]> git.proxmox.com Git - mirror_zfs-debian.git/blob - lib/libzfs/libzfs_import.c
projects / mirror_zfs-debian.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
Imported Upstream version 0.6.4.2
[mirror_zfs-debian.git] / lib / libzfs / libzfs_import.c
1 /*
2 * CDDL HEADER START
3 *
4 * The contents of this file are subject to the terms of the
5 * Common Development and Distribution License (the "License").
6 * You may not use this file except in compliance with the License.
7 *
8 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
9 * or http://www.opensolaris.org/os/licensing.
10 * See the License for the specific language governing permissions
11 * and limitations under the License.
12 *
13 * When distributing Covered Code, include this CDDL HEADER in each
14 * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
15 * If applicable, add the following below this CDDL HEADER, with the
16 * fields enclosed by brackets "[]" replaced with your own identifying
17 * information: Portions Copyright [yyyy] [name of copyright owner]
18 *
19 * CDDL HEADER END
20 */
21 /*
22 * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
23 * Copyright 2011 Nexenta Systems, Inc. All rights reserved.
24 * Copyright (c) 2012 by Delphix. All rights reserved.
25 */
26
27 /*
28 * Pool import support functions.
29 *
30 * To import a pool, we rely on reading the configuration information from the
31 * ZFS label of each device. If we successfully read the label, then we
32 * organize the configuration information in the following hierarchy:
33 *
34 * pool guid -> toplevel vdev guid -> label txg
35 *
36 * Duplicate entries matching this same tuple will be discarded. Once we have
37 * examined every device, we pick the best label txg config for each toplevel
38 * vdev. We then arrange these toplevel vdevs into a complete pool config, and
39 * update any paths that have changed. Finally, we attempt to import the pool
40 * using our derived config, and record the results.
41 */
42
43 #include <ctype.h>
44 #include <devid.h>
45 #include <dirent.h>
46 #include <errno.h>
47 #include <libintl.h>
48 #include <stddef.h>
49 #include <stdlib.h>
50 #include <string.h>
51 #include <sys/stat.h>
52 #include <unistd.h>
53 #include <fcntl.h>
54 #include <sys/vtoc.h>
55 #include <sys/dktp/fdisk.h>
56 #include <sys/efi_partition.h>
57
58 #include <sys/vdev_impl.h>
59 #ifdef HAVE_LIBBLKID
60 #include <blkid/blkid.h>
61 #endif
62
63 #include "libzfs.h"
64 #include "libzfs_impl.h"
65
66 /*
67 * Intermediate structures used to gather configuration information.
68 */
69 typedef struct config_entry {
70 uint64_t ce_txg;
71 nvlist_t *ce_config;
72 struct config_entry *ce_next;
73 } config_entry_t;
74
75 typedef struct vdev_entry {
76 uint64_t ve_guid;
77 config_entry_t *ve_configs;
78 struct vdev_entry *ve_next;
79 } vdev_entry_t;
80
81 typedef struct pool_entry {
82 uint64_t pe_guid;
83 vdev_entry_t *pe_vdevs;
84 struct pool_entry *pe_next;
85 } pool_entry_t;
86
87 typedef struct name_entry {
88 char *ne_name;
89 uint64_t ne_guid;
90 uint64_t ne_order;
91 uint64_t ne_num_labels;
92 struct name_entry *ne_next;
93 } name_entry_t;
94
95 typedef struct pool_list {
96 pool_entry_t *pools;
97 name_entry_t *names;
98 } pool_list_t;
99
100 static char *
101 get_devid(const char *path)
102 {
103 int fd;
104 ddi_devid_t devid;
105 char *minor, *ret;
106
107 if ((fd = open(path, O_RDONLY)) < 0)
108 return (NULL);
109
110 minor = NULL;
111 ret = NULL;
112 if (devid_get(fd, &devid) == 0) {
113 if (devid_get_minor_name(fd, &minor) == 0)
114 ret = devid_str_encode(devid, minor);
115 if (minor != NULL)
116 devid_str_free(minor);
117 devid_free(devid);
118 }
119 (void) close(fd);
120
121 return (ret);
122 }
123
124
125 /*
126 * Go through and fix up any path and/or devid information for the given vdev
127 * configuration.
128 */
129 static int
130 fix_paths(nvlist_t *nv, name_entry_t *names)
131 {
132 nvlist_t **child;
133 uint_t c, children;
134 uint64_t guid;
135 name_entry_t *ne, *best;
136 char *path, *devid;
137
138 if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN,
139 &child, &children) == 0) {
140 for (c = 0; c < children; c++)
141 if (fix_paths(child[c], names) != 0)
142 return (-1);
143 return (0);
144 }
145
146 /*
147 * This is a leaf (file or disk) vdev. In either case, go through
148 * the name list and see if we find a matching guid. If so, replace
149 * the path and see if we can calculate a new devid.
150 *
151 * There may be multiple names associated with a particular guid, in
152 * which case we have overlapping partitions or multiple paths to the
153 * same disk. In this case we prefer to use the path name which
154 * matches the ZPOOL_CONFIG_PATH. If no matching entry is found we
155 * use the lowest order device which corresponds to the first match
156 * while traversing the ZPOOL_IMPORT_PATH search path.
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 best = NULL;
163 for (ne = names; ne != NULL; ne = ne->ne_next) {
164 if (ne->ne_guid == guid) {
165
166 if (path == NULL) {
167 best = ne;
168 break;
169 }
170
171 if ((strlen(path) == strlen(ne->ne_name)) &&
172 strncmp(path, ne->ne_name, strlen(path)) == 0) {
173 best = ne;
174 break;
175 }
176
177 if (best == NULL) {
178 best = ne;
179 continue;
180 }
181
182 /* Prefer paths with move vdev labels. */
183 if (ne->ne_num_labels > best->ne_num_labels) {
184 best = ne;
185 continue;
186 }
187
188 /* Prefer paths earlier in the search order. */
189 if (best->ne_num_labels == best->ne_num_labels &&
190 ne->ne_order < best->ne_order) {
191 best = ne;
192 continue;
193 }
194 }
195 }
196
197 if (best == NULL)
198 return (0);
199
200 if (nvlist_add_string(nv, ZPOOL_CONFIG_PATH, best->ne_name) != 0)
201 return (-1);
202
203 if ((devid = get_devid(best->ne_name)) == NULL) {
204 (void) nvlist_remove_all(nv, ZPOOL_CONFIG_DEVID);
205 } else {
206 if (nvlist_add_string(nv, ZPOOL_CONFIG_DEVID, devid) != 0)
207 return (-1);
208 devid_str_free(devid);
209 }
210
211 return (0);
212 }
213
214 /*
215 * Add the given configuration to the list of known devices.
216 */
217 static int
218 add_config(libzfs_handle_t *hdl, pool_list_t *pl, const char *path,
219 int order, int num_labels, nvlist_t *config)
220 {
221 uint64_t pool_guid, vdev_guid, top_guid, txg, state;
222 pool_entry_t *pe;
223 vdev_entry_t *ve;
224 config_entry_t *ce;
225 name_entry_t *ne;
226
227 /*
228 * If this is a hot spare not currently in use or level 2 cache
229 * device, add it to the list of names to translate, but don't do
230 * anything else.
231 */
232 if (nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_STATE,
233 &state) == 0 &&
234 (state == POOL_STATE_SPARE || state == POOL_STATE_L2CACHE) &&
235 nvlist_lookup_uint64(config, ZPOOL_CONFIG_GUID, &vdev_guid) == 0) {
236 if ((ne = zfs_alloc(hdl, sizeof (name_entry_t))) == NULL)
237 return (-1);
238
239 if ((ne->ne_name = zfs_strdup(hdl, path)) == NULL) {
240 free(ne);
241 return (-1);
242 }
243 ne->ne_guid = vdev_guid;
244 ne->ne_order = order;
245 ne->ne_num_labels = num_labels;
246 ne->ne_next = pl->names;
247 pl->names = ne;
248 return (0);
249 }
250
251 /*
252 * If we have a valid config but cannot read any of these fields, then
253 * it means we have a half-initialized label. In vdev_label_init()
254 * we write a label with txg == 0 so that we can identify the device
255 * in case the user refers to the same disk later on. If we fail to
256 * create the pool, we'll be left with a label in this state
257 * which should not be considered part of a valid pool.
258 */
259 if (nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID,
260 &pool_guid) != 0 ||
261 nvlist_lookup_uint64(config, ZPOOL_CONFIG_GUID,
262 &vdev_guid) != 0 ||
263 nvlist_lookup_uint64(config, ZPOOL_CONFIG_TOP_GUID,
264 &top_guid) != 0 ||
265 nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_TXG,
266 &txg) != 0 || txg == 0) {
267 nvlist_free(config);
268 return (0);
269 }
270
271 /*
272 * First, see if we know about this pool. If not, then add it to the
273 * list of known pools.
274 */
275 for (pe = pl->pools; pe != NULL; pe = pe->pe_next) {
276 if (pe->pe_guid == pool_guid)
277 break;
278 }
279
280 if (pe == NULL) {
281 if ((pe = zfs_alloc(hdl, sizeof (pool_entry_t))) == NULL) {
282 nvlist_free(config);
283 return (-1);
284 }
285 pe->pe_guid = pool_guid;
286 pe->pe_next = pl->pools;
287 pl->pools = pe;
288 }
289
290 /*
291 * Second, see if we know about this toplevel vdev. Add it if its
292 * missing.
293 */
294 for (ve = pe->pe_vdevs; ve != NULL; ve = ve->ve_next) {
295 if (ve->ve_guid == top_guid)
296 break;
297 }
298
299 if (ve == NULL) {
300 if ((ve = zfs_alloc(hdl, sizeof (vdev_entry_t))) == NULL) {
301 nvlist_free(config);
302 return (-1);
303 }
304 ve->ve_guid = top_guid;
305 ve->ve_next = pe->pe_vdevs;
306 pe->pe_vdevs = ve;
307 }
308
309 /*
310 * Third, see if we have a config with a matching transaction group. If
311 * so, then we do nothing. Otherwise, add it to the list of known
312 * configs.
313 */
314 for (ce = ve->ve_configs; ce != NULL; ce = ce->ce_next) {
315 if (ce->ce_txg == txg)
316 break;
317 }
318
319 if (ce == NULL) {
320 if ((ce = zfs_alloc(hdl, sizeof (config_entry_t))) == NULL) {
321 nvlist_free(config);
322 return (-1);
323 }
324 ce->ce_txg = txg;
325 ce->ce_config = config;
326 ce->ce_next = ve->ve_configs;
327 ve->ve_configs = ce;
328 } else {
329 nvlist_free(config);
330 }
331
332 /*
333 * At this point we've successfully added our config to the list of
334 * known configs. The last thing to do is add the vdev guid -> path
335 * mappings so that we can fix up the configuration as necessary before
336 * doing the import.
337 */
338 if ((ne = zfs_alloc(hdl, sizeof (name_entry_t))) == NULL)
339 return (-1);
340
341 if ((ne->ne_name = zfs_strdup(hdl, path)) == NULL) {
342 free(ne);
343 return (-1);
344 }
345
346 ne->ne_guid = vdev_guid;
347 ne->ne_order = order;
348 ne->ne_num_labels = num_labels;
349 ne->ne_next = pl->names;
350 pl->names = ne;
351
352 return (0);
353 }
354
355 /*
356 * Returns true if the named pool matches the given GUID.
357 */
358 static int
359 pool_active(libzfs_handle_t *hdl, const char *name, uint64_t guid,
360 boolean_t *isactive)
361 {
362 zpool_handle_t *zhp;
363 uint64_t theguid;
364
365 if (zpool_open_silent(hdl, name, &zhp) != 0)
366 return (-1);
367
368 if (zhp == NULL) {
369 *isactive = B_FALSE;
370 return (0);
371 }
372
373 verify(nvlist_lookup_uint64(zhp->zpool_config, ZPOOL_CONFIG_POOL_GUID,
374 &theguid) == 0);
375
376 zpool_close(zhp);
377
378 *isactive = (theguid == guid);
379 return (0);
380 }
381
382 static nvlist_t *
383 refresh_config(libzfs_handle_t *hdl, nvlist_t *config)
384 {
385 nvlist_t *nvl;
386 zfs_cmd_t zc = {"\0"};
387 int err;
388
389 if (zcmd_write_conf_nvlist(hdl, &zc, config) != 0)
390 return (NULL);
391
392 if (zcmd_alloc_dst_nvlist(hdl, &zc,
393 zc.zc_nvlist_conf_size * 2) != 0) {
394 zcmd_free_nvlists(&zc);
395 return (NULL);
396 }
397
398 while ((err = ioctl(hdl->libzfs_fd, ZFS_IOC_POOL_TRYIMPORT,
399 &zc)) != 0 && errno == ENOMEM) {
400 if (zcmd_expand_dst_nvlist(hdl, &zc) != 0) {
401 zcmd_free_nvlists(&zc);
402 return (NULL);
403 }
404 }
405
406 if (err) {
407 zcmd_free_nvlists(&zc);
408 return (NULL);
409 }
410
411 if (zcmd_read_dst_nvlist(hdl, &zc, &nvl) != 0) {
412 zcmd_free_nvlists(&zc);
413 return (NULL);
414 }
415
416 zcmd_free_nvlists(&zc);
417 return (nvl);
418 }
419
420 /*
421 * Determine if the vdev id is a hole in the namespace.
422 */
423 boolean_t
424 vdev_is_hole(uint64_t *hole_array, uint_t holes, uint_t id)
425 {
426 int c;
427
428 for (c = 0; c < holes; c++) {
429
430 /* Top-level is a hole */
431 if (hole_array[c] == id)
432 return (B_TRUE);
433 }
434 return (B_FALSE);
435 }
436
437 /*
438 * Convert our list of pools into the definitive set of configurations. We
439 * start by picking the best config for each toplevel vdev. Once that's done,
440 * we assemble the toplevel vdevs into a full config for the pool. We make a
441 * pass to fix up any incorrect paths, and then add it to the main list to
442 * return to the user.
443 */
444 static nvlist_t *
445 get_configs(libzfs_handle_t *hdl, pool_list_t *pl, boolean_t active_ok)
446 {
447 pool_entry_t *pe;
448 vdev_entry_t *ve;
449 config_entry_t *ce;
450 nvlist_t *ret = NULL, *config = NULL, *tmp = NULL, *nvtop, *nvroot;
451 nvlist_t **spares, **l2cache;
452 uint_t i, nspares, nl2cache;
453 boolean_t config_seen;
454 uint64_t best_txg;
455 char *name, *hostname = NULL;
456 uint64_t guid;
457 uint_t children = 0;
458 nvlist_t **child = NULL;
459 uint_t holes;
460 uint64_t *hole_array, max_id;
461 uint_t c;
462 boolean_t isactive;
463 uint64_t hostid;
464 nvlist_t *nvl;
465 boolean_t valid_top_config = B_FALSE;
466
467 if (nvlist_alloc(&ret, 0, 0) != 0)
468 goto nomem;
469
470 for (pe = pl->pools; pe != NULL; pe = pe->pe_next) {
471 uint64_t id, max_txg = 0;
472
473 if (nvlist_alloc(&config, NV_UNIQUE_NAME, 0) != 0)
474 goto nomem;
475 config_seen = B_FALSE;
476
477 /*
478 * Iterate over all toplevel vdevs. Grab the pool configuration
479 * from the first one we find, and then go through the rest and
480 * add them as necessary to the 'vdevs' member of the config.
481 */
482 for (ve = pe->pe_vdevs; ve != NULL; ve = ve->ve_next) {
483
484 /*
485 * Determine the best configuration for this vdev by
486 * selecting the config with the latest transaction
487 * group.
488 */
489 best_txg = 0;
490 for (ce = ve->ve_configs; ce != NULL;
491 ce = ce->ce_next) {
492
493 if (ce->ce_txg > best_txg) {
494 tmp = ce->ce_config;
495 best_txg = ce->ce_txg;
496 }
497 }
498
499 /*
500 * We rely on the fact that the max txg for the
501 * pool will contain the most up-to-date information
502 * about the valid top-levels in the vdev namespace.
503 */
504 if (best_txg > max_txg) {
505 (void) nvlist_remove(config,
506 ZPOOL_CONFIG_VDEV_CHILDREN,
507 DATA_TYPE_UINT64);
508 (void) nvlist_remove(config,
509 ZPOOL_CONFIG_HOLE_ARRAY,
510 DATA_TYPE_UINT64_ARRAY);
511
512 max_txg = best_txg;
513 hole_array = NULL;
514 holes = 0;
515 max_id = 0;
516 valid_top_config = B_FALSE;
517
518 if (nvlist_lookup_uint64(tmp,
519 ZPOOL_CONFIG_VDEV_CHILDREN, &max_id) == 0) {
520 verify(nvlist_add_uint64(config,
521 ZPOOL_CONFIG_VDEV_CHILDREN,
522 max_id) == 0);
523 valid_top_config = B_TRUE;
524 }
525
526 if (nvlist_lookup_uint64_array(tmp,
527 ZPOOL_CONFIG_HOLE_ARRAY, &hole_array,
528 &holes) == 0) {
529 verify(nvlist_add_uint64_array(config,
530 ZPOOL_CONFIG_HOLE_ARRAY,
531 hole_array, holes) == 0);
532 }
533 }
534
535 if (!config_seen) {
536 /*
537 * Copy the relevant pieces of data to the pool
538 * configuration:
539 *
540 * version
541 * pool guid
542 * name
543 * comment (if available)
544 * pool state
545 * hostid (if available)
546 * hostname (if available)
547 */
548 uint64_t state, version;
549 char *comment = NULL;
550
551 version = fnvlist_lookup_uint64(tmp,
552 ZPOOL_CONFIG_VERSION);
553 fnvlist_add_uint64(config,
554 ZPOOL_CONFIG_VERSION, version);
555 guid = fnvlist_lookup_uint64(tmp,
556 ZPOOL_CONFIG_POOL_GUID);
557 fnvlist_add_uint64(config,
558 ZPOOL_CONFIG_POOL_GUID, guid);
559 name = fnvlist_lookup_string(tmp,
560 ZPOOL_CONFIG_POOL_NAME);
561 fnvlist_add_string(config,
562 ZPOOL_CONFIG_POOL_NAME, name);
563
564 if (nvlist_lookup_string(tmp,
565 ZPOOL_CONFIG_COMMENT, &comment) == 0)
566 fnvlist_add_string(config,
567 ZPOOL_CONFIG_COMMENT, comment);
568
569 state = fnvlist_lookup_uint64(tmp,
570 ZPOOL_CONFIG_POOL_STATE);
571 fnvlist_add_uint64(config,
572 ZPOOL_CONFIG_POOL_STATE, state);
573
574 hostid = 0;
575 if (nvlist_lookup_uint64(tmp,
576 ZPOOL_CONFIG_HOSTID, &hostid) == 0) {
577 fnvlist_add_uint64(config,
578 ZPOOL_CONFIG_HOSTID, hostid);
579 hostname = fnvlist_lookup_string(tmp,
580 ZPOOL_CONFIG_HOSTNAME);
581 fnvlist_add_string(config,
582 ZPOOL_CONFIG_HOSTNAME, hostname);
583 }
584
585 config_seen = B_TRUE;
586 }
587
588 /*
589 * Add this top-level vdev to the child array.
590 */
591 verify(nvlist_lookup_nvlist(tmp,
592 ZPOOL_CONFIG_VDEV_TREE, &nvtop) == 0);
593 verify(nvlist_lookup_uint64(nvtop, ZPOOL_CONFIG_ID,
594 &id) == 0);
595
596 if (id >= children) {
597 nvlist_t **newchild;
598
599 newchild = zfs_alloc(hdl, (id + 1) *
600 sizeof (nvlist_t *));
601 if (newchild == NULL)
602 goto nomem;
603
604 for (c = 0; c < children; c++)
605 newchild[c] = child[c];
606
607 free(child);
608 child = newchild;
609 children = id + 1;
610 }
611 if (nvlist_dup(nvtop, &child[id], 0) != 0)
612 goto nomem;
613
614 }
615
616 /*
617 * If we have information about all the top-levels then
618 * clean up the nvlist which we've constructed. This
619 * means removing any extraneous devices that are
620 * beyond the valid range or adding devices to the end
621 * of our array which appear to be missing.
622 */
623 if (valid_top_config) {
624 if (max_id < children) {
625 for (c = max_id; c < children; c++)
626 nvlist_free(child[c]);
627 children = max_id;
628 } else if (max_id > children) {
629 nvlist_t **newchild;
630
631 newchild = zfs_alloc(hdl, (max_id) *
632 sizeof (nvlist_t *));
633 if (newchild == NULL)
634 goto nomem;
635
636 for (c = 0; c < children; c++)
637 newchild[c] = child[c];
638
639 free(child);
640 child = newchild;
641 children = max_id;
642 }
643 }
644
645 verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID,
646 &guid) == 0);
647
648 /*
649 * The vdev namespace may contain holes as a result of
650 * device removal. We must add them back into the vdev
651 * tree before we process any missing devices.
652 */
653 if (holes > 0) {
654 ASSERT(valid_top_config);
655
656 for (c = 0; c < children; c++) {
657 nvlist_t *holey;
658
659 if (child[c] != NULL ||
660 !vdev_is_hole(hole_array, holes, c))
661 continue;
662
663 if (nvlist_alloc(&holey, NV_UNIQUE_NAME,
664 0) != 0)
665 goto nomem;
666
667 /*
668 * Holes in the namespace are treated as
669 * "hole" top-level vdevs and have a
670 * special flag set on them.
671 */
672 if (nvlist_add_string(holey,
673 ZPOOL_CONFIG_TYPE,
674 VDEV_TYPE_HOLE) != 0 ||
675 nvlist_add_uint64(holey,
676 ZPOOL_CONFIG_ID, c) != 0 ||
677 nvlist_add_uint64(holey,
678 ZPOOL_CONFIG_GUID, 0ULL) != 0)
679 goto nomem;
680 child[c] = holey;
681 }
682 }
683
684 /*
685 * Look for any missing top-level vdevs. If this is the case,
686 * create a faked up 'missing' vdev as a placeholder. We cannot
687 * simply compress the child array, because the kernel performs
688 * certain checks to make sure the vdev IDs match their location
689 * in the configuration.
690 */
691 for (c = 0; c < children; c++) {
692 if (child[c] == NULL) {
693 nvlist_t *missing;
694 if (nvlist_alloc(&missing, NV_UNIQUE_NAME,
695 0) != 0)
696 goto nomem;
697 if (nvlist_add_string(missing,
698 ZPOOL_CONFIG_TYPE,
699 VDEV_TYPE_MISSING) != 0 ||
700 nvlist_add_uint64(missing,
701 ZPOOL_CONFIG_ID, c) != 0 ||
702 nvlist_add_uint64(missing,
703 ZPOOL_CONFIG_GUID, 0ULL) != 0) {
704 nvlist_free(missing);
705 goto nomem;
706 }
707 child[c] = missing;
708 }
709 }
710
711 /*
712 * Put all of this pool's top-level vdevs into a root vdev.
713 */
714 if (nvlist_alloc(&nvroot, NV_UNIQUE_NAME, 0) != 0)
715 goto nomem;
716 if (nvlist_add_string(nvroot, ZPOOL_CONFIG_TYPE,
717 VDEV_TYPE_ROOT) != 0 ||
718 nvlist_add_uint64(nvroot, ZPOOL_CONFIG_ID, 0ULL) != 0 ||
719 nvlist_add_uint64(nvroot, ZPOOL_CONFIG_GUID, guid) != 0 ||
720 nvlist_add_nvlist_array(nvroot, ZPOOL_CONFIG_CHILDREN,
721 child, children) != 0) {
722 nvlist_free(nvroot);
723 goto nomem;
724 }
725
726 for (c = 0; c < children; c++)
727 nvlist_free(child[c]);
728 free(child);
729 children = 0;
730 child = NULL;
731
732 /*
733 * Go through and fix up any paths and/or devids based on our
734 * known list of vdev GUID -> path mappings.
735 */
736 if (fix_paths(nvroot, pl->names) != 0) {
737 nvlist_free(nvroot);
738 goto nomem;
739 }
740
741 /*
742 * Add the root vdev to this pool's configuration.
743 */
744 if (nvlist_add_nvlist(config, ZPOOL_CONFIG_VDEV_TREE,
745 nvroot) != 0) {
746 nvlist_free(nvroot);
747 goto nomem;
748 }
749 nvlist_free(nvroot);
750
751 /*
752 * zdb uses this path to report on active pools that were
753 * imported or created using -R.
754 */
755 if (active_ok)
756 goto add_pool;
757
758 /*
759 * Determine if this pool is currently active, in which case we
760 * can't actually import it.
761 */
762 verify(nvlist_lookup_string(config, ZPOOL_CONFIG_POOL_NAME,
763 &name) == 0);
764 verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID,
765 &guid) == 0);
766
767 if (pool_active(hdl, name, guid, &isactive) != 0)
768 goto error;
769
770 if (isactive) {
771 nvlist_free(config);
772 config = NULL;
773 continue;
774 }
775
776 if ((nvl = refresh_config(hdl, config)) == NULL) {
777 nvlist_free(config);
778 config = NULL;
779 continue;
780 }
781
782 nvlist_free(config);
783 config = nvl;
784
785 /*
786 * Go through and update the paths for spares, now that we have
787 * them.
788 */
789 verify(nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE,
790 &nvroot) == 0);
791 if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_SPARES,
792 &spares, &nspares) == 0) {
793 for (i = 0; i < nspares; i++) {
794 if (fix_paths(spares[i], pl->names) != 0)
795 goto nomem;
796 }
797 }
798
799 /*
800 * Update the paths for l2cache devices.
801 */
802 if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_L2CACHE,
803 &l2cache, &nl2cache) == 0) {
804 for (i = 0; i < nl2cache; i++) {
805 if (fix_paths(l2cache[i], pl->names) != 0)
806 goto nomem;
807 }
808 }
809
810 /*
811 * Restore the original information read from the actual label.
812 */
813 (void) nvlist_remove(config, ZPOOL_CONFIG_HOSTID,
814 DATA_TYPE_UINT64);
815 (void) nvlist_remove(config, ZPOOL_CONFIG_HOSTNAME,
816 DATA_TYPE_STRING);
817 if (hostid != 0) {
818 verify(nvlist_add_uint64(config, ZPOOL_CONFIG_HOSTID,
819 hostid) == 0);
820 verify(nvlist_add_string(config, ZPOOL_CONFIG_HOSTNAME,
821 hostname) == 0);
822 }
823
824 add_pool:
825 /*
826 * Add this pool to the list of configs.
827 */
828 verify(nvlist_lookup_string(config, ZPOOL_CONFIG_POOL_NAME,
829 &name) == 0);
830 if (nvlist_add_nvlist(ret, name, config) != 0)
831 goto nomem;
832
833 nvlist_free(config);
834 config = NULL;
835 }
836
837 return (ret);
838
839 nomem:
840 (void) no_memory(hdl);
841 error:
842 nvlist_free(config);
843 nvlist_free(ret);
844 for (c = 0; c < children; c++)
845 nvlist_free(child[c]);
846 free(child);
847
848 return (NULL);
849 }
850
851 /*
852 * Return the offset of the given label.
853 */
854 static uint64_t
855 label_offset(uint64_t size, int l)
856 {
857 ASSERT(P2PHASE_TYPED(size, sizeof (vdev_label_t), uint64_t) == 0);
858 return (l * sizeof (vdev_label_t) + (l < VDEV_LABELS / 2 ?
859 0 : size - VDEV_LABELS * sizeof (vdev_label_t)));
860 }
861
862 /*
863 * Given a file descriptor, read the label information and return an nvlist
864 * describing the configuration, if there is one. The number of valid
865 * labels found will be returned in num_labels when non-NULL.
866 */
867 int
868 zpool_read_label(int fd, nvlist_t **config, int *num_labels)
869 {
870 struct stat64 statbuf;
871 int l, count = 0;
872 vdev_label_t *label;
873 nvlist_t *expected_config = NULL;
874 uint64_t expected_guid = 0, size;
875
876 *config = NULL;
877
878 if (fstat64_blk(fd, &statbuf) == -1)
879 return (0);
880 size = P2ALIGN_TYPED(statbuf.st_size, sizeof (vdev_label_t), uint64_t);
881
882 if ((label = malloc(sizeof (vdev_label_t))) == NULL)
883 return (-1);
884
885 for (l = 0; l < VDEV_LABELS; l++) {
886 uint64_t state, guid, txg;
887
888 if (pread64(fd, label, sizeof (vdev_label_t),
889 label_offset(size, l)) != sizeof (vdev_label_t))
890 continue;
891
892 if (nvlist_unpack(label->vl_vdev_phys.vp_nvlist,
893 sizeof (label->vl_vdev_phys.vp_nvlist), config, 0) != 0)
894 continue;
895
896 if (nvlist_lookup_uint64(*config, ZPOOL_CONFIG_GUID,
897 &guid) != 0 || guid == 0) {
898 nvlist_free(*config);
899 continue;
900 }
901
902 if (nvlist_lookup_uint64(*config, ZPOOL_CONFIG_POOL_STATE,
903 &state) != 0 || state > POOL_STATE_L2CACHE) {
904 nvlist_free(*config);
905 continue;
906 }
907
908 if (state != POOL_STATE_SPARE && state != POOL_STATE_L2CACHE &&
909 (nvlist_lookup_uint64(*config, ZPOOL_CONFIG_POOL_TXG,
910 &txg) != 0 || txg == 0)) {
911 nvlist_free(*config);
912 continue;
913 }
914
915 if (expected_guid) {
916 if (expected_guid == guid)
917 count++;
918
919 nvlist_free(*config);
920 } else {
921 expected_config = *config;
922 expected_guid = guid;
923 count++;
924 }
925 }
926
927 if (num_labels != NULL)
928 *num_labels = count;
929
930 free(label);
931 *config = expected_config;
932
933 return (0);
934 }
935
936 /*
937 * Given a file descriptor, clear (zero) the label information. This function
938 * is used in the appliance stack as part of the ZFS sysevent module and
939 * to implement the "zpool labelclear" command.
940 */
941 int
942 zpool_clear_label(int fd)
943 {
944 struct stat64 statbuf;
945 int l;
946 vdev_label_t *label;
947 uint64_t size;
948
949 if (fstat64_blk(fd, &statbuf) == -1)
950 return (0);
951 size = P2ALIGN_TYPED(statbuf.st_size, sizeof (vdev_label_t), uint64_t);
952
953 if ((label = calloc(sizeof (vdev_label_t), 1)) == NULL)
954 return (-1);
955
956 for (l = 0; l < VDEV_LABELS; l++) {
957 if (pwrite64(fd, label, sizeof (vdev_label_t),
958 label_offset(size, l)) != sizeof (vdev_label_t)) {
959 free(label);
960 return (-1);
961 }
962 }
963
964 free(label);
965 return (0);
966 }
967
968 #ifdef HAVE_LIBBLKID
969 /*
970 * Use libblkid to quickly search for zfs devices
971 */
972 static int
973 zpool_find_import_blkid(libzfs_handle_t *hdl, pool_list_t *pools)
974 {
975 blkid_cache cache;
976 blkid_dev_iterate iter;
977 blkid_dev dev;
978 const char *devname;
979 nvlist_t *config;
980 int fd, err, num_labels;
981
982 err = blkid_get_cache(&cache, NULL);
983 if (err != 0) {
984 (void) zfs_error_fmt(hdl, EZFS_BADCACHE,
985 dgettext(TEXT_DOMAIN, "blkid_get_cache() %d"), err);
986 goto err_blkid1;
987 }
988
989 err = blkid_probe_all(cache);
990 if (err != 0) {
991 (void) zfs_error_fmt(hdl, EZFS_BADCACHE,
992 dgettext(TEXT_DOMAIN, "blkid_probe_all() %d"), err);
993 goto err_blkid2;
994 }
995
996 iter = blkid_dev_iterate_begin(cache);
997 if (iter == NULL) {
998 (void) zfs_error_fmt(hdl, EZFS_BADCACHE,
999 dgettext(TEXT_DOMAIN, "blkid_dev_iterate_begin()"));
1000 goto err_blkid2;
1001 }
1002
1003 err = blkid_dev_set_search(iter, "TYPE", "zfs_member");
1004 if (err != 0) {
1005 (void) zfs_error_fmt(hdl, EZFS_BADCACHE,
1006 dgettext(TEXT_DOMAIN, "blkid_dev_set_search() %d"), err);
1007 goto err_blkid3;
1008 }
1009
1010 while (blkid_dev_next(iter, &dev) == 0) {
1011 devname = blkid_dev_devname(dev);
1012 if ((fd = open64(devname, O_RDONLY)) < 0)
1013 continue;
1014
1015 err = zpool_read_label(fd, &config, &num_labels);
1016 (void) close(fd);
1017
1018 if (err != 0) {
1019 (void) no_memory(hdl);
1020 goto err_blkid3;
1021 }
1022
1023 if (config != NULL) {
1024 err = add_config(hdl, pools, devname, 0,
1025 num_labels, config);
1026 if (err != 0)
1027 goto err_blkid3;
1028 }
1029 }
1030
1031 err_blkid3:
1032 blkid_dev_iterate_end(iter);
1033 err_blkid2:
1034 blkid_put_cache(cache);
1035 err_blkid1:
1036 return (err);
1037 }
1038 #endif /* HAVE_LIBBLKID */
1039
1040 char *
1041 zpool_default_import_path[DEFAULT_IMPORT_PATH_SIZE] = {
1042 "/dev/disk/by-vdev", /* Custom rules, use first if they exist */
1043 "/dev/mapper", /* Use multipath devices before components */
1044 "/dev/disk/by-uuid", /* Single unique entry and persistent */
1045 "/dev/disk/by-id", /* May be multiple entries and persistent */
1046 "/dev/disk/by-path", /* Encodes physical location and persistent */
1047 "/dev/disk/by-label", /* Custom persistent labels */
1048 "/dev" /* UNSAFE device names will change */
1049 };
1050
1051 /*
1052 * Given a list of directories to search, find all pools stored on disk. This
1053 * includes partial pools which are not available to import. If no args are
1054 * given (argc is 0), then the default directory (/dev/dsk) is searched.
1055 * poolname or guid (but not both) are provided by the caller when trying
1056 * to import a specific pool.
1057 */
1058 static nvlist_t *
1059 zpool_find_import_impl(libzfs_handle_t *hdl, importargs_t *iarg)
1060 {
1061 int i, num_labels, dirs = iarg->paths;
1062 DIR *dirp = NULL;
1063 struct dirent64 *dp;
1064 char path[MAXPATHLEN];
1065 char *end, **dir = iarg->path;
1066 size_t pathleft;
1067 struct stat64 statbuf;
1068 nvlist_t *ret = NULL, *config;
1069 int fd;
1070 pool_list_t pools = { 0 };
1071 pool_entry_t *pe, *penext;
1072 vdev_entry_t *ve, *venext;
1073 config_entry_t *ce, *cenext;
1074 name_entry_t *ne, *nenext;
1075
1076 verify(iarg->poolname == NULL || iarg->guid == 0);
1077
1078 if (dirs == 0) {
1079 #ifdef HAVE_LIBBLKID
1080 /* Use libblkid to scan all device for their type */
1081 if (zpool_find_import_blkid(hdl, &pools) == 0)
1082 goto skip_scanning;
1083
1084 (void) zfs_error_fmt(hdl, EZFS_BADCACHE,
1085 dgettext(TEXT_DOMAIN, "blkid failure falling back "
1086 "to manual probing"));
1087 #endif /* HAVE_LIBBLKID */
1088
1089 dir = zpool_default_import_path;
1090 dirs = DEFAULT_IMPORT_PATH_SIZE;
1091 }
1092
1093 /*
1094 * Go through and read the label configuration information from every
1095 * possible device, organizing the information according to pool GUID
1096 * and toplevel GUID.
1097 */
1098 for (i = 0; i < dirs; i++) {
1099 char *rdsk;
1100 int dfd;
1101
1102 /* use realpath to normalize the path */
1103 if (realpath(dir[i], path) == 0) {
1104
1105 /* it is safe to skip missing search paths */
1106 if (errno == ENOENT)
1107 continue;
1108
1109 zfs_error_aux(hdl, strerror(errno));
1110 (void) zfs_error_fmt(hdl, EZFS_BADPATH,
1111 dgettext(TEXT_DOMAIN, "cannot open '%s'"), dir[i]);
1112 goto error;
1113 }
1114 end = &path[strlen(path)];
1115 *end++ = '/';
1116 *end = 0;
1117 pathleft = &path[sizeof (path)] - end;
1118
1119 /*
1120 * Using raw devices instead of block devices when we're
1121 * reading the labels skips a bunch of slow operations during
1122 * close(2) processing, so we replace /dev/dsk with /dev/rdsk.
1123 */
1124 if (strcmp(path, "/dev/dsk/") == 0)
1125 rdsk = "/dev/rdsk/";
1126 else
1127 rdsk = path;
1128
1129 if ((dfd = open64(rdsk, O_RDONLY)) < 0 ||
1130 (dirp = fdopendir(dfd)) == NULL) {
1131 zfs_error_aux(hdl, strerror(errno));
1132 (void) zfs_error_fmt(hdl, EZFS_BADPATH,
1133 dgettext(TEXT_DOMAIN, "cannot open '%s'"),
1134 rdsk);
1135 goto error;
1136 }
1137
1138 /*
1139 * This is not MT-safe, but we have no MT consumers of libzfs
1140 */
1141 while ((dp = readdir64(dirp)) != NULL) {
1142 const char *name = dp->d_name;
1143 if (name[0] == '.' &&
1144 (name[1] == 0 || (name[1] == '.' && name[2] == 0)))
1145 continue;
1146
1147 /*
1148 * Skip checking devices with well known prefixes:
1149 * watchdog - A special close is required to avoid
1150 * triggering it and resetting the system.
1151 * fuse - Fuse control device.
1152 * ppp - Generic PPP driver.
1153 * tty* - Generic serial interface.
1154 * vcs* - Virtual console memory.
1155 * parport* - Parallel port interface.
1156 * lp* - Printer interface.
1157 * fd* - Floppy interface.
1158 * hpet - High Precision Event Timer, crashes qemu
1159 * when accessed from a virtual machine.
1160 * core - Symlink to /proc/kcore, causes a crash
1161 * when access from Xen dom0.
1162 */
1163 if ((strncmp(name, "watchdog", 8) == 0) ||
1164 (strncmp(name, "fuse", 4) == 0) ||
1165 (strncmp(name, "ppp", 3) == 0) ||
1166 (strncmp(name, "tty", 3) == 0) ||
1167 (strncmp(name, "vcs", 3) == 0) ||
1168 (strncmp(name, "parport", 7) == 0) ||
1169 (strncmp(name, "lp", 2) == 0) ||
1170 (strncmp(name, "fd", 2) == 0) ||
1171 (strncmp(name, "hpet", 4) == 0) ||
1172 (strncmp(name, "core", 4) == 0))
1173 continue;
1174
1175 /*
1176 * Ignore failed stats. We only want regular
1177 * files and block devices.
1178 */
1179 if ((fstatat64(dfd, name, &statbuf, 0) != 0) ||
1180 (!S_ISREG(statbuf.st_mode) &&
1181 !S_ISBLK(statbuf.st_mode)))
1182 continue;
1183
1184 if ((fd = openat64(dfd, name, O_RDONLY)) < 0)
1185 continue;
1186
1187 if ((zpool_read_label(fd, &config, &num_labels))) {
1188 (void) close(fd);
1189 (void) no_memory(hdl);
1190 goto error;
1191 }
1192
1193 (void) close(fd);
1194
1195 if (config != NULL) {
1196 boolean_t matched = B_TRUE;
1197 char *pname;
1198
1199 if ((iarg->poolname != NULL) &&
1200 (nvlist_lookup_string(config,
1201 ZPOOL_CONFIG_POOL_NAME, &pname) == 0)) {
1202
1203 if (strcmp(iarg->poolname, pname))
1204 matched = B_FALSE;
1205
1206 } else if (iarg->guid != 0) {
1207 uint64_t this_guid;
1208
1209 matched = nvlist_lookup_uint64(config,
1210 ZPOOL_CONFIG_POOL_GUID,
1211 &this_guid) == 0 &&
1212 iarg->guid == this_guid;
1213 }
1214 if (!matched) {
1215 nvlist_free(config);
1216 config = NULL;
1217 continue;
1218 }
1219 /* use the non-raw path for the config */
1220 (void) strlcpy(end, name, pathleft);
1221 if (add_config(hdl, &pools, path, i+1,
1222 num_labels, config))
1223 goto error;
1224 }
1225 }
1226
1227 (void) closedir(dirp);
1228 dirp = NULL;
1229 }
1230
1231 #ifdef HAVE_LIBBLKID
1232 skip_scanning:
1233 #endif
1234 ret = get_configs(hdl, &pools, iarg->can_be_active);
1235
1236 error:
1237 for (pe = pools.pools; pe != NULL; pe = penext) {
1238 penext = pe->pe_next;
1239 for (ve = pe->pe_vdevs; ve != NULL; ve = venext) {
1240 venext = ve->ve_next;
1241 for (ce = ve->ve_configs; ce != NULL; ce = cenext) {
1242 cenext = ce->ce_next;
1243 if (ce->ce_config)
1244 nvlist_free(ce->ce_config);
1245 free(ce);
1246 }
1247 free(ve);
1248 }
1249 free(pe);
1250 }
1251
1252 for (ne = pools.names; ne != NULL; ne = nenext) {
1253 nenext = ne->ne_next;
1254 if (ne->ne_name)
1255 free(ne->ne_name);
1256 free(ne);
1257 }
1258
1259 if (dirp)
1260 (void) closedir(dirp);
1261
1262 return (ret);
1263 }
1264
1265 nvlist_t *
1266 zpool_find_import(libzfs_handle_t *hdl, int argc, char **argv)
1267 {
1268 importargs_t iarg = { 0 };
1269
1270 iarg.paths = argc;
1271 iarg.path = argv;
1272
1273 return (zpool_find_import_impl(hdl, &iarg));
1274 }
1275
1276 /*
1277 * Given a cache file, return the contents as a list of importable pools.
1278 * poolname or guid (but not both) are provided by the caller when trying
1279 * to import a specific pool.
1280 */
1281 nvlist_t *
1282 zpool_find_import_cached(libzfs_handle_t *hdl, const char *cachefile,
1283 char *poolname, uint64_t guid)
1284 {
1285 char *buf;
1286 int fd;
1287 struct stat64 statbuf;
1288 nvlist_t *raw, *src, *dst;
1289 nvlist_t *pools;
1290 nvpair_t *elem;
1291 char *name;
1292 uint64_t this_guid;
1293 boolean_t active;
1294
1295 verify(poolname == NULL || guid == 0);
1296
1297 if ((fd = open(cachefile, O_RDONLY)) < 0) {
1298 zfs_error_aux(hdl, "%s", strerror(errno));
1299 (void) zfs_error(hdl, EZFS_BADCACHE,
1300 dgettext(TEXT_DOMAIN, "failed to open cache file"));
1301 return (NULL);
1302 }
1303
1304 if (fstat64(fd, &statbuf) != 0) {
1305 zfs_error_aux(hdl, "%s", strerror(errno));
1306 (void) close(fd);
1307 (void) zfs_error(hdl, EZFS_BADCACHE,
1308 dgettext(TEXT_DOMAIN, "failed to get size of cache file"));
1309 return (NULL);
1310 }
1311
1312 if ((buf = zfs_alloc(hdl, statbuf.st_size)) == NULL) {
1313 (void) close(fd);
1314 return (NULL);
1315 }
1316
1317 if (read(fd, buf, statbuf.st_size) != statbuf.st_size) {
1318 (void) close(fd);
1319 free(buf);
1320 (void) zfs_error(hdl, EZFS_BADCACHE,
1321 dgettext(TEXT_DOMAIN,
1322 "failed to read cache file contents"));
1323 return (NULL);
1324 }
1325
1326 (void) close(fd);
1327
1328 if (nvlist_unpack(buf, statbuf.st_size, &raw, 0) != 0) {
1329 free(buf);
1330 (void) zfs_error(hdl, EZFS_BADCACHE,
1331 dgettext(TEXT_DOMAIN,
1332 "invalid or corrupt cache file contents"));
1333 return (NULL);
1334 }
1335
1336 free(buf);
1337
1338 /*
1339 * Go through and get the current state of the pools and refresh their
1340 * state.
1341 */
1342 if (nvlist_alloc(&pools, 0, 0) != 0) {
1343 (void) no_memory(hdl);
1344 nvlist_free(raw);
1345 return (NULL);
1346 }
1347
1348 elem = NULL;
1349 while ((elem = nvlist_next_nvpair(raw, elem)) != NULL) {
1350 src = fnvpair_value_nvlist(elem);
1351
1352 name = fnvlist_lookup_string(src, ZPOOL_CONFIG_POOL_NAME);
1353 if (poolname != NULL && strcmp(poolname, name) != 0)
1354 continue;
1355
1356 this_guid = fnvlist_lookup_uint64(src, ZPOOL_CONFIG_POOL_GUID);
1357 if (guid != 0 && guid != this_guid)
1358 continue;
1359
1360 if (pool_active(hdl, name, this_guid, &active) != 0) {
1361 nvlist_free(raw);
1362 nvlist_free(pools);
1363 return (NULL);
1364 }
1365
1366 if (active)
1367 continue;
1368
1369 if ((dst = refresh_config(hdl, src)) == NULL) {
1370 nvlist_free(raw);
1371 nvlist_free(pools);
1372 return (NULL);
1373 }
1374
1375 if (nvlist_add_nvlist(pools, nvpair_name(elem), dst) != 0) {
1376 (void) no_memory(hdl);
1377 nvlist_free(dst);
1378 nvlist_free(raw);
1379 nvlist_free(pools);
1380 return (NULL);
1381 }
1382 nvlist_free(dst);
1383 }
1384
1385 nvlist_free(raw);
1386 return (pools);
1387 }
1388
1389 static int
1390 name_or_guid_exists(zpool_handle_t *zhp, void *data)
1391 {
1392 importargs_t *import = data;
1393 int found = 0;
1394
1395 if (import->poolname != NULL) {
1396 char *pool_name;
1397
1398 verify(nvlist_lookup_string(zhp->zpool_config,
1399 ZPOOL_CONFIG_POOL_NAME, &pool_name) == 0);
1400 if (strcmp(pool_name, import->poolname) == 0)
1401 found = 1;
1402 } else {
1403 uint64_t pool_guid;
1404
1405 verify(nvlist_lookup_uint64(zhp->zpool_config,
1406 ZPOOL_CONFIG_POOL_GUID, &pool_guid) == 0);
1407 if (pool_guid == import->guid)
1408 found = 1;
1409 }
1410
1411 zpool_close(zhp);
1412 return (found);
1413 }
1414
1415 nvlist_t *
1416 zpool_search_import(libzfs_handle_t *hdl, importargs_t *import)
1417 {
1418 verify(import->poolname == NULL || import->guid == 0);
1419
1420 if (import->unique)
1421 import->exists = zpool_iter(hdl, name_or_guid_exists, import);
1422
1423 if (import->cachefile != NULL)
1424 return (zpool_find_import_cached(hdl, import->cachefile,
1425 import->poolname, import->guid));
1426
1427 return (zpool_find_import_impl(hdl, import));
1428 }
1429
1430 boolean_t
1431 find_guid(nvlist_t *nv, uint64_t guid)
1432 {
1433 uint64_t tmp;
1434 nvlist_t **child;
1435 uint_t c, children;
1436
1437 verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID, &tmp) == 0);
1438 if (tmp == guid)
1439 return (B_TRUE);
1440
1441 if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN,
1442 &child, &children) == 0) {
1443 for (c = 0; c < children; c++)
1444 if (find_guid(child[c], guid))
1445 return (B_TRUE);
1446 }
1447
1448 return (B_FALSE);
1449 }
1450
1451 typedef struct aux_cbdata {
1452 const char *cb_type;
1453 uint64_t cb_guid;
1454 zpool_handle_t *cb_zhp;
1455 } aux_cbdata_t;
1456
1457 static int
1458 find_aux(zpool_handle_t *zhp, void *data)
1459 {
1460 aux_cbdata_t *cbp = data;
1461 nvlist_t **list;
1462 uint_t i, count;
1463 uint64_t guid;
1464 nvlist_t *nvroot;
1465
1466 verify(nvlist_lookup_nvlist(zhp->zpool_config, ZPOOL_CONFIG_VDEV_TREE,
1467 &nvroot) == 0);
1468
1469 if (nvlist_lookup_nvlist_array(nvroot, cbp->cb_type,
1470 &list, &count) == 0) {
1471 for (i = 0; i < count; i++) {
1472 verify(nvlist_lookup_uint64(list[i],
1473 ZPOOL_CONFIG_GUID, &guid) == 0);
1474 if (guid == cbp->cb_guid) {
1475 cbp->cb_zhp = zhp;
1476 return (1);
1477 }
1478 }
1479 }
1480
1481 zpool_close(zhp);
1482 return (0);
1483 }
1484
1485 /*
1486 * Determines if the pool is in use. If so, it returns true and the state of
1487 * the pool as well as the name of the pool. Both strings are allocated and
1488 * must be freed by the caller.
1489 */
1490 int
1491 zpool_in_use(libzfs_handle_t *hdl, int fd, pool_state_t *state, char **namestr,
1492 boolean_t *inuse)
1493 {
1494 nvlist_t *config;
1495 char *name;
1496 boolean_t ret;
1497 uint64_t guid, vdev_guid;
1498 zpool_handle_t *zhp;
1499 nvlist_t *pool_config;
1500 uint64_t stateval, isspare;
1501 aux_cbdata_t cb = { 0 };
1502 boolean_t isactive;
1503
1504 *inuse = B_FALSE;
1505
1506 if (zpool_read_label(fd, &config, NULL) != 0) {
1507 (void) no_memory(hdl);
1508 return (-1);
1509 }
1510
1511 if (config == NULL)
1512 return (0);
1513
1514 verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_STATE,
1515 &stateval) == 0);
1516 verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_GUID,
1517 &vdev_guid) == 0);
1518
1519 if (stateval != POOL_STATE_SPARE && stateval != POOL_STATE_L2CACHE) {
1520 verify(nvlist_lookup_string(config, ZPOOL_CONFIG_POOL_NAME,
1521 &name) == 0);
1522 verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID,
1523 &guid) == 0);
1524 }
1525
1526 switch (stateval) {
1527 case POOL_STATE_EXPORTED:
1528 /*
1529 * A pool with an exported state may in fact be imported
1530 * read-only, so check the in-core state to see if it's
1531 * active and imported read-only. If it is, set
1532 * its state to active.
1533 */
1534 if (pool_active(hdl, name, guid, &isactive) == 0 && isactive &&
1535 (zhp = zpool_open_canfail(hdl, name)) != NULL &&
1536 zpool_get_prop_int(zhp, ZPOOL_PROP_READONLY, NULL))
1537 stateval = POOL_STATE_ACTIVE;
1538
1539 ret = B_TRUE;
1540 break;
1541
1542 case POOL_STATE_ACTIVE:
1543 /*
1544 * For an active pool, we have to determine if it's really part
1545 * of a currently active pool (in which case the pool will exist
1546 * and the guid will be the same), or whether it's part of an
1547 * active pool that was disconnected without being explicitly
1548 * exported.
1549 */
1550 if (pool_active(hdl, name, guid, &isactive) != 0) {
1551 nvlist_free(config);
1552 return (-1);
1553 }
1554
1555 if (isactive) {
1556 /*
1557 * Because the device may have been removed while
1558 * offlined, we only report it as active if the vdev is
1559 * still present in the config. Otherwise, pretend like
1560 * it's not in use.
1561 */
1562 if ((zhp = zpool_open_canfail(hdl, name)) != NULL &&
1563 (pool_config = zpool_get_config(zhp, NULL))
1564 != NULL) {
1565 nvlist_t *nvroot;
1566
1567 verify(nvlist_lookup_nvlist(pool_config,
1568 ZPOOL_CONFIG_VDEV_TREE, &nvroot) == 0);
1569 ret = find_guid(nvroot, vdev_guid);
1570 } else {
1571 ret = B_FALSE;
1572 }
1573
1574 /*
1575 * If this is an active spare within another pool, we
1576 * treat it like an unused hot spare. This allows the
1577 * user to create a pool with a hot spare that currently
1578 * in use within another pool. Since we return B_TRUE,
1579 * libdiskmgt will continue to prevent generic consumers
1580 * from using the device.
1581 */
1582 if (ret && nvlist_lookup_uint64(config,
1583 ZPOOL_CONFIG_IS_SPARE, &isspare) == 0 && isspare)
1584 stateval = POOL_STATE_SPARE;
1585
1586 if (zhp != NULL)
1587 zpool_close(zhp);
1588 } else {
1589 stateval = POOL_STATE_POTENTIALLY_ACTIVE;
1590 ret = B_TRUE;
1591 }
1592 break;
1593
1594 case POOL_STATE_SPARE:
1595 /*
1596 * For a hot spare, it can be either definitively in use, or
1597 * potentially active. To determine if it's in use, we iterate
1598 * over all pools in the system and search for one with a spare
1599 * with a matching guid.
1600 *
1601 * Due to the shared nature of spares, we don't actually report
1602 * the potentially active case as in use. This means the user
1603 * can freely create pools on the hot spares of exported pools,
1604 * but to do otherwise makes the resulting code complicated, and
1605 * we end up having to deal with this case anyway.
1606 */
1607 cb.cb_zhp = NULL;
1608 cb.cb_guid = vdev_guid;
1609 cb.cb_type = ZPOOL_CONFIG_SPARES;
1610 if (zpool_iter(hdl, find_aux, &cb) == 1) {
1611 name = (char *)zpool_get_name(cb.cb_zhp);
1612 ret = TRUE;
1613 } else {
1614 ret = FALSE;
1615 }
1616 break;
1617
1618 case POOL_STATE_L2CACHE:
1619
1620 /*
1621 * Check if any pool is currently using this l2cache device.
1622 */
1623 cb.cb_zhp = NULL;
1624 cb.cb_guid = vdev_guid;
1625 cb.cb_type = ZPOOL_CONFIG_L2CACHE;
1626 if (zpool_iter(hdl, find_aux, &cb) == 1) {
1627 name = (char *)zpool_get_name(cb.cb_zhp);
1628 ret = TRUE;
1629 } else {
1630 ret = FALSE;
1631 }
1632 break;
1633
1634 default:
1635 ret = B_FALSE;
1636 }
1637
1638
1639 if (ret) {
1640 if ((*namestr = zfs_strdup(hdl, name)) == NULL) {
1641 if (cb.cb_zhp)
1642 zpool_close(cb.cb_zhp);
1643 nvlist_free(config);
1644 return (-1);
1645 }
1646 *state = (pool_state_t)stateval;
1647 }
1648
1649 if (cb.cb_zhp)
1650 zpool_close(cb.cb_zhp);
1651
1652 nvlist_free(config);
1653 *inuse = ret;
1654 return (0);
1655 }
Mirror of ZFSOnLinux for Debian