Print this page
1693 persistent 'comment' field for a zpool
| Split |
Close |
| Expand all |
| Collapse all |
--- old/usr/src/uts/common/fs/zfs/spa_config.c
+++ new/usr/src/uts/common/fs/zfs/spa_config.c
1 1 /*
2 2 * CDDL HEADER START
3 3 *
4 4 * The contents of this file are subject to the terms of the
5 5 * Common Development and Distribution License (the "License").
6 6 * You may not use this file except in compliance with the License.
7 7 *
8 8 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
9 9 * or http://www.opensolaris.org/os/licensing.
10 10 * See the License for the specific language governing permissions
11 11 * and limitations under the License.
12 12 *
13 13 * When distributing Covered Code, include this CDDL HEADER in each
|
↓ open down ↓ |
13 lines elided |
↑ open up ↑ |
14 14 * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
15 15 * If applicable, add the following below this CDDL HEADER, with the
16 16 * fields enclosed by brackets "[]" replaced with your own identifying
17 17 * information: Portions Copyright [yyyy] [name of copyright owner]
18 18 *
19 19 * CDDL HEADER END
20 20 */
21 21
22 22 /*
23 23 * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
24 + * Copyright 2011 Nexenta Systems, Inc. All rights reserved.
25 + * Copyright (c) 2011 by Delphix. All rights reserved.
24 26 */
25 27
26 28 #include <sys/spa.h>
27 29 #include <sys/spa_impl.h>
28 30 #include <sys/nvpair.h>
29 31 #include <sys/uio.h>
30 32 #include <sys/fs/zfs.h>
31 33 #include <sys/vdev_impl.h>
32 34 #include <sys/zfs_ioctl.h>
33 35 #include <sys/utsname.h>
34 36 #include <sys/systeminfo.h>
35 37 #include <sys/sunddi.h>
36 38 #ifdef _KERNEL
37 39 #include <sys/kobj.h>
38 40 #include <sys/zone.h>
39 41 #endif
40 42
41 43 /*
42 44 * Pool configuration repository.
43 45 *
44 46 * Pool configuration is stored as a packed nvlist on the filesystem. By
45 47 * default, all pools are stored in /etc/zfs/zpool.cache and loaded on boot
46 48 * (when the ZFS module is loaded). Pools can also have the 'cachefile'
47 49 * property set that allows them to be stored in an alternate location until
48 50 * the control of external software.
49 51 *
50 52 * For each cache file, we have a single nvlist which holds all the
51 53 * configuration information. When the module loads, we read this information
52 54 * from /etc/zfs/zpool.cache and populate the SPA namespace. This namespace is
53 55 * maintained independently in spa.c. Whenever the namespace is modified, or
54 56 * the configuration of a pool is changed, we call spa_config_sync(), which
55 57 * walks through all the active pools and writes the configuration to disk.
56 58 */
57 59
58 60 static uint64_t spa_config_generation = 1;
59 61
60 62 /*
61 63 * This can be overridden in userland to preserve an alternate namespace for
62 64 * userland pools when doing testing.
63 65 */
64 66 const char *spa_config_path = ZPOOL_CACHE;
65 67
66 68 /*
67 69 * Called when the module is first loaded, this routine loads the configuration
68 70 * file into the SPA namespace. It does not actually open or load the pools; it
69 71 * only populates the namespace.
70 72 */
71 73 void
72 74 spa_config_load(void)
73 75 {
74 76 void *buf = NULL;
75 77 nvlist_t *nvlist, *child;
76 78 nvpair_t *nvpair;
77 79 char *pathname;
78 80 struct _buf *file;
79 81 uint64_t fsize;
80 82
81 83 /*
82 84 * Open the configuration file.
83 85 */
84 86 pathname = kmem_alloc(MAXPATHLEN, KM_SLEEP);
85 87
86 88 (void) snprintf(pathname, MAXPATHLEN, "%s%s",
87 89 (rootdir != NULL) ? "./" : "", spa_config_path);
88 90
89 91 file = kobj_open_file(pathname);
90 92
91 93 kmem_free(pathname, MAXPATHLEN);
92 94
93 95 if (file == (struct _buf *)-1)
94 96 return;
95 97
96 98 if (kobj_get_filesize(file, &fsize) != 0)
97 99 goto out;
98 100
99 101 buf = kmem_alloc(fsize, KM_SLEEP);
100 102
101 103 /*
102 104 * Read the nvlist from the file.
103 105 */
104 106 if (kobj_read_file(file, buf, fsize, 0) < 0)
105 107 goto out;
106 108
107 109 /*
108 110 * Unpack the nvlist.
109 111 */
110 112 if (nvlist_unpack(buf, fsize, &nvlist, KM_SLEEP) != 0)
111 113 goto out;
112 114
113 115 /*
114 116 * Iterate over all elements in the nvlist, creating a new spa_t for
115 117 * each one with the specified configuration.
116 118 */
117 119 mutex_enter(&spa_namespace_lock);
118 120 nvpair = NULL;
119 121 while ((nvpair = nvlist_next_nvpair(nvlist, nvpair)) != NULL) {
120 122 if (nvpair_type(nvpair) != DATA_TYPE_NVLIST)
121 123 continue;
122 124
123 125 VERIFY(nvpair_value_nvlist(nvpair, &child) == 0);
124 126
125 127 if (spa_lookup(nvpair_name(nvpair)) != NULL)
126 128 continue;
127 129 (void) spa_add(nvpair_name(nvpair), child, NULL);
128 130 }
129 131 mutex_exit(&spa_namespace_lock);
130 132
131 133 nvlist_free(nvlist);
132 134
133 135 out:
134 136 if (buf != NULL)
135 137 kmem_free(buf, fsize);
136 138
137 139 kobj_close_file(file);
138 140 }
139 141
140 142 static void
141 143 spa_config_write(spa_config_dirent_t *dp, nvlist_t *nvl)
142 144 {
143 145 size_t buflen;
144 146 char *buf;
145 147 vnode_t *vp;
146 148 int oflags = FWRITE | FTRUNC | FCREAT | FOFFMAX;
147 149 char *temp;
148 150
149 151 /*
150 152 * If the nvlist is empty (NULL), then remove the old cachefile.
151 153 */
152 154 if (nvl == NULL) {
153 155 (void) vn_remove(dp->scd_path, UIO_SYSSPACE, RMFILE);
154 156 return;
155 157 }
156 158
157 159 /*
158 160 * Pack the configuration into a buffer.
159 161 */
160 162 VERIFY(nvlist_size(nvl, &buflen, NV_ENCODE_XDR) == 0);
161 163
162 164 buf = kmem_alloc(buflen, KM_SLEEP);
163 165 temp = kmem_zalloc(MAXPATHLEN, KM_SLEEP);
164 166
165 167 VERIFY(nvlist_pack(nvl, &buf, &buflen, NV_ENCODE_XDR,
166 168 KM_SLEEP) == 0);
167 169
168 170 /*
169 171 * Write the configuration to disk. We need to do the traditional
170 172 * 'write to temporary file, sync, move over original' to make sure we
171 173 * always have a consistent view of the data.
172 174 */
173 175 (void) snprintf(temp, MAXPATHLEN, "%s.tmp", dp->scd_path);
174 176
175 177 if (vn_open(temp, UIO_SYSSPACE, oflags, 0644, &vp, CRCREAT, 0) == 0) {
176 178 if (vn_rdwr(UIO_WRITE, vp, buf, buflen, 0, UIO_SYSSPACE,
177 179 0, RLIM64_INFINITY, kcred, NULL) == 0 &&
178 180 VOP_FSYNC(vp, FSYNC, kcred, NULL) == 0) {
179 181 (void) vn_rename(temp, dp->scd_path, UIO_SYSSPACE);
180 182 }
181 183 (void) VOP_CLOSE(vp, oflags, 1, 0, kcred, NULL);
182 184 VN_RELE(vp);
183 185 }
184 186
185 187 (void) vn_remove(temp, UIO_SYSSPACE, RMFILE);
186 188
187 189 kmem_free(buf, buflen);
188 190 kmem_free(temp, MAXPATHLEN);
189 191 }
190 192
191 193 /*
192 194 * Synchronize pool configuration to disk. This must be called with the
193 195 * namespace lock held.
194 196 */
195 197 void
196 198 spa_config_sync(spa_t *target, boolean_t removing, boolean_t postsysevent)
197 199 {
198 200 spa_config_dirent_t *dp, *tdp;
199 201 nvlist_t *nvl;
200 202
201 203 ASSERT(MUTEX_HELD(&spa_namespace_lock));
202 204
203 205 if (rootdir == NULL || !(spa_mode_global & FWRITE))
204 206 return;
205 207
206 208 /*
207 209 * Iterate over all cachefiles for the pool, past or present. When the
208 210 * cachefile is changed, the new one is pushed onto this list, allowing
209 211 * us to update previous cachefiles that no longer contain this pool.
