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 /*
23 * Copyright 2015 Nexenta Systems, Inc. All rights reserved.
24 * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
25 * Copyright (c) 2013 by Delphix. All rights reserved.
26 * Copyright 2015 RackTop Systems.
27 */
28
29 /*
30 * Pool import support functions.
31 *
32 * To import a pool, we rely on reading the configuration information from the
33 * ZFS label of each device. If we successfully read the label, then we
34 * organize the configuration information in the following hierarchy:
35 *
36 * pool guid -> toplevel vdev guid -> label txg
37 *
38 * Duplicate entries matching this same tuple will be discarded. Once we have
39 * examined every device, we pick the best label txg config for each toplevel
40 * vdev. We then arrange these toplevel vdevs into a complete pool config, and
41 * update any paths that have changed. Finally, we attempt to import the pool
42 * using our derived config, and record the results.
43 */
44
45 #include <ctype.h>
46 #include <devid.h>
47 #include <dirent.h>
48 #include <errno.h>
49 #include <libintl.h>
50 #include <stddef.h>
51 #include <stdlib.h>
52 #include <string.h>
53 #include <sys/stat.h>
54 #include <unistd.h>
55 #include <fcntl.h>
56 #include <sys/vtoc.h>
57 #include <sys/dktp/fdisk.h>
58 #include <sys/efi_partition.h>
59 #include <thread_pool.h>
60
61 #include <sys/vdev_impl.h>
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 struct name_entry *ne_next;
91 } name_entry_t;
92
93 typedef struct pool_list {
94 pool_entry_t *pools;
95 name_entry_t *names;
96 } pool_list_t;
97
98 static char *
99 get_devid(const char *path)
100 {
101 int fd;
102 ddi_devid_t devid;
103 char *minor, *ret;
104
105 if ((fd = open(path, O_RDONLY)) < 0)
106 return (NULL);
107
108 minor = NULL;
109 ret = NULL;
110 if (devid_get(fd, &devid) == 0) {
111 if (devid_get_minor_name(fd, &minor) == 0)
112 ret = devid_str_encode(devid, minor);
113 if (minor != NULL)
114 devid_str_free(minor);
115 devid_free(devid);
116 }
117 (void) close(fd);
118
119 return (ret);
120 }
121
122
123 /*
124 * Go through and fix up any path and/or devid information for the given vdev
125 * configuration.
126 */
127 static int
128 fix_paths(nvlist_t *nv, name_entry_t *names)
129 {
130 nvlist_t **child;
131 uint_t c, children;
132 uint64_t guid;
133 name_entry_t *ne, *best;
134 char *path, *devid;
135 int matched;
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 slices or multiple paths to the same
152 * disk. If this is the case, then we want to pick the path that is
153 * the most similar to the original, where "most similar" is the number
154 * of matching characters starting from the end of the path. This will
155 * preserve slice numbers even if the disks have been reorganized, and
156 * will also catch preferred disk names if multiple paths exist.
157 */
158 verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID, &guid) == 0);
159 if (nvlist_lookup_string(nv, ZPOOL_CONFIG_PATH, &path) != 0)
160 path = NULL;
161
162 matched = 0;
163 best = NULL;
164 for (ne = names; ne != NULL; ne = ne->ne_next) {
165 if (ne->ne_guid == guid) {
166 const char *src, *dst;
167 int count;
168
169 if (path == NULL) {
170 best = ne;
171 break;
172 }
173
174 src = ne->ne_name + strlen(ne->ne_name) - 1;
175 dst = path + strlen(path) - 1;
176 for (count = 0; src >= ne->ne_name && dst >= path;
177 src--, dst--, count++)
178 if (*src != *dst)
179 break;
180
181 /*
182 * At this point, 'count' is the number of characters
183 * matched from the end.
184 */
185 if (count > matched || best == NULL) {
186 best = ne;
187 matched = count;
188 }
189 }
190 }
191
192 if (best == NULL)
193 return (0);
194
195 if (nvlist_add_string(nv, ZPOOL_CONFIG_PATH, best->ne_name) != 0)
196 return (-1);
197
198 if ((devid = get_devid(best->ne_name)) == NULL) {
199 (void) nvlist_remove_all(nv, ZPOOL_CONFIG_DEVID);
200 } else {
201 if (nvlist_add_string(nv, ZPOOL_CONFIG_DEVID, devid) != 0) {
202 devid_str_free(devid);
203 return (-1);
204 }
205 devid_str_free(devid);
206 }
207
208 return (0);
209 }
210
211 /*
212 * Add the given configuration to the list of known devices.
213 */
214 static int
215 add_config(libzfs_handle_t *hdl, pool_list_t *pl, const char *path,
216 nvlist_t *config)
217 {
218 uint64_t pool_guid, vdev_guid, top_guid, txg, state;
219 pool_entry_t *pe;
220 vdev_entry_t *ve;
221 config_entry_t *ce;
222 name_entry_t *ne;
223
224 /*
225 * If this is a hot spare not currently in use or level 2 cache
226 * device, add it to the list of names to translate, but don't do
227 * anything else.
228 */
229 if (nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_STATE,
230 &state) == 0 &&
231 (state == POOL_STATE_SPARE || state == POOL_STATE_L2CACHE) &&
232 nvlist_lookup_uint64(config, ZPOOL_CONFIG_GUID, &vdev_guid) == 0) {
233 if ((ne = zfs_alloc(hdl, sizeof (name_entry_t))) == NULL)
234 return (-1);
235
236 if ((ne->ne_name = zfs_strdup(hdl, path)) == NULL) {
237 free(ne);
238 return (-1);
239 }
240 ne->ne_guid = vdev_guid;
241 ne->ne_next = pl->names;
242 pl->names = ne;
243 nvlist_free(config);
244 return (0);
245 }
246
247 /*
248 * If we have a valid config but cannot read any of these fields, then
249 * it means we have a half-initialized label. In vdev_label_init()
250 * we write a label with txg == 0 so that we can identify the device
251 * in case the user refers to the same disk later on. If we fail to
252 * create the pool, we'll be left with a label in this state
253 * which should not be considered part of a valid pool.
254 */
255 if (nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID,
256 &pool_guid) != 0 ||
257 nvlist_lookup_uint64(config, ZPOOL_CONFIG_GUID,
258 &vdev_guid) != 0 ||
259 nvlist_lookup_uint64(config, ZPOOL_CONFIG_TOP_GUID,
260 &top_guid) != 0 ||
261 nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_TXG,
262 &txg) != 0 || txg == 0) {
263 nvlist_free(config);
264 return (0);
265 }
266
267 /*
268 * First, see if we know about this pool. If not, then add it to the
269 * list of known pools.
270 */
271 for (pe = pl->pools; pe != NULL; pe = pe->pe_next) {
272 if (pe->pe_guid == pool_guid)
273 break;
274 }
275
276 if (pe == NULL) {
277 if ((pe = zfs_alloc(hdl, sizeof (pool_entry_t))) == NULL)
278 return (-1);
279 pe->pe_guid = pool_guid;
280 pe->pe_next = pl->pools;
281 pl->pools = pe;
282 }
283
284 /*
285 * Second, see if we know about this toplevel vdev. Add it if its
286 * missing.
