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