]> git.proxmox.com Git - mirror_zfs-debian.git/blob - lib/libzfs/libzfs_import.c
Retire zpool_id infrastructure
[mirror_zfs-debian.git] / lib / libzfs / libzfs_import.c
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
58 #include <sys/vdev_impl.h>
59 #ifdef HAVE_LIBBLKID
60 #include <blkid/blkid.h>
61 #endif
62
63 #include "libzfs.h"
64 #include "libzfs_impl.h"
65
66 /*
67 * Intermediate structures used to gather configuration information.
68 */
69 typedef struct config_entry {
70 uint64_t ce_txg;
71 nvlist_t *ce_config;
72 struct config_entry *ce_next;
73 } config_entry_t;
74
75 typedef struct vdev_entry {
76 uint64_t ve_guid;
77 config_entry_t *ve_configs;
78 struct vdev_entry *ve_next;
79 } vdev_entry_t;
80
81 typedef struct pool_entry {
82 uint64_t pe_guid;
83 vdev_entry_t *pe_vdevs;
84 struct pool_entry *pe_next;
85 } pool_entry_t;
86
87 typedef struct name_entry {
88 char *ne_name;
89 uint64_t ne_guid;
90 uint64_t ne_order;
91 struct name_entry *ne_next;
92 } name_entry_t;
93
94 typedef struct pool_list {
95 pool_entry_t *pools;
96 name_entry_t *names;
97 } pool_list_t;
98
99 static char *
100 get_devid(const char *path)
101 {
102 int fd;
103 ddi_devid_t devid;
104 char *minor, *ret;
105
106 if ((fd = open(path, O_RDONLY)) < 0)
107 return (NULL);
108
109 minor = NULL;
110 ret = NULL;
111 if (devid_get(fd, &devid) == 0) {
112 if (devid_get_minor_name(fd, &minor) == 0)
113 ret = devid_str_encode(devid, minor);
114 if (minor != NULL)
115 devid_str_free(minor);
116 devid_free(devid);
117 }
118 (void) close(fd);
119
120 return (ret);
121 }
122
123
124 /*
125 * Go through and fix up any path and/or devid information for the given vdev
126 * configuration.
127 */
128 static int
129 fix_paths(nvlist_t *nv, name_entry_t *names)
130 {
131 nvlist_t **child;
132 uint_t c, children;
133 uint64_t guid;
134 name_entry_t *ne, *best;
135 char *path, *devid;
136
137 if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN,
138 &child, &children) == 0) {
139 for (c = 0; c < children; c++)
140 if (fix_paths(child[c], names) != 0)
141 return (-1);
142 return (0);
143 }
144
145 /*
146 * This is a leaf (file or disk) vdev. In either case, go through
147 * the name list and see if we find a matching guid. If so, replace
148 * the path and see if we can calculate a new devid.
149 *
150 * There may be multiple names associated with a particular guid, in
151 * which case we have overlapping partitions or multiple paths to the
152 * same disk. In this case we prefer to use the path name which
153 * matches the ZPOOL_CONFIG_PATH. If no matching entry is found we
154 * use the lowest order device which corresponds to the first match
155 * while traversing the ZPOOL_IMPORT_PATH search path.
156 */
157 verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID, &guid) == 0);
158 if (nvlist_lookup_string(nv, ZPOOL_CONFIG_PATH, &path) != 0)
159 path = NULL;
160
161 best = NULL;
162 for (ne = names; ne != NULL; ne = ne->ne_next) {
163 if (ne->ne_guid == guid) {
164
165 if (path == NULL) {
166 best = ne;
167 break;
168 }
169
170 if ((strlen(path) == strlen(ne->ne_name)) &&
171 !strncmp(path, ne->ne_name, strlen(path))) {
172 best = ne;
173 break;
174 }
175
176 if (best == NULL || ne->ne_order < best->ne_order)
177 best = ne;
178 }
179 }
180
181 if (best == NULL)
182 return (0);
183
184 if (nvlist_add_string(nv, ZPOOL_CONFIG_PATH, best->ne_name) != 0)
185 return (-1);
186
187 if ((devid = get_devid(best->ne_name)) == NULL) {
188 (void) nvlist_remove_all(nv, ZPOOL_CONFIG_DEVID);
189 } else {
190 if (nvlist_add_string(nv, ZPOOL_CONFIG_DEVID, devid) != 0)
191 return (-1);
192 devid_str_free(devid);
193 }
194
195 return (0);
196 }
197
198 /*
199 * Add the given configuration to the list of known devices.
200 */
201 static int
202 add_config(libzfs_handle_t *hdl, pool_list_t *pl, const char *path,
203 int order, nvlist_t *config)
204 {
205 uint64_t pool_guid, vdev_guid, top_guid, txg, state;
206 pool_entry_t *pe;
207 vdev_entry_t *ve;
208 config_entry_t *ce;
209 name_entry_t *ne;
210
211 /*
212 * If this is a hot spare not currently in use or level 2 cache
213 * device, add it to the list of names to translate, but don't do
214 * anything else.
215 */
216 if (nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_STATE,
217 &state) == 0 &&
218 (state == POOL_STATE_SPARE || state == POOL_STATE_L2CACHE) &&
219 nvlist_lookup_uint64(config, ZPOOL_CONFIG_GUID, &vdev_guid) == 0) {
220 if ((ne = zfs_alloc(hdl, sizeof (name_entry_t))) == NULL)
221 return (-1);
222
223 if ((ne->ne_name = zfs_strdup(hdl, path)) == NULL) {
224 free(ne);
225 return (-1);
226 }
227 ne->ne_guid = vdev_guid;
228 ne->ne_order = order;
229 ne->ne_next = pl->names;
230 pl->names = ne;
231 return (0);
232 }
233
234 /*
235 * If we have a valid config but cannot read any of these fields, then
236 * it means we have a half-initialized label. In vdev_label_init()
237 * we write a label with txg == 0 so that we can identify the device
238 * in case the user refers to the same disk later on. If we fail to
239 * create the pool, we'll be left with a label in this state
240 * which should not be considered part of a valid pool.
241 */
242 if (nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID,
243 &pool_guid) != 0 ||
244 nvlist_lookup_uint64(config, ZPOOL_CONFIG_GUID,
245 &vdev_guid) != 0 ||
246 nvlist_lookup_uint64(config, ZPOOL_CONFIG_TOP_GUID,
247 &top_guid) != 0 ||
248 nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_TXG,
249 &txg) != 0 || txg == 0) {
250 nvlist_free(config);
251 return (0);
252 }
253
254 /*
255 * First, see if we know about this pool. If not, then add it to the
256 * list of known pools.
