The Design and Implementation of the FreeBSD Operating System, Second Edition
Now available: The Design and Implementation of the FreeBSD Operating System (Second Edition)


[ source navigation ] [ diff markup ] [ identifier search ] [ freetext search ] [ file search ] [ list types ] [ track identifier ]

FreeBSD/Linux Kernel Cross Reference
sys/kern/vfs_mount.c

Version: -  FREEBSD  -  FREEBSD-13-STABLE  -  FREEBSD-13-0  -  FREEBSD-12-STABLE  -  FREEBSD-12-0  -  FREEBSD-11-STABLE  -  FREEBSD-11-0  -  FREEBSD-10-STABLE  -  FREEBSD-10-0  -  FREEBSD-9-STABLE  -  FREEBSD-9-0  -  FREEBSD-8-STABLE  -  FREEBSD-8-0  -  FREEBSD-7-STABLE  -  FREEBSD-7-0  -  FREEBSD-6-STABLE  -  FREEBSD-6-0  -  FREEBSD-5-STABLE  -  FREEBSD-5-0  -  FREEBSD-4-STABLE  -  FREEBSD-3-STABLE  -  FREEBSD22  -  l41  -  OPENBSD  -  linux-2.6  -  MK84  -  PLAN9  -  xnu-8792 
SearchContext: -  none  -  3  -  10 

    1 /*-
    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.
   11  *
   12  * Redistribution and use in source and binary forms, with or without
   13  * modification, are permitted provided that the following conditions
   14  * are met:
   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.
   23  *
   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
   34  * SUCH DAMAGE.
   35  */
   36 
   37 #include <sys/cdefs.h>
   38 __FBSDID("$FreeBSD$");
   39 
   40 #include <sys/param.h>
   41 #include <sys/conf.h>
   42 #include <sys/fcntl.h>
   43 #include <sys/jail.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>
   50 #include <sys/priv.h>
   51 #include <sys/proc.h>
   52 #include <sys/filedesc.h>
   53 #include <sys/reboot.h>
   54 #include <sys/sbuf.h>
   55 #include <sys/syscallsubr.h>
   56 #include <sys/sysproto.h>
   57 #include <sys/sx.h>
   58 #include <sys/sysctl.h>
   59 #include <sys/sysent.h>
   60 #include <sys/systm.h>
   61 #include <sys/vnode.h>
   62 #include <vm/uma.h>
   63 
   64 #include <geom/geom.h>
   65 
   66 #include <machine/stdarg.h>
   67 
   68 #include <security/audit/audit.h>
   69 #include <security/mac/mac_framework.h>
   70 
   71 #define VFS_MOUNTARG_SIZE_MAX   (1024 * 64)
   72 
   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);
   76 
   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");
   80 
   81 MALLOC_DEFINE(M_MOUNT, "mount", "vfs mount structure");
   82 static uma_zone_t mount_zone;
   83 
   84 /* List of mounted filesystems. */
   85 struct mntlist mountlist = TAILQ_HEAD_INITIALIZER(mountlist);
   86 
   87 /* For any iteration/modification of mountlist */
   88 struct mtx mountlist_mtx;
   89 MTX_SYSINIT(mountlist, &mountlist_mtx, "mountlist", MTX_DEF);
   90 
   91 /*
   92  * Global opts, taken by all filesystems
   93  */
   94 static const char *global_opts[] = {
   95         "errmsg",
   96         "fstype",
   97         "fspath",
   98         "ro",
   99         "rw",
  100         "nosuid",
  101         "noexec",
  102         NULL
  103 };
  104 
  105 static int
  106 mount_init(void *mem, int size, int flags)
  107 {
  108         struct mount *mp;
  109 
  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);
  113         return (0);
  114 }
  115 
  116 static void
  117 mount_fini(void *mem, int size)
  118 {
  119         struct mount *mp;
  120 
  121         mp = (struct mount *)mem;
  122         lockdestroy(&mp->mnt_explock);
  123         mtx_destroy(&mp->mnt_mtx);
  124 }
  125 
  126 static void
  127 vfs_mount_init(void *dummy __unused)
  128 {
  129 
  130         mount_zone = uma_zcreate("Mountpoints", sizeof(struct mount), NULL,
  131             NULL, mount_init, mount_fini, UMA_ALIGN_PTR, UMA_ZONE_NOFREE);
  132 }
  133 SYSINIT(vfs_mount, SI_SUB_VFS, SI_ORDER_ANY, vfs_mount_init, NULL);
  134 
  135 /*
  136  * ---------------------------------------------------------------------
  137  * Functions for building and sanitizing the mount options
  138  */
  139 
  140 /* Remove one mount option. */
  141 static void
  142 vfs_freeopt(struct vfsoptlist *opts, struct vfsopt *opt)
  143 {
  144 
  145         TAILQ_REMOVE(opts, opt, link);
  146         free(opt->name, M_MOUNT);
  147         if (opt->value != NULL)
  148                 free(opt->value, M_MOUNT);
  149         free(opt, M_MOUNT);
  150 }
  151 
  152 /* Release all resources related to the mount options. */
  153 void
  154 vfs_freeopts(struct vfsoptlist *opts)
  155 {
  156         struct vfsopt *opt;
  157 
  158         while (!TAILQ_EMPTY(opts)) {
  159                 opt = TAILQ_FIRST(opts);
  160                 vfs_freeopt(opts, opt);
  161         }
  162         free(opts, M_MOUNT);
  163 }
  164 
  165 void
  166 vfs_deleteopt(struct vfsoptlist *opts, const char *name)
  167 {
  168         struct vfsopt *opt, *temp;
  169 
  170         if (opts == NULL)
  171                 return;
  172         TAILQ_FOREACH_SAFE(opt, opts, link, temp)  {
  173                 if (strcmp(opt->name, name) == 0)
  174                         vfs_freeopt(opts, opt);
  175         }
  176 }
  177 
  178 static int
  179 vfs_isopt_ro(const char *opt)
  180 {
  181 
  182         if (strcmp(opt, "ro") == 0 || strcmp(opt, "rdonly") == 0 ||
  183             strcmp(opt, "norw") == 0)
  184                 return (1);
  185         return (0);
  186 }
  187 
  188 static int
  189 vfs_isopt_rw(const char *opt)
  190 {
  191 
  192         if (strcmp(opt, "rw") == 0 || strcmp(opt, "noro") == 0)
  193                 return (1);
  194         return (0);
  195 }
  196 
  197 /*
  198  * Check if options are equal (with or without the "no" prefix).
  199  */
  200 static int
  201 vfs_equalopts(const char *opt1, const char *opt2)
  202 {
  203         char *p;
  204 
  205         /* "opt" vs. "opt" or "noopt" vs. "noopt" */
  206         if (strcmp(opt1, opt2) == 0)
  207                 return (1);
  208         /* "noopt" vs. "opt" */
  209         if (strncmp(opt1, "no", 2) == 0 && strcmp(opt1 + 2, opt2) == 0)
  210                 return (1);
  211         /* "opt" vs. "noopt" */
  212         if (strncmp(opt2, "no", 2) == 0 && strcmp(opt1, opt2 + 2) == 0)
  213                 return (1);
  214         while ((p = strchr(opt1, '.')) != NULL &&
  215             !strncmp(opt1, opt2, ++p - opt1)) {
  216                 opt2 += p - opt1;
  217                 opt1 = p;
  218                 /* "foo.noopt" vs. "foo.opt" */
  219                 if (strncmp(opt1, "no", 2) == 0 && strcmp(opt1 + 2, opt2) == 0)
  220                         return (1);
  221                 /* "foo.opt" vs. "foo.noopt" */
  222                 if (strncmp(opt2, "no", 2) == 0 && strcmp(opt1, opt2 + 2) == 0)
  223                         return (1);
  224         }
  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)))
  228                 return (1);
  229         return (0);
  230 }
  231 
  232 /*
  233  * If a mount option is specified several times,
  234  * (with or without the "no" prefix) only keep
  235  * the last occurence of it.
  236  */
  237 static void
  238 vfs_sanitizeopts(struct vfsoptlist *opts)
  239 {
  240         struct vfsopt *opt, *opt2, *tmp;
  241 
  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);
  248                                 opt2 = tmp;
  249                         } else {
  250                                 opt2 = TAILQ_PREV(opt2, vfsoptlist, link);
  251                         }
  252                 }
  253         }
  254 }
  255 
  256 /*
  257  * Build a linked list of mount options from a struct uio.
