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/fcntl.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>
55 #include <sys/syscallsubr.h>
56 #include <sys/sysproto.h>
58 #include <sys/sysctl.h>
59 #include <sys/sysent.h>
60 #include <sys/systm.h>
61 #include <sys/vnode.h>
64 #include <geom/geom.h>
66 #include <machine/stdarg.h>
68 #include <security/audit/audit.h>
69 #include <security/mac/mac_framework.h>
71 #define VFS_MOUNTARG_SIZE_MAX (1024 * 64)
73 static int vfs_domount(struct thread *td, const char *fstype, char *fspath,
74 uint64_t fsflags, struct vfsoptlist **optlist);
75 static void free_mntarg(struct mntarg *ma);
77 static int usermount = 0;
78 SYSCTL_INT(_vfs, OID_AUTO, usermount, CTLFLAG_RW, &usermount, 0,
79 "Unprivileged users may mount and unmount file systems");
81 static bool default_autoro = false;
82 SYSCTL_BOOL(_vfs, OID_AUTO, default_autoro, CTLFLAG_RW, &default_autoro, 0,
83 "Retry failed r/w mount as r/o if no explicit ro/rw option is specified");
85 MALLOC_DEFINE(M_MOUNT, "mount", "vfs mount structure");
86 MALLOC_DEFINE(M_STATFS, "statfs", "statfs structure");
87 static uma_zone_t mount_zone;
89 /* List of mounted filesystems. */
90 struct mntlist mountlist = TAILQ_HEAD_INITIALIZER(mountlist);
92 /* For any iteration/modification of mountlist */
93 struct mtx mountlist_mtx;
94 MTX_SYSINIT(mountlist, &mountlist_mtx, "mountlist", MTX_DEF);
97 * Global opts, taken by all filesystems
99 static const char *global_opts[] = {
111 mount_init(void *mem, int size, int flags)
115 mp = (struct mount *)mem;
116 mtx_init(&mp->mnt_mtx, "struct mount mtx", NULL, MTX_DEF);
117 lockinit(&mp->mnt_explock, PVFS, "explock", 0, 0);
122 mount_fini(void *mem, int size)
126 mp = (struct mount *)mem;
127 lockdestroy(&mp->mnt_explock);
128 mtx_destroy(&mp->mnt_mtx);
132 vfs_mount_init(void *dummy __unused)
135 mount_zone = uma_zcreate("Mountpoints", sizeof(struct mount), NULL,
136 NULL, mount_init, mount_fini, UMA_ALIGN_PTR, UMA_ZONE_NOFREE);
138 SYSINIT(vfs_mount, SI_SUB_VFS, SI_ORDER_ANY, vfs_mount_init, NULL);
141 * ---------------------------------------------------------------------
142 * Functions for building and sanitizing the mount options
145 /* Remove one mount option. */
147 vfs_freeopt(struct vfsoptlist *opts, struct vfsopt *opt)
150 TAILQ_REMOVE(opts, opt, link);
151 free(opt->name, M_MOUNT);
152 if (opt->value != NULL)
153 free(opt->value, M_MOUNT);
157 /* Release all resources related to the mount options. */
159 vfs_freeopts(struct vfsoptlist *opts)
163 while (!TAILQ_EMPTY(opts)) {
164 opt = TAILQ_FIRST(opts);
165 vfs_freeopt(opts, opt);
171 vfs_deleteopt(struct vfsoptlist *opts, const char *name)
173 struct vfsopt *opt, *temp;
177 TAILQ_FOREACH_SAFE(opt, opts, link, temp) {
178 if (strcmp(opt->name, name) == 0)
179 vfs_freeopt(opts, opt);
184 vfs_isopt_ro(const char *opt)
187 if (strcmp(opt, "ro") == 0 || strcmp(opt, "rdonly") == 0 ||
188 strcmp(opt, "norw") == 0)
194 vfs_isopt_rw(const char *opt)
197 if (strcmp(opt, "rw") == 0 || strcmp(opt, "noro") == 0)
203 * Check if options are equal (with or without the "no" prefix).
206 vfs_equalopts(const char *opt1, const char *opt2)
210 /* "opt" vs. "opt" or "noopt" vs. "noopt" */
211 if (strcmp(opt1, opt2) == 0)
213 /* "noopt" vs. "opt" */
214 if (strncmp(opt1, "no", 2) == 0 && strcmp(opt1 + 2, opt2) == 0)
216 /* "opt" vs. "noopt" */
217 if (strncmp(opt2, "no", 2) == 0 && strcmp(opt1, opt2 + 2) == 0)
219 while ((p = strchr(opt1, '.')) != NULL &&
220 !strncmp(opt1, opt2, ++p - opt1)) {
223 /* "foo.noopt" vs. "foo.opt" */
224 if (strncmp(opt1, "no", 2) == 0 && strcmp(opt1 + 2, opt2) == 0)
226 /* "foo.opt" vs. "foo.noopt" */
227 if (strncmp(opt2, "no", 2) == 0 && strcmp(opt1, opt2 + 2) == 0)
230 /* "ro" / "rdonly" / "norw" / "rw" / "noro" */
231 if ((vfs_isopt_ro(opt1) || vfs_isopt_rw(opt1)) &&
232 (vfs_isopt_ro(opt2) || vfs_isopt_rw(opt2)))
238 * If a mount option is specified several times,
239 * (with or without the "no" prefix) only keep
240 * the last occurrence of it.
243 vfs_sanitizeopts(struct vfsoptlist *opts)
245 struct vfsopt *opt, *opt2, *tmp;
247 TAILQ_FOREACH_REVERSE(opt, opts, vfsoptlist, link) {
248 opt2 = TAILQ_PREV(opt, vfsoptlist, link);
249 while (opt2 != NULL) {
250 if (vfs_equalopts(opt->name, opt2->name)) {
251 tmp = TAILQ_PREV(opt2, vfsoptlist, link);
252 vfs_freeopt(opts, opt2);
255 opt2 = TAILQ_PREV(opt2, vfsoptlist, link);
262 * Build a linked list of mount options from a struct uio.
265 vfs_buildopts(struct uio *auio, struct vfsoptlist **options)
267 struct vfsoptlist *opts;
269 size_t memused, namelen, optlen;
270 unsigned int i, iovcnt;
273 opts = malloc(sizeof(struct vfsoptlist), M_MOUNT, M_WAITOK);
276 iovcnt = auio->uio_iovcnt;
277 for (i = 0; i < iovcnt; i += 2) {
278 namelen = auio->uio_iov[i].iov_len;
279 optlen = auio->uio_iov[i + 1].iov_len;
280 memused += sizeof(struct vfsopt) + optlen + namelen;
282 * Avoid consuming too much memory, and attempts to overflow
285 if (memused > VFS_MOUNTARG_SIZE_MAX ||
286 optlen > VFS_MOUNTARG_SIZE_MAX ||
287 namelen > VFS_MOUNTARG_SIZE_MAX) {
292 opt = malloc(sizeof(struct vfsopt), M_MOUNT, M_WAITOK);
293 opt->name = malloc(namelen, M_MOUNT, M_WAITOK);
300 * Do this early, so jumps to "bad" will free the current
303 TAILQ_INSERT_TAIL(opts, opt, link);
305 if (auio->uio_segflg == UIO_SYSSPACE) {
306 bcopy(auio->uio_iov[i].iov_base, opt->name, namelen);
308 error = copyin(auio->uio_iov[i].iov_base, opt->name,
313 /* Ensure names are null-terminated strings. */
314 if (namelen == 0 || opt->name[namelen - 1] != '\0') {
320 opt->value = malloc(optlen, M_MOUNT, M_WAITOK);
321 if (auio->uio_segflg == UIO_SYSSPACE) {
322 bcopy(auio->uio_iov[i + 1].iov_base, opt->value,
325 error = copyin(auio->uio_iov[i + 1].iov_base,
332 vfs_sanitizeopts(opts);
341 * Merge the old mount options with the new ones passed
342 * in the MNT_UPDATE case.
344 * XXX: This function will keep a "nofoo" option in the new
345 * options. E.g, if the option's canonical name is "foo",
346 * "nofoo" ends up in the mount point's active options.
349 vfs_mergeopts(struct vfsoptlist *toopts, struct vfsoptlist *oldopts)
351 struct vfsopt *opt, *new;
353 TAILQ_FOREACH(opt, oldopts, link) {
354 new = malloc(sizeof(struct vfsopt), M_MOUNT, M_WAITOK);
355 new->name = strdup(opt->name, M_MOUNT);
357 new->value = malloc(opt->len, M_MOUNT, M_WAITOK);
358 bcopy(opt->value, new->value, opt->len);
362 new->seen = opt->seen;
363 TAILQ_INSERT_HEAD(toopts, new, link);
365 vfs_sanitizeopts(toopts);
369 * Mount a filesystem.
371 #ifndef _SYS_SYSPROTO_H_
379 sys_nmount(struct thread *td, struct nmount_args *uap)
387 * Mount flags are now 64-bits. On 32-bit archtectures only
388 * 32-bits are passed in, but from here on everything handles
389 * 64-bit flags correctly.
393 AUDIT_ARG_FFLAGS(flags);
394 CTR4(KTR_VFS, "%s: iovp %p with iovcnt %d and flags %d", __func__,
395 uap->iovp, uap->iovcnt, flags);
398 * Filter out MNT_ROOTFS. We do not want clients of nmount() in
399 * userspace to set this flag, but we must filter it out if we want
400 * MNT_UPDATE on the root file system to work.
401 * MNT_ROOTFS should only be set by the kernel when mounting its
404 flags &= ~MNT_ROOTFS;
406 iovcnt = uap->iovcnt;
408 * Check that we have an even number of iovec's
409 * and that we have at least two options.
411 if ((iovcnt & 1) || (iovcnt < 4)) {
412 CTR2(KTR_VFS, "%s: failed for invalid iovcnt %d", __func__,
417 error = copyinuio(uap->iovp, iovcnt, &auio);
419 CTR2(KTR_VFS, "%s: failed for invalid uio op with %d errno",
423 error = vfs_donmount(td, flags, auio);
430 * ---------------------------------------------------------------------
431 * Various utility functions
435 vfs_ref(struct mount *mp)
438 CTR2(KTR_VFS, "%s: mp %p", __func__, mp);
445 vfs_rel(struct mount *mp)
448 CTR2(KTR_VFS, "%s: mp %p", __func__, mp);
455 * Allocate and initialize the mount point struct.
458 vfs_mount_alloc(struct vnode *vp, struct vfsconf *vfsp, const char *fspath,
463 mp = uma_zalloc(mount_zone, M_WAITOK);
464 bzero(&mp->mnt_startzero,
465 __rangeof(struct mount, mnt_startzero, mnt_endzero));
466 TAILQ_INIT(&mp->mnt_nvnodelist);
467 mp->mnt_nvnodelistsize = 0;
468 TAILQ_INIT(&mp->mnt_activevnodelist);
469 mp->mnt_activevnodelistsize = 0;
471 (void) vfs_busy(mp, MBF_NOWAIT);
472 atomic_add_acq_int(&vfsp->vfc_refcount, 1);
473 mp->mnt_op = vfsp->vfc_vfsops;
475 mp->mnt_stat.f_type = vfsp->vfc_typenum;
477 strlcpy(mp->mnt_stat.f_fstypename, vfsp->vfc_name, MFSNAMELEN);
478 mp->mnt_vnodecovered = vp;
479 mp->mnt_cred = crdup(cred);
480 mp->mnt_stat.f_owner = cred->cr_uid;
481 strlcpy(mp->mnt_stat.f_mntonname, fspath, MNAMELEN);
482 mp->mnt_iosize_max = DFLTPHYS;
485 mac_mount_create(cred, mp);
487 arc4rand(&mp->mnt_hashseed, sizeof mp->mnt_hashseed, 0);
488 TAILQ_INIT(&mp->mnt_uppers);
493 * Destroy the mount struct previously allocated by vfs_mount_alloc().
