?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
3090 vdev_reopen() during reguid causes vdev to be treated as corrupt
[unleashed.git] / usr / src / lib / libzfs / common / libzfs_import.c
blob36d7420c3bb1d7193490f396cda1151df1b7e463
1 /*
2 * CDDL HEADER START
3 *
4 * The contents of this file are subject to the terms of the
5 * Common Development and Distribution License (the "License").
6 * You may not use this file except in compliance with the License.
7 *
8 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
9 * or http://www.opensolaris.org/os/licensing.
10 * See the License for the specific language governing permissions
11 * and limitations under the License.
12 *
13 * When distributing Covered Code, include this CDDL HEADER in each
14 * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
15 * If applicable, add the following below this CDDL HEADER, with the
16 * fields enclosed by brackets "[]" replaced with your own identifying
17 * information: Portions Copyright [yyyy] [name of copyright owner]
18 *
19 * CDDL HEADER END
20 */
21 /*
22 * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
23 * Copyright 2011 Nexenta Systems, Inc. All rights reserved.
24 * Copyright (c) 2012 by Delphix. All rights reserved.
25 */
26
27 /*
28 * Pool import support functions.
29 *
30 * To import a pool, we rely on reading the configuration information from the
31 * ZFS label of each device. If we successfully read the label, then we
32 * organize the configuration information in the following hierarchy:
33 *
34 * pool guid -> toplevel vdev guid -> label txg
35 *
36 * Duplicate entries matching this same tuple will be discarded. Once we have
37 * examined every device, we pick the best label txg config for each toplevel
38 * vdev. We then arrange these toplevel vdevs into a complete pool config, and
39 * update any paths that have changed. Finally, we attempt to import the pool
40 * using our derived config, and record the results.
41 */
42
43 #include <ctype.h>
44 #include <devid.h>
45 #include <dirent.h>
46 #include <errno.h>
47 #include <libintl.h>
48 #include <stddef.h>
49 #include <stdlib.h>
50 #include <string.h>
51 #include <sys/stat.h>
52 #include <unistd.h>
53 #include <fcntl.h>
54 #include <sys/vtoc.h>
55 #include <sys/dktp/fdisk.h>
56 #include <sys/efi_partition.h>
57 #include <thread_pool.h>
58
59 #include <sys/vdev_impl.h>
60
61 #include "libzfs.h"
62 #include "libzfs_impl.h"
63
64 /*
65 * Intermediate structures used to gather configuration information.
66 */
67 typedef struct config_entry {
68 uint64_t ce_txg;
69 nvlist_t *ce_config;
70 struct config_entry *ce_next;
71 } config_entry_t;
72
73 typedef struct vdev_entry {
74 uint64_t ve_guid;
75 config_entry_t *ve_configs;
76 struct vdev_entry *ve_next;
77 } vdev_entry_t;
78
79 typedef struct pool_entry {
80 uint64_t pe_guid;
81 vdev_entry_t *pe_vdevs;
82 struct pool_entry *pe_next;
83 } pool_entry_t;
84
85 typedef struct name_entry {
86 char *ne_name;
87 uint64_t ne_guid;
88 struct name_entry *ne_next;
89 } name_entry_t;
90
91 typedef struct pool_list {
92 pool_entry_t *pools;
93 name_entry_t *names;
94 } pool_list_t;
95
96 static char *
97 get_devid(const char *path)
98 {
99 int fd;
100 ddi_devid_t devid;
101 char *minor, *ret;
102
103 if ((fd = open(path, O_RDONLY)) < 0)
104 return (NULL);
105
106 minor = NULL;
107 ret = NULL;
108 if (devid_get(fd, &devid) == 0) {
109 if (devid_get_minor_name(fd, &minor) == 0)
110 ret = devid_str_encode(devid, minor);
111 if (minor != NULL)
112 devid_str_free(minor);
113 devid_free(devid);
114 }
115 (void) close(fd);
116
117 return (ret);
118 }
119
120
121 /*
122 * Go through and fix up any path and/or devid information for the given vdev
123 * configuration.
124 */
125 static int
126 fix_paths(nvlist_t *nv, name_entry_t *names)
127 {
128 nvlist_t **child;
129 uint_t c, children;
130 uint64_t guid;
131 name_entry_t *ne, *best;
132 char *path, *devid;
133 int matched;
134
135 if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN,
136 &child, &children) == 0) {
137 for (c = 0; c < children; c++)
138 if (fix_paths(child[c], names) != 0)
139 return (-1);
140 return (0);
141 }
142
143 /*
144 * This is a leaf (file or disk) vdev. In either case, go through
145 * the name list and see if we find a matching guid. If so, replace
146 * the path and see if we can calculate a new devid.
147 *
148 * There may be multiple names associated with a particular guid, in
149 * which case we have overlapping slices or multiple paths to the same
150 * disk. If this is the case, then we want to pick the path that is
151 * the most similar to the original, where "most similar" is the number
152 * of matching characters starting from the end of the path. This will
153 * preserve slice numbers even if the disks have been reorganized, and
154 * will also catch preferred disk names if multiple paths exist.
155 */
156 verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID, &guid) == 0);
157 if (nvlist_lookup_string(nv, ZPOOL_CONFIG_PATH, &path) != 0)
158 path = NULL;
159
160 matched = 0;
161 best = NULL;
162 for (ne = names; ne != NULL; ne = ne->ne_next) {
163 if (ne->ne_guid == guid) {
164 const char *src, *dst;
165 int count;
166
167 if (path == NULL) {
168 best = ne;
169 break;
170 }
171
172 src = ne->ne_name + strlen(ne->ne_name) - 1;
173 dst = path + strlen(path) - 1;
174 for (count = 0; src >= ne->ne_name && dst >= path;
175 src--, dst--, count++)
176 if (*src != *dst)
177 break;
178
179 /*
180 * At this point, 'count' is the number of characters
181 * matched from the end.
182 */
183 if (count > matched || best == NULL) {
184 best = ne;
185 matched = count;
186 }
187 }
188 }
189
190 if (best == NULL)
191 return (0);
192
193 if (nvlist_add_string(nv, ZPOOL_CONFIG_PATH, best->ne_name) != 0)
194 return (-1);
195
196 if ((devid = get_devid(best->ne_name)) == NULL) {
197 (void) nvlist_remove_all(nv, ZPOOL_CONFIG_DEVID);
198 } else {
199 if (nvlist_add_string(nv, ZPOOL_CONFIG_DEVID, devid) != 0)
200 return (-1);
201 devid_str_free(devid);
202 }
203
204 return (0);
205 }
206
207 /*
208 * Add the given configuration to the list of known devices.
209 */
210 static int
211 add_config(libzfs_handle_t *hdl, pool_list_t *pl, const char *path,
212 nvlist_t *config)
213 {
214 uint64_t pool_guid, vdev_guid, top_guid, txg, state;
215 pool_entry_t *pe;
216 vdev_entry_t *ve;
217 config_entry_t *ce;
218 name_entry_t *ne;
219
220 /*
221 * If this is a hot spare not currently in use or level 2 cache
222 * device, add it to the list of names to translate, but don't do
223 * anything else.