257 */
258 for (pe = pl->pools; pe != NULL; pe = pe->pe_next) {
259 if (pe->pe_guid == pool_guid)
260 break;
261 }
262
263 if (pe == NULL) {
264 if ((pe = zfs_alloc(hdl, sizeof (pool_entry_t))) == NULL) {
265 nvlist_free(config);
266 return (-1);
267 }
268 pe->pe_guid = pool_guid;
269 pe->pe_next = pl->pools;
270 pl->pools = pe;
271 }
272
273 /*
274 * Second, see if we know about this toplevel vdev. Add it if its
275 * missing.
276 */
277 for (ve = pe->pe_vdevs; ve != NULL; ve = ve->ve_next) {
278 if (ve->ve_guid == top_guid)
279 break;
280 }
281
282 if (ve == NULL) {
283 if ((ve = zfs_alloc(hdl, sizeof (vdev_entry_t))) == NULL) {
284 nvlist_free(config);
285 return (-1);
286 }
287 ve->ve_guid = top_guid;
288 ve->ve_next = pe->pe_vdevs;
289 pe->pe_vdevs = ve;
290 }
291
292 /*
293 * Third, see if we have a config with a matching transaction group. If
294 * so, then we do nothing. Otherwise, add it to the list of known
295 * configs.
296 */
297 for (ce = ve->ve_configs; ce != NULL; ce = ce->ce_next) {
298 if (ce->ce_txg == txg)
299 break;
300 }
301
302 if (ce == NULL) {
303 if ((ce = zfs_alloc(hdl, sizeof (config_entry_t))) == NULL) {
304 nvlist_free(config);
305 return (-1);
306 }
307 ce->ce_txg = txg;
308 ce->ce_config = config;
309 ce->ce_next = ve->ve_configs;
310 ve->ve_configs = ce;
311 } else {
312 nvlist_free(config);
313 }
314
315 /*
316 * At this point we've successfully added our config to the list of
317 * known configs. The last thing to do is add the vdev guid -> path
318 * mappings so that we can fix up the configuration as necessary before
319 * doing the import.
320 */
321 if ((ne = zfs_alloc(hdl, sizeof (name_entry_t))) == NULL)
322 return (-1);
323
324 if ((ne->ne_name = zfs_strdup(hdl, path)) == NULL) {
325 free(ne);
326 return (-1);
327 }
328
329 ne->ne_guid = vdev_guid;
330 ne->ne_order = order;
331 ne->ne_next = pl->names;
332 pl->names = ne;
333
334 return (0);
335 }
336
337 /*
338 * Returns true if the named pool matches the given GUID.
339 */
340 static int
341 pool_active(libzfs_handle_t *hdl, const char *name, uint64_t guid,
342 boolean_t *isactive)
343 {
344 zpool_handle_t *zhp;
345 uint64_t theguid;
346
347 if (zpool_open_silent(hdl, name, &zhp) != 0)
348 return (-1);
349
350 if (zhp == NULL) {
351 *isactive = B_FALSE;
352 return (0);
353 }
354
355 verify(nvlist_lookup_uint64(zhp->zpool_config, ZPOOL_CONFIG_POOL_GUID,
356 &theguid) == 0);
357
358 zpool_close(zhp);
359
360 *isactive = (theguid == guid);
361 return (0);
362 }
363
364 static nvlist_t *
365 refresh_config(libzfs_handle_t *hdl, nvlist_t *config)
366 {
367 nvlist_t *nvl;
368 zfs_cmd_t zc = { "\0", "\0", "\0", "\0", 0 };
369 int err;
370
371 if (zcmd_write_conf_nvlist(hdl, &zc, config) != 0)
372 return (NULL);
373
374 if (zcmd_alloc_dst_nvlist(hdl, &zc,
375 zc.zc_nvlist_conf_size * 2) != 0) {
376 zcmd_free_nvlists(&zc);
377 return (NULL);
378 }
379
380 while ((err = ioctl(hdl->libzfs_fd, ZFS_IOC_POOL_TRYIMPORT,
381 &zc)) != 0 && errno == ENOMEM) {
382 if (zcmd_expand_dst_nvlist(hdl, &zc) != 0) {
383 zcmd_free_nvlists(&zc);
384 return (NULL);
385 }
386 }
387
388 if (err) {
389 zcmd_free_nvlists(&zc);
390 return (NULL);
391 }
392
393 if (zcmd_read_dst_nvlist(hdl, &zc, &nvl) != 0) {
394 zcmd_free_nvlists(&zc);
395 return (NULL);
396 }
397
398 zcmd_free_nvlists(&zc);
399 return (nvl);
400 }
401
402 /*
403 * Determine if the vdev id is a hole in the namespace.
404 */
405 boolean_t
406 vdev_is_hole(uint64_t *hole_array, uint_t holes, uint_t id)
407 {
408 int c;
409
410 for (c = 0; c < holes; c++) {
411
412 /* Top-level is a hole */
413 if (hole_array[c] == id)
414 return (B_TRUE);
415 }
416 return (B_FALSE);
417 }
418
419 /*
420 * Convert our list of pools into the definitive set of configurations. We
421 * start by picking the best config for each toplevel vdev. Once that's done,
422 * we assemble the toplevel vdevs into a full config for the pool. We make a
423 * pass to fix up any incorrect paths, and then add it to the main list to
424 * return to the user.
425 */
426 static nvlist_t *
427 get_configs(libzfs_handle_t *hdl, pool_list_t *pl, boolean_t active_ok)
428 {
429 pool_entry_t *pe;
430 vdev_entry_t *ve;
431 config_entry_t *ce;
432 nvlist_t *ret = NULL, *config = NULL, *tmp = NULL, *nvtop, *nvroot;
433 nvlist_t **spares, **l2cache;
434 uint_t i, nspares, nl2cache;
435 boolean_t config_seen;
436 uint64_t best_txg;
437 char *name, *hostname = NULL;
438 uint64_t guid;
439 uint_t children = 0;
440 nvlist_t **child = NULL;
441 uint_t holes;
442 uint64_t *hole_array, max_id;
443 uint_t c;
444 boolean_t isactive;
445 uint64_t hostid;
446 nvlist_t *nvl;
447 boolean_t found_one = B_FALSE;
448 boolean_t valid_top_config = B_FALSE;
449
450 if (nvlist_alloc(&ret, 0, 0) != 0)
451 goto nomem;
452
453 for (pe = pl->pools; pe != NULL; pe = pe->pe_next) {
454 uint64_t id, max_txg = 0;
455
456 if (nvlist_alloc(&config, NV_UNIQUE_NAME, 0) != 0)
457 goto nomem;
458 config_seen = B_FALSE;
459
460 /*
461 * Iterate over all toplevel vdevs. Grab the pool configuration
462 * from the first one we find, and then go through the rest and
463 * add them as necessary to the 'vdevs' member of the config.