496 vfs_mount_destroy(struct mount *mp)
500 mp->mnt_kern_flag |= MNTK_REFEXPIRE;
501 if (mp->mnt_kern_flag & MNTK_MWAIT) {
502 mp->mnt_kern_flag &= ~MNTK_MWAIT;
506 msleep(mp, MNT_MTX(mp), PVFS, "mntref", 0);
507 KASSERT(mp->mnt_ref == 0,
508 ("%s: invalid refcount in the drain path @ %s:%d", __func__,
509 __FILE__, __LINE__));
510 if (mp->mnt_writeopcount != 0)
511 panic("vfs_mount_destroy: nonzero writeopcount");
512 if (mp->mnt_secondary_writes != 0)
513 panic("vfs_mount_destroy: nonzero secondary_writes");
514 atomic_subtract_rel_int(&mp->mnt_vfc->vfc_refcount, 1);
515 if (!TAILQ_EMPTY(&mp->mnt_nvnodelist)) {
518 TAILQ_FOREACH(vp, &mp->mnt_nvnodelist, v_nmntvnodes)
519 vn_printf(vp, "dangling vnode ");
520 panic("unmount: dangling vnode");
522 KASSERT(TAILQ_EMPTY(&mp->mnt_uppers), ("mnt_uppers"));
523 if (mp->mnt_nvnodelistsize != 0)
524 panic("vfs_mount_destroy: nonzero nvnodelistsize");
525 if (mp->mnt_activevnodelistsize != 0)
526 panic("vfs_mount_destroy: nonzero activevnodelistsize");
527 if (mp->mnt_lockref != 0)
528 panic("vfs_mount_destroy: nonzero lock refcount");
530 if (mp->mnt_vnodecovered != NULL)
531 vrele(mp->mnt_vnodecovered);
533 mac_mount_destroy(mp);
535 if (mp->mnt_opt != NULL)
536 vfs_freeopts(mp->mnt_opt);
537 crfree(mp->mnt_cred);
538 uma_zfree(mount_zone, mp);
542 vfs_should_downgrade_to_ro_mount(uint64_t fsflags, int error)
544 /* This is an upgrade of an exisiting mount. */
545 if ((fsflags & MNT_UPDATE) != 0)
547 /* This is already an R/O mount. */
548 if ((fsflags & MNT_RDONLY) != 0)
552 case ENODEV: /* generic, geom, ... */
553 case EACCES: /* cam/scsi, ... */
554 case EROFS: /* md, mmcsd, ... */
556 * These errors can be returned by the storage layer to signal
557 * that the media is read-only. No harm in the R/O mount
558 * attempt if the error was returned for some other reason.
567 vfs_donmount(struct thread *td, uint64_t fsflags, struct uio *fsoptions)
569 struct vfsoptlist *optlist;
570 struct vfsopt *opt, *tmp_opt;
571 char *fstype, *fspath, *errmsg;
572 int error, fstypelen, fspathlen, errmsg_len, errmsg_pos;
575 errmsg = fspath = NULL;
576 errmsg_len = fspathlen = 0;
578 autoro = default_autoro;
580 error = vfs_buildopts(fsoptions, &optlist);
584 if (vfs_getopt(optlist, "errmsg", (void **)&errmsg, &errmsg_len) == 0)
585 errmsg_pos = vfs_getopt_pos(optlist, "errmsg");
588 * We need these two options before the others,
589 * and they are mandatory for any filesystem.
590 * Ensure they are NUL terminated as well.
593 error = vfs_getopt(optlist, "fstype", (void **)&fstype, &fstypelen);
594 if (error || fstypelen <= 0 || fstype[fstypelen - 1] != '\0') {
597 strncpy(errmsg, "Invalid fstype", errmsg_len);
601 error = vfs_getopt(optlist, "fspath", (void **)&fspath, &fspathlen);
602 if (error || fspathlen <= 0 || fspath[fspathlen - 1] != '\0') {
605 strncpy(errmsg, "Invalid fspath", errmsg_len);
610 * We need to see if we have the "update" option
611 * before we call vfs_domount(), since vfs_domount() has special
612 * logic based on MNT_UPDATE. This is very important
613 * when we want to update the root filesystem.
