module/os/freebsd/spl/spl_vfs.c

   1 /*
   2  * Copyright (c) 2006-2007 Pawel Jakub Dawidek <pjd@FreeBSD.org>
   3  * All rights reserved.
   4  *
   5  * Redistribution and use in source and binary forms, with or without
   6  * modification, are permitted provided that the following conditions
   7  * are met:
   8  * 1. Redistributions of source code must retain the above copyright
   9  *    notice, this list of conditions and the following disclaimer.
  10  * 2. Redistributions in binary form must reproduce the above copyright
  11  *    notice, this list of conditions and the following disclaimer in the
  12  *    documentation and/or other materials provided with the distribution.
  13  *
  14  * THIS SOFTWARE IS PROVIDED BY THE AUTHORS AND CONTRIBUTORS ``AS IS'' AND
  15  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
  16  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  17  * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHORS OR CONTRIBUTORS BE LIABLE
  18  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
  19  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
  20  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
  21  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
  22  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
  23  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
  24  * SUCH DAMAGE.
  25  */
  26
  27 #include <sys/cdefs.h>
  28 __FBSDID("$FreeBSD$");
  29
  30 #include <sys/types.h>
  31 #include <sys/param.h>
  32 #include <sys/kernel.h>
  33 #include <sys/systm.h>
  34 #include <sys/malloc.h>
  35 #include <sys/mount.h>
  36 #include <sys/cred.h>
  37 #include <sys/vfs.h>
  38 #include <sys/priv.h>
  39 #include <sys/libkern.h>
  40
  41 #include <sys/mutex.h>
  42 #include <sys/vnode.h>
  43 #include <sys/taskq.h>
  44
  45 #include <sys/ccompat.h>
  46
  47 MALLOC_DECLARE(M_MOUNT);
  48
  49 void
  50 vfs_setmntopt(vfs_t *vfsp, const char *name, const char *arg,
  51     int flags __unused)
  52 {
  53         struct vfsopt *opt;
  54         size_t namesize;
  55         int locked;
  56
  57         if (!(locked = mtx_owned(MNT_MTX(vfsp))))
  58                 MNT_ILOCK(vfsp);
  59
  60         if (vfsp->mnt_opt == NULL) {
  61                 void *opts;
  62
  63                 MNT_IUNLOCK(vfsp);
  64                 opts = malloc(sizeof (*vfsp->mnt_opt), M_MOUNT, M_WAITOK);
  65                 MNT_ILOCK(vfsp);
  66                 if (vfsp->mnt_opt == NULL) {
  67                         vfsp->mnt_opt = opts;
  68                         TAILQ_INIT(vfsp->mnt_opt);
  69                 } else {
  70                         free(opts, M_MOUNT);
  71                 }
  72         }
  73
  74         MNT_IUNLOCK(vfsp);
  75
  76         opt = malloc(sizeof (*opt), M_MOUNT, M_WAITOK);
  77         namesize = strlen(name) + 1;
  78         opt->name = malloc(namesize, M_MOUNT, M_WAITOK);
  79         strlcpy(opt->name, name, namesize);
  80         opt->pos = -1;
  81         opt->seen = 1;
  82         if (arg == NULL) {
  83                 opt->value = NULL;
  84                 opt->len = 0;
  85         } else {
  86                 opt->len = strlen(arg) + 1;
  87                 opt->value = malloc(opt->len, M_MOUNT, M_WAITOK);
  88                 memcpy(opt->value, arg, opt->len);
  89         }
  90
  91         MNT_ILOCK(vfsp);
  92         TAILQ_INSERT_TAIL(vfsp->mnt_opt, opt, link);
  93         if (!locked)
  94                 MNT_IUNLOCK(vfsp);
  95 }
  96
  97 void
  98 vfs_clearmntopt(vfs_t *vfsp, const char *name)
  99 {
 100         int locked;
 101
 102         if (!(locked = mtx_owned(MNT_MTX(vfsp))))
 103                 MNT_ILOCK(vfsp);
 104         vfs_deleteopt(vfsp->mnt_opt, name);
 105         if (!locked)
 106                 MNT_IUNLOCK(vfsp);
 107 }
 108
 109 int
 110 vfs_optionisset(const vfs_t *vfsp, const char *opt, char **argp)
 111 {
 112         struct vfsoptlist *opts = vfsp->mnt_optnew;
 113         int error;
 114
 115         if (opts == NULL)
 116                 return (0);
 117         error = vfs_getopt(opts, opt, (void **)argp, NULL);
 118         return (error != 0 ? 0 : 1);
 119 }
 120
 121 int
 122 mount_snapshot(kthread_t *td, vnode_t **vpp, const char *fstype, char *fspath,
 123     char *fspec, int fsflags)
 124 {
 125         struct vfsconf *vfsp;
 126         struct mount *mp;
 127         vnode_t *vp, *mvp;
 128         int error;
 129
 130         ASSERT_VOP_ELOCKED(*vpp, "mount_snapshot");
 131
 132         vp = *vpp;
 133         *vpp = NULL;
 134         error = 0;
 135
 136         /*
 137          * Be ultra-paranoid about making sure the type and fspath
 138          * variables will fit in our mp buffers, including the
 139          * terminating NUL.
 140          */
 141         if (strlen(fstype) >= MFSNAMELEN || strlen(fspath) >= MNAMELEN)
 142                 error = ENAMETOOLONG;
 143         if (error == 0 && (vfsp = vfs_byname_kld(fstype, td, &error)) == NULL)
 144                 error = ENODEV;
 145         if (error == 0 && vp->v_type != VDIR)
 146                 error = ENOTDIR;
 147         /*
 148          * We need vnode lock to protect v_mountedhere and vnode interlock
 149          * to protect v_iflag.
 150          */
 151         if (error == 0) {
 152                 VI_LOCK(vp);
 153                 if ((vp->v_iflag & VI_MOUNT) == 0 && vp->v_mountedhere == NULL)
 154                         vp->v_iflag |= VI_MOUNT;
 155                 else
 156                         error = EBUSY;
 157                 VI_UNLOCK(vp);
 158         }
 159         if (error != 0) {
 160                 vput(vp);
 161                 return (error);
 162         }
 163         vn_seqc_write_begin(vp);
 164         VOP_UNLOCK1(vp);
 165
 166         /*
 167          * Allocate and initialize the filesystem.
 168          * We don't want regular user that triggered snapshot mount to be able
 169          * to unmount it, so pass credentials of the parent mount.
 170          */
 171         mp = vfs_mount_alloc(vp, vfsp, fspath, vp->v_mount->mnt_cred);
 172
 173         mp->mnt_optnew = NULL;
 174         vfs_setmntopt(mp, "from", fspec, 0);
 175         mp->mnt_optnew = mp->mnt_opt;
 176         mp->mnt_opt = NULL;
 177
 178         /*
 179          * Set the mount level flags.
 180          */
 181         mp->mnt_flag = fsflags & MNT_UPDATEMASK;
 182         /*
 183          * Snapshots are always read-only.
 184          */
 185         mp->mnt_flag |= MNT_RDONLY;
 186         /*
 187          * We don't want snapshots to allow access to vulnerable setuid
 188          * programs, so we turn off setuid when mounting snapshots.
 189          */
 190         mp->mnt_flag |= MNT_NOSUID;
 191         /*
 192          * We don't want snapshots to be visible in regular
 193          * mount(8) and df(1) output.
 194          */
 195         mp->mnt_flag |= MNT_IGNORE;
 196
 197         error = VFS_MOUNT(mp);
 198         if (error != 0) {
 199                 /*
 200                  * Clear VI_MOUNT and decrement the use count "atomically",
 201                  * under the vnode lock.  This is not strictly required,
 202                  * but makes it easier to reason about the life-cycle and
 203                  * ownership of the covered vnode.
 204                  */
 205                 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
 206                 VI_LOCK(vp);
 207                 vp->v_iflag &= ~VI_MOUNT;
 208                 VI_UNLOCK(vp);
 209                 vn_seqc_write_end(vp);
 210                 vput(vp);
 211                 vfs_unbusy(mp);
 212                 vfs_freeopts(mp->mnt_optnew);
 213                 mp->mnt_vnodecovered = NULL;
 214                 vfs_mount_destroy(mp);
 215                 return (error);
 216         }
 217
 218         if (mp->mnt_opt != NULL)
 219                 vfs_freeopts(mp->mnt_opt);
 220         mp->mnt_opt = mp->mnt_optnew;
 221         (void) VFS_STATFS(mp, &mp->mnt_stat);
 222
 223         /*
 224          * Prevent external consumers of mount options from reading
 225          * mnt_optnew.
 226          */
 227         mp->mnt_optnew = NULL;
 228
 229         vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
 230 #ifdef FREEBSD_NAMECACHE
 231         cache_purge(vp);
 232 #endif
 233         VI_LOCK(vp);
 234         vp->v_iflag &= ~VI_MOUNT;
 235 #ifdef VIRF_MOUNTPOINT
 236         vn_irflag_set_locked(vp, VIRF_MOUNTPOINT);
 237 #endif
 238         vp->v_mountedhere = mp;
 239         VI_UNLOCK(vp);
 240         /* Put the new filesystem on the mount list. */
 241         mtx_lock(&mountlist_mtx);
 242         TAILQ_INSERT_TAIL(&mountlist, mp, mnt_list);
 243         mtx_unlock(&mountlist_mtx);
 244         vfs_event_signal(NULL, VQ_MOUNT, 0);
 245         if (VFS_ROOT(mp, LK_EXCLUSIVE, &mvp))
 246                 panic("mount: lost mount");
 247         vn_seqc_write_end(vp);
 248         VOP_UNLOCK1(vp);
 249 #if __FreeBSD_version >= 1300048
 250         vfs_op_exit(mp);
 251 #endif
 252         vfs_unbusy(mp);
 253         *vpp = mvp;
 254         return (0);
 255 }
 256
 257 /*
 258  * Like vn_rele() except if we are going to call VOP_INACTIVE() then do it
 259  * asynchronously using a taskq. This can avoid deadlocks caused by re-entering
 260  * the file system as a result of releasing the vnode. Note, file systems
 261  * already have to handle the race where the vnode is incremented before the
 262  * inactive routine is called and does its locking.
 263  *
 264  * Warning: Excessive use of this routine can lead to performance problems.
 265  * This is because taskqs throttle back allocation if too many are created.
 266  */
 267 void
 268 vn_rele_async(vnode_t *vp, taskq_t *taskq)
 269 {
 270         VERIFY3U(vp->v_usecount, >, 0);
 271         if (refcount_release_if_not_last(&vp->v_usecount)) {
 272 #if __FreeBSD_version < 1300045
 273                 vdrop(vp);
 274 #endif
 275                 return;
 276         }
 277         VERIFY3U(taskq_dispatch((taskq_t *)taskq,
 278             (task_func_t *)vrele, vp, TQ_SLEEP), !=, 0);
 279 }