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 (c) 2012 by Delphix. All rights reserved.
24 * Copyright 2014 Nexenta Systems, Inc. All rights reserved.
25 * Copyright 2014 RackTop Systems.
26 */
27
28 /*
29 * Pool import support functions.
30 *
31 * To import a pool, we rely on reading the configuration information from the
32 * ZFS label of each device. If we successfully read the label, then we
33 * organize the configuration information in the following hierarchy:
34 *
35 * pool guid -> toplevel vdev guid -> label txg
36 *
37 * Duplicate entries matching this same tuple will be discarded. Once we have
38 * examined every device, we pick the best label txg config for each toplevel
39 * vdev. We then arrange these toplevel vdevs into a complete pool config, and
40 * update any paths that have changed. Finally, we attempt to import the pool
41 * using our derived config, and record the results.
42 */
43
44 #include <ctype.h>
45 #include <devid.h>
46 #include <dirent.h>
47 #include <errno.h>
48 #include <libintl.h>
49 #include <stddef.h>
50 #include <stdlib.h>
51 #include <string.h>
52 #include <sys/stat.h>
53 #include <unistd.h>
54 #include <fcntl.h>
55 #include <sys/vtoc.h>
56 #include <sys/dktp/fdisk.h>
57 #include <sys/efi_partition.h>
58 #include <thread_pool.h>
59
60 #include <sys/vdev_impl.h>
61
62 #include "libzfs.h"
63 #include "libzfs_impl.h"
64
65 /*
66 * Intermediate structures used to gather configuration information.
67 */
68 typedef struct config_entry {
69 uint64_t ce_txg;
70 nvlist_t *ce_config;
71 struct config_entry *ce_next;
72 } config_entry_t;
73
74 typedef struct vdev_entry {
75 uint64_t ve_guid;
76 config_entry_t *ve_configs;
77 struct vdev_entry *ve_next;
78 } vdev_entry_t;
79
80 typedef struct pool_entry {
81 uint64_t pe_guid;
82 vdev_entry_t *pe_vdevs;
83 struct pool_entry *pe_next;
84 } pool_entry_t;
85
86 typedef struct name_entry {
87 char *ne_name;
88 uint64_t ne_guid;
89 struct name_entry *ne_next;
90 } name_entry_t;
91
92 typedef struct pool_list {
93 pool_entry_t *pools;
94 name_entry_t *names;
95 } pool_list_t;
96
97 static char *
98 get_devid(const char *path)
99 {
100 int fd;
101 ddi_devid_t devid;
102 char *minor, *ret;
103
104 if ((fd = open(path, O_RDONLY)) < 0)
105 return (NULL);
106
107 minor = NULL;
108 ret = NULL;
109 if (devid_get(fd, &devid) == 0) {
110 if (devid_get_minor_name(fd, &minor) == 0)
111 ret = devid_str_encode(devid, minor);
112 if (minor != NULL)
113 devid_str_free(minor);
114 devid_free(devid);
115 }
116 (void) close(fd);
117
118 return (ret);
119 }
120
121
122 /*
123 * Go through and fix up any path and/or devid information for the given vdev
124 * configuration.
125 */
126 static int
127 fix_paths(nvlist_t *nv, name_entry_t *names)
128 {
129 nvlist_t **child;
130 uint_t c, children;
131 uint64_t guid;
132 name_entry_t *ne, *best;
133 char *path, *devid;
134 int matched;
135
136 if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN,
137 &child, &children) == 0) {
138 for (c = 0; c < children; c++)
139 if (fix_paths(child[c], names) != 0)
140 return (-1);
141 return (0);
142 }
143
144 /*
145 * This is a leaf (file or disk) vdev. In either case, go through
146 * the name list and see if we find a matching guid. If so, replace
147 * the path and see if we can calculate a new devid.
148 *
149 * There may be multiple names associated with a particular guid, in
150 * which case we have overlapping slices or multiple paths to the same
151 * disk. If this is the case, then we want to pick the path that is
152 * the most similar to the original, where "most similar" is the number
153 * of matching characters starting from the end of the path. This will
154 * preserve slice numbers even if the disks have been reorganized, and
155 * will also catch preferred disk names if multiple paths exist.
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 matched = 0;
162 best = NULL;
163 for (ne = names; ne != NULL; ne = ne->ne_next) {
164 if (ne->ne_guid == guid) {
165 const char *src, *dst;
166 int count;
167
168 if (path == NULL) {
169 best = ne;
170 break;
171 }
172
173 src = ne->ne_name + strlen(ne->ne_name) - 1;
174 dst = path + strlen(path) - 1;
175 for (count = 0; src >= ne->ne_name && dst >= path;
176 src--, dst--, count++)
177 if (*src != *dst)
178 break;
179
180 /*
181 * At this point, 'count' is the number of characters
182 * matched from the end.
183 */
184 if (count > matched || best == NULL) {
185 best = ne;
186 matched = count;
187 }
188 }
189 }
190
191 if (best == NULL)
192 return (0);
193
194 if (nvlist_add_string(nv, ZPOOL_CONFIG_PATH, best->ne_name) != 0)
195 return (-1);
196
197 if ((devid = get_devid(best->ne_name)) == NULL) {
198 (void) nvlist_remove_all(nv, ZPOOL_CONFIG_DEVID);
199 } else {
200 if (nvlist_add_string(nv, ZPOOL_CONFIG_DEVID, devid) != 0)
201 return (-1);
202 devid_str_free(devid);
203 }
204
205 return (0);
206 }
207
208 /*
209 * Add the given configuration to the list of known devices.
210 */
211 static int
212 add_config(libzfs_handle_t *hdl, pool_list_t *pl, const char *path,
213 nvlist_t *config)
214 {
215 uint64_t pool_guid, vdev_guid, top_guid, txg, state;
216 pool_entry_t *pe;
217 vdev_entry_t *ve;
218 config_entry_t *ce;
219 name_entry_t *ne;
220
221 /*
222 * If this is a hot spare not currently in use or level 2 cache
223 * device, add it to the list of names to translate, but don't do
224 * anything else.
