]>
Commit | Line | Data |
---|---|---|
34dc7c2f BB |
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 | /* | |
428870ff | 22 | * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved. |
26ef0cc7 | 23 | * Copyright (c) 2011, 2015 by Delphix. All rights reserved. |
34dc7c2f BB |
24 | */ |
25 | ||
34dc7c2f BB |
26 | #include <sys/zfs_context.h> |
27 | #include <sys/spa.h> | |
5853fe79 | 28 | #include <sys/spa_impl.h> |
34dc7c2f BB |
29 | #include <sys/vdev_file.h> |
30 | #include <sys/vdev_impl.h> | |
31 | #include <sys/zio.h> | |
32 | #include <sys/fs/zfs.h> | |
b128c09f | 33 | #include <sys/fm/fs/zfs.h> |
34dc7c2f BB |
34 | |
35 | /* | |
36 | * Virtual device vector for files. | |
37 | */ | |
38 | ||
428870ff BB |
39 | static void |
40 | vdev_file_hold(vdev_t *vd) | |
41 | { | |
42 | ASSERT(vd->vdev_path != NULL); | |
43 | } | |
44 | ||
45 | static void | |
46 | vdev_file_rele(vdev_t *vd) | |
47 | { | |
48 | ASSERT(vd->vdev_path != NULL); | |
49 | } | |
50 | ||
34dc7c2f | 51 | static int |
1bd201e7 CS |
52 | vdev_file_open(vdev_t *vd, uint64_t *psize, uint64_t *max_psize, |
53 | uint64_t *ashift) | |
34dc7c2f BB |
54 | { |
55 | vdev_file_t *vf; | |
56 | vnode_t *vp; | |
b128c09f | 57 | vattr_t vattr; |
34dc7c2f BB |
58 | int error; |
59 | ||
fb40095f RY |
60 | /* Rotational optimizations only make sense on block devices */ |
61 | vd->vdev_nonrot = B_TRUE; | |
62 | ||
34dc7c2f BB |
63 | /* |
64 | * We must have a pathname, and it must be absolute. | |
65 | */ | |
66 | if (vd->vdev_path == NULL || vd->vdev_path[0] != '/') { | |
67 | vd->vdev_stat.vs_aux = VDEV_AUX_BAD_LABEL; | |
2e528b49 | 68 | return (SET_ERROR(EINVAL)); |
34dc7c2f BB |
69 | } |
70 | ||
428870ff BB |
71 | /* |
72 | * Reopen the device if it's not currently open. Otherwise, | |
73 | * just update the physical size of the device. | |
74 | */ | |
75 | if (vd->vdev_tsd != NULL) { | |
76 | ASSERT(vd->vdev_reopening); | |
77 | vf = vd->vdev_tsd; | |
78 | goto skip_open; | |
79 | } | |
80 | ||
79c76d5b | 81 | vf = vd->vdev_tsd = kmem_zalloc(sizeof (vdev_file_t), KM_SLEEP); |
34dc7c2f BB |
82 | |
83 | /* | |
84 | * We always open the files from the root of the global zone, even if | |
85 | * we're in a local zone. If the user has gotten to this point, the | |
86 | * administrator has already decided that the pool should be available | |
87 | * to local zone users, so the underlying devices should be as well. | |
88 | */ | |
89 | ASSERT(vd->vdev_path != NULL && vd->vdev_path[0] == '/'); | |
90 | error = vn_openat(vd->vdev_path + 1, UIO_SYSSPACE, | |
fb5f0bc8 | 91 | spa_mode(vd->vdev_spa) | FOFFMAX, 0, &vp, 0, 0, rootdir, -1); |
34dc7c2f BB |
92 | |
93 | if (error) { | |
94 | vd->vdev_stat.vs_aux = VDEV_AUX_OPEN_FAILED; | |
