optnew;
922 * Prevent external consumers of mount options from reading mnt_optnew.
924 mp->mnt_optnew = NULL;
927 if ((mp->mnt_flag & MNT_ASYNC) != 0 &&
928 (mp->mnt_kern_flag & MNTK_NOASYNC) == 0)
929 mp->mnt_kern_flag |= MNTK_ASYNC;
931 mp->mnt_kern_flag &= ~MNTK_ASYNC;
934 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
937 vp->v_iflag &= ~VI_MOUNT;
939 vp->v_mountedhere = mp;
940 /* Place the new filesystem at the end of the mount list. */
941 mtx_lock(&mountlist_mtx);
942 TAILQ_INSERT_TAIL(&mountlist, mp, mnt_list);
943 mtx_unlock(&mountlist_mtx);
944 vfs_event_signal(NULL, VQ_MOUNT, 0);
945 vn_lock(newdp, LK_EXCLUSIVE | LK_RETRY);
947 EVENTHANDLER_DIRECT_INVOKE(vfs_mounted, mp, newdp, td);
948 VOP_UNLOCK(newdp, 0);
949 mountcheckdirs(vp, newdp);
951 if ((mp->mnt_flag & MNT_RDONLY) == 0)
952 vfs_allocate_syncvnode(mp);
958 * vfs_domount_update(): update of mounted file system
962 struct thread *td, /* Calling thread. */
963 struct vnode *vp, /* Mount point vnode. */
964 uint64_t fsflags, /* Flags common to all filesystems. */
965 struct vfsoptlist **optlist /* Options local to the filesystem. */
968 struct export_args export;
971 int error, export_error, len;
974 ASSERT_VOP_ELOCKED(vp, __func__);
975 KASSERT((fsflags & MNT_UPDATE) != 0, ("MNT_UPDATE should be here"));
978 if ((vp->v_vflag & VV_ROOT) == 0) {
979 if (vfs_copyopt(*optlist, "export", &export, sizeof(export))
989 * We only allow the filesystem to be reloaded if it
990 * is currently mounted read-only.
993 if ((fsflags & MNT_RELOAD) != 0 && (flag & MNT_RDONLY) == 0) {
995 return (EOPNOTSUPP); /* Needs translation */
998 * Only privileged root, or (if MNT_USER is set) the user that
999 * did the original mount is permitted to update it.
1001 error = vfs_suser(mp, td);
1006 if (vfs_busy(mp, MBF_NOWAIT)) {
1011 if ((vp->v_iflag & VI_MOUNT) != 0 || vp->v_mountedhere != NULL) {
1017 vp->v_iflag |= VI_MOUNT;
1022 if ((mp->mnt_kern_flag & MNTK_UNMOUNT) != 0) {
1027 mp->mnt_flag &= ~MNT_UPDATEMASK;
1028 mp->mnt_flag |= fsflags & (MNT_RELOAD | MNT_FORCE | MNT_UPDATE |
1029 MNT_SNAPSHOT | MNT_ROOTFS | MNT_UPDATEMASK | MNT_RDONLY);
1030 if ((mp->mnt_flag & MNT_ASYNC) == 0)
1031 mp->mnt_kern_flag &= ~MNTK_ASYNC;
1033 mp->mnt_optnew = *optlist;
1034 vfs_mergeopts(mp->mnt_optnew, mp->mnt_opt);
1037 * Mount the filesystem.
1038 * XXX The final recipients of VFS_MOUNT just overwrite the ndp they
1039 * get. No freeing of cn_pnbuf.
1041 error = VFS_MOUNT(mp);
1044 /* Process the export option. */
1045 if (error == 0 && vfs_getopt(mp->mnt_optnew, "export", &bufp,
1047 /* Assume that there is only 1 ABI for each length. */
1049 case (sizeof(struct oexport_args)):
1050 bzero(&export, sizeof(export));
1052 case (sizeof(export)):
1053 bcopy(bufp, &export, len);
1054 export_error = vfs_export(mp, &export);
1057 export_error = EINVAL;
1064 mp->mnt_flag &= ~(MNT_UPDATE | MNT_RELOAD | MNT_FORCE |
1068 * If we fail, restore old mount flags. MNT_QUOTA is special,
1069 * because it is not part of MNT_UPDATEMASK, but it could have
1070 * changed in the meantime if quotactl(2) was called.
