Print this page
3742 zfs comments need cleaner, more consistent style
Submitted by: Will Andrews <willa@spectralogic.com>
Submitted by: Alan Somers <alans@spectralogic.com>
Reviewed by: Matthew Ahrens <mahrens@delphix.com>
Reviewed by: George Wilson <george.wilson@delphix.com>
Reviewed by: Eric Schrock <eric.schrock@delphix.com>
| 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
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 24 * Copyright 2011 Nexenta Systems, Inc. All rights reserved.
25 25 * Copyright (c) 2012 by Delphix. All rights reserved.
26 26 */
27 27
28 28 #include <sys/spa.h>
29 29 #include <sys/spa_impl.h>
30 30 #include <sys/nvpair.h>
31 31 #include <sys/uio.h>
32 32 #include <sys/fs/zfs.h>
33 33 #include <sys/vdev_impl.h>
34 34 #include <sys/zfs_ioctl.h>
35 35 #include <sys/utsname.h>
36 36 #include <sys/systeminfo.h>
37 37 #include <sys/sunddi.h>
38 38 #include <sys/zfeature.h>
39 39 #ifdef _KERNEL
40 40 #include <sys/kobj.h>
41 41 #include <sys/zone.h>
42 42 #endif
43 43
44 44 /*
45 45 * Pool configuration repository.
46 46 *
47 47 * Pool configuration is stored as a packed nvlist on the filesystem. By
48 48 * default, all pools are stored in /etc/zfs/zpool.cache and loaded on boot
49 49 * (when the ZFS module is loaded). Pools can also have the 'cachefile'
50 50 * property set that allows them to be stored in an alternate location until
51 51 * the control of external software.
52 52 *
53 53 * For each cache file, we have a single nvlist which holds all the
54 54 * configuration information. When the module loads, we read this information
55 55 * from /etc/zfs/zpool.cache and populate the SPA namespace. This namespace is
56 56 * maintained independently in spa.c. Whenever the namespace is modified, or
57 57 * the configuration of a pool is changed, we call spa_config_sync(), which
58 58 * walks through all the active pools and writes the configuration to disk.
59 59 */
60 60
61 61 static uint64_t spa_config_generation = 1;
62 62
63 63 /*
64 64 * This can be overridden in userland to preserve an alternate namespace for
65 65 * userland pools when doing testing.
66 66 */
67 67 const char *spa_config_path = ZPOOL_CACHE;
68 68
69 69 /*
70 70 * Called when the module is first loaded, this routine loads the configuration
71 71 * file into the SPA namespace. It does not actually open or load the pools; it
72 72 * only populates the namespace.
73 73 */
74 74 void
75 75 spa_config_load(void)
76 76 {
77 77 void *buf = NULL;
78 78 nvlist_t *nvlist, *child;
79 79 nvpair_t *nvpair;
80 80 char *pathname;
81 81 struct _buf *file;
82 82 uint64_t fsize;
83 83
84 84 /*
85 85 * Open the configuration file.
86 86 */
87 87 pathname = kmem_alloc(MAXPATHLEN, KM_SLEEP);
88 88
89 89 (void) snprintf(pathname, MAXPATHLEN, "%s%s",
90 90 (rootdir != NULL) ? "./" : "", spa_config_path);
91 91
92 92 file = kobj_open_file(pathname);
93 93
94 94 kmem_free(pathname, MAXPATHLEN);
95 95
96 96 if (file == (struct _buf *)-1)
97 97 return;
98 98
99 99 if (kobj_get_filesize(file, &fsize) != 0)
100 100 goto out;
101 101
102 102 buf = kmem_alloc(fsize, KM_SLEEP);
103 103
104 104 /*
105 105 * Read the nvlist from the file.
106 106 */
107 107 if (kobj_read_file(file, buf, fsize, 0) < 0)
108 108 goto out;
109 109
110 110 /*
111 111 * Unpack the nvlist.
112 112 */
113 113 if (nvlist_unpack(buf, fsize, &nvlist, KM_SLEEP) != 0)
114 114 goto out;
115 115
116 116 /*
117 117 * Iterate over all elements in the nvlist, creating a new spa_t for
118 118 * each one with the specified configuration.
