Print this page
4626 libzfs memleak in zpool_in_use()
Reviewed by: Tony Nguyen <tony.nguyen@nexenta.com>
Reviewed by: Saso Kiselkov <saso.kiselkov@nexenta.com>
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 * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
23 23 * Copyright 2011 Nexenta Systems, Inc. All rights reserved.
24 24 * Copyright (c) 2012 by Delphix. All rights reserved.
25 25 */
26 26
27 27 /*
28 28 * Pool import support functions.
29 29 *
30 30 * To import a pool, we rely on reading the configuration information from the
31 31 * ZFS label of each device. If we successfully read the label, then we
32 32 * organize the configuration information in the following hierarchy:
33 33 *
34 34 * pool guid -> toplevel vdev guid -> label txg
35 35 *
36 36 * Duplicate entries matching this same tuple will be discarded. Once we have
37 37 * examined every device, we pick the best label txg config for each toplevel
38 38 * vdev. We then arrange these toplevel vdevs into a complete pool config, and
39 39 * update any paths that have changed. Finally, we attempt to import the pool
40 40 * using our derived config, and record the results.
41 41 */
42 42
43 43 #include <ctype.h>
44 44 #include <devid.h>
45 45 #include <dirent.h>
46 46 #include <errno.h>
47 47 #include <libintl.h>
48 48 #include <stddef.h>
49 49 #include <stdlib.h>
50 50 #include <string.h>
51 51 #include <sys/stat.h>
52 52 #include <unistd.h>
53 53 #include <fcntl.h>
54 54 #include <sys/vtoc.h>
55 55 #include <sys/dktp/fdisk.h>
56 56 #include <sys/efi_partition.h>
57 57 #include <thread_pool.h>
58 58
59 59 #include <sys/vdev_impl.h>
60 60
61 61 #include "libzfs.h"
62 62 #include "libzfs_impl.h"
63 63
64 64 /*
65 65 * Intermediate structures used to gather configuration information.
66 66 */
67 67 typedef struct config_entry {
68 68 uint64_t ce_txg;
69 69 nvlist_t *ce_config;
70 70 struct config_entry *ce_next;
71 71 } config_entry_t;
72 72
73 73 typedef struct vdev_entry {
74 74 uint64_t ve_guid;
75 75 config_entry_t *ve_configs;
76 76 struct vdev_entry *ve_next;
77 77 } vdev_entry_t;
78 78
79 79 typedef struct pool_entry {
80 80 uint64_t pe_guid;
81 81 vdev_entry_t *pe_vdevs;
82 82 struct pool_entry *pe_next;
83 83 } pool_entry_t;
84 84
85 85 typedef struct name_entry {
86 86 char *ne_name;
87 87 uint64_t ne_guid;
88 88 struct name_entry *ne_next;
89 89 } name_entry_t;
90 90
91 91 typedef struct pool_list {
92 92 pool_entry_t *pools;
93 93 name_entry_t *names;
94 94 } pool_list_t;
95 95
96 96 static char *
97 97 get_devid(const char *path)
98 98 {
99 99 int fd;
100 100 ddi_devid_t devid;
101 101 char *minor, *ret;
102 102
103 103 if ((fd = open(path, O_RDONLY)) < 0)
104 104 return (NULL);
105 105
106 106 minor = NULL;
107 107 ret = NULL;
108 108 if (devid_get(fd, &devid) == 0) {
109 109 if (devid_get_minor_name(fd, &minor) == 0)
110 110 ret = devid_str_encode(devid, minor);
111 111 if (minor != NULL)
112 112 devid_str_free(minor);
113 113 devid_free(devid);
114 114 }
115 115 (void) close(fd);
116 116
117 117 return (ret);
118 118 }
119 119
120 120
121 121 /*
122 122 * Go through and fix up any path and/or devid information for the given vdev
123 123 * configuration.
124 124 */
125 125 static int
126 126 fix_paths(nvlist_t *nv, name_entry_t *names)
127 127 {
128 128 nvlist_t **child;
129 129 uint_t c, children;
130 130 uint64_t guid;
131 131 name_entry_t *ne, *best;
132 132 char *path, *devid;
133 133 int matched;
134 134
135 135 if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN,
136 136 &child, &children) == 0) {
137 137 for (c = 0; c < children; c++)
138 138 if (fix_paths(child[c], names) != 0)
139 139 return (-1);
140 140 return (0);
141 141 }
142 142
143 143 /*
144 144 * This is a leaf (file or disk) vdev. In either case, go through
145 145 * the name list and see if we find a matching guid. If so, replace
146 146 * the path and see if we can calculate a new devid.
147 147 *
148 148 * There may be multiple names associated with a particular guid, in
149 149 * which case we have overlapping slices or multiple paths to the same
150 150 * disk. If this is the case, then we want to pick the path that is
151 151 * the most similar to the original, where "most similar" is the number
152 152 * of matching characters starting from the end of the path. This will
153 153 * preserve slice numbers even if the disks have been reorganized, and
154 154 * will also catch preferred disk names if multiple paths exist.
155 155 */
156 156 verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID, &guid) == 0);
157 157 if (nvlist_lookup_string(nv, ZPOOL_CONFIG_PATH, &path) != 0)
158 158 path = NULL;
159 159
160 160 matched = 0;
161 161 best = NULL;
162 162 for (ne = names; ne != NULL; ne = ne->ne_next) {
163 163 if (ne->ne_guid == guid) {
164 164 const char *src, *dst;
165 165 int count;
166 166
167 167 if (path == NULL) {
168 168 best = ne;
169 169 break;
170 170 }
171 171
172 172 src = ne->ne_name + strlen(ne->ne_name) - 1;
173 173 dst = path + strlen(path) - 1;
174 174 for (count = 0; src >= ne->ne_name && dst >= path;
175 175 src--, dst--, count++)
176 176 if (*src != *dst)
177 177 break;
178 178
179 179 /*
180 180 * At this point, 'count' is the number of characters
181 181 * matched from the end.
182 182 */
183 183 if (count > matched || best == NULL) {
184 184 best = ne;
185 185 matched = count;
186 186 }
187 187 }
188 188 }
189 189
190 190 if (best == NULL)
191 191 return (0);
192 192
193 193 if (nvlist_add_string(nv, ZPOOL_CONFIG_PATH, best->ne_name) != 0)
194 194 return (-1);
195 195
196 196 if ((devid = get_devid(best->ne_name)) == NULL) {
197 197 (void) nvlist_remove_all(nv, ZPOOL_CONFIG_DEVID);
198 198 } else {
199 199 if (nvlist_add_string(nv, ZPOOL_CONFIG_DEVID, devid) != 0)
200 200 return (-1);
201 201 devid_str_free(devid);
202 202 }
203 203
204 204 return (0);
205 205 }
206 206
207 207 /*
208 208 * Add the given configuration to the list of known devices.
209 209 */
210 210 static int
211 211 add_config(libzfs_handle_t *hdl, pool_list_t *pl, const char *path,
212 212 nvlist_t *config)
213 213 {
214 214 uint64_t pool_guid, vdev_guid, top_guid, txg, state;
215 215 pool_entry_t *pe;
216 216 vdev_entry_t *ve;
217 217 config_entry_t *ce;
218 218 name_entry_t *ne;
219 219
220 220 /*
221 221 * If this is a hot spare not currently in use or level 2 cache
222 222 * device, add it to the list of names to translate, but don't do
223 223 * anything else.
224 224 */
225 225 if (nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_STATE,
226 226 &state) == 0 &&
227 227 (state == POOL_STATE_SPARE || state == POOL_STATE_L2CACHE) &&
228 228 nvlist_lookup_uint64(config, ZPOOL_CONFIG_GUID, &vdev_guid) == 0) {
229 229 if ((ne = zfs_alloc(hdl, sizeof (name_entry_t))) == NULL)
230 230 return (-1);
231 231
232 232 if ((ne->ne_name = zfs_strdup(hdl, path)) == NULL) {
233 233 free(ne);
234 234 return (-1);
235 235 }
236 236 ne->ne_guid = vdev_guid;
237 237 ne->ne_next = pl->names;
238 238 pl->names = ne;
239 239 return (0);
240 240 }
241 241
242 242 /*
243 243 * If we have a valid config but cannot read any of these fields, then
244 244 * it means we have a half-initialized label. In vdev_label_init()
245 245 * we write a label with txg == 0 so that we can identify the device
246 246 * in case the user refers to the same disk later on. If we fail to
247 247 * create the pool, we'll be left with a label in this state
248 248 * which should not be considered part of a valid pool.