1071 * All in all we want current value of MNT_QUOTA, not the old
1074 mp->mnt_flag = (mp->mnt_flag & MNT_QUOTA) | (flag & ~MNT_QUOTA);
1076 if ((mp->mnt_flag & MNT_ASYNC) != 0 &&
1077 (mp->mnt_kern_flag & MNTK_NOASYNC) == 0)
1078 mp->mnt_kern_flag |= MNTK_ASYNC;
1080 mp->mnt_kern_flag &= ~MNTK_ASYNC;
1086 if (mp->mnt_opt != NULL)
1087 vfs_freeopts(mp->mnt_opt);
1088 mp->mnt_opt = mp->mnt_optnew;
1090 (void)VFS_STATFS(mp, &mp->mnt_stat);
1092 * Prevent external consumers of mount options from reading
1095 mp->mnt_optnew = NULL;
1097 if ((mp->mnt_flag & MNT_RDONLY) == 0)
1098 vfs_allocate_syncvnode(mp);
1100 vfs_deallocate_syncvnode(mp);
1104 vp->v_iflag &= ~VI_MOUNT;
1107 return (error != 0 ? error : export_error);
1111 * vfs_domount(): actually attempt a filesystem mount.
1115 struct thread *td, /* Calling thread. */
1116 const char *fstype, /* Filesystem type. */
1117 char *fspath, /* Mount path. */
1118 uint64_t fsflags, /* Flags common to all filesystems. */
1119 struct vfsoptlist **optlist /* Options local to the filesystem. */
1122 struct vfsconf *vfsp;
1123 struct nameidata nd;
1129 * Be ultra-paranoid about making sure the type and fspath
1130 * variables will fit in our mp buffers, including the
1133 if (strlen(fstype) >= MFSNAMELEN || strlen(fspath) >= MNAMELEN)
1134 return (ENAMETOOLONG);
1136 if (jailed(td->td_ucred) || usermount == 0) {
1137 if ((error = priv_check(td, PRIV_VFS_MOUNT)) != 0)
1142 * Do not allow NFS export or MNT_SUIDDIR by unprivileged users.
1144 if (fsflags & MNT_EXPORTED) {
1145 error = priv_check(td, PRIV_VFS_MOUNT_EXPORTED);
1149 if (fsflags & MNT_SUIDDIR) {
1150 error = priv_check(td, PRIV_VFS_MOUNT_SUIDDIR);
1155 * Silently enforce MNT_NOSUID and MNT_USER for unprivileged users.
1157 if ((fsflags & (MNT_NOSUID | MNT_USER)) != (MNT_NOSUID | MNT_USER)) {
1158 if (priv_check(td, PRIV_VFS_MOUNT_NONUSER) != 0)
1159 fsflags |= MNT_NOSUID | MNT_USER;
1162 /* Load KLDs before we lock the covered vnode to avoid reversals. */
1164 if ((fsflags & MNT_UPDATE) == 0) {
1165 /* Don't try to load KLDs if we're mounting the root. */
1166 if (fsflags & MNT_ROOTFS)
1167 vfsp = vfs_byname(fstype);
1169 vfsp = vfs_byname_kld(fstype, td, &error);
1175 * Get vnode to be covered or mount point's vnode in case of MNT_UPDATE.
1177 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1,
1178 UIO_SYSSPACE, fspath, td);
1182 NDFREE(&nd, NDF_ONLY_PNBUF);
1184 if ((fsflags & MNT_UPDATE) == 0) {
1185 pathbuf = malloc(MNAMELEN, M_TEMP, M_WAITOK);
1186 strcpy(pathbuf, fspath);
1187 error = vn_path_to_global_path(td, vp, pathbuf, MNAMELEN);
1188 /* debug.disablefullpath == 1 results in ENODEV */
1189 if (error == 0 || error == ENODEV) {
1190 error = vfs_domount_first(td, vfsp, pathbuf, vp,
1193 free(pathbuf, M_TEMP);
1195 error = vfs_domount_update(td, vp, fsflags, optlist);
1201 * Unmount a filesystem.
1203 * Note: unmount takes a path to the vnode mounted on as argument, not
1204 * special file (as before).
1206 #ifndef _SYS_SYSPROTO_H_
1207 struct unmount_args {
1214 sys_unmount(struct thread *td, struct unmount_args *uap)
1216 struct nameidata nd;
1219 int error, id0, id1;
1221 AUDIT_ARG_VALUE(uap->flags);
1222 if (jailed(td->td_ucred) || usermount == 0) {
1223 error = priv_check(td, PRIV_VFS_UNMOUNT);
1228 pathbuf = malloc(MNAMELEN, M_TEMP, M_WAITOK);
1229 error = copyinstr(uap->path, pathbuf, MNAMELEN, NULL);
1231 free(pathbuf, M_TEMP);
1234 if (uap->flags & MNT_BYFSID) {
1235 AUDIT_ARG_TEXT(pathbuf);
1236 /* Decode the filesystem ID. */
1237 if (sscanf(pathbuf, "FSID:%d:%d", &id0, &id1) != 2) {
1238 free(pathbuf, M_TEMP);
1242 mtx_lock(&mountlist_mtx);
1243 TAILQ_FOREACH_REVERSE(mp, &mountlist, mntlist, mnt_list) {
1244 if (mp->mnt_stat.f_fsid.val[0] == id0 &&
1245 mp->mnt_stat.f_fsid.val[1] == id1) {
1250 mtx_unlock(&mountlist_mtx);
1253 * Try to find global path for path argument.
1255 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1,
1256 UIO_SYSSPACE, pathbuf, td);
1257 if (namei(&nd) == 0) {
1258 NDFREE(&nd, NDF_ONLY_PNBUF);
1259 error = vn_path_to_global_path(td, nd.ni_vp, pathbuf,
1261 if (error == 0 || error == ENODEV)
1264 mtx_lock(&mountlist_mtx);
1265 TAILQ_FOREACH_REVERSE(mp, &mountlist, mntlist, mnt_list) {
1266 if (strcmp(mp->mnt_stat.f_mntonname, pathbuf) == 0) {
1271 mtx_unlock(&mountlist_mtx);
1273 free(pathbuf, M_TEMP);
1276 * Previously we returned ENOENT for a nonexistent path and
1277 * EINVAL for a non-mountpoint. We cannot tell these apart
1278 * now, so in the !MNT_BYFSID case return the more likely
1279 * EINVAL for compatibility.
1281 return ((uap->flags & MNT_BYFSID) ? ENOENT : EINVAL);
1285 * Don't allow unmounting the root filesystem.
1287 if (mp->mnt_flag & MNT_ROOTFS) {
1291 error = dounmount(mp, uap->flags, td);
1296 * Return error if any of the vnodes, ignoring the root vnode
1297 * and the syncer vnode, have non-zero usecount.
1299 * This function is purely advisory - it can return false positives
1303 vfs_check_usecounts(struct mount *mp)
1305 struct vnode *vp, *mvp;
1307 MNT_VNODE_FOREACH_ALL(vp, mp, mvp) {
1308 if ((vp->v_vflag & VV_ROOT) == 0 && vp->v_type != VNON &&
1309 vp->v_usecount != 0) {
1311 MNT_VNODE_FOREACH_ALL_ABORT(mp, mvp);
1321 dounmount_cleanup(struct mount *mp, struct vnode *coveredvp, int mntkflags)
1324 mtx_assert(MNT_MTX(mp), MA_OWNED);
1325 mp->mnt_kern_flag &= ~mntkflags;
1326 if ((mp->mnt_kern_flag & MNTK_MWAIT) != 0) {
1327 mp->mnt_kern_flag &= ~MNTK_MWAIT;
1331 if (coveredvp != NULL) {
1332 VOP_UNLOCK(coveredvp, 0);
1335 vn_finished_write(mp);
1339 * Do the actual filesystem unmount.
