]>
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> |
a6255b7f | 34 | #include <sys/abd.h> |
34dc7c2f BB |
35 | |
36 | /* | |
37 | * Virtual device vector for files. | |
38 | */ | |
39 | ||
da8f51e1 CC |
40 | static taskq_t *vdev_file_taskq; |
41 | ||
428870ff BB |
42 | static void |
43 | vdev_file_hold(vdev_t *vd) | |
44 | { | |
45 | ASSERT(vd->vdev_path != NULL); | |
46 | } | |
47 | ||
48 | static void | |
49 | vdev_file_rele(vdev_t *vd) | |
50 | { | |
51 | ASSERT(vd->vdev_path != NULL); | |
52 | } | |
53 | ||
34dc7c2f | 54 | static int |
1bd201e7 CS |
55 | vdev_file_open(vdev_t *vd, uint64_t *psize, uint64_t *max_psize, |
56 | uint64_t *ashift) | |
34dc7c2f BB |
57 | { |
58 | vdev_file_t *vf; | |
59 | vnode_t *vp; | |
b128c09f | 60 | vattr_t vattr; |
34dc7c2f BB |
61 | int error; |
62 | ||
fb40095f RY |
63 | /* Rotational optimizations only make sense on block devices */ |
64 | vd->vdev_nonrot = B_TRUE; | |
65 | ||
34dc7c2f BB |
66 | /* |
67 | * We must have a pathname, and it must be absolute. | |
68 | */ | |
69 | if (vd->vdev_path == NULL || vd->vdev_path[0] != '/') { | |
70 | vd->vdev_stat.vs_aux = VDEV_AUX_BAD_LABEL; | |
2e528b49 | 71 | return (SET_ERROR(EINVAL)); |
34dc7c2f BB |
72 | } |
73 | ||
428870ff BB |
74 | /* |
75 | * Reopen the device if it's not currently open. Otherwise, | |
76 | * just update the physical size of the device. | |
77 | */ | |
78 | if (vd->vdev_tsd != NULL) { | |
79 | ASSERT(vd->vdev_reopening); | |
80 | vf = vd->vdev_tsd; | |
81 | goto skip_open; | |
82 | } | |
83 | ||
79c76d5b | 84 | vf = vd->vdev_tsd = kmem_zalloc(sizeof (vdev_file_t), KM_SLEEP); |
34dc7c2f BB |
85 | |
86 | /* | |
87 | * We always open the files from the root of the global zone, even if | |
88 | * we're in a local zone. If the user has gotten to this point, the | |
89 | * administrator has already decided that the pool should be available | |
90 | * to local zone users, so the underlying devices should be as well. | |
91 | */ | |
92 | ASSERT(vd->vdev_path != NULL && vd->vdev_path[0] == '/'); | |
93 | error = vn_openat(vd->vdev_path + 1, UIO_SYSSPACE, | |
fb5f0bc8 | 94 | spa_mode(vd->vdev_spa) | FOFFMAX, 0, &vp, 0, 0, rootdir, -1); |
34dc7c2f BB |
95 | |
96 | if (error) { | |
97 | vd->vdev_stat.vs_aux = VDEV_AUX_OPEN_FAILED; | |
98 | return (error); | |
99 | } | |
100 | ||
101 | vf->vf_vnode = vp; | |
102 | ||
103 | #ifdef _KERNEL | |
104 | /* | |
105 | * Make sure it's a regular file. | |
106 | */ | |
107 | if (vp->v_type != VREG) { | |
108 | vd->vdev_stat.vs_aux = VDEV_AUX_OPEN_FAILED; | |
2e528b49 | 109 | return (SET_ERROR(ENODEV)); |
34dc7c2f BB |
110 | } |
111 | #endif | |
428870ff BB |
112 | |
113 | skip_open: | |
34dc7c2f BB |
114 | /* |
115 | * Determine the physical size of the file. | |
116 | */ | |
117 | vattr.va_mask = AT_SIZE; | |
118 | error = VOP_GETATTR(vf->vf_vnode, &vattr, 0, kcred, NULL); | |
119 | if (error) { | |
120 | vd->vdev_stat.vs_aux = VDEV_AUX_OPEN_FAILED; | |
121 | return (error); | |
122 | } | |
123 | ||
1bd201e7 | 124 | *max_psize = *psize = vattr.va_size; |
34dc7c2f BB |
125 | *ashift = SPA_MINBLOCKSHIFT; |
126 | ||
127 | return (0); | |
128 | } | |
129 | ||
130 | static void | |
131 | vdev_file_close(vdev_t *vd) | |
132 | { | |
133 | vdev_file_t *vf = vd->vdev_tsd; | |
134 | ||
428870ff | 135 | if (vd->vdev_reopening || vf == NULL) |
34dc7c2f BB |
136 | return; |
137 | ||
138 | if (vf->vf_vnode != NULL) { | |
139 | (void) VOP_PUTPAGE(vf->vf_vnode, 0, 0, B_INVAL, kcred, NULL); | |
fb5f0bc8 BB |
140 | (void) VOP_CLOSE(vf->vf_vnode, spa_mode(vd->vdev_spa), 1, 0, |
141 | kcred, NULL); | |
34dc7c2f BB |
142 | } |
143 | ||
428870ff | 144 | vd->vdev_delayed_close = B_FALSE; |
34dc7c2f BB |
145 | kmem_free(vf, sizeof (vdev_file_t)); |
146 | vd->vdev_tsd = NULL; | |
147 | } | |
148 | ||
5853fe79 GW |
149 | static void |
150 | vdev_file_io_strategy(void *arg) | |
34dc7c2f | 151 | { |
5853fe79 | 152 | zio_t *zio = (zio_t *)arg; |
34dc7c2f | 153 | vdev_t *vd = zio->io_vd; |
5853fe79 GW |
154 | vdev_file_t *vf = vd->vdev_tsd; |
155 | ssize_t resid; | |
a6255b7f DQ |
156 | void *buf; |
157 | ||
158 | if (zio->io_type == ZIO_TYPE_READ) | |
159 | buf = abd_borrow_buf(zio->io_abd, zio->io_size); | |
160 | else | |
161 | buf = abd_borrow_buf_copy(zio->io_abd, zio->io_size); | |
34dc7c2f | 162 | |
5853fe79 | 163 | zio->io_error = vn_rdwr(zio->io_type == ZIO_TYPE_READ ? |
a6255b7f DQ |
164 | UIO_READ : UIO_WRITE, vf->vf_vnode, buf, zio->io_size, |
165 | zio->io_offset, UIO_SYSSPACE, 0, RLIM64_INFINITY, kcred, &resid); | |
166 | ||
167 | if (zio->io_type == ZIO_TYPE_READ) | |
168 | abd_return_buf_copy(zio->io_abd, buf, zio->io_size); | |
169 | else | |
170 | abd_return_buf(zio->io_abd, buf, zio->io_size); | |
5853fe79 GW |
171 | |
172 | if (resid != 0 && zio->io_error == 0) | |
2e528b49 | 173 | zio->io_error = SET_ERROR(ENOSPC); |
3adfc400 | 174 | |
26ef0cc7 | 175 | zio_delay_interrupt(zio); |
5853fe79 GW |
176 | } |
177 | ||
92119cc2 BB |
178 | static void |
179 | vdev_file_io_fsync(void *arg) | |
180 | { | |
181 | zio_t *zio = (zio_t *)arg; | |
182 | vdev_file_t *vf = zio->io_vd->vdev_tsd; | |
183 | ||
184 | zio->io_error = VOP_FSYNC(vf->vf_vnode, FSYNC | FDSYNC, kcred, NULL); | |
185 | ||
186 | zio_interrupt(zio); | |
187 | } | |
188 | ||
98b25418 | 189 | static void |
5853fe79 GW |
190 | vdev_file_io_start(zio_t *zio) |
191 | { | |
5853fe79 GW |
192 | vdev_t *vd = zio->io_vd; |
193 | vdev_file_t *vf = vd->vdev_tsd; | |
34dc7c2f | 194 | |
3adfc400 | 195 | if (zio->io_type == ZIO_TYPE_IOCTL) { |
5853fe79 GW |
196 | /* XXPOLICY */ |
197 | if (!vdev_readable(vd)) { | |
2e528b49 | 198 | zio->io_error = SET_ERROR(ENXIO); |
98b25418 GW |
199 | zio_interrupt(zio); |
200 | return; | |
5853fe79 GW |
201 | } |
202 | ||
34dc7c2f BB |
203 | switch (zio->io_cmd) { |
204 | case DKIOCFLUSHWRITECACHE: | |
f9a1ac4d H |
205 | |
206 | if (zfs_nocacheflush) | |
207 | break; | |
208 | ||
92119cc2 BB |
209 | /* |
210 | * We cannot safely call vfs_fsync() when PF_FSTRANS | |
211 | * is set in the current context. Filesystems like | |
212 | * XFS include sanity checks to verify it is not | |
213 | * already set, see xfs_vm_writepage(). Therefore | |
214 | * the sync must be dispatched to a different context. | |
215 | */ | |
e624cd19 | 216 | if (__spl_pf_fstrans_check()) { |
da8f51e1 | 217 | VERIFY3U(taskq_dispatch(vdev_file_taskq, |
48d3eb40 BB |
218 | vdev_file_io_fsync, zio, TQ_SLEEP), !=, |
219 | TASKQID_INVALID); | |
98b25418 | 220 | return; |
92119cc2 BB |
221 | } |
222 | ||
34dc7c2f BB |
223 | zio->io_error = VOP_FSYNC(vf->vf_vnode, FSYNC | FDSYNC, |
224 | kcred, NULL); | |
34dc7c2f BB |
225 | break; |
226 | default: | |
2e528b49 | 227 | zio->io_error = SET_ERROR(ENOTSUP); |
34dc7c2f BB |
228 | } |
229 | ||
98b25418 GW |
230 | zio_execute(zio); |
231 | return; | |
34dc7c2f BB |
232 | } |
233 | ||
26ef0cc7 TH |
234 | zio->io_target_timestamp = zio_handle_io_delay(zio); |
235 | ||
da8f51e1 | 236 | VERIFY3U(taskq_dispatch(vdev_file_taskq, vdev_file_io_strategy, zio, |
48d3eb40 | 237 | TQ_SLEEP), !=, TASKQID_INVALID); |
34dc7c2f BB |
238 | } |
239 | ||
b128c09f BB |
240 | /* ARGSUSED */ |
241 | static void | |
34dc7c2f BB |
242 | vdev_file_io_done(zio_t *zio) |
243 | { | |
34dc7c2f BB |
244 | } |
245 | ||
246 | vdev_ops_t vdev_file_ops = { | |
247 | vdev_file_open, | |
248 | vdev_file_close, | |
34dc7c2f BB |
249 | vdev_default_asize, |
250 | vdev_file_io_start, | |
251 | vdev_file_io_done, | |
252 | NULL, | |
3d6da72d | 253 | NULL, |
428870ff BB |
254 | vdev_file_hold, |
255 | vdev_file_rele, | |
34dc7c2f BB |
256 | VDEV_TYPE_FILE, /* name of this vdev type */ |
257 | B_TRUE /* leaf vdev */ | |
258 | }; | |
259 | ||
da8f51e1 CC |
260 | void |
261 | vdev_file_init(void) | |
262 | { | |
263 | vdev_file_taskq = taskq_create("z_vdev_file", MAX(boot_ncpus, 16), | |
264 | minclsyspri, boot_ncpus, INT_MAX, TASKQ_DYNAMIC); | |
265 | ||
266 | VERIFY(vdev_file_taskq); | |
267 | } | |
268 | ||
269 | void | |
270 | vdev_file_fini(void) | |
271 | { | |
272 | taskq_destroy(vdev_file_taskq); | |
273 | } | |
274 | ||
34dc7c2f BB |
275 | /* |
276 | * From userland we access disks just like files. | |
277 | */ | |
278 | #ifndef _KERNEL | |
279 | ||
280 | vdev_ops_t vdev_disk_ops = { | |
281 | vdev_file_open, | |
282 | vdev_file_close, | |
34dc7c2f BB |
283 | vdev_default_asize, |
284 | vdev_file_io_start, | |
285 | vdev_file_io_done, | |
286 | NULL, | |
3d6da72d | 287 | NULL, |
428870ff BB |
288 | vdev_file_hold, |
289 | vdev_file_rele, | |
34dc7c2f BB |
290 | VDEV_TYPE_DISK, /* name of this vdev type */ |
291 | B_TRUE /* leaf vdev */ | |
292 | }; | |
293 | ||
294 | #endif |