]> git.proxmox.com Git - mirror_zfs-debian.git/blame - lib/libzfs/libzfs_import.c
Improve `zpool import` search behavior
[mirror_zfs-debian.git] / lib / libzfs / libzfs_import.c
CommitLineData
34dc7c2f
BB
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/*
572e2857 22 * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
d96eb2b1
DM
23 * Copyright 2011 Nexenta Systems, Inc. All rights reserved.
24 * Copyright (c) 2011 by Delphix. All rights reserved.
34dc7c2f
BB
25 */
26
34dc7c2f
BB
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
428870ff 43#include <ctype.h>
34dc7c2f
BB
44#include <devid.h>
45#include <dirent.h>
46#include <errno.h>
47#include <libintl.h>
428870ff 48#include <stddef.h>
34dc7c2f
BB
49#include <stdlib.h>
50#include <string.h>
51#include <sys/stat.h>
52#include <unistd.h>
53#include <fcntl.h>
428870ff
BB
54#include <sys/vtoc.h>
55#include <sys/dktp/fdisk.h>
56#include <sys/efi_partition.h>
34dc7c2f
BB
57
58#include <sys/vdev_impl.h>
d603ed6c
BB
59#ifdef HAVE_LIBBLKID
60#include <blkid/blkid.h>
61#endif
34dc7c2f
BB
62
63#include "libzfs.h"
64#include "libzfs_impl.h"
65
66/*
67 * Intermediate structures used to gather configuration information.
68 */
69typedef struct config_entry {
70 uint64_t ce_txg;
71 nvlist_t *ce_config;
72 struct config_entry *ce_next;
73} config_entry_t;
74
75typedef 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
81typedef 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
87typedef struct name_entry {
88 char *ne_name;
89 uint64_t ne_guid;
44867b6d 90 uint64_t ne_order;
34dc7c2f
BB
91 struct name_entry *ne_next;
92} name_entry_t;
93
94typedef struct pool_list {
95 pool_entry_t *pools;
96 name_entry_t *names;
97} pool_list_t;
98
99static char *
100get_devid(const char *path)
101{
102 int fd;
103 ddi_devid_t devid;
104 char *minor, *ret;
105
106 if ((fd = open(path, O_RDONLY)) < 0)
107 return (NULL);
108
109 minor = NULL;
110 ret = NULL;
111 if (devid_get(fd, &devid) == 0) {
112 if (devid_get_minor_name(fd, &minor) == 0)
113 ret = devid_str_encode(devid, minor);
114 if (minor != NULL)
115 devid_str_free(minor);
116 devid_free(devid);
117 }
118 (void) close(fd);
119
120 return (ret);
121}
122
123
124/*
125 * Go through and fix up any path and/or devid information for the given vdev
126 * configuration.
127 */
128static int
129fix_paths(nvlist_t *nv, name_entry_t *names)
130{
131 nvlist_t **child;
132 uint_t c, children;
133 uint64_t guid;
134 name_entry_t *ne, *best;
135 char *path, *devid;
34dc7c2f
BB
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
44867b6d
BB
151 * which case we have overlapping partitions or multiple paths to the
152 * same disk. In this case we prefer to use the path name which
153 * matches the ZPOOL_CONFIG_PATH. If no matching entry is found we
154 * use the lowest order device which corresponds to the first match
155 * while traversing the ZPOOL_IMPORT_PATH search path.
34dc7c2f
BB
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
34dc7c2f
BB
161 best = NULL;
162 for (ne = names; ne != NULL; ne = ne->ne_next) {
163 if (ne->ne_guid == guid) {
34dc7c2f
BB
164
165 if (path == NULL) {
166 best = ne;
167 break;
168 }
169
44867b6d
BB
170 if ((strlen(path) == strlen(ne->ne_name)) &&
171 !strncmp(path, ne->ne_name, strlen(path))) {
34dc7c2f 172 best = ne;
44867b6d 173 break;
34dc7c2f 174 }
44867b6d
BB
175
176 if (ne->ne_order < best->ne_order || best == NULL)
177 best = ne;
34dc7c2f
BB
178 }
179 }
180
181 if (best == NULL)
182 return (0);
183
184 if (nvlist_add_string(nv, ZPOOL_CONFIG_PATH, best->ne_name) != 0)
185 return (-1);
186
187 if ((devid = get_devid(best->ne_name)) == NULL) {
188 (void) nvlist_remove_all(nv, ZPOOL_CONFIG_DEVID);
189 } else {
190 if (nvlist_add_string(nv, ZPOOL_CONFIG_DEVID, devid) != 0)
191 return (-1);
192 devid_str_free(devid);
193 }
194
195 return (0);
196}
197
198/*
199 * Add the given configuration to the list of known devices.
200 */
201static int
202add_config(libzfs_handle_t *hdl, pool_list_t *pl, const char *path,
44867b6d 203 int order, nvlist_t *config)
34dc7c2f
BB
204{
205 uint64_t pool_guid, vdev_guid, top_guid, txg, state;
206 pool_entry_t *pe;
207 vdev_entry_t *ve;
208 config_entry_t *ce;
209 name_entry_t *ne;
210
211 /*
212 * If this is a hot spare not currently in use or level 2 cache
213 * device, add it to the list of names to translate, but don't do
214 * anything else.
215 */
216 if (nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_STATE,
217 &state) == 0 &&
218 (state == POOL_STATE_SPARE || state == POOL_STATE_L2CACHE) &&
219 nvlist_lookup_uint64(config, ZPOOL_CONFIG_GUID, &vdev_guid) == 0) {
220 if ((ne = zfs_alloc(hdl, sizeof (name_entry_t))) == NULL)
221 return (-1);
222
223 if ((ne->ne_name = zfs_strdup(hdl, path)) == NULL) {
224 free(ne);
225 return (-1);
226 }
227 ne->ne_guid = vdev_guid;
44867b6d 228 ne->ne_order = order;
34dc7c2f
BB
229 ne->ne_next = pl->names;
230 pl->names = ne;
231 return (0);
232 }
233
234 /*
235 * If we have a valid config but cannot read any of these fields, then
236 * it means we have a half-initialized label. In vdev_label_init()
237 * we write a label with txg == 0 so that we can identify the device
238 * in case the user refers to the same disk later on. If we fail to
239 * create the pool, we'll be left with a label in this state
240 * which should not be considered part of a valid pool.
241 */
242 if (nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID,
243 &pool_guid) != 0 ||
244 nvlist_lookup_uint64(config, ZPOOL_CONFIG_GUID,
245 &vdev_guid) != 0 ||
246 nvlist_lookup_uint64(config, ZPOOL_CONFIG_TOP_GUID,
247 &top_guid) != 0 ||
248 nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_TXG,
249 &txg) != 0 || txg == 0) {
250 nvlist_free(config);
251 return (0);
252 }
253
254 /*
255 * First, see if we know about this pool. If not, then add it to the
256 * list of known pools.
257 */
258 for (pe = pl->pools; pe != NULL; pe = pe->pe_next) {
259 if (pe->pe_guid == pool_guid)
260 break;
261 }
262
263 if (pe == NULL) {
264 if ((pe = zfs_alloc(hdl, sizeof (pool_entry_t))) == NULL) {
265 nvlist_free(config);
266 return (-1);
267 }
268 pe->pe_guid = pool_guid;
269 pe->pe_next = pl->pools;
270 pl->pools = pe;
271 }
272
273 /*
274 * Second, see if we know about this toplevel vdev. Add it if its
275 * missing.
276 */
277 for (ve = pe->pe_vdevs; ve != NULL; ve = ve->ve_next) {
278 if (ve->ve_guid == top_guid)
279 break;
280 }
281
282 if (ve == NULL) {
283 if ((ve = zfs_alloc(hdl, sizeof (vdev_entry_t))) == NULL) {
284 nvlist_free(config);
285 return (-1);
286 }
287 ve->ve_guid = top_guid;
288 ve->ve_next = pe->pe_vdevs;
289 pe->pe_vdevs = ve;
290 }
291
292 /*
293 * Third, see if we have a config with a matching transaction group. If
294 * so, then we do nothing. Otherwise, add it to the list of known
295 * configs.
296 */
297 for (ce = ve->ve_configs; ce != NULL; ce = ce->ce_next) {
298 if (ce->ce_txg == txg)
299 break;
300 }
301
302 if (ce == NULL) {
303 if ((ce = zfs_alloc(hdl, sizeof (config_entry_t))) == NULL) {
304 nvlist_free(config);
305 return (-1);
306 }
307 ce->ce_txg = txg;
308 ce->ce_config = config;
309 ce->ce_next = ve->ve_configs;
310 ve->ve_configs = ce;
311 } else {
312 nvlist_free(config);
313 }
314
315 /*
316 * At this point we've successfully added our config to the list of
317 * known configs. The last thing to do is add the vdev guid -> path
318 * mappings so that we can fix up the configuration as necessary before
319 * doing the import.
320 */
321 if ((ne = zfs_alloc(hdl, sizeof (name_entry_t))) == NULL)
322 return (-1);
323
324 if ((ne->ne_name = zfs_strdup(hdl, path)) == NULL) {
325 free(ne);
326 return (-1);
327 }
328
329 ne->ne_guid = vdev_guid;
44867b6d 330 ne->ne_order = order;
34dc7c2f
BB
331 ne->ne_next = pl->names;
332 pl->names = ne;
333
334 return (0);
335}
336
337/*
338 * Returns true if the named pool matches the given GUID.
339 */
340static int
341pool_active(libzfs_handle_t *hdl, const char *name, uint64_t guid,
342 boolean_t *isactive)
343{
344 zpool_handle_t *zhp;
345 uint64_t theguid;
346
347 if (zpool_open_silent(hdl, name, &zhp) != 0)
348 return (-1);
349
350 if (zhp == NULL) {
351 *isactive = B_FALSE;
352 return (0);
353 }
354
355 verify(nvlist_lookup_uint64(zhp->zpool_config, ZPOOL_CONFIG_POOL_GUID,
356 &theguid) == 0);
357
358 zpool_close(zhp);
359
360 *isactive = (theguid == guid);
361 return (0);
362}
363
364static nvlist_t *
365refresh_config(libzfs_handle_t *hdl, nvlist_t *config)
366{
367 nvlist_t *nvl;
2598c001 368 zfs_cmd_t zc = { "\0", "\0", "\0", "\0", 0 };
34dc7c2f
BB
369 int err;
370
371 if (zcmd_write_conf_nvlist(hdl, &zc, config) != 0)
372 return (NULL);
373
374 if (zcmd_alloc_dst_nvlist(hdl, &zc,
375 zc.zc_nvlist_conf_size * 2) != 0) {
376 zcmd_free_nvlists(&zc);
377 return (NULL);
378 }
379
380 while ((err = ioctl(hdl->libzfs_fd, ZFS_IOC_POOL_TRYIMPORT,
381 &zc)) != 0 && errno == ENOMEM) {
382 if (zcmd_expand_dst_nvlist(hdl, &zc) != 0) {
383 zcmd_free_nvlists(&zc);
384 return (NULL);
385 }
386 }
387
388 if (err) {
34dc7c2f
BB
389 zcmd_free_nvlists(&zc);
390 return (NULL);
391 }
392
393 if (zcmd_read_dst_nvlist(hdl, &zc, &nvl) != 0) {
394 zcmd_free_nvlists(&zc);
395 return (NULL);
396 }
397
398 zcmd_free_nvlists(&zc);
399 return (nvl);
400}
401
428870ff
BB
402/*
403 * Determine if the vdev id is a hole in the namespace.
404 */
405boolean_t
406vdev_is_hole(uint64_t *hole_array, uint_t holes, uint_t id)
407{
d6320ddb
BB
408 int c;
409
410 for (c = 0; c < holes; c++) {
428870ff
BB
411
412 /* Top-level is a hole */
413 if (hole_array[c] == id)
414 return (B_TRUE);
415 }
416 return (B_FALSE);
417}
418
34dc7c2f
BB
419/*
420 * Convert our list of pools into the definitive set of configurations. We
421 * start by picking the best config for each toplevel vdev. Once that's done,
422 * we assemble the toplevel vdevs into a full config for the pool. We make a
423 * pass to fix up any incorrect paths, and then add it to the main list to
424 * return to the user.
425 */
426static nvlist_t *
427get_configs(libzfs_handle_t *hdl, pool_list_t *pl, boolean_t active_ok)
428{
429 pool_entry_t *pe;
430 vdev_entry_t *ve;
431 config_entry_t *ce;
d4ed6673 432 nvlist_t *ret = NULL, *config = NULL, *tmp = NULL, *nvtop, *nvroot;
34dc7c2f
BB
433 nvlist_t **spares, **l2cache;
434 uint_t i, nspares, nl2cache;
435 boolean_t config_seen;
436 uint64_t best_txg;
d96eb2b1 437 char *name, *hostname, *comment;
34dc7c2f
BB
438 uint64_t version, guid;
439 uint_t children = 0;
440 nvlist_t **child = NULL;
428870ff
BB
441 uint_t holes;
442 uint64_t *hole_array, max_id;
34dc7c2f
BB
443 uint_t c;
444 boolean_t isactive;
445 uint64_t hostid;
446 nvlist_t *nvl;
b128c09f 447 boolean_t found_one = B_FALSE;
428870ff 448 boolean_t valid_top_config = B_FALSE;
34dc7c2f
BB
449
450 if (nvlist_alloc(&ret, 0, 0) != 0)
451 goto nomem;
452
453 for (pe = pl->pools; pe != NULL; pe = pe->pe_next) {
428870ff 454 uint64_t id, max_txg = 0;
34dc7c2f
BB
455
456 if (nvlist_alloc(&config, NV_UNIQUE_NAME, 0) != 0)
457 goto nomem;
458 config_seen = B_FALSE;
459
460 /*
461 * Iterate over all toplevel vdevs. Grab the pool configuration
462 * from the first one we find, and then go through the rest and
463 * add them as necessary to the 'vdevs' member of the config.
464 */
465 for (ve = pe->pe_vdevs; ve != NULL; ve = ve->ve_next) {
466
467 /*
468 * Determine the best configuration for this vdev by
469 * selecting the config with the latest transaction
470 * group.
471 */
472 best_txg = 0;
473 for (ce = ve->ve_configs; ce != NULL;
474 ce = ce->ce_next) {
475
476 if (ce->ce_txg > best_txg) {
477 tmp = ce->ce_config;
478 best_txg = ce->ce_txg;
479 }
480 }
481
428870ff
BB
482 /*
483 * We rely on the fact that the max txg for the
484 * pool will contain the most up-to-date information
485 * about the valid top-levels in the vdev namespace.
486 */
487 if (best_txg > max_txg) {
488 (void) nvlist_remove(config,
489 ZPOOL_CONFIG_VDEV_CHILDREN,
490 DATA_TYPE_UINT64);
491 (void) nvlist_remove(config,
492 ZPOOL_CONFIG_HOLE_ARRAY,
493 DATA_TYPE_UINT64_ARRAY);
494
495 max_txg = best_txg;
496 hole_array = NULL;
497 holes = 0;
498 max_id = 0;
499 valid_top_config = B_FALSE;
500
501 if (nvlist_lookup_uint64(tmp,
502 ZPOOL_CONFIG_VDEV_CHILDREN, &max_id) == 0) {
503 verify(nvlist_add_uint64(config,
504 ZPOOL_CONFIG_VDEV_CHILDREN,
505 max_id) == 0);
506 valid_top_config = B_TRUE;
507 }
508
509 if (nvlist_lookup_uint64_array(tmp,
510 ZPOOL_CONFIG_HOLE_ARRAY, &hole_array,
511 &holes) == 0) {
512 verify(nvlist_add_uint64_array(config,
513 ZPOOL_CONFIG_HOLE_ARRAY,
514 hole_array, holes) == 0);
515 }
516 }
517
34dc7c2f
BB
518 if (!config_seen) {
519 /*
520 * Copy the relevant pieces of data to the pool
521 * configuration:
522 *
523 * version
524 * pool guid
525 * name
d96eb2b1 526 * comment (if available)
34dc7c2f
BB
527 * pool state
528 * hostid (if available)
529 * hostname (if available)
530 */
531 uint64_t state;
532
533 verify(nvlist_lookup_uint64(tmp,
534 ZPOOL_CONFIG_VERSION, &version) == 0);
535 if (nvlist_add_uint64(config,
536 ZPOOL_CONFIG_VERSION, version) != 0)
537 goto nomem;
538 verify(nvlist_lookup_uint64(tmp,
539 ZPOOL_CONFIG_POOL_GUID, &guid) == 0);
540 if (nvlist_add_uint64(config,
541 ZPOOL_CONFIG_POOL_GUID, guid) != 0)
542 goto nomem;
543 verify(nvlist_lookup_string(tmp,
544 ZPOOL_CONFIG_POOL_NAME, &name) == 0);
545 if (nvlist_add_string(config,
546 ZPOOL_CONFIG_POOL_NAME, name) != 0)
547 goto nomem;
d96eb2b1
DM
548
549 /*
550 * COMMENT is optional, don't bail if it's not
551 * there, instead, set it to NULL.
552 */
553 if (nvlist_lookup_string(tmp,
554 ZPOOL_CONFIG_COMMENT, &comment) != 0)
555 comment = NULL;
556 else if (nvlist_add_string(config,
557 ZPOOL_CONFIG_COMMENT, comment) != 0)
558 goto nomem;
559
34dc7c2f
BB
560 verify(nvlist_lookup_uint64(tmp,
561 ZPOOL_CONFIG_POOL_STATE, &state) == 0);
562 if (nvlist_add_uint64(config,
563 ZPOOL_CONFIG_POOL_STATE, state) != 0)
564 goto nomem;
d96eb2b1 565
34dc7c2f
BB
566 hostid = 0;
567 if (nvlist_lookup_uint64(tmp,
568 ZPOOL_CONFIG_HOSTID, &hostid) == 0) {
569 if (nvlist_add_uint64(config,
570 ZPOOL_CONFIG_HOSTID, hostid) != 0)
571 goto nomem;
572 verify(nvlist_lookup_string(tmp,
573 ZPOOL_CONFIG_HOSTNAME,
574 &hostname) == 0);
575 if (nvlist_add_string(config,
576 ZPOOL_CONFIG_HOSTNAME,
577 hostname) != 0)
578 goto nomem;
579 }
580
581 config_seen = B_TRUE;
582 }
583
584 /*
585 * Add this top-level vdev to the child array.
586 */
587 verify(nvlist_lookup_nvlist(tmp,
588 ZPOOL_CONFIG_VDEV_TREE, &nvtop) == 0);
589 verify(nvlist_lookup_uint64(nvtop, ZPOOL_CONFIG_ID,
590 &id) == 0);
428870ff 591
34dc7c2f
BB
592 if (id >= children) {
593 nvlist_t **newchild;
594
595 newchild = zfs_alloc(hdl, (id + 1) *
596 sizeof (nvlist_t *));
597 if (newchild == NULL)
598 goto nomem;
599
600 for (c = 0; c < children; c++)
601 newchild[c] = child[c];
602
603 free(child);
604 child = newchild;
605 children = id + 1;
606 }
607 if (nvlist_dup(nvtop, &child[id], 0) != 0)
608 goto nomem;
609
610 }
611
428870ff
BB
612 /*
613 * If we have information about all the top-levels then
614 * clean up the nvlist which we've constructed. This
615 * means removing any extraneous devices that are
616 * beyond the valid range or adding devices to the end
617 * of our array which appear to be missing.
618 */
619 if (valid_top_config) {
620 if (max_id < children) {
621 for (c = max_id; c < children; c++)
622 nvlist_free(child[c]);
623 children = max_id;
624 } else if (max_id > children) {
625 nvlist_t **newchild;
626
627 newchild = zfs_alloc(hdl, (max_id) *
628 sizeof (nvlist_t *));
629 if (newchild == NULL)
630 goto nomem;
631
632 for (c = 0; c < children; c++)
633 newchild[c] = child[c];
634
635 free(child);
636 child = newchild;
637 children = max_id;
638 }
639 }
640
34dc7c2f
BB
641 verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID,
642 &guid) == 0);
643
428870ff
BB
644 /*
645 * The vdev namespace may contain holes as a result of
646 * device removal. We must add them back into the vdev
647 * tree before we process any missing devices.
648 */
649 if (holes > 0) {
650 ASSERT(valid_top_config);
651
652 for (c = 0; c < children; c++) {
653 nvlist_t *holey;
654
655 if (child[c] != NULL ||
656 !vdev_is_hole(hole_array, holes, c))
657 continue;
658
659 if (nvlist_alloc(&holey, NV_UNIQUE_NAME,
660 0) != 0)
661 goto nomem;
662
663 /*
664 * Holes in the namespace are treated as
665 * "hole" top-level vdevs and have a
666 * special flag set on them.
667 */
668 if (nvlist_add_string(holey,
669 ZPOOL_CONFIG_TYPE,
670 VDEV_TYPE_HOLE) != 0 ||
671 nvlist_add_uint64(holey,
672 ZPOOL_CONFIG_ID, c) != 0 ||
673 nvlist_add_uint64(holey,
674 ZPOOL_CONFIG_GUID, 0ULL) != 0)
675 goto nomem;
676 child[c] = holey;
677 }
678 }
679
34dc7c2f
BB
680 /*
681 * Look for any missing top-level vdevs. If this is the case,
682 * create a faked up 'missing' vdev as a placeholder. We cannot
683 * simply compress the child array, because the kernel performs
684 * certain checks to make sure the vdev IDs match their location
685 * in the configuration.
686 */
428870ff 687 for (c = 0; c < children; c++) {
34dc7c2f
BB
688 if (child[c] == NULL) {
689 nvlist_t *missing;
690 if (nvlist_alloc(&missing, NV_UNIQUE_NAME,
691 0) != 0)
692 goto nomem;
693 if (nvlist_add_string(missing,
694 ZPOOL_CONFIG_TYPE,
695 VDEV_TYPE_MISSING) != 0 ||
696 nvlist_add_uint64(missing,
697 ZPOOL_CONFIG_ID, c) != 0 ||
698 nvlist_add_uint64(missing,
699 ZPOOL_CONFIG_GUID, 0ULL) != 0) {
700 nvlist_free(missing);
701 goto nomem;
702 }
703 child[c] = missing;
704 }
428870ff 705 }
34dc7c2f
BB
706
707 /*
708 * Put all of this pool's top-level vdevs into a root vdev.
709 */
710 if (nvlist_alloc(&nvroot, NV_UNIQUE_NAME, 0) != 0)
711 goto nomem;
712 if (nvlist_add_string(nvroot, ZPOOL_CONFIG_TYPE,
713 VDEV_TYPE_ROOT) != 0 ||
714 nvlist_add_uint64(nvroot, ZPOOL_CONFIG_ID, 0ULL) != 0 ||
715 nvlist_add_uint64(nvroot, ZPOOL_CONFIG_GUID, guid) != 0 ||
716 nvlist_add_nvlist_array(nvroot, ZPOOL_CONFIG_CHILDREN,
717 child, children) != 0) {
718 nvlist_free(nvroot);
719 goto nomem;
720 }
721
722 for (c = 0; c < children; c++)
723 nvlist_free(child[c]);
724 free(child);
725 children = 0;
726 child = NULL;
727
728 /*
729 * Go through and fix up any paths and/or devids based on our
730 * known list of vdev GUID -> path mappings.
731 */
732 if (fix_paths(nvroot, pl->names) != 0) {
733 nvlist_free(nvroot);
734 goto nomem;
735 }
736
737 /*
738 * Add the root vdev to this pool's configuration.
739 */
740 if (nvlist_add_nvlist(config, ZPOOL_CONFIG_VDEV_TREE,
741 nvroot) != 0) {
742 nvlist_free(nvroot);
743 goto nomem;
744 }
745 nvlist_free(nvroot);
746
747 /*
748 * zdb uses this path to report on active pools that were
749 * imported or created using -R.
750 */
751 if (active_ok)
752 goto add_pool;
753
754 /*
755 * Determine if this pool is currently active, in which case we
756 * can't actually import it.
757 */
758 verify(nvlist_lookup_string(config, ZPOOL_CONFIG_POOL_NAME,
759 &name) == 0);
760 verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID,
761 &guid) == 0);
762
763 if (pool_active(hdl, name, guid, &isactive) != 0)
764 goto error;
765
766 if (isactive) {
767 nvlist_free(config);
768 config = NULL;
769 continue;
770 }
771
428870ff
BB
772 if ((nvl = refresh_config(hdl, config)) == NULL) {
773 nvlist_free(config);
774 config = NULL;
775 continue;
776 }
34dc7c2f
BB
777
778 nvlist_free(config);
779 config = nvl;
780
781 /*
782 * Go through and update the paths for spares, now that we have
783 * them.
784 */
785 verify(nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE,
786 &nvroot) == 0);
787 if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_SPARES,
788 &spares, &nspares) == 0) {
789 for (i = 0; i < nspares; i++) {
790 if (fix_paths(spares[i], pl->names) != 0)
791 goto nomem;
792 }
793 }
794
795 /*
796 * Update the paths for l2cache devices.
797 */
798 if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_L2CACHE,
799 &l2cache, &nl2cache) == 0) {
800 for (i = 0; i < nl2cache; i++) {
801 if (fix_paths(l2cache[i], pl->names) != 0)
802 goto nomem;
803 }
804 }
805
806 /*
807 * Restore the original information read from the actual label.
808 */
809 (void) nvlist_remove(config, ZPOOL_CONFIG_HOSTID,
810 DATA_TYPE_UINT64);
811 (void) nvlist_remove(config, ZPOOL_CONFIG_HOSTNAME,
812 DATA_TYPE_STRING);
813 if (hostid != 0) {
814 verify(nvlist_add_uint64(config, ZPOOL_CONFIG_HOSTID,
815 hostid) == 0);
816 verify(nvlist_add_string(config, ZPOOL_CONFIG_HOSTNAME,
817 hostname) == 0);
818 }
819
820add_pool:
821 /*
822 * Add this pool to the list of configs.
823 */
824 verify(nvlist_lookup_string(config, ZPOOL_CONFIG_POOL_NAME,
825 &name) == 0);
826 if (nvlist_add_nvlist(ret, name, config) != 0)
827 goto nomem;
828
b128c09f 829 found_one = B_TRUE;
34dc7c2f
BB
830 nvlist_free(config);
831 config = NULL;
832 }
833
b128c09f
BB
834 if (!found_one) {
835 nvlist_free(ret);
836 ret = NULL;
837 }
838
34dc7c2f
BB
839 return (ret);
840
841nomem:
842 (void) no_memory(hdl);
843error:
844 nvlist_free(config);
845 nvlist_free(ret);
846 for (c = 0; c < children; c++)
847 nvlist_free(child[c]);
848 free(child);
849
850 return (NULL);
851}
852
853/*
854 * Return the offset of the given label.
855 */
856static uint64_t
857label_offset(uint64_t size, int l)
858{
859 ASSERT(P2PHASE_TYPED(size, sizeof (vdev_label_t), uint64_t) == 0);
860 return (l * sizeof (vdev_label_t) + (l < VDEV_LABELS / 2 ?
861 0 : size - VDEV_LABELS * sizeof (vdev_label_t)));
862}
863
864/*
865 * Given a file descriptor, read the label information and return an nvlist
866 * describing the configuration, if there is one.
867 */
868int
869zpool_read_label(int fd, nvlist_t **config)
870{
871 struct stat64 statbuf;
872 int l;
873 vdev_label_t *label;
874 uint64_t state, txg, size;
875
876 *config = NULL;
877
878 if (fstat64(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++) {
b128c09f 886 if (pread64(fd, label, sizeof (vdev_label_t),
34dc7c2f
BB
887 label_offset(size, l)) != sizeof (vdev_label_t))
888 continue;
889
890 if (nvlist_unpack(label->vl_vdev_phys.vp_nvlist,
891 sizeof (label->vl_vdev_phys.vp_nvlist), config, 0) != 0)
892 continue;
893
894 if (nvlist_lookup_uint64(*config, ZPOOL_CONFIG_POOL_STATE,
895 &state) != 0 || state > POOL_STATE_L2CACHE) {
896 nvlist_free(*config);
897 continue;
898 }
899
900 if (state != POOL_STATE_SPARE && state != POOL_STATE_L2CACHE &&
901 (nvlist_lookup_uint64(*config, ZPOOL_CONFIG_POOL_TXG,
902 &txg) != 0 || txg == 0)) {
903 nvlist_free(*config);
904 continue;
905 }
906
907 free(label);
908 return (0);
909 }
910
911 free(label);
912 *config = NULL;
913 return (0);
914}
915
d603ed6c
BB
916#ifdef HAVE_LIBBLKID
917/*
918 * Use libblkid to quickly search for zfs devices
919 */
428870ff 920static int
d603ed6c 921zpool_find_import_blkid(libzfs_handle_t *hdl, pool_list_t *pools)
428870ff 922{
d603ed6c
BB
923 blkid_cache cache;
924 blkid_dev_iterate iter;
925 blkid_dev dev;
926 const char *devname;
428870ff 927 nvlist_t *config;
d603ed6c 928 int fd, err;
428870ff 929
d603ed6c
BB
930 err = blkid_get_cache(&cache, NULL);
931 if (err != 0) {
932 (void) zfs_error_fmt(hdl, EZFS_BADCACHE,
933 dgettext(TEXT_DOMAIN, "blkid_get_cache() %d"), err);
934 goto err_blkid1;
428870ff
BB
935 }
936
d603ed6c
BB
937 err = blkid_probe_all(cache);
938 if (err != 0) {
939 (void) zfs_error_fmt(hdl, EZFS_BADCACHE,
940 dgettext(TEXT_DOMAIN, "blkid_probe_all() %d"), err);
941 goto err_blkid2;
428870ff 942 }
428870ff 943
d603ed6c
BB
944 iter = blkid_dev_iterate_begin(cache);
945 if (iter == NULL) {
946 (void) zfs_error_fmt(hdl, EZFS_BADCACHE,
947 dgettext(TEXT_DOMAIN, "blkid_dev_iterate_begin()"));
948 goto err_blkid2;
949 }
428870ff 950
d603ed6c
BB
951 err = blkid_dev_set_search(iter, "TYPE", "zfs");
952 if (err != 0) {
953 (void) zfs_error_fmt(hdl, EZFS_BADCACHE,
954 dgettext(TEXT_DOMAIN, "blkid_dev_set_search() %d"), err);
955 goto err_blkid3;
428870ff 956 }
428870ff 957
d603ed6c
BB
958 while (blkid_dev_next(iter, &dev) == 0) {
959 devname = blkid_dev_devname(dev);
960 if ((fd = open64(devname, O_RDONLY)) < 0)
961 continue;
428870ff 962
d603ed6c
BB
963 err = zpool_read_label(fd, &config);
964 (void) close(fd);
428870ff 965
d603ed6c
BB
966 if (err != 0) {
967 (void) no_memory(hdl);
968 goto err_blkid3;
969 }
428870ff 970
d603ed6c 971 if (config != NULL) {
44867b6d 972 err = add_config(hdl, pools, devname, 0, config);
d603ed6c
BB
973 if (err != 0)
974 goto err_blkid3;
975 }
428870ff
BB
976 }
977
d603ed6c
BB
978err_blkid3:
979 blkid_dev_iterate_end(iter);
980err_blkid2:
981 blkid_put_cache(cache);
982err_blkid1:
983 return err;
428870ff 984}
d603ed6c 985#endif /* HAVE_LIBBLKID */
428870ff 986
44867b6d
BB
987#define DEFAULT_IMPORT_PATH_SIZE 8
988
989static char *
990zpool_default_import_path[DEFAULT_IMPORT_PATH_SIZE] = {
991 "/dev/disk/by-vdev", /* Custom rules, use first if they exist */
992 "/dev/disk/zpool", /* Custom rules, use first if they exist */
993 "/dev/mapper", /* Use multipath devices before components */
994 "/dev/disk/by-uuid", /* Single unique entry and persistent */
995 "/dev/disk/by-id", /* May be multiple entries and persistent */
996 "/dev/disk/by-path", /* Encodes physical location and persistent */
997 "/dev/disk/by-label", /* Custom persistent labels */
998 "/dev" /* UNSAFE device names will change */
999};
1000
34dc7c2f
BB
1001/*
1002 * Given a list of directories to search, find all pools stored on disk. This
1003 * includes partial pools which are not available to import. If no args are
1004 * given (argc is 0), then the default directory (/dev/dsk) is searched.
b128c09f
BB
1005 * poolname or guid (but not both) are provided by the caller when trying
1006 * to import a specific pool.
34dc7c2f 1007 */
b128c09f 1008static nvlist_t *
428870ff 1009zpool_find_import_impl(libzfs_handle_t *hdl, importargs_t *iarg)
34dc7c2f 1010{
428870ff 1011 int i, dirs = iarg->paths;
34dc7c2f
BB
1012 DIR *dirp = NULL;
1013 struct dirent64 *dp;
1014 char path[MAXPATHLEN];
428870ff 1015 char *end, **dir = iarg->path;
34dc7c2f 1016 size_t pathleft;
d603ed6c
BB
1017 struct stat64 statbuf;
1018 nvlist_t *ret = NULL, *config;
d603ed6c 1019 int fd;
34dc7c2f
BB
1020 pool_list_t pools = { 0 };
1021 pool_entry_t *pe, *penext;
1022 vdev_entry_t *ve, *venext;
1023 config_entry_t *ce, *cenext;
1024 name_entry_t *ne, *nenext;
d603ed6c
BB
1025
1026 verify(iarg->poolname == NULL || iarg->guid == 0);
34dc7c2f 1027
428870ff 1028 if (dirs == 0) {
d603ed6c
BB
1029#ifdef HAVE_LIBBLKID
1030 /* Use libblkid to scan all device for their type */
1031 if (zpool_find_import_blkid(hdl, &pools) == 0)
1032 goto skip_scanning;
1033
1034 (void) zfs_error_fmt(hdl, EZFS_BADCACHE,
1035 dgettext(TEXT_DOMAIN, "blkid failure falling back "
1036 "to manual probing"));
1037#endif /* HAVE_LIBBLKID */
44867b6d
BB
1038
1039 dir = zpool_default_import_path;
1040 dirs = DEFAULT_IMPORT_PATH_SIZE;
34dc7c2f
BB
1041 }
1042
1043 /*
1044 * Go through and read the label configuration information from every
1045 * possible device, organizing the information according to pool GUID
1046 * and toplevel GUID.
1047 */
428870ff 1048 for (i = 0; i < dirs; i++) {
34dc7c2f
BB
1049 char *rdsk;
1050 int dfd;
1051
1052 /* use realpath to normalize the path */
428870ff 1053 if (realpath(dir[i], path) == 0) {
44867b6d
BB
1054
1055 /* it is safe to skip missing search paths */
1056 if (errno == ENOENT)
1057 continue;
1058
1059 zfs_error_aux(hdl, strerror(errno));
34dc7c2f 1060 (void) zfs_error_fmt(hdl, EZFS_BADPATH,
428870ff 1061 dgettext(TEXT_DOMAIN, "cannot open '%s'"), dir[i]);
34dc7c2f
BB
1062 goto error;
1063 }
1064 end = &path[strlen(path)];
1065 *end++ = '/';
1066 *end = 0;
1067 pathleft = &path[sizeof (path)] - end;
1068
1069 /*
1070 * Using raw devices instead of block devices when we're
1071 * reading the labels skips a bunch of slow operations during
1072 * close(2) processing, so we replace /dev/dsk with /dev/rdsk.
1073 */
1074 if (strcmp(path, "/dev/dsk/") == 0)
1075 rdsk = "/dev/rdsk/";
1076 else
1077 rdsk = path;
1078
1079 if ((dfd = open64(rdsk, O_RDONLY)) < 0 ||
1080 (dirp = fdopendir(dfd)) == NULL) {
1081 zfs_error_aux(hdl, strerror(errno));
1082 (void) zfs_error_fmt(hdl, EZFS_BADPATH,
1083 dgettext(TEXT_DOMAIN, "cannot open '%s'"),
1084 rdsk);
1085 goto error;
1086 }
1087
1088 /*
1089 * This is not MT-safe, but we have no MT consumers of libzfs
1090 */
1091 while ((dp = readdir64(dirp)) != NULL) {
1092 const char *name = dp->d_name;
1093 if (name[0] == '.' &&
1094 (name[1] == 0 || (name[1] == '.' && name[2] == 0)))
1095 continue;
1096
d603ed6c
BB
1097 /*
1098 * Skip checking devices with well known prefixes:
1099 * watchdog - A special close is required to avoid
1100 * triggering it and resetting the system.
1101 * fuse - Fuse control device.
1102 * ppp - Generic PPP driver.
1103 * tty* - Generic serial interface.
1104 * vcs* - Virtual console memory.
1105 * parport* - Parallel port interface.
1106 * lp* - Printer interface.
1107 * fd* - Floppy interface.
f03e41e8
BB
1108 * hpet - High Precision Event Timer, crashes qemu
1109 * when accessed from a virtual machine.
1110 * core - Symlink to /proc/kcore, causes a crash
1111 * when access from Xen dom0.
d603ed6c
BB
1112 */
1113 if ((strncmp(name, "watchdog", 8) == 0) ||
1114 (strncmp(name, "fuse", 4) == 0) ||
1115 (strncmp(name, "ppp", 3) == 0) ||
1116 (strncmp(name, "tty", 3) == 0) ||
1117 (strncmp(name, "vcs", 3) == 0) ||
1118 (strncmp(name, "parport", 7) == 0) ||
1119 (strncmp(name, "lp", 2) == 0) ||
5e7affae 1120 (strncmp(name, "fd", 2) == 0) ||
f03e41e8
BB
1121 (strncmp(name, "hpet", 4) == 0) ||
1122 (strncmp(name, "core", 4) == 0))
d603ed6c
BB
1123 continue;
1124
1125 /*
1126 * Ignore failed stats. We only want regular
f03e41e8 1127 * files and block devices.
d603ed6c 1128 */
f03e41e8 1129 if ((fstatat64(dfd, name, &statbuf, 0) != 0) ||
d603ed6c 1130 (!S_ISREG(statbuf.st_mode) &&
f03e41e8
BB
1131 !S_ISBLK(statbuf.st_mode)))
1132 continue;
1133
1134 if ((fd = openat64(dfd, name, O_RDONLY)) < 0)
d603ed6c 1135 continue;
d603ed6c
BB
1136
1137 if ((zpool_read_label(fd, &config)) != 0) {
1138 (void) close(fd);
1139 (void) no_memory(hdl);
1140 goto error;
1141 }
1142
1143 (void) close(fd);
1144
1145 if (config != NULL) {
b128c09f
BB
1146 boolean_t matched = B_TRUE;
1147
428870ff 1148 if (iarg->poolname != NULL) {
b128c09f
BB
1149 char *pname;
1150
1151 matched = nvlist_lookup_string(config,
1152 ZPOOL_CONFIG_POOL_NAME,
1153 &pname) == 0 &&
428870ff
BB
1154 strcmp(iarg->poolname, pname) == 0;
1155 } else if (iarg->guid != 0) {
b128c09f
BB
1156 uint64_t this_guid;
1157
1158 matched = nvlist_lookup_uint64(config,
1159 ZPOOL_CONFIG_POOL_GUID,
1160 &this_guid) == 0 &&
428870ff 1161 iarg->guid == this_guid;
b128c09f
BB
1162 }
1163 if (!matched) {
1164 nvlist_free(config);
1165 config = NULL;
1166 continue;
1167 }
34dc7c2f 1168 /* use the non-raw path for the config */
d603ed6c 1169 (void) strlcpy(end, name, pathleft);
44867b6d 1170 if (add_config(hdl, &pools, path, i+1, config))
34dc7c2f
BB
1171 goto error;
1172 }
1173 }
1174
1175 (void) closedir(dirp);
1176 dirp = NULL;
1177 }
1178
d603ed6c
BB
1179#ifdef HAVE_LIBBLKID
1180skip_scanning:
1181#endif
428870ff 1182 ret = get_configs(hdl, &pools, iarg->can_be_active);
34dc7c2f
BB
1183
1184error:
1185 for (pe = pools.pools; pe != NULL; pe = penext) {
1186 penext = pe->pe_next;
1187 for (ve = pe->pe_vdevs; ve != NULL; ve = venext) {
1188 venext = ve->ve_next;
1189 for (ce = ve->ve_configs; ce != NULL; ce = cenext) {
1190 cenext = ce->ce_next;
1191 if (ce->ce_config)
1192 nvlist_free(ce->ce_config);
1193 free(ce);
1194 }
1195 free(ve);
1196 }
1197 free(pe);
1198 }
1199
1200 for (ne = pools.names; ne != NULL; ne = nenext) {
1201 nenext = ne->ne_next;
1202 if (ne->ne_name)
1203 free(ne->ne_name);
1204 free(ne);
1205 }
1206
1207 if (dirp)
1208 (void) closedir(dirp);
1209
1210 return (ret);
1211}
1212
b128c09f
BB
1213nvlist_t *
1214zpool_find_import(libzfs_handle_t *hdl, int argc, char **argv)
1215{
428870ff 1216 importargs_t iarg = { 0 };
b128c09f 1217
428870ff
BB
1218 iarg.paths = argc;
1219 iarg.path = argv;
b128c09f 1220
428870ff 1221 return (zpool_find_import_impl(hdl, &iarg));
b128c09f
BB
1222}
1223
34dc7c2f
BB
1224/*
1225 * Given a cache file, return the contents as a list of importable pools.
b128c09f
BB
1226 * poolname or guid (but not both) are provided by the caller when trying
1227 * to import a specific pool.
34dc7c2f
BB
1228 */
1229nvlist_t *
1230zpool_find_import_cached(libzfs_handle_t *hdl, const char *cachefile,
b128c09f 1231 char *poolname, uint64_t guid)
34dc7c2f
BB
1232{
1233 char *buf;
1234 int fd;
1235 struct stat64 statbuf;
1236 nvlist_t *raw, *src, *dst;
1237 nvlist_t *pools;
1238 nvpair_t *elem;
1239 char *name;
b128c09f 1240 uint64_t this_guid;
34dc7c2f
BB
1241 boolean_t active;
1242
b128c09f
BB
1243 verify(poolname == NULL || guid == 0);
1244
34dc7c2f
BB
1245 if ((fd = open(cachefile, O_RDONLY)) < 0) {
1246 zfs_error_aux(hdl, "%s", strerror(errno));
1247 (void) zfs_error(hdl, EZFS_BADCACHE,
1248 dgettext(TEXT_DOMAIN, "failed to open cache file"));
1249 return (NULL);
1250 }
1251
1252 if (fstat64(fd, &statbuf) != 0) {
1253 zfs_error_aux(hdl, "%s", strerror(errno));
1254 (void) close(fd);
1255 (void) zfs_error(hdl, EZFS_BADCACHE,
1256 dgettext(TEXT_DOMAIN, "failed to get size of cache file"));
1257 return (NULL);
1258 }
1259
1260 if ((buf = zfs_alloc(hdl, statbuf.st_size)) == NULL) {
1261 (void) close(fd);
1262 return (NULL);
1263 }
1264
1265 if (read(fd, buf, statbuf.st_size) != statbuf.st_size) {
1266 (void) close(fd);
1267 free(buf);
1268 (void) zfs_error(hdl, EZFS_BADCACHE,
1269 dgettext(TEXT_DOMAIN,
1270 "failed to read cache file contents"));
1271 return (NULL);
1272 }
1273
1274 (void) close(fd);
1275
1276 if (nvlist_unpack(buf, statbuf.st_size, &raw, 0) != 0) {
1277 free(buf);
1278 (void) zfs_error(hdl, EZFS_BADCACHE,
1279 dgettext(TEXT_DOMAIN,
1280 "invalid or corrupt cache file contents"));
1281 return (NULL);
1282 }
1283
1284 free(buf);
1285
1286 /*
1287 * Go through and get the current state of the pools and refresh their
1288 * state.
1289 */
1290 if (nvlist_alloc(&pools, 0, 0) != 0) {
1291 (void) no_memory(hdl);
1292 nvlist_free(raw);
1293 return (NULL);
1294 }
1295
1296 elem = NULL;
1297 while ((elem = nvlist_next_nvpair(raw, elem)) != NULL) {
1298 verify(nvpair_value_nvlist(elem, &src) == 0);
1299
1300 verify(nvlist_lookup_string(src, ZPOOL_CONFIG_POOL_NAME,
1301 &name) == 0);
b128c09f
BB
1302 if (poolname != NULL && strcmp(poolname, name) != 0)
1303 continue;
1304
34dc7c2f 1305 verify(nvlist_lookup_uint64(src, ZPOOL_CONFIG_POOL_GUID,
b128c09f
BB
1306 &this_guid) == 0);
1307 if (guid != 0) {
1308 verify(nvlist_lookup_uint64(src, ZPOOL_CONFIG_POOL_GUID,
1309 &this_guid) == 0);
1310 if (guid != this_guid)
1311 continue;
1312 }
34dc7c2f 1313
b128c09f
BB
1314 if (pool_active(hdl, name, this_guid, &active) != 0) {
1315 nvlist_free(raw);
1316 nvlist_free(pools);
1317 return (NULL);
1318 }
34dc7c2f 1319
b128c09f
BB
1320 if (active)
1321 continue;
34dc7c2f 1322
b128c09f
BB
1323 if ((dst = refresh_config(hdl, src)) == NULL) {
1324 nvlist_free(raw);
1325 nvlist_free(pools);
1326 return (NULL);
1327 }
34dc7c2f 1328
b128c09f
BB
1329 if (nvlist_add_nvlist(pools, nvpair_name(elem), dst) != 0) {
1330 (void) no_memory(hdl);
34dc7c2f 1331 nvlist_free(dst);
b128c09f
BB
1332 nvlist_free(raw);
1333 nvlist_free(pools);
1334 return (NULL);
34dc7c2f 1335 }
b128c09f 1336 nvlist_free(dst);
34dc7c2f
BB
1337 }
1338
1339 nvlist_free(raw);
1340 return (pools);
1341}
1342
428870ff
BB
1343static int
1344name_or_guid_exists(zpool_handle_t *zhp, void *data)
1345{
1346 importargs_t *import = data;
1347 int found = 0;
1348
1349 if (import->poolname != NULL) {
1350 char *pool_name;
1351
1352 verify(nvlist_lookup_string(zhp->zpool_config,
1353 ZPOOL_CONFIG_POOL_NAME, &pool_name) == 0);
1354 if (strcmp(pool_name, import->poolname) == 0)
1355 found = 1;
1356 } else {
1357 uint64_t pool_guid;
1358
1359 verify(nvlist_lookup_uint64(zhp->zpool_config,
1360 ZPOOL_CONFIG_POOL_GUID, &pool_guid) == 0);
1361 if (pool_guid == import->guid)
1362 found = 1;
1363 }
1364
1365 zpool_close(zhp);
1366 return (found);
1367}
1368
1369nvlist_t *
1370zpool_search_import(libzfs_handle_t *hdl, importargs_t *import)
1371{
1372 verify(import->poolname == NULL || import->guid == 0);
1373
1374 if (import->unique)
1375 import->exists = zpool_iter(hdl, name_or_guid_exists, import);
1376
1377 if (import->cachefile != NULL)
1378 return (zpool_find_import_cached(hdl, import->cachefile,
1379 import->poolname, import->guid));
1380
1381 return (zpool_find_import_impl(hdl, import));
1382}
34dc7c2f
BB
1383
1384boolean_t
1385find_guid(nvlist_t *nv, uint64_t guid)
1386{
1387 uint64_t tmp;
1388 nvlist_t **child;
1389 uint_t c, children;
1390
1391 verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID, &tmp) == 0);
1392 if (tmp == guid)
1393 return (B_TRUE);
1394
1395 if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN,
1396 &child, &children) == 0) {
1397 for (c = 0; c < children; c++)
1398 if (find_guid(child[c], guid))
1399 return (B_TRUE);
1400 }
1401
1402 return (B_FALSE);
1403}
1404
1405typedef struct aux_cbdata {
1406 const char *cb_type;
1407 uint64_t cb_guid;
1408 zpool_handle_t *cb_zhp;
1409} aux_cbdata_t;
1410
1411static int
1412find_aux(zpool_handle_t *zhp, void *data)
1413{
1414 aux_cbdata_t *cbp = data;
1415 nvlist_t **list;
1416 uint_t i, count;
1417 uint64_t guid;
1418 nvlist_t *nvroot;
1419
1420 verify(nvlist_lookup_nvlist(zhp->zpool_config, ZPOOL_CONFIG_VDEV_TREE,
1421 &nvroot) == 0);
1422
1423 if (nvlist_lookup_nvlist_array(nvroot, cbp->cb_type,
1424 &list, &count) == 0) {
1425 for (i = 0; i < count; i++) {
1426 verify(nvlist_lookup_uint64(list[i],
1427 ZPOOL_CONFIG_GUID, &guid) == 0);
1428 if (guid == cbp->cb_guid) {
1429 cbp->cb_zhp = zhp;
1430 return (1);
1431 }
1432 }
1433 }
1434
1435 zpool_close(zhp);
1436 return (0);
1437}
1438
1439/*
1440 * Determines if the pool is in use. If so, it returns true and the state of
1441 * the pool as well as the name of the pool. Both strings are allocated and
1442 * must be freed by the caller.
1443 */
1444int
1445zpool_in_use(libzfs_handle_t *hdl, int fd, pool_state_t *state, char **namestr,
1446 boolean_t *inuse)
1447{
1448 nvlist_t *config;
1449 char *name;
1450 boolean_t ret;
1451 uint64_t guid, vdev_guid;
1452 zpool_handle_t *zhp;
1453 nvlist_t *pool_config;
1454 uint64_t stateval, isspare;
1455 aux_cbdata_t cb = { 0 };
1456 boolean_t isactive;
1457
1458 *inuse = B_FALSE;
1459
1460 if (zpool_read_label(fd, &config) != 0) {
1461 (void) no_memory(hdl);
1462 return (-1);
1463 }
1464
1465 if (config == NULL)
1466 return (0);
1467
1468 verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_STATE,
1469 &stateval) == 0);
1470 verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_GUID,
1471 &vdev_guid) == 0);
1472
1473 if (stateval != POOL_STATE_SPARE && stateval != POOL_STATE_L2CACHE) {
1474 verify(nvlist_lookup_string(config, ZPOOL_CONFIG_POOL_NAME,
1475 &name) == 0);
1476 verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID,
1477 &guid) == 0);
1478 }
1479
1480 switch (stateval) {
1481 case POOL_STATE_EXPORTED:
572e2857
BB
1482 /*
1483 * A pool with an exported state may in fact be imported
1484 * read-only, so check the in-core state to see if it's
1485 * active and imported read-only. If it is, set
1486 * its state to active.
1487 */
1488 if (pool_active(hdl, name, guid, &isactive) == 0 && isactive &&
1489 (zhp = zpool_open_canfail(hdl, name)) != NULL &&
1490 zpool_get_prop_int(zhp, ZPOOL_PROP_READONLY, NULL))
1491 stateval = POOL_STATE_ACTIVE;
1492
34dc7c2f
BB
1493 ret = B_TRUE;
1494 break;
1495
1496 case POOL_STATE_ACTIVE:
1497 /*
1498 * For an active pool, we have to determine if it's really part
1499 * of a currently active pool (in which case the pool will exist
1500 * and the guid will be the same), or whether it's part of an
1501 * active pool that was disconnected without being explicitly
1502 * exported.
1503 */
1504 if (pool_active(hdl, name, guid, &isactive) != 0) {
1505 nvlist_free(config);
1506 return (-1);
1507 }
1508
1509 if (isactive) {
1510 /*
1511 * Because the device may have been removed while
1512 * offlined, we only report it as active if the vdev is
1513 * still present in the config. Otherwise, pretend like
1514 * it's not in use.
1515 */
1516 if ((zhp = zpool_open_canfail(hdl, name)) != NULL &&
1517 (pool_config = zpool_get_config(zhp, NULL))
1518 != NULL) {
1519 nvlist_t *nvroot;
1520
1521 verify(nvlist_lookup_nvlist(pool_config,
1522 ZPOOL_CONFIG_VDEV_TREE, &nvroot) == 0);
1523 ret = find_guid(nvroot, vdev_guid);
1524 } else {
1525 ret = B_FALSE;
1526 }
1527
1528 /*
1529 * If this is an active spare within another pool, we
1530 * treat it like an unused hot spare. This allows the
1531 * user to create a pool with a hot spare that currently
1532 * in use within another pool. Since we return B_TRUE,
1533 * libdiskmgt will continue to prevent generic consumers
1534 * from using the device.
1535 */
1536 if (ret && nvlist_lookup_uint64(config,
1537 ZPOOL_CONFIG_IS_SPARE, &isspare) == 0 && isspare)
1538 stateval = POOL_STATE_SPARE;
1539
1540 if (zhp != NULL)
1541 zpool_close(zhp);
1542 } else {
1543 stateval = POOL_STATE_POTENTIALLY_ACTIVE;
1544 ret = B_TRUE;
1545 }
1546 break;
1547
1548 case POOL_STATE_SPARE:
1549 /*
1550 * For a hot spare, it can be either definitively in use, or
1551 * potentially active. To determine if it's in use, we iterate
1552 * over all pools in the system and search for one with a spare
1553 * with a matching guid.
1554 *
1555 * Due to the shared nature of spares, we don't actually report
1556 * the potentially active case as in use. This means the user
1557 * can freely create pools on the hot spares of exported pools,
1558 * but to do otherwise makes the resulting code complicated, and
1559 * we end up having to deal with this case anyway.
1560 */
1561 cb.cb_zhp = NULL;
1562 cb.cb_guid = vdev_guid;
1563 cb.cb_type = ZPOOL_CONFIG_SPARES;
1564 if (zpool_iter(hdl, find_aux, &cb) == 1) {
1565 name = (char *)zpool_get_name(cb.cb_zhp);
1566 ret = TRUE;
1567 } else {
1568 ret = FALSE;
1569 }
1570 break;
1571
1572 case POOL_STATE_L2CACHE:
1573
1574 /*
1575 * Check if any pool is currently using this l2cache device.
1576 */
1577 cb.cb_zhp = NULL;
1578 cb.cb_guid = vdev_guid;
1579 cb.cb_type = ZPOOL_CONFIG_L2CACHE;
1580 if (zpool_iter(hdl, find_aux, &cb) == 1) {
1581 name = (char *)zpool_get_name(cb.cb_zhp);
1582 ret = TRUE;
1583 } else {
1584 ret = FALSE;
1585 }
1586 break;
1587
1588 default:
1589 ret = B_FALSE;
1590 }
1591
1592
1593 if (ret) {
1594 if ((*namestr = zfs_strdup(hdl, name)) == NULL) {
1595 if (cb.cb_zhp)
1596 zpool_close(cb.cb_zhp);
1597 nvlist_free(config);
1598 return (-1);
1599 }
1600 *state = (pool_state_t)stateval;
1601 }
1602
1603 if (cb.cb_zhp)
1604 zpool_close(cb.cb_zhp);
1605
1606 nvlist_free(config);
1607 *inuse = ret;
1608 return (0);
1609}