module/zfs/vdev_file.c

   1 /*
   2  * CDDL HEADER START
   3  *
   4  * The contents of this file are subject to the terms of the
   5  * Common Development and Distribution License (the "License").
   6  * You may not use this file except in compliance with the License.
   7  *
   8  * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
   9  * or http://www.opensolaris.org/os/licensing.
  10  * See the License for the specific language governing permissions
  11  * and limitations under the License.
  12  *
  13  * When distributing Covered Code, include this CDDL HEADER in each
  14  * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
  15  * If applicable, add the following below this CDDL HEADER, with the
  16  * fields enclosed by brackets "[]" replaced with your own identifying
  17  * information: Portions Copyright [yyyy] [name of copyright owner]
  18  *
  19  * CDDL HEADER END
  20  */
  21 /*
  22  * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
  23  * Copyright (c) 2011, 2014 by Delphix. All rights reserved.
  24  */
  25
  26 #include <sys/zfs_context.h>
  27 #include <sys/spa.h>
  28 #include <sys/spa_impl.h>
  29 #include <sys/vdev_file.h>
  30 #include <sys/vdev_impl.h>
  31 #include <sys/zio.h>
  32 #include <sys/fs/zfs.h>
  33 #include <sys/fm/fs/zfs.h>
  34
  35 /*
  36  * Virtual device vector for files.
  37  */
  38
  39 static void
  40 vdev_file_hold(vdev_t *vd)
  41 {
  42         ASSERT(vd->vdev_path != NULL);
  43 }
  44
  45 static void
  46 vdev_file_rele(vdev_t *vd)
  47 {
  48         ASSERT(vd->vdev_path != NULL);
  49 }
  50
  51 static int
  52 vdev_file_open(vdev_t *vd, uint64_t *psize, uint64_t *max_psize,
  53     uint64_t *ashift)
  54 {
  55         vdev_file_t *vf;
  56         vnode_t *vp;
  57         vattr_t vattr;
  58         int error;
  59
  60         /*
  61          * We must have a pathname, and it must be absolute.
  62          */
  63         if (vd->vdev_path == NULL || vd->vdev_path[0] != '/') {
  64                 vd->vdev_stat.vs_aux = VDEV_AUX_BAD_LABEL;
  65                 return (SET_ERROR(EINVAL));
  66         }
  67
  68         /*
  69          * Reopen the device if it's not currently open.  Otherwise,
  70          * just update the physical size of the device.
  71          */
  72         if (vd->vdev_tsd != NULL) {
  73                 ASSERT(vd->vdev_reopening);
  74                 vf = vd->vdev_tsd;
  75                 goto skip_open;
  76         }
  77
  78         vf = vd->vdev_tsd = kmem_zalloc(sizeof (vdev_file_t), KM_SLEEP);
  79
  80         /*
  81          * We always open the files from the root of the global zone, even if
  82          * we're in a local zone.  If the user has gotten to this point, the
  83          * administrator has already decided that the pool should be available
  84          * to local zone users, so the underlying devices should be as well.
  85          */
  86         ASSERT(vd->vdev_path != NULL && vd->vdev_path[0] == '/');
  87         error = vn_openat(vd->vdev_path + 1, UIO_SYSSPACE,
  88             spa_mode(vd->vdev_spa) | FOFFMAX, 0, &vp, 0, 0, rootdir, -1);
  89
  90         if (error) {
  91                 vd->vdev_stat.vs_aux = VDEV_AUX_OPEN_FAILED;
  92                 return (error);
  93         }
  94
  95         vf->vf_vnode = vp;
  96
  97 #ifdef _KERNEL
  98         /*
  99          * Make sure it's a regular file.
 100          */
 101         if (vp->v_type != VREG) {
 102                 vd->vdev_stat.vs_aux = VDEV_AUX_OPEN_FAILED;
 103                 return (SET_ERROR(ENODEV));
 104         }
 105 #endif
 106
 107 skip_open:
 108         /*
 109          * Determine the physical size of the file.
 110          */
 111         vattr.va_mask = AT_SIZE;
 112         error = VOP_GETATTR(vf->vf_vnode, &vattr, 0, kcred, NULL);
 113         if (error) {
 114                 vd->vdev_stat.vs_aux = VDEV_AUX_OPEN_FAILED;
 115                 return (error);
 116         }
 117
 118         *max_psize = *psize = vattr.va_size;
 119         *ashift = SPA_MINBLOCKSHIFT;
 120
 121         return (0);
 122 }
 123
 124 static void
 125 vdev_file_close(vdev_t *vd)
 126 {
 127         vdev_file_t *vf = vd->vdev_tsd;
 128
 129         if (vd->vdev_reopening || vf == NULL)
 130                 return;
 131
 132         if (vf->vf_vnode != NULL) {
 133                 (void) VOP_PUTPAGE(vf->vf_vnode, 0, 0, B_INVAL, kcred, NULL);
 134                 (void) VOP_CLOSE(vf->vf_vnode, spa_mode(vd->vdev_spa), 1, 0,
 135                     kcred, NULL);
 136         }
 137
 138         vd->vdev_delayed_close = B_FALSE;
 139         kmem_free(vf, sizeof (vdev_file_t));
 140         vd->vdev_tsd = NULL;
 141 }
 142
 143 static void
 144 vdev_file_io_strategy(void *arg)
 145 {
 146         zio_t *zio = (zio_t *)arg;
 147         vdev_t *vd = zio->io_vd;
 148         vdev_file_t *vf = vd->vdev_tsd;
 149         ssize_t resid;
 150
 151         zio->io_error = vn_rdwr(zio->io_type == ZIO_TYPE_READ ?
 152             UIO_READ : UIO_WRITE, vf->vf_vnode, zio->io_data,
 153             zio->io_size, zio->io_offset, UIO_SYSSPACE,
 154             0, RLIM64_INFINITY, kcred, &resid);
 155
 156         if (resid != 0 && zio->io_error == 0)
 157                 zio->io_error = SET_ERROR(ENOSPC);
 158
 159         zio_interrupt(zio);
 160 }
 161
 162 static void
 163 vdev_file_io_fsync(void *arg)
 164 {
 165         zio_t *zio = (zio_t *)arg;
 166         vdev_file_t *vf = zio->io_vd->vdev_tsd;
 167
 168         zio->io_error = VOP_FSYNC(vf->vf_vnode, FSYNC | FDSYNC, kcred, NULL);
 169
 170         zio_interrupt(zio);
 171 }
 172
 173 static void
 174 vdev_file_io_start(zio_t *zio)
 175 {
 176         vdev_t *vd = zio->io_vd;
 177         vdev_file_t *vf = vd->vdev_tsd;
 178
 179         if (zio->io_type == ZIO_TYPE_IOCTL) {
 180                 /* XXPOLICY */
 181                 if (!vdev_readable(vd)) {
 182                         zio->io_error = SET_ERROR(ENXIO);
 183                         zio_interrupt(zio);
 184                         return;
 185                 }
 186
 187                 switch (zio->io_cmd) {
 188                 case DKIOCFLUSHWRITECACHE:
 189
 190                         if (zfs_nocacheflush)
 191                                 break;
 192
 193                         /*
 194                          * We cannot safely call vfs_fsync() when PF_FSTRANS
 195                          * is set in the current context.  Filesystems like
 196                          * XFS include sanity checks to verify it is not
 197                          * already set, see xfs_vm_writepage().  Therefore
 198                          * the sync must be dispatched to a different context.
 199                          */
 200                         if (spl_fstrans_check()) {
 201                                 VERIFY3U(taskq_dispatch(system_taskq,
 202                                     vdev_file_io_fsync, zio, TQ_SLEEP), !=, 0);
 203                                 return;
 204                         }
 205
 206                         zio->io_error = VOP_FSYNC(vf->vf_vnode, FSYNC | FDSYNC,
 207                             kcred, NULL);
 208                         break;
 209                 default:
 210                         zio->io_error = SET_ERROR(ENOTSUP);
 211                 }
 212
 213                 zio_execute(zio);
 214                 return;
 215         }
 216
 217         VERIFY3U(taskq_dispatch(system_taskq, vdev_file_io_strategy, zio,
 218             TQ_SLEEP), !=, 0);
 219 }
 220
 221 /* ARGSUSED */
 222 static void
 223 vdev_file_io_done(zio_t *zio)
 224 {
 225 }
 226
 227 vdev_ops_t vdev_file_ops = {
 228         vdev_file_open,
 229         vdev_file_close,
 230         vdev_default_asize,
 231         vdev_file_io_start,
 232         vdev_file_io_done,
 233         NULL,
 234         vdev_file_hold,
 235         vdev_file_rele,
 236         VDEV_TYPE_FILE,         /* name of this vdev type */
 237         B_TRUE                  /* leaf vdev */
 238 };
 239
 240 /*
 241  * From userland we access disks just like files.
 242  */
 243 #ifndef _KERNEL
 244
 245 vdev_ops_t vdev_disk_ops = {
 246         vdev_file_open,
 247         vdev_file_close,
 248         vdev_default_asize,
 249         vdev_file_io_start,
 250         vdev_file_io_done,
 251         NULL,
 252         vdev_file_hold,
 253         vdev_file_rele,
 254         VDEV_TYPE_DISK,         /* name of this vdev type */
 255         B_TRUE                  /* leaf vdev */
 256 };
 257
 258 #endif