]> git.proxmox.com Git - mirror_zfs-debian.git/blob - lib/libzfs/libzfs_pool.c
fd734d8b48ede7af2de5629cedc671c9878dd359
[mirror_zfs-debian.git] / lib / libzfs / libzfs_pool.c
1 /*
2 * CDDL HEADER START
3 *
4 * The contents of this file are subject to the terms of the
5 * Common Development and Distribution License (the "License").
6 * You may not use this file except in compliance with the License.
7 *
8 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
9 * or http://www.opensolaris.org/os/licensing.
10 * See the License for the specific language governing permissions
11 * and limitations under the License.
12 *
13 * When distributing Covered Code, include this CDDL HEADER in each
14 * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
15 * If applicable, add the following below this CDDL HEADER, with the
16 * fields enclosed by brackets "[]" replaced with your own identifying
17 * information: Portions Copyright [yyyy] [name of copyright owner]
18 *
19 * CDDL HEADER END
20 */
21
22 /*
23 * Copyright 2009 Sun Microsystems, Inc. All rights reserved.
24 * Use is subject to license terms.
25 */
26
27 #include <alloca.h>
28 #include <assert.h>
29 #include <ctype.h>
30 #include <errno.h>
31 #include <devid.h>
32 #include <dirent.h>
33 #include <fcntl.h>
34 #include <libintl.h>
35 #include <stdio.h>
36 #include <stdlib.h>
37 #include <strings.h>
38 #include <unistd.h>
39 #include <zone.h>
40 #include <sys/efi_partition.h>
41 #include <sys/vtoc.h>
42 #include <sys/zfs_ioctl.h>
43 #include <sys/zio.h>
44 #include <strings.h>
45 #include <dlfcn.h>
46
47 #include "zfs_namecheck.h"
48 #include "zfs_prop.h"
49 #include "libzfs_impl.h"
50
51 static int read_efi_label(nvlist_t *config, diskaddr_t *sb);
52
53 #if defined(__i386) || defined(__amd64)
54 #define BOOTCMD "installgrub(1M)"
55 #else
56 #define BOOTCMD "installboot(1M)"
57 #endif
58
59 #define DISK_ROOT "/dev/dsk"
60 #define RDISK_ROOT "/dev/rdsk"
61 #define BACKUP_SLICE "s2"
62
63 /*
64 * ====================================================================
65 * zpool property functions
66 * ====================================================================
67 */
68
69 static int
70 zpool_get_all_props(zpool_handle_t *zhp)
71 {
72 zfs_cmd_t zc = { 0 };
73 libzfs_handle_t *hdl = zhp->zpool_hdl;
74
75 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
76
77 if (zcmd_alloc_dst_nvlist(hdl, &zc, 0) != 0)
78 return (-1);
79
80 while (ioctl(hdl->libzfs_fd, ZFS_IOC_POOL_GET_PROPS, &zc) != 0) {
81 if (errno == ENOMEM) {
82 if (zcmd_expand_dst_nvlist(hdl, &zc) != 0) {
83 zcmd_free_nvlists(&zc);
84 return (-1);
85 }
86 } else {
87 zcmd_free_nvlists(&zc);
88 return (-1);
89 }
90 }
91
92 if (zcmd_read_dst_nvlist(hdl, &zc, &zhp->zpool_props) != 0) {
93 zcmd_free_nvlists(&zc);
94 return (-1);
95 }
96
97 zcmd_free_nvlists(&zc);
98
99 return (0);
100 }
101
102 static int
103 zpool_props_refresh(zpool_handle_t *zhp)
104 {
105 nvlist_t *old_props;
106
107 old_props = zhp->zpool_props;
108
109 if (zpool_get_all_props(zhp) != 0)
110 return (-1);
111
112 nvlist_free(old_props);
113 return (0);
114 }
115
116 static char *
117 zpool_get_prop_string(zpool_handle_t *zhp, zpool_prop_t prop,
118 zprop_source_t *src)
119 {
120 nvlist_t *nv, *nvl;
121 uint64_t ival;
122 char *value;
123 zprop_source_t source;
124
125 nvl = zhp->zpool_props;
126 if (nvlist_lookup_nvlist(nvl, zpool_prop_to_name(prop), &nv) == 0) {
127 verify(nvlist_lookup_uint64(nv, ZPROP_SOURCE, &ival) == 0);
128 source = ival;
129 verify(nvlist_lookup_string(nv, ZPROP_VALUE, &value) == 0);
130 } else {
131 source = ZPROP_SRC_DEFAULT;
132 if ((value = (char *)zpool_prop_default_string(prop)) == NULL)
133 value = "-";
134 }
135
136 if (src)
137 *src = source;
138
139 return (value);
140 }
141
142 uint64_t
143 zpool_get_prop_int(zpool_handle_t *zhp, zpool_prop_t prop, zprop_source_t *src)
144 {
145 nvlist_t *nv, *nvl;
146 uint64_t value;
147 zprop_source_t source;
148
149 if (zhp->zpool_props == NULL && zpool_get_all_props(zhp)) {
150 /*
151 * zpool_get_all_props() has most likely failed because
152 * the pool is faulted, but if all we need is the top level
153 * vdev's guid then get it from the zhp config nvlist.
154 */
155 if ((prop == ZPOOL_PROP_GUID) &&
156 (nvlist_lookup_nvlist(zhp->zpool_config,
157 ZPOOL_CONFIG_VDEV_TREE, &nv) == 0) &&
158 (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID, &value)
159 == 0)) {
160 return (value);
161 }
162 return (zpool_prop_default_numeric(prop));
163 }
164
165 nvl = zhp->zpool_props;
166 if (nvlist_lookup_nvlist(nvl, zpool_prop_to_name(prop), &nv) == 0) {
167 verify(nvlist_lookup_uint64(nv, ZPROP_SOURCE, &value) == 0);
168 source = value;
169 verify(nvlist_lookup_uint64(nv, ZPROP_VALUE, &value) == 0);
170 } else {
171 source = ZPROP_SRC_DEFAULT;
172 value = zpool_prop_default_numeric(prop);
173 }
174
175 if (src)
176 *src = source;
177
178 return (value);
179 }
180
181 /*
182 * Map VDEV STATE to printed strings.
183 */
184 char *
185 zpool_state_to_name(vdev_state_t state, vdev_aux_t aux)
186 {
187 switch (state) {
188 case VDEV_STATE_CLOSED:
189 case VDEV_STATE_OFFLINE:
190 return (gettext("OFFLINE"));
191 case VDEV_STATE_REMOVED:
192 return (gettext("REMOVED"));
193 case VDEV_STATE_CANT_OPEN:
194 if (aux == VDEV_AUX_CORRUPT_DATA || aux == VDEV_AUX_BAD_LOG)
195 return (gettext("FAULTED"));
196 else
197 return (gettext("UNAVAIL"));
198 case VDEV_STATE_FAULTED:
199 return (gettext("FAULTED"));
200 case VDEV_STATE_DEGRADED:
201 return (gettext("DEGRADED"));
202 case VDEV_STATE_HEALTHY:
203 return (gettext("ONLINE"));
204 }
205
206 return (gettext("UNKNOWN"));
207 }
208
209 /*
210 * Get a zpool property value for 'prop' and return the value in
211 * a pre-allocated buffer.
212 */
213 int
214 zpool_get_prop(zpool_handle_t *zhp, zpool_prop_t prop, char *buf, size_t len,
215 zprop_source_t *srctype)
216 {
217 uint64_t intval;
218 const char *strval;
219 zprop_source_t src = ZPROP_SRC_NONE;
220 nvlist_t *nvroot;
221 vdev_stat_t *vs;
222 uint_t vsc;
223
224 if (zpool_get_state(zhp) == POOL_STATE_UNAVAIL) {
225 switch (prop) {
226 case ZPOOL_PROP_NAME:
227 (void) strlcpy(buf, zpool_get_name(zhp), len);
228 break;
229
230 case ZPOOL_PROP_HEALTH:
231 (void) strlcpy(buf, "FAULTED", len);
232 break;
233
234 case ZPOOL_PROP_GUID:
235 intval = zpool_get_prop_int(zhp, prop, &src);
236 (void) snprintf(buf, len, "%llu", intval);
237 break;
238
239 case ZPOOL_PROP_ALTROOT:
240 case ZPOOL_PROP_CACHEFILE:
241 if (zhp->zpool_props != NULL ||
242 zpool_get_all_props(zhp) == 0) {
243 (void) strlcpy(buf,
244 zpool_get_prop_string(zhp, prop, &src),
245 len);
246 if (srctype != NULL)
247 *srctype = src;
248 return (0);
249 }
250 /* FALLTHROUGH */
251 default:
252 (void) strlcpy(buf, "-", len);
253 break;
254 }
255
256 if (srctype != NULL)
257 *srctype = src;
258 return (0);
259 }
260
261 if (zhp->zpool_props == NULL && zpool_get_all_props(zhp) &&
262 prop != ZPOOL_PROP_NAME)
263 return (-1);
264
265 switch (zpool_prop_get_type(prop)) {
266 case PROP_TYPE_STRING:
267 (void) strlcpy(buf, zpool_get_prop_string(zhp, prop, &src),
268 len);
269 break;
270
271 case PROP_TYPE_NUMBER:
272 intval = zpool_get_prop_int(zhp, prop, &src);
273
274 switch (prop) {
275 case ZPOOL_PROP_SIZE:
276 case ZPOOL_PROP_USED:
277 case ZPOOL_PROP_AVAILABLE:
278 (void) zfs_nicenum(intval, buf, len);
279 break;
280
281 case ZPOOL_PROP_CAPACITY:
282 (void) snprintf(buf, len, "%llu%%",
283 (u_longlong_t)intval);
284 break;
285
286 case ZPOOL_PROP_HEALTH:
287 verify(nvlist_lookup_nvlist(zpool_get_config(zhp, NULL),
288 ZPOOL_CONFIG_VDEV_TREE, &nvroot) == 0);
289 verify(nvlist_lookup_uint64_array(nvroot,
290 ZPOOL_CONFIG_STATS, (uint64_t **)&vs, &vsc) == 0);
291
292 (void) strlcpy(buf, zpool_state_to_name(intval,
293 vs->vs_aux), len);
294 break;
295 default:
296 (void) snprintf(buf, len, "%llu", intval);
297 }
298 break;
299
300 case PROP_TYPE_INDEX:
301 intval = zpool_get_prop_int(zhp, prop, &src);
302 if (zpool_prop_index_to_string(prop, intval, &strval)
303 != 0)
304 return (-1);
305 (void) strlcpy(buf, strval, len);
306 break;
307
308 default:
309 abort();
310 }
311
312 if (srctype)
313 *srctype = src;
314
315 return (0);
316 }
317
318 /*
319 * Check if the bootfs name has the same pool name as it is set to.
320 * Assuming bootfs is a valid dataset name.
321 */
322 static boolean_t
323 bootfs_name_valid(const char *pool, char *bootfs)
324 {
325 int len = strlen(pool);
326
327 if (!zfs_name_valid(bootfs, ZFS_TYPE_FILESYSTEM|ZFS_TYPE_SNAPSHOT))
328 return (B_FALSE);
329
330 if (strncmp(pool, bootfs, len) == 0 &&
331 (bootfs[len] == '/' || bootfs[len] == '\0'))
332 return (B_TRUE);
333
334 return (B_FALSE);
335 }
336
337 /*
338 * Inspect the configuration to determine if any of the devices contain
339 * an EFI label.
340 */
341 static boolean_t
342 pool_uses_efi(nvlist_t *config)
343 {
344 nvlist_t **child;
345 uint_t c, children;
346
347 if (nvlist_lookup_nvlist_array(config, ZPOOL_CONFIG_CHILDREN,
348 &child, &children) != 0)
349 return (read_efi_label(config, NULL) >= 0);
350
351 for (c = 0; c < children; c++) {
352 if (pool_uses_efi(child[c]))
353 return (B_TRUE);
354 }
355 return (B_FALSE);
356 }
357
358 static boolean_t
359 pool_is_bootable(zpool_handle_t *zhp)
360 {
361 char bootfs[ZPOOL_MAXNAMELEN];
362
363 return (zpool_get_prop(zhp, ZPOOL_PROP_BOOTFS, bootfs,
364 sizeof (bootfs), NULL) == 0 && strncmp(bootfs, "-",
365 sizeof (bootfs)) != 0);
366 }
367
368
369 /*
370 * Given an nvlist of zpool properties to be set, validate that they are
371 * correct, and parse any numeric properties (index, boolean, etc) if they are
372 * specified as strings.
373 */
374 static nvlist_t *
375 zpool_valid_proplist(libzfs_handle_t *hdl, const char *poolname,
376 nvlist_t *props, uint64_t version, boolean_t create_or_import, char *errbuf)
377 {
378 nvpair_t *elem;
379 nvlist_t *retprops;
380 zpool_prop_t prop;
381 char *strval;
382 uint64_t intval;
383 char *slash;
384 struct stat64 statbuf;
385 zpool_handle_t *zhp;
386 nvlist_t *nvroot;
387
388 if (nvlist_alloc(&retprops, NV_UNIQUE_NAME, 0) != 0) {
389 (void) no_memory(hdl);
390 return (NULL);
391 }
392
393 elem = NULL;
394 while ((elem = nvlist_next_nvpair(props, elem)) != NULL) {
395 const char *propname = nvpair_name(elem);
396
397 /*
398 * Make sure this property is valid and applies to this type.