615 TAILQ_FOREACH_SAFE(opt, optlist, link, tmp_opt) {
616 if (strcmp(opt->name, "update") == 0) {
617 fsflags |= MNT_UPDATE;
618 vfs_freeopt(optlist, opt);
620 else if (strcmp(opt->name, "async") == 0)
621 fsflags |= MNT_ASYNC;
622 else if (strcmp(opt->name, "force") == 0) {
623 fsflags |= MNT_FORCE;
624 vfs_freeopt(optlist, opt);
626 else if (strcmp(opt->name, "reload") == 0) {
627 fsflags |= MNT_RELOAD;
628 vfs_freeopt(optlist, opt);
630 else if (strcmp(opt->name, "multilabel") == 0)
631 fsflags |= MNT_MULTILABEL;
632 else if (strcmp(opt->name, "noasync") == 0)
633 fsflags &= ~MNT_ASYNC;
634 else if (strcmp(opt->name, "noatime") == 0)
635 fsflags |= MNT_NOATIME;
636 else if (strcmp(opt->name, "atime") == 0) {
637 free(opt->name, M_MOUNT);
638 opt->name = strdup("nonoatime", M_MOUNT);
640 else if (strcmp(opt->name, "noclusterr") == 0)
641 fsflags |= MNT_NOCLUSTERR;
642 else if (strcmp(opt->name, "clusterr") == 0) {
643 free(opt->name, M_MOUNT);
644 opt->name = strdup("nonoclusterr", M_MOUNT);
646 else if (strcmp(opt->name, "noclusterw") == 0)
647 fsflags |= MNT_NOCLUSTERW;
648 else if (strcmp(opt->name, "clusterw") == 0) {
649 free(opt->name, M_MOUNT);
650 opt->name = strdup("nonoclusterw", M_MOUNT);
652 else if (strcmp(opt->name, "noexec") == 0)
653 fsflags |= MNT_NOEXEC;
654 else if (strcmp(opt->name, "exec") == 0) {
655 free(opt->name, M_MOUNT);
656 opt->name = strdup("nonoexec", M_MOUNT);
658 else if (strcmp(opt->name, "nosuid") == 0)
659 fsflags |= MNT_NOSUID;
660 else if (strcmp(opt->name, "suid") == 0) {
661 free(opt->name, M_MOUNT);
662 opt->name = strdup("nonosuid", M_MOUNT);
664 else if (strcmp(opt->name, "nosymfollow") == 0)
665 fsflags |= MNT_NOSYMFOLLOW;
666 else if (strcmp(opt->name, "symfollow") == 0) {
667 free(opt->name, M_MOUNT);
668 opt->name = strdup("nonosymfollow", M_MOUNT);
670 else if (strcmp(opt->name, "noro") == 0) {
671 fsflags &= ~MNT_RDONLY;
674 else if (strcmp(opt->name, "rw") == 0) {
675 fsflags &= ~MNT_RDONLY;
678 else if (strcmp(opt->name, "ro") == 0) {
679 fsflags |= MNT_RDONLY;
682 else if (strcmp(opt->name, "rdonly") == 0) {
683 free(opt->name, M_MOUNT);
684 opt->name = strdup("ro", M_MOUNT);
685 fsflags |= MNT_RDONLY;
688 else if (strcmp(opt->name, "autoro") == 0) {
689 vfs_freeopt(optlist, opt);
692 else if (strcmp(opt->name, "suiddir") == 0)
693 fsflags |= MNT_SUIDDIR;
694 else if (strcmp(opt->name, "sync") == 0)
695 fsflags |= MNT_SYNCHRONOUS;
696 else if (strcmp(opt->name, "union") == 0)
697 fsflags |= MNT_UNION;
698 else if (strcmp(opt->name, "automounted") == 0) {
699 fsflags |= MNT_AUTOMOUNTED;
700 vfs_freeopt(optlist, opt);
705 * Be ultra-paranoid about making sure the type and fspath
706 * variables will fit in our mp buffers, including the
709 if (fstypelen > MFSNAMELEN || fspathlen > MNAMELEN) {
710 error = ENAMETOOLONG;
714 error = vfs_domount(td, fstype, fspath, fsflags, &optlist);
717 * See if we can mount in the read-only mode if the error code suggests
718 * that it could be possible and the mount options allow for that.
719 * Never try it if "[no]{ro|rw}" has been explicitly requested and not
720 * overridden by "autoro".
722 if (autoro && vfs_should_downgrade_to_ro_mount(fsflags, error)) {
723 printf("%s: R/W mount failed, possibly R/O media,"
724 " trying R/O mount\n", __func__);
725 fsflags |= MNT_RDONLY;
726 error = vfs_domount(td, fstype, fspath, fsflags, &optlist);
729 /* copyout the errmsg */
730 if (errmsg_pos != -1 && ((2 * errmsg_pos + 1) < fsoptions->uio_iovcnt)
731 && errmsg_len > 0 && errmsg != NULL) {
732 if (fsoptions->uio_segflg == UIO_SYSSPACE) {
734 fsoptions->uio_iov[2 * errmsg_pos + 1].iov_base,
735 fsoptions->uio_iov[2 * errmsg_pos + 1].iov_len);
738 fsoptions->uio_iov[2 * errmsg_pos + 1].iov_base,
739 fsoptions->uio_iov[2 * errmsg_pos + 1].iov_len);
744 vfs_freeopts(optlist);
751 #ifndef _SYS_SYSPROTO_H_
761 sys_mount(struct thread *td, struct mount_args *uap)
764 struct vfsconf *vfsp = NULL;
765 struct mntarg *ma = NULL;
770 * Mount flags are now 64-bits. On 32-bit architectures only
771 * 32-bits are passed in, but from here on everything handles
772 * 64-bit flags correctly.
776 AUDIT_ARG_FFLAGS(flags);
779 * Filter out MNT_ROOTFS. We do not want clients of mount() in
780 * userspace to set this flag, but we must filter it out if we want
781 * MNT_UPDATE on the root file system to work.
782 * MNT_ROOTFS should only be set by the kernel when mounting its
785 flags &= ~MNT_ROOTFS;
787 fstype = malloc(MFSNAMELEN, M_TEMP, M_WAITOK);
788 error = copyinstr(uap->type, fstype, MFSNAMELEN, NULL);
790 free(fstype, M_TEMP);
794 AUDIT_ARG_TEXT(fstype);
795 vfsp = vfs_byname_kld(fstype, td, &error);
796 free(fstype, M_TEMP);
799 if (vfsp->vfc_vfsops->vfs_cmount == NULL)
802 ma = mount_argsu(ma, "fstype", uap->type, MFSNAMELEN);
803 ma = mount_argsu(ma, "fspath", uap->path, MNAMELEN);
804 ma = mount_argb(ma, flags & MNT_RDONLY, "noro");
805 ma = mount_argb(ma, !(flags & MNT_NOSUID), "nosuid");
806 ma = mount_argb(ma, !(flags & MNT_NOEXEC), "noexec");
808 error = vfsp->vfc_vfsops->vfs_cmount(ma, uap->data, flags);
813 * vfs_domount_first(): first file system mount (not update)
817 struct thread *td, /* Calling thread. */
818 struct vfsconf *vfsp, /* File system type. */
819 char *fspath, /* Mount path. */
820 struct vnode *vp, /* Vnode to be covered. */
821 uint64_t fsflags, /* Flags common to all filesystems. */
822 struct vfsoptlist **optlist /* Options local to the filesystem. */
830 ASSERT_VOP_ELOCKED(vp, __func__);
831 KASSERT((fsflags & MNT_UPDATE) == 0, ("MNT_UPDATE shouldn't be here"));
834 * If the user is not root, ensure that they own the directory
835 * onto which we are attempting to mount.