225 */
226 if (nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_STATE,
227 &state) == 0 &&
228 (state == POOL_STATE_SPARE || state == POOL_STATE_L2CACHE) &&
229 nvlist_lookup_uint64(config, ZPOOL_CONFIG_GUID, &vdev_guid) == 0) {
230 if ((ne = zfs_alloc(hdl, sizeof (name_entry_t))) == NULL)
231 return (-1);
232
233 if ((ne->ne_name = zfs_strdup(hdl, path)) == NULL) {
234 free(ne);
235 return (-1);
236 }
237 ne->ne_guid = vdev_guid;
238 ne->ne_next = pl->names;
239 pl->names = ne;
240 return (0);
241 }
242
243 /*
244 * If we have a valid config but cannot read any of these fields, then
245 * it means we have a half-initialized label. In vdev_label_init()
246 * we write a label with txg == 0 so that we can identify the device
247 * in case the user refers to the same disk later on. If we fail to
248 * create the pool, we'll be left with a label in this state
249 * which should not be considered part of a valid pool.
250 */
251 if (nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID,
252 &pool_guid) != 0 ||
253 nvlist_lookup_uint64(config, ZPOOL_CONFIG_GUID,
254 &vdev_guid) != 0 ||
255 nvlist_lookup_uint64(config, ZPOOL_CONFIG_TOP_GUID,
256 &top_guid) != 0 ||
257 nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_TXG,
258 &txg) != 0 || txg == 0) {
259 nvlist_free(config);
260 return (0);
261 }
262
263 /*
264 * First, see if we know about this pool. If not, then add it to the
265 * list of known pools.
266 */
267 for (pe = pl->pools; pe != NULL; pe = pe->pe_next) {
268 if (pe->pe_guid == pool_guid)
269 break;
270 }
271
272 if (pe == NULL) {
273 if ((pe = zfs_alloc(hdl, sizeof (pool_entry_t))) == NULL) {
274 nvlist_free(config);
275 return (-1);
276 }
277 pe->pe_guid = pool_guid;
278 pe->pe_next = pl->pools;
279 pl->pools = pe;
280 }
281
282 /*
283 * Second, see if we know about this toplevel vdev. Add it if its
284 * missing.
285 */
286 for (ve = pe->pe_vdevs; ve != NULL; ve = ve->ve_next) {
287 if (ve->ve_guid == top_guid)
288 break;
289 }
290
291 if (ve == NULL) {
292 if ((ve = zfs_alloc(hdl, sizeof (vdev_entry_t))) == NULL) {
293 nvlist_free(config);
294 return (-1);
295 }
296 ve->ve_guid = top_guid;
297 ve->ve_next = pe->pe_vdevs;
298 pe->pe_vdevs = ve;
299 }
300
301 /*
302 * Third, see if we have a config with a matching transaction group. If
303 * so, then we do nothing. Otherwise, add it to the list of known
304 * configs.
305 */
306 for (ce = ve->ve_configs; ce != NULL; ce = ce->ce_next) {
307 if (ce->ce_txg == txg)
308 break;
309 }
310
311 if (ce == NULL) {
312 if ((ce = zfs_alloc(hdl, sizeof (config_entry_t))) == NULL) {
313 nvlist_free(config);
314 return (-1);
315 }
316 ce->ce_txg = txg;
317 ce->ce_config = config;
318 ce->ce_next = ve->ve_configs;
319 ve->ve_configs = ce;
320 } else {
321 nvlist_free(config);
322 }
323
324 /*
325 * At this point we've successfully added our config to the list of
326 * known configs. The last thing to do is add the vdev guid -> path
327 * mappings so that we can fix up the configuration as necessary before
328 * doing the import.
329 */
330 if ((ne = zfs_alloc(hdl, sizeof (name_entry_t))) == NULL)
331 return (-1);
332
333 if ((ne->ne_name = zfs_strdup(hdl, path)) == NULL) {
334 free(ne);
335 return (-1);
336 }
337
338 ne->ne_guid = vdev_guid;
339 ne->ne_next = pl->names;
340 pl->names = ne;
341
342 return (0);
343 }
344
345 /*
346 * Returns true if the named pool matches the given GUID.
347 */
348 static int
349 pool_active(libzfs_handle_t *hdl, const char *name, uint64_t guid,
350 boolean_t *isactive)
351 {
352 zpool_handle_t *zhp;
353 uint64_t theguid;
354
355 if (zpool_open_silent(hdl, name, &zhp) != 0)
356 return (-1);
357
358 if (zhp == NULL) {
359 *isactive = B_FALSE;
360 return (0);
361 }
362
363 verify(nvlist_lookup_uint64(zhp->zpool_config, ZPOOL_CONFIG_POOL_GUID,
364 &theguid) == 0);
365
366 zpool_close(zhp);
367
368 *isactive = (theguid == guid);
369 return (0);
370 }
371
372 static nvlist_t *
373 refresh_config(libzfs_handle_t *hdl, nvlist_t *config)
374 {
375 nvlist_t *nvl;
376 zfs_cmd_t zc = { 0 };
377 int err;
378
379 if (zcmd_write_conf_nvlist(hdl, &zc, config) != 0)
380 return (NULL);
381
382 if (zcmd_alloc_dst_nvlist(hdl, &zc,
383 zc.zc_nvlist_conf_size * 2) != 0) {
384 zcmd_free_nvlists(&zc);
385 return (NULL);
386 }
387
388 while ((err = ioctl(hdl->libzfs_fd, ZFS_IOC_POOL_TRYIMPORT,
389 &zc)) != 0 && errno == ENOMEM) {
390 if (zcmd_expand_dst_nvlist(hdl, &zc) != 0) {
391 zcmd_free_nvlists(&zc);
392 return (NULL);
393 }
394 }
395
396 if (err) {
397 zcmd_free_nvlists(&zc);
398 return (NULL);
399 }
400
401 if (zcmd_read_dst_nvlist(hdl, &zc, &nvl) != 0) {
402 zcmd_free_nvlists(&zc);
403 return (NULL);
404 }
405
406 zcmd_free_nvlists(&zc);
407 return (nvl);
408 }
409
410 /*
411 * Determine if the vdev id is a hole in the namespace.
