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.
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.
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]
22 * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
23 * Copyright (c) 2011, 2016 by Delphix. All rights reserved.
26 #include <sys/zfs_context.h>
28 #include <sys/vdev_file.h>
29 #include <sys/vdev_impl.h>
31 #include <sys/fs/zfs.h>
32 #include <sys/fm/fs/zfs.h>
36 * Virtual device vector for files.
39 static taskq_t *vdev_file_taskq;
44 vdev_file_taskq = taskq_create("z_vdev_file", MAX(max_ncpus, 16),
45 minclsyspri, max_ncpus, INT_MAX, 0);
51 taskq_destroy(vdev_file_taskq);
55 vdev_file_hold(vdev_t *vd)
57 ASSERT(vd->vdev_path != NULL);
61 vdev_file_rele(vdev_t *vd)
63 ASSERT(vd->vdev_path != NULL);
67 vdev_file_open(vdev_t *vd, uint64_t *psize, uint64_t *max_psize,
68 uint64_t *logical_ashift, uint64_t *physical_ashift)
76 * We must have a pathname, and it must be absolute.
78 if (vd->vdev_path == NULL || vd->vdev_path[0] != '/') {
79 vd->vdev_stat.vs_aux = VDEV_AUX_BAD_LABEL;
80 return (SET_ERROR(EINVAL));
84 * Reopen the device if it's not currently open. Otherwise,
85 * just update the physical size of the device.
87 if (vd->vdev_tsd != NULL) {
88 ASSERT(vd->vdev_reopening);
94 vf = vd->vdev_tsd = kmem_zalloc(sizeof (vdev_file_t), KM_SLEEP);
97 * We always open the files from the root of the global zone, even if
98 * we're in a local zone. If the user has gotten to this point, the
99 * administrator has already decided that the pool should be available
100 * to local zone users, so the underlying devices should be as well.
102 ASSERT(vd->vdev_path != NULL && vd->vdev_path[0] == '/');
103 error = vn_openat(vd->vdev_path + 1, UIO_SYSSPACE,
104 spa_mode(vd->vdev_spa) | FOFFMAX, 0, &vp, 0, 0, rootdir, -1);
107 vd->vdev_stat.vs_aux = VDEV_AUX_OPEN_FAILED;
108 kmem_free(vd->vdev_tsd, sizeof (vdev_file_t));
117 * Make sure it's a regular file.
119 if (vp->v_type != VREG) {
121 (void) VOP_CLOSE(vp, spa_mode(vd->vdev_spa), 1, 0, kcred, NULL);
123 vd->vdev_stat.vs_aux = VDEV_AUX_OPEN_FAILED;
125 kmem_free(vd->vdev_tsd, sizeof (vdev_file_t));
128 return (SET_ERROR(ENODEV));
134 * Determine the physical size of the file.
136 vattr.va_mask = AT_SIZE;
137 vn_lock(vp, LK_SHARED | LK_RETRY);
138 error = VOP_GETATTR(vp, &vattr, kcred);
141 (void) VOP_CLOSE(vp, spa_mode(vd->vdev_spa), 1, 0, kcred, NULL);
142 vd->vdev_stat.vs_aux = VDEV_AUX_OPEN_FAILED;
143 kmem_free(vd->vdev_tsd, sizeof (vdev_file_t));
148 vd->vdev_notrim = B_TRUE;
150 *max_psize = *psize = vattr.va_size;
151 *logical_ashift = SPA_MINBLOCKSHIFT;
152 *physical_ashift = SPA_MINBLOCKSHIFT;
158 vdev_file_close(vdev_t *vd)
160 vdev_file_t *vf = vd->vdev_tsd;
162 if (vd->vdev_reopening || vf == NULL)
165 if (vf->vf_vnode != NULL) {
166 (void) VOP_CLOSE(vf->vf_vnode, spa_mode(vd->vdev_spa), 1, 0,
170 vd->vdev_delayed_close = B_FALSE;
171 kmem_free(vf, sizeof (vdev_file_t));
176 * Implements the interrupt side for file vdev types. This routine will be
177 * called when the I/O completes allowing us to transfer the I/O to the
178 * interrupt taskqs. For consistency, the code structure mimics disk vdev
182 vdev_file_io_intr(zio_t *zio)
184 zio_delay_interrupt(zio);
188 vdev_file_io_strategy(void *arg)
191 vdev_t *vd = zio->io_vd;
200 ASSERT(zio->io_type == ZIO_TYPE_READ || zio->io_type == ZIO_TYPE_WRITE);
201 if (zio->io_type == ZIO_TYPE_READ) {
202 addr = abd_borrow_buf(zio->io_abd, zio->io_size);
204 addr = abd_borrow_buf_copy(zio->io_abd, zio->io_size);
207 zio->io_error = vn_rdwr(zio->io_type == ZIO_TYPE_READ ?
208 UIO_READ : UIO_WRITE, vp, addr, zio->io_size,
209 zio->io_offset, UIO_SYSSPACE, 0, RLIM64_INFINITY, kcred, &resid);
211 if (zio->io_type == ZIO_TYPE_READ) {
212 abd_return_buf_copy(zio->io_abd, addr, zio->io_size);
214 abd_return_buf(zio->io_abd, addr, zio->io_size);
217 if (resid != 0 && zio->io_error == 0)
218 zio->io_error = ENOSPC;
220 vdev_file_io_intr(zio);
224 vdev_file_io_start(zio_t *zio)
226 vdev_t *vd = zio->io_vd;
227 vdev_file_t *vf = vd->vdev_tsd;
229 if (zio->io_type == ZIO_TYPE_IOCTL) {
231 if (!vdev_readable(vd)) {
232 zio->io_error = SET_ERROR(ENXIO);
237 switch (zio->io_cmd) {
238 case DKIOCFLUSHWRITECACHE:
239 zio->io_error = VOP_FSYNC(vf->vf_vnode, FSYNC | FDSYNC,
243 zio->io_error = SET_ERROR(ENOTSUP);
250 ASSERT(zio->io_type == ZIO_TYPE_READ || zio->io_type == ZIO_TYPE_WRITE);
251 zio->io_target_timestamp = zio_handle_io_delay(zio);
253 VERIFY3U(taskq_dispatch(vdev_file_taskq, vdev_file_io_strategy, zio,
259 vdev_file_io_done(zio_t *zio)
263 vdev_ops_t vdev_file_ops = {
275 VDEV_TYPE_FILE, /* name of this vdev type */
276 B_TRUE /* leaf vdev */
280 * From userland we access disks just like files.
284 vdev_ops_t vdev_disk_ops = {
296 VDEV_TYPE_DISK, /* name of this vdev type */
297 B_TRUE /* leaf vdev */