1342 dounmount(struct mount *mp, int flags, struct thread *td)
1344 struct vnode *coveredvp;
1346 uint64_t async_flag;
1349 if ((coveredvp = mp->mnt_vnodecovered) != NULL) {
1350 mnt_gen_r = mp->mnt_gen;
1353 vn_lock(coveredvp, LK_EXCLUSIVE | LK_INTERLOCK | LK_RETRY);
1355 * Check for mp being unmounted while waiting for the
1356 * covered vnode lock.
1358 if (coveredvp->v_mountedhere != mp ||
1359 coveredvp->v_mountedhere->mnt_gen != mnt_gen_r) {
1360 VOP_UNLOCK(coveredvp, 0);
1368 * Only privileged root, or (if MNT_USER is set) the user that did the
1369 * original mount is permitted to unmount this filesystem.
1371 error = vfs_suser(mp, td);
1373 if (coveredvp != NULL) {
1374 VOP_UNLOCK(coveredvp, 0);
1381 vn_start_write(NULL, &mp, V_WAIT | V_MNTREF);
1383 if ((mp->mnt_kern_flag & MNTK_UNMOUNT) != 0 ||
1384 (mp->mnt_flag & MNT_UPDATE) != 0 ||
1385 !TAILQ_EMPTY(&mp->mnt_uppers)) {
1386 dounmount_cleanup(mp, coveredvp, 0);
1389 mp->mnt_kern_flag |= MNTK_UNMOUNT | MNTK_NOINSMNTQ;
1390 if (flags & MNT_NONBUSY) {
1392 error = vfs_check_usecounts(mp);
1395 dounmount_cleanup(mp, coveredvp, MNTK_UNMOUNT |
1400 /* Allow filesystems to detect that a forced unmount is in progress. */
1401 if (flags & MNT_FORCE) {
1402 mp->mnt_kern_flag |= MNTK_UNMOUNTF;
1405 * Must be done after setting MNTK_UNMOUNTF and before
1406 * waiting for mnt_lockref to become 0.
1412 if (mp->mnt_lockref) {
1413 mp->mnt_kern_flag |= MNTK_DRAINING;
1414 error = msleep(&mp->mnt_lockref, MNT_MTX(mp), PVFS,
1418 KASSERT(mp->mnt_lockref == 0,
1419 ("%s: invalid lock refcount in the drain path @ %s:%d",
1420 __func__, __FILE__, __LINE__));
1422 ("%s: invalid return value for msleep in the drain path @ %s:%d",
1423 __func__, __FILE__, __LINE__));
1425 if (mp->mnt_flag & MNT_EXPUBLIC)
1426 vfs_setpublicfs(NULL, NULL, NULL);
1429 * From now, we can claim that the use reference on the
1430 * coveredvp is ours, and the ref can be released only by
1431 * successfull unmount by us, or left for later unmount
1432 * attempt. The previously acquired hold reference is no
1433 * longer needed to protect the vnode from reuse.
1435 if (coveredvp != NULL)
1438 vfs_msync(mp, MNT_WAIT);
1440 async_flag = mp->mnt_flag & MNT_ASYNC;
1441 mp->mnt_flag &= ~MNT_ASYNC;
1442 mp->mnt_kern_flag &= ~MNTK_ASYNC;
1444 cache_purgevfs(mp, false); /* remove cache entries for this file sys */
1445 vfs_deallocate_syncvnode(mp);
1446 if ((mp->mnt_flag & MNT_RDONLY) != 0 || (flags & MNT_FORCE) != 0 ||
1447 (error = VFS_SYNC(mp, MNT_WAIT)) == 0)
1448 error = VFS_UNMOUNT(mp, flags);
1449 vn_finished_write(mp);
1451 * If we failed to flush the dirty blocks for this mount point,
1452 * undo all the cdir/rdir and rootvnode changes we made above.
1453 * Unless we failed to do so because the device is reporting that
1454 * it doesn't exist anymore.
1456 if (error && error != ENXIO) {
1458 mp->mnt_kern_flag &= ~MNTK_NOINSMNTQ;
1459 if ((mp->mnt_flag & MNT_RDONLY) == 0) {
1461 vfs_allocate_syncvnode(mp);
1464 mp->mnt_kern_flag &= ~(MNTK_UNMOUNT | MNTK_UNMOUNTF);
1465 mp->mnt_flag |= async_flag;
1466 if ((mp->mnt_flag & MNT_ASYNC) != 0 &&
1467 (mp->mnt_kern_flag & MNTK_NOASYNC) == 0)
1468 mp->mnt_kern_flag |= MNTK_ASYNC;
1469 if (mp->mnt_kern_flag & MNTK_MWAIT) {
1470 mp->mnt_kern_flag &= ~MNTK_MWAIT;
1475 VOP_UNLOCK(coveredvp, 0);
1478 mtx_lock(&mountlist_mtx);
1479 TAILQ_REMOVE(&mountlist, mp, mnt_list);
1480 mtx_unlock(&mountlist_mtx);
1481 EVENTHANDLER_DIRECT_INVOKE(vfs_unmounted, mp, td);
1482 if (coveredvp != NULL) {
1483 coveredvp->v_mountedhere = NULL;
1484 VOP_UNLOCK(coveredvp, 0);
1486 vfs_event_signal(NULL, VQ_UNMOUNT, 0);
1487 if (rootvnode != NULL && mp == rootvnode->v_mount) {
1491 if (mp == rootdevmp)
1493 vfs_mount_destroy(mp);
1498 * Report errors during filesystem mounting.
1501 vfs_mount_error(struct mount *mp, const char *fmt, ...)
1503 struct vfsoptlist *moptlist = mp->mnt_optnew;
1508 error = vfs_getopt(moptlist, "errmsg", (void **)&errmsg, &len);
1509 if (error || errmsg == NULL || len <= 0)
1513 vsnprintf(errmsg, (size_t)len, fmt, ap);
1518 vfs_opterror(struct vfsoptlist *opts, const char *fmt, ...)
1524 error = vfs_getopt(opts, "errmsg", (void **)&errmsg, &len);
1525 if (error || errmsg == NULL || len <= 0)
1529 vsnprintf(errmsg, (size_t)len, fmt, ap);
1534 * ---------------------------------------------------------------------
1535 * Functions for querying mount options/arguments from filesystems.
1539 * Check that no unknown options are given
1542 vfs_filteropt(struct vfsoptlist *opts, const char **legal)
1546 const char **t, *p, *q;
1549 TAILQ_FOREACH(opt, opts, link) {
1552 if (p[0] == 'n' && p[1] == 'o')
1554 for(t = global_opts; *t != NULL; t++) {
1555 if (strcmp(*t, p) == 0)
1558 if (strcmp(*t, q) == 0)
1564 for(t = legal; *t != NULL; t++) {
1565 if (strcmp(*t, p) == 0)
1568 if (strcmp(*t, q) == 0)
1574 snprintf(errmsg, sizeof(errmsg),
1575 "mount option <%s> is unknown", p);
1579 TAILQ_FOREACH(opt, opts, link) {
1580 if (strcmp(opt->name, "errmsg") == 0) {
1581 strncpy((char *)opt->value, errmsg, opt->len);
1586 printf("%s\n", errmsg);
1592 * Get a mount option by its name.
1594 * Return 0 if the option was found, ENOENT otherwise.
1595 * If len is non-NULL it will be filled with the length
1596 * of the option. If buf is non-NULL, it will be filled
1597 * with the address of the option.
1600 vfs_getopt(struct vfsoptlist *opts, const char *name, void **buf, int *len)
1604 KASSERT(opts != NULL, ("vfs_getopt: caller passed 'opts' as NULL"));
1606 TAILQ_FOREACH(opt, opts, link) {
1607 if (strcmp(name, opt->name) == 0) {
1620 vfs_getopt_pos(struct vfsoptlist *opts, const char *name)
1627 TAILQ_FOREACH(opt, opts, link) {
1628 if (strcmp(name, opt->name) == 0) {
1637 vfs_getopt_size(struct vfsoptlist *opts, const char *name, off_t *value)
1639 char *opt_value, *vtp;
1643 error = vfs_getopt(opts, name, (void **)&opt_value, &opt_len);
1646 if (opt_len == 0 || opt_value == NULL)
1648 if (opt_value[0] == '\0' || opt_value[opt_len - 1] != '\0')
1650 iv = strtoq(opt_value, &vtp, 0);
1651 if (vtp == opt_value || (vtp[0] != '\0' && vtp[1] != '\0'))
1678 vfs_getopts(struct vfsoptlist *opts, const char *name, int *error)
1683 TAILQ_FOREACH(opt, opts, link) {
1684 if (strcmp(name, opt->name) != 0)
1687 if (opt->len == 0 ||
1688 ((char *)opt->value)[opt->len - 1] != '\0') {
1692 return (opt->value);
1699 vfs_flagopt(struct vfsoptlist *opts, const char *name, uint64_t *w,
1704 TAILQ_FOREACH(opt, opts, link) {
1705 if (strcmp(name, opt->name) == 0) {
1718 vfs_scanopt(struct vfsoptlist *opts, const char *name, const char *fmt, ...)
1724 KASSERT(opts != NULL, ("vfs_getopt: caller passed 'opts' as NULL"));
1726 TAILQ_FOREACH(opt, opts, link) {
1727 if (strcmp(name, opt->name) != 0)
1730 if (opt->len == 0 || opt->value == NULL)
1732 if (((char *)opt->value)[opt->len - 1] != '\0')
1735 ret = vsscanf(opt->value, fmt, ap);
1743 vfs_setopt(struct vfsoptlist *opts, const char *name, void *value, int len)
1747 TAILQ_FOREACH(opt, opts, link) {
1748 if (strcmp(name, opt->name) != 0)
1751 if (opt->value == NULL)
1754 if (opt->len != len)
1756 bcopy(value, opt->value, len);
1764 vfs_setopt_part(struct vfsoptlist *opts, const char *name, void *value, int len)
1768 TAILQ_FOREACH(opt, opts, link) {
1769 if (strcmp(name, opt->name) != 0)
1772 if (opt->value == NULL)
1778 bcopy(value, opt->value, len);
1786 vfs_setopts(struct vfsoptlist *opts, const char *name, const char *value)
1790 TAILQ_FOREACH(opt, opts, link) {
1791 if (strcmp(name, opt->name) != 0)
1794 if (opt->value == NULL)
1795 opt->len = strlen(value) + 1;
1796 else if (strlcpy(opt->value, value, opt->len) >= opt->len)
1804 * Find and copy a mount option.
1806 * The size of the buffer has to be specified
1807 * in len, if it is not the same length as the
1808 * mount option, EINVAL is returned.
1809 * Returns ENOENT if the option is not found.
1812 vfs_copyopt(struct vfsoptlist *opts, const char *name, void *dest, int len)
1816 KASSERT(opts != NULL, ("vfs_copyopt: caller passed 'opts' as NULL"));
1818 TAILQ_FOREACH(opt, opts, link) {
1819 if (strcmp(name, opt->name) == 0) {
1821 if (len != opt->len)
1823 bcopy(opt->value, dest, opt->len);
1831 __vfs_statfs(struct mount *mp, struct statfs *sbp)
1835 error = mp->mnt_op->vfs_statfs(mp, &mp->mnt_stat);
1836 if (sbp != &mp->mnt_stat)
1837 *sbp = mp->mnt_stat;
1842 vfs_mountedfrom(struct mount *mp, const char *from)
1845 bzero(mp->mnt_stat.f_mntfromname, sizeof mp->mnt_stat.f_mntfromname);
1846 strlcpy(mp->mnt_stat.f_mntfromname, from,
1847 sizeof mp->mnt_stat.f_mntfromname);
1851 * ---------------------------------------------------------------------
1852 * This is the api for building mount args and mounting filesystems from
1853 * inside the kernel.
1855 * The API works by accumulation of individual args. First error is
1858 * XXX: should be documented in new manpage kernel_mount(9)
1861 /* A memory allocation which must be freed when we are done */
1863 SLIST_ENTRY(mntaarg) next;
1866 /* The header for the mount arguments */
1871 SLIST_HEAD(, mntaarg) list;
1875 * Add a boolean argument.
1877 * flag is the boolean value.
1878 * name must start with "no".
1881 mount_argb(struct mntarg *ma, int flag, const char *name)
1884 KASSERT(name[0] == 'n' && name[1] == 'o',
1885 ("mount_argb(...,%s): name must start with 'no'", name));
1887 return (mount_arg(ma, name + (flag ? 2 : 0), NULL, 0));
1891 * Add an argument printf style
1894 mount_argf(struct mntarg *ma, const char *name, const char *fmt, ...)
