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/vdev_file.h>
29 #include <sys/vdev_impl.h>
30 #include <sys/zio.h>
31 #include <sys/fs/zfs.h>
32 #include <sys/fm/fs/zfs.h>
33
34 /*
35 * Virtual device vector for files.
36 */
37
38 static void
vdev_file_hold(vdev_t * vd)39 vdev_file_hold(vdev_t *vd)
40 {
41 ASSERT(vd->vdev_path != NULL);
42 }
43
44 static void
vdev_file_rele(vdev_t * vd)45 vdev_file_rele(vdev_t *vd)
46 {
47 ASSERT(vd->vdev_path != NULL);
48 }
49
50 static int
vdev_file_open(vdev_t * vd,uint64_t * psize,uint64_t * max_psize,uint64_t * logical_ashift,uint64_t * physical_ashift)51 vdev_file_open(vdev_t *vd, uint64_t *psize, uint64_t *max_psize,
52 uint64_t *logical_ashift, uint64_t *physical_ashift)
53 {
54 vdev_file_t *vf;
55 vnode_t *vp;
56 vattr_t vattr;
57 int error;
58
59 /*
60 * We must have a pathname, and it must be absolute.
61 */
62 if (vd->vdev_path == NULL || vd->vdev_path[0] != '/') {
63 vd->vdev_stat.vs_aux = VDEV_AUX_BAD_LABEL;
64 return (SET_ERROR(EINVAL));
65 }
66
67 /*
68 * Reopen the device if it's not currently open. Otherwise,
69 * just update the physical size of the device.
70 */
71 if (vd->vdev_tsd != NULL) {
72 ASSERT(vd->vdev_reopening);
73 vf = vd->vdev_tsd;
74 vp = vf->vf_vnode;
75 goto skip_open;
76 }
77
78 vf = vd->vdev_tsd = kmem_zalloc(sizeof (vdev_file_t), KM_SLEEP);
79
80 /*
81 * We always open the files from the root of the global zone, even if
82 * we're in a local zone. If the user has gotten to this point, the
83 * administrator has already decided that the pool should be available
84 * to local zone users, so the underlying devices should be as well.
85 */
86 ASSERT(vd->vdev_path != NULL && vd->vdev_path[0] == '/');
87 error = vn_openat(vd->vdev_path + 1, UIO_SYSSPACE,
88 spa_mode(vd->vdev_spa) | FOFFMAX, 0, &vp, 0, 0, rootdir, -1);
89
90 if (error) {
91 vd->vdev_stat.vs_aux = VDEV_AUX_OPEN_FAILED;
92 kmem_free(vd->vdev_tsd, sizeof (vdev_file_t));
93 vd->vdev_tsd = NULL;
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 (void) VOP_CLOSE(vp, spa_mode(vd->vdev_spa), 1, 0, kcred, NULL);
105 kmem_free(vd->vdev_tsd, sizeof (vdev_file_t));
106 vd->vdev_tsd = NULL;
107 vd->vdev_stat.vs_aux = VDEV_AUX_OPEN_FAILED;
108 return (SET_ERROR(ENODEV));
109 }
110 #endif /* _KERNEL */
111
112 skip_open:
113 /*
114 * Determine the physical size of the file.
115 */
116 vattr.va_mask = AT_SIZE;
117 #ifdef __FreeBSD__
118 vn_lock(vp, LK_SHARED | LK_RETRY);
119 error = VOP_GETATTR(vp, &vattr, kcred);
120 VOP_UNLOCK(vp, 0);
121 #endif
122 #ifdef __NetBSD__
123 error = VOP_GETATTR(vp, &vattr, 0, kcred, NULL);
124 #endif
125 if (error) {
126 (void) VOP_CLOSE(vp, spa_mode(vd->vdev_spa), 1, 0, kcred, NULL);
127 vd->vdev_stat.vs_aux = VDEV_AUX_OPEN_FAILED;
128 kmem_free(vd->vdev_tsd, sizeof (vdev_file_t));
129 vd->vdev_tsd = NULL;
130 return (error);
131 }
132
133 vd->vdev_notrim = B_TRUE;
134
135 *max_psize = *psize = vattr.va_size;
136 *logical_ashift = SPA_MINBLOCKSHIFT;
137 *physical_ashift = SPA_MINBLOCKSHIFT;
138
139 return (0);
140 }
141
142 static void
vdev_file_close(vdev_t * vd)143 vdev_file_close(vdev_t *vd)
144 {
145 vdev_file_t *vf = vd->vdev_tsd;
146
147 if (vd->vdev_reopening || vf == NULL)
148 return;
149
150 if (vf->vf_vnode != NULL) {
151 (void) VOP_CLOSE(vf->vf_vnode, spa_mode(vd->vdev_spa), 1, 0,
152 kcred, NULL);
153 }
154
155 vd->vdev_delayed_close = B_FALSE;
156 kmem_free(vf, sizeof (vdev_file_t));
157 vd->vdev_tsd = NULL;
158 }
159
160 static void
vdev_file_io_start(zio_t * zio)161 vdev_file_io_start(zio_t *zio)
162 {
163 vdev_t *vd = zio->io_vd;
164 vdev_file_t *vf;
165 vnode_t *vp;
166 ssize_t resid;
167
168 if (!vdev_readable(vd)) {
169 zio->io_error = SET_ERROR(ENXIO);
170 zio_interrupt(zio);
171 return;
172 }
173
174 vf = vd->vdev_tsd;
175 vp = vf->vf_vnode;
176
177 if (zio->io_type == ZIO_TYPE_IOCTL) {
178 switch (zio->io_cmd) {
179 case DKIOCFLUSHWRITECACHE:
180 zio->io_error = VOP_FSYNC(vp, FSYNC | FDSYNC,
181 kcred, NULL);
182 break;
183 default:
184 zio->io_error = SET_ERROR(ENOTSUP);
185 }
186
187 zio_execute(zio);
188 return;
189 }
190
191 ASSERT(zio->io_type == ZIO_TYPE_READ || zio->io_type == ZIO_TYPE_WRITE);
192 zio->io_target_timestamp = zio_handle_io_delay(zio);
193
194 zio->io_error = vn_rdwr(zio->io_type == ZIO_TYPE_READ ?
195 UIO_READ : UIO_WRITE, vp, zio->io_data, zio->io_size,
196 zio->io_offset, UIO_SYSSPACE, 0, RLIM64_INFINITY, kcred, &resid);
197
198 if (resid != 0 && zio->io_error == 0)
199 zio->io_error = ENOSPC;
200
201 zio_delay_interrupt(zio);
202
203 #ifdef illumos
204 VERIFY3U(taskq_dispatch(system_taskq, vdev_file_io_strategy, bp,
205 TQ_SLEEP), !=, 0);
206 #endif
207 }
208
209 /* ARGSUSED */
210 static void
vdev_file_io_done(zio_t * zio)211 vdev_file_io_done(zio_t *zio)
212 {
213 }
214
215 vdev_ops_t vdev_file_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_FILE, /* name of this vdev type */
225 B_TRUE /* leaf vdev */
226 };
227
228 /*
229 * From userland we access disks just like files.
230 */
231 #ifndef _KERNEL
232
233 vdev_ops_t vdev_disk_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_DISK, /* name of this vdev type */
243 B_TRUE /* leaf vdev */
244 };
245
246 #endif
247