119 119 */
120 120 mutex_enter(&spa_namespace_lock);
121 121 nvpair = NULL;
122 122 while ((nvpair = nvlist_next_nvpair(nvlist, nvpair)) != NULL) {
123 123 if (nvpair_type(nvpair) != DATA_TYPE_NVLIST)
124 124 continue;
125 125
126 126 VERIFY(nvpair_value_nvlist(nvpair, &child) == 0);
127 127
128 128 if (spa_lookup(nvpair_name(nvpair)) != NULL)
129 129 continue;
130 130 (void) spa_add(nvpair_name(nvpair), child, NULL);
131 131 }
132 132 mutex_exit(&spa_namespace_lock);
133 133
134 134 nvlist_free(nvlist);
135 135
136 136 out:
137 137 if (buf != NULL)
138 138 kmem_free(buf, fsize);
139 139
140 140 kobj_close_file(file);
141 141 }
142 142
143 143 static void
144 144 spa_config_write(spa_config_dirent_t *dp, nvlist_t *nvl)
145 145 {
146 146 size_t buflen;
147 147 char *buf;
148 148 vnode_t *vp;
149 149 int oflags = FWRITE | FTRUNC | FCREAT | FOFFMAX;
150 150 char *temp;
151 151
152 152 /*
153 153 * If the nvlist is empty (NULL), then remove the old cachefile.
154 154 */
155 155 if (nvl == NULL) {
156 156 (void) vn_remove(dp->scd_path, UIO_SYSSPACE, RMFILE);
157 157 return;
158 158 }
159 159
160 160 /*
161 161 * Pack the configuration into a buffer.
162 162 */
163 163 VERIFY(nvlist_size(nvl, &buflen, NV_ENCODE_XDR) == 0);
164 164
165 165 buf = kmem_alloc(buflen, KM_SLEEP);
166 166 temp = kmem_zalloc(MAXPATHLEN, KM_SLEEP);
167 167
168 168 VERIFY(nvlist_pack(nvl, &buf, &buflen, NV_ENCODE_XDR,
169 169 KM_SLEEP) == 0);
170 170
171 171 /*
172 172 * Write the configuration to disk. We need to do the traditional
173 173 * 'write to temporary file, sync, move over original' to make sure we
174 174 * always have a consistent view of the data.
175 175 */
176 176 (void) snprintf(temp, MAXPATHLEN, "%s.tmp", dp->scd_path);
177 177
178 178 if (vn_open(temp, UIO_SYSSPACE, oflags, 0644, &vp, CRCREAT, 0) == 0) {
179 179 if (vn_rdwr(UIO_WRITE, vp, buf, buflen, 0, UIO_SYSSPACE,
180 180 0, RLIM64_INFINITY, kcred, NULL) == 0 &&
181 181 VOP_FSYNC(vp, FSYNC, kcred, NULL) == 0) {
182 182 (void) vn_rename(temp, dp->scd_path, UIO_SYSSPACE);
183 183 }
184 184 (void) VOP_CLOSE(vp, oflags, 1, 0, kcred, NULL);
185 185 VN_RELE(vp);
186 186 }
187 187
188 188 (void) vn_remove(temp, UIO_SYSSPACE, RMFILE);
189 189
190 190 kmem_free(buf, buflen);
191 191 kmem_free(temp, MAXPATHLEN);
192 192 }
193 193
194 194 /*
195 195 * Synchronize pool configuration to disk. This must be called with the
196 196 * namespace lock held.
197 197 */
198 198 void
199 199 spa_config_sync(spa_t *target, boolean_t removing, boolean_t postsysevent)
200 200 {
201 201 spa_config_dirent_t *dp, *tdp;
202 202 nvlist_t *nvl;
203 203
204 204 ASSERT(MUTEX_HELD(&spa_namespace_lock));
205 205
206 206 if (rootdir == NULL || !(spa_mode_global & FWRITE))
207 207 return;
208 208
209 209 /*
210 210 * Iterate over all cachefiles for the pool, past or present. When the
211 211 * cachefile is changed, the new one is pushed onto this list, allowing
212 212 * us to update previous cachefiles that no longer contain this pool.
213 213 */
214 214 for (dp = list_head(&target->spa_config_list); dp != NULL;
215 215 dp = list_next(&target->spa_config_list, dp)) {
216 216 spa_t *spa = NULL;
217 217 if (dp->scd_path == NULL)
218 218 continue;
219 219
220 220 /*
221 221 * Iterate over all pools, adding any matching pools to 'nvl'.
222 222 */
223 223 nvl = NULL;
224 224 while ((spa = spa_next(spa)) != NULL) {
225 225 /*
226 226 * Skip over our own pool if we're about to remove
227 227 * ourselves from the spa namespace or any pool that
228 228 * is readonly. Since we cannot guarantee that a
229 229 * readonly pool would successfully import upon reboot,
230 230 * we don't allow them to be written to the cache file.
231 231 */
232 232 if ((spa == target && removing) ||
233 233 !spa_writeable(spa))
234 234 continue;
235 235
236 236 mutex_enter(&spa->spa_props_lock);
237 237 tdp = list_head(&spa->spa_config_list);
238 238 if (spa->spa_config == NULL ||
239 239 tdp->scd_path == NULL ||
240 240 strcmp(tdp->scd_path, dp->scd_path) != 0) {
241 241 mutex_exit(&spa->spa_props_lock);
242 242 continue;
243 243 }
244 244
245 245 if (nvl == NULL)
246 246 VERIFY(nvlist_alloc(&nvl, NV_UNIQUE_NAME,
247 247 KM_SLEEP) == 0);
248 248
249 249 VERIFY(nvlist_add_nvlist(nvl, spa->spa_name,
250 250 spa->spa_config) == 0);
251 251 mutex_exit(&spa->spa_props_lock);
252 252 }
253 253
254 254 spa_config_write(dp, nvl);
255 255 nvlist_free(nvl);
256 256 }
257 257
258 258 /*
259 259 * Remove any config entries older than the current one.
260 260 */
261 261 dp = list_head(&target->spa_config_list);
262 262 while ((tdp = list_next(&target->spa_config_list, dp)) != NULL) {
263 263 list_remove(&target->spa_config_list, tdp);
264 264 if (tdp->scd_path != NULL)
265 265 spa_strfree(tdp->scd_path);
266 266 kmem_free(tdp, sizeof (spa_config_dirent_t));
267 267 }
268 268
269 269 spa_config_generation++;
270 270
271 271 if (postsysevent)
272 272 spa_event_notify(target, NULL, ESC_ZFS_CONFIG_SYNC);
273 273 }
274 274
275 275 /*
276 276 * Sigh. Inside a local zone, we don't have access to /etc/zfs/zpool.cache,
277 277 * and we don't want to allow the local zone to see all the pools anyway.
278 278 * So we have to invent the ZFS_IOC_CONFIG ioctl to grab the configuration
279 279 * information for all pool visible within the zone.
280 280 */
281 281 nvlist_t *
282 282 spa_all_configs(uint64_t *generation)
283 283 {
284 284 nvlist_t *pools;
285 285 spa_t *spa = NULL;
286 286
287 287 if (*generation == spa_config_generation)
288 288 return (NULL);
289 289
290 290 VERIFY(nvlist_alloc(&pools, NV_UNIQUE_NAME, KM_SLEEP) == 0);
291 291
292 292 mutex_enter(&spa_namespace_lock);
293 293 while ((spa = spa_next(spa)) != NULL) {
294 294 if (INGLOBALZONE(curproc) ||
295 295 zone_dataset_visible(spa_name(spa), NULL)) {
296 296 mutex_enter(&spa->spa_props_lock);
297 297 VERIFY(nvlist_add_nvlist(pools, spa_name(spa),
298 298 spa->spa_config) == 0);
299 299 mutex_exit(&spa->spa_props_lock);
300 300 }
301 301 }
302 302 *generation = spa_config_generation;
303 303 mutex_exit(&spa_namespace_lock);
304 304
305 305 return (pools);
306 306 }
307 307
308 308 void
309 309 spa_config_set(spa_t *spa, nvlist_t *config)
|
↓ open down ↓ |
309 lines elided |
↑ open up ↑ |
310 310 {
311 311 mutex_enter(&spa->spa_props_lock);
312 312 if (spa->spa_config != NULL)
313 313 nvlist_free(spa->spa_config);
314 314 spa->spa_config = config;
315 315 mutex_exit(&spa->spa_props_lock);
316 316 }
317 317
318 318 /*
319 319 * Generate the pool's configuration based on the current in-core state.
320 + *
320 321 * We infer whether to generate a complete config or just one top-level config
321 322 * based on whether vd is the root vdev.
322 323 */
323 324 nvlist_t *
324 325 spa_config_generate(spa_t *spa, vdev_t *vd, uint64_t txg, int getstats)
325 326 {
326 327 nvlist_t *config, *nvroot;
327 328 vdev_t *rvd = spa->spa_root_vdev;
328 329 unsigned long hostid = 0;
329 330 boolean_t locked = B_FALSE;
330 331 uint64_t split_guid;
331 332
332 333 if (vd == NULL) {
333 334 vd = rvd;
334 335 locked = B_TRUE;
335 336 spa_config_enter(spa, SCL_CONFIG | SCL_STATE, FTAG, RW_READER);
336 337 }
337 338
338 339 ASSERT(spa_config_held(spa, SCL_CONFIG | SCL_STATE, RW_READER) ==
339 340 (SCL_CONFIG | SCL_STATE));
340 341
341 342 /*
342 343 * If txg is -1, report the current value of spa->spa_config_txg.
343 344 */
344 345 if (txg == -1ULL)
345 346 txg = spa->spa_config_txg;
346 347
347 348 VERIFY(nvlist_alloc(&config, NV_UNIQUE_NAME, KM_SLEEP) == 0);
348 349
349 350 VERIFY(nvlist_add_uint64(config, ZPOOL_CONFIG_VERSION,
350 351 spa_version(spa)) == 0);
351 352 VERIFY(nvlist_add_string(config, ZPOOL_CONFIG_POOL_NAME,
352 353 spa_name(spa)) == 0);
353 354 VERIFY(nvlist_add_uint64(config, ZPOOL_CONFIG_POOL_STATE,
354 355 spa_state(spa)) == 0);
355 356 VERIFY(nvlist_add_uint64(config, ZPOOL_CONFIG_POOL_TXG,
356 357 txg) == 0);
357 358 VERIFY(nvlist_add_uint64(config, ZPOOL_CONFIG_POOL_GUID,
358 359 spa_guid(spa)) == 0);
359 360 VERIFY(spa->spa_comment == NULL || nvlist_add_string(config,
360 361 ZPOOL_CONFIG_COMMENT, spa->spa_comment) == 0);
361 362
362 363
363 364 #ifdef _KERNEL
364 365 hostid = zone_get_hostid(NULL);
365 366 #else /* _KERNEL */
366 367 /*
367 368 * We're emulating the system's hostid in userland, so we can't use
368 369 * zone_get_hostid().
369 370 */
370 371 (void) ddi_strtoul(hw_serial, NULL, 10, &hostid);
371 372 #endif /* _KERNEL */
372 373 if (hostid != 0) {
373 374 VERIFY(nvlist_add_uint64(config, ZPOOL_CONFIG_HOSTID,
374 375 hostid) == 0);
375 376 }
376 377 VERIFY(nvlist_add_string(config, ZPOOL_CONFIG_HOSTNAME,
377 378 utsname.nodename) == 0);
378 379
379 380 if (vd != rvd) {
380 381 VERIFY(nvlist_add_uint64(config, ZPOOL_CONFIG_TOP_GUID,
381 382 vd->vdev_top->vdev_guid) == 0);
382 383 VERIFY(nvlist_add_uint64(config, ZPOOL_CONFIG_GUID,
383 384 vd->vdev_guid) == 0);
384 385 if (vd->vdev_isspare)
385 386 VERIFY(nvlist_add_uint64(config, ZPOOL_CONFIG_IS_SPARE,
386 387 1ULL) == 0);
387 388 if (vd->vdev_islog)
388 389 VERIFY(nvlist_add_uint64(config, ZPOOL_CONFIG_IS_LOG,
389 390 1ULL) == 0);
390 391 vd = vd->vdev_top; /* label contains top config */
391 392 } else {
392 393 /*
393 394 * Only add the (potentially large) split information
394 395 * in the mos config, and not in the vdev labels
395 396 */
396 397 if (spa->spa_config_splitting != NULL)
397 398 VERIFY(nvlist_add_nvlist(config, ZPOOL_CONFIG_SPLIT,
398 399 spa->spa_config_splitting) == 0);
399 400 }
400 401
401 402 /*
402 403 * Add the top-level config. We even add this on pools which
403 404 * don't support holes in the namespace.
404 405 */
405 406 vdev_top_config_generate(spa, config);
406 407
407 408 /*
408 409 * If we're splitting, record the original pool's guid.
409 410 */
410 411 if (spa->spa_config_splitting != NULL &&
411 412 nvlist_lookup_uint64(spa->spa_config_splitting,
412 413 ZPOOL_CONFIG_SPLIT_GUID, &split_guid) == 0) {
413 414 VERIFY(nvlist_add_uint64(config, ZPOOL_CONFIG_SPLIT_GUID,
414 415 split_guid) == 0);
415 416 }
416 417
417 418 nvroot = vdev_config_generate(spa, vd, getstats, 0);
418 419 VERIFY(nvlist_add_nvlist(config, ZPOOL_CONFIG_VDEV_TREE, nvroot) == 0);
419 420 nvlist_free(nvroot);
420 421
421 422 /*
422 423 * Store what's necessary for reading the MOS in the label.
423 424 */
424 425 VERIFY(nvlist_add_nvlist(config, ZPOOL_CONFIG_FEATURES_FOR_READ,
425 426 spa->spa_label_features) == 0);
426 427
427 428 if (getstats && spa_load_state(spa) == SPA_LOAD_NONE) {
428 429 ddt_histogram_t *ddh;
429 430 ddt_stat_t *dds;
430 431 ddt_object_t *ddo;
431 432
432 433 ddh = kmem_zalloc(sizeof (ddt_histogram_t), KM_SLEEP);
433 434 ddt_get_dedup_histogram(spa, ddh);
434 435 VERIFY(nvlist_add_uint64_array(config,
435 436 ZPOOL_CONFIG_DDT_HISTOGRAM,
436 437 (uint64_t *)ddh, sizeof (*ddh) / sizeof (uint64_t)) == 0);
437 438 kmem_free(ddh, sizeof (ddt_histogram_t));
438 439
439 440 ddo = kmem_zalloc(sizeof (ddt_object_t), KM_SLEEP);
440 441 ddt_get_dedup_object_stats(spa, ddo);
441 442 VERIFY(nvlist_add_uint64_array(config,
442 443 ZPOOL_CONFIG_DDT_OBJ_STATS,
443 444 (uint64_t *)ddo, sizeof (*ddo) / sizeof (uint64_t)) == 0);
444 445 kmem_free(ddo, sizeof (ddt_object_t));
445 446
446 447 dds = kmem_zalloc(sizeof (ddt_stat_t), KM_SLEEP);
447 448 ddt_get_dedup_stats(spa, dds);
448 449 VERIFY(nvlist_add_uint64_array(config,
449 450 ZPOOL_CONFIG_DDT_STATS,
450 451 (uint64_t *)dds, sizeof (*dds) / sizeof (uint64_t)) == 0);
451 452 kmem_free(dds, sizeof (ddt_stat_t));
452 453 }
453 454
454 455 if (locked)
455 456 spa_config_exit(spa, SCL_CONFIG | SCL_STATE, FTAG);
456 457
457 458 return (config);
458 459 }
459 460
460 461 /*
461 462 * Update all disk labels, generate a fresh config based on the current
462 463 * in-core state, and sync the global config cache (do not sync the config
463 464 * cache if this is a booting rootpool).
464 465 */
465 466 void
466 467 spa_config_update(spa_t *spa, int what)
467 468 {
468 469 vdev_t *rvd = spa->spa_root_vdev;
469 470 uint64_t txg;
470 471 int c;
471 472
472 473 ASSERT(MUTEX_HELD(&spa_namespace_lock));
473 474
474 475 spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER);
475 476 txg = spa_last_synced_txg(spa) + 1;
476 477 if (what == SPA_CONFIG_UPDATE_POOL) {
477 478 vdev_config_dirty(rvd);
478 479 } else {
479 480 /*
480 481 * If we have top-level vdevs that were added but have
481 482 * not yet been prepared for allocation, do that now.
482 483 * (It's safe now because the config cache is up to date,
483 484 * so it will be able to translate the new DVAs.)
484 485 * See comments in spa_vdev_add() for full details.
485 486 */
486 487 for (c = 0; c < rvd->vdev_children; c++) {
487 488 vdev_t *tvd = rvd->vdev_child[c];
488 489 if (tvd->vdev_ms_array == 0)
489 490 vdev_metaslab_set_size(tvd);
490 491 vdev_expand(tvd, txg);
491 492 }
492 493 }
493 494 spa_config_exit(spa, SCL_ALL, FTAG);
494 495
495 496 /*
496 497 * Wait for the mosconfig to be regenerated and synced.
497 498 */
498 499 txg_wait_synced(spa->spa_dsl_pool, txg);
499 500
500 501 /*
501 502 * Update the global config cache to reflect the new mosconfig.
502 503 */
503 504 if (!spa->spa_is_root)
504 505 spa_config_sync(spa, B_FALSE, what != SPA_CONFIG_UPDATE_POOL);
505 506
506 507 if (what == SPA_CONFIG_UPDATE_POOL)
507 508 spa_config_update(spa, SPA_CONFIG_UPDATE_VDEVS);
508 509 }
|
↓ open down ↓ |
179 lines elided |
↑ open up ↑ |
XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX