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 MALLOC_DEFINE(M_MOUNT, "mount", "vfs mount structure");
82 static uma_zone_t mount_zone;
84 /* List of mounted filesystems. */
85 struct mntlist mountlist = TAILQ_HEAD_INITIALIZER(mountlist);
87 /* For any iteration/modification of mountlist */
88 struct mtx mountlist_mtx;
89 MTX_SYSINIT(mountlist, &mountlist_mtx, "mountlist", MTX_DEF);
92 * Global opts, taken by all filesystems
94 static const char *global_opts[] = {
106 mount_init(void *mem, int size, int flags)
110 mp = (struct mount *)mem;
111 mtx_init(&mp->mnt_mtx, "struct mount mtx", NULL, MTX_DEF);
112 lockinit(&mp->mnt_explock, PVFS, "explock", 0, 0);
117 mount_fini(void *mem, int size)
121 mp = (struct mount *)mem;
122 lockdestroy(&mp->mnt_explock);
123 mtx_destroy(&mp->mnt_mtx);
127 vfs_mount_init(void *dummy __unused)
130 mount_zone = uma_zcreate("Mountpoints", sizeof(struct mount), NULL,
131 NULL, mount_init, mount_fini, UMA_ALIGN_PTR, UMA_ZONE_NOFREE);
133 SYSINIT(vfs_mount, SI_SUB_VFS, SI_ORDER_ANY, vfs_mount_init, NULL);
136 * ---------------------------------------------------------------------
137 * Functions for building and sanitizing the mount options
140 /* Remove one mount option. */
142 vfs_freeopt(struct vfsoptlist *opts, struct vfsopt *opt)
145 TAILQ_REMOVE(opts, opt, link);
146 free(opt->name, M_MOUNT);
147 if (opt->value != NULL)
148 free(opt->value, M_MOUNT);
152 /* Release all resources related to the mount options. */
154 vfs_freeopts(struct vfsoptlist *opts)
158 while (!TAILQ_EMPTY(opts)) {
159 opt = TAILQ_FIRST(opts);
160 vfs_freeopt(opts, opt);
166 vfs_deleteopt(struct vfsoptlist *opts, const char *name)
168 struct vfsopt *opt, *temp;
172 TAILQ_FOREACH_SAFE(opt, opts, link, temp) {
173 if (strcmp(opt->name, name) == 0)
174 vfs_freeopt(opts, opt);
179 vfs_isopt_ro(const char *opt)
182 if (strcmp(opt, "ro") == 0 || strcmp(opt, "rdonly") == 0 ||
183 strcmp(opt, "norw") == 0)
189 vfs_isopt_rw(const char *opt)
192 if (strcmp(opt, "rw") == 0 || strcmp(opt, "noro") == 0)
198 * Check if options are equal (with or without the "no" prefix).
201 vfs_equalopts(const char *opt1, const char *opt2)
205 /* "opt" vs. "opt" or "noopt" vs. "noopt" */
206 if (strcmp(opt1, opt2) == 0)
208 /* "noopt" vs. "opt" */
209 if (strncmp(opt1, "no", 2) == 0 && strcmp(opt1 + 2, opt2) == 0)
211 /* "opt" vs. "noopt" */
212 if (strncmp(opt2, "no", 2) == 0 && strcmp(opt1, opt2 + 2) == 0)
214 while ((p = strchr(opt1, '.')) != NULL &&
215 !strncmp(opt1, opt2, ++p - opt1)) {
218 /* "foo.noopt" vs. "foo.opt" */
219 if (strncmp(opt1, "no", 2) == 0 && strcmp(opt1 + 2, opt2) == 0)
221 /* "foo.opt" vs. "foo.noopt" */
222 if (strncmp(opt2, "no", 2) == 0 && strcmp(opt1, opt2 + 2) == 0)
225 /* "ro" / "rdonly" / "norw" / "rw" / "noro" */
226 if ((vfs_isopt_ro(opt1) || vfs_isopt_rw(opt1)) &&
227 (vfs_isopt_ro(opt2) || vfs_isopt_rw(opt2)))
233 * If a mount option is specified several times,
234 * (with or without the "no" prefix) only keep
235 * the last occurrence of it.
238 vfs_sanitizeopts(struct vfsoptlist *opts)
240 struct vfsopt *opt, *opt2, *tmp;
242 TAILQ_FOREACH_REVERSE(opt, opts, vfsoptlist, link) {
243 opt2 = TAILQ_PREV(opt, vfsoptlist, link);
244 while (opt2 != NULL) {
245 if (vfs_equalopts(opt->name, opt2->name)) {
246 tmp = TAILQ_PREV(opt2, vfsoptlist, link);
247 vfs_freeopt(opts, opt2);
250 opt2 = TAILQ_PREV(opt2, vfsoptlist, link);
257 * Build a linked list of mount options from a struct uio.
260 vfs_buildopts(struct uio *auio, struct vfsoptlist **options)
262 struct vfsoptlist *opts;
264 size_t memused, namelen, optlen;
265 unsigned int i, iovcnt;
268 opts = malloc(sizeof(struct vfsoptlist), M_MOUNT, M_WAITOK);
271 iovcnt = auio->uio_iovcnt;
272 for (i = 0; i < iovcnt; i += 2) {
273 namelen = auio->uio_iov[i].iov_len;
274 optlen = auio->uio_iov[i + 1].iov_len;
275 memused += sizeof(struct vfsopt) + optlen + namelen;
277 * Avoid consuming too much memory, and attempts to overflow
280 if (memused > VFS_MOUNTARG_SIZE_MAX ||
281 optlen > VFS_MOUNTARG_SIZE_MAX ||
282 namelen > VFS_MOUNTARG_SIZE_MAX) {
287 opt = malloc(sizeof(struct vfsopt), M_MOUNT, M_WAITOK);
288 opt->name = malloc(namelen, M_MOUNT, M_WAITOK);
295 * Do this early, so jumps to "bad" will free the current
298 TAILQ_INSERT_TAIL(opts, opt, link);
300 if (auio->uio_segflg == UIO_SYSSPACE) {
301 bcopy(auio->uio_iov[i].iov_base, opt->name, namelen);
303 error = copyin(auio->uio_iov[i].iov_base, opt->name,
308 /* Ensure names are null-terminated strings. */
309 if (namelen == 0 || opt->name[namelen - 1] != '\0') {
315 opt->value = malloc(optlen, M_MOUNT, M_WAITOK);
316 if (auio->uio_segflg == UIO_SYSSPACE) {
317 bcopy(auio->uio_iov[i + 1].iov_base, opt->value,
320 error = copyin(auio->uio_iov[i + 1].iov_base,
327 vfs_sanitizeopts(opts);
336 * Merge the old mount options with the new ones passed
337 * in the MNT_UPDATE case.
339 * XXX: This function will keep a "nofoo" option in the new
340 * options. E.g, if the option's canonical name is "foo",
341 * "nofoo" ends up in the mount point's active options.
344 vfs_mergeopts(struct vfsoptlist *toopts, struct vfsoptlist *oldopts)
346 struct vfsopt *opt, *new;
348 TAILQ_FOREACH(opt, oldopts, link) {
349 new = malloc(sizeof(struct vfsopt), M_MOUNT, M_WAITOK);
350 new->name = strdup(opt->name, M_MOUNT);
352 new->value = malloc(opt->len, M_MOUNT, M_WAITOK);
353 bcopy(opt->value, new->value, opt->len);
357 new->seen = opt->seen;
358 TAILQ_INSERT_HEAD(toopts, new, link);
360 vfs_sanitizeopts(toopts);
364 * Mount a filesystem.
369 struct nmount_args /* {
381 * Mount flags are now 64-bits. On 32-bit archtectures only
382 * 32-bits are passed in, but from here on everything handles
383 * 64-bit flags correctly.
387 AUDIT_ARG_FFLAGS(flags);
388 CTR4(KTR_VFS, "%s: iovp %p with iovcnt %d and flags %d", __func__,
389 uap->iovp, uap->iovcnt, flags);
392 * Filter out MNT_ROOTFS. We do not want clients of nmount() in
393 * userspace to set this flag, but we must filter it out if we want
394 * MNT_UPDATE on the root file system to work.
395 * MNT_ROOTFS should only be set by the kernel when mounting its
398 flags &= ~MNT_ROOTFS;
400 iovcnt = uap->iovcnt;
402 * Check that we have an even number of iovec's
403 * and that we have at least two options.
405 if ((iovcnt & 1) || (iovcnt < 4)) {
406 CTR2(KTR_VFS, "%s: failed for invalid iovcnt %d", __func__,
411 error = copyinuio(uap->iovp, iovcnt, &auio);
413 CTR2(KTR_VFS, "%s: failed for invalid uio op with %d errno",
417 error = vfs_donmount(td, flags, auio);
424 * ---------------------------------------------------------------------
425 * Various utility functions
429 vfs_ref(struct mount *mp)
432 CTR2(KTR_VFS, "%s: mp %p", __func__, mp);
439 vfs_rel(struct mount *mp)
442 CTR2(KTR_VFS, "%s: mp %p", __func__, mp);
449 * Allocate and initialize the mount point struct.
452 vfs_mount_alloc(struct vnode *vp, struct vfsconf *vfsp, const char *fspath,
457 mp = uma_zalloc(mount_zone, M_WAITOK);
458 bzero(&mp->mnt_startzero,
459 __rangeof(struct mount, mnt_startzero, mnt_endzero));
460 TAILQ_INIT(&mp->mnt_nvnodelist);
461 mp->mnt_nvnodelistsize = 0;
462 TAILQ_INIT(&mp->mnt_activevnodelist);
463 mp->mnt_activevnodelistsize = 0;
465 (void) vfs_busy(mp, MBF_NOWAIT);
466 atomic_add_acq_int(&vfsp->vfc_refcount, 1);
467 mp->mnt_op = vfsp->vfc_vfsops;
469 mp->mnt_stat.f_type = vfsp->vfc_typenum;
471 strlcpy(mp->mnt_stat.f_fstypename, vfsp->vfc_name, MFSNAMELEN);
472 mp->mnt_vnodecovered = vp;
473 mp->mnt_cred = crdup(cred);
474 mp->mnt_stat.f_owner = cred->cr_uid;
475 strlcpy(mp->mnt_stat.f_mntonname, fspath, MNAMELEN);
476 mp->mnt_iosize_max = DFLTPHYS;
479 mac_mount_create(cred, mp);
481 arc4rand(&mp->mnt_hashseed, sizeof mp->mnt_hashseed, 0);
482 TAILQ_INIT(&mp->mnt_uppers);
487 * Destroy the mount struct previously allocated by vfs_mount_alloc().
490 vfs_mount_destroy(struct mount *mp)
494 mp->mnt_kern_flag |= MNTK_REFEXPIRE;
495 if (mp->mnt_kern_flag & MNTK_MWAIT) {
496 mp->mnt_kern_flag &= ~MNTK_MWAIT;
500 msleep(mp, MNT_MTX(mp), PVFS, "mntref", 0);
501 KASSERT(mp->mnt_ref == 0,
502 ("%s: invalid refcount in the drain path @ %s:%d", __func__,
503 __FILE__, __LINE__));
504 if (mp->mnt_writeopcount != 0)
505 panic("vfs_mount_destroy: nonzero writeopcount");
506 if (mp->mnt_secondary_writes != 0)
507 panic("vfs_mount_destroy: nonzero secondary_writes");
508 atomic_subtract_rel_int(&mp->mnt_vfc->vfc_refcount, 1);
509 if (!TAILQ_EMPTY(&mp->mnt_nvnodelist)) {
512 TAILQ_FOREACH(vp, &mp->mnt_nvnodelist, v_nmntvnodes)
514 panic("unmount: dangling vnode");
516 KASSERT(TAILQ_EMPTY(&mp->mnt_uppers), ("mnt_uppers"));
517 if (mp->mnt_nvnodelistsize != 0)
518 panic("vfs_mount_destroy: nonzero nvnodelistsize");
519 if (mp->mnt_activevnodelistsize != 0)
520 panic("vfs_mount_destroy: nonzero activevnodelistsize");
521 if (mp->mnt_lockref != 0)
522 panic("vfs_mount_destroy: nonzero lock refcount");
524 if (mp->mnt_vnodecovered != NULL)
525 vrele(mp->mnt_vnodecovered);
527 mac_mount_destroy(mp);
529 if (mp->mnt_opt != NULL)
530 vfs_freeopts(mp->mnt_opt);
531 crfree(mp->mnt_cred);
532 uma_zfree(mount_zone, mp);
536 vfs_donmount(struct thread *td, uint64_t fsflags, struct uio *fsoptions)
538 struct vfsoptlist *optlist;
539 struct vfsopt *opt, *tmp_opt;
540 char *fstype, *fspath, *errmsg;
541 int error, fstypelen, fspathlen, errmsg_len, errmsg_pos;
543 errmsg = fspath = NULL;
544 errmsg_len = fspathlen = 0;
547 error = vfs_buildopts(fsoptions, &optlist);
551 if (vfs_getopt(optlist, "errmsg", (void **)&errmsg, &errmsg_len) == 0)
552 errmsg_pos = vfs_getopt_pos(optlist, "errmsg");
555 * We need these two options before the others,
556 * and they are mandatory for any filesystem.
557 * Ensure they are NUL terminated as well.
560 error = vfs_getopt(optlist, "fstype", (void **)&fstype, &fstypelen);
561 if (error || fstype[fstypelen - 1] != '\0') {
564 strncpy(errmsg, "Invalid fstype", errmsg_len);
568 error = vfs_getopt(optlist, "fspath", (void **)&fspath, &fspathlen);
569 if (error || fspath[fspathlen - 1] != '\0') {
572 strncpy(errmsg, "Invalid fspath", errmsg_len);
577 * We need to see if we have the "update" option
578 * before we call vfs_domount(), since vfs_domount() has special
579 * logic based on MNT_UPDATE. This is very important
580 * when we want to update the root filesystem.
582 TAILQ_FOREACH_SAFE(opt, optlist, link, tmp_opt) {
583 if (strcmp(opt->name, "update") == 0) {
584 fsflags |= MNT_UPDATE;
585 vfs_freeopt(optlist, opt);
587 else if (strcmp(opt->name, "async") == 0)
588 fsflags |= MNT_ASYNC;
589 else if (strcmp(opt->name, "force") == 0) {
590 fsflags |= MNT_FORCE;
591 vfs_freeopt(optlist, opt);
593 else if (strcmp(opt->name, "reload") == 0) {
594 fsflags |= MNT_RELOAD;
595 vfs_freeopt(optlist, opt);
597 else if (strcmp(opt->name, "multilabel") == 0)
598 fsflags |= MNT_MULTILABEL;
599 else if (strcmp(opt->name, "noasync") == 0)
600 fsflags &= ~MNT_ASYNC;
601 else if (strcmp(opt->name, "noatime") == 0)
602 fsflags |= MNT_NOATIME;
603 else if (strcmp(opt->name, "atime") == 0) {
604 free(opt->name, M_MOUNT);
605 opt->name = strdup("nonoatime", M_MOUNT);
607 else if (strcmp(opt->name, "noclusterr") == 0)
608 fsflags |= MNT_NOCLUSTERR;
609 else if (strcmp(opt->name, "clusterr") == 0) {
610 free(opt->name, M_MOUNT);
611 opt->name = strdup("nonoclusterr", M_MOUNT);
613 else if (strcmp(opt->name, "noclusterw") == 0)
614 fsflags |= MNT_NOCLUSTERW;
615 else if (strcmp(opt->name, "clusterw") == 0) {
616 free(opt->name, M_MOUNT);
617 opt->name = strdup("nonoclusterw", M_MOUNT);
619 else if (strcmp(opt->name, "noexec") == 0)
620 fsflags |= MNT_NOEXEC;
621 else if (strcmp(opt->name, "exec") == 0) {
622 free(opt->name, M_MOUNT);
623 opt->name = strdup("nonoexec", M_MOUNT);
625 else if (strcmp(opt->name, "nosuid") == 0)
626 fsflags |= MNT_NOSUID;
627 else if (strcmp(opt->name, "suid") == 0) {
628 free(opt->name, M_MOUNT);
629 opt->name = strdup("nonosuid", M_MOUNT);
631 else if (strcmp(opt->name, "nosymfollow") == 0)
632 fsflags |= MNT_NOSYMFOLLOW;
633 else if (strcmp(opt->name, "symfollow") == 0) {
634 free(opt->name, M_MOUNT);
635 opt->name = strdup("nonosymfollow", M_MOUNT);
637 else if (strcmp(opt->name, "noro") == 0)
638 fsflags &= ~MNT_RDONLY;
639 else if (strcmp(opt->name, "rw") == 0)
640 fsflags &= ~MNT_RDONLY;
641 else if (strcmp(opt->name, "ro") == 0)
642 fsflags |= MNT_RDONLY;
643 else if (strcmp(opt->name, "rdonly") == 0) {
644 free(opt->name, M_MOUNT);
645 opt->name = strdup("ro", M_MOUNT);
646 fsflags |= MNT_RDONLY;
648 else if (strcmp(opt->name, "suiddir") == 0)
649 fsflags |= MNT_SUIDDIR;
650 else if (strcmp(opt->name, "sync") == 0)
651 fsflags |= MNT_SYNCHRONOUS;
652 else if (strcmp(opt->name, "union") == 0)
653 fsflags |= MNT_UNION;
654 else if (strcmp(opt->name, "automounted") == 0) {
655 fsflags |= MNT_AUTOMOUNTED;
656 vfs_freeopt(optlist, opt);
661 * Be ultra-paranoid about making sure the type and fspath
662 * variables will fit in our mp buffers, including the
665 if (fstypelen > MFSNAMELEN || fspathlen > MNAMELEN) {
666 error = ENAMETOOLONG;
670 error = vfs_domount(td, fstype, fspath, fsflags, &optlist);
672 /* copyout the errmsg */
673 if (errmsg_pos != -1 && ((2 * errmsg_pos + 1) < fsoptions->uio_iovcnt)
674 && errmsg_len > 0 && errmsg != NULL) {
675 if (fsoptions->uio_segflg == UIO_SYSSPACE) {
677 fsoptions->uio_iov[2 * errmsg_pos + 1].iov_base,
678 fsoptions->uio_iov[2 * errmsg_pos + 1].iov_len);
681 fsoptions->uio_iov[2 * errmsg_pos + 1].iov_base,
682 fsoptions->uio_iov[2 * errmsg_pos + 1].iov_len);
687 vfs_freeopts(optlist);
694 #ifndef _SYS_SYSPROTO_H_
706 struct mount_args /* {
714 struct vfsconf *vfsp = NULL;
715 struct mntarg *ma = NULL;
720 * Mount flags are now 64-bits. On 32-bit architectures only
721 * 32-bits are passed in, but from here on everything handles
722 * 64-bit flags correctly.
726 AUDIT_ARG_FFLAGS(flags);
729 * Filter out MNT_ROOTFS. We do not want clients of mount() in
730 * userspace to set this flag, but we must filter it out if we want
731 * MNT_UPDATE on the root file system to work.
732 * MNT_ROOTFS should only be set by the kernel when mounting its
735 flags &= ~MNT_ROOTFS;
737 fstype = malloc(MFSNAMELEN, M_TEMP, M_WAITOK);
738 error = copyinstr(uap->type, fstype, MFSNAMELEN, NULL);
740 free(fstype, M_TEMP);
744 AUDIT_ARG_TEXT(fstype);
745 vfsp = vfs_byname_kld(fstype, td, &error);
746 free(fstype, M_TEMP);
749 if (vfsp->vfc_vfsops->vfs_cmount == NULL)
752 ma = mount_argsu(ma, "fstype", uap->type, MFSNAMELEN);
753 ma = mount_argsu(ma, "fspath", uap->path, MNAMELEN);
754 ma = mount_argb(ma, flags & MNT_RDONLY, "noro");
755 ma = mount_argb(ma, !(flags & MNT_NOSUID), "nosuid");
756 ma = mount_argb(ma, !(flags & MNT_NOEXEC), "noexec");
758 error = vfsp->vfc_vfsops->vfs_cmount(ma, uap->data, flags);
763 * vfs_domount_first(): first file system mount (not update)
767 struct thread *td, /* Calling thread. */
768 struct vfsconf *vfsp, /* File system type. */
769 char *fspath, /* Mount path. */
770 struct vnode *vp, /* Vnode to be covered. */
771 uint64_t fsflags, /* Flags common to all filesystems. */
772 struct vfsoptlist **optlist /* Options local to the filesystem. */
780 ASSERT_VOP_ELOCKED(vp, __func__);
781 KASSERT((fsflags & MNT_UPDATE) == 0, ("MNT_UPDATE shouldn't be here"));
784 * If the user is not root, ensure that they own the directory
785 * onto which we are attempting to mount.
787 error = VOP_GETATTR(vp, &va, td->td_ucred);
788 if (error == 0 && va.va_uid != td->td_ucred->cr_uid)
789 error = priv_check_cred(td->td_ucred, PRIV_VFS_ADMIN, 0);
791 error = vinvalbuf(vp, V_SAVE, 0, 0);
792 if (error == 0 && vp->v_type != VDIR)
796 if ((vp->v_iflag & VI_MOUNT) == 0 && vp->v_mountedhere == NULL)
797 vp->v_iflag |= VI_MOUNT;
808 /* Allocate and initialize the filesystem. */
809 mp = vfs_mount_alloc(vp, vfsp, fspath, td->td_ucred);
810 /* XXXMAC: pass to vfs_mount_alloc? */
811 mp->mnt_optnew = *optlist;
812 /* Set the mount level flags. */
813 mp->mnt_flag = (fsflags & (MNT_UPDATEMASK | MNT_ROOTFS | MNT_RDONLY));
816 * Mount the filesystem.
817 * XXX The final recipients of VFS_MOUNT just overwrite the ndp they
818 * get. No freeing of cn_pnbuf.
820 error = VFS_MOUNT(mp);
823 mp->mnt_vnodecovered = NULL;
824 vfs_mount_destroy(mp);
826 vp->v_iflag &= ~VI_MOUNT;
832 if (mp->mnt_opt != NULL)
833 vfs_freeopts(mp->mnt_opt);
834 mp->mnt_opt = mp->mnt_optnew;
836 (void)VFS_STATFS(mp, &mp->mnt_stat);
839 * Prevent external consumers of mount options from reading mnt_optnew.
841 mp->mnt_optnew = NULL;
844 if ((mp->mnt_flag & MNT_ASYNC) != 0 &&
845 (mp->mnt_kern_flag & MNTK_NOASYNC) == 0)
846 mp->mnt_kern_flag |= MNTK_ASYNC;
848 mp->mnt_kern_flag &= ~MNTK_ASYNC;
851 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
854 vp->v_iflag &= ~VI_MOUNT;
856 vp->v_mountedhere = mp;
857 /* Place the new filesystem at the end of the mount list. */
858 mtx_lock(&mountlist_mtx);
859 TAILQ_INSERT_TAIL(&mountlist, mp, mnt_list);
860 mtx_unlock(&mountlist_mtx);
861 vfs_event_signal(NULL, VQ_MOUNT, 0);
862 if (VFS_ROOT(mp, LK_EXCLUSIVE, &newdp))
863 panic("mount: lost mount");
865 EVENTHANDLER_INVOKE(vfs_mounted, mp, newdp, td);
866 VOP_UNLOCK(newdp, 0);
867 mountcheckdirs(vp, newdp);
869 if ((mp->mnt_flag & MNT_RDONLY) == 0)
870 vfs_allocate_syncvnode(mp);
876 * vfs_domount_update(): update of mounted file system
880 struct thread *td, /* Calling thread. */
881 struct vnode *vp, /* Mount point vnode. */
882 uint64_t fsflags, /* Flags common to all filesystems. */
883 struct vfsoptlist **optlist /* Options local to the filesystem. */
886 struct oexport_args oexport;
887 struct export_args export;
889 int error, export_error;
892 ASSERT_VOP_ELOCKED(vp, __func__);
893 KASSERT((fsflags & MNT_UPDATE) != 0, ("MNT_UPDATE should be here"));
896 if ((vp->v_vflag & VV_ROOT) == 0) {
897 if (vfs_copyopt(*optlist, "export", &export, sizeof(export))
907 * We only allow the filesystem to be reloaded if it
908 * is currently mounted read-only.
911 if ((fsflags & MNT_RELOAD) != 0 && (flag & MNT_RDONLY) == 0) {
913 return (EOPNOTSUPP); /* Needs translation */
916 * Only privileged root, or (if MNT_USER is set) the user that
917 * did the original mount is permitted to update it.
919 error = vfs_suser(mp, td);
924 if (vfs_busy(mp, MBF_NOWAIT)) {
929 if ((vp->v_iflag & VI_MOUNT) != 0 || vp->v_mountedhere != NULL) {
935 vp->v_iflag |= VI_MOUNT;
940 if ((mp->mnt_kern_flag & MNTK_UNMOUNT) != 0) {
945 mp->mnt_flag &= ~MNT_UPDATEMASK;
946 mp->mnt_flag |= fsflags & (MNT_RELOAD | MNT_FORCE | MNT_UPDATE |
947 MNT_SNAPSHOT | MNT_ROOTFS | MNT_UPDATEMASK | MNT_RDONLY);
948 if ((mp->mnt_flag & MNT_ASYNC) == 0)
949 mp->mnt_kern_flag &= ~MNTK_ASYNC;
951 mp->mnt_optnew = *optlist;
952 vfs_mergeopts(mp->mnt_optnew, mp->mnt_opt);
955 * Mount the filesystem.
