]>
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. |
2e528b49 | 23 | * Copyright (c) 2013 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 | ||
bc25c932 CC |
39 | static taskq_t *vdev_file_taskq; |
40 | ||
428870ff BB |
41 | static void |
42 | vdev_file_hold(vdev_t *vd) | |
43 | { | |
44 | ASSERT(vd->vdev_path != NULL); | |
45 | } | |
46 | ||
47 | static void | |
48 | vdev_file_rele(vdev_t *vd) | |
49 | { | |
50 | ASSERT(vd->vdev_path != NULL); | |
51 | } | |
52 | ||
34dc7c2f | 53 | static int |
1bd201e7 CS |
54 | vdev_file_open(vdev_t *vd, uint64_t *psize, uint64_t *max_psize, |
55 | uint64_t *ashift) | |
34dc7c2f BB |
56 | { |
57 | vdev_file_t *vf; | |
58 | vnode_t *vp; | |
b128c09f | 59 | vattr_t vattr; |
34dc7c2f BB |
60 | int error; |
61 | ||
62 | /* | |
63 | * We must have a pathname, and it must be absolute. | |
64 | */ | |
65 | if (vd->vdev_path == NULL || vd->vdev_path[0] != '/') { | |
66 | vd->vdev_stat.vs_aux = VDEV_AUX_BAD_LABEL; | |
2e528b49 | 67 | return (SET_ERROR(EINVAL)); |
34dc7c2f BB |
68 | } |
69 | ||
428870ff BB |
70 | /* |
71 | * Reopen the device if it's not currently open. Otherwise, | |
72 | * just update the physical size of the device. | |
73 | */ | |
74 | if (vd->vdev_tsd != NULL) { | |
75 | ASSERT(vd->vdev_reopening); | |
76 | vf = vd->vdev_tsd; | |
77 | goto skip_open; | |
78 | } | |
79 | ||
b8d06fca | 80 | vf = vd->vdev_tsd = kmem_zalloc(sizeof (vdev_file_t), KM_PUSHPAGE); |
34dc7c2f BB |
81 | |
82 | /* | |
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. | |
87 | */ | |
88 | ASSERT(vd->vdev_path != NULL && vd->vdev_path[0] == '/'); | |
89 | error = vn_openat(vd->vdev_path + 1, UIO_SYSSPACE, | |
fb5f0bc8 | 90 | spa_mode(vd->vdev_spa) | FOFFMAX, 0, &vp, 0, 0, rootdir, -1); |
34dc7c2f BB |
91 | |
92 | if (error) { | |
93 | vd->vdev_stat.vs_aux = VDEV_AUX_OPEN_FAILED; | |
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 | vd->vdev_stat.vs_aux = VDEV_AUX_OPEN_FAILED; | |
2e528b49 | 105 | return (SET_ERROR(ENODEV)); |
34dc7c2f BB |
106 | } |
107 | #endif | |
428870ff BB |
108 | |
109 | skip_open: | |
34dc7c2f BB |
110 | /* |
111 | * Determine the physical size of the file. | |
112 | */ | |
113 | vattr.va_mask = AT_SIZE; | |
114 | error = VOP_GETATTR(vf->vf_vnode, &vattr, 0, kcred, NULL); | |
115 | if (error) { | |
116 | vd->vdev_stat.vs_aux = VDEV_AUX_OPEN_FAILED; | |
117 | return (error); | |
118 | } | |
119 | ||
1bd201e7 | 120 | *max_psize = *psize = vattr.va_size; |
34dc7c2f BB |
121 | *ashift = SPA_MINBLOCKSHIFT; |
122 | ||
123 | return (0); | |
124 | } | |
125 | ||
126 | static void | |
127 | vdev_file_close(vdev_t *vd) | |
128 | { | |
129 | vdev_file_t *vf = vd->vdev_tsd; | |
130 | ||
428870ff | 131 | if (vd->vdev_reopening || vf == NULL) |
34dc7c2f BB |
132 | return; |
133 | ||
134 | if (vf->vf_vnode != NULL) { | |
135 | (void) VOP_PUTPAGE(vf->vf_vnode, 0, 0, B_INVAL, kcred, NULL); | |
fb5f0bc8 BB |
136 | (void) VOP_CLOSE(vf->vf_vnode, spa_mode(vd->vdev_spa), 1, 0, |
137 | kcred, NULL); | |
34dc7c2f BB |
138 | } |
139 | ||
428870ff | 140 | vd->vdev_delayed_close = B_FALSE; |
34dc7c2f BB |
141 | kmem_free(vf, sizeof (vdev_file_t)); |
142 | vd->vdev_tsd = NULL; | |
143 | } | |
144 | ||
5853fe79 GW |
145 | static void |
146 | vdev_file_io_strategy(void *arg) | |
34dc7c2f | 147 | { |
5853fe79 | 148 | zio_t *zio = (zio_t *)arg; |
34dc7c2f | 149 | vdev_t *vd = zio->io_vd; |
5853fe79 GW |
150 | vdev_file_t *vf = vd->vdev_tsd; |
151 | ssize_t resid; | |
34dc7c2f | 152 | |
5853fe79 GW |
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); | |
157 | ||
158 | if (resid != 0 && zio->io_error == 0) | |
2e528b49 | 159 | zio->io_error = SET_ERROR(ENOSPC); |
3adfc400 | 160 | |
5853fe79 GW |
161 | zio_interrupt(zio); |
162 | } | |
163 | ||
92119cc2 BB |
164 | static void |
165 | vdev_file_io_fsync(void *arg) | |
166 | { | |
167 | zio_t *zio = (zio_t *)arg; | |
168 | vdev_file_t *vf = zio->io_vd->vdev_tsd; | |
169 | ||
170 | zio->io_error = VOP_FSYNC(vf->vf_vnode, FSYNC | FDSYNC, kcred, NULL); | |
171 | ||
172 | zio_interrupt(zio); | |
173 | } | |
174 | ||
5853fe79 GW |
175 | static int |
176 | vdev_file_io_start(zio_t *zio) | |
177 | { | |
5853fe79 GW |
178 | vdev_t *vd = zio->io_vd; |
179 | vdev_file_t *vf = vd->vdev_tsd; | |
34dc7c2f | 180 | |
3adfc400 | 181 | if (zio->io_type == ZIO_TYPE_IOCTL) { |
5853fe79 GW |
182 | /* XXPOLICY */ |
183 | if (!vdev_readable(vd)) { | |
2e528b49 | 184 | zio->io_error = SET_ERROR(ENXIO); |
5853fe79 GW |
185 | return (ZIO_PIPELINE_CONTINUE); |
186 | } | |
187 | ||
34dc7c2f BB |
188 | switch (zio->io_cmd) { |
189 | case DKIOCFLUSHWRITECACHE: | |
f9a1ac4d H |
190 | |
191 | if (zfs_nocacheflush) | |
192 | break; | |
193 | ||
92119cc2 BB |
194 | /* |
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. | |
200 | */ | |
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); | |
205 | } | |
206 | ||
34dc7c2f BB |
207 | zio->io_error = VOP_FSYNC(vf->vf_vnode, FSYNC | FDSYNC, |
208 | kcred, NULL); | |
34dc7c2f BB |
209 | break; |
210 | default: | |
2e528b49 | 211 | zio->io_error = SET_ERROR(ENOTSUP); |
34dc7c2f BB |
212 | } |
213 | ||
214 | return (ZIO_PIPELINE_CONTINUE); | |
215 | } | |
216 | ||
bc25c932 | 217 | VERIFY3U(taskq_dispatch(vdev_file_taskq, vdev_file_io_strategy, zio, |
04aa2de8 | 218 | TQ_PUSHPAGE), !=, 0); |
34dc7c2f BB |
219 | |
220 | return (ZIO_PIPELINE_STOP); | |
221 | } | |
222 | ||
b128c09f BB |
223 | /* ARGSUSED */ |
224 | static void | |
34dc7c2f BB |
225 | vdev_file_io_done(zio_t *zio) |
226 | { | |
34dc7c2f BB |
227 | } |
228 | ||
229 | vdev_ops_t vdev_file_ops = { | |
230 | vdev_file_open, | |
231 | vdev_file_close, | |
34dc7c2f BB |
232 | vdev_default_asize, |
233 | vdev_file_io_start, | |
234 | vdev_file_io_done, | |
235 | NULL, | |
428870ff BB |
236 | vdev_file_hold, |
237 | vdev_file_rele, | |
34dc7c2f BB |
238 | VDEV_TYPE_FILE, /* name of this vdev type */ |
239 | B_TRUE /* leaf vdev */ | |
240 | }; | |
241 | ||
bc25c932 CC |
242 | void |
243 | vdev_file_init(void) | |
244 | { | |
245 | vdev_file_taskq = taskq_create("vdev_file_taskq", 100, minclsyspri, | |
246 | max_ncpus, INT_MAX, TASKQ_PREPOPULATE | TASKQ_THREADS_CPU_PCT); | |
247 | ||
248 | VERIFY(vdev_file_taskq); | |
249 | } | |
250 | ||
251 | void | |
252 | vdev_file_fini(void) | |
253 | { | |
254 | taskq_destroy(vdev_file_taskq); | |
255 | } | |
256 | ||
34dc7c2f BB |
257 | /* |
258 | * From userland we access disks just like files. | |
259 | */ | |
260 | #ifndef _KERNEL | |
261 | ||
262 | vdev_ops_t vdev_disk_ops = { | |
263 | vdev_file_open, | |
264 | vdev_file_close, | |
34dc7c2f BB |
265 | vdev_default_asize, |
266 | vdev_file_io_start, | |
267 | vdev_file_io_done, | |
268 | NULL, | |
428870ff BB |
269 | vdev_file_hold, |
270 | vdev_file_rele, | |
34dc7c2f BB |
271 | VDEV_TYPE_DISK, /* name of this vdev type */ |
272 | B_TRUE /* leaf vdev */ | |
273 | }; | |
274 | ||
275 | #endif |