249 249 */
250 250 if (nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID,
251 251 &pool_guid) != 0 ||
252 252 nvlist_lookup_uint64(config, ZPOOL_CONFIG_GUID,
253 253 &vdev_guid) != 0 ||
254 254 nvlist_lookup_uint64(config, ZPOOL_CONFIG_TOP_GUID,
255 255 &top_guid) != 0 ||
256 256 nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_TXG,
257 257 &txg) != 0 || txg == 0) {
258 258 nvlist_free(config);
259 259 return (0);
260 260 }
261 261
262 262 /*
263 263 * First, see if we know about this pool. If not, then add it to the
264 264 * list of known pools.
265 265 */
266 266 for (pe = pl->pools; pe != NULL; pe = pe->pe_next) {
267 267 if (pe->pe_guid == pool_guid)
268 268 break;
269 269 }
270 270
271 271 if (pe == NULL) {
272 272 if ((pe = zfs_alloc(hdl, sizeof (pool_entry_t))) == NULL) {
273 273 nvlist_free(config);
274 274 return (-1);
275 275 }
276 276 pe->pe_guid = pool_guid;
277 277 pe->pe_next = pl->pools;
278 278 pl->pools = pe;
279 279 }
280 280
281 281 /*
282 282 * Second, see if we know about this toplevel vdev. Add it if its
283 283 * missing.
284 284 */
285 285 for (ve = pe->pe_vdevs; ve != NULL; ve = ve->ve_next) {
286 286 if (ve->ve_guid == top_guid)
287 287 break;
288 288 }
289 289
290 290 if (ve == NULL) {
291 291 if ((ve = zfs_alloc(hdl, sizeof (vdev_entry_t))) == NULL) {
292 292 nvlist_free(config);
293 293 return (-1);
294 294 }
295 295 ve->ve_guid = top_guid;
296 296 ve->ve_next = pe->pe_vdevs;
297 297 pe->pe_vdevs = ve;
298 298 }
299 299
300 300 /*
301 301 * Third, see if we have a config with a matching transaction group. If
302 302 * so, then we do nothing. Otherwise, add it to the list of known
303 303 * configs.
304 304 */
305 305 for (ce = ve->ve_configs; ce != NULL; ce = ce->ce_next) {
306 306 if (ce->ce_txg == txg)
307 307 break;
308 308 }
309 309
310 310 if (ce == NULL) {
311 311 if ((ce = zfs_alloc(hdl, sizeof (config_entry_t))) == NULL) {
312 312 nvlist_free(config);
313 313 return (-1);
314 314 }
315 315 ce->ce_txg = txg;
316 316 ce->ce_config = config;
317 317 ce->ce_next = ve->ve_configs;
318 318 ve->ve_configs = ce;
319 319 } else {
320 320 nvlist_free(config);
321 321 }
322 322
323 323 /*
324 324 * At this point we've successfully added our config to the list of
325 325 * known configs. The last thing to do is add the vdev guid -> path
326 326 * mappings so that we can fix up the configuration as necessary before
327 327 * doing the import.
328 328 */
329 329 if ((ne = zfs_alloc(hdl, sizeof (name_entry_t))) == NULL)
330 330 return (-1);
331 331
332 332 if ((ne->ne_name = zfs_strdup(hdl, path)) == NULL) {
333 333 free(ne);
334 334 return (-1);
335 335 }
336 336
337 337 ne->ne_guid = vdev_guid;
338 338 ne->ne_next = pl->names;
339 339 pl->names = ne;
340 340
341 341 return (0);
342 342 }
343 343
344 344 /*
345 345 * Returns true if the named pool matches the given GUID.
346 346 */
347 347 static int
348 348 pool_active(libzfs_handle_t *hdl, const char *name, uint64_t guid,
349 349 boolean_t *isactive)
350 350 {
351 351 zpool_handle_t *zhp;
352 352 uint64_t theguid;
353 353
354 354 if (zpool_open_silent(hdl, name, &zhp) != 0)
355 355 return (-1);
356 356
357 357 if (zhp == NULL) {
358 358 *isactive = B_FALSE;
359 359 return (0);
360 360 }
361 361
362 362 verify(nvlist_lookup_uint64(zhp->zpool_config, ZPOOL_CONFIG_POOL_GUID,
363 363 &theguid) == 0);
364 364
365 365 zpool_close(zhp);
366 366
367 367 *isactive = (theguid == guid);
368 368 return (0);
369 369 }
370 370
371 371 static nvlist_t *
372 372 refresh_config(libzfs_handle_t *hdl, nvlist_t *config)
373 373 {
374 374 nvlist_t *nvl;
375 375 zfs_cmd_t zc = { 0 };
376 376 int err;
377 377
378 378 if (zcmd_write_conf_nvlist(hdl, &zc, config) != 0)
379 379 return (NULL);
380 380
381 381 if (zcmd_alloc_dst_nvlist(hdl, &zc,
382 382 zc.zc_nvlist_conf_size * 2) != 0) {
383 383 zcmd_free_nvlists(&zc);
384 384 return (NULL);
385 385 }
386 386
387 387 while ((err = ioctl(hdl->libzfs_fd, ZFS_IOC_POOL_TRYIMPORT,
388 388 &zc)) != 0 && errno == ENOMEM) {
389 389 if (zcmd_expand_dst_nvlist(hdl, &zc) != 0) {
390 390 zcmd_free_nvlists(&zc);
391 391 return (NULL);
392 392 }
393 393 }
394 394
395 395 if (err) {
396 396 zcmd_free_nvlists(&zc);
397 397 return (NULL);
398 398 }
399 399
400 400 if (zcmd_read_dst_nvlist(hdl, &zc, &nvl) != 0) {
401 401 zcmd_free_nvlists(&zc);
402 402 return (NULL);
403 403 }
404 404
405 405 zcmd_free_nvlists(&zc);
406 406 return (nvl);
407 407 }
408 408
409 409 /*
410 410 * Determine if the vdev id is a hole in the namespace.
411 411 */
412 412 boolean_t
413 413 vdev_is_hole(uint64_t *hole_array, uint_t holes, uint_t id)
414 414 {
415 415 for (int c = 0; c < holes; c++) {
416 416
417 417 /* Top-level is a hole */
418 418 if (hole_array[c] == id)
419 419 return (B_TRUE);
420 420 }
421 421 return (B_FALSE);
422 422 }
423 423
424 424 /*
425 425 * Convert our list of pools into the definitive set of configurations. We
426 426 * start by picking the best config for each toplevel vdev. Once that's done,
427 427 * we assemble the toplevel vdevs into a full config for the pool. We make a
428 428 * pass to fix up any incorrect paths, and then add it to the main list to
429 429 * return to the user.
430 430 */
431 431 static nvlist_t *
432 432 get_configs(libzfs_handle_t *hdl, pool_list_t *pl, boolean_t active_ok)
433 433 {
434 434 pool_entry_t *pe;
435 435 vdev_entry_t *ve;
436 436 config_entry_t *ce;
437 437 nvlist_t *ret = NULL, *config = NULL, *tmp, *nvtop, *nvroot;
438 438 nvlist_t **spares, **l2cache;
439 439 uint_t i, nspares, nl2cache;
440 440 boolean_t config_seen;
441 441 uint64_t best_txg;
442 442 char *name, *hostname;
443 443 uint64_t guid;
444 444 uint_t children = 0;
445 445 nvlist_t **child = NULL;
446 446 uint_t holes;
447 447 uint64_t *hole_array, max_id;
448 448 uint_t c;
449 449 boolean_t isactive;
450 450 uint64_t hostid;
451 451 nvlist_t *nvl;
452 452 boolean_t found_one = B_FALSE;
453 453 boolean_t valid_top_config = B_FALSE;
454 454
455 455 if (nvlist_alloc(&ret, 0, 0) != 0)
456 456 goto nomem;
457 457
458 458 for (pe = pl->pools; pe != NULL; pe = pe->pe_next) {
459 459 uint64_t id, max_txg = 0;
460 460
461 461 if (nvlist_alloc(&config, NV_UNIQUE_NAME, 0) != 0)
462 462 goto nomem;
463 463 config_seen = B_FALSE;
464 464
465 465 /*
466 466 * Iterate over all toplevel vdevs. Grab the pool configuration
467 467 * from the first one we find, and then go through the rest and
468 468 * add them as necessary to the 'vdevs' member of the config.