837 error = VOP_GETATTR(vp, &va, td->td_ucred);
838 if (error == 0 && va.va_uid != td->td_ucred->cr_uid)
839 error = priv_check_cred(td->td_ucred, PRIV_VFS_ADMIN, 0);
841 error = vinvalbuf(vp, V_SAVE, 0, 0);
842 if (error == 0 && vp->v_type != VDIR)
846 if ((vp->v_iflag & VI_MOUNT) == 0 && vp->v_mountedhere == NULL)
847 vp->v_iflag |= VI_MOUNT;
858 /* Allocate and initialize the filesystem. */
859 mp = vfs_mount_alloc(vp, vfsp, fspath, td->td_ucred);
860 /* XXXMAC: pass to vfs_mount_alloc? */
861 mp->mnt_optnew = *optlist;
862 /* Set the mount level flags. */
863 mp->mnt_flag = (fsflags & (MNT_UPDATEMASK | MNT_ROOTFS | MNT_RDONLY));
866 * Mount the filesystem.
867 * XXX The final recipients of VFS_MOUNT just overwrite the ndp they
868 * get. No freeing of cn_pnbuf.
871 if ((error = VFS_MOUNT(mp)) != 0 ||
872 (error1 = VFS_STATFS(mp, &mp->mnt_stat)) != 0 ||
873 (error1 = VFS_ROOT(mp, LK_EXCLUSIVE, &newdp)) != 0) {
876 if ((error1 = VFS_UNMOUNT(mp, 0)) != 0)
877 printf("VFS_UNMOUNT returned %d\n", error1);
880 mp->mnt_vnodecovered = NULL;
881 vfs_mount_destroy(mp);
883 vp->v_iflag &= ~VI_MOUNT;
888 VOP_UNLOCK(newdp, 0);
890 if (mp->mnt_opt != NULL)
891 vfs_freeopts(mp->mnt_opt);
892 mp->mnt_opt = mp->mnt_optnew;
896 * Prevent external consumers of mount options from reading mnt_optnew.
898 mp->mnt_optnew = NULL;
901 if ((mp->mnt_flag & MNT_ASYNC) != 0 &&
902 (mp->mnt_kern_flag & MNTK_NOASYNC) == 0)
903 mp->mnt_kern_flag |= MNTK_ASYNC;
905 mp->mnt_kern_flag &= ~MNTK_ASYNC;
908 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
911 vp->v_iflag &= ~VI_MOUNT;
913 vp->v_mountedhere = mp;
914 /* Place the new filesystem at the end of the mount list. */
915 mtx_lock(&mountlist_mtx);
916 TAILQ_INSERT_TAIL(&mountlist, mp, mnt_list);
917 mtx_unlock(&mountlist_mtx);
918 vfs_event_signal(NULL, VQ_MOUNT, 0);
919 vn_lock(newdp, LK_EXCLUSIVE | LK_RETRY);
921 EVENTHANDLER_INVOKE(vfs_mounted, mp, newdp, td);
922 VOP_UNLOCK(newdp, 0);
923 mountcheckdirs(vp, newdp);
925 if ((mp->mnt_flag & MNT_RDONLY) == 0)
926 vfs_allocate_syncvnode(mp);
932 * vfs_domount_update(): update of mounted file system
936 struct thread *td, /* Calling thread. */
937 struct vnode *vp, /* Mount point vnode. */
938 uint64_t fsflags, /* Flags common to all filesystems. */
939 struct vfsoptlist **optlist /* Options local to the filesystem. */
942 struct export_args export;
945 int error, export_error, len;
948 ASSERT_VOP_ELOCKED(vp, __func__);
949 KASSERT((fsflags & MNT_UPDATE) != 0, ("MNT_UPDATE should be here"));
952 if ((vp->v_vflag & VV_ROOT) == 0) {
953 if (vfs_copyopt(*optlist, "export", &export, sizeof(export))
963 * We only allow the filesystem to be reloaded if it
964 * is currently mounted read-only.
967 if ((fsflags & MNT_RELOAD) != 0 && (flag & MNT_RDONLY) == 0) {
969 return (EOPNOTSUPP); /* Needs translation */
972 * Only privileged root, or (if MNT_USER is set) the user that
973 * did the original mount is permitted to update it.
975 error = vfs_suser(mp, td);
980 if (vfs_busy(mp, MBF_NOWAIT)) {
985 if ((vp->v_iflag & VI_MOUNT) != 0 || vp->v_mountedhere != NULL) {
991 vp->v_iflag |= VI_MOUNT;
996 if ((mp->mnt_kern_flag & MNTK_UNMOUNT) != 0) {
1001 mp->mnt_flag &= ~MNT_UPDATEMASK;
1002 mp->mnt_flag |= fsflags & (MNT_RELOAD | MNT_FORCE | MNT_UPDATE |
1003 MNT_SNAPSHOT | MNT_ROOTFS | MNT_UPDATEMASK | MNT_RDONLY);
1004 if ((mp->mnt_flag & MNT_ASYNC) == 0)
1005 mp->mnt_kern_flag &= ~MNTK_ASYNC;
1007 mp->mnt_optnew = *optlist;
1008 vfs_mergeopts(mp->mnt_optnew, mp->mnt_opt);
1011 * Mount the filesystem.
1012 * XXX The final recipients of VFS_MOUNT just overwrite the ndp they
1013 * get. No freeing of cn_pnbuf.