224 */
225 if (nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_STATE,
226 &state) == 0 &&
227 (state == POOL_STATE_SPARE || state == POOL_STATE_L2CACHE) &&
228 nvlist_lookup_uint64(config, ZPOOL_CONFIG_GUID, &vdev_guid) == 0) {
229 if ((ne = zfs_alloc(hdl, sizeof (name_entry_t))) == NULL)
230 return (-1);
231
232 if ((ne->ne_name = zfs_strdup(hdl, path)) == NULL) {
233 free(ne);
234 return (-1);
235 }
236 ne->ne_guid = vdev_guid;
237 ne->ne_next = pl->names;
238 pl->names = ne;
239 return (0);
240 }
241
242 /*
243 * If we have a valid config but cannot read any of these fields, then
244 * it means we have a half-initialized label. In vdev_label_init()
245 * we write a label with txg == 0 so that we can identify the device
246 * in case the user refers to the same disk later on. If we fail to
247 * create the pool, we'll be left with a label in this state
248 * which should not be considered part of a valid pool.
249 */
250 if (nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID,
251 &pool_guid) != 0 ||
252 nvlist_lookup_uint64(config, ZPOOL_CONFIG_GUID,
253 &vdev_guid) != 0 ||
254 nvlist_lookup_uint64(config, ZPOOL_CONFIG_TOP_GUID,
255 &top_guid) != 0 ||
256 nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_TXG,
257 &txg) != 0 || txg == 0) {
258 nvlist_free(config);
259 return (0);
260 }
261
262 /*
263 * First, see if we know about this pool. If not, then add it to the
264 * list of known pools.
265 */
266 for (pe = pl->pools; pe != NULL; pe = pe->pe_next) {
267 if (pe->pe_guid == pool_guid)
268 break;
269 }
270
271 if (pe == NULL) {
272 if ((pe = zfs_alloc(hdl, sizeof (pool_entry_t))) == NULL) {
273 nvlist_free(config);
274 return (-1);
275 }
276 pe->pe_guid = pool_guid;
277 pe->pe_next = pl->pools;
278 pl->pools = pe;
279 }
280
281 /*
282 * Second, see if we know about this toplevel vdev. Add it if its
283 * missing.
284 */
285 for (ve = pe->pe_vdevs; ve != NULL; ve = ve->ve_next) {
286 if (ve->ve_guid == top_guid)
287 break;
288 }
289
290 if (ve == NULL) {
291 if ((ve = zfs_alloc(hdl, sizeof (vdev_entry_t))) == NULL) {
292 nvlist_free(config);
293 return (-1);
294 }
295 ve->ve_guid = top_guid;
296 ve->ve_next = pe->pe_vdevs;
297 pe->pe_vdevs = ve;
298 }
299
300 /*
301 * Third, see if we have a config with a matching transaction group. If
302 * so, then we do nothing. Otherwise, add it to the list of known
303 * configs.
304 */
305 for (ce = ve->ve_configs; ce != NULL; ce = ce->ce_next) {
306 if (ce->ce_txg == txg)
307 break;
308 }
309
310 if (ce == NULL) {
311 if ((ce = zfs_alloc(hdl, sizeof (config_entry_t))) == NULL) {
312 nvlist_free(config);
313 return (-1);
314 }
315 ce->ce_txg = txg;
316 ce->ce_config = config;
317 ce->ce_next = ve->ve_configs;
318 ve->ve_configs = ce;
319 } else {
320 nvlist_free(config);
321 }
322
323 /*
324 * At this point we've successfully added our config to the list of
325 * known configs. The last thing to do is add the vdev guid -> path
326 * mappings so that we can fix up the configuration as necessary before
327 * doing the import.
328 */
329 if ((ne = zfs_alloc(hdl, sizeof (name_entry_t))) == NULL)
330 return (-1);
331
332 if ((ne->ne_name = zfs_strdup(hdl, path)) == NULL) {
333 free(ne);
334 return (-1);
335 }
336
337 ne->ne_guid = vdev_guid;
338 ne->ne_next = pl->names;
339 pl->names = ne;
340
341 return (0);
342 }
343
344 /*
345 * Returns true if the named pool matches the given GUID.
346 */
347 static int
348 pool_active(libzfs_handle_t *hdl, const char *name, uint64_t guid,
349 boolean_t *isactive)
350 {
351 zpool_handle_t *zhp;
352 uint64_t theguid;
353
354 if (zpool_open_silent(hdl, name, &zhp) != 0)
355 return (-1);
356
357 if (zhp == NULL) {
358 *isactive = B_FALSE;
359 return (0);
360 }
361
362 verify(nvlist_lookup_uint64(zhp->zpool_config, ZPOOL_CONFIG_POOL_GUID,
363 &theguid) == 0);
364
365 zpool_close(zhp);
366
367 *isactive = (theguid == guid);
368 return (0);
369 }
370
371 static nvlist_t *
372 refresh_config(libzfs_handle_t *hdl, nvlist_t *config)
373 {
374 nvlist_t *nvl;
375 zfs_cmd_t zc = { 0 };
376 int err;
377
378 if (zcmd_write_conf_nvlist(hdl, &zc, config) != 0)
379 return (NULL);
380
381 if (zcmd_alloc_dst_nvlist(hdl, &zc,
382 zc.zc_nvlist_conf_size * 2) != 0) {
383 zcmd_free_nvlists(&zc);
384 return (NULL);
385 }
386
387 while ((err = ioctl(hdl->libzfs_fd, ZFS_IOC_POOL_TRYIMPORT,
388 &zc)) != 0 && errno == ENOMEM) {
389 if (zcmd_expand_dst_nvlist(hdl, &zc) != 0) {
390 zcmd_free_nvlists(&zc);
391 return (NULL);
392 }
393 }
394
395 if (err) {
396 zcmd_free_nvlists(&zc);
397 return (NULL);
398 }
399
400 if (zcmd_read_dst_nvlist(hdl, &zc, &nvl) != 0) {
401 zcmd_free_nvlists(&zc);
402 return (NULL);
403 }
404
405 zcmd_free_nvlists(&zc);
406 return (nvl);
407 }
408
409 /*
410 * Determine if the vdev id is a hole in the namespace.
411 */
412 boolean_t
413 vdev_is_hole(uint64_t *hole_array, uint_t holes, uint_t id)
414 {
415 for (int c = 0; c < holes; c++) {
416
417 /* Top-level is a hole */
418 if (hole_array[c] == id)
419 return (B_TRUE);
420 }
421 return (B_FALSE);
422 }
423
424 /*
425 * Convert our list of pools into the definitive set of configurations. We
426 * start by picking the best config for each toplevel vdev. Once that's done,
427 * we assemble the toplevel vdevs into a full config for the pool. We make a
428 * pass to fix up any incorrect paths, and then add it to the main list to
429 * return to the user.
430 */
431 static nvlist_t *
432 get_configs(libzfs_handle_t *hdl, pool_list_t *pl, boolean_t active_ok)
433 {
434 pool_entry_t *pe;
435 vdev_entry_t *ve;
436 config_entry_t *ce;
437 nvlist_t *ret = NULL, *config = NULL, *tmp, *nvtop, *nvroot;
438 nvlist_t **spares, **l2cache;
439 uint_t i, nspares, nl2cache;
440 boolean_t config_seen;
441 uint64_t best_txg;
442 char *name, *hostname;
443 uint64_t guid;
444 uint_t children = 0;
445 nvlist_t **child = NULL;
446 uint_t holes;
447 uint64_t *hole_array, max_id;
448 uint_t c;
449 boolean_t isactive;
450 uint64_t hostid;
451 nvlist_t *nvl;
452 boolean_t found_one = B_FALSE;
453 boolean_t valid_top_config = B_FALSE;
454
455 if (nvlist_alloc(&ret, 0, 0) != 0)
456 goto nomem;
457
458 for (pe = pl->pools; pe != NULL; pe = pe->pe_next) {
459 uint64_t id, max_txg = 0;
460
461 if (nvlist_alloc(&config, NV_UNIQUE_NAME, 0) != 0)
462 goto nomem;
463 config_seen = B_FALSE;
464
465 /*
466 * Iterate over all toplevel vdevs. Grab the pool configuration
467 * from the first one we find, and then go through the rest and
468 * add them as necessary to the 'vdevs' member of the config.
469 */
470 for (ve = pe->pe_vdevs; ve != NULL; ve = ve->ve_next) {
471
472 /*
473 * Determine the best configuration for this vdev by
474 * selecting the config with the latest transaction
475 * group.
476 */
477 best_txg = 0;
478 for (ce = ve->ve_configs; ce != NULL;
479 ce = ce->ce_next) {
480
481 if (ce->ce_txg > best_txg) {
482 tmp = ce->ce_config;
483 best_txg = ce->ce_txg;
484 }
485 }
486
487 /*
488 * We rely on the fact that the max txg for the
489 * pool will contain the most up-to-date information
490 * about the valid top-levels in the vdev namespace.
491 */
492 if (best_txg > max_txg) {
493 (void) nvlist_remove(config,
494 ZPOOL_CONFIG_VDEV_CHILDREN,
495 DATA_TYPE_UINT64);
496 (void) nvlist_remove(config,
497 ZPOOL_CONFIG_HOLE_ARRAY,
498 DATA_TYPE_UINT64_ARRAY);
499
500 max_txg = best_txg;
501 hole_array = NULL;
502 holes = 0;
503 max_id = 0;
504 valid_top_config = B_FALSE;
505
506 if (nvlist_lookup_uint64(tmp,
507 ZPOOL_CONFIG_VDEV_CHILDREN, &max_id) == 0) {
508 verify(nvlist_add_uint64(config,
509 ZPOOL_CONFIG_VDEV_CHILDREN,
510 max_id) == 0);
511 valid_top_config = B_TRUE;
512 }
513
514 if (nvlist_lookup_uint64_array(tmp,
515 ZPOOL_CONFIG_HOLE_ARRAY, &hole_array,
516 &holes) == 0) {
517 verify(nvlist_add_uint64_array(config,
518 ZPOOL_CONFIG_HOLE_ARRAY,
519 hole_array, holes) == 0);
520 }
521 }
522
523 if (!config_seen) {
524 /*
525 * Copy the relevant pieces of data to the pool
526 * configuration:
527 *
528 * version
529 * pool guid
530 * name
531 * pool txg (if available)
532 * comment (if available)
533 * pool state
534 * hostid (if available)
535 * hostname (if available)
536 */
537 uint64_t state, version, pool_txg;
538 char *comment = NULL;
539
540 version = fnvlist_lookup_uint64(tmp,
541 ZPOOL_CONFIG_VERSION);
542 fnvlist_add_uint64(config,
543 ZPOOL_CONFIG_VERSION, version);
544 guid = fnvlist_lookup_uint64(tmp,
545 ZPOOL_CONFIG_POOL_GUID);
546 fnvlist_add_uint64(config,
547 ZPOOL_CONFIG_POOL_GUID, guid);
548 name = fnvlist_lookup_string(tmp,
549 ZPOOL_CONFIG_POOL_NAME);
550 fnvlist_add_string(config,
551 ZPOOL_CONFIG_POOL_NAME, name);
552
553 if (nvlist_lookup_uint64(tmp,
554 ZPOOL_CONFIG_POOL_TXG, &pool_txg) == 0)
555 fnvlist_add_uint64(config,
556 ZPOOL_CONFIG_POOL_TXG, pool_txg);
557
558 if (nvlist_lookup_string(tmp,
559 ZPOOL_CONFIG_COMMENT, &comment) == 0)
560 fnvlist_add_string(config,
561 ZPOOL_CONFIG_COMMENT, comment);
562
563 state = fnvlist_lookup_uint64(tmp,
564 ZPOOL_CONFIG_POOL_STATE);
565 fnvlist_add_uint64(config,
566 ZPOOL_CONFIG_POOL_STATE, state);
567
568 hostid = 0;
569 if (nvlist_lookup_uint64(tmp,
570 ZPOOL_CONFIG_HOSTID, &hostid) == 0) {
571 fnvlist_add_uint64(config,
572 ZPOOL_CONFIG_HOSTID, hostid);
573 hostname = fnvlist_lookup_string(tmp,
574 ZPOOL_CONFIG_HOSTNAME);
575 fnvlist_add_string(config,
576 ZPOOL_CONFIG_HOSTNAME, hostname);
577 }
578
579 config_seen = B_TRUE;
580 }
581
582 /*
583 * Add this top-level vdev to the child array.
584 */
585 verify(nvlist_lookup_nvlist(tmp,
586 ZPOOL_CONFIG_VDEV_TREE, &nvtop) == 0);
587 verify(nvlist_lookup_uint64(nvtop, ZPOOL_CONFIG_ID,
588 &id) == 0);
589
590 if (id >= children) {
591 nvlist_t **newchild;
592
593 newchild = zfs_alloc(hdl, (id + 1) *
594 sizeof (nvlist_t *));
595 if (newchild == NULL)
596 goto nomem;
597
598 for (c = 0; c < children; c++)
599 newchild[c] = child[c];
600
601 free(child);
602 child = newchild;
603 children = id + 1;
604 }
605 if (nvlist_dup(nvtop, &child[id], 0) != 0)
606 goto nomem;
607
608 }
609
610 /*
611 * If we have information about all the top-levels then
612 * clean up the nvlist which we've constructed. This
613 * means removing any extraneous devices that are
614 * beyond the valid range or adding devices to the end
615 * of our array which appear to be missing.
616 */
617 if (valid_top_config) {
618 if (max_id < children) {
619 for (c = max_id; c < children; c++)
620 nvlist_free(child[c]);
621 children = max_id;
622 } else if (max_id > children) {
623 nvlist_t **newchild;
624
625 newchild = zfs_alloc(hdl, (max_id) *
626 sizeof (nvlist_t *));
627 if (newchild == NULL)
628 goto nomem;
629
630 for (c = 0; c < children; c++)
631 newchild[c] = child[c];
632
633 free(child);
634 child = newchild;
635 children = max_id;
636 }
637 }
638
639 verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID,
640 &guid) == 0);
641
642 /*
643 * The vdev namespace may contain holes as a result of
644 * device removal. We must add them back into the vdev
645 * tree before we process any missing devices.
646 */
647 if (holes > 0) {
648 ASSERT(valid_top_config);
649
650 for (c = 0; c < children; c++) {
651 nvlist_t *holey;
652
653 if (child[c] != NULL ||
654 !vdev_is_hole(hole_array, holes, c))
655 continue;
656
657 if (nvlist_alloc(&holey, NV_UNIQUE_NAME,
658 0) != 0)
659 goto nomem;
660
661 /*
662 * Holes in the namespace are treated as
663 * "hole" top-level vdevs and have a
664 * special flag set on them.
