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: releng/9.1/sys/kern/vfs_mount.c 235626 2012-05-18 19:48:38Z mckusick $");
   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         return (mp);
  483 }
  484 
  485 /*
  486  * Destroy the mount struct previously allocated by vfs_mount_alloc().
  487  */
  488 void
  489 vfs_mount_destroy(struct mount *mp)
  490 {
  491 
  492         MNT_ILOCK(mp);
  493         mp->mnt_kern_flag |= MNTK_REFEXPIRE;
  494         if (mp->mnt_kern_flag & MNTK_MWAIT) {
  495                 mp->mnt_kern_flag &= ~MNTK_MWAIT;
  496                 wakeup(mp);
  497         }
  498         while (mp->mnt_ref)
  499                 msleep(mp, MNT_MTX(mp), PVFS, "mntref", 0);
  500         KASSERT(mp->mnt_ref == 0,
  501             ("%s: invalid refcount in the drain path @ %s:%d", __func__,
  502             __FILE__, __LINE__));
  503         if (mp->mnt_writeopcount != 0)
  504                 panic("vfs_mount_destroy: nonzero writeopcount");
  505         if (mp->mnt_secondary_writes != 0)
  506                 panic("vfs_mount_destroy: nonzero secondary_writes");
  507         mp->mnt_vfc->vfc_refcount--;
  508         if (!TAILQ_EMPTY(&mp->mnt_nvnodelist)) {
  509                 struct vnode *vp;
  510 
  511                 TAILQ_FOREACH(vp, &mp->mnt_nvnodelist, v_nmntvnodes)
  512                         vprint("", vp);
  513                 panic("unmount: dangling vnode");
  514         }
  515         if (mp->mnt_nvnodelistsize != 0)
  516                 panic("vfs_mount_destroy: nonzero nvnodelistsize");
  517         if (mp->mnt_activevnodelistsize != 0)
  518                 panic("vfs_mount_destroy: nonzero activevnodelistsize");
  519         if (mp->mnt_lockref != 0)
  520                 panic("vfs_mount_destroy: nonzero lock refcount");
  521         MNT_IUNLOCK(mp);
  522 #ifdef MAC
  523         mac_mount_destroy(mp);
  524 #endif
  525         if (mp->mnt_opt != NULL)
  526                 vfs_freeopts(mp->mnt_opt);
  527         crfree(mp->mnt_cred);
  528         uma_zfree(mount_zone, mp);
  529 }
  530 
  531 int
  532 vfs_donmount(struct thread *td, uint64_t fsflags, struct uio *fsoptions)
  533 {
  534         struct vfsoptlist *optlist;
  535         struct vfsopt *opt, *tmp_opt;
  536         char *fstype, *fspath, *errmsg;
  537         int error, fstypelen, fspathlen, errmsg_len, errmsg_pos;
  538 
  539         errmsg = fspath = NULL;
  540         errmsg_len = fspathlen = 0;
  541         errmsg_pos = -1;
  542 
  543         error = vfs_buildopts(fsoptions, &optlist);
  544         if (error)
  545                 return (error);
  546 
  547         if (vfs_getopt(optlist, "errmsg", (void **)&errmsg, &errmsg_len) == 0)
  548                 errmsg_pos = vfs_getopt_pos(optlist, "errmsg");
  549 
  550         /*
  551          * We need these two options before the others,
  552          * and they are mandatory for any filesystem.
  553          * Ensure they are NUL terminated as well.
  554          */
  555         fstypelen = 0;
  556         error = vfs_getopt(optlist, "fstype", (void **)&fstype, &fstypelen);
  557         if (error || fstype[fstypelen - 1] != '\0') {
  558                 error = EINVAL;
  559                 if (errmsg != NULL)
  560                         strncpy(errmsg, "Invalid fstype", errmsg_len);
  561                 goto bail;
  562         }
  563         fspathlen = 0;
  564         error = vfs_getopt(optlist, "fspath", (void **)&fspath, &fspathlen);
  565         if (error || fspath[fspathlen - 1] != '\0') {
  566                 error = EINVAL;
  567                 if (errmsg != NULL)
  568                         strncpy(errmsg, "Invalid fspath", errmsg_len);
  569                 goto bail;
  570         }
  571 
  572         /*
  573          * We need to see if we have the "update" option
  574          * before we call vfs_domount(), since vfs_domount() has special
  575          * logic based on MNT_UPDATE.  This is very important
  576          * when we want to update the root filesystem.
  577          */
  578         TAILQ_FOREACH_SAFE(opt, optlist, link, tmp_opt) {
  579                 if (strcmp(opt->name, "update") == 0) {
  580                         fsflags |= MNT_UPDATE;
  581                         vfs_freeopt(optlist, opt);
  582                 }
  583                 else if (strcmp(opt->name, "async") == 0)
  584                         fsflags |= MNT_ASYNC;
  585                 else if (strcmp(opt->name, "force") == 0) {
  586                         fsflags |= MNT_FORCE;
  587                         vfs_freeopt(optlist, opt);
  588                 }
  589                 else if (strcmp(opt->name, "reload") == 0) {
  590                         fsflags |= MNT_RELOAD;
  591                         vfs_freeopt(optlist, opt);
  592                 }
  593                 else if (strcmp(opt->name, "multilabel") == 0)
  594                         fsflags |= MNT_MULTILABEL;
  595                 else if (strcmp(opt->name, "noasync") == 0)
  596                         fsflags &= ~MNT_ASYNC;
  597                 else if (strcmp(opt->name, "noatime") == 0)
  598                         fsflags |= MNT_NOATIME;
  599                 else if (strcmp(opt->name, "atime") == 0) {
  600                         free(opt->name, M_MOUNT);
  601                         opt->name = strdup("nonoatime", M_MOUNT);
  602                 }
  603                 else if (strcmp(opt->name, "noclusterr") == 0)
  604                         fsflags |= MNT_NOCLUSTERR;
  605                 else if (strcmp(opt->name, "clusterr") == 0) {
  606                         free(opt->name, M_MOUNT);
  607                         opt->name = strdup("nonoclusterr", M_MOUNT);
  608                 }
  609                 else if (strcmp(opt->name, "noclusterw") == 0)
  610                         fsflags |= MNT_NOCLUSTERW;
  611                 else if (strcmp(opt->name, "clusterw") == 0) {
  612                         free(opt->name, M_MOUNT);
  613                         opt->name = strdup("nonoclusterw", M_MOUNT);
  614                 }
  615                 else if (strcmp(opt->name, "noexec") == 0)
  616                         fsflags |= MNT_NOEXEC;
  617                 else if (strcmp(opt->name, "exec") == 0) {
  618                         free(opt->name, M_MOUNT);
  619                         opt->name = strdup("nonoexec", M_MOUNT);
  620                 }
  621                 else if (strcmp(opt->name, "nosuid") == 0)
  622                         fsflags |= MNT_NOSUID;
  623                 else if (strcmp(opt->name, "suid") == 0) {
  624                         free(opt->name, M_MOUNT);
  625                         opt->name = strdup("nonosuid", M_MOUNT);
  626                 }
  627                 else if (strcmp(opt->name, "nosymfollow") == 0)
  628                         fsflags |= MNT_NOSYMFOLLOW;
  629                 else if (strcmp(opt->name, "symfollow") == 0) {
  630                         free(opt->name, M_MOUNT);
  631                         opt->name = strdup("nonosymfollow", M_MOUNT);
  632                 }
  633                 else if (strcmp(opt->name, "noro") == 0)
  634                         fsflags &= ~MNT_RDONLY;
  635                 else if (strcmp(opt->name, "rw") == 0)
  636                         fsflags &= ~MNT_RDONLY;
  637                 else if (strcmp(opt->name, "ro") == 0)
  638                         fsflags |= MNT_RDONLY;
  639                 else if (strcmp(opt->name, "rdonly") == 0) {
  640                         free(opt->name, M_MOUNT);
  641                         opt->name = strdup("ro", M_MOUNT);
  642                         fsflags |= MNT_RDONLY;
  643                 }
  644                 else if (strcmp(opt->name, "suiddir") == 0)
  645                         fsflags |= MNT_SUIDDIR;
  646                 else if (strcmp(opt->name, "sync") == 0)
  647                         fsflags |= MNT_SYNCHRONOUS;
  648                 else if (strcmp(opt->name, "union") == 0)
  649                         fsflags |= MNT_UNION;
  650         }
  651 
  652         /*
  653          * Be ultra-paranoid about making sure the type and fspath
  654          * variables will fit in our mp buffers, including the
  655          * terminating NUL.
  656          */
  657         if (fstypelen >= MFSNAMELEN - 1 || fspathlen >= MNAMELEN - 1) {
  658                 error = ENAMETOOLONG;
  659                 goto bail;
  660         }
  661 
  662         error = vfs_domount(td, fstype, fspath, fsflags, &optlist);
  663 bail:
  664         /* copyout the errmsg */
  665         if (errmsg_pos != -1 && ((2 * errmsg_pos + 1) < fsoptions->uio_iovcnt)
  666             && errmsg_len > 0 && errmsg != NULL) {
  667                 if (fsoptions->uio_segflg == UIO_SYSSPACE) {
  668                         bcopy(errmsg,
  669                             fsoptions->uio_iov[2 * errmsg_pos + 1].iov_base,
  670                             fsoptions->uio_iov[2 * errmsg_pos + 1].iov_len);
  671                 } else {
  672                         copyout(errmsg,
  673                             fsoptions->uio_iov[2 * errmsg_pos + 1].iov_base,
  674                             fsoptions->uio_iov[2 * errmsg_pos + 1].iov_len);
  675                 }
  676         }
  677 
  678         if (optlist != NULL)
  679                 vfs_freeopts(optlist);
  680         return (error);
  681 }
  682 
  683 /*
  684  * Old mount API.
  685  */
  686 #ifndef _SYS_SYSPROTO_H_
  687 struct mount_args {
  688         char    *type;
  689         char    *path;
  690         int     flags;
  691         caddr_t data;
  692 };
  693 #endif
  694 /* ARGSUSED */
  695 int
  696 sys_mount(td, uap)
  697         struct thread *td;
  698         struct mount_args /* {
  699                 char *type;
  700                 char *path;
  701                 int flags;
  702                 caddr_t data;
  703         } */ *uap;
  704 {
  705         char *fstype;
  706         struct vfsconf *vfsp = NULL;
  707         struct mntarg *ma = NULL;
  708         uint64_t flags;
  709         int error;
  710 
  711         /*
  712          * Mount flags are now 64-bits. On 32-bit archtectures only
  713          * 32-bits are passed in, but from here on everything handles
  714          * 64-bit flags correctly.
  715          */
  716         flags = uap->flags;
  717 
  718         AUDIT_ARG_FFLAGS(flags);
  719 
  720         /*
  721          * Filter out MNT_ROOTFS.  We do not want clients of mount() in
  722          * userspace to set this flag, but we must filter it out if we want
  723          * MNT_UPDATE on the root file system to work.
  724          * MNT_ROOTFS should only be set by the kernel when mounting its
  725          * root file system.
  726          */
  727         flags &= ~MNT_ROOTFS;
  728 
  729         fstype = malloc(MFSNAMELEN, M_TEMP, M_WAITOK);
  730         error = copyinstr(uap->type, fstype, MFSNAMELEN, NULL);
  731         if (error) {
  732                 free(fstype, M_TEMP);
  733                 return (error);
  734         }
  735 
  736         AUDIT_ARG_TEXT(fstype);
  737         mtx_lock(&Giant);
  738         vfsp = vfs_byname_kld(fstype, td, &error);
  739         free(fstype, M_TEMP);
  740         if (vfsp == NULL) {
  741                 mtx_unlock(&Giant);
  742                 return (ENOENT);
  743         }
  744         if (vfsp->vfc_vfsops->vfs_cmount == NULL) {
  745                 mtx_unlock(&Giant);
  746                 return (EOPNOTSUPP);
  747         }
  748 
  749         ma = mount_argsu(ma, "fstype", uap->type, MNAMELEN);
  750         ma = mount_argsu(ma, "fspath", uap->path, MNAMELEN);
  751         ma = mount_argb(ma, flags & MNT_RDONLY, "noro");
  752         ma = mount_argb(ma, !(flags & MNT_NOSUID), "nosuid");
  753         ma = mount_argb(ma, !(flags & MNT_NOEXEC), "noexec");
  754 
  755         error = vfsp->vfc_vfsops->vfs_cmount(ma, uap->data, flags);
  756         mtx_unlock(&Giant);
  757         return (error);
  758 }
  759 
  760 /*
  761  * vfs_domount_first(): first file system mount (not update)
  762  */
  763 static int
  764 vfs_domount_first(
  765         struct thread *td,              /* Calling thread. */
  766         struct vfsconf *vfsp,           /* File system type. */
  767         char *fspath,                   /* Mount path. */
  768         struct vnode *vp,               /* Vnode to be covered. */
  769         uint64_t fsflags,               /* Flags common to all filesystems. */
  770         struct vfsoptlist **optlist     /* Options local to the filesystem. */
  771         )
  772 {
  773         struct vattr va;
  774         struct mount *mp;
  775         struct vnode *newdp;
  776         int error;
  777 
  778         mtx_assert(&Giant, MA_OWNED);
  779         ASSERT_VOP_ELOCKED(vp, __func__);
  780         KASSERT((fsflags & MNT_UPDATE) == 0, ("MNT_UPDATE shouldn't be here"));
  781 
  782         /*
  783          * If the user is not root, ensure that they own the directory
  784          * onto which we are attempting to mount.
  785          */
  786         error = VOP_GETATTR(vp, &va, td->td_ucred);
  787         if (error == 0 && va.va_uid != td->td_ucred->cr_uid)
  788                 error = priv_check_cred(td->td_ucred, PRIV_VFS_ADMIN, 0);
  789         if (error == 0)
  790                 error = vinvalbuf(vp, V_SAVE, 0, 0);
  791         if (error == 0 && vp->v_type != VDIR)
  792                 error = ENOTDIR;
  793         if (error == 0) {
  794                 VI_LOCK(vp);
  795                 if ((vp->v_iflag & VI_MOUNT) == 0 && vp->v_mountedhere == NULL)
  796                         vp->v_iflag |= VI_MOUNT;
  797                 else
  798                         error = EBUSY;
  799                 VI_UNLOCK(vp);
  800         }
  801         if (error != 0) {
  802                 vput(vp);
  803                 return (error);
  804         }
  805         VOP_UNLOCK(vp, 0);
  806 
  807         /* Allocate and initialize the filesystem. */
  808         mp = vfs_mount_alloc(vp, vfsp, fspath, td->td_ucred);
  809         /* XXXMAC: pass to vfs_mount_alloc? */
  810         mp->mnt_optnew = *optlist;
  811         /* Set the mount level flags. */
  812         mp->mnt_flag = (fsflags & (MNT_UPDATEMASK | MNT_ROOTFS | MNT_RDONLY));
  813 
  814         /*
  815          * Mount the filesystem.
  816          * XXX The final recipients of VFS_MOUNT just overwrite the ndp they
  817          * get.  No freeing of cn_pnbuf.
  818          */
  819         error = VFS_MOUNT(mp);
  820         if (error != 0) {
  821                 vfs_unbusy(mp);
  822                 vfs_mount_destroy(mp);
  823                 VI_LOCK(vp);
  824                 vp->v_iflag &= ~VI_MOUNT;
  825                 VI_UNLOCK(vp);
  826                 vrele(vp);
  827                 return (error);
  828         }
  829 
  830         if (mp->mnt_opt != NULL)
  831                 vfs_freeopts(mp->mnt_opt);
  832         mp->mnt_opt = mp->mnt_optnew;
  833         *optlist = NULL;
  834         (void)VFS_STATFS(mp, &mp->mnt_stat);
  835 
  836         /*
  837          * Prevent external consumers of mount options from reading mnt_optnew.
  838          */
  839         mp->mnt_optnew = NULL;
  840 
  841         MNT_ILOCK(mp);
  842         if ((mp->mnt_flag & MNT_ASYNC) != 0 &&
  843             (mp->mnt_kern_flag & MNTK_NOASYNC) == 0)
  844                 mp->mnt_kern_flag |= MNTK_ASYNC;
  845         else
  846                 mp->mnt_kern_flag &= ~MNTK_ASYNC;
  847         MNT_IUNLOCK(mp);
  848 
  849         vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
  850         cache_purge(vp);
  851         VI_LOCK(vp);
  852         vp->v_iflag &= ~VI_MOUNT;
  853         VI_UNLOCK(vp);
  854         vp->v_mountedhere = mp;
  855         /* Place the new filesystem at the end of the mount list. */
  856         mtx_lock(&mountlist_mtx);
  857         TAILQ_INSERT_TAIL(&mountlist, mp, mnt_list);
  858         mtx_unlock(&mountlist_mtx);
  859         vfs_event_signal(NULL, VQ_MOUNT, 0);
  860         if (VFS_ROOT(mp, LK_EXCLUSIVE, &newdp))
  861                 panic("mount: lost mount");
  862         VOP_UNLOCK(newdp, 0);
  863         VOP_UNLOCK(vp, 0);
  864         mountcheckdirs(vp, newdp);
  865         vrele(newdp);
  866         if ((mp->mnt_flag & MNT_RDONLY) == 0)
  867                 vfs_allocate_syncvnode(mp);
  868         vfs_unbusy(mp);
  869         return (0);
  870 }
  871 
  872 /*
  873  * vfs_domount_update(): update of mounted file system
  874  */
  875 static int
  876 vfs_domount_update(
  877         struct thread *td,              /* Calling thread. */
  878         struct vnode *vp,               /* Mount point vnode. */
  879         uint64_t fsflags,               /* Flags common to all filesystems. */
  880         struct vfsoptlist **optlist     /* Options local to the filesystem. */
  881         )
  882 {
  883         struct oexport_args oexport;
  884         struct export_args export;
  885         struct mount *mp;
  886         int error, export_error;
  887         uint64_t flag;
  888 
  889         mtx_assert(&Giant, MA_OWNED);
  890         ASSERT_VOP_ELOCKED(vp, __func__);
  891         KASSERT((fsflags & MNT_UPDATE) != 0, ("MNT_UPDATE should be here"));
  892 
  893         if ((vp->v_vflag & VV_ROOT) == 0) {
  894                 vput(vp);
  895                 return (EINVAL);
  896         }
  897         mp = vp->v_mount;
  898         /*
  899          * We only allow the filesystem to be reloaded if it
  900          * is currently mounted read-only.
  901          */
  902         flag = mp->mnt_flag;
  903         if ((fsflags & MNT_RELOAD) != 0 && (flag & MNT_RDONLY) == 0) {
  904                 vput(vp);
  905                 return (EOPNOTSUPP);    /* Needs translation */
  906         }
  907         /*
  908          * Only privileged root, or (if MNT_USER is set) the user that
  909          * did the original mount is permitted to update it.
  910          */
  911         error = vfs_suser(mp, td);
  912         if (error != 0) {
  913                 vput(vp);
  914                 return (error);
  915         }
  916         if (vfs_busy(mp, MBF_NOWAIT)) {
  917                 vput(vp);
  918                 return (EBUSY);
  919         }
  920         VI_LOCK(vp);
  921         if ((vp->v_iflag & VI_MOUNT) != 0 || vp->v_mountedhere != NULL) {
  922                 VI_UNLOCK(vp);
  923                 vfs_unbusy(mp);
  924                 vput(vp);
  925                 return (EBUSY);
  926         }
  927         vp->v_iflag |= VI_MOUNT;
  928         VI_UNLOCK(vp);
  929         VOP_UNLOCK(vp, 0);
  930 
  931         MNT_ILOCK(mp);
  932         mp->mnt_flag &= ~MNT_UPDATEMASK;
  933         mp->mnt_flag |= fsflags & (MNT_RELOAD | MNT_FORCE | MNT_UPDATE |
  934             MNT_SNAPSHOT | MNT_ROOTFS | MNT_UPDATEMASK | MNT_RDONLY);
  935         if ((mp->mnt_flag & MNT_ASYNC) == 0)
  936                 mp->mnt_kern_flag &= ~MNTK_ASYNC;
  937         MNT_IUNLOCK(mp);
  938         mp->mnt_optnew = *optlist;
  939         vfs_mergeopts(mp->mnt_optnew, mp->mnt_opt);
  940 
  941         /*
  942          * Mount the filesystem.
  943          * XXX The final recipients of VFS_MOUNT just overwrite the ndp they
  944          * get.  No freeing of cn_pnbuf.
  945          */
  946         error = VFS_MOUNT(mp);
  947 
  948         export_error = 0;
  949         if (error == 0) {
  950                 /* Process the export option. */
  951                 if (vfs_copyopt(mp->mnt_optnew, "export", &export,
  952                     sizeof(export)) == 0) {
  953                         export_error = vfs_export(mp, &export);
  954                 } else if (vfs_copyopt(mp->mnt_optnew, "export", &oexport,
  955                     sizeof(oexport)) == 0) {
  956                         export.ex_flags = oexport.ex_flags;
  957                         export.ex_root = oexport.ex_root;
  958                         export.ex_anon = oexport.ex_anon;
  959                         export.ex_addr = oexport.ex_addr;
  960                         export.ex_addrlen = oexport.ex_addrlen;
  961                         export.ex_mask = oexport.ex_mask;
  962                         export.ex_masklen = oexport.ex_masklen;
  963                         export.ex_indexfile = oexport.ex_indexfile;
  964                         export.ex_numsecflavors = 0;
  965                         export_error = vfs_export(mp, &export);
  966                 }
  967         }
  968 
  969         MNT_ILOCK(mp);
  970         if (error == 0) {
  971                 mp->mnt_flag &= ~(MNT_UPDATE | MNT_RELOAD | MNT_FORCE |
  972                     MNT_SNAPSHOT);
  973         } else {
  974                 /*
  975                  * If we fail, restore old mount flags. MNT_QUOTA is special,
  976                  * because it is not part of MNT_UPDATEMASK, but it could have
  977                  * changed in the meantime if quotactl(2) was called.
  978                  * All in all we want current value of MNT_QUOTA, not the old
  979                  * one.
  980                  */
  981                 mp->mnt_flag = (mp->mnt_flag & MNT_QUOTA) | (flag & ~MNT_QUOTA);
  982         }
  983         if ((mp->mnt_flag & MNT_ASYNC) != 0 &&
  984             (mp->mnt_kern_flag & MNTK_NOASYNC) == 0)
  985                 mp->mnt_kern_flag |= MNTK_ASYNC;
  986         else
  987                 mp->mnt_kern_flag &= ~MNTK_ASYNC;
  988         MNT_IUNLOCK(mp);
  989 
  990         if (error != 0)
  991                 goto end;
  992 
  993         if (mp->mnt_opt != NULL)
  994                 vfs_freeopts(mp->mnt_opt);
  995         mp->mnt_opt = mp->mnt_optnew;
  996         *optlist = NULL;
  997         (void)VFS_STATFS(mp, &mp->mnt_stat);
  998         /*
  999          * Prevent external consumers of mount options from reading
 1000          * mnt_optnew.
 1001          */
 1002         mp->mnt_optnew = NULL;
 1003 
 1004         if ((mp->mnt_flag & MNT_RDONLY) == 0)
 1005                 vfs_allocate_syncvnode(mp);
 1006         else
 1007                 vfs_deallocate_syncvnode(mp);
 1008 end:
 1009         vfs_unbusy(mp);
 1010         VI_LOCK(vp);
 1011         vp->v_iflag &= ~VI_MOUNT;
 1012         VI_UNLOCK(vp);
 1013         vrele(vp);
 1014         return (error != 0 ? error : export_error);
 1015 }
 1016 
 1017 /*
 1018  * vfs_domount(): actually attempt a filesystem mount.
 1019  */
 1020 static int
 1021 vfs_domount(
 1022         struct thread *td,              /* Calling thread. */
 1023         const char *fstype,             /* Filesystem type. */
 1024         char *fspath,                   /* Mount path. */
 1025         uint64_t fsflags,               /* Flags common to all filesystems. */
 1026         struct vfsoptlist **optlist     /* Options local to the filesystem. */
 1027         )
 1028 {
 1029         struct vfsconf *vfsp;
 1030         struct nameidata nd;
 1031         struct vnode *vp;
 1032         char *pathbuf;
 1033         int error;
 1034 
 1035         /*
 1036          * Be ultra-paranoid about making sure the type and fspath
 1037          * variables will fit in our mp buffers, including the
 1038          * terminating NUL.
 1039          */
 1040         if (strlen(fstype) >= MFSNAMELEN || strlen(fspath) >= MNAMELEN)
 1041                 return (ENAMETOOLONG);
 1042 
 1043         if (jailed(td->td_ucred) || usermount == 0) {
 1044                 if ((error = priv_check(td, PRIV_VFS_MOUNT)) != 0)
 1045                         return (error);
 1046         }
 1047 
 1048         /*
 1049          * Do not allow NFS export or MNT_SUIDDIR by unprivileged users.
 1050          */
 1051         if (fsflags & MNT_EXPORTED) {
 1052                 error = priv_check(td, PRIV_VFS_MOUNT_EXPORTED);
 1053                 if (error)
 1054                         return (error);
 1055         }
 1056         if (fsflags & MNT_SUIDDIR) {
 1057                 error = priv_check(td, PRIV_VFS_MOUNT_SUIDDIR);
 1058                 if (error)
 1059                         return (error);
 1060         }
 1061         /*
 1062          * Silently enforce MNT_NOSUID and MNT_USER for unprivileged users.
 1063          */
 1064         if ((fsflags & (MNT_NOSUID | MNT_USER)) != (MNT_NOSUID | MNT_USER)) {
 1065                 if (priv_check(td, PRIV_VFS_MOUNT_NONUSER) != 0)
 1066                         fsflags |= MNT_NOSUID | MNT_USER;
 1067         }
 1068 
 1069         /* Load KLDs before we lock the covered vnode to avoid reversals. */
 1070         vfsp = NULL;
 1071         if ((fsflags & MNT_UPDATE) == 0) {
 1072                 /* Don't try to load KLDs if we're mounting the root. */
 1073                 if (fsflags & MNT_ROOTFS)
 1074                         vfsp = vfs_byname(fstype);
 1075                 else
 1076                         vfsp = vfs_byname_kld(fstype, td, &error);
 1077                 if (vfsp == NULL)
 1078                         return (ENODEV);
 1079                 if (jailed(td->td_ucred) && !(vfsp->vfc_flags & VFCF_JAIL))
 1080                         return (EPERM);
 1081         }
 1082 
 1083         /*
 1084          * Get vnode to be covered or mount point's vnode in case of MNT_UPDATE.
 1085          */
 1086         NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | MPSAFE | AUDITVNODE1,
 1087             UIO_SYSSPACE, fspath, td);
 1088         error = namei(&nd);
 1089         if (error != 0)
 1090                 return (error);
 1091         if (!NDHASGIANT(&nd))
 1092                 mtx_lock(&Giant);
 1093         NDFREE(&nd, NDF_ONLY_PNBUF);
 1094         vp = nd.ni_vp;
 1095         if ((fsflags & MNT_UPDATE) == 0) {
 1096                 pathbuf = malloc(MNAMELEN, M_TEMP, M_WAITOK);
 1097                 strcpy(pathbuf, fspath);
 1098                 error = vn_path_to_global_path(td, vp, pathbuf, MNAMELEN);
 1099                 /* debug.disablefullpath == 1 results in ENODEV */
 1100                 if (error == 0 || error == ENODEV) {
 1101                         error = vfs_domount_first(td, vfsp, pathbuf, vp,
 1102                             fsflags, optlist);
 1103                 }
 1104                 free(pathbuf, M_TEMP);
 1105         } else
 1106                 error = vfs_domount_update(td, vp, fsflags, optlist);
 1107         mtx_unlock(&Giant);
 1108 
 1109         ASSERT_VI_UNLOCKED(vp, __func__);
 1110         ASSERT_VOP_UNLOCKED(vp, __func__);
 1111 
 1112         return (error);
 1113 }
 1114 
 1115 /*
 1116  * Unmount a filesystem.
 1117  *
 1118  * Note: unmount takes a path to the vnode mounted on as argument, not
 1119  * special file (as before).
 1120  */
 1121 #ifndef _SYS_SYSPROTO_H_
 1122 struct unmount_args {
 1123         char    *path;
 1124         int     flags;
 1125 };
 1126 #endif
 1127 /* ARGSUSED */
 1128 int
 1129 sys_unmount(td, uap)
 1130         struct thread *td;
 1131         register struct unmount_args /* {
 1132                 char *path;
 1133                 int flags;
 1134         } */ *uap;
 1135 {
 1136         struct nameidata nd;
 1137         struct mount *mp;
 1138         char *pathbuf;
 1139         int error, id0, id1, vfslocked;
 1140 
 1141         AUDIT_ARG_VALUE(uap->flags);
 1142         if (jailed(td->td_ucred) || usermount == 0) {
 1143                 error = priv_check(td, PRIV_VFS_UNMOUNT);
 1144                 if (error)
 1145                         return (error);
 1146         }
 1147 
 1148         pathbuf = malloc(MNAMELEN, M_TEMP, M_WAITOK);
 1149         error = copyinstr(uap->path, pathbuf, MNAMELEN, NULL);
 1150         if (error) {
 1151                 free(pathbuf, M_TEMP);
 1152                 return (error);
 1153         }
 1154         mtx_lock(&Giant);
 1155         if (uap->flags & MNT_BYFSID) {
 1156                 AUDIT_ARG_TEXT(pathbuf);
 1157                 /* Decode the filesystem ID. */
 1158                 if (sscanf(pathbuf, "FSID:%d:%d", &id0, &id1) != 2) {
 1159                         mtx_unlock(&Giant);
 1160                         free(pathbuf, M_TEMP);
 1161                         return (EINVAL);
 1162                 }
 1163 
 1164                 mtx_lock(&mountlist_mtx);
 1165                 TAILQ_FOREACH_REVERSE(mp, &mountlist, mntlist, mnt_list) {
 1166                         if (mp->mnt_stat.f_fsid.val[0] == id0 &&
 1167                             mp->mnt_stat.f_fsid.val[1] == id1)
 1168                                 break;
 1169                 }
 1170                 mtx_unlock(&mountlist_mtx);
 1171         } else {
 1172                 AUDIT_ARG_UPATH1(td, pathbuf);
 1173                 /*
 1174                  * Try to find global path for path argument.
 1175                  */
 1176                 NDINIT(&nd, LOOKUP,
 1177                     FOLLOW | LOCKLEAF | MPSAFE | AUDITVNODE1,
 1178                     UIO_SYSSPACE, pathbuf, td);
 1179                 if (namei(&nd) == 0) {
 1180                         vfslocked = NDHASGIANT(&nd);
 1181                         NDFREE(&nd, NDF_ONLY_PNBUF);
 1182                         error = vn_path_to_global_path(td, nd.ni_vp, pathbuf,
 1183                             MNAMELEN);
 1184                         if (error == 0 || error == ENODEV)
 1185                                 vput(nd.ni_vp);
 1186                         VFS_UNLOCK_GIANT(vfslocked);
 1187                 }
 1188                 mtx_lock(&mountlist_mtx);
 1189                 TAILQ_FOREACH_REVERSE(mp, &mountlist, mntlist, mnt_list) {
 1190                         if (strcmp(mp->mnt_stat.f_mntonname, pathbuf) == 0)
 1191                                 break;
 1192                 }
 1193                 mtx_unlock(&mountlist_mtx);
 1194         }
 1195         free(pathbuf, M_TEMP);
 1196         if (mp == NULL) {
 1197                 /*
 1198                  * Previously we returned ENOENT for a nonexistent path and
 1199                  * EINVAL for a non-mountpoint.  We cannot tell these apart
 1200                  * now, so in the !MNT_BYFSID case return the more likely
 1201                  * EINVAL for compatibility.
 1202                  */
 1203                 mtx_unlock(&Giant);
 1204                 return ((uap->flags & MNT_BYFSID) ? ENOENT : EINVAL);
 1205         }
 1206 
 1207         /*
 1208          * Don't allow unmounting the root filesystem.
 1209          */
 1210         if (mp->mnt_flag & MNT_ROOTFS) {
 1211                 mtx_unlock(&Giant);
 1212                 return (EINVAL);
 1213         }
 1214         error = dounmount(mp, uap->flags, td);
 1215         mtx_unlock(&Giant);
 1216         return (error);
 1217 }
 1218 
 1219 /*
 1220  * Do the actual filesystem unmount.
 1221  */
 1222 int
 1223 dounmount(mp, flags, td)
 1224         struct mount *mp;
 1225         int flags;
 1226         struct thread *td;
 1227 {
 1228         struct vnode *coveredvp, *fsrootvp;
 1229         int error;
 1230         uint64_t async_flag;
 1231         int mnt_gen_r;
 1232 
 1233         mtx_assert(&Giant, MA_OWNED);
 1234 
 1235         if ((coveredvp = mp->mnt_vnodecovered) != NULL) {
 1236                 mnt_gen_r = mp->mnt_gen;
 1237                 VI_LOCK(coveredvp);
 1238                 vholdl(coveredvp);
 1239                 vn_lock(coveredvp, LK_EXCLUSIVE | LK_INTERLOCK | LK_RETRY);
 1240                 vdrop(coveredvp);
 1241                 /*
 1242                  * Check for mp being unmounted while waiting for the
 1243                  * covered vnode lock.
 1244                  */
 1245                 if (coveredvp->v_mountedhere != mp ||
 1246                     coveredvp->v_mountedhere->mnt_gen != mnt_gen_r) {
 1247                         VOP_UNLOCK(coveredvp, 0);
 1248                         return (EBUSY);
 1249                 }
 1250         }
 1251         /*
 1252          * Only privileged root, or (if MNT_USER is set) the user that did the
 1253          * original mount is permitted to unmount this filesystem.
 1254          */
 1255         error = vfs_suser(mp, td);
 1256         if (error) {
 1257                 if (coveredvp)
 1258                         VOP_UNLOCK(coveredvp, 0);
 1259                 return (error);
 1260         }
 1261 
 1262         MNT_ILOCK(mp);
 1263         if (mp->mnt_kern_flag & MNTK_UNMOUNT) {
 1264                 MNT_IUNLOCK(mp);
 1265                 if (coveredvp)
 1266                         VOP_UNLOCK(coveredvp, 0);
 1267                 return (EBUSY);
 1268         }
 1269         mp->mnt_kern_flag |= MNTK_UNMOUNT | MNTK_NOINSMNTQ;
 1270         /* Allow filesystems to detect that a forced unmount is in progress. */
 1271         if (flags & MNT_FORCE)
 1272                 mp->mnt_kern_flag |= MNTK_UNMOUNTF;
 1273         error = 0;
 1274         if (mp->mnt_lockref) {
 1275                 mp->mnt_kern_flag |= MNTK_DRAINING;
 1276                 error = msleep(&mp->mnt_lockref, MNT_MTX(mp), PVFS,
 1277                     "mount drain", 0);
 1278         }
 1279         MNT_IUNLOCK(mp);
 1280         KASSERT(mp->mnt_lockref == 0,
 1281             ("%s: invalid lock refcount in the drain path @ %s:%d",
 1282             __func__, __FILE__, __LINE__));
 1283         KASSERT(error == 0,
 1284             ("%s: invalid return value for msleep in the drain path @ %s:%d",
 1285             __func__, __FILE__, __LINE__));
 1286         vn_start_write(NULL, &mp, V_WAIT);
 1287 
 1288         if (mp->mnt_flag & MNT_EXPUBLIC)
 1289                 vfs_setpublicfs(NULL, NULL, NULL);
 1290 
 1291         vfs_msync(mp, MNT_WAIT);
 1292         MNT_ILOCK(mp);
 1293         async_flag = mp->mnt_flag & MNT_ASYNC;
 1294         mp->mnt_flag &= ~MNT_ASYNC;
 1295         mp->mnt_kern_flag &= ~MNTK_ASYNC;
 1296         MNT_IUNLOCK(mp);
 1297         cache_purgevfs(mp);     /* remove cache entries for this file sys */
 1298         vfs_deallocate_syncvnode(mp);
 1299         /*
 1300          * For forced unmounts, move process cdir/rdir refs on the fs root
 1301          * vnode to the covered vnode.  For non-forced unmounts we want
 1302          * such references to cause an EBUSY error.
 1303          */
 1304         if ((flags & MNT_FORCE) &&
 1305             VFS_ROOT(mp, LK_EXCLUSIVE, &fsrootvp) == 0) {
 1306                 if (mp->mnt_vnodecovered != NULL)
 1307                         mountcheckdirs(fsrootvp, mp->mnt_vnodecovered);
 1308                 if (fsrootvp == rootvnode) {
 1309                         vrele(rootvnode);
 1310                         rootvnode = NULL;
 1311                 }
 1312                 vput(fsrootvp);
 1313         }
 1314         if (((mp->mnt_flag & MNT_RDONLY) ||
 1315              (error = VFS_SYNC(mp, MNT_WAIT)) == 0) || (flags & MNT_FORCE) != 0)
 1316                 error = VFS_UNMOUNT(mp, flags);
 1317         vn_finished_write(mp);
 1318         /*
 1319          * If we failed to flush the dirty blocks for this mount point,
 1320          * undo all the cdir/rdir and rootvnode changes we made above.
 1321          * Unless we failed to do so because the device is reporting that
 1322          * it doesn't exist anymore.
 1323          */
 1324         if (error && error != ENXIO) {
 1325                 if ((flags & MNT_FORCE) &&
 1326                     VFS_ROOT(mp, LK_EXCLUSIVE, &fsrootvp) == 0) {
 1327                         if (mp->mnt_vnodecovered != NULL)
 1328                                 mountcheckdirs(mp->mnt_vnodecovered, fsrootvp);
 1329                         if (rootvnode == NULL) {
 1330                                 rootvnode = fsrootvp;
 1331                                 vref(rootvnode);
 1332                         }
 1333                         vput(fsrootvp);
 1334                 }
 1335                 MNT_ILOCK(mp);
 1336                 mp->mnt_kern_flag &= ~MNTK_NOINSMNTQ;
 1337                 if ((mp->mnt_flag & MNT_RDONLY) == 0) {
 1338                         MNT_IUNLOCK(mp);
 1339                         vfs_allocate_syncvnode(mp);
 1340                         MNT_ILOCK(mp);
 1341                 }
 1342                 mp->mnt_kern_flag &= ~(MNTK_UNMOUNT | MNTK_UNMOUNTF);
 1343                 mp->mnt_flag |= async_flag;
 1344                 if ((mp->mnt_flag & MNT_ASYNC) != 0 &&
 1345                     (mp->mnt_kern_flag & MNTK_NOASYNC) == 0)
 1346                         mp->mnt_kern_flag |= MNTK_ASYNC;
 1347                 if (mp->mnt_kern_flag & MNTK_MWAIT) {
 1348                         mp->mnt_kern_flag &= ~MNTK_MWAIT;
 1349                         wakeup(mp);
 1350                 }
 1351                 MNT_IUNLOCK(mp);
 1352                 if (coveredvp)
 1353                         VOP_UNLOCK(coveredvp, 0);
 1354                 return (error);
 1355         }
 1356         mtx_lock(&mountlist_mtx);
 1357         TAILQ_REMOVE(&mountlist, mp, mnt_list);
 1358         mtx_unlock(&mountlist_mtx);
 1359         if (coveredvp != NULL) {
 1360                 coveredvp->v_mountedhere = NULL;
 1361                 vput(coveredvp);
 1362         }
 1363         vfs_event_signal(NULL, VQ_UNMOUNT, 0);
 1364         vfs_mount_destroy(mp);
 1365         return (0);
 1366 }
 1367 
 1368 /*
 1369  * Report errors during filesystem mounting.
 1370  */
 1371 void
 1372 vfs_mount_error(struct mount *mp, const char *fmt, ...)
 1373 {
 1374         struct vfsoptlist *moptlist = mp->mnt_optnew;
 1375         va_list ap;
 1376         int error, len;
 1377         char *errmsg;
 1378 
 1379         error = vfs_getopt(moptlist, "errmsg", (void **)&errmsg, &len);
 1380         if (error || errmsg == NULL || len <= 0)
 1381                 return;
 1382 
 1383         va_start(ap, fmt);
 1384         vsnprintf(errmsg, (size_t)len, fmt, ap);
 1385         va_end(ap);
 1386 }
 1387 
 1388 void
 1389 vfs_opterror(struct vfsoptlist *opts, const char *fmt, ...)
 1390 {
 1391         va_list ap;
 1392         int error, len;
 1393         char *errmsg;
 1394 
 1395         error = vfs_getopt(opts, "errmsg", (void **)&errmsg, &len);
 1396         if (error || errmsg == NULL || len <= 0)
 1397                 return;
 1398 
 1399         va_start(ap, fmt);
 1400         vsnprintf(errmsg, (size_t)len, fmt, ap);
 1401         va_end(ap);
 1402 }
 1403 
 1404 /*
 1405  * ---------------------------------------------------------------------
 1406  * Functions for querying mount options/arguments from filesystems.
 1407  */
 1408 
 1409 /*
 1410  * Check that no unknown options are given
 1411  */
 1412 int
 1413 vfs_filteropt(struct vfsoptlist *opts, const char **legal)
 1414 {
 1415         struct vfsopt *opt;
 1416         char errmsg[255];
 1417         const char **t, *p, *q;
 1418         int ret = 0;
 1419 
 1420         TAILQ_FOREACH(opt, opts, link) {
 1421                 p = opt->name;
 1422                 q = NULL;
 1423                 if (p[0] == 'n' && p[1] == 'o')
 1424                         q = p + 2;
 1425                 for(t = global_opts; *t != NULL; t++) {
 1426                         if (strcmp(*t, p) == 0)
 1427                                 break;
 1428                         if (q != NULL) {
 1429                                 if (strcmp(*t, q) == 0)
 1430                                         break;
 1431                         }
 1432                 }
 1433                 if (*t != NULL)
 1434                         continue;
 1435                 for(t = legal; *t != NULL; t++) {
 1436                         if (strcmp(*t, p) == 0)
 1437                                 break;
 1438                         if (q != NULL) {
 1439                                 if (strcmp(*t, q) == 0)
 1440                                         break;
 1441                         }
 1442                 }
 1443                 if (*t != NULL)
 1444                         continue;
 1445                 snprintf(errmsg, sizeof(errmsg),
 1446                     "mount option <%s> is unknown", p);
 1447                 ret = EINVAL;
 1448         }
 1449         if (ret != 0) {
 1450                 TAILQ_FOREACH(opt, opts, link) {
 1451                         if (strcmp(opt->name, "errmsg") == 0) {
 1452                                 strncpy((char *)opt->value, errmsg, opt->len);
 1453                                 break;
 1454                         }
 1455                 }
 1456                 if (opt == NULL)
 1457                         printf("%s\n", errmsg);
 1458         }
 1459         return (ret);
 1460 }
 1461 
 1462 /*
 1463  * Get a mount option by its name.
 1464  *
 1465  * Return 0 if the option was found, ENOENT otherwise.
 1466  * If len is non-NULL it will be filled with the length
 1467  * of the option. If buf is non-NULL, it will be filled
 1468  * with the address of the option.
 1469  */
 1470 int
 1471 vfs_getopt(opts, name, buf, len)
 1472         struct vfsoptlist *opts;
 1473         const char *name;
 1474         void **buf;
 1475         int *len;
 1476 {
 1477         struct vfsopt *opt;
 1478 
 1479         KASSERT(opts != NULL, ("vfs_getopt: caller passed 'opts' as NULL"));
 1480 
 1481         TAILQ_FOREACH(opt, opts, link) {
 1482                 if (strcmp(name, opt->name) == 0) {
 1483                         opt->seen = 1;
 1484                         if (len != NULL)
 1485                                 *len = opt->len;
 1486                         if (buf != NULL)
 1487                                 *buf = opt->value;
 1488                         return (0);
 1489                 }
 1490         }
 1491         return (ENOENT);
 1492 }
 1493 
 1494 int
 1495 vfs_getopt_pos(struct vfsoptlist *opts, const char *name)
 1496 {
 1497         struct vfsopt *opt;
 1498 
 1499         if (opts == NULL)
 1500                 return (-1);
 1501 
 1502         TAILQ_FOREACH(opt, opts, link) {
 1503                 if (strcmp(name, opt->name) == 0) {
 1504                         opt->seen = 1;
 1505                         return (opt->pos);
 1506                 }
 1507         }
 1508         return (-1);
 1509 }
 1510 
 1511 char *
 1512 vfs_getopts(struct vfsoptlist *opts, const char *name, int *error)
 1513 {
 1514         struct vfsopt *opt;
 1515 
 1516         *error = 0;
 1517         TAILQ_FOREACH(opt, opts, link) {
 1518                 if (strcmp(name, opt->name) != 0)
 1519                         continue;
 1520                 opt->seen = 1;
 1521                 if (opt->len == 0 ||
 1522                     ((char *)opt->value)[opt->len - 1] != '\0') {
 1523                         *error = EINVAL;
 1524                         return (NULL);
 1525                 }
 1526                 return (opt->value);
 1527         }
 1528         *error = ENOENT;
 1529         return (NULL);
 1530 }
 1531 
 1532 int
 1533 vfs_flagopt(struct vfsoptlist *opts, const char *name, uint64_t *w,
 1534         uint64_t val)
 1535 {
 1536         struct vfsopt *opt;
 1537 
 1538         TAILQ_FOREACH(opt, opts, link) {
 1539                 if (strcmp(name, opt->name) == 0) {
 1540                         opt->seen = 1;
 1541                         if (w != NULL)
 1542                                 *w |= val;
 1543                         return (1);
 1544                 }
 1545         }
 1546         if (w != NULL)
 1547                 *w &= ~val;
 1548         return (0);
 1549 }
 1550 
 1551 int
 1552 vfs_scanopt(struct vfsoptlist *opts, const char *name, const char *fmt, ...)
 1553 {
 1554         va_list ap;
 1555         struct vfsopt *opt;
 1556         int ret;
 1557 
 1558         KASSERT(opts != NULL, ("vfs_getopt: caller passed 'opts' as NULL"));
 1559 
 1560         TAILQ_FOREACH(opt, opts, link) {
 1561                 if (strcmp(name, opt->name) != 0)
 1562                         continue;
 1563                 opt->seen = 1;
 1564                 if (opt->len == 0 || opt->value == NULL)
 1565                         return (0);
 1566                 if (((char *)opt->value)[opt->len - 1] != '\0')
 1567                         return (0);
 1568                 va_start(ap, fmt);
 1569                 ret = vsscanf(opt->value, fmt, ap);
 1570                 va_end(ap);
 1571                 return (ret);
 1572         }
 1573         return (0);
 1574 }
 1575 
 1576 int
 1577 vfs_setopt(struct vfsoptlist *opts, const char *name, void *value, int len)
 1578 {
 1579         struct vfsopt *opt;
 1580 
 1581         TAILQ_FOREACH(opt, opts, link) {
 1582                 if (strcmp(name, opt->name) != 0)
 1583                         continue;
 1584                 opt->seen = 1;
 1585                 if (opt->value == NULL)
 1586                         opt->len = len;
 1587                 else {
 1588                         if (opt->len != len)
 1589                                 return (EINVAL);
 1590                         bcopy(value, opt->value, len);
 1591                 }
 1592                 return (0);
 1593         }
 1594         return (ENOENT);
 1595 }
 1596 
 1597 int
 1598 vfs_setopt_part(struct vfsoptlist *opts, const char *name, void *value, int len)
 1599 {
 1600         struct vfsopt *opt;
 1601 
 1602         TAILQ_FOREACH(opt, opts, link) {
 1603                 if (strcmp(name, opt->name) != 0)
 1604                         continue;
 1605                 opt->seen = 1;
 1606                 if (opt->value == NULL)
 1607                         opt->len = len;
 1608                 else {
 1609                         if (opt->len < len)
 1610                                 return (EINVAL);
 1611                         opt->len = len;
 1612                         bcopy(value, opt->value, len);
 1613                 }
 1614                 return (0);
 1615         }
 1616         return (ENOENT);
 1617 }
 1618 
 1619 int
 1620 vfs_setopts(struct vfsoptlist *opts, const char *name, const char *value)
 1621 {
 1622         struct vfsopt *opt;
 1623 
 1624         TAILQ_FOREACH(opt, opts, link) {
 1625                 if (strcmp(name, opt->name) != 0)
 1626                         continue;
 1627                 opt->seen = 1;
 1628                 if (opt->value == NULL)
 1629                         opt->len = strlen(value) + 1;
 1630                 else if (strlcpy(opt->value, value, opt->len) >= opt->len)
 1631                         return (EINVAL);
 1632                 return (0);
 1633         }
 1634         return (ENOENT);
 1635 }
 1636 
 1637 /*
 1638  * Find and copy a mount option.
 1639  *
 1640  * The size of the buffer has to be specified
 1641  * in len, if it is not the same length as the
 1642  * mount option, EINVAL is returned.
 1643  * Returns ENOENT if the option is not found.
 1644  */
 1645 int
 1646 vfs_copyopt(opts, name, dest, len)
 1647         struct vfsoptlist *opts;
 1648         const char *name;
 1649         void *dest;
 1650         int len;
 1651 {
 1652         struct vfsopt *opt;
 1653 
 1654         KASSERT(opts != NULL, ("vfs_copyopt: caller passed 'opts' as NULL"));
 1655 
 1656         TAILQ_FOREACH(opt, opts, link) {
 1657                 if (strcmp(name, opt->name) == 0) {
 1658                         opt->seen = 1;
 1659                         if (len != opt->len)
 1660                                 return (EINVAL);
 1661                         bcopy(opt->value, dest, opt->len);
 1662                         return (0);
 1663                 }
 1664         }
 1665         return (ENOENT);
 1666 }
 1667 
 1668 /*
 1669  * These are helper functions for filesystems to traverse all
 1670  * their vnodes.  See MNT_VNODE_FOREACH() in sys/mount.h.
 1671  *
 1672  * This interface has been deprecated in favor of MNT_VNODE_FOREACH_ALL.
 1673  */
 1674 
 1675 MALLOC_DECLARE(M_VNODE_MARKER);
 1676 
 1677 struct vnode *
 1678 __mnt_vnode_next(struct vnode **mvp, struct mount *mp)
 1679 {
 1680         struct vnode *vp;
 1681 
 1682         mtx_assert(MNT_MTX(mp), MA_OWNED);
 1683 
 1684         KASSERT((*mvp)->v_mount == mp, ("marker vnode mount list mismatch"));
 1685         if (should_yield()) {
 1686                 MNT_IUNLOCK(mp);
 1687                 kern_yield(PRI_UNCHANGED);
 1688                 MNT_ILOCK(mp);
 1689         }
 1690         vp = TAILQ_NEXT(*mvp, v_nmntvnodes);
 1691         while (vp != NULL && vp->v_type == VMARKER)
 1692                 vp = TAILQ_NEXT(vp, v_nmntvnodes);
 1693 
 1694         /* Check if we are done */
 1695         if (vp == NULL) {
 1696                 __mnt_vnode_markerfree(mvp, mp);
 1697                 return (NULL);
 1698         }
 1699         TAILQ_REMOVE(&mp->mnt_nvnodelist, *mvp, v_nmntvnodes);
 1700         TAILQ_INSERT_AFTER(&mp->mnt_nvnodelist, vp, *mvp, v_nmntvnodes);
 1701         return (vp);
 1702 }
 1703 
 1704 struct vnode *
 1705 __mnt_vnode_first(struct vnode **mvp, struct mount *mp)
 1706 {
 1707         struct vnode *vp;
 1708 
 1709         mtx_assert(MNT_MTX(mp), MA_OWNED);
 1710 
 1711         vp = TAILQ_FIRST(&mp->mnt_nvnodelist);
 1712         while (vp != NULL && vp->v_type == VMARKER)
 1713                 vp = TAILQ_NEXT(vp, v_nmntvnodes);
 1714 
 1715         /* Check if we are done */
 1716         if (vp == NULL) {
 1717                 *mvp = NULL;
 1718                 return (NULL);
 1719         }
 1720         MNT_REF(mp);
 1721         MNT_IUNLOCK(mp);
 1722         *mvp = (struct vnode *) malloc(sizeof(struct vnode),
 1723                                        M_VNODE_MARKER,
 1724                                        M_WAITOK | M_ZERO);
 1725         MNT_ILOCK(mp);
 1726         (*mvp)->v_type = VMARKER;
 1727 
 1728         vp = TAILQ_FIRST(&mp->mnt_nvnodelist);
 1729         while (vp != NULL && vp->v_type == VMARKER)
 1730                 vp = TAILQ_NEXT(vp, v_nmntvnodes);
 1731 
 1732         /* Check if we are done */
 1733         if (vp == NULL) {
 1734                 MNT_IUNLOCK(mp);
 1735                 free(*mvp, M_VNODE_MARKER);
 1736                 MNT_ILOCK(mp);
 1737                 *mvp = NULL;
 1738                 MNT_REL(mp);
 1739                 return (NULL);
 1740         }
 1741         (*mvp)->v_mount = mp;
 1742         TAILQ_INSERT_AFTER(&mp->mnt_nvnodelist, vp, *mvp, v_nmntvnodes);
 1743         return (vp);
 1744 }
 1745 
 1746 
 1747 void
 1748 __mnt_vnode_markerfree(struct vnode **mvp, struct mount *mp)
 1749 {
 1750 
 1751         if (*mvp == NULL)
 1752                 return;
 1753 
 1754         mtx_assert(MNT_MTX(mp), MA_OWNED);
 1755 
 1756         KASSERT((*mvp)->v_mount == mp, ("marker vnode mount list mismatch"));
 1757         TAILQ_REMOVE(&mp->mnt_nvnodelist, *mvp, v_nmntvnodes);
 1758         MNT_IUNLOCK(mp);
 1759         free(*mvp, M_VNODE_MARKER);
 1760         MNT_ILOCK(mp);
 1761         *mvp = NULL;
 1762         MNT_REL(mp);
 1763 }
 1764 
 1765 int
 1766 __vfs_statfs(struct mount *mp, struct statfs *sbp)
 1767 {
 1768         int error;
 1769 
 1770         error = mp->mnt_op->vfs_statfs(mp, &mp->mnt_stat);
 1771         if (sbp != &mp->mnt_stat)
 1772                 *sbp = mp->mnt_stat;
 1773         return (error);
 1774 }
 1775 
 1776 void
 1777 vfs_mountedfrom(struct mount *mp, const char *from)
 1778 {
 1779 
 1780         bzero(mp->mnt_stat.f_mntfromname, sizeof mp->mnt_stat.f_mntfromname);
 1781         strlcpy(mp->mnt_stat.f_mntfromname, from,
 1782             sizeof mp->mnt_stat.f_mntfromname);
 1783 }
 1784 
 1785 /*
 1786  * ---------------------------------------------------------------------
 1787  * This is the api for building mount args and mounting filesystems from
 1788  * inside the kernel.
 1789  *
 1790  * The API works by accumulation of individual args.  First error is
 1791  * latched.
 1792  *
 1793  * XXX: should be documented in new manpage kernel_mount(9)
 1794  */
 1795 
 1796 /* A memory allocation which must be freed when we are done */
 1797 struct mntaarg {
 1798         SLIST_ENTRY(mntaarg)    next;
 1799 };
 1800 
 1801 /* The header for the mount arguments */
 1802 struct mntarg {
 1803         struct iovec *v;
 1804         int len;
 1805         int error;
 1806         SLIST_HEAD(, mntaarg)   list;
 1807 };
 1808 
 1809 /*
 1810  * Add a boolean argument.
 1811  *
 1812  * flag is the boolean value.
 1813  * name must start with "no".
 1814  */
 1815 struct mntarg *
 1816 mount_argb(struct mntarg *ma, int flag, const char *name)
 1817 {
 1818 
 1819         KASSERT(name[0] == 'n' && name[1] == 'o',
 1820             ("mount_argb(...,%s): name must start with 'no'", name));
 1821 
 1822         return (mount_arg(ma, name + (flag ? 2 : 0), NULL, 0));
 1823 }
 1824 
 1825 /*
 1826  * Add an argument printf style
 1827  */
 1828 struct mntarg *
 1829 mount_argf(struct mntarg *ma, const char *name, const char *fmt, ...)
 1830 {
 1831         va_list ap;
 1832         struct mntaarg *maa;
 1833         struct sbuf *sb;
 1834         int len;
 1835 
 1836         if (ma == NULL) {
 1837                 ma = malloc(sizeof *ma, M_MOUNT, M_WAITOK | M_ZERO);
 1838                 SLIST_INIT(&ma->list);
 1839         }
 1840         if (ma->error)
 1841                 return (ma);
 1842 
 1843         ma->v = realloc(ma->v, sizeof *ma->v * (ma->len + 2),
 1844             M_MOUNT, M_WAITOK);
 1845         ma->v[ma->len].iov_base = (void *)(uintptr_t)name;
 1846         ma->v[ma->len].iov_len = strlen(name) + 1;
 1847         ma->len++;
 1848 
 1849         sb = sbuf_new_auto();
 1850         va_start(ap, fmt);
 1851         sbuf_vprintf(sb, fmt, ap);
 1852         va_end(ap);
 1853         sbuf_finish(sb);
 1854         len = sbuf_len(sb) + 1;
 1855         maa = malloc(sizeof *maa + len, M_MOUNT, M_WAITOK | M_ZERO);
 1856         SLIST_INSERT_HEAD(&ma->list, maa, next);
 1857         bcopy(sbuf_data(sb), maa + 1, len);
 1858         sbuf_delete(sb);
 1859 
 1860         ma->v[ma->len].iov_base = maa + 1;
 1861         ma->v[ma->len].iov_len = len;
 1862         ma->len++;
 1863 
 1864         return (ma);
 1865 }
 1866 
 1867 /*
 1868  * Add an argument which is a userland string.
 1869  */
 1870 struct mntarg *
 1871 mount_argsu(struct mntarg *ma, const char *name, const void *val, int len)
 1872 {
 1873         struct mntaarg *maa;
 1874         char *tbuf;
 1875 
 1876         if (val == NULL)
 1877                 return (ma);
 1878         if (ma == NULL) {
 1879                 ma = malloc(sizeof *ma, M_MOUNT, M_WAITOK | M_ZERO);
 1880                 SLIST_INIT(&ma->list);
 1881         }
 1882         if (ma->error)
 1883                 return (ma);
 1884         maa = malloc(sizeof *maa + len, M_MOUNT, M_WAITOK | M_ZERO);
 1885         SLIST_INSERT_HEAD(&ma->list, maa, next);
 1886         tbuf = (void *)(maa + 1);
 1887         ma->error = copyinstr(val, tbuf, len, NULL);
 1888         return (mount_arg(ma, name, tbuf, -1));
 1889 }
 1890 
 1891 /*
 1892  * Plain argument.
 1893  *
 1894  * If length is -1, treat value as a C string.
 1895  */
 1896 struct mntarg *
 1897 mount_arg(struct mntarg *ma, const char *name, const void *val, int len)
 1898 {
 1899 
 1900         if (ma == NULL) {
 1901                 ma = malloc(sizeof *ma, M_MOUNT, M_WAITOK | M_ZERO);
 1902                 SLIST_INIT(&ma->list);
 1903         }
 1904         if (ma->error)
 1905                 return (ma);
 1906 
 1907         ma->v = realloc(ma->v, sizeof *ma->v * (ma->len + 2),
 1908             M_MOUNT, M_WAITOK);
 1909         ma->v[ma->len].iov_base = (void *)(uintptr_t)name;
 1910         ma->v[ma->len].iov_len = strlen(name) + 1;
 1911         ma->len++;
 1912 
 1913         ma->v[ma->len].iov_base = (void *)(uintptr_t)val;
 1914         if (len < 0)
 1915                 ma->v[ma->len].iov_len = strlen(val) + 1;
 1916         else
 1917                 ma->v[ma->len].iov_len = len;
 1918         ma->len++;
 1919         return (ma);
 1920 }
 1921 
 1922 /*
 1923  * Free a mntarg structure
 1924  */
 1925 static void
 1926 free_mntarg(struct mntarg *ma)
 1927 {
 1928         struct mntaarg *maa;
 1929 
 1930         while (!SLIST_EMPTY(&ma->list)) {
 1931                 maa = SLIST_FIRST(&ma->list);
 1932                 SLIST_REMOVE_HEAD(&ma->list, next);
 1933                 free(maa, M_MOUNT);
 1934         }
 1935         free(ma->v, M_MOUNT);
 1936         free(ma, M_MOUNT);
 1937 }
 1938 
 1939 /*
 1940  * Mount a filesystem
 1941  */
 1942 int
 1943 kernel_mount(struct mntarg *ma, uint64_t flags)
 1944 {
 1945         struct uio auio;
 1946         int error;
 1947 
 1948         KASSERT(ma != NULL, ("kernel_mount NULL ma"));
 1949         KASSERT(ma->v != NULL, ("kernel_mount NULL ma->v"));
 1950         KASSERT(!(ma->len & 1), ("kernel_mount odd ma->len (%d)", ma->len));
 1951 
 1952         auio.uio_iov = ma->v;
 1953         auio.uio_iovcnt = ma->len;
 1954         auio.uio_segflg = UIO_SYSSPACE;
 1955 
 1956         error = ma->error;
 1957         if (!error)
 1958                 error = vfs_donmount(curthread, flags, &auio);
 1959         free_mntarg(ma);
 1960         return (error);
 1961 }
 1962 
 1963 /*
 1964  * A printflike function to mount a filesystem.
 1965  */
 1966 int
 1967 kernel_vmount(int flags, ...)
 1968 {
 1969         struct mntarg *ma = NULL;
 1970         va_list ap;
 1971         const char *cp;
 1972         const void *vp;
 1973         int error;
 1974 
 1975         va_start(ap, flags);
 1976         for (;;) {
 1977                 cp = va_arg(ap, const char *);
 1978                 if (cp == NULL)
 1979                         break;
 1980                 vp = va_arg(ap, const void *);
 1981                 ma = mount_arg(ma, cp, vp, (vp != NULL ? -1 : 0));
 1982         }
 1983         va_end(ap);
 1984 
 1985         error = kernel_mount(ma, flags);
 1986         return (error);
 1987 }
 1988 
 1989 void
 1990 vfs_oexport_conv(const struct oexport_args *oexp, struct export_args *exp)
 1991 {
 1992 
 1993         bcopy(oexp, exp, sizeof(*oexp));
 1994         exp->ex_numsecflavors = 0;
 1995 }

Cache object: c91b8eb4c34176307b6dcaf0ea684a8d


[ 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.