469 469 */
470 470 for (ve = pe->pe_vdevs; ve != NULL; ve = ve->ve_next) {
471 471
472 472 /*
473 473 * Determine the best configuration for this vdev by
474 474 * selecting the config with the latest transaction
475 475 * group.
476 476 */
477 477 best_txg = 0;
478 478 for (ce = ve->ve_configs; ce != NULL;
479 479 ce = ce->ce_next) {
480 480
481 481 if (ce->ce_txg > best_txg) {
482 482 tmp = ce->ce_config;
483 483 best_txg = ce->ce_txg;
484 484 }
485 485 }
486 486
487 487 /*
488 488 * We rely on the fact that the max txg for the
489 489 * pool will contain the most up-to-date information
490 490 * about the valid top-levels in the vdev namespace.
491 491 */
492 492 if (best_txg > max_txg) {
493 493 (void) nvlist_remove(config,
494 494 ZPOOL_CONFIG_VDEV_CHILDREN,
495 495 DATA_TYPE_UINT64);
496 496 (void) nvlist_remove(config,
497 497 ZPOOL_CONFIG_HOLE_ARRAY,
498 498 DATA_TYPE_UINT64_ARRAY);
499 499
500 500 max_txg = best_txg;
501 501 hole_array = NULL;
502 502 holes = 0;
503 503 max_id = 0;
504 504 valid_top_config = B_FALSE;
505 505
506 506 if (nvlist_lookup_uint64(tmp,
507 507 ZPOOL_CONFIG_VDEV_CHILDREN, &max_id) == 0) {
508 508 verify(nvlist_add_uint64(config,
509 509 ZPOOL_CONFIG_VDEV_CHILDREN,
510 510 max_id) == 0);
511 511 valid_top_config = B_TRUE;
512 512 }
513 513
514 514 if (nvlist_lookup_uint64_array(tmp,
515 515 ZPOOL_CONFIG_HOLE_ARRAY, &hole_array,
516 516 &holes) == 0) {
517 517 verify(nvlist_add_uint64_array(config,
518 518 ZPOOL_CONFIG_HOLE_ARRAY,
519 519 hole_array, holes) == 0);
520 520 }
521 521 }
522 522
523 523 if (!config_seen) {
524 524 /*
525 525 * Copy the relevant pieces of data to the pool
526 526 * configuration:
527 527 *
528 528 * version
529 529 * pool guid
530 530 * name
531 531 * comment (if available)
532 532 * pool state
533 533 * hostid (if available)
534 534 * hostname (if available)
535 535 */
536 536 uint64_t state, version;
537 537 char *comment = NULL;
538 538
539 539 version = fnvlist_lookup_uint64(tmp,
540 540 ZPOOL_CONFIG_VERSION);
541 541 fnvlist_add_uint64(config,
542 542 ZPOOL_CONFIG_VERSION, version);
543 543 guid = fnvlist_lookup_uint64(tmp,
544 544 ZPOOL_CONFIG_POOL_GUID);
545 545 fnvlist_add_uint64(config,
546 546 ZPOOL_CONFIG_POOL_GUID, guid);
547 547 name = fnvlist_lookup_string(tmp,
548 548 ZPOOL_CONFIG_POOL_NAME);
549 549 fnvlist_add_string(config,
550 550 ZPOOL_CONFIG_POOL_NAME, name);
551 551
552 552 if (nvlist_lookup_string(tmp,
553 553 ZPOOL_CONFIG_COMMENT, &comment) == 0)
554 554 fnvlist_add_string(config,
555 555 ZPOOL_CONFIG_COMMENT, comment);
556 556
557 557 state = fnvlist_lookup_uint64(tmp,
558 558 ZPOOL_CONFIG_POOL_STATE);
559 559 fnvlist_add_uint64(config,
560 560 ZPOOL_CONFIG_POOL_STATE, state);
561 561
562 562 hostid = 0;
563 563 if (nvlist_lookup_uint64(tmp,
564 564 ZPOOL_CONFIG_HOSTID, &hostid) == 0) {
565 565 fnvlist_add_uint64(config,
566 566 ZPOOL_CONFIG_HOSTID, hostid);
567 567 hostname = fnvlist_lookup_string(tmp,
568 568 ZPOOL_CONFIG_HOSTNAME);
569 569 fnvlist_add_string(config,
570 570 ZPOOL_CONFIG_HOSTNAME, hostname);
571 571 }
572 572
573 573 config_seen = B_TRUE;
574 574 }
575 575
576 576 /*
577 577 * Add this top-level vdev to the child array.
578 578 */
579 579 verify(nvlist_lookup_nvlist(tmp,
580 580 ZPOOL_CONFIG_VDEV_TREE, &nvtop) == 0);
581 581 verify(nvlist_lookup_uint64(nvtop, ZPOOL_CONFIG_ID,
582 582 &id) == 0);
583 583
584 584 if (id >= children) {
585 585 nvlist_t **newchild;
586 586
587 587 newchild = zfs_alloc(hdl, (id + 1) *
588 588 sizeof (nvlist_t *));
589 589 if (newchild == NULL)
590 590 goto nomem;
591 591
592 592 for (c = 0; c < children; c++)
593 593 newchild[c] = child[c];
594 594
595 595 free(child);
596 596 child = newchild;
597 597 children = id + 1;
598 598 }
599 599 if (nvlist_dup(nvtop, &child[id], 0) != 0)
600 600 goto nomem;
601 601
602 602 }
603 603
604 604 /*
605 605 * If we have information about all the top-levels then
606 606 * clean up the nvlist which we've constructed. This
607 607 * means removing any extraneous devices that are
608 608 * beyond the valid range or adding devices to the end
609 609 * of our array which appear to be missing.
610 610 */
611 611 if (valid_top_config) {
612 612 if (max_id < children) {
613 613 for (c = max_id; c < children; c++)
614 614 nvlist_free(child[c]);
615 615 children = max_id;
616 616 } else if (max_id > children) {
617 617 nvlist_t **newchild;
618 618
619 619 newchild = zfs_alloc(hdl, (max_id) *
620 620 sizeof (nvlist_t *));
621 621 if (newchild == NULL)
622 622 goto nomem;
623 623
624 624 for (c = 0; c < children; c++)
625 625 newchild[c] = child[c];
626 626
627 627 free(child);
628 628 child = newchild;
629 629 children = max_id;
630 630 }
631 631 }
632 632
633 633 verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID,
634 634 &guid) == 0);
635 635
636 636 /*
637 637 * The vdev namespace may contain holes as a result of
638 638 * device removal. We must add them back into the vdev
639 639 * tree before we process any missing devices.
640 640 */
641 641 if (holes > 0) {
642 642 ASSERT(valid_top_config);
643 643
644 644 for (c = 0; c < children; c++) {
645 645 nvlist_t *holey;
646 646
647 647 if (child[c] != NULL ||
648 648 !vdev_is_hole(hole_array, holes, c))
649 649 continue;
650 650
651 651 if (nvlist_alloc(&holey, NV_UNIQUE_NAME,
652 652 0) != 0)
653 653 goto nomem;
654 654
655 655 /*
656 656 * Holes in the namespace are treated as
657 657 * "hole" top-level vdevs and have a
658 658 * special flag set on them.
659 659 */
660 660 if (nvlist_add_string(holey,
661 661 ZPOOL_CONFIG_TYPE,
662 662 VDEV_TYPE_HOLE) != 0 ||
663 663 nvlist_add_uint64(holey,
664 664 ZPOOL_CONFIG_ID, c) != 0 ||
665 665 nvlist_add_uint64(holey,
666 666 ZPOOL_CONFIG_GUID, 0ULL) != 0)
667 667 goto nomem;
668 668 child[c] = holey;
669 669 }
670 670 }
671 671
672 672 /*
673 673 * Look for any missing top-level vdevs. If this is the case,
674 674 * create a faked up 'missing' vdev as a placeholder. We cannot
675 675 * simply compress the child array, because the kernel performs
676 676 * certain checks to make sure the vdev IDs match their location
677 677 * in the configuration.
678 678 */
679 679 for (c = 0; c < children; c++) {
680 680 if (child[c] == NULL) {
681 681 nvlist_t *missing;
682 682 if (nvlist_alloc(&missing, NV_UNIQUE_NAME,
683 683 0) != 0)
684 684 goto nomem;
685 685 if (nvlist_add_string(missing,
686 686 ZPOOL_CONFIG_TYPE,
687 687 VDEV_TYPE_MISSING) != 0 ||
688 688 nvlist_add_uint64(missing,
689 689 ZPOOL_CONFIG_ID, c) != 0 ||
690 690 nvlist_add_uint64(missing,
691 691 ZPOOL_CONFIG_GUID, 0ULL) != 0) {
692 692 nvlist_free(missing);
693 693 goto nomem;
694 694 }
695 695 child[c] = missing;
696 696 }
697 697 }
698 698
699 699 /*
700 700 * Put all of this pool's top-level vdevs into a root vdev.
701 701 */
702 702 if (nvlist_alloc(&nvroot, NV_UNIQUE_NAME, 0) != 0)
703 703 goto nomem;
704 704 if (nvlist_add_string(nvroot, ZPOOL_CONFIG_TYPE,
705 705 VDEV_TYPE_ROOT) != 0 ||
706 706 nvlist_add_uint64(nvroot, ZPOOL_CONFIG_ID, 0ULL) != 0 ||
707 707 nvlist_add_uint64(nvroot, ZPOOL_CONFIG_GUID, guid) != 0 ||
708 708 nvlist_add_nvlist_array(nvroot, ZPOOL_CONFIG_CHILDREN,
709 709 child, children) != 0) {
710 710 nvlist_free(nvroot);
711 711 goto nomem;
712 712 }
713 713
714 714 for (c = 0; c < children; c++)
715 715 nvlist_free(child[c]);
716 716 free(child);
717 717 children = 0;
718 718 child = NULL;
719 719
720 720 /*
721 721 * Go through and fix up any paths and/or devids based on our
722 722 * known list of vdev GUID -> path mappings.
723 723 */
724 724 if (fix_paths(nvroot, pl->names) != 0) {
725 725 nvlist_free(nvroot);
726 726 goto nomem;
727 727 }
728 728
729 729 /*
730 730 * Add the root vdev to this pool's configuration.
731 731 */
732 732 if (nvlist_add_nvlist(config, ZPOOL_CONFIG_VDEV_TREE,
733 733 nvroot) != 0) {
734 734 nvlist_free(nvroot);
735 735 goto nomem;
736 736 }
737 737 nvlist_free(nvroot);
738 738
739 739 /*
740 740 * zdb uses this path to report on active pools that were
741 741 * imported or created using -R.
742 742 */
743 743 if (active_ok)
744 744 goto add_pool;
745 745
746 746 /*
747 747 * Determine if this pool is currently active, in which case we
748 748 * can't actually import it.
749 749 */
750 750 verify(nvlist_lookup_string(config, ZPOOL_CONFIG_POOL_NAME,
751 751 &name) == 0);
752 752 verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID,
753 753 &guid) == 0);
754 754
755 755 if (pool_active(hdl, name, guid, &isactive) != 0)
756 756 goto error;
757 757
758 758 if (isactive) {
759 759 nvlist_free(config);
760 760 config = NULL;
761 761 continue;
762 762 }
763 763
764 764 if ((nvl = refresh_config(hdl, config)) == NULL) {
765 765 nvlist_free(config);
766 766 config = NULL;
767 767 continue;
768 768 }
769 769
770 770 nvlist_free(config);
771 771 config = nvl;
772 772
773 773 /*
774 774 * Go through and update the paths for spares, now that we have
775 775 * them.
776 776 */
777 777 verify(nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE,
778 778 &nvroot) == 0);
779 779 if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_SPARES,
780 780 &spares, &nspares) == 0) {
781 781 for (i = 0; i < nspares; i++) {
782 782 if (fix_paths(spares[i], pl->names) != 0)
783 783 goto nomem;
784 784 }
785 785 }
786 786
787 787 /*
788 788 * Update the paths for l2cache devices.
789 789 */
790 790 if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_L2CACHE,
791 791 &l2cache, &nl2cache) == 0) {
792 792 for (i = 0; i < nl2cache; i++) {
793 793 if (fix_paths(l2cache[i], pl->names) != 0)
794 794 goto nomem;
795 795 }
796 796 }
797 797
798 798 /*
799 799 * Restore the original information read from the actual label.
800 800 */
801 801 (void) nvlist_remove(config, ZPOOL_CONFIG_HOSTID,
802 802 DATA_TYPE_UINT64);
803 803 (void) nvlist_remove(config, ZPOOL_CONFIG_HOSTNAME,
804 804 DATA_TYPE_STRING);
805 805 if (hostid != 0) {
806 806 verify(nvlist_add_uint64(config, ZPOOL_CONFIG_HOSTID,
807 807 hostid) == 0);
808 808 verify(nvlist_add_string(config, ZPOOL_CONFIG_HOSTNAME,
809 809 hostname) == 0);
810 810 }
811 811
812 812 add_pool:
813 813 /*
814 814 * Add this pool to the list of configs.
815 815 */
816 816 verify(nvlist_lookup_string(config, ZPOOL_CONFIG_POOL_NAME,
817 817 &name) == 0);
818 818 if (nvlist_add_nvlist(ret, name, config) != 0)
819 819 goto nomem;
820 820
821 821 found_one = B_TRUE;
822 822 nvlist_free(config);
823 823 config = NULL;
824 824 }
825 825
826 826 if (!found_one) {
827 827 nvlist_free(ret);
828 828 ret = NULL;
829 829 }
830 830
831 831 return (ret);
832 832
833 833 nomem:
834 834 (void) no_memory(hdl);
835 835 error:
836 836 nvlist_free(config);
837 837 nvlist_free(ret);
838 838 for (c = 0; c < children; c++)
839 839 nvlist_free(child[c]);
840 840 free(child);
841 841
842 842 return (NULL);
843 843 }
844 844
845 845 /*
846 846 * Return the offset of the given label.
847 847 */
848 848 static uint64_t
849 849 label_offset(uint64_t size, int l)
850 850 {
851 851 ASSERT(P2PHASE_TYPED(size, sizeof (vdev_label_t), uint64_t) == 0);
852 852 return (l * sizeof (vdev_label_t) + (l < VDEV_LABELS / 2 ?
853 853 0 : size - VDEV_LABELS * sizeof (vdev_label_t)));
854 854 }
855 855
856 856 /*
857 857 * Given a file descriptor, read the label information and return an nvlist
858 858 * describing the configuration, if there is one.
859 859 */
860 860 int
861 861 zpool_read_label(int fd, nvlist_t **config)
862 862 {
863 863 struct stat64 statbuf;
864 864 int l;
865 865 vdev_label_t *label;
866 866 uint64_t state, txg, size;
867 867
868 868 *config = NULL;
869 869
870 870 if (fstat64(fd, &statbuf) == -1)
871 871 return (0);
872 872 size = P2ALIGN_TYPED(statbuf.st_size, sizeof (vdev_label_t), uint64_t);
873 873
874 874 if ((label = malloc(sizeof (vdev_label_t))) == NULL)
875 875 return (-1);
876 876
877 877 for (l = 0; l < VDEV_LABELS; l++) {
878 878 if (pread64(fd, label, sizeof (vdev_label_t),
879 879 label_offset(size, l)) != sizeof (vdev_label_t))
880 880 continue;
881 881
882 882 if (nvlist_unpack(label->vl_vdev_phys.vp_nvlist,
883 883 sizeof (label->vl_vdev_phys.vp_nvlist), config, 0) != 0)
884 884 continue;
885 885
886 886 if (nvlist_lookup_uint64(*config, ZPOOL_CONFIG_POOL_STATE,
887 887 &state) != 0 || state > POOL_STATE_L2CACHE) {
888 888 nvlist_free(*config);
889 889 continue;
890 890 }
891 891
892 892 if (state != POOL_STATE_SPARE && state != POOL_STATE_L2CACHE &&
893 893 (nvlist_lookup_uint64(*config, ZPOOL_CONFIG_POOL_TXG,
894 894 &txg) != 0 || txg == 0)) {
895 895 nvlist_free(*config);
896 896 continue;
897 897 }
898 898
899 899 free(label);
900 900 return (0);
901 901 }
902 902
903 903 free(label);
904 904 *config = NULL;
905 905 return (0);
906 906 }
907 907
908 908 typedef struct rdsk_node {
909 909 char *rn_name;
910 910 int rn_dfd;
911 911 libzfs_handle_t *rn_hdl;
912 912 nvlist_t *rn_config;
913 913 avl_tree_t *rn_avl;
914 914 avl_node_t rn_node;
915 915 boolean_t rn_nozpool;
916 916 } rdsk_node_t;
917 917
918 918 static int
919 919 slice_cache_compare(const void *arg1, const void *arg2)
920 920 {
921 921 const char *nm1 = ((rdsk_node_t *)arg1)->rn_name;
922 922 const char *nm2 = ((rdsk_node_t *)arg2)->rn_name;
923 923 char *nm1slice, *nm2slice;
924 924 int rv;
925 925
926 926 /*
927 927 * slices zero and two are the most likely to provide results,
928 928 * so put those first
929 929 */
930 930 nm1slice = strstr(nm1, "s0");
931 931 nm2slice = strstr(nm2, "s0");
932 932 if (nm1slice && !nm2slice) {
933 933 return (-1);
934 934 }
935 935 if (!nm1slice && nm2slice) {
936 936 return (1);
937 937 }
938 938 nm1slice = strstr(nm1, "s2");
939 939 nm2slice = strstr(nm2, "s2");
940 940 if (nm1slice && !nm2slice) {
941 941 return (-1);
942 942 }
943 943 if (!nm1slice && nm2slice) {
944 944 return (1);
945 945 }
946 946
947 947 rv = strcmp(nm1, nm2);
948 948 if (rv == 0)
949 949 return (0);
950 950 return (rv > 0 ? 1 : -1);
951 951 }
952 952
953 953 static void
954 954 check_one_slice(avl_tree_t *r, char *diskname, uint_t partno,
955 955 diskaddr_t size, uint_t blksz)
956 956 {
957 957 rdsk_node_t tmpnode;
958 958 rdsk_node_t *node;
959 959 char sname[MAXNAMELEN];
960 960
961 961 tmpnode.rn_name = &sname[0];
962 962 (void) snprintf(tmpnode.rn_name, MAXNAMELEN, "%s%u",
963 963 diskname, partno);
964 964 /*
965 965 * protect against division by zero for disk labels that
966 966 * contain a bogus sector size
967 967 */
968 968 if (blksz == 0)
969 969 blksz = DEV_BSIZE;
970 970 /* too small to contain a zpool? */
971 971 if ((size < (SPA_MINDEVSIZE / blksz)) &&
972 972 (node = avl_find(r, &tmpnode, NULL)))
973 973 node->rn_nozpool = B_TRUE;
974 974 }
975 975
976 976 static void
977 977 nozpool_all_slices(avl_tree_t *r, const char *sname)
978 978 {
979 979 char diskname[MAXNAMELEN];
980 980 char *ptr;
981 981 int i;
982 982
983 983 (void) strncpy(diskname, sname, MAXNAMELEN);
984 984 if (((ptr = strrchr(diskname, 's')) == NULL) &&
985 985 ((ptr = strrchr(diskname, 'p')) == NULL))
986 986 return;
987 987 ptr[0] = 's';
988 988 ptr[1] = '\0';
989 989 for (i = 0; i < NDKMAP; i++)
990 990 check_one_slice(r, diskname, i, 0, 1);
991 991 ptr[0] = 'p';
992 992 for (i = 0; i <= FD_NUMPART; i++)
993 993 check_one_slice(r, diskname, i, 0, 1);
994 994 }
995 995
996 996 static void
997 997 check_slices(avl_tree_t *r, int fd, const char *sname)
998 998 {
999 999 struct extvtoc vtoc;
1000 1000 struct dk_gpt *gpt;
1001 1001 char diskname[MAXNAMELEN];
1002 1002 char *ptr;
1003 1003 int i;
1004 1004
1005 1005 (void) strncpy(diskname, sname, MAXNAMELEN);
1006 1006 if ((ptr = strrchr(diskname, 's')) == NULL || !isdigit(ptr[1]))
1007 1007 return;
1008 1008 ptr[1] = '\0';
1009 1009
1010 1010 if (read_extvtoc(fd, &vtoc) >= 0) {
1011 1011 for (i = 0; i < NDKMAP; i++)
1012 1012 check_one_slice(r, diskname, i,
1013 1013 vtoc.v_part[i].p_size, vtoc.v_sectorsz);
1014 1014 } else if (efi_alloc_and_read(fd, &gpt) >= 0) {
1015 1015 /*
1016 1016 * on x86 we'll still have leftover links that point
1017 1017 * to slices s[9-15], so use NDKMAP instead
1018 1018 */
1019 1019 for (i = 0; i < NDKMAP; i++)
1020 1020 check_one_slice(r, diskname, i,
1021 1021 gpt->efi_parts[i].p_size, gpt->efi_lbasize);
1022 1022 /* nodes p[1-4] are never used with EFI labels */
1023 1023 ptr[0] = 'p';
1024 1024 for (i = 1; i <= FD_NUMPART; i++)
1025 1025 check_one_slice(r, diskname, i, 0, 1);
1026 1026 efi_free(gpt);
1027 1027 }
1028 1028 }
1029 1029
1030 1030 static void
1031 1031 zpool_open_func(void *arg)
1032 1032 {
1033 1033 rdsk_node_t *rn = arg;
1034 1034 struct stat64 statbuf;
1035 1035 nvlist_t *config;
1036 1036 int fd;
1037 1037
1038 1038 if (rn->rn_nozpool)
1039 1039 return;
1040 1040 if ((fd = openat64(rn->rn_dfd, rn->rn_name, O_RDONLY)) < 0) {
1041 1041 /* symlink to a device that's no longer there */
1042 1042 if (errno == ENOENT)
1043 1043 nozpool_all_slices(rn->rn_avl, rn->rn_name);
1044 1044 return;
1045 1045 }
1046 1046 /*
1047 1047 * Ignore failed stats. We only want regular
1048 1048 * files, character devs and block devs.
1049 1049 */
1050 1050 if (fstat64(fd, &statbuf) != 0 ||
1051 1051 (!S_ISREG(statbuf.st_mode) &&
1052 1052 !S_ISCHR(statbuf.st_mode) &&
1053 1053 !S_ISBLK(statbuf.st_mode))) {
1054 1054 (void) close(fd);
1055 1055 return;
1056 1056 }
1057 1057 /* this file is too small to hold a zpool */
1058 1058 if (S_ISREG(statbuf.st_mode) &&
1059 1059 statbuf.st_size < SPA_MINDEVSIZE) {
1060 1060 (void) close(fd);
1061 1061 return;
1062 1062 } else if (!S_ISREG(statbuf.st_mode)) {
1063 1063 /*
1064 1064 * Try to read the disk label first so we don't have to
1065 1065 * open a bunch of minor nodes that can't have a zpool.
1066 1066 */
1067 1067 check_slices(rn->rn_avl, fd, rn->rn_name);
1068 1068 }
1069 1069
1070 1070 if ((zpool_read_label(fd, &config)) != 0) {
1071 1071 (void) close(fd);
1072 1072 (void) no_memory(rn->rn_hdl);
1073 1073 return;
1074 1074 }
1075 1075 (void) close(fd);
1076 1076
1077 1077
1078 1078 rn->rn_config = config;
1079 1079 if (config != NULL) {
1080 1080 assert(rn->rn_nozpool == B_FALSE);
1081 1081 }
1082 1082 }
1083 1083
1084 1084 /*
1085 1085 * Given a file descriptor, clear (zero) the label information. This function
1086 1086 * is currently only used in the appliance stack as part of the ZFS sysevent
1087 1087 * module.
1088 1088 */
1089 1089 int
1090 1090 zpool_clear_label(int fd)
1091 1091 {
1092 1092 struct stat64 statbuf;
1093 1093 int l;
1094 1094 vdev_label_t *label;
1095 1095 uint64_t size;
1096 1096
1097 1097 if (fstat64(fd, &statbuf) == -1)
1098 1098 return (0);
1099 1099 size = P2ALIGN_TYPED(statbuf.st_size, sizeof (vdev_label_t), uint64_t);
1100 1100
1101 1101 if ((label = calloc(sizeof (vdev_label_t), 1)) == NULL)
1102 1102 return (-1);
1103 1103
1104 1104 for (l = 0; l < VDEV_LABELS; l++) {
1105 1105 if (pwrite64(fd, label, sizeof (vdev_label_t),
1106 1106 label_offset(size, l)) != sizeof (vdev_label_t))
1107 1107 return (-1);
1108 1108 }
1109 1109
1110 1110 free(label);
1111 1111 return (0);
1112 1112 }
1113 1113
1114 1114 /*
1115 1115 * Given a list of directories to search, find all pools stored on disk. This
1116 1116 * includes partial pools which are not available to import. If no args are
1117 1117 * given (argc is 0), then the default directory (/dev/dsk) is searched.
1118 1118 * poolname or guid (but not both) are provided by the caller when trying
1119 1119 * to import a specific pool.
1120 1120 */
1121 1121 static nvlist_t *
1122 1122 zpool_find_import_impl(libzfs_handle_t *hdl, importargs_t *iarg)
1123 1123 {
1124 1124 int i, dirs = iarg->paths;
1125 1125 DIR *dirp = NULL;
1126 1126 struct dirent64 *dp;
1127 1127 char path[MAXPATHLEN];
1128 1128 char *end, **dir = iarg->path;
1129 1129 size_t pathleft;
1130 1130 nvlist_t *ret = NULL;
1131 1131 static char *default_dir = "/dev/dsk";
1132 1132 pool_list_t pools = { 0 };
1133 1133 pool_entry_t *pe, *penext;
1134 1134 vdev_entry_t *ve, *venext;
1135 1135 config_entry_t *ce, *cenext;
1136 1136 name_entry_t *ne, *nenext;
1137 1137 avl_tree_t slice_cache;
1138 1138 rdsk_node_t *slice;
1139 1139 void *cookie;
1140 1140
1141 1141 if (dirs == 0) {
1142 1142 dirs = 1;
1143 1143 dir = &default_dir;
1144 1144 }
1145 1145
1146 1146 /*
1147 1147 * Go through and read the label configuration information from every
1148 1148 * possible device, organizing the information according to pool GUID
1149 1149 * and toplevel GUID.
1150 1150 */
1151 1151 for (i = 0; i < dirs; i++) {
1152 1152 tpool_t *t;
1153 1153 char *rdsk;
1154 1154 int dfd;
1155 1155
1156 1156 /* use realpath to normalize the path */
1157 1157 if (realpath(dir[i], path) == 0) {
1158 1158 (void) zfs_error_fmt(hdl, EZFS_BADPATH,
1159 1159 dgettext(TEXT_DOMAIN, "cannot open '%s'"), dir[i]);
1160 1160 goto error;
1161 1161 }
1162 1162 end = &path[strlen(path)];
1163 1163 *end++ = '/';
1164 1164 *end = 0;
1165 1165 pathleft = &path[sizeof (path)] - end;
1166 1166
1167 1167 /*
1168 1168 * Using raw devices instead of block devices when we're
1169 1169 * reading the labels skips a bunch of slow operations during
1170 1170 * close(2) processing, so we replace /dev/dsk with /dev/rdsk.
1171 1171 */
1172 1172 if (strcmp(path, "/dev/dsk/") == 0)
1173 1173 rdsk = "/dev/rdsk/";
1174 1174 else
1175 1175 rdsk = path;
1176 1176
1177 1177 if ((dfd = open64(rdsk, O_RDONLY)) < 0 ||
1178 1178 (dirp = fdopendir(dfd)) == NULL) {
1179 1179 zfs_error_aux(hdl, strerror(errno));
1180 1180 (void) zfs_error_fmt(hdl, EZFS_BADPATH,
1181 1181 dgettext(TEXT_DOMAIN, "cannot open '%s'"),
1182 1182 rdsk);
1183 1183 goto error;
1184 1184 }
1185 1185
1186 1186 avl_create(&slice_cache, slice_cache_compare,
1187 1187 sizeof (rdsk_node_t), offsetof(rdsk_node_t, rn_node));
1188 1188 /*
1189 1189 * This is not MT-safe, but we have no MT consumers of libzfs
1190 1190 */
1191 1191 while ((dp = readdir64(dirp)) != NULL) {
1192 1192 const char *name = dp->d_name;
1193 1193 if (name[0] == '.' &&
1194 1194 (name[1] == 0 || (name[1] == '.' && name[2] == 0)))
1195 1195 continue;
1196 1196
1197 1197 slice = zfs_alloc(hdl, sizeof (rdsk_node_t));
1198 1198 slice->rn_name = zfs_strdup(hdl, name);
1199 1199 slice->rn_avl = &slice_cache;
1200 1200 slice->rn_dfd = dfd;
1201 1201 slice->rn_hdl = hdl;
1202 1202 slice->rn_nozpool = B_FALSE;
1203 1203 avl_add(&slice_cache, slice);
1204 1204 }
1205 1205 /*
1206 1206 * create a thread pool to do all of this in parallel;
1207 1207 * rn_nozpool is not protected, so this is racy in that
1208 1208 * multiple tasks could decide that the same slice can
1209 1209 * not hold a zpool, which is benign. Also choose
1210 1210 * double the number of processors; we hold a lot of
1211 1211 * locks in the kernel, so going beyond this doesn't
1212 1212 * buy us much.
1213 1213 */
1214 1214 t = tpool_create(1, 2 * sysconf(_SC_NPROCESSORS_ONLN),
1215 1215 0, NULL);
1216 1216 for (slice = avl_first(&slice_cache); slice;
1217 1217 (slice = avl_walk(&slice_cache, slice,
1218 1218 AVL_AFTER)))
1219 1219 (void) tpool_dispatch(t, zpool_open_func, slice);
1220 1220 tpool_wait(t);
1221 1221 tpool_destroy(t);
1222 1222
1223 1223 cookie = NULL;
1224 1224 while ((slice = avl_destroy_nodes(&slice_cache,
1225 1225 &cookie)) != NULL) {
1226 1226 if (slice->rn_config != NULL) {
1227 1227 nvlist_t *config = slice->rn_config;
1228 1228 boolean_t matched = B_TRUE;
1229 1229
1230 1230 if (iarg->poolname != NULL) {
1231 1231 char *pname;
1232 1232
1233 1233 matched = nvlist_lookup_string(config,
1234 1234 ZPOOL_CONFIG_POOL_NAME,
1235 1235 &pname) == 0 &&
1236 1236 strcmp(iarg->poolname, pname) == 0;
1237 1237 } else if (iarg->guid != 0) {
1238 1238 uint64_t this_guid;
1239 1239
1240 1240 matched = nvlist_lookup_uint64(config,
1241 1241 ZPOOL_CONFIG_POOL_GUID,
1242 1242 &this_guid) == 0 &&
1243 1243 iarg->guid == this_guid;
1244 1244 }
1245 1245 if (!matched) {
1246 1246 nvlist_free(config);
1247 1247 config = NULL;
1248 1248 continue;
1249 1249 }
1250 1250 /* use the non-raw path for the config */
1251 1251 (void) strlcpy(end, slice->rn_name, pathleft);
1252 1252 if (add_config(hdl, &pools, path, config) != 0)
1253 1253 goto error;
1254 1254 }
1255 1255 free(slice->rn_name);
1256 1256 free(slice);
1257 1257 }
1258 1258 avl_destroy(&slice_cache);
1259 1259
1260 1260 (void) closedir(dirp);
1261 1261 dirp = NULL;
1262 1262 }
1263 1263
1264 1264 ret = get_configs(hdl, &pools, iarg->can_be_active);
1265 1265
1266 1266 error:
1267 1267 for (pe = pools.pools; pe != NULL; pe = penext) {
1268 1268 penext = pe->pe_next;
1269 1269 for (ve = pe->pe_vdevs; ve != NULL; ve = venext) {
1270 1270 venext = ve->ve_next;
1271 1271 for (ce = ve->ve_configs; ce != NULL; ce = cenext) {
1272 1272 cenext = ce->ce_next;
1273 1273 if (ce->ce_config)
1274 1274 nvlist_free(ce->ce_config);
1275 1275 free(ce);
1276 1276 }
1277 1277 free(ve);
1278 1278 }
1279 1279 free(pe);
1280 1280 }
1281 1281
1282 1282 for (ne = pools.names; ne != NULL; ne = nenext) {
1283 1283 nenext = ne->ne_next;
1284 1284 if (ne->ne_name)
1285 1285 free(ne->ne_name);
1286 1286 free(ne);
1287 1287 }
1288 1288
1289 1289 if (dirp)
1290 1290 (void) closedir(dirp);
1291 1291
1292 1292 return (ret);
1293 1293 }
1294 1294
1295 1295 nvlist_t *
1296 1296 zpool_find_import(libzfs_handle_t *hdl, int argc, char **argv)
1297 1297 {
1298 1298 importargs_t iarg = { 0 };
1299 1299
1300 1300 iarg.paths = argc;
1301 1301 iarg.path = argv;
1302 1302
1303 1303 return (zpool_find_import_impl(hdl, &iarg));
1304 1304 }
1305 1305
1306 1306 /*
1307 1307 * Given a cache file, return the contents as a list of importable pools.