665 */
666 if (nvlist_add_string(holey,
667 ZPOOL_CONFIG_TYPE,
668 VDEV_TYPE_HOLE) != 0 ||
669 nvlist_add_uint64(holey,
670 ZPOOL_CONFIG_ID, c) != 0 ||
671 nvlist_add_uint64(holey,
672 ZPOOL_CONFIG_GUID, 0ULL) != 0)
673 goto nomem;
674 child[c] = holey;
675 }
676 }
677
678 /*
679 * Look for any missing top-level vdevs. If this is the case,
680 * create a faked up 'missing' vdev as a placeholder. We cannot
681 * simply compress the child array, because the kernel performs
682 * certain checks to make sure the vdev IDs match their location
683 * in the configuration.
684 */
685 for (c = 0; c < children; c++) {
686 if (child[c] == NULL) {
687 nvlist_t *missing;
688 if (nvlist_alloc(&missing, NV_UNIQUE_NAME,
689 0) != 0)
690 goto nomem;
691 if (nvlist_add_string(missing,
692 ZPOOL_CONFIG_TYPE,
693 VDEV_TYPE_MISSING) != 0 ||
694 nvlist_add_uint64(missing,
695 ZPOOL_CONFIG_ID, c) != 0 ||
696 nvlist_add_uint64(missing,
697 ZPOOL_CONFIG_GUID, 0ULL) != 0) {
698 nvlist_free(missing);
699 goto nomem;
700 }
701 child[c] = missing;
702 }
703 }
704
705 /*
706 * Put all of this pool's top-level vdevs into a root vdev.
707 */
708 if (nvlist_alloc(&nvroot, NV_UNIQUE_NAME, 0) != 0)
709 goto nomem;
710 if (nvlist_add_string(nvroot, ZPOOL_CONFIG_TYPE,
711 VDEV_TYPE_ROOT) != 0 ||
712 nvlist_add_uint64(nvroot, ZPOOL_CONFIG_ID, 0ULL) != 0 ||
713 nvlist_add_uint64(nvroot, ZPOOL_CONFIG_GUID, guid) != 0 ||
714 nvlist_add_nvlist_array(nvroot, ZPOOL_CONFIG_CHILDREN,
715 child, children) != 0) {
716 nvlist_free(nvroot);
717 goto nomem;
718 }
719
720 for (c = 0; c < children; c++)
721 nvlist_free(child[c]);
722 free(child);
723 children = 0;
724 child = NULL;
725
726 /*
727 * Go through and fix up any paths and/or devids based on our
728 * known list of vdev GUID -> path mappings.
729 */
730 if (fix_paths(nvroot, pl->names) != 0) {
731 nvlist_free(nvroot);
732 goto nomem;
733 }
734
735 /*
736 * Add the root vdev to this pool's configuration.
737 */
738 if (nvlist_add_nvlist(config, ZPOOL_CONFIG_VDEV_TREE,
739 nvroot) != 0) {
740 nvlist_free(nvroot);
741 goto nomem;
742 }
743 nvlist_free(nvroot);
744
745 /*
746 * zdb uses this path to report on active pools that were
747 * imported or created using -R.
748 */
749 if (active_ok)
750 goto add_pool;
751
752 /*
753 * Determine if this pool is currently active, in which case we
754 * can't actually import it.
755 */
756 verify(nvlist_lookup_string(config, ZPOOL_CONFIG_POOL_NAME,
757 &name) == 0);
758 verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID,
759 &guid) == 0);
760
761 if (pool_active(hdl, name, guid, &isactive) != 0)
762 goto error;
763
764 if (isactive) {
765 nvlist_free(config);
766 config = NULL;
767 continue;
768 }
769
770 if ((nvl = refresh_config(hdl, config)) == NULL) {
771 nvlist_free(config);
772 config = NULL;
773 continue;
774 }
775
776 nvlist_free(config);
777 config = nvl;
778
779 /*
780 * Go through and update the paths for spares, now that we have
781 * them.
782 */
783 verify(nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE,
784 &nvroot) == 0);
785 if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_SPARES,
786 &spares, &nspares) == 0) {
787 for (i = 0; i < nspares; i++) {
788 if (fix_paths(spares[i], pl->names) != 0)
789 goto nomem;
790 }
791 }
792
793 /*
794 * Update the paths for l2cache devices.
795 */
796 if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_L2CACHE,
797 &l2cache, &nl2cache) == 0) {
798 for (i = 0; i < nl2cache; i++) {
799 if (fix_paths(l2cache[i], pl->names) != 0)
800 goto nomem;
801 }
802 }
803
804 /*
805 * Restore the original information read from the actual label.
806 */
807 (void) nvlist_remove(config, ZPOOL_CONFIG_HOSTID,
808 DATA_TYPE_UINT64);
809 (void) nvlist_remove(config, ZPOOL_CONFIG_HOSTNAME,
810 DATA_TYPE_STRING);
811 if (hostid != 0) {
812 verify(nvlist_add_uint64(config, ZPOOL_CONFIG_HOSTID,
813 hostid) == 0);
814 verify(nvlist_add_string(config, ZPOOL_CONFIG_HOSTNAME,
815 hostname) == 0);
816 }
817
818 add_pool:
819 /*
820 * Add this pool to the list of configs.
821 */
822 verify(nvlist_lookup_string(config, ZPOOL_CONFIG_POOL_NAME,
823 &name) == 0);
824 if (nvlist_add_nvlist(ret, name, config) != 0)
825 goto nomem;
826
827 found_one = B_TRUE;
828 nvlist_free(config);
829 config = NULL;
830 }
831
832 if (!found_one) {
833 nvlist_free(ret);
834 ret = NULL;
835 }
836
837 return (ret);
838
839 nomem:
840 (void) no_memory(hdl);
841 error:
842 nvlist_free(config);
843 nvlist_free(ret);
844 for (c = 0; c < children; c++)
845 nvlist_free(child[c]);
846 free(child);
847
848 return (NULL);
849 }
850
851 /*
852 * Return the offset of the given label.
853 */
854 static uint64_t
855 label_offset(uint64_t size, int l)
856 {
857 ASSERT(P2PHASE_TYPED(size, sizeof (vdev_label_t), uint64_t) == 0);
858 return (l * sizeof (vdev_label_t) + (l < VDEV_LABELS / 2 ?
859 0 : size - VDEV_LABELS * sizeof (vdev_label_t)));
860 }
861
862 /*
863 * Given a file descriptor, read the label information and return an nvlist
864 * describing the configuration, if there is one.