464 */
465 for (ve = pe->pe_vdevs; ve != NULL; ve = ve->ve_next) {
466
467 /*
468 * Determine the best configuration for this vdev by
469 * selecting the config with the latest transaction
470 * group.
471 */
472 best_txg = 0;
473 for (ce = ve->ve_configs; ce != NULL;
474 ce = ce->ce_next) {
475
476 if (ce->ce_txg > best_txg) {
477 tmp = ce->ce_config;
478 best_txg = ce->ce_txg;
479 }
480 }
481
482 /*
483 * We rely on the fact that the max txg for the
484 * pool will contain the most up-to-date information
485 * about the valid top-levels in the vdev namespace.
486 */
487 if (best_txg > max_txg) {
488 (void) nvlist_remove(config,
489 ZPOOL_CONFIG_VDEV_CHILDREN,
490 DATA_TYPE_UINT64);
491 (void) nvlist_remove(config,
492 ZPOOL_CONFIG_HOLE_ARRAY,
493 DATA_TYPE_UINT64_ARRAY);
494
495 max_txg = best_txg;
496 hole_array = NULL;
497 holes = 0;
498 max_id = 0;
499 valid_top_config = B_FALSE;
500
501 if (nvlist_lookup_uint64(tmp,
502 ZPOOL_CONFIG_VDEV_CHILDREN, &max_id) == 0) {
503 verify(nvlist_add_uint64(config,
504 ZPOOL_CONFIG_VDEV_CHILDREN,
505 max_id) == 0);
506 valid_top_config = B_TRUE;
507 }
508
509 if (nvlist_lookup_uint64_array(tmp,
510 ZPOOL_CONFIG_HOLE_ARRAY, &hole_array,
511 &holes) == 0) {
512 verify(nvlist_add_uint64_array(config,
513 ZPOOL_CONFIG_HOLE_ARRAY,
514 hole_array, holes) == 0);
515 }
516 }
517
518 if (!config_seen) {
519 /*
520 * Copy the relevant pieces of data to the pool
521 * configuration:
522 *
523 * version
524 * pool guid
525 * name
526 * pool txg (if available)
527 * comment (if available)
528 * pool state
529 * hostid (if available)
530 * hostname (if available)
531 */
532 uint64_t state, version, pool_txg;
533 char *comment = NULL;
534
535 version = fnvlist_lookup_uint64(tmp,
536 ZPOOL_CONFIG_VERSION);
537 fnvlist_add_uint64(config,
538 ZPOOL_CONFIG_VERSION, version);
539 guid = fnvlist_lookup_uint64(tmp,
540 ZPOOL_CONFIG_POOL_GUID);
541 fnvlist_add_uint64(config,
542 ZPOOL_CONFIG_POOL_GUID, guid);
543 name = fnvlist_lookup_string(tmp,
544 ZPOOL_CONFIG_POOL_NAME);
545 fnvlist_add_string(config,
546 ZPOOL_CONFIG_POOL_NAME, name);
547
548 if (nvlist_lookup_uint64(tmp,
549 ZPOOL_CONFIG_POOL_TXG, &pool_txg) == 0)
550 fnvlist_add_uint64(config,
551 ZPOOL_CONFIG_POOL_TXG, pool_txg);
552
553 if (nvlist_lookup_string(tmp,
554 ZPOOL_CONFIG_COMMENT, &comment) == 0)
555 fnvlist_add_string(config,
556 ZPOOL_CONFIG_COMMENT, comment);
557
558 state = fnvlist_lookup_uint64(tmp,
559 ZPOOL_CONFIG_POOL_STATE);
560 fnvlist_add_uint64(config,
561 ZPOOL_CONFIG_POOL_STATE, state);
562
563 hostid = 0;
564 if (nvlist_lookup_uint64(tmp,
565 ZPOOL_CONFIG_HOSTID, &hostid) == 0) {
566 fnvlist_add_uint64(config,
567 ZPOOL_CONFIG_HOSTID, hostid);
568 hostname = fnvlist_lookup_string(tmp,
569 ZPOOL_CONFIG_HOSTNAME);
570 fnvlist_add_string(config,
571 ZPOOL_CONFIG_HOSTNAME, hostname);
572 }
573
574 config_seen = B_TRUE;
575 }
576
577 /*
578 * Add this top-level vdev to the child array.
579 */
580 verify(nvlist_lookup_nvlist(tmp,
581 ZPOOL_CONFIG_VDEV_TREE, &nvtop) == 0);
582 verify(nvlist_lookup_uint64(nvtop, ZPOOL_CONFIG_ID,
583 &id) == 0);
584
585 if (id >= children) {
586 nvlist_t **newchild;
587
588 newchild = zfs_alloc(hdl, (id + 1) *
589 sizeof (nvlist_t *));
590 if (newchild == NULL)
591 goto nomem;
592
593 for (c = 0; c < children; c++)
594 newchild[c] = child[c];
595
596 free(child);
597 child = newchild;
598 children = id + 1;
599 }
600 if (nvlist_dup(nvtop, &child[id], 0) != 0)
601 goto nomem;
602
603 }
604
605 /*
606 * If we have information about all the top-levels then
607 * clean up the nvlist which we've constructed. This
608 * means removing any extraneous devices that are
609 * beyond the valid range or adding devices to the end
610 * of our array which appear to be missing.
611 */
612 if (valid_top_config) {
613 if (max_id < children) {
614 for (c = max_id; c < children; c++)
615 nvlist_free(child[c]);
616 children = max_id;
617 } else if (max_id > children) {
618 nvlist_t **newchild;
619
620 newchild = zfs_alloc(hdl, (max_id) *
621 sizeof (nvlist_t *));
622 if (newchild == NULL)
623 goto nomem;
624
625 for (c = 0; c < children; c++)
626 newchild[c] = child[c];
627
628 free(child);
629 child = newchild;
630 children = max_id;
631 }
632 }
633
634 verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID,
635 &guid) == 0);
636
637 /*
638 * The vdev namespace may contain holes as a result of
639 * device removal. We must add them back into the vdev
640 * tree before we process any missing devices.
641 */
642 if (holes > 0) {
643 ASSERT(valid_top_config);
644
645 for (c = 0; c < children; c++) {
646 nvlist_t *holey;
647
648 if (child[c] != NULL ||
649 !vdev_is_hole(hole_array, holes, c))
650 continue;
651
652 if (nvlist_alloc(&holey, NV_UNIQUE_NAME,
653 0) != 0)
654 goto nomem;
655
656 /*
657 * Holes in the namespace are treated as
658 * "hole" top-level vdevs and have a
659 * special flag set on them.
660 */
661 if (nvlist_add_string(holey,
662 ZPOOL_CONFIG_TYPE,
663 VDEV_TYPE_HOLE) != 0 ||
664 nvlist_add_uint64(holey,
665 ZPOOL_CONFIG_ID, c) != 0 ||
666 nvlist_add_uint64(holey,
667 ZPOOL_CONFIG_GUID, 0ULL) != 0)
668 goto nomem;
669 child[c] = holey;
670 }
671 }
672
673 /*
674 * Look for any missing top-level vdevs. If this is the case,
675 * create a faked up 'missing' vdev as a placeholder. We cannot
676 * simply compress the child array, because the kernel performs
677 * certain checks to make sure the vdev IDs match their location
678 * in the configuration.
679 */
680 for (c = 0; c < children; c++) {
681 if (child[c] == NULL) {
682 nvlist_t *missing;
683 if (nvlist_alloc(&missing, NV_UNIQUE_NAME,
684 0) != 0)
685 goto nomem;
686 if (nvlist_add_string(missing,
687 ZPOOL_CONFIG_TYPE,
688 VDEV_TYPE_MISSING) != 0 ||
689 nvlist_add_uint64(missing,
690 ZPOOL_CONFIG_ID, c) != 0 ||
691 nvlist_add_uint64(missing,
692 ZPOOL_CONFIG_GUID, 0ULL) != 0) {
693 nvlist_free(missing);
694 goto nomem;
695 }
696 child[c] = missing;
697 }
698 }
699
700 /*
701 * Put all of this pool's top-level vdevs into a root vdev.
702 */
703 if (nvlist_alloc(&nvroot, NV_UNIQUE_NAME, 0) != 0)
704 goto nomem;
705 if (nvlist_add_string(nvroot, ZPOOL_CONFIG_TYPE,
706 VDEV_TYPE_ROOT) != 0 ||
707 nvlist_add_uint64(nvroot, ZPOOL_CONFIG_ID, 0ULL) != 0 ||
708 nvlist_add_uint64(nvroot, ZPOOL_CONFIG_GUID, guid) != 0 ||
709 nvlist_add_nvlist_array(nvroot, ZPOOL_CONFIG_CHILDREN,
710 child, children) != 0) {
711 nvlist_free(nvroot);
712 goto nomem;
713 }
714
715 for (c = 0; c < children; c++)
716 nvlist_free(child[c]);
717 free(child);
718 children = 0;
719 child = NULL;
720
721 /*
722 * Go through and fix up any paths and/or devids based on our
723 * known list of vdev GUID -> path mappings.
724 */
725 if (fix_paths(nvroot, pl->names) != 0) {
726 nvlist_free(nvroot);
727 goto nomem;
728 }
729
730 /*
731 * Add the root vdev to this pool's configuration.
732 */
733 if (nvlist_add_nvlist(config, ZPOOL_CONFIG_VDEV_TREE,
734 nvroot) != 0) {
735 nvlist_free(nvroot);
736 goto nomem;
737 }
738 nvlist_free(nvroot);
739
740 /*
741 * zdb uses this path to report on active pools that were
742 * imported or created using -R.
743 */
744 if (active_ok)
745 goto add_pool;
746
747 /*
748 * Determine if this pool is currently active, in which case we
749 * can't actually import it.
750 */
751 verify(nvlist_lookup_string(config, ZPOOL_CONFIG_POOL_NAME,
752 &name) == 0);
753 verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID,
754 &guid) == 0);
755
756 if (pool_active(hdl, name, guid, &isactive) != 0)
757 goto error;
758
759 if (isactive) {
760 nvlist_free(config);
761 config = NULL;
762 continue;
763 }
764
765 if ((nvl = refresh_config(hdl, config)) == NULL) {
766 nvlist_free(config);
767 config = NULL;
768 continue;
769 }
770
771 nvlist_free(config);
772 config = nvl;
773
774 /*
775 * Go through and update the paths for spares, now that we have
776 * them.
777 */
778 verify(nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE,
779 &nvroot) == 0);
780 if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_SPARES,
781 &spares, &nspares) == 0) {
782 for (i = 0; i < nspares; i++) {
783 if (fix_paths(spares[i], pl->names) != 0)
784 goto nomem;
785 }
786 }
787
788 /*
789 * Update the paths for l2cache devices.
790 */
791 if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_L2CACHE,
792 &l2cache, &nl2cache) == 0) {
793 for (i = 0; i < nl2cache; i++) {
794 if (fix_paths(l2cache[i], pl->names) != 0)
795 goto nomem;
796 }
797 }
798
799 /*
800 * Restore the original information read from the actual label.
801 */
802 (void) nvlist_remove(config, ZPOOL_CONFIG_HOSTID,
803 DATA_TYPE_UINT64);
804 (void) nvlist_remove(config, ZPOOL_CONFIG_HOSTNAME,
805 DATA_TYPE_STRING);
806 if (hostid != 0) {
807 verify(nvlist_add_uint64(config, ZPOOL_CONFIG_HOSTID,
808 hostid) == 0);
809 verify(nvlist_add_string(config, ZPOOL_CONFIG_HOSTNAME,
810 hostname) == 0);
811 }
812
813 add_pool:
814 /*
815 * Add this pool to the list of configs.
816 */
817 verify(nvlist_lookup_string(config, ZPOOL_CONFIG_POOL_NAME,
818 &name) == 0);
819 if (nvlist_add_nvlist(ret, name, config) != 0)
820 goto nomem;
821
822 found_one = B_TRUE;
823 nvlist_free(config);
824 config = NULL;
825 }
826
827 if (!found_one) {
828 nvlist_free(ret);
829 ret = NULL;
830 }
831
832 return (ret);
833
834 nomem:
835 (void) no_memory(hdl);
836 error:
837 nvlist_free(config);
838 nvlist_free(ret);
839 for (c = 0; c < children; c++)
840 nvlist_free(child[c]);
841 free(child);
842
843 return (NULL);
844 }
845
846 /*
847 * Return the offset of the given label.
848 */
849 static uint64_t
850 label_offset(uint64_t size, int l)
851 {
852 ASSERT(P2PHASE_TYPED(size, sizeof (vdev_label_t), uint64_t) == 0);
853 return (l * sizeof (vdev_label_t) + (l < VDEV_LABELS / 2 ?
854 0 : size - VDEV_LABELS * sizeof (vdev_label_t)));
855 }
856
857 /*
858 * Given a file descriptor, read the label information and return an nvlist
859 * describing the configuration, if there is one.