1015 error = VFS_MOUNT(mp);
1018 /* Process the export option. */
1019 if (error == 0 && vfs_getopt(mp->mnt_optnew, "export", &bufp,
1021 /* Assume that there is only 1 ABI for each length. */
1023 case (sizeof(struct oexport_args)):
1024 bzero(&export, sizeof(export));
1026 case (sizeof(export)):
1027 bcopy(bufp, &export, len);
1028 export_error = vfs_export(mp, &export);
1031 export_error = EINVAL;
1038 mp->mnt_flag &= ~(MNT_UPDATE | MNT_RELOAD | MNT_FORCE |
1042 * If we fail, restore old mount flags. MNT_QUOTA is special,
1043 * because it is not part of MNT_UPDATEMASK, but it could have
1044 * changed in the meantime if quotactl(2) was called.
1045 * All in all we want current value of MNT_QUOTA, not the old
1048 mp->mnt_flag = (mp->mnt_flag & MNT_QUOTA) | (flag & ~MNT_QUOTA);
1050 if ((mp->mnt_flag & MNT_ASYNC) != 0 &&
1051 (mp->mnt_kern_flag & MNTK_NOASYNC) == 0)
1052 mp->mnt_kern_flag |= MNTK_ASYNC;
1054 mp->mnt_kern_flag &= ~MNTK_ASYNC;
1060 if (mp->mnt_opt != NULL)
1061 vfs_freeopts(mp->mnt_opt);
1062 mp->mnt_opt = mp->mnt_optnew;
1064 (void)VFS_STATFS(mp, &mp->mnt_stat);
1066 * Prevent external consumers of mount options from reading
1069 mp->mnt_optnew = NULL;
1071 if ((mp->mnt_flag & MNT_RDONLY) == 0)
1072 vfs_allocate_syncvnode(mp);
1074 vfs_deallocate_syncvnode(mp);
1078 vp->v_iflag &= ~VI_MOUNT;
1081 return (error != 0 ? error : export_error);
1085 * vfs_domount(): actually attempt a filesystem mount.
1089 struct thread *td, /* Calling thread. */
1090 const char *fstype, /* Filesystem type. */
1091 char *fspath, /* Mount path. */
1092 uint64_t fsflags, /* Flags common to all filesystems. */
1093 struct vfsoptlist **optlist /* Options local to the filesystem. */
1096 struct vfsconf *vfsp;
1097 struct nameidata nd;
1103 * Be ultra-paranoid about making sure the type and fspath
1104 * variables will fit in our mp buffers, including the
1107 if (strlen(fstype) >= MFSNAMELEN || strlen(fspath) >= MNAMELEN)
1108 return (ENAMETOOLONG);
1110 if (jailed(td->td_ucred) || usermount == 0) {
1111 if ((error = priv_check(td, PRIV_VFS_MOUNT)) != 0)
1116 * Do not allow NFS export or MNT_SUIDDIR by unprivileged users.
1118 if (fsflags & MNT_EXPORTED) {
1119 error = priv_check(td, PRIV_VFS_MOUNT_EXPORTED);
1123 if (fsflags & MNT_SUIDDIR) {
1124 error = priv_check(td, PRIV_VFS_MOUNT_SUIDDIR);
1129 * Silently enforce MNT_NOSUID and MNT_USER for unprivileged users.
1131 if ((fsflags & (MNT_NOSUID | MNT_USER)) != (MNT_NOSUID | MNT_USER)) {
1132 if (priv_check(td, PRIV_VFS_MOUNT_NONUSER) != 0)
1133 fsflags |= MNT_NOSUID | MNT_USER;
1136 /* Load KLDs before we lock the covered vnode to avoid reversals. */
1138 if ((fsflags & MNT_UPDATE) == 0) {
1139 /* Don't try to load KLDs if we're mounting the root. */
1140 if (fsflags & MNT_ROOTFS)
1141 vfsp = vfs_byname(fstype);
1143 vfsp = vfs_byname_kld(fstype, td, &error);
1146 if (jailed(td->td_ucred) && !(vfsp->vfc_flags & VFCF_JAIL))
1151 * Get vnode to be covered or mount point's vnode in case of MNT_UPDATE.
1153 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1,
1154 UIO_SYSSPACE, fspath, td);
1158 NDFREE(&nd, NDF_ONLY_PNBUF);
1160 if ((fsflags & MNT_UPDATE) == 0) {
1161 pathbuf = malloc(MNAMELEN, M_TEMP, M_WAITOK);
1162 strcpy(pathbuf, fspath);
1163 error = vn_path_to_global_path(td, vp, pathbuf, MNAMELEN);
1164 /* debug.disablefullpath == 1 results in ENODEV */
1165 if (error == 0 || error == ENODEV) {
1166 error = vfs_domount_first(td, vfsp, pathbuf, vp,
1169 free(pathbuf, M_TEMP);
1171 error = vfs_domount_update(td, vp, fsflags, optlist);
1177 * Unmount a filesystem.
1179 * Note: unmount takes a path to the vnode mounted on as argument, not
1180 * special file (as before).
1182 #ifndef _SYS_SYSPROTO_H_
1183 struct unmount_args {
1190 sys_unmount(struct thread *td, struct unmount_args *uap)
1192 struct nameidata nd;
1195 int error, id0, id1;
1197 AUDIT_ARG_VALUE(uap->flags);
1198 if (jailed(td->td_ucred) || usermount == 0) {
1199 error = priv_check(td, PRIV_VFS_UNMOUNT);
1204 pathbuf = malloc(MNAMELEN, M_TEMP, M_WAITOK);
1205 error = copyinstr(uap->path, pathbuf, MNAMELEN, NULL);
1207 free(pathbuf, M_TEMP);
1210 if (uap->flags & MNT_BYFSID) {
1211 AUDIT_ARG_TEXT(pathbuf);
1212 /* Decode the filesystem ID. */
1213 if (sscanf(pathbuf, "FSID:%d:%d", &id0, &id1) != 2) {
1214 free(pathbuf, M_TEMP);
1218 mtx_lock(&mountlist_mtx);
1219 TAILQ_FOREACH_REVERSE(mp, &mountlist, mntlist, mnt_list) {
1220 if (mp->mnt_stat.f_fsid.val[0] == id0 &&
1221 mp->mnt_stat.f_fsid.val[1] == id1) {
1226 mtx_unlock(&mountlist_mtx);
1229 * Try to find global path for path argument.