  258  */
  259 int
  260 vfs_buildopts(struct uio *auio, struct vfsoptlist **options)
  261 {
  262         struct vfsoptlist *opts;
  263         struct vfsopt *opt;
  264         size_t memused, namelen, optlen;
  265         unsigned int i, iovcnt;
  266         int error;
  267 
  268         opts = malloc(sizeof(struct vfsoptlist), M_MOUNT, M_WAITOK);
  269         TAILQ_INIT(opts);
  270         memused = 0;
  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;
  276                 /*
  277                  * Avoid consuming too much memory, and attempts to overflow
  278                  * memused.
  279                  */
  280                 if (memused > VFS_MOUNTARG_SIZE_MAX ||
  281                     optlen > VFS_MOUNTARG_SIZE_MAX ||
  282                     namelen > VFS_MOUNTARG_SIZE_MAX) {
  283                         error = EINVAL;
  284                         goto bad;
  285                 }
  286 
  287                 opt = malloc(sizeof(struct vfsopt), M_MOUNT, M_WAITOK);
  288                 opt->name = malloc(namelen, M_MOUNT, M_WAITOK);
  289                 opt->value = NULL;
  290                 opt->len = 0;
  291                 opt->pos = i / 2;
  292                 opt->seen = 0;
  293 
  294                 /*
  295                  * Do this early, so jumps to "bad" will free the current
  296                  * option.
  297                  */
  298                 TAILQ_INSERT_TAIL(opts, opt, link);
  299 
  300                 if (auio->uio_segflg == UIO_SYSSPACE) {
  301                         bcopy(auio->uio_iov[i].iov_base, opt->name, namelen);
  302                 } else {
  303                         error = copyin(auio->uio_iov[i].iov_base, opt->name,
  304                             namelen);
  305                         if (error)
  306                                 goto bad;
  307                 }
  308                 /* Ensure names are null-terminated strings. */
  309                 if (namelen == 0 || opt->name[namelen - 1] != '\0') {
  310                         error = EINVAL;
  311                         goto bad;
  312                 }
  313                 if (optlen != 0) {
  314                         opt->len = optlen;
  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,
  318                                     optlen);
  319                         } else {
  320                                 error = copyin(auio->uio_iov[i + 1].iov_base,
  321                                     opt->value, optlen);
  322                                 if (error)
  323                                         goto bad;
  324                         }
  325                 }
  326         }
  327         vfs_sanitizeopts(opts);
  328         *options = opts;
  329         return (0);
  330 bad:
  331         vfs_freeopts(opts);
  332         return (error);
  333 }
  334 
  335 /*
  336  * Merge the old mount options with the new ones passed
  337  * in the MNT_UPDATE case.
  338  *
  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.
  342  */
  343 static void
  344 vfs_mergeopts(struct vfsoptlist *toopts, struct vfsoptlist *oldopts)
  345 {
  346         struct vfsopt *opt, *new;
  347 
  348         TAILQ_FOREACH(opt, oldopts, link) {
  349                 new = malloc(sizeof(struct vfsopt), M_MOUNT, M_WAITOK);
  350                 new->name = strdup(opt->name, M_MOUNT);
  351                 if (opt->len != 0) {
  352                         new->value = malloc(opt->len, M_MOUNT, M_WAITOK);
  353                         bcopy(opt->value, new->value, opt->len);
  354                 } else
  355                         new->value = NULL;
  356                 new->len = opt->len;
  357                 new->seen = opt->seen;
  358                 TAILQ_INSERT_HEAD(toopts, new, link);
  359         }
  360         vfs_sanitizeopts(toopts);
  361 }
  362 
  363 /*
  364  * Mount a filesystem.
  365  */
  366 int
  367 sys_nmount(td, uap)
  368         struct thread *td;
  369         struct nmount_args /* {
  370                 struct iovec *iovp;
  371                 unsigned int iovcnt;
  372                 int flags;
  373         } */ *uap;
  374 {
  375         struct uio *auio;
  376         int error;
  377         u_int iovcnt;
  378         uint64_t flags;
  379 
  380         /*
  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.
  384          */
  385         flags = uap->flags;
  386 
  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);
  390 
  391         /*
  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
  396          * root file system.
  397          */
  398         flags &= ~MNT_ROOTFS;
  399 
  400         iovcnt = uap->iovcnt;
  401         /*
  402          * Check that we have an even number of iovec's
  403          * and that we have at least two options.
  404          */
  405         if ((iovcnt & 1) || (iovcnt < 4)) {
  406                 CTR2(KTR_VFS, "%s: failed for invalid iovcnt %d", __func__,
  407                     uap->iovcnt);
  408                 return (EINVAL);
  409         }
  410 
  411         error = copyinuio(uap->iovp, iovcnt, &auio);
  412         if (error) {
  413                 CTR2(KTR_VFS, "%s: failed for invalid uio op with %d errno",
  414                     __func__, error);
  415                 return (error);
  416         }
  417         error = vfs_donmount(td, flags, auio);
  418 
  419         free(auio, M_IOV);
  420         return (error);
  421 }
  422 
  423 /*
  424  * ---------------------------------------------------------------------
  425  * Various utility functions
  426  */
  427 
  428 void
  429 vfs_ref(struct mount *mp)
  430 {
  431 
  432         CTR2(KTR_VFS, "%s: mp %p", __func__, mp);
  433         MNT_ILOCK(mp);
  434         MNT_REF(mp);
  435         MNT_IUNLOCK(mp);
  436 }
  437 
  438 void
  439 vfs_rel(struct mount *mp)
  440 {
  441 
  442         CTR2(KTR_VFS, "%s: mp %p", __func__, mp);
  443         MNT_ILOCK(mp);
  444         MNT_REL(mp);
  445         MNT_IUNLOCK(mp);
  446 }
  447 
  448 /*
  449  * Allocate and initialize the mount point struct.
  450  */
  451 struct mount *
  452 vfs_mount_alloc(struct vnode *vp, struct vfsconf *vfsp, const char *fspath,
  453     struct ucred *cred)
  454 {
  455         struct mount *mp;
  456 
  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;
  464         mp->mnt_ref = 0;
  465         (void) vfs_busy(mp, MBF_NOWAIT);
  466         mp->mnt_op = vfsp->vfc_vfsops;
  467         mp->mnt_vfc = vfsp;
  468         vfsp->vfc_refcount++;   /* XXX Unlocked */
  469         mp->mnt_stat.f_type = vfsp->vfc_typenum;
  470         mp->mnt_gen++;
  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;
  477 #ifdef MAC
  478         mac_mount_init(mp);
  479         mac_mount_create(cred, mp);
  480 #endif
  481         arc4rand(&mp->mnt_hashseed, sizeof mp->mnt_hashseed, 0);
  482         TAILQ_INIT(&mp->mnt_uppers);
  483         return (mp);
  484 }
  485 
  486 /*
  487  * Destroy the mount struct previously allocated by vfs_mount_alloc().
  488  */
  489 void
  490 vfs_mount_destroy(struct mount *mp)
  491 {
  492 
  493         MNT_ILOCK(mp);
  494         mp->mnt_kern_flag |= MNTK_REFEXPIRE;
  495         if (mp->mnt_kern_flag & MNTK_MWAIT) {
  496                 mp->mnt_kern_flag &= ~MNTK_MWAIT;
  497                 wakeup(mp);
  498         }
  499         while (mp->mnt_ref)
  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)) {
  510                 struct vnode *vp;
  511 
  512                 TAILQ_FOREACH(vp, &mp->mnt_nvnodelist, v_nmntvnodes)
  513                         vprint("", vp);
  514                 panic("unmount: dangling vnode");
  515         }
  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");
  523         MNT_IUNLOCK(mp);
  524 #ifdef MAC
  525         mac_mount_destroy(mp);
  526 #endif
  527         if (mp->mnt_opt != NULL)
  528                 vfs_freeopts(mp->mnt_opt);
  529         crfree(mp->mnt_cred);
  530         uma_zfree(mount_zone, mp);
  531 }
  532 
  533 int
  534 vfs_donmount(struct thread *td, uint64_t fsflags, struct uio *fsoptions)
  535 {
  536         struct vfsoptlist *optlist;
  537         struct vfsopt *opt, *tmp_opt;
  538         char *fstype, *fspath, *errmsg;
  539         int error, fstypelen, fspathlen, errmsg_len, errmsg_pos;
  540 
  541         errmsg = fspath = NULL;
  542         errmsg_len = fspathlen = 0;
  543         errmsg_pos = -1;
  544 
  545         error = vfs_buildopts(fsoptions, &optlist);
  546         if (error)
  547                 return (error);
  548 
  549         if (vfs_getopt(optlist, "errmsg", (void **)&errmsg, &errmsg_len) == 0)
  550                 errmsg_pos = vfs_getopt_pos(optlist, "errmsg");
  551 
  552         /*
  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.
  556          */
  557         fstypelen = 0;
  558         error = vfs_getopt(optlist, "fstype", (void **)&fstype, &fstypelen);
  559         if (error || fstype[fstypelen - 1] != '\0') {
  560                 error = EINVAL;
  561                 if (errmsg != NULL)
  562                         strncpy(errmsg, "Invalid fstype", errmsg_len);
  563                 goto bail;
  564         }
  565         fspathlen = 0;
  566         error = vfs_getopt(optlist, "fspath", (void **)&fspath, &fspathlen);
  567         if (error || fspath[fspathlen - 1] != '\0') {
  568                 error = EINVAL;
  569                 if (errmsg != NULL)
  570                         strncpy(errmsg, "Invalid fspath", errmsg_len);
  571                 goto bail;
  572         }
  573 
  574         /*
  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.
  579          */
  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);
  584                 }
  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);
  590                 }
  591                 else if (strcmp(opt->name, "reload") == 0) {
  592                         fsflags |= MNT_RELOAD;
  593                         vfs_freeopt(optlist, opt);
  594                 }
  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);
  604                 }
  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);
  610                 }
  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);
  616                 }
  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);
  622                 }
  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);
  628                 }
  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);
  634                 }
  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;
  645                 }
  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;
  652         }
  653 
  654         /*
  655          * Be ultra-paranoid about making sure the type and fspath
  656          * variables will fit in our mp buffers, including the
  657          * terminating NUL.
  658          */
  659         if (fstypelen >= MFSNAMELEN - 1 || fspathlen >= MNAMELEN - 1) {
  660                 error = ENAMETOOLONG;
  661                 goto bail;
  662         }
  663 
  664         error = vfs_domount(td, fstype, fspath, fsflags, &optlist);
  665 bail:
  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) {
  670                         bcopy(errmsg,
  671                             fsoptions->uio_iov[2 * errmsg_pos + 1].iov_base,
  672                             fsoptions->uio_iov[2 * errmsg_pos + 1].iov_len);
  673                 } else {
  674                         copyout(errmsg,
  675                             fsoptions->uio_iov[2 * errmsg_pos + 1].iov_base,
  676                             fsoptions->uio_iov[2 * errmsg_pos + 1].iov_len);
  677                 }
  678         }
  679 
  680         if (optlist != NULL)
  681                 vfs_freeopts(optlist);
  682         return (error);
  683 }
  684 
  685 /*
  686  * Old mount API.
  687  */
  688 #ifndef _SYS_SYSPROTO_H_
  689 struct mount_args {
  690         char    *type;
  691         char    *path;
  692         int     flags;
  693         caddr_t data;
  694 };
  695 #endif
  696 /* ARGSUSED */
  697 int
  698 sys_mount(td, uap)
  699         struct thread *td;
  700         struct mount_args /* {
  701                 char *type;
  702                 char *path;
  703                 int flags;
  704                 caddr_t data;
  705         } */ *uap;
  706 {
  707         char *fstype;
  708         struct vfsconf *vfsp = NULL;
  709         struct mntarg *ma = NULL;
  710         uint64_t flags;
  711         int error;
  712 
  713         /*
  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.
  717          */
  718         flags = uap->flags;
  719 
  720         AUDIT_ARG_FFLAGS(flags);
  721 
  722         /*
  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
  727          * root file system.
  728          */
  729         flags &= ~MNT_ROOTFS;
  730 
  731         fstype = malloc(MFSNAMELEN, M_TEMP, M_WAITOK);
  732         error = copyinstr(uap->type, fstype, MFSNAMELEN, NULL);
  733         if (error) {
  734                 free(fstype, M_TEMP);
  735                 return (error);
  736         }
  737 
  738         AUDIT_ARG_TEXT(fstype);
  739         mtx_lock(&Giant);
  740         vfsp = vfs_byname_kld(fstype, td, &error);
  741         free(fstype, M_TEMP);
  742         if (vfsp == NULL) {
  743                 mtx_unlock(&Giant);
  744                 return (ENOENT);
  745         }
  746         if (vfsp->vfc_vfsops->vfs_cmount == NULL) {
  747                 mtx_unlock(&Giant);
  748                 return (EOPNOTSUPP);
  749         }
  750 
  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");
  756 
  757         error = vfsp->vfc_vfsops->vfs_cmount(ma, uap->data, flags);
  758         mtx_unlock(&Giant);
  759         return (error);
  760 }
  761 
  762 /*
  763  * vfs_domount_first(): first file system mount (not update)
  764  */
  765 static int
  766 vfs_domount_first(
  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. */
  773         )
  774 {
  775         struct vattr va;
  776         struct mount *mp;
  777         struct vnode *newdp;
  778         int error;
  779 
  780         mtx_assert(&Giant, MA_OWNED);
  781         ASSERT_VOP_ELOCKED(vp, __func__);
  782         KASSERT((fsflags & MNT_UPDATE) == 0, ("MNT_UPDATE shouldn't be here"));
  783 
  784         /*
  785          * If the user is not root, ensure that they own the directory
  786          * onto which we are attempting to mount.
  787          */
  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);
  791         if (error == 0)
  792                 error = vinvalbuf(vp, V_SAVE, 0, 0);
  793         if (error == 0 && vp->v_type != VDIR)
  794                 error = ENOTDIR;
  795         if (error == 0) {
  796                 VI_LOCK(vp);
  797                 if ((vp->v_iflag & VI_MOUNT) == 0 && vp->v_mountedhere == NULL)
  798                         vp->v_iflag |= VI_MOUNT;
  799                 else
  800                         error = EBUSY;
  801                 VI_UNLOCK(vp);
  802         }
  803         if (error != 0) {
  804                 vput(vp);
  805                 return (error);
  806         }
  807         VOP_UNLOCK(vp, 0);
  808 
  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));
  815 
  816         /*
  817          * Mount the filesystem.
  818          * XXX The final recipients of VFS_MOUNT just overwrite the ndp they
  819          * get.  No freeing of cn_pnbuf.
  820          */
  821         error = VFS_MOUNT(mp);
  822         if (error != 0) {
  823                 vfs_unbusy(mp);
  824                 vfs_mount_destroy(mp);
  825                 VI_LOCK(vp);
  826                 vp->v_iflag &= ~VI_MOUNT;
  827                 VI_UNLOCK(vp);
  828                 vrele(vp);
  829                 return (error);
  830         }
  831 
  832         if (mp->mnt_opt != NULL)
  833                 vfs_freeopts(mp->mnt_opt);
  834         mp->mnt_opt = mp->mnt_optnew;
  835         *optlist = NULL;
  836         (void)VFS_STATFS(mp, &mp->mnt_stat);
  837 
  838         /*
  839          * Prevent external consumers of mount options from reading mnt_optnew.
  840          */
  841         mp->mnt_optnew = NULL;
  842 
  843         MNT_ILOCK(mp);
  844         if ((mp->mnt_flag & MNT_ASYNC) != 0 &&
  845             (mp->mnt_kern_flag & MNTK_NOASYNC) == 0)
  846                 mp->mnt_kern_flag |= MNTK_ASYNC;
  847         else
  848                 mp->mnt_kern_flag &= ~MNTK_ASYNC;
  849         MNT_IUNLOCK(mp);
  850 
  851         vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
  852         cache_purge(vp);
  853         VI_LOCK(vp);
  854         vp->v_iflag &= ~VI_MOUNT;
  855         VI_UNLOCK(vp);
  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);
  865         VOP_UNLOCK(vp, 0);
  866         mountcheckdirs(vp, newdp);
  867         vrele(newdp);
  868         if ((mp->mnt_flag & MNT_RDONLY) == 0)
  869                 vfs_allocate_syncvnode(mp);
  870         vfs_unbusy(mp);
  871         return (0);
  872 }
  873 
  874 /*
  875  * vfs_domount_update(): update of mounted file system
  876  */
  877 static int
  878 vfs_domount_update(
  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. */
  883         )
  884 {
  885         struct oexport_args oexport;
  886         struct export_args export;
  887         struct mount *mp;
  888         int error, export_error;
  889         uint64_t flag;
  890 
  891         mtx_assert(&Giant, MA_OWNED);
  892         ASSERT_VOP_ELOCKED(vp, __func__);
  893         KASSERT((fsflags & MNT_UPDATE) != 0, ("MNT_UPDATE should be here"));
  894 
  895         if ((vp->v_vflag & VV_ROOT) == 0) {
  896                 vput(vp);
  897                 return (EINVAL);
  898         }
  899         mp = vp->v_mount;
  900         /*
  901          * We only allow the filesystem to be reloaded if it
  902          * is currently mounted read-only.
  903          */
  904         flag = mp->mnt_flag;
  905         if ((fsflags & MNT_RELOAD) != 0 && (flag & MNT_RDONLY) == 0) {
  906                 vput(vp);
  907                 return (EOPNOTSUPP);    /* Needs translation */
  908         }
  909         /*
  910          * Only privileged root, or (if MNT_USER is set) the user that
  911          * did the original mount is permitted to update it.
  912          */
  913         error = vfs_suser(mp, td);
  914         if (error != 0) {
  915                 vput(vp);
  916                 return (error);
  917         }
  918         if (vfs_busy(mp, MBF_NOWAIT)) {
  919                 vput(vp);
  920                 return (EBUSY);
  921         }
  922         VI_LOCK(vp);
  923         if ((vp->v_iflag & VI_MOUNT) != 0 || vp->v_mountedhere != NULL) {
  924                 VI_UNLOCK(vp);
  925                 vfs_unbusy(mp);
  926                 vput(vp);
  927                 return (EBUSY);
  928         }
  929         vp->v_iflag |= VI_MOUNT;
  930         VI_UNLOCK(vp);
  931         VOP_UNLOCK(vp, 0);
  932 
  933         MNT_ILOCK(mp);
  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;
  939         MNT_IUNLOCK(mp);
  940         mp->mnt_optnew = *optlist;
  941         vfs_mergeopts(mp->mnt_optnew, mp->mnt_opt);
  942 
  943         /*
  944          * Mount the filesystem.
  945          * XXX The final recipients of VFS_MOUNT just overwrite the ndp they
  946          * get.  No freeing of cn_pnbuf.
  947          */
  948         error = VFS_MOUNT(mp);
  949 
  950         export_error = 0;
  951         if (error == 0) {
  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);
  968                 }
  969         }
  970 
  971         MNT_ILOCK(mp);
  972         if (error == 0) {
  973                 mp->mnt_flag &= ~(MNT_UPDATE | MNT_RELOAD | MNT_FORCE |
  974                     MNT_SNAPSHOT);
  975         } else {
  976                 /*
  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
  981                  * one.
  982                  */
  983                 mp->mnt_flag = (mp->mnt_flag & MNT_QUOTA) | (flag & ~MNT_QUOTA);
  984         }
  985         if ((mp->mnt_flag & MNT_ASYNC) != 0 &&
  986             (mp->mnt_kern_flag & MNTK_NOASYNC) == 0)
  987                 mp->mnt_kern_flag |= MNTK_ASYNC;
  988         else
  989                 mp->mnt_kern_flag &= ~MNTK_ASYNC;
  990         MNT_IUNLOCK(mp);
  991 
  992         if (error != 0)
  993                 goto end;
  994 
  995         if (mp->mnt_opt != NULL)
  996                 vfs_freeopts(mp->mnt_opt);
  997         mp->mnt_opt = mp->mnt_optnew;
  998         *optlist = NULL;
  999         (void)VFS_STATFS(mp, &mp->mnt_stat);
 1000         /*
 1001          * Prevent external consumers of mount options from reading
 1002          * mnt_optnew.
 1003          */
 1004         mp->mnt_optnew = NULL;
 1005 
 1006         if ((mp->mnt_flag & MNT_RDONLY) == 0)
 1007                 vfs_allocate_syncvnode(mp);
 1008         else
 1009                 vfs_deallocate_syncvnode(mp);
 1010 end:
 1011         vfs_unbusy(mp);
 1012         VI_LOCK(vp);
 1013         vp->v_iflag &= ~VI_MOUNT;
 1014         VI_UNLOCK(vp);
 1015         vrele(vp);
 1016         return (error != 0 ? error : export_error);
 1017 }
 1018 
 1019 /*
 1020  * vfs_domount(): actually attempt a filesystem mount.
 1021  */
 1022 static int
 1023 vfs_domount(
 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. */
 1029         )
 1030 {
 1031         struct vfsconf *vfsp;
 1032         struct nameidata nd;
 1033         struct vnode *vp;
 1034         char *pathbuf;
 1035         int error;
 1036 
 1037         /*
 1038          * Be ultra-paranoid about making sure the type and fspath
 1039          * variables will fit in our mp buffers, including the
 1040          * terminating NUL.
 1041          */
 1042         if (strlen(fstype) >= MFSNAMELEN || strlen(fspath) >= MNAMELEN)
 1043                 return (ENAMETOOLONG);
 1044 
 1045         if (jailed(td->td_ucred) || usermount == 0) {
 1046                 if ((error = priv_check(td, PRIV_VFS_MOUNT)) != 0)
 1047                         return (error);
 1048         }
 1049 
 1050         /*
 1051          * Do not allow NFS export or MNT_SUIDDIR by unprivileged users.
 1052          */
 1053         if (fsflags & MNT_EXPORTED) {
 1054                 error = priv_check(td, PRIV_VFS_MOUNT_EXPORTED);
 1055                 if (error)
 1056                         return (error);
 1057         }
 1058         if (fsflags & MNT_SUIDDIR) {
 1059                 error = priv_check(td, PRIV_VFS_MOUNT_SUIDDIR);
 1060                 if (error)
 1061                         return (error);
 1062         }
 1063         /*
 1064          * Silently enforce MNT_NOSUID and MNT_USER for unprivileged users.
 1065          */
 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;
 1069         }
 1070 
 1071         /* Load KLDs before we lock the covered vnode to avoid reversals. */
 1072         vfsp = NULL;
 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);
 1077                 else
 1078                         vfsp = vfs_byname_kld(fstype, td, &error);
 1079                 if (vfsp == NULL)
 1080                         return (ENODEV);
 1081                 if (jailed(td->td_ucred) && !(vfsp->vfc_flags & VFCF_JAIL))
 1082                         return (EPERM);
 1083         }
 1084 
 1085         /*
 1086          * Get vnode to be covered or mount point's vnode in case of MNT_UPDATE.
 1087          */
 1088         NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | MPSAFE | AUDITVNODE1,
 1089             UIO_SYSSPACE, fspath, td);
 1090         error = namei(&nd);
 1091         if (error != 0)
 1092                 return (error);
 1093         if (!NDHASGIANT(&nd))
 1094                 mtx_lock(&Giant);
 1095         NDFREE(&nd, NDF_ONLY_PNBUF);
 1096         vp = nd.ni_vp;
 1097         if ((fsflags & MNT_UPDATE) == 0) {
 1098                 pathbuf = malloc(MNAMELEN, M_TEMP, M_WAITOK);
 1099                 strcpy(pathbuf, fspath);
 1100                 error = vn_path_to_global_path(td, vp, pathbuf, MNAMELEN);
 1101                 /* debug.disablefullpath == 1 results in ENODEV */
 1102                 if (error == 0 || error == ENODEV) {
 1103                         error = vfs_domount_first(td, vfsp, pathbuf, vp,
 1104                             fsflags, optlist);
 1105                 }
 1106                 free(pathbuf, M_TEMP);
 1107         } else
 1108                 error = vfs_domount_update(td, vp, fsflags, optlist);
 1109         mtx_unlock(&Giant);
 1110 
 1111         ASSERT_VI_UNLOCKED(vp, __func__);
 1112         ASSERT_VOP_UNLOCKED(vp, __func__);
 1113 
 1114         return (error);
 1115 }
 1116 
 1117 /*
 1118  * Unmount a filesystem.
 1119  *
 1120  * Note: unmount takes a path to the vnode mounted on as argument, not
 1121  * special file (as before).
 1122  */
 1123 #ifndef _SYS_SYSPROTO_H_
 1124 struct unmount_args {
 1125         char    *path;
 1126         int     flags;
 1127 };
 1128 #endif
 1129 /* ARGSUSED */
 1130 int
 1131 sys_unmount(td, uap)
 1132         struct thread *td;
 1133         register struct unmount_args /* {
 1134                 char *path;
 1135                 int flags;
 1136         } */ *uap;
 1137 {
 1138         struct nameidata nd;
 1139         struct mount *mp;
 1140         char *pathbuf;
 1141         int error, id0, id1, vfslocked;
 1142 
 1143         AUDIT_ARG_VALUE(uap->flags);
 1144         if (jailed(td->td_ucred) || usermount == 0) {
 1145                 error = priv_check(td, PRIV_VFS_UNMOUNT);
 1146                 if (error)
 1147                         return (error);
 1148         }
 1149 
 1150         pathbuf = malloc(MNAMELEN, M_TEMP, M_WAITOK);
 1151         error = copyinstr(uap->path, pathbuf, MNAMELEN, NULL);
 1152         if (error) {
 1153                 free(pathbuf, M_TEMP);
 1154                 return (error);
 1155         }
 1156         mtx_lock(&Giant);
 1157         if (uap->flags & MNT_BYFSID) {
 1158                 AUDIT_ARG_TEXT(pathbuf);
 1159                 /* Decode the filesystem ID. */
 1160                 if (sscanf(pathbuf, "FSID:%d:%d", &id0, &id1) != 2) {
 1161                         mtx_unlock(&Giant);
 1162                         free(pathbuf, M_TEMP);
 1163                         return (EINVAL);
 1164                 }
 1165 
 1166                 mtx_lock(&mountlist_mtx);
 1167                 TAILQ_FOREACH_REVERSE(mp, &mountlist, mntlist, mnt_list) {
 1168                         if (mp->mnt_stat.f_fsid.val[0] == id0 &&
 1169                             mp->mnt_stat.f_fsid.val[1] == id1)
 1170                                 break;
 1171                 }
 1172                 mtx_unlock(&mountlist_mtx);
 1173         } else {
 1174                 /*
 1175                  * Try to find global path for path argument.
 1176                  */
 1177                 NDINIT(&nd, LOOKUP,
 1178                     FOLLOW | LOCKLEAF | MPSAFE | AUDITVNODE1,
 1179                     UIO_SYSSPACE, pathbuf, td);
 1180                 if (namei(&nd) == 0) {
 1181                         vfslocked = NDHASGIANT(&nd);
 1182                         NDFREE(&nd, NDF_ONLY_PNBUF);
 1183                         error = vn_path_to_global_path(td, nd.ni_vp, pathbuf,
 1184                             MNAMELEN);
 1185                         if (error == 0 || error == ENODEV)
 1186                                 vput(nd.ni_vp);
 1187                         VFS_UNLOCK_GIANT(vfslocked);
 1188                 }
 1189                 mtx_lock(&mountlist_mtx);
 1190                 TAILQ_FOREACH_REVERSE(mp, &mountlist, mntlist, mnt_list) {
 1191                         if (strcmp(mp->mnt_stat.f_mntonname, pathbuf) == 0)
 1192                                 break;
 1193                 }
 1194                 mtx_unlock(&mountlist_mtx);
 1195         }
 1196         free(pathbuf, M_TEMP);
 1197         if (mp == NULL) {
 1198                 /*
 1199                  * Previously we returned ENOENT for a nonexistent path and
 1200                  * EINVAL for a non-mountpoint.  We cannot tell these apart
 1201                  * now, so in the !MNT_BYFSID case return the more likely
 1202                  * EINVAL for compatibility.
 1203                  */
 1204                 mtx_unlock(&Giant);
 1205                 return ((uap->flags & MNT_BYFSID) ? ENOENT : EINVAL);
 1206         }
 1207 
 1208         /*
 1209          * Don't allow unmounting the root filesystem.
 1210          */
 1211         if (mp->mnt_flag & MNT_ROOTFS) {
 1212                 mtx_unlock(&Giant);
 1213                 return (EINVAL);
 1214         }
 1215         error = dounmount(mp, uap->flags, td);
 1216         mtx_unlock(&Giant);
 1217         return (error);
 1218 }
 1219 
 1220 /*
 1221  * Do the actual filesystem unmount.
 1222  */
 1223 int
 1224 dounmount(mp, flags, td)
 1225         struct mount *mp;
 1226         int flags;
 1227         struct thread *td;
 1228 {
 1229         struct vnode *coveredvp, *fsrootvp;
 1230         int error;
 1231         uint64_t async_flag;
 1232         int mnt_gen_r;
 1233 
 1234         mtx_assert(&Giant, MA_OWNED);
 1235 
 1236         if ((coveredvp = mp->mnt_vnodecovered) != NULL) {
 1237                 mnt_gen_r = mp->mnt_gen;
 1238                 VI_LOCK(coveredvp);
 1239                 vholdl(coveredvp);
 1240                 vn_lock(coveredvp, LK_EXCLUSIVE | LK_INTERLOCK | LK_RETRY);
 1241                 vdrop(coveredvp);
 1242                 /*
 1243                  * Check for mp being unmounted while waiting for the
 1244                  * covered vnode lock.
 1245                  */
 1246                 if (coveredvp->v_mountedhere != mp ||
 1247                     coveredvp->v_mountedhere->mnt_gen != mnt_gen_r) {
 1248                         VOP_UNLOCK(coveredvp, 0);
 1249                         return (EBUSY);
 1250                 }
 1251         }
 1252         /*
 1253          * Only privileged root, or (if MNT_USER is set) the user that did the
 1254          * original mount is permitted to unmount this filesystem.
 1255          */
 1256         error = vfs_suser(mp, td);
 1257         if (error) {
 1258                 if (coveredvp)
 1259                         VOP_UNLOCK(coveredvp, 0);
 1260                 return (error);
 1261         }
 1262 
 1263         vn_start_write(NULL, &mp, V_WAIT);
 1264         MNT_ILOCK(mp);
 1265         if ((mp->mnt_kern_flag & MNTK_UNMOUNT) != 0 ||
 1266             !TAILQ_EMPTY(&mp->mnt_uppers)) {
 1267                 MNT_IUNLOCK(mp);
 1268                 if (coveredvp)
 1269                         VOP_UNLOCK(coveredvp, 0);
 1270                 vn_finished_write(mp);
 1271                 return (EBUSY);
 1272         }
 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;
 1277         error = 0;
 1278         if (mp->mnt_lockref) {
 1279                 mp->mnt_kern_flag |= MNTK_DRAINING;
 1280                 error = msleep(&mp->mnt_lockref, MNT_MTX(mp), PVFS,
 1281                     "mount drain", 0);
 1282         }
 1283         MNT_IUNLOCK(mp);
 1284         KASSERT(mp->mnt_lockref == 0,
 1285             ("%s: invalid lock refcount in the drain path @ %s:%d",
 1286             __func__, __FILE__, __LINE__));
 1287         KASSERT(error == 0,
 1288             ("%s: invalid return value for msleep in the drain path @ %s:%d",
 1289             __func__, __FILE__, __LINE__));
 1290 
 1291         if (mp->mnt_flag & MNT_EXPUBLIC)
 1292                 vfs_setpublicfs(NULL, NULL, NULL);
 1293 
 1294         vfs_msync(mp, MNT_WAIT);
 1295         MNT_ILOCK(mp);
 1296         async_flag = mp->mnt_flag & MNT_ASYNC;
 1297         mp->mnt_flag &= ~MNT_ASYNC;
 1298         mp->mnt_kern_flag &= ~MNTK_ASYNC;
 1299         MNT_IUNLOCK(mp);
 1300         cache_purgevfs(mp);     /* remove cache entries for this file sys */
 1301         vfs_deallocate_syncvnode(mp);
 1302         /*
 1303          * For forced unmounts, move process cdir/rdir refs on the fs root
 1304          * vnode to the covered vnode.  For non-forced unmounts we want
 1305          * such references to cause an EBUSY error.
 1306          */
 1307         if ((flags & MNT_FORCE) &&
 1308             VFS_ROOT(mp, LK_EXCLUSIVE, &fsrootvp) == 0) {
 1309                 if (mp->mnt_vnodecovered != NULL &&
 1310                     (mp->mnt_flag & MNT_IGNORE) == 0)
 1311                         mountcheckdirs(fsrootvp, mp->mnt_vnodecovered);
 1312                 if (fsrootvp == rootvnode) {
 1313                         vrele(rootvnode);
 1314                         rootvnode = NULL;
 1315                 }
 1316                 vput(fsrootvp);
 1317         }
 1318         if (((mp->mnt_flag & MNT_RDONLY) ||
 1319              (error = VFS_SYNC(mp, MNT_WAIT)) == 0) || (flags & MNT_FORCE) != 0)
 1320                 error = VFS_UNMOUNT(mp, flags);
 1321         vn_finished_write(mp);
 1322         /*
 1323          * If we failed to flush the dirty blocks for this mount point,
 1324          * undo all the cdir/rdir and rootvnode changes we made above.
 1325          * Unless we failed to do so because the device is reporting that
 1326          * it doesn't exist anymore.
 1327          */
 1328         if (error && error != ENXIO) {
 1329                 if ((flags & MNT_FORCE) &&
 1330                     VFS_ROOT(mp, LK_EXCLUSIVE, &fsrootvp) == 0) {
 1331                         if (mp->mnt_vnodecovered != NULL &&
 1332                             (mp->mnt_flag & MNT_IGNORE) == 0)
 1333                                 mountcheckdirs(mp->mnt_vnodecovered, fsrootvp);
 1334                         if (rootvnode == NULL) {
 1335                                 rootvnode = fsrootvp;
 1336                                 vref(rootvnode);
 1337                         }
 1338                         vput(fsrootvp);
 1339                 }
 1340                 MNT_ILOCK(mp);
 1341                 mp->mnt_kern_flag &= ~MNTK_NOINSMNTQ;
 1342                 if ((mp->mnt_flag & MNT_RDONLY) == 0) {
 1343                         MNT_IUNLOCK(mp);
 1344                         vfs_allocate_syncvnode(mp);
 1345                         MNT_ILOCK(mp);
 1346                 }
 1347                 mp->mnt_kern_flag &= ~(MNTK_UNMOUNT | MNTK_UNMOUNTF);
 1348                 mp->mnt_flag |= async_flag;
 1349                 if ((mp->mnt_flag & MNT_ASYNC) != 0 &&
 1350                     (mp->mnt_kern_flag & MNTK_NOASYNC) == 0)
 1351                         mp->mnt_kern_flag |= MNTK_ASYNC;
 1352                 if (mp->mnt_kern_flag & MNTK_MWAIT) {
 1353                         mp->mnt_kern_flag &= ~MNTK_MWAIT;
 1354                         wakeup(mp);
 1355                 }
 1356                 MNT_IUNLOCK(mp);
 1357                 if (coveredvp)
 1358                         VOP_UNLOCK(coveredvp, 0);
 1359                 return (error);
 1360         }
 1361         mtx_lock(&mountlist_mtx);
 1362         TAILQ_REMOVE(&mountlist, mp, mnt_list);
 1363         mtx_unlock(&mountlist_mtx);
 1364         if (coveredvp != NULL) {
 1365                 coveredvp->v_mountedhere = NULL;
 1366                 vput(coveredvp);
 1367         }
 1368         vfs_event_signal(NULL, VQ_UNMOUNT, 0);
 1369         vfs_mount_destroy(mp);
 1370         return (0);
 1371 }
 1372 
 1373 /*
 1374  * Report errors during filesystem mounting.
 1375  */
 1376 void
 1377 vfs_mount_error(struct mount *mp, const char *fmt, ...)
 1378 {
 1379         struct vfsoptlist *moptlist = mp->mnt_optnew;
 1380         va_list ap;
 1381         int error, len;
 1382         char *errmsg;
 1383 
 1384         error = vfs_getopt(moptlist, "errmsg", (void **)&errmsg, &len);
 1385         if (error || errmsg == NULL || len <= 0)
 1386                 return;
 1387 
 1388         va_start(ap, fmt);
 1389         vsnprintf(errmsg, (size_t)len, fmt, ap);
 1390         va_end(ap);
 1391 }
 1392 
 1393 void
 1394 vfs_opterror(struct vfsoptlist *opts, const char *fmt, ...)
 1395 {
 1396         va_list ap;
 1397         int error, len;
 1398         char *errmsg;
 1399 
 1400         error = vfs_getopt(opts, "errmsg", (void **)&errmsg, &len);
 1401         if (error || errmsg == NULL || len <= 0)
 1402                 return;
 1403 
 1404         va_start(ap, fmt);
 1405         vsnprintf(errmsg, (size_t)len, fmt, ap);
 1406         va_end(ap);
 1407 }
 1408 
 1409 /*
 1410  * ---------------------------------------------------------------------
 1411  * Functions for querying mount options/arguments from filesystems.
 1412  */
 1413 
 1414 /*
 1415  * Check that no unknown options are given
 1416  */
 1417 int
 1418 vfs_filteropt(struct vfsoptlist *opts, const char **legal)
 1419 {
 1420         struct vfsopt *opt;
 1421         char errmsg[255];
 1422         const char **t, *p, *q;
 1423         int ret = 0;
 1424 
 1425         TAILQ_FOREACH(opt, opts, link) {
 1426                 p = opt->name;
 1427                 q = NULL;
 1428                 if (p[0] == 'n' && p[1] == 'o')
 1429                         q = p + 2;
 1430                 for(t = global_opts; *t != NULL; t++) {
 1431                         if (strcmp(*t, p) == 0)
 1432                                 break;
 1433                         if (q != NULL) {
 1434                                 if (strcmp(*t, q) == 0)
 1435                                         break;
 1436                         }
 1437                 }
 1438                 if (*t != NULL)
 1439                         continue;
 1440                 for(t = legal; *t != NULL; t++) {
 1441                         if (strcmp(*t, p) == 0)
 1442                                 break;
 1443                         if (q != NULL) {
 1444                                 if (strcmp(*t, q) == 0)
 1445                                         break;
 1446                         }
 1447                 }
 1448                 if (*t != NULL)
 1449                         continue;
 1450                 snprintf(errmsg, sizeof(errmsg),
 1451                     "mount option <%s> is unknown", p);
 1452                 ret = EINVAL;
 1453         }
 1454         if (ret != 0) {
 1455                 TAILQ_FOREACH(opt, opts, link) {
 1456                         if (strcmp(opt->name, "errmsg") == 0) {
 1457                                 strncpy((char *)opt->value, errmsg, opt->len);
 1458                                 break;
 1459                         }
 1460                 }
 1461                 if (opt == NULL)
 1462                         printf("%s\n", errmsg);
 1463         }
 1464         return (ret);
 1465 }
 1466 
 1467 /*
 1468  * Get a mount option by its name.
 1469  *
 1470  * Return 0 if the option was found, ENOENT otherwise.
 1471  * If len is non-NULL it will be filled with the length
 1472  * of the option. If buf is non-NULL, it will be filled
 1473  * with the address of the option.
 1474  */
 1475 int
 1476 vfs_getopt(opts, name, buf, len)
 1477         struct vfsoptlist *opts;
 1478         const char *name;
 1479         void **buf;
 1480         int *len;
 1481 {
 1482         struct vfsopt *opt;
 1483 
 1484         KASSERT(opts != NULL, ("vfs_getopt: caller passed 'opts' as NULL"));
 1485 
 1486         TAILQ_FOREACH(opt, opts, link) {
 1487                 if (strcmp(name, opt->name) == 0) {
 1488                         opt->seen = 1;
 1489                         if (len != NULL)
 1490                                 *len = opt->len;
 1491                         if (buf != NULL)
 1492                                 *buf = opt->value;
 1493                         return (0);
 1494                 }
 1495         }
 1496         return (ENOENT);
 1497 }
 1498 
 1499 int
 1500 vfs_getopt_pos(struct vfsoptlist *opts, const char *name)
 1501 {
 1502         struct vfsopt *opt;
 1503 
 1504         if (opts == NULL)
 1505                 return (-1);
 1506 
 1507         TAILQ_FOREACH(opt, opts, link) {
 1508                 if (strcmp(name, opt->name) == 0) {
 1509                         opt->seen = 1;
 1510                         return (opt->pos);
 1511                 }
 1512         }
 1513         return (-1);
 1514 }
 1515 
 1516 char *
 1517 vfs_getopts(struct vfsoptlist *opts, const char *name, int *error)
 1518 {
 1519         struct vfsopt *opt;
 1520 
 1521         *error = 0;
 1522         TAILQ_FOREACH(opt, opts, link) {
 1523                 if (strcmp(name, opt->name) != 0)
 1524                         continue;
 1525                 opt->seen = 1;
 1526                 if (opt->len == 0 ||
 1527                     ((char *)opt->value)[opt->len - 1] != '\0') {
 1528                         *error = EINVAL;
 1529                         return (NULL);
 1530                 }
 1531                 return (opt->value);
 1532         }
 1533         *error = ENOENT;
 1534         return (NULL);
 1535 }
 1536 
 1537 int
 1538 vfs_flagopt(struct vfsoptlist *opts, const char *name, uint64_t *w,
 1539         uint64_t val)
 1540 {
 1541         struct vfsopt *opt;
 1542 
 1543         TAILQ_FOREACH(opt, opts, link) {
 1544                 if (strcmp(name, opt->name) == 0) {
 1545                         opt->seen = 1;
 1546                         if (w != NULL)
 1547                                 *w |= val;
 1548                         return (1);
 1549                 }
 1550         }
 1551         if (w != NULL)
 1552                 *w &= ~val;
 1553         return (0);
 1554 }
 1555 
 1556 int
 1557 vfs_scanopt(struct vfsoptlist *opts, const char *name, const char *fmt, ...)
 1558 {
 1559         va_list ap;
 1560         struct vfsopt *opt;
 1561         int ret;
 1562 
 1563         KASSERT(opts != NULL, ("vfs_getopt: caller passed 'opts' as NULL"));
 1564 
 1565         TAILQ_FOREACH(opt, opts, link) {
 1566                 if (strcmp(name, opt->name) != 0)
 1567                         continue;
 1568                 opt->seen = 1;
 1569                 if (opt->len == 0 || opt->value == NULL)
 1570                         return (0);
 1571                 if (((char *)opt->value)[opt->len - 1] != '\0')
 1572                         return (0);
 1573                 va_start(ap, fmt);
 1574                 ret = vsscanf(opt->value, fmt, ap);
 1575                 va_end(ap);
 1576                 return (ret);
 1577         }
 1578         return (0);
 1579 }
 1580 
 1581 int
 1582 vfs_setopt(struct vfsoptlist *opts, const char *name, void *value, int len)
 1583 {
 1584         struct vfsopt *opt;
 1585 
 1586         TAILQ_FOREACH(opt, opts, link) {
 1587                 if (strcmp(name, opt->name) != 0)
 1588                         continue;
 1589                 opt->seen = 1;
 1590                 if (opt->value == NULL)
 1591                         opt->len = len;
 1592                 else {
 1593                         if (opt->len != len)
 1594                                 return (EINVAL);
 1595                         bcopy(value, opt->value, len);
 1596                 }
 1597                 return (0);
 1598         }
 1599         return (ENOENT);
 1600 }
 1601 
 1602 int
 1603 vfs_setopt_part(struct vfsoptlist *opts, const char *name, void *value, int len)
 1604 {
 1605         struct vfsopt *opt;
 1606 
 1607         TAILQ_FOREACH(opt, opts, link) {
 1608                 if (strcmp(name, opt->name) != 0)
 1609                         continue;
 1610                 opt->seen = 1;
 1611                 if (opt->value == NULL)
 1612                         opt->len = len;
 1613                 else {
 1614                         if (opt->len < len)
 1615                                 return (EINVAL);
 1616                         opt->len = len;
 1617                         bcopy(value, opt->value, len);
 1618                 }
 1619                 return (0);
 1620         }
 1621         return (ENOENT);
 1622 }
 1623 
 1624 int
 1625 vfs_setopts(struct vfsoptlist *opts, const char *name, const char *value)
 1626 {
 1627         struct vfsopt *opt;
 1628 
 1629         TAILQ_FOREACH(opt, opts, link) {
 1630                 if (strcmp(name, opt->name) != 0)
 1631                         continue;
 1632                 opt->seen = 1;
 1633                 if (opt->value == NULL)
 1634                         opt->len = strlen(value) + 1;
 1635                 else if (strlcpy(opt->value, value, opt->len) >= opt->len)
 1636                         return (EINVAL);
 1637                 return (0);
 1638         }
 1639         return (ENOENT);
 1640 }
 1641 
 1642 /*
 1643  * Find and copy a mount option.
 1644  *
 1645  * The size of the buffer has to be specified
 1646  * in len, if it is not the same length as the
 1647  * mount option, EINVAL is returned.
 1648  * Returns ENOENT if the option is not found.
 1649  */
 1650 int
 1651 vfs_copyopt(opts, name, dest, len)
 1652         struct vfsoptlist *opts;
 1653         const char *name;
 1654         void *dest;
 1655         int len;
 1656 {
 1657         struct vfsopt *opt;
 1658 
 1659         KASSERT(opts != NULL, ("vfs_copyopt: caller passed 'opts' as NULL"));
 1660 
 1661         TAILQ_FOREACH(opt, opts, link) {
 1662                 if (strcmp(name, opt->name) == 0) {
 1663                         opt->seen = 1;
 1664                         if (len != opt->len)
 1665                                 return (EINVAL);
 1666                         bcopy(opt->value, dest, opt->len);
 1667                         return (0);
 1668                 }
 1669         }
 1670         return (ENOENT);
 1671 }
 1672 
 1673 /*
 1674  * These are helper functions for filesystems to traverse all
 1675  * their vnodes.  See MNT_VNODE_FOREACH() in sys/mount.h.
 1676  *
 1677  * This interface has been deprecated in favor of MNT_VNODE_FOREACH_ALL.
 1678  */
 1679 
 1680 MALLOC_DECLARE(M_VNODE_MARKER);
 1681 
 1682 struct vnode *
 1683 __mnt_vnode_next(struct vnode **mvp, struct mount *mp)
 1684 {
 1685         struct vnode *vp;
 1686 
 1687         mtx_assert(MNT_MTX(mp), MA_OWNED);
 1688 
 1689         KASSERT((*mvp)->v_mount == mp, ("marker vnode mount list mismatch"));
 1690         if (should_yield()) {
 1691                 MNT_IUNLOCK(mp);
 1692                 kern_yield(PRI_UNCHANGED);
 1693                 MNT_ILOCK(mp);
 1694         }
 1695         vp = TAILQ_NEXT(*mvp, v_nmntvnodes);
 1696         while (vp != NULL && vp->v_type == VMARKER)
 1697                 vp = TAILQ_NEXT(vp, v_nmntvnodes);
 1698 
 1699         /* Check if we are done */
 1700         if (vp == NULL) {
 1701                 __mnt_vnode_markerfree(mvp, mp);
 1702                 return (NULL);
 1703         }
 1704         TAILQ_REMOVE(&mp->mnt_nvnodelist, *mvp, v_nmntvnodes);
 1705         TAILQ_INSERT_AFTER(&mp->mnt_nvnodelist, vp, *mvp, v_nmntvnodes);
 1706         return (vp);
 1707 }
 1708 
 1709 struct vnode *
 1710 __mnt_vnode_first(struct vnode **mvp, struct mount *mp)
 1711 {
 1712         struct vnode *vp;
 1713 
 1714         mtx_assert(MNT_MTX(mp), MA_OWNED);
 1715 
 1716         vp = TAILQ_FIRST(&mp->mnt_nvnodelist);
 1717         while (vp != NULL && vp->v_type == VMARKER)
 1718                 vp = TAILQ_NEXT(vp, v_nmntvnodes);
 1719 
 1720         /* Check if we are done */
 1721         if (vp == NULL) {
 1722                 *mvp = NULL;
 1723                 return (NULL);
 1724         }
 1725         MNT_REF(mp);
 1726         MNT_IUNLOCK(mp);
 1727         *mvp = (struct vnode *) malloc(sizeof(struct vnode),
 1728                                        M_VNODE_MARKER,
 1729                                        M_WAITOK | M_ZERO);
 1730         MNT_ILOCK(mp);
 1731         (*mvp)->v_type = VMARKER;
 1732 
 1733         vp = TAILQ_FIRST(&mp->mnt_nvnodelist);
 1734         while (vp != NULL && vp->v_type == VMARKER)
 1735                 vp = TAILQ_NEXT(vp, v_nmntvnodes);
 1736 
 1737         /* Check if we are done */
 1738         if (vp == NULL) {
 1739                 MNT_IUNLOCK(mp);
 1740                 free(*mvp, M_VNODE_MARKER);
 1741                 MNT_ILOCK(mp);
 1742                 *mvp = NULL;
 1743                 MNT_REL(mp);
 1744                 return (NULL);
 1745         }
 1746         (*mvp)->v_mount = mp;
 1747         TAILQ_INSERT_AFTER(&mp->mnt_nvnodelist, vp, *mvp, v_nmntvnodes);
 1748         return (vp);
 1749 }
 1750 
 1751 
 1752 void
 1753 __mnt_vnode_markerfree(struct vnode **mvp, struct mount *mp)
 1754 {
 1755 
 1756         if (*mvp == NULL)
 1757                 return;
 1758 
 1759         mtx_assert(MNT_MTX(mp), MA_OWNED);
 1760 
 1761         KASSERT((*mvp)->v_mount == mp, ("marker vnode mount list mismatch"));
 1762         TAILQ_REMOVE(&mp->mnt_nvnodelist, *mvp, v_nmntvnodes);
 1763         MNT_IUNLOCK(mp);
 1764         free(*mvp, M_VNODE_MARKER);
 1765         MNT_ILOCK(mp);
 1766         *mvp = NULL;
 1767         MNT_REL(mp);
 1768 }
 1769 
 1770 int
 1771 __vfs_statfs(struct mount *mp, struct statfs *sbp)
 1772 {
 1773         int error;
 1774 
 1775         error = mp->mnt_op->vfs_statfs(mp, &mp->mnt_stat);
 1776         if (sbp != &mp->mnt_stat)
 1777                 *sbp = mp->mnt_stat;
 1778         return (error);
 1779 }
 1780 
 1781 void
 1782 vfs_mountedfrom(struct mount *mp, const char *from)
 1783 {
 1784 
 1785         bzero(mp->mnt_stat.f_mntfromname, sizeof mp->mnt_stat.f_mntfromname);
 1786         strlcpy(mp->mnt_stat.f_mntfromname, from,
 1787             sizeof mp->mnt_stat.f_mntfromname);
 1788 }
 1789 
 1790 /*
 1791  * ---------------------------------------------------------------------
 1792  * This is the api for building mount args and mounting filesystems from
 1793  * inside the kernel.
 1794  *
 1795  * The API works by accumulation of individual args.  First error is
 1796  * latched.
 1797  *
 1798  * XXX: should be documented in new manpage kernel_mount(9)
 1799  */
 1800 
 1801 /* A memory allocation which must be freed when we are done */
 1802 struct mntaarg {
 1803         SLIST_ENTRY(mntaarg)    next;
 1804 };
 1805 
 1806 /* The header for the mount arguments */
 1807 struct mntarg {
 1808         struct iovec *v;
 1809         int len;
 1810         int error;
 1811         SLIST_HEAD(, mntaarg)   list;
 1812 };
 1813 
 1814 /*
 1815  * Add a boolean argument.
 1816  *
 1817  * flag is the boolean value.
 1818  * name must start with "no".
 1819  */
 1820 struct mntarg *
 1821 mount_argb(struct mntarg *ma, int flag, const char *name)
 1822 {
 1823 
 1824         KASSERT(name[0] == 'n' && name[1] == 'o',
 1825             ("mount_argb(...,%s): name must start with 'no'", name));
 1826 
 1827         return (mount_arg(ma, name + (flag ? 2 : 0), NULL, 0));
 1828 }
 1829 
 1830 /*
 1831  * Add an argument printf style
 1832  */
 1833 struct mntarg *
 1834 mount_argf(struct mntarg *ma, const char *name, const char *fmt, ...)
 1835 {
 1836         va_list ap;
 1837         struct mntaarg *maa;
 1838         struct sbuf *sb;
 1839         int len;
 1840 
 1841         if (ma == NULL) {
 1842                 ma = malloc(sizeof *ma, M_MOUNT, M_WAITOK | M_ZERO);
 1843                 SLIST_INIT(&ma->list);
 1844         }
 1845         if (ma->error)
 1846                 return (ma);
 1847 
 1848         ma->v = realloc(ma->v, sizeof *ma->v * (ma->len + 2),
 1849             M_MOUNT, M_WAITOK);
 1850         ma->v[ma->len].iov_base = (void *)(uintptr_t)name;
 1851         ma->v[ma->len].iov_len = strlen(name) + 1;
 1852         ma->len++;
 1853 
 1854         sb = sbuf_new_auto();
 1855         va_start(ap, fmt);
 1856         sbuf_vprintf(sb, fmt, ap);
 1857         va_end(ap);
 1858         sbuf_finish(sb);
 1859         len = sbuf_len(sb) + 1;
 1860         maa = malloc(sizeof *maa + len, M_MOUNT, M_WAITOK | M_ZERO);
 1861         SLIST_INSERT_HEAD(&ma->list, maa, next);
 1862         bcopy(sbuf_data(sb), maa + 1, len);
 1863         sbuf_delete(sb);
 1864 
 1865         ma->v[ma->len].iov_base = maa + 1;
 1866         ma->v[ma->len].iov_len = len;
 1867         ma->len++;
 1868 
 1869         return (ma);
 1870 }
 1871 
 1872 /*
 1873  * Add an argument which is a userland string.
 1874  */
 1875 struct mntarg *
 1876 mount_argsu(struct mntarg *ma, const char *name, const void *val, int len)
 1877 {
 1878         struct mntaarg *maa;
 1879         char *tbuf;
 1880 
 1881         if (val == NULL)
 1882                 return (ma);
 1883         if (ma == NULL) {
 1884                 ma = malloc(sizeof *ma, M_MOUNT, M_WAITOK | M_ZERO);
 1885                 SLIST_INIT(&ma->list);
 1886         }
 1887         if (ma->error)
 1888                 return (ma);
 1889         maa = malloc(sizeof *maa + len, M_MOUNT, M_WAITOK | M_ZERO);
 1890         SLIST_INSERT_HEAD(&ma->list, maa, next);
 1891         tbuf = (void *)(maa + 1);
 1892         ma->error = copyinstr(val, tbuf, len, NULL);
 1893         return (mount_arg(ma, name, tbuf, -1));
 1894 }
 1895 
 1896 /*
 1897  * Plain argument.
 1898  *
 1899  * If length is -1, treat value as a C string.
 1900  */
 1901 struct mntarg *
 1902 mount_arg(struct mntarg *ma, const char *name, const void *val, int len)
 1903 {
 1904 
 1905         if (ma == NULL) {
 1906                 ma = malloc(sizeof *ma, M_MOUNT, M_WAITOK | M_ZERO);
 1907                 SLIST_INIT(&ma->list);
 1908         }
 1909         if (ma->error)
 1910                 return (ma);
 1911 
 1912         ma->v = realloc(ma->v, sizeof *ma->v * (ma->len + 2),
 1913             M_MOUNT, M_WAITOK);
 1914         ma->v[ma->len].iov_base = (void *)(uintptr_t)name;
 1915         ma->v[ma->len].iov_len = strlen(name) + 1;
 1916         ma->len++;
 1917 
 1918         ma->v[ma->len].iov_base = (void *)(uintptr_t)val;
 1919         if (len < 0)
 1920                 ma->v[ma->len].iov_len = strlen(val) + 1;
 1921         else
 1922                 ma->v[ma->len].iov_len = len;
 1923         ma->len++;
 1924         return (ma);
 1925 }
 1926 
 1927 /*
 1928  * Free a mntarg structure
 1929  */
 1930 static void
 1931 free_mntarg(struct mntarg *ma)
 1932 {
 1933         struct mntaarg *maa;
 1934 
 1935         while (!SLIST_EMPTY(&ma->list)) {
 1936                 maa = SLIST_FIRST(&ma->list);
 1937                 SLIST_REMOVE_HEAD(&ma->list, next);
 1938                 free(maa, M_MOUNT);
 1939         }
 1940         free(ma->v, M_MOUNT);
 1941         free(ma, M_MOUNT);
 1942 }
 1943 
 1944 /*
 1945  * Mount a filesystem
 1946  */
 1947 int
 1948 kernel_mount(struct mntarg *ma, uint64_t flags)
 1949 {
 1950         struct uio auio;
 1951         int error;
 1952 
 1953         KASSERT(ma != NULL, ("kernel_mount NULL ma"));
 1954         KASSERT(ma->v != NULL, ("kernel_mount NULL ma->v"));
 1955         KASSERT(!(ma->len & 1), ("kernel_mount odd ma->len (%d)", ma->len));
 1956 
 1957         auio.uio_iov = ma->v;
 1958         auio.uio_iovcnt = ma->len;
 1959         auio.uio_segflg = UIO_SYSSPACE;
 1960 
 1961         error = ma->error;
 1962         if (!error)
 1963                 error = vfs_donmount(curthread, flags, &auio);
 1964         free_mntarg(ma);
 1965         return (error);
 1966 }
 1967 
 1968 /*
 1969  * A printflike function to mount a filesystem.
 1970  */
 1971 int
 1972 kernel_vmount(int flags, ...)
 1973 {
 1974         struct mntarg *ma = NULL;
 1975         va_list ap;
 1976         const char *cp;
 1977         const void *vp;
 1978         int error;
 1979 
 1980         va_start(ap, flags);
 1981         for (;;) {
 1982                 cp = va_arg(ap, const char *);
 1983                 if (cp == NULL)
 1984                         break;
 1985                 vp = va_arg(ap, const void *);
 1986                 ma = mount_arg(ma, cp, vp, (vp != NULL ? -1 : 0));
 1987         }
 1988         va_end(ap);
 1989 
 1990         error = kernel_mount(ma, flags);
 1991         return (error);
 1992 }
 1993 
 1994 void
 1995 vfs_oexport_conv(const struct oexport_args *oexp, struct export_args *exp)
 1996 {
 1997 
 1998         bcopy(oexp, exp, sizeof(*oexp));
 1999         exp->ex_numsecflavors = 0;
 2000 }

Cache object: ff5f007199c33ecb4d7331e44e3505ed


[ source navigation ] [ diff markup ] [ identifier search ] [ freetext search ] [ file search ] [ list types ] [ track identifier ]


This page is part of the FreeBSD/Linux Linux Kernel Cross-Reference, and was automatically generated using a modified version of the LXR engine.