1897 struct mntaarg *maa;
1902 ma = malloc(sizeof *ma, M_MOUNT, M_WAITOK | M_ZERO);
1903 SLIST_INIT(&ma->list);
1908 ma->v = realloc(ma->v, sizeof *ma->v * (ma->len + 2),
1910 ma->v[ma->len].iov_base = (void *)(uintptr_t)name;
1911 ma->v[ma->len].iov_len = strlen(name) + 1;
1914 sb = sbuf_new_auto();
1916 sbuf_vprintf(sb, fmt, ap);
1919 len = sbuf_len(sb) + 1;
1920 maa = malloc(sizeof *maa + len, M_MOUNT, M_WAITOK | M_ZERO);
1921 SLIST_INSERT_HEAD(&ma->list, maa, next);
1922 bcopy(sbuf_data(sb), maa + 1, len);
1925 ma->v[ma->len].iov_base = maa + 1;
1926 ma->v[ma->len].iov_len = len;
1933 * Add an argument which is a userland string.
1936 mount_argsu(struct mntarg *ma, const char *name, const void *val, int len)
1938 struct mntaarg *maa;
1944 ma = malloc(sizeof *ma, M_MOUNT, M_WAITOK | M_ZERO);
1945 SLIST_INIT(&ma->list);
1949 maa = malloc(sizeof *maa + len, M_MOUNT, M_WAITOK | M_ZERO);
1950 SLIST_INSERT_HEAD(&ma->list, maa, next);
1951 tbuf = (void *)(maa + 1);
1952 ma->error = copyinstr(val, tbuf, len, NULL);
1953 return (mount_arg(ma, name, tbuf, -1));
1959 * If length is -1, treat value as a C string.
1962 mount_arg(struct mntarg *ma, const char *name, const void *val, int len)
1966 ma = malloc(sizeof *ma, M_MOUNT, M_WAITOK | M_ZERO);
1967 SLIST_INIT(&ma->list);
1972 ma->v = realloc(ma->v, sizeof *ma->v * (ma->len + 2),
1974 ma->v[ma->len].iov_base = (void *)(uintptr_t)name;
1975 ma->v[ma->len].iov_len = strlen(name) + 1;
1978 ma->v[ma->len].iov_base = (void *)(uintptr_t)val;
1980 ma->v[ma->len].iov_len = strlen(val) + 1;
1982 ma->v[ma->len].iov_len = len;
1988 * Free a mntarg structure
1991 free_mntarg(struct mntarg *ma)
1993 struct mntaarg *maa;
1995 while (!SLIST_EMPTY(&ma->list)) {
1996 maa = SLIST_FIRST(&ma->list);
1997 SLIST_REMOVE_HEAD(&ma->list, next);
2000 free(ma->v, M_MOUNT);
2005 * Mount a filesystem
2008 kernel_mount(struct mntarg *ma, uint64_t flags)
2013 KASSERT(ma != NULL, ("kernel_mount NULL ma"));
2014 KASSERT(ma->v != NULL, ("kernel_mount NULL ma->v"));
2015 KASSERT(!(ma->len & 1), ("kernel_mount odd ma->len (%d)", ma->len));
2017 auio.uio_iov = ma->v;
2018 auio.uio_iovcnt = ma->len;
2019 auio.uio_segflg = UIO_SYSSPACE;
2023 error = vfs_donmount(curthread, flags, &auio);
2029 * A printflike function to mount a filesystem.
2032 kernel_vmount(int flags, ...)
2034 struct mntarg *ma = NULL;
2040 va_start(ap, flags);
2042 cp = va_arg(ap, const char *);
2045 vp = va_arg(ap, const void *);
2046 ma = mount_arg(ma, cp, vp, (vp != NULL ? -1 : 0));
2050 error = kernel_mount(ma, flags);
2055 vfs_oexport_conv(const struct oexport_args *oexp, struct export_args *exp)
2058 bcopy(oexp, exp, sizeof(*oexp));
2059 exp->ex_numsecflavors = 0;