860 */
861 int
862 zpool_read_label(int fd, nvlist_t **config)
863 {
864 struct stat64 statbuf;
865 int l;
866 vdev_label_t *label;
867 uint64_t state, txg, size;
868
869 *config = NULL;
870
871 if (fstat64(fd, &statbuf) == -1)
872 return (0);
873 size = P2ALIGN_TYPED(statbuf.st_size, sizeof (vdev_label_t), uint64_t);
874
875 if ((label = malloc(sizeof (vdev_label_t))) == NULL)
876 return (-1);
877
878 for (l = 0; l < VDEV_LABELS; l++) {
879 if (pread64(fd, label, sizeof (vdev_label_t),
880 label_offset(size, l)) != sizeof (vdev_label_t))
881 continue;
882
883 if (nvlist_unpack(label->vl_vdev_phys.vp_nvlist,
884 sizeof (label->vl_vdev_phys.vp_nvlist), config, 0) != 0)
885 continue;
886
887 if (nvlist_lookup_uint64(*config, ZPOOL_CONFIG_POOL_STATE,
888 &state) != 0 || state > POOL_STATE_L2CACHE) {
889 nvlist_free(*config);
890 continue;
891 }
892
893 if (state != POOL_STATE_SPARE && state != POOL_STATE_L2CACHE &&
894 (nvlist_lookup_uint64(*config, ZPOOL_CONFIG_POOL_TXG,
895 &txg) != 0 || txg == 0)) {
896 nvlist_free(*config);
897 continue;
898 }
899
900 free(label);
901 return (0);
902 }
903
904 free(label);
905 *config = NULL;
906 return (0);
907 }
908
909 #ifdef HAVE_LIBBLKID
910 /*
911 * Use libblkid to quickly search for zfs devices
912 */
913 static int
914 zpool_find_import_blkid(libzfs_handle_t *hdl, pool_list_t *pools)
915 {
916 blkid_cache cache;
917 blkid_dev_iterate iter;
918 blkid_dev dev;
919 const char *devname;
920 nvlist_t *config;
921 int fd, err;
922
923 err = blkid_get_cache(&cache, NULL);
924 if (err != 0) {
925 (void) zfs_error_fmt(hdl, EZFS_BADCACHE,
926 dgettext(TEXT_DOMAIN, "blkid_get_cache() %d"), err);
927 goto err_blkid1;
928 }
929
930 err = blkid_probe_all(cache);
931 if (err != 0) {
932 (void) zfs_error_fmt(hdl, EZFS_BADCACHE,
933 dgettext(TEXT_DOMAIN, "blkid_probe_all() %d"), err);
934 goto err_blkid2;
935 }
936
937 iter = blkid_dev_iterate_begin(cache);
938 if (iter == NULL) {
939 (void) zfs_error_fmt(hdl, EZFS_BADCACHE,
940 dgettext(TEXT_DOMAIN, "blkid_dev_iterate_begin()"));
941 goto err_blkid2;
942 }
943
944 err = blkid_dev_set_search(iter, "TYPE", "zfs");
945 if (err != 0) {
946 (void) zfs_error_fmt(hdl, EZFS_BADCACHE,
947 dgettext(TEXT_DOMAIN, "blkid_dev_set_search() %d"), err);
948 goto err_blkid3;
949 }
950
951 while (blkid_dev_next(iter, &dev) == 0) {
952 devname = blkid_dev_devname(dev);
953 if ((fd = open64(devname, O_RDONLY)) < 0)
954 continue;
955
956 err = zpool_read_label(fd, &config);
957 (void) close(fd);
958
959 if (err != 0) {
960 (void) no_memory(hdl);
961 goto err_blkid3;
962 }
963
964 if (config != NULL) {
965 err = add_config(hdl, pools, devname, 0, config);
966 if (err != 0)
967 goto err_blkid3;
968 }
969 }
970
971 err_blkid3:
972 blkid_dev_iterate_end(iter);
973 err_blkid2:
974 blkid_put_cache(cache);
975 err_blkid1:
976 return err;
977 }
978 #endif /* HAVE_LIBBLKID */
979
980 char *
981 zpool_default_import_path[DEFAULT_IMPORT_PATH_SIZE] = {
982 "/dev/disk/by-vdev", /* Custom rules, use first if they exist */
983 "/dev/mapper", /* Use multipath devices before components */
984 "/dev/disk/by-uuid", /* Single unique entry and persistent */
985 "/dev/disk/by-id", /* May be multiple entries and persistent */
986 "/dev/disk/by-path", /* Encodes physical location and persistent */
987 "/dev/disk/by-label", /* Custom persistent labels */
988 "/dev" /* UNSAFE device names will change */
989 };
990
991 /*
992 * Given a list of directories to search, find all pools stored on disk. This
993 * includes partial pools which are not available to import. If no args are
994 * given (argc is 0), then the default directory (/dev/dsk) is searched.
995 * poolname or guid (but not both) are provided by the caller when trying
996 * to import a specific pool.
997 */
998 static nvlist_t *
999 zpool_find_import_impl(libzfs_handle_t *hdl, importargs_t *iarg)
1000 {
1001 int i, dirs = iarg->paths;
1002 DIR *dirp = NULL;
1003 struct dirent64 *dp;
1004 char path[MAXPATHLEN];
1005 char *end, **dir = iarg->path;
1006 size_t pathleft;
1007 struct stat64 statbuf;
1008 nvlist_t *ret = NULL, *config;
1009 int fd;
1010 pool_list_t pools = { 0 };
1011 pool_entry_t *pe, *penext;
1012 vdev_entry_t *ve, *venext;
1013 config_entry_t *ce, *cenext;
1014 name_entry_t *ne, *nenext;
1015
1016 verify(iarg->poolname == NULL || iarg->guid == 0);
1017
1018 if (dirs == 0) {
1019 #ifdef HAVE_LIBBLKID
1020 /* Use libblkid to scan all device for their type */
1021 if (zpool_find_import_blkid(hdl, &pools) == 0)
1022 goto skip_scanning;
1023
1024 (void) zfs_error_fmt(hdl, EZFS_BADCACHE,
1025 dgettext(TEXT_DOMAIN, "blkid failure falling back "
1026 "to manual probing"));
1027 #endif /* HAVE_LIBBLKID */
1028
1029 dir = zpool_default_import_path;
1030 dirs = DEFAULT_IMPORT_PATH_SIZE;
1031 }
1032
1033 /*
1034 * Go through and read the label configuration information from every
1035 * possible device, organizing the information according to pool GUID
1036 * and toplevel GUID.
