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) 2013 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 taskq_t *vdev_file_taskq;
  40
  41 static void
  42 vdev_file_hold(vdev_t *vd)
  43 {
  44         ASSERT(vd->vdev_path != NULL);
  45 }
  46
  47 static void
  48 vdev_file_rele(vdev_t *vd)
  49 {
  50         ASSERT(vd->vdev_path != NULL);
  51 }
  52
  53 static int
  54 vdev_file_open(vdev_t *vd, uint64_t *psize, uint64_t *max_psize,
  55     uint64_t *ashift)
  56 {
  57         vdev_file_t *vf;
  58         vnode_t *vp;
  59         vattr_t vattr;
  60         int error;
  61
  62         /*
  63          * We must have a pathname, and it must be absolute.
  64          */
  65         if (vd->vdev_path == NULL || vd->vdev_path[0] != '/') {
  66                 vd->vdev_stat.vs_aux = VDEV_AUX_BAD_LABEL;
  67                 return (SET_ERROR(EINVAL));
  68         }
  69
  70         /*
  71          * Reopen the device if it's not currently open.  Otherwise,
  72          * just update the physical size of the device.
  73          */
  74         if (vd->vdev_tsd != NULL) {
  75                 ASSERT(vd->vdev_reopening);
  76                 vf = vd->vdev_tsd;
  77                 goto skip_open;
  78         }
  79
  80         vf = vd->vdev_tsd = kmem_zalloc(sizeof (vdev_file_t), KM_PUSHPAGE);
  81
  82         /*
  83          * We always open the files from the root of the global zone, even if
  84          * we're in a local zone.  If the user has gotten to this point, the
  85          * administrator has already decided that the pool should be available
  86          * to local zone users, so the underlying devices should be as well.
  87          */
  88         ASSERT(vd->vdev_path != NULL && vd->vdev_path[0] == '/');
  89         error = vn_openat(vd->vdev_path + 1, UIO_SYSSPACE,
  90             spa_mode(vd->vdev_spa) | FOFFMAX, 0, &vp, 0, 0, rootdir, -1);
  91
  92         if (error) {
  93                 vd->vdev_stat.vs_aux = VDEV_AUX_OPEN_FAILED;
  94                 return (error);
  95         }
  96
  97         vf->vf_vnode = vp;
  98
  99 #ifdef _KERNEL
 100         /*
 101          * Make sure it's a regular file.
 102          */
 103         if (vp->v_type != VREG) {
 104                 vd->vdev_stat.vs_aux = VDEV_AUX_OPEN_FAILED;
 105                 return (SET_ERROR(ENODEV));
 106         }
 107 #endif
 108
 109 skip_open:
 110         /*
 111          * Determine the physical size of the file.
 112          */
 113         vattr.va_mask = AT_SIZE;
 114         error = VOP_GETATTR(vf->vf_vnode, &vattr, 0, kcred, NULL);
 115         if (error) {
 116                 vd->vdev_stat.vs_aux = VDEV_AUX_OPEN_FAILED;
 117                 return (error);
 118         }
 119
 120         *max_psize = *psize = vattr.va_size;
 121         *ashift = SPA_MINBLOCKSHIFT;
 122
 123         return (0);
 124 }
 125
 126 static void
 127 vdev_file_close(vdev_t *vd)
 128 {
 129         vdev_file_t *vf = vd->vdev_tsd;
 130
 131         if (vd->vdev_reopening || vf == NULL)
 132                 return;
 133
 134         if (vf->vf_vnode != NULL) {
 135                 (void) VOP_PUTPAGE(vf->vf_vnode, 0, 0, B_INVAL, kcred, NULL);
 136                 (void) VOP_CLOSE(vf->vf_vnode, spa_mode(vd->vdev_spa), 1, 0,
 137                     kcred, NULL);
 138         }
 139
 140         vd->vdev_delayed_close = B_FALSE;
 141         kmem_free(vf, sizeof (vdev_file_t));
 142         vd->vdev_tsd = NULL;
 143 }
 144
 145 static void
 146 vdev_file_io_strategy(void *arg)
 147 {
 148         zio_t *zio = (zio_t *)arg;
 149         vdev_t *vd = zio->io_vd;
 150         vdev_file_t *vf = vd->vdev_tsd;
 151         ssize_t resid;
 152
 153         zio->io_error = vn_rdwr(zio->io_type == ZIO_TYPE_READ ?
 154             UIO_READ : UIO_WRITE, vf->vf_vnode, zio->io_data,
 155             zio->io_size, zio->io_offset, UIO_SYSSPACE,
 156             0, RLIM64_INFINITY, kcred, &resid);
 157
 158         if (resid != 0 && zio->io_error == 0)
 159                 zio->io_error = SET_ERROR(ENOSPC);
 160
 161         zio_interrupt(zio);
 162 }
 163
 164 static int
 165 vdev_file_io_start(zio_t *zio)
 166 {
 167         vdev_t *vd = zio->io_vd;
 168         vdev_file_t *vf = vd->vdev_tsd;
 169
 170         if (zio->io_type == ZIO_TYPE_IOCTL) {
 171                 /* XXPOLICY */
 172                 if (!vdev_readable(vd)) {
 173                         zio->io_error = SET_ERROR(ENXIO);
 174                         return (ZIO_PIPELINE_CONTINUE);
 175                 }
 176
 177                 switch (zio->io_cmd) {
 178                 case DKIOCFLUSHWRITECACHE:
 179
 180                         if (zfs_nocacheflush)
 181                                 break;
 182
 183                         zio->io_error = VOP_FSYNC(vf->vf_vnode, FSYNC | FDSYNC,
 184                             kcred, NULL);
 185                         break;
 186                 default:
 187                         zio->io_error = SET_ERROR(ENOTSUP);
 188                 }
 189
 190                 return (ZIO_PIPELINE_CONTINUE);
 191         }
 192
 193         VERIFY3U(taskq_dispatch(vdev_file_taskq, vdev_file_io_strategy, zio,
 194             TQ_PUSHPAGE), !=, 0);
 195
 196         return (ZIO_PIPELINE_STOP);
 197 }
 198
 199 /* ARGSUSED */
 200 static void
 201 vdev_file_io_done(zio_t *zio)
 202 {
 203 }
 204
 205 vdev_ops_t vdev_file_ops = {
 206         vdev_file_open,
 207         vdev_file_close,
 208         vdev_default_asize,
 209         vdev_file_io_start,
 210         vdev_file_io_done,
 211         NULL,
 212         vdev_file_hold,
 213         vdev_file_rele,
 214         VDEV_TYPE_FILE,         /* name of this vdev type */
 215         B_TRUE                  /* leaf vdev */
 216 };
 217
 218 void
 219 vdev_file_init(void)
 220 {
 221         vdev_file_taskq = taskq_create("vdev_file_taskq", 100, minclsyspri,
 222             max_ncpus, INT_MAX, TASKQ_PREPOPULATE | TASKQ_THREADS_CPU_PCT);
 223
 224         VERIFY(vdev_file_taskq);
 225 }
 226
 227 void
 228 vdev_file_fini(void)
 229 {
 230         taskq_destroy(vdev_file_taskq);
 231 }
 232
 233 /*
 234  * From userland we access disks just like files.
 235  */
 236 #ifndef _KERNEL
 237
 238 vdev_ops_t vdev_disk_ops = {
 239         vdev_file_open,
 240         vdev_file_close,
 241         vdev_default_asize,
 242         vdev_file_io_start,
 243         vdev_file_io_done,
 244         NULL,
 245         vdev_file_hold,
 246         vdev_file_rele,
 247         VDEV_TYPE_DISK,         /* name of this vdev type */
 248         B_TRUE                  /* leaf vdev */
 249 };
 250
 251 #endif