Print this page
NEX-5736 implement autoreplace matching based on FRU slot number
NEX-6200 hot spares are not reactivated after reinserting into enclosure
NEX-9403 need to update FRU for spare and l2cache devices
NEX-9404 remove lofi autoreplace support from syseventd
NEX-9409 hotsparing doesn't work for vdevs without FRU
NEX-9424 zfs`vdev_online() needs better notification about state changes
Portions contributed by: Alek Pinchuk <alek@nexenta.com>
Portions contributed by: Josef 'Jeff' Sipek <josef.sipek@nexenta.com>
Reviewed by: Roman Strashkin <roman.strashkin@nexenta.com>
Reviewed by: Steve Peng <steve.peng@nexenta.com>
Reviewed by: Sanjay Nadkarni <sanjay.nadkarni@nexenta.com>
NEX-9228 libzfs`add_config() leaks config nvl when reading spare/l2cache devices
Reviewed by: Hans Rosenfeld <hans.rosenfeld@nexenta.com>
Reviewed by: Roman Strashkin <roman.strashkin@nexenta.com>
Reviewed by: Sanjay Nadkarni <sanjay.nadkarni@nexenta.com>
1778 Assertion failed: rn->rn_nozpool == B_FALSE, file ../common/libzfs_import.c, line 1077, function zpool_open_func
Reviewed by: Matthew Ahrens <mahrens@delphix.com>
Reviewed by: George Wilson <george@delphix.com>
Reviewed by: Richard Elling <richard.elling@richardelling.com>
Approved by: Gordon Ross <gordon.ross@nexenta.com>
NEX-2997 Memory leaks in libzfs import implementation
Reviewed by: Dan Fields <dan.fields@nexenta.com>
NEX-1557 Parallel mount during HA Failover sometimes doesn't share the dataset, causes shares to go offline
| Split |
Close |
| Expand all |
| Collapse all |
--- old/usr/src/lib/libzfs/common/libzfs_import.c
+++ new/usr/src/lib/libzfs/common/libzfs_import.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 (c) 2012, 2016 by Delphix. All rights reserved.
25 25 * Copyright 2015 RackTop Systems.
26 26 * Copyright 2017 Nexenta Systems, Inc.
27 27 */
28 28
29 29 /*
30 30 * Pool import support functions.
31 31 *
32 32 * To import a pool, we rely on reading the configuration information from the
33 33 * ZFS label of each device. If we successfully read the label, then we
34 34 * organize the configuration information in the following hierarchy:
35 35 *
36 36 * pool guid -> toplevel vdev guid -> label txg
37 37 *
38 38 * Duplicate entries matching this same tuple will be discarded. Once we have
39 39 * examined every device, we pick the best label txg config for each toplevel
40 40 * vdev. We then arrange these toplevel vdevs into a complete pool config, and
41 41 * update any paths that have changed. Finally, we attempt to import the pool
42 42 * using our derived config, and record the results.
43 43 */
44 44
45 45 #include <ctype.h>
46 46 #include <devid.h>
47 47 #include <dirent.h>
48 48 #include <errno.h>
49 49 #include <libintl.h>
50 50 #include <stddef.h>
51 51 #include <stdlib.h>
52 52 #include <string.h>
53 53 #include <sys/stat.h>
54 54 #include <unistd.h>
55 55 #include <fcntl.h>
56 56 #include <sys/vtoc.h>
57 57 #include <sys/dktp/fdisk.h>
58 58 #include <sys/efi_partition.h>
59 59 #include <thread_pool.h>
60 60
61 61 #include <sys/vdev_impl.h>
62 62
63 63 #include "libzfs.h"
64 64 #include "libzfs_impl.h"
65 65
66 66 /*
67 67 * Intermediate structures used to gather configuration information.
68 68 */
69 69 typedef struct config_entry {
70 70 uint64_t ce_txg;
71 71 nvlist_t *ce_config;
72 72 struct config_entry *ce_next;
73 73 } config_entry_t;
74 74
75 75 typedef struct vdev_entry {
76 76 uint64_t ve_guid;
77 77 config_entry_t *ve_configs;
78 78 struct vdev_entry *ve_next;
79 79 } vdev_entry_t;
80 80
81 81 typedef struct pool_entry {
82 82 uint64_t pe_guid;
83 83 vdev_entry_t *pe_vdevs;
84 84 struct pool_entry *pe_next;
85 85 } pool_entry_t;
86 86
87 87 typedef struct name_entry {
88 88 char *ne_name;
89 89 uint64_t ne_guid;
90 90 struct name_entry *ne_next;
91 91 } name_entry_t;
92 92
93 93 typedef struct pool_list {
94 94 pool_entry_t *pools;
95 95 name_entry_t *names;
96 96 } pool_list_t;
97 97
98 98 /*
99 99 * Go through and fix up any path and/or devid information for the given vdev
100 100 * configuration.
101 101 */
102 102 static int
103 103 fix_paths(nvlist_t *nv, name_entry_t *names)
104 104 {
105 105 nvlist_t **child;
106 106 uint_t c, children;
107 107 uint64_t guid;
108 108 name_entry_t *ne, *best;
109 109 char *path, *devid;
110 110 int matched;
111 111
112 112 if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN,
113 113 &child, &children) == 0) {
114 114 for (c = 0; c < children; c++)
115 115 if (fix_paths(child[c], names) != 0)
116 116 return (-1);
117 117 return (0);
118 118 }
119 119
120 120 /*
121 121 * This is a leaf (file or disk) vdev. In either case, go through
122 122 * the name list and see if we find a matching guid. If so, replace
123 123 * the path and see if we can calculate a new devid.
124 124 *
125 125 * There may be multiple names associated with a particular guid, in
126 126 * which case we have overlapping slices or multiple paths to the same
127 127 * disk. If this is the case, then we want to pick the path that is
128 128 * the most similar to the original, where "most similar" is the number
129 129 * of matching characters starting from the end of the path. This will
130 130 * preserve slice numbers even if the disks have been reorganized, and
131 131 * will also catch preferred disk names if multiple paths exist.
132 132 */
133 133 verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID, &guid) == 0);
134 134 if (nvlist_lookup_string(nv, ZPOOL_CONFIG_PATH, &path) != 0)
135 135 path = NULL;
136 136
137 137 matched = 0;
138 138 best = NULL;
139 139 for (ne = names; ne != NULL; ne = ne->ne_next) {
140 140 if (ne->ne_guid == guid) {
141 141 const char *src, *dst;
142 142 int count;
143 143
144 144 if (path == NULL) {
145 145 best = ne;
146 146 break;
147 147 }
148 148
149 149 src = ne->ne_name + strlen(ne->ne_name) - 1;
150 150 dst = path + strlen(path) - 1;
151 151 for (count = 0; src >= ne->ne_name && dst >= path;
152 152 src--, dst--, count++)
153 153 if (*src != *dst)
154 154 break;
155 155
156 156 /*
157 157 * At this point, 'count' is the number of characters
158 158 * matched from the end.
159 159 */
160 160 if (count > matched || best == NULL) {
161 161 best = ne;
162 162 matched = count;
163 163 }
164 164 }
165 165 }
166 166
167 167 if (best == NULL)
168 168 return (0);
169 169
170 170 if (nvlist_add_string(nv, ZPOOL_CONFIG_PATH, best->ne_name) != 0)
171 171 return (-1);
172 172
173 173 if ((devid = devid_str_from_path(best->ne_name)) == NULL) {
174 174 (void) nvlist_remove_all(nv, ZPOOL_CONFIG_DEVID);
175 175 } else {
176 176 if (nvlist_add_string(nv, ZPOOL_CONFIG_DEVID, devid) != 0) {
177 177 devid_str_free(devid);
178 178 return (-1);
179 179 }
180 180 devid_str_free(devid);
181 181 }
182 182
183 183 return (0);
184 184 }
185 185
186 186 /*
187 187 * Add the given configuration to the list of known devices.
188 188 */
189 189 static int
190 190 add_config(libzfs_handle_t *hdl, pool_list_t *pl, const char *path,
191 191 nvlist_t *config)
192 192 {
193 193 uint64_t pool_guid, vdev_guid, top_guid, txg, state;
194 194 pool_entry_t *pe;
195 195 vdev_entry_t *ve;
196 196 config_entry_t *ce;
197 197 name_entry_t *ne;
198 198
199 199 /*
200 200 * If this is a hot spare not currently in use or level 2 cache
201 201 * device, add it to the list of names to translate, but don't do
202 202 * anything else.
203 203 */
204 204 if (nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_STATE,
205 205 &state) == 0 &&
206 206 (state == POOL_STATE_SPARE || state == POOL_STATE_L2CACHE) &&
207 207 nvlist_lookup_uint64(config, ZPOOL_CONFIG_GUID, &vdev_guid) == 0) {
208 208 if ((ne = zfs_alloc(hdl, sizeof (name_entry_t))) == NULL)
209 209 return (-1);
210 210
211 211 if ((ne->ne_name = zfs_strdup(hdl, path)) == NULL) {
212 212 free(ne);
213 213 return (-1);
214 214 }
215 215
216 216 ne->ne_guid = vdev_guid;
217 217 ne->ne_next = pl->names;
218 218 pl->names = ne;
219 219
220 220 nvlist_free(config);
221 221 return (0);
222 222 }
223 223
224 224 /*
225 225 * If we have a valid config but cannot read any of these fields, then
226 226 * it means we have a half-initialized label. In vdev_label_init()
227 227 * we write a label with txg == 0 so that we can identify the device
228 228 * in case the user refers to the same disk later on. If we fail to
229 229 * create the pool, we'll be left with a label in this state
230 230 * which should not be considered part of a valid pool.
231 231 */
232 232 if (nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID,
233 233 &pool_guid) != 0 ||
234 234 nvlist_lookup_uint64(config, ZPOOL_CONFIG_GUID,
235 235 &vdev_guid) != 0 ||
236 236 nvlist_lookup_uint64(config, ZPOOL_CONFIG_TOP_GUID,
237 237 &top_guid) != 0 ||
238 238 nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_TXG,
239 239 &txg) != 0 || txg == 0) {
240 240 nvlist_free(config);
241 241 return (0);
242 242 }
243 243
244 244 /*
245 245 * First, see if we know about this pool. If not, then add it to the
246 246 * list of known pools.
247 247 */
248 248 for (pe = pl->pools; pe != NULL; pe = pe->pe_next) {
249 249 if (pe->pe_guid == pool_guid)
250 250 break;
251 251 }
252 252
253 253 if (pe == NULL) {
254 254 if ((pe = zfs_alloc(hdl, sizeof (pool_entry_t))) == NULL) {
255 255 nvlist_free(config);
256 256 return (-1);
257 257 }
258 258 pe->pe_guid = pool_guid;
259 259 pe->pe_next = pl->pools;
260 260 pl->pools = pe;
261 261 }
262 262
263 263 /*
264 264 * Second, see if we know about this toplevel vdev. Add it if its
265 265 * missing.
266 266 */
267 267 for (ve = pe->pe_vdevs; ve != NULL; ve = ve->ve_next) {
268 268 if (ve->ve_guid == top_guid)
269 269 break;
270 270 }
271 271
272 272 if (ve == NULL) {
273 273 if ((ve = zfs_alloc(hdl, sizeof (vdev_entry_t))) == NULL) {
274 274 nvlist_free(config);
275 275 return (-1);
276 276 }
277 277 ve->ve_guid = top_guid;
278 278 ve->ve_next = pe->pe_vdevs;
279 279 pe->pe_vdevs = ve;
280 280 }
281 281
282 282 /*
283 283 * Third, see if we have a config with a matching transaction group. If
284 284 * so, then we do nothing. Otherwise, add it to the list of known
285 285 * configs.
286 286 */
287 287 for (ce = ve->ve_configs; ce != NULL; ce = ce->ce_next) {
288 288 if (ce->ce_txg == txg)
289 289 break;
290 290 }
291 291
292 292 if (ce == NULL) {
293 293 if ((ce = zfs_alloc(hdl, sizeof (config_entry_t))) == NULL) {
294 294 nvlist_free(config);
295 295 return (-1);
296 296 }
297 297 ce->ce_txg = txg;
298 298 ce->ce_config = config;
299 299 ce->ce_next = ve->ve_configs;
300 300 ve->ve_configs = ce;
301 301 } else {
302 302 nvlist_free(config);
303 303 }
304 304
305 305 /*
306 306 * At this point we've successfully added our config to the list of
307 307 * known configs. The last thing to do is add the vdev guid -> path
308 308 * mappings so that we can fix up the configuration as necessary before
309 309 * doing the import.
310 310 */
311 311 if ((ne = zfs_alloc(hdl, sizeof (name_entry_t))) == NULL)
312 312 return (-1);
313 313
314 314 if ((ne->ne_name = zfs_strdup(hdl, path)) == NULL) {
315 315 free(ne);
316 316 return (-1);
317 317 }
318 318
319 319 ne->ne_guid = vdev_guid;
320 320 ne->ne_next = pl->names;
321 321 pl->names = ne;
322 322
323 323 return (0);
324 324 }
325 325
326 326 /*
327 327 * Returns true if the named pool matches the given GUID.
328 328 */
329 329 static int
330 330 pool_active(libzfs_handle_t *hdl, const char *name, uint64_t guid,
331 331 boolean_t *isactive)
332 332 {
333 333 zpool_handle_t *zhp;
334 334 uint64_t theguid;
335 335
336 336 if (zpool_open_silent(hdl, name, &zhp) != 0)
337 337 return (-1);
338 338
339 339 if (zhp == NULL) {
340 340 *isactive = B_FALSE;
341 341 return (0);
342 342 }
343 343
344 344 verify(nvlist_lookup_uint64(zhp->zpool_config, ZPOOL_CONFIG_POOL_GUID,
345 345 &theguid) == 0);
346 346
347 347 zpool_close(zhp);
348 348
349 349 *isactive = (theguid == guid);
350 350 return (0);
351 351 }
352 352
353 353 static nvlist_t *
354 354 refresh_config(libzfs_handle_t *hdl, nvlist_t *config)
355 355 {
356 356 nvlist_t *nvl;
357 357 zfs_cmd_t zc = { 0 };
358 358 int err, dstbuf_size;
359 359
360 360 if (zcmd_write_conf_nvlist(hdl, &zc, config) != 0)
361 361 return (NULL);
362 362
363 363 dstbuf_size = MAX(CONFIG_BUF_MINSIZE, zc.zc_nvlist_conf_size * 4);
364 364
365 365 if (zcmd_alloc_dst_nvlist(hdl, &zc, dstbuf_size) != 0) {
366 366 zcmd_free_nvlists(&zc);
367 367 return (NULL);
368 368 }
369 369
370 370 while ((err = ioctl(hdl->libzfs_fd, ZFS_IOC_POOL_TRYIMPORT,
371 371 &zc)) != 0 && errno == ENOMEM) {
372 372 if (zcmd_expand_dst_nvlist(hdl, &zc) != 0) {
373 373 zcmd_free_nvlists(&zc);
374 374 return (NULL);
375 375 }
376 376 }
377 377
378 378 if (err) {
379 379 zcmd_free_nvlists(&zc);
380 380 return (NULL);
381 381 }
382 382
383 383 if (zcmd_read_dst_nvlist(hdl, &zc, &nvl) != 0) {
384 384 zcmd_free_nvlists(&zc);
385 385 return (NULL);
386 386 }
387 387
388 388 zcmd_free_nvlists(&zc);
389 389 return (nvl);
390 390 }
391 391
392 392 /*
393 393 * Determine if the vdev id is a hole in the namespace.
394 394 */
395 395 boolean_t
396 396 vdev_is_hole(uint64_t *hole_array, uint_t holes, uint_t id)
397 397 {
398 398 for (int c = 0; c < holes; c++) {
399 399
400 400 /* Top-level is a hole */
401 401 if (hole_array[c] == id)
402 402 return (B_TRUE);
403 403 }
404 404 return (B_FALSE);
|
↓ open down ↓ |
404 lines elided |
↑ open up ↑ |
405 405 }
406 406
407 407 /*
408 408 * Convert our list of pools into the definitive set of configurations. We
409 409 * start by picking the best config for each toplevel vdev. Once that's done,
410 410 * we assemble the toplevel vdevs into a full config for the pool. We make a
411 411 * pass to fix up any incorrect paths, and then add it to the main list to
412 412 * return to the user.