1037 */
1038 for (i = 0; i < dirs; i++) {
1039 char *rdsk;
1040 int dfd;
1041
1042 /* use realpath to normalize the path */
1043 if (realpath(dir[i], path) == 0) {
1044
1045 /* it is safe to skip missing search paths */
1046 if (errno == ENOENT)
1047 continue;
1048
1049 zfs_error_aux(hdl, strerror(errno));
1050 (void) zfs_error_fmt(hdl, EZFS_BADPATH,
1051 dgettext(TEXT_DOMAIN, "cannot open '%s'"), dir[i]);
1052 goto error;
1053 }
1054 end = &path[strlen(path)];
1055 *end++ = '/';
1056 *end = 0;
1057 pathleft = &path[sizeof (path)] - end;
1058
1059 /*
1060 * Using raw devices instead of block devices when we're
1061 * reading the labels skips a bunch of slow operations during
1062 * close(2) processing, so we replace /dev/dsk with /dev/rdsk.
1063 */
1064 if (strcmp(path, "/dev/dsk/") == 0)
1065 rdsk = "/dev/rdsk/";
1066 else
1067 rdsk = path;
1068
1069 if ((dfd = open64(rdsk, O_RDONLY)) < 0 ||
1070 (dirp = fdopendir(dfd)) == NULL) {
1071 zfs_error_aux(hdl, strerror(errno));
1072 (void) zfs_error_fmt(hdl, EZFS_BADPATH,
1073 dgettext(TEXT_DOMAIN, "cannot open '%s'"),
1074 rdsk);
1075 goto error;
1076 }
1077
1078 /*
1079 * This is not MT-safe, but we have no MT consumers of libzfs
1080 */
1081 while ((dp = readdir64(dirp)) != NULL) {
1082 const char *name = dp->d_name;
1083 if (name[0] == '.' &&
1084 (name[1] == 0 || (name[1] == '.' && name[2] == 0)))
1085 continue;
1086
1087 /*
1088 * Skip checking devices with well known prefixes:
1089 * watchdog - A special close is required to avoid
1090 * triggering it and resetting the system.
1091 * fuse - Fuse control device.
1092 * ppp - Generic PPP driver.
1093 * tty* - Generic serial interface.
1094 * vcs* - Virtual console memory.
1095 * parport* - Parallel port interface.
1096 * lp* - Printer interface.
1097 * fd* - Floppy interface.
1098 * hpet - High Precision Event Timer, crashes qemu
1099 * when accessed from a virtual machine.
1100 * core - Symlink to /proc/kcore, causes a crash
1101 * when access from Xen dom0.
1102 */
1103 if ((strncmp(name, "watchdog", 8) == 0) ||
1104 (strncmp(name, "fuse", 4) == 0) ||
1105 (strncmp(name, "ppp", 3) == 0) ||
1106 (strncmp(name, "tty", 3) == 0) ||
1107 (strncmp(name, "vcs", 3) == 0) ||
1108 (strncmp(name, "parport", 7) == 0) ||
1109 (strncmp(name, "lp", 2) == 0) ||
1110 (strncmp(name, "fd", 2) == 0) ||
1111 (strncmp(name, "hpet", 4) == 0) ||
1112 (strncmp(name, "core", 4) == 0))
1113 continue;
1114
1115 /*
1116 * Ignore failed stats. We only want regular
1117 * files and block devices.
1118 */
1119 if ((fstatat64(dfd, name, &statbuf, 0) != 0) ||
1120 (!S_ISREG(statbuf.st_mode) &&
1121 !S_ISBLK(statbuf.st_mode)))
1122 continue;
1123
1124 if ((fd = openat64(dfd, name, O_RDONLY)) < 0)
1125 continue;
1126
1127 if ((zpool_read_label(fd, &config)) != 0) {
1128 (void) close(fd);
1129 (void) no_memory(hdl);
1130 goto error;
1131 }
1132
1133 (void) close(fd);
1134
1135 if (config != NULL) {
1136 boolean_t matched = B_TRUE;
1137 char *pname;
1138
1139 if ((iarg->poolname != NULL) &&
1140 (nvlist_lookup_string(config,
1141 ZPOOL_CONFIG_POOL_NAME, &pname) == 0)) {
1142
1143 if (strcmp(iarg->poolname, pname))
1144 matched = B_FALSE;
1145
1146 } else if (iarg->guid != 0) {
1147 uint64_t this_guid;
1148
1149 matched = nvlist_lookup_uint64(config,
1150 ZPOOL_CONFIG_POOL_GUID,
1151 &this_guid) == 0 &&
1152 iarg->guid == this_guid;
1153 }
1154 if (!matched) {
1155 nvlist_free(config);
1156 config = NULL;
1157 continue;
1158 }
1159 /* use the non-raw path for the config */
1160 (void) strlcpy(end, name, pathleft);
1161 if (add_config(hdl, &pools, path, i+1, config))
1162 goto error;
1163 }
1164 }
1165
1166 (void) closedir(dirp);
1167 dirp = NULL;
1168 }
1169
1170 #ifdef HAVE_LIBBLKID
1171 skip_scanning:
1172 #endif
1173 ret = get_configs(hdl, &pools, iarg->can_be_active);
1174
1175 error:
1176 for (pe = pools.pools; pe != NULL; pe = penext) {
1177 penext = pe->pe_next;
1178 for (ve = pe->pe_vdevs; ve != NULL; ve = venext) {
1179 venext = ve->ve_next;
1180 for (ce = ve->ve_configs; ce != NULL; ce = cenext) {
1181 cenext = ce->ce_next;
1182 if (ce->ce_config)
1183 nvlist_free(ce->ce_config);
1184 free(ce);
1185 }
1186 free(ve);
1187 }
1188 free(pe);
1189 }
1190
1191 for (ne = pools.names; ne != NULL; ne = nenext) {
1192 nenext = ne->ne_next;
1193 if (ne->ne_name)
1194 free(ne->ne_name);
1195 free(ne);
1196 }
1197
1198 if (dirp)
1199 (void) closedir(dirp);
1200
1201 return (ret);
1202 }
1203
1204 nvlist_t *
1205 zpool_find_import(libzfs_handle_t *hdl, int argc, char **argv)
1206 {
1207 importargs_t iarg = { 0 };
1208
1209 iarg.paths = argc;
1210 iarg.path = argv;
1211
1212 return (zpool_find_import_impl(hdl, &iarg));
1213 }
1214
1215 /*
1216 * Given a cache file, return the contents as a list of importable pools.
1217 * poolname or guid (but not both) are provided by the caller when trying
1218 * to import a specific pool.
