]>
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 | /* | |
0fdd8d64 | 22 | * Copyright 2015 Nexenta Systems, Inc. All rights reserved. |
572e2857 | 23 | * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved. |
3bc7e0fb | 24 | * Copyright (c) 2012 by Delphix. All rights reserved. |
ee42b3d6 | 25 | * Copyright 2015 RackTop Systems. |
39fc0cb5 | 26 | * Copyright (c) 2016, Intel Corporation. |
34dc7c2f BB |
27 | */ |
28 | ||
34dc7c2f BB |
29 | /* |
30 | * Pool import support functions. | |
31 | * | |
32 | * To import a pool, we rely on reading the configuration information from the | |
33 | * ZFS label of each device. If we successfully read the label, then we | |
34 | * organize the configuration information in the following hierarchy: | |
35 | * | |
36 | * pool guid -> toplevel vdev guid -> label txg | |
37 | * | |
38 | * Duplicate entries matching this same tuple will be discarded. Once we have | |
39 | * examined every device, we pick the best label txg config for each toplevel | |
40 | * vdev. We then arrange these toplevel vdevs into a complete pool config, and | |
41 | * update any paths that have changed. Finally, we attempt to import the pool | |
42 | * using our derived config, and record the results. | |
43 | */ | |
44 | ||
428870ff | 45 | #include <ctype.h> |
34dc7c2f BB |
46 | #include <devid.h> |
47 | #include <dirent.h> | |
48 | #include <errno.h> | |
49 | #include <libintl.h> | |
39fc0cb5 DB |
50 | #ifdef HAVE_LIBUDEV |
51 | #include <libudev.h> | |
52 | #include <sched.h> | |
53 | #endif | |
428870ff | 54 | #include <stddef.h> |
34dc7c2f BB |
55 | #include <stdlib.h> |
56 | #include <string.h> | |
57 | #include <sys/stat.h> | |
58 | #include <unistd.h> | |
59 | #include <fcntl.h> | |
428870ff BB |
60 | #include <sys/vtoc.h> |
61 | #include <sys/dktp/fdisk.h> | |
62 | #include <sys/efi_partition.h> | |
34dc7c2f | 63 | #include <sys/vdev_impl.h> |
d603ed6c | 64 | #include <blkid/blkid.h> |
34dc7c2f BB |
65 | #include "libzfs.h" |
66 | #include "libzfs_impl.h" | |
1bbd8770 | 67 | #include <libzfs.h> |
34dc7c2f BB |
68 | |
69 | /* | |
70 | * Intermediate structures used to gather configuration information. | |
71 | */ | |
72 | typedef struct config_entry { | |
73 | uint64_t ce_txg; | |
74 | nvlist_t *ce_config; | |
75 | struct config_entry *ce_next; | |
76 | } config_entry_t; | |
77 | ||
78 | typedef struct vdev_entry { | |
79 | uint64_t ve_guid; | |
80 | config_entry_t *ve_configs; | |
81 | struct vdev_entry *ve_next; | |
82 | } vdev_entry_t; | |
83 | ||
84 | typedef struct pool_entry { | |
85 | uint64_t pe_guid; | |
86 | vdev_entry_t *pe_vdevs; | |
87 | struct pool_entry *pe_next; | |
88 | } pool_entry_t; | |
89 | ||
90 | typedef struct name_entry { | |
91 | char *ne_name; | |
92 | uint64_t ne_guid; | |
44867b6d | 93 | uint64_t ne_order; |
7d90f569 | 94 | uint64_t ne_num_labels; |
34dc7c2f BB |
95 | struct name_entry *ne_next; |
96 | } name_entry_t; | |
97 | ||
98 | typedef struct pool_list { | |
99 | pool_entry_t *pools; | |
100 | name_entry_t *names; | |
101 | } pool_list_t; | |
102 | ||
325414e4 BB |
103 | #define DEV_BYID_PATH "/dev/disk/by-id/" |
104 | ||
39fc0cb5 DB |
105 | /* |
106 | * Linux persistent device strings for vdev labels | |
107 | * | |
108 | * based on libudev for consistency with libudev disk add/remove events | |
109 | */ | |
110 | #ifdef HAVE_LIBUDEV | |
111 | ||
39fc0cb5 DB |
112 | typedef struct vdev_dev_strs { |
113 | char vds_devid[128]; | |
114 | char vds_devphys[128]; | |
115 | } vdev_dev_strs_t; | |
116 | ||
117 | /* | |
118 | * Obtain the persistent device id string (describes what) | |
119 | * | |
d02ca379 | 120 | * used by ZED vdev matching for auto-{online,expand,replace} |
39fc0cb5 | 121 | */ |
d02ca379 DB |
122 | int |
123 | zfs_device_get_devid(struct udev_device *dev, char *bufptr, size_t buflen) | |
39fc0cb5 DB |
124 | { |
125 | struct udev_list_entry *entry; | |
126 | const char *bus; | |
127 | char devbyid[MAXPATHLEN]; | |
128 | ||
129 | /* The bus based by-id path is preferred */ | |
130 | bus = udev_device_get_property_value(dev, "ID_BUS"); | |
131 | ||
132 | if (bus == NULL) { | |
133 | const char *dm_uuid; | |
134 | ||
135 | /* | |
136 | * For multipath nodes use the persistent uuid based identifier | |
137 | * | |
138 | * Example: /dev/disk/by-id/dm-uuid-mpath-35000c5006304de3f | |
139 | */ | |
140 | dm_uuid = udev_device_get_property_value(dev, "DM_UUID"); | |
141 | if (dm_uuid != NULL) { | |
142 | (void) snprintf(bufptr, buflen, "dm-uuid-%s", dm_uuid); | |
143 | return (0); | |
144 | } | |
145 | return (ENODATA); | |
146 | } | |
147 | ||
148 | /* | |
149 | * locate the bus specific by-id link | |
150 | */ | |
151 | (void) snprintf(devbyid, sizeof (devbyid), "%s%s-", DEV_BYID_PATH, bus); | |
152 | entry = udev_device_get_devlinks_list_entry(dev); | |
153 | while (entry != NULL) { | |
154 | const char *name; | |
155 | ||
156 | name = udev_list_entry_get_name(entry); | |
157 | if (strncmp(name, devbyid, strlen(devbyid)) == 0) { | |
158 | name += strlen(DEV_BYID_PATH); | |
159 | (void) strlcpy(bufptr, name, buflen); | |
160 | return (0); | |
161 | } | |
162 | entry = udev_list_entry_get_next(entry); | |
163 | } | |
164 | ||
165 | return (ENODATA); | |
166 | } | |
167 | ||
168 | /* | |
169 | * Obtain the persistent physical location string (describes where) | |
170 | * | |
d02ca379 | 171 | * used by ZED vdev matching for auto-{online,expand,replace} |
39fc0cb5 | 172 | */ |
d02ca379 DB |
173 | int |
174 | zfs_device_get_physical(struct udev_device *dev, char *bufptr, size_t buflen) | |
39fc0cb5 | 175 | { |
6078881a | 176 | const char *physpath = NULL; |
39fc0cb5 DB |
177 | |
178 | /* | |
6078881a TH |
179 | * Normal disks use ID_PATH for their physical path. Device mapper |
180 | * devices are virtual and don't have a physical path. For them we | |
181 | * use ID_VDEV instead, which is setup via the /etc/vdev_id.conf file. | |
182 | * ID_VDEV provides a persistent path to a virtual device. If you | |
183 | * don't have vdev_id.conf setup, you cannot use multipath autoreplace. | |
39fc0cb5 | 184 | */ |
6078881a TH |
185 | if (!((physpath = udev_device_get_property_value(dev, "ID_PATH")) && |
186 | physpath[0])) { | |
187 | if (!((physpath = | |
188 | udev_device_get_property_value(dev, "ID_VDEV")) && | |
189 | physpath[0])) { | |
190 | return (ENODATA); | |
191 | } | |
39fc0cb5 DB |
192 | } |
193 | ||
6078881a TH |
194 | (void) strlcpy(bufptr, physpath, buflen); |
195 | ||
196 | return (0); | |
197 | } | |
198 | ||
199 | boolean_t | |
200 | udev_is_mpath(struct udev_device *dev) | |
201 | { | |
202 | return udev_device_get_property_value(dev, "DM_UUID") && | |
02730c33 | 203 | udev_device_get_property_value(dev, "MPATH_SBIN_PATH"); |
39fc0cb5 DB |
204 | } |
205 | ||
206 | /* | |
207 | * A disk is considered a multipath whole disk when: | |
208 | * DEVNAME key value has "dm-" | |
209 | * DM_NAME key value has "mpath" prefix | |
210 | * DM_UUID key exists | |
211 | * ID_PART_TABLE_TYPE key does not exist or is not gpt | |
212 | */ | |
213 | static boolean_t | |
214 | udev_mpath_whole_disk(struct udev_device *dev) | |
215 | { | |
6078881a | 216 | const char *devname, *type, *uuid; |
39fc0cb5 DB |
217 | |
218 | devname = udev_device_get_property_value(dev, "DEVNAME"); | |
39fc0cb5 DB |
219 | type = udev_device_get_property_value(dev, "ID_PART_TABLE_TYPE"); |
220 | uuid = udev_device_get_property_value(dev, "DM_UUID"); | |
221 | ||
222 | if ((devname != NULL && strncmp(devname, "/dev/dm-", 8) == 0) && | |
39fc0cb5 DB |
223 | ((type == NULL) || (strcmp(type, "gpt") != 0)) && |
224 | (uuid != NULL)) { | |
225 | return (B_TRUE); | |
226 | } | |
227 | ||
228 | return (B_FALSE); | |
229 | } | |
230 | ||
231 | /* | |
232 | * Check if a disk is effectively a multipath whole disk | |
233 | */ | |
234 | boolean_t | |
235 | is_mpath_whole_disk(const char *path) | |
236 | { | |
237 | struct udev *udev; | |
238 | struct udev_device *dev = NULL; | |
239 | char nodepath[MAXPATHLEN]; | |
240 | char *sysname; | |
241 | boolean_t wholedisk = B_FALSE; | |
242 | ||
243 | if (realpath(path, nodepath) == NULL) | |
244 | return (B_FALSE); | |
245 | sysname = strrchr(nodepath, '/') + 1; | |
246 | if (strncmp(sysname, "dm-", 3) != 0) | |
247 | return (B_FALSE); | |
248 | if ((udev = udev_new()) == NULL) | |
249 | return (B_FALSE); | |
250 | if ((dev = udev_device_new_from_subsystem_sysname(udev, "block", | |
251 | sysname)) == NULL) { | |
252 | udev_device_unref(dev); | |
253 | return (B_FALSE); | |
254 | } | |
255 | ||
256 | wholedisk = udev_mpath_whole_disk(dev); | |
257 | ||
258 | udev_device_unref(dev); | |
259 | return (wholedisk); | |
260 | } | |
261 | ||
262 | static int | |
263 | udev_device_is_ready(struct udev_device *dev) | |
264 | { | |
265 | #ifdef HAVE_LIBUDEV_UDEV_DEVICE_GET_IS_INITIALIZED | |
266 | return (udev_device_get_is_initialized(dev)); | |
267 | #else | |
268 | /* wait for DEVLINKS property to be initialized */ | |
269 | return (udev_device_get_property_value(dev, "DEVLINKS") != NULL); | |
270 | #endif | |
271 | } | |
272 | ||
2d82ea8b BB |
273 | /* |
274 | * Wait up to timeout_ms for udev to set up the device node. The device is | |
275 | * considered ready when libudev determines it has been initialized, all of | |
276 | * the device links have been verified to exist, and it has been allowed to | |
277 | * settle. At this point the device the device can be accessed reliably. | |
278 | * Depending on the complexity of the udev rules this process could take | |
279 | * several seconds. | |
280 | */ | |
281 | int | |
282 | zpool_label_disk_wait(char *path, int timeout_ms) | |
283 | { | |
284 | struct udev *udev; | |
285 | struct udev_device *dev = NULL; | |
286 | char nodepath[MAXPATHLEN]; | |
287 | char *sysname = NULL; | |
288 | int ret = ENODEV; | |
289 | int settle_ms = 50; | |
290 | long sleep_ms = 10; | |
291 | hrtime_t start, settle; | |
292 | ||
293 | if ((udev = udev_new()) == NULL) | |
294 | return (ENXIO); | |
295 | ||
296 | start = gethrtime(); | |
297 | settle = 0; | |
298 | ||
299 | do { | |
300 | if (sysname == NULL) { | |
301 | if (realpath(path, nodepath) != NULL) { | |
302 | sysname = strrchr(nodepath, '/') + 1; | |
303 | } else { | |
304 | (void) usleep(sleep_ms * MILLISEC); | |
305 | continue; | |
306 | } | |
307 | } | |
308 | ||
