]>
Commit | Line | Data |
---|---|---|
60101509 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 | /* | |
22 | * Copyright (C) 2008-2010 Lawrence Livermore National Security, LLC. | |
23 | * Produced at Lawrence Livermore National Laboratory (cf, DISCLAIMER). | |
24 | * Rewritten for Linux by Brian Behlendorf <behlendorf1@llnl.gov>. | |
25 | * LLNL-CODE-403049. | |
26 | * | |
27 | * ZFS volume emulation driver. | |
28 | * | |
29 | * Makes a DMU object look like a volume of arbitrary size, up to 2^64 bytes. | |
30 | * Volumes are accessed through the symbolic links named: | |
31 | * | |
32 | * /dev/<pool_name>/<dataset_name> | |
33 | * | |
34 | * Volumes are persistent through reboot and module load. No user command | |
35 | * needs to be run before opening and using a device. | |
36 | */ | |
37 | ||
a08ee875 | 38 | #include <sys/dbuf.h> |
60101509 BB |
39 | #include <sys/dmu_traverse.h> |
40 | #include <sys/dsl_dataset.h> | |
41 | #include <sys/dsl_prop.h> | |
42 | #include <sys/zap.h> | |
e10b0808 | 43 | #include <sys/zfeature.h> |
60101509 BB |
44 | #include <sys/zil_impl.h> |
45 | #include <sys/zio.h> | |
46 | #include <sys/zfs_rlock.h> | |
47 | #include <sys/zfs_znode.h> | |
48 | #include <sys/zvol.h> | |
61e90960 | 49 | #include <linux/blkdev_compat.h> |
60101509 | 50 | |
74497b7a | 51 | unsigned int zvol_inhibit_dev = 0; |
60101509 | 52 | unsigned int zvol_major = ZVOL_MAJOR; |
e10b0808 | 53 | unsigned int zvol_prefetch_bytes = (128 * 1024); |
7c0e5708 | 54 | unsigned long zvol_max_discard_blocks = 16384; |
60101509 | 55 | |
60101509 BB |
56 | static kmutex_t zvol_state_lock; |
57 | static list_t zvol_state_list; | |
58 | static char *zvol_tag = "zvol_tag"; | |
59 | ||
60 | /* | |
61 | * The in-core state of each volume. | |
62 | */ | |
63 | typedef struct zvol_state { | |
4c0d8e50 | 64 | char zv_name[MAXNAMELEN]; /* name */ |
a08ee875 LG |
65 | uint64_t zv_volsize; /* advertised space */ |
66 | uint64_t zv_volblocksize; /* volume block size */ | |
60101509 BB |
67 | objset_t *zv_objset; /* objset handle */ |
68 | uint32_t zv_flags; /* ZVOL_* flags */ | |
69 | uint32_t zv_open_count; /* open counts */ | |
70 | uint32_t zv_changed; /* disk changed */ | |
71 | zilog_t *zv_zilog; /* ZIL handle */ | |
72 | znode_t zv_znode; /* for range locking */ | |
73 | dmu_buf_t *zv_dbuf; /* bonus handle */ | |
74 | dev_t zv_dev; /* device id */ | |
75 | struct gendisk *zv_disk; /* generic disk */ | |
76 | struct request_queue *zv_queue; /* request queue */ | |
77 | spinlock_t zv_lock; /* request queue lock */ | |
78 | list_node_t zv_next; /* next zvol_state_t linkage */ | |
79 | } zvol_state_t; | |
80 | ||
81 | #define ZVOL_RDONLY 0x1 | |
82 | ||
83 | /* | |
84 | * Find the next available range of ZVOL_MINORS minor numbers. The | |
85 | * zvol_state_list is kept in ascending minor order so we simply need | |
86 | * to scan the list for the first gap in the sequence. This allows us | |
87 | * to recycle minor number as devices are created and removed. | |
88 | */ | |
89 | static int | |
90 | zvol_find_minor(unsigned *minor) | |
91 | { | |
92 | zvol_state_t *zv; | |
93 | ||
94 | *minor = 0; | |
95 | ASSERT(MUTEX_HELD(&zvol_state_lock)); | |
96 | for (zv = list_head(&zvol_state_list); zv != NULL; | |
a08ee875 | 97 | zv = list_next(&zvol_state_list, zv), *minor += ZVOL_MINORS) { |
60101509 BB |
98 | if (MINOR(zv->zv_dev) != MINOR(*minor)) |
99 | break; | |
100 | } | |
101 | ||
102 | /* All minors are in use */ | |
103 | if (*minor >= (1 << MINORBITS)) | |
a08ee875 | 104 | return (SET_ERROR(ENXIO)); |
60101509 | 105 | |
a08ee875 | 106 | return (0); |
60101509 BB |
107 | } |
108 | ||
109 | /* | |
110 | * Find a zvol_state_t given the full major+minor dev_t. | |
111 | */ | |
112 | static zvol_state_t * | |
113 | zvol_find_by_dev(dev_t dev) | |
114 | { | |
115 | zvol_state_t *zv; | |
116 | ||
117 | ASSERT(MUTEX_HELD(&zvol_state_lock)); | |
118 | for (zv = list_head(&zvol_state_list); zv != NULL; | |
a08ee875 | 119 | zv = list_next(&zvol_state_list, zv)) { |
60101509 | 120 | if (zv->zv_dev == dev) |
a08ee875 | 121 | return (zv); |
60101509 BB |
122 | } |
123 | ||
a08ee875 | 124 | return (NULL); |
60101509 BB |
125 | } |
126 | ||
127 | /* | |
128 | * Find a zvol_state_t given the name provided at zvol_alloc() time. | |
129 | */ | |
130 | static zvol_state_t * | |
131 | zvol_find_by_name(const char *name) | |
132 | { | |
133 | zvol_state_t *zv; | |
134 | ||
135 | ASSERT(MUTEX_HELD(&zvol_state_lock)); | |
136 | for (zv = list_head(&zvol_state_list); zv != NULL; | |
a08ee875 LG |
137 | zv = list_next(&zvol_state_list, zv)) { |
138 | if (strncmp(zv->zv_name, name, MAXNAMELEN) == 0) | |
139 | return (zv); | |
60101509 BB |
140 | } |
141 | ||
a08ee875 | 142 | return (NULL); |
60101509 BB |
143 | } |
144 | ||
6c285672 JL |
145 | |
146 | /* | |
147 | * Given a path, return TRUE if path is a ZVOL. | |
148 | */ | |
149 | boolean_t | |
150 | zvol_is_zvol(const char *device) | |
151 | { | |
152 | struct block_device *bdev; | |
153 | unsigned int major; | |
154 | ||
155 | bdev = lookup_bdev(device); | |
156 | if (IS_ERR(bdev)) | |
157 | return (B_FALSE); | |
158 | ||
159 | major = MAJOR(bdev->bd_dev); | |
160 | bdput(bdev); | |
161 | ||
162 | if (major == zvol_major) | |
a08ee875 | 163 | return (B_TRUE); |
6c285672 JL |
164 | |
165 | return (B_FALSE); | |
166 | } | |
167 | ||
60101509 BB |
168 | /* |
169 | * ZFS_IOC_CREATE callback handles dmu zvol and zap object creation. | |
170 | */ | |
171 | void | |
172 | zvol_create_cb(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx) | |
173 | { | |
174 | zfs_creat_t *zct = arg; | |
175 | nvlist_t *nvprops = zct->zct_props; | |
176 | int error; | |
177 | uint64_t volblocksize, volsize; | |
178 | ||
179 | VERIFY(nvlist_lookup_uint64(nvprops, | |
180 | zfs_prop_to_name(ZFS_PROP_VOLSIZE), &volsize) == 0); | |
181 | if (nvlist_lookup_uint64(nvprops, | |
182 | zfs_prop_to_name(ZFS_PROP_VOLBLOCKSIZE), &volblocksize) != 0) | |
183 | volblocksize = zfs_prop_default_numeric(ZFS_PROP_VOLBLOCKSIZE); | |
184 | ||
185 | /* | |
186 | * These properties must be removed from the list so the generic | |
187 | * property setting step won't apply to them. | |
188 | */ | |
189 | VERIFY(nvlist_remove_all(nvprops, | |
190 | zfs_prop_to_name(ZFS_PROP_VOLSIZE)) == 0); | |
191 | (void) nvlist_remove_all(nvprops, | |
192 | zfs_prop_to_name(ZFS_PROP_VOLBLOCKSIZE)); | |
193 | ||
194 | error = dmu_object_claim(os, ZVOL_OBJ, DMU_OT_ZVOL, volblocksize, | |
195 | DMU_OT_NONE, 0, tx); | |
196 | ASSERT(error == 0); | |
197 | ||
198 | error = zap_create_claim(os, ZVOL_ZAP_OBJ, DMU_OT_ZVOL_PROP, | |
199 | DMU_OT_NONE, 0, tx); | |
200 | ASSERT(error == 0); | |
201 | ||
202 | error = zap_update(os, ZVOL_ZAP_OBJ, "size", 8, 1, &volsize, tx); | |
203 | ASSERT(error == 0); | |
204 | } | |
205 | ||
206 | /* | |
207 | * ZFS_IOC_OBJSET_STATS entry point. | |
208 | */ | |
209 | int | |
210 | zvol_get_stats(objset_t *os, nvlist_t *nv) | |
211 | { | |
212 | int error; | |
213 | dmu_object_info_t *doi; | |
214 | uint64_t val; | |
215 | ||
216 | error = zap_lookup(os, ZVOL_ZAP_OBJ, "size", 8, 1, &val); | |
217 | if (error) | |
a08ee875 | 218 | return (SET_ERROR(error)); |
60101509 BB |
219 | |
220 | dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_VOLSIZE, val); | |
a08ee875 | 221 | doi = kmem_alloc(sizeof (dmu_object_info_t), KM_SLEEP); |
60101509 BB |
222 | error = dmu_object_info(os, ZVOL_OBJ, doi); |
223 | ||
224 | if (error == 0) { | |
225 | dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_VOLBLOCKSIZE, | |
226 | doi->doi_data_block_size); | |
227 | } | |
228 | ||
a08ee875 | 229 | kmem_free(doi, sizeof (dmu_object_info_t)); |
60101509 | 230 | |
a08ee875 LG |
231 | return (SET_ERROR(error)); |
232 | } | |
233 | ||
234 | static void | |
235 | zvol_size_changed(zvol_state_t *zv, uint64_t volsize) | |
236 | { | |
237 | struct block_device *bdev; | |
238 | ||
239 | bdev = bdget_disk(zv->zv_disk, 0); | |
240 | if (bdev == NULL) | |
241 | return; | |
242 | /* | |
243 | * 2.6.28 API change | |
244 | * Added check_disk_size_change() helper function. | |
245 | */ | |
246 | #ifdef HAVE_CHECK_DISK_SIZE_CHANGE | |
247 | set_capacity(zv->zv_disk, volsize >> 9); | |
248 | zv->zv_volsize = volsize; | |
249 | check_disk_size_change(zv->zv_disk, bdev); | |
250 | #else | |
251 | zv->zv_volsize = volsize; | |
252 | zv->zv_changed = 1; | |
253 | (void) check_disk_change(bdev); | |
254 | #endif /* HAVE_CHECK_DISK_SIZE_CHANGE */ | |
255 | ||
256 | bdput(bdev); | |
60101509 BB |
257 | } |
258 | ||
259 | /* | |
260 | * Sanity check volume size. | |
261 | */ | |
262 | int | |
263 | zvol_check_volsize(uint64_t volsize, uint64_t blocksize) | |
264 | { | |
265 | if (volsize == 0) | |
a08ee875 | 266 | return (SET_ERROR(EINVAL)); |
60101509 BB |
267 | |
268 | if (volsize % blocksize != 0) | |
a08ee875 | 269 | return (SET_ERROR(EINVAL)); |
60101509 BB |
270 | |
271 | #ifdef _ILP32 | |
272 | if (volsize - 1 > MAXOFFSET_T) | |
a08ee875 | 273 | return (SET_ERROR(EOVERFLOW)); |
60101509 BB |
274 | #endif |
275 | return (0); | |
276 | } | |
277 | ||
278 | /* | |
279 | * Ensure the zap is flushed then inform the VFS of the capacity change. | |
280 | */ | |
281 | static int | |
a08ee875 | 282 | zvol_update_volsize(uint64_t volsize, objset_t *os) |
60101509 | 283 | { |
60101509 BB |
284 | dmu_tx_t *tx; |
285 | int error; | |
286 | ||
287 | ASSERT(MUTEX_HELD(&zvol_state_lock)); | |
288 | ||
df554c14 | 289 | tx = dmu_tx_create(os); |
60101509 BB |
290 | dmu_tx_hold_zap(tx, ZVOL_ZAP_OBJ, TRUE, NULL); |
291 | error = dmu_tx_assign(tx, TXG_WAIT); | |
292 | if (error) { | |
293 | dmu_tx_abort(tx); | |
a08ee875 | 294 | return (SET_ERROR(error)); |
60101509 BB |
295 | } |
296 | ||
df554c14 | 297 | error = zap_update(os, ZVOL_ZAP_OBJ, "size", 8, 1, |
60101509 BB |
298 | &volsize, tx); |
299 | dmu_tx_commit(tx); | |
300 | ||
a08ee875 LG |
301 | if (error == 0) |
302 | error = dmu_free_long_range(os, | |
303 | ZVOL_OBJ, volsize, DMU_OBJECT_END); | |
60101509 | 304 | |
a08ee875 LG |
305 | return (error); |
306 | } | |
60101509 | 307 | |
a08ee875 LG |
308 | static int |
309 | zvol_update_live_volsize(zvol_state_t *zv, uint64_t volsize) | |
310 | { | |
311 | zvol_size_changed(zv, volsize); | |
60101509 | 312 | |
a08ee875 LG |
313 | /* |
314 | * We should post a event here describing the expansion. However, | |
315 | * the zfs_ereport_post() interface doesn't nicely support posting | |
316 | * events for zvols, it assumes events relate to vdevs or zios. | |
317 | */ | |
60101509 BB |
318 | |
319 | return (0); | |
320 | } | |
321 | ||
322 | /* | |
323 | * Set ZFS_PROP_VOLSIZE set entry point. | |
324 | */ | |
325 | int | |
326 | zvol_set_volsize(const char *name, uint64_t volsize) | |
327 | { | |
a08ee875 | 328 | zvol_state_t *zv = NULL; |
60101509 | 329 | objset_t *os = NULL; |
60101509 | 330 | int error; |
a08ee875 LG |
331 | dmu_object_info_t *doi; |
332 | uint64_t readonly; | |
333 | boolean_t owned = B_FALSE; | |
60101509 | 334 | |
a08ee875 LG |
335 | error = dsl_prop_get_integer(name, |
336 | zfs_prop_to_name(ZFS_PROP_READONLY), &readonly, NULL); | |
337 | if (error != 0) | |
338 | return (SET_ERROR(error)); | |
339 | if (readonly) | |
340 | return (SET_ERROR(EROFS)); | |
60101509 | 341 | |
a08ee875 | 342 | mutex_enter(&zvol_state_lock); |
60101509 | 343 | zv = zvol_find_by_name(name); |
60101509 | 344 | |
a08ee875 LG |
345 | if (zv == NULL || zv->zv_objset == NULL) { |
346 | if ((error = dmu_objset_own(name, DMU_OST_ZVOL, B_FALSE, | |
347 | FTAG, &os)) != 0) { | |
348 | mutex_exit(&zvol_state_lock); | |
349 | return (SET_ERROR(error)); | |
350 | } | |
351 | owned = B_TRUE; | |
352 | if (zv != NULL) | |
353 | zv->zv_objset = os; | |
354 | } else { | |
355 | os = zv->zv_objset; | |
356 | } | |
60101509 | 357 | |
a08ee875 | 358 | doi = kmem_alloc(sizeof (dmu_object_info_t), KM_SLEEP); |
60101509 | 359 | |
a08ee875 LG |
360 | if ((error = dmu_object_info(os, ZVOL_OBJ, doi)) || |
361 | (error = zvol_check_volsize(volsize, doi->doi_data_block_size))) | |
362 | goto out; | |
60101509 | 363 | |
a08ee875 LG |
364 | error = zvol_update_volsize(volsize, os); |
365 | kmem_free(doi, sizeof (dmu_object_info_t)); | |
60101509 | 366 | |
a08ee875 LG |
367 | if (error == 0 && zv != NULL) |
368 | error = zvol_update_live_volsize(zv, volsize); | |
60101509 | 369 | out: |
a08ee875 LG |
370 | if (owned) { |
371 | dmu_objset_disown(os, FTAG); | |
372 | if (zv != NULL) | |
373 | zv->zv_objset = NULL; | |
374 | } | |
60101509 | 375 | mutex_exit(&zvol_state_lock); |
60101509 BB |
376 | return (error); |
377 | } | |
378 | ||
379 | /* | |
380 | * Sanity check volume block size. | |
381 | */ | |
382 | int | |
e10b0808 | 383 | zvol_check_volblocksize(const char *name, uint64_t volblocksize) |
60101509 | 384 | { |
e10b0808 AX |
385 | /* Record sizes above 128k need the feature to be enabled */ |
386 | if (volblocksize > SPA_OLD_MAXBLOCKSIZE) { | |
387 | spa_t *spa; | |
388 | int error; | |
389 | ||
390 | if ((error = spa_open(name, &spa, FTAG)) != 0) | |
391 | return (error); | |
392 | ||
393 | if (!spa_feature_is_enabled(spa, SPA_FEATURE_LARGE_BLOCKS)) { | |
394 | spa_close(spa, FTAG); | |
395 | return (SET_ERROR(ENOTSUP)); | |
396 | } | |
397 | ||
398 | /* | |
399 | * We don't allow setting the property above 1MB, | |
400 | * unless the tunable has been changed. | |
401 | */ | |
402 | if (volblocksize > zfs_max_recordsize) | |
403 | return (SET_ERROR(EDOM)); | |
404 | ||
405 | spa_close(spa, FTAG); | |
406 | } | |
407 | ||
60101509 BB |
408 | if (volblocksize < SPA_MINBLOCKSIZE || |
409 | volblocksize > SPA_MAXBLOCKSIZE || | |
410 | !ISP2(volblocksize)) | |
a08ee875 | 411 | return (SET_ERROR(EDOM)); |
60101509 BB |
412 | |
413 | return (0); | |
414 | } | |
415 | ||
416 | /* | |
417 | * Set ZFS_PROP_VOLBLOCKSIZE set entry point. | |
418 | */ | |
419 | int | |
420 | zvol_set_volblocksize(const char *name, uint64_t volblocksize) | |
421 | { | |
422 | zvol_state_t *zv; | |
423 | dmu_tx_t *tx; | |
424 | int error; | |
425 | ||
426 | mutex_enter(&zvol_state_lock); | |
427 | ||
428 | zv = zvol_find_by_name(name); | |
429 | if (zv == NULL) { | |
a08ee875 | 430 | error = SET_ERROR(ENXIO); |
60101509 BB |
431 | goto out; |
432 | } | |
433 | ||
a08ee875 LG |
434 | if (zv->zv_flags & ZVOL_RDONLY) { |
435 | error = SET_ERROR(EROFS); | |
60101509 BB |
436 | goto out; |
437 | } | |
438 | ||
439 | tx = dmu_tx_create(zv->zv_objset); | |
440 | dmu_tx_hold_bonus(tx, ZVOL_OBJ); | |
441 | error = dmu_tx_assign(tx, TXG_WAIT); | |
442 | if (error) { | |
443 | dmu_tx_abort(tx); | |
444 | } else { | |
445 | error = dmu_object_set_blocksize(zv->zv_objset, ZVOL_OBJ, | |
446 | volblocksize, 0, tx); | |
447 | if (error == ENOTSUP) | |
a08ee875 | 448 | error = SET_ERROR(EBUSY); |
60101509 BB |
449 | dmu_tx_commit(tx); |
450 | if (error == 0) | |
451 | zv->zv_volblocksize = volblocksize; | |
452 | } | |
453 | out: | |
454 | mutex_exit(&zvol_state_lock); | |
455 | ||
a08ee875 | 456 | return (SET_ERROR(error)); |
60101509 BB |
457 | } |
458 | ||
459 | /* | |
460 | * Replay a TX_WRITE ZIL transaction that didn't get committed | |
461 | * after a system failure | |
462 | */ | |
463 | static int | |
464 | zvol_replay_write(zvol_state_t *zv, lr_write_t *lr, boolean_t byteswap) | |
465 | { | |
466 | objset_t *os = zv->zv_objset; | |
467 | char *data = (char *)(lr + 1); /* data follows lr_write_t */ | |
468 | uint64_t off = lr->lr_offset; | |
469 | uint64_t len = lr->lr_length; | |
470 | dmu_tx_t *tx; | |
471 | int error; | |
472 | ||
473 | if (byteswap) | |
474 | byteswap_uint64_array(lr, sizeof (*lr)); | |
475 | ||
476 | tx = dmu_tx_create(os); | |
477 | dmu_tx_hold_write(tx, ZVOL_OBJ, off, len); | |
478 | error = dmu_tx_assign(tx, TXG_WAIT); | |
479 | if (error) { | |
480 | dmu_tx_abort(tx); | |
481 | } else { | |
482 | dmu_write(os, ZVOL_OBJ, off, len, data, tx); | |
483 | dmu_tx_commit(tx); | |
484 | } | |
485 | ||
a08ee875 | 486 | return (SET_ERROR(error)); |
60101509 BB |
487 | } |
488 | ||
489 | static int | |
490 | zvol_replay_err(zvol_state_t *zv, lr_t *lr, boolean_t byteswap) | |
491 | { | |
a08ee875 | 492 | return (SET_ERROR(ENOTSUP)); |
60101509 BB |
493 | } |
494 | ||
495 | /* | |
496 | * Callback vectors for replaying records. | |
497 | * Only TX_WRITE is needed for zvol. | |
498 | */ | |
b01615d5 RY |
499 | zil_replay_func_t zvol_replay_vector[TX_MAX_TYPE] = { |
500 | (zil_replay_func_t)zvol_replay_err, /* no such transaction type */ | |
501 | (zil_replay_func_t)zvol_replay_err, /* TX_CREATE */ | |
502 | (zil_replay_func_t)zvol_replay_err, /* TX_MKDIR */ | |
503 | (zil_replay_func_t)zvol_replay_err, /* TX_MKXATTR */ | |
504 | (zil_replay_func_t)zvol_replay_err, /* TX_SYMLINK */ | |
505 | (zil_replay_func_t)zvol_replay_err, /* TX_REMOVE */ | |
506 | (zil_replay_func_t)zvol_replay_err, /* TX_RMDIR */ | |
507 | (zil_replay_func_t)zvol_replay_err, /* TX_LINK */ | |
508 | (zil_replay_func_t)zvol_replay_err, /* TX_RENAME */ | |
509 | (zil_replay_func_t)zvol_replay_write, /* TX_WRITE */ | |
510 | (zil_replay_func_t)zvol_replay_err, /* TX_TRUNCATE */ | |
511 | (zil_replay_func_t)zvol_replay_err, /* TX_SETATTR */ | |
512 | (zil_replay_func_t)zvol_replay_err, /* TX_ACL */ | |
60101509 BB |
513 | }; |
514 | ||
515 | /* | |
516 | * zvol_log_write() handles synchronous writes using TX_WRITE ZIL transactions. | |
517 | * | |
518 | * We store data in the log buffers if it's small enough. | |
519 | * Otherwise we will later flush the data out via dmu_sync(). | |
520 | */ | |
521 | ssize_t zvol_immediate_write_sz = 32768; | |
522 | ||
523 | static void | |
a08ee875 LG |
524 | zvol_log_write(zvol_state_t *zv, dmu_tx_t *tx, uint64_t offset, |
525 | uint64_t size, int sync) | |
60101509 BB |
526 | { |
527 | uint32_t blocksize = zv->zv_volblocksize; | |
528 | zilog_t *zilog = zv->zv_zilog; | |
529 | boolean_t slogging; | |
ab85f845 | 530 | ssize_t immediate_write_sz; |
60101509 BB |
531 | |
532 | if (zil_replaying(zilog, tx)) | |
533 | return; | |
534 | ||
ab85f845 ED |
535 | immediate_write_sz = (zilog->zl_logbias == ZFS_LOGBIAS_THROUGHPUT) |
536 | ? 0 : zvol_immediate_write_sz; | |
537 | slogging = spa_has_slogs(zilog->zl_spa) && | |
538 | (zilog->zl_logbias == ZFS_LOGBIAS_LATENCY); | |
60101509 BB |
539 | |
540 | while (size) { | |
541 | itx_t *itx; | |
542 | lr_write_t *lr; | |
543 | ssize_t len; | |
544 | itx_wr_state_t write_state; | |
545 | ||
546 | /* | |
547 | * Unlike zfs_log_write() we can be called with | |
548 | * up to DMU_MAX_ACCESS/2 (5MB) writes. | |
549 | */ | |
ab85f845 | 550 | if (blocksize > immediate_write_sz && !slogging && |
60101509 BB |
551 | size >= blocksize && offset % blocksize == 0) { |
552 | write_state = WR_INDIRECT; /* uses dmu_sync */ | |
553 | len = blocksize; | |
554 | } else if (sync) { | |
555 | write_state = WR_COPIED; | |
556 | len = MIN(ZIL_MAX_LOG_DATA, size); | |
557 | } else { | |
558 | write_state = WR_NEED_COPY; | |
559 | len = MIN(ZIL_MAX_LOG_DATA, size); | |
560 | } | |
561 | ||
562 | itx = zil_itx_create(TX_WRITE, sizeof (*lr) + | |
563 | (write_state == WR_COPIED ? len : 0)); | |
564 | lr = (lr_write_t *)&itx->itx_lr; | |
565 | if (write_state == WR_COPIED && dmu_read(zv->zv_objset, | |
566 | ZVOL_OBJ, offset, len, lr+1, DMU_READ_NO_PREFETCH) != 0) { | |
567 | zil_itx_destroy(itx); | |
568 | itx = zil_itx_create(TX_WRITE, sizeof (*lr)); | |
569 | lr = (lr_write_t *)&itx->itx_lr; | |
570 | write_state = WR_NEED_COPY; | |
571 | } | |
572 | ||
573 | itx->itx_wr_state = write_state; | |
574 | if (write_state == WR_NEED_COPY) | |
575 | itx->itx_sod += len; | |
576 | lr->lr_foid = ZVOL_OBJ; | |
577 | lr->lr_offset = offset; | |
578 | lr->lr_length = len; | |
579 | lr->lr_blkoff = 0; | |
580 | BP_ZERO(&lr->lr_blkptr); | |
581 | ||
582 | itx->itx_private = zv; | |
583 | itx->itx_sync = sync; | |
584 | ||
585 | (void) zil_itx_assign(zilog, itx, tx); | |
586 | ||
587 | offset += len; | |
588 | size -= len; | |
589 | } | |
590 | } | |
591 | ||
e10b0808 AX |
592 | static int |
593 | zvol_write(struct bio *bio) | |
60101509 | 594 | { |
e10b0808 AX |
595 | zvol_state_t *zv = bio->bi_bdev->bd_disk->private_data; |
596 | uint64_t offset = BIO_BI_SECTOR(bio) << 9; | |
597 | uint64_t size = BIO_BI_SIZE(bio); | |
60101509 BB |
598 | int error = 0; |
599 | dmu_tx_t *tx; | |
600 | rl_t *rl; | |
601 | ||
e10b0808 | 602 | if (bio->bi_rw & VDEV_REQ_FLUSH) |
b18019d2 ED |
603 | zil_commit(zv->zv_zilog, ZVOL_OBJ); |
604 | ||
605 | /* | |
606 | * Some requests are just for flush and nothing else. | |
607 | */ | |
e10b0808 | 608 | if (size == 0) |
8630650a | 609 | goto out; |
b18019d2 | 610 | |
60101509 BB |
611 | rl = zfs_range_lock(&zv->zv_znode, offset, size, RL_WRITER); |
612 | ||
613 | tx = dmu_tx_create(zv->zv_objset); | |
614 | dmu_tx_hold_write(tx, ZVOL_OBJ, offset, size); | |
615 | ||
616 | /* This will only fail for ENOSPC */ | |
617 | error = dmu_tx_assign(tx, TXG_WAIT); | |
618 | if (error) { | |
619 | dmu_tx_abort(tx); | |
620 | zfs_range_unlock(rl); | |
8630650a | 621 | goto out; |
60101509 BB |
622 | } |
623 | ||
e10b0808 | 624 | error = dmu_write_bio(zv->zv_objset, ZVOL_OBJ, bio, tx); |
60101509 | 625 | if (error == 0) |
b18019d2 | 626 | zvol_log_write(zv, tx, offset, size, |
e10b0808 | 627 | !!(bio->bi_rw & VDEV_REQ_FUA)); |
60101509 BB |
628 | |
629 | dmu_tx_commit(tx); | |
630 | zfs_range_unlock(rl); | |
631 | ||
e10b0808 | 632 | if ((bio->bi_rw & VDEV_REQ_FUA) || |
b18019d2 | 633 | zv->zv_objset->os_sync == ZFS_SYNC_ALWAYS) |
60101509 BB |
634 | zil_commit(zv->zv_zilog, ZVOL_OBJ); |
635 | ||
8630650a | 636 | out: |
e10b0808 | 637 | return (error); |
60101509 BB |
638 | } |
639 | ||
e10b0808 AX |
640 | static int |
641 | zvol_discard(struct bio *bio) | |
30930fba | 642 | { |
e10b0808 AX |
643 | zvol_state_t *zv = bio->bi_bdev->bd_disk->private_data; |
644 | uint64_t start = BIO_BI_SECTOR(bio) << 9; | |
645 | uint64_t size = BIO_BI_SIZE(bio); | |
646 | uint64_t end = start + size; | |
30930fba ED |
647 | int error; |
648 | rl_t *rl; | |
649 | ||
e10b0808 AX |
650 | if (end > zv->zv_volsize) |
651 | return (SET_ERROR(EIO)); | |
30930fba | 652 | |
089fa91b | 653 | /* |
e10b0808 AX |
654 | * Align the request to volume block boundaries when REQ_SECURE is |
655 | * available, but not requested. If we don't, then this will force | |
656 | * dnode_free_range() to zero out the unaligned parts, which is slow | |
657 | * (read-modify-write) and useless since we are not freeing any space | |
658 | * by doing so. Kernels that do not support REQ_SECURE (2.6.32 through | |
659 | * 2.6.35) will not receive this optimization. | |
089fa91b | 660 | */ |
e10b0808 AX |
661 | #ifdef REQ_SECURE |
662 | if (!(bio->bi_rw & REQ_SECURE)) { | |
663 | start = P2ROUNDUP(start, zv->zv_volblocksize); | |
664 | end = P2ALIGN(end, zv->zv_volblocksize); | |
665 | size = end - start; | |
30930fba | 666 | } |
e10b0808 | 667 | #endif |
30930fba | 668 | |
e10b0808 AX |
669 | if (start >= end) |
670 | return (0); | |
671 | ||
672 | rl = zfs_range_lock(&zv->zv_znode, start, size, RL_WRITER); | |
30930fba | 673 | |
e10b0808 | 674 | error = dmu_free_long_range(zv->zv_objset, ZVOL_OBJ, start, size); |
30930fba ED |
675 | |
676 | /* | |
677 | * TODO: maybe we should add the operation to the log. | |
678 | */ | |
679 | ||
680 | zfs_range_unlock(rl); | |
e10b0808 AX |
681 | |
682 | return (error); | |
30930fba | 683 | } |
30930fba | 684 | |
e10b0808 AX |
685 | static int |
686 | zvol_read(struct bio *bio) | |
60101509 | 687 | { |
e10b0808 AX |
688 | zvol_state_t *zv = bio->bi_bdev->bd_disk->private_data; |
689 | uint64_t offset = BIO_BI_SECTOR(bio) << 9; | |
690 | uint64_t len = BIO_BI_SIZE(bio); | |
60101509 BB |
691 | int error; |
692 | rl_t *rl; | |
693 | ||
e10b0808 AX |
694 | if (len == 0) |
695 | return (0); | |
b18019d2 | 696 | |
60101509 | 697 | |
e10b0808 AX |
698 | rl = zfs_range_lock(&zv->zv_znode, offset, len, RL_READER); |
699 | ||
700 | error = dmu_read_bio(zv->zv_objset, ZVOL_OBJ, bio); | |
60101509 BB |
701 | |
702 | zfs_range_unlock(rl); | |
703 | ||
704 | /* convert checksum errors into IO errors */ | |
705 | if (error == ECKSUM) | |
a08ee875 | 706 | error = SET_ERROR(EIO); |
60101509 | 707 | |
e10b0808 | 708 | return (error); |
60101509 BB |
709 | } |
710 | ||
e10b0808 AX |
711 | static MAKE_REQUEST_FN_RET |
712 | zvol_request(struct request_queue *q, struct bio *bio) | |
60101509 BB |
713 | { |
714 | zvol_state_t *zv = q->queuedata; | |
e10b0808 AX |
715 | fstrans_cookie_t cookie = spl_fstrans_mark(); |
716 | uint64_t offset = BIO_BI_SECTOR(bio); | |
717 | unsigned int sectors = bio_sectors(bio); | |
718 | int rw = bio_data_dir(bio); | |
719 | #ifdef HAVE_GENERIC_IO_ACCT | |
720 | unsigned long start = jiffies; | |
721 | #endif | |
722 | int error = 0; | |
60101509 | 723 | |
e10b0808 AX |
724 | if (bio_has_data(bio) && offset + sectors > |
725 | get_capacity(zv->zv_disk)) { | |
726 | printk(KERN_INFO | |
727 | "%s: bad access: block=%llu, count=%lu\n", | |
728 | zv->zv_disk->disk_name, | |
729 | (long long unsigned)offset, | |
730 | (long unsigned)sectors); | |
731 | error = SET_ERROR(EIO); | |
732 | goto out1; | |
733 | } | |
60101509 | 734 | |
e10b0808 | 735 | generic_start_io_acct(rw, sectors, &zv->zv_disk->part0); |
60101509 | 736 | |
e10b0808 AX |
737 | if (rw == WRITE) { |
738 | if (unlikely(zv->zv_flags & ZVOL_RDONLY)) { | |
739 | error = SET_ERROR(EROFS); | |
740 | goto out2; | |
741 | } | |
30930fba | 742 | |
e10b0808 AX |
743 | if (bio->bi_rw & VDEV_REQ_DISCARD) { |
744 | error = zvol_discard(bio); | |
745 | goto out2; | |
60101509 | 746 | } |
e10b0808 AX |
747 | |
748 | error = zvol_write(bio); | |
749 | } else | |
750 | error = zvol_read(bio); | |
751 | ||
752 | out2: | |
753 | generic_end_io_acct(rw, &zv->zv_disk->part0, start); | |
754 | out1: | |
755 | BIO_END_IO(bio, -error); | |
756 | spl_fstrans_unmark(cookie); | |
757 | #ifdef HAVE_MAKE_REQUEST_FN_RET_INT | |
758 | return (0); | |
94a40997 AX |
759 | #elif defined(HAVE_MAKE_REQUEST_FN_RET_QC) |
760 | return (BLK_QC_T_NONE); | |
e10b0808 | 761 | #endif |
60101509 BB |
762 | } |
763 | ||
764 | static void | |
765 | zvol_get_done(zgd_t *zgd, int error) | |
766 | { | |
767 | if (zgd->zgd_db) | |
768 | dmu_buf_rele(zgd->zgd_db, zgd); | |
769 | ||
770 | zfs_range_unlock(zgd->zgd_rl); | |
771 | ||
772 | if (error == 0 && zgd->zgd_bp) | |
773 | zil_add_block(zgd->zgd_zilog, zgd->zgd_bp); | |
774 | ||
775 | kmem_free(zgd, sizeof (zgd_t)); | |
776 | } | |
777 | ||
778 | /* | |
779 | * Get data to generate a TX_WRITE intent log record. | |
780 | */ | |
781 | static int | |
782 | zvol_get_data(void *arg, lr_write_t *lr, char *buf, zio_t *zio) | |
783 | { | |
784 | zvol_state_t *zv = arg; | |
785 | objset_t *os = zv->zv_objset; | |
a08ee875 | 786 | uint64_t object = ZVOL_OBJ; |
60101509 BB |
787 | uint64_t offset = lr->lr_offset; |
788 | uint64_t size = lr->lr_length; | |
a08ee875 | 789 | blkptr_t *bp = &lr->lr_blkptr; |
60101509 BB |
790 | dmu_buf_t *db; |
791 | zgd_t *zgd; | |
792 | int error; | |
793 | ||
794 | ASSERT(zio != NULL); | |
795 | ASSERT(size != 0); | |
796 | ||
ea04106b | 797 | zgd = (zgd_t *)kmem_zalloc(sizeof (zgd_t), KM_SLEEP); |
60101509 BB |
798 | zgd->zgd_zilog = zv->zv_zilog; |
799 | zgd->zgd_rl = zfs_range_lock(&zv->zv_znode, offset, size, RL_READER); | |
800 | ||
801 | /* | |
802 | * Write records come in two flavors: immediate and indirect. | |
803 | * For small writes it's cheaper to store the data with the | |
804 | * log record (immediate); for large writes it's cheaper to | |
805 | * sync the data and get a pointer to it (indirect) so that | |
806 | * we don't have to write the data twice. | |
807 | */ | |
808 | if (buf != NULL) { /* immediate write */ | |
a08ee875 | 809 | error = dmu_read(os, object, offset, size, buf, |
60101509 BB |
810 | DMU_READ_NO_PREFETCH); |
811 | } else { | |
812 | size = zv->zv_volblocksize; | |
813 | offset = P2ALIGN_TYPED(offset, size, uint64_t); | |
a08ee875 | 814 | error = dmu_buf_hold(os, object, offset, zgd, &db, |
60101509 BB |
815 | DMU_READ_NO_PREFETCH); |
816 | if (error == 0) { | |
a08ee875 LG |
817 | blkptr_t *obp = dmu_buf_get_blkptr(db); |
818 | if (obp) { | |
819 | ASSERT(BP_IS_HOLE(bp)); | |
820 | *bp = *obp; | |
821 | } | |
822 | ||
60101509 BB |
823 | zgd->zgd_db = db; |
824 | zgd->zgd_bp = &lr->lr_blkptr; | |
825 | ||
826 | ASSERT(db != NULL); | |
827 | ASSERT(db->db_offset == offset); | |
828 | ASSERT(db->db_size == size); | |
829 | ||
830 | error = dmu_sync(zio, lr->lr_common.lrc_txg, | |
831 | zvol_get_done, zgd); | |
832 | ||
833 | if (error == 0) | |
834 | return (0); | |
835 | } | |
836 | } | |
837 | ||
838 | zvol_get_done(zgd, error); | |
839 | ||
a08ee875 | 840 | return (SET_ERROR(error)); |
60101509 BB |
841 | } |
842 | ||
843 | /* | |
844 | * The zvol_state_t's are inserted in increasing MINOR(dev_t) order. | |
845 | */ | |
846 | static void | |
847 | zvol_insert(zvol_state_t *zv_insert) | |
848 | { | |
849 | zvol_state_t *zv = NULL; | |
850 | ||
851 | ASSERT(MUTEX_HELD(&zvol_state_lock)); | |
852 | ASSERT3U(MINOR(zv_insert->zv_dev) & ZVOL_MINOR_MASK, ==, 0); | |
853 | for (zv = list_head(&zvol_state_list); zv != NULL; | |
a08ee875 | 854 | zv = list_next(&zvol_state_list, zv)) { |
60101509 BB |
855 | if (MINOR(zv->zv_dev) > MINOR(zv_insert->zv_dev)) |
856 | break; | |
857 | } | |
858 | ||
859 | list_insert_before(&zvol_state_list, zv, zv_insert); | |
860 | } | |
861 | ||
862 | /* | |
863 | * Simply remove the zvol from to list of zvols. | |
864 | */ | |
865 | static void | |
866 | zvol_remove(zvol_state_t *zv_remove) | |
867 | { | |
868 | ASSERT(MUTEX_HELD(&zvol_state_lock)); | |
869 | list_remove(&zvol_state_list, zv_remove); | |
870 | } | |
871 | ||
872 | static int | |
873 | zvol_first_open(zvol_state_t *zv) | |
874 | { | |
875 | objset_t *os; | |
876 | uint64_t volsize; | |
65d56083 | 877 | int locked = 0; |
60101509 BB |
878 | int error; |
879 | uint64_t ro; | |
880 | ||
65d56083 BB |
881 | /* |
882 | * In all other cases the spa_namespace_lock is taken before the | |
883 | * bdev->bd_mutex lock. But in this case the Linux __blkdev_get() | |
884 | * function calls fops->open() with the bdev->bd_mutex lock held. | |
885 | * | |
886 | * To avoid a potential lock inversion deadlock we preemptively | |
887 | * try to take the spa_namespace_lock(). Normally it will not | |
888 | * be contended and this is safe because spa_open_common() handles | |
889 | * the case where the caller already holds the spa_namespace_lock. | |
890 | * | |
891 | * When it is contended we risk a lock inversion if we were to | |
892 | * block waiting for the lock. Luckily, the __blkdev_get() | |
893 | * function allows us to return -ERESTARTSYS which will result in | |
894 | * bdev->bd_mutex being dropped, reacquired, and fops->open() being | |
895 | * called again. This process can be repeated safely until both | |
896 | * locks are acquired. | |
897 | */ | |
898 | if (!mutex_owned(&spa_namespace_lock)) { | |
899 | locked = mutex_tryenter(&spa_namespace_lock); | |
900 | if (!locked) | |
a08ee875 | 901 | return (-SET_ERROR(ERESTARTSYS)); |
65d56083 BB |
902 | } |
903 | ||
ea04106b AX |
904 | error = dsl_prop_get_integer(zv->zv_name, "readonly", &ro, NULL); |
905 | if (error) | |
906 | goto out_mutex; | |
907 | ||
60101509 BB |
908 | /* lie and say we're read-only */ |
909 | error = dmu_objset_own(zv->zv_name, DMU_OST_ZVOL, 1, zvol_tag, &os); | |
910 | if (error) | |
babf3f9b | 911 | goto out_mutex; |
60101509 BB |
912 | |
913 | error = zap_lookup(os, ZVOL_ZAP_OBJ, "size", 8, 1, &volsize); | |
914 | if (error) { | |
babf3f9b MM |
915 | dmu_objset_disown(os, zvol_tag); |
916 | goto out_mutex; | |
60101509 BB |
917 | } |
918 | ||
919 | zv->zv_objset = os; | |
920 | error = dmu_bonus_hold(os, ZVOL_OBJ, zvol_tag, &zv->zv_dbuf); | |
921 | if (error) { | |
babf3f9b MM |
922 | dmu_objset_disown(os, zvol_tag); |
923 | goto out_mutex; | |
60101509 BB |
924 | } |
925 | ||
926 | set_capacity(zv->zv_disk, volsize >> 9); | |
927 | zv->zv_volsize = volsize; | |
928 | zv->zv_zilog = zil_open(os, zvol_get_data); | |
929 | ||
a4430fce GW |
930 | if (ro || dmu_objset_is_snapshot(os) || |
931 | !spa_writeable(dmu_objset_spa(os))) { | |
babf3f9b MM |
932 | set_disk_ro(zv->zv_disk, 1); |
933 | zv->zv_flags |= ZVOL_RDONLY; | |
60101509 | 934 | } else { |
babf3f9b MM |
935 | set_disk_ro(zv->zv_disk, 0); |
936 | zv->zv_flags &= ~ZVOL_RDONLY; | |
60101509 BB |
937 | } |
938 | ||
babf3f9b MM |
939 | out_mutex: |
940 | if (locked) | |
941 | mutex_exit(&spa_namespace_lock); | |
942 | ||
a08ee875 | 943 | return (SET_ERROR(-error)); |
60101509 BB |
944 | } |
945 | ||
946 | static void | |
947 | zvol_last_close(zvol_state_t *zv) | |
948 | { | |
949 | zil_close(zv->zv_zilog); | |
950 | zv->zv_zilog = NULL; | |
04434775 | 951 | |
60101509 BB |
952 | dmu_buf_rele(zv->zv_dbuf, zvol_tag); |
953 | zv->zv_dbuf = NULL; | |
04434775 MA |
954 | |
955 | /* | |
956 | * Evict cached data | |
957 | */ | |
958 | if (dsl_dataset_is_dirty(dmu_objset_ds(zv->zv_objset)) && | |
959 | !(zv->zv_flags & ZVOL_RDONLY)) | |
960 | txg_wait_synced(dmu_objset_pool(zv->zv_objset), 0); | |
961 | (void) dmu_objset_evict_dbufs(zv->zv_objset); | |
962 | ||
60101509 BB |
963 | dmu_objset_disown(zv->zv_objset, zvol_tag); |
964 | zv->zv_objset = NULL; | |
965 | } | |
966 | ||
967 | static int | |
968 | zvol_open(struct block_device *bdev, fmode_t flag) | |
969 | { | |
970 | zvol_state_t *zv = bdev->bd_disk->private_data; | |
971 | int error = 0, drop_mutex = 0; | |
972 | ||
973 | /* | |
974 | * If the caller is already holding the mutex do not take it | |
975 | * again, this will happen as part of zvol_create_minor(). | |
976 | * Once add_disk() is called the device is live and the kernel | |
977 | * will attempt to open it to read the partition information. | |
978 | */ | |
979 | if (!mutex_owned(&zvol_state_lock)) { | |
980 | mutex_enter(&zvol_state_lock); | |
981 | drop_mutex = 1; | |
982 | } | |
983 | ||
984 | ASSERT3P(zv, !=, NULL); | |
985 | ||
986 | if (zv->zv_open_count == 0) { | |
987 | error = zvol_first_open(zv); | |
988 | if (error) | |
989 | goto out_mutex; | |
990 | } | |
991 | ||
a08ee875 | 992 | if ((flag & FMODE_WRITE) && (zv->zv_flags & ZVOL_RDONLY)) { |
60101509 BB |
993 | error = -EROFS; |
994 | goto out_open_count; | |
995 | } | |
996 | ||
997 | zv->zv_open_count++; | |
998 | ||
999 | out_open_count: | |
1000 | if (zv->zv_open_count == 0) | |
1001 | zvol_last_close(zv); | |
1002 | ||
1003 | out_mutex: | |
1004 | if (drop_mutex) | |
1005 | mutex_exit(&zvol_state_lock); | |
1006 | ||
1007 | check_disk_change(bdev); | |
1008 | ||
a08ee875 | 1009 | return (SET_ERROR(error)); |
60101509 BB |
1010 | } |
1011 | ||
c06d4368 AX |
1012 | #ifdef HAVE_BLOCK_DEVICE_OPERATIONS_RELEASE_VOID |
1013 | static void | |
1014 | #else | |
60101509 | 1015 | static int |
c06d4368 | 1016 | #endif |
60101509 BB |
1017 | zvol_release(struct gendisk *disk, fmode_t mode) |
1018 | { | |
1019 | zvol_state_t *zv = disk->private_data; | |
1020 | int drop_mutex = 0; | |
1021 | ||
1022 | if (!mutex_owned(&zvol_state_lock)) { | |
1023 | mutex_enter(&zvol_state_lock); | |
1024 | drop_mutex = 1; | |
1025 | } | |
1026 | ||
ea04106b AX |
1027 | if (zv->zv_open_count > 0) { |
1028 | zv->zv_open_count--; | |
1029 | if (zv->zv_open_count == 0) | |
1030 | zvol_last_close(zv); | |
1031 | } | |
60101509 BB |
1032 | |
1033 | if (drop_mutex) | |
1034 | mutex_exit(&zvol_state_lock); | |
1035 | ||
c06d4368 | 1036 | #ifndef HAVE_BLOCK_DEVICE_OPERATIONS_RELEASE_VOID |
60101509 | 1037 | return (0); |
c06d4368 | 1038 | #endif |
60101509 BB |
1039 | } |
1040 | ||
1041 | static int | |
1042 | zvol_ioctl(struct block_device *bdev, fmode_t mode, | |
a08ee875 | 1043 | unsigned int cmd, unsigned long arg) |
60101509 BB |
1044 | { |
1045 | zvol_state_t *zv = bdev->bd_disk->private_data; | |
1046 | int error = 0; | |
1047 | ||
1048 | if (zv == NULL) | |
a08ee875 | 1049 | return (SET_ERROR(-ENXIO)); |
60101509 BB |
1050 | |
1051 | switch (cmd) { | |
1052 | case BLKFLSBUF: | |
1053 | zil_commit(zv->zv_zilog, ZVOL_OBJ); | |
1054 | break; | |
4c0d8e50 FN |
1055 | case BLKZNAME: |
1056 | error = copy_to_user((void *)arg, zv->zv_name, MAXNAMELEN); | |
1057 | break; | |
60101509 BB |
1058 | |
1059 | default: | |
1060 | error = -ENOTTY; | |
1061 | break; | |
1062 | ||
1063 | } | |
1064 | ||
a08ee875 | 1065 | return (SET_ERROR(error)); |
60101509 BB |
1066 | } |
1067 | ||
1068 | #ifdef CONFIG_COMPAT | |
1069 | static int | |
1070 | zvol_compat_ioctl(struct block_device *bdev, fmode_t mode, | |
a08ee875 | 1071 | unsigned cmd, unsigned long arg) |
60101509 | 1072 | { |
a08ee875 | 1073 | return (zvol_ioctl(bdev, mode, cmd, arg)); |
60101509 BB |
1074 | } |
1075 | #else | |
a08ee875 | 1076 | #define zvol_compat_ioctl NULL |
60101509 BB |
1077 | #endif |
1078 | ||
1079 | static int zvol_media_changed(struct gendisk *disk) | |
1080 | { | |
1081 | zvol_state_t *zv = disk->private_data; | |
1082 | ||
a08ee875 | 1083 | return (zv->zv_changed); |
60101509 BB |
1084 | } |
1085 | ||
1086 | static int zvol_revalidate_disk(struct gendisk *disk) | |
1087 | { | |
1088 | zvol_state_t *zv = disk->private_data; | |
1089 | ||
1090 | zv->zv_changed = 0; | |
1091 | set_capacity(zv->zv_disk, zv->zv_volsize >> 9); | |
1092 | ||
a08ee875 | 1093 | return (0); |
60101509 BB |
1094 | } |
1095 | ||
1096 | /* | |
1097 | * Provide a simple virtual geometry for legacy compatibility. For devices | |
1098 | * smaller than 1 MiB a small head and sector count is used to allow very | |
1099 | * tiny devices. For devices over 1 Mib a standard head and sector count | |
1100 | * is used to keep the cylinders count reasonable. | |
1101 | */ | |
1102 | static int | |
1103 | zvol_getgeo(struct block_device *bdev, struct hd_geometry *geo) | |
1104 | { | |
1105 | zvol_state_t *zv = bdev->bd_disk->private_data; | |
1106 | sector_t sectors = get_capacity(zv->zv_disk); | |
1107 | ||
1108 | if (sectors > 2048) { | |
1109 | geo->heads = 16; | |
1110 | geo->sectors = 63; | |
1111 | } else { | |
1112 | geo->heads = 2; | |
1113 | geo->sectors = 4; | |
1114 | } | |
1115 | ||
1116 | geo->start = 0; | |
1117 | geo->cylinders = sectors / (geo->heads * geo->sectors); | |
1118 | ||
a08ee875 | 1119 | return (0); |
60101509 BB |
1120 | } |
1121 | ||
1122 | static struct kobject * | |
1123 | zvol_probe(dev_t dev, int *part, void *arg) | |
1124 | { | |
1125 | zvol_state_t *zv; | |
1126 | struct kobject *kobj; | |
1127 | ||
1128 | mutex_enter(&zvol_state_lock); | |
1129 | zv = zvol_find_by_dev(dev); | |
23a61ccc | 1130 | kobj = zv ? get_disk(zv->zv_disk) : NULL; |
60101509 BB |
1131 | mutex_exit(&zvol_state_lock); |
1132 | ||
a08ee875 | 1133 | return (kobj); |
60101509 BB |
1134 | } |
1135 | ||
1136 | #ifdef HAVE_BDEV_BLOCK_DEVICE_OPERATIONS | |
1137 | static struct block_device_operations zvol_ops = { | |
a08ee875 LG |
1138 | .open = zvol_open, |
1139 | .release = zvol_release, | |
1140 | .ioctl = zvol_ioctl, | |
1141 | .compat_ioctl = zvol_compat_ioctl, | |
1142 | .media_changed = zvol_media_changed, | |
1143 | .revalidate_disk = zvol_revalidate_disk, | |
1144 | .getgeo = zvol_getgeo, | |
1145 | .owner = THIS_MODULE, | |
60101509 BB |
1146 | }; |
1147 | ||
1148 | #else /* HAVE_BDEV_BLOCK_DEVICE_OPERATIONS */ | |
1149 | ||
1150 | static int | |
1151 | zvol_open_by_inode(struct inode *inode, struct file *file) | |
1152 | { | |
a08ee875 | 1153 | return (zvol_open(inode->i_bdev, file->f_mode)); |
60101509 BB |
1154 | } |
1155 | ||
1156 | static int | |
1157 | zvol_release_by_inode(struct inode *inode, struct file *file) | |
1158 | { | |
a08ee875 | 1159 | return (zvol_release(inode->i_bdev->bd_disk, file->f_mode)); |
60101509 BB |
1160 | } |
1161 | ||
1162 | static int | |
1163 | zvol_ioctl_by_inode(struct inode *inode, struct file *file, | |
a08ee875 | 1164 | unsigned int cmd, unsigned long arg) |
60101509 | 1165 | { |
b1c58213 | 1166 | if (file == NULL || inode == NULL) |
a08ee875 LG |
1167 | return (SET_ERROR(-EINVAL)); |
1168 | ||
1169 | return (zvol_ioctl(inode->i_bdev, file->f_mode, cmd, arg)); | |
60101509 BB |
1170 | } |
1171 | ||
a08ee875 | 1172 | #ifdef CONFIG_COMPAT |
60101509 BB |
1173 | static long |
1174 | zvol_compat_ioctl_by_inode(struct file *file, | |
a08ee875 | 1175 | unsigned int cmd, unsigned long arg) |
60101509 | 1176 | { |
b1c58213 | 1177 | if (file == NULL) |
a08ee875 LG |
1178 | return (SET_ERROR(-EINVAL)); |
1179 | ||
1180 | return (zvol_compat_ioctl(file->f_dentry->d_inode->i_bdev, | |
1181 | file->f_mode, cmd, arg)); | |
60101509 | 1182 | } |
a08ee875 LG |
1183 | #else |
1184 | #define zvol_compat_ioctl_by_inode NULL | |
1185 | #endif | |
60101509 BB |
1186 | |
1187 | static struct block_device_operations zvol_ops = { | |
a08ee875 LG |
1188 | .open = zvol_open_by_inode, |
1189 | .release = zvol_release_by_inode, | |
1190 | .ioctl = zvol_ioctl_by_inode, | |
1191 | .compat_ioctl = zvol_compat_ioctl_by_inode, | |
1192 | .media_changed = zvol_media_changed, | |
1193 | .revalidate_disk = zvol_revalidate_disk, | |
1194 | .getgeo = zvol_getgeo, | |
1195 | .owner = THIS_MODULE, | |
60101509 BB |
1196 | }; |
1197 | #endif /* HAVE_BDEV_BLOCK_DEVICE_OPERATIONS */ | |
1198 | ||
1199 | /* | |
1200 | * Allocate memory for a new zvol_state_t and setup the required | |
1201 | * request queue and generic disk structures for the block device. | |
1202 | */ | |
1203 | static zvol_state_t * | |
1204 | zvol_alloc(dev_t dev, const char *name) | |
1205 | { | |
1206 | zvol_state_t *zv; | |
1207 | ||
ea04106b | 1208 | zv = kmem_zalloc(sizeof (zvol_state_t), KM_SLEEP); |
c06d4368 AX |
1209 | |
1210 | spin_lock_init(&zv->zv_lock); | |
1211 | list_link_init(&zv->zv_next); | |
60101509 | 1212 | |
e10b0808 | 1213 | zv->zv_queue = blk_alloc_queue(GFP_ATOMIC); |
60101509 BB |
1214 | if (zv->zv_queue == NULL) |
1215 | goto out_kmem; | |
1216 | ||
e10b0808 | 1217 | blk_queue_make_request(zv->zv_queue, zvol_request); |
7bd04f2d | 1218 | |
b18019d2 ED |
1219 | #ifdef HAVE_BLK_QUEUE_FLUSH |
1220 | blk_queue_flush(zv->zv_queue, VDEV_REQ_FLUSH | VDEV_REQ_FUA); | |
1221 | #else | |
1222 | blk_queue_ordered(zv->zv_queue, QUEUE_ORDERED_DRAIN, NULL); | |
1223 | #endif /* HAVE_BLK_QUEUE_FLUSH */ | |
1224 | ||
60101509 BB |
1225 | zv->zv_disk = alloc_disk(ZVOL_MINORS); |
1226 | if (zv->zv_disk == NULL) | |
1227 | goto out_queue; | |
1228 | ||
1229 | zv->zv_queue->queuedata = zv; | |
1230 | zv->zv_dev = dev; | |
1231 | zv->zv_open_count = 0; | |
4c0d8e50 | 1232 | strlcpy(zv->zv_name, name, MAXNAMELEN); |
60101509 BB |
1233 | |
1234 | mutex_init(&zv->zv_znode.z_range_lock, NULL, MUTEX_DEFAULT, NULL); | |
1235 | avl_create(&zv->zv_znode.z_range_avl, zfs_range_compare, | |
1236 | sizeof (rl_t), offsetof(rl_t, r_node)); | |
3c4988c8 BB |
1237 | zv->zv_znode.z_is_zvol = TRUE; |
1238 | ||
60101509 BB |
1239 | zv->zv_disk->major = zvol_major; |
1240 | zv->zv_disk->first_minor = (dev & MINORMASK); | |
1241 | zv->zv_disk->fops = &zvol_ops; | |
1242 | zv->zv_disk->private_data = zv; | |
1243 | zv->zv_disk->queue = zv->zv_queue; | |
4c0d8e50 FN |
1244 | snprintf(zv->zv_disk->disk_name, DISK_NAME_LEN, "%s%d", |
1245 | ZVOL_DEV_NAME, (dev & MINORMASK)); | |
60101509 | 1246 | |
a08ee875 | 1247 | return (zv); |
60101509 BB |
1248 | |
1249 | out_queue: | |
1250 | blk_cleanup_queue(zv->zv_queue); | |
1251 | out_kmem: | |
1252 | kmem_free(zv, sizeof (zvol_state_t)); | |
c06d4368 | 1253 | |
a08ee875 | 1254 | return (NULL); |
60101509 BB |
1255 | } |
1256 | ||
1257 | /* | |
1258 | * Cleanup then free a zvol_state_t which was created by zvol_alloc(). | |
1259 | */ | |
1260 | static void | |
1261 | zvol_free(zvol_state_t *zv) | |
1262 | { | |
1263 | avl_destroy(&zv->zv_znode.z_range_avl); | |
1264 | mutex_destroy(&zv->zv_znode.z_range_lock); | |
1265 | ||
1266 | del_gendisk(zv->zv_disk); | |
1267 | blk_cleanup_queue(zv->zv_queue); | |
1268 | put_disk(zv->zv_disk); | |
1269 | ||
1270 | kmem_free(zv, sizeof (zvol_state_t)); | |
1271 | } | |
1272 | ||
1273 | static int | |
0b4d1b58 ED |
1274 | __zvol_snapdev_hidden(const char *name) |
1275 | { | |
a08ee875 LG |
1276 | uint64_t snapdev; |
1277 | char *parent; | |
1278 | char *atp; | |
1279 | int error = 0; | |
1280 | ||
ea04106b | 1281 | parent = kmem_alloc(MAXPATHLEN, KM_SLEEP); |
a08ee875 LG |
1282 | (void) strlcpy(parent, name, MAXPATHLEN); |
1283 | ||
1284 | if ((atp = strrchr(parent, '@')) != NULL) { | |
1285 | *atp = '\0'; | |
1286 | error = dsl_prop_get_integer(parent, "snapdev", &snapdev, NULL); | |
1287 | if ((error == 0) && (snapdev == ZFS_SNAPDEV_HIDDEN)) | |
1288 | error = SET_ERROR(ENODEV); | |
1289 | } | |
1290 | ||
1291 | kmem_free(parent, MAXPATHLEN); | |
1292 | ||
1293 | return (SET_ERROR(error)); | |
0b4d1b58 ED |
1294 | } |
1295 | ||
1296 | static int | |
1297 | __zvol_create_minor(const char *name, boolean_t ignore_snapdev) | |
60101509 BB |
1298 | { |
1299 | zvol_state_t *zv; | |
1300 | objset_t *os; | |
1301 | dmu_object_info_t *doi; | |
1302 | uint64_t volsize; | |
e10b0808 | 1303 | uint64_t len; |
60101509 BB |
1304 | unsigned minor = 0; |
1305 | int error = 0; | |
1306 | ||
1307 | ASSERT(MUTEX_HELD(&zvol_state_lock)); | |
1308 | ||
1309 | zv = zvol_find_by_name(name); | |
1310 | if (zv) { | |
a08ee875 | 1311 | error = SET_ERROR(EEXIST); |
60101509 BB |
1312 | goto out; |
1313 | } | |
1314 | ||
0b4d1b58 ED |
1315 | if (ignore_snapdev == B_FALSE) { |
1316 | error = __zvol_snapdev_hidden(name); | |
1317 | if (error) | |
1318 | goto out; | |
1319 | } | |
1320 | ||
ea04106b | 1321 | doi = kmem_alloc(sizeof (dmu_object_info_t), KM_SLEEP); |
60101509 BB |
1322 | |
1323 | error = dmu_objset_own(name, DMU_OST_ZVOL, B_TRUE, zvol_tag, &os); | |
1324 | if (error) | |
1325 | goto out_doi; | |
1326 | ||
1327 | error = dmu_object_info(os, ZVOL_OBJ, doi); | |
1328 | if (error) | |
1329 | goto out_dmu_objset_disown; | |
1330 | ||
1331 | error = zap_lookup(os, ZVOL_ZAP_OBJ, "size", 8, 1, &volsize); | |
1332 | if (error) | |
1333 | goto out_dmu_objset_disown; | |
1334 | ||
1335 | error = zvol_find_minor(&minor); | |
1336 | if (error) | |
1337 | goto out_dmu_objset_disown; | |
1338 | ||
1339 | zv = zvol_alloc(MKDEV(zvol_major, minor), name); | |
1340 | if (zv == NULL) { | |
a08ee875 | 1341 | error = SET_ERROR(EAGAIN); |
60101509 BB |
1342 | goto out_dmu_objset_disown; |
1343 | } | |
1344 | ||
1345 | if (dmu_objset_is_snapshot(os)) | |
1346 | zv->zv_flags |= ZVOL_RDONLY; | |
1347 | ||
1348 | zv->zv_volblocksize = doi->doi_data_block_size; | |
1349 | zv->zv_volsize = volsize; | |
1350 | zv->zv_objset = os; | |
1351 | ||
1352 | set_capacity(zv->zv_disk, zv->zv_volsize >> 9); | |
1353 | ||
e10b0808 | 1354 | blk_queue_max_hw_sectors(zv->zv_queue, (DMU_MAX_ACCESS / 4) >> 9); |
34037afe ED |
1355 | blk_queue_max_segments(zv->zv_queue, UINT16_MAX); |
1356 | blk_queue_max_segment_size(zv->zv_queue, UINT_MAX); | |
1357 | blk_queue_physical_block_size(zv->zv_queue, zv->zv_volblocksize); | |
1358 | blk_queue_io_opt(zv->zv_queue, zv->zv_volblocksize); | |
7c0e5708 ED |
1359 | blk_queue_max_discard_sectors(zv->zv_queue, |
1360 | (zvol_max_discard_blocks * zv->zv_volblocksize) >> 9); | |
ee5fd0bb | 1361 | blk_queue_discard_granularity(zv->zv_queue, zv->zv_volblocksize); |
30930fba | 1362 | queue_flag_set_unlocked(QUEUE_FLAG_DISCARD, zv->zv_queue); |
e10b0808 | 1363 | #ifdef QUEUE_FLAG_NONROT |
34037afe ED |
1364 | queue_flag_set_unlocked(QUEUE_FLAG_NONROT, zv->zv_queue); |
1365 | #endif | |
e10b0808 AX |
1366 | #ifdef QUEUE_FLAG_ADD_RANDOM |
1367 | queue_flag_clear_unlocked(QUEUE_FLAG_ADD_RANDOM, zv->zv_queue); | |
1368 | #endif | |
34037afe | 1369 | |
a4430fce GW |
1370 | if (spa_writeable(dmu_objset_spa(os))) { |
1371 | if (zil_replay_disable) | |
1372 | zil_destroy(dmu_objset_zil(os), B_FALSE); | |
1373 | else | |
1374 | zil_replay(os, zv, zvol_replay_vector); | |
1375 | } | |
60101509 | 1376 | |
e10b0808 AX |
1377 | /* |
1378 | * When udev detects the addition of the device it will immediately | |
1379 | * invoke blkid(8) to determine the type of content on the device. | |
1380 | * Prefetching the blocks commonly scanned by blkid(8) will speed | |
1381 | * up this process. | |
1382 | */ | |
1383 | len = MIN(MAX(zvol_prefetch_bytes, 0), SPA_MAXBLOCKSIZE); | |
1384 | if (len > 0) { | |
1385 | dmu_prefetch(os, ZVOL_OBJ, 0, len); | |
1386 | dmu_prefetch(os, ZVOL_OBJ, volsize - len, len); | |
1387 | } | |
1388 | ||
f74a147c | 1389 | zv->zv_objset = NULL; |
60101509 BB |
1390 | out_dmu_objset_disown: |
1391 | dmu_objset_disown(os, zvol_tag); | |
60101509 | 1392 | out_doi: |
a08ee875 | 1393 | kmem_free(doi, sizeof (dmu_object_info_t)); |
60101509 BB |
1394 | out: |
1395 | ||
1396 | if (error == 0) { | |
1397 | zvol_insert(zv); | |
1398 | add_disk(zv->zv_disk); | |
1399 | } | |
1400 | ||
a08ee875 | 1401 | return (SET_ERROR(error)); |
60101509 BB |
1402 | } |
1403 | ||
1404 | /* | |
1405 | * Create a block device minor node and setup the linkage between it | |
1406 | * and the specified volume. Once this function returns the block | |
1407 | * device is live and ready for use. | |
1408 | */ | |
1409 | int | |
1410 | zvol_create_minor(const char *name) | |
1411 | { | |
1412 | int error; | |
1413 | ||
1414 | mutex_enter(&zvol_state_lock); | |
0b4d1b58 | 1415 | error = __zvol_create_minor(name, B_FALSE); |
60101509 BB |
1416 | mutex_exit(&zvol_state_lock); |
1417 | ||
a08ee875 | 1418 | return (SET_ERROR(error)); |
60101509 BB |
1419 | } |
1420 | ||
1421 | static int | |
1422 | __zvol_remove_minor(const char *name) | |
1423 | { | |
1424 | zvol_state_t *zv; | |
1425 | ||
1426 | ASSERT(MUTEX_HELD(&zvol_state_lock)); | |
1427 | ||
1428 | zv = zvol_find_by_name(name); | |
1429 | if (zv == NULL) | |
a08ee875 | 1430 | return (SET_ERROR(ENXIO)); |
60101509 BB |
1431 | |
1432 | if (zv->zv_open_count > 0) | |
a08ee875 | 1433 | return (SET_ERROR(EBUSY)); |
60101509 BB |
1434 | |
1435 | zvol_remove(zv); | |
1436 | zvol_free(zv); | |
1437 | ||
1438 | return (0); | |
1439 | } | |
1440 | ||
1441 | /* | |
1442 | * Remove a block device minor node for the specified volume. | |
1443 | */ | |
1444 | int | |
1445 | zvol_remove_minor(const char *name) | |
1446 | { | |
1447 | int error; | |
1448 | ||
1449 | mutex_enter(&zvol_state_lock); | |
1450 | error = __zvol_remove_minor(name); | |
1451 | mutex_exit(&zvol_state_lock); | |
1452 | ||
a08ee875 LG |
1453 | return (SET_ERROR(error)); |
1454 | } | |
1455 | ||
1456 | /* | |
1457 | * Rename a block device minor mode for the specified volume. | |
1458 | */ | |
1459 | static void | |
1460 | __zvol_rename_minor(zvol_state_t *zv, const char *newname) | |
1461 | { | |
1462 | int readonly = get_disk_ro(zv->zv_disk); | |
1463 | ||
1464 | ASSERT(MUTEX_HELD(&zvol_state_lock)); | |
1465 | ||
1466 | strlcpy(zv->zv_name, newname, sizeof (zv->zv_name)); | |
1467 | ||
1468 | /* | |
1469 | * The block device's read-only state is briefly changed causing | |
1470 | * a KOBJ_CHANGE uevent to be issued. This ensures udev detects | |
1471 | * the name change and fixes the symlinks. This does not change | |
1472 | * ZVOL_RDONLY in zv->zv_flags so the actual read-only state never | |
1473 | * changes. This would normally be done using kobject_uevent() but | |
1474 | * that is a GPL-only symbol which is why we need this workaround. | |
1475 | */ | |
1476 | set_disk_ro(zv->zv_disk, !readonly); | |
1477 | set_disk_ro(zv->zv_disk, readonly); | |
60101509 BB |
1478 | } |
1479 | ||
1480 | static int | |
a08ee875 | 1481 | zvol_create_minors_cb(const char *dsname, void *arg) |
60101509 | 1482 | { |
a08ee875 | 1483 | (void) zvol_create_minor(dsname); |
60101509 | 1484 | |
d5674448 | 1485 | return (0); |
60101509 BB |
1486 | } |
1487 | ||
1488 | /* | |
a08ee875 | 1489 | * Create minors for specified dataset including children and snapshots. |
60101509 BB |
1490 | */ |
1491 | int | |
a08ee875 | 1492 | zvol_create_minors(const char *name) |
60101509 | 1493 | { |
60101509 BB |
1494 | int error = 0; |
1495 | ||
a08ee875 LG |
1496 | if (!zvol_inhibit_dev) |
1497 | error = dmu_objset_find((char *)name, zvol_create_minors_cb, | |
1498 | NULL, DS_FIND_CHILDREN | DS_FIND_SNAPSHOTS); | |
1499 | ||
1500 | return (SET_ERROR(error)); | |
1501 | } | |
1502 | ||
1503 | /* | |
1504 | * Remove minors for specified dataset including children and snapshots. | |
1505 | */ | |
1506 | void | |
1507 | zvol_remove_minors(const char *name) | |
1508 | { | |
1509 | zvol_state_t *zv, *zv_next; | |
1510 | int namelen = ((name) ? strlen(name) : 0); | |
1511 | ||
74497b7a | 1512 | if (zvol_inhibit_dev) |
a08ee875 | 1513 | return; |
74497b7a | 1514 | |
60101509 | 1515 | mutex_enter(&zvol_state_lock); |
a08ee875 LG |
1516 | |
1517 | for (zv = list_head(&zvol_state_list); zv != NULL; zv = zv_next) { | |
1518 | zv_next = list_next(&zvol_state_list, zv); | |
1519 | ||
1520 | if (name == NULL || strcmp(zv->zv_name, name) == 0 || | |
1521 | (strncmp(zv->zv_name, name, namelen) == 0 && | |
1522 | zv->zv_name[namelen] == '/')) { | |
1523 | zvol_remove(zv); | |
1524 | zvol_free(zv); | |
60101509 | 1525 | } |
60101509 | 1526 | } |
60101509 | 1527 | |
a08ee875 | 1528 | mutex_exit(&zvol_state_lock); |
60101509 BB |
1529 | } |
1530 | ||
1531 | /* | |
a08ee875 | 1532 | * Rename minors for specified dataset including children and snapshots. |
60101509 BB |
1533 | */ |
1534 | void | |
a08ee875 | 1535 | zvol_rename_minors(const char *oldname, const char *newname) |
60101509 BB |
1536 | { |
1537 | zvol_state_t *zv, *zv_next; | |
a08ee875 LG |
1538 | int oldnamelen, newnamelen; |
1539 | char *name; | |
60101509 | 1540 | |
74497b7a DH |
1541 | if (zvol_inhibit_dev) |
1542 | return; | |
1543 | ||
a08ee875 LG |
1544 | oldnamelen = strlen(oldname); |
1545 | newnamelen = strlen(newname); | |
ea04106b | 1546 | name = kmem_alloc(MAXNAMELEN, KM_SLEEP); |
60101509 BB |
1547 | |
1548 | mutex_enter(&zvol_state_lock); | |
a08ee875 | 1549 | |
60101509 BB |
1550 | for (zv = list_head(&zvol_state_list); zv != NULL; zv = zv_next) { |
1551 | zv_next = list_next(&zvol_state_list, zv); | |
1552 | ||
a08ee875 LG |
1553 | if (strcmp(zv->zv_name, oldname) == 0) { |
1554 | __zvol_rename_minor(zv, newname); | |
1555 | } else if (strncmp(zv->zv_name, oldname, oldnamelen) == 0 && | |
1556 | (zv->zv_name[oldnamelen] == '/' || | |
1557 | zv->zv_name[oldnamelen] == '@')) { | |
1558 | snprintf(name, MAXNAMELEN, "%s%c%s", newname, | |
1559 | zv->zv_name[oldnamelen], | |
1560 | zv->zv_name + oldnamelen + 1); | |
1561 | __zvol_rename_minor(zv, name); | |
60101509 BB |
1562 | } |
1563 | } | |
a08ee875 | 1564 | |
60101509 | 1565 | mutex_exit(&zvol_state_lock); |
a08ee875 LG |
1566 | |
1567 | kmem_free(name, MAXNAMELEN); | |
60101509 BB |
1568 | } |
1569 | ||
0b4d1b58 ED |
1570 | static int |
1571 | snapdev_snapshot_changed_cb(const char *dsname, void *arg) { | |
1572 | uint64_t snapdev = *(uint64_t *) arg; | |
1573 | ||
1574 | if (strchr(dsname, '@') == NULL) | |
a08ee875 | 1575 | return (0); |
0b4d1b58 ED |
1576 | |
1577 | switch (snapdev) { | |
1578 | case ZFS_SNAPDEV_VISIBLE: | |
1579 | mutex_enter(&zvol_state_lock); | |
1580 | (void) __zvol_create_minor(dsname, B_TRUE); | |
1581 | mutex_exit(&zvol_state_lock); | |
1582 | break; | |
1583 | case ZFS_SNAPDEV_HIDDEN: | |
1584 | (void) zvol_remove_minor(dsname); | |
1585 | break; | |
1586 | } | |
a08ee875 LG |
1587 | |
1588 | return (0); | |
0b4d1b58 ED |
1589 | } |
1590 | ||
1591 | int | |
1592 | zvol_set_snapdev(const char *dsname, uint64_t snapdev) { | |
1593 | (void) dmu_objset_find((char *) dsname, snapdev_snapshot_changed_cb, | |
1594 | &snapdev, DS_FIND_SNAPSHOTS | DS_FIND_CHILDREN); | |
1595 | /* caller should continue to modify snapdev property */ | |
1596 | return (-1); | |
1597 | } | |
1598 | ||
60101509 BB |
1599 | int |
1600 | zvol_init(void) | |
1601 | { | |
1602 | int error; | |
1603 | ||
c06d4368 | 1604 | list_create(&zvol_state_list, sizeof (zvol_state_t), |
a08ee875 LG |
1605 | offsetof(zvol_state_t, zv_next)); |
1606 | ||
c06d4368 AX |
1607 | mutex_init(&zvol_state_lock, NULL, MUTEX_DEFAULT, NULL); |
1608 | ||
60101509 BB |
1609 | error = register_blkdev(zvol_major, ZVOL_DRIVER); |
1610 | if (error) { | |
1611 | printk(KERN_INFO "ZFS: register_blkdev() failed %d\n", error); | |
e10b0808 | 1612 | goto out; |
60101509 BB |
1613 | } |
1614 | ||
1615 | blk_register_region(MKDEV(zvol_major, 0), 1UL << MINORBITS, | |
a08ee875 | 1616 | THIS_MODULE, zvol_probe, NULL, NULL); |
60101509 | 1617 | |
c06d4368 | 1618 | return (0); |
60101509 | 1619 | |
e10b0808 | 1620 | out: |
c06d4368 AX |
1621 | mutex_destroy(&zvol_state_lock); |
1622 | list_destroy(&zvol_state_list); | |
60101509 | 1623 | |
a08ee875 | 1624 | return (SET_ERROR(error)); |
60101509 BB |
1625 | } |
1626 | ||
1627 | void | |
1628 | zvol_fini(void) | |
1629 | { | |
1630 | zvol_remove_minors(NULL); | |
1631 | blk_unregister_region(MKDEV(zvol_major, 0), 1UL << MINORBITS); | |
1632 | unregister_blkdev(zvol_major, ZVOL_DRIVER); | |
60101509 BB |
1633 | mutex_destroy(&zvol_state_lock); |
1634 | list_destroy(&zvol_state_list); | |
1635 | } | |
1636 | ||
74497b7a DH |
1637 | module_param(zvol_inhibit_dev, uint, 0644); |
1638 | MODULE_PARM_DESC(zvol_inhibit_dev, "Do not create zvol device nodes"); | |
1639 | ||
30a9524e | 1640 | module_param(zvol_major, uint, 0444); |
60101509 BB |
1641 | MODULE_PARM_DESC(zvol_major, "Major number for zvol device"); |
1642 | ||
7c0e5708 | 1643 | module_param(zvol_max_discard_blocks, ulong, 0444); |
a08ee875 | 1644 | MODULE_PARM_DESC(zvol_max_discard_blocks, "Max number of blocks to discard"); |
e10b0808 AX |
1645 | |
1646 | module_param(zvol_prefetch_bytes, uint, 0644); | |
1647 | MODULE_PARM_DESC(zvol_prefetch_bytes, "Prefetch N bytes at zvol start+end"); |