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 occurence 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 mp->mnt_op = vfsp->vfc_vfsops;
468 vfsp->vfc_refcount++; /* XXX Unlocked */
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 mp->mnt_vfc->vfc_refcount--;
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");
525 mac_mount_destroy(mp);
527 if (mp->mnt_opt != NULL)
528 vfs_freeopts(mp->mnt_opt);
529 crfree(mp->mnt_cred);
530 uma_zfree(mount_zone, mp);
534 vfs_donmount(struct thread *td, uint64_t fsflags, struct uio *fsoptions)
536 struct vfsoptlist *optlist;
537 struct vfsopt *opt, *tmp_opt;
538 char *fstype, *fspath, *errmsg;
539 int error, fstypelen, fspathlen, errmsg_len, errmsg_pos;
541 errmsg = fspath = NULL;
542 errmsg_len = fspathlen = 0;
545 error = vfs_buildopts(fsoptions, &optlist);
549 if (vfs_getopt(optlist, "errmsg", (void **)&errmsg, &errmsg_len) == 0)
550 errmsg_pos = vfs_getopt_pos(optlist, "errmsg");
553 * We need these two options before the others,
554 * and they are mandatory for any filesystem.
555 * Ensure they are NUL terminated as well.
558 error = vfs_getopt(optlist, "fstype", (void **)&fstype, &fstypelen);
559 if (error || fstype[fstypelen - 1] != '\0') {
562 strncpy(errmsg, "Invalid fstype", errmsg_len);
566 error = vfs_getopt(optlist, "fspath", (void **)&fspath, &fspathlen);
567 if (error || fspath[fspathlen - 1] != '\0') {
570 strncpy(errmsg, "Invalid fspath", errmsg_len);
575 * We need to see if we have the "update" option
576 * before we call vfs_domount(), since vfs_domount() has special
577 * logic based on MNT_UPDATE. This is very important
578 * when we want to update the root filesystem.
580 TAILQ_FOREACH_SAFE(opt, optlist, link, tmp_opt) {
581 if (strcmp(opt->name, "update") == 0) {
582 fsflags |= MNT_UPDATE;
583 vfs_freeopt(optlist, opt);
585 else if (strcmp(opt->name, "async") == 0)
586 fsflags |= MNT_ASYNC;
587 else if (strcmp(opt->name, "force") == 0) {
588 fsflags |= MNT_FORCE;
589 vfs_freeopt(optlist, opt);
591 else if (strcmp(opt->name, "reload") == 0) {
592 fsflags |= MNT_RELOAD;
593 vfs_freeopt(optlist, opt);
595 else if (strcmp(opt->name, "multilabel") == 0)
596 fsflags |= MNT_MULTILABEL;
597 else if (strcmp(opt->name, "noasync") == 0)
598 fsflags &= ~MNT_ASYNC;
599 else if (strcmp(opt->name, "noatime") == 0)
600 fsflags |= MNT_NOATIME;
601 else if (strcmp(opt->name, "atime") == 0) {
602 free(opt->name, M_MOUNT);
603 opt->name = strdup("nonoatime", M_MOUNT);
605 else if (strcmp(opt->name, "noclusterr") == 0)
606 fsflags |= MNT_NOCLUSTERR;
607 else if (strcmp(opt->name, "clusterr") == 0) {
608 free(opt->name, M_MOUNT);
609 opt->name = strdup("nonoclusterr", M_MOUNT);
611 else if (strcmp(opt->name, "noclusterw") == 0)
612 fsflags |= MNT_NOCLUSTERW;
613 else if (strcmp(opt->name, "clusterw") == 0) {
614 free(opt->name, M_MOUNT);
615 opt->name = strdup("nonoclusterw", M_MOUNT);
617 else if (strcmp(opt->name, "noexec") == 0)
618 fsflags |= MNT_NOEXEC;
619 else if (strcmp(opt->name, "exec") == 0) {
620 free(opt->name, M_MOUNT);
621 opt->name = strdup("nonoexec", M_MOUNT);
623 else if (strcmp(opt->name, "nosuid") == 0)
624 fsflags |= MNT_NOSUID;
625 else if (strcmp(opt->name, "suid") == 0) {
626 free(opt->name, M_MOUNT);
627 opt->name = strdup("nonosuid", M_MOUNT);
629 else if (strcmp(opt->name, "nosymfollow") == 0)
630 fsflags |= MNT_NOSYMFOLLOW;
631 else if (strcmp(opt->name, "symfollow") == 0) {
632 free(opt->name, M_MOUNT);
633 opt->name = strdup("nonosymfollow", M_MOUNT);
635 else if (strcmp(opt->name, "noro") == 0)
636 fsflags &= ~MNT_RDONLY;
637 else if (strcmp(opt->name, "rw") == 0)
638 fsflags &= ~MNT_RDONLY;
639 else if (strcmp(opt->name, "ro") == 0)
640 fsflags |= MNT_RDONLY;
641 else if (strcmp(opt->name, "rdonly") == 0) {
642 free(opt->name, M_MOUNT);
643 opt->name = strdup("ro", M_MOUNT);
644 fsflags |= MNT_RDONLY;
646 else if (strcmp(opt->name, "suiddir") == 0)
647 fsflags |= MNT_SUIDDIR;
648 else if (strcmp(opt->name, "sync") == 0)
649 fsflags |= MNT_SYNCHRONOUS;
650 else if (strcmp(opt->name, "union") == 0)
651 fsflags |= MNT_UNION;
655 * Be ultra-paranoid about making sure the type and fspath
656 * variables will fit in our mp buffers, including the
659 if (fstypelen >= MFSNAMELEN - 1 || fspathlen >= MNAMELEN - 1) {
660 error = ENAMETOOLONG;
664 error = vfs_domount(td, fstype, fspath, fsflags, &optlist);
666 /* copyout the errmsg */
667 if (errmsg_pos != -1 && ((2 * errmsg_pos + 1) < fsoptions->uio_iovcnt)
668 && errmsg_len > 0 && errmsg != NULL) {
669 if (fsoptions->uio_segflg == UIO_SYSSPACE) {
671 fsoptions->uio_iov[2 * errmsg_pos + 1].iov_base,
672 fsoptions->uio_iov[2 * errmsg_pos + 1].iov_len);
675 fsoptions->uio_iov[2 * errmsg_pos + 1].iov_base,
676 fsoptions->uio_iov[2 * errmsg_pos + 1].iov_len);
681 vfs_freeopts(optlist);
688 #ifndef _SYS_SYSPROTO_H_
700 struct mount_args /* {
708 struct vfsconf *vfsp = NULL;
709 struct mntarg *ma = NULL;
714 * Mount flags are now 64-bits. On 32-bit architectures only
715 * 32-bits are passed in, but from here on everything handles
716 * 64-bit flags correctly.
720 AUDIT_ARG_FFLAGS(flags);
723 * Filter out MNT_ROOTFS. We do not want clients of mount() in
724 * userspace to set this flag, but we must filter it out if we want
725 * MNT_UPDATE on the root file system to work.
726 * MNT_ROOTFS should only be set by the kernel when mounting its
729 flags &= ~MNT_ROOTFS;
731 fstype = malloc(MFSNAMELEN, M_TEMP, M_WAITOK);
732 error = copyinstr(uap->type, fstype, MFSNAMELEN, NULL);
734 free(fstype, M_TEMP);
738 AUDIT_ARG_TEXT(fstype);
740 vfsp = vfs_byname_kld(fstype, td, &error);
741 free(fstype, M_TEMP);
746 if (vfsp->vfc_vfsops->vfs_cmount == NULL) {
751 ma = mount_argsu(ma, "fstype", uap->type, MNAMELEN);
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);
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 mtx_assert(&Giant, MA_OWNED);
781 ASSERT_VOP_ELOCKED(vp, __func__);
782 KASSERT((fsflags & MNT_UPDATE) == 0, ("MNT_UPDATE shouldn't be here"));
785 * If the user is not root, ensure that they own the directory
786 * onto which we are attempting to mount.
788 error = VOP_GETATTR(vp, &va, td->td_ucred);
789 if (error == 0 && va.va_uid != td->td_ucred->cr_uid)
790 error = priv_check_cred(td->td_ucred, PRIV_VFS_ADMIN, 0);
792 error = vinvalbuf(vp, V_SAVE, 0, 0);
793 if (error == 0 && vp->v_type != VDIR)
797 if ((vp->v_iflag & VI_MOUNT) == 0 && vp->v_mountedhere == NULL)
798 vp->v_iflag |= VI_MOUNT;
809 /* Allocate and initialize the filesystem. */
810 mp = vfs_mount_alloc(vp, vfsp, fspath, td->td_ucred);
811 /* XXXMAC: pass to vfs_mount_alloc? */
812 mp->mnt_optnew = *optlist;
813 /* Set the mount level flags. */
814 mp->mnt_flag = (fsflags & (MNT_UPDATEMASK | MNT_ROOTFS | MNT_RDONLY));
817 * Mount the filesystem.
818 * XXX The final recipients of VFS_MOUNT just overwrite the ndp they
819 * get. No freeing of cn_pnbuf.
821 error = VFS_MOUNT(mp);
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");
864 VOP_UNLOCK(newdp, 0);
866 mountcheckdirs(vp, newdp);
868 if ((mp->mnt_flag & MNT_RDONLY) == 0)
869 vfs_allocate_syncvnode(mp);
875 * vfs_domount_update(): update of mounted file system
879 struct thread *td, /* Calling thread. */
880 struct vnode *vp, /* Mount point vnode. */
881 uint64_t fsflags, /* Flags common to all filesystems. */
882 struct vfsoptlist **optlist /* Options local to the filesystem. */
885 struct oexport_args oexport;
886 struct export_args export;
888 int error, export_error;
891 mtx_assert(&Giant, MA_OWNED);
892 ASSERT_VOP_ELOCKED(vp, __func__);
893 KASSERT((fsflags & MNT_UPDATE) != 0, ("MNT_UPDATE should be here"));
895 if ((vp->v_vflag & VV_ROOT) == 0) {
901 * We only allow the filesystem to be reloaded if it
902 * is currently mounted read-only.