309 | dev = udev_device_new_from_subsystem_sysname(udev, | |
310 | "block", sysname); | |
311 | if ((dev != NULL) && udev_device_is_ready(dev)) { | |
312 | struct udev_list_entry *links, *link; | |
313 | ||
314 | ret = 0; | |
315 | links = udev_device_get_devlinks_list_entry(dev); | |
316 | ||
317 | udev_list_entry_foreach(link, links) { | |
318 | struct stat64 statbuf; | |
319 | const char *name; | |
320 | ||
321 | name = udev_list_entry_get_name(link); | |
322 | errno = 0; | |
323 | if (stat64(name, &statbuf) == 0 && errno == 0) | |
324 | continue; | |
325 | ||
326 | settle = 0; | |
327 | ret = ENODEV; | |
328 | break; | |
329 | } | |
330 | ||
331 | if (ret == 0) { | |
332 | if (settle == 0) { | |
333 | settle = gethrtime(); | |
334 | } else if (NSEC2MSEC(gethrtime() - settle) >= | |
335 | settle_ms) { | |
336 | udev_device_unref(dev); | |
337 | break; | |
338 | } | |
339 | } | |
340 | } | |
341 | ||
342 | udev_device_unref(dev); | |
343 | (void) usleep(sleep_ms * MILLISEC); | |
344 | ||
345 | } while (NSEC2MSEC(gethrtime() - start) < timeout_ms); | |
346 | ||
347 | udev_unref(udev); | |
348 | ||
349 | return (ret); | |
350 | } | |
351 | ||
352 | ||
39fc0cb5 DB |
353 | /* |
354 | * Encode the persistent devices strings | |
355 | * used for the vdev disk label | |
356 | */ | |
357 | static int | |
358 | encode_device_strings(const char *path, vdev_dev_strs_t *ds, | |
359 | boolean_t wholedisk) | |
34dc7c2f | 360 | { |
39fc0cb5 DB |
361 | struct udev *udev; |
362 | struct udev_device *dev = NULL; | |
363 | char nodepath[MAXPATHLEN]; | |
364 | char *sysname; | |
365 | int ret = ENODEV; | |
366 | hrtime_t start; | |
367 | ||
368 | if ((udev = udev_new()) == NULL) | |
369 | return (ENXIO); | |
370 | ||
371 | /* resolve path to a runtime device node instance */ | |
372 | if (realpath(path, nodepath) == NULL) | |
373 | goto no_dev; | |
374 | ||
375 | sysname = strrchr(nodepath, '/') + 1; | |
376 | ||
377 | /* | |
378 | * Wait up to 3 seconds for udev to set up the device node context | |
379 | */ | |
380 | start = gethrtime(); | |
381 | do { | |
382 | dev = udev_device_new_from_subsystem_sysname(udev, "block", | |
383 | sysname); | |
384 | if (dev == NULL) | |
385 | goto no_dev; | |
386 | if (udev_device_is_ready(dev)) | |
387 | break; /* udev ready */ | |
388 | ||
389 | udev_device_unref(dev); | |
390 | dev = NULL; | |
391 | ||
392 | if (NSEC2MSEC(gethrtime() - start) < 10) | |
393 | (void) sched_yield(); /* yield/busy wait up to 10ms */ | |
394 | else | |
395 | (void) usleep(10 * MILLISEC); | |
396 | ||
397 | } while (NSEC2MSEC(gethrtime() - start) < (3 * MILLISEC)); | |
398 | ||
399 | if (dev == NULL) | |
400 | goto no_dev; | |
401 | ||
402 | /* | |
403 | * Only whole disks require extra device strings | |
404 | */ | |
405 | if (!wholedisk && !udev_mpath_whole_disk(dev)) | |
406 | goto no_dev; | |
407 | ||
d02ca379 | 408 | ret = zfs_device_get_devid(dev, ds->vds_devid, sizeof (ds->vds_devid)); |
39fc0cb5 DB |
409 | if (ret != 0) |
410 | goto no_dev_ref; | |
411 | ||
412 | /* physical location string (optional) */ | |
d02ca379 | 413 | if (zfs_device_get_physical(dev, ds->vds_devphys, |
39fc0cb5 DB |
414 | sizeof (ds->vds_devphys)) != 0) { |
415 | ds->vds_devphys[0] = '\0'; /* empty string --> not available */ | |
34dc7c2f | 416 | } |
39fc0cb5 DB |
417 | |
418 | no_dev_ref: | |
419 | udev_device_unref(dev); | |
420 | no_dev: | |
421 | udev_unref(udev); | |
34dc7c2f BB |
422 | |
423 | return (ret); | |
424 | } | |
425 | ||
39fc0cb5 DB |
426 | /* |
427 | * Update a leaf vdev's persistent device strings (Linux only) | |
428 | * | |
429 | * - only applies for a dedicated leaf vdev (aka whole disk) | |
430 | * - updated during pool create|add|attach|import | |
431 | * - used for matching device matching during auto-{online,expand,replace} | |
432 | * - stored in a leaf disk config label (i.e. alongside 'path' NVP) | |
433 | * - these strings are currently not used in kernel (i.e. for vdev_disk_open) | |
434 | * | |
435 | * single device node example: | |
436 | * devid: 'scsi-MG03SCA300_350000494a8cb3d67-part1' | |
437 | * phys_path: 'pci-0000:04:00.0-sas-0x50000394a8cb3d67-lun-0' | |
438 | * | |
439 | * multipath device node example: | |
440 | * devid: 'dm-uuid-mpath-35000c5006304de3f' | |
1bbd8770 TH |
441 | * |
442 | * We also store the enclosure sysfs path for turning on enclosure LEDs | |
443 | * (if applicable): | |
444 | * vdev_enc_sysfs_path: '/sys/class/enclosure/11:0:1:0/SLOT 4' | |
39fc0cb5 DB |
445 | */ |
446 | void | |
447 | update_vdev_config_dev_strs(nvlist_t *nv) | |
448 | { | |
449 | vdev_dev_strs_t vds; | |
450 | char *env, *type, *path; | |
451 | uint64_t wholedisk = 0; | |
1bbd8770 | 452 | char *upath, *spath; |
39fc0cb5 DB |
453 | |
454 | /* | |
455 | * For the benefit of legacy ZFS implementations, allow | |
456 | * for opting out of devid strings in the vdev label. | |
457 | * | |
458 | * example use: | |
459 | * env ZFS_VDEV_DEVID_OPT_OUT=YES zpool import dozer | |
460 | * | |
461 | * explanation: | |
462 | * Older ZFS on Linux implementations had issues when attempting to | |
463 | * display pool config VDEV names if a "devid" NVP value is present | |
464 | * in the pool's config. | |
465 | * | |
466 | * For example, a pool that originated on illumos platform would | |
467 | * have a devid value in the config and "zpool status" would fail | |
468 | * when listing the config. | |
469 | * | |
470 | * A pool can be stripped of any "devid" values on import or | |
471 | * prevented from adding them on zpool create|add by setting | |
472 | * ZFS_VDEV_DEVID_OPT_OUT. | |
473 | */ | |
474 | env = getenv("ZFS_VDEV_DEVID_OPT_OUT"); | |
475 | if (env && (strtoul(env, NULL, 0) > 0 || | |
476 | !strncasecmp(env, "YES", 3) || !strncasecmp(env, "ON", 2))) { | |
477 | (void) nvlist_remove_all(nv, ZPOOL_CONFIG_DEVID); | |
478 | (void) nvlist_remove_all(nv, ZPOOL_CONFIG_PHYS_PATH); | |
1bbd8770 | 479 | (void) nvlist_remove_all(nv, ZPOOL_CONFIG_VDEV_ENC_SYSFS_PATH); |
39fc0cb5 DB |
480 | return; |
481 | } | |
482 | ||
483 | if (nvlist_lookup_string(nv, ZPOOL_CONFIG_TYPE, &type) != 0 || | |
484 | strcmp(type, VDEV_TYPE_DISK) != 0) { | |
485 | return; | |
486 | } | |
487 | if (nvlist_lookup_string(nv, ZPOOL_CONFIG_PATH, &path) != 0) | |
488 | return; | |
489 | (void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_WHOLE_DISK, &wholedisk); | |
490 | ||
491 | /* | |
492 | * Update device string values in config nvlist | |
493 | */ | |
494 | if (encode_device_strings(path, &vds, (boolean_t)wholedisk) == 0) { | |
495 | (void) nvlist_add_string(nv, ZPOOL_CONFIG_DEVID, vds.vds_devid); | |
496 | if (vds.vds_devphys[0] != '\0') { | |
497 | (void) nvlist_add_string(nv, ZPOOL_CONFIG_PHYS_PATH, | |
498 | vds.vds_devphys); | |
499 | } | |
1bbd8770 TH |
500 | |
501 | /* Add enclosure sysfs path (if disk is in an enclosure) */ | |
502 | upath = zfs_get_underlying_path(path); | |
503 | spath = zfs_get_enclosure_sysfs_path(upath); | |
504 | if (spath) | |
505 | nvlist_add_string(nv, ZPOOL_CONFIG_VDEV_ENC_SYSFS_PATH, | |
506 | spath); | |
507 | free(upath); | |
508 | free(spath); | |
39fc0cb5 DB |
509 | } else { |
510 | /* clear out any stale entries */ | |
511 | (void) nvlist_remove_all(nv, ZPOOL_CONFIG_DEVID); | |
512 | (void) nvlist_remove_all(nv, ZPOOL_CONFIG_PHYS_PATH); | |
1bbd8770 | 513 | (void) nvlist_remove_all(nv, ZPOOL_CONFIG_VDEV_ENC_SYSFS_PATH); |
39fc0cb5 DB |
514 | } |
515 | } | |
516 | #else | |
517 | ||
518 | boolean_t | |
519 | is_mpath_whole_disk(const char *path) | |
520 | { | |
521 | return (B_FALSE); | |
522 | } | |
523 | ||
2d82ea8b BB |
524 | /* |
525 | * Wait up to timeout_ms for udev to set up the device node. The device is | |
526 | * considered ready when the provided path have been verified to exist and | |
527 | * it has been allowed to settle. At this point the device the device can | |
528 | * be accessed reliably. Depending on the complexity of the udev rules thisi | |
529 | * process could take several seconds. | |
530 | */ | |
531 | int | |
532 | zpool_label_disk_wait(char *path, int timeout_ms) | |
533 | { | |
534 | int settle_ms = 50; | |
535 | long sleep_ms = 10; | |
536 | hrtime_t start, settle; | |
537 | struct stat64 statbuf; | |
538 | ||
539 | start = gethrtime(); | |
540 | settle = 0; | |
541 | ||
542 | do { | |
543 | errno = 0; | |
544 | if ((stat64(path, &statbuf) == 0) && (errno == 0)) { | |
545 | if (settle == 0) | |
546 | settle = gethrtime(); | |
547 | else if (NSEC2MSEC(gethrtime() - settle) >= settle_ms) | |
548 | return (0); | |
549 | } else if (errno != ENOENT) { | |
550 | return (errno); | |
551 | } | |
552 | ||
553 | usleep(sleep_ms * MILLISEC); | |
554 | } while (NSEC2MSEC(gethrtime() - start) < timeout_ms); | |
555 | ||
556 | return (ENODEV); | |
557 | } | |
558 | ||
39fc0cb5 DB |
559 | void |
560 | update_vdev_config_dev_strs(nvlist_t *nv) | |
561 | { | |
562 | } | |
563 | ||
564 | #endif /* HAVE_LIBUDEV */ | |
565 | ||
34dc7c2f BB |
566 | /* |
567 | * Go through and fix up any path and/or devid information for the given vdev | |
568 | * configuration. | |
569 | */ | |
570 | static int | |
571 | fix_paths(nvlist_t *nv, name_entry_t *names) | |
572 | { | |
573 | nvlist_t **child; | |
574 | uint_t c, children; | |
575 | uint64_t guid; | |
576 | name_entry_t *ne, *best; | |
39fc0cb5 | 577 | char *path; |
34dc7c2f BB |
578 | |
579 | if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN, | |
580 | &child, &children) == 0) { | |
581 | for (c = 0; c < children; c++) | |
582 | if (fix_paths(child[c], names) != 0) | |
583 | return (-1); | |
584 | return (0); | |
585 | } | |
586 | ||
587 | /* | |
588 | * This is a leaf (file or disk) vdev. In either case, go through | |
589 | * the name list and see if we find a matching guid. If so, replace | |
590 | * the path and see if we can calculate a new devid. | |
591 | * | |
592 | * There may be multiple names associated with a particular guid, in | |
44867b6d BB |
593 | * which case we have overlapping partitions or multiple paths to the |
594 | * same disk. In this case we prefer to use the path name which | |
595 | * matches the ZPOOL_CONFIG_PATH. If no matching entry is found we | |
596 | * use the lowest order device which corresponds to the first match | |
597 | * while traversing the ZPOOL_IMPORT_PATH search path. | |
34dc7c2f BB |
598 | */ |
599 | verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID, &guid) == 0); | |
600 | if (nvlist_lookup_string(nv, ZPOOL_CONFIG_PATH, &path) != 0) | |
601 | path = NULL; | |
602 | ||
34dc7c2f BB |
603 | best = NULL; |
604 | for (ne = names; ne != NULL; ne = ne->ne_next) { | |
605 | if (ne->ne_guid == guid) { | |
34dc7c2f BB |
606 | if (path == NULL) { |
607 | best = ne; | |
608 | break; | |
609 | } | |
610 | ||
44867b6d | 611 | if ((strlen(path) == strlen(ne->ne_name)) && |
d1d7e268 | 612 | strncmp(path, ne->ne_name, strlen(path)) == 0) { |
34dc7c2f | 613 | best = ne; |
44867b6d | 614 | break; |
34dc7c2f | 615 | } |
44867b6d | 616 | |
7d90f569 | 617 | if (best == NULL) { |
44867b6d | 618 | best = ne; |
7d90f569 BB |
619 | continue; |
620 | } | |
621 | ||
622 | /* Prefer paths with move vdev labels. */ | |
623 | if (ne->ne_num_labels > best->ne_num_labels) { | |
624 | best = ne; | |
625 | continue; | |
626 | } | |
627 | ||
628 | /* Prefer paths earlier in the search order. */ | |
20c901dc | 629 | if (ne->ne_num_labels == best->ne_num_labels && |
7d90f569 BB |
630 | ne->ne_order < best->ne_order) { |
631 | best = ne; | |
632 | continue; | |
633 | } | |
34dc7c2f BB |
634 | } |
635 | } | |
636 | ||
637 | if (best == NULL) | |
638 | return (0); | |
639 | ||
640 | if (nvlist_add_string(nv, ZPOOL_CONFIG_PATH, best->ne_name) != 0) | |
641 | return (-1); | |
642 | ||
39fc0cb5 DB |
643 | /* Linux only - update ZPOOL_CONFIG_DEVID and ZPOOL_CONFIG_PHYS_PATH */ |
644 | update_vdev_config_dev_strs(nv); | |
34dc7c2f BB |
645 | |
646 | return (0); | |
647 | } | |
648 | ||
649 | /* | |
650 | * Add the given configuration to the list of known devices. | |
651 | */ | |
652 | static int | |
653 | add_config(libzfs_handle_t *hdl, pool_list_t *pl, const char *path, | |
7d90f569 | 654 | int order, int num_labels, nvlist_t *config) |
34dc7c2f BB |
655 | { |
656 | uint64_t pool_guid, vdev_guid, top_guid, txg, state; | |
657 | pool_entry_t *pe; | |
658 | vdev_entry_t *ve; | |
659 | config_entry_t *ce; | |
660 | name_entry_t *ne; | |
661 | ||
662 | /* | |
663 | * If this is a hot spare not currently in use or level 2 cache | |
664 | * device, add it to the list of names to translate, but don't do | |
665 | * anything else. | |
666 | */ | |
667 | if (nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_STATE, | |
668 | &state) == 0 && | |
669 | (state == POOL_STATE_SPARE || state == POOL_STATE_L2CACHE) && | |
670 | nvlist_lookup_uint64(config, ZPOOL_CONFIG_GUID, &vdev_guid) == 0) { | |
e24e62a9 | 671 | if ((ne = zfs_alloc(hdl, sizeof (name_entry_t))) == NULL) { |
672 | nvlist_free(config); | |
34dc7c2f | 673 | return (-1); |
e24e62a9 | 674 | } |
34dc7c2f BB |
675 | |
676 | if ((ne->ne_name = zfs_strdup(hdl, path)) == NULL) { | |
677 | free(ne); | |
e24e62a9 | 678 | nvlist_free(config); |
34dc7c2f BB |
679 | return (-1); |
680 | } | |
681 | ne->ne_guid = vdev_guid; | |
44867b6d | 682 | ne->ne_order = order; |
7d90f569 | 683 | ne->ne_num_labels = num_labels; |
34dc7c2f BB |
684 | ne->ne_next = pl->names; |
685 | pl->names = ne; | |
e24e62a9 | 686 | nvlist_free(config); |
34dc7c2f BB |
687 | return (0); |
688 | } | |
689 | ||
690 | /* | |
691 | * If we have a valid config but cannot read any of these fields, then | |
692 | * it means we have a half-initialized label. In vdev_label_init() | |
693 | * we write a label with txg == 0 so that we can identify the device | |
694 | * in case the user refers to the same disk later on. If we fail to | |
695 | * create the pool, we'll be left with a label in this state | |
696 | * which should not be considered part of a valid pool. | |
697 | */ | |
698 | if (nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID, | |
699 | &pool_guid) != 0 || | |
700 | nvlist_lookup_uint64(config, ZPOOL_CONFIG_GUID, | |
701 | &vdev_guid) != 0 || | |
702 | nvlist_lookup_uint64(config, ZPOOL_CONFIG_TOP_GUID, | |
703 | &top_guid) != 0 || | |
704 | nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_TXG, | |
705 | &txg) != 0 || txg == 0) { | |
706 | nvlist_free(config); | |
707 | return (0); | |
708 | } | |
709 | ||
710 | /* | |
711 | * First, see if we know about this pool. If not, then add it to the | |
712 | * list of known pools. | |
713 | */ | |
714 | for (pe = pl->pools; pe != NULL; pe = pe->pe_next) { | |
715 | if (pe->pe_guid == pool_guid) | |
716 | break; | |
717 | } | |
718 | ||
719 | if (pe == NULL) { | |
720 | if ((pe = zfs_alloc(hdl, sizeof (pool_entry_t))) == NULL) { | |
721 | nvlist_free(config); | |
722 | return (-1); | |
723 | } | |
724 | pe->pe_guid = pool_guid; | |
725 | pe->pe_next = pl->pools; | |
726 | pl->pools = pe; | |
727 | } | |
728 | ||
729 | /* | |
730 | * Second, see if we know about this toplevel vdev. Add it if its | |
731 | * missing. | |
732 | */ | |
733 | for (ve = pe->pe_vdevs; ve != NULL; ve = ve->ve_next) { | |
734 | if (ve->ve_guid == top_guid) | |
735 | break; | |
736 | } | |
737 | ||
738 | if (ve == NULL) { | |
739 | if ((ve = zfs_alloc(hdl, sizeof (vdev_entry_t))) == NULL) { | |
740 | nvlist_free(config); | |
741 | return (-1); | |
742 | } | |
743 | ve->ve_guid = top_guid; | |
744 | ve->ve_next = pe->pe_vdevs; | |
745 | pe->pe_vdevs = ve; | |
746 | } | |
747 | ||
748 | /* | |
749 | * Third, see if we have a config with a matching transaction group. If | |
750 | * so, then we do nothing. Otherwise, add it to the list of known | |
751 | * configs. | |
752 | */ | |
753 | for (ce = ve->ve_configs; ce != NULL; ce = ce->ce_next) { | |
754 | if (ce->ce_txg == txg) | |
755 | break; | |
756 | } | |
757 | ||
758 | if (ce == NULL) { | |
759 | if ((ce = zfs_alloc(hdl, sizeof (config_entry_t))) == NULL) { | |
760 | nvlist_free(config); | |
761 | return (-1); | |
762 | } | |
763 | ce->ce_txg = txg; | |
764 | ce->ce_config = config; | |
765 | ce->ce_next = ve->ve_configs; | |
766 | ve->ve_configs = ce; | |
767 | } else { | |
768 | nvlist_free(config); | |
769 | } | |
770 | ||
771 | /* | |
772 | * At this point we've successfully added our config to the list of | |
773 | * known configs. The last thing to do is add the vdev guid -> path | |
774 | * mappings so that we can fix up the configuration as necessary before | |
775 | * doing the import. | |
776 | */ | |
777 | if ((ne = zfs_alloc(hdl, sizeof (name_entry_t))) == NULL) | |
778 | return (-1); | |
779 | ||
780 | if ((ne->ne_name = zfs_strdup(hdl, path)) == NULL) { | |
781 | free(ne); | |
782 | return (-1); | |
783 | } | |
784 | ||
785 | ne->ne_guid = vdev_guid; | |
44867b6d | 786 | ne->ne_order = order; |
7d90f569 | 787 | ne->ne_num_labels = num_labels; |
34dc7c2f BB |
788 | ne->ne_next = pl->names; |
789 | pl->names = ne; | |
790 | ||
791 | return (0); | |
792 | } | |
793 | ||
794 | /* | |
795 | * Returns true if the named pool matches the given GUID. | |
796 | */ | |
797 | static int | |
798 | pool_active(libzfs_handle_t *hdl, const char *name, uint64_t guid, | |
799 | boolean_t *isactive) | |
800 | { | |
801 | zpool_handle_t *zhp; | |
802 | uint64_t theguid; | |
803 | ||
804 | if (zpool_open_silent(hdl, name, &zhp) != 0) | |
805 | return (-1); | |
806 | ||
807 | if (zhp == NULL) { | |
808 | *isactive = B_FALSE; | |
809 | return (0); | |
810 | } | |
811 | ||
812 | verify(nvlist_lookup_uint64(zhp->zpool_config, ZPOOL_CONFIG_POOL_GUID, | |
813 | &theguid) == 0); | |
814 | ||
815 | zpool_close(zhp); | |
816 | ||
817 | *isactive = (theguid == guid); | |
818 | return (0); | |
819 | } | |
820 | ||
821 | static nvlist_t * | |
822 | refresh_config(libzfs_handle_t *hdl, nvlist_t *config) | |
823 | { | |
824 | nvlist_t *nvl; | |
13fe0198 | 825 | zfs_cmd_t zc = {"\0"}; |
34dc7c2f BB |
826 | int err; |
827 | ||
828 | if (zcmd_write_conf_nvlist(hdl, &zc, config) != 0) | |
829 | return (NULL); | |
830 | ||
831 | if (zcmd_alloc_dst_nvlist(hdl, &zc, | |
832 | zc.zc_nvlist_conf_size * 2) != 0) { | |
833 | zcmd_free_nvlists(&zc); | |
834 | return (NULL); | |
835 | } | |
836 | ||
837 | while ((err = ioctl(hdl->libzfs_fd, ZFS_IOC_POOL_TRYIMPORT, | |
838 | &zc)) != 0 && errno == ENOMEM) { | |
839 | if (zcmd_expand_dst_nvlist(hdl, &zc) != 0) { | |
840 | zcmd_free_nvlists(&zc); | |
841 | return (NULL); | |
842 | } | |
843 | } | |
844 | ||
845 | if (err) { | |
34dc7c2f BB |
846 | zcmd_free_nvlists(&zc); |
847 | return (NULL); | |
848 | } | |
849 | ||
850 | if (zcmd_read_dst_nvlist(hdl, &zc, &nvl) != 0) { | |
851 | zcmd_free_nvlists(&zc); | |
852 | return (NULL); | |
853 | } | |
854 | ||
855 | zcmd_free_nvlists(&zc); | |
856 | return (nvl); | |
857 | } | |
858 | ||
428870ff BB |
859 | /* |
860 | * Determine if the vdev id is a hole in the namespace. | |
861 | */ | |
862 | boolean_t | |
863 | vdev_is_hole(uint64_t *hole_array, uint_t holes, uint_t id) | |
864 | { | |
d6320ddb BB |
865 | int c; |
866 | ||
867 | for (c = 0; c < holes; c++) { | |
428870ff BB |
868 | |
869 | /* Top-level is a hole */ | |
870 | if (hole_array[c] == id) | |
871 | return (B_TRUE); | |
872 | } | |
873 | return (B_FALSE); | |
874 | } | |
875 | ||
34dc7c2f BB |
876 | /* |
877 | * Convert our list of pools into the definitive set of configurations. We | |
878 | * start by picking the best config for each toplevel vdev. Once that's done, | |
879 | * we assemble the toplevel vdevs into a full config for the pool. We make a | |
880 | * pass to fix up any incorrect paths, and then add it to the main list to | |
881 | * return to the user. | |
882 | */ | |
883 | static nvlist_t * | |
884 | get_configs(libzfs_handle_t *hdl, pool_list_t *pl, boolean_t active_ok) | |
885 | { | |
886 | pool_entry_t *pe; | |
887 | vdev_entry_t *ve; | |
888 | config_entry_t *ce; | |
d4ed6673 | 889 | nvlist_t *ret = NULL, *config = NULL, *tmp = NULL, *nvtop, *nvroot; |
34dc7c2f BB |
890 | nvlist_t **spares, **l2cache; |
891 | uint_t i, nspares, nl2cache; | |
892 | boolean_t config_seen; | |
893 | uint64_t best_txg; | |
3bc7e0fb GW |
894 | char *name, *hostname = NULL; |
895 | uint64_t guid; | |
34dc7c2f BB |
896 | uint_t children = 0; |
897 | nvlist_t **child = NULL; | |
428870ff BB |
898 | uint_t holes; |
899 | uint64_t *hole_array, max_id; | |
34dc7c2f BB |
900 | uint_t c; |
901 | boolean_t isactive; | |
902 | uint64_t hostid; | |
903 | nvlist_t *nvl; | |
428870ff | 904 | boolean_t valid_top_config = B_FALSE; |
34dc7c2f BB |
905 | |
906 | if (nvlist_alloc(&ret, 0, 0) != 0) | |
907 | goto nomem; | |
908 | ||
909 | for (pe = pl->pools; pe != NULL; pe = pe->pe_next) { | |
428870ff | 910 | uint64_t id, max_txg = 0; |
34dc7c2f BB |
911 | |
912 | if (nvlist_alloc(&config, NV_UNIQUE_NAME, 0) != 0) | |
913 | goto nomem; | |
914 | config_seen = B_FALSE; | |
915 | ||
916 | /* | |
917 | * Iterate over all toplevel vdevs. Grab the pool configuration | |
918 | * from the first one we find, and then go through the rest and | |
919 | * add them as necessary to the 'vdevs' member of the config. | |
920 | */ | |
921 | for (ve = pe->pe_vdevs; ve != NULL; ve = ve->ve_next) { | |
922 | ||
923 | /* | |
924 | * Determine the best configuration for this vdev by | |
925 | * selecting the config with the latest transaction | |
926 | * group. | |
927 | */ | |
928 | best_txg = 0; | |
929 | for (ce = ve->ve_configs; ce != NULL; | |
930 | ce = ce->ce_next) { | |
931 | ||
932 | if (ce->ce_txg > best_txg) { | |
933 | tmp = ce->ce_config; | |
934 | best_txg = ce->ce_txg; | |
935 | } | |
936 | } | |
937 | ||
428870ff BB |
938 | /* |
939 | * We rely on the fact that the max txg for the | |
940 | * pool will contain the most up-to-date information | |
941 | * about the valid top-levels in the vdev namespace. | |
942 | */ | |
943 | if (best_txg > max_txg) { | |
944 | (void) nvlist_remove(config, | |
945 | ZPOOL_CONFIG_VDEV_CHILDREN, | |
946 | DATA_TYPE_UINT64); | |
947 | (void) nvlist_remove(config, | |
948 | ZPOOL_CONFIG_HOLE_ARRAY, | |
949 | DATA_TYPE_UINT64_ARRAY); | |
950 | ||
951 | max_txg = best_txg; | |
952 | hole_array = NULL; | |
953 | holes = 0; | |
954 | max_id = 0; | |
955 | valid_top_config = B_FALSE; | |
956 | ||
957 | if (nvlist_lookup_uint64(tmp, | |
958 | ZPOOL_CONFIG_VDEV_CHILDREN, &max_id) == 0) { | |
959 | verify(nvlist_add_uint64(config, | |
960 | ZPOOL_CONFIG_VDEV_CHILDREN, | |
961 | max_id) == 0); | |
962 | valid_top_config = B_TRUE; | |
963 | } | |
964 | ||
965 | if (nvlist_lookup_uint64_array(tmp, | |
966 | ZPOOL_CONFIG_HOLE_ARRAY, &hole_array, | |
967 | &holes) == 0) { | |
968 | verify(nvlist_add_uint64_array(config, | |
969 | ZPOOL_CONFIG_HOLE_ARRAY, | |
970 | hole_array, holes) == 0); | |
971 | } | |
972 | } | |
973 | ||
34dc7c2f BB |
974 | if (!config_seen) { |
975 | /* | |
976 | * Copy the relevant pieces of data to the pool | |
977 | * configuration: | |
978 | * | |
979 | * version | |
3bc7e0fb GW |
980 | * pool guid |
981 | * name | |
d96eb2b1 | 982 | * comment (if available) |
3bc7e0fb | 983 | * pool state |
34dc7c2f BB |
984 | * hostid (if available) |
985 | * hostname (if available) | |
986 | */ | |
295304be | 987 | uint64_t state, version; |
3bc7e0fb GW |
988 | char *comment = NULL; |
989 | ||
990 | version = fnvlist_lookup_uint64(tmp, | |
991 | ZPOOL_CONFIG_VERSION); | |
992 | fnvlist_add_uint64(config, | |
993 | ZPOOL_CONFIG_VERSION, version); | |
994 | guid = fnvlist_lookup_uint64(tmp, | |
995 | ZPOOL_CONFIG_POOL_GUID); | |
996 | fnvlist_add_uint64(config, | |
997 | ZPOOL_CONFIG_POOL_GUID, guid); | |
998 | name = fnvlist_lookup_string(tmp, | |
999 | ZPOOL_CONFIG_POOL_NAME); | |
1000 | fnvlist_add_string(config, | |
1001 | ZPOOL_CONFIG_POOL_NAME, name); | |
34dc7c2f | 1002 | |
d96eb2b1 | 1003 | if (nvlist_lookup_string(tmp, |
3bc7e0fb GW |
1004 | ZPOOL_CONFIG_COMMENT, &comment) == 0) |
1005 | fnvlist_add_string(config, | |
1006 | ZPOOL_CONFIG_COMMENT, comment); | |
d96eb2b1 | 1007 | |
3bc7e0fb GW |
1008 | state = fnvlist_lookup_uint64(tmp, |
1009 | ZPOOL_CONFIG_POOL_STATE); | |
1010 | fnvlist_add_uint64(config, | |
1011 | ZPOOL_CONFIG_POOL_STATE, state); | |
d96eb2b1 | 1012 | |
34dc7c2f BB |
1013 | hostid = 0; |
1014 | if (nvlist_lookup_uint64(tmp, | |
1015 | ZPOOL_CONFIG_HOSTID, &hostid) == 0) { | |
3bc7e0fb GW |
1016 | fnvlist_add_uint64(config, |
1017 | ZPOOL_CONFIG_HOSTID, hostid); | |
1018 | hostname = fnvlist_lookup_string(tmp, | |
1019 | ZPOOL_CONFIG_HOSTNAME); | |
1020 | fnvlist_add_string(config, | |
1021 | ZPOOL_CONFIG_HOSTNAME, hostname); | |
34dc7c2f BB |
1022 | } |
1023 | ||
1024 | config_seen = B_TRUE; | |
1025 | } | |
1026 | ||
1027 | /* | |
1028 | * Add this top-level vdev to the child array. | |
1029 | */ | |
1030 | verify(nvlist_lookup_nvlist(tmp, | |
1031 | ZPOOL_CONFIG_VDEV_TREE, &nvtop) == 0); | |
1032 | verify(nvlist_lookup_uint64(nvtop, ZPOOL_CONFIG_ID, | |
1033 | &id) == 0); | |
428870ff | 1034 | |
34dc7c2f BB |
1035 | if (id >= children) { |
1036 | nvlist_t **newchild; | |
1037 | ||
1038 | newchild = zfs_alloc(hdl, (id + 1) * | |
1039 | sizeof (nvlist_t *)); | |
1040 | if (newchild == NULL) | |
1041 | goto nomem; | |
1042 | ||
1043 | for (c = 0; c < children; c++) | |
1044 | newchild[c] = child[c]; | |
1045 | ||
1046 | free(child); | |
1047 | child = newchild; | |
1048 | children = id + 1; | |
1049 | } | |
1050 | if (nvlist_dup(nvtop, &child[id], 0) != 0) | |
1051 | goto nomem; | |
1052 | ||
1053 | } | |
1054 | ||
428870ff BB |
1055 | /* |
1056 | * If we have information about all the top-levels then | |
1057 | * clean up the nvlist which we've constructed. This | |
1058 | * means removing any extraneous devices that are | |
1059 | * beyond the valid range or adding devices to the end | |
1060 | * of our array which appear to be missing. | |
1061 | */ | |
1062 | if (valid_top_config) { | |
1063 | if (max_id < children) { | |
1064 | for (c = max_id; c < children; c++) | |
1065 | nvlist_free(child[c]); | |
1066 | children = max_id; | |
1067 | } else if (max_id > children) { | |
1068 | nvlist_t **newchild; | |
1069 | ||
1070 | newchild = zfs_alloc(hdl, (max_id) * | |
1071 | sizeof (nvlist_t *)); | |
1072 | if (newchild == NULL) | |
1073 | goto nomem; | |
1074 | ||
1075 | for (c = 0; c < children; c++) | |
1076 | newchild[c] = child[c]; | |
1077 | ||
1078 | free(child); | |
1079 | child = newchild; | |
1080 | children = max_id; | |
1081 | } | |
1082 | } | |
1083 | ||
34dc7c2f BB |
1084 | verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID, |
1085 | &guid) == 0); | |
1086 | ||
428870ff BB |
1087 | /* |
1088 | * The vdev namespace may contain holes as a result of | |
1089 | * device removal. We must add them back into the vdev | |
1090 | * tree before we process any missing devices. | |
1091 | */ | |
1092 | if (holes > 0) { | |
1093 | ASSERT(valid_top_config); | |
1094 | ||
1095 | for (c = 0; c < children; c++) { | |
1096 | nvlist_t *holey; | |
1097 | ||
1098 | if (child[c] != NULL || | |
1099 | !vdev_is_hole(hole_array, holes, c)) | |
1100 | continue; | |
1101 | ||
1102 | if (nvlist_alloc(&holey, NV_UNIQUE_NAME, | |
1103 | 0) != 0) | |
1104 | goto nomem; | |
1105 | ||
1106 | /* | |
1107 | * Holes in the namespace are treated as | |
1108 | * "hole" top-level vdevs and have a | |
1109 | * special flag set on them. | |
1110 | */ | |
1111 | if (nvlist_add_string(holey, | |
1112 | ZPOOL_CONFIG_TYPE, | |
1113 | VDEV_TYPE_HOLE) != 0 || | |
1114 | nvlist_add_uint64(holey, | |
1115 | ZPOOL_CONFIG_ID, c) != 0 || | |
1116 | nvlist_add_uint64(holey, | |
0fdd8d64 MT |
1117 | ZPOOL_CONFIG_GUID, 0ULL) != 0) { |
1118 | nvlist_free(holey); | |
428870ff | 1119 | goto nomem; |
0fdd8d64 | 1120 | } |
428870ff BB |
1121 | child[c] = holey; |
1122 | } | |
1123 | } | |
1124 | ||
34dc7c2f BB |
1125 | /* |
1126 | * Look for any missing top-level vdevs. If this is the case, | |
1127 | * create a faked up 'missing' vdev as a placeholder. We cannot | |
1128 | * simply compress the child array, because the kernel performs | |
1129 | * certain checks to make sure the vdev IDs match their location | |
1130 | * in the configuration. | |
1131 | */ | |
428870ff | 1132 | for (c = 0; c < children; c++) { |
34dc7c2f BB |
1133 | if (child[c] == NULL) { |
1134 | nvlist_t *missing; | |
1135 | if (nvlist_alloc(&missing, NV_UNIQUE_NAME, | |
1136 | 0) != 0) | |
1137 | goto nomem; | |
1138 | if (nvlist_add_string(missing, | |
1139 | ZPOOL_CONFIG_TYPE, | |
1140 | VDEV_TYPE_MISSING) != 0 || | |
1141 | nvlist_add_uint64(missing, | |
1142 | ZPOOL_CONFIG_ID, c) != 0 || | |
1143 | nvlist_add_uint64(missing, | |
1144 | ZPOOL_CONFIG_GUID, 0ULL) != 0) { | |
1145 | nvlist_free(missing); | |
1146 | goto nomem; | |
1147 | } | |
1148 | child[c] = missing; | |
1149 | } | |
428870ff | 1150 | } |
34dc7c2f BB |
1151 | |
1152 | /* | |
1153 | * Put all of this pool's top-level vdevs into a root vdev. | |
1154 | */ | |
1155 | if (nvlist_alloc(&nvroot, NV_UNIQUE_NAME, 0) != 0) | |
1156 | goto nomem; | |
1157 | if (nvlist_add_string(nvroot, ZPOOL_CONFIG_TYPE, | |
1158 | VDEV_TYPE_ROOT) != 0 || | |
1159 | nvlist_add_uint64(nvroot, ZPOOL_CONFIG_ID, 0ULL) != 0 || | |
1160 | nvlist_add_uint64(nvroot, ZPOOL_CONFIG_GUID, guid) != 0 || | |
1161 | nvlist_add_nvlist_array(nvroot, ZPOOL_CONFIG_CHILDREN, | |
1162 | child, children) != 0) { | |
1163 | nvlist_free(nvroot); | |
1164 | goto nomem; | |
1165 | } | |
1166 | ||
1167 | for (c = 0; c < children; c++) | |
1168 | nvlist_free(child[c]); | |
1169 | free(child); | |
1170 | children = 0; | |
1171 | child = NULL; | |
1172 | ||
1173 | /* | |
1174 | * Go through and fix up any paths and/or devids based on our | |
1175 | * known list of vdev GUID -> path mappings. | |
1176 | */ | |
1177 | if (fix_paths(nvroot, pl->names) != 0) { | |
1178 | nvlist_free(nvroot); | |
1179 | goto nomem; | |
1180 | } | |
1181 | ||
1182 | /* | |
1183 | * Add the root vdev to this pool's configuration. | |
1184 | */ | |
1185 | if (nvlist_add_nvlist(config, ZPOOL_CONFIG_VDEV_TREE, | |
1186 | nvroot) != 0) { | |
1187 | nvlist_free(nvroot); | |
1188 | goto nomem; | |
1189 | } | |
1190 | nvlist_free(nvroot); | |
1191 | ||
1192 | /* | |
1193 | * zdb uses this path to report on active pools that were | |
1194 | * imported or created using -R. | |
1195 | */ | |
1196 | if (active_ok) | |
1197 | goto add_pool; | |
1198 | ||
1199 | /* | |
1200 | * Determine if this pool is currently active, in which case we | |
1201 | * can't actually import it. | |
1202 | */ | |
1203 | verify(nvlist_lookup_string(config, ZPOOL_CONFIG_POOL_NAME, | |
1204 | &name) == 0); | |
1205 | verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID, | |
1206 | &guid) == 0); | |
1207 | ||
1208 | if (pool_active(hdl, name, guid, &isactive) != 0) | |
1209 | goto error; | |
1210 | ||
1211 | if (isactive) { | |
1212 | nvlist_free(config); | |
1213 | config = NULL; | |
1214 | continue; | |
1215 | } | |
1216 | ||
428870ff BB |
1217 | if ((nvl = refresh_config(hdl, config)) == NULL) { |
1218 | nvlist_free(config); | |
1219 | config = NULL; | |
1220 | continue; | |
1221 | } | |
34dc7c2f BB |
1222 | |
1223 | nvlist_free(config); | |
1224 | config = nvl; | |
1225 | ||
1226 | /* | |
1227 | * Go through and update the paths for spares, now that we have | |
1228 | * them. | |
1229 | */ | |
1230 | verify(nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE, | |
1231 | &nvroot) == 0); | |
1232 | if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_SPARES, | |
1233 | &spares, &nspares) == 0) { | |
1234 | for (i = 0; i < nspares; i++) { | |
1235 | if (fix_paths(spares[i], pl->names) != 0) | |
1236 | goto nomem; | |
1237 | } | |
1238 | } | |
1239 | ||
1240 | /* | |
1241 | * Update the paths for l2cache devices. | |
1242 | */ | |
1243 | if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_L2CACHE, | |
1244 | &l2cache, &nl2cache) == 0) { | |
1245 | for (i = 0; i < nl2cache; i++) { | |
1246 | if (fix_paths(l2cache[i], pl->names) != 0) | |
1247 | goto nomem; | |
1248 | } | |
1249 | } | |
1250 | ||
1251 | /* | |
1252 | * Restore the original information read from the actual label. | |
1253 | */ | |
1254 | (void) nvlist_remove(config, ZPOOL_CONFIG_HOSTID, | |
1255 | DATA_TYPE_UINT64); | |
1256 | (void) nvlist_remove(config, ZPOOL_CONFIG_HOSTNAME, | |
1257 | DATA_TYPE_STRING); | |
1258 | if (hostid != 0) { | |
1259 | verify(nvlist_add_uint64(config, ZPOOL_CONFIG_HOSTID, | |
1260 | hostid) == 0); | |
1261 | verify(nvlist_add_string(config, ZPOOL_CONFIG_HOSTNAME, | |
1262 | hostname) == 0); | |
1263 | } | |
1264 | ||
1265 | add_pool: | |
1266 | /* | |
1267 | * Add this pool to the list of configs. | |
1268 | */ | |
1269 | verify(nvlist_lookup_string(config, ZPOOL_CONFIG_POOL_NAME, | |
1270 | &name) == 0); | |
1271 | if (nvlist_add_nvlist(ret, name, config) != 0) | |
1272 | goto nomem; | |
1273 | ||
1274 | nvlist_free(config); | |
1275 | config = NULL; | |
1276 | } | |
1277 | ||
1278 | return (ret); | |
1279 | ||
1280 | nomem: | |
1281 | (void) no_memory(hdl); | |
1282 | error: | |
1283 | nvlist_free(config); | |
1284 | nvlist_free(ret); | |
1285 | for (c = 0; c < children; c++) | |
1286 | nvlist_free(child[c]); | |
1287 | free(child); | |
1288 | ||
1289 | return (NULL); | |
1290 | } | |
1291 | ||
1292 | /* | |
1293 | * Return the offset of the given label. | |
1294 | */ | |
1295 | static uint64_t | |
1296 | label_offset(uint64_t size, int l) | |
1297 | { | |
1298 | ASSERT(P2PHASE_TYPED(size, sizeof (vdev_label_t), uint64_t) == 0); | |
1299 | return (l * sizeof (vdev_label_t) + (l < VDEV_LABELS / 2 ? | |
1300 | 0 : size - VDEV_LABELS * sizeof (vdev_label_t))); | |
1301 | } | |
1302 | ||
1303 | /* | |
1304 | * Given a file descriptor, read the label information and return an nvlist | |
7d90f569 BB |
1305 | * describing the configuration, if there is one. The number of valid |
1306 | * labels found will be returned in num_labels when non-NULL. | |
34dc7c2f BB |
1307 | */ |
1308 | int | |
7d90f569 | 1309 | zpool_read_label(int fd, nvlist_t **config, int *num_labels) |
34dc7c2f BB |
1310 | { |
1311 | struct stat64 statbuf; | |
7d90f569 | 1312 | int l, count = 0; |
34dc7c2f | 1313 | vdev_label_t *label; |
7d90f569 BB |
1314 | nvlist_t *expected_config = NULL; |
1315 | uint64_t expected_guid = 0, size; | |
83bf769d | 1316 | int error; |
34dc7c2f BB |
1317 | |
1318 | *config = NULL; | |
1319 | ||
ff3510c1 | 1320 | if (fstat64_blk(fd, &statbuf) == -1) |
34dc7c2f BB |
1321 | return (0); |
1322 | size = P2ALIGN_TYPED(statbuf.st_size, sizeof (vdev_label_t), uint64_t); | |
1323 | ||
83bf769d BB |
1324 | error = posix_memalign((void **)&label, PAGESIZE, sizeof (*label)); |
1325 | if (error) | |
34dc7c2f BB |
1326 | return (-1); |
1327 | ||
1328 | for (l = 0; l < VDEV_LABELS; l++) { | |
7d90f569 BB |
1329 | uint64_t state, guid, txg; |
1330 | ||
b128c09f | 1331 | if (pread64(fd, label, sizeof (vdev_label_t), |
34dc7c2f BB |
1332 | label_offset(size, l)) != sizeof (vdev_label_t)) |
1333 | continue; | |
1334 | ||
1335 | if (nvlist_unpack(label->vl_vdev_phys.vp_nvlist, | |
1336 | sizeof (label->vl_vdev_phys.vp_nvlist), config, 0) != 0) | |
1337 | continue; | |
1338 | ||
7d90f569 BB |
1339 | if (nvlist_lookup_uint64(*config, ZPOOL_CONFIG_GUID, |
1340 | &guid) != 0 || guid == 0) { | |
1341 | nvlist_free(*config); | |
1342 | continue; | |
1343 | } | |
1344 | ||
34dc7c2f BB |
1345 | if (nvlist_lookup_uint64(*config, ZPOOL_CONFIG_POOL_STATE, |
1346 | &state) != 0 || state > POOL_STATE_L2CACHE) { | |
1347 | nvlist_free(*config); | |
1348 | continue; | |
1349 | } | |
1350 | ||
1351 | if (state != POOL_STATE_SPARE && state != POOL_STATE_L2CACHE && | |
1352 | (nvlist_lookup_uint64(*config, ZPOOL_CONFIG_POOL_TXG, | |
1353 | &txg) != 0 || txg == 0)) { | |
1354 | nvlist_free(*config); | |
1355 | continue; | |
1356 | } | |
1357 | ||
7d90f569 BB |
1358 | if (expected_guid) { |
1359 | if (expected_guid == guid) | |
1360 | count++; | |
1361 | ||
1362 | nvlist_free(*config); | |
1363 | } else { | |
1364 | expected_config = *config; | |
1365 | expected_guid = guid; | |
1366 | count++; | |
1367 | } | |
34dc7c2f BB |
1368 | } |
1369 | ||
7d90f569 BB |
1370 | if (num_labels != NULL) |
1371 | *num_labels = count; | |
1372 | ||
34dc7c2f | 1373 | free(label); |
7d90f569 BB |
1374 | *config = expected_config; |
1375 | ||
34dc7c2f BB |
1376 | return (0); |
1377 | } | |
1378 | ||
519129ff | 1379 | typedef struct rdsk_node { |
8a39abaa BB |
1380 | char *rn_name; /* Full path to device */ |
1381 | int rn_order; /* Preferred order (low to high) */ | |
1382 | int rn_num_labels; /* Number of valid labels */ | |
83bf769d | 1383 | uint64_t rn_vdev_guid; /* Expected vdev guid when set */ |
519129ff | 1384 | libzfs_handle_t *rn_hdl; |
8a39abaa | 1385 | nvlist_t *rn_config; /* Label config */ |
519129ff BB |
1386 | avl_tree_t *rn_avl; |
1387 | avl_node_t rn_node; | |
8a39abaa BB |
1388 | kmutex_t *rn_lock; |
1389 | boolean_t rn_labelpaths; | |
519129ff BB |
1390 | } rdsk_node_t; |
1391 | ||
83bf769d BB |
1392 | /* |
1393 | * Sorted by vdev guid and full path to allow for multiple entries with | |
1394 | * the same full path name. This is required because it's possible to | |
1395 | * have multiple block devices with labels that refer to the same | |
1396 | * ZPOOL_CONFIG_PATH yet have different vdev guids. In this case both | |
1397 | * entries need to be added to the cache. Scenarios where this can occur | |
1398 | * include overwritten pool labels, devices which are visible from multiple | |
1399 | * hosts and multipath devices. | |
1400 | */ | |
519129ff BB |
1401 | static int |
1402 | slice_cache_compare(const void *arg1, const void *arg2) | |
1403 | { | |
1404 | const char *nm1 = ((rdsk_node_t *)arg1)->rn_name; | |
1405 | const char *nm2 = ((rdsk_node_t *)arg2)->rn_name; | |
83bf769d BB |
1406 | uint64_t guid1 = ((rdsk_node_t *)arg1)->rn_vdev_guid; |
1407 | uint64_t guid2 = ((rdsk_node_t *)arg2)->rn_vdev_guid; | |
519129ff BB |
1408 | int rv; |
1409 | ||
83bf769d BB |
1410 | rv = AVL_CMP(guid1, guid2); |
1411 | if (rv) | |
1412 | return (rv); | |
519129ff | 1413 | |
83bf769d | 1414 | return (AVL_ISIGN(strcmp(nm1, nm2))); |
519129ff BB |
1415 | } |
1416 | ||
8fc5674c NB |
1417 | static boolean_t |
1418 | is_watchdog_dev(char *dev) | |
1419 | { | |
1420 | /* For 'watchdog' dev */ | |
1421 | if (strcmp(dev, "watchdog") == 0) | |
1422 | return (B_TRUE); | |
1423 | ||
1424 | /* For 'watchdog<digit><whatever> */ | |
1425 | if (strstr(dev, "watchdog") == dev && isdigit(dev[8])) | |
1426 | return (B_TRUE); | |
1427 | ||
1428 | return (B_FALSE); | |
1429 | } | |
1430 | ||
8a39abaa BB |
1431 | static int |
1432 | label_paths_impl(libzfs_handle_t *hdl, nvlist_t *nvroot, uint64_t pool_guid, | |
1433 | uint64_t vdev_guid, char **path, char **devid) | |
1434 | { | |
1435 | nvlist_t **child; | |
1436 | uint_t c, children; | |
1437 | uint64_t guid; | |
1438 | char *val; | |
1439 | int error; | |
1440 | ||
1441 | if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_CHILDREN, | |
1442 | &child, &children) == 0) { | |
1443 | for (c = 0; c < children; c++) { | |
1444 | error = label_paths_impl(hdl, child[c], | |
1445 | pool_guid, vdev_guid, path, devid); | |
1446 | if (error) | |
1447 | return (error); | |
1448 | } | |
1449 | return (0); | |
1450 | } | |
1451 | ||
1452 | if (nvroot == NULL) | |
1453 | return (0); | |
1454 | ||
1455 | error = nvlist_lookup_uint64(nvroot, ZPOOL_CONFIG_GUID, &guid); | |
1456 | if ((error != 0) || (guid != vdev_guid)) | |
1457 | return (0); | |
1458 | ||
1459 | error = nvlist_lookup_string(nvroot, ZPOOL_CONFIG_PATH, &val); | |
1460 | if (error == 0) | |
1461 | *path = val; | |
1462 | ||
1463 | error = nvlist_lookup_string(nvroot, ZPOOL_CONFIG_DEVID, &val); | |
1464 | if (error == 0) | |
1465 | *devid = val; | |
1466 | ||
1467 | return (0); | |
1468 | } | |
1469 | ||
1470 | /* | |
1471 | * Given a disk label fetch the ZPOOL_CONFIG_PATH and ZPOOL_CONFIG_DEVID | |
1472 | * and store these strings as config_path and devid_path respectively. | |
1473 | * The returned pointers are only valid as long as label remains valid. | |
1474 | */ | |
1475 | static int | |
1476 | label_paths(libzfs_handle_t *hdl, nvlist_t *label, char **path, char **devid) | |
1477 | { | |
1478 | nvlist_t *nvroot; | |
1479 | uint64_t pool_guid; | |
1480 | uint64_t vdev_guid; | |
1481 | ||
1482 | *path = NULL; | |
1483 | *devid = NULL; | |
1484 | ||
1485 | if (nvlist_lookup_nvlist(label, ZPOOL_CONFIG_VDEV_TREE, &nvroot) || | |
1486 | nvlist_lookup_uint64(label, ZPOOL_CONFIG_POOL_GUID, &pool_guid) || | |
1487 | nvlist_lookup_uint64(label, ZPOOL_CONFIG_GUID, &vdev_guid)) | |
1488 | return (ENOENT); | |
1489 | ||
1490 | return (label_paths_impl(hdl, nvroot, pool_guid, vdev_guid, path, | |
1491 | devid)); | |
1492 | } | |
1493 | ||
519129ff BB |
1494 | static void |
1495 | zpool_open_func(void *arg) | |
1496 | { | |
1497 | rdsk_node_t *rn = arg; | |
8a39abaa | 1498 | libzfs_handle_t *hdl = rn->rn_hdl; |
519129ff BB |
1499 | struct stat64 statbuf; |
1500 | nvlist_t *config; | |
8a39abaa | 1501 | char *bname, *dupname; |
83bf769d | 1502 | uint64_t vdev_guid = 0; |
8a39abaa | 1503 | int error; |
519129ff BB |
1504 | int num_labels; |
1505 | int fd; | |
1506 | ||
519129ff BB |
1507 | /* |
1508 | * Skip devices with well known prefixes there can be side effects | |
1509 | * when opening devices which need to be avoided. | |
1510 | * | |
519129ff | 1511 | * hpet - High Precision Event Timer |
519129ff BB |
1512 | * watchdog - Watchdog must be closed in a special way. |
1513 | */ | |
8a39abaa BB |
1514 | dupname = zfs_strdup(hdl, rn->rn_name); |
1515 | bname = basename(dupname); | |
1516 | error = ((strcmp(bname, "hpet") == 0) || is_watchdog_dev(bname)); | |
1517 | free(dupname); | |
1518 | if (error) | |
519129ff BB |
1519 | return; |
1520 | ||
1521 | /* | |
1522 | * Ignore failed stats. We only want regular files and block devices. | |
1523 | */ | |
8a39abaa | 1524 | if (stat64(rn->rn_name, &statbuf) != 0 || |
519129ff BB |
1525 | (!S_ISREG(statbuf.st_mode) && !S_ISBLK(statbuf.st_mode))) |
1526 | return; | |
1527 | ||
83bf769d BB |
1528 | /* |
1529 | * Preferentially open using O_DIRECT to bypass the block device | |
1530 | * cache which may be stale for multipath devices. An EINVAL errno | |
1531 | * indicates O_DIRECT is unsupported so fallback to just O_RDONLY. | |
1532 | */ | |
1533 | fd = open(rn->rn_name, O_RDONLY | O_DIRECT); | |
1534 | if ((fd < 0) && (errno == EINVAL)) | |
1535 | fd = open(rn->rn_name, O_RDONLY); | |
1536 | ||
1537 | if (fd < 0) | |
519129ff | 1538 | return; |
8a39abaa | 1539 | |
519129ff | 1540 | /* |
8a39abaa | 1541 | * This file is too small to hold a zpool |
519129ff | 1542 | */ |
83bf769d | 1543 | if (S_ISREG(statbuf.st_mode) && statbuf.st_size < SPA_MINDEVSIZE) { |
519129ff BB |
1544 | (void) close(fd); |
1545 | return; | |
519129ff BB |
1546 | } |
1547 | ||
83bf769d BB |
1548 | error = zpool_read_label(fd, &config, &num_labels); |
1549 | if (error != 0) { | |
519129ff | 1550 | (void) close(fd); |
519129ff BB |
1551 | return; |
1552 | } | |
1553 | ||
1554 | if (num_labels == 0) { | |
1555 | (void) close(fd); | |
1556 | nvlist_free(config); | |
1557 | return; | |
1558 | } | |
1559 | ||
83bf769d BB |
1560 | /* |
1561 | * Check that the vdev is for the expected guid. Additional entries | |
1562 | * are speculatively added based on the paths stored in the labels. | |
1563 | * Entries with valid paths but incorrect guids must be removed. | |
1564 | */ | |
1565 | error = nvlist_lookup_uint64(config, ZPOOL_CONFIG_GUID, &vdev_guid); | |
1566 | if (error || (rn->rn_vdev_guid && rn->rn_vdev_guid != vdev_guid)) { | |
1567 | (void) close(fd); | |
1568 | nvlist_free(config); | |
1569 | return; | |
1570 | } | |
1571 | ||
519129ff BB |
1572 | (void) close(fd); |
1573 | ||
1574 | rn->rn_config = config; | |
1575 | rn->rn_num_labels = num_labels; | |
8a39abaa BB |
1576 | |
1577 | /* | |
1578 | * Add additional entries for paths described by this label. | |
1579 | */ | |
1580 | if (rn->rn_labelpaths) { | |
1581 | char *path = NULL; | |
1582 | char *devid = NULL; | |
1583 | rdsk_node_t *slice; | |
1584 | avl_index_t where; | |
1585 | int error; | |
1586 | ||
1587 | if (label_paths(rn->rn_hdl, rn->rn_config, &path, &devid)) | |
1588 | return; | |
1589 | ||
1590 | /* | |
1591 | * Allow devlinks to stabilize so all paths are available. | |
1592 | */ | |
1593 | zpool_label_disk_wait(rn->rn_name, DISK_LABEL_WAIT); | |
1594 | ||
1595 | if (path != NULL) { | |
1596 | slice = zfs_alloc(hdl, sizeof (rdsk_node_t)); | |
1597 | slice->rn_name = zfs_strdup(hdl, path); | |
83bf769d | 1598 | slice->rn_vdev_guid = vdev_guid; |
8a39abaa BB |
1599 | slice->rn_avl = rn->rn_avl; |
1600 | slice->rn_hdl = hdl; | |
83bf769d | 1601 | slice->rn_order = IMPORT_ORDER_PREFERRED_1; |
8a39abaa BB |
1602 | slice->rn_labelpaths = B_FALSE; |
1603 | mutex_enter(rn->rn_lock); | |
1604 | if (avl_find(rn->rn_avl, slice, &where)) { | |
1605 | mutex_exit(rn->rn_lock); | |
1606 | free(slice->rn_name); | |
1607 | free(slice); | |
1608 | } else { | |
1609 | avl_insert(rn->rn_avl, slice, where); | |
1610 | mutex_exit(rn->rn_lock); | |
1611 | zpool_open_func(slice); | |
1612 | } | |
1613 | } | |
1614 | ||
1615 | if (devid != NULL) { | |
1616 | slice = zfs_alloc(hdl, sizeof (rdsk_node_t)); | |
1617 | error = asprintf(&slice->rn_name, "%s%s", | |
1618 | DEV_BYID_PATH, devid); | |
1619 | if (error == -1) { | |
1620 | free(slice); | |
1621 | return; | |
1622 | } | |
1623 | ||
83bf769d | 1624 | slice->rn_vdev_guid = vdev_guid; |
8a39abaa BB |
1625 | slice->rn_avl = rn->rn_avl; |
1626 | slice->rn_hdl = hdl; | |
83bf769d | 1627 | slice->rn_order = IMPORT_ORDER_PREFERRED_2; |
8a39abaa BB |
1628 | slice->rn_labelpaths = B_FALSE; |
1629 | mutex_enter(rn->rn_lock); | |
1630 | if (avl_find(rn->rn_avl, slice, &where)) { | |
1631 | mutex_exit(rn->rn_lock); | |
1632 | free(slice->rn_name); | |
1633 | free(slice); | |
1634 | } else { | |
1635 | avl_insert(rn->rn_avl, slice, where); | |
1636 | mutex_exit(rn->rn_lock); | |
1637 | zpool_open_func(slice); | |
1638 | } | |
1639 | } | |
1640 | } | |
519129ff BB |
1641 | } |
1642 | ||
51a3ae72 DK |
1643 | /* |
1644 | * Given a file descriptor, clear (zero) the label information. This function | |
131cc95c DK |
1645 | * is used in the appliance stack as part of the ZFS sysevent module and |
1646 | * to implement the "zpool labelclear" command. | |
51a3ae72 DK |
1647 | */ |
1648 | int | |
1649 | zpool_clear_label(int fd) | |
1650 | { | |
1651 | struct stat64 statbuf; | |
1652 | int l; | |
1653 | vdev_label_t *label; | |
1654 | uint64_t size; | |
1655 | ||
1656 | if (fstat64_blk(fd, &statbuf) == -1) | |
1657 | return (0); | |
1658 | size = P2ALIGN_TYPED(statbuf.st_size, sizeof (vdev_label_t), uint64_t); | |
1659 | ||
1660 | if ((label = calloc(sizeof (vdev_label_t), 1)) == NULL) | |
1661 | return (-1); | |
1662 | ||
1663 | for (l = 0; l < VDEV_LABELS; l++) { | |
1664 | if (pwrite64(fd, label, sizeof (vdev_label_t), | |
4def05f8 RY |
1665 | label_offset(size, l)) != sizeof (vdev_label_t)) { |
1666 | free(label); | |
51a3ae72 | 1667 | return (-1); |
4def05f8 | 1668 | } |
51a3ae72 DK |
1669 | } |
1670 | ||
1671 | free(label); | |
1672 | return (0); | |
1673 | } | |
1674 | ||
d603ed6c | 1675 | /* |
8a39abaa BB |
1676 | * Scan a list of directories for zfs devices. |
1677 | */ | |
1678 | static int | |
1679 | zpool_find_import_scan(libzfs_handle_t *hdl, kmutex_t *lock, | |
1680 | avl_tree_t **slice_cache, char **dir, int dirs) | |
1681 | { | |
1682 | avl_tree_t *cache; | |
1683 | rdsk_node_t *slice; | |
1684 | void *cookie; | |
1685 | int i, error; | |
1686 | ||
1687 | *slice_cache = NULL; | |
1688 | cache = zfs_alloc(hdl, sizeof (avl_tree_t)); | |
1689 | avl_create(cache, slice_cache_compare, sizeof (rdsk_node_t), | |
1690 | offsetof(rdsk_node_t, rn_node)); | |
1691 | ||
1692 | for (i = 0; i < dirs; i++) { | |
1693 | char path[MAXPATHLEN]; | |
1694 | struct dirent64 *dp; | |
1695 | DIR *dirp; | |
1696 | ||
1697 | if (realpath(dir[i], path) == NULL) { | |
1698 | error = errno; | |
1699 | if (error == ENOENT) | |
1700 | continue; | |
1701 | ||
1702 | zfs_error_aux(hdl, strerror(error)); | |
1703 | (void) zfs_error_fmt(hdl, EZFS_BADPATH, dgettext( | |
1704 | TEXT_DOMAIN, "cannot resolve path '%s'"), dir[i]); | |
1705 | goto error; | |
1706 | } | |
1707 | ||
1708 | dirp = opendir(path); | |
1709 | if (dirp == NULL) { | |
1710 | error = errno; | |
1711 | zfs_error_aux(hdl, strerror(error)); | |
1712 | (void) zfs_error_fmt(hdl, EZFS_BADPATH, | |
1713 | dgettext(TEXT_DOMAIN, "cannot open '%s'"), path); | |
1714 | goto error; | |
1715 | } | |
1716 | ||
1717 | while ((dp = readdir64(dirp)) != NULL) { | |
1718 | const char *name = dp->d_name; | |
1719 | if (name[0] == '.' && | |
1720 | (name[1] == 0 || (name[1] == '.' && name[2] == 0))) | |
1721 | continue; | |
1722 | ||
1723 | slice = zfs_alloc(hdl, sizeof (rdsk_node_t)); | |
1724 | error = asprintf(&slice->rn_name, "%s/%s", path, name); | |
1725 | if (error == -1) { | |
1726 | free(slice); | |
1727 | continue; | |
1728 | } | |
83bf769d | 1729 | slice->rn_vdev_guid = 0; |
8a39abaa BB |
1730 | slice->rn_lock = lock; |
1731 | slice->rn_avl = cache; | |
1732 | slice->rn_hdl = hdl; | |
83bf769d | 1733 | slice->rn_order = i + IMPORT_ORDER_SCAN_OFFSET; |
8a39abaa BB |
1734 | slice->rn_labelpaths = B_FALSE; |
1735 | mutex_enter(lock); | |
1736 | avl_add(cache, slice); | |
1737 | mutex_exit(lock); | |
1738 | } | |
1739 | ||
1740 | (void) closedir(dirp); | |
1741 | } | |
1742 | ||
1743 | *slice_cache = cache; | |
1744 | return (0); | |
1745 | ||
1746 | error: | |
78867a0a | 1747 | cookie = NULL; |
8a39abaa BB |
1748 | while ((slice = avl_destroy_nodes(cache, &cookie)) != NULL) { |
1749 | free(slice->rn_name); | |
1750 | free(slice); | |
1751 | } | |
1752 | free(cache); | |
1753 | ||
1754 | return (error); | |
1755 | } | |
1756 | ||
1757 | /* | |
1758 | * Use libblkid to quickly enumerate all known zfs devices. | |
d603ed6c | 1759 | */ |
428870ff | 1760 | static int |
8a39abaa BB |
1761 | zpool_find_import_blkid(libzfs_handle_t *hdl, kmutex_t *lock, |
1762 | avl_tree_t **slice_cache) | |
428870ff | 1763 | { |
8a39abaa | 1764 | rdsk_node_t *slice; |
d603ed6c BB |
1765 | blkid_cache cache; |
1766 | blkid_dev_iterate iter; | |
1767 | blkid_dev dev; | |
83bf769d | 1768 | avl_index_t where; |
8a39abaa | 1769 | int error; |
428870ff | 1770 | |
8a39abaa BB |
1771 | *slice_cache = NULL; |
1772 | ||
1773 | error = blkid_get_cache(&cache, NULL); | |
1774 | if (error != 0) | |
1775 | return (error); | |
428870ff | 1776 | |
8a39abaa BB |
1777 | error = blkid_probe_all_new(cache); |
1778 | if (error != 0) { | |
1779 | blkid_put_cache(cache); | |
1780 | return (error); | |
428870ff | 1781 | } |
428870ff | 1782 | |
d603ed6c BB |
1783 | iter = blkid_dev_iterate_begin(cache); |
1784 | if (iter == NULL) { | |
8a39abaa BB |
1785 | blkid_put_cache(cache); |
1786 | return (EINVAL); | |
d603ed6c | 1787 | } |
428870ff | 1788 | |
8a39abaa BB |
1789 | error = blkid_dev_set_search(iter, "TYPE", "zfs_member"); |
1790 | if (error != 0) { | |
1791 | blkid_dev_iterate_end(iter); | |
1792 | blkid_put_cache(cache); | |
1793 | return (error); | |
428870ff | 1794 | } |
428870ff | 1795 | |
8a39abaa BB |
1796 | *slice_cache = zfs_alloc(hdl, sizeof (avl_tree_t)); |
1797 | avl_create(*slice_cache, slice_cache_compare, sizeof (rdsk_node_t), | |
1798 | offsetof(rdsk_node_t, rn_node)); | |
428870ff | 1799 | |
8a39abaa BB |
1800 | while (blkid_dev_next(iter, &dev) == 0) { |
1801 | slice = zfs_alloc(hdl, sizeof (rdsk_node_t)); | |
1802 | slice->rn_name = zfs_strdup(hdl, blkid_dev_devname(dev)); | |
83bf769d | 1803 | slice->rn_vdev_guid = 0; |
8a39abaa BB |
1804 | slice->rn_lock = lock; |
1805 | slice->rn_avl = *slice_cache; | |
1806 | slice->rn_hdl = hdl; | |
8a39abaa | 1807 | slice->rn_labelpaths = B_TRUE; |
83bf769d BB |
1808 | |
1809 | error = zfs_path_order(slice->rn_name, &slice->rn_order); | |
1810 | if (error == 0) | |
1811 | slice->rn_order += IMPORT_ORDER_SCAN_OFFSET; | |
1812 | else | |
1813 | slice->rn_order = IMPORT_ORDER_DEFAULT; | |
1814 | ||
8a39abaa | 1815 | mutex_enter(lock); |
83bf769d BB |
1816 | if (avl_find(*slice_cache, slice, &where)) { |
1817 | free(slice->rn_name); | |
1818 | free(slice); | |
1819 | } else { | |
1820 | avl_insert(*slice_cache, slice, where); | |
1821 | } | |
8a39abaa | 1822 | mutex_exit(lock); |
428870ff BB |
1823 | } |
1824 | ||
d603ed6c | 1825 | blkid_dev_iterate_end(iter); |
d603ed6c | 1826 | blkid_put_cache(cache); |
8a39abaa BB |
1827 | |
1828 | return (0); | |
428870ff BB |
1829 | } |
1830 | ||
eac47204 | 1831 | char * |
44867b6d BB |
1832 | zpool_default_import_path[DEFAULT_IMPORT_PATH_SIZE] = { |
1833 | "/dev/disk/by-vdev", /* Custom rules, use first if they exist */ | |
44867b6d | 1834 | "/dev/mapper", /* Use multipath devices before components */ |
95003f70 TC |
1835 | "/dev/disk/by-partlabel", /* Single unique entry set by user */ |
1836 | "/dev/disk/by-partuuid", /* Generated partition uuid */ | |
1837 | "/dev/disk/by-label", /* Custom persistent labels */ | |
44867b6d BB |
1838 | "/dev/disk/by-uuid", /* Single unique entry and persistent */ |
1839 | "/dev/disk/by-id", /* May be multiple entries and persistent */ | |
1840 | "/dev/disk/by-path", /* Encodes physical location and persistent */ | |
44867b6d BB |
1841 | "/dev" /* UNSAFE device names will change */ |
1842 | }; | |
1843 | ||
34dc7c2f BB |
1844 | /* |
1845 | * Given a list of directories to search, find all pools stored on disk. This | |
1846 | * includes partial pools which are not available to import. If no args are | |
1847 | * given (argc is 0), then the default directory (/dev/dsk) is searched. | |
b128c09f BB |
1848 | * poolname or guid (but not both) are provided by the caller when trying |
1849 | * to import a specific pool. | |
34dc7c2f | 1850 | */ |
b128c09f | 1851 | static nvlist_t * |
428870ff | 1852 | zpool_find_import_impl(libzfs_handle_t *hdl, importargs_t *iarg) |
34dc7c2f | 1853 | { |
519129ff | 1854 | nvlist_t *ret = NULL; |
34dc7c2f BB |
1855 | pool_list_t pools = { 0 }; |
1856 | pool_entry_t *pe, *penext; | |
1857 | vdev_entry_t *ve, *venext; | |
1858 | config_entry_t *ce, *cenext; | |
1859 | name_entry_t *ne, *nenext; | |
8a39abaa BB |
1860 | kmutex_t lock; |
1861 | avl_tree_t *cache; | |
519129ff BB |
1862 | rdsk_node_t *slice; |
1863 | void *cookie; | |
8a39abaa | 1864 | taskq_t *t; |
d603ed6c BB |
1865 | |
1866 | verify(iarg->poolname == NULL || iarg->guid == 0); | |
8a39abaa | 1867 | mutex_init(&lock, NULL, MUTEX_DEFAULT, NULL); |
34dc7c2f | 1868 | |
7d11e37e | 1869 | /* |
8a39abaa BB |
1870 | * Locate pool member vdevs using libblkid or by directory scanning. |
1871 | * On success a newly allocated AVL tree which is populated with an | |
1872 | * entry for each discovered vdev will be returned as the cache. | |
1873 | * It's the callers responsibility to consume and destroy this tree. | |
7d11e37e | 1874 | */ |
8a39abaa BB |
1875 | if (iarg->scan || iarg->paths != 0) { |
1876 | int dirs = iarg->paths; | |
1877 | char **dir = iarg->path; | |
d603ed6c | 1878 | |
8a39abaa BB |
1879 | if (dirs == 0) { |
1880 | dir = zpool_default_import_path; | |
1881 | dirs = DEFAULT_IMPORT_PATH_SIZE; | |
1882 | } | |
1883 | ||
1884 | if (zpool_find_import_scan(hdl, &lock, &cache, dir, dirs) != 0) | |
1885 | return (NULL); | |
1886 | } else { | |
1887 | if (zpool_find_import_blkid(hdl, &lock, &cache) != 0) | |
1888 | return (NULL); | |
34dc7c2f BB |
1889 | } |
1890 | ||
1891 | /* | |
8a39abaa BB |
1892 | * Create a thread pool to parallelize the process of reading and |
1893 | * validating labels, a large number of threads can be used due to | |
1894 | * minimal contention. | |
34dc7c2f | 1895 | */ |
8a39abaa BB |
1896 | t = taskq_create("z_import", 2 * boot_ncpus, defclsyspri, |
1897 | 2 * boot_ncpus, INT_MAX, TASKQ_PREPOPULATE); | |
44867b6d | 1898 | |
8a39abaa BB |
1899 | for (slice = avl_first(cache); slice; |
1900 | (slice = avl_walk(cache, slice, AVL_AFTER))) | |
1901 | (void) taskq_dispatch(t, zpool_open_func, slice, TQ_SLEEP); | |
44867b6d | 1902 | |
8a39abaa BB |
1903 | taskq_wait(t); |
1904 | taskq_destroy(t); | |
34dc7c2f | 1905 | |
8a39abaa BB |
1906 | /* |
1907 | * Process the cache filtering out any entries which are not | |
1908 | * for the specificed pool then adding matching label configs. | |
1909 | */ | |
1910 | cookie = NULL; | |
1911 | while ((slice = avl_destroy_nodes(cache, &cookie)) != NULL) { | |
1912 | if (slice->rn_config != NULL) { | |
1913 | nvlist_t *config = slice->rn_config; | |
1914 | boolean_t matched = B_TRUE; | |
c30d8ded | 1915 | int fd; |
8a39abaa BB |
1916 | |
1917 | if (iarg->poolname != NULL) { | |
1918 | char *pname; | |
1919 | ||
1920 | matched = nvlist_lookup_string(config, | |
1921 | ZPOOL_CONFIG_POOL_NAME, &pname) == 0 && | |
1922 | strcmp(iarg->poolname, pname) == 0; | |
1923 | } else if (iarg->guid != 0) { | |
1924 | uint64_t this_guid; | |
1925 | ||
1926 | matched = nvlist_lookup_uint64(config, | |
1927 | ZPOOL_CONFIG_POOL_GUID, &this_guid) == 0 && | |
1928 | iarg->guid == this_guid; | |
1929 | } | |
1930 | if (!matched) { | |
1931 | nvlist_free(config); | |
1932 | } else { | |
c30d8ded BB |
1933 | /* |
1934 | * Verify all remaining entries can be opened | |
1935 | * exclusively. This will prune all underlying | |
1936 | * multipath devices which otherwise could | |
1937 | * result in the vdev appearing as UNAVAIL. | |
1938 | */ | |
1939 | fd = open(slice->rn_name, O_RDONLY | O_EXCL); | |
1940 | if (fd >= 0) { | |
1941 | close(fd); | |
1942 | add_config(hdl, &pools, | |
1943 | slice->rn_name, slice->rn_order, | |
1944 | slice->rn_num_labels, config); | |
1945 | } else { | |
1946 | nvlist_free(config); | |
1947 | } | |
34dc7c2f BB |
1948 | } |
1949 | } | |
8a39abaa BB |
1950 | free(slice->rn_name); |
1951 | free(slice); | |
34dc7c2f | 1952 | } |
8a39abaa BB |
1953 | avl_destroy(cache); |
1954 | free(cache); | |
1955 | mutex_destroy(&lock); | |
34dc7c2f | 1956 | |
428870ff | 1957 | ret = get_configs(hdl, &pools, iarg->can_be_active); |
34dc7c2f | 1958 | |
34dc7c2f BB |
1959 | for (pe = pools.pools; pe != NULL; pe = penext) { |
1960 | penext = pe->pe_next; | |
1961 | for (ve = pe->pe_vdevs; ve != NULL; ve = venext) { | |
1962 | venext = ve->ve_next; | |
1963 | for (ce = ve->ve_configs; ce != NULL; ce = cenext) { | |
1964 | cenext = ce->ce_next; | |
8a5fc748 | 1965 | nvlist_free(ce->ce_config); |
34dc7c2f BB |
1966 | free(ce); |
1967 | } | |
1968 | free(ve); | |
1969 | } | |
1970 | free(pe); | |
1971 | } | |
1972 | ||
1973 | for (ne = pools.names; ne != NULL; ne = nenext) { | |
1974 | nenext = ne->ne_next; | |
0fdd8d64 | 1975 | free(ne->ne_name); |
34dc7c2f BB |
1976 | free(ne); |
1977 | } | |
1978 | ||
34dc7c2f BB |
1979 | return (ret); |
1980 | } | |
1981 | ||
b128c09f BB |
1982 | nvlist_t * |
1983 | zpool_find_import(libzfs_handle_t *hdl, int argc, char **argv) | |
1984 | { | |
428870ff | 1985 | importargs_t iarg = { 0 }; |
b128c09f | 1986 | |
428870ff BB |
1987 | iarg.paths = argc; |
1988 | iarg.path = argv; | |
b128c09f | 1989 | |
428870ff | 1990 | return (zpool_find_import_impl(hdl, &iarg)); |
b128c09f BB |
1991 | } |
1992 | ||
34dc7c2f BB |
1993 | /* |
1994 | * Given a cache file, return the contents as a list of importable pools. | |
b128c09f BB |
1995 | * poolname or guid (but not both) are provided by the caller when trying |
1996 | * to import a specific pool. | |
34dc7c2f BB |
1997 | */ |
1998 | nvlist_t * | |
1999 | zpool_find_import_cached(libzfs_handle_t *hdl, const char *cachefile, | |
b128c09f | 2000 | char *poolname, uint64_t guid) |
34dc7c2f BB |
2001 | { |
2002 | char *buf; | |
2003 | int fd; | |
2004 | struct stat64 statbuf; | |
2005 | nvlist_t *raw, *src, *dst; | |
2006 | nvlist_t *pools; | |
2007 | nvpair_t *elem; | |
2008 | char *name; | |
b128c09f | 2009 | uint64_t this_guid; |
34dc7c2f BB |
2010 | boolean_t active; |
2011 | ||
b128c09f BB |
2012 | verify(poolname == NULL || guid == 0); |
2013 | ||
34dc7c2f BB |
2014 | if ((fd = open(cachefile, O_RDONLY)) < 0) { |
2015 | zfs_error_aux(hdl, "%s", strerror(errno)); | |
2016 | (void) zfs_error(hdl, EZFS_BADCACHE, | |
2017 | dgettext(TEXT_DOMAIN, "failed to open cache file")); | |
2018 | return (NULL); | |
2019 | } | |
2020 | ||
2021 | if (fstat64(fd, &statbuf) != 0) { | |
2022 | zfs_error_aux(hdl, "%s", strerror(errno)); | |
2023 | (void) close(fd); | |
2024 | (void) zfs_error(hdl, EZFS_BADCACHE, | |
2025 | dgettext(TEXT_DOMAIN, "failed to get size of cache file")); | |
2026 | return (NULL); | |
2027 | } | |
2028 | ||
2029 | if ((buf = zfs_alloc(hdl, statbuf.st_size)) == NULL) { | |
2030 | (void) close(fd); | |
2031 | return (NULL); | |
2032 | } | |
2033 | ||
2034 | if (read(fd, buf, statbuf.st_size) != statbuf.st_size) { | |
2035 | (void) close(fd); | |
2036 | free(buf); | |
2037 | (void) zfs_error(hdl, EZFS_BADCACHE, | |
2038 | dgettext(TEXT_DOMAIN, | |
2039 | "failed to read cache file contents")); | |
2040 | return (NULL); | |
2041 | } | |
2042 | ||
2043 | (void) close(fd); | |
2044 | ||
2045 | if (nvlist_unpack(buf, statbuf.st_size, &raw, 0) != 0) { | |
2046 | free(buf); | |
2047 | (void) zfs_error(hdl, EZFS_BADCACHE, | |
2048 | dgettext(TEXT_DOMAIN, | |
2049 | "invalid or corrupt cache file contents")); | |
2050 | return (NULL); | |
2051 | } | |
2052 | ||
2053 | free(buf); | |
2054 | ||
2055 | /* | |
2056 | * Go through and get the current state of the pools and refresh their | |
2057 | * state. | |
2058 | */ | |
2059 | if (nvlist_alloc(&pools, 0, 0) != 0) { | |
2060 | (void) no_memory(hdl); | |
2061 | nvlist_free(raw); | |
2062 | return (NULL); | |
2063 | } | |
2064 | ||
2065 | elem = NULL; | |
2066 | while ((elem = nvlist_next_nvpair(raw, elem)) != NULL) { | |
ab2894e6 | 2067 | src = fnvpair_value_nvlist(elem); |
34dc7c2f | 2068 | |
ab2894e6 | 2069 | name = fnvlist_lookup_string(src, ZPOOL_CONFIG_POOL_NAME); |
b128c09f BB |
2070 | if (poolname != NULL && strcmp(poolname, name) != 0) |
2071 | continue; | |
2072 | ||
ab2894e6 MA |
2073 | this_guid = fnvlist_lookup_uint64(src, ZPOOL_CONFIG_POOL_GUID); |
2074 | if (guid != 0 && guid != this_guid) | |
2075 | continue; | |
34dc7c2f | 2076 | |
b128c09f BB |
2077 | if (pool_active(hdl, name, this_guid, &active) != 0) { |
2078 | nvlist_free(raw); | |
2079 | nvlist_free(pools); | |
2080 | return (NULL); | |
2081 | } | |
34dc7c2f | 2082 | |
b128c09f BB |
2083 | if (active) |
2084 | continue; | |
34dc7c2f | 2085 | |
b128c09f BB |
2086 | if ((dst = refresh_config(hdl, src)) == NULL) { |
2087 | nvlist_free(raw); | |
2088 | nvlist_free(pools); | |
2089 | return (NULL); | |
2090 | } | |
34dc7c2f | 2091 | |
b128c09f BB |
2092 | if (nvlist_add_nvlist(pools, nvpair_name(elem), dst) != 0) { |
2093 | (void) no_memory(hdl); | |
34dc7c2f | 2094 | nvlist_free(dst); |
b128c09f BB |
2095 | nvlist_free(raw); |
2096 | nvlist_free(pools); | |
2097 | return (NULL); | |
34dc7c2f | 2098 | } |
b128c09f | 2099 | nvlist_free(dst); |
34dc7c2f BB |
2100 | } |
2101 | ||
2102 | nvlist_free(raw); | |
2103 | return (pools); | |
2104 | } | |
2105 | ||
428870ff BB |
2106 | static int |
2107 | name_or_guid_exists(zpool_handle_t *zhp, void *data) | |
2108 | { | |
2109 | importargs_t *import = data; | |
2110 | int found = 0; | |
2111 | ||
2112 | if (import->poolname != NULL) { | |
2113 | char *pool_name; | |
2114 | ||
2115 | verify(nvlist_lookup_string(zhp->zpool_config, | |
2116 | ZPOOL_CONFIG_POOL_NAME, &pool_name) == 0); | |
2117 | if (strcmp(pool_name, import->poolname) == 0) | |
2118 | found = 1; | |
2119 | } else { | |
2120 | uint64_t pool_guid; | |
2121 | ||
2122 | verify(nvlist_lookup_uint64(zhp->zpool_config, | |
2123 | ZPOOL_CONFIG_POOL_GUID, &pool_guid) == 0); | |
2124 | if (pool_guid == import->guid) | |
2125 | found = 1; | |
2126 | } | |
2127 | ||
2128 | zpool_close(zhp); | |
2129 | return (found); | |
2130 | } | |
2131 | ||
2132 | nvlist_t * | |
2133 | zpool_search_import(libzfs_handle_t *hdl, importargs_t *import) | |
2134 | { | |
2135 | verify(import->poolname == NULL || import->guid == 0); | |
2136 | ||
2137 | if (import->unique) | |
2138 | import->exists = zpool_iter(hdl, name_or_guid_exists, import); | |
2139 | ||
2140 | if (import->cachefile != NULL) | |
2141 | return (zpool_find_import_cached(hdl, import->cachefile, | |
2142 | import->poolname, import->guid)); | |
2143 | ||
2144 | return (zpool_find_import_impl(hdl, import)); | |
2145 | } | |
34dc7c2f BB |
2146 | |
2147 | boolean_t | |
2148 | find_guid(nvlist_t *nv, uint64_t guid) | |
2149 | { | |
2150 | uint64_t tmp; | |
2151 | nvlist_t **child; | |
2152 | uint_t c, children; | |
2153 | ||
2154 | verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID, &tmp) == 0); | |
2155 | if (tmp == guid) | |
2156 | return (B_TRUE); | |
2157 | ||
2158 | if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN, | |
2159 | &child, &children) == 0) { | |
2160 | for (c = 0; c < children; c++) | |
2161 | if (find_guid(child[c], guid)) | |
2162 | return (B_TRUE); | |
2163 | } | |
2164 | ||
2165 | return (B_FALSE); | |
2166 | } | |
2167 | ||
2168 | typedef struct aux_cbdata { | |
2169 | const char *cb_type; | |
2170 | uint64_t cb_guid; | |
2171 | zpool_handle_t *cb_zhp; | |
2172 | } aux_cbdata_t; | |
2173 | ||
2174 | static int | |
2175 | find_aux(zpool_handle_t *zhp, void *data) | |
2176 | { | |
2177 | aux_cbdata_t *cbp = data; | |
2178 | nvlist_t **list; | |
2179 | uint_t i, count; | |
2180 | uint64_t guid; | |
2181 | nvlist_t *nvroot; | |
2182 | ||
2183 | verify(nvlist_lookup_nvlist(zhp->zpool_config, ZPOOL_CONFIG_VDEV_TREE, | |
2184 | &nvroot) == 0); | |
2185 | ||
2186 | if (nvlist_lookup_nvlist_array(nvroot, cbp->cb_type, | |
2187 | &list, &count) == 0) { | |
2188 | for (i = 0; i < count; i++) { | |
2189 | verify(nvlist_lookup_uint64(list[i], | |
2190 | ZPOOL_CONFIG_GUID, &guid) == 0); | |
2191 | if (guid == cbp->cb_guid) { | |
2192 | cbp->cb_zhp = zhp; | |
2193 | return (1); | |
2194 | } | |
2195 | } | |
2196 | } | |
2197 | ||
2198 | zpool_close(zhp); | |
2199 | return (0); | |
2200 | } | |
2201 | ||
2202 | /* | |
2203 | * Determines if the pool is in use. If so, it returns true and the state of | |
2204 | * the pool as well as the name of the pool. Both strings are allocated and | |
2205 | * must be freed by the caller. | |
2206 | */ | |
2207 | int | |
2208 | zpool_in_use(libzfs_handle_t *hdl, int fd, pool_state_t *state, char **namestr, | |
2209 | boolean_t *inuse) | |
2210 | { | |
2211 | nvlist_t *config; | |
2212 | char *name; | |
2213 | boolean_t ret; | |
2214 | uint64_t guid, vdev_guid; | |
2215 | zpool_handle_t *zhp; | |
2216 | nvlist_t *pool_config; | |
2217 | uint64_t stateval, isspare; | |
2218 | aux_cbdata_t cb = { 0 }; | |
2219 | boolean_t isactive; | |
2220 | ||
2221 | *inuse = B_FALSE; | |
2222 | ||
7d90f569 | 2223 | if (zpool_read_label(fd, &config, NULL) != 0) { |
34dc7c2f BB |
2224 | (void) no_memory(hdl); |
2225 | return (-1); | |
2226 | } | |
2227 | ||
2228 | if (config == NULL) | |
2229 | return (0); | |
2230 | ||
2231 | verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_STATE, | |
2232 | &stateval) == 0); | |
2233 | verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_GUID, | |
2234 | &vdev_guid) == 0); | |
2235 | ||
2236 | if (stateval != POOL_STATE_SPARE && stateval != POOL_STATE_L2CACHE) { | |
2237 | verify(nvlist_lookup_string(config, ZPOOL_CONFIG_POOL_NAME, | |
2238 | &name) == 0); | |
2239 | verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID, | |
2240 | &guid) == 0); | |
2241 | } | |
2242 | ||
2243 | switch (stateval) { | |
2244 | case POOL_STATE_EXPORTED: | |
572e2857 BB |
2245 | /* |
2246 | * A pool with an exported state may in fact be imported | |
2247 | * read-only, so check the in-core state to see if it's | |
2248 | * active and imported read-only. If it is, set | |
2249 | * its state to active. | |
2250 | */ | |
2251 | if (pool_active(hdl, name, guid, &isactive) == 0 && isactive && | |
02f8fe42 JJS |
2252 | (zhp = zpool_open_canfail(hdl, name)) != NULL) { |
2253 | if (zpool_get_prop_int(zhp, ZPOOL_PROP_READONLY, NULL)) | |
2254 | stateval = POOL_STATE_ACTIVE; | |
2255 | ||
2256 | /* | |
2257 | * All we needed the zpool handle for is the | |
2258 | * readonly prop check. | |
2259 | */ | |
2260 | zpool_close(zhp); | |
2261 | } | |
572e2857 | 2262 | |
34dc7c2f BB |
2263 | ret = B_TRUE; |
2264 | break; | |
2265 | ||
2266 | case POOL_STATE_ACTIVE: | |
2267 | /* | |
2268 | * For an active pool, we have to determine if it's really part | |
2269 | * of a currently active pool (in which case the pool will exist | |
2270 | * and the guid will be the same), or whether it's part of an | |
2271 | * active pool that was disconnected without being explicitly | |
2272 | * exported. | |
2273 | */ | |
2274 | if (pool_active(hdl, name, guid, &isactive) != 0) { | |
2275 | nvlist_free(config); | |
2276 | return (-1); | |
2277 | } | |
2278 | ||
2279 | if (isactive) { | |
2280 | /* | |
2281 | * Because the device may have been removed while | |
2282 | * offlined, we only report it as active if the vdev is | |
2283 | * still present in the config. Otherwise, pretend like | |
2284 | * it's not in use. | |
2285 | */ | |
2286 | if ((zhp = zpool_open_canfail(hdl, name)) != NULL && | |
2287 | (pool_config = zpool_get_config(zhp, NULL)) | |
2288 | != NULL) { | |
2289 | nvlist_t *nvroot; | |
2290 | ||
2291 | verify(nvlist_lookup_nvlist(pool_config, | |
2292 | ZPOOL_CONFIG_VDEV_TREE, &nvroot) == 0); | |
2293 | ret = find_guid(nvroot, vdev_guid); | |
2294 | } else { | |
2295 | ret = B_FALSE; | |
2296 | } | |
2297 | ||
2298 | /* | |
2299 | * If this is an active spare within another pool, we | |
2300 | * treat it like an unused hot spare. This allows the | |
2301 | * user to create a pool with a hot spare that currently | |
2302 | * in use within another pool. Since we return B_TRUE, | |
2303 | * libdiskmgt will continue to prevent generic consumers | |
2304 | * from using the device. | |
2305 | */ | |
2306 | if (ret && nvlist_lookup_uint64(config, | |
2307 | ZPOOL_CONFIG_IS_SPARE, &isspare) == 0 && isspare) | |
2308 | stateval = POOL_STATE_SPARE; | |
2309 | ||
2310 | if (zhp != NULL) | |
2311 | zpool_close(zhp); | |
2312 | } else { | |
2313 | stateval = POOL_STATE_POTENTIALLY_ACTIVE; | |
2314 | ret = B_TRUE; | |
2315 | } | |
2316 | break; | |
2317 | ||
2318 | case POOL_STATE_SPARE: | |
2319 | /* | |
2320 | * For a hot spare, it can be either definitively in use, or | |
2321 | * potentially active. To determine if it's in use, we iterate | |
2322 | * over all pools in the system and search for one with a spare | |
2323 | * with a matching guid. | |
2324 | * | |
2325 | * Due to the shared nature of spares, we don't actually report | |
2326 | * the potentially active case as in use. This means the user | |
2327 | * can freely create pools on the hot spares of exported pools, | |
2328 | * but to do otherwise makes the resulting code complicated, and | |
2329 | * we end up having to deal with this case anyway. | |
2330 | */ | |
2331 | cb.cb_zhp = NULL; | |
2332 | cb.cb_guid = vdev_guid; | |
2333 | cb.cb_type = ZPOOL_CONFIG_SPARES; | |
2334 | if (zpool_iter(hdl, find_aux, &cb) == 1) { | |
2335 | name = (char *)zpool_get_name(cb.cb_zhp); | |
0fdd8d64 | 2336 | ret = B_TRUE; |
34dc7c2f | 2337 | } else { |
0fdd8d64 | 2338 | ret = B_FALSE; |
34dc7c2f BB |
2339 | } |
2340 | break; | |
2341 | ||
2342 | case POOL_STATE_L2CACHE: | |
2343 | ||
2344 | /* | |
2345 | * Check if any pool is currently using this l2cache device. | |
2346 | */ | |
2347 | cb.cb_zhp = NULL; | |
2348 | cb.cb_guid = vdev_guid; | |
2349 | cb.cb_type = ZPOOL_CONFIG_L2CACHE; | |
2350 | if (zpool_iter(hdl, find_aux, &cb) == 1) { | |
2351 | name = (char *)zpool_get_name(cb.cb_zhp); | |
0fdd8d64 | 2352 | ret = B_TRUE; |
34dc7c2f | 2353 | } else { |
0fdd8d64 | 2354 | ret = B_FALSE; |
34dc7c2f BB |
2355 | } |
2356 | break; | |
2357 | ||
2358 | default: | |
2359 | ret = B_FALSE; | |
2360 | } | |
2361 | ||
2362 | ||
2363 | if (ret) { | |
2364 | if ((*namestr = zfs_strdup(hdl, name)) == NULL) { | |
2365 | if (cb.cb_zhp) | |
2366 | zpool_close(cb.cb_zhp); | |
2367 | nvlist_free(config); | |
2368 | return (-1); | |
2369 | } | |
2370 | *state = (pool_state_t)stateval; | |
2371 | } | |
2372 | ||
2373 | if (cb.cb_zhp) | |
2374 | zpool_close(cb.cb_zhp); | |
2375 | ||
2376 | nvlist_free(config); | |
2377 | *inuse = ret; | |
2378 | return (0); | |
2379 | } |