]>
Commit | Line | Data |
---|---|---|
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. |
3bc7e0fb | 23 | * Copyright (c) 2012 by Delphix. All rights reserved. |
02f8fe42 | 24 | * Copyright 2014 Nexenta Systems, Inc. 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 | */ | |
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; | |
44867b6d | 90 | uint64_t ne_order; |
7d90f569 | 91 | uint64_t ne_num_labels; |
34dc7c2f BB |
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; | |
34dc7c2f BB |
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 | |
44867b6d BB |
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. | |
34dc7c2f BB |
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 | ||
34dc7c2f BB |
162 | best = NULL; |
163 | for (ne = names; ne != NULL; ne = ne->ne_next) { | |
164 | if (ne->ne_guid == guid) { | |
34dc7c2f BB |
165 | |
166 | if (path == NULL) { | |
167 | best = ne; | |
168 | break; | |
169 | } | |
170 | ||
44867b6d | 171 | if ((strlen(path) == strlen(ne->ne_name)) && |
d1d7e268 | 172 | strncmp(path, ne->ne_name, strlen(path)) == 0) { |
34dc7c2f | 173 | best = ne; |
44867b6d | 174 | break; |
34dc7c2f | 175 | } |
44867b6d | 176 | |
7d90f569 | 177 | if (best == NULL) { |
44867b6d | 178 | best = ne; |
7d90f569 BB |
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 | } | |
34dc7c2f BB |
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, | |
7d90f569 | 219 | int order, int num_labels, nvlist_t *config) |
34dc7c2f BB |
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; | |
44867b6d | 244 | ne->ne_order = order; |
7d90f569 | 245 | ne->ne_num_labels = num_labels; |
34dc7c2f BB |
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; | |
44867b6d | 347 | ne->ne_order = order; |
7d90f569 | 348 | ne->ne_num_labels = num_labels; |
34dc7c2f BB |
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; | |
13fe0198 | 386 | zfs_cmd_t zc = {"\0"}; |
34dc7c2f BB |
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) { | |
34dc7c2f BB |
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 | ||
428870ff BB |
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 | { | |
d6320ddb BB |
426 | int c; |
427 | ||
428 | for (c = 0; c < holes; c++) { | |
428870ff BB |
429 | |
430 | /* Top-level is a hole */ | |
431 | if (hole_array[c] == id) | |
432 | return (B_TRUE); | |
433 | } | |
434 | return (B_FALSE); | |
435 | } | |
436 | ||
34dc7c2f BB |
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; | |
d4ed6673 | 450 | nvlist_t *ret = NULL, *config = NULL, *tmp = NULL, *nvtop, *nvroot; |
34dc7c2f BB |
451 | nvlist_t **spares, **l2cache; |
452 | uint_t i, nspares, nl2cache; | |
453 | boolean_t config_seen; | |
454 | uint64_t best_txg; | |
3bc7e0fb GW |
455 | char *name, *hostname = NULL; |
456 | uint64_t guid; | |
34dc7c2f BB |
457 | uint_t children = 0; |
458 | nvlist_t **child = NULL; | |
428870ff BB |
459 | uint_t holes; |
460 | uint64_t *hole_array, max_id; | |
34dc7c2f BB |
461 | uint_t c; |
462 | boolean_t isactive; | |
463 | uint64_t hostid; | |
464 | nvlist_t *nvl; | |
428870ff | 465 | boolean_t valid_top_config = B_FALSE; |
34dc7c2f BB |
466 | |
467 | if (nvlist_alloc(&ret, 0, 0) != 0) | |
468 | goto nomem; | |
469 | ||
470 | for (pe = pl->pools; pe != NULL; pe = pe->pe_next) { | |
428870ff | 471 | uint64_t id, max_txg = 0; |
34dc7c2f BB |
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 | ||
428870ff BB |
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 | ||
34dc7c2f BB |
535 | if (!config_seen) { |
536 | /* | |
537 | * Copy the relevant pieces of data to the pool | |
538 | * configuration: | |
539 | * | |
540 | * version | |
3bc7e0fb GW |
541 | * pool guid |
542 | * name | |
d96eb2b1 | 543 | * comment (if available) |
3bc7e0fb | 544 | * pool state |
34dc7c2f BB |
545 | * hostid (if available) |
546 | * hostname (if available) | |
547 | */ | |
295304be | 548 | uint64_t state, version; |
3bc7e0fb GW |
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); | |
34dc7c2f | 563 | |
d96eb2b1 | 564 | if (nvlist_lookup_string(tmp, |
3bc7e0fb GW |
565 | ZPOOL_CONFIG_COMMENT, &comment) == 0) |
566 | fnvlist_add_string(config, | |
567 | ZPOOL_CONFIG_COMMENT, comment); | |
d96eb2b1 | 568 | |
3bc7e0fb GW |
569 | state = fnvlist_lookup_uint64(tmp, |
570 | ZPOOL_CONFIG_POOL_STATE); | |
571 | fnvlist_add_uint64(config, | |
572 | ZPOOL_CONFIG_POOL_STATE, state); | |
d96eb2b1 | 573 | |
34dc7c2f BB |
574 | hostid = 0; |
575 | if (nvlist_lookup_uint64(tmp, | |
576 | ZPOOL_CONFIG_HOSTID, &hostid) == 0) { | |
3bc7e0fb GW |
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); | |
34dc7c2f BB |
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); | |
428870ff | 595 | |
34dc7c2f BB |
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 | ||
428870ff BB |
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 | ||
34dc7c2f BB |
645 | verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID, |
646 | &guid) == 0); | |
647 | ||
428870ff BB |
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 | ||
34dc7c2f BB |
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 | */ | |
428870ff | 691 | for (c = 0; c < children; c++) { |
34dc7c2f BB |
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 | } | |
428870ff | 709 | } |
34dc7c2f BB |
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 | ||
428870ff BB |
776 | if ((nvl = refresh_config(hdl, config)) == NULL) { |
777 | nvlist_free(config); | |
778 | config = NULL; | |
779 | continue; | |
780 | } | |
34dc7c2f BB |
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 | |
7d90f569 BB |
864 | * describing the configuration, if there is one. The number of valid |
865 | * labels found will be returned in num_labels when non-NULL. | |
34dc7c2f BB |
866 | */ |
867 | int | |
7d90f569 | 868 | zpool_read_label(int fd, nvlist_t **config, int *num_labels) |
34dc7c2f BB |
869 | { |
870 | struct stat64 statbuf; | |
7d90f569 | 871 | int l, count = 0; |
34dc7c2f | 872 | vdev_label_t *label; |
7d90f569 BB |
873 | nvlist_t *expected_config = NULL; |
874 | uint64_t expected_guid = 0, size; | |
34dc7c2f BB |
875 | |
876 | *config = NULL; | |
877 | ||
ff3510c1 | 878 | if (fstat64_blk(fd, &statbuf) == -1) |
34dc7c2f BB |
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++) { | |
7d90f569 BB |
886 | uint64_t state, guid, txg; |
887 | ||
b128c09f | 888 | if (pread64(fd, label, sizeof (vdev_label_t), |
34dc7c2f BB |
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 | ||
7d90f569 BB |
896 | if (nvlist_lookup_uint64(*config, ZPOOL_CONFIG_GUID, |
897 | &guid) != 0 || guid == 0) { | |
898 | nvlist_free(*config); | |
899 | continue; | |
900 | } | |
901 | ||
34dc7c2f BB |
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 | ||
7d90f569 BB |
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 | } | |
34dc7c2f BB |
925 | } |
926 | ||
7d90f569 BB |
927 | if (num_labels != NULL) |
928 | *num_labels = count; | |
929 | ||
34dc7c2f | 930 | free(label); |
7d90f569 BB |
931 | *config = expected_config; |
932 | ||
34dc7c2f BB |
933 | return (0); |
934 | } | |
935 | ||
519129ff BB |
936 | typedef struct rdsk_node { |
937 | char *rn_name; | |
938 | int rn_num_labels; | |
939 | int rn_dfd; | |
940 | libzfs_handle_t *rn_hdl; | |
941 | nvlist_t *rn_config; | |
942 | avl_tree_t *rn_avl; | |
943 | avl_node_t rn_node; | |
944 | boolean_t rn_nozpool; | |
945 | } rdsk_node_t; | |
946 | ||
947 | static int | |
948 | slice_cache_compare(const void *arg1, const void *arg2) | |
949 | { | |
950 | const char *nm1 = ((rdsk_node_t *)arg1)->rn_name; | |
951 | const char *nm2 = ((rdsk_node_t *)arg2)->rn_name; | |
952 | char *nm1slice, *nm2slice; | |
953 | int rv; | |
954 | ||
955 | /* | |
956 | * partitions one and three (slices zero and two) are the most | |
957 | * likely to provide results, so put those first | |
958 | */ | |
959 | nm1slice = strstr(nm1, "part1"); | |
960 | nm2slice = strstr(nm2, "part1"); | |
961 | if (nm1slice && !nm2slice) { | |
962 | return (-1); | |
963 | } | |
964 | if (!nm1slice && nm2slice) { | |
965 | return (1); | |
966 | } | |
967 | nm1slice = strstr(nm1, "part3"); | |
968 | nm2slice = strstr(nm2, "part3"); | |
969 | if (nm1slice && !nm2slice) { | |
970 | return (-1); | |
971 | } | |
972 | if (!nm1slice && nm2slice) { | |
973 | return (1); | |
974 | } | |
975 | ||
976 | rv = strcmp(nm1, nm2); | |
977 | if (rv == 0) | |
978 | return (0); | |
979 | return (rv > 0 ? 1 : -1); | |
980 | } | |
981 | ||
982 | #ifndef __linux__ | |
983 | static void | |
984 | check_one_slice(avl_tree_t *r, char *diskname, uint_t partno, | |
985 | diskaddr_t size, uint_t blksz) | |
986 | { | |
987 | rdsk_node_t tmpnode; | |
988 | rdsk_node_t *node; | |
989 | char sname[MAXNAMELEN]; | |
990 | ||
991 | tmpnode.rn_name = &sname[0]; | |
992 | (void) snprintf(tmpnode.rn_name, MAXNAMELEN, "%s%u", | |
993 | diskname, partno); | |
994 | /* too small to contain a zpool? */ | |
995 | if ((size < (SPA_MINDEVSIZE / blksz)) && | |
996 | (node = avl_find(r, &tmpnode, NULL))) | |
997 | node->rn_nozpool = B_TRUE; | |
998 | } | |
999 | #endif | |
1000 | ||
1001 | static void | |
1002 | nozpool_all_slices(avl_tree_t *r, const char *sname) | |
1003 | { | |
1004 | #ifndef __linux__ | |
1005 | char diskname[MAXNAMELEN]; | |
1006 | char *ptr; | |
1007 | int i; | |
1008 | ||
1009 | (void) strncpy(diskname, sname, MAXNAMELEN); | |
1010 | if (((ptr = strrchr(diskname, 's')) == NULL) && | |
1011 | ((ptr = strrchr(diskname, 'p')) == NULL)) | |
1012 | return; | |
1013 | ptr[0] = 's'; | |
1014 | ptr[1] = '\0'; | |
1015 | for (i = 0; i < NDKMAP; i++) | |
1016 | check_one_slice(r, diskname, i, 0, 1); | |
1017 | ptr[0] = 'p'; | |
1018 | for (i = 0; i <= FD_NUMPART; i++) | |
1019 | check_one_slice(r, diskname, i, 0, 1); | |
1020 | #endif | |
1021 | } | |
1022 | ||
1023 | static void | |
1024 | check_slices(avl_tree_t *r, int fd, const char *sname) | |
1025 | { | |
1026 | #ifndef __linux__ | |
1027 | struct extvtoc vtoc; | |
1028 | struct dk_gpt *gpt; | |
1029 | char diskname[MAXNAMELEN]; | |
1030 | char *ptr; | |
1031 | int i; | |
1032 | ||
1033 | (void) strncpy(diskname, sname, MAXNAMELEN); | |
1034 | if ((ptr = strrchr(diskname, 's')) == NULL || !isdigit(ptr[1])) | |
1035 | return; | |
1036 | ptr[1] = '\0'; | |
1037 | ||
1038 | if (read_extvtoc(fd, &vtoc) >= 0) { | |
1039 | for (i = 0; i < NDKMAP; i++) | |
1040 | check_one_slice(r, diskname, i, | |
1041 | vtoc.v_part[i].p_size, vtoc.v_sectorsz); | |
1042 | } else if (efi_alloc_and_read(fd, &gpt) >= 0) { | |
1043 | /* | |
1044 | * on x86 we'll still have leftover links that point | |
1045 | * to slices s[9-15], so use NDKMAP instead | |
1046 | */ | |
1047 | for (i = 0; i < NDKMAP; i++) | |
1048 | check_one_slice(r, diskname, i, | |
1049 | gpt->efi_parts[i].p_size, gpt->efi_lbasize); | |
1050 | /* nodes p[1-4] are never used with EFI labels */ | |
1051 | ptr[0] = 'p'; | |
1052 | for (i = 1; i <= FD_NUMPART; i++) | |
1053 | check_one_slice(r, diskname, i, 0, 1); | |
1054 | efi_free(gpt); | |
1055 | } | |
1056 | #endif | |
1057 | } | |
1058 | ||
1059 | static void | |
1060 | zpool_open_func(void *arg) | |
1061 | { | |
1062 | rdsk_node_t *rn = arg; | |
1063 | struct stat64 statbuf; | |
1064 | nvlist_t *config; | |
1065 | int num_labels; | |
1066 | int fd; | |
1067 | ||
1068 | if (rn->rn_nozpool) | |
1069 | return; | |
1070 | #ifdef __linux__ | |
1071 | /* | |
1072 | * Skip devices with well known prefixes there can be side effects | |
1073 | * when opening devices which need to be avoided. | |
1074 | * | |
1075 | * core - Symlink to /proc/kcore | |
1076 | * fd* - Floppy interface. | |
1077 | * fuse - Fuse control device. | |
1078 | * hpet - High Precision Event Timer | |
1079 | * lp* - Printer interface. | |
1080 | * parport* - Parallel port interface. | |
1081 | * ppp - Generic PPP driver. | |
1082 | * random - Random device | |
1083 | * rtc - Real Time Clock | |
1084 | * tty* - Generic serial interface. | |
1085 | * urandom - Random device. | |
1086 | * usbmon* - USB IO monitor. | |
1087 | * vcs* - Virtual console memory. | |
1088 | * watchdog - Watchdog must be closed in a special way. | |
1089 | */ | |
1090 | if ((strncmp(rn->rn_name, "core", 4) == 0) || | |
1091 | (strncmp(rn->rn_name, "fd", 2) == 0) || | |
1092 | (strncmp(rn->rn_name, "fuse", 4) == 0) || | |
1093 | (strncmp(rn->rn_name, "hpet", 4) == 0) || | |
1094 | (strncmp(rn->rn_name, "lp", 2) == 0) || | |
1095 | (strncmp(rn->rn_name, "parport", 7) == 0) || | |
1096 | (strncmp(rn->rn_name, "ppp", 3) == 0) || | |
1097 | (strncmp(rn->rn_name, "random", 6) == 0) || | |
1098 | (strncmp(rn->rn_name, "rtc", 3) == 0) || | |
1099 | (strncmp(rn->rn_name, "tty", 3) == 0) || | |
1100 | (strncmp(rn->rn_name, "urandom", 7) == 0) || | |
1101 | (strncmp(rn->rn_name, "usbmon", 6) == 0) || | |
1102 | (strncmp(rn->rn_name, "vcs", 3) == 0) || | |
1103 | (strncmp(rn->rn_name, "watchdog", 8) == 0)) | |
1104 | return; | |
1105 | ||
1106 | /* | |
1107 | * Ignore failed stats. We only want regular files and block devices. | |
1108 | */ | |
1109 | if (fstatat64(rn->rn_dfd, rn->rn_name, &statbuf, 0) != 0 || | |
1110 | (!S_ISREG(statbuf.st_mode) && !S_ISBLK(statbuf.st_mode))) | |
1111 | return; | |
1112 | ||
1113 | if ((fd = openat64(rn->rn_dfd, rn->rn_name, O_RDONLY)) < 0) { | |
1114 | /* symlink to a device that's no longer there */ | |
1115 | if (errno == ENOENT) | |
1116 | nozpool_all_slices(rn->rn_avl, rn->rn_name); | |
1117 | return; | |
1118 | } | |
1119 | #else | |
1120 | if ((fd = openat64(rn->rn_dfd, rn->rn_name, O_RDONLY)) < 0) { | |
1121 | /* symlink to a device that's no longer there */ | |
1122 | if (errno == ENOENT) | |
1123 | nozpool_all_slices(rn->rn_avl, rn->rn_name); | |
1124 | return; | |
1125 | } | |
1126 | /* | |
1127 | * Ignore failed stats. We only want regular | |
1128 | * files, character devs and block devs. | |
1129 | */ | |
1130 | if (fstat64(fd, &statbuf) != 0 || | |
1131 | (!S_ISREG(statbuf.st_mode) && | |
1132 | !S_ISCHR(statbuf.st_mode) && | |
1133 | !S_ISBLK(statbuf.st_mode))) { | |
1134 | (void) close(fd); | |
1135 | return; | |
1136 | } | |
1137 | #endif | |
1138 | /* this file is too small to hold a zpool */ | |
1139 | if (S_ISREG(statbuf.st_mode) && | |
1140 | statbuf.st_size < SPA_MINDEVSIZE) { | |
1141 | (void) close(fd); | |
1142 | return; | |
1143 | } else if (!S_ISREG(statbuf.st_mode)) { | |
1144 | /* | |
1145 | * Try to read the disk label first so we don't have to | |
1146 | * open a bunch of minor nodes that can't have a zpool. | |
1147 | */ | |
1148 | check_slices(rn->rn_avl, fd, rn->rn_name); | |
1149 | } | |
1150 | ||
1151 | if ((zpool_read_label(fd, &config, &num_labels)) != 0) { | |
1152 | (void) close(fd); | |
1153 | (void) no_memory(rn->rn_hdl); | |
1154 | return; | |
1155 | } | |
1156 | ||
1157 | if (num_labels == 0) { | |
1158 | (void) close(fd); | |
1159 | nvlist_free(config); | |
1160 | return; | |
1161 | } | |
1162 | ||
1163 | (void) close(fd); | |
1164 | ||
1165 | rn->rn_config = config; | |
1166 | rn->rn_num_labels = num_labels; | |
1167 | if (config != NULL) { | |
1168 | assert(rn->rn_nozpool == B_FALSE); | |
1169 | } | |
1170 | } | |
1171 | ||
51a3ae72 DK |
1172 | /* |
1173 | * Given a file descriptor, clear (zero) the label information. This function | |
131cc95c DK |
1174 | * is used in the appliance stack as part of the ZFS sysevent module and |
1175 | * to implement the "zpool labelclear" command. | |
51a3ae72 DK |
1176 | */ |
1177 | int | |
1178 | zpool_clear_label(int fd) | |
1179 | { | |
1180 | struct stat64 statbuf; | |
1181 | int l; | |
1182 | vdev_label_t *label; | |
1183 | uint64_t size; | |
1184 | ||
1185 | if (fstat64_blk(fd, &statbuf) == -1) | |
1186 | return (0); | |
1187 | size = P2ALIGN_TYPED(statbuf.st_size, sizeof (vdev_label_t), uint64_t); | |
1188 | ||
1189 | if ((label = calloc(sizeof (vdev_label_t), 1)) == NULL) | |
1190 | return (-1); | |
1191 | ||
1192 | for (l = 0; l < VDEV_LABELS; l++) { | |
1193 | if (pwrite64(fd, label, sizeof (vdev_label_t), | |
4def05f8 RY |
1194 | label_offset(size, l)) != sizeof (vdev_label_t)) { |
1195 | free(label); | |
51a3ae72 | 1196 | return (-1); |
4def05f8 | 1197 | } |
51a3ae72 DK |
1198 | } |
1199 | ||
1200 | free(label); | |
1201 | return (0); | |
1202 | } | |
1203 | ||
d603ed6c BB |
1204 | #ifdef HAVE_LIBBLKID |
1205 | /* | |
1206 | * Use libblkid to quickly search for zfs devices | |
1207 | */ | |
428870ff | 1208 | static int |
d603ed6c | 1209 | zpool_find_import_blkid(libzfs_handle_t *hdl, pool_list_t *pools) |
428870ff | 1210 | { |
d603ed6c BB |
1211 | blkid_cache cache; |
1212 | blkid_dev_iterate iter; | |
1213 | blkid_dev dev; | |
1214 | const char *devname; | |
428870ff | 1215 | nvlist_t *config; |
7d90f569 | 1216 | int fd, err, num_labels; |
428870ff | 1217 | |
d603ed6c BB |
1218 | err = blkid_get_cache(&cache, NULL); |
1219 | if (err != 0) { | |
1220 | (void) zfs_error_fmt(hdl, EZFS_BADCACHE, | |
1221 | dgettext(TEXT_DOMAIN, "blkid_get_cache() %d"), err); | |
1222 | goto err_blkid1; | |
428870ff BB |
1223 | } |
1224 | ||
d603ed6c BB |
1225 | err = blkid_probe_all(cache); |
1226 | if (err != 0) { | |
1227 | (void) zfs_error_fmt(hdl, EZFS_BADCACHE, | |
1228 | dgettext(TEXT_DOMAIN, "blkid_probe_all() %d"), err); | |
1229 | goto err_blkid2; | |
428870ff | 1230 | } |
428870ff | 1231 | |
d603ed6c BB |
1232 | iter = blkid_dev_iterate_begin(cache); |
1233 | if (iter == NULL) { | |
1234 | (void) zfs_error_fmt(hdl, EZFS_BADCACHE, | |
1235 | dgettext(TEXT_DOMAIN, "blkid_dev_iterate_begin()")); | |
1236 | goto err_blkid2; | |
1237 | } | |
428870ff | 1238 | |
1db7b9be | 1239 | err = blkid_dev_set_search(iter, "TYPE", "zfs_member"); |
d603ed6c BB |
1240 | if (err != 0) { |
1241 | (void) zfs_error_fmt(hdl, EZFS_BADCACHE, | |
1242 | dgettext(TEXT_DOMAIN, "blkid_dev_set_search() %d"), err); | |
1243 | goto err_blkid3; | |
428870ff | 1244 | } |
428870ff | 1245 | |
d603ed6c BB |
1246 | while (blkid_dev_next(iter, &dev) == 0) { |
1247 | devname = blkid_dev_devname(dev); | |
1248 | if ((fd = open64(devname, O_RDONLY)) < 0) | |
1249 | continue; | |
428870ff | 1250 | |
7d90f569 | 1251 | err = zpool_read_label(fd, &config, &num_labels); |
d603ed6c | 1252 | (void) close(fd); |
428870ff | 1253 | |
d603ed6c BB |
1254 | if (err != 0) { |
1255 | (void) no_memory(hdl); | |
1256 | goto err_blkid3; | |
1257 | } | |
428870ff | 1258 | |
d603ed6c | 1259 | if (config != NULL) { |
7d90f569 BB |
1260 | err = add_config(hdl, pools, devname, 0, |
1261 | num_labels, config); | |
d603ed6c BB |
1262 | if (err != 0) |
1263 | goto err_blkid3; | |
1264 | } | |
428870ff BB |
1265 | } |
1266 | ||
d603ed6c BB |
1267 | err_blkid3: |
1268 | blkid_dev_iterate_end(iter); | |
1269 | err_blkid2: | |
1270 | blkid_put_cache(cache); | |
1271 | err_blkid1: | |
d1d7e268 | 1272 | return (err); |
428870ff | 1273 | } |
d603ed6c | 1274 | #endif /* HAVE_LIBBLKID */ |
428870ff | 1275 | |
eac47204 | 1276 | char * |
44867b6d BB |
1277 | zpool_default_import_path[DEFAULT_IMPORT_PATH_SIZE] = { |
1278 | "/dev/disk/by-vdev", /* Custom rules, use first if they exist */ | |
44867b6d BB |
1279 | "/dev/mapper", /* Use multipath devices before components */ |
1280 | "/dev/disk/by-uuid", /* Single unique entry and persistent */ | |
1281 | "/dev/disk/by-id", /* May be multiple entries and persistent */ | |
1282 | "/dev/disk/by-path", /* Encodes physical location and persistent */ | |
1283 | "/dev/disk/by-label", /* Custom persistent labels */ | |
1284 | "/dev" /* UNSAFE device names will change */ | |
1285 | }; | |
1286 | ||
34dc7c2f BB |
1287 | /* |
1288 | * Given a list of directories to search, find all pools stored on disk. This | |
1289 | * includes partial pools which are not available to import. If no args are | |
1290 | * given (argc is 0), then the default directory (/dev/dsk) is searched. | |
b128c09f BB |
1291 | * poolname or guid (but not both) are provided by the caller when trying |
1292 | * to import a specific pool. | |
34dc7c2f | 1293 | */ |
b128c09f | 1294 | static nvlist_t * |
428870ff | 1295 | zpool_find_import_impl(libzfs_handle_t *hdl, importargs_t *iarg) |
34dc7c2f | 1296 | { |
519129ff | 1297 | int i, dirs = iarg->paths; |
34dc7c2f BB |
1298 | DIR *dirp = NULL; |
1299 | struct dirent64 *dp; | |
1300 | char path[MAXPATHLEN]; | |
428870ff | 1301 | char *end, **dir = iarg->path; |
34dc7c2f | 1302 | size_t pathleft; |
519129ff | 1303 | nvlist_t *ret = NULL; |
34dc7c2f BB |
1304 | pool_list_t pools = { 0 }; |
1305 | pool_entry_t *pe, *penext; | |
1306 | vdev_entry_t *ve, *venext; | |
1307 | config_entry_t *ce, *cenext; | |
1308 | name_entry_t *ne, *nenext; | |
519129ff BB |
1309 | avl_tree_t slice_cache; |
1310 | rdsk_node_t *slice; | |
1311 | void *cookie; | |
d603ed6c BB |
1312 | |
1313 | verify(iarg->poolname == NULL || iarg->guid == 0); | |
34dc7c2f | 1314 | |
428870ff | 1315 | if (dirs == 0) { |
d603ed6c BB |
1316 | #ifdef HAVE_LIBBLKID |
1317 | /* Use libblkid to scan all device for their type */ | |
1318 | if (zpool_find_import_blkid(hdl, &pools) == 0) | |
1319 | goto skip_scanning; | |
1320 | ||
1321 | (void) zfs_error_fmt(hdl, EZFS_BADCACHE, | |
1322 | dgettext(TEXT_DOMAIN, "blkid failure falling back " | |
1323 | "to manual probing")); | |
1324 | #endif /* HAVE_LIBBLKID */ | |
44867b6d BB |
1325 | |
1326 | dir = zpool_default_import_path; | |
1327 | dirs = DEFAULT_IMPORT_PATH_SIZE; | |
34dc7c2f BB |
1328 | } |
1329 | ||
1330 | /* | |
1331 | * Go through and read the label configuration information from every | |
1332 | * possible device, organizing the information according to pool GUID | |
1333 | * and toplevel GUID. | |
1334 | */ | |
428870ff | 1335 | for (i = 0; i < dirs; i++) { |
519129ff | 1336 | taskq_t *t; |
34dc7c2f BB |
1337 | char *rdsk; |
1338 | int dfd; | |
1339 | ||
1340 | /* use realpath to normalize the path */ | |
428870ff | 1341 | if (realpath(dir[i], path) == 0) { |
44867b6d BB |
1342 | |
1343 | /* it is safe to skip missing search paths */ | |
1344 | if (errno == ENOENT) | |
1345 | continue; | |
1346 | ||
1347 | zfs_error_aux(hdl, strerror(errno)); | |
34dc7c2f | 1348 | (void) zfs_error_fmt(hdl, EZFS_BADPATH, |
428870ff | 1349 | dgettext(TEXT_DOMAIN, "cannot open '%s'"), dir[i]); |
34dc7c2f BB |
1350 | goto error; |
1351 | } | |
1352 | end = &path[strlen(path)]; | |
1353 | *end++ = '/'; | |
1354 | *end = 0; | |
1355 | pathleft = &path[sizeof (path)] - end; | |
1356 | ||
1357 | /* | |
1358 | * Using raw devices instead of block devices when we're | |
1359 | * reading the labels skips a bunch of slow operations during | |
1360 | * close(2) processing, so we replace /dev/dsk with /dev/rdsk. | |
1361 | */ | |
1362 | if (strcmp(path, "/dev/dsk/") == 0) | |
1363 | rdsk = "/dev/rdsk/"; | |
1364 | else | |
1365 | rdsk = path; | |
1366 | ||
1367 | if ((dfd = open64(rdsk, O_RDONLY)) < 0 || | |
1368 | (dirp = fdopendir(dfd)) == NULL) { | |
1369 | zfs_error_aux(hdl, strerror(errno)); | |
1370 | (void) zfs_error_fmt(hdl, EZFS_BADPATH, | |
1371 | dgettext(TEXT_DOMAIN, "cannot open '%s'"), | |
1372 | rdsk); | |
1373 | goto error; | |
1374 | } | |
1375 | ||
519129ff BB |
1376 | avl_create(&slice_cache, slice_cache_compare, |
1377 | sizeof (rdsk_node_t), offsetof(rdsk_node_t, rn_node)); | |
1378 | ||
34dc7c2f BB |
1379 | /* |
1380 | * This is not MT-safe, but we have no MT consumers of libzfs | |
1381 | */ | |
1382 | while ((dp = readdir64(dirp)) != NULL) { | |
1383 | const char *name = dp->d_name; | |
1384 | if (name[0] == '.' && | |
1385 | (name[1] == 0 || (name[1] == '.' && name[2] == 0))) | |
1386 | continue; | |
1387 | ||
519129ff BB |
1388 | slice = zfs_alloc(hdl, sizeof (rdsk_node_t)); |
1389 | slice->rn_name = zfs_strdup(hdl, name); | |
1390 | slice->rn_avl = &slice_cache; | |
1391 | slice->rn_dfd = dfd; | |
1392 | slice->rn_hdl = hdl; | |
1393 | slice->rn_nozpool = B_FALSE; | |
1394 | avl_add(&slice_cache, slice); | |
1395 | } | |
1396 | /* | |
1397 | * create a thread pool to do all of this in parallel; | |
1398 | * rn_nozpool is not protected, so this is racy in that | |
1399 | * multiple tasks could decide that the same slice can | |
1400 | * not hold a zpool, which is benign. Also choose | |
1401 | * double the number of processors; we hold a lot of | |
1402 | * locks in the kernel, so going beyond this doesn't | |
1403 | * buy us much. | |
1404 | */ | |
1405 | thread_init(); | |
1406 | t = taskq_create("z_import", 2 * boot_ncpus, defclsyspri, | |
1407 | 2 * boot_ncpus, INT_MAX, TASKQ_PREPOPULATE); | |
1408 | for (slice = avl_first(&slice_cache); slice; | |
1409 | (slice = avl_walk(&slice_cache, slice, | |
1410 | AVL_AFTER))) | |
1411 | (void) taskq_dispatch(t, zpool_open_func, slice, | |
1412 | TQ_SLEEP); | |
1413 | taskq_wait(t); | |
1414 | taskq_destroy(t); | |
1415 | thread_fini(); | |
1416 | ||
1417 | cookie = NULL; | |
1418 | while ((slice = avl_destroy_nodes(&slice_cache, | |
1419 | &cookie)) != NULL) { | |
1420 | if (slice->rn_config != NULL) { | |
1421 | nvlist_t *config = slice->rn_config; | |
b128c09f | 1422 | boolean_t matched = B_TRUE; |
30b937ee | 1423 | |
519129ff BB |
1424 | if (iarg->poolname != NULL) { |
1425 | char *pname; | |
b128c09f | 1426 | |
519129ff BB |
1427 | matched = nvlist_lookup_string(config, |
1428 | ZPOOL_CONFIG_POOL_NAME, | |
1429 | &pname) == 0 && | |
1430 | strcmp(iarg->poolname, pname) == 0; | |
428870ff | 1431 | } else if (iarg->guid != 0) { |
b128c09f BB |
1432 | uint64_t this_guid; |
1433 | ||
1434 | matched = nvlist_lookup_uint64(config, | |
1435 | ZPOOL_CONFIG_POOL_GUID, | |
1436 | &this_guid) == 0 && | |
428870ff | 1437 | iarg->guid == this_guid; |
b128c09f BB |
1438 | } |
1439 | if (!matched) { | |
1440 | nvlist_free(config); | |
1441 | config = NULL; | |
1442 | continue; | |
1443 | } | |
34dc7c2f | 1444 | /* use the non-raw path for the config */ |
519129ff | 1445 | (void) strlcpy(end, slice->rn_name, pathleft); |
7d90f569 | 1446 | if (add_config(hdl, &pools, path, i+1, |
519129ff | 1447 | slice->rn_num_labels, config)) |
34dc7c2f BB |
1448 | goto error; |
1449 | } | |
519129ff BB |
1450 | free(slice->rn_name); |
1451 | free(slice); | |
34dc7c2f | 1452 | } |
519129ff | 1453 | avl_destroy(&slice_cache); |
34dc7c2f BB |
1454 | |
1455 | (void) closedir(dirp); | |
1456 | dirp = NULL; | |
1457 | } | |
1458 | ||
d603ed6c BB |
1459 | #ifdef HAVE_LIBBLKID |
1460 | skip_scanning: | |
1461 | #endif | |
428870ff | 1462 | ret = get_configs(hdl, &pools, iarg->can_be_active); |
34dc7c2f BB |
1463 | |
1464 | error: | |
1465 | for (pe = pools.pools; pe != NULL; pe = penext) { | |
1466 | penext = pe->pe_next; | |
1467 | for (ve = pe->pe_vdevs; ve != NULL; ve = venext) { | |
1468 | venext = ve->ve_next; | |
1469 | for (ce = ve->ve_configs; ce != NULL; ce = cenext) { | |
1470 | cenext = ce->ce_next; | |
1471 | if (ce->ce_config) | |
1472 | nvlist_free(ce->ce_config); | |
1473 | free(ce); | |
1474 | } | |
1475 | free(ve); | |
1476 | } | |
1477 | free(pe); | |
1478 | } | |
1479 | ||
1480 | for (ne = pools.names; ne != NULL; ne = nenext) { | |
1481 | nenext = ne->ne_next; | |
1482 | if (ne->ne_name) | |
1483 | free(ne->ne_name); | |
1484 | free(ne); | |
1485 | } | |
1486 | ||
1487 | if (dirp) | |
1488 | (void) closedir(dirp); | |
1489 | ||
1490 | return (ret); | |
1491 | } | |
1492 | ||
b128c09f BB |
1493 | nvlist_t * |
1494 | zpool_find_import(libzfs_handle_t *hdl, int argc, char **argv) | |
1495 | { | |
428870ff | 1496 | importargs_t iarg = { 0 }; |
b128c09f | 1497 | |
428870ff BB |
1498 | iarg.paths = argc; |
1499 | iarg.path = argv; | |
b128c09f | 1500 | |
428870ff | 1501 | return (zpool_find_import_impl(hdl, &iarg)); |
b128c09f BB |
1502 | } |
1503 | ||
34dc7c2f BB |
1504 | /* |
1505 | * Given a cache file, return the contents as a list of importable pools. | |
b128c09f BB |
1506 | * poolname or guid (but not both) are provided by the caller when trying |
1507 | * to import a specific pool. | |
34dc7c2f BB |
1508 | */ |
1509 | nvlist_t * | |
1510 | zpool_find_import_cached(libzfs_handle_t *hdl, const char *cachefile, | |
b128c09f | 1511 | char *poolname, uint64_t guid) |
34dc7c2f BB |
1512 | { |
1513 | char *buf; | |
1514 | int fd; | |
1515 | struct stat64 statbuf; | |
1516 | nvlist_t *raw, *src, *dst; | |
1517 | nvlist_t *pools; | |
1518 | nvpair_t *elem; | |
1519 | char *name; | |
b128c09f | 1520 | uint64_t this_guid; |
34dc7c2f BB |
1521 | boolean_t active; |
1522 | ||
b128c09f BB |
1523 | verify(poolname == NULL || guid == 0); |
1524 | ||
34dc7c2f BB |
1525 | if ((fd = open(cachefile, O_RDONLY)) < 0) { |
1526 | zfs_error_aux(hdl, "%s", strerror(errno)); | |
1527 | (void) zfs_error(hdl, EZFS_BADCACHE, | |
1528 | dgettext(TEXT_DOMAIN, "failed to open cache file")); | |
1529 | return (NULL); | |
1530 | } | |
1531 | ||
1532 | if (fstat64(fd, &statbuf) != 0) { | |
1533 | zfs_error_aux(hdl, "%s", strerror(errno)); | |
1534 | (void) close(fd); | |
1535 | (void) zfs_error(hdl, EZFS_BADCACHE, | |
1536 | dgettext(TEXT_DOMAIN, "failed to get size of cache file")); | |
1537 | return (NULL); | |
1538 | } | |
1539 | ||
1540 | if ((buf = zfs_alloc(hdl, statbuf.st_size)) == NULL) { | |
1541 | (void) close(fd); | |
1542 | return (NULL); | |
1543 | } | |
1544 | ||
1545 | if (read(fd, buf, statbuf.st_size) != statbuf.st_size) { | |
1546 | (void) close(fd); | |
1547 | free(buf); | |
1548 | (void) zfs_error(hdl, EZFS_BADCACHE, | |
1549 | dgettext(TEXT_DOMAIN, | |
1550 | "failed to read cache file contents")); | |
1551 | return (NULL); | |
1552 | } | |
1553 | ||
1554 | (void) close(fd); | |
1555 | ||
1556 | if (nvlist_unpack(buf, statbuf.st_size, &raw, 0) != 0) { | |
1557 | free(buf); | |
1558 | (void) zfs_error(hdl, EZFS_BADCACHE, | |
1559 | dgettext(TEXT_DOMAIN, | |
1560 | "invalid or corrupt cache file contents")); | |
1561 | return (NULL); | |
1562 | } | |
1563 | ||
1564 | free(buf); | |
1565 | ||
1566 | /* | |
1567 | * Go through and get the current state of the pools and refresh their | |
1568 | * state. | |
1569 | */ | |
1570 | if (nvlist_alloc(&pools, 0, 0) != 0) { | |
1571 | (void) no_memory(hdl); | |
1572 | nvlist_free(raw); | |
1573 | return (NULL); | |
1574 | } | |
1575 | ||
1576 | elem = NULL; | |
1577 | while ((elem = nvlist_next_nvpair(raw, elem)) != NULL) { | |
ab2894e6 | 1578 | src = fnvpair_value_nvlist(elem); |
34dc7c2f | 1579 | |
ab2894e6 | 1580 | name = fnvlist_lookup_string(src, ZPOOL_CONFIG_POOL_NAME); |
b128c09f BB |
1581 | if (poolname != NULL && strcmp(poolname, name) != 0) |
1582 | continue; | |
1583 | ||
ab2894e6 MA |
1584 | this_guid = fnvlist_lookup_uint64(src, ZPOOL_CONFIG_POOL_GUID); |
1585 | if (guid != 0 && guid != this_guid) | |
1586 | continue; | |
34dc7c2f | 1587 | |
b128c09f BB |
1588 | if (pool_active(hdl, name, this_guid, &active) != 0) { |
1589 | nvlist_free(raw); | |
1590 | nvlist_free(pools); | |
1591 | return (NULL); | |
1592 | } | |
34dc7c2f | 1593 | |
b128c09f BB |
1594 | if (active) |
1595 | continue; | |
34dc7c2f | 1596 | |
b128c09f BB |
1597 | if ((dst = refresh_config(hdl, src)) == NULL) { |
1598 | nvlist_free(raw); | |
1599 | nvlist_free(pools); | |
1600 | return (NULL); | |
1601 | } | |
34dc7c2f | 1602 | |
b128c09f BB |
1603 | if (nvlist_add_nvlist(pools, nvpair_name(elem), dst) != 0) { |
1604 | (void) no_memory(hdl); | |
34dc7c2f | 1605 | nvlist_free(dst); |
b128c09f BB |
1606 | nvlist_free(raw); |
1607 | nvlist_free(pools); | |
1608 | return (NULL); | |
34dc7c2f | 1609 | } |
b128c09f | 1610 | nvlist_free(dst); |
34dc7c2f BB |
1611 | } |
1612 | ||
1613 | nvlist_free(raw); | |
1614 | return (pools); | |
1615 | } | |
1616 | ||
428870ff BB |
1617 | static int |
1618 | name_or_guid_exists(zpool_handle_t *zhp, void *data) | |
1619 | { | |
1620 | importargs_t *import = data; | |
1621 | int found = 0; | |
1622 | ||
1623 | if (import->poolname != NULL) { | |
1624 | char *pool_name; | |
1625 | ||
1626 | verify(nvlist_lookup_string(zhp->zpool_config, | |
1627 | ZPOOL_CONFIG_POOL_NAME, &pool_name) == 0); | |
1628 | if (strcmp(pool_name, import->poolname) == 0) | |
1629 | found = 1; | |
1630 | } else { | |
1631 | uint64_t pool_guid; | |
1632 | ||
1633 | verify(nvlist_lookup_uint64(zhp->zpool_config, | |
1634 | ZPOOL_CONFIG_POOL_GUID, &pool_guid) == 0); | |
1635 | if (pool_guid == import->guid) | |
1636 | found = 1; | |
1637 | } | |
1638 | ||
1639 | zpool_close(zhp); | |
1640 | return (found); | |
1641 | } | |
1642 | ||
1643 | nvlist_t * | |
1644 | zpool_search_import(libzfs_handle_t *hdl, importargs_t *import) | |
1645 | { | |
1646 | verify(import->poolname == NULL || import->guid == 0); | |
1647 | ||
1648 | if (import->unique) | |
1649 | import->exists = zpool_iter(hdl, name_or_guid_exists, import); | |
1650 | ||
1651 | if (import->cachefile != NULL) | |
1652 | return (zpool_find_import_cached(hdl, import->cachefile, | |
1653 | import->poolname, import->guid)); | |
1654 | ||
1655 | return (zpool_find_import_impl(hdl, import)); | |
1656 | } | |
34dc7c2f BB |
1657 | |
1658 | boolean_t | |
1659 | find_guid(nvlist_t *nv, uint64_t guid) | |
1660 | { | |
1661 | uint64_t tmp; | |
1662 | nvlist_t **child; | |
1663 | uint_t c, children; | |
1664 | ||
1665 | verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID, &tmp) == 0); | |
1666 | if (tmp == guid) | |
1667 | return (B_TRUE); | |
1668 | ||
1669 | if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN, | |
1670 | &child, &children) == 0) { | |
1671 | for (c = 0; c < children; c++) | |
1672 | if (find_guid(child[c], guid)) | |
1673 | return (B_TRUE); | |
1674 | } | |
1675 | ||
1676 | return (B_FALSE); | |
1677 | } | |
1678 | ||
1679 | typedef struct aux_cbdata { | |
1680 | const char *cb_type; | |
1681 | uint64_t cb_guid; | |
1682 | zpool_handle_t *cb_zhp; | |
1683 | } aux_cbdata_t; | |
1684 | ||
1685 | static int | |
1686 | find_aux(zpool_handle_t *zhp, void *data) | |
1687 | { | |
1688 | aux_cbdata_t *cbp = data; | |
1689 | nvlist_t **list; | |
1690 | uint_t i, count; | |
1691 | uint64_t guid; | |
1692 | nvlist_t *nvroot; | |
1693 | ||
1694 | verify(nvlist_lookup_nvlist(zhp->zpool_config, ZPOOL_CONFIG_VDEV_TREE, | |
1695 | &nvroot) == 0); | |
1696 | ||
1697 | if (nvlist_lookup_nvlist_array(nvroot, cbp->cb_type, | |
1698 | &list, &count) == 0) { | |
1699 | for (i = 0; i < count; i++) { | |
1700 | verify(nvlist_lookup_uint64(list[i], | |
1701 | ZPOOL_CONFIG_GUID, &guid) == 0); | |
1702 | if (guid == cbp->cb_guid) { | |
1703 | cbp->cb_zhp = zhp; | |
1704 | return (1); | |
1705 | } | |
1706 | } | |
1707 | } | |
1708 | ||
1709 | zpool_close(zhp); | |
1710 | return (0); | |
1711 | } | |
1712 | ||
1713 | /* | |
1714 | * Determines if the pool is in use. If so, it returns true and the state of | |
1715 | * the pool as well as the name of the pool. Both strings are allocated and | |
1716 | * must be freed by the caller. | |
1717 | */ | |
1718 | int | |
1719 | zpool_in_use(libzfs_handle_t *hdl, int fd, pool_state_t *state, char **namestr, | |
1720 | boolean_t *inuse) | |
1721 | { | |
1722 | nvlist_t *config; | |
1723 | char *name; | |
1724 | boolean_t ret; | |
1725 | uint64_t guid, vdev_guid; | |
1726 | zpool_handle_t *zhp; | |
1727 | nvlist_t *pool_config; | |
1728 | uint64_t stateval, isspare; | |
1729 | aux_cbdata_t cb = { 0 }; | |
1730 | boolean_t isactive; | |
1731 | ||
1732 | *inuse = B_FALSE; | |
1733 | ||
7d90f569 | 1734 | if (zpool_read_label(fd, &config, NULL) != 0) { |
34dc7c2f BB |
1735 | (void) no_memory(hdl); |
1736 | return (-1); | |
1737 | } | |
1738 | ||
1739 | if (config == NULL) | |
1740 | return (0); | |
1741 | ||
1742 | verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_STATE, | |
1743 | &stateval) == 0); | |
1744 | verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_GUID, | |
1745 | &vdev_guid) == 0); | |
1746 | ||
1747 | if (stateval != POOL_STATE_SPARE && stateval != POOL_STATE_L2CACHE) { | |
1748 | verify(nvlist_lookup_string(config, ZPOOL_CONFIG_POOL_NAME, | |
1749 | &name) == 0); | |
1750 | verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID, | |
1751 | &guid) == 0); | |
1752 | } | |
1753 | ||
1754 | switch (stateval) { | |
1755 | case POOL_STATE_EXPORTED: | |
572e2857 BB |
1756 | /* |
1757 | * A pool with an exported state may in fact be imported | |
1758 | * read-only, so check the in-core state to see if it's | |
1759 | * active and imported read-only. If it is, set | |
1760 | * its state to active. | |
1761 | */ | |
1762 | if (pool_active(hdl, name, guid, &isactive) == 0 && isactive && | |
02f8fe42 JJS |
1763 | (zhp = zpool_open_canfail(hdl, name)) != NULL) { |
1764 | if (zpool_get_prop_int(zhp, ZPOOL_PROP_READONLY, NULL)) | |
1765 | stateval = POOL_STATE_ACTIVE; | |
1766 | ||
1767 | /* | |
1768 | * All we needed the zpool handle for is the | |
1769 | * readonly prop check. | |
1770 | */ | |
1771 | zpool_close(zhp); | |
1772 | } | |
572e2857 | 1773 | |
34dc7c2f BB |
1774 | ret = B_TRUE; |
1775 | break; | |
1776 | ||
1777 | case POOL_STATE_ACTIVE: | |
1778 | /* | |
1779 | * For an active pool, we have to determine if it's really part | |
1780 | * of a currently active pool (in which case the pool will exist | |
1781 | * and the guid will be the same), or whether it's part of an | |
1782 | * active pool that was disconnected without being explicitly | |
1783 | * exported. | |
1784 | */ | |
1785 | if (pool_active(hdl, name, guid, &isactive) != 0) { | |
1786 | nvlist_free(config); | |
1787 | return (-1); | |
1788 | } | |
1789 | ||
1790 | if (isactive) { | |
1791 | /* | |
1792 | * Because the device may have been removed while | |
1793 | * offlined, we only report it as active if the vdev is | |
1794 | * still present in the config. Otherwise, pretend like | |
1795 | * it's not in use. | |
1796 | */ | |
1797 | if ((zhp = zpool_open_canfail(hdl, name)) != NULL && | |
1798 | (pool_config = zpool_get_config(zhp, NULL)) | |
1799 | != NULL) { | |
1800 | nvlist_t *nvroot; | |
1801 | ||
1802 | verify(nvlist_lookup_nvlist(pool_config, | |
1803 | ZPOOL_CONFIG_VDEV_TREE, &nvroot) == 0); | |
1804 | ret = find_guid(nvroot, vdev_guid); | |
1805 | } else { | |
1806 | ret = B_FALSE; | |
1807 | } | |
1808 | ||
1809 | /* | |
1810 | * If this is an active spare within another pool, we | |
1811 | * treat it like an unused hot spare. This allows the | |
1812 | * user to create a pool with a hot spare that currently | |
1813 | * in use within another pool. Since we return B_TRUE, | |
1814 | * libdiskmgt will continue to prevent generic consumers | |
1815 | * from using the device. | |
1816 | */ | |
1817 | if (ret && nvlist_lookup_uint64(config, | |
1818 | ZPOOL_CONFIG_IS_SPARE, &isspare) == 0 && isspare) | |
1819 | stateval = POOL_STATE_SPARE; | |
1820 | ||
1821 | if (zhp != NULL) | |
1822 | zpool_close(zhp); | |
1823 | } else { | |
1824 | stateval = POOL_STATE_POTENTIALLY_ACTIVE; | |
1825 | ret = B_TRUE; | |
1826 | } | |
1827 | break; | |
1828 | ||
1829 | case POOL_STATE_SPARE: | |
1830 | /* | |
1831 | * For a hot spare, it can be either definitively in use, or | |
1832 | * potentially active. To determine if it's in use, we iterate | |
1833 | * over all pools in the system and search for one with a spare | |
1834 | * with a matching guid. | |
1835 | * | |
1836 | * Due to the shared nature of spares, we don't actually report | |
1837 | * the potentially active case as in use. This means the user | |
1838 | * can freely create pools on the hot spares of exported pools, | |
1839 | * but to do otherwise makes the resulting code complicated, and | |
1840 | * we end up having to deal with this case anyway. | |
1841 | */ | |
1842 | cb.cb_zhp = NULL; | |
1843 | cb.cb_guid = vdev_guid; | |
1844 | cb.cb_type = ZPOOL_CONFIG_SPARES; | |
1845 | if (zpool_iter(hdl, find_aux, &cb) == 1) { | |
1846 | name = (char *)zpool_get_name(cb.cb_zhp); | |
1847 | ret = TRUE; | |
1848 | } else { | |
1849 | ret = FALSE; | |
1850 | } | |
1851 | break; | |
1852 | ||
1853 | case POOL_STATE_L2CACHE: | |
1854 | ||
1855 | /* | |
1856 | * Check if any pool is currently using this l2cache device. | |
1857 | */ | |
1858 | cb.cb_zhp = NULL; | |
1859 | cb.cb_guid = vdev_guid; | |
1860 | cb.cb_type = ZPOOL_CONFIG_L2CACHE; | |
1861 | if (zpool_iter(hdl, find_aux, &cb) == 1) { | |
1862 | name = (char *)zpool_get_name(cb.cb_zhp); | |
1863 | ret = TRUE; | |
1864 | } else { | |
1865 | ret = FALSE; | |
1866 | } | |
1867 | break; | |
1868 | ||
1869 | default: | |
1870 | ret = B_FALSE; | |
1871 | } | |
1872 | ||
1873 | ||
1874 | if (ret) { | |
1875 | if ((*namestr = zfs_strdup(hdl, name)) == NULL) { | |
1876 | if (cb.cb_zhp) | |
1877 | zpool_close(cb.cb_zhp); | |
1878 | nvlist_free(config); | |
1879 | return (-1); | |
1880 | } | |
1881 | *state = (pool_state_t)stateval; | |
1882 | } | |
1883 | ||
1884 | if (cb.cb_zhp) | |
1885 | zpool_close(cb.cb_zhp); | |
1886 | ||
1887 | nvlist_free(config); | |
1888 | *inuse = ret; | |
1889 | return (0); | |
1890 | } |