1231 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1,
1232 UIO_SYSSPACE, pathbuf, td);
1233 if (namei(&nd) == 0) {
1234 NDFREE(&nd, NDF_ONLY_PNBUF);
1235 error = vn_path_to_global_path(td, nd.ni_vp, pathbuf,
1237 if (error == 0 || error == ENODEV)
1240 mtx_lock(&mountlist_mtx);
1241 TAILQ_FOREACH_REVERSE(mp, &mountlist, mntlist, mnt_list) {
1242 if (strcmp(mp->mnt_stat.f_mntonname, pathbuf) == 0) {
1247 mtx_unlock(&mountlist_mtx);
1249 free(pathbuf, M_TEMP);
1252 * Previously we returned ENOENT for a nonexistent path and
1253 * EINVAL for a non-mountpoint. We cannot tell these apart
1254 * now, so in the !MNT_BYFSID case return the more likely
1255 * EINVAL for compatibility.
1257 return ((uap->flags & MNT_BYFSID) ? ENOENT : EINVAL);
1261 * Don't allow unmounting the root filesystem.
1263 if (mp->mnt_flag & MNT_ROOTFS) {
1267 error = dounmount(mp, uap->flags, td);
1272 * Return error if any of the vnodes, ignoring the root vnode
1273 * and the syncer vnode, have non-zero usecount.
1275 * This function is purely advisory - it can return false positives
1279 vfs_check_usecounts(struct mount *mp)
1281 struct vnode *vp, *mvp;
1283 MNT_VNODE_FOREACH_ALL(vp, mp, mvp) {
1284 if ((vp->v_vflag & VV_ROOT) == 0 && vp->v_type != VNON &&
1285 vp->v_usecount != 0) {
1287 MNT_VNODE_FOREACH_ALL_ABORT(mp, mvp);
1297 dounmount_cleanup(struct mount *mp, struct vnode *coveredvp, int mntkflags)
1300 mtx_assert(MNT_MTX(mp), MA_OWNED);
1301 mp->mnt_kern_flag &= ~mntkflags;
1302 if ((mp->mnt_kern_flag & MNTK_MWAIT) != 0) {
1303 mp->mnt_kern_flag &= ~MNTK_MWAIT;
1307 if (coveredvp != NULL) {
1308 VOP_UNLOCK(coveredvp, 0);
1311 vn_finished_write(mp);
1315 * Do the actual filesystem unmount.
1318 dounmount(struct mount *mp, int flags, struct thread *td)
1320 struct vnode *coveredvp, *fsrootvp;
1322 uint64_t async_flag;
1325 if ((coveredvp = mp->mnt_vnodecovered) != NULL) {
1326 mnt_gen_r = mp->mnt_gen;
1329 vn_lock(coveredvp, LK_EXCLUSIVE | LK_INTERLOCK | LK_RETRY);
1331 * Check for mp being unmounted while waiting for the
1332 * covered vnode lock.
1334 if (coveredvp->v_mountedhere != mp ||
1335 coveredvp->v_mountedhere->mnt_gen != mnt_gen_r) {
1336 VOP_UNLOCK(coveredvp, 0);
1344 * Only privileged root, or (if MNT_USER is set) the user that did the
1345 * original mount is permitted to unmount this filesystem.
1347 error = vfs_suser(mp, td);
1349 if (coveredvp != NULL) {
1350 VOP_UNLOCK(coveredvp, 0);
1357 vn_start_write(NULL, &mp, V_WAIT | V_MNTREF);
1359 if ((mp->mnt_kern_flag & MNTK_UNMOUNT) != 0 ||
1360 (mp->mnt_flag & MNT_UPDATE) != 0 ||
1361 !TAILQ_EMPTY(&mp->mnt_uppers)) {
1362 dounmount_cleanup(mp, coveredvp, 0);
1365 mp->mnt_kern_flag |= MNTK_UNMOUNT | MNTK_NOINSMNTQ;
1366 if (flags & MNT_NONBUSY) {
1368 error = vfs_check_usecounts(mp);
1371 dounmount_cleanup(mp, coveredvp, MNTK_UNMOUNT |
1376 /* Allow filesystems to detect that a forced unmount is in progress. */
1377 if (flags & MNT_FORCE) {
1378 mp->mnt_kern_flag |= MNTK_UNMOUNTF;
1381 * Must be done after setting MNTK_UNMOUNTF and before
1382 * waiting for mnt_lockref to become 0.
1388 if (mp->mnt_lockref) {
1389 mp->mnt_kern_flag |= MNTK_DRAINING;
1390 error = msleep(&mp->mnt_lockref, MNT_MTX(mp), PVFS,
1394 KASSERT(mp->mnt_lockref == 0,
1395 ("%s: invalid lock refcount in the drain path @ %s:%d",
1396 __func__, __FILE__, __LINE__));
1398 ("%s: invalid return value for msleep in the drain path @ %s:%d",
1399 __func__, __FILE__, __LINE__));
1401 if (mp->mnt_flag & MNT_EXPUBLIC)
1402 vfs_setpublicfs(NULL, NULL, NULL);
1405 * From now, we can claim that the use reference on the
1406 * coveredvp is ours, and the ref can be released only by
1407 * successfull unmount by us, or left for later unmount
1408 * attempt. The previously acquired hold reference is no
1409 * longer needed to protect the vnode from reuse.
1411 if (coveredvp != NULL)
1414 vfs_msync(mp, MNT_WAIT);
1416 async_flag = mp->mnt_flag & MNT_ASYNC;
1417 mp->mnt_flag &= ~MNT_ASYNC;
1418 mp->mnt_kern_flag &= ~MNTK_ASYNC;
1420 cache_purgevfs(mp, false); /* remove cache entries for this file sys */
1421 vfs_deallocate_syncvnode(mp);
1423 * For forced unmounts, move process cdir/rdir refs on the fs root
1424 * vnode to the covered vnode. For non-forced unmounts we want
1425 * such references to cause an EBUSY error.