865 */
866 int
867 zpool_read_label(int fd, nvlist_t **config)
868 {
869 struct stat64 statbuf;
870 int l;
871 vdev_label_t *label;
872 uint64_t state, txg, size;
873
874 *config = NULL;
875
876 if (fstat64(fd, &statbuf) == -1)
877 return (0);
878 size = P2ALIGN_TYPED(statbuf.st_size, sizeof (vdev_label_t), uint64_t);
879
880 if ((label = malloc(sizeof (vdev_label_t))) == NULL)
881 return (-1);
882
883 for (l = 0; l < VDEV_LABELS; l++) {
884 if (pread64(fd, label, sizeof (vdev_label_t),
885 label_offset(size, l)) != sizeof (vdev_label_t))
886 continue;
887
888 if (nvlist_unpack(label->vl_vdev_phys.vp_nvlist,
889 sizeof (label->vl_vdev_phys.vp_nvlist), config, 0) != 0)
890 continue;
891
892 if (nvlist_lookup_uint64(*config, ZPOOL_CONFIG_POOL_STATE,
893 &state) != 0 || state > POOL_STATE_L2CACHE) {
894 nvlist_free(*config);
895 continue;
896 }
897
898 if (state != POOL_STATE_SPARE && state != POOL_STATE_L2CACHE &&
899 (nvlist_lookup_uint64(*config, ZPOOL_CONFIG_POOL_TXG,
900 &txg) != 0 || txg == 0)) {
901 nvlist_free(*config);
902 continue;
903 }
904
905 free(label);
906 return (0);
907 }
908
909 free(label);
910 *config = NULL;
911 return (0);
912 }
913
914 typedef struct rdsk_node {
915 char *rn_name;
916 int rn_dfd;
917 libzfs_handle_t *rn_hdl;
918 nvlist_t *rn_config;
919 avl_tree_t *rn_avl;
920 avl_node_t rn_node;
921 boolean_t rn_nozpool;
922 } rdsk_node_t;
923
924 static int
925 slice_cache_compare(const void *arg1, const void *arg2)
926 {
927 const char *nm1 = ((rdsk_node_t *)arg1)->rn_name;
928 const char *nm2 = ((rdsk_node_t *)arg2)->rn_name;
929 char *nm1slice, *nm2slice;
930 int rv;
931
932 /*
933 * slices zero and two are the most likely to provide results,
934 * so put those first
935 */
936 nm1slice = strstr(nm1, "s0");
937 nm2slice = strstr(nm2, "s0");
938 if (nm1slice && !nm2slice) {
939 return (-1);
940 }
941 if (!nm1slice && nm2slice) {
942 return (1);
943 }
944 nm1slice = strstr(nm1, "s2");
945 nm2slice = strstr(nm2, "s2");
946 if (nm1slice && !nm2slice) {
947 return (-1);
948 }
949 if (!nm1slice && nm2slice) {
950 return (1);
951 }
952
953 rv = strcmp(nm1, nm2);
954 if (rv == 0)
955 return (0);
956 return (rv > 0 ? 1 : -1);
957 }
958
959 static void
960 check_one_slice(avl_tree_t *r, char *diskname, uint_t partno,
961 diskaddr_t size, uint_t blksz)
962 {
963 rdsk_node_t tmpnode;
964 rdsk_node_t *node;
965 char sname[MAXNAMELEN];
966
967 tmpnode.rn_name = &sname[0];
968 (void) snprintf(tmpnode.rn_name, MAXNAMELEN, "%s%u",
969 diskname, partno);
970 /*
971 * protect against division by zero for disk labels that
972 * contain a bogus sector size
973 */
974 if (blksz == 0)
975 blksz = DEV_BSIZE;
976 /* too small to contain a zpool? */
977 if ((size < (SPA_MINDEVSIZE / blksz)) &&
978 (node = avl_find(r, &tmpnode, NULL)))
979 node->rn_nozpool = B_TRUE;
980 }
981
982 static void
983 nozpool_all_slices(avl_tree_t *r, const char *sname)
984 {
985 char diskname[MAXNAMELEN];
986 char *ptr;
987 int i;
988
989 (void) strncpy(diskname, sname, MAXNAMELEN);
990 if (((ptr = strrchr(diskname, 's')) == NULL) &&
991 ((ptr = strrchr(diskname, 'p')) == NULL))
992 return;
993 ptr[0] = 's';
994 ptr[1] = '\0';
995 for (i = 0; i < NDKMAP; i++)
996 check_one_slice(r, diskname, i, 0, 1);
997 ptr[0] = 'p';
998 for (i = 0; i <= FD_NUMPART; i++)
999 check_one_slice(r, diskname, i, 0, 1);
1000 }
1001
1002 static void
1003 check_slices(avl_tree_t *r, int fd, const char *sname)
1004 {
1005 struct extvtoc vtoc;
1006 struct dk_gpt *gpt;
1007 char diskname[MAXNAMELEN];
1008 char *ptr;
1009 int i;
1010
1011 (void) strncpy(diskname, sname, MAXNAMELEN);
1012 if ((ptr = strrchr(diskname, 's')) == NULL || !isdigit(ptr[1]))
1013 return;
1014 ptr[1] = '\0';
1015
1016 if (read_extvtoc(fd, &vtoc) >= 0) {
1017 for (i = 0; i < NDKMAP; i++)
1018 check_one_slice(r, diskname, i,
1019 vtoc.v_part[i].p_size, vtoc.v_sectorsz);
1020 } else if (efi_alloc_and_read(fd, &gpt) >= 0) {
1021 /*
1022 * on x86 we'll still have leftover links that point
1023 * to slices s[9-15], so use NDKMAP instead
1024 */
1025 for (i = 0; i < NDKMAP; i++)
1026 check_one_slice(r, diskname, i,
1027 gpt->efi_parts[i].p_size, gpt->efi_lbasize);
1028 /* nodes p[1-4] are never used with EFI labels */
1029 ptr[0] = 'p';
1030 for (i = 1; i <= FD_NUMPART; i++)
1031 check_one_slice(r, diskname, i, 0, 1);
1032 efi_free(gpt);
1033 }
1034 }
1035
1036 static void
1037 zpool_open_func(void *arg)
1038 {
1039 rdsk_node_t *rn = arg;
1040 struct stat64 statbuf;
1041 nvlist_t *config;
1042 int fd;
1043
1044 if (rn->rn_nozpool)
1045 return;
1046 if ((fd = openat64(rn->rn_dfd, rn->rn_name, O_RDONLY)) < 0) {
1047 /* symlink to a device that's no longer there */
1048 if (errno == ENOENT)
1049 nozpool_all_slices(rn->rn_avl, rn->rn_name);
1050 return;
1051 }
1052 /*
1053 * Ignore failed stats. We only want regular
1054 * files, character devs and block devs.
1055 */
1056 if (fstat64(fd, &statbuf) != 0 ||
1057 (!S_ISREG(statbuf.st_mode) &&
1058 !S_ISCHR(statbuf.st_mode) &&
1059 !S_ISBLK(statbuf.st_mode))) {
1060 (void) close(fd);
1061 return;
1062 }
1063 /* this file is too small to hold a zpool */
1064 if (S_ISREG(statbuf.st_mode) &&
1065 statbuf.st_size < SPA_MINDEVSIZE) {
1066 (void) close(fd);
1067 return;
1068 } else if (!S_ISREG(statbuf.st_mode)) {
1069 /*
1070 * Try to read the disk label first so we don't have to
1071 * open a bunch of minor nodes that can't have a zpool.
1072 */
1073 check_slices(rn->rn_avl, fd, rn->rn_name);
1074 }
1075
1076 if ((zpool_read_label(fd, &config)) != 0) {
1077 (void) close(fd);
1078 (void) no_memory(rn->rn_hdl);
1079 return;
1080 }
1081 (void) close(fd);
1082
1083
1084 rn->rn_config = config;
1085 if (config != NULL) {
1086 assert(rn->rn_nozpool == B_FALSE);
1087 }
1088 }
1089
1090 /*
1091 * Given a file descriptor, clear (zero) the label information. This function
1092 * is currently only used in the appliance stack as part of the ZFS sysevent
1093 * module.
1094 */
1095 int
1096 zpool_clear_label(int fd)
1097 {
1098 struct stat64 statbuf;
1099 int l;
1100 vdev_label_t *label;
1101 uint64_t size;
1102
1103 if (fstat64(fd, &statbuf) == -1)
1104 return (0);
1105 size = P2ALIGN_TYPED(statbuf.st_size, sizeof (vdev_label_t), uint64_t);
1106
1107 if ((label = calloc(sizeof (vdev_label_t), 1)) == NULL)
1108 return (-1);
1109
1110 for (l = 0; l < VDEV_LABELS; l++) {
1111 if (pwrite64(fd, label, sizeof (vdev_label_t),
1112 label_offset(size, l)) != sizeof (vdev_label_t))
1113 return (-1);
1114 }
1115
1116 free(label);
1117 return (0);
1118 }
1119
1120 /*
1121 * Given a list of directories to search, find all pools stored on disk. This
1122 * includes partial pools which are not available to import. If no args are
1123 * given (argc is 0), then the default directory (/dev/dsk) is searched.
1124 * poolname or guid (but not both) are provided by the caller when trying
1125 * to import a specific pool.
1126 */
1127 static nvlist_t *
1128 zpool_find_import_impl(libzfs_handle_t *hdl, importargs_t *iarg)
1129 {
1130 int i, dirs = iarg->paths;
1131 DIR *dirp = NULL;
1132 struct dirent64 *dp;
1133 char path[MAXPATHLEN];
1134 char *end, **dir = iarg->path;
1135 size_t pathleft;
1136 nvlist_t *ret = NULL;
1137 static char *default_dir = "/dev/dsk";
1138 pool_list_t pools = { 0 };
1139 pool_entry_t *pe, *penext;
1140 vdev_entry_t *ve, *venext;
1141 config_entry_t *ce, *cenext;
1142 name_entry_t *ne, *nenext;
1143 avl_tree_t slice_cache;
1144 rdsk_node_t *slice;
1145 void *cookie;
1146
1147 if (dirs == 0) {
1148 dirs = 1;
1149 dir = &default_dir;
1150 }
1151
1152 /*
1153 * Go through and read the label configuration information from every
1154 * possible device, organizing the information according to pool GUID
1155 * and toplevel GUID.
1156 */
1157 for (i = 0; i < dirs; i++) {
1158 tpool_t *t;
1159 char *rdsk;
1160 int dfd;
1161
1162 /* use realpath to normalize the path */
1163 if (realpath(dir[i], path) == 0) {
1164 (void) zfs_error_fmt(hdl, EZFS_BADPATH,
1165 dgettext(TEXT_DOMAIN, "cannot open '%s'"), dir[i]);
1166 goto error;
1167 }
1168 end = &path[strlen(path)];
1169 *end++ = '/';
1170 *end = 0;
1171 pathleft = &path[sizeof (path)] - end;
1172
1173 /*
1174 * Using raw devices instead of block devices when we're
1175 * reading the labels skips a bunch of slow operations during
1176 * close(2) processing, so we replace /dev/dsk with /dev/rdsk.
1177 */
1178 if (strcmp(path, "/dev/dsk/") == 0)
1179 rdsk = "/dev/rdsk/";
1180 else
1181 rdsk = path;
1182
1183 if ((dfd = open64(rdsk, O_RDONLY)) < 0 ||
1184 (dirp = fdopendir(dfd)) == NULL) {
1185 zfs_error_aux(hdl, strerror(errno));
1186 (void) zfs_error_fmt(hdl, EZFS_BADPATH,
1187 dgettext(TEXT_DOMAIN, "cannot open '%s'"),
1188 rdsk);
1189 goto error;
1190 }
1191
1192 avl_create(&slice_cache, slice_cache_compare,
1193 sizeof (rdsk_node_t), offsetof(rdsk_node_t, rn_node));
1194 /*
1195 * This is not MT-safe, but we have no MT consumers of libzfs
1196 */
1197 while ((dp = readdir64(dirp)) != NULL) {
1198 const char *name = dp->d_name;
1199 if (name[0] == '.' &&
1200 (name[1] == 0 || (name[1] == '.' && name[2] == 0)))
1201 continue;
1202
1203 slice = zfs_alloc(hdl, sizeof (rdsk_node_t));
1204 slice->rn_name = zfs_strdup(hdl, name);
1205 slice->rn_avl = &slice_cache;
1206 slice->rn_dfd = dfd;
1207 slice->rn_hdl = hdl;
1208 slice->rn_nozpool = B_FALSE;
1209 avl_add(&slice_cache, slice);
1210 }
1211 /*
1212 * create a thread pool to do all of this in parallel;
1213 * rn_nozpool is not protected, so this is racy in that
1214 * multiple tasks could decide that the same slice can
1215 * not hold a zpool, which is benign. Also choose
1216 * double the number of processors; we hold a lot of
1217 * locks in the kernel, so going beyond this doesn't
1218 * buy us much.
1219 */
1220 t = tpool_create(1, 2 * sysconf(_SC_NPROCESSORS_ONLN),
1221 0, NULL);
1222 for (slice = avl_first(&slice_cache); slice;
1223 (slice = avl_walk(&slice_cache, slice,
1224 AVL_AFTER)))
1225 (void) tpool_dispatch(t, zpool_open_func, slice);
1226 tpool_wait(t);
1227 tpool_destroy(t);
1228
1229 cookie = NULL;
1230 while ((slice = avl_destroy_nodes(&slice_cache,
1231 &cookie)) != NULL) {
1232 if (slice->rn_config != NULL) {
1233 nvlist_t *config = slice->rn_config;
1234 boolean_t matched = B_TRUE;
1235
1236 if (iarg->poolname != NULL) {
1237 char *pname;
1238
1239 matched = nvlist_lookup_string(config,
1240 ZPOOL_CONFIG_POOL_NAME,
1241 &pname) == 0 &&
1242 strcmp(iarg->poolname, pname) == 0;
1243 } else if (iarg->guid != 0) {
1244 uint64_t this_guid;
1245
1246 matched = nvlist_lookup_uint64(config,
1247 ZPOOL_CONFIG_POOL_GUID,
1248 &this_guid) == 0 &&
1249 iarg->guid == this_guid;
1250 }
1251 if (!matched) {
1252 nvlist_free(config);
1253 config = NULL;
1254 continue;
1255 }
1256 /* use the non-raw path for the config */
1257 (void) strlcpy(end, slice->rn_name, pathleft);
1258 if (add_config(hdl, &pools, path, config) != 0)
1259 goto error;
1260 }
1261 free(slice->rn_name);
1262 free(slice);
1263 }
1264 avl_destroy(&slice_cache);
1265
1266 (void) closedir(dirp);
1267 dirp = NULL;
1268 }
1269
1270 ret = get_configs(hdl, &pools, iarg->can_be_active);
1271
1272 error:
1273 for (pe = pools.pools; pe != NULL; pe = penext) {
1274 penext = pe->pe_next;
1275 for (ve = pe->pe_vdevs; ve != NULL; ve = venext) {
1276 venext = ve->ve_next;
1277 for (ce = ve->ve_configs; ce != NULL; ce = cenext) {
1278 cenext = ce->ce_next;
1279 if (ce->ce_config)
1280 nvlist_free(ce->ce_config);
1281 free(ce);
1282 }
1283 free(ve);
1284 }
1285 free(pe);
1286 }
1287
1288 for (ne = pools.names; ne != NULL; ne = nenext) {
1289 nenext = ne->ne_next;
1290 if (ne->ne_name)
1291 free(ne->ne_name);
1292 free(ne);
1293 }
1294
1295 if (dirp)
1296 (void) closedir(dirp);
1297
1298 return (ret);
1299 }
1300
1301 nvlist_t *
1302 zpool_find_import(libzfs_handle_t *hdl, int argc, char **argv)
1303 {
1304 importargs_t iarg = { 0 };
1305
1306 iarg.paths = argc;
1307 iarg.path = argv;
1308
1309 return (zpool_find_import_impl(hdl, &iarg));
1310 }
1311
1312 /*
1313 * Given a cache file, return the contents as a list of importable pools.
