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  -  FREEBSD11  -  FREEBSD10  -  FREEBSD9  -  FREEBSD92  -  FREEBSD91  -  FREEBSD90  -  FREEBSD8  -  FREEBSD82  -  FREEBSD81  -  FREEBSD80  -  FREEBSD7  -  FREEBSD74  -  FREEBSD73  -  FREEBSD72  -  FREEBSD71  -  FREEBSD70  -  FREEBSD6  -  FREEBSD64  -  FREEBSD63  -  FREEBSD62  -  FREEBSD61  -  FREEBSD60  -  FREEBSD5  -  FREEBSD55  -  FREEBSD54  -  FREEBSD53  -  FREEBSD52  -  FREEBSD51  -  FREEBSD50  -  FREEBSD4  -  FREEBSD3  -  FREEBSD22  -  linux-2.6  -  linux-2.4.22  -  MK83  -  MK84  -  PLAN9  -  DFBSD  -  NETBSD  -  NETBSD5  -  NETBSD4  -  NETBSD3  -  NETBSD20  -  OPENBSD  -  xnu-517  -  xnu-792  -  xnu-792.6.70  -  xnu-1228  -  xnu-1456.1.26  -  xnu-1699.24.8  -  xnu-2050.18.24  -  OPENSOLARIS  -  minix-3-1-1 
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: stable/10/sys/kern/vfs_mount.c 332754 2018-04-19 05:52:47Z avg $");
   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 static int      default_autoro = false;
   82 SYSCTL_INT(_vfs, OID_AUTO, default_autoro, CTLFLAG_RW, &default_autoro, 0,
   83     "Retry failed r/w mount as r/o if no explicit ro/rw option is specified");
   84 
   85 MALLOC_DEFINE(M_MOUNT, "mount", "vfs mount structure");
   86 static uma_zone_t mount_zone;
   87 
   88 /* List of mounted filesystems. */
   89 struct mntlist mountlist = TAILQ_HEAD_INITIALIZER(mountlist);
   90 
   91 /* For any iteration/modification of mountlist */
   92 struct mtx mountlist_mtx;
   93 MTX_SYSINIT(mountlist, &mountlist_mtx, "mountlist", MTX_DEF);
   94 
   95 /*
   96  * Global opts, taken by all filesystems
   97  */
   98 static const char *global_opts[] = {
   99         "errmsg",
  100         "fstype",
  101         "fspath",
  102         "ro",
  103         "rw",
  104         "nosuid",
  105         "noexec",
  106         NULL
  107 };
  108 
  109 static int
  110 mount_init(void *mem, int size, int flags)
  111 {
  112         struct mount *mp;
  113 
  114         mp = (struct mount *)mem;
  115         mtx_init(&mp->mnt_mtx, "struct mount mtx", NULL, MTX_DEF);
  116         lockinit(&mp->mnt_explock, PVFS, "explock", 0, 0);
  117         return (0);
  118 }
  119 
  120 static void
  121 mount_fini(void *mem, int size)
  122 {
  123         struct mount *mp;
  124 
  125         mp = (struct mount *)mem;
  126         lockdestroy(&mp->mnt_explock);
  127         mtx_destroy(&mp->mnt_mtx);
  128 }
  129 
  130 static void
  131 vfs_mount_init(void *dummy __unused)
  132 {
  133 
  134         mount_zone = uma_zcreate("Mountpoints", sizeof(struct mount), NULL,
  135             NULL, mount_init, mount_fini, UMA_ALIGN_PTR, UMA_ZONE_NOFREE);
  136 }
  137 SYSINIT(vfs_mount, SI_SUB_VFS, SI_ORDER_ANY, vfs_mount_init, NULL);
  138 
  139 /*
  140  * ---------------------------------------------------------------------
  141  * Functions for building and sanitizing the mount options
  142  */
  143 
  144 /* Remove one mount option. */
  145 static void
  146 vfs_freeopt(struct vfsoptlist *opts, struct vfsopt *opt)
  147 {
  148 
  149         TAILQ_REMOVE(opts, opt, link);
  150         free(opt->name, M_MOUNT);
  151         if (opt->value != NULL)
  152                 free(opt->value, M_MOUNT);
  153         free(opt, M_MOUNT);
  154 }
  155 
  156 /* Release all resources related to the mount options. */
  157 void
  158 vfs_freeopts(struct vfsoptlist *opts)
  159 {
  160         struct vfsopt *opt;
  161 
  162         while (!TAILQ_EMPTY(opts)) {
  163                 opt = TAILQ_FIRST(opts);
  164                 vfs_freeopt(opts, opt);
  165         }
  166         free(opts, M_MOUNT);
  167 }
  168 
  169 void
  170 vfs_deleteopt(struct vfsoptlist *opts, const char *name)
  171 {
  172         struct vfsopt *opt, *temp;
  173 
  174         if (opts == NULL)
  175                 return;
  176         TAILQ_FOREACH_SAFE(opt, opts, link, temp)  {
  177                 if (strcmp(opt->name, name) == 0)
  178                         vfs_freeopt(opts, opt);
  179         }
  180 }
  181 
  182 static int
  183 vfs_isopt_ro(const char *opt)
  184 {
  185 
  186         if (strcmp(opt, "ro") == 0 || strcmp(opt, "rdonly") == 0 ||
  187             strcmp(opt, "norw") == 0)
  188                 return (1);
  189         return (0);
  190 }
  191 
  192 static int
  193 vfs_isopt_rw(const char *opt)
  194 {
  195 
  196         if (strcmp(opt, "rw") == 0 || strcmp(opt, "noro") == 0)
  197                 return (1);
  198         return (0);
  199 }
  200 
  201 /*
  202  * Check if options are equal (with or without the "no" prefix).
  203  */
  204 static int
  205 vfs_equalopts(const char *opt1, const char *opt2)
  206 {
  207         char *p;
  208 
  209         /* "opt" vs. "opt" or "noopt" vs. "noopt" */
  210         if (strcmp(opt1, opt2) == 0)
  211                 return (1);
  212         /* "noopt" vs. "opt" */
  213         if (strncmp(opt1, "no", 2) == 0 && strcmp(opt1 + 2, opt2) == 0)
  214                 return (1);
  215         /* "opt" vs. "noopt" */
  216         if (strncmp(opt2, "no", 2) == 0 && strcmp(opt1, opt2 + 2) == 0)
  217                 return (1);
  218         while ((p = strchr(opt1, '.')) != NULL &&
  219             !strncmp(opt1, opt2, ++p - opt1)) {
  220                 opt2 += p - opt1;
  221                 opt1 = p;
  222                 /* "foo.noopt" vs. "foo.opt" */
  223                 if (strncmp(opt1, "no", 2) == 0 && strcmp(opt1 + 2, opt2) == 0)
  224                         return (1);
  225                 /* "foo.opt" vs. "foo.noopt" */
  226                 if (strncmp(opt2, "no", 2) == 0 && strcmp(opt1, opt2 + 2) == 0)
  227                         return (1);
  228         }
  229         /* "ro" / "rdonly" / "norw" / "rw" / "noro" */
  230         if ((vfs_isopt_ro(opt1) || vfs_isopt_rw(opt1)) &&
  231             (vfs_isopt_ro(opt2) || vfs_isopt_rw(opt2)))
  232                 return (1);
  233         return (0);
  234 }
  235 
  236 /*
  237  * If a mount option is specified several times,
  238  * (with or without the "no" prefix) only keep
  239  * the last occurrence of it.
  240  */
  241 static void
  242 vfs_sanitizeopts(struct vfsoptlist *opts)
  243 {
  244         struct vfsopt *opt, *opt2, *tmp;
  245 
  246         TAILQ_FOREACH_REVERSE(opt, opts, vfsoptlist, link) {
  247                 opt2 = TAILQ_PREV(opt, vfsoptlist, link);
  248                 while (opt2 != NULL) {
  249                         if (vfs_equalopts(opt->name, opt2->name)) {
  250                                 tmp = TAILQ_PREV(opt2, vfsoptlist, link);
  251                                 vfs_freeopt(opts, opt2);
  252                                 opt2 = tmp;
  253                         } else {
  254                                 opt2 = TAILQ_PREV(opt2, vfsoptlist, link);
  255                         }
  256                 }
  257         }
  258 }
  259 
  260 /*
  261  * Build a linked list of mount options from a struct uio.
  262  */
  263 int
  264 vfs_buildopts(struct uio *auio, struct vfsoptlist **options)
  265 {
  266         struct vfsoptlist *opts;
  267         struct vfsopt *opt;
  268         size_t memused, namelen, optlen;
  269         unsigned int i, iovcnt;
  270         int error;
  271 
  272         opts = malloc(sizeof(struct vfsoptlist), M_MOUNT, M_WAITOK);
  273         TAILQ_INIT(opts);
  274         memused = 0;
  275         iovcnt = auio->uio_iovcnt;
  276         for (i = 0; i < iovcnt; i += 2) {
  277                 namelen = auio->uio_iov[i].iov_len;
  278                 optlen = auio->uio_iov[i + 1].iov_len;
  279                 memused += sizeof(struct vfsopt) + optlen + namelen;
  280                 /*
  281                  * Avoid consuming too much memory, and attempts to overflow
  282                  * memused.
  283                  */
  284                 if (memused > VFS_MOUNTARG_SIZE_MAX ||
  285                     optlen > VFS_MOUNTARG_SIZE_MAX ||
  286                     namelen > VFS_MOUNTARG_SIZE_MAX) {
  287                         error = EINVAL;
  288                         goto bad;
  289                 }
  290 
  291                 opt = malloc(sizeof(struct vfsopt), M_MOUNT, M_WAITOK);
  292                 opt->name = malloc(namelen, M_MOUNT, M_WAITOK);
  293                 opt->value = NULL;
  294                 opt->len = 0;
  295                 opt->pos = i / 2;
  296                 opt->seen = 0;
  297 
  298                 /*
  299                  * Do this early, so jumps to "bad" will free the current
  300                  * option.
  301                  */
  302                 TAILQ_INSERT_TAIL(opts, opt, link);
  303 
  304                 if (auio->uio_segflg == UIO_SYSSPACE) {
  305                         bcopy(auio->uio_iov[i].iov_base, opt->name, namelen);
  306                 } else {
  307                         error = copyin(auio->uio_iov[i].iov_base, opt->name,
  308                             namelen);
  309                         if (error)
  310                                 goto bad;
  311                 }
  312                 /* Ensure names are null-terminated strings. */
  313                 if (namelen == 0 || opt->name[namelen - 1] != '\0') {
  314                         error = EINVAL;
  315                         goto bad;
  316                 }
  317                 if (optlen != 0) {
  318                         opt->len = optlen;
  319                         opt->value = malloc(optlen, M_MOUNT, M_WAITOK);
  320                         if (auio->uio_segflg == UIO_SYSSPACE) {
  321                                 bcopy(auio->uio_iov[i + 1].iov_base, opt->value,
  322                                     optlen);
  323                         } else {
  324                                 error = copyin(auio->uio_iov[i + 1].iov_base,
  325                                     opt->value, optlen);
  326                                 if (error)
  327                                         goto bad;
  328                         }
  329                 }
  330         }
  331         vfs_sanitizeopts(opts);
  332         *options = opts;
  333         return (0);
  334 bad:
  335         vfs_freeopts(opts);
  336         return (error);
  337 }
  338 
  339 /*
  340  * Merge the old mount options with the new ones passed
  341  * in the MNT_UPDATE case.
  342  *
  343  * XXX: This function will keep a "nofoo" option in the new
  344  * options.  E.g, if the option's canonical name is "foo",
  345  * "nofoo" ends up in the mount point's active options.
  346  */
  347 static void
  348 vfs_mergeopts(struct vfsoptlist *toopts, struct vfsoptlist *oldopts)
  349 {
  350         struct vfsopt *opt, *new;
  351 
  352         TAILQ_FOREACH(opt, oldopts, link) {
  353                 new = malloc(sizeof(struct vfsopt), M_MOUNT, M_WAITOK);
  354                 new->name = strdup(opt->name, M_MOUNT);
  355                 if (opt->len != 0) {
  356                         new->value = malloc(opt->len, M_MOUNT, M_WAITOK);
  357                         bcopy(opt->value, new->value, opt->len);
  358                 } else
  359                         new->value = NULL;
  360                 new->len = opt->len;
  361                 new->seen = opt->seen;
  362                 TAILQ_INSERT_HEAD(toopts, new, link);
  363         }
  364         vfs_sanitizeopts(toopts);
  365 }
  366 
  367 /*
  368  * Mount a filesystem.
  369  */
  370 int
  371 sys_nmount(td, uap)
  372         struct thread *td;
  373         struct nmount_args /* {
  374                 struct iovec *iovp;
  375                 unsigned int iovcnt;
  376                 int flags;
  377         } */ *uap;
  378 {
  379         struct uio *auio;
  380         int error;
  381         u_int iovcnt;
  382         uint64_t flags;
  383 
  384         /*
  385          * Mount flags are now 64-bits. On 32-bit archtectures only
  386          * 32-bits are passed in, but from here on everything handles
  387          * 64-bit flags correctly.
  388          */
  389         flags = uap->flags;
  390 
  391         AUDIT_ARG_FFLAGS(flags);
  392         CTR4(KTR_VFS, "%s: iovp %p with iovcnt %d and flags %d", __func__,
  393             uap->iovp, uap->iovcnt, flags);
  394 
  395         /*
  396          * Filter out MNT_ROOTFS.  We do not want clients of nmount() in
  397          * userspace to set this flag, but we must filter it out if we want
  398          * MNT_UPDATE on the root file system to work.
  399          * MNT_ROOTFS should only be set by the kernel when mounting its
  400          * root file system.
  401          */
  402         flags &= ~MNT_ROOTFS;
  403 
  404         iovcnt = uap->iovcnt;
  405         /*
  406          * Check that we have an even number of iovec's
  407          * and that we have at least two options.
  408          */
  409         if ((iovcnt & 1) || (iovcnt < 4)) {
  410                 CTR2(KTR_VFS, "%s: failed for invalid iovcnt %d", __func__,
  411                     uap->iovcnt);
  412                 return (EINVAL);
  413         }
  414 
  415         error = copyinuio(uap->iovp, iovcnt, &auio);
  416         if (error) {
  417                 CTR2(KTR_VFS, "%s: failed for invalid uio op with %d errno",
  418                     __func__, error);
  419                 return (error);
  420         }
  421         error = vfs_donmount(td, flags, auio);
  422 
  423         free(auio, M_IOV);
  424         return (error);
  425 }
  426 
  427 /*
  428  * ---------------------------------------------------------------------
  429  * Various utility functions
  430  */
  431 
  432 void
  433 vfs_ref(struct mount *mp)
  434 {
  435 
  436         CTR2(KTR_VFS, "%s: mp %p", __func__, mp);
  437         MNT_ILOCK(mp);
  438         MNT_REF(mp);
  439         MNT_IUNLOCK(mp);
  440 }
  441 
  442 void
  443 vfs_rel(struct mount *mp)
  444 {
  445 
  446         CTR2(KTR_VFS, "%s: mp %p", __func__, mp);
  447         MNT_ILOCK(mp);
  448         MNT_REL(mp);
  449         MNT_IUNLOCK(mp);
  450 }
  451 
  452 /*
  453  * Allocate and initialize the mount point struct.
  454  */
  455 struct mount *
  456 vfs_mount_alloc(struct vnode *vp, struct vfsconf *vfsp, const char *fspath,
  457     struct ucred *cred)
  458 {
  459         struct mount *mp;
  460 
  461         mp = uma_zalloc(mount_zone, M_WAITOK);
  462         bzero(&mp->mnt_startzero,
  463             __rangeof(struct mount, mnt_startzero, mnt_endzero));
  464         TAILQ_INIT(&mp->mnt_nvnodelist);
  465         mp->mnt_nvnodelistsize = 0;
  466         TAILQ_INIT(&mp->mnt_activevnodelist);
  467         mp->mnt_activevnodelistsize = 0;
  468         mp->mnt_ref = 0;
  469         (void) vfs_busy(mp, MBF_NOWAIT);
  470         atomic_add_acq_int(&vfsp->vfc_refcount, 1);
  471         mp->mnt_op = vfsp->vfc_vfsops;
  472         mp->mnt_vfc = vfsp;
  473         mp->mnt_stat.f_type = vfsp->vfc_typenum;
  474         mp->mnt_gen++;
  475         strlcpy(mp->mnt_stat.f_fstypename, vfsp->vfc_name, MFSNAMELEN);
  476         mp->mnt_vnodecovered = vp;
  477         mp->mnt_cred = crdup(cred);
  478         mp->mnt_stat.f_owner = cred->cr_uid;
  479         strlcpy(mp->mnt_stat.f_mntonname, fspath, MNAMELEN);
  480         mp->mnt_iosize_max = DFLTPHYS;
  481 #ifdef MAC
  482         mac_mount_init(mp);
  483         mac_mount_create(cred, mp);
  484 #endif
  485         arc4rand(&mp->mnt_hashseed, sizeof mp->mnt_hashseed, 0);
  486         TAILQ_INIT(&mp->mnt_uppers);
  487         return (mp);
  488 }
  489 
  490 /*
  491  * Destroy the mount struct previously allocated by vfs_mount_alloc().
  492  */
  493 void
  494 vfs_mount_destroy(struct mount *mp)
  495 {
  496 
  497         MNT_ILOCK(mp);
  498         mp->mnt_kern_flag |= MNTK_REFEXPIRE;
  499         if (mp->mnt_kern_flag & MNTK_MWAIT) {
  500                 mp->mnt_kern_flag &= ~MNTK_MWAIT;
  501                 wakeup(mp);
  502         }
  503         while (mp->mnt_ref)
  504                 msleep(mp, MNT_MTX(mp), PVFS, "mntref", 0);
  505         KASSERT(mp->mnt_ref == 0,
  506             ("%s: invalid refcount in the drain path @ %s:%d", __func__,
  507             __FILE__, __LINE__));
  508         if (mp->mnt_writeopcount != 0)
  509                 panic("vfs_mount_destroy: nonzero writeopcount");
  510         if (mp->mnt_secondary_writes != 0)
  511                 panic("vfs_mount_destroy: nonzero secondary_writes");
  512         atomic_subtract_rel_int(&mp->mnt_vfc->vfc_refcount, 1);
  513         if (!TAILQ_EMPTY(&mp->mnt_nvnodelist)) {
  514                 struct vnode *vp;
  515 
  516                 TAILQ_FOREACH(vp, &mp->mnt_nvnodelist, v_nmntvnodes)
  517                         vprint("", vp);
  518                 panic("unmount: dangling vnode");
  519         }
  520         KASSERT(TAILQ_EMPTY(&mp->mnt_uppers), ("mnt_uppers"));
  521         if (mp->mnt_nvnodelistsize != 0)
  522                 panic("vfs_mount_destroy: nonzero nvnodelistsize");
  523         if (mp->mnt_activevnodelistsize != 0)
  524                 panic("vfs_mount_destroy: nonzero activevnodelistsize");
  525         if (mp->mnt_lockref != 0)
  526                 panic("vfs_mount_destroy: nonzero lock refcount");
  527         MNT_IUNLOCK(mp);
  528         if (mp->mnt_vnodecovered != NULL)
  529                 vrele(mp->mnt_vnodecovered);
  530 #ifdef MAC
  531         mac_mount_destroy(mp);
  532 #endif
  533         if (mp->mnt_opt != NULL)
  534                 vfs_freeopts(mp->mnt_opt);
  535         crfree(mp->mnt_cred);
  536         uma_zfree(mount_zone, mp);
  537 }
  538 
  539 static bool
  540 vfs_should_downgrade_to_ro_mount(uint64_t fsflags, int error)
  541 {
  542         /* This is an upgrade of an exisiting mount. */
  543         if ((fsflags & MNT_UPDATE) != 0)
  544                 return (false);
  545         /* This is already an R/O mount. */
  546         if ((fsflags & MNT_RDONLY) != 0)
  547                 return (false);
  548 
  549         switch (error) {
  550         case ENODEV:    /* generic, geom, ... */
  551         case EACCES:    /* cam/scsi, ... */
  552         case EROFS:     /* md, mmcsd, ... */
  553                 /*
  554                  * These errors can be returned by the storage layer to signal
  555                  * that the media is read-only.  No harm in the R/O mount
  556                  * attempt if the error was returned for some other reason.
  557                  */
  558                 return (true);
  559         default:
  560                 return (false);
  561         }
  562 }
  563 
  564 int
  565 vfs_donmount(struct thread *td, uint64_t fsflags, struct uio *fsoptions)
  566 {
  567         struct vfsoptlist *optlist;
  568         struct vfsopt *opt, *tmp_opt;
  569         char *fstype, *fspath, *errmsg;
  570         int error, fstypelen, fspathlen, errmsg_len, errmsg_pos;
  571         bool autoro;
  572 
  573         errmsg = fspath = NULL;
  574         errmsg_len = fspathlen = 0;
  575         errmsg_pos = -1;
  576         autoro = default_autoro;
  577 
  578         error = vfs_buildopts(fsoptions, &optlist);
  579         if (error)
  580                 return (error);
  581 
  582         if (vfs_getopt(optlist, "errmsg", (void **)&errmsg, &errmsg_len) == 0)
  583                 errmsg_pos = vfs_getopt_pos(optlist, "errmsg");
  584 
  585         /*
  586          * We need these two options before the others,
  587          * and they are mandatory for any filesystem.
  588          * Ensure they are NUL terminated as well.
  589          */
  590         fstypelen = 0;
  591         error = vfs_getopt(optlist, "fstype", (void **)&fstype, &fstypelen);
  592         if (error || fstype[fstypelen - 1] != '\0') {
  593                 error = EINVAL;
  594                 if (errmsg != NULL)
  595                         strncpy(errmsg, "Invalid fstype", errmsg_len);
  596                 goto bail;
  597         }
  598         fspathlen = 0;
  599         error = vfs_getopt(optlist, "fspath", (void **)&fspath, &fspathlen);
  600         if (error || fspath[fspathlen - 1] != '\0') {
  601                 error = EINVAL;
  602                 if (errmsg != NULL)
  603                         strncpy(errmsg, "Invalid fspath", errmsg_len);
  604                 goto bail;
  605         }
  606 
  607         /*
  608          * We need to see if we have the "update" option
  609          * before we call vfs_domount(), since vfs_domount() has special
  610          * logic based on MNT_UPDATE.  This is very important
  611          * when we want to update the root filesystem.
  612          */
  613         TAILQ_FOREACH_SAFE(opt, optlist, link, tmp_opt) {
  614                 if (strcmp(opt->name, "update") == 0) {
  615                         fsflags |= MNT_UPDATE;
  616                         vfs_freeopt(optlist, opt);
  617                 }
  618                 else if (strcmp(opt->name, "async") == 0)
  619                         fsflags |= MNT_ASYNC;
  620                 else if (strcmp(opt->name, "force") == 0) {
  621                         fsflags |= MNT_FORCE;
  622                         vfs_freeopt(optlist, opt);
  623                 }
  624                 else if (strcmp(opt->name, "reload") == 0) {
  625                         fsflags |= MNT_RELOAD;
  626                         vfs_freeopt(optlist, opt);
  627                 }
  628                 else if (strcmp(opt->name, "multilabel") == 0)
  629                         fsflags |= MNT_MULTILABEL;
  630                 else if (strcmp(opt->name, "noasync") == 0)
  631                         fsflags &= ~MNT_ASYNC;
  632                 else if (strcmp(opt->name, "noatime") == 0)
  633                         fsflags |= MNT_NOATIME;
  634                 else if (strcmp(opt->name, "atime") == 0) {
  635                         free(opt->name, M_MOUNT);
  636                         opt->name = strdup("nonoatime", M_MOUNT);
  637                 }
  638                 else if (strcmp(opt->name, "noclusterr") == 0)
  639                         fsflags |= MNT_NOCLUSTERR;
  640                 else if (strcmp(opt->name, "clusterr") == 0) {
  641                         free(opt->name, M_MOUNT);
  642                         opt->name = strdup("nonoclusterr", M_MOUNT);
  643                 }
  644                 else if (strcmp(opt->name, "noclusterw") == 0)
  645                         fsflags |= MNT_NOCLUSTERW;
  646                 else if (strcmp(opt->name, "clusterw") == 0) {
  647                         free(opt->name, M_MOUNT);
  648                         opt->name = strdup("nonoclusterw", M_MOUNT);
  649                 }
  650                 else if (strcmp(opt->name, "noexec") == 0)
  651                         fsflags |= MNT_NOEXEC;
  652                 else if (strcmp(opt->name, "exec") == 0) {
  653                         free(opt->name, M_MOUNT);
  654                         opt->name = strdup("nonoexec", M_MOUNT);
  655                 }
  656                 else if (strcmp(opt->name, "nosuid") == 0)
  657                         fsflags |= MNT_NOSUID;
  658                 else if (strcmp(opt->name, "suid") == 0) {
  659                         free(opt->name, M_MOUNT);
  660                         opt->name = strdup("nonosuid", M_MOUNT);
  661                 }
  662                 else if (strcmp(opt->name, "nosymfollow") == 0)
  663                         fsflags |= MNT_NOSYMFOLLOW;
  664                 else if (strcmp(opt->name, "symfollow") == 0) {
  665                         free(opt->name, M_MOUNT);
  666                         opt->name = strdup("nonosymfollow", M_MOUNT);
  667                 }
  668                 else if (strcmp(opt->name, "noro") == 0) {
  669                         fsflags &= ~MNT_RDONLY;
  670                         autoro = false;
  671                 }
  672                 else if (strcmp(opt->name, "rw") == 0) {
  673                         fsflags &= ~MNT_RDONLY;
  674                         autoro = false;
  675                 }
  676                 else if (strcmp(opt->name, "ro") == 0) {
  677                         fsflags |= MNT_RDONLY;
  678                         autoro = false;
  679                 }
  680                 else if (strcmp(opt->name, "rdonly") == 0) {
  681                         free(opt->name, M_MOUNT);
  682                         opt->name = strdup("ro", M_MOUNT);
  683                         fsflags |= MNT_RDONLY;
  684                         autoro = false;
  685                 }
  686                 else if (strcmp(opt->name, "autoro") == 0) {
  687                         vfs_freeopt(optlist, opt);
  688                         autoro = true;
  689                 }
  690                 else if (strcmp(opt->name, "suiddir") == 0)
  691                         fsflags |= MNT_SUIDDIR;
  692                 else if (strcmp(opt->name, "sync") == 0)
  693                         fsflags |= MNT_SYNCHRONOUS;
  694                 else if (strcmp(opt->name, "union") == 0)
  695                         fsflags |= MNT_UNION;
  696                 else if (strcmp(opt->name, "automounted") == 0) {
  697                         fsflags |= MNT_AUTOMOUNTED;
  698                         vfs_freeopt(optlist, opt);
  699                 }
  700         }
  701 
  702         /*
  703          * Be ultra-paranoid about making sure the type and fspath
  704          * variables will fit in our mp buffers, including the
  705          * terminating NUL.
  706          */
  707         if (fstypelen > MFSNAMELEN || fspathlen > MNAMELEN) {
  708                 error = ENAMETOOLONG;
  709                 goto bail;
  710         }
  711 
  712         error = vfs_domount(td, fstype, fspath, fsflags, &optlist);
  713 
  714         /*
  715          * See if we can mount in the read-only mode if the error code suggests
  716          * that it could be possible and the mount options allow for that.
  717          * Never try it if "[no]{ro|rw}" has been explicitly requested and not
  718          * overridden by "autoro".
  719          */
  720         if (autoro && vfs_should_downgrade_to_ro_mount(fsflags, error)) {
  721                 printf("%s: R/W mount failed, possibly R/O media,"
  722                     " trying R/O mount\n", __func__);
  723                 fsflags |= MNT_RDONLY;
  724                 error = vfs_domount(td, fstype, fspath, fsflags, &optlist);
  725         }
  726 bail:
  727         /* copyout the errmsg */
  728         if (errmsg_pos != -1 && ((2 * errmsg_pos + 1) < fsoptions->uio_iovcnt)
  729             && errmsg_len > 0 && errmsg != NULL) {
  730                 if (fsoptions->uio_segflg == UIO_SYSSPACE) {
  731                         bcopy(errmsg,
  732                             fsoptions->uio_iov[2 * errmsg_pos + 1].iov_base,
  733                             fsoptions->uio_iov[2 * errmsg_pos + 1].iov_len);
  734                 } else {
  735                         copyout(errmsg,
  736                             fsoptions->uio_iov[2 * errmsg_pos + 1].iov_base,
  737                             fsoptions->uio_iov[2 * errmsg_pos + 1].iov_len);
  738                 }
  739         }
  740 
  741         if (optlist != NULL)
  742                 vfs_freeopts(optlist);
  743         return (error);
  744 }
  745 
  746 /*
  747  * Old mount API.
  748  */
  749 #ifndef _SYS_SYSPROTO_H_
  750 struct mount_args {
  751         char    *type;
  752         char    *path;
  753         int     flags;
  754         caddr_t data;
  755 };
  756 #endif
  757 /* ARGSUSED */
  758 int
  759 sys_mount(td, uap)
  760         struct thread *td;
  761         struct mount_args /* {
  762                 char *type;
  763                 char *path;
  764                 int flags;
  765                 caddr_t data;
  766         } */ *uap;
  767 {
  768         char *fstype;
  769         struct vfsconf *vfsp = NULL;
  770         struct mntarg *ma = NULL;
  771         uint64_t flags;
  772         int error;
  773 
  774         /*
  775          * Mount flags are now 64-bits. On 32-bit architectures only
  776          * 32-bits are passed in, but from here on everything handles
  777          * 64-bit flags correctly.
  778          */
  779         flags = uap->flags;
  780 
  781         AUDIT_ARG_FFLAGS(flags);
  782 
  783         /*
  784          * Filter out MNT_ROOTFS.  We do not want clients of mount() in
  785          * userspace to set this flag, but we must filter it out if we want
  786          * MNT_UPDATE on the root file system to work.
  787          * MNT_ROOTFS should only be set by the kernel when mounting its
  788          * root file system.
  789          */
  790         flags &= ~MNT_ROOTFS;
  791 
  792         fstype = malloc(MFSNAMELEN, M_TEMP, M_WAITOK);
  793         error = copyinstr(uap->type, fstype, MFSNAMELEN, NULL);
  794         if (error) {
  795                 free(fstype, M_TEMP);
  796                 return (error);
  797         }
  798 
  799         AUDIT_ARG_TEXT(fstype);
  800         vfsp = vfs_byname_kld(fstype, td, &error);
  801         free(fstype, M_TEMP);
  802         if (vfsp == NULL)
  803                 return (ENOENT);
  804         if (vfsp->vfc_vfsops->vfs_cmount == NULL)
  805                 return (EOPNOTSUPP);
  806 
  807         ma = mount_argsu(ma, "fstype", uap->type, MFSNAMELEN);
  808         ma = mount_argsu(ma, "fspath", uap->path, MNAMELEN);
  809         ma = mount_argb(ma, flags & MNT_RDONLY, "noro");
  810         ma = mount_argb(ma, !(flags & MNT_NOSUID), "nosuid");
  811         ma = mount_argb(ma, !(flags & MNT_NOEXEC), "noexec");
  812 
  813         error = vfsp->vfc_vfsops->vfs_cmount(ma, uap->data, flags);
  814         return (error);
  815 }
  816 
  817 /*
  818  * vfs_domount_first(): first file system mount (not update)
  819  */
  820 static int
  821 vfs_domount_first(
  822         struct thread *td,              /* Calling thread. */
  823         struct vfsconf *vfsp,           /* File system type. */
  824         char *fspath,                   /* Mount path. */
  825         struct vnode *vp,               /* Vnode to be covered. */
  826         uint64_t fsflags,               /* Flags common to all filesystems. */
  827         struct vfsoptlist **optlist     /* Options local to the filesystem. */
  828         )
  829 {
  830         struct vattr va;
  831         struct mount *mp;
  832         struct vnode *newdp;
  833         int error;
  834 
  835         ASSERT_VOP_ELOCKED(vp, __func__);
  836         KASSERT((fsflags & MNT_UPDATE) == 0, ("MNT_UPDATE shouldn't be here"));
  837 
  838         /*
  839          * If the user is not root, ensure that they own the directory
  840          * onto which we are attempting to mount.
  841          */
  842         error = VOP_GETATTR(vp, &va, td->td_ucred);
  843         if (error == 0 && va.va_uid != td->td_ucred->cr_uid)
  844                 error = priv_check_cred(td->td_ucred, PRIV_VFS_ADMIN, 0);
  845         if (error == 0)
  846                 error = vinvalbuf(vp, V_SAVE, 0, 0);
  847         if (error == 0 && vp->v_type != VDIR)
  848                 error = ENOTDIR;
  849         if (error == 0) {
  850                 VI_LOCK(vp);
  851                 if ((vp->v_iflag & VI_MOUNT) == 0 && vp->v_mountedhere == NULL)
  852                         vp->v_iflag |= VI_MOUNT;
  853                 else
  854                         error = EBUSY;
  855                 VI_UNLOCK(vp);
  856         }
  857         if (error != 0) {
  858                 vput(vp);
  859                 return (error);
  860         }
  861         VOP_UNLOCK(vp, 0);
  862 
  863         /* Allocate and initialize the filesystem. */
  864         mp = vfs_mount_alloc(vp, vfsp, fspath, td->td_ucred);
  865         /* XXXMAC: pass to vfs_mount_alloc? */
  866         mp->mnt_optnew = *optlist;
  867         /* Set the mount level flags. */
  868         mp->mnt_flag = (fsflags & (MNT_UPDATEMASK | MNT_ROOTFS | MNT_RDONLY));
  869 
  870         /*
  871          * Mount the filesystem.
  872          * XXX The final recipients of VFS_MOUNT just overwrite the ndp they
  873          * get.  No freeing of cn_pnbuf.
  874          */
  875         error = VFS_MOUNT(mp);
  876         if (error != 0) {
  877                 vfs_unbusy(mp);
  878                 mp->mnt_vnodecovered = NULL;
  879                 vfs_mount_destroy(mp);
  880                 VI_LOCK(vp);
  881                 vp->v_iflag &= ~VI_MOUNT;
  882                 VI_UNLOCK(vp);
  883                 vrele(vp);
  884                 return (error);
  885         }
  886 
  887         if (mp->mnt_opt != NULL)
  888                 vfs_freeopts(mp->mnt_opt);
  889         mp->mnt_opt = mp->mnt_optnew;
  890         *optlist = NULL;
  891         (void)VFS_STATFS(mp, &mp->mnt_stat);
  892 
  893         /*
  894          * Prevent external consumers of mount options from reading mnt_optnew.
  895          */
  896         mp->mnt_optnew = NULL;
  897 
  898         MNT_ILOCK(mp);
  899         if ((mp->mnt_flag & MNT_ASYNC) != 0 &&
  900             (mp->mnt_kern_flag & MNTK_NOASYNC) == 0)
  901                 mp->mnt_kern_flag |= MNTK_ASYNC;
  902         else
  903                 mp->mnt_kern_flag &= ~MNTK_ASYNC;
  904         MNT_IUNLOCK(mp);
  905 
  906         vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
  907         cache_purge(vp);
  908         VI_LOCK(vp);
  909         vp->v_iflag &= ~VI_MOUNT;
  910         VI_UNLOCK(vp);
  911         vp->v_mountedhere = mp;
  912         /* Place the new filesystem at the end of the mount list. */
  913         mtx_lock(&mountlist_mtx);
  914         TAILQ_INSERT_TAIL(&mountlist, mp, mnt_list);
  915         mtx_unlock(&mountlist_mtx);
  916         vfs_event_signal(NULL, VQ_MOUNT, 0);
  917         if (VFS_ROOT(mp, LK_EXCLUSIVE, &newdp))
  918                 panic("mount: lost mount");
  919         VOP_UNLOCK(vp, 0);
  920         EVENTHANDLER_INVOKE(vfs_mounted, mp, newdp, td);
  921         VOP_UNLOCK(newdp, 0);
  922         mountcheckdirs(vp, newdp);
  923         vrele(newdp);
  924         if ((mp->mnt_flag & MNT_RDONLY) == 0)
  925                 vfs_allocate_syncvnode(mp);
  926         vfs_unbusy(mp);
  927         return (0);
  928 }
  929 
  930 /*
  931  * vfs_domount_update(): update of mounted file system
  932  */
  933 static int
  934 vfs_domount_update(
  935         struct thread *td,              /* Calling thread. */
  936         struct vnode *vp,               /* Mount point vnode. */
  937         uint64_t fsflags,               /* Flags common to all filesystems. */
  938         struct vfsoptlist **optlist     /* Options local to the filesystem. */
  939         )
  940 {
  941         struct oexport_args oexport;
  942         struct export_args export;
  943         struct mount *mp;
  944         int error, export_error;
  945         uint64_t flag;
  946 
  947         ASSERT_VOP_ELOCKED(vp, __func__);
  948         KASSERT((fsflags & MNT_UPDATE) != 0, ("MNT_UPDATE should be here"));
  949         mp = vp->v_mount;
  950 
  951         if ((vp->v_vflag & VV_ROOT) == 0) {
  952                 if (vfs_copyopt(*optlist, "export", &export, sizeof(export))
  953                     == 0)
  954                         error = EXDEV;
  955                 else
  956                         error = EINVAL;
  957                 vput(vp);
  958                 return (error);
  959         }
  960 
  961         /*
  962          * We only allow the filesystem to be reloaded if it
  963          * is currently mounted read-only.
  964          */
  965         flag = mp->mnt_flag;
  966         if ((fsflags & MNT_RELOAD) != 0 && (flag & MNT_RDONLY) == 0) {
  967                 vput(vp);
  968                 return (EOPNOTSUPP);    /* Needs translation */
  969         }
  970         /*
  971          * Only privileged root, or (if MNT_USER is set) the user that
  972          * did the original mount is permitted to update it.
  973          */
  974         error = vfs_suser(mp, td);
  975         if (error != 0) {
  976                 vput(vp);
  977                 return (error);
  978         }
  979         if (vfs_busy(mp, MBF_NOWAIT)) {
  980                 vput(vp);
  981                 return (EBUSY);
  982         }
  983         VI_LOCK(vp);
  984         if ((vp->v_iflag & VI_MOUNT) != 0 || vp->v_mountedhere != NULL) {
  985                 VI_UNLOCK(vp);
  986                 vfs_unbusy(mp);
  987                 vput(vp);
  988                 return (EBUSY);
  989         }
  990         vp->v_iflag |= VI_MOUNT;
  991         VI_UNLOCK(vp);
  992         VOP_UNLOCK(vp, 0);
  993 
  994         MNT_ILOCK(mp);
  995         if ((mp->mnt_kern_flag & MNTK_UNMOUNT) != 0) {
  996                 MNT_IUNLOCK(mp);
  997                 error = EBUSY;
  998                 goto end;
  999         }
 1000         mp->mnt_flag &= ~MNT_UPDATEMASK;
 1001         mp->mnt_flag |= fsflags & (MNT_RELOAD | MNT_FORCE | MNT_UPDATE |
 1002             MNT_SNAPSHOT | MNT_ROOTFS | MNT_UPDATEMASK | MNT_RDONLY);
 1003         if ((mp->mnt_flag & MNT_ASYNC) == 0)
 1004                 mp->mnt_kern_flag &= ~MNTK_ASYNC;
 1005         MNT_IUNLOCK(mp);
 1006         mp->mnt_optnew = *optlist;
 1007         vfs_mergeopts(mp->mnt_optnew, mp->mnt_opt);
 1008 
 1009         /*
 1010          * Mount the filesystem.
 1011          * XXX The final recipients of VFS_MOUNT just overwrite the ndp they
 1012          * get.  No freeing of cn_pnbuf.
 1013          */
 1014         error = VFS_MOUNT(mp);
 1015 
 1016         export_error = 0;
 1017         if (error == 0) {
 1018                 /* Process the export option. */
 1019                 if (vfs_copyopt(mp->mnt_optnew, "export", &export,
 1020                     sizeof(export)) == 0) {
 1021                         export_error = vfs_export(mp, &export);
 1022                 } else if (vfs_copyopt(mp->mnt_optnew, "export", &oexport,
 1023                     sizeof(oexport)) == 0) {
 1024                         export.ex_flags = oexport.ex_flags;
 1025                         export.ex_root = oexport.ex_root;
 1026                         export.ex_anon = oexport.ex_anon;
 1027                         export.ex_addr = oexport.ex_addr;
 1028                         export.ex_addrlen = oexport.ex_addrlen;
 1029                         export.ex_mask = oexport.ex_mask;
 1030                         export.ex_masklen = oexport.ex_masklen;
 1031                         export.ex_indexfile = oexport.ex_indexfile;
 1032                         export.ex_numsecflavors = 0;
 1033                         export_error = vfs_export(mp, &export);
 1034                 }
 1035         }
 1036 
 1037         MNT_ILOCK(mp);
 1038         if (error == 0) {
 1039                 mp->mnt_flag &= ~(MNT_UPDATE | MNT_RELOAD | MNT_FORCE |
 1040                     MNT_SNAPSHOT);
 1041         } else {
 1042                 /*
 1043                  * If we fail, restore old mount flags. MNT_QUOTA is special,
 1044                  * because it is not part of MNT_UPDATEMASK, but it could have
 1045                  * changed in the meantime if quotactl(2) was called.
 1046                  * All in all we want current value of MNT_QUOTA, not the old
 1047                  * one.
 1048                  */
 1049                 mp->mnt_flag = (mp->mnt_flag & MNT_QUOTA) | (flag & ~MNT_QUOTA);
 1050         }
 1051         if ((mp->mnt_flag & MNT_ASYNC) != 0 &&
 1052             (mp->mnt_kern_flag & MNTK_NOASYNC) == 0)
 1053                 mp->mnt_kern_flag |= MNTK_ASYNC;
 1054         else
 1055                 mp->mnt_kern_flag &= ~MNTK_ASYNC;
 1056         MNT_IUNLOCK(mp);
 1057 
 1058         if (error != 0)
 1059                 goto end;
 1060 
 1061         if (mp->mnt_opt != NULL)
 1062                 vfs_freeopts(mp->mnt_opt);
 1063         mp->mnt_opt = mp->mnt_optnew;
 1064         *optlist = NULL;
 1065         (void)VFS_STATFS(mp, &mp->mnt_stat);
 1066         /*
 1067          * Prevent external consumers of mount options from reading
 1068          * mnt_optnew.
 1069          */
 1070         mp->mnt_optnew = NULL;
 1071 
 1072         if ((mp->mnt_flag & MNT_RDONLY) == 0)
 1073                 vfs_allocate_syncvnode(mp);
 1074         else
 1075                 vfs_deallocate_syncvnode(mp);
 1076 end:
 1077         vfs_unbusy(mp);
 1078         VI_LOCK(vp);
 1079         vp->v_iflag &= ~VI_MOUNT;
 1080         VI_UNLOCK(vp);
 1081         vrele(vp);
 1082         return (error != 0 ? error : export_error);
 1083 }
 1084 
 1085 /*
 1086  * vfs_domount(): actually attempt a filesystem mount.
 1087  */
 1088 static int
 1089 vfs_domount(
 1090         struct thread *td,              /* Calling thread. */
 1091         const char *fstype,             /* Filesystem type. */
 1092         char *fspath,                   /* Mount path. */
 1093         uint64_t fsflags,               /* Flags common to all filesystems. */
 1094         struct vfsoptlist **optlist     /* Options local to the filesystem. */
 1095         )
 1096 {
 1097         struct vfsconf *vfsp;
 1098         struct nameidata nd;
 1099         struct vnode *vp;
 1100         char *pathbuf;
 1101         int error;
 1102 
 1103         /*
 1104          * Be ultra-paranoid about making sure the type and fspath
 1105          * variables will fit in our mp buffers, including the
 1106          * terminating NUL.
 1107          */
 1108         if (strlen(fstype) >= MFSNAMELEN || strlen(fspath) >= MNAMELEN)
 1109                 return (ENAMETOOLONG);
 1110 
 1111         if (jailed(td->td_ucred) || usermount == 0) {
 1112                 if ((error = priv_check(td, PRIV_VFS_MOUNT)) != 0)
 1113                         return (error);
 1114         }
 1115 
 1116         /*
 1117          * Do not allow NFS export or MNT_SUIDDIR by unprivileged users.
 1118          */
 1119         if (fsflags & MNT_EXPORTED) {
 1120                 error = priv_check(td, PRIV_VFS_MOUNT_EXPORTED);
 1121                 if (error)
 1122                         return (error);
 1123         }
 1124         if (fsflags & MNT_SUIDDIR) {
 1125                 error = priv_check(td, PRIV_VFS_MOUNT_SUIDDIR);
 1126                 if (error)
 1127                         return (error);
 1128         }
 1129         /*
 1130          * Silently enforce MNT_NOSUID and MNT_USER for unprivileged users.
 1131          */
 1132         if ((fsflags & (MNT_NOSUID | MNT_USER)) != (MNT_NOSUID | MNT_USER)) {
 1133                 if (priv_check(td, PRIV_VFS_MOUNT_NONUSER) != 0)
 1134                         fsflags |= MNT_NOSUID | MNT_USER;
 1135         }
 1136 
 1137         /* Load KLDs before we lock the covered vnode to avoid reversals. */
 1138         vfsp = NULL;
 1139         if ((fsflags & MNT_UPDATE) == 0) {
 1140                 /* Don't try to load KLDs if we're mounting the root. */
 1141                 if (fsflags & MNT_ROOTFS)
 1142                         vfsp = vfs_byname(fstype);
 1143                 else
 1144                         vfsp = vfs_byname_kld(fstype, td, &error);
 1145                 if (vfsp == NULL)
 1146                         return (ENODEV);
 1147                 if (jailed(td->td_ucred) && !(vfsp->vfc_flags & VFCF_JAIL))
 1148                         return (EPERM);
 1149         }
 1150 
 1151         /*
 1152          * Get vnode to be covered or mount point's vnode in case of MNT_UPDATE.
 1153          */
 1154         NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1,
 1155             UIO_SYSSPACE, fspath, td);
 1156         error = namei(&nd);
 1157         if (error != 0)
 1158                 return (error);
 1159         NDFREE(&nd, NDF_ONLY_PNBUF);
 1160         vp = nd.ni_vp;
 1161         if ((fsflags & MNT_UPDATE) == 0) {
 1162                 pathbuf = malloc(MNAMELEN, M_TEMP, M_WAITOK);
 1163                 strcpy(pathbuf, fspath);
 1164                 error = vn_path_to_global_path(td, vp, pathbuf, MNAMELEN);
 1165                 /* debug.disablefullpath == 1 results in ENODEV */
 1166                 if (error == 0 || error == ENODEV) {
 1167                         error = vfs_domount_first(td, vfsp, pathbuf, vp,
 1168                             fsflags, optlist);
 1169                 }
 1170                 free(pathbuf, M_TEMP);
 1171         } else
 1172                 error = vfs_domount_update(td, vp, fsflags, optlist);
 1173 
 1174         return (error);
 1175 }
 1176 
 1177 /*
 1178  * Unmount a filesystem.
 1179  *
 1180  * Note: unmount takes a path to the vnode mounted on as argument, not
 1181  * special file (as before).
 1182  */
 1183 #ifndef _SYS_SYSPROTO_H_
 1184 struct unmount_args {
 1185         char    *path;
 1186         int     flags;
 1187 };
 1188 #endif
 1189 /* ARGSUSED */
 1190 int
 1191 sys_unmount(struct thread *td, struct unmount_args *uap)
 1192 {
 1193         struct nameidata nd;
 1194         struct mount *mp;
 1195         char *pathbuf;
 1196         int error, id0, id1;
 1197 
 1198         AUDIT_ARG_VALUE(uap->flags);
 1199         if (jailed(td->td_ucred) || usermount == 0) {
 1200                 error = priv_check(td, PRIV_VFS_UNMOUNT);
 1201                 if (error)
 1202                         return (error);
 1203         }
 1204 
 1205         pathbuf = malloc(MNAMELEN, M_TEMP, M_WAITOK);
 1206         error = copyinstr(uap->path, pathbuf, MNAMELEN, NULL);
 1207         if (error) {
 1208                 free(pathbuf, M_TEMP);
 1209                 return (error);
 1210         }
 1211         if (uap->flags & MNT_BYFSID) {
 1212                 AUDIT_ARG_TEXT(pathbuf);
 1213                 /* Decode the filesystem ID. */
 1214                 if (sscanf(pathbuf, "FSID:%d:%d", &id0, &id1) != 2) {
 1215                         free(pathbuf, M_TEMP);
 1216                         return (EINVAL);
 1217                 }
 1218 
 1219                 mtx_lock(&mountlist_mtx);
 1220                 TAILQ_FOREACH_REVERSE(mp, &mountlist, mntlist, mnt_list) {
 1221                         if (mp->mnt_stat.f_fsid.val[0] == id0 &&
 1222                             mp->mnt_stat.f_fsid.val[1] == id1) {
 1223                                 vfs_ref(mp);
 1224                                 break;
 1225                         }
 1226                 }
 1227                 mtx_unlock(&mountlist_mtx);
 1228         } else {
 1229                 /*
 1230                  * Try to find global path for path argument.
 1231                  */
 1232                 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1,
 1233                     UIO_SYSSPACE, pathbuf, td);
 1234                 if (namei(&nd) == 0) {
 1235                         NDFREE(&nd, NDF_ONLY_PNBUF);
 1236                         error = vn_path_to_global_path(td, nd.ni_vp, pathbuf,
 1237                             MNAMELEN);
 1238                         if (error == 0 || error == ENODEV)
 1239                                 vput(nd.ni_vp);
 1240                 }
 1241                 mtx_lock(&mountlist_mtx);
 1242                 TAILQ_FOREACH_REVERSE(mp, &mountlist, mntlist, mnt_list) {
 1243                         if (strcmp(mp->mnt_stat.f_mntonname, pathbuf) == 0) {
 1244                                 vfs_ref(mp);
 1245                                 break;
 1246                         }
 1247                 }
 1248                 mtx_unlock(&mountlist_mtx);
 1249         }
 1250         free(pathbuf, M_TEMP);
 1251         if (mp == NULL) {
 1252                 /*
 1253                  * Previously we returned ENOENT for a nonexistent path and
 1254                  * EINVAL for a non-mountpoint.  We cannot tell these apart
 1255                  * now, so in the !MNT_BYFSID case return the more likely
 1256                  * EINVAL for compatibility.
 1257                  */
 1258                 return ((uap->flags & MNT_BYFSID) ? ENOENT : EINVAL);
 1259         }
 1260 
 1261         /*
 1262          * Don't allow unmounting the root filesystem.
 1263          */
 1264         if (mp->mnt_flag & MNT_ROOTFS) {
 1265                 vfs_rel(mp);
 1266                 return (EINVAL);
 1267         }
 1268         error = dounmount(mp, uap->flags, td);
 1269         return (error);
 1270 }
 1271 
 1272 /*
 1273  * Do the actual filesystem unmount.
 1274  */
 1275 int
 1276 dounmount(struct mount *mp, int flags, struct thread *td)
 1277 {
 1278         struct vnode *coveredvp, *fsrootvp;
 1279         int error;
 1280         uint64_t async_flag;
 1281         int mnt_gen_r;
 1282 
 1283         if ((coveredvp = mp->mnt_vnodecovered) != NULL) {
 1284                 mnt_gen_r = mp->mnt_gen;
 1285                 VI_LOCK(coveredvp);
 1286                 vholdl(coveredvp);
 1287                 vn_lock(coveredvp, LK_EXCLUSIVE | LK_INTERLOCK | LK_RETRY);
 1288                 /*
 1289                  * Check for mp being unmounted while waiting for the
 1290                  * covered vnode lock.
 1291                  */
 1292                 if (coveredvp->v_mountedhere != mp ||
 1293                     coveredvp->v_mountedhere->mnt_gen != mnt_gen_r) {
 1294                         VOP_UNLOCK(coveredvp, 0);
 1295                         vdrop(coveredvp);
 1296                         vfs_rel(mp);
 1297                         return (EBUSY);
 1298                 }
 1299         }
 1300 
 1301         /*
 1302          * Only privileged root, or (if MNT_USER is set) the user that did the
 1303          * original mount is permitted to unmount this filesystem.
 1304          */
 1305         error = vfs_suser(mp, td);
 1306         if (error != 0) {
 1307                 if (coveredvp != NULL) {
 1308                         VOP_UNLOCK(coveredvp, 0);
 1309                         vdrop(coveredvp);
 1310                 }
 1311                 vfs_rel(mp);
 1312                 return (error);
 1313         }
 1314 
 1315         vn_start_write(NULL, &mp, V_WAIT | V_MNTREF);
 1316         MNT_ILOCK(mp);
 1317         if ((mp->mnt_kern_flag & MNTK_UNMOUNT) != 0 ||
 1318             (mp->mnt_flag & MNT_UPDATE) != 0 ||
 1319             !TAILQ_EMPTY(&mp->mnt_uppers)) {
 1320                 MNT_IUNLOCK(mp);
 1321                 if (coveredvp != NULL) {
 1322                         VOP_UNLOCK(coveredvp, 0);
 1323                         vdrop(coveredvp);
 1324                 }
 1325                 vn_finished_write(mp);
 1326                 return (EBUSY);
 1327         }
 1328         mp->mnt_kern_flag |= MNTK_UNMOUNT | MNTK_NOINSMNTQ;
 1329         /* Allow filesystems to detect that a forced unmount is in progress. */
 1330         if (flags & MNT_FORCE) {
 1331                 mp->mnt_kern_flag |= MNTK_UNMOUNTF;
 1332                 MNT_IUNLOCK(mp);
 1333                 /*
 1334                  * Must be done after setting MNTK_UNMOUNTF and before
 1335                  * waiting for mnt_lockref to become 0.
 1336                  */
 1337                 VFS_PURGE(mp);
 1338                 MNT_ILOCK(mp);
 1339         }
 1340         error = 0;
 1341         if (mp->mnt_lockref) {
 1342                 mp->mnt_kern_flag |= MNTK_DRAINING;
 1343                 error = msleep(&mp->mnt_lockref, MNT_MTX(mp), PVFS,
 1344                     "mount drain", 0);
 1345         }
 1346         MNT_IUNLOCK(mp);
 1347         KASSERT(mp->mnt_lockref == 0,
 1348             ("%s: invalid lock refcount in the drain path @ %s:%d",
 1349             __func__, __FILE__, __LINE__));
 1350         KASSERT(error == 0,
 1351             ("%s: invalid return value for msleep in the drain path @ %s:%d",
 1352             __func__, __FILE__, __LINE__));
 1353 
 1354         if (mp->mnt_flag & MNT_EXPUBLIC)
 1355                 vfs_setpublicfs(NULL, NULL, NULL);
 1356 
 1357         /*
 1358          * From now, we can claim that the use reference on the
 1359          * coveredvp is ours, and the ref can be released only by
 1360          * successfull unmount by us, or left for later unmount
 1361          * attempt.  The previously acquired hold reference is no
 1362          * longer needed to protect the vnode from reuse.
 1363          */
 1364         if (coveredvp != NULL)
 1365                 vdrop(coveredvp);
 1366 
 1367         vfs_msync(mp, MNT_WAIT);
 1368         MNT_ILOCK(mp);
 1369         async_flag = mp->mnt_flag & MNT_ASYNC;
 1370         mp->mnt_flag &= ~MNT_ASYNC;
 1371         mp->mnt_kern_flag &= ~MNTK_ASYNC;
 1372         MNT_IUNLOCK(mp);
 1373         cache_purgevfs(mp);     /* remove cache entries for this file sys */
 1374         vfs_deallocate_syncvnode(mp);
 1375         /*
 1376          * For forced unmounts, move process cdir/rdir refs on the fs root
 1377          * vnode to the covered vnode.  For non-forced unmounts we want
 1378          * such references to cause an EBUSY error.
 1379          */
 1380         if ((flags & MNT_FORCE) &&
 1381             VFS_ROOT(mp, LK_EXCLUSIVE, &fsrootvp) == 0) {
 1382                 if (mp->mnt_vnodecovered != NULL &&
 1383                     (mp->mnt_flag & MNT_IGNORE) == 0)
 1384                         mountcheckdirs(fsrootvp, mp->mnt_vnodecovered);
 1385                 if (fsrootvp == rootvnode) {
 1386                         vrele(rootvnode);
 1387                         rootvnode = NULL;
 1388                 }
 1389                 vput(fsrootvp);
 1390         }
 1391         if ((mp->mnt_flag & MNT_RDONLY) != 0 || (flags & MNT_FORCE) != 0 ||
 1392             (error = VFS_SYNC(mp, MNT_WAIT)) == 0)
 1393                 error = VFS_UNMOUNT(mp, flags);
 1394         vn_finished_write(mp);
 1395         /*
 1396          * If we failed to flush the dirty blocks for this mount point,
 1397          * undo all the cdir/rdir and rootvnode changes we made above.
 1398          * Unless we failed to do so because the device is reporting that
 1399          * it doesn't exist anymore.
 1400          */
 1401         if (error && error != ENXIO) {
 1402                 if ((flags & MNT_FORCE) &&
 1403                     VFS_ROOT(mp, LK_EXCLUSIVE, &fsrootvp) == 0) {
 1404                         if (mp->mnt_vnodecovered != NULL &&
 1405                             (mp->mnt_flag & MNT_IGNORE) == 0)
 1406                                 mountcheckdirs(mp->mnt_vnodecovered, fsrootvp);
 1407                         if (rootvnode == NULL) {
 1408                                 rootvnode = fsrootvp;
 1409                                 vref(rootvnode);
 1410                         }
 1411                         vput(fsrootvp);
 1412                 }
 1413                 MNT_ILOCK(mp);
 1414                 mp->mnt_kern_flag &= ~MNTK_NOINSMNTQ;
 1415                 if ((mp->mnt_flag & MNT_RDONLY) == 0) {
 1416                         MNT_IUNLOCK(mp);
 1417                         vfs_allocate_syncvnode(mp);
 1418                         MNT_ILOCK(mp);
 1419                 }
 1420                 mp->mnt_kern_flag &= ~(MNTK_UNMOUNT | MNTK_UNMOUNTF);
 1421                 mp->mnt_flag |= async_flag;
 1422                 if ((mp->mnt_flag & MNT_ASYNC) != 0 &&
 1423                     (mp->mnt_kern_flag & MNTK_NOASYNC) == 0)
 1424                         mp->mnt_kern_flag |= MNTK_ASYNC;
 1425                 if (mp->mnt_kern_flag & MNTK_MWAIT) {
 1426                         mp->mnt_kern_flag &= ~MNTK_MWAIT;
 1427                         wakeup(mp);
 1428                 }
 1429                 MNT_IUNLOCK(mp);
 1430                 if (coveredvp)
 1431                         VOP_UNLOCK(coveredvp, 0);
 1432                 return (error);
 1433         }
 1434         mtx_lock(&mountlist_mtx);
 1435         TAILQ_REMOVE(&mountlist, mp, mnt_list);
 1436         mtx_unlock(&mountlist_mtx);
 1437         EVENTHANDLER_INVOKE(vfs_unmounted, mp, td);
 1438         if (coveredvp != NULL) {
 1439                 coveredvp->v_mountedhere = NULL;
 1440                 VOP_UNLOCK(coveredvp, 0);
 1441         }
 1442         vfs_event_signal(NULL, VQ_UNMOUNT, 0);
 1443         if (mp == rootdevmp)
 1444                 rootdevmp = NULL;
 1445         vfs_mount_destroy(mp);
 1446         return (0);
 1447 }
 1448 
 1449 /*
 1450  * Report errors during filesystem mounting.
 1451  */
 1452 void
 1453 vfs_mount_error(struct mount *mp, const char *fmt, ...)
 1454 {
 1455         struct vfsoptlist *moptlist = mp->mnt_optnew;
 1456         va_list ap;
 1457         int error, len;
 1458         char *errmsg;
 1459 
 1460         error = vfs_getopt(moptlist, "errmsg", (void **)&errmsg, &len);
 1461         if (error || errmsg == NULL || len <= 0)
 1462                 return;
 1463 
 1464         va_start(ap, fmt);
 1465         vsnprintf(errmsg, (size_t)len, fmt, ap);
 1466         va_end(ap);
 1467 }
 1468 
 1469 void
 1470 vfs_opterror(struct vfsoptlist *opts, const char *fmt, ...)
 1471 {
 1472         va_list ap;
 1473         int error, len;
 1474         char *errmsg;
 1475 
 1476         error = vfs_getopt(opts, "errmsg", (void **)&errmsg, &len);
 1477         if (error || errmsg == NULL || len <= 0)
 1478                 return;
 1479 
 1480         va_start(ap, fmt);
 1481         vsnprintf(errmsg, (size_t)len, fmt, ap);
 1482         va_end(ap);
 1483 }
 1484 
 1485 /*
 1486  * ---------------------------------------------------------------------
 1487  * Functions for querying mount options/arguments from filesystems.
 1488  */
 1489 
 1490 /*
 1491  * Check that no unknown options are given
 1492  */
 1493 int
 1494 vfs_filteropt(struct vfsoptlist *opts, const char **legal)
 1495 {
 1496         struct vfsopt *opt;
 1497         char errmsg[255];
 1498         const char **t, *p, *q;
 1499         int ret = 0;
 1500 
 1501         TAILQ_FOREACH(opt, opts, link) {
 1502                 p = opt->name;
 1503                 q = NULL;
 1504                 if (p[0] == 'n' && p[1] == 'o')
 1505                         q = p + 2;
 1506                 for(t = global_opts; *t != NULL; t++) {
 1507                         if (strcmp(*t, p) == 0)
 1508                                 break;
 1509                         if (q != NULL) {
 1510                                 if (strcmp(*t, q) == 0)
 1511                                         break;
 1512                         }
 1513                 }
 1514                 if (*t != NULL)
 1515                         continue;
 1516                 for(t = legal; *t != NULL; t++) {
 1517                         if (strcmp(*t, p) == 0)
 1518                                 break;
 1519                         if (q != NULL) {
 1520                                 if (strcmp(*t, q) == 0)
 1521                                         break;
 1522                         }
 1523                 }
 1524                 if (*t != NULL)
 1525                         continue;
 1526                 snprintf(errmsg, sizeof(errmsg),
 1527                     "mount option <%s> is unknown", p);
 1528                 ret = EINVAL;
 1529         }
 1530         if (ret != 0) {
 1531                 TAILQ_FOREACH(opt, opts, link) {
 1532                         if (strcmp(opt->name, "errmsg") == 0) {
 1533                                 strncpy((char *)opt->value, errmsg, opt->len);
 1534                                 break;
 1535                         }
 1536                 }
 1537                 if (opt == NULL)
 1538                         printf("%s\n", errmsg);
 1539         }
 1540         return (ret);
 1541 }
 1542 
 1543 /*
 1544  * Get a mount option by its name.
 1545  *
 1546  * Return 0 if the option was found, ENOENT otherwise.
 1547  * If len is non-NULL it will be filled with the length
 1548  * of the option. If buf is non-NULL, it will be filled
 1549  * with the address of the option.
 1550  */
 1551 int
 1552 vfs_getopt(opts, name, buf, len)
 1553         struct vfsoptlist *opts;
 1554         const char *name;
 1555         void **buf;
 1556         int *len;
 1557 {
 1558         struct vfsopt *opt;
 1559 
 1560         KASSERT(opts != NULL, ("vfs_getopt: caller passed 'opts' as NULL"));
 1561 
 1562         TAILQ_FOREACH(opt, opts, link) {
 1563                 if (strcmp(name, opt->name) == 0) {
 1564                         opt->seen = 1;
 1565                         if (len != NULL)
 1566                                 *len = opt->len;
 1567                         if (buf != NULL)
 1568                                 *buf = opt->value;
 1569                         return (0);
 1570                 }
 1571         }
 1572         return (ENOENT);
 1573 }
 1574 
 1575 int
 1576 vfs_getopt_pos(struct vfsoptlist *opts, const char *name)
 1577 {
 1578         struct vfsopt *opt;
 1579 
 1580         if (opts == NULL)
 1581                 return (-1);
 1582 
 1583         TAILQ_FOREACH(opt, opts, link) {
 1584                 if (strcmp(name, opt->name) == 0) {
 1585                         opt->seen = 1;
 1586                         return (opt->pos);
 1587                 }
 1588         }
 1589         return (-1);
 1590 }
 1591 
 1592 int
 1593 vfs_getopt_size(struct vfsoptlist *opts, const char *name, off_t *value)
 1594 {
 1595         char *opt_value, *vtp;
 1596         quad_t iv;
 1597         int error, opt_len;
 1598 
 1599         error = vfs_getopt(opts, name, (void **)&opt_value, &opt_len);
 1600         if (error != 0)
 1601                 return (error);
 1602         if (opt_len == 0 || opt_value == NULL)
 1603                 return (EINVAL);
 1604         if (opt_value[0] == '\0' || opt_value[opt_len - 1] != '\0')
 1605                 return (EINVAL);
 1606         iv = strtoq(opt_value, &vtp, 0);
 1607         if (vtp == opt_value || (vtp[0] != '\0' && vtp[1] != '\0'))
 1608                 return (EINVAL);
 1609         if (iv < 0)
 1610                 return (EINVAL);
 1611         switch (vtp[0]) {
 1612         case 't':
 1613         case 'T':
 1614                 iv *= 1024;
 1615         case 'g':
 1616         case 'G':
 1617                 iv *= 1024;
 1618         case 'm':
 1619         case 'M':
 1620                 iv *= 1024;
 1621         case 'k':
 1622         case 'K':
 1623                 iv *= 1024;
 1624         case '\0':
 1625                 break;
 1626         default:
 1627                 return (EINVAL);
 1628         }
 1629         *value = iv;
 1630 
 1631         return (0);
 1632 }
 1633 
 1634 char *
 1635 vfs_getopts(struct vfsoptlist *opts, const char *name, int *error)
 1636 {
 1637         struct vfsopt *opt;
 1638 
 1639         *error = 0;
 1640         TAILQ_FOREACH(opt, opts, link) {
 1641                 if (strcmp(name, opt->name) != 0)
 1642                         continue;
 1643                 opt->seen = 1;
 1644                 if (opt->len == 0 ||
 1645                     ((char *)opt->value)[opt->len - 1] != '\0') {
 1646                         *error = EINVAL;
 1647                         return (NULL);
 1648                 }
 1649                 return (opt->value);
 1650         }
 1651         *error = ENOENT;
 1652         return (NULL);
 1653 }
 1654 
 1655 int
 1656 vfs_flagopt(struct vfsoptlist *opts, const char *name, uint64_t *w,
 1657         uint64_t val)
 1658 {
 1659         struct vfsopt *opt;
 1660 
 1661         TAILQ_FOREACH(opt, opts, link) {
 1662                 if (strcmp(name, opt->name) == 0) {
 1663                         opt->seen = 1;
 1664                         if (w != NULL)
 1665                                 *w |= val;
 1666                         return (1);
 1667                 }
 1668         }
 1669         if (w != NULL)
 1670                 *w &= ~val;
 1671         return (0);
 1672 }
 1673 
 1674 int
 1675 vfs_scanopt(struct vfsoptlist *opts, const char *name, const char *fmt, ...)
 1676 {
 1677         va_list ap;
 1678         struct vfsopt *opt;
 1679         int ret;
 1680 
 1681         KASSERT(opts != NULL, ("vfs_getopt: caller passed 'opts' as NULL"));
 1682 
 1683         TAILQ_FOREACH(opt, opts, link) {
 1684                 if (strcmp(name, opt->name) != 0)
 1685                         continue;
 1686                 opt->seen = 1;
 1687                 if (opt->len == 0 || opt->value == NULL)
 1688                         return (0);
 1689                 if (((char *)opt->value)[opt->len - 1] != '\0')
 1690                         return (0);
 1691                 va_start(ap, fmt);
 1692                 ret = vsscanf(opt->value, fmt, ap);
 1693                 va_end(ap);
 1694                 return (ret);
 1695         }
 1696         return (0);
 1697 }
 1698 
 1699 int
 1700 vfs_setopt(struct vfsoptlist *opts, const char *name, void *value, int len)
 1701 {
 1702         struct vfsopt *opt;
 1703 
 1704         TAILQ_FOREACH(opt, opts, link) {
 1705                 if (strcmp(name, opt->name) != 0)
 1706                         continue;
 1707                 opt->seen = 1;
 1708                 if (opt->value == NULL)
 1709                         opt->len = len;
 1710                 else {
 1711                         if (opt->len != len)
 1712                                 return (EINVAL);
 1713                         bcopy(value, opt->value, len);
 1714                 }
 1715                 return (0);
 1716         }
 1717         return (ENOENT);
 1718 }
 1719 
 1720 int
 1721 vfs_setopt_part(struct vfsoptlist *opts, const char *name, void *value, int len)
 1722 {
 1723         struct vfsopt *opt;
 1724 
 1725         TAILQ_FOREACH(opt, opts, link) {
 1726                 if (strcmp(name, opt->name) != 0)
 1727                         continue;
 1728                 opt->seen = 1;
 1729                 if (opt->value == NULL)
 1730                         opt->len = len;
 1731                 else {
 1732                         if (opt->len < len)
 1733                                 return (EINVAL);
 1734                         opt->len = len;
 1735                         bcopy(value, opt->value, len);
 1736                 }
 1737                 return (0);
 1738         }
 1739         return (ENOENT);
 1740 }
 1741 
 1742 int
 1743 vfs_setopts(struct vfsoptlist *opts, const char *name, const char *value)
 1744 {
 1745         struct vfsopt *opt;
 1746 
 1747         TAILQ_FOREACH(opt, opts, link) {
 1748                 if (strcmp(name, opt->name) != 0)
 1749                         continue;
 1750                 opt->seen = 1;
 1751                 if (opt->value == NULL)
 1752                         opt->len = strlen(value) + 1;
 1753                 else if (strlcpy(opt->value, value, opt->len) >= opt->len)
 1754                         return (EINVAL);
 1755                 return (0);
 1756         }
 1757         return (ENOENT);
 1758 }
 1759 
 1760 /*
 1761  * Find and copy a mount option.
 1762  *
 1763  * The size of the buffer has to be specified
 1764  * in len, if it is not the same length as the
 1765  * mount option, EINVAL is returned.
 1766  * Returns ENOENT if the option is not found.
 1767  */
 1768 int
 1769 vfs_copyopt(opts, name, dest, len)
 1770         struct vfsoptlist *opts;
 1771         const char *name;
 1772         void *dest;
 1773         int len;
 1774 {
 1775         struct vfsopt *opt;
 1776 
 1777         KASSERT(opts != NULL, ("vfs_copyopt: caller passed 'opts' as NULL"));
 1778 
 1779         TAILQ_FOREACH(opt, opts, link) {
 1780                 if (strcmp(name, opt->name) == 0) {
 1781                         opt->seen = 1;
 1782                         if (len != opt->len)
 1783                                 return (EINVAL);
 1784                         bcopy(opt->value, dest, opt->len);
 1785                         return (0);
 1786                 }
 1787         }
 1788         return (ENOENT);
 1789 }
 1790 
 1791 int
 1792 __vfs_statfs(struct mount *mp, struct statfs *sbp)
 1793 {
 1794         int error;
 1795 
 1796         error = mp->mnt_op->vfs_statfs(mp, &mp->mnt_stat);
 1797         if (sbp != &mp->mnt_stat)
 1798                 *sbp = mp->mnt_stat;
 1799         return (error);
 1800 }
 1801 
 1802 void
 1803 vfs_mountedfrom(struct mount *mp, const char *from)
 1804 {
 1805 
 1806         bzero(mp->mnt_stat.f_mntfromname, sizeof mp->mnt_stat.f_mntfromname);
 1807         strlcpy(mp->mnt_stat.f_mntfromname, from,
 1808             sizeof mp->mnt_stat.f_mntfromname);
 1809 }
 1810 
 1811 /*
 1812  * ---------------------------------------------------------------------
 1813  * This is the api for building mount args and mounting filesystems from
 1814  * inside the kernel.
 1815  *
 1816  * The API works by accumulation of individual args.  First error is
 1817  * latched.
 1818  *
 1819  * XXX: should be documented in new manpage kernel_mount(9)
 1820  */
 1821 
 1822 /* A memory allocation which must be freed when we are done */
 1823 struct mntaarg {
 1824         SLIST_ENTRY(mntaarg)    next;
 1825 };
 1826 
 1827 /* The header for the mount arguments */
 1828 struct mntarg {
 1829         struct iovec *v;
 1830         int len;
 1831         int error;
 1832         SLIST_HEAD(, mntaarg)   list;
 1833 };
 1834 
 1835 /*
 1836  * Add a boolean argument.
 1837  *
 1838  * flag is the boolean value.
 1839  * name must start with "no".
 1840  */
 1841 struct mntarg *
 1842 mount_argb(struct mntarg *ma, int flag, const char *name)
 1843 {
 1844 
 1845         KASSERT(name[0] == 'n' && name[1] == 'o',
 1846             ("mount_argb(...,%s): name must start with 'no'", name));
 1847 
 1848         return (mount_arg(ma, name + (flag ? 2 : 0), NULL, 0));
 1849 }
 1850 
 1851 /*
 1852  * Add an argument printf style
 1853  */
 1854 struct mntarg *
 1855 mount_argf(struct mntarg *ma, const char *name, const char *fmt, ...)
 1856 {
 1857         va_list ap;
 1858         struct mntaarg *maa;
 1859         struct sbuf *sb;
 1860         int len;
 1861 
 1862         if (ma == NULL) {
 1863                 ma = malloc(sizeof *ma, M_MOUNT, M_WAITOK | M_ZERO);
 1864                 SLIST_INIT(&ma->list);
 1865         }
 1866         if (ma->error)
 1867                 return (ma);
 1868 
 1869         ma->v = realloc(ma->v, sizeof *ma->v * (ma->len + 2),
 1870             M_MOUNT, M_WAITOK);
 1871         ma->v[ma->len].iov_base = (void *)(uintptr_t)name;
 1872         ma->v[ma->len].iov_len = strlen(name) + 1;
 1873         ma->len++;
 1874 
 1875         sb = sbuf_new_auto();
 1876         va_start(ap, fmt);
 1877         sbuf_vprintf(sb, fmt, ap);
 1878         va_end(ap);
 1879         sbuf_finish(sb);
 1880         len = sbuf_len(sb) + 1;
 1881         maa = malloc(sizeof *maa + len, M_MOUNT, M_WAITOK | M_ZERO);
 1882         SLIST_INSERT_HEAD(&ma->list, maa, next);
 1883         bcopy(sbuf_data(sb), maa + 1, len);
 1884         sbuf_delete(sb);
 1885 
 1886         ma->v[ma->len].iov_base = maa + 1;
 1887         ma->v[ma->len].iov_len = len;
 1888         ma->len++;
 1889 
 1890         return (ma);
 1891 }
 1892 
 1893 /*
 1894  * Add an argument which is a userland string.
 1895  */
 1896 struct mntarg *
 1897 mount_argsu(struct mntarg *ma, const char *name, const void *val, int len)
 1898 {
 1899         struct mntaarg *maa;
 1900         char *tbuf;
 1901 
 1902         if (val == NULL)
 1903                 return (ma);
 1904         if (ma == NULL) {
 1905                 ma = malloc(sizeof *ma, M_MOUNT, M_WAITOK | M_ZERO);
 1906                 SLIST_INIT(&ma->list);
 1907         }
 1908         if (ma->error)
 1909                 return (ma);
 1910         maa = malloc(sizeof *maa + len, M_MOUNT, M_WAITOK | M_ZERO);
 1911         SLIST_INSERT_HEAD(&ma->list, maa, next);
 1912         tbuf = (void *)(maa + 1);
 1913         ma->error = copyinstr(val, tbuf, len, NULL);
 1914         return (mount_arg(ma, name, tbuf, -1));
 1915 }
 1916 
 1917 /*
 1918  * Plain argument.
 1919  *
 1920  * If length is -1, treat value as a C string.
 1921  */
 1922 struct mntarg *
 1923 mount_arg(struct mntarg *ma, const char *name, const void *val, int len)
 1924 {
 1925 
 1926         if (ma == NULL) {
 1927                 ma = malloc(sizeof *ma, M_MOUNT, M_WAITOK | M_ZERO);
 1928                 SLIST_INIT(&ma->list);
 1929         }
 1930         if (ma->error)
 1931                 return (ma);
 1932 
 1933         ma->v = realloc(ma->v, sizeof *ma->v * (ma->len + 2),
 1934             M_MOUNT, M_WAITOK);
 1935         ma->v[ma->len].iov_base = (void *)(uintptr_t)name;
 1936         ma->v[ma->len].iov_len = strlen(name) + 1;
 1937         ma->len++;
 1938 
 1939         ma->v[ma->len].iov_base = (void *)(uintptr_t)val;
 1940         if (len < 0)
 1941                 ma->v[ma->len].iov_len = strlen(val) + 1;
 1942         else
 1943                 ma->v[ma->len].iov_len = len;
 1944         ma->len++;
 1945         return (ma);
 1946 }
 1947 
 1948 /*
 1949  * Free a mntarg structure
 1950  */
 1951 static void
 1952 free_mntarg(struct mntarg *ma)
 1953 {
 1954         struct mntaarg *maa;
 1955 
 1956         while (!SLIST_EMPTY(&ma->list)) {
 1957                 maa = SLIST_FIRST(&ma->list);
 1958                 SLIST_REMOVE_HEAD(&ma->list, next);
 1959                 free(maa, M_MOUNT);
 1960         }
 1961         free(ma->v, M_MOUNT);
 1962         free(ma, M_MOUNT);
 1963 }
 1964 
 1965 /*
 1966  * Mount a filesystem
 1967  */
 1968 int
 1969 kernel_mount(struct mntarg *ma, uint64_t flags)
 1970 {
 1971         struct uio auio;
 1972         int error;
 1973 
 1974         KASSERT(ma != NULL, ("kernel_mount NULL ma"));
 1975         KASSERT(ma->v != NULL, ("kernel_mount NULL ma->v"));
 1976         KASSERT(!(ma->len & 1), ("kernel_mount odd ma->len (%d)", ma->len));
 1977 
 1978         auio.uio_iov = ma->v;
 1979         auio.uio_iovcnt = ma->len;
 1980         auio.uio_segflg = UIO_SYSSPACE;
 1981 
 1982         error = ma->error;
 1983         if (!error)
 1984                 error = vfs_donmount(curthread, flags, &auio);
 1985         free_mntarg(ma);
 1986         return (error);
 1987 }
 1988 
 1989 /*
 1990  * A printflike function to mount a filesystem.
 1991  */
 1992 int
 1993 kernel_vmount(int flags, ...)
 1994 {
 1995         struct mntarg *ma = NULL;
 1996         va_list ap;
 1997         const char *cp;
 1998         const void *vp;
 1999         int error;
 2000 
 2001         va_start(ap, flags);
 2002         for (;;) {
 2003                 cp = va_arg(ap, const char *);
 2004                 if (cp == NULL)
 2005                         break;
 2006                 vp = va_arg(ap, const void *);
 2007                 ma = mount_arg(ma, cp, vp, (vp != NULL ? -1 : 0));
 2008         }
 2009         va_end(ap);
 2010 
 2011         error = kernel_mount(ma, flags);
 2012         return (error);
 2013 }
 2014 
 2015 void
 2016 vfs_oexport_conv(const struct oexport_args *oexp, struct export_args *exp)
 2017 {
 2018 
 2019         bcopy(oexp, exp, sizeof(*oexp));
 2020         exp->ex_numsecflavors = 0;
 2021 }

Cache object: 493d7172f58f6ccc77c51c73117af14c


[ 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.