95 | return (error); | |
96 | } | |
97 | ||
98 | vf->vf_vnode = vp; | |
99 | ||
100 | #ifdef _KERNEL | |
101 | /* | |
102 | * Make sure it's a regular file. | |
103 | */ | |
104 | if (vp->v_type != VREG) { | |
105 | vd->vdev_stat.vs_aux = VDEV_AUX_OPEN_FAILED; | |
2e528b49 | 106 | return (SET_ERROR(ENODEV)); |
34dc7c2f BB |
107 | } |
108 | #endif | |
428870ff BB |
109 | |
110 | skip_open: | |
34dc7c2f BB |
111 | /* |
112 | * Determine the physical size of the file. | |
113 | */ | |
114 | vattr.va_mask = AT_SIZE; | |
115 | error = VOP_GETATTR(vf->vf_vnode, &vattr, 0, kcred, NULL); | |
116 | if (error) { | |
117 | vd->vdev_stat.vs_aux = VDEV_AUX_OPEN_FAILED; | |
118 | return (error); | |
119 | } | |
120 | ||
1bd201e7 | 121 | *max_psize = *psize = vattr.va_size; |
34dc7c2f BB |
122 | *ashift = SPA_MINBLOCKSHIFT; |
123 | ||
124 | return (0); | |
125 | } | |
126 | ||
127 | static void | |
128 | vdev_file_close(vdev_t *vd) | |
129 | { | |
130 | vdev_file_t *vf = vd->vdev_tsd; | |
131 | ||
428870ff | 132 | if (vd->vdev_reopening || vf == NULL) |
34dc7c2f BB |
133 | return; |
134 | ||
135 | if (vf->vf_vnode != NULL) { | |
136 | (void) VOP_PUTPAGE(vf->vf_vnode, 0, 0, B_INVAL, kcred, NULL); | |
fb5f0bc8 BB |
137 | (void) VOP_CLOSE(vf->vf_vnode, spa_mode(vd->vdev_spa), 1, 0, |
138 | kcred, NULL); | |
34dc7c2f BB |
139 | } |
140 | ||
428870ff | 141 | vd->vdev_delayed_close = B_FALSE; |
34dc7c2f BB |
142 | kmem_free(vf, sizeof (vdev_file_t)); |
143 | vd->vdev_tsd = NULL; | |
144 | } | |
145 | ||
5853fe79 GW |
146 | static void |
147 | vdev_file_io_strategy(void *arg) | |
34dc7c2f | 148 | { |
5853fe79 | 149 | zio_t *zio = (zio_t *)arg; |
34dc7c2f | 150 | vdev_t *vd = zio->io_vd; |
5853fe79 GW |
151 | vdev_file_t *vf = vd->vdev_tsd; |
152 | ssize_t resid; | |
34dc7c2f | 153 | |
5853fe79 GW |
154 | zio->io_error = vn_rdwr(zio->io_type == ZIO_TYPE_READ ? |
155 | UIO_READ : UIO_WRITE, vf->vf_vnode, zio->io_data, | |
156 | zio->io_size, zio->io_offset, UIO_SYSSPACE, | |
157 | 0, RLIM64_INFINITY, kcred, &resid); | |
158 | ||
159 | if (resid != 0 && zio->io_error == 0) | |
2e528b49 | 160 | zio->io_error = SET_ERROR(ENOSPC); |
3adfc400 | 161 | |
26ef0cc7 | 162 | zio_delay_interrupt(zio); |
5853fe79 GW |
163 | } |
164 | ||
92119cc2 BB |
165 | static void |
166 | vdev_file_io_fsync(void *arg) | |
167 | { | |
168 | zio_t *zio = (zio_t *)arg; | |
169 | vdev_file_t *vf = zio->io_vd->vdev_tsd; | |
170 | ||
171 | zio->io_error = VOP_FSYNC(vf->vf_vnode, FSYNC | FDSYNC, kcred, NULL); | |
172 | ||
173 | zio_interrupt(zio); | |
174 | } | |
175 | ||
98b25418 | 176 | static void |
5853fe79 GW |
177 | vdev_file_io_start(zio_t *zio) |
178 | { | |
5853fe79 GW |
179 | vdev_t *vd = zio->io_vd; |
180 | vdev_file_t *vf = vd->vdev_tsd; | |
34dc7c2f | 181 | |
3adfc400 | 182 | if (zio->io_type == ZIO_TYPE_IOCTL) { |
5853fe79 GW |
183 | /* XXPOLICY */ |
184 | if (!vdev_readable(vd)) { | |
2e528b49 | 185 | zio->io_error = SET_ERROR(ENXIO); |
98b25418 GW |
186 | zio_interrupt(zio); |
187 | return; | |
5853fe79 GW |
188 | } |
189 | ||
34dc7c2f BB |
190 | switch (zio->io_cmd) { |
191 | case DKIOCFLUSHWRITECACHE: | |
f9a1ac4d H |
192 | |
193 | if (zfs_nocacheflush) | |
194 | break; | |
195 | ||
92119cc2 BB |
196 | /* |
197 | * We cannot safely call vfs_fsync() when PF_FSTRANS | |
198 | * is set in the current context. Filesystems like | |
199 | * XFS include sanity checks to verify it is not | |
200 | * already set, see xfs_vm_writepage(). Therefore | |
201 | * the sync must be dispatched to a different context. | |
202 | */ | |
203 | if (spl_fstrans_check()) { | |
aa9af22c | 204 | VERIFY3U(taskq_dispatch(system_taskq, |
92119cc2 | 205 | vdev_file_io_fsync, zio, TQ_SLEEP), !=, 0); |
98b25418 | 206 | return; |
92119cc2 BB |
207 | } |
208 | ||
34dc7c2f BB |
209 | zio->io_error = VOP_FSYNC(vf->vf_vnode, FSYNC | FDSYNC, |
210 | kcred, NULL); | |
34dc7c2f BB |
211 | break; |
212 | default: | |
2e528b49 | 213 | zio->io_error = SET_ERROR(ENOTSUP); |
34dc7c2f BB |
214 | } |
215 | ||
98b25418 GW |
216 | zio_execute(zio); |
217 | return; | |
34dc7c2f BB |
218 | } |
219 | ||
26ef0cc7 TH |
220 | zio->io_target_timestamp = zio_handle_io_delay(zio); |
221 | ||
aa9af22c | 222 | VERIFY3U(taskq_dispatch(system_taskq, vdev_file_io_strategy, zio, |
79c76d5b | 223 | TQ_SLEEP), !=, 0); |
34dc7c2f BB |
224 | } |
225 | ||
b128c09f BB |
226 | /* ARGSUSED */ |
227 | static void | |
34dc7c2f BB |
228 | vdev_file_io_done(zio_t *zio) |
229 | { | |
34dc7c2f BB |
230 | } |
231 | ||
232 | vdev_ops_t vdev_file_ops = { | |
233 | vdev_file_open, | |
234 | vdev_file_close, | |
34dc7c2f BB |
235 | vdev_default_asize, |
236 | vdev_file_io_start, | |
237 | vdev_file_io_done, | |
238 | NULL, | |
428870ff BB |
239 | vdev_file_hold, |
240 | vdev_file_rele, | |
34dc7c2f BB |
241 | VDEV_TYPE_FILE, /* name of this vdev type */ |
242 | B_TRUE /* leaf vdev */ | |
243 | }; | |
244 | ||
245 | /* | |
246 | * From userland we access disks just like files. | |
247 | */ | |
248 | #ifndef _KERNEL | |
249 | ||
250 | vdev_ops_t vdev_disk_ops = { | |
251 | vdev_file_open, | |
252 | vdev_file_close, | |
34dc7c2f BB |
253 | vdev_default_asize, |
254 | vdev_file_io_start, | |
255 | vdev_file_io_done, | |
256 | NULL, | |
428870ff BB |
257 | vdev_file_hold, |
258 | vdev_file_rele, | |
34dc7c2f BB |
259 | VDEV_TYPE_DISK, /* name of this vdev type */ |
260 | B_TRUE /* leaf vdev */ | |
261 | }; | |
262 | ||
263 | #endif |