1219 */
1220 nvlist_t *
1221 zpool_find_import_cached(libzfs_handle_t *hdl, const char *cachefile,
1222 char *poolname, uint64_t guid)
1223 {
1224 char *buf;
1225 int fd;
1226 struct stat64 statbuf;
1227 nvlist_t *raw, *src, *dst;
1228 nvlist_t *pools;
1229 nvpair_t *elem;
1230 char *name;
1231 uint64_t this_guid;
1232 boolean_t active;
1233
1234 verify(poolname == NULL || guid == 0);
1235
1236 if ((fd = open(cachefile, O_RDONLY)) < 0) {
1237 zfs_error_aux(hdl, "%s", strerror(errno));
1238 (void) zfs_error(hdl, EZFS_BADCACHE,
1239 dgettext(TEXT_DOMAIN, "failed to open cache file"));
1240 return (NULL);
1241 }
1242
1243 if (fstat64(fd, &statbuf) != 0) {
1244 zfs_error_aux(hdl, "%s", strerror(errno));
1245 (void) close(fd);
1246 (void) zfs_error(hdl, EZFS_BADCACHE,
1247 dgettext(TEXT_DOMAIN, "failed to get size of cache file"));
1248 return (NULL);
1249 }
1250
1251 if ((buf = zfs_alloc(hdl, statbuf.st_size)) == NULL) {
1252 (void) close(fd);
1253 return (NULL);
1254 }
1255
1256 if (read(fd, buf, statbuf.st_size) != statbuf.st_size) {
1257 (void) close(fd);
1258 free(buf);
1259 (void) zfs_error(hdl, EZFS_BADCACHE,
1260 dgettext(TEXT_DOMAIN,
1261 "failed to read cache file contents"));
1262 return (NULL);
1263 }
1264
1265 (void) close(fd);
1266
1267 if (nvlist_unpack(buf, statbuf.st_size, &raw, 0) != 0) {
1268 free(buf);
1269 (void) zfs_error(hdl, EZFS_BADCACHE,
1270 dgettext(TEXT_DOMAIN,
1271 "invalid or corrupt cache file contents"));
1272 return (NULL);
1273 }
1274
1275 free(buf);
1276
1277 /*
1278 * Go through and get the current state of the pools and refresh their
1279 * state.
1280 */
1281 if (nvlist_alloc(&pools, 0, 0) != 0) {
1282 (void) no_memory(hdl);
1283 nvlist_free(raw);
1284 return (NULL);
1285 }
1286
1287 elem = NULL;
1288 while ((elem = nvlist_next_nvpair(raw, elem)) != NULL) {
1289 verify(nvpair_value_nvlist(elem, &src) == 0);
1290
1291 verify(nvlist_lookup_string(src, ZPOOL_CONFIG_POOL_NAME,
1292 &name) == 0);
1293 if (poolname != NULL && strcmp(poolname, name) != 0)
1294 continue;
1295
1296 verify(nvlist_lookup_uint64(src, ZPOOL_CONFIG_POOL_GUID,
1297 &this_guid) == 0);
1298 if (guid != 0) {
1299 verify(nvlist_lookup_uint64(src, ZPOOL_CONFIG_POOL_GUID,
1300 &this_guid) == 0);
1301 if (guid != this_guid)
1302 continue;
1303 }
1304
1305 if (pool_active(hdl, name, this_guid, &active) != 0) {
1306 nvlist_free(raw);
1307 nvlist_free(pools);
1308 return (NULL);
1309 }
1310
1311 if (active)
1312 continue;
1313
1314 if ((dst = refresh_config(hdl, src)) == NULL) {
1315 nvlist_free(raw);
1316 nvlist_free(pools);
1317 return (NULL);
1318 }
1319
1320 if (nvlist_add_nvlist(pools, nvpair_name(elem), dst) != 0) {
1321 (void) no_memory(hdl);
1322 nvlist_free(dst);
1323 nvlist_free(raw);
1324 nvlist_free(pools);
1325 return (NULL);
1326 }
1327 nvlist_free(dst);
1328 }
1329
1330 nvlist_free(raw);
1331 return (pools);
1332 }
1333
1334 static int
1335 name_or_guid_exists(zpool_handle_t *zhp, void *data)
1336 {
1337 importargs_t *import = data;
1338 int found = 0;
1339
1340 if (import->poolname != NULL) {
1341 char *pool_name;
1342
1343 verify(nvlist_lookup_string(zhp->zpool_config,
1344 ZPOOL_CONFIG_POOL_NAME, &pool_name) == 0);
1345 if (strcmp(pool_name, import->poolname) == 0)
1346 found = 1;
1347 } else {
1348 uint64_t pool_guid;
1349
1350 verify(nvlist_lookup_uint64(zhp->zpool_config,
1351 ZPOOL_CONFIG_POOL_GUID, &pool_guid) == 0);
1352 if (pool_guid == import->guid)
1353 found = 1;
1354 }
1355
1356 zpool_close(zhp);
1357 return (found);
1358 }
1359
1360 nvlist_t *
1361 zpool_search_import(libzfs_handle_t *hdl, importargs_t *import)
1362 {
1363 verify(import->poolname == NULL || import->guid == 0);
1364
1365 if (import->unique)
1366 import->exists = zpool_iter(hdl, name_or_guid_exists, import);
1367
1368 if (import->cachefile != NULL)
1369 return (zpool_find_import_cached(hdl, import->cachefile,
1370 import->poolname, import->guid));
1371
1372 return (zpool_find_import_impl(hdl, import));
1373 }
1374
1375 boolean_t
1376 find_guid(nvlist_t *nv, uint64_t guid)
1377 {
1378 uint64_t tmp;
1379 nvlist_t **child;
1380 uint_t c, children;
1381
1382 verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID, &tmp) == 0);
1383 if (tmp == guid)
1384 return (B_TRUE);
1385
1386 if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN,
1387 &child, &children) == 0) {
1388 for (c = 0; c < children; c++)
1389 if (find_guid(child[c], guid))
1390 return (B_TRUE);
1391 }
1392
1393 return (B_FALSE);
1394 }
1395
1396 typedef struct aux_cbdata {
1397 const char *cb_type;
1398 uint64_t cb_guid;
1399 zpool_handle_t *cb_zhp;
1400 } aux_cbdata_t;
1401
1402 static int
1403 find_aux(zpool_handle_t *zhp, void *data)
1404 {
1405 aux_cbdata_t *cbp = data;
1406 nvlist_t **list;
1407 uint_t i, count;
1408 uint64_t guid;
1409 nvlist_t *nvroot;
1410
1411 verify(nvlist_lookup_nvlist(zhp->zpool_config, ZPOOL_CONFIG_VDEV_TREE,
1412 &nvroot) == 0);
1413
1414 if (nvlist_lookup_nvlist_array(nvroot, cbp->cb_type,
1415 &list, &count) == 0) {
1416 for (i = 0; i < count; i++) {
1417 verify(nvlist_lookup_uint64(list[i],
1418 ZPOOL_CONFIG_GUID, &guid) == 0);
1419 if (guid == cbp->cb_guid) {
1420 cbp->cb_zhp = zhp;
1421 return (1);
1422 }
1423 }
1424 }
1425
1426 zpool_close(zhp);
1427 return (0);
1428 }
1429
1430 /*
1431 * Determines if the pool is in use. If so, it returns true and the state of
1432 * the pool as well as the name of the pool. Both strings are allocated and
1433 * must be freed by the caller.
1434 */
1435 int
1436 zpool_in_use(libzfs_handle_t *hdl, int fd, pool_state_t *state, char **namestr,
1437 boolean_t *inuse)
1438 {
1439 nvlist_t *config;
1440 char *name;
1441 boolean_t ret;
1442 uint64_t guid, vdev_guid;
1443 zpool_handle_t *zhp;
1444 nvlist_t *pool_config;
1445 uint64_t stateval, isspare;
1446 aux_cbdata_t cb = { 0 };
1447 boolean_t isactive;
1448
1449 *inuse = B_FALSE;
1450
1451 if (zpool_read_label(fd, &config) != 0) {
1452 (void) no_memory(hdl);
1453 return (-1);
1454 }
1455
1456 if (config == NULL)
1457 return (0);
1458
1459 verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_STATE,
1460 &stateval) == 0);
1461 verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_GUID,
1462 &vdev_guid) == 0);
1463
1464 if (stateval != POOL_STATE_SPARE && stateval != POOL_STATE_L2CACHE) {
1465 verify(nvlist_lookup_string(config, ZPOOL_CONFIG_POOL_NAME,
1466 &name) == 0);
1467 verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID,
1468 &guid) == 0);
1469 }
1470
1471 switch (stateval) {
1472 case POOL_STATE_EXPORTED:
1473 /*
1474 * A pool with an exported state may in fact be imported
1475 * read-only, so check the in-core state to see if it's
1476 * active and imported read-only. If it is, set
1477 * its state to active.
1478 */
1479 if (pool_active(hdl, name, guid, &isactive) == 0 && isactive &&
1480 (zhp = zpool_open_canfail(hdl, name)) != NULL &&
1481 zpool_get_prop_int(zhp, ZPOOL_PROP_READONLY, NULL))
1482 stateval = POOL_STATE_ACTIVE;
1483
1484 ret = B_TRUE;
1485 break;
1486
1487 case POOL_STATE_ACTIVE:
1488 /*
1489 * For an active pool, we have to determine if it's really part
1490 * of a currently active pool (in which case the pool will exist
1491 * and the guid will be the same), or whether it's part of an
1492 * active pool that was disconnected without being explicitly
1493 * exported.
1494 */
1495 if (pool_active(hdl, name, guid, &isactive) != 0) {
1496 nvlist_free(config);
1497 return (-1);
1498 }
1499
1500 if (isactive) {
1501 /*
1502 * Because the device may have been removed while
1503 * offlined, we only report it as active if the vdev is
1504 * still present in the config. Otherwise, pretend like
1505 * it's not in use.
1506 */
1507 if ((zhp = zpool_open_canfail(hdl, name)) != NULL &&
1508 (pool_config = zpool_get_config(zhp, NULL))
1509 != NULL) {
1510 nvlist_t *nvroot;
1511
1512 verify(nvlist_lookup_nvlist(pool_config,
1513 ZPOOL_CONFIG_VDEV_TREE, &nvroot) == 0);
1514 ret = find_guid(nvroot, vdev_guid);
1515 } else {
1516 ret = B_FALSE;
1517 }
1518
1519 /*
1520 * If this is an active spare within another pool, we
1521 * treat it like an unused hot spare. This allows the
1522 * user to create a pool with a hot spare that currently
1523 * in use within another pool. Since we return B_TRUE,
1524 * libdiskmgt will continue to prevent generic consumers
1525 * from using the device.
1526 */
1527 if (ret && nvlist_lookup_uint64(config,
1528 ZPOOL_CONFIG_IS_SPARE, &isspare) == 0 && isspare)
1529 stateval = POOL_STATE_SPARE;
1530
1531 if (zhp != NULL)
1532 zpool_close(zhp);
1533 } else {
1534 stateval = POOL_STATE_POTENTIALLY_ACTIVE;
1535 ret = B_TRUE;
1536 }
1537 break;
1538
1539 case POOL_STATE_SPARE:
1540 /*
1541 * For a hot spare, it can be either definitively in use, or
1542 * potentially active. To determine if it's in use, we iterate
1543 * over all pools in the system and search for one with a spare
1544 * with a matching guid.
1545 *
1546 * Due to the shared nature of spares, we don't actually report
1547 * the potentially active case as in use. This means the user
1548 * can freely create pools on the hot spares of exported pools,
1549 * but to do otherwise makes the resulting code complicated, and
1550 * we end up having to deal with this case anyway.
1551 */
1552 cb.cb_zhp = NULL;
1553 cb.cb_guid = vdev_guid;
1554 cb.cb_type = ZPOOL_CONFIG_SPARES;
1555 if (zpool_iter(hdl, find_aux, &cb) == 1) {
1556 name = (char *)zpool_get_name(cb.cb_zhp);
1557 ret = TRUE;
1558 } else {
1559 ret = FALSE;
1560 }
1561 break;
1562
1563 case POOL_STATE_L2CACHE:
1564
1565 /*
1566 * Check if any pool is currently using this l2cache device.
1567 */
1568 cb.cb_zhp = NULL;
1569 cb.cb_guid = vdev_guid;
1570 cb.cb_type = ZPOOL_CONFIG_L2CACHE;
1571 if (zpool_iter(hdl, find_aux, &cb) == 1) {
1572 name = (char *)zpool_get_name(cb.cb_zhp);
1573 ret = TRUE;
1574 } else {
1575 ret = FALSE;
1576 }
1577 break;
1578
1579 default:
1580 ret = B_FALSE;
1581 }
1582
1583
1584 if (ret) {
1585 if ((*namestr = zfs_strdup(hdl, name)) == NULL) {
1586 if (cb.cb_zhp)
1587 zpool_close(cb.cb_zhp);
1588 nvlist_free(config);
1589 return (-1);
1590 }
1591 *state = (pool_state_t)stateval;
1592 }
1593
1594 if (cb.cb_zhp)
1595 zpool_close(cb.cb_zhp);
1596
1597 nvlist_free(config);
1598 *inuse = ret;
1599 return (0);
1600 }