905 if ((fsflags & MNT_RELOAD) != 0 && (flag & MNT_RDONLY) == 0) {
907 return (EOPNOTSUPP); /* Needs translation */
910 * Only privileged root, or (if MNT_USER is set) the user that
911 * did the original mount is permitted to update it.
913 error = vfs_suser(mp, td);
918 if (vfs_busy(mp, MBF_NOWAIT)) {
923 if ((vp->v_iflag & VI_MOUNT) != 0 || vp->v_mountedhere != NULL) {
929 vp->v_iflag |= VI_MOUNT;
934 mp->mnt_flag &= ~MNT_UPDATEMASK;
935 mp->mnt_flag |= fsflags & (MNT_RELOAD | MNT_FORCE | MNT_UPDATE |
936 MNT_SNAPSHOT | MNT_ROOTFS | MNT_UPDATEMASK | MNT_RDONLY);
937 if ((mp->mnt_flag & MNT_ASYNC) == 0)
938 mp->mnt_kern_flag &= ~MNTK_ASYNC;
940 mp->mnt_optnew = *optlist;
941 vfs_mergeopts(mp->mnt_optnew, mp->mnt_opt);
944 * Mount the filesystem.
945 * XXX The final recipients of VFS_MOUNT just overwrite the ndp they
946 * get. No freeing of cn_pnbuf.
948 error = VFS_MOUNT(mp);
952 /* Process the export option. */
953 if (vfs_copyopt(mp->mnt_optnew, "export", &export,
954 sizeof(export)) == 0) {
955 export_error = vfs_export(mp, &export);
956 } else if (vfs_copyopt(mp->mnt_optnew, "export", &oexport,
957 sizeof(oexport)) == 0) {
958 export.ex_flags = oexport.ex_flags;
959 export.ex_root = oexport.ex_root;
960 export.ex_anon = oexport.ex_anon;
961 export.ex_addr = oexport.ex_addr;
962 export.ex_addrlen = oexport.ex_addrlen;
963 export.ex_mask = oexport.ex_mask;
964 export.ex_masklen = oexport.ex_masklen;
965 export.ex_indexfile = oexport.ex_indexfile;
966 export.ex_numsecflavors = 0;
967 export_error = vfs_export(mp, &export);
973 mp->mnt_flag &= ~(MNT_UPDATE | MNT_RELOAD | MNT_FORCE |
977 * If we fail, restore old mount flags. MNT_QUOTA is special,
978 * because it is not part of MNT_UPDATEMASK, but it could have
979 * changed in the meantime if quotactl(2) was called.
980 * All in all we want current value of MNT_QUOTA, not the old
983 mp->mnt_flag = (mp->mnt_flag & MNT_QUOTA) | (flag & ~MNT_QUOTA);
985 if ((mp->mnt_flag & MNT_ASYNC) != 0 &&
986 (mp->mnt_kern_flag & MNTK_NOASYNC) == 0)
987 mp->mnt_kern_flag |= MNTK_ASYNC;
989 mp->mnt_kern_flag &= ~MNTK_ASYNC;
995 if (mp->mnt_opt != NULL)
996 vfs_freeopts(mp->mnt_opt);
997 mp->mnt_opt = mp->mnt_optnew;
999 (void)VFS_STATFS(mp, &mp->mnt_stat);
1001 * Prevent external consumers of mount options from reading
1004 mp->mnt_optnew = NULL;
1006 if ((mp->mnt_flag & MNT_RDONLY) == 0)
1007 vfs_allocate_syncvnode(mp);
1009 vfs_deallocate_syncvnode(mp);
1013 vp->v_iflag &= ~VI_MOUNT;
1016 return (error != 0 ? error : export_error);
1020 * vfs_domount(): actually attempt a filesystem mount.
1024 struct thread *td, /* Calling thread. */
1025 const char *fstype, /* Filesystem type. */
1026 char *fspath, /* Mount path. */
1027 uint64_t fsflags, /* Flags common to all filesystems. */
1028 struct vfsoptlist **optlist /* Options local to the filesystem. */
1031 struct vfsconf *vfsp;
1032 struct nameidata nd;
1038 * Be ultra-paranoid about making sure the type and fspath
1039 * variables will fit in our mp buffers, including the
1042 if (strlen(fstype) >= MFSNAMELEN || strlen(fspath) >= MNAMELEN)
1043 return (ENAMETOOLONG);
1045 if (jailed(td->td_ucred) || usermount == 0) {
1046 if ((error = priv_check(td, PRIV_VFS_MOUNT)) != 0)
1051 * Do not allow NFS export or MNT_SUIDDIR by unprivileged users.
1053 if (fsflags & MNT_EXPORTED) {
1054 error = priv_check(td, PRIV_VFS_MOUNT_EXPORTED);
1058 if (fsflags & MNT_SUIDDIR) {
1059 error = priv_check(td, PRIV_VFS_MOUNT_SUIDDIR);
1064 * Silently enforce MNT_NOSUID and MNT_USER for unprivileged users.
1066 if ((fsflags & (MNT_NOSUID | MNT_USER)) != (MNT_NOSUID | MNT_USER)) {
1067 if (priv_check(td, PRIV_VFS_MOUNT_NONUSER) != 0)
1068 fsflags |= MNT_NOSUID | MNT_USER;
1071 /* Load KLDs before we lock the covered vnode to avoid reversals. */
1073 if ((fsflags & MNT_UPDATE) == 0) {
1074 /* Don't try to load KLDs if we're mounting the root. */
1075 if (fsflags & MNT_ROOTFS)
1076 vfsp = vfs_byname(fstype);
1078 vfsp = vfs_byname_kld(fstype, td, &error);
1081 if (jailed(td->td_ucred) && !(vfsp->vfc_flags & VFCF_JAIL))
1086 * Get vnode to be covered or mount point's vnode in case of MNT_UPDATE.
1088 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1,
1089 UIO_SYSSPACE, fspath, td);
1094 NDFREE(&nd, NDF_ONLY_PNBUF);
1096 if ((fsflags & MNT_UPDATE) == 0) {
1097 pathbuf = malloc(MNAMELEN, M_TEMP, M_WAITOK);
1098 strcpy(pathbuf, fspath);
1099 error = vn_path_to_global_path(td, vp, pathbuf, MNAMELEN);
1100 /* debug.disablefullpath == 1 results in ENODEV */
1101 if (error == 0 || error == ENODEV) {
1102 error = vfs_domount_first(td, vfsp, pathbuf, vp,
1105 free(pathbuf, M_TEMP);
1107 error = vfs_domount_update(td, vp, fsflags, optlist);
1110 ASSERT_VI_UNLOCKED(vp, __func__);
1111 ASSERT_VOP_UNLOCKED(vp, __func__);
1117 * Unmount a filesystem.
1119 * Note: unmount takes a path to the vnode mounted on as argument, not
1120 * special file (as before).
1122 #ifndef _SYS_SYSPROTO_H_
1123 struct unmount_args {
1130 sys_unmount(td, uap)
1132 register struct unmount_args /* {
1137 struct nameidata nd;
1140 int error, id0, id1;
1142 AUDIT_ARG_VALUE(uap->flags);
1143 if (jailed(td->td_ucred) || usermount == 0) {
1144 error = priv_check(td, PRIV_VFS_UNMOUNT);
1149 pathbuf = malloc(MNAMELEN, M_TEMP, M_WAITOK);
1150 error = copyinstr(uap->path, pathbuf, MNAMELEN, NULL);
1152 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) {
1161 free(pathbuf, M_TEMP);
1165 mtx_lock(&mountlist_mtx);
1166 TAILQ_FOREACH_REVERSE(mp, &mountlist, mntlist, mnt_list) {
1167 if (mp->mnt_stat.f_fsid.val[0] == id0 &&
1168 mp->mnt_stat.f_fsid.val[1] == id1)
1171 mtx_unlock(&mountlist_mtx);
1174 * Try to find global path for path argument.
1176 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1,
1177 UIO_SYSSPACE, pathbuf, td);
1178 if (namei(&nd) == 0) {
1179 NDFREE(&nd, NDF_ONLY_PNBUF);
1180 error = vn_path_to_global_path(td, nd.ni_vp, pathbuf,
1182 if (error == 0 || error == ENODEV)
1185 mtx_lock(&mountlist_mtx);
1186 TAILQ_FOREACH_REVERSE(mp, &mountlist, mntlist, mnt_list) {
1187 if (strcmp(mp->mnt_stat.f_mntonname, pathbuf) == 0)
1190 mtx_unlock(&mountlist_mtx);
1192 free(pathbuf, M_TEMP);
1195 * Previously we returned ENOENT for a nonexistent path and
1196 * EINVAL for a non-mountpoint. We cannot tell these apart
1197 * now, so in the !MNT_BYFSID case return the more likely
1198 * EINVAL for compatibility.
1201 return ((uap->flags & MNT_BYFSID) ? ENOENT : EINVAL);
1205 * Don't allow unmounting the root filesystem.
1207 if (mp->mnt_flag & MNT_ROOTFS) {
1211 error = dounmount(mp, uap->flags, td);
1217 * Do the actual filesystem unmount.
1220 dounmount(mp, flags, td)
1225 struct vnode *coveredvp, *fsrootvp;
1227 uint64_t async_flag;
1230 mtx_assert(&Giant, MA_OWNED);
1232 if ((coveredvp = mp->mnt_vnodecovered) != NULL) {
1233 mnt_gen_r = mp->mnt_gen;
1236 vn_lock(coveredvp, LK_EXCLUSIVE | LK_INTERLOCK | LK_RETRY);
1239 * Check for mp being unmounted while waiting for the
1240 * covered vnode lock.
1242 if (coveredvp->v_mountedhere != mp ||
1243 coveredvp->v_mountedhere->mnt_gen != mnt_gen_r) {
1244 VOP_UNLOCK(coveredvp, 0);
1249 * Only privileged root, or (if MNT_USER is set) the user that did the
1250 * original mount is permitted to unmount this filesystem.
1252 error = vfs_suser(mp, td);
1255 VOP_UNLOCK(coveredvp, 0);
1260 if ((mp->mnt_kern_flag & MNTK_UNMOUNT) != 0 ||
1261 !TAILQ_EMPTY(&mp->mnt_uppers)) {
1264 VOP_UNLOCK(coveredvp, 0);
1267 mp->mnt_kern_flag |= MNTK_UNMOUNT | MNTK_NOINSMNTQ;
1268 /* Allow filesystems to detect that a forced unmount is in progress. */
1269 if (flags & MNT_FORCE)
1270 mp->mnt_kern_flag |= MNTK_UNMOUNTF;
1272 if (mp->mnt_lockref) {
1273 mp->mnt_kern_flag |= MNTK_DRAINING;
1274 error = msleep(&mp->mnt_lockref, MNT_MTX(mp), PVFS,
1278 KASSERT(mp->mnt_lockref == 0,
1279 ("%s: invalid lock refcount in the drain path @ %s:%d",
1280 __func__, __FILE__, __LINE__));
1282 ("%s: invalid return value for msleep in the drain path @ %s:%d",
1283 __func__, __FILE__, __LINE__));
1284 vn_start_write(NULL, &mp, V_WAIT);
1286 if (mp->mnt_flag & MNT_EXPUBLIC)
1287 vfs_setpublicfs(NULL, NULL, NULL);
1289 vfs_msync(mp, MNT_WAIT);
1291 async_flag = mp->mnt_flag & MNT_ASYNC;
1292 mp->mnt_flag &= ~MNT_ASYNC;
1293 mp->mnt_kern_flag &= ~MNTK_ASYNC;
1295 cache_purgevfs(mp); /* remove cache entries for this file sys */
1296 vfs_deallocate_syncvnode(mp);
1298 * For forced unmounts, move process cdir/rdir refs on the fs root
1299 * vnode to the covered vnode. For non-forced unmounts we want
1300 * such references to cause an EBUSY error.
1302 if ((flags & MNT_FORCE) &&
1303 VFS_ROOT(mp, LK_EXCLUSIVE, &fsrootvp) == 0) {
1304 if (mp->mnt_vnodecovered != NULL)
1305 mountcheckdirs(fsrootvp, mp->mnt_vnodecovered);
1306 if (fsrootvp == rootvnode) {
1312 if (((mp->mnt_flag & MNT_RDONLY) ||
1313 (error = VFS_SYNC(mp, MNT_WAIT)) == 0) || (flags & MNT_FORCE) != 0)
1314 error = VFS_UNMOUNT(mp, flags);
1315 vn_finished_write(mp);
1317 * If we failed to flush the dirty blocks for this mount point,
1318 * undo all the cdir/rdir and rootvnode changes we made above.
1319 * Unless we failed to do so because the device is reporting that
1320 * it doesn't exist anymore.
1322 if (error && error != ENXIO) {
1323 if ((flags & MNT_FORCE) &&
1324 VFS_ROOT(mp, LK_EXCLUSIVE, &fsrootvp) == 0) {
1325 if (mp->mnt_vnodecovered != NULL)
1326 mountcheckdirs(mp->mnt_vnodecovered, fsrootvp);
1327 if (rootvnode == NULL) {
1328 rootvnode = fsrootvp;
1334 mp->mnt_kern_flag &= ~MNTK_NOINSMNTQ;
1335 if ((mp->mnt_flag & MNT_RDONLY) == 0) {
1337 vfs_allocate_syncvnode(mp);
1340 mp->mnt_kern_flag &= ~(MNTK_UNMOUNT | MNTK_UNMOUNTF);
1341 mp->mnt_flag |= async_flag;
1342 if ((mp->mnt_flag & MNT_ASYNC) != 0 &&
1343 (mp->mnt_kern_flag & MNTK_NOASYNC) == 0)
1344 mp->mnt_kern_flag |= MNTK_ASYNC;
1345 if (mp->mnt_kern_flag & MNTK_MWAIT) {
1346 mp->mnt_kern_flag &= ~MNTK_MWAIT;
1351 VOP_UNLOCK(coveredvp, 0);
1354 mtx_lock(&mountlist_mtx);
1355 TAILQ_REMOVE(&mountlist, mp, mnt_list);
1356 mtx_unlock(&mountlist_mtx);
1357 if (coveredvp != NULL) {
1358 coveredvp->v_mountedhere = NULL;
1361 vfs_event_signal(NULL, VQ_UNMOUNT, 0);
1362 vfs_mount_destroy(mp);
1367 * Report errors during filesystem mounting.