1308 1308 * poolname or guid (but not both) are provided by the caller when trying
1309 1309 * to import a specific pool.
1310 1310 */
1311 1311 nvlist_t *
1312 1312 zpool_find_import_cached(libzfs_handle_t *hdl, const char *cachefile,
1313 1313 char *poolname, uint64_t guid)
1314 1314 {
1315 1315 char *buf;
1316 1316 int fd;
1317 1317 struct stat64 statbuf;
1318 1318 nvlist_t *raw, *src, *dst;
1319 1319 nvlist_t *pools;
1320 1320 nvpair_t *elem;
1321 1321 char *name;
1322 1322 uint64_t this_guid;
1323 1323 boolean_t active;
1324 1324
1325 1325 verify(poolname == NULL || guid == 0);
1326 1326
1327 1327 if ((fd = open(cachefile, O_RDONLY)) < 0) {
1328 1328 zfs_error_aux(hdl, "%s", strerror(errno));
1329 1329 (void) zfs_error(hdl, EZFS_BADCACHE,
1330 1330 dgettext(TEXT_DOMAIN, "failed to open cache file"));
1331 1331 return (NULL);
1332 1332 }
1333 1333
1334 1334 if (fstat64(fd, &statbuf) != 0) {
1335 1335 zfs_error_aux(hdl, "%s", strerror(errno));
1336 1336 (void) close(fd);
1337 1337 (void) zfs_error(hdl, EZFS_BADCACHE,
1338 1338 dgettext(TEXT_DOMAIN, "failed to get size of cache file"));
1339 1339 return (NULL);
1340 1340 }
1341 1341
1342 1342 if ((buf = zfs_alloc(hdl, statbuf.st_size)) == NULL) {
1343 1343 (void) close(fd);
1344 1344 return (NULL);
1345 1345 }
1346 1346
1347 1347 if (read(fd, buf, statbuf.st_size) != statbuf.st_size) {
1348 1348 (void) close(fd);
1349 1349 free(buf);
1350 1350 (void) zfs_error(hdl, EZFS_BADCACHE,
1351 1351 dgettext(TEXT_DOMAIN,
1352 1352 "failed to read cache file contents"));
1353 1353 return (NULL);
1354 1354 }
1355 1355
1356 1356 (void) close(fd);
1357 1357
1358 1358 if (nvlist_unpack(buf, statbuf.st_size, &raw, 0) != 0) {
1359 1359 free(buf);
1360 1360 (void) zfs_error(hdl, EZFS_BADCACHE,
1361 1361 dgettext(TEXT_DOMAIN,
1362 1362 "invalid or corrupt cache file contents"));
1363 1363 return (NULL);
1364 1364 }
1365 1365
1366 1366 free(buf);
1367 1367
1368 1368 /*
1369 1369 * Go through and get the current state of the pools and refresh their
1370 1370 * state.
1371 1371 */
1372 1372 if (nvlist_alloc(&pools, 0, 0) != 0) {
1373 1373 (void) no_memory(hdl);
1374 1374 nvlist_free(raw);
1375 1375 return (NULL);
1376 1376 }
1377 1377
1378 1378 elem = NULL;
1379 1379 while ((elem = nvlist_next_nvpair(raw, elem)) != NULL) {
1380 1380 verify(nvpair_value_nvlist(elem, &src) == 0);
1381 1381
1382 1382 verify(nvlist_lookup_string(src, ZPOOL_CONFIG_POOL_NAME,
1383 1383 &name) == 0);
1384 1384 if (poolname != NULL && strcmp(poolname, name) != 0)
1385 1385 continue;
1386 1386
1387 1387 verify(nvlist_lookup_uint64(src, ZPOOL_CONFIG_POOL_GUID,
1388 1388 &this_guid) == 0);
1389 1389 if (guid != 0) {
1390 1390 verify(nvlist_lookup_uint64(src, ZPOOL_CONFIG_POOL_GUID,
1391 1391 &this_guid) == 0);
1392 1392 if (guid != this_guid)
1393 1393 continue;
1394 1394 }
1395 1395
1396 1396 if (pool_active(hdl, name, this_guid, &active) != 0) {
1397 1397 nvlist_free(raw);
1398 1398 nvlist_free(pools);
1399 1399 return (NULL);
1400 1400 }
1401 1401
1402 1402 if (active)
1403 1403 continue;
1404 1404
1405 1405 if ((dst = refresh_config(hdl, src)) == NULL) {
1406 1406 nvlist_free(raw);
1407 1407 nvlist_free(pools);
1408 1408 return (NULL);
1409 1409 }
1410 1410
1411 1411 if (nvlist_add_nvlist(pools, nvpair_name(elem), dst) != 0) {
1412 1412 (void) no_memory(hdl);
1413 1413 nvlist_free(dst);
1414 1414 nvlist_free(raw);
1415 1415 nvlist_free(pools);
1416 1416 return (NULL);
1417 1417 }
1418 1418 nvlist_free(dst);
1419 1419 }
1420 1420
1421 1421 nvlist_free(raw);
1422 1422 return (pools);
1423 1423 }
1424 1424
1425 1425 static int
1426 1426 name_or_guid_exists(zpool_handle_t *zhp, void *data)
1427 1427 {
1428 1428 importargs_t *import = data;
1429 1429 int found = 0;
1430 1430
1431 1431 if (import->poolname != NULL) {
1432 1432 char *pool_name;
1433 1433
1434 1434 verify(nvlist_lookup_string(zhp->zpool_config,
1435 1435 ZPOOL_CONFIG_POOL_NAME, &pool_name) == 0);
1436 1436 if (strcmp(pool_name, import->poolname) == 0)
1437 1437 found = 1;
1438 1438 } else {
1439 1439 uint64_t pool_guid;
1440 1440
1441 1441 verify(nvlist_lookup_uint64(zhp->zpool_config,
1442 1442 ZPOOL_CONFIG_POOL_GUID, &pool_guid) == 0);
1443 1443 if (pool_guid == import->guid)
1444 1444 found = 1;
1445 1445 }
1446 1446
1447 1447 zpool_close(zhp);
1448 1448 return (found);
1449 1449 }
1450 1450
1451 1451 nvlist_t *
1452 1452 zpool_search_import(libzfs_handle_t *hdl, importargs_t *import)
1453 1453 {
1454 1454 verify(import->poolname == NULL || import->guid == 0);
1455 1455
1456 1456 if (import->unique)
1457 1457 import->exists = zpool_iter(hdl, name_or_guid_exists, import);
1458 1458
1459 1459 if (import->cachefile != NULL)
1460 1460 return (zpool_find_import_cached(hdl, import->cachefile,
1461 1461 import->poolname, import->guid));
1462 1462
1463 1463 return (zpool_find_import_impl(hdl, import));
1464 1464 }
1465 1465
1466 1466 boolean_t
1467 1467 find_guid(nvlist_t *nv, uint64_t guid)
1468 1468 {
1469 1469 uint64_t tmp;
1470 1470 nvlist_t **child;
1471 1471 uint_t c, children;
1472 1472
1473 1473 verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID, &tmp) == 0);
1474 1474 if (tmp == guid)
1475 1475 return (B_TRUE);
1476 1476
1477 1477 if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN,
1478 1478 &child, &children) == 0) {
1479 1479 for (c = 0; c < children; c++)
1480 1480 if (find_guid(child[c], guid))
1481 1481 return (B_TRUE);
1482 1482 }
1483 1483
1484 1484 return (B_FALSE);
1485 1485 }
1486 1486
1487 1487 typedef struct aux_cbdata {
1488 1488 const char *cb_type;
1489 1489 uint64_t cb_guid;
1490 1490 zpool_handle_t *cb_zhp;
1491 1491 } aux_cbdata_t;
1492 1492
1493 1493 static int
1494 1494 find_aux(zpool_handle_t *zhp, void *data)
1495 1495 {
1496 1496 aux_cbdata_t *cbp = data;
1497 1497 nvlist_t **list;
1498 1498 uint_t i, count;
1499 1499 uint64_t guid;
1500 1500 nvlist_t *nvroot;
1501 1501
1502 1502 verify(nvlist_lookup_nvlist(zhp->zpool_config, ZPOOL_CONFIG_VDEV_TREE,
1503 1503 &nvroot) == 0);
1504 1504
1505 1505 if (nvlist_lookup_nvlist_array(nvroot, cbp->cb_type,
1506 1506 &list, &count) == 0) {
1507 1507 for (i = 0; i < count; i++) {
1508 1508 verify(nvlist_lookup_uint64(list[i],
1509 1509 ZPOOL_CONFIG_GUID, &guid) == 0);
1510 1510 if (guid == cbp->cb_guid) {
1511 1511 cbp->cb_zhp = zhp;
1512 1512 return (1);
1513 1513 }
1514 1514 }
1515 1515 }
1516 1516
1517 1517 zpool_close(zhp);
1518 1518 return (0);
1519 1519 }
1520 1520
1521 1521 /*
1522 1522 * Determines if the pool is in use. If so, it returns true and the state of
1523 1523 * the pool as well as the name of the pool. Both strings are allocated and
1524 1524 * must be freed by the caller.