1427 if ((flags & MNT_FORCE) &&
1428 VFS_ROOT(mp, LK_EXCLUSIVE, &fsrootvp) == 0) {
1429 if (mp->mnt_vnodecovered != NULL &&
1430 (mp->mnt_flag & MNT_IGNORE) == 0)
1431 mountcheckdirs(fsrootvp, mp->mnt_vnodecovered);
1432 if (fsrootvp == rootvnode) {
1438 if ((mp->mnt_flag & MNT_RDONLY) != 0 || (flags & MNT_FORCE) != 0 ||
1439 (error = VFS_SYNC(mp, MNT_WAIT)) == 0)
1440 error = VFS_UNMOUNT(mp, flags);
1441 vn_finished_write(mp);
1443 * If we failed to flush the dirty blocks for this mount point,
1444 * undo all the cdir/rdir and rootvnode changes we made above.
1445 * Unless we failed to do so because the device is reporting that
1446 * it doesn't exist anymore.
1448 if (error && error != ENXIO) {
1449 if ((flags & MNT_FORCE) &&
1450 VFS_ROOT(mp, LK_EXCLUSIVE, &fsrootvp) == 0) {
1451 if (mp->mnt_vnodecovered != NULL &&
1452 (mp->mnt_flag & MNT_IGNORE) == 0)
1453 mountcheckdirs(mp->mnt_vnodecovered, fsrootvp);
1454 if (rootvnode == NULL) {
1455 rootvnode = fsrootvp;
1461 mp->mnt_kern_flag &= ~MNTK_NOINSMNTQ;
1462 if ((mp->mnt_flag & MNT_RDONLY) == 0) {
1464 vfs_allocate_syncvnode(mp);
1467 mp->mnt_kern_flag &= ~(MNTK_UNMOUNT | MNTK_UNMOUNTF);
1468 mp->mnt_flag |= async_flag;
1469 if ((mp->mnt_flag & MNT_ASYNC) != 0 &&
1470 (mp->mnt_kern_flag & MNTK_NOASYNC) == 0)
1471 mp->mnt_kern_flag |= MNTK_ASYNC;
1472 if (mp->mnt_kern_flag & MNTK_MWAIT) {
1473 mp->mnt_kern_flag &= ~MNTK_MWAIT;
1478 VOP_UNLOCK(coveredvp, 0);
1481 mtx_lock(&mountlist_mtx);
1482 TAILQ_REMOVE(&mountlist, mp, mnt_list);
1483 mtx_unlock(&mountlist_mtx);
1484 EVENTHANDLER_INVOKE(vfs_unmounted, mp, td);
1485 if (coveredvp != NULL) {
1486 coveredvp->v_mountedhere = NULL;
1487 VOP_UNLOCK(coveredvp, 0);
1489 vfs_event_signal(NULL, VQ_UNMOUNT, 0);
1490 if (mp == rootdevmp)
1492 vfs_mount_destroy(mp);
1497 * Report errors during filesystem mounting.
1500 vfs_mount_error(struct mount *mp, const char *fmt, ...)
1502 struct vfsoptlist *moptlist = mp->mnt_optnew;
1507 error = vfs_getopt(moptlist, "errmsg", (void **)&errmsg, &len);
1508 if (error || errmsg == NULL || len <= 0)
1512 vsnprintf(errmsg, (size_t)len, fmt, ap);
1517 vfs_opterror(struct vfsoptlist *opts, const char *fmt, ...)
1523 error = vfs_getopt(opts, "errmsg", (void **)&errmsg, &len);
1524 if (error || errmsg == NULL || len <= 0)
1528 vsnprintf(errmsg, (size_t)len, fmt, ap);
1533 * ---------------------------------------------------------------------
1534 * Functions for querying mount options/arguments from filesystems.
1538 * Check that no unknown options are given
1541 vfs_filteropt(struct vfsoptlist *opts, const char **legal)
1545 const char **t, *p, *q;
1548 TAILQ_FOREACH(opt, opts, link) {
1551 if (p[0] == 'n' && p[1] == 'o')
1553 for(t = global_opts; *t != NULL; t++) {
1554 if (strcmp(*t, p) == 0)
1557 if (strcmp(*t, q) == 0)
1563 for(t = legal; *t != NULL; t++) {
1564 if (strcmp(*t, p) == 0)
1567 if (strcmp(*t, q) == 0)
1573 snprintf(errmsg, sizeof(errmsg),
1574 "mount option <%s> is unknown", p);
1578 TAILQ_FOREACH(opt, opts, link) {
1579 if (strcmp(opt->name, "errmsg") == 0) {
1580 strncpy((char *)opt->value, errmsg, opt->len);
1585 printf("%s\n", errmsg);
1591 * Get a mount option by its name.
1593 * Return 0 if the option was found, ENOENT otherwise.
1594 * If len is non-NULL it will be filled with the length
1595 * of the option. If buf is non-NULL, it will be filled
1596 * with the address of the option.
1599 vfs_getopt(struct vfsoptlist *opts, const char *name, void **buf, int *len)
1603 KASSERT(opts != NULL, ("vfs_getopt: caller passed 'opts' as NULL"));
1605 TAILQ_FOREACH(opt, opts, link) {
1606 if (strcmp(name, opt->name) == 0) {
1619 vfs_getopt_pos(struct vfsoptlist *opts, const char *name)
1626 TAILQ_FOREACH(opt, opts, link) {
1627 if (strcmp(name, opt->name) == 0) {
1636 vfs_getopt_size(struct vfsoptlist *opts, const char *name, off_t *value)
1638 char *opt_value, *vtp;
1642 error = vfs_getopt(opts, name, (void **)&opt_value, &opt_len);
1645 if (opt_len == 0 || opt_value == NULL)
1647 if (opt_value[0] == '\0' || opt_value[opt_len - 1] != '\0')
1649 iv = strtoq(opt_value, &vtp, 0);
1650 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;