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) 2013 by Delphix. All rights reserved.
26 #include <sys/zfs_context.h>
28 #include <sys/spa_impl.h>
29 #include <sys/vdev_file.h>
30 #include <sys/vdev_impl.h>
32 #include <sys/fs/zfs.h>
33 #include <sys/fm/fs/zfs.h>
36 * Virtual device vector for files.
39 static taskq_t *vdev_file_taskq;
42 vdev_file_hold(vdev_t *vd)
44 ASSERT(vd->vdev_path != NULL);
48 vdev_file_rele(vdev_t *vd)
50 ASSERT(vd->vdev_path != NULL);
54 vdev_file_open(vdev_t *vd, uint64_t *psize, uint64_t *max_psize,
63 * We must have a pathname, and it must be absolute.
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));
71 * Reopen the device if it's not currently open. Otherwise,
72 * just update the physical size of the device.
74 if (vd->vdev_tsd != NULL) {
75 ASSERT(vd->vdev_reopening);
80 vf = vd->vdev_tsd = kmem_zalloc(sizeof (vdev_file_t), KM_SLEEP);
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.
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);
93 vd->vdev_stat.vs_aux = VDEV_AUX_OPEN_FAILED;
101 * Make sure it's a regular file.
103 if (vp->v_type != VREG) {
104 vd->vdev_stat.vs_aux = VDEV_AUX_OPEN_FAILED;
105 return (SET_ERROR(ENODEV));
111 * Determine the physical size of the file.
113 vattr.va_mask = AT_SIZE;
114 error = VOP_GETATTR(vf->vf_vnode, &vattr, 0, kcred, NULL);
116 vd->vdev_stat.vs_aux = VDEV_AUX_OPEN_FAILED;
120 *max_psize = *psize = vattr.va_size;
121 *ashift = SPA_MINBLOCKSHIFT;
127 vdev_file_close(vdev_t *vd)
129 vdev_file_t *vf = vd->vdev_tsd;
131 if (vd->vdev_reopening || vf == NULL)
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,
140 vd->vdev_delayed_close = B_FALSE;
141 kmem_free(vf, sizeof (vdev_file_t));
146 vdev_file_io_strategy(void *arg)
148 zio_t *zio = (zio_t *)arg;
149 vdev_t *vd = zio->io_vd;
150 vdev_file_t *vf = vd->vdev_tsd;
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);
158 if (resid != 0 && zio->io_error == 0)
159 zio->io_error = SET_ERROR(ENOSPC);
165 vdev_file_io_fsync(void *arg)
167 zio_t *zio = (zio_t *)arg;
168 vdev_file_t *vf = zio->io_vd->vdev_tsd;
170 zio->io_error = VOP_FSYNC(vf->vf_vnode, FSYNC | FDSYNC, kcred, NULL);
176 vdev_file_io_start(zio_t *zio)
178 vdev_t *vd = zio->io_vd;
179 vdev_file_t *vf = vd->vdev_tsd;
181 if (zio->io_type == ZIO_TYPE_IOCTL) {
183 if (!vdev_readable(vd)) {
184 zio->io_error = SET_ERROR(ENXIO);
185 return (ZIO_PIPELINE_CONTINUE);
188 switch (zio->io_cmd) {
189 case DKIOCFLUSHWRITECACHE:
191 if (zfs_nocacheflush)
195 * We cannot safely call vfs_fsync() when PF_FSTRANS
196 * is set in the current context. Filesystems like
197 * XFS include sanity checks to verify it is not
198 * already set, see xfs_vm_writepage(). Therefore
199 * the sync must be dispatched to a different context.
201 if (spl_fstrans_check()) {
202 VERIFY3U(taskq_dispatch(vdev_file_taskq,
203 vdev_file_io_fsync, zio, TQ_SLEEP), !=, 0);
204 return (ZIO_PIPELINE_STOP);
207 zio->io_error = VOP_FSYNC(vf->vf_vnode, FSYNC | FDSYNC,
211 zio->io_error = SET_ERROR(ENOTSUP);
214 return (ZIO_PIPELINE_CONTINUE);
217 VERIFY3U(taskq_dispatch(vdev_file_taskq, vdev_file_io_strategy, zio,
220 return (ZIO_PIPELINE_STOP);
225 vdev_file_io_done(zio_t *zio)
229 vdev_ops_t vdev_file_ops = {
238 VDEV_TYPE_FILE, /* name of this vdev type */
239 B_TRUE /* leaf vdev */
245 vdev_file_taskq = taskq_create("vdev_file_taskq", 100, minclsyspri,
246 max_ncpus, INT_MAX, TASKQ_PREPOPULATE | TASKQ_THREADS_CPU_PCT);
248 VERIFY(vdev_file_taskq);
254 taskq_destroy(vdev_file_taskq);
258 * From userland we access disks just like files.
262 vdev_ops_t vdev_disk_ops = {
271 VDEV_TYPE_DISK, /* name of this vdev type */
272 B_TRUE /* leaf vdev */