413 413 */
414 414 static nvlist_t *
415 -get_configs(libzfs_handle_t *hdl, pool_list_t *pl, boolean_t active_ok,
416 - nvlist_t *policy)
415 +get_configs(libzfs_handle_t *hdl, pool_list_t *pl, boolean_t active_ok)
417 416 {
418 417 pool_entry_t *pe;
419 418 vdev_entry_t *ve;
420 419 config_entry_t *ce;
421 420 nvlist_t *ret = NULL, *config = NULL, *tmp = NULL, *nvtop, *nvroot;
422 421 nvlist_t **spares, **l2cache;
423 422 uint_t i, nspares, nl2cache;
424 423 boolean_t config_seen;
425 424 uint64_t best_txg;
426 425 char *name, *hostname = NULL;
427 426 uint64_t guid;
428 427 uint_t children = 0;
429 428 nvlist_t **child = NULL;
430 429 uint_t holes;
431 430 uint64_t *hole_array, max_id;
432 431 uint_t c;
433 432 boolean_t isactive;
434 433 uint64_t hostid;
435 434 nvlist_t *nvl;
436 435 boolean_t found_one = B_FALSE;
437 436 boolean_t valid_top_config = B_FALSE;
438 437
439 438 if (nvlist_alloc(&ret, 0, 0) != 0)
440 439 goto nomem;
441 440
442 441 for (pe = pl->pools; pe != NULL; pe = pe->pe_next) {
443 442 uint64_t id, max_txg = 0;
444 443
445 444 if (nvlist_alloc(&config, NV_UNIQUE_NAME, 0) != 0)
446 445 goto nomem;
447 446 config_seen = B_FALSE;
448 447
449 448 /*
450 449 * Iterate over all toplevel vdevs. Grab the pool configuration
451 450 * from the first one we find, and then go through the rest and
452 451 * add them as necessary to the 'vdevs' member of the config.
453 452 */
454 453 for (ve = pe->pe_vdevs; ve != NULL; ve = ve->ve_next) {
455 454
456 455 /*
457 456 * Determine the best configuration for this vdev by
458 457 * selecting the config with the latest transaction
459 458 * group.
460 459 */
461 460 best_txg = 0;
462 461 for (ce = ve->ve_configs; ce != NULL;
463 462 ce = ce->ce_next) {
464 463
465 464 if (ce->ce_txg > best_txg) {
466 465 tmp = ce->ce_config;
467 466 best_txg = ce->ce_txg;
468 467 }
469 468 }
470 469
471 470 /*
472 471 * We rely on the fact that the max txg for the
473 472 * pool will contain the most up-to-date information
474 473 * about the valid top-levels in the vdev namespace.
475 474 */
476 475 if (best_txg > max_txg) {
477 476 (void) nvlist_remove(config,
478 477 ZPOOL_CONFIG_VDEV_CHILDREN,
479 478 DATA_TYPE_UINT64);
480 479 (void) nvlist_remove(config,
481 480 ZPOOL_CONFIG_HOLE_ARRAY,
482 481 DATA_TYPE_UINT64_ARRAY);
483 482
484 483 max_txg = best_txg;
485 484 hole_array = NULL;
486 485 holes = 0;
487 486 max_id = 0;
488 487 valid_top_config = B_FALSE;
489 488
490 489 if (nvlist_lookup_uint64(tmp,
491 490 ZPOOL_CONFIG_VDEV_CHILDREN, &max_id) == 0) {
492 491 verify(nvlist_add_uint64(config,
493 492 ZPOOL_CONFIG_VDEV_CHILDREN,
494 493 max_id) == 0);
495 494 valid_top_config = B_TRUE;
496 495 }
497 496
498 497 if (nvlist_lookup_uint64_array(tmp,
499 498 ZPOOL_CONFIG_HOLE_ARRAY, &hole_array,
500 499 &holes) == 0) {
501 500 verify(nvlist_add_uint64_array(config,
502 501 ZPOOL_CONFIG_HOLE_ARRAY,
503 502 hole_array, holes) == 0);
504 503 }
505 504 }
506 505
507 506 if (!config_seen) {
508 507 /*
509 508 * Copy the relevant pieces of data to the pool
510 509 * configuration:
511 510 *
512 511 * version
513 512 * pool guid
514 513 * name
515 514 * comment (if available)
516 515 * pool state
517 516 * hostid (if available)
518 517 * hostname (if available)
519 518 */
520 519 uint64_t state, version;
521 520 char *comment = NULL;
522 521
523 522 version = fnvlist_lookup_uint64(tmp,
524 523 ZPOOL_CONFIG_VERSION);
525 524 fnvlist_add_uint64(config,
526 525 ZPOOL_CONFIG_VERSION, version);
527 526 guid = fnvlist_lookup_uint64(tmp,
528 527 ZPOOL_CONFIG_POOL_GUID);
529 528 fnvlist_add_uint64(config,
530 529 ZPOOL_CONFIG_POOL_GUID, guid);
531 530 name = fnvlist_lookup_string(tmp,
532 531 ZPOOL_CONFIG_POOL_NAME);
533 532 fnvlist_add_string(config,
534 533 ZPOOL_CONFIG_POOL_NAME, name);
535 534
536 535 if (nvlist_lookup_string(tmp,
537 536 ZPOOL_CONFIG_COMMENT, &comment) == 0)
538 537 fnvlist_add_string(config,
539 538 ZPOOL_CONFIG_COMMENT, comment);
540 539
541 540 state = fnvlist_lookup_uint64(tmp,
542 541 ZPOOL_CONFIG_POOL_STATE);
543 542 fnvlist_add_uint64(config,
544 543 ZPOOL_CONFIG_POOL_STATE, state);
545 544
546 545 hostid = 0;
547 546 if (nvlist_lookup_uint64(tmp,
548 547 ZPOOL_CONFIG_HOSTID, &hostid) == 0) {
549 548 fnvlist_add_uint64(config,
550 549 ZPOOL_CONFIG_HOSTID, hostid);
551 550 hostname = fnvlist_lookup_string(tmp,
552 551 ZPOOL_CONFIG_HOSTNAME);
553 552 fnvlist_add_string(config,
554 553 ZPOOL_CONFIG_HOSTNAME, hostname);
555 554 }
556 555
557 556 config_seen = B_TRUE;
558 557 }
559 558
560 559 /*
561 560 * Add this top-level vdev to the child array.
562 561 */
563 562 verify(nvlist_lookup_nvlist(tmp,
564 563 ZPOOL_CONFIG_VDEV_TREE, &nvtop) == 0);
565 564 verify(nvlist_lookup_uint64(nvtop, ZPOOL_CONFIG_ID,
566 565 &id) == 0);
567 566
568 567 if (id >= children) {
569 568 nvlist_t **newchild;
570 569
571 570 newchild = zfs_alloc(hdl, (id + 1) *
572 571 sizeof (nvlist_t *));
573 572 if (newchild == NULL)
574 573 goto nomem;
575 574
576 575 for (c = 0; c < children; c++)
577 576 newchild[c] = child[c];
578 577
579 578 free(child);
580 579 child = newchild;
581 580 children = id + 1;
582 581 }
583 582 if (nvlist_dup(nvtop, &child[id], 0) != 0)
584 583 goto nomem;
585 584
586 585 }
587 586
588 587 /*
589 588 * If we have information about all the top-levels then
590 589 * clean up the nvlist which we've constructed. This
591 590 * means removing any extraneous devices that are
592 591 * beyond the valid range or adding devices to the end
593 592 * of our array which appear to be missing.
594 593 */
595 594 if (valid_top_config) {
596 595 if (max_id < children) {
597 596 for (c = max_id; c < children; c++)
598 597 nvlist_free(child[c]);
599 598 children = max_id;
600 599 } else if (max_id > children) {
601 600 nvlist_t **newchild;
602 601
603 602 newchild = zfs_alloc(hdl, (max_id) *
604 603 sizeof (nvlist_t *));
605 604 if (newchild == NULL)
606 605 goto nomem;
607 606
608 607 for (c = 0; c < children; c++)
609 608 newchild[c] = child[c];
610 609
611 610 free(child);
612 611 child = newchild;
613 612 children = max_id;
614 613 }
615 614 }
616 615
617 616 verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID,
618 617 &guid) == 0);
619 618
620 619 /*
621 620 * The vdev namespace may contain holes as a result of
622 621 * device removal. We must add them back into the vdev
623 622 * tree before we process any missing devices.
624 623 */
625 624 if (holes > 0) {
626 625 ASSERT(valid_top_config);
627 626
628 627 for (c = 0; c < children; c++) {
629 628 nvlist_t *holey;
630 629
631 630 if (child[c] != NULL ||
632 631 !vdev_is_hole(hole_array, holes, c))
633 632 continue;
634 633
635 634 if (nvlist_alloc(&holey, NV_UNIQUE_NAME,
636 635 0) != 0)
637 636 goto nomem;
638 637
639 638 /*
640 639 * Holes in the namespace are treated as
641 640 * "hole" top-level vdevs and have a
642 641 * special flag set on them.
643 642 */
644 643 if (nvlist_add_string(holey,
645 644 ZPOOL_CONFIG_TYPE,
646 645 VDEV_TYPE_HOLE) != 0 ||
647 646 nvlist_add_uint64(holey,
648 647 ZPOOL_CONFIG_ID, c) != 0 ||
649 648 nvlist_add_uint64(holey,
650 649 ZPOOL_CONFIG_GUID, 0ULL) != 0) {
651 650 nvlist_free(holey);
652 651 goto nomem;
653 652 }
654 653 child[c] = holey;
655 654 }
656 655 }
657 656
658 657 /*
659 658 * Look for any missing top-level vdevs. If this is the case,
660 659 * create a faked up 'missing' vdev as a placeholder. We cannot
661 660 * simply compress the child array, because the kernel performs
662 661 * certain checks to make sure the vdev IDs match their location
663 662 * in the configuration.
664 663 */
665 664 for (c = 0; c < children; c++) {
666 665 if (child[c] == NULL) {
667 666 nvlist_t *missing;
668 667 if (nvlist_alloc(&missing, NV_UNIQUE_NAME,
669 668 0) != 0)
670 669 goto nomem;
671 670 if (nvlist_add_string(missing,
672 671 ZPOOL_CONFIG_TYPE,
673 672 VDEV_TYPE_MISSING) != 0 ||
674 673 nvlist_add_uint64(missing,
675 674 ZPOOL_CONFIG_ID, c) != 0 ||
676 675 nvlist_add_uint64(missing,
677 676 ZPOOL_CONFIG_GUID, 0ULL) != 0) {
678 677 nvlist_free(missing);
679 678 goto nomem;
680 679 }
681 680 child[c] = missing;
682 681 }
683 682 }
684 683
685 684 /*
686 685 * Put all of this pool's top-level vdevs into a root vdev.
687 686 */
688 687 if (nvlist_alloc(&nvroot, NV_UNIQUE_NAME, 0) != 0)
689 688 goto nomem;
690 689 if (nvlist_add_string(nvroot, ZPOOL_CONFIG_TYPE,
691 690 VDEV_TYPE_ROOT) != 0 ||
692 691 nvlist_add_uint64(nvroot, ZPOOL_CONFIG_ID, 0ULL) != 0 ||
693 692 nvlist_add_uint64(nvroot, ZPOOL_CONFIG_GUID, guid) != 0 ||
694 693 nvlist_add_nvlist_array(nvroot, ZPOOL_CONFIG_CHILDREN,
695 694 child, children) != 0) {
696 695 nvlist_free(nvroot);
697 696 goto nomem;
698 697 }
699 698
700 699 for (c = 0; c < children; c++)
701 700 nvlist_free(child[c]);
702 701 free(child);
703 702 children = 0;
704 703 child = NULL;
705 704
706 705 /*
707 706 * Go through and fix up any paths and/or devids based on our
708 707 * known list of vdev GUID -> path mappings.
709 708 */
710 709 if (fix_paths(nvroot, pl->names) != 0) {
711 710 nvlist_free(nvroot);
712 711 goto nomem;
713 712 }
714 713
715 714 /*
716 715 * Add the root vdev to this pool's configuration.
717 716 */
718 717 if (nvlist_add_nvlist(config, ZPOOL_CONFIG_VDEV_TREE,
719 718 nvroot) != 0) {
720 719 nvlist_free(nvroot);
721 720 goto nomem;
722 721 }
723 722 nvlist_free(nvroot);
724 723
725 724 /*
726 725 * zdb uses this path to report on active pools that were
727 726 * imported or created using -R.
728 727 */
729 728 if (active_ok)
730 729 goto add_pool;
731 730
732 731 /*
733 732 * Determine if this pool is currently active, in which case we
734 733 * can't actually import it.
735 734 */
736 735 verify(nvlist_lookup_string(config, ZPOOL_CONFIG_POOL_NAME,
737 736 &name) == 0);
738 737 verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID,
739 738 &guid) == 0);
|
↓ open down ↓ |
313 lines elided |
↑ open up ↑ |
740 739
741 740 if (pool_active(hdl, name, guid, &isactive) != 0)
742 741 goto error;
743 742
744 743 if (isactive) {
745 744 nvlist_free(config);
746 745 config = NULL;
747 746 continue;
748 747 }
749 748
750 - if (policy != NULL) {
751 - if (nvlist_add_nvlist(config, ZPOOL_REWIND_POLICY,
752 - policy) != 0)
753 - goto nomem;
754 - }
755 -
756 749 if ((nvl = refresh_config(hdl, config)) == NULL) {
757 750 nvlist_free(config);
758 751 config = NULL;
759 752 continue;
760 753 }
761 754
762 755 nvlist_free(config);
763 756 config = nvl;
764 757
765 758 /*
766 759 * Go through and update the paths for spares, now that we have
767 760 * them.
768 761 */
769 762 verify(nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE,
770 763 &nvroot) == 0);
771 764 if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_SPARES,
772 765 &spares, &nspares) == 0) {
773 766 for (i = 0; i < nspares; i++) {
774 767 if (fix_paths(spares[i], pl->names) != 0)
775 768 goto nomem;
776 769 }
777 770 }
778 771
779 772 /*
780 773 * Update the paths for l2cache devices.
781 774 */
782 775 if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_L2CACHE,
783 776 &l2cache, &nl2cache) == 0) {
784 777 for (i = 0; i < nl2cache; i++) {
785 778 if (fix_paths(l2cache[i], pl->names) != 0)
786 779 goto nomem;
787 780 }
788 781 }
789 782
790 783 /*
791 784 * Restore the original information read from the actual label.
792 785 */
793 786 (void) nvlist_remove(config, ZPOOL_CONFIG_HOSTID,
794 787 DATA_TYPE_UINT64);
795 788 (void) nvlist_remove(config, ZPOOL_CONFIG_HOSTNAME,
796 789 DATA_TYPE_STRING);
797 790 if (hostid != 0) {
798 791 verify(nvlist_add_uint64(config, ZPOOL_CONFIG_HOSTID,
799 792 hostid) == 0);
800 793 verify(nvlist_add_string(config, ZPOOL_CONFIG_HOSTNAME,
801 794 hostname) == 0);
802 795 }
803 796
804 797 add_pool:
805 798 /*
806 799 * Add this pool to the list of configs.
807 800 */
808 801 verify(nvlist_lookup_string(config, ZPOOL_CONFIG_POOL_NAME,
809 802 &name) == 0);
810 803 if (nvlist_add_nvlist(ret, name, config) != 0)
811 804 goto nomem;
812 805
813 806 found_one = B_TRUE;
814 807 nvlist_free(config);
815 808 config = NULL;
816 809 }
817 810
818 811 if (!found_one) {
819 812 nvlist_free(ret);
820 813 ret = NULL;
821 814 }
822 815
823 816 return (ret);
824 817
825 818 nomem:
826 819 (void) no_memory(hdl);
827 820 error:
828 821 nvlist_free(config);
829 822 nvlist_free(ret);
830 823 for (c = 0; c < children; c++)
831 824 nvlist_free(child[c]);
832 825 free(child);
833 826
834 827 return (NULL);
835 828 }
836 829
837 830 /*
838 831 * Return the offset of the given label.
839 832 */
840 833 static uint64_t
841 834 label_offset(uint64_t size, int l)
842 835 {
843 836 ASSERT(P2PHASE_TYPED(size, sizeof (vdev_label_t), uint64_t) == 0);
844 837 return (l * sizeof (vdev_label_t) + (l < VDEV_LABELS / 2 ?
845 838 0 : size - VDEV_LABELS * sizeof (vdev_label_t)));
846 839 }
847 840
848 841 /*
849 842 * Given a file descriptor, read the label information and return an nvlist
850 843 * describing the configuration, if there is one.
851 844 * Return 0 on success, or -1 on failure
852 845 */
853 846 int
854 847 zpool_read_label(int fd, nvlist_t **config)
855 848 {
856 849 struct stat64 statbuf;
|
↓ open down ↓ |
91 lines elided |
↑ open up ↑ |
857 850 int l;
858 851 vdev_label_t *label;
859 852 uint64_t state, txg, size;
860 853
861 854 *config = NULL;
862 855
863 856 if (fstat64(fd, &statbuf) == -1)
864 857 return (-1);
865 858 size = P2ALIGN_TYPED(statbuf.st_size, sizeof (vdev_label_t), uint64_t);
866 859
867 - if ((label = malloc(sizeof (vdev_label_t))) == NULL)
860 + if ((label = calloc(sizeof (vdev_label_t), 1)) == NULL)
868 861 return (-1);
869 862
870 863 for (l = 0; l < VDEV_LABELS; l++) {
871 864 if (pread64(fd, label, sizeof (vdev_label_t),
872 865 label_offset(size, l)) != sizeof (vdev_label_t))
873 866 continue;
874 867
875 868 if (nvlist_unpack(label->vl_vdev_phys.vp_nvlist,
876 869 sizeof (label->vl_vdev_phys.vp_nvlist), config, 0) != 0)
877 870 continue;
878 871
879 872 if (nvlist_lookup_uint64(*config, ZPOOL_CONFIG_POOL_STATE,
880 873 &state) != 0 || state > POOL_STATE_L2CACHE) {
881 874 nvlist_free(*config);
882 875 continue;
883 876 }
884 877
885 878 if (state != POOL_STATE_SPARE && state != POOL_STATE_L2CACHE &&
886 879 (nvlist_lookup_uint64(*config, ZPOOL_CONFIG_POOL_TXG,
887 880 &txg) != 0 || txg == 0)) {
888 881 nvlist_free(*config);
889 882 continue;
890 883 }
891 884
892 885 free(label);
893 886 return (0);
894 887 }
895 888
896 889 free(label);
897 890 *config = NULL;
898 891 return (-1);
899 892 }
900 893
901 894 typedef struct rdsk_node {
902 895 char *rn_name;
903 896 int rn_dfd;
904 897 libzfs_handle_t *rn_hdl;
905 898 nvlist_t *rn_config;
906 899 avl_tree_t *rn_avl;
907 900 avl_node_t rn_node;
908 901 boolean_t rn_nozpool;
909 902 } rdsk_node_t;
910 903
911 904 static int
912 905 slice_cache_compare(const void *arg1, const void *arg2)
913 906 {
914 907 const char *nm1 = ((rdsk_node_t *)arg1)->rn_name;
915 908 const char *nm2 = ((rdsk_node_t *)arg2)->rn_name;
916 909 char *nm1slice, *nm2slice;
917 910 int rv;
918 911
919 912 /*
920 913 * slices zero and two are the most likely to provide results,
921 914 * so put those first
922 915 */
923 916 nm1slice = strstr(nm1, "s0");
924 917 nm2slice = strstr(nm2, "s0");
925 918 if (nm1slice && !nm2slice) {
926 919 return (-1);
927 920 }
928 921 if (!nm1slice && nm2slice) {
929 922 return (1);
930 923 }
931 924 nm1slice = strstr(nm1, "s2");
932 925 nm2slice = strstr(nm2, "s2");
933 926 if (nm1slice && !nm2slice) {
934 927 return (-1);
935 928 }
936 929 if (!nm1slice && nm2slice) {
937 930 return (1);
938 931 }
939 932
940 933 rv = strcmp(nm1, nm2);
941 934 if (rv == 0)
942 935 return (0);
943 936 return (rv > 0 ? 1 : -1);
944 937 }
945 938
946 939 static void
947 940 check_one_slice(avl_tree_t *r, char *diskname, uint_t partno,
948 941 diskaddr_t size, uint_t blksz)
949 942 {
950 943 rdsk_node_t tmpnode;
951 944 rdsk_node_t *node;
952 945 char sname[MAXNAMELEN];
953 946
954 947 tmpnode.rn_name = &sname[0];
955 948 (void) snprintf(tmpnode.rn_name, MAXNAMELEN, "%s%u",
956 949 diskname, partno);
957 950 /*
958 951 * protect against division by zero for disk labels that
959 952 * contain a bogus sector size
960 953 */
961 954 if (blksz == 0)
962 955 blksz = DEV_BSIZE;
963 956 /* too small to contain a zpool? */
964 957 if ((size < (SPA_MINDEVSIZE / blksz)) &&
965 958 (node = avl_find(r, &tmpnode, NULL)))
966 959 node->rn_nozpool = B_TRUE;
967 960 }
968 961
969 962 static void
970 963 nozpool_all_slices(avl_tree_t *r, const char *sname)
971 964 {
972 965 char diskname[MAXNAMELEN];
973 966 char *ptr;
974 967 int i;
975 968
976 969 (void) strncpy(diskname, sname, MAXNAMELEN);
977 970 if (((ptr = strrchr(diskname, 's')) == NULL) &&
978 971 ((ptr = strrchr(diskname, 'p')) == NULL))
979 972 return;
980 973 ptr[0] = 's';
981 974 ptr[1] = '\0';
982 975 for (i = 0; i < NDKMAP; i++)
983 976 check_one_slice(r, diskname, i, 0, 1);
984 977 ptr[0] = 'p';
985 978 for (i = 0; i <= FD_NUMPART; i++)
986 979 check_one_slice(r, diskname, i, 0, 1);
987 980 }
988 981
989 982 static void
990 983 check_slices(avl_tree_t *r, int fd, const char *sname)
991 984 {
992 985 struct extvtoc vtoc;
993 986 struct dk_gpt *gpt;
994 987 char diskname[MAXNAMELEN];
995 988 char *ptr;
996 989 int i;
997 990
998 991 (void) strncpy(diskname, sname, MAXNAMELEN);
999 992 if ((ptr = strrchr(diskname, 's')) == NULL || !isdigit(ptr[1]))
1000 993 return;
1001 994 ptr[1] = '\0';
1002 995
1003 996 if (read_extvtoc(fd, &vtoc) >= 0) {
1004 997 for (i = 0; i < NDKMAP; i++)
1005 998 check_one_slice(r, diskname, i,
1006 999 vtoc.v_part[i].p_size, vtoc.v_sectorsz);
1007 1000 } else if (efi_alloc_and_read(fd, &gpt) >= 0) {
1008 1001 /*
1009 1002 * on x86 we'll still have leftover links that point
1010 1003 * to slices s[9-15], so use NDKMAP instead
1011 1004 */
1012 1005 for (i = 0; i < NDKMAP; i++)
1013 1006 check_one_slice(r, diskname, i,
1014 1007 gpt->efi_parts[i].p_size, gpt->efi_lbasize);
1015 1008 /* nodes p[1-4] are never used with EFI labels */
1016 1009 ptr[0] = 'p';
1017 1010 for (i = 1; i <= FD_NUMPART; i++)
1018 1011 check_one_slice(r, diskname, i, 0, 1);
1019 1012 efi_free(gpt);
1020 1013 }
1021 1014 }
1022 1015
1023 1016 static void
1024 1017 zpool_open_func(void *arg)
1025 1018 {
1026 1019 rdsk_node_t *rn = arg;
1027 1020 struct stat64 statbuf;
1028 1021 nvlist_t *config;
1029 1022 int fd;
1030 1023
1031 1024 if (rn->rn_nozpool)
1032 1025 return;
1033 1026 if ((fd = openat64(rn->rn_dfd, rn->rn_name, O_RDONLY)) < 0) {
1034 1027 /* symlink to a device that's no longer there */
1035 1028 if (errno == ENOENT)
1036 1029 nozpool_all_slices(rn->rn_avl, rn->rn_name);
1037 1030 return;
1038 1031 }
1039 1032 /*
1040 1033 * Ignore failed stats. We only want regular
1041 1034 * files, character devs and block devs.
1042 1035 */
1043 1036 if (fstat64(fd, &statbuf) != 0 ||
1044 1037 (!S_ISREG(statbuf.st_mode) &&
1045 1038 !S_ISCHR(statbuf.st_mode) &&
1046 1039 !S_ISBLK(statbuf.st_mode))) {
1047 1040 (void) close(fd);
1048 1041 return;
1049 1042 }
1050 1043 /* this file is too small to hold a zpool */
1051 1044 if (S_ISREG(statbuf.st_mode) &&
1052 1045 statbuf.st_size < SPA_MINDEVSIZE) {
1053 1046 (void) close(fd);
1054 1047 return;
1055 1048 } else if (!S_ISREG(statbuf.st_mode)) {
1056 1049 /*
1057 1050 * Try to read the disk label first so we don't have to
1058 1051 * open a bunch of minor nodes that can't have a zpool.
1059 1052 */
1060 1053 check_slices(rn->rn_avl, fd, rn->rn_name);
1061 1054 }
1062 1055
1063 1056 if ((zpool_read_label(fd, &config)) != 0 && errno == ENOMEM) {
1064 1057 (void) close(fd);
1065 1058 (void) no_memory(rn->rn_hdl);
1066 1059 return;
1067 1060 }
1068 1061 (void) close(fd);
1069 1062
1070 1063 rn->rn_config = config;
1071 1064 }
1072 1065
1073 1066 /*
1074 1067 * Given a file descriptor, clear (zero) the label information.
1075 1068 */
1076 1069 int
1077 1070 zpool_clear_label(int fd)
1078 1071 {
1079 1072 struct stat64 statbuf;
1080 1073 int l;
1081 1074 vdev_label_t *label;
1082 1075 uint64_t size;
1083 1076
1084 1077 if (fstat64(fd, &statbuf) == -1)
1085 1078 return (0);
1086 1079 size = P2ALIGN_TYPED(statbuf.st_size, sizeof (vdev_label_t), uint64_t);
1087 1080
1088 1081 if ((label = calloc(sizeof (vdev_label_t), 1)) == NULL)
1089 1082 return (-1);
1090 1083
1091 1084 for (l = 0; l < VDEV_LABELS; l++) {
1092 1085 if (pwrite64(fd, label, sizeof (vdev_label_t),
1093 1086 label_offset(size, l)) != sizeof (vdev_label_t)) {
1094 1087 free(label);
1095 1088 return (-1);
1096 1089 }
1097 1090 }
1098 1091
1099 1092 free(label);
1100 1093 return (0);
1101 1094 }
1102 1095
1103 1096 /*
1104 1097 * Given a list of directories to search, find all pools stored on disk. This
1105 1098 * includes partial pools which are not available to import. If no args are
1106 1099 * given (argc is 0), then the default directory (/dev/dsk) is searched.
1107 1100 * poolname or guid (but not both) are provided by the caller when trying
1108 1101 * to import a specific pool.
1109 1102 */
1110 1103 static nvlist_t *
1111 1104 zpool_find_import_impl(libzfs_handle_t *hdl, importargs_t *iarg)
1112 1105 {
1113 1106 int i, dirs = iarg->paths;
1114 1107 struct dirent64 *dp;
1115 1108 char path[MAXPATHLEN];
1116 1109 char *end, **dir = iarg->path;
1117 1110 size_t pathleft;
1118 1111 nvlist_t *ret = NULL;
1119 1112 static char *default_dir = ZFS_DISK_ROOT;
1120 1113 pool_list_t pools = { 0 };
1121 1114 pool_entry_t *pe, *penext;
1122 1115 vdev_entry_t *ve, *venext;
1123 1116 config_entry_t *ce, *cenext;
1124 1117 name_entry_t *ne, *nenext;
1125 1118 avl_tree_t slice_cache;
1126 1119 rdsk_node_t *slice;
1127 1120 void *cookie;
1128 1121
1129 1122 if (dirs == 0) {
1130 1123 dirs = 1;
1131 1124 dir = &default_dir;
1132 1125 }
1133 1126
1134 1127 /*
1135 1128 * Go through and read the label configuration information from every
1136 1129 * possible device, organizing the information according to pool GUID
1137 1130 * and toplevel GUID.
1138 1131 */
1139 1132 for (i = 0; i < dirs; i++) {
1140 1133 tpool_t *t;
1141 1134 char rdsk[MAXPATHLEN];
1142 1135 int dfd;
1143 1136 boolean_t config_failed = B_FALSE;
1144 1137 DIR *dirp;
1145 1138
1146 1139 /* use realpath to normalize the path */
1147 1140 if (realpath(dir[i], path) == 0) {
1148 1141 (void) zfs_error_fmt(hdl, EZFS_BADPATH,
1149 1142 dgettext(TEXT_DOMAIN, "cannot open '%s'"), dir[i]);
1150 1143 goto error;
1151 1144 }
1152 1145 end = &path[strlen(path)];
1153 1146 *end++ = '/';
1154 1147 *end = 0;
1155 1148 pathleft = &path[sizeof (path)] - end;
1156 1149
1157 1150 /*
1158 1151 * Using raw devices instead of block devices when we're
1159 1152 * reading the labels skips a bunch of slow operations during
1160 1153 * close(2) processing, so we replace /dev/dsk with /dev/rdsk.
1161 1154 */
1162 1155 if (strcmp(path, ZFS_DISK_ROOTD) == 0)
1163 1156 (void) strlcpy(rdsk, ZFS_RDISK_ROOTD, sizeof (rdsk));
1164 1157 else
1165 1158 (void) strlcpy(rdsk, path, sizeof (rdsk));
1166 1159
1167 1160 if ((dfd = open64(rdsk, O_RDONLY)) < 0 ||
1168 1161 (dirp = fdopendir(dfd)) == NULL) {
1169 1162 if (dfd >= 0)
1170 1163 (void) close(dfd);
1171 1164 zfs_error_aux(hdl, strerror(errno));
1172 1165 (void) zfs_error_fmt(hdl, EZFS_BADPATH,
1173 1166 dgettext(TEXT_DOMAIN, "cannot open '%s'"),
1174 1167 rdsk);
1175 1168 goto error;
1176 1169 }
1177 1170
1178 1171 avl_create(&slice_cache, slice_cache_compare,
1179 1172 sizeof (rdsk_node_t), offsetof(rdsk_node_t, rn_node));
1180 1173 /*
1181 1174 * This is not MT-safe, but we have no MT consumers of libzfs
1182 1175 */
1183 1176 while ((dp = readdir64(dirp)) != NULL) {
1184 1177 const char *name = dp->d_name;
1185 1178 if (name[0] == '.' &&
1186 1179 (name[1] == 0 || (name[1] == '.' && name[2] == 0)))
1187 1180 continue;
1188 1181
1189 1182 slice = zfs_alloc(hdl, sizeof (rdsk_node_t));
1190 1183 slice->rn_name = zfs_strdup(hdl, name);
1191 1184 slice->rn_avl = &slice_cache;
1192 1185 slice->rn_dfd = dfd;
1193 1186 slice->rn_hdl = hdl;
1194 1187 slice->rn_nozpool = B_FALSE;
1195 1188 avl_add(&slice_cache, slice);
1196 1189 }
1197 1190 /*
1198 1191 * create a thread pool to do all of this in parallel;
1199 1192 * rn_nozpool is not protected, so this is racy in that
1200 1193 * multiple tasks could decide that the same slice can
1201 1194 * not hold a zpool, which is benign. Also choose
1202 1195 * double the number of processors; we hold a lot of
1203 1196 * locks in the kernel, so going beyond this doesn't
1204 1197 * buy us much.
1205 1198 */
1206 1199 t = tpool_create(1, 2 * sysconf(_SC_NPROCESSORS_ONLN),
1207 1200 0, NULL);
1208 1201 for (slice = avl_first(&slice_cache); slice;
1209 1202 (slice = avl_walk(&slice_cache, slice,
1210 1203 AVL_AFTER)))
1211 1204 (void) tpool_dispatch(t, zpool_open_func, slice);
1212 1205 tpool_wait(t);
1213 1206 tpool_destroy(t);
1214 1207
1215 1208 cookie = NULL;
1216 1209 while ((slice = avl_destroy_nodes(&slice_cache,
1217 1210 &cookie)) != NULL) {
1218 1211 if (slice->rn_config != NULL && !config_failed) {
1219 1212 nvlist_t *config = slice->rn_config;
1220 1213 boolean_t matched = B_TRUE;
1221 1214
1222 1215 if (iarg->poolname != NULL) {
1223 1216 char *pname;
1224 1217
1225 1218 matched = nvlist_lookup_string(config,
1226 1219 ZPOOL_CONFIG_POOL_NAME,
1227 1220 &pname) == 0 &&
1228 1221 strcmp(iarg->poolname, pname) == 0;
1229 1222 } else if (iarg->guid != 0) {
1230 1223 uint64_t this_guid;
1231 1224
1232 1225 matched = nvlist_lookup_uint64(config,
1233 1226 ZPOOL_CONFIG_POOL_GUID,
1234 1227 &this_guid) == 0 &&
1235 1228 iarg->guid == this_guid;
1236 1229 }
1237 1230 if (!matched) {
1238 1231 nvlist_free(config);
1239 1232 } else {
1240 1233 /*
1241 1234 * use the non-raw path for the config
1242 1235 */
1243 1236 (void) strlcpy(end, slice->rn_name,
1244 1237 pathleft);
1245 1238 if (add_config(hdl, &pools, path,
1246 1239 config) != 0)
1247 1240 config_failed = B_TRUE;
1248 1241 }
1249 1242 }
1250 1243 free(slice->rn_name);
|
↓ open down ↓ |
373 lines elided |
↑ open up ↑ |
1251 1244 free(slice);
1252 1245 }
1253 1246 avl_destroy(&slice_cache);
1254 1247
1255 1248 (void) closedir(dirp);
1256 1249
1257 1250 if (config_failed)
1258 1251 goto error;
1259 1252 }
1260 1253
1261 - ret = get_configs(hdl, &pools, iarg->can_be_active, iarg->policy);
1254 + ret = get_configs(hdl, &pools, iarg->can_be_active);
1262 1255
1263 1256 error:
1264 1257 for (pe = pools.pools; pe != NULL; pe = penext) {
1265 1258 penext = pe->pe_next;
1266 1259 for (ve = pe->pe_vdevs; ve != NULL; ve = venext) {
1267 1260 venext = ve->ve_next;
1268 1261 for (ce = ve->ve_configs; ce != NULL; ce = cenext) {
1269 1262 cenext = ce->ce_next;
1270 1263 nvlist_free(ce->ce_config);
1271 1264 free(ce);
1272 1265 }
1273 1266 free(ve);
1274 1267 }
1275 1268 free(pe);
1276 1269 }
1277 1270
1278 1271 for (ne = pools.names; ne != NULL; ne = nenext) {
1279 1272 nenext = ne->ne_next;
1280 1273 free(ne->ne_name);
1281 1274 free(ne);
1282 1275 }
1283 1276
1284 1277 return (ret);
1285 1278 }
1286 1279
1287 1280 nvlist_t *
1288 1281 zpool_find_import(libzfs_handle_t *hdl, int argc, char **argv)
1289 1282 {
1290 1283 importargs_t iarg = { 0 };
1291 1284
1292 1285 iarg.paths = argc;
1293 1286 iarg.path = argv;
1294 1287
1295 1288 return (zpool_find_import_impl(hdl, &iarg));
1296 1289 }
1297 1290
1298 1291 /*
1299 1292 * Given a cache file, return the contents as a list of importable pools.
1300 1293 * poolname or guid (but not both) are provided by the caller when trying
1301 1294 * to import a specific pool.
1302 1295 */
1303 1296 nvlist_t *
1304 1297 zpool_find_import_cached(libzfs_handle_t *hdl, const char *cachefile,
1305 1298 char *poolname, uint64_t guid)
1306 1299 {
1307 1300 char *buf;
1308 1301 int fd;
1309 1302 struct stat64 statbuf;
1310 1303 nvlist_t *raw, *src, *dst;
1311 1304 nvlist_t *pools;
1312 1305 nvpair_t *elem;
1313 1306 char *name;
1314 1307 uint64_t this_guid;
1315 1308 boolean_t active;
1316 1309
1317 1310 verify(poolname == NULL || guid == 0);
1318 1311
1319 1312 if ((fd = open(cachefile, O_RDONLY)) < 0) {
1320 1313 zfs_error_aux(hdl, "%s", strerror(errno));
1321 1314 (void) zfs_error(hdl, EZFS_BADCACHE,
1322 1315 dgettext(TEXT_DOMAIN, "failed to open cache file"));
1323 1316 return (NULL);
1324 1317 }
1325 1318
1326 1319 if (fstat64(fd, &statbuf) != 0) {
1327 1320 zfs_error_aux(hdl, "%s", strerror(errno));
1328 1321 (void) close(fd);
1329 1322 (void) zfs_error(hdl, EZFS_BADCACHE,
1330 1323 dgettext(TEXT_DOMAIN, "failed to get size of cache file"));
1331 1324 return (NULL);
1332 1325 }
1333 1326
1334 1327 if ((buf = zfs_alloc(hdl, statbuf.st_size)) == NULL) {
1335 1328 (void) close(fd);
1336 1329 return (NULL);
1337 1330 }
1338 1331
1339 1332 if (read(fd, buf, statbuf.st_size) != statbuf.st_size) {
1340 1333 (void) close(fd);
1341 1334 free(buf);
1342 1335 (void) zfs_error(hdl, EZFS_BADCACHE,
1343 1336 dgettext(TEXT_DOMAIN,
1344 1337 "failed to read cache file contents"));
1345 1338 return (NULL);
1346 1339 }
1347 1340
1348 1341 (void) close(fd);
1349 1342
1350 1343 if (nvlist_unpack(buf, statbuf.st_size, &raw, 0) != 0) {
1351 1344 free(buf);
1352 1345 (void) zfs_error(hdl, EZFS_BADCACHE,
1353 1346 dgettext(TEXT_DOMAIN,
1354 1347 "invalid or corrupt cache file contents"));
1355 1348 return (NULL);
1356 1349 }
1357 1350
1358 1351 free(buf);
1359 1352
1360 1353 /*
1361 1354 * Go through and get the current state of the pools and refresh their
1362 1355 * state.
1363 1356 */
1364 1357 if (nvlist_alloc(&pools, 0, 0) != 0) {
1365 1358 (void) no_memory(hdl);
1366 1359 nvlist_free(raw);
1367 1360 return (NULL);
1368 1361 }
1369 1362
1370 1363 elem = NULL;
1371 1364 while ((elem = nvlist_next_nvpair(raw, elem)) != NULL) {
1372 1365 src = fnvpair_value_nvlist(elem);
1373 1366
1374 1367 name = fnvlist_lookup_string(src, ZPOOL_CONFIG_POOL_NAME);
1375 1368 if (poolname != NULL && strcmp(poolname, name) != 0)
1376 1369 continue;
1377 1370
1378 1371 this_guid = fnvlist_lookup_uint64(src, ZPOOL_CONFIG_POOL_GUID);
1379 1372 if (guid != 0 && guid != this_guid)
1380 1373 continue;
|
↓ open down ↓ |
109 lines elided |
↑ open up ↑ |
1381 1374
1382 1375 if (pool_active(hdl, name, this_guid, &active) != 0) {
1383 1376 nvlist_free(raw);
1384 1377 nvlist_free(pools);
1385 1378 return (NULL);
1386 1379 }
1387 1380
1388 1381 if (active)
1389 1382 continue;
1390 1383
1391 - if (nvlist_add_string(src, ZPOOL_CONFIG_CACHEFILE,
1392 - cachefile) != 0) {
1393 - (void) no_memory(hdl);
1394 - nvlist_free(raw);
1395 - nvlist_free(pools);
1396 - return (NULL);
1397 - }
1398 -
1399 1384 if ((dst = refresh_config(hdl, src)) == NULL) {
1400 1385 nvlist_free(raw);
1401 1386 nvlist_free(pools);
1402 1387 return (NULL);
1403 1388 }
1404 1389
1405 1390 if (nvlist_add_nvlist(pools, nvpair_name(elem), dst) != 0) {
1406 1391 (void) no_memory(hdl);
1407 1392 nvlist_free(dst);
1408 1393 nvlist_free(raw);
1409 1394 nvlist_free(pools);
1410 1395 return (NULL);
1411 1396 }
1412 1397 nvlist_free(dst);
1413 1398 }
1414 1399
1415 1400 nvlist_free(raw);
1416 1401 return (pools);
1417 1402 }
1418 1403
1419 1404 static int
1420 1405 name_or_guid_exists(zpool_handle_t *zhp, void *data)
1421 1406 {
1422 1407 importargs_t *import = data;
1423 1408 int found = 0;
1424 1409
1425 1410 if (import->poolname != NULL) {
1426 1411 char *pool_name;
1427 1412
1428 1413 verify(nvlist_lookup_string(zhp->zpool_config,
1429 1414 ZPOOL_CONFIG_POOL_NAME, &pool_name) == 0);
1430 1415 if (strcmp(pool_name, import->poolname) == 0)
1431 1416 found = 1;
1432 1417 } else {
1433 1418 uint64_t pool_guid;
1434 1419
1435 1420 verify(nvlist_lookup_uint64(zhp->zpool_config,
1436 1421 ZPOOL_CONFIG_POOL_GUID, &pool_guid) == 0);
1437 1422 if (pool_guid == import->guid)
1438 1423 found = 1;
1439 1424 }
1440 1425
1441 1426 zpool_close(zhp);
1442 1427 return (found);
1443 1428 }
1444 1429
1445 1430 nvlist_t *
1446 1431 zpool_search_import(libzfs_handle_t *hdl, importargs_t *import)
1447 1432 {
1448 1433 verify(import->poolname == NULL || import->guid == 0);
1449 1434
1450 1435 if (import->unique)
1451 1436 import->exists = zpool_iter(hdl, name_or_guid_exists, import);
1452 1437
1453 1438 if (import->cachefile != NULL)
1454 1439 return (zpool_find_import_cached(hdl, import->cachefile,
1455 1440 import->poolname, import->guid));
1456 1441
1457 1442 return (zpool_find_import_impl(hdl, import));
1458 1443 }
1459 1444
1460 1445 boolean_t
1461 1446 find_guid(nvlist_t *nv, uint64_t guid)
1462 1447 {
1463 1448 uint64_t tmp;
1464 1449 nvlist_t **child;
1465 1450 uint_t c, children;
1466 1451
1467 1452 verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID, &tmp) == 0);
1468 1453 if (tmp == guid)
1469 1454 return (B_TRUE);
1470 1455
1471 1456 if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN,
1472 1457 &child, &children) == 0) {
1473 1458 for (c = 0; c < children; c++)
1474 1459 if (find_guid(child[c], guid))
1475 1460 return (B_TRUE);
1476 1461 }
1477 1462
1478 1463 return (B_FALSE);
1479 1464 }
1480 1465
1481 1466 typedef struct aux_cbdata {
1482 1467 const char *cb_type;
1483 1468 uint64_t cb_guid;
1484 1469 zpool_handle_t *cb_zhp;
1485 1470 } aux_cbdata_t;
1486 1471
1487 1472 static int
1488 1473 find_aux(zpool_handle_t *zhp, void *data)
1489 1474 {
1490 1475 aux_cbdata_t *cbp = data;
1491 1476 nvlist_t **list;
1492 1477 uint_t i, count;
1493 1478 uint64_t guid;
1494 1479 nvlist_t *nvroot;
1495 1480
1496 1481 verify(nvlist_lookup_nvlist(zhp->zpool_config, ZPOOL_CONFIG_VDEV_TREE,
1497 1482 &nvroot) == 0);
1498 1483
1499 1484 if (nvlist_lookup_nvlist_array(nvroot, cbp->cb_type,
1500 1485 &list, &count) == 0) {
1501 1486 for (i = 0; i < count; i++) {
1502 1487 verify(nvlist_lookup_uint64(list[i],
1503 1488 ZPOOL_CONFIG_GUID, &guid) == 0);
1504 1489 if (guid == cbp->cb_guid) {
1505 1490 cbp->cb_zhp = zhp;
1506 1491 return (1);
1507 1492 }
1508 1493 }
1509 1494 }
1510 1495
1511 1496 zpool_close(zhp);
1512 1497 return (0);
1513 1498 }
1514 1499
1515 1500 /*
1516 1501 * Determines if the pool is in use. If so, it returns true and the state of
1517 1502 * the pool as well as the name of the pool. Both strings are allocated and
1518 1503 * must be freed by the caller.
1519 1504 */
1520 1505 int
1521 1506 zpool_in_use(libzfs_handle_t *hdl, int fd, pool_state_t *state, char **namestr,
1522 1507 boolean_t *inuse)
1523 1508 {
1524 1509 nvlist_t *config;
1525 1510 char *name;
1526 1511 boolean_t ret;
1527 1512 uint64_t guid, vdev_guid;
1528 1513 zpool_handle_t *zhp;
1529 1514 nvlist_t *pool_config;
1530 1515 uint64_t stateval, isspare;
1531 1516 aux_cbdata_t cb = { 0 };
1532 1517 boolean_t isactive;
1533 1518
1534 1519 *inuse = B_FALSE;
1535 1520
1536 1521 if (zpool_read_label(fd, &config) != 0 && errno == ENOMEM) {
1537 1522 (void) no_memory(hdl);
1538 1523 return (-1);
1539 1524 }
1540 1525
1541 1526 if (config == NULL)
1542 1527 return (0);
1543 1528
1544 1529 verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_STATE,
1545 1530 &stateval) == 0);
1546 1531 verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_GUID,
1547 1532 &vdev_guid) == 0);
1548 1533
1549 1534 if (stateval != POOL_STATE_SPARE && stateval != POOL_STATE_L2CACHE) {
1550 1535 verify(nvlist_lookup_string(config, ZPOOL_CONFIG_POOL_NAME,
1551 1536 &name) == 0);
1552 1537 verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID,
1553 1538 &guid) == 0);
1554 1539 }
1555 1540
1556 1541 switch (stateval) {
1557 1542 case POOL_STATE_EXPORTED:
1558 1543 /*
1559 1544 * A pool with an exported state may in fact be imported
1560 1545 * read-only, so check the in-core state to see if it's
1561 1546 * active and imported read-only. If it is, set
1562 1547 * its state to active.
1563 1548 */
1564 1549 if (pool_active(hdl, name, guid, &isactive) == 0 && isactive &&
1565 1550 (zhp = zpool_open_canfail(hdl, name)) != NULL) {
1566 1551 if (zpool_get_prop_int(zhp, ZPOOL_PROP_READONLY, NULL))
1567 1552 stateval = POOL_STATE_ACTIVE;
1568 1553
1569 1554 /*
1570 1555 * All we needed the zpool handle for is the
1571 1556 * readonly prop check.
1572 1557 */
1573 1558 zpool_close(zhp);
1574 1559 }
1575 1560
1576 1561 ret = B_TRUE;
1577 1562 break;
1578 1563
1579 1564 case POOL_STATE_ACTIVE:
1580 1565 /*
1581 1566 * For an active pool, we have to determine if it's really part
1582 1567 * of a currently active pool (in which case the pool will exist
1583 1568 * and the guid will be the same), or whether it's part of an
1584 1569 * active pool that was disconnected without being explicitly
1585 1570 * exported.
1586 1571 */
1587 1572 if (pool_active(hdl, name, guid, &isactive) != 0) {
1588 1573 nvlist_free(config);
1589 1574 return (-1);
1590 1575 }
1591 1576
1592 1577 if (isactive) {
1593 1578 /*
1594 1579 * Because the device may have been removed while
1595 1580 * offlined, we only report it as active if the vdev is
1596 1581 * still present in the config. Otherwise, pretend like
1597 1582 * it's not in use.
1598 1583 */
1599 1584 if ((zhp = zpool_open_canfail(hdl, name)) != NULL &&
1600 1585 (pool_config = zpool_get_config(zhp, NULL))
1601 1586 != NULL) {
1602 1587 nvlist_t *nvroot;
1603 1588
1604 1589 verify(nvlist_lookup_nvlist(pool_config,
1605 1590 ZPOOL_CONFIG_VDEV_TREE, &nvroot) == 0);
1606 1591 ret = find_guid(nvroot, vdev_guid);
1607 1592 } else {
1608 1593 ret = B_FALSE;
1609 1594 }
1610 1595
1611 1596 /*
1612 1597 * If this is an active spare within another pool, we
1613 1598 * treat it like an unused hot spare. This allows the
1614 1599 * user to create a pool with a hot spare that currently
1615 1600 * in use within another pool. Since we return B_TRUE,
1616 1601 * libdiskmgt will continue to prevent generic consumers
1617 1602 * from using the device.
1618 1603 */
1619 1604 if (ret && nvlist_lookup_uint64(config,
1620 1605 ZPOOL_CONFIG_IS_SPARE, &isspare) == 0 && isspare)
1621 1606 stateval = POOL_STATE_SPARE;
1622 1607
1623 1608 if (zhp != NULL)
1624 1609 zpool_close(zhp);
1625 1610 } else {
1626 1611 stateval = POOL_STATE_POTENTIALLY_ACTIVE;
1627 1612 ret = B_TRUE;
1628 1613 }
1629 1614 break;
1630 1615
1631 1616 case POOL_STATE_SPARE:
1632 1617 /*
1633 1618 * For a hot spare, it can be either definitively in use, or
1634 1619 * potentially active. To determine if it's in use, we iterate
1635 1620 * over all pools in the system and search for one with a spare
1636 1621 * with a matching guid.
1637 1622 *
1638 1623 * Due to the shared nature of spares, we don't actually report
1639 1624 * the potentially active case as in use. This means the user
1640 1625 * can freely create pools on the hot spares of exported pools,
1641 1626 * but to do otherwise makes the resulting code complicated, and
1642 1627 * we end up having to deal with this case anyway.
1643 1628 */
1644 1629 cb.cb_zhp = NULL;
1645 1630 cb.cb_guid = vdev_guid;
1646 1631 cb.cb_type = ZPOOL_CONFIG_SPARES;
1647 1632 if (zpool_iter(hdl, find_aux, &cb) == 1) {
1648 1633 name = (char *)zpool_get_name(cb.cb_zhp);
1649 1634 ret = B_TRUE;
1650 1635 } else {
1651 1636 ret = B_FALSE;
1652 1637 }
1653 1638 break;
1654 1639
1655 1640 case POOL_STATE_L2CACHE:
1656 1641
1657 1642 /*
1658 1643 * Check if any pool is currently using this l2cache device.
1659 1644 */
1660 1645 cb.cb_zhp = NULL;
1661 1646 cb.cb_guid = vdev_guid;
1662 1647 cb.cb_type = ZPOOL_CONFIG_L2CACHE;
1663 1648 if (zpool_iter(hdl, find_aux, &cb) == 1) {
1664 1649 name = (char *)zpool_get_name(cb.cb_zhp);
1665 1650 ret = B_TRUE;
1666 1651 } else {
1667 1652 ret = B_FALSE;
1668 1653 }
1669 1654 break;
1670 1655
1671 1656 default:
1672 1657 ret = B_FALSE;
1673 1658 }
1674 1659
1675 1660
1676 1661 if (ret) {
1677 1662 if ((*namestr = zfs_strdup(hdl, name)) == NULL) {
1678 1663 if (cb.cb_zhp)
1679 1664 zpool_close(cb.cb_zhp);
1680 1665 nvlist_free(config);
1681 1666 return (-1);
1682 1667 }
1683 1668 *state = (pool_state_t)stateval;
1684 1669 }
1685 1670
1686 1671 if (cb.cb_zhp)
1687 1672 zpool_close(cb.cb_zhp);
1688 1673
1689 1674 nvlist_free(config);
1690 1675 *inuse = ret;
1691 1676 return (0);
1692 1677 }
|
↓ open down ↓ |
284 lines elided |
↑ open up ↑ |
XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX