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 */
24
25#include <sys/zfs_context.h>
26#include <sys/spa.h>
27#include <sys/vdev_file.h>
28#include <sys/vdev_impl.h>
29#include <sys/zio.h>
30#include <sys/fs/zfs.h>
31#include <sys/fm/fs/zfs.h>
32
33/*
34 * Virtual device vector for files.
35 */
36
37static void
38vdev_file_hold(vdev_t *vd)
39{
40 ASSERT(vd->vdev_path != NULL);
41}
42
43static void
44vdev_file_rele(vdev_t *vd)
45{
46 ASSERT(vd->vdev_path != NULL);
47}
48
49static int
50vdev_file_open(vdev_t *vd, uint64_t *psize, uint64_t *ashift)
51{
52 vdev_file_t *vf;
53 vnode_t *vp;
54 vattr_t vattr;
55 int error, vfslocked;
56
57 /*
58 * We must have a pathname, and it must be absolute.
59 */
60 if (vd->vdev_path == NULL || vd->vdev_path[0] != '/') {
61 vd->vdev_stat.vs_aux = VDEV_AUX_BAD_LABEL;
62 return (EINVAL);
63 }
64
65 /*
66 * Reopen the device if it's not currently open. Otherwise,
67 * just update the physical size of the device.
68 */
69 if (vd->vdev_tsd != NULL) {
70 ASSERT(vd->vdev_reopening);
71 vf = vd->vdev_tsd;
72 vp = vf->vf_vnode;
73 goto skip_open;
74 }
75
76 vf = vd->vdev_tsd = kmem_zalloc(sizeof (vdev_file_t), KM_SLEEP);
77
78 /*
79 * We always open the files from the root of the global zone, even if
80 * we're in a local zone. If the user has gotten to this point, the
81 * administrator has already decided that the pool should be available
82 * to local zone users, so the underlying devices should be as well.
83 */
84 ASSERT(vd->vdev_path != NULL && vd->vdev_path[0] == '/');
85 error = vn_openat(vd->vdev_path + 1, UIO_SYSSPACE,
86 spa_mode(vd->vdev_spa) | FOFFMAX, 0, &vp, 0, 0, rootdir, -1);
87
88 if (error) {
89 vd->vdev_stat.vs_aux = VDEV_AUX_OPEN_FAILED;
90 kmem_free(vd->vdev_tsd, sizeof (vdev_file_t));
91 vd->vdev_tsd = NULL;
92 return (error);
93 }
94
95 vf->vf_vnode = vp;
96
97#ifdef _KERNEL
98 /*
99 * Make sure it's a regular file.
100 */
101 if (vp->v_type != VREG) {
102 (void) VOP_CLOSE(vp, spa_mode(vd->vdev_spa), 1, 0, kcred, NULL);
103 vd->vdev_stat.vs_aux = VDEV_AUX_OPEN_FAILED;
104 kmem_free(vd->vdev_tsd, sizeof (vdev_file_t));
105 vd->vdev_tsd = NULL;
106 return (ENODEV);
107 }
108#endif
109
110skip_open:
111 /*
112 * Determine the physical size of the file.
113 */
114 vattr.va_mask = AT_SIZE;
115 vfslocked = VFS_LOCK_GIANT(vp->v_mount);
116 vn_lock(vp, LK_SHARED | LK_RETRY);
117 error = VOP_GETATTR(vp, &vattr, kcred);
118 VOP_UNLOCK(vp, 0);
119 VFS_UNLOCK_GIANT(vfslocked);
120 if (error) {
121 (void) VOP_CLOSE(vp, spa_mode(vd->vdev_spa), 1, 0, kcred, NULL);
122 vd->vdev_stat.vs_aux = VDEV_AUX_OPEN_FAILED;
123 kmem_free(vd->vdev_tsd, sizeof (vdev_file_t));
124 vd->vdev_tsd = NULL;
125 return (error);
126 }
127
128 *psize = vattr.va_size;
129 *ashift = SPA_MINBLOCKSHIFT;
130
131 return (0);
132}
133
134static void
135vdev_file_close(vdev_t *vd)
136{
137 vdev_file_t *vf = vd->vdev_tsd;
138
139 if (vd->vdev_reopening || vf == NULL)
140 return;
141
142 if (vf->vf_vnode != NULL) {
143 (void) VOP_CLOSE(vf->vf_vnode, spa_mode(vd->vdev_spa), 1, 0,
144 kcred, NULL);
145 }
146
147 vd->vdev_delayed_close = B_FALSE;
148 kmem_free(vf, sizeof (vdev_file_t));
149 vd->vdev_tsd = NULL;
150}
151
152static int
153vdev_file_io_start(zio_t *zio)
154{
155 vdev_t *vd = zio->io_vd;
156 vdev_file_t *vf = vd->vdev_tsd;
157 vnode_t *vp = vf->vf_vnode;
158 ssize_t resid;
159
160 if (zio->io_type == ZIO_TYPE_IOCTL) {
161 /* XXPOLICY */
162 if (!vdev_readable(vd)) {
163 zio->io_error = ENXIO;
164 return (ZIO_PIPELINE_CONTINUE);
165 }
166
167 switch (zio->io_cmd) {
168 case DKIOCFLUSHWRITECACHE:
169 zio->io_error = VOP_FSYNC(vp, FSYNC | FDSYNC,
170 kcred, NULL);
171 break;
172 default:
173 zio->io_error = ENOTSUP;
174 }
175
176 return (ZIO_PIPELINE_CONTINUE);
177 }
178
179 zio->io_error = vn_rdwr(zio->io_type == ZIO_TYPE_READ ?
180 UIO_READ : UIO_WRITE, vp, zio->io_data, zio->io_size,
181 zio->io_offset, UIO_SYSSPACE, 0, RLIM64_INFINITY, kcred, &resid);
182
183 if (resid != 0 && zio->io_error == 0)
184 zio->io_error = ENOSPC;
185
186 zio_interrupt(zio);
187
188 return (ZIO_PIPELINE_STOP);
189}
190
191/* ARGSUSED */
192static void
193vdev_file_io_done(zio_t *zio)
194{
195}
196
197vdev_ops_t vdev_file_ops = {
198 vdev_file_open,
199 vdev_file_close,
200 vdev_default_asize,
201 vdev_file_io_start,
202 vdev_file_io_done,
203 NULL,
204 vdev_file_hold,
205 vdev_file_rele,
206 VDEV_TYPE_FILE, /* name of this vdev type */
207 B_TRUE /* leaf vdev */
208};
209
210/*
211 * From userland we access disks just like files.
212 */
213#ifndef _KERNEL
214
215vdev_ops_t vdev_disk_ops = {
216 vdev_file_open,
217 vdev_file_close,
218 vdev_default_asize,
219 vdev_file_io_start,
220 vdev_file_io_done,
221 NULL,
222 vdev_file_hold,
223 vdev_file_rele,
224 VDEV_TYPE_DISK, /* name of this vdev type */
225 B_TRUE /* leaf vdev */
226};
227
228#endif