956 * XXX The final recipients of VFS_MOUNT just overwrite the ndp they
957 * get. No freeing of cn_pnbuf.
959 error = VFS_MOUNT(mp);
963 /* Process the export option. */
964 if (vfs_copyopt(mp->mnt_optnew, "export", &export,
965 sizeof(export)) == 0) {
966 export_error = vfs_export(mp, &export);
967 } else if (vfs_copyopt(mp->mnt_optnew, "export", &oexport,
968 sizeof(oexport)) == 0) {
969 export.ex_flags = oexport.ex_flags;
970 export.ex_root = oexport.ex_root;
971 export.ex_anon = oexport.ex_anon;
972 export.ex_addr = oexport.ex_addr;
973 export.ex_addrlen = oexport.ex_addrlen;
974 export.ex_mask = oexport.ex_mask;
975 export.ex_masklen = oexport.ex_masklen;
976 export.ex_indexfile = oexport.ex_indexfile;
977 export.ex_numsecflavors = 0;
978 export_error = vfs_export(mp, &export);
984 mp->mnt_flag &= ~(MNT_UPDATE | MNT_RELOAD | MNT_FORCE |
988 * If we fail, restore old mount flags. MNT_QUOTA is special,
989 * because it is not part of MNT_UPDATEMASK, but it could have
990 * changed in the meantime if quotactl(2) was called.
991 * All in all we want current value of MNT_QUOTA, not the old
994 mp->mnt_flag = (mp->mnt_flag & MNT_QUOTA) | (flag & ~MNT_QUOTA);
996 if ((mp->mnt_flag & MNT_ASYNC) != 0 &&
997 (mp->mnt_kern_flag & MNTK_NOASYNC) == 0)
998 mp->mnt_kern_flag |= MNTK_ASYNC;
1000 mp->mnt_kern_flag &= ~MNTK_ASYNC;
1006 if (mp->mnt_opt != NULL)
1007 vfs_freeopts(mp->mnt_opt);
1008 mp->mnt_opt = mp->mnt_optnew;
1010 (void)VFS_STATFS(mp, &mp->mnt_stat);
1012 * Prevent external consumers of mount options from reading
1015 mp->mnt_optnew = NULL;
1017 if ((mp->mnt_flag & MNT_RDONLY) == 0)
1018 vfs_allocate_syncvnode(mp);
1020 vfs_deallocate_syncvnode(mp);
1024 vp->v_iflag &= ~VI_MOUNT;
1027 return (error != 0 ? error : export_error);
1031 * vfs_domount(): actually attempt a filesystem mount.
1035 struct thread *td, /* Calling thread. */
1036 const char *fstype, /* Filesystem type. */
1037 char *fspath, /* Mount path. */
1038 uint64_t fsflags, /* Flags common to all filesystems. */
1039 struct vfsoptlist **optlist /* Options local to the filesystem. */
1042 struct vfsconf *vfsp;
1043 struct nameidata nd;
1049 * Be ultra-paranoid about making sure the type and fspath
1050 * variables will fit in our mp buffers, including the
1053 if (strlen(fstype) >= MFSNAMELEN || strlen(fspath) >= MNAMELEN)
1054 return (ENAMETOOLONG);
1056 if (jailed(td->td_ucred) || usermount == 0) {
1057 if ((error = priv_check(td, PRIV_VFS_MOUNT)) != 0)
1062 * Do not allow NFS export or MNT_SUIDDIR by unprivileged users.
1064 if (fsflags & MNT_EXPORTED) {
1065 error = priv_check(td, PRIV_VFS_MOUNT_EXPORTED);
1069 if (fsflags & MNT_SUIDDIR) {
1070 error = priv_check(td, PRIV_VFS_MOUNT_SUIDDIR);
1075 * Silently enforce MNT_NOSUID and MNT_USER for unprivileged users.
1077 if ((fsflags & (MNT_NOSUID | MNT_USER)) != (MNT_NOSUID | MNT_USER)) {
1078 if (priv_check(td, PRIV_VFS_MOUNT_NONUSER) != 0)
1079 fsflags |= MNT_NOSUID | MNT_USER;
1082 /* Load KLDs before we lock the covered vnode to avoid reversals. */
1084 if ((fsflags & MNT_UPDATE) == 0) {
1085 /* Don't try to load KLDs if we're mounting the root. */
1086 if (fsflags & MNT_ROOTFS)
1087 vfsp = vfs_byname(fstype);
1089 vfsp = vfs_byname_kld(fstype, td, &error);
1092 if (jailed(td->td_ucred) && !(vfsp->vfc_flags & VFCF_JAIL))
1097 * Get vnode to be covered or mount point's vnode in case of MNT_UPDATE.
1099 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1,
1100 UIO_SYSSPACE, fspath, td);
1104 NDFREE(&nd, NDF_ONLY_PNBUF);
1106 if ((fsflags & MNT_UPDATE) == 0) {
1107 pathbuf = malloc(MNAMELEN, M_TEMP, M_WAITOK);
1108 strcpy(pathbuf, fspath);
1109 error = vn_path_to_global_path(td, vp, pathbuf, MNAMELEN);
1110 /* debug.disablefullpath == 1 results in ENODEV */
1111 if (error == 0 || error == ENODEV) {
1112 error = vfs_domount_first(td, vfsp, pathbuf, vp,
1115 free(pathbuf, M_TEMP);
1117 error = vfs_domount_update(td, vp, fsflags, optlist);
1123 * Unmount a filesystem.
1125 * Note: unmount takes a path to the vnode mounted on as argument, not
1126 * special file (as before).
1128 #ifndef _SYS_SYSPROTO_H_
1129 struct unmount_args {
1136 sys_unmount(struct thread *td, struct unmount_args *uap)
1138 struct nameidata nd;
1141 int error, id0, id1;
1143 AUDIT_ARG_VALUE(uap->flags);
1144 if (jailed(td->td_ucred) || usermount == 0) {
1145 error = priv_check(td, PRIV_VFS_UNMOUNT);
1150 pathbuf = malloc(MNAMELEN, M_TEMP, M_WAITOK);
1151 error = copyinstr(uap->path, pathbuf, MNAMELEN, NULL);
1153 free(pathbuf, M_TEMP);
1156 if (uap->flags & MNT_BYFSID) {
1157 AUDIT_ARG_TEXT(pathbuf);
1158 /* Decode the filesystem ID. */
1159 if (sscanf(pathbuf, "FSID:%d:%d", &id0, &id1) != 2) {
1160 free(pathbuf, M_TEMP);
1164 mtx_lock(&mountlist_mtx);
1165 TAILQ_FOREACH_REVERSE(mp, &mountlist, mntlist, mnt_list) {
1166 if (mp->mnt_stat.f_fsid.val[0] == id0 &&
1167 mp->mnt_stat.f_fsid.val[1] == id1) {
1172 mtx_unlock(&mountlist_mtx);
1175 * Try to find global path for path argument.
1177 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1,
1178 UIO_SYSSPACE, pathbuf, td);
1179 if (namei(&nd) == 0) {
1180 NDFREE(&nd, NDF_ONLY_PNBUF);
1181 error = vn_path_to_global_path(td, nd.ni_vp, pathbuf,
1183 if (error == 0 || error == ENODEV)
1186 mtx_lock(&mountlist_mtx);
1187 TAILQ_FOREACH_REVERSE(mp, &mountlist, mntlist, mnt_list) {
1188 if (strcmp(mp->mnt_stat.f_mntonname, pathbuf) == 0) {
1193 mtx_unlock(&mountlist_mtx);
1195 free(pathbuf, M_TEMP);
1198 * Previously we returned ENOENT for a nonexistent path and
1199 * EINVAL for a non-mountpoint. We cannot tell these apart
1200 * now, so in the !MNT_BYFSID case return the more likely
1201 * EINVAL for compatibility.
1203 return ((uap->flags & MNT_BYFSID) ? ENOENT : EINVAL);
1207 * Don't allow unmounting the root filesystem.
1209 if (mp->mnt_flag & MNT_ROOTFS) {
1213 error = dounmount(mp, uap->flags, td);
1218 * Do the actual filesystem unmount.
1221 dounmount(struct mount *mp, int flags, struct thread *td)
1223 struct vnode *coveredvp, *fsrootvp;
1225 uint64_t async_flag;
1228 if ((coveredvp = mp->mnt_vnodecovered) != NULL) {
1229 mnt_gen_r = mp->mnt_gen;
1232 vn_lock(coveredvp, LK_EXCLUSIVE | LK_INTERLOCK | LK_RETRY);
1234 * Check for mp being unmounted while waiting for the
1235 * covered vnode lock.
1237 if (coveredvp->v_mountedhere != mp ||
1238 coveredvp->v_mountedhere->mnt_gen != mnt_gen_r) {
1239 VOP_UNLOCK(coveredvp, 0);
1247 * Only privileged root, or (if MNT_USER is set) the user that did the
1248 * original mount is permitted to unmount this filesystem.
1250 error = vfs_suser(mp, td);
1252 if (coveredvp != NULL) {
1253 VOP_UNLOCK(coveredvp, 0);
1260 vn_start_write(NULL, &mp, V_WAIT | V_MNTREF);
1262 if ((mp->mnt_kern_flag & MNTK_UNMOUNT) != 0 ||
1263 (mp->mnt_flag & MNT_UPDATE) != 0 ||
1264 !TAILQ_EMPTY(&mp->mnt_uppers)) {
1266 if (coveredvp != NULL) {
1267 VOP_UNLOCK(coveredvp, 0);
1270 vn_finished_write(mp);
1273 mp->mnt_kern_flag |= MNTK_UNMOUNT | MNTK_NOINSMNTQ;
1274 /* Allow filesystems to detect that a forced unmount is in progress. */
1275 if (flags & MNT_FORCE) {
1276 mp->mnt_kern_flag |= MNTK_UNMOUNTF;
1279 * Must be done after setting MNTK_UNMOUNTF and before
1280 * waiting for mnt_lockref to become 0.
1286 if (mp->mnt_lockref) {
1287 mp->mnt_kern_flag |= MNTK_DRAINING;
1288 error = msleep(&mp->mnt_lockref, MNT_MTX(mp), PVFS,
1292 KASSERT(mp->mnt_lockref == 0,
1293 ("%s: invalid lock refcount in the drain path @ %s:%d",
1294 __func__, __FILE__, __LINE__));
1296 ("%s: invalid return value for msleep in the drain path @ %s:%d",
1297 __func__, __FILE__, __LINE__));
1299 if (mp->mnt_flag & MNT_EXPUBLIC)
1300 vfs_setpublicfs(NULL, NULL, NULL);
1303 * From now, we can claim that the use reference on the
1304 * coveredvp is ours, and the ref can be released only by
1305 * successfull unmount by us, or left for later unmount
1306 * attempt. The previously acquired hold reference is no
1307 * longer needed to protect the vnode from reuse.
1309 if (coveredvp != NULL)
1312 vfs_msync(mp, MNT_WAIT);
1314 async_flag = mp->mnt_flag & MNT_ASYNC;
1315 mp->mnt_flag &= ~MNT_ASYNC;
1316 mp->mnt_kern_flag &= ~MNTK_ASYNC;
1318 cache_purgevfs(mp); /* remove cache entries for this file sys */
1319 vfs_deallocate_syncvnode(mp);
1321 * For forced unmounts, move process cdir/rdir refs on the fs root
1322 * vnode to the covered vnode. For non-forced unmounts we want
1323 * such references to cause an EBUSY error.
1325 if ((flags & MNT_FORCE) &&
1326 VFS_ROOT(mp, LK_EXCLUSIVE, &fsrootvp) == 0) {
1327 if (mp->mnt_vnodecovered != NULL &&
1328 (mp->mnt_flag & MNT_IGNORE) == 0)
1329 mountcheckdirs(fsrootvp, mp->mnt_vnodecovered);
1330 if (fsrootvp == rootvnode) {
1336 if ((mp->mnt_flag & MNT_RDONLY) != 0 || (flags & MNT_FORCE) != 0 ||
1337 (error = VFS_SYNC(mp, MNT_WAIT)) == 0)
1338 error = VFS_UNMOUNT(mp, flags);
1339 vn_finished_write(mp);
1341 * If we failed to flush the dirty blocks for this mount point,
1342 * undo all the cdir/rdir and rootvnode changes we made above.
1343 * Unless we failed to do so because the device is reporting that
1344 * it doesn't exist anymore.
1346 if (error && error != ENXIO) {
1347 if ((flags & MNT_FORCE) &&
1348 VFS_ROOT(mp, LK_EXCLUSIVE, &fsrootvp) == 0) {
1349 if (mp->mnt_vnodecovered != NULL &&
1350 (mp->mnt_flag & MNT_IGNORE) == 0)
1351 mountcheckdirs(mp->mnt_vnodecovered, fsrootvp);
1352 if (rootvnode == NULL) {
1353 rootvnode = fsrootvp;
1359 mp->mnt_kern_flag &= ~MNTK_NOINSMNTQ;
1360 if ((mp->mnt_flag & MNT_RDONLY) == 0) {
1362 vfs_allocate_syncvnode(mp);
1365 mp->mnt_kern_flag &= ~(MNTK_UNMOUNT | MNTK_UNMOUNTF);
1366 mp->mnt_flag |= async_flag;
1367 if ((mp->mnt_flag & MNT_ASYNC) != 0 &&
1368 (mp->mnt_kern_flag & MNTK_NOASYNC) == 0)
1369 mp->mnt_kern_flag |= MNTK_ASYNC;
1370 if (mp->mnt_kern_flag & MNTK_MWAIT) {
1371 mp->mnt_kern_flag &= ~MNTK_MWAIT;
1376 VOP_UNLOCK(coveredvp, 0);
1379 mtx_lock(&mountlist_mtx);
1380 TAILQ_REMOVE(&mountlist, mp, mnt_list);
1381 mtx_unlock(&mountlist_mtx);
1382 EVENTHANDLER_INVOKE(vfs_unmounted, mp, td);
1383 if (coveredvp != NULL) {
1384 coveredvp->v_mountedhere = NULL;
1385 VOP_UNLOCK(coveredvp, 0);
1387 vfs_event_signal(NULL, VQ_UNMOUNT, 0);
1388 if (mp == rootdevmp)
1390 vfs_mount_destroy(mp);
1395 * Report errors during filesystem mounting.
1398 vfs_mount_error(struct mount *mp, const char *fmt, ...)
1400 struct vfsoptlist *moptlist = mp->mnt_optnew;
1405 error = vfs_getopt(moptlist, "errmsg", (void **)&errmsg, &len);
1406 if (error || errmsg == NULL || len <= 0)
1410 vsnprintf(errmsg, (size_t)len, fmt, ap);
1415 vfs_opterror(struct vfsoptlist *opts, const char *fmt, ...)
1421 error = vfs_getopt(opts, "errmsg", (void **)&errmsg, &len);
1422 if (error || errmsg == NULL || len <= 0)
1426 vsnprintf(errmsg, (size_t)len, fmt, ap);
1431 * ---------------------------------------------------------------------
1432 * Functions for querying mount options/arguments from filesystems.
1436 * Check that no unknown options are given
1439 vfs_filteropt(struct vfsoptlist *opts, const char **legal)
1443 const char **t, *p, *q;
1446 TAILQ_FOREACH(opt, opts, link) {
1449 if (p[0] == 'n' && p[1] == 'o')
1451 for(t = global_opts; *t != NULL; t++) {
1452 if (strcmp(*t, p) == 0)
1455 if (strcmp(*t, q) == 0)
1461 for(t = legal; *t != NULL; t++) {
1462 if (strcmp(*t, p) == 0)
1465 if (strcmp(*t, q) == 0)
1471 snprintf(errmsg, sizeof(errmsg),
1472 "mount option <%s> is unknown", p);
1476 TAILQ_FOREACH(opt, opts, link) {
1477 if (strcmp(opt->name, "errmsg") == 0) {
1478 strncpy((char *)opt->value, errmsg, opt->len);
1483 printf("%s\n", errmsg);
1489 * Get a mount option by its name.
1491 * Return 0 if the option was found, ENOENT otherwise.
1492 * If len is non-NULL it will be filled with the length
1493 * of the option. If buf is non-NULL, it will be filled
1494 * with the address of the option.
1497 vfs_getopt(opts, name, buf, len)
1498 struct vfsoptlist *opts;
1505 KASSERT(opts != NULL, ("vfs_getopt: caller passed 'opts' as NULL"));
1507 TAILQ_FOREACH(opt, opts, link) {
1508 if (strcmp(name, opt->name) == 0) {
1521 vfs_getopt_pos(struct vfsoptlist *opts, const char *name)
1528 TAILQ_FOREACH(opt, opts, link) {
1529 if (strcmp(name, opt->name) == 0) {
1538 vfs_getopt_size(struct vfsoptlist *opts, const char *name, off_t *value)
1540 char *opt_value, *vtp;
1544 error = vfs_getopt(opts, name, (void **)&opt_value, &opt_len);
1547 if (opt_len == 0 || opt_value == NULL)
1549 if (opt_value[0] == '\0' || opt_value[opt_len - 1] != '\0')
1551 iv = strtoq(opt_value, &vtp, 0);
1552 if (vtp == opt_value || (vtp[0] != '\0' && vtp[1] != '\0'))
1580 vfs_getopts(struct vfsoptlist *opts, const char *name, int *error)
1585 TAILQ_FOREACH(opt, opts, link) {
1586 if (strcmp(name, opt->name) != 0)
1589 if (opt->len == 0 ||
1590 ((char *)opt->value)[opt->len - 1] != '\0') {
1594 return (opt->value);
1601 vfs_flagopt(struct vfsoptlist *opts, const char *name, uint64_t *w,
1606 TAILQ_FOREACH(opt, opts, link) {
1607 if (strcmp(name, opt->name) == 0) {
1620 vfs_scanopt(struct vfsoptlist *opts, const char *name, const char *fmt, ...)
1626 KASSERT(opts != NULL, ("vfs_getopt: caller passed 'opts' as NULL"));
1628 TAILQ_FOREACH(opt, opts, link) {
1629 if (strcmp(name, opt->name) != 0)
1632 if (opt->len == 0 || opt->value == NULL)
1634 if (((char *)opt->value)[opt->len - 1] != '\0')
1637 ret = vsscanf(opt->value, fmt, ap);
1645 vfs_setopt(struct vfsoptlist *opts, const char *name, void *value, int len)
1649 TAILQ_FOREACH(opt, opts, link) {
1650 if (strcmp(name, opt->name) != 0)
1653 if (opt->value == NULL)
1656 if (opt->len != len)
1658 bcopy(value, opt->value, len);
1666 vfs_setopt_part(struct vfsoptlist *opts, const char *name, void *value, int len)
1670 TAILQ_FOREACH(opt, opts, link) {
1671 if (strcmp(name, opt->name) != 0)
1674 if (opt->value == NULL)
1680 bcopy(value, opt->value, len);
1688 vfs_setopts(struct vfsoptlist *opts, const char *name, const char *value)
1692 TAILQ_FOREACH(opt, opts, link) {
1693 if (strcmp(name, opt->name) != 0)
1696 if (opt->value == NULL)
1697 opt->len = strlen(value) + 1;
1698 else if (strlcpy(opt->value, value, opt->len) >= opt->len)
1706 * Find and copy a mount option.
1708 * The size of the buffer has to be specified
1709 * in len, if it is not the same length as the
1710 * mount option, EINVAL is returned.
1711 * Returns ENOENT if the option is not found.
1714 vfs_copyopt(opts, name, dest, len)
1715 struct vfsoptlist *opts;
1722 KASSERT(opts != NULL, ("vfs_copyopt: caller passed 'opts' as NULL"));
1724 TAILQ_FOREACH(opt, opts, link) {
1725 if (strcmp(name, opt->name) == 0) {
1727 if (len != opt->len)
1729 bcopy(opt->value, dest, opt->len);
1737 __vfs_statfs(struct mount *mp, struct statfs *sbp)
1741 error = mp->mnt_op->vfs_statfs(mp, &mp->mnt_stat);
1742 if (sbp != &mp->mnt_stat)
1743 *sbp = mp->mnt_stat;
1748 vfs_mountedfrom(struct mount *mp, const char *from)
1751 bzero(mp->mnt_stat.f_mntfromname, sizeof mp->mnt_stat.f_mntfromname);
1752 strlcpy(mp->mnt_stat.f_mntfromname, from,
1753 sizeof mp->mnt_stat.f_mntfromname);
1757 * ---------------------------------------------------------------------
1758 * This is the api for building mount args and mounting filesystems from
1759 * inside the kernel.
1761 * The API works by accumulation of individual args. First error is
1764 * XXX: should be documented in new manpage kernel_mount(9)
1767 /* A memory allocation which must be freed when we are done */
1769 SLIST_ENTRY(mntaarg) next;
1772 /* The header for the mount arguments */
1777 SLIST_HEAD(, mntaarg) list;
1781 * Add a boolean argument.
1783 * flag is the boolean value.
1784 * name must start with "no".
1787 mount_argb(struct mntarg *ma, int flag, const char *name)
1790 KASSERT(name[0] == 'n' && name[1] == 'o',
1791 ("mount_argb(...,%s): name must start with 'no'", name));
1793 return (mount_arg(ma, name + (flag ? 2 : 0), NULL, 0));
1797 * Add an argument printf style
1800 mount_argf(struct mntarg *ma, const char *name, const char *fmt, ...)
1803 struct mntaarg *maa;
1808 ma = malloc(sizeof *ma, M_MOUNT, M_WAITOK | M_ZERO);
1809 SLIST_INIT(&ma->list);
1814 ma->v = realloc(ma->v, sizeof *ma->v * (ma->len + 2),
1816 ma->v[ma->len].iov_base = (void *)(uintptr_t)name;
1817 ma->v[ma->len].iov_len = strlen(name) + 1;
1820 sb = sbuf_new_auto();
1822 sbuf_vprintf(sb, fmt, ap);
1825 len = sbuf_len(sb) + 1;
1826 maa = malloc(sizeof *maa + len, M_MOUNT, M_WAITOK | M_ZERO);
1827 SLIST_INSERT_HEAD(&ma->list, maa, next);
1828 bcopy(sbuf_data(sb), maa + 1, len);
1831 ma->v[ma->len].iov_base = maa + 1;
1832 ma->v[ma->len].iov_len = len;
1839 * Add an argument which is a userland string.
1842 mount_argsu(struct mntarg *ma, const char *name, const void *val, int len)
1844 struct mntaarg *maa;
1850 ma = malloc(sizeof *ma, M_MOUNT, M_WAITOK | M_ZERO);
1851 SLIST_INIT(&ma->list);
1855 maa = malloc(sizeof *maa + len, M_MOUNT, M_WAITOK | M_ZERO);
1856 SLIST_INSERT_HEAD(&ma->list, maa, next);
1857 tbuf = (void *)(maa + 1);
1858 ma->error = copyinstr(val, tbuf, len, NULL);
1859 return (mount_arg(ma, name, tbuf, -1));
1865 * If length is -1, treat value as a C string.
1868 mount_arg(struct mntarg *ma, const char *name, const void *val, int len)
1872 ma = malloc(sizeof *ma, M_MOUNT, M_WAITOK | M_ZERO);
1873 SLIST_INIT(&ma->list);
1878 ma->v = realloc(ma->v, sizeof *ma->v * (ma->len + 2),
1880 ma->v[ma->len].iov_base = (void *)(uintptr_t)name;
1881 ma->v[ma->len].iov_len = strlen(name) + 1;
1884 ma->v[ma->len].iov_base = (void *)(uintptr_t)val;
1886 ma->v[ma->len].iov_len = strlen(val) + 1;
1888 ma->v[ma->len].iov_len = len;
1894 * Free a mntarg structure
1897 free_mntarg(struct mntarg *ma)
1899 struct mntaarg *maa;
1901 while (!SLIST_EMPTY(&ma->list)) {
1902 maa = SLIST_FIRST(&ma->list);
1903 SLIST_REMOVE_HEAD(&ma->list, next);
1906 free(ma->v, M_MOUNT);
1911 * Mount a filesystem
1914 kernel_mount(struct mntarg *ma, uint64_t flags)
1919 KASSERT(ma != NULL, ("kernel_mount NULL ma"));
1920 KASSERT(ma->v != NULL, ("kernel_mount NULL ma->v"));
1921 KASSERT(!(ma->len & 1), ("kernel_mount odd ma->len (%d)", ma->len));
1923 auio.uio_iov = ma->v;
1924 auio.uio_iovcnt = ma->len;
1925 auio.uio_segflg = UIO_SYSSPACE;
1929 error = vfs_donmount(curthread, flags, &auio);
1935 * A printflike function to mount a filesystem.
1938 kernel_vmount(int flags, ...)
1940 struct mntarg *ma = NULL;
1946 va_start(ap, flags);
1948 cp = va_arg(ap, const char *);
1951 vp = va_arg(ap, const void *);
1952 ma = mount_arg(ma, cp, vp, (vp != NULL ? -1 : 0));
1956 error = kernel_mount(ma, flags);
1961 vfs_oexport_conv(const struct oexport_args *oexp, struct export_args *exp)
1964 bcopy(oexp, exp, sizeof(*oexp));
1965 exp->ex_numsecflavors = 0;