1370 vfs_mount_error(struct mount *mp, const char *fmt, ...)
1372 struct vfsoptlist *moptlist = mp->mnt_optnew;
1377 error = vfs_getopt(moptlist, "errmsg", (void **)&errmsg, &len);
1378 if (error || errmsg == NULL || len <= 0)
1382 vsnprintf(errmsg, (size_t)len, fmt, ap);
1387 vfs_opterror(struct vfsoptlist *opts, const char *fmt, ...)
1393 error = vfs_getopt(opts, "errmsg", (void **)&errmsg, &len);
1394 if (error || errmsg == NULL || len <= 0)
1398 vsnprintf(errmsg, (size_t)len, fmt, ap);
1403 * ---------------------------------------------------------------------
1404 * Functions for querying mount options/arguments from filesystems.
1408 * Check that no unknown options are given
1411 vfs_filteropt(struct vfsoptlist *opts, const char **legal)
1415 const char **t, *p, *q;
1418 TAILQ_FOREACH(opt, opts, link) {
1421 if (p[0] == 'n' && p[1] == 'o')
1423 for(t = global_opts; *t != NULL; t++) {
1424 if (strcmp(*t, p) == 0)
1427 if (strcmp(*t, q) == 0)
1433 for(t = legal; *t != NULL; t++) {
1434 if (strcmp(*t, p) == 0)
1437 if (strcmp(*t, q) == 0)
1443 snprintf(errmsg, sizeof(errmsg),
1444 "mount option <%s> is unknown", p);
1448 TAILQ_FOREACH(opt, opts, link) {
1449 if (strcmp(opt->name, "errmsg") == 0) {
1450 strncpy((char *)opt->value, errmsg, opt->len);
1455 printf("%s\n", errmsg);
1461 * Get a mount option by its name.
1463 * Return 0 if the option was found, ENOENT otherwise.
1464 * If len is non-NULL it will be filled with the length
1465 * of the option. If buf is non-NULL, it will be filled
1466 * with the address of the option.
1469 vfs_getopt(opts, name, buf, len)
1470 struct vfsoptlist *opts;
1477 KASSERT(opts != NULL, ("vfs_getopt: caller passed 'opts' as NULL"));
1479 TAILQ_FOREACH(opt, opts, link) {
1480 if (strcmp(name, opt->name) == 0) {
1493 vfs_getopt_pos(struct vfsoptlist *opts, const char *name)
1500 TAILQ_FOREACH(opt, opts, link) {
1501 if (strcmp(name, opt->name) == 0) {
1510 vfs_getopt_size(struct vfsoptlist *opts, const char *name, off_t *value)
1512 char *opt_value, *vtp;
1516 error = vfs_getopt(opts, name, (void **)&opt_value, &opt_len);
1519 if (opt_len == 0 || opt_value == NULL)
1521 if (opt_value[0] == '\0' || opt_value[opt_len - 1] != '\0')
1523 iv = strtoq(opt_value, &vtp, 0);
1524 if (vtp == opt_value || (vtp[0] != '\0' && vtp[1] != '\0'))
1552 vfs_getopts(struct vfsoptlist *opts, const char *name, int *error)
1557 TAILQ_FOREACH(opt, opts, link) {
1558 if (strcmp(name, opt->name) != 0)
1561 if (opt->len == 0 ||
1562 ((char *)opt->value)[opt->len - 1] != '\0') {
1566 return (opt->value);
1573 vfs_flagopt(struct vfsoptlist *opts, const char *name, uint64_t *w,
1578 TAILQ_FOREACH(opt, opts, link) {
1579 if (strcmp(name, opt->name) == 0) {
1592 vfs_scanopt(struct vfsoptlist *opts, const char *name, const char *fmt, ...)
1598 KASSERT(opts != NULL, ("vfs_getopt: caller passed 'opts' as NULL"));
1600 TAILQ_FOREACH(opt, opts, link) {
1601 if (strcmp(name, opt->name) != 0)
1604 if (opt->len == 0 || opt->value == NULL)
1606 if (((char *)opt->value)[opt->len - 1] != '\0')
1609 ret = vsscanf(opt->value, fmt, ap);
1617 vfs_setopt(struct vfsoptlist *opts, const char *name, void *value, int len)
1621 TAILQ_FOREACH(opt, opts, link) {
1622 if (strcmp(name, opt->name) != 0)
1625 if (opt->value == NULL)
1628 if (opt->len != len)
1630 bcopy(value, opt->value, len);
1638 vfs_setopt_part(struct vfsoptlist *opts, const char *name, void *value, int len)
1642 TAILQ_FOREACH(opt, opts, link) {
1643 if (strcmp(name, opt->name) != 0)
1646 if (opt->value == NULL)
1652 bcopy(value, opt->value, len);
1660 vfs_setopts(struct vfsoptlist *opts, const char *name, const char *value)
1664 TAILQ_FOREACH(opt, opts, link) {
1665 if (strcmp(name, opt->name) != 0)
1668 if (opt->value == NULL)
1669 opt->len = strlen(value) + 1;
1670 else if (strlcpy(opt->value, value, opt->len) >= opt->len)
1678 * Find and copy a mount option.
1680 * The size of the buffer has to be specified
1681 * in len, if it is not the same length as the
1682 * mount option, EINVAL is returned.
1683 * Returns ENOENT if the option is not found.
1686 vfs_copyopt(opts, name, dest, len)
1687 struct vfsoptlist *opts;
1694 KASSERT(opts != NULL, ("vfs_copyopt: caller passed 'opts' as NULL"));
1696 TAILQ_FOREACH(opt, opts, link) {
1697 if (strcmp(name, opt->name) == 0) {
1699 if (len != opt->len)
1701 bcopy(opt->value, dest, opt->len);
1709 * These are helper functions for filesystems to traverse all
1710 * their vnodes. See MNT_VNODE_FOREACH() in sys/mount.h.
1712 * This interface has been deprecated in favor of MNT_VNODE_FOREACH_ALL.
1715 MALLOC_DECLARE(M_VNODE_MARKER);
1718 __mnt_vnode_next(struct vnode **mvp, struct mount *mp)
1722 mtx_assert(MNT_MTX(mp), MA_OWNED);
1724 KASSERT((*mvp)->v_mount == mp, ("marker vnode mount list mismatch"));
1725 if (should_yield()) {
1727 kern_yield(PRI_UNCHANGED);
1730 vp = TAILQ_NEXT(*mvp, v_nmntvnodes);
1731 while (vp != NULL && vp->v_type == VMARKER)
1732 vp = TAILQ_NEXT(vp, v_nmntvnodes);
1734 /* Check if we are done */
1736 __mnt_vnode_markerfree(mvp, mp);
1739 TAILQ_REMOVE(&mp->mnt_nvnodelist, *mvp, v_nmntvnodes);
1740 TAILQ_INSERT_AFTER(&mp->mnt_nvnodelist, vp, *mvp, v_nmntvnodes);
1745 __mnt_vnode_first(struct vnode **mvp, struct mount *mp)
1749 mtx_assert(MNT_MTX(mp), MA_OWNED);
1751 vp = TAILQ_FIRST(&mp->mnt_nvnodelist);
1752 while (vp != NULL && vp->v_type == VMARKER)
1753 vp = TAILQ_NEXT(vp, v_nmntvnodes);
1755 /* Check if we are done */
1762 *mvp = (struct vnode *) malloc(sizeof(struct vnode),
1766 (*mvp)->v_type = VMARKER;
1768 vp = TAILQ_FIRST(&mp->mnt_nvnodelist);
1769 while (vp != NULL && vp->v_type == VMARKER)
1770 vp = TAILQ_NEXT(vp, v_nmntvnodes);
1772 /* Check if we are done */
1775 free(*mvp, M_VNODE_MARKER);
1781 (*mvp)->v_mount = mp;
1782 TAILQ_INSERT_AFTER(&mp->mnt_nvnodelist, vp, *mvp, v_nmntvnodes);
1788 __mnt_vnode_markerfree(struct vnode **mvp, struct mount *mp)
1794 mtx_assert(MNT_MTX(mp), MA_OWNED);
1796 KASSERT((*mvp)->v_mount == mp, ("marker vnode mount list mismatch"));
1797 TAILQ_REMOVE(&mp->mnt_nvnodelist, *mvp, v_nmntvnodes);
1799 free(*mvp, M_VNODE_MARKER);
1806 __vfs_statfs(struct mount *mp, struct statfs *sbp)
1810 error = mp->mnt_op->vfs_statfs(mp, &mp->mnt_stat);
1811 if (sbp != &mp->mnt_stat)
1812 *sbp = mp->mnt_stat;
1817 vfs_mountedfrom(struct mount *mp, const char *from)
1820 bzero(mp->mnt_stat.f_mntfromname, sizeof mp->mnt_stat.f_mntfromname);
1821 strlcpy(mp->mnt_stat.f_mntfromname, from,
1822 sizeof mp->mnt_stat.f_mntfromname);
1826 * ---------------------------------------------------------------------
1827 * This is the api for building mount args and mounting filesystems from
1828 * inside the kernel.
1830 * The API works by accumulation of individual args. First error is
1833 * XXX: should be documented in new manpage kernel_mount(9)
1836 /* A memory allocation which must be freed when we are done */
1838 SLIST_ENTRY(mntaarg) next;
1841 /* The header for the mount arguments */
1846 SLIST_HEAD(, mntaarg) list;
1850 * Add a boolean argument.
1852 * flag is the boolean value.
1853 * name must start with "no".
1856 mount_argb(struct mntarg *ma, int flag, const char *name)
1859 KASSERT(name[0] == 'n' && name[1] == 'o',
1860 ("mount_argb(...,%s): name must start with 'no'", name));
1862 return (mount_arg(ma, name + (flag ? 2 : 0), NULL, 0));
1866 * Add an argument printf style
1869 mount_argf(struct mntarg *ma, const char *name, const char *fmt, ...)
1872 struct mntaarg *maa;
1877 ma = malloc(sizeof *ma, M_MOUNT, M_WAITOK | M_ZERO);
1878 SLIST_INIT(&ma->list);
1883 ma->v = realloc(ma->v, sizeof *ma->v * (ma->len + 2),
1885 ma->v[ma->len].iov_base = (void *)(uintptr_t)name;
1886 ma->v[ma->len].iov_len = strlen(name) + 1;
1889 sb = sbuf_new_auto();
1891 sbuf_vprintf(sb, fmt, ap);
1894 len = sbuf_len(sb) + 1;
1895 maa = malloc(sizeof *maa + len, M_MOUNT, M_WAITOK | M_ZERO);
1896 SLIST_INSERT_HEAD(&ma->list, maa, next);
1897 bcopy(sbuf_data(sb), maa + 1, len);
1900 ma->v[ma->len].iov_base = maa + 1;
1901 ma->v[ma->len].iov_len = len;
1908 * Add an argument which is a userland string.
1911 mount_argsu(struct mntarg *ma, const char *name, const void *val, int len)
1913 struct mntaarg *maa;
1919 ma = malloc(sizeof *ma, M_MOUNT, M_WAITOK | M_ZERO);
1920 SLIST_INIT(&ma->list);
1924 maa = malloc(sizeof *maa + len, M_MOUNT, M_WAITOK | M_ZERO);
1925 SLIST_INSERT_HEAD(&ma->list, maa, next);
1926 tbuf = (void *)(maa + 1);
1927 ma->error = copyinstr(val, tbuf, len, NULL);
1928 return (mount_arg(ma, name, tbuf, -1));
1934 * If length is -1, treat value as a C string.
1937 mount_arg(struct mntarg *ma, const char *name, const void *val, int len)
1941 ma = malloc(sizeof *ma, M_MOUNT, M_WAITOK | M_ZERO);
1942 SLIST_INIT(&ma->list);
1947 ma->v = realloc(ma->v, sizeof *ma->v * (ma->len + 2),
1949 ma->v[ma->len].iov_base = (void *)(uintptr_t)name;
1950 ma->v[ma->len].iov_len = strlen(name) + 1;
1953 ma->v[ma->len].iov_base = (void *)(uintptr_t)val;
1955 ma->v[ma->len].iov_len = strlen(val) + 1;
1957 ma->v[ma->len].iov_len = len;
1963 * Free a mntarg structure
1966 free_mntarg(struct mntarg *ma)
1968 struct mntaarg *maa;
1970 while (!SLIST_EMPTY(&ma->list)) {
1971 maa = SLIST_FIRST(&ma->list);
1972 SLIST_REMOVE_HEAD(&ma->list, next);
1975 free(ma->v, M_MOUNT);
1980 * Mount a filesystem
1983 kernel_mount(struct mntarg *ma, uint64_t flags)
1988 KASSERT(ma != NULL, ("kernel_mount NULL ma"));
1989 KASSERT(ma->v != NULL, ("kernel_mount NULL ma->v"));
1990 KASSERT(!(ma->len & 1), ("kernel_mount odd ma->len (%d)", ma->len));
1992 auio.uio_iov = ma->v;
1993 auio.uio_iovcnt = ma->len;
1994 auio.uio_segflg = UIO_SYSSPACE;
1998 error = vfs_donmount(curthread, flags, &auio);
2004 * A printflike function to mount a filesystem.
2007 kernel_vmount(int flags, ...)
2009 struct mntarg *ma = NULL;
2015 va_start(ap, flags);
2017 cp = va_arg(ap, const char *);
2020 vp = va_arg(ap, const void *);
2021 ma = mount_arg(ma, cp, vp, (vp != NULL ? -1 : 0));
2025 error = kernel_mount(ma, flags);
2030 vfs_oexport_conv(const struct oexport_args *oexp, struct export_args *exp)
2033 bcopy(oexp, exp, sizeof(*oexp));
2034 exp->ex_numsecflavors = 0;