1314 * poolname or guid (but not both) are provided by the caller when trying
1315 * to import a specific pool.
1316 */
1317 nvlist_t *
1318 zpool_find_import_cached(libzfs_handle_t *hdl, const char *cachefile,
1319 char *poolname, uint64_t guid)
1320 {
1321 char *buf;
1322 int fd;
1323 struct stat64 statbuf;
1324 nvlist_t *raw, *src, *dst;
1325 nvlist_t *pools;
1326 nvpair_t *elem;
1327 char *name;
1328 uint64_t this_guid;
1329 boolean_t active;
1330
1331 verify(poolname == NULL || guid == 0);
1332
1333 if ((fd = open(cachefile, O_RDONLY)) < 0) {
1334 zfs_error_aux(hdl, "%s", strerror(errno));
1335 (void) zfs_error(hdl, EZFS_BADCACHE,
1336 dgettext(TEXT_DOMAIN, "failed to open cache file"));
1337 return (NULL);
1338 }
1339
1340 if (fstat64(fd, &statbuf) != 0) {
1341 zfs_error_aux(hdl, "%s", strerror(errno));
1342 (void) close(fd);
1343 (void) zfs_error(hdl, EZFS_BADCACHE,
1344 dgettext(TEXT_DOMAIN, "failed to get size of cache file"));
1345 return (NULL);
1346 }
1347
1348 if ((buf = zfs_alloc(hdl, statbuf.st_size)) == NULL) {
1349 (void) close(fd);
1350 return (NULL);
1351 }
1352
1353 if (read(fd, buf, statbuf.st_size) != statbuf.st_size) {
1354 (void) close(fd);
1355 free(buf);
1356 (void) zfs_error(hdl, EZFS_BADCACHE,
1357 dgettext(TEXT_DOMAIN,
1358 "failed to read cache file contents"));
1359 return (NULL);
1360 }
1361
1362 (void) close(fd);
1363
1364 if (nvlist_unpack(buf, statbuf.st_size, &raw, 0) != 0) {
1365 free(buf);
1366 (void) zfs_error(hdl, EZFS_BADCACHE,
1367 dgettext(TEXT_DOMAIN,
1368 "invalid or corrupt cache file contents"));
1369 return (NULL);
1370 }
1371
1372 free(buf);
1373
1374 /*
1375 * Go through and get the current state of the pools and refresh their
1376 * state.
1377 */
1378 if (nvlist_alloc(&pools, 0, 0) != 0) {
1379 (void) no_memory(hdl);
1380 nvlist_free(raw);
1381 return (NULL);
1382 }
1383
1384 elem = NULL;
1385 while ((elem = nvlist_next_nvpair(raw, elem)) != NULL) {
1386 verify(nvpair_value_nvlist(elem, &src) == 0);
1387
1388 verify(nvlist_lookup_string(src, ZPOOL_CONFIG_POOL_NAME,
1389 &name) == 0);
1390 if (poolname != NULL && strcmp(poolname, name) != 0)
1391 continue;
1392
1393 verify(nvlist_lookup_uint64(src, ZPOOL_CONFIG_POOL_GUID,
1394 &this_guid) == 0);
1395 if (guid != 0) {
1396 verify(nvlist_lookup_uint64(src, ZPOOL_CONFIG_POOL_GUID,
1397 &this_guid) == 0);
1398 if (guid != this_guid)
1399 continue;
1400 }
1401
1402 if (pool_active(hdl, name, this_guid, &active) != 0) {
1403 nvlist_free(raw);
1404 nvlist_free(pools);
1405 return (NULL);
1406 }
1407
1408 if (active)
1409 continue;
1410
1411 if ((dst = refresh_config(hdl, src)) == NULL) {
1412 nvlist_free(raw);
1413 nvlist_free(pools);
1414 return (NULL);
1415 }
1416
1417 if (nvlist_add_nvlist(pools, nvpair_name(elem), dst) != 0) {
1418 (void) no_memory(hdl);
1419 nvlist_free(dst);
1420 nvlist_free(raw);
1421 nvlist_free(pools);
1422 return (NULL);
1423 }
1424 nvlist_free(dst);
1425 }
1426
1427 nvlist_free(raw);
1428 return (pools);
1429 }
1430
1431 static int
1432 name_or_guid_exists(zpool_handle_t *zhp, void *data)
1433 {
1434 importargs_t *import = data;
1435 int found = 0;
1436
1437 if (import->poolname != NULL) {
1438 char *pool_name;
1439
1440 verify(nvlist_lookup_string(zhp->zpool_config,
1441 ZPOOL_CONFIG_POOL_NAME, &pool_name) == 0);
1442 if (strcmp(pool_name, import->poolname) == 0)
1443 found = 1;
1444 } else {
1445 uint64_t pool_guid;
1446
1447 verify(nvlist_lookup_uint64(zhp->zpool_config,
1448 ZPOOL_CONFIG_POOL_GUID, &pool_guid) == 0);
1449 if (pool_guid == import->guid)
1450 found = 1;
1451 }
1452
1453 zpool_close(zhp);
1454 return (found);
1455 }
1456
1457 nvlist_t *
1458 zpool_search_import(libzfs_handle_t *hdl, importargs_t *import)
1459 {
1460 verify(import->poolname == NULL || import->guid == 0);
1461
1462 if (import->unique)
1463 import->exists = zpool_iter(hdl, name_or_guid_exists, import);
1464
1465 if (import->cachefile != NULL)
1466 return (zpool_find_import_cached(hdl, import->cachefile,
1467 import->poolname, import->guid));
1468
1469 return (zpool_find_import_impl(hdl, import));
1470 }
1471
1472 boolean_t
1473 find_guid(nvlist_t *nv, uint64_t guid)
1474 {
1475 uint64_t tmp;
1476 nvlist_t **child;
1477 uint_t c, children;
1478
1479 verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID, &tmp) == 0);
1480 if (tmp == guid)
1481 return (B_TRUE);
1482
1483 if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN,
1484 &child, &children) == 0) {
1485 for (c = 0; c < children; c++)
1486 if (find_guid(child[c], guid))
1487 return (B_TRUE);
1488 }
1489
1490 return (B_FALSE);
1491 }
1492
1493 typedef struct aux_cbdata {
1494 const char *cb_type;
1495 uint64_t cb_guid;
1496 zpool_handle_t *cb_zhp;
1497 } aux_cbdata_t;
1498
1499 static int
1500 find_aux(zpool_handle_t *zhp, void *data)
1501 {
1502 aux_cbdata_t *cbp = data;
1503 nvlist_t **list;
1504 uint_t i, count;
1505 uint64_t guid;
1506 nvlist_t *nvroot;
1507
1508 verify(nvlist_lookup_nvlist(zhp->zpool_config, ZPOOL_CONFIG_VDEV_TREE,
1509 &nvroot) == 0);
1510
1511 if (nvlist_lookup_nvlist_array(nvroot, cbp->cb_type,
1512 &list, &count) == 0) {
1513 for (i = 0; i < count; i++) {
1514 verify(nvlist_lookup_uint64(list[i],
1515 ZPOOL_CONFIG_GUID, &guid) == 0);
1516 if (guid == cbp->cb_guid) {
1517 cbp->cb_zhp = zhp;
1518 return (1);
1519 }
1520 }
1521 }
1522
1523 zpool_close(zhp);
1524 return (0);
1525 }
1526
1527 /*
1528 * Determines if the pool is in use. If so, it returns true and the state of
1529 * the pool as well as the name of the pool. Both strings are allocated and
1530 * must be freed by the caller.
1531 */
1532 int
1533 zpool_in_use(libzfs_handle_t *hdl, int fd, pool_state_t *state, char **namestr,
1534 boolean_t *inuse)
1535 {
1536 nvlist_t *config;
1537 char *name;
1538 boolean_t ret;
1539 uint64_t guid, vdev_guid;
1540 zpool_handle_t *zhp;
1541 nvlist_t *pool_config;
1542 uint64_t stateval, isspare;
1543 aux_cbdata_t cb = { 0 };
1544 boolean_t isactive;
1545
1546 *inuse = B_FALSE;
1547
1548 if (zpool_read_label(fd, &config) != 0) {
1549 (void) no_memory(hdl);
1550 return (-1);
1551 }
1552
1553 if (config == NULL)
1554 return (0);
1555
1556 verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_STATE,
1557 &stateval) == 0);
1558 verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_GUID,
1559 &vdev_guid) == 0);
1560
1561 if (stateval != POOL_STATE_SPARE && stateval != POOL_STATE_L2CACHE) {
1562 verify(nvlist_lookup_string(config, ZPOOL_CONFIG_POOL_NAME,
1563 &name) == 0);
1564 verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID,
1565 &guid) == 0);
1566 }
1567
1568 switch (stateval) {
1569 case POOL_STATE_EXPORTED:
1570 /*
1571 * A pool with an exported state may in fact be imported
1572 * read-only, so check the in-core state to see if it's
1573 * active and imported read-only. If it is, set
1574 * its state to active.
1575 */
1576 if (pool_active(hdl, name, guid, &isactive) == 0 && isactive &&
1577 (zhp = zpool_open_canfail(hdl, name)) != NULL &&
1578 zpool_get_prop_int(zhp, ZPOOL_PROP_READONLY, NULL))
1579 stateval = POOL_STATE_ACTIVE;
1580
1581 ret = B_TRUE;
1582 break;
1583
1584 case POOL_STATE_ACTIVE:
1585 /*
1586 * For an active pool, we have to determine if it's really part
1587 * of a currently active pool (in which case the pool will exist
1588 * and the guid will be the same), or whether it's part of an
1589 * active pool that was disconnected without being explicitly
1590 * exported.
1591 */
1592 if (pool_active(hdl, name, guid, &isactive) != 0) {
1593 nvlist_free(config);
1594 return (-1);
1595 }
1596
1597 if (isactive) {
1598 /*
1599 * Because the device may have been removed while
1600 * offlined, we only report it as active if the vdev is
1601 * still present in the config. Otherwise, pretend like
1602 * it's not in use.
1603 */
1604 if ((zhp = zpool_open_canfail(hdl, name)) != NULL &&
1605 (pool_config = zpool_get_config(zhp, NULL))
1606 != NULL) {
1607 nvlist_t *nvroot;
1608
1609 verify(nvlist_lookup_nvlist(pool_config,
1610 ZPOOL_CONFIG_VDEV_TREE, &nvroot) == 0);
1611 ret = find_guid(nvroot, vdev_guid);
1612 } else {
1613 ret = B_FALSE;
1614 }
1615
1616 /*
1617 * If this is an active spare within another pool, we
1618 * treat it like an unused hot spare. This allows the
1619 * user to create a pool with a hot spare that currently
1620 * in use within another pool. Since we return B_TRUE,
1621 * libdiskmgt will continue to prevent generic consumers
1622 * from using the device.
1623 */
1624 if (ret && nvlist_lookup_uint64(config,
1625 ZPOOL_CONFIG_IS_SPARE, &isspare) == 0 && isspare)
1626 stateval = POOL_STATE_SPARE;
1627
1628 if (zhp != NULL)
1629 zpool_close(zhp);
1630 } else {
1631 stateval = POOL_STATE_POTENTIALLY_ACTIVE;
1632 ret = B_TRUE;
1633 }
1634 break;
1635
1636 case POOL_STATE_SPARE:
1637 /*
1638 * For a hot spare, it can be either definitively in use, or
1639 * potentially active. To determine if it's in use, we iterate
1640 * over all pools in the system and search for one with a spare
1641 * with a matching guid.
1642 *
1643 * Due to the shared nature of spares, we don't actually report
1644 * the potentially active case as in use. This means the user
1645 * can freely create pools on the hot spares of exported pools,
1646 * but to do otherwise makes the resulting code complicated, and
1647 * we end up having to deal with this case anyway.
1648 */
1649 cb.cb_zhp = NULL;
1650 cb.cb_guid = vdev_guid;
1651 cb.cb_type = ZPOOL_CONFIG_SPARES;
1652 if (zpool_iter(hdl, find_aux, &cb) == 1) {
1653 name = (char *)zpool_get_name(cb.cb_zhp);
1654 ret = TRUE;
1655 } else {
1656 ret = FALSE;
1657 }
1658 break;
1659
1660 case POOL_STATE_L2CACHE:
1661
1662 /*
1663 * Check if any pool is currently using this l2cache device.
1664 */
1665 cb.cb_zhp = NULL;
1666 cb.cb_guid = vdev_guid;
1667 cb.cb_type = ZPOOL_CONFIG_L2CACHE;
1668 if (zpool_iter(hdl, find_aux, &cb) == 1) {
1669 name = (char *)zpool_get_name(cb.cb_zhp);
1670 ret = TRUE;
1671 } else {
1672 ret = FALSE;
1673 }
1674 break;
1675
1676 default:
1677 ret = B_FALSE;
1678 }
1679
1680
1681 if (ret) {
1682 if ((*namestr = zfs_strdup(hdl, name)) == NULL) {
1683 if (cb.cb_zhp)
1684 zpool_close(cb.cb_zhp);
1685 nvlist_free(config);
1686 return (-1);
1687 }
1688 *state = (pool_state_t)stateval;
1689 }
1690
1691 if (cb.cb_zhp)
1692 zpool_close(cb.cb_zhp);
1693
1694 nvlist_free(config);
1695 *inuse = ret;
1696 return (0);
1697 }
Unleashed OS