287 */
288 for (ve = pe->pe_vdevs; ve != NULL; ve = ve->ve_next) {
289 if (ve->ve_guid == top_guid)
290 break;
291 }
292
293 if (ve == NULL) {
294 if ((ve = zfs_alloc(hdl, sizeof (vdev_entry_t))) == NULL)
295 return (-1);
296 ve->ve_guid = top_guid;
297 ve->ve_next = pe->pe_vdevs;
298 pe->pe_vdevs = ve;
299 }
300
301 /*
302 * Third, add the vdev guid -> path mappings so that we can fix up
303 * the configuration as necessary before doing the import.
304 */
305 if ((ne = zfs_alloc(hdl, sizeof (name_entry_t))) == NULL)
306 return (-1);
307
308 if ((ne->ne_name = zfs_strdup(hdl, path)) == NULL) {
309 free(ne);
310 return (-1);
311 }
312
313 ne->ne_guid = vdev_guid;
314 ne->ne_next = pl->names;
315 pl->names = ne;
316
317 /*
318 * Finally, see if we have a config with a matching transaction
319 * group. If so, then we do nothing. Otherwise, add it to the list
320 * of known configs.
321 */
322 for (ce = ve->ve_configs; ce != NULL; ce = ce->ce_next) {
323 if (ce->ce_txg == txg)
324 break;
325 }
326
327 if (ce == NULL) {
328 if ((ce = zfs_alloc(hdl, sizeof (config_entry_t))) == NULL)
329 return (-1);
330 ce->ce_txg = txg;
331 ce->ce_config = config;
332 ce->ce_next = ve->ve_configs;
333 ve->ve_configs = ce;
334 } else {
335 nvlist_free(config);
336 }
337
338 return (0);
339 }
340
341 /*
342 * Returns true if the named pool matches the given GUID.
343 */
344 static int
345 pool_active(libzfs_handle_t *hdl, const char *name, uint64_t guid,
346 boolean_t *isactive)
347 {
348 zpool_handle_t *zhp;
349 uint64_t theguid;
350
351 if (zpool_open_silent(hdl, name, &zhp) != 0)
352 return (-1);
353
354 if (zhp == NULL) {
355 *isactive = B_FALSE;
356 return (0);
357 }
358
359 verify(nvlist_lookup_uint64(zhp->zpool_config, ZPOOL_CONFIG_POOL_GUID,
360 &theguid) == 0);
361
362 zpool_close(zhp);
363
364 *isactive = (theguid == guid);
365 return (0);
366 }
367
368 static nvlist_t *
369 refresh_config(libzfs_handle_t *hdl, nvlist_t *config)
370 {
371 nvlist_t *nvl;
372 zfs_cmd_t zc = { 0 };
373 int err;
374
375 if (zcmd_write_conf_nvlist(hdl, &zc, config) != 0)
376 return (NULL);
377
378 if (zcmd_alloc_dst_nvlist(hdl, &zc,
379 zc.zc_nvlist_conf_size * 2) != 0) {
380 zcmd_free_nvlists(&zc);
381 return (NULL);
382 }
383
384 while ((err = ioctl(hdl->libzfs_fd, ZFS_IOC_POOL_TRYIMPORT,
385 &zc)) != 0 && errno == ENOMEM) {
386 if (zcmd_expand_dst_nvlist(hdl, &zc) != 0) {
387 zcmd_free_nvlists(&zc);
388 return (NULL);
389 }
390 }
391
392 if (err) {
393 zcmd_free_nvlists(&zc);
394 return (NULL);
395 }
396
397 if (zcmd_read_dst_nvlist(hdl, &zc, &nvl) != 0) {
398 zcmd_free_nvlists(&zc);
399 return (NULL);
400 }
401
402 zcmd_free_nvlists(&zc);
403 return (nvl);
404 }
405
406 /*
407 * Determine if the vdev id is a hole in the namespace.
408 */
409 boolean_t
410 vdev_is_hole(uint64_t *hole_array, uint_t holes, uint_t id)
411 {
412 for (int c = 0; c < holes; c++) {
413
414 /* Top-level is a hole */
415 if (hole_array[c] == id)
416 return (B_TRUE);
417 }
418 return (B_FALSE);
419 }
420
421 /*
422 * Convert our list of pools into the definitive set of configurations. We
423 * start by picking the best config for each toplevel vdev. Once that's done,
424 * we assemble the toplevel vdevs into a full config for the pool. We make a
425 * pass to fix up any incorrect paths, and then add it to the main list to
426 * return to the user.
427 */
428 static nvlist_t *
429 get_configs(libzfs_handle_t *hdl, pool_list_t *pl, boolean_t active_ok)
430 {
431 pool_entry_t *pe;
432 vdev_entry_t *ve;
433 config_entry_t *ce;
434 nvlist_t *ret = NULL, *config = NULL, *tmp, *nvtop, *nvroot;
435 nvlist_t **spares, **l2cache;
436 uint_t i, nspares, nl2cache;
437 boolean_t config_seen;
438 uint64_t best_txg;
439 char *name, *hostname;
440 uint64_t guid;
441 uint_t children = 0;
442 nvlist_t **child = NULL;
443 uint_t holes;
444 uint64_t *hole_array, max_id;
445 uint_t c;
446 boolean_t isactive;
447 uint64_t hostid;
448 nvlist_t *nvl;
449 boolean_t found_one = B_FALSE;
450 boolean_t valid_top_config = B_FALSE;
451
452 if (nvlist_alloc(&ret, 0, 0) != 0)
453 goto nomem;
454
455 for (pe = pl->pools; pe != NULL; pe = pe->pe_next) {
456 uint64_t id, max_txg = 0;
457
458 if (nvlist_alloc(&config, NV_UNIQUE_NAME, 0) != 0)
459 goto nomem;
460 config_seen = B_FALSE;
461
462 /*
463 * Iterate over all toplevel vdevs. Grab the pool configuration
464 * from the first one we find, and then go through the rest and
465 * add them as necessary to the 'vdevs' member of the config.
466 */
467 for (ve = pe->pe_vdevs; ve != NULL; ve = ve->ve_next) {
468
469 /*
470 * Determine the best configuration for this vdev by
471 * selecting the config with the latest transaction
472 * group.
473 */
474 best_txg = 0;
475 for (ce = ve->ve_configs; ce != NULL;
476 ce = ce->ce_next) {
477
478 if (ce->ce_txg > best_txg) {
479 tmp = ce->ce_config;
480 best_txg = ce->ce_txg;
481 }
482 }
483
484 /*
485 * We rely on the fact that the max txg for the
486 * pool will contain the most up-to-date information
487 * about the valid top-levels in the vdev namespace.
488 */
489 if (best_txg > max_txg) {
490 (void) nvlist_remove(config,
491 ZPOOL_CONFIG_VDEV_CHILDREN,
492 DATA_TYPE_UINT64);
493 (void) nvlist_remove(config,
494 ZPOOL_CONFIG_HOLE_ARRAY,
495 DATA_TYPE_UINT64_ARRAY);
496
497 max_txg = best_txg;
498 hole_array = NULL;
499 holes = 0;
500 max_id = 0;
501 valid_top_config = B_FALSE;
502
503 if (nvlist_lookup_uint64(tmp,
504 ZPOOL_CONFIG_VDEV_CHILDREN, &max_id) == 0) {
505 verify(nvlist_add_uint64(config,
506 ZPOOL_CONFIG_VDEV_CHILDREN,
507 max_id) == 0);
508 valid_top_config = B_TRUE;
509 }
510
511 if (nvlist_lookup_uint64_array(tmp,
512 ZPOOL_CONFIG_HOLE_ARRAY, &hole_array,
513 &holes) == 0) {
514 verify(nvlist_add_uint64_array(config,
515 ZPOOL_CONFIG_HOLE_ARRAY,
516 hole_array, holes) == 0);
517 }
518 }
519
520 if (!config_seen) {
521 /*
522 * Copy the relevant pieces of data to the pool
523 * configuration:
524 *
525 * version
526 * pool guid
527 * name
528 * comment (if available)
529 * pool state
530 * hostid (if available)
531 * hostname (if available)
532 */
533 uint64_t state, version;
534 char *comment = NULL;
535
536 version = fnvlist_lookup_uint64(tmp,
537 ZPOOL_CONFIG_VERSION);
538 fnvlist_add_uint64(config,
539 ZPOOL_CONFIG_VERSION, version);
540 guid = fnvlist_lookup_uint64(tmp,
541 ZPOOL_CONFIG_POOL_GUID);
542 fnvlist_add_uint64(config,
543 ZPOOL_CONFIG_POOL_GUID, guid);
544 name = fnvlist_lookup_string(tmp,
545 ZPOOL_CONFIG_POOL_NAME);
546 fnvlist_add_string(config,
547 ZPOOL_CONFIG_POOL_NAME, name);
548
549 if (nvlist_lookup_string(tmp,
550 ZPOOL_CONFIG_COMMENT, &comment) == 0)
551 fnvlist_add_string(config,
552 ZPOOL_CONFIG_COMMENT, comment);
553
554 state = fnvlist_lookup_uint64(tmp,
555 ZPOOL_CONFIG_POOL_STATE);
556 fnvlist_add_uint64(config,
557 ZPOOL_CONFIG_POOL_STATE, state);
558
559 hostid = 0;
560 if (nvlist_lookup_uint64(tmp,
561 ZPOOL_CONFIG_HOSTID, &hostid) == 0) {
562 fnvlist_add_uint64(config,
563 ZPOOL_CONFIG_HOSTID, hostid);
564 hostname = fnvlist_lookup_string(tmp,
565 ZPOOL_CONFIG_HOSTNAME);
566 fnvlist_add_string(config,
567 ZPOOL_CONFIG_HOSTNAME, hostname);
568 }
569
570 config_seen = B_TRUE;
571 }
572
573 /*
574 * Add this top-level vdev to the child array.
575 */
576 verify(nvlist_lookup_nvlist(tmp,
577 ZPOOL_CONFIG_VDEV_TREE, &nvtop) == 0);
578 verify(nvlist_lookup_uint64(nvtop, ZPOOL_CONFIG_ID,
579 &id) == 0);
580
581 if (id >= children) {
582 nvlist_t **newchild;
583
584 newchild = zfs_alloc(hdl, (id + 1) *
585 sizeof (nvlist_t *));
586 if (newchild == NULL)
587 goto nomem;
588
589 for (c = 0; c < children; c++)
590 newchild[c] = child[c];
591
592 free(child);
593 child = newchild;
594 children = id + 1;
595 }
596 if (nvlist_dup(nvtop, &child[id], 0) != 0)
597 goto nomem;
598
599 }
600
601 /*
602 * If we have information about all the top-levels then
603 * clean up the nvlist which we've constructed. This
604 * means removing any extraneous devices that are
605 * beyond the valid range or adding devices to the end
606 * of our array which appear to be missing.
607 */
608 if (valid_top_config) {
609 if (max_id < children) {
610 for (c = max_id; c < children; c++)
611 nvlist_free(child[c]);
612 children = max_id;
613 } else if (max_id > children) {
614 nvlist_t **newchild;
615
616 newchild = zfs_alloc(hdl, (max_id) *
617 sizeof (nvlist_t *));
618 if (newchild == NULL)
619 goto nomem;
620
621 for (c = 0; c < children; c++)
622 newchild[c] = child[c];
623
624 free(child);
625 child = newchild;
626 children = max_id;
627 }
628 }
629
630 verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID,
631 &guid) == 0);
632
633 /*
634 * The vdev namespace may contain holes as a result of
635 * device removal. We must add them back into the vdev
636 * tree before we process any missing devices.
637 */
638 if (holes > 0) {
639 ASSERT(valid_top_config);
640
641 for (c = 0; c < children; c++) {
642 nvlist_t *holey;
643
644 if (child[c] != NULL ||
645 !vdev_is_hole(hole_array, holes, c))
646 continue;
647
648 if (nvlist_alloc(&holey, NV_UNIQUE_NAME,
649 0) != 0)
650 goto nomem;
651
652 /*
653 * Holes in the namespace are treated as
654 * "hole" top-level vdevs and have a
655 * special flag set on them.
656 */
657 if (nvlist_add_string(holey,
658 ZPOOL_CONFIG_TYPE,
659 VDEV_TYPE_HOLE) != 0 ||
660 nvlist_add_uint64(holey,
661 ZPOOL_CONFIG_ID, c) != 0 ||
662 nvlist_add_uint64(holey,
663 ZPOOL_CONFIG_GUID, 0ULL) != 0) {
664 nvlist_free(holey);
665 goto nomem;
666 }
667 child[c] = holey;
668 }
669 }
670
671 /*
672 * Look for any missing top-level vdevs. If this is the case,
673 * create a faked up 'missing' vdev as a placeholder. We cannot
674 * simply compress the child array, because the kernel performs
675 * certain checks to make sure the vdev IDs match their location
676 * in the configuration.
677 */
678 for (c = 0; c < children; c++) {
679 if (child[c] == NULL) {
680 nvlist_t *missing;
681 if (nvlist_alloc(&missing, NV_UNIQUE_NAME,
682 0) != 0)
683 goto nomem;
684 if (nvlist_add_string(missing,
685 ZPOOL_CONFIG_TYPE,
686 VDEV_TYPE_MISSING) != 0 ||
687 nvlist_add_uint64(missing,
688 ZPOOL_CONFIG_ID, c) != 0 ||
689 nvlist_add_uint64(missing,
690 ZPOOL_CONFIG_GUID, 0ULL) != 0) {
691 nvlist_free(missing);
692 goto nomem;
693 }
694 child[c] = missing;
695 }
696 }
697
698 /*
699 * Put all of this pool's top-level vdevs into a root vdev.
700 */
701 if (nvlist_alloc(&nvroot, NV_UNIQUE_NAME, 0) != 0)
702 goto nomem;
703 if (nvlist_add_string(nvroot, ZPOOL_CONFIG_TYPE,
704 VDEV_TYPE_ROOT) != 0 ||
705 nvlist_add_uint64(nvroot, ZPOOL_CONFIG_ID, 0ULL) != 0 ||
706 nvlist_add_uint64(nvroot, ZPOOL_CONFIG_GUID, guid) != 0 ||
707 nvlist_add_nvlist_array(nvroot, ZPOOL_CONFIG_CHILDREN,
708 child, children) != 0) {
709 nvlist_free(nvroot);
710 goto nomem;
711 }
712
713 for (c = 0; c < children; c++)
714 nvlist_free(child[c]);
715 free(child);
716 children = 0;
717 child = NULL;
718
719 /*
720 * Go through and fix up any paths and/or devids based on our
721 * known list of vdev GUID -> path mappings.
722 */
723 if (fix_paths(nvroot, pl->names) != 0) {
724 nvlist_free(nvroot);
725 goto nomem;
726 }
727
728 /*
729 * Add the root vdev to this pool's configuration.
730 */
731 if (nvlist_add_nvlist(config, ZPOOL_CONFIG_VDEV_TREE,
732 nvroot) != 0) {
733 nvlist_free(nvroot);
734 goto nomem;
735 }
736 nvlist_free(nvroot);
737
738 /*
739 * zdb uses this path to report on active pools that were
740 * imported or created using -R.
741 */
742 if (active_ok)
743 goto add_pool;
744
745 /*
746 * Determine if this pool is currently active, in which case we
747 * can't actually import it.
748 */
749 verify(nvlist_lookup_string(config, ZPOOL_CONFIG_POOL_NAME,
750 &name) == 0);
751 verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID,
752 &guid) == 0);
753
754 if (pool_active(hdl, name, guid, &isactive) != 0)
755 goto error;
756
757 if (isactive) {
758 nvlist_free(config);
759 config = NULL;
760 continue;
761 }
762
763 if ((nvl = refresh_config(hdl, config)) == NULL) {
764 nvlist_free(config);
765 config = NULL;
766 continue;
767 }
768
769 nvlist_free(config);
770 config = nvl;
771
772 /*
773 * Go through and update the paths for spares, now that we have
774 * them.
775 */
776 verify(nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE,
777 &nvroot) == 0);
778 if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_SPARES,
779 &spares, &nspares) == 0) {
780 for (i = 0; i < nspares; i++) {
781 if (fix_paths(spares[i], pl->names) != 0)
782 goto nomem;
783 }
784 }
785
786 /*
787 * Update the paths for l2cache devices.
788 */
789 if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_L2CACHE,
790 &l2cache, &nl2cache) == 0) {
791 for (i = 0; i < nl2cache; i++) {
792 if (fix_paths(l2cache[i], pl->names) != 0)
793 goto nomem;
794 }
795 }
796
797 /*
798 * Restore the original information read from the actual label.
799 */
800 (void) nvlist_remove(config, ZPOOL_CONFIG_HOSTID,
801 DATA_TYPE_UINT64);
802 (void) nvlist_remove(config, ZPOOL_CONFIG_HOSTNAME,
803 DATA_TYPE_STRING);
804 if (hostid != 0) {
805 verify(nvlist_add_uint64(config, ZPOOL_CONFIG_HOSTID,
806 hostid) == 0);
807 verify(nvlist_add_string(config, ZPOOL_CONFIG_HOSTNAME,
808 hostname) == 0);
809 }
810
811 add_pool:
812 /*
813 * Add this pool to the list of configs.
814 */
815 verify(nvlist_lookup_string(config, ZPOOL_CONFIG_POOL_NAME,
816 &name) == 0);
817 if (nvlist_add_nvlist(ret, name, config) != 0)
818 goto nomem;
819
820 found_one = B_TRUE;
821 nvlist_free(config);
822 config = NULL;
823 }
824
825 if (!found_one) {
826 nvlist_free(ret);
827 ret = NULL;
828 }
829
830 return (ret);
831
832 nomem:
833 (void) no_memory(hdl);
834 error:
835 nvlist_free(config);
836 nvlist_free(ret);
837 for (c = 0; c < children; c++)
838 nvlist_free(child[c]);
839 free(child);
840
841 return (NULL);
842 }
843
844 /*
845 * Return the offset of the given label.
846 */
847 static uint64_t
848 label_offset(uint64_t size, int l)
849 {
850 ASSERT(P2PHASE_TYPED(size, sizeof (vdev_label_t), uint64_t) == 0);
851 return (l * sizeof (vdev_label_t) + (l < VDEV_LABELS / 2 ?
852 0 : size - VDEV_LABELS * sizeof (vdev_label_t)));
853 }
854
855 /*
856 * Given a file descriptor, read the label information and return an nvlist
857 * describing the configuration, if there is one.
858 */
859 int
860 zpool_read_label(int fd, nvlist_t **config)
861 {
862 struct stat64 statbuf;
863 int l;
864 vdev_label_t *label;
865 uint64_t state, txg, size;
866
867 *config = NULL;
868
869 if (fstat64(fd, &statbuf) == -1)
870 return (0);
871 size = P2ALIGN_TYPED(statbuf.st_size, sizeof (vdev_label_t), uint64_t);
872
873 if ((label = malloc(sizeof (vdev_label_t))) == NULL)
874 return (-1);
875
876 for (l = 0; l < VDEV_LABELS; l++) {
877 if (pread64(fd, label, sizeof (vdev_label_t),
878 label_offset(size, l)) != sizeof (vdev_label_t))
879 continue;
880
881 if (nvlist_unpack(label->vl_vdev_phys.vp_nvlist,
882 sizeof (label->vl_vdev_phys.vp_nvlist), config, 0) != 0)
883 continue;
884
885 if (nvlist_lookup_uint64(*config, ZPOOL_CONFIG_POOL_STATE,
886 &state) != 0 || state > POOL_STATE_L2CACHE) {
887 nvlist_free(*config);
888 continue;
889 }
890
891 if (state != POOL_STATE_SPARE && state != POOL_STATE_L2CACHE &&
892 (nvlist_lookup_uint64(*config, ZPOOL_CONFIG_POOL_TXG,
893 &txg) != 0 || txg == 0)) {
894 nvlist_free(*config);
895 continue;
896 }
897
898 free(label);
899 return (0);
900 }
901
902 free(label);
903 *config = NULL;
904 return (0);
905 }
906
907 typedef struct rdsk_node {
908 char *rn_name;
909 int rn_dfd;
910 libzfs_handle_t *rn_hdl;
911 nvlist_t *rn_config;
912 avl_tree_t *rn_avl;
913 avl_node_t rn_node;
914 boolean_t rn_nozpool;
915 } rdsk_node_t;
916
917 static int
918 slice_cache_compare(const void *arg1, const void *arg2)
919 {
920 const char *nm1 = ((rdsk_node_t *)arg1)->rn_name;
921 const char *nm2 = ((rdsk_node_t *)arg2)->rn_name;
922 char *nm1slice, *nm2slice;
923 int rv;
924
925 /*
926 * slices zero and two are the most likely to provide results,
927 * so put those first
928 */
929 nm1slice = strstr(nm1, "s0");
930 nm2slice = strstr(nm2, "s0");
931 if (nm1slice && !nm2slice) {
932 return (-1);
933 }
934 if (!nm1slice && nm2slice) {
935 return (1);
936 }
937 nm1slice = strstr(nm1, "s2");
938 nm2slice = strstr(nm2, "s2");
939 if (nm1slice && !nm2slice) {
940 return (-1);
941 }
942 if (!nm1slice && nm2slice) {
943 return (1);
944 }
945
946 rv = strcmp(nm1, nm2);
947 if (rv == 0)
948 return (0);
949 return (rv > 0 ? 1 : -1);
950 }
951
952 static void
953 check_one_slice(avl_tree_t *r, char *diskname, uint_t partno,
954 diskaddr_t size, uint_t blksz)
955 {
956 rdsk_node_t tmpnode;
957 rdsk_node_t *node;
958 char sname[MAXNAMELEN];
959
960 tmpnode.rn_name = &sname[0];
961 (void) snprintf(tmpnode.rn_name, MAXNAMELEN, "%s%u",
962 diskname, partno);
963 /*
964 * protect against division by zero for disk labels that
965 * contain a bogus sector size
966 */
967 if (blksz == 0)
968 blksz = DEV_BSIZE;
969 /* too small to contain a zpool? */
970 if ((size < (SPA_MINDEVSIZE / blksz)) &&
971 (node = avl_find(r, &tmpnode, NULL)))
972 node->rn_nozpool = B_TRUE;
973 }
974
975 static void
976 nozpool_all_slices(avl_tree_t *r, const char *sname)
977 {
978 char diskname[MAXNAMELEN];
979 char *ptr;
980 int i;
981
982 (void) strncpy(diskname, sname, MAXNAMELEN);
983 if (((ptr = strrchr(diskname, 's')) == NULL) &&
984 ((ptr = strrchr(diskname, 'p')) == NULL))
985 return;
986 ptr[0] = 's';
987 ptr[1] = '\0';
988 for (i = 0; i < NDKMAP; i++)
989 check_one_slice(r, diskname, i, 0, 1);
990 ptr[0] = 'p';
991 for (i = 0; i <= FD_NUMPART; i++)
992 check_one_slice(r, diskname, i, 0, 1);
993 }
994
995 static void
996 check_slices(avl_tree_t *r, int fd, const char *sname)
997 {
998 struct extvtoc vtoc;
999 struct dk_gpt *gpt;
1000 char diskname[MAXNAMELEN];
1001 char *ptr;
1002 int i;
1003
1004 (void) strncpy(diskname, sname, MAXNAMELEN);
1005 if ((ptr = strrchr(diskname, 's')) == NULL || !isdigit(ptr[1]))
1006 return;
1007 ptr[1] = '\0';
1008
1009 if (read_extvtoc(fd, &vtoc) >= 0) {
1010 for (i = 0; i < NDKMAP; i++)
1011 check_one_slice(r, diskname, i,
1012 vtoc.v_part[i].p_size, vtoc.v_sectorsz);
1013 } else if (efi_alloc_and_read(fd, &gpt) >= 0) {
1014 /*
1015 * on x86 we'll still have leftover links that point
1016 * to slices s[9-15], so use NDKMAP instead
1017 */
1018 for (i = 0; i < NDKMAP; i++)
1019 check_one_slice(r, diskname, i,
1020 gpt->efi_parts[i].p_size, gpt->efi_lbasize);
1021 /* nodes p[1-4] are never used with EFI labels */
1022 ptr[0] = 'p';
1023 for (i = 1; i <= FD_NUMPART; i++)
1024 check_one_slice(r, diskname, i, 0, 1);
1025 efi_free(gpt);
1026 }
1027 }
1028
1029 static void
1030 zpool_open_func(void *arg)
1031 {
1032 rdsk_node_t *rn = arg;
1033 struct stat64 statbuf;
1034 nvlist_t *config;
1035 int fd;
1036
1037 if (rn->rn_nozpool)
1038 return;
1039 if ((fd = openat64(rn->rn_dfd, rn->rn_name, O_RDONLY)) < 0) {
1040 /* symlink to a device that's no longer there */
1041 if (errno == ENOENT)
1042 nozpool_all_slices(rn->rn_avl, rn->rn_name);
1043 return;
1044 }
1045 /*
1046 * Ignore failed stats. We only want regular
1047 * files, character devs and block devs.
1048 */
1049 if (fstat64(fd, &statbuf) != 0 ||
1050 (!S_ISREG(statbuf.st_mode) &&
1051 !S_ISCHR(statbuf.st_mode) &&
1052 !S_ISBLK(statbuf.st_mode))) {
1053 (void) close(fd);
1054 return;
1055 }
1056 /* this file is too small to hold a zpool */
1057 if (S_ISREG(statbuf.st_mode) &&
1058 statbuf.st_size < SPA_MINDEVSIZE) {
1059 (void) close(fd);
1060 return;
1061 } else if (!S_ISREG(statbuf.st_mode)) {
1062 /*
1063 * Try to read the disk label first so we don't have to
1064 * open a bunch of minor nodes that can't have a zpool.
1065 */
1066 check_slices(rn->rn_avl, fd, rn->rn_name);
1067 }
1068
1069 if ((zpool_read_label(fd, &config)) != 0) {
1070 (void) close(fd);
1071 (void) no_memory(rn->rn_hdl);
1072 return;
1073 }
1074 (void) close(fd);
1075
1076 rn->rn_config = config;
1077 }
1078
1079 /*
1080 * Given a file descriptor, clear (zero) the label information. This function
1081 * is currently only used in the appliance stack as part of the ZFS sysevent
1082 * module.
1083 */
1084 int
1085 zpool_clear_label(int fd)
1086 {
1087 struct stat64 statbuf;
1088 int l;
1089 vdev_label_t *label;
1090 uint64_t size;
1091
1092 if (fstat64(fd, &statbuf) == -1)
1093 return (0);
1094 size = P2ALIGN_TYPED(statbuf.st_size, sizeof (vdev_label_t), uint64_t);
1095
1096 if ((label = calloc(sizeof (vdev_label_t), 1)) == NULL)
1097 return (-1);
1098
1099 for (l = 0; l < VDEV_LABELS; l++) {
1100 if (pwrite64(fd, label, sizeof (vdev_label_t),
1101 label_offset(size, l)) != sizeof (vdev_label_t)) {
1102 free(label);
1103 return (-1);
1104 }
1105 }
1106
1107 free(label);
1108 return (0);
1109 }
1110
1111 /*
1112 * Given a list of directories to search, find all pools stored on disk. This
1113 * includes partial pools which are not available to import. If no args are
1114 * given (argc is 0), then the default directory (/dev/dsk) is searched.
1115 * poolname or guid (but not both) are provided by the caller when trying
1116 * to import a specific pool.
1117 */
1118 static nvlist_t *
1119 zpool_find_import_impl(libzfs_handle_t *hdl, importargs_t *iarg)
1120 {
1121 int i, dirs = iarg->paths;
1122 struct dirent64 *dp;
1123 char path[MAXPATHLEN];
1124 char *end, **dir = iarg->path;
1125 size_t pathleft;
1126 nvlist_t *ret = NULL;
1127 static char *default_dir = "/dev/dsk";
1128 pool_list_t pools = { 0 };
1129 pool_entry_t *pe, *penext;
1130 vdev_entry_t *ve, *venext;
1131 config_entry_t *ce, *cenext;
1132 name_entry_t *ne, *nenext;
1133 avl_tree_t slice_cache;
1134 rdsk_node_t *slice;
1135 void *cookie;
1136
1137 if (dirs == 0) {
1138 dirs = 1;
1139 dir = &default_dir;
1140 }
1141
1142 /*
1143 * Go through and read the label configuration information from every
1144 * possible device, organizing the information according to pool GUID
1145 * and toplevel GUID.
1146 */
1147 for (i = 0; i < dirs; i++) {
1148 tpool_t *t;
1149 char *rdsk;
1150 int dfd;
1151 boolean_t config_failed = B_FALSE;
1152 DIR *dirp;
1153
1154 /* use realpath to normalize the path */
1155 if (realpath(dir[i], path) == 0) {
1156 (void) zfs_error_fmt(hdl, EZFS_BADPATH,
1157 dgettext(TEXT_DOMAIN, "cannot open '%s'"), dir[i]);
1158 goto error;
1159 }
1160 end = &path[strlen(path)];
1161 *end++ = '/';
1162 *end = 0;
1163 pathleft = &path[sizeof (path)] - end;
1164
1165 /*
1166 * Using raw devices instead of block devices when we're
1167 * reading the labels skips a bunch of slow operations during
1168 * close(2) processing, so we replace /dev/dsk with /dev/rdsk.
1169 */
1170 if (strcmp(path, "/dev/dsk/") == 0)
1171 rdsk = "/dev/rdsk/";
1172 else
1173 rdsk = path;
1174
1175 if ((dfd = open64(rdsk, O_RDONLY)) < 0 ||
1176 (dirp = fdopendir(dfd)) == NULL) {
1177 if (dfd >= 0)
1178 (void) close(dfd);
1179 zfs_error_aux(hdl, strerror(errno));
1180 (void) zfs_error_fmt(hdl, EZFS_BADPATH,
1181 dgettext(TEXT_DOMAIN, "cannot open '%s'"),
1182 rdsk);
1183 goto error;
1184 }
1185
1186 avl_create(&slice_cache, slice_cache_compare,
1187 sizeof (rdsk_node_t), offsetof(rdsk_node_t, rn_node));
1188 /*
1189 * This is not MT-safe, but we have no MT consumers of libzfs
1190 */
1191 while ((dp = readdir64(dirp)) != NULL) {
1192 const char *name = dp->d_name;
1193 if (name[0] == '.' &&
1194 (name[1] == 0 || (name[1] == '.' && name[2] == 0)))
1195 continue;
1196
1197 slice = zfs_alloc(hdl, sizeof (rdsk_node_t));
1198 slice->rn_name = zfs_strdup(hdl, name);
1199 slice->rn_avl = &slice_cache;
1200 slice->rn_dfd = dfd;
1201 slice->rn_hdl = hdl;
1202 slice->rn_nozpool = B_FALSE;
1203 avl_add(&slice_cache, slice);
1204 }
1205 /*
1206 * create a thread pool to do all of this in parallel;
1207 * rn_nozpool is not protected, so this is racy in that
1208 * multiple tasks could decide that the same slice can
1209 * not hold a zpool, which is benign. Also choose
1210 * double the number of processors; we hold a lot of
1211 * locks in the kernel, so going beyond this doesn't
1212 * buy us much.
1213 */
1214 t = tpool_create(1, 2 * sysconf(_SC_NPROCESSORS_ONLN),
1215 0, NULL);
1216 for (slice = avl_first(&slice_cache); slice;
1217 (slice = avl_walk(&slice_cache, slice,
1218 AVL_AFTER)))
1219 (void) tpool_dispatch(t, zpool_open_func, slice);
1220 tpool_wait(t);
1221 tpool_destroy(t);
1222
1223 cookie = NULL;
1224 while ((slice = avl_destroy_nodes(&slice_cache,
1225 &cookie)) != NULL) {
1226 if (slice->rn_config != NULL && !config_failed) {
1227 nvlist_t *config = slice->rn_config;
1228 boolean_t matched = B_TRUE;
1229
1230 if (iarg->poolname != NULL) {
1231 char *pname;
1232
1233 matched = nvlist_lookup_string(config,
1234 ZPOOL_CONFIG_POOL_NAME,
1235 &pname) == 0 &&
1236 strcmp(iarg->poolname, pname) == 0;
1237 } else if (iarg->guid != 0) {
1238 uint64_t this_guid;
1239
1240 matched = nvlist_lookup_uint64(config,
1241 ZPOOL_CONFIG_POOL_GUID,
1242 &this_guid) == 0 &&
1243 iarg->guid == this_guid;
1244 }
1245 if (!matched) {
1246 nvlist_free(config);
1247 } else {
1248 /*
1249 * use the non-raw path for the config
1250 */
1251 (void) strlcpy(end, slice->rn_name,
1252 pathleft);
1253 if (add_config(hdl, &pools, path,
1254 config) != 0) {
1255 nvlist_free(config);
1256 config_failed = B_TRUE;
1257 }
1258 }
1259 }
1260 free(slice->rn_name);
1261 free(slice);
1262 }
1263 avl_destroy(&slice_cache);
1264
1265 (void) closedir(dirp);
1266
1267 if (config_failed)
1268 goto error;
1269 }
1270
1271 ret = get_configs(hdl, &pools, iarg->can_be_active);
1272
1273 error:
1274 for (pe = pools.pools; pe != NULL; pe = penext) {
1275 penext = pe->pe_next;
1276 for (ve = pe->pe_vdevs; ve != NULL; ve = venext) {
1277 venext = ve->ve_next;
1278 for (ce = ve->ve_configs; ce != NULL; ce = cenext) {
1279 cenext = ce->ce_next;
1280 if (ce->ce_config)
1281 nvlist_free(ce->ce_config);
1282 free(ce);
1283 }
1284 free(ve);
1285 }
1286 free(pe);
1287 }
1288
1289 for (ne = pools.names; ne != NULL; ne = nenext) {
1290 nenext = ne->ne_next;
1291 free(ne->ne_name);
1292 free(ne);
1293 }
1294
1295 return (ret);
1296 }
1297
1298 nvlist_t *
1299 zpool_find_import(libzfs_handle_t *hdl, int argc, char **argv)
1300 {
1301 importargs_t iarg = { 0 };
1302
1303 iarg.paths = argc;
1304 iarg.path = argv;
1305
1306 return (zpool_find_import_impl(hdl, &iarg));
1307 }
1308
1309 /*
1310 * Given a cache file, return the contents as a list of importable pools.
1311 * poolname or guid (but not both) are provided by the caller when trying
1312 * to import a specific pool.
1313 */
1314 nvlist_t *
1315 zpool_find_import_cached(libzfs_handle_t *hdl, const char *cachefile,
1316 char *poolname, uint64_t guid)
1317 {
1318 char *buf;
1319 int fd;
1320 struct stat64 statbuf;
1321 nvlist_t *raw, *src, *dst;
1322 nvlist_t *pools;
1323 nvpair_t *elem;
1324 char *name;
1325 uint64_t this_guid;
1326 boolean_t active;
1327
1328 verify(poolname == NULL || guid == 0);
1329
1330 if ((fd = open(cachefile, O_RDONLY)) < 0) {
1331 zfs_error_aux(hdl, "%s", strerror(errno));
1332 (void) zfs_error(hdl, EZFS_BADCACHE,
1333 dgettext(TEXT_DOMAIN, "failed to open cache file"));
1334 return (NULL);
1335 }
1336
1337 if (fstat64(fd, &statbuf) != 0) {
1338 zfs_error_aux(hdl, "%s", strerror(errno));
1339 (void) close(fd);
1340 (void) zfs_error(hdl, EZFS_BADCACHE,
1341 dgettext(TEXT_DOMAIN, "failed to get size of cache file"));
1342 return (NULL);
1343 }
1344
1345 if ((buf = zfs_alloc(hdl, statbuf.st_size)) == NULL) {
1346 (void) close(fd);
1347 return (NULL);
1348 }
1349
1350 if (read(fd, buf, statbuf.st_size) != statbuf.st_size) {
1351 (void) close(fd);
1352 free(buf);
1353 (void) zfs_error(hdl, EZFS_BADCACHE,
1354 dgettext(TEXT_DOMAIN,
1355 "failed to read cache file contents"));
1356 return (NULL);
1357 }
1358
1359 (void) close(fd);
1360
1361 if (nvlist_unpack(buf, statbuf.st_size, &raw, 0) != 0) {
1362 free(buf);
1363 (void) zfs_error(hdl, EZFS_BADCACHE,
1364 dgettext(TEXT_DOMAIN,
1365 "invalid or corrupt cache file contents"));
1366 return (NULL);
1367 }
1368
1369 free(buf);
1370
1371 /*
1372 * Go through and get the current state of the pools and refresh their
1373 * state.
1374 */
1375 if (nvlist_alloc(&pools, 0, 0) != 0) {
1376 (void) no_memory(hdl);
1377 nvlist_free(raw);
1378 return (NULL);
1379 }
1380
1381 elem = NULL;
1382 while ((elem = nvlist_next_nvpair(raw, elem)) != NULL) {
1383 src = fnvpair_value_nvlist(elem);
1384
1385 name = fnvlist_lookup_string(src, ZPOOL_CONFIG_POOL_NAME);
1386 if (poolname != NULL && strcmp(poolname, name) != 0)
1387 continue;
1388
1389 this_guid = fnvlist_lookup_uint64(src, ZPOOL_CONFIG_POOL_GUID);
1390 if (guid != 0 && guid != this_guid)
1391 continue;
1392
1393 if (pool_active(hdl, name, this_guid, &active) != 0) {
1394 nvlist_free(raw);
1395 nvlist_free(pools);
1396 return (NULL);
1397 }
1398
1399 if (active)
1400 continue;
1401
1402 if ((dst = refresh_config(hdl, src)) == NULL) {
1403 nvlist_free(raw);
1404 nvlist_free(pools);
1405 return (NULL);
1406 }
1407
1408 if (nvlist_add_nvlist(pools, nvpair_name(elem), dst) != 0) {
1409 (void) no_memory(hdl);
1410 nvlist_free(dst);
1411 nvlist_free(raw);
1412 nvlist_free(pools);
1413 return (NULL);
1414 }
1415 nvlist_free(dst);
1416 }
1417
1418 nvlist_free(raw);
1419 return (pools);
1420 }
1421
1422 static int
1423 name_or_guid_exists(zpool_handle_t *zhp, void *data)
1424 {
1425 importargs_t *import = data;
1426 int found = 0;
1427
1428 if (import->poolname != NULL) {
1429 char *pool_name;
1430
1431 verify(nvlist_lookup_string(zhp->zpool_config,
1432 ZPOOL_CONFIG_POOL_NAME, &pool_name) == 0);
1433 if (strcmp(pool_name, import->poolname) == 0)
1434 found = 1;
1435 } else {
1436 uint64_t pool_guid;
1437
1438 verify(nvlist_lookup_uint64(zhp->zpool_config,
1439 ZPOOL_CONFIG_POOL_GUID, &pool_guid) == 0);
1440 if (pool_guid == import->guid)
1441 found = 1;
1442 }
1443
1444 zpool_close(zhp);
1445 return (found);
1446 }
1447
1448 nvlist_t *
1449 zpool_search_import(libzfs_handle_t *hdl, importargs_t *import)
1450 {
1451 verify(import->poolname == NULL || import->guid == 0);
1452
1453 if (import->unique)
1454 import->exists = zpool_iter(hdl, name_or_guid_exists, import);
1455
1456 if (import->cachefile != NULL)
1457 return (zpool_find_import_cached(hdl, import->cachefile,
1458 import->poolname, import->guid));
1459
1460 return (zpool_find_import_impl(hdl, import));
1461 }
1462
1463 boolean_t
1464 find_guid(nvlist_t *nv, uint64_t guid)
1465 {
1466 uint64_t tmp;
1467 nvlist_t **child;
1468 uint_t c, children;
1469
1470 verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID, &tmp) == 0);
1471 if (tmp == guid)
1472 return (B_TRUE);
1473
1474 if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN,
1475 &child, &children) == 0) {
1476 for (c = 0; c < children; c++)
1477 if (find_guid(child[c], guid))
1478 return (B_TRUE);
1479 }
1480
1481 return (B_FALSE);
1482 }
1483
1484 typedef struct aux_cbdata {
1485 const char *cb_type;
1486 uint64_t cb_guid;
1487 zpool_handle_t *cb_zhp;
1488 } aux_cbdata_t;
1489
1490 static int
1491 find_aux(zpool_handle_t *zhp, void *data)
1492 {
1493 aux_cbdata_t *cbp = data;
1494 nvlist_t **list;
1495 uint_t i, count;
1496 uint64_t guid;
1497 nvlist_t *nvroot;
1498
1499 verify(nvlist_lookup_nvlist(zhp->zpool_config, ZPOOL_CONFIG_VDEV_TREE,
1500 &nvroot) == 0);
1501
1502 if (nvlist_lookup_nvlist_array(nvroot, cbp->cb_type,
1503 &list, &count) == 0) {
1504 for (i = 0; i < count; i++) {
1505 verify(nvlist_lookup_uint64(list[i],
1506 ZPOOL_CONFIG_GUID, &guid) == 0);
1507 if (guid == cbp->cb_guid) {
1508 cbp->cb_zhp = zhp;
1509 return (1);
1510 }
1511 }
1512 }
1513
1514 zpool_close(zhp);
1515 return (0);
1516 }
1517
1518 /*
1519 * Determines if the pool is in use. If so, it returns true and the state of
1520 * the pool as well as the name of the pool. Both strings are allocated and
1521 * must be freed by the caller.
1522 */
1523 int
1524 zpool_in_use(libzfs_handle_t *hdl, int fd, pool_state_t *state, char **namestr,
1525 boolean_t *inuse)
1526 {
1527 nvlist_t *config;
1528 char *name;
1529 boolean_t ret;
1530 uint64_t guid, vdev_guid;
1531 zpool_handle_t *zhp;
1532 nvlist_t *pool_config;
1533 uint64_t stateval, isspare;
1534 aux_cbdata_t cb = { 0 };
1535 boolean_t isactive;
1536
1537 *inuse = B_FALSE;
1538
1539 if (zpool_read_label(fd, &config) != 0) {
1540 (void) no_memory(hdl);
1541 return (-1);
1542 }
1543
1544 if (config == NULL)
1545 return (0);
1546
1547 verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_STATE,
1548 &stateval) == 0);
1549 verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_GUID,
1550 &vdev_guid) == 0);
1551
1552 if (stateval != POOL_STATE_SPARE && stateval != POOL_STATE_L2CACHE) {
1553 verify(nvlist_lookup_string(config, ZPOOL_CONFIG_POOL_NAME,
1554 &name) == 0);
1555 verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID,
1556 &guid) == 0);
1557 }
1558
1559 switch (stateval) {
1560 case POOL_STATE_EXPORTED:
1561 /*
1562 * A pool with an exported state may in fact be imported
1563 * read-only, so check the in-core state to see if it's
1564 * active and imported read-only. If it is, set
1565 * its state to active.
1566 */
1567 if (pool_active(hdl, name, guid, &isactive) == 0 && isactive &&
1568 (zhp = zpool_open_canfail(hdl, name)) != NULL) {
1569 if (zpool_get_prop_int(zhp, ZPOOL_PROP_READONLY, NULL))
1570 stateval = POOL_STATE_ACTIVE;
1571
1572 /*
1573 * All we needed the zpool handle for is the
1574 * readonly prop check.
1575 */
1576 zpool_close(zhp);
1577 }
1578
1579 ret = B_TRUE;
1580 break;
1581
1582 case POOL_STATE_ACTIVE:
1583 /*
1584 * For an active pool, we have to determine if it's really part
1585 * of a currently active pool (in which case the pool will exist
1586 * and the guid will be the same), or whether it's part of an
1587 * active pool that was disconnected without being explicitly
1588 * exported.
1589 */
1590 if (pool_active(hdl, name, guid, &isactive) != 0) {
1591 nvlist_free(config);
1592 return (-1);
1593 }
1594
1595 if (isactive) {
1596 /*
1597 * Because the device may have been removed while
1598 * offlined, we only report it as active if the vdev is
1599 * still present in the config. Otherwise, pretend like
1600 * it's not in use.
1601 */
1602 if ((zhp = zpool_open_canfail(hdl, name)) != NULL &&
1603 (pool_config = zpool_get_config(zhp, NULL))
1604 != NULL) {
1605 nvlist_t *nvroot;
1606
1607 verify(nvlist_lookup_nvlist(pool_config,
1608 ZPOOL_CONFIG_VDEV_TREE, &nvroot) == 0);
1609 ret = find_guid(nvroot, vdev_guid);
1610 } else {
1611 ret = B_FALSE;
1612 }
1613
1614 /*
1615 * If this is an active spare within another pool, we
1616 * treat it like an unused hot spare. This allows the
1617 * user to create a pool with a hot spare that currently
1618 * in use within another pool. Since we return B_TRUE,
1619 * libdiskmgt will continue to prevent generic consumers
1620 * from using the device.
1621 */
1622 if (ret && nvlist_lookup_uint64(config,
1623 ZPOOL_CONFIG_IS_SPARE, &isspare) == 0 && isspare)
1624 stateval = POOL_STATE_SPARE;
1625
1626 if (zhp != NULL)
1627 zpool_close(zhp);
1628 } else {
1629 stateval = POOL_STATE_POTENTIALLY_ACTIVE;
1630 ret = B_TRUE;
1631 }
1632 break;
1633
1634 case POOL_STATE_SPARE:
1635 /*
1636 * For a hot spare, it can be either definitively in use, or
1637 * potentially active. To determine if it's in use, we iterate
1638 * over all pools in the system and search for one with a spare
1639 * with a matching guid.
1640 *
1641 * Due to the shared nature of spares, we don't actually report
1642 * the potentially active case as in use. This means the user
1643 * can freely create pools on the hot spares of exported pools,
1644 * but to do otherwise makes the resulting code complicated, and
1645 * we end up having to deal with this case anyway.
1646 */
1647 cb.cb_zhp = NULL;
1648 cb.cb_guid = vdev_guid;
1649 cb.cb_type = ZPOOL_CONFIG_SPARES;
1650 if (zpool_iter(hdl, find_aux, &cb) == 1) {
1651 name = (char *)zpool_get_name(cb.cb_zhp);
1652 ret = B_TRUE;
1653 } else {
1654 ret = B_FALSE;
1655 }
1656 break;
1657
1658 case POOL_STATE_L2CACHE:
1659
1660 /*
1661 * Check if any pool is currently using this l2cache device.
1662 */
1663 cb.cb_zhp = NULL;
1664 cb.cb_guid = vdev_guid;
1665 cb.cb_type = ZPOOL_CONFIG_L2CACHE;
1666 if (zpool_iter(hdl, find_aux, &cb) == 1) {
1667 name = (char *)zpool_get_name(cb.cb_zhp);
1668 ret = B_TRUE;
1669 } else {
1670 ret = B_FALSE;
1671 }
1672 break;
1673
1674 default:
1675 ret = B_FALSE;
1676 }
1677
1678
1679 if (ret) {
1680 if ((*namestr = zfs_strdup(hdl, name)) == NULL) {
1681 if (cb.cb_zhp)
1682 zpool_close(cb.cb_zhp);
1683 nvlist_free(config);
1684 return (-1);
1685 }
1686 *state = (pool_state_t)stateval;
1687 }
1688
1689 if (cb.cb_zhp)
1690 zpool_close(cb.cb_zhp);
1691
1692 nvlist_free(config);
1693 *inuse = ret;
1694 return (0);
1695 }