412 */
413 boolean_t
414 vdev_is_hole(uint64_t *hole_array, uint_t holes, uint_t id)
415 {
416 for (int c = 0; c < holes; c++) {
417
418 /* Top-level is a hole */
419 if (hole_array[c] == id)
420 return (B_TRUE);
421 }
422 return (B_FALSE);
423 }
424
425 /*
426 * Convert our list of pools into the definitive set of configurations. We
427 * start by picking the best config for each toplevel vdev. Once that's done,
428 * we assemble the toplevel vdevs into a full config for the pool. We make a
429 * pass to fix up any incorrect paths, and then add it to the main list to
430 * return to the user.
431 */
432 static nvlist_t *
433 get_configs(libzfs_handle_t *hdl, pool_list_t *pl, boolean_t active_ok)
434 {
435 pool_entry_t *pe;
436 vdev_entry_t *ve;
437 config_entry_t *ce;
438 nvlist_t *ret = NULL, *config = NULL, *tmp = NULL, *nvtop, *nvroot;
439 nvlist_t **spares, **l2cache;
440 uint_t i, nspares, nl2cache;
441 boolean_t config_seen;
442 uint64_t best_txg;
443 char *name, *hostname = NULL;
444 uint64_t guid;
445 uint_t children = 0;
446 nvlist_t **child = NULL;
447 uint_t holes;
448 uint64_t *hole_array, max_id;
449 uint_t c;
450 boolean_t isactive;
451 uint64_t hostid;
452 nvlist_t *nvl;
453 boolean_t found_one = B_FALSE;
454 boolean_t valid_top_config = B_FALSE;
455
456 if (nvlist_alloc(&ret, 0, 0) != 0)
457 goto nomem;
458
459 for (pe = pl->pools; pe != NULL; pe = pe->pe_next) {
460 uint64_t id, max_txg = 0;
461
462 if (nvlist_alloc(&config, NV_UNIQUE_NAME, 0) != 0)
463 goto nomem;
464 config_seen = B_FALSE;
465
466 /*
467 * Iterate over all toplevel vdevs. Grab the pool configuration
468 * from the first one we find, and then go through the rest and
469 * add them as necessary to the 'vdevs' member of the config.
470 */
471 for (ve = pe->pe_vdevs; ve != NULL; ve = ve->ve_next) {
472
473 /*
474 * Determine the best configuration for this vdev by
475 * selecting the config with the latest transaction
476 * group.
477 */
478 best_txg = 0;
479 for (ce = ve->ve_configs; ce != NULL;
480 ce = ce->ce_next) {
481
482 if (ce->ce_txg > best_txg) {
483 tmp = ce->ce_config;
484 best_txg = ce->ce_txg;
485 }
486 }
487
488 /*
489 * We rely on the fact that the max txg for the
490 * pool will contain the most up-to-date information
491 * about the valid top-levels in the vdev namespace.
492 */
493 if (best_txg > max_txg) {
494 (void) nvlist_remove(config,
495 ZPOOL_CONFIG_VDEV_CHILDREN,
496 DATA_TYPE_UINT64);
497 (void) nvlist_remove(config,
498 ZPOOL_CONFIG_HOLE_ARRAY,
499 DATA_TYPE_UINT64_ARRAY);
500
501 max_txg = best_txg;
502 hole_array = NULL;
503 holes = 0;
504 max_id = 0;
505 valid_top_config = B_FALSE;
506
507 if (nvlist_lookup_uint64(tmp,
508 ZPOOL_CONFIG_VDEV_CHILDREN, &max_id) == 0) {
509 verify(nvlist_add_uint64(config,
510 ZPOOL_CONFIG_VDEV_CHILDREN,
511 max_id) == 0);
512 valid_top_config = B_TRUE;
513 }
514
515 if (nvlist_lookup_uint64_array(tmp,
516 ZPOOL_CONFIG_HOLE_ARRAY, &hole_array,
517 &holes) == 0) {
518 verify(nvlist_add_uint64_array(config,
519 ZPOOL_CONFIG_HOLE_ARRAY,
520 hole_array, holes) == 0);
521 }
522 }
523
524 if (!config_seen) {
525 /*
526 * Copy the relevant pieces of data to the pool
527 * configuration:
528 *
529 * version
530 * pool guid
531 * name
532 * comment (if available)
533 * pool state
534 * hostid (if available)
535 * hostname (if available)
536 */
537 uint64_t state, version;
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_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 typedef struct slice_node {
910 char *sn_name;
911 nvlist_t *sn_config;
912 boolean_t sn_nozpool;
913 int sn_partno;
914 struct disk_node *sn_disk;
915 struct slice_node *sn_next;
916 } slice_node_t;
917
918 typedef struct disk_node {
919 char *dn_name;
920 int dn_dfd;
921 libzfs_handle_t *dn_hdl;
922 nvlist_t *dn_config;
923 struct slice_node *dn_slices;
924 struct disk_node *dn_next;
925 } disk_node_t;
926
927 #ifdef sparc
928 #define WHOLE_DISK "s2"
929 #else
930 #define WHOLE_DISK "p0"
931 #endif
932
933 /*
934 * This function splits the slice from the device name. Currently it supports
935 * VTOC slices (s[0-16]) and DOS/FDISK partitions (p[0-4]). If this function
936 * is updated to support other slice types then the check_slices function will
937 * also need to be updated.
938 */
939 static boolean_t
940 get_disk_slice(libzfs_handle_t *hdl, char *disk, char **slice, int *partno)
941 {
942 char *p;
943
944 if ((p = strrchr(disk, 's')) == NULL &&
945 (p = strrchr(disk, 'p')) == NULL)
946 return (B_FALSE);
947
948 if (!isdigit(p[1]))
949 return (B_FALSE);
950
951 *slice = zfs_strdup(hdl, p);
952 *partno = atoi(p + 1);
953
954 p = '\0';
955 return (B_TRUE);
956 }
957
958 static void
959 check_one_slice(slice_node_t *slice, diskaddr_t size, uint_t blksz)
960 {
961 /*
962 * protect against division by zero for disk labels that
963 * contain a bogus sector size
964 */
965 if (blksz == 0)
966 blksz = DEV_BSIZE;
967 /* too small to contain a zpool? */
968 if (size < (SPA_MINDEVSIZE / blksz))
969 slice->sn_nozpool = B_TRUE;
970 }
971
972 static void
973 check_slices(slice_node_t *slices, int fd)
974 {
975 struct extvtoc vtoc;
976 struct dk_gpt *gpt;
977 slice_node_t *slice;
978 diskaddr_t size;
979
980 if (read_extvtoc(fd, &vtoc) >= 0) {
981 for (slice = slices; slice; slice = slice->sn_next) {
982 if (slice->sn_name[0] == 'p')
983 continue;
984 size = vtoc.v_part[slice->sn_partno].p_size;
985 check_one_slice(slice, size, vtoc.v_sectorsz);
986 }
987 } else if (efi_alloc_and_read(fd, &gpt) >= 0) {
988 for (slice = slices; slice; slice = slice->sn_next) {
989 /* nodes p[1-4] are never used with EFI labels */
990 if (slice->sn_name[0] == 'p') {
991 if (slice->sn_partno > 0)
992 slice->sn_nozpool = B_TRUE;
993 continue;
994 }
995 size = gpt->efi_parts[slice->sn_partno].p_size;
996 check_one_slice(slice, size, gpt->efi_lbasize);
997 }
998 efi_free(gpt);
999 }
1000 }
1001
1002 static void
1003 zpool_open_func(void *arg)
1004 {
1005 disk_node_t *disk = arg;
1006 struct stat64 statbuf;
1007 slice_node_t *slice;
1008 nvlist_t *config;
1009 char *devname;
1010 int fd;
1011
1012 /*
1013 * If the disk has no slices we open it directly, otherwise we try
1014 * to open the whole disk slice.
1015 */
1016 if (disk->dn_slices == NULL)
1017 devname = strdup(disk->dn_name);
1018 else
1019 (void) asprintf(&devname, "%s" WHOLE_DISK, disk->dn_name);
1020
1021 if (devname == NULL) {
1022 (void) no_memory(disk->dn_hdl);
1023 return;
1024 }
1025
1026 if ((fd = openat64(disk->dn_dfd, devname, O_RDONLY)) < 0) {
1027 free(devname);
1028 return;
1029 }
1030 /*
1031 * Ignore failed stats. We only want regular
1032 * files, character devs and block devs.
1033 */
1034 if (fstat64(fd, &statbuf) != 0 ||
1035 (!S_ISREG(statbuf.st_mode) &&
1036 !S_ISCHR(statbuf.st_mode) &&
1037 !S_ISBLK(statbuf.st_mode))) {
1038 (void) close(fd);
1039 free(devname);
1040 return;
1041 }
1042 /* this file is too small to hold a zpool */
1043 if (S_ISREG(statbuf.st_mode) && statbuf.st_size < SPA_MINDEVSIZE) {
1044 (void) close(fd);
1045 free(devname);
1046 return;
1047 } else if (!S_ISREG(statbuf.st_mode) && disk->dn_slices != NULL) {
1048 /*
1049 * Try to read the disk label first so we don't have to
1050 * open a bunch of minor nodes that can't have a zpool.
1051 */
1052 check_slices(disk->dn_slices, fd);
1053 }
1054
1055 /*
1056 * If we're working with the device directly (it has no slices)
1057 * then we can just read the config and we're done.
1058 */
1059 if (disk->dn_slices == NULL) {
1060 if (zpool_read_label(fd, &config) != 0) {
1061 (void) no_memory(disk->dn_hdl);
1062 (void) close(fd);
1063 free(devname);
1064 return;
1065 }
1066 disk->dn_config = config;
1067 (void) close(fd);
1068 free(devname);
1069 return;
1070 }
1071
1072 (void) close(fd);
1073 free(devname);
1074
1075 /*
1076 * Go through and read the label off each slice. The check_slices
1077 * function has already performed some basic checks and set the
1078 * sn_nozpool flag on any slices which just can't contain a zpool.
1079 */
1080 for (slice = disk->dn_slices; slice; slice = slice->sn_next) {
1081 if (slice->sn_nozpool == B_TRUE)
1082 continue;
1083
1084 (void) asprintf(&devname, "%s%s", disk->dn_name,
1085 slice->sn_name);
1086
1087 if (devname == NULL) {
1088 (void) no_memory(disk->dn_hdl);
1089 free(devname);
1090 return;
1091 }
1092
1093 if ((fd = openat64(disk->dn_dfd, devname, O_RDONLY)) < 0) {
1094 free(devname);
1095 continue;
1096 }
1097
1098 if ((zpool_read_label(fd, &config)) != 0) {
1099 (void) no_memory(disk->dn_hdl);
1100 (void) close(fd);
1101 free(devname);
1102 return;
1103 }
1104
1105 slice->sn_config = config;
1106 (void) close(fd);
1107 free(devname);
1108 }
1109 }
1110
1111 /*
1112 * Given a file descriptor, clear (zero) the label information. This function
1113 * is currently only used in the appliance stack as part of the ZFS sysevent
1114 * module.
1115 */
1116 int
1117 zpool_clear_label(int fd)
1118 {
1119 struct stat64 statbuf;
1120 int l;
1121 vdev_label_t *label;
1122 uint64_t size;
1123
1124 if (fstat64(fd, &statbuf) == -1)
1125 return (0);
1126 size = P2ALIGN_TYPED(statbuf.st_size, sizeof (vdev_label_t), uint64_t);
1127
1128 if ((label = calloc(sizeof (vdev_label_t), 1)) == NULL)
1129 return (-1);
1130
1131 for (l = 0; l < VDEV_LABELS; l++) {
1132 if (pwrite64(fd, label, sizeof (vdev_label_t),
1133 label_offset(size, l)) != sizeof (vdev_label_t))
1134 return (-1);
1135 }
1136
1137 free(label);
1138 return (0);
1139 }
1140
1141 /*
1142 * Given a list of directories to search, find all pools stored on disk. This
1143 * includes partial pools which are not available to import. If no args are
1144 * given (argc is 0), then the default directory (/dev/dsk) is searched.
1145 * poolname or guid (but not both) are provided by the caller when trying
1146 * to import a specific pool.
1147 */
1148 static nvlist_t *
1149 zpool_find_import_impl(libzfs_handle_t *hdl, importargs_t *iarg)
1150 {
1151 int i, dirs = iarg->paths;
1152 DIR *dirp = NULL;
1153 struct dirent64 *dp;
1154 char path[MAXPATHLEN];
1155 char *end, **dir = iarg->path;
1156 size_t pathleft;
1157 nvlist_t *ret = NULL;
1158 static char *default_dir = "/dev/dsk";
1159 pool_list_t pools = { 0 };
1160 pool_entry_t *pe, *penext;
1161 vdev_entry_t *ve, *venext;
1162 config_entry_t *ce, *cenext;
1163 name_entry_t *ne, *nenext;
1164 void *cookie;
1165
1166 if (dirs == 0) {
1167 dirs = 1;
1168 dir = &default_dir;
1169 }
1170
1171 /*
1172 * Go through and read the label configuration information from every
1173 * possible device, organizing the information according to pool GUID
1174 * and toplevel GUID.
1175 */
1176 for (i = 0; i < dirs; i++) {
1177 tpool_t *t;
1178 char *rdsk;
1179 int dfd;
1180 disk_node_t *disks = NULL, *curdisk = NULL;
1181 slice_node_t *curslice = NULL;
1182
1183 /* use realpath to normalize the path */
1184 if (realpath(dir[i], path) == 0) {
1185 (void) zfs_error_fmt(hdl, EZFS_BADPATH,
1186 dgettext(TEXT_DOMAIN, "cannot open '%s'"), dir[i]);
1187 goto error;
1188 }
1189 end = &path[strlen(path)];
1190 *end++ = '/';
1191 *end = 0;
1192 pathleft = &path[sizeof (path)] - end;
1193
1194 /*
1195 * Using raw devices instead of block devices when we're
1196 * reading the labels skips a bunch of slow operations during
1197 * close(2) processing, so we replace /dev/dsk with /dev/rdsk.
1198 */
1199 if (strcmp(path, "/dev/dsk/") == 0)
1200 rdsk = "/dev/rdsk/";
1201 else
1202 rdsk = path;
1203
1204 if ((dfd = open64(rdsk, O_RDONLY)) < 0 ||
1205 (dirp = fdopendir(dfd)) == NULL) {
1206 zfs_error_aux(hdl, strerror(errno));
1207 (void) zfs_error_fmt(hdl, EZFS_BADPATH,
1208 dgettext(TEXT_DOMAIN, "cannot open '%s'"),
1209 rdsk);
1210 goto error;
1211 }
1212
1213 /*
1214 * This is not MT-safe, but we have no MT consumers of libzfs
1215 */
1216 while ((dp = readdir64(dirp)) != NULL) {
1217 boolean_t isslice;
1218 char *name, *sname;
1219 int partno;
1220
1221 if (dp->d_name[0] == '.' && (dp->d_name[1] == '\0' ||
1222 (dp->d_name[1] == '.' && dp->d_name[2] == '\0')))
1223 continue;
1224
1225 name = zfs_strdup(hdl, dp->d_name);
1226
1227 /*
1228 * We create a new disk node every time we encounter
1229 * a disk with no slices or the disk name changes.
1230 */
1231 isslice = get_disk_slice(hdl, name, &sname, &partno);
1232 if (isslice == B_FALSE || curdisk == NULL ||
1233 strcmp(curdisk->dn_name, name) != 0) {
1234 disk_node_t *newdisk;
1235
1236 newdisk = zfs_alloc(hdl, sizeof (disk_node_t));
1237 newdisk->dn_name = name;
1238 newdisk->dn_dfd = dfd;
1239 newdisk->dn_hdl = hdl;
1240
1241 if (curdisk != NULL)
1242 curdisk->dn_next = newdisk;
1243 else
1244 disks = newdisk;
1245
1246 curdisk = newdisk;
1247 curslice = NULL;
1248 }
1249
1250 assert(curdisk != NULL);
1251
1252 /*
1253 * Add a new slice node to the current disk node.
1254 * We do this for all slices including zero slices.
1255 */
1256 if (isslice == B_TRUE) {
1257 slice_node_t *newslice;
1258
1259 newslice = zfs_alloc(hdl,
1260 sizeof (slice_node_t));
1261 newslice->sn_name = sname;
1262 newslice->sn_partno = partno;
1263 newslice->sn_disk = curdisk;
1264
1265 if (curslice != NULL)
1266 curslice->sn_next = newslice;
1267 else
1268 curdisk->dn_slices = newslice;
1269
1270 curslice = newslice;
1271 }
1272 }
1273 /*
1274 * create a thread pool to do all of this in parallel;
1275 * choose double the number of processors; we hold a lot
1276 * of locks in the kernel, so going beyond this doesn't
1277 * buy us much. Each disk (and any slices it might have)
1278 * is handled inside a single thread.
1279 */
1280 t = tpool_create(1, 2 * sysconf(_SC_NPROCESSORS_ONLN),
1281 0, NULL);
1282 for (curdisk = disks; curdisk; curdisk = curdisk->dn_next)
1283 (void) tpool_dispatch(t, zpool_open_func, curdisk);
1284 tpool_wait(t);
1285 tpool_destroy(t);
1286
1287 curdisk = disks;
1288 while (curdisk != NULL) {
1289 nvlist_t *config;
1290 disk_node_t *prevdisk;
1291
1292 /*
1293 * If the device has slices we examine the config on
1294 * each of those. If not we use the config directly
1295 * from the device instead.
1296 */
1297 curslice = curdisk->dn_slices;
1298
1299 if (curslice != NULL)
1300 config = curslice->sn_config;
1301 else
1302 config = curdisk->dn_config;
1303
1304 do {
1305 boolean_t matched = B_TRUE;
1306
1307 if (config == NULL)
1308 goto next;
1309
1310 if (iarg->poolname != NULL) {
1311 char *pname;
1312
1313 matched = nvlist_lookup_string(config,
1314 ZPOOL_CONFIG_POOL_NAME,
1315 &pname) == 0 &&
1316 strcmp(iarg->poolname, pname) == 0;
1317 } else if (iarg->guid != 0) {
1318 uint64_t this_guid;
1319
1320 matched = nvlist_lookup_uint64(config,
1321 ZPOOL_CONFIG_POOL_GUID,
1322 &this_guid) == 0 &&
1323 iarg->guid == this_guid;
1324 }
1325
1326 if (!matched) {
1327 nvlist_free(config);
1328 goto next;
1329 }
1330
1331 /* use the non-raw path for the config */
1332 if (curslice != NULL)
1333 (void) snprintf(end, pathleft, "%s%s",
1334 curdisk->dn_name,
1335 curslice->sn_name);
1336 else
1337 (void) strlcpy(end, curdisk->dn_name,
1338 pathleft);
1339 if (add_config(hdl, &pools, path, config) != 0)
1340 goto error;
1341
1342 next:
1343 /*
1344 * If we're looking at slices free this one
1345 * and go move onto the next.
1346 */
1347 if (curslice != NULL) {
1348 slice_node_t *prevslice;
1349
1350 prevslice = curslice;
1351 curslice = curslice->sn_next;
1352
1353 free(prevslice->sn_name);
1354 free(prevslice);
1355
1356 if (curslice != NULL) {
1357 config = curslice->sn_config;
1358 }
1359 }
1360 } while (curslice != NULL);
1361
1362 /*
1363 * Free this disk and move onto the next one.
1364 */
1365 prevdisk = curdisk;
1366 curdisk = curdisk->dn_next;
1367
1368 free(prevdisk->dn_name);
1369 free(prevdisk);
1370 }
1371
1372 (void) closedir(dirp);
1373 dirp = NULL;
1374 }
1375
1376 ret = get_configs(hdl, &pools, iarg->can_be_active);
1377
1378 error:
1379 for (pe = pools.pools; pe != NULL; pe = penext) {
1380 penext = pe->pe_next;
1381 for (ve = pe->pe_vdevs; ve != NULL; ve = venext) {
1382 venext = ve->ve_next;
1383 for (ce = ve->ve_configs; ce != NULL; ce = cenext) {
1384 cenext = ce->ce_next;
1385 if (ce->ce_config)
1386 nvlist_free(ce->ce_config);
1387 free(ce);
1388 }
1389 free(ve);
1390 }
1391 free(pe);
1392 }
1393
1394 for (ne = pools.names; ne != NULL; ne = nenext) {
1395 nenext = ne->ne_next;
1396 if (ne->ne_name)
1397 free(ne->ne_name);
1398 free(ne);
1399 }
1400
1401 if (dirp)
1402 (void) closedir(dirp);
1403
1404 return (ret);
1405 }
1406
1407 nvlist_t *
1408 zpool_find_import(libzfs_handle_t *hdl, int argc, char **argv)
1409 {
1410 importargs_t iarg = { 0 };
1411
1412 iarg.paths = argc;
1413 iarg.path = argv;
1414
1415 return (zpool_find_import_impl(hdl, &iarg));
1416 }
1417
1418 /*
1419 * Given a cache file, return the contents as a list of importable pools.
1420 * poolname or guid (but not both) are provided by the caller when trying
1421 * to import a specific pool.
1422 */
1423 nvlist_t *
1424 zpool_find_import_cached(libzfs_handle_t *hdl, const char *cachefile,
1425 char *poolname, uint64_t guid)
1426 {
1427 char *buf;
1428 int fd;
1429 struct stat64 statbuf;
1430 nvlist_t *raw, *src, *dst;
1431 nvlist_t *pools;
1432 nvpair_t *elem;
1433 char *name;
1434 uint64_t this_guid;
1435 boolean_t active;
1436
1437 verify(poolname == NULL || guid == 0);
1438
1439 if ((fd = open(cachefile, O_RDONLY)) < 0) {
1440 zfs_error_aux(hdl, "%s", strerror(errno));
1441 (void) zfs_error(hdl, EZFS_BADCACHE,
1442 dgettext(TEXT_DOMAIN, "failed to open cache file"));
1443 return (NULL);
1444 }
1445
1446 if (fstat64(fd, &statbuf) != 0) {
1447 zfs_error_aux(hdl, "%s", strerror(errno));
1448 (void) close(fd);
1449 (void) zfs_error(hdl, EZFS_BADCACHE,
1450 dgettext(TEXT_DOMAIN, "failed to get size of cache file"));
1451 return (NULL);
1452 }
1453
1454 if ((buf = zfs_alloc(hdl, statbuf.st_size)) == NULL) {
1455 (void) close(fd);
1456 return (NULL);
1457 }
1458
1459 if (read(fd, buf, statbuf.st_size) != statbuf.st_size) {
1460 (void) close(fd);
1461 free(buf);
1462 (void) zfs_error(hdl, EZFS_BADCACHE,
1463 dgettext(TEXT_DOMAIN,
1464 "failed to read cache file contents"));
1465 return (NULL);
1466 }
1467
1468 (void) close(fd);
1469
1470 if (nvlist_unpack(buf, statbuf.st_size, &raw, 0) != 0) {
1471 free(buf);
1472 (void) zfs_error(hdl, EZFS_BADCACHE,
1473 dgettext(TEXT_DOMAIN,
1474 "invalid or corrupt cache file contents"));
1475 return (NULL);
1476 }
1477
1478 free(buf);
1479
1480 /*
1481 * Go through and get the current state of the pools and refresh their
1482 * state.
1483 */
1484 if (nvlist_alloc(&pools, 0, 0) != 0) {
1485 (void) no_memory(hdl);
1486 nvlist_free(raw);
1487 return (NULL);
1488 }
1489
1490 elem = NULL;
1491 while ((elem = nvlist_next_nvpair(raw, elem)) != NULL) {
1492 verify(nvpair_value_nvlist(elem, &src) == 0);
1493
1494 verify(nvlist_lookup_string(src, ZPOOL_CONFIG_POOL_NAME,
1495 &name) == 0);
1496 if (poolname != NULL && strcmp(poolname, name) != 0)
1497 continue;
1498
1499 verify(nvlist_lookup_uint64(src, ZPOOL_CONFIG_POOL_GUID,
1500 &this_guid) == 0);
1501 if (guid != 0) {
1502 verify(nvlist_lookup_uint64(src, ZPOOL_CONFIG_POOL_GUID,
1503 &this_guid) == 0);
1504 if (guid != this_guid)
1505 continue;
1506 }
1507
1508 if (pool_active(hdl, name, this_guid, &active) != 0) {
1509 nvlist_free(raw);
1510 nvlist_free(pools);
1511 return (NULL);
1512 }
1513
1514 if (active)
1515 continue;
1516
1517 if ((dst = refresh_config(hdl, src)) == NULL) {
1518 nvlist_free(raw);
1519 nvlist_free(pools);
1520 return (NULL);
1521 }
1522
1523 if (nvlist_add_nvlist(pools, nvpair_name(elem), dst) != 0) {
1524 (void) no_memory(hdl);
1525 nvlist_free(dst);
1526 nvlist_free(raw);
1527 nvlist_free(pools);
1528 return (NULL);
1529 }
1530 nvlist_free(dst);
1531 }
1532
1533 nvlist_free(raw);
1534 return (pools);
1535 }
1536
1537 static int
1538 name_or_guid_exists(zpool_handle_t *zhp, void *data)
1539 {
1540 importargs_t *import = data;
1541 int found = 0;
1542
1543 if (import->poolname != NULL) {
1544 char *pool_name;
1545
1546 verify(nvlist_lookup_string(zhp->zpool_config,
1547 ZPOOL_CONFIG_POOL_NAME, &pool_name) == 0);
1548 if (strcmp(pool_name, import->poolname) == 0)
1549 found = 1;
1550 } else {
1551 uint64_t pool_guid;
1552
1553 verify(nvlist_lookup_uint64(zhp->zpool_config,
1554 ZPOOL_CONFIG_POOL_GUID, &pool_guid) == 0);
1555 if (pool_guid == import->guid)
1556 found = 1;
1557 }
1558
1559 zpool_close(zhp);
1560 return (found);
1561 }
1562
1563 nvlist_t *
1564 zpool_search_import(libzfs_handle_t *hdl, importargs_t *import)
1565 {
1566 verify(import->poolname == NULL || import->guid == 0);
1567
1568 if (import->unique)
1569 import->exists = zpool_iter(hdl, name_or_guid_exists, import);
1570
1571 if (import->cachefile != NULL)
1572 return (zpool_find_import_cached(hdl, import->cachefile,
1573 import->poolname, import->guid));
1574
1575 return (zpool_find_import_impl(hdl, import));
1576 }
1577
1578 boolean_t
1579 find_guid(nvlist_t *nv, uint64_t guid)
1580 {
1581 uint64_t tmp;
1582 nvlist_t **child;
1583 uint_t c, children;
1584
1585 verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID, &tmp) == 0);
1586 if (tmp == guid)
1587 return (B_TRUE);
1588
1589 if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN,
1590 &child, &children) == 0) {
1591 for (c = 0; c < children; c++)
1592 if (find_guid(child[c], guid))
1593 return (B_TRUE);
1594 }
1595
1596 return (B_FALSE);
1597 }
1598
1599 typedef struct aux_cbdata {
1600 const char *cb_type;
1601 uint64_t cb_guid;
1602 zpool_handle_t *cb_zhp;
1603 } aux_cbdata_t;
1604
1605 static int
1606 find_aux(zpool_handle_t *zhp, void *data)
1607 {
1608 aux_cbdata_t *cbp = data;
1609 nvlist_t **list;
1610 uint_t i, count;
1611 uint64_t guid;
1612 nvlist_t *nvroot;
1613
1614 verify(nvlist_lookup_nvlist(zhp->zpool_config, ZPOOL_CONFIG_VDEV_TREE,
1615 &nvroot) == 0);
1616
1617 if (nvlist_lookup_nvlist_array(nvroot, cbp->cb_type,
1618 &list, &count) == 0) {
1619 for (i = 0; i < count; i++) {
1620 verify(nvlist_lookup_uint64(list[i],
1621 ZPOOL_CONFIG_GUID, &guid) == 0);
1622 if (guid == cbp->cb_guid) {
1623 cbp->cb_zhp = zhp;
1624 return (1);
1625 }
1626 }
1627 }
1628
1629 zpool_close(zhp);
1630 return (0);
1631 }
1632
1633 /*
1634 * Determines if the pool is in use. If so, it returns true and the state of
1635 * the pool as well as the name of the pool. Both strings are allocated and
1636 * must be freed by the caller.
1637 */
1638 int
1639 zpool_in_use(libzfs_handle_t *hdl, int fd, pool_state_t *state, char **namestr,
1640 boolean_t *inuse)
1641 {
1642 nvlist_t *config;
1643 char *name;
1644 boolean_t ret;
1645 uint64_t guid, vdev_guid;
1646 zpool_handle_t *zhp;
1647 nvlist_t *pool_config;
1648 uint64_t stateval, isspare;
1649 aux_cbdata_t cb = { 0 };
1650 boolean_t isactive;
1651
1652 *inuse = B_FALSE;
1653
1654 if (zpool_read_label(fd, &config) != 0) {
1655 (void) no_memory(hdl);
1656 return (-1);
1657 }
1658
1659 if (config == NULL)
1660 return (0);
1661
1662 verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_STATE,
1663 &stateval) == 0);
1664 verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_GUID,
1665 &vdev_guid) == 0);
1666
1667 if (stateval != POOL_STATE_SPARE && stateval != POOL_STATE_L2CACHE) {
1668 verify(nvlist_lookup_string(config, ZPOOL_CONFIG_POOL_NAME,
1669 &name) == 0);
1670 verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID,
1671 &guid) == 0);
1672 }
1673
1674 switch (stateval) {
1675 case POOL_STATE_EXPORTED:
1676 /*
1677 * A pool with an exported state may in fact be imported
1678 * read-only, so check the in-core state to see if it's
1679 * active and imported read-only. If it is, set
1680 * its state to active.
1681 */
1682 if (pool_active(hdl, name, guid, &isactive) == 0 && isactive &&
1683 (zhp = zpool_open_canfail(hdl, name)) != NULL) {
1684 if (zpool_get_prop_int(zhp, ZPOOL_PROP_READONLY, NULL))
1685 stateval = POOL_STATE_ACTIVE;
1686
1687 /*
1688 * All we needed the zpool handle for is the
1689 * readonly prop check.
1690 */
1691 zpool_close(zhp);
1692 }
1693
1694 ret = B_TRUE;
1695 break;
1696
1697 case POOL_STATE_ACTIVE:
1698 /*
1699 * For an active pool, we have to determine if it's really part
1700 * of a currently active pool (in which case the pool will exist
1701 * and the guid will be the same), or whether it's part of an
1702 * active pool that was disconnected without being explicitly
1703 * exported.
1704 */
1705 if (pool_active(hdl, name, guid, &isactive) != 0) {
1706 nvlist_free(config);
1707 return (-1);
1708 }
1709
1710 if (isactive) {
1711 /*
1712 * Because the device may have been removed while
1713 * offlined, we only report it as active if the vdev is
1714 * still present in the config. Otherwise, pretend like
1715 * it's not in use.
1716 */
1717 if ((zhp = zpool_open_canfail(hdl, name)) != NULL &&
1718 (pool_config = zpool_get_config(zhp, NULL))
1719 != NULL) {
1720 nvlist_t *nvroot;
1721
1722 verify(nvlist_lookup_nvlist(pool_config,
1723 ZPOOL_CONFIG_VDEV_TREE, &nvroot) == 0);
1724 ret = find_guid(nvroot, vdev_guid);
1725 } else {
1726 ret = B_FALSE;
1727 }
1728
1729 /*
1730 * If this is an active spare within another pool, we
1731 * treat it like an unused hot spare. This allows the
1732 * user to create a pool with a hot spare that currently
1733 * in use within another pool. Since we return B_TRUE,
1734 * libdiskmgt will continue to prevent generic consumers
1735 * from using the device.
1736 */
1737 if (ret && nvlist_lookup_uint64(config,
1738 ZPOOL_CONFIG_IS_SPARE, &isspare) == 0 && isspare)
1739 stateval = POOL_STATE_SPARE;
1740
1741 if (zhp != NULL)
1742 zpool_close(zhp);
1743 } else {
1744 stateval = POOL_STATE_POTENTIALLY_ACTIVE;
1745 ret = B_TRUE;
1746 }
1747 break;
1748
1749 case POOL_STATE_SPARE:
1750 /*
1751 * For a hot spare, it can be either definitively in use, or
1752 * potentially active. To determine if it's in use, we iterate
1753 * over all pools in the system and search for one with a spare
1754 * with a matching guid.
1755 *
1756 * Due to the shared nature of spares, we don't actually report
1757 * the potentially active case as in use. This means the user
1758 * can freely create pools on the hot spares of exported pools,
1759 * but to do otherwise makes the resulting code complicated, and
1760 * we end up having to deal with this case anyway.
1761 */
1762 cb.cb_zhp = NULL;
1763 cb.cb_guid = vdev_guid;
1764 cb.cb_type = ZPOOL_CONFIG_SPARES;
1765 if (zpool_iter(hdl, find_aux, &cb) == 1) {
1766 name = (char *)zpool_get_name(cb.cb_zhp);
1767 ret = TRUE;
1768 } else {
1769 ret = FALSE;
1770 }
1771 break;
1772
1773 case POOL_STATE_L2CACHE:
1774
1775 /*
1776 * Check if any pool is currently using this l2cache device.
1777 */
1778 cb.cb_zhp = NULL;
1779 cb.cb_guid = vdev_guid;
1780 cb.cb_type = ZPOOL_CONFIG_L2CACHE;
1781 if (zpool_iter(hdl, find_aux, &cb) == 1) {
1782 name = (char *)zpool_get_name(cb.cb_zhp);
1783 ret = TRUE;
1784 } else {
1785 ret = FALSE;
1786 }
1787 break;
1788
1789 default:
1790 ret = B_FALSE;
1791 }
1792
1793
1794 if (ret) {
1795 if ((*namestr = zfs_strdup(hdl, name)) == NULL) {
1796 if (cb.cb_zhp)
1797 zpool_close(cb.cb_zhp);
1798 nvlist_free(config);
1799 return (-1);
1800 }
1801 *state = (pool_state_t)stateval;
1802 }
1803
1804 if (cb.cb_zhp)
1805 zpool_close(cb.cb_zhp);
1806
1807 nvlist_free(config);
1808 *inuse = ret;
1809 return (0);
1810 }