1525 1525 */
1526 1526 int
1527 1527 zpool_in_use(libzfs_handle_t *hdl, int fd, pool_state_t *state, char **namestr,
1528 1528 boolean_t *inuse)
1529 1529 {
1530 1530 nvlist_t *config;
1531 1531 char *name;
1532 1532 boolean_t ret;
1533 1533 uint64_t guid, vdev_guid;
1534 1534 zpool_handle_t *zhp;
1535 1535 nvlist_t *pool_config;
1536 1536 uint64_t stateval, isspare;
1537 1537 aux_cbdata_t cb = { 0 };
1538 1538 boolean_t isactive;
1539 1539
1540 1540 *inuse = B_FALSE;
1541 1541
1542 1542 if (zpool_read_label(fd, &config) != 0) {
1543 1543 (void) no_memory(hdl);
1544 1544 return (-1);
1545 1545 }
1546 1546
1547 1547 if (config == NULL)
1548 1548 return (0);
1549 1549
1550 1550 verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_STATE,
1551 1551 &stateval) == 0);
1552 1552 verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_GUID,
1553 1553 &vdev_guid) == 0);
1554 1554
1555 1555 if (stateval != POOL_STATE_SPARE && stateval != POOL_STATE_L2CACHE) {
1556 1556 verify(nvlist_lookup_string(config, ZPOOL_CONFIG_POOL_NAME,
1557 1557 &name) == 0);
1558 1558 verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID,
1559 1559 &guid) == 0);
1560 1560 }
↓ open down ↓ |
1560 lines elided |
↑ open up ↑ |
1561 1561
1562 1562 switch (stateval) {
1563 1563 case POOL_STATE_EXPORTED:
1564 1564 /*
1565 1565 * A pool with an exported state may in fact be imported
1566 1566 * read-only, so check the in-core state to see if it's
1567 1567 * active and imported read-only. If it is, set
1568 1568 * its state to active.
1569 1569 */
1570 1570 if (pool_active(hdl, name, guid, &isactive) == 0 && isactive &&
1571 - (zhp = zpool_open_canfail(hdl, name)) != NULL &&
1572 - zpool_get_prop_int(zhp, ZPOOL_PROP_READONLY, NULL))
1573 - stateval = POOL_STATE_ACTIVE;
1571 + (zhp = zpool_open_canfail(hdl, name)) != NULL) {
1572 + if (zpool_get_prop_int(zhp, ZPOOL_PROP_READONLY, NULL))
1573 + stateval = POOL_STATE_ACTIVE;
1574 +
1575 + zpool_close(zhp);
1576 + }
1574 1577
1575 1578 ret = B_TRUE;
1576 1579 break;
1577 1580
1578 1581 case POOL_STATE_ACTIVE:
1579 1582 /*
1580 1583 * For an active pool, we have to determine if it's really part
1581 1584 * of a currently active pool (in which case the pool will exist
1582 1585 * and the guid will be the same), or whether it's part of an
1583 1586 * active pool that was disconnected without being explicitly
1584 1587 * exported.
1585 1588 */
1586 1589 if (pool_active(hdl, name, guid, &isactive) != 0) {
1587 1590 nvlist_free(config);
1588 1591 return (-1);
1589 1592 }
1590 1593
1591 1594 if (isactive) {
1592 1595 /*
1593 1596 * Because the device may have been removed while
1594 1597 * offlined, we only report it as active if the vdev is
1595 1598 * still present in the config. Otherwise, pretend like
1596 1599 * it's not in use.
1597 1600 */
1598 1601 if ((zhp = zpool_open_canfail(hdl, name)) != NULL &&
1599 1602 (pool_config = zpool_get_config(zhp, NULL))
1600 1603 != NULL) {
1601 1604 nvlist_t *nvroot;
1602 1605
1603 1606 verify(nvlist_lookup_nvlist(pool_config,
1604 1607 ZPOOL_CONFIG_VDEV_TREE, &nvroot) == 0);
1605 1608 ret = find_guid(nvroot, vdev_guid);
1606 1609 } else {
1607 1610 ret = B_FALSE;
1608 1611 }
1609 1612
1610 1613 /*
1611 1614 * If this is an active spare within another pool, we
1612 1615 * treat it like an unused hot spare. This allows the
1613 1616 * user to create a pool with a hot spare that currently
1614 1617 * in use within another pool. Since we return B_TRUE,
1615 1618 * libdiskmgt will continue to prevent generic consumers
1616 1619 * from using the device.
1617 1620 */
1618 1621 if (ret && nvlist_lookup_uint64(config,
1619 1622 ZPOOL_CONFIG_IS_SPARE, &isspare) == 0 && isspare)
1620 1623 stateval = POOL_STATE_SPARE;
1621 1624
1622 1625 if (zhp != NULL)
1623 1626 zpool_close(zhp);
1624 1627 } else {
1625 1628 stateval = POOL_STATE_POTENTIALLY_ACTIVE;
1626 1629 ret = B_TRUE;
1627 1630 }
1628 1631 break;
1629 1632
1630 1633 case POOL_STATE_SPARE:
1631 1634 /*
1632 1635 * For a hot spare, it can be either definitively in use, or
1633 1636 * potentially active. To determine if it's in use, we iterate
1634 1637 * over all pools in the system and search for one with a spare
1635 1638 * with a matching guid.
1636 1639 *
1637 1640 * Due to the shared nature of spares, we don't actually report
1638 1641 * the potentially active case as in use. This means the user
1639 1642 * can freely create pools on the hot spares of exported pools,
1640 1643 * but to do otherwise makes the resulting code complicated, and
1641 1644 * we end up having to deal with this case anyway.
1642 1645 */
1643 1646 cb.cb_zhp = NULL;
1644 1647 cb.cb_guid = vdev_guid;
1645 1648 cb.cb_type = ZPOOL_CONFIG_SPARES;
1646 1649 if (zpool_iter(hdl, find_aux, &cb) == 1) {
1647 1650 name = (char *)zpool_get_name(cb.cb_zhp);
1648 1651 ret = TRUE;
1649 1652 } else {
1650 1653 ret = FALSE;
1651 1654 }
1652 1655 break;
1653 1656
1654 1657 case POOL_STATE_L2CACHE:
1655 1658
1656 1659 /*
1657 1660 * Check if any pool is currently using this l2cache device.
1658 1661 */
1659 1662 cb.cb_zhp = NULL;
1660 1663 cb.cb_guid = vdev_guid;
1661 1664 cb.cb_type = ZPOOL_CONFIG_L2CACHE;
1662 1665 if (zpool_iter(hdl, find_aux, &cb) == 1) {
1663 1666 name = (char *)zpool_get_name(cb.cb_zhp);
1664 1667 ret = TRUE;
1665 1668 } else {
1666 1669 ret = FALSE;
1667 1670 }
1668 1671 break;
1669 1672
1670 1673 default:
1671 1674 ret = B_FALSE;
1672 1675 }
1673 1676
1674 1677
1675 1678 if (ret) {
1676 1679 if ((*namestr = zfs_strdup(hdl, name)) == NULL) {
1677 1680 if (cb.cb_zhp)
1678 1681 zpool_close(cb.cb_zhp);
1679 1682 nvlist_free(config);
1680 1683 return (-1);
1681 1684 }
1682 1685 *state = (pool_state_t)stateval;
1683 1686 }
1684 1687
1685 1688 if (cb.cb_zhp)
1686 1689 zpool_close(cb.cb_zhp);
1687 1690
1688 1691 nvlist_free(config);
1689 1692 *inuse = ret;
1690 1693 return (0);
1691 1694 }
↓ open down ↓ |
108 lines elided |
↑ open up ↑ |
XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX