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, 2015 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 #include <sys/abd.h>
  35
  36 /*
  37  * Virtual device vector for files.
  38  */
  39
  40 static void
  41 vdev_file_hold(vdev_t *vd)
  42 {
  43         ASSERT(vd->vdev_path != NULL);
  44 }
  45
  46 static void
  47 vdev_file_rele(vdev_t *vd)
  48 {
  49         ASSERT(vd->vdev_path != NULL);
  50 }
  51
  52 static int
  53 vdev_file_open(vdev_t *vd, uint64_t *psize, uint64_t *max_psize,
  54     uint64_t *ashift)
  55 {
  56         vdev_file_t *vf;
  57         vnode_t *vp;
  58         vattr_t vattr;
  59         int error;
  60
  61         /* Rotational optimizations only make sense on block devices */
  62         vd->vdev_nonrot = B_TRUE;
  63
  64         /*
  65          * We must have a pathname, and it must be absolute.
  66          */
  67         if (vd->vdev_path == NULL || vd->vdev_path[0] != '/') {
  68                 vd->vdev_stat.vs_aux = VDEV_AUX_BAD_LABEL;
  69                 return (SET_ERROR(EINVAL));
  70         }
  71
  72         /*
  73          * Reopen the device if it's not currently open.  Otherwise,
  74          * just update the physical size of the device.
  75          */
  76         if (vd->vdev_tsd != NULL) {
  77                 ASSERT(vd->vdev_reopening);
  78                 vf = vd->vdev_tsd;
  79                 goto skip_open;
  80         }
  81
  82         vf = vd->vdev_tsd = kmem_zalloc(sizeof (vdev_file_t), KM_SLEEP);
  83
  84         /*
  85          * We always open the files from the root of the global zone, even if
  86          * we're in a local zone.  If the user has gotten to this point, the
  87          * administrator has already decided that the pool should be available
  88          * to local zone users, so the underlying devices should be as well.
  89          */
  90         ASSERT(vd->vdev_path != NULL && vd->vdev_path[0] == '/');
  91         error = vn_openat(vd->vdev_path + 1, UIO_SYSSPACE,
  92             spa_mode(vd->vdev_spa) | FOFFMAX, 0, &vp, 0, 0, rootdir, -1);
  93
  94         if (error) {
  95                 vd->vdev_stat.vs_aux = VDEV_AUX_OPEN_FAILED;
  96                 return (error);
  97         }
  98
  99         vf->vf_vnode = vp;
 100
 101 #ifdef _KERNEL
 102         /*
 103          * Make sure it's a regular file.
 104          */
 105         if (vp->v_type != VREG) {
 106                 vd->vdev_stat.vs_aux = VDEV_AUX_OPEN_FAILED;
 107                 return (SET_ERROR(ENODEV));
 108         }
 109 #endif
 110
 111 skip_open:
 112         /*
 113          * Determine the physical size of the file.
 114          */
 115         vattr.va_mask = AT_SIZE;
 116         error = VOP_GETATTR(vf->vf_vnode, &vattr, 0, kcred, NULL);
 117         if (error) {
 118                 vd->vdev_stat.vs_aux = VDEV_AUX_OPEN_FAILED;
 119                 return (error);
 120         }
 121
 122         *max_psize = *psize = vattr.va_size;
 123         *ashift = SPA_MINBLOCKSHIFT;
 124
 125         return (0);
 126 }
 127
 128 static void
 129 vdev_file_close(vdev_t *vd)
 130 {
 131         vdev_file_t *vf = vd->vdev_tsd;
 132
 133         if (vd->vdev_reopening || vf == NULL)
 134                 return;
 135
 136         if (vf->vf_vnode != NULL) {
 137                 (void) VOP_PUTPAGE(vf->vf_vnode, 0, 0, B_INVAL, kcred, NULL);
 138                 (void) VOP_CLOSE(vf->vf_vnode, spa_mode(vd->vdev_spa), 1, 0,
 139                     kcred, NULL);
 140         }
 141
 142         vd->vdev_delayed_close = B_FALSE;
 143         kmem_free(vf, sizeof (vdev_file_t));
 144         vd->vdev_tsd = NULL;
 145 }
 146
 147 static void
 148 vdev_file_io_strategy(void *arg)
 149 {
 150         zio_t *zio = (zio_t *)arg;
 151         vdev_t *vd = zio->io_vd;
 152         vdev_file_t *vf = vd->vdev_tsd;
 153         ssize_t resid;
 154         void *buf;
 155
 156         if (zio->io_type == ZIO_TYPE_READ)
 157                 buf = abd_borrow_buf(zio->io_abd, zio->io_size);
 158         else
 159                 buf = abd_borrow_buf_copy(zio->io_abd, zio->io_size);
 160
 161         zio->io_error = vn_rdwr(zio->io_type == ZIO_TYPE_READ ?
 162             UIO_READ : UIO_WRITE, vf->vf_vnode, buf, zio->io_size,
 163             zio->io_offset, UIO_SYSSPACE, 0, RLIM64_INFINITY, kcred, &resid);
 164
 165         if (zio->io_type == ZIO_TYPE_READ)
 166                 abd_return_buf_copy(zio->io_abd, buf, zio->io_size);
 167         else
 168                 abd_return_buf(zio->io_abd, buf, zio->io_size);
 169
 170         if (resid != 0 && zio->io_error == 0)
 171                 zio->io_error = SET_ERROR(ENOSPC);
 172
 173         zio_delay_interrupt(zio);
 174 }
 175
 176 static void
 177 vdev_file_io_fsync(void *arg)
 178 {
 179         zio_t *zio = (zio_t *)arg;
 180         vdev_file_t *vf = zio->io_vd->vdev_tsd;
 181
 182         zio->io_error = VOP_FSYNC(vf->vf_vnode, FSYNC | FDSYNC, kcred, NULL);
 183
 184         zio_interrupt(zio);
 185 }
 186
 187 static void
 188 vdev_file_io_start(zio_t *zio)
 189 {
 190         vdev_t *vd = zio->io_vd;
 191         vdev_file_t *vf = vd->vdev_tsd;
 192
 193         if (zio->io_type == ZIO_TYPE_IOCTL) {
 194                 /* XXPOLICY */
 195                 if (!vdev_readable(vd)) {
 196                         zio->io_error = SET_ERROR(ENXIO);
 197                         zio_interrupt(zio);
 198                         return;
 199                 }
 200
 201                 switch (zio->io_cmd) {
 202                 case DKIOCFLUSHWRITECACHE:
 203
 204                         if (zfs_nocacheflush)
 205                                 break;
 206
 207                         /*
 208                          * We cannot safely call vfs_fsync() when PF_FSTRANS
 209                          * is set in the current context.  Filesystems like
 210                          * XFS include sanity checks to verify it is not
 211                          * already set, see xfs_vm_writepage().  Therefore
 212                          * the sync must be dispatched to a different context.
 213                          */
 214                         if (spl_fstrans_check()) {
 215                                 VERIFY3U(taskq_dispatch(system_taskq,
 216                                     vdev_file_io_fsync, zio, TQ_SLEEP), !=,
 217                                     TASKQID_INVALID);
 218                                 return;
 219                         }
 220
 221                         zio->io_error = VOP_FSYNC(vf->vf_vnode, FSYNC | FDSYNC,
 222                             kcred, NULL);
 223                         break;
 224                 default:
 225                         zio->io_error = SET_ERROR(ENOTSUP);
 226                 }
 227
 228                 zio_execute(zio);
 229                 return;
 230         }
 231
 232         zio->io_target_timestamp = zio_handle_io_delay(zio);
 233
 234         VERIFY3U(taskq_dispatch(system_taskq, vdev_file_io_strategy, zio,
 235             TQ_SLEEP), !=, TASKQID_INVALID);
 236 }
 237
 238 /* ARGSUSED */
 239 static void
 240 vdev_file_io_done(zio_t *zio)
 241 {
 242 }
 243
 244 vdev_ops_t vdev_file_ops = {
 245         vdev_file_open,
 246         vdev_file_close,
 247         vdev_default_asize,
 248         vdev_file_io_start,
 249         vdev_file_io_done,
 250         NULL,
 251         vdev_file_hold,
 252         vdev_file_rele,
 253         VDEV_TYPE_FILE,         /* name of this vdev type */
 254         B_TRUE                  /* leaf vdev */
 255 };
 256
 257 /*
 258  * From userland we access disks just like files.
 259  */
 260 #ifndef _KERNEL
 261
 262 vdev_ops_t vdev_disk_ops = {
 263         vdev_file_open,
 264         vdev_file_close,
 265         vdev_default_asize,
 266         vdev_file_io_start,
 267         vdev_file_io_done,
 268         NULL,
 269         vdev_file_hold,
 270         vdev_file_rele,
 271         VDEV_TYPE_DISK,         /* name of this vdev type */
 272         B_TRUE                  /* leaf vdev */
 273 };
 274
 275 #endif