210 212 */
211 213 for (dp = list_head(&target->spa_config_list); dp != NULL;
212 214 dp = list_next(&target->spa_config_list, dp)) {
213 215 spa_t *spa = NULL;
214 216 if (dp->scd_path == NULL)
215 217 continue;
216 218
217 219 /*
218 220 * Iterate over all pools, adding any matching pools to 'nvl'.
219 221 */
220 222 nvl = NULL;
221 223 while ((spa = spa_next(spa)) != NULL) {
222 224 if (spa == target && removing)
223 225 continue;
224 226
225 227 mutex_enter(&spa->spa_props_lock);
226 228 tdp = list_head(&spa->spa_config_list);
227 229 if (spa->spa_config == NULL ||
228 230 tdp->scd_path == NULL ||
229 231 strcmp(tdp->scd_path, dp->scd_path) != 0) {
230 232 mutex_exit(&spa->spa_props_lock);
231 233 continue;
232 234 }
233 235
234 236 if (nvl == NULL)
235 237 VERIFY(nvlist_alloc(&nvl, NV_UNIQUE_NAME,
236 238 KM_SLEEP) == 0);
237 239
238 240 VERIFY(nvlist_add_nvlist(nvl, spa->spa_name,
239 241 spa->spa_config) == 0);
240 242 mutex_exit(&spa->spa_props_lock);
241 243 }
242 244
243 245 spa_config_write(dp, nvl);
244 246 nvlist_free(nvl);
245 247 }
246 248
247 249 /*
248 250 * Remove any config entries older than the current one.
249 251 */
250 252 dp = list_head(&target->spa_config_list);
251 253 while ((tdp = list_next(&target->spa_config_list, dp)) != NULL) {
252 254 list_remove(&target->spa_config_list, tdp);
253 255 if (tdp->scd_path != NULL)
254 256 spa_strfree(tdp->scd_path);
255 257 kmem_free(tdp, sizeof (spa_config_dirent_t));
256 258 }
257 259
258 260 spa_config_generation++;
259 261
260 262 if (postsysevent)
261 263 spa_event_notify(target, NULL, ESC_ZFS_CONFIG_SYNC);
262 264 }
263 265
264 266 /*
265 267 * Sigh. Inside a local zone, we don't have access to /etc/zfs/zpool.cache,
266 268 * and we don't want to allow the local zone to see all the pools anyway.
267 269 * So we have to invent the ZFS_IOC_CONFIG ioctl to grab the configuration
268 270 * information for all pool visible within the zone.
269 271 */
270 272 nvlist_t *
271 273 spa_all_configs(uint64_t *generation)
272 274 {
273 275 nvlist_t *pools;
274 276 spa_t *spa = NULL;
275 277
276 278 if (*generation == spa_config_generation)
277 279 return (NULL);
278 280
279 281 VERIFY(nvlist_alloc(&pools, NV_UNIQUE_NAME, KM_SLEEP) == 0);
280 282
281 283 mutex_enter(&spa_namespace_lock);
282 284 while ((spa = spa_next(spa)) != NULL) {
283 285 if (INGLOBALZONE(curproc) ||
284 286 zone_dataset_visible(spa_name(spa), NULL)) {
285 287 mutex_enter(&spa->spa_props_lock);
286 288 VERIFY(nvlist_add_nvlist(pools, spa_name(spa),
287 289 spa->spa_config) == 0);
288 290 mutex_exit(&spa->spa_props_lock);
289 291 }
290 292 }
291 293 *generation = spa_config_generation;
292 294 mutex_exit(&spa_namespace_lock);
293 295
294 296 return (pools);
295 297 }
296 298
297 299 void
298 300 spa_config_set(spa_t *spa, nvlist_t *config)
299 301 {
300 302 mutex_enter(&spa->spa_props_lock);
301 303 if (spa->spa_config != NULL)
302 304 nvlist_free(spa->spa_config);
303 305 spa->spa_config = config;
304 306 mutex_exit(&spa->spa_props_lock);
305 307 }
306 308
307 309 /*
308 310 * Generate the pool's configuration based on the current in-core state.
309 311 * We infer whether to generate a complete config or just one top-level config
310 312 * based on whether vd is the root vdev.
311 313 */
312 314 nvlist_t *
313 315 spa_config_generate(spa_t *spa, vdev_t *vd, uint64_t txg, int getstats)
314 316 {
315 317 nvlist_t *config, *nvroot;
316 318 vdev_t *rvd = spa->spa_root_vdev;
317 319 unsigned long hostid = 0;
318 320 boolean_t locked = B_FALSE;
319 321 uint64_t split_guid;
320 322
321 323 if (vd == NULL) {
322 324 vd = rvd;
323 325 locked = B_TRUE;
324 326 spa_config_enter(spa, SCL_CONFIG | SCL_STATE, FTAG, RW_READER);
325 327 }
326 328
327 329 ASSERT(spa_config_held(spa, SCL_CONFIG | SCL_STATE, RW_READER) ==
328 330 (SCL_CONFIG | SCL_STATE));
329 331
330 332 /*
331 333 * If txg is -1, report the current value of spa->spa_config_txg.
332 334 */
333 335 if (txg == -1ULL)
334 336 txg = spa->spa_config_txg;
335 337
336 338 VERIFY(nvlist_alloc(&config, NV_UNIQUE_NAME, KM_SLEEP) == 0);
337 339
|
↓ open down ↓ |
304 lines elided |
↑ open up ↑ |
338 340 VERIFY(nvlist_add_uint64(config, ZPOOL_CONFIG_VERSION,
339 341 spa_version(spa)) == 0);
340 342 VERIFY(nvlist_add_string(config, ZPOOL_CONFIG_POOL_NAME,
341 343 spa_name(spa)) == 0);
342 344 VERIFY(nvlist_add_uint64(config, ZPOOL_CONFIG_POOL_STATE,
343 345 spa_state(spa)) == 0);
344 346 VERIFY(nvlist_add_uint64(config, ZPOOL_CONFIG_POOL_TXG,
345 347 txg) == 0);
346 348 VERIFY(nvlist_add_uint64(config, ZPOOL_CONFIG_POOL_GUID,
347 349 spa_guid(spa)) == 0);
350 + VERIFY(spa->spa_comment == NULL || nvlist_add_string(config,
351 + ZPOOL_CONFIG_COMMENT, spa->spa_comment) == 0);
352 +
353 +
348 354 #ifdef _KERNEL
349 355 hostid = zone_get_hostid(NULL);
350 356 #else /* _KERNEL */
351 357 /*
352 358 * We're emulating the system's hostid in userland, so we can't use
353 359 * zone_get_hostid().
354 360 */
355 361 (void) ddi_strtoul(hw_serial, NULL, 10, &hostid);
356 362 #endif /* _KERNEL */
357 363 if (hostid != 0) {
358 364 VERIFY(nvlist_add_uint64(config, ZPOOL_CONFIG_HOSTID,
359 365 hostid) == 0);
360 366 }
361 367 VERIFY(nvlist_add_string(config, ZPOOL_CONFIG_HOSTNAME,
362 368 utsname.nodename) == 0);
363 369
364 370 if (vd != rvd) {
365 371 VERIFY(nvlist_add_uint64(config, ZPOOL_CONFIG_TOP_GUID,
366 372 vd->vdev_top->vdev_guid) == 0);
367 373 VERIFY(nvlist_add_uint64(config, ZPOOL_CONFIG_GUID,
368 374 vd->vdev_guid) == 0);
369 375 if (vd->vdev_isspare)
370 376 VERIFY(nvlist_add_uint64(config, ZPOOL_CONFIG_IS_SPARE,
371 377 1ULL) == 0);
372 378 if (vd->vdev_islog)
373 379 VERIFY(nvlist_add_uint64(config, ZPOOL_CONFIG_IS_LOG,
374 380 1ULL) == 0);
375 381 vd = vd->vdev_top; /* label contains top config */
376 382 } else {
377 383 /*
378 384 * Only add the (potentially large) split information
379 385 * in the mos config, and not in the vdev labels
380 386 */
381 387 if (spa->spa_config_splitting != NULL)
382 388 VERIFY(nvlist_add_nvlist(config, ZPOOL_CONFIG_SPLIT,
383 389 spa->spa_config_splitting) == 0);
384 390 }
385 391
386 392 /*
387 393 * Add the top-level config. We even add this on pools which
388 394 * don't support holes in the namespace.
389 395 */
390 396 vdev_top_config_generate(spa, config);
391 397
392 398 /*
393 399 * If we're splitting, record the original pool's guid.
394 400 */
395 401 if (spa->spa_config_splitting != NULL &&
396 402 nvlist_lookup_uint64(spa->spa_config_splitting,
397 403 ZPOOL_CONFIG_SPLIT_GUID, &split_guid) == 0) {
398 404 VERIFY(nvlist_add_uint64(config, ZPOOL_CONFIG_SPLIT_GUID,
399 405 split_guid) == 0);
400 406 }
401 407
402 408 nvroot = vdev_config_generate(spa, vd, getstats, 0);
403 409 VERIFY(nvlist_add_nvlist(config, ZPOOL_CONFIG_VDEV_TREE, nvroot) == 0);
404 410 nvlist_free(nvroot);
405 411
406 412 if (getstats && spa_load_state(spa) == SPA_LOAD_NONE) {
407 413 ddt_histogram_t *ddh;
408 414 ddt_stat_t *dds;
409 415 ddt_object_t *ddo;
410 416
411 417 ddh = kmem_zalloc(sizeof (ddt_histogram_t), KM_SLEEP);
412 418 ddt_get_dedup_histogram(spa, ddh);
413 419 VERIFY(nvlist_add_uint64_array(config,
414 420 ZPOOL_CONFIG_DDT_HISTOGRAM,
415 421 (uint64_t *)ddh, sizeof (*ddh) / sizeof (uint64_t)) == 0);
416 422 kmem_free(ddh, sizeof (ddt_histogram_t));
417 423
418 424 ddo = kmem_zalloc(sizeof (ddt_object_t), KM_SLEEP);
419 425 ddt_get_dedup_object_stats(spa, ddo);
420 426 VERIFY(nvlist_add_uint64_array(config,
421 427 ZPOOL_CONFIG_DDT_OBJ_STATS,
422 428 (uint64_t *)ddo, sizeof (*ddo) / sizeof (uint64_t)) == 0);
423 429 kmem_free(ddo, sizeof (ddt_object_t));
424 430
425 431 dds = kmem_zalloc(sizeof (ddt_stat_t), KM_SLEEP);
426 432 ddt_get_dedup_stats(spa, dds);
427 433 VERIFY(nvlist_add_uint64_array(config,
428 434 ZPOOL_CONFIG_DDT_STATS,
429 435 (uint64_t *)dds, sizeof (*dds) / sizeof (uint64_t)) == 0);
430 436 kmem_free(dds, sizeof (ddt_stat_t));
431 437 }
432 438
433 439 if (locked)
434 440 spa_config_exit(spa, SCL_CONFIG | SCL_STATE, FTAG);
435 441
436 442 return (config);
437 443 }
438 444
439 445 /*
440 446 * Update all disk labels, generate a fresh config based on the current
441 447 * in-core state, and sync the global config cache (do not sync the config
442 448 * cache if this is a booting rootpool).
443 449 */
444 450 void
445 451 spa_config_update(spa_t *spa, int what)
446 452 {
447 453 vdev_t *rvd = spa->spa_root_vdev;
448 454 uint64_t txg;
449 455 int c;
450 456
451 457 ASSERT(MUTEX_HELD(&spa_namespace_lock));
452 458
453 459 spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER);
454 460 txg = spa_last_synced_txg(spa) + 1;
455 461 if (what == SPA_CONFIG_UPDATE_POOL) {
456 462 vdev_config_dirty(rvd);
457 463 } else {
458 464 /*
459 465 * If we have top-level vdevs that were added but have
460 466 * not yet been prepared for allocation, do that now.
461 467 * (It's safe now because the config cache is up to date,
462 468 * so it will be able to translate the new DVAs.)
463 469 * See comments in spa_vdev_add() for full details.
464 470 */
465 471 for (c = 0; c < rvd->vdev_children; c++) {
466 472 vdev_t *tvd = rvd->vdev_child[c];
467 473 if (tvd->vdev_ms_array == 0)
468 474 vdev_metaslab_set_size(tvd);
469 475 vdev_expand(tvd, txg);
470 476 }
471 477 }
472 478 spa_config_exit(spa, SCL_ALL, FTAG);
473 479
474 480 /*
475 481 * Wait for the mosconfig to be regenerated and synced.
476 482 */
477 483 txg_wait_synced(spa->spa_dsl_pool, txg);
478 484
479 485 /*
480 486 * Update the global config cache to reflect the new mosconfig.
481 487 */
482 488 if (!spa->spa_is_root)
483 489 spa_config_sync(spa, B_FALSE, what != SPA_CONFIG_UPDATE_POOL);
484 490
485 491 if (what == SPA_CONFIG_UPDATE_POOL)
486 492 spa_config_update(spa, SPA_CONFIG_UPDATE_VDEVS);
487 493 }
|
↓ open down ↓ |
130 lines elided |
↑ open up ↑ |
XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX