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 * 3. 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 MALLOC_DEFINE(M_STATFS, "statfs", "statfs structure");
83 static uma_zone_t mount_zone;
85 /* List of mounted filesystems. */
86 struct mntlist mountlist = TAILQ_HEAD_INITIALIZER(mountlist);
88 /* For any iteration/modification of mountlist */
89 struct mtx mountlist_mtx;
90 MTX_SYSINIT(mountlist, &mountlist_mtx, "mountlist", MTX_DEF);
93 * Global opts, taken by all filesystems
95 static const char *global_opts[] = {
107 mount_init(void *mem, int size, int flags)
111 mp = (struct mount *)mem;
112 mtx_init(&mp->mnt_mtx, "struct mount mtx", NULL, MTX_DEF);
113 mtx_init(&mp->mnt_listmtx, "struct mount vlist mtx", NULL, MTX_DEF);
114 lockinit(&mp->mnt_explock, PVFS, "explock", 0, 0);
119 mount_fini(void *mem, int size)
123 mp = (struct mount *)mem;
124 lockdestroy(&mp->mnt_explock);
125 mtx_destroy(&mp->mnt_listmtx);
126 mtx_destroy(&mp->mnt_mtx);
130 vfs_mount_init(void *dummy __unused)
133 mount_zone = uma_zcreate("Mountpoints", sizeof(struct mount), NULL,
134 NULL, mount_init, mount_fini, UMA_ALIGN_PTR, UMA_ZONE_NOFREE);
136 SYSINIT(vfs_mount, SI_SUB_VFS, SI_ORDER_ANY, vfs_mount_init, NULL);
139 * ---------------------------------------------------------------------
140 * Functions for building and sanitizing the mount options
143 /* Remove one mount option. */
145 vfs_freeopt(struct vfsoptlist *opts, struct vfsopt *opt)
148 TAILQ_REMOVE(opts, opt, link);
149 free(opt->name, M_MOUNT);
150 if (opt->value != NULL)
151 free(opt->value, M_MOUNT);
155 /* Release all resources related to the mount options. */
157 vfs_freeopts(struct vfsoptlist *opts)
161 while (!TAILQ_EMPTY(opts)) {
162 opt = TAILQ_FIRST(opts);
163 vfs_freeopt(opts, opt);
169 vfs_deleteopt(struct vfsoptlist *opts, const char *name)
171 struct vfsopt *opt, *temp;
175 TAILQ_FOREACH_SAFE(opt, opts, link, temp) {
176 if (strcmp(opt->name, name) == 0)
177 vfs_freeopt(opts, opt);
182 vfs_isopt_ro(const char *opt)
185 if (strcmp(opt, "ro") == 0 || strcmp(opt, "rdonly") == 0 ||
186 strcmp(opt, "norw") == 0)
192 vfs_isopt_rw(const char *opt)
195 if (strcmp(opt, "rw") == 0 || strcmp(opt, "noro") == 0)
201 * Check if options are equal (with or without the "no" prefix).
204 vfs_equalopts(const char *opt1, const char *opt2)
208 /* "opt" vs. "opt" or "noopt" vs. "noopt" */
209 if (strcmp(opt1, opt2) == 0)
211 /* "noopt" vs. "opt" */
212 if (strncmp(opt1, "no", 2) == 0 && strcmp(opt1 + 2, opt2) == 0)
214 /* "opt" vs. "noopt" */
215 if (strncmp(opt2, "no", 2) == 0 && strcmp(opt1, opt2 + 2) == 0)
217 while ((p = strchr(opt1, '.')) != NULL &&
218 !strncmp(opt1, opt2, ++p - opt1)) {
221 /* "foo.noopt" vs. "foo.opt" */
222 if (strncmp(opt1, "no", 2) == 0 && strcmp(opt1 + 2, opt2) == 0)
224 /* "foo.opt" vs. "foo.noopt" */
225 if (strncmp(opt2, "no", 2) == 0 && strcmp(opt1, opt2 + 2) == 0)
228 /* "ro" / "rdonly" / "norw" / "rw" / "noro" */
229 if ((vfs_isopt_ro(opt1) || vfs_isopt_rw(opt1)) &&
230 (vfs_isopt_ro(opt2) || vfs_isopt_rw(opt2)))
236 * If a mount option is specified several times,
237 * (with or without the "no" prefix) only keep
238 * the last occurrence of it.
241 vfs_sanitizeopts(struct vfsoptlist *opts)
243 struct vfsopt *opt, *opt2, *tmp;
245 TAILQ_FOREACH_REVERSE(opt, opts, vfsoptlist, link) {
246 opt2 = TAILQ_PREV(opt, vfsoptlist, link);
247 while (opt2 != NULL) {
248 if (vfs_equalopts(opt->name, opt2->name)) {
249 tmp = TAILQ_PREV(opt2, vfsoptlist, link);
250 vfs_freeopt(opts, opt2);
253 opt2 = TAILQ_PREV(opt2, vfsoptlist, link);
260 * Build a linked list of mount options from a struct uio.
263 vfs_buildopts(struct uio *auio, struct vfsoptlist **options)
265 struct vfsoptlist *opts;
267 size_t memused, namelen, optlen;
268 unsigned int i, iovcnt;
271 opts = malloc(sizeof(struct vfsoptlist), M_MOUNT, M_WAITOK);
274 iovcnt = auio->uio_iovcnt;
275 for (i = 0; i < iovcnt; i += 2) {
276 namelen = auio->uio_iov[i].iov_len;
277 optlen = auio->uio_iov[i + 1].iov_len;
278 memused += sizeof(struct vfsopt) + optlen + namelen;
280 * Avoid consuming too much memory, and attempts to overflow
283 if (memused > VFS_MOUNTARG_SIZE_MAX ||
284 optlen > VFS_MOUNTARG_SIZE_MAX ||
285 namelen > VFS_MOUNTARG_SIZE_MAX) {
290 opt = malloc(sizeof(struct vfsopt), M_MOUNT, M_WAITOK);
291 opt->name = malloc(namelen, M_MOUNT, M_WAITOK);
298 * Do this early, so jumps to "bad" will free the current
301 TAILQ_INSERT_TAIL(opts, opt, link);
303 if (auio->uio_segflg == UIO_SYSSPACE) {
304 bcopy(auio->uio_iov[i].iov_base, opt->name, namelen);
306 error = copyin(auio->uio_iov[i].iov_base, opt->name,
311 /* Ensure names are null-terminated strings. */
312 if (namelen == 0 || opt->name[namelen - 1] != '\0') {
318 opt->value = malloc(optlen, M_MOUNT, M_WAITOK);
319 if (auio->uio_segflg == UIO_SYSSPACE) {
320 bcopy(auio->uio_iov[i + 1].iov_base, opt->value,
323 error = copyin(auio->uio_iov[i + 1].iov_base,
330 vfs_sanitizeopts(opts);
339 * Merge the old mount options with the new ones passed
340 * in the MNT_UPDATE case.
342 * XXX: This function will keep a "nofoo" option in the new
343 * options. E.g, if the option's canonical name is "foo",
344 * "nofoo" ends up in the mount point's active options.
347 vfs_mergeopts(struct vfsoptlist *toopts, struct vfsoptlist *oldopts)
349 struct vfsopt *opt, *new;
351 TAILQ_FOREACH(opt, oldopts, link) {
352 new = malloc(sizeof(struct vfsopt), M_MOUNT, M_WAITOK);
353 new->name = strdup(opt->name, M_MOUNT);
355 new->value = malloc(opt->len, M_MOUNT, M_WAITOK);
356 bcopy(opt->value, new->value, opt->len);
360 new->seen = opt->seen;
361 TAILQ_INSERT_HEAD(toopts, new, link);
363 vfs_sanitizeopts(toopts);
367 * Mount a filesystem.
369 #ifndef _SYS_SYSPROTO_H_
377 sys_nmount(struct thread *td, struct nmount_args *uap)
385 * Mount flags are now 64-bits. On 32-bit archtectures only
386 * 32-bits are passed in, but from here on everything handles
387 * 64-bit flags correctly.
391 AUDIT_ARG_FFLAGS(flags);
392 CTR4(KTR_VFS, "%s: iovp %p with iovcnt %d and flags %d", __func__,
393 uap->iovp, uap->iovcnt, flags);
396 * Filter out MNT_ROOTFS. We do not want clients of nmount() in
397 * userspace to set this flag, but we must filter it out if we want
398 * MNT_UPDATE on the root file system to work.
399 * MNT_ROOTFS should only be set by the kernel when mounting its
402 flags &= ~MNT_ROOTFS;
404 iovcnt = uap->iovcnt;
406 * Check that we have an even number of iovec's
407 * and that we have at least two options.
409 if ((iovcnt & 1) || (iovcnt < 4)) {
410 CTR2(KTR_VFS, "%s: failed for invalid iovcnt %d", __func__,
415 error = copyinuio(uap->iovp, iovcnt, &auio);
417 CTR2(KTR_VFS, "%s: failed for invalid uio op with %d errno",
421 error = vfs_donmount(td, flags, auio);
428 * ---------------------------------------------------------------------
429 * Various utility functions
433 vfs_ref(struct mount *mp)
436 CTR2(KTR_VFS, "%s: mp %p", __func__, mp);
443 vfs_rel(struct mount *mp)
446 CTR2(KTR_VFS, "%s: mp %p", __func__, mp);
453 * Allocate and initialize the mount point struct.
456 vfs_mount_alloc(struct vnode *vp, struct vfsconf *vfsp, const char *fspath,
461 mp = uma_zalloc(mount_zone, M_WAITOK);
462 bzero(&mp->mnt_startzero,
463 __rangeof(struct mount, mnt_startzero, mnt_endzero));
464 TAILQ_INIT(&mp->mnt_nvnodelist);
465 mp->mnt_nvnodelistsize = 0;
466 TAILQ_INIT(&mp->mnt_activevnodelist);
467 mp->mnt_activevnodelistsize = 0;
468 TAILQ_INIT(&mp->mnt_tmpfreevnodelist);
469 mp->mnt_tmpfreevnodelistsize = 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_tmpfreevnodelistsize != 0)
528 panic("vfs_mount_destroy: nonzero tmpfreevnodelistsize");
529 if (mp->mnt_lockref != 0)
530 panic("vfs_mount_destroy: nonzero lock refcount");
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_donmount(struct thread *td, uint64_t fsflags, struct uio *fsoptions)
544 struct vfsoptlist *optlist;
545 struct vfsopt *opt, *tmp_opt;
546 char *fstype, *fspath, *errmsg;
547 int error, fstypelen, fspathlen, errmsg_len, errmsg_pos;
549 errmsg = fspath = NULL;
550 errmsg_len = fspathlen = 0;
553 error = vfs_buildopts(fsoptions, &optlist);
557 if (vfs_getopt(optlist, "errmsg", (void **)&errmsg, &errmsg_len) == 0)
558 errmsg_pos = vfs_getopt_pos(optlist, "errmsg");
561 * We need these two options before the others,
562 * and they are mandatory for any filesystem.
563 * Ensure they are NUL terminated as well.
566 error = vfs_getopt(optlist, "fstype", (void **)&fstype, &fstypelen);
567 if (error || fstype[fstypelen - 1] != '\0') {
570 strncpy(errmsg, "Invalid fstype", errmsg_len);
574 error = vfs_getopt(optlist, "fspath", (void **)&fspath, &fspathlen);
575 if (error || fspath[fspathlen - 1] != '\0') {
578 strncpy(errmsg, "Invalid fspath", errmsg_len);
583 * We need to see if we have the "update" option
584 * before we call vfs_domount(), since vfs_domount() has special
585 * logic based on MNT_UPDATE. This is very important
586 * when we want to update the root filesystem.
588 TAILQ_FOREACH_SAFE(opt, optlist, link, tmp_opt) {
589 if (strcmp(opt->name, "update") == 0) {
590 fsflags |= MNT_UPDATE;
591 vfs_freeopt(optlist, opt);
593 else if (strcmp(opt->name, "async") == 0)
594 fsflags |= MNT_ASYNC;
595 else if (strcmp(opt->name, "force") == 0) {
596 fsflags |= MNT_FORCE;
597 vfs_freeopt(optlist, opt);
599 else if (strcmp(opt->name, "reload") == 0) {
600 fsflags |= MNT_RELOAD;
601 vfs_freeopt(optlist, opt);
603 else if (strcmp(opt->name, "multilabel") == 0)
604 fsflags |= MNT_MULTILABEL;
605 else if (strcmp(opt->name, "noasync") == 0)
606 fsflags &= ~MNT_ASYNC;
607 else if (strcmp(opt->name, "noatime") == 0)
608 fsflags |= MNT_NOATIME;
609 else if (strcmp(opt->name, "atime") == 0) {
610 free(opt->name, M_MOUNT);
611 opt->name = strdup("nonoatime", M_MOUNT);
613 else if (strcmp(opt->name, "noclusterr") == 0)
614 fsflags |= MNT_NOCLUSTERR;
615 else if (strcmp(opt->name, "clusterr") == 0) {
616 free(opt->name, M_MOUNT);
617 opt->name = strdup("nonoclusterr", M_MOUNT);
619 else if (strcmp(opt->name, "noclusterw") == 0)
620 fsflags |= MNT_NOCLUSTERW;
621 else if (strcmp(opt->name, "clusterw") == 0) {
622 free(opt->name, M_MOUNT);
623 opt->name = strdup("nonoclusterw", M_MOUNT);
625 else if (strcmp(opt->name, "noexec") == 0)
626 fsflags |= MNT_NOEXEC;
627 else if (strcmp(opt->name, "exec") == 0) {
628 free(opt->name, M_MOUNT);
629 opt->name = strdup("nonoexec", M_MOUNT);
631 else if (strcmp(opt->name, "nosuid") == 0)
632 fsflags |= MNT_NOSUID;
633 else if (strcmp(opt->name, "suid") == 0) {
634 free(opt->name, M_MOUNT);
635 opt->name = strdup("nonosuid", M_MOUNT);
637 else if (strcmp(opt->name, "nosymfollow") == 0)
638 fsflags |= MNT_NOSYMFOLLOW;
639 else if (strcmp(opt->name, "symfollow") == 0) {
640 free(opt->name, M_MOUNT);
641 opt->name = strdup("nonosymfollow", M_MOUNT);
643 else if (strcmp(opt->name, "noro") == 0)
644 fsflags &= ~MNT_RDONLY;
645 else if (strcmp(opt->name, "rw") == 0)
646 fsflags &= ~MNT_RDONLY;
647 else if (strcmp(opt->name, "ro") == 0)
648 fsflags |= MNT_RDONLY;
649 else if (strcmp(opt->name, "rdonly") == 0) {
650 free(opt->name, M_MOUNT);
651 opt->name = strdup("ro", M_MOUNT);
652 fsflags |= MNT_RDONLY;
654 else if (strcmp(opt->name, "suiddir") == 0)
655 fsflags |= MNT_SUIDDIR;
656 else if (strcmp(opt->name, "sync") == 0)
657 fsflags |= MNT_SYNCHRONOUS;
658 else if (strcmp(opt->name, "union") == 0)
659 fsflags |= MNT_UNION;
660 else if (strcmp(opt->name, "automounted") == 0) {
661 fsflags |= MNT_AUTOMOUNTED;
662 vfs_freeopt(optlist, opt);
667 * Be ultra-paranoid about making sure the type and fspath
668 * variables will fit in our mp buffers, including the
671 if (fstypelen > MFSNAMELEN || fspathlen > MNAMELEN) {
672 error = ENAMETOOLONG;
676 error = vfs_domount(td, fstype, fspath, fsflags, &optlist);
678 /* copyout the errmsg */
679 if (errmsg_pos != -1 && ((2 * errmsg_pos + 1) < fsoptions->uio_iovcnt)
680 && errmsg_len > 0 && errmsg != NULL) {
681 if (fsoptions->uio_segflg == UIO_SYSSPACE) {
683 fsoptions->uio_iov[2 * errmsg_pos + 1].iov_base,
684 fsoptions->uio_iov[2 * errmsg_pos + 1].iov_len);
687 fsoptions->uio_iov[2 * errmsg_pos + 1].iov_base,
688 fsoptions->uio_iov[2 * errmsg_pos + 1].iov_len);
693 vfs_freeopts(optlist);
700 #ifndef _SYS_SYSPROTO_H_
710 sys_mount(struct thread *td, struct mount_args *uap)
713 struct vfsconf *vfsp = NULL;
714 struct mntarg *ma = NULL;
719 * Mount flags are now 64-bits. On 32-bit architectures only
720 * 32-bits are passed in, but from here on everything handles
721 * 64-bit flags correctly.
725 AUDIT_ARG_FFLAGS(flags);
728 * Filter out MNT_ROOTFS. We do not want clients of mount() in
729 * userspace to set this flag, but we must filter it out if we want
730 * MNT_UPDATE on the root file system to work.
731 * MNT_ROOTFS should only be set by the kernel when mounting its
734 flags &= ~MNT_ROOTFS;
736 fstype = malloc(MFSNAMELEN, M_TEMP, M_WAITOK);
737 error = copyinstr(uap->type, fstype, MFSNAMELEN, NULL);
739 free(fstype, M_TEMP);
743 AUDIT_ARG_TEXT(fstype);
744 vfsp = vfs_byname_kld(fstype, td, &error);
745 free(fstype, M_TEMP);
748 if (vfsp->vfc_vfsops->vfs_cmount == NULL)
751 ma = mount_argsu(ma, "fstype", uap->type, MFSNAMELEN);
752 ma = mount_argsu(ma, "fspath", uap->path, MNAMELEN);
753 ma = mount_argb(ma, flags & MNT_RDONLY, "noro");
754 ma = mount_argb(ma, !(flags & MNT_NOSUID), "nosuid");
755 ma = mount_argb(ma, !(flags & MNT_NOEXEC), "noexec");
757 error = vfsp->vfc_vfsops->vfs_cmount(ma, uap->data, flags);
762 * vfs_domount_first(): first file system mount (not update)
766 struct thread *td, /* Calling thread. */
767 struct vfsconf *vfsp, /* File system type. */
768 char *fspath, /* Mount path. */
769 struct vnode *vp, /* Vnode to be covered. */
770 uint64_t fsflags, /* Flags common to all filesystems. */
771 struct vfsoptlist **optlist /* Options local to the filesystem. */
779 ASSERT_VOP_ELOCKED(vp, __func__);
780 KASSERT((fsflags & MNT_UPDATE) == 0, ("MNT_UPDATE shouldn't be here"));
783 * If the user is not root, ensure that they own the directory
784 * onto which we are attempting to mount.
786 error = VOP_GETATTR(vp, &va, td->td_ucred);
787 if (error == 0 && va.va_uid != td->td_ucred->cr_uid)
788 error = priv_check_cred(td->td_ucred, PRIV_VFS_ADMIN, 0);
790 error = vinvalbuf(vp, V_SAVE, 0, 0);
791 if (error == 0 && vp->v_type != VDIR)
795 if ((vp->v_iflag & VI_MOUNT) == 0 && vp->v_mountedhere == NULL)
796 vp->v_iflag |= VI_MOUNT;
807 /* Allocate and initialize the filesystem. */
808 mp = vfs_mount_alloc(vp, vfsp, fspath, td->td_ucred);
809 /* XXXMAC: pass to vfs_mount_alloc? */
810 mp->mnt_optnew = *optlist;
811 /* Set the mount level flags. */
812 mp->mnt_flag = (fsflags & (MNT_UPDATEMASK | MNT_ROOTFS | MNT_RDONLY));
815 * Mount the filesystem.
816 * XXX The final recipients of VFS_MOUNT just overwrite the ndp they
817 * get. No freeing of cn_pnbuf.
819 error = VFS_MOUNT(mp);
822 vfs_mount_destroy(mp);
824 vp->v_iflag &= ~VI_MOUNT;
830 if (mp->mnt_opt != NULL)
831 vfs_freeopts(mp->mnt_opt);
832 mp->mnt_opt = mp->mnt_optnew;
834 (void)VFS_STATFS(mp, &mp->mnt_stat);
837 * Prevent external consumers of mount options from reading mnt_optnew.
839 mp->mnt_optnew = NULL;
842 if ((mp->mnt_flag & MNT_ASYNC) != 0 &&
843 (mp->mnt_kern_flag & MNTK_NOASYNC) == 0)
844 mp->mnt_kern_flag |= MNTK_ASYNC;
846 mp->mnt_kern_flag &= ~MNTK_ASYNC;
849 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
852 vp->v_iflag &= ~VI_MOUNT;
854 vp->v_mountedhere = mp;
855 /* Place the new filesystem at the end of the mount list. */
856 mtx_lock(&mountlist_mtx);
857 TAILQ_INSERT_TAIL(&mountlist, mp, mnt_list);
858 mtx_unlock(&mountlist_mtx);
859 vfs_event_signal(NULL, VQ_MOUNT, 0);
860 if (VFS_ROOT(mp, LK_EXCLUSIVE, &newdp))
861 panic("mount: lost mount");
863 EVENTHANDLER_INVOKE(vfs_mounted, mp, newdp, td);
864 VOP_UNLOCK(newdp, 0);
865 mountcheckdirs(vp, newdp);
867 if ((mp->mnt_flag & MNT_RDONLY) == 0)
868 vfs_allocate_syncvnode(mp);
874 * vfs_domount_update(): update of mounted file system
878 struct thread *td, /* Calling thread. */
879 struct vnode *vp, /* Mount point vnode. */
880 uint64_t fsflags, /* Flags common to all filesystems. */
881 struct vfsoptlist **optlist /* Options local to the filesystem. */
884 struct export_args export;
887 int error, export_error, len;
890 ASSERT_VOP_ELOCKED(vp, __func__);
891 KASSERT((fsflags & MNT_UPDATE) != 0, ("MNT_UPDATE should be here"));
894 if ((vp->v_vflag & VV_ROOT) == 0) {
895 if (vfs_copyopt(*optlist, "export", &export, sizeof(export))
905 * We only allow the filesystem to be reloaded if it
906 * is currently mounted read-only.
909 if ((fsflags & MNT_RELOAD) != 0 && (flag & MNT_RDONLY) == 0) {
911 return (EOPNOTSUPP); /* Needs translation */
914 * Only privileged root, or (if MNT_USER is set) the user that
915 * did the original mount is permitted to update it.
917 error = vfs_suser(mp, td);
922 if (vfs_busy(mp, MBF_NOWAIT)) {
927 if ((vp->v_iflag & VI_MOUNT) != 0 || vp->v_mountedhere != NULL) {
933 vp->v_iflag |= VI_MOUNT;
938 if ((mp->mnt_kern_flag & MNTK_UNMOUNT) != 0) {
943 mp->mnt_flag &= ~MNT_UPDATEMASK;
944 mp->mnt_flag |= fsflags & (MNT_RELOAD | MNT_FORCE | MNT_UPDATE |
945 MNT_SNAPSHOT | MNT_ROOTFS | MNT_UPDATEMASK | MNT_RDONLY);
946 if ((mp->mnt_flag & MNT_ASYNC) == 0)
947 mp->mnt_kern_flag &= ~MNTK_ASYNC;
949 mp->mnt_optnew = *optlist;
950 vfs_mergeopts(mp->mnt_optnew, mp->mnt_opt);
953 * Mount the filesystem.
954 * XXX The final recipients of VFS_MOUNT just overwrite the ndp they
955 * get. No freeing of cn_pnbuf.
957 error = VFS_MOUNT(mp);
960 /* Process the export option. */
961 if (error == 0 && vfs_getopt(mp->mnt_optnew, "export", &bufp,
963 /* Assume that there is only 1 ABI for each length. */
965 case (sizeof(struct oexport_args)):
966 bzero(&export, sizeof(export));
968 case (sizeof(export)):
969 bcopy(bufp, &export, len);
970 export_error = vfs_export(mp, &export);
973 export_error = EINVAL;
980 mp->mnt_flag &= ~(MNT_UPDATE | MNT_RELOAD | MNT_FORCE |
984 * If we fail, restore old mount flags. MNT_QUOTA is special,
985 * because it is not part of MNT_UPDATEMASK, but it could have
986 * changed in the meantime if quotactl(2) was called.
987 * All in all we want current value of MNT_QUOTA, not the old
990 mp->mnt_flag = (mp->mnt_flag & MNT_QUOTA) | (flag & ~MNT_QUOTA);
992 if ((mp->mnt_flag & MNT_ASYNC) != 0 &&
993 (mp->mnt_kern_flag & MNTK_NOASYNC) == 0)
994 mp->mnt_kern_flag |= MNTK_ASYNC;
996 mp->mnt_kern_flag &= ~MNTK_ASYNC;
1002 if (mp->mnt_opt != NULL)
1003 vfs_freeopts(mp->mnt_opt);
1004 mp->mnt_opt = mp->mnt_optnew;
1006 (void)VFS_STATFS(mp, &mp->mnt_stat);
1008 * Prevent external consumers of mount options from reading
1011 mp->mnt_optnew = NULL;
1013 if ((mp->mnt_flag & MNT_RDONLY) == 0)
1014 vfs_allocate_syncvnode(mp);
1016 vfs_deallocate_syncvnode(mp);
1020 vp->v_iflag &= ~VI_MOUNT;
1023 return (error != 0 ? error : export_error);
1027 * vfs_domount(): actually attempt a filesystem mount.
1031 struct thread *td, /* Calling thread. */
1032 const char *fstype, /* Filesystem type. */
1033 char *fspath, /* Mount path. */
1034 uint64_t fsflags, /* Flags common to all filesystems. */
1035 struct vfsoptlist **optlist /* Options local to the filesystem. */
1038 struct vfsconf *vfsp;
1039 struct nameidata nd;
1045 * Be ultra-paranoid about making sure the type and fspath
1046 * variables will fit in our mp buffers, including the
1049 if (strlen(fstype) >= MFSNAMELEN || strlen(fspath) >= MNAMELEN)
1050 return (ENAMETOOLONG);
1052 if (jailed(td->td_ucred) || usermount == 0) {
1053 if ((error = priv_check(td, PRIV_VFS_MOUNT)) != 0)
1058 * Do not allow NFS export or MNT_SUIDDIR by unprivileged users.
1060 if (fsflags & MNT_EXPORTED) {
1061 error = priv_check(td, PRIV_VFS_MOUNT_EXPORTED);
1065 if (fsflags & MNT_SUIDDIR) {
1066 error = priv_check(td, PRIV_VFS_MOUNT_SUIDDIR);
1071 * Silently enforce MNT_NOSUID and MNT_USER for unprivileged users.
1073 if ((fsflags & (MNT_NOSUID | MNT_USER)) != (MNT_NOSUID | MNT_USER)) {
1074 if (priv_check(td, PRIV_VFS_MOUNT_NONUSER) != 0)
1075 fsflags |= MNT_NOSUID | MNT_USER;
1078 /* Load KLDs before we lock the covered vnode to avoid reversals. */
1080 if ((fsflags & MNT_UPDATE) == 0) {
1081 /* Don't try to load KLDs if we're mounting the root. */
1082 if (fsflags & MNT_ROOTFS)
1083 vfsp = vfs_byname(fstype);
1085 vfsp = vfs_byname_kld(fstype, td, &error);
1088 if (jailed(td->td_ucred) && !(vfsp->vfc_flags & VFCF_JAIL))
1093 * Get vnode to be covered or mount point's vnode in case of MNT_UPDATE.
1095 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1,
1096 UIO_SYSSPACE, fspath, td);
1100 NDFREE(&nd, NDF_ONLY_PNBUF);
1102 if ((fsflags & MNT_UPDATE) == 0) {
1103 pathbuf = malloc(MNAMELEN, M_TEMP, M_WAITOK);
1104 strcpy(pathbuf, fspath);
1105 error = vn_path_to_global_path(td, vp, pathbuf, MNAMELEN);
1106 /* debug.disablefullpath == 1 results in ENODEV */
1107 if (error == 0 || error == ENODEV) {
1108 error = vfs_domount_first(td, vfsp, pathbuf, vp,
1111 free(pathbuf, M_TEMP);
1113 error = vfs_domount_update(td, vp, fsflags, optlist);
1119 * Unmount a filesystem.
1121 * Note: unmount takes a path to the vnode mounted on as argument, not
1122 * special file (as before).
1124 #ifndef _SYS_SYSPROTO_H_
1125 struct unmount_args {
1132 sys_unmount(struct thread *td, struct unmount_args *uap)
1134 struct nameidata nd;
1137 int error, id0, id1;
1139 AUDIT_ARG_VALUE(uap->flags);
1140 if (jailed(td->td_ucred) || usermount == 0) {
1141 error = priv_check(td, PRIV_VFS_UNMOUNT);
1146 pathbuf = malloc(MNAMELEN, M_TEMP, M_WAITOK);
1147 error = copyinstr(uap->path, pathbuf, MNAMELEN, NULL);
1149 free(pathbuf, M_TEMP);
1152 if (uap->flags & MNT_BYFSID) {
1153 AUDIT_ARG_TEXT(pathbuf);
1154 /* Decode the filesystem ID. */
1155 if (sscanf(pathbuf, "FSID:%d:%d", &id0, &id1) != 2) {
1156 free(pathbuf, M_TEMP);
1160 mtx_lock(&mountlist_mtx);
1161 TAILQ_FOREACH_REVERSE(mp, &mountlist, mntlist, mnt_list) {
1162 if (mp->mnt_stat.f_fsid.val[0] == id0 &&
1163 mp->mnt_stat.f_fsid.val[1] == id1) {
1168 mtx_unlock(&mountlist_mtx);
1171 * Try to find global path for path argument.
1173 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1,
1174 UIO_SYSSPACE, pathbuf, td);
1175 if (namei(&nd) == 0) {
1176 NDFREE(&nd, NDF_ONLY_PNBUF);
1177 error = vn_path_to_global_path(td, nd.ni_vp, pathbuf,
1179 if (error == 0 || error == ENODEV)
1182 mtx_lock(&mountlist_mtx);
1183 TAILQ_FOREACH_REVERSE(mp, &mountlist, mntlist, mnt_list) {
1184 if (strcmp(mp->mnt_stat.f_mntonname, pathbuf) == 0) {
1189 mtx_unlock(&mountlist_mtx);
1191 free(pathbuf, M_TEMP);
1194 * Previously we returned ENOENT for a nonexistent path and
1195 * EINVAL for a non-mountpoint. We cannot tell these apart
1196 * now, so in the !MNT_BYFSID case return the more likely
1197 * EINVAL for compatibility.
1199 return ((uap->flags & MNT_BYFSID) ? ENOENT : EINVAL);
1203 * Don't allow unmounting the root filesystem.
1205 if (mp->mnt_flag & MNT_ROOTFS) {
1209 error = dounmount(mp, uap->flags, td);
1214 * Return error if any of the vnodes, ignoring the root vnode
1215 * and the syncer vnode, have non-zero usecount.
1217 * This function is purely advisory - it can return false positives
1221 vfs_check_usecounts(struct mount *mp)
1223 struct vnode *vp, *mvp;
1225 MNT_VNODE_FOREACH_ALL(vp, mp, mvp) {
1226 if ((vp->v_vflag & VV_ROOT) == 0 && vp->v_type != VNON &&
1227 vp->v_usecount != 0) {
1229 MNT_VNODE_FOREACH_ALL_ABORT(mp, mvp);
1239 dounmount_cleanup(struct mount *mp, struct vnode *coveredvp, int mntkflags)
1242 mtx_assert(MNT_MTX(mp), MA_OWNED);
1243 mp->mnt_kern_flag &= ~mntkflags;
1244 if ((mp->mnt_kern_flag & MNTK_MWAIT) != 0) {
1245 mp->mnt_kern_flag &= ~MNTK_MWAIT;
1249 if (coveredvp != NULL) {
1250 VOP_UNLOCK(coveredvp, 0);
1253 vn_finished_write(mp);
1257 * Do the actual filesystem unmount.
1260 dounmount(struct mount *mp, int flags, struct thread *td)
1262 struct vnode *coveredvp, *fsrootvp;
1264 uint64_t async_flag;
1267 if ((coveredvp = mp->mnt_vnodecovered) != NULL) {
1268 mnt_gen_r = mp->mnt_gen;
1271 vn_lock(coveredvp, LK_EXCLUSIVE | LK_INTERLOCK | LK_RETRY);
1273 * Check for mp being unmounted while waiting for the
1274 * covered vnode lock.
1276 if (coveredvp->v_mountedhere != mp ||
1277 coveredvp->v_mountedhere->mnt_gen != mnt_gen_r) {
1278 VOP_UNLOCK(coveredvp, 0);
1286 * Only privileged root, or (if MNT_USER is set) the user that did the
1287 * original mount is permitted to unmount this filesystem.
1289 error = vfs_suser(mp, td);
1291 if (coveredvp != NULL) {
1292 VOP_UNLOCK(coveredvp, 0);
1299 vn_start_write(NULL, &mp, V_WAIT | V_MNTREF);
1301 if ((mp->mnt_kern_flag & MNTK_UNMOUNT) != 0 ||
1302 (mp->mnt_flag & MNT_UPDATE) != 0 ||
1303 !TAILQ_EMPTY(&mp->mnt_uppers)) {
1304 dounmount_cleanup(mp, coveredvp, 0);
1307 mp->mnt_kern_flag |= MNTK_UNMOUNT | MNTK_NOINSMNTQ;
1308 if (flags & MNT_NONBUSY) {
1310 error = vfs_check_usecounts(mp);
1313 dounmount_cleanup(mp, coveredvp, MNTK_UNMOUNT |
1318 /* Allow filesystems to detect that a forced unmount is in progress. */
1319 if (flags & MNT_FORCE) {
1320 mp->mnt_kern_flag |= MNTK_UNMOUNTF;
1323 * Must be done after setting MNTK_UNMOUNTF and before
1324 * waiting for mnt_lockref to become 0.
1330 if (mp->mnt_lockref) {
1331 mp->mnt_kern_flag |= MNTK_DRAINING;
1332 error = msleep(&mp->mnt_lockref, MNT_MTX(mp), PVFS,
1336 KASSERT(mp->mnt_lockref == 0,
1337 ("%s: invalid lock refcount in the drain path @ %s:%d",
1338 __func__, __FILE__, __LINE__));
1340 ("%s: invalid return value for msleep in the drain path @ %s:%d",
1341 __func__, __FILE__, __LINE__));
1343 if (mp->mnt_flag & MNT_EXPUBLIC)
1344 vfs_setpublicfs(NULL, NULL, NULL);
1347 * From now, we can claim that the use reference on the
1348 * coveredvp is ours, and the ref can be released only by
1349 * successfull unmount by us, or left for later unmount
1350 * attempt. The previously acquired hold reference is no
1351 * longer needed to protect the vnode from reuse.
1353 if (coveredvp != NULL)
1356 vfs_msync(mp, MNT_WAIT);
1358 async_flag = mp->mnt_flag & MNT_ASYNC;
1359 mp->mnt_flag &= ~MNT_ASYNC;
1360 mp->mnt_kern_flag &= ~MNTK_ASYNC;
1362 cache_purgevfs(mp, false); /* remove cache entries for this file sys */
1363 vfs_deallocate_syncvnode(mp);
1365 * For forced unmounts, move process cdir/rdir refs on the fs root
1366 * vnode to the covered vnode. For non-forced unmounts we want
1367 * such references to cause an EBUSY error.
1369 if ((flags & MNT_FORCE) &&
1370 VFS_ROOT(mp, LK_EXCLUSIVE, &fsrootvp) == 0) {
1371 if (mp->mnt_vnodecovered != NULL &&
1372 (mp->mnt_flag & MNT_IGNORE) == 0)
1373 mountcheckdirs(fsrootvp, mp->mnt_vnodecovered);
1374 if (fsrootvp == rootvnode) {
1380 if ((mp->mnt_flag & MNT_RDONLY) != 0 || (flags & MNT_FORCE) != 0 ||
1381 (error = VFS_SYNC(mp, MNT_WAIT)) == 0)
1382 error = VFS_UNMOUNT(mp, flags);
1383 vn_finished_write(mp);
1385 * If we failed to flush the dirty blocks for this mount point,
1386 * undo all the cdir/rdir and rootvnode changes we made above.
1387 * Unless we failed to do so because the device is reporting that
1388 * it doesn't exist anymore.
1390 if (error && error != ENXIO) {
1391 if ((flags & MNT_FORCE) &&
1392 VFS_ROOT(mp, LK_EXCLUSIVE, &fsrootvp) == 0) {
1393 if (mp->mnt_vnodecovered != NULL &&
1394 (mp->mnt_flag & MNT_IGNORE) == 0)
1395 mountcheckdirs(mp->mnt_vnodecovered, fsrootvp);
1396 if (rootvnode == NULL) {
1397 rootvnode = fsrootvp;
1403 mp->mnt_kern_flag &= ~MNTK_NOINSMNTQ;
1404 if ((mp->mnt_flag & MNT_RDONLY) == 0) {
1406 vfs_allocate_syncvnode(mp);
1409 mp->mnt_kern_flag &= ~(MNTK_UNMOUNT | MNTK_UNMOUNTF);
1410 mp->mnt_flag |= async_flag;
1411 if ((mp->mnt_flag & MNT_ASYNC) != 0 &&
1412 (mp->mnt_kern_flag & MNTK_NOASYNC) == 0)
1413 mp->mnt_kern_flag |= MNTK_ASYNC;
1414 if (mp->mnt_kern_flag & MNTK_MWAIT) {
1415 mp->mnt_kern_flag &= ~MNTK_MWAIT;
1420 VOP_UNLOCK(coveredvp, 0);
1423 mtx_lock(&mountlist_mtx);
1424 TAILQ_REMOVE(&mountlist, mp, mnt_list);
1425 mtx_unlock(&mountlist_mtx);
1426 EVENTHANDLER_INVOKE(vfs_unmounted, mp, td);
1427 if (coveredvp != NULL) {
1428 coveredvp->v_mountedhere = NULL;
1431 vfs_event_signal(NULL, VQ_UNMOUNT, 0);
1432 if (mp == rootdevmp)
1434 vfs_mount_destroy(mp);
1439 * Report errors during filesystem mounting.
1442 vfs_mount_error(struct mount *mp, const char *fmt, ...)
1444 struct vfsoptlist *moptlist = mp->mnt_optnew;
1449 error = vfs_getopt(moptlist, "errmsg", (void **)&errmsg, &len);
1450 if (error || errmsg == NULL || len <= 0)
1454 vsnprintf(errmsg, (size_t)len, fmt, ap);
1459 vfs_opterror(struct vfsoptlist *opts, const char *fmt, ...)
1465 error = vfs_getopt(opts, "errmsg", (void **)&errmsg, &len);
1466 if (error || errmsg == NULL || len <= 0)
1470 vsnprintf(errmsg, (size_t)len, fmt, ap);
1475 * ---------------------------------------------------------------------
1476 * Functions for querying mount options/arguments from filesystems.
1480 * Check that no unknown options are given
1483 vfs_filteropt(struct vfsoptlist *opts, const char **legal)
1487 const char **t, *p, *q;
1490 TAILQ_FOREACH(opt, opts, link) {
1493 if (p[0] == 'n' && p[1] == 'o')
1495 for(t = global_opts; *t != NULL; t++) {
1496 if (strcmp(*t, p) == 0)
1499 if (strcmp(*t, q) == 0)
1505 for(t = legal; *t != NULL; t++) {
1506 if (strcmp(*t, p) == 0)
1509 if (strcmp(*t, q) == 0)
1515 snprintf(errmsg, sizeof(errmsg),
1516 "mount option <%s> is unknown", p);
1520 TAILQ_FOREACH(opt, opts, link) {
1521 if (strcmp(opt->name, "errmsg") == 0) {
1522 strncpy((char *)opt->value, errmsg, opt->len);
1527 printf("%s\n", errmsg);
1533 * Get a mount option by its name.
1535 * Return 0 if the option was found, ENOENT otherwise.
1536 * If len is non-NULL it will be filled with the length
1537 * of the option. If buf is non-NULL, it will be filled
1538 * with the address of the option.
1541 vfs_getopt(struct vfsoptlist *opts, const char *name, void **buf, int *len)
1545 KASSERT(opts != NULL, ("vfs_getopt: caller passed 'opts' as NULL"));
1547 TAILQ_FOREACH(opt, opts, link) {
1548 if (strcmp(name, opt->name) == 0) {
1561 vfs_getopt_pos(struct vfsoptlist *opts, const char *name)
1568 TAILQ_FOREACH(opt, opts, link) {
1569 if (strcmp(name, opt->name) == 0) {
1578 vfs_getopt_size(struct vfsoptlist *opts, const char *name, off_t *value)
1580 char *opt_value, *vtp;
1584 error = vfs_getopt(opts, name, (void **)&opt_value, &opt_len);
1587 if (opt_len == 0 || opt_value == NULL)
1589 if (opt_value[0] == '\0' || opt_value[opt_len - 1] != '\0')
1591 iv = strtoq(opt_value, &vtp, 0);
1592 if (vtp == opt_value || (vtp[0] != '\0' && vtp[1] != '\0'))
1620 vfs_getopts(struct vfsoptlist *opts, const char *name, int *error)
1625 TAILQ_FOREACH(opt, opts, link) {
1626 if (strcmp(name, opt->name) != 0)
1629 if (opt->len == 0 ||
1630 ((char *)opt->value)[opt->len - 1] != '\0') {
1634 return (opt->value);
1641 vfs_flagopt(struct vfsoptlist *opts, const char *name, uint64_t *w,
1646 TAILQ_FOREACH(opt, opts, link) {
1647 if (strcmp(name, opt->name) == 0) {
1660 vfs_scanopt(struct vfsoptlist *opts, const char *name, const char *fmt, ...)
1666 KASSERT(opts != NULL, ("vfs_getopt: caller passed 'opts' as NULL"));
1668 TAILQ_FOREACH(opt, opts, link) {
1669 if (strcmp(name, opt->name) != 0)
1672 if (opt->len == 0 || opt->value == NULL)
1674 if (((char *)opt->value)[opt->len - 1] != '\0')
1677 ret = vsscanf(opt->value, fmt, ap);
1685 vfs_setopt(struct vfsoptlist *opts, const char *name, void *value, int len)
1689 TAILQ_FOREACH(opt, opts, link) {
1690 if (strcmp(name, opt->name) != 0)
1693 if (opt->value == NULL)
1696 if (opt->len != len)
1698 bcopy(value, opt->value, len);
1706 vfs_setopt_part(struct vfsoptlist *opts, const char *name, void *value, int len)
1710 TAILQ_FOREACH(opt, opts, link) {
1711 if (strcmp(name, opt->name) != 0)
1714 if (opt->value == NULL)
1720 bcopy(value, opt->value, len);
1728 vfs_setopts(struct vfsoptlist *opts, const char *name, const char *value)
1732 TAILQ_FOREACH(opt, opts, link) {
1733 if (strcmp(name, opt->name) != 0)
1736 if (opt->value == NULL)
1737 opt->len = strlen(value) + 1;
1738 else if (strlcpy(opt->value, value, opt->len) >= opt->len)
1746 * Find and copy a mount option.
1748 * The size of the buffer has to be specified
1749 * in len, if it is not the same length as the
1750 * mount option, EINVAL is returned.
1751 * Returns ENOENT if the option is not found.
1754 vfs_copyopt(struct vfsoptlist *opts, const char *name, void *dest, int len)
1758 KASSERT(opts != NULL, ("vfs_copyopt: caller passed 'opts' as NULL"));
1760 TAILQ_FOREACH(opt, opts, link) {
1761 if (strcmp(name, opt->name) == 0) {
1763 if (len != opt->len)
1765 bcopy(opt->value, dest, opt->len);
1773 __vfs_statfs(struct mount *mp, struct statfs *sbp)
1777 error = mp->mnt_op->vfs_statfs(mp, &mp->mnt_stat);
1778 if (sbp != &mp->mnt_stat)
1779 *sbp = mp->mnt_stat;
1784 vfs_mountedfrom(struct mount *mp, const char *from)
1787 bzero(mp->mnt_stat.f_mntfromname, sizeof mp->mnt_stat.f_mntfromname);
1788 strlcpy(mp->mnt_stat.f_mntfromname, from,
1789 sizeof mp->mnt_stat.f_mntfromname);
1793 * ---------------------------------------------------------------------
1794 * This is the api for building mount args and mounting filesystems from
1795 * inside the kernel.
1797 * The API works by accumulation of individual args. First error is
1800 * XXX: should be documented in new manpage kernel_mount(9)
1803 /* A memory allocation which must be freed when we are done */
1805 SLIST_ENTRY(mntaarg) next;
1808 /* The header for the mount arguments */
1813 SLIST_HEAD(, mntaarg) list;
1817 * Add a boolean argument.
1819 * flag is the boolean value.
1820 * name must start with "no".
1823 mount_argb(struct mntarg *ma, int flag, const char *name)
1826 KASSERT(name[0] == 'n' && name[1] == 'o',
1827 ("mount_argb(...,%s): name must start with 'no'", name));
1829 return (mount_arg(ma, name + (flag ? 2 : 0), NULL, 0));
1833 * Add an argument printf style
1836 mount_argf(struct mntarg *ma, const char *name, const char *fmt, ...)
1839 struct mntaarg *maa;
1844 ma = malloc(sizeof *ma, M_MOUNT, M_WAITOK | M_ZERO);
1845 SLIST_INIT(&ma->list);
1850 ma->v = realloc(ma->v, sizeof *ma->v * (ma->len + 2),
1852 ma->v[ma->len].iov_base = (void *)(uintptr_t)name;
1853 ma->v[ma->len].iov_len = strlen(name) + 1;
1856 sb = sbuf_new_auto();
1858 sbuf_vprintf(sb, fmt, ap);
1861 len = sbuf_len(sb) + 1;
1862 maa = malloc(sizeof *maa + len, M_MOUNT, M_WAITOK | M_ZERO);
1863 SLIST_INSERT_HEAD(&ma->list, maa, next);
1864 bcopy(sbuf_data(sb), maa + 1, len);
1867 ma->v[ma->len].iov_base = maa + 1;
1868 ma->v[ma->len].iov_len = len;
1875 * Add an argument which is a userland string.
1878 mount_argsu(struct mntarg *ma, const char *name, const void *val, int len)
1880 struct mntaarg *maa;
1886 ma = malloc(sizeof *ma, M_MOUNT, M_WAITOK | M_ZERO);
1887 SLIST_INIT(&ma->list);
1891 maa = malloc(sizeof *maa + len, M_MOUNT, M_WAITOK | M_ZERO);
1892 SLIST_INSERT_HEAD(&ma->list, maa, next);
1893 tbuf = (void *)(maa + 1);
1894 ma->error = copyinstr(val, tbuf, len, NULL);
1895 return (mount_arg(ma, name, tbuf, -1));
1901 * If length is -1, treat value as a C string.
1904 mount_arg(struct mntarg *ma, const char *name, const void *val, int len)
1908 ma = malloc(sizeof *ma, M_MOUNT, M_WAITOK | M_ZERO);
1909 SLIST_INIT(&ma->list);
1914 ma->v = realloc(ma->v, sizeof *ma->v * (ma->len + 2),
1916 ma->v[ma->len].iov_base = (void *)(uintptr_t)name;
1917 ma->v[ma->len].iov_len = strlen(name) + 1;
1920 ma->v[ma->len].iov_base = (void *)(uintptr_t)val;
1922 ma->v[ma->len].iov_len = strlen(val) + 1;
1924 ma->v[ma->len].iov_len = len;
1930 * Free a mntarg structure
1933 free_mntarg(struct mntarg *ma)
1935 struct mntaarg *maa;
1937 while (!SLIST_EMPTY(&ma->list)) {
1938 maa = SLIST_FIRST(&ma->list);
1939 SLIST_REMOVE_HEAD(&ma->list, next);
1942 free(ma->v, M_MOUNT);
1947 * Mount a filesystem
1950 kernel_mount(struct mntarg *ma, uint64_t flags)
1955 KASSERT(ma != NULL, ("kernel_mount NULL ma"));
1956 KASSERT(ma->v != NULL, ("kernel_mount NULL ma->v"));
1957 KASSERT(!(ma->len & 1), ("kernel_mount odd ma->len (%d)", ma->len));
1959 auio.uio_iov = ma->v;
1960 auio.uio_iovcnt = ma->len;
1961 auio.uio_segflg = UIO_SYSSPACE;
1965 error = vfs_donmount(curthread, flags, &auio);
1971 * A printflike function to mount a filesystem.
1974 kernel_vmount(int flags, ...)
1976 struct mntarg *ma = NULL;
1982 va_start(ap, flags);
1984 cp = va_arg(ap, const char *);
1987 vp = va_arg(ap, const void *);
1988 ma = mount_arg(ma, cp, vp, (vp != NULL ? -1 : 0));
1992 error = kernel_mount(ma, flags);
1997 vfs_oexport_conv(const struct oexport_args *oexp, struct export_args *exp)
2000 bcopy(oexp, exp, sizeof(*oexp));
2001 exp->ex_numsecflavors = 0;