399 */
400 if ((prop = zpool_name_to_prop(propname)) == ZPROP_INVAL) {
401 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
402 "invalid property '%s'"), propname);
403 (void) zfs_error(hdl, EZFS_BADPROP, errbuf);
404 goto error;
405 }
406
407 if (zpool_prop_readonly(prop)) {
408 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "'%s' "
409 "is readonly"), propname);
410 (void) zfs_error(hdl, EZFS_PROPREADONLY, errbuf);
411 goto error;
412 }
413
414 if (zprop_parse_value(hdl, elem, prop, ZFS_TYPE_POOL, retprops,
415 &strval, &intval, errbuf) != 0)
416 goto error;
417
418 /*
419 * Perform additional checking for specific properties.
420 */
421 switch (prop) {
422 case ZPOOL_PROP_VERSION:
423 if (intval < version || intval > SPA_VERSION) {
424 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
425 "property '%s' number %d is invalid."),
426 propname, intval);
427 (void) zfs_error(hdl, EZFS_BADVERSION, errbuf);
428 goto error;
429 }
430 break;
431
432 case ZPOOL_PROP_BOOTFS:
433 if (create_or_import) {
434 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
435 "property '%s' cannot be set at creation "
436 "or import time"), propname);
437 (void) zfs_error(hdl, EZFS_BADPROP, errbuf);
438 goto error;
439 }
440
441 if (version < SPA_VERSION_BOOTFS) {
442 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
443 "pool must be upgraded to support "
444 "'%s' property"), propname);
445 (void) zfs_error(hdl, EZFS_BADVERSION, errbuf);
446 goto error;
447 }
448
449 /*
450 * bootfs property value has to be a dataset name and
451 * the dataset has to be in the same pool as it sets to.
452 */
453 if (strval[0] != '\0' && !bootfs_name_valid(poolname,
454 strval)) {
455 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "'%s' "
456 "is an invalid name"), strval);
457 (void) zfs_error(hdl, EZFS_INVALIDNAME, errbuf);
458 goto error;
459 }
460
461 if ((zhp = zpool_open_canfail(hdl, poolname)) == NULL) {
462 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
463 "could not open pool '%s'"), poolname);
464 (void) zfs_error(hdl, EZFS_OPENFAILED, errbuf);
465 goto error;
466 }
467 verify(nvlist_lookup_nvlist(zpool_get_config(zhp, NULL),
468 ZPOOL_CONFIG_VDEV_TREE, &nvroot) == 0);
469
470 /*
471 * bootfs property cannot be set on a disk which has
472 * been EFI labeled.
473 */
474 if (pool_uses_efi(nvroot)) {
475 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
476 "property '%s' not supported on "
477 "EFI labeled devices"), propname);
478 (void) zfs_error(hdl, EZFS_POOL_NOTSUP, errbuf);
479 zpool_close(zhp);
480 goto error;
481 }
482 zpool_close(zhp);
483 break;
484
485 case ZPOOL_PROP_ALTROOT:
486 if (!create_or_import) {
487 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
488 "property '%s' can only be set during pool "
489 "creation or import"), propname);
490 (void) zfs_error(hdl, EZFS_BADPROP, errbuf);
491 goto error;
492 }
493
494 if (strval[0] != '/') {
495 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
496 "bad alternate root '%s'"), strval);
497 (void) zfs_error(hdl, EZFS_BADPATH, errbuf);
498 goto error;
499 }
500 break;
501
502 case ZPOOL_PROP_CACHEFILE:
503 if (strval[0] == '\0')
504 break;
505
506 if (strcmp(strval, "none") == 0)
507 break;
508
509 if (strval[0] != '/') {
510 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
511 "property '%s' must be empty, an "
512 "absolute path, or 'none'"), propname);
513 (void) zfs_error(hdl, EZFS_BADPATH, errbuf);
514 goto error;
515 }
516
517 slash = strrchr(strval, '/');
518
519 if (slash[1] == '\0' || strcmp(slash, "/.") == 0 ||
520 strcmp(slash, "/..") == 0) {
521 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
522 "'%s' is not a valid file"), strval);
523 (void) zfs_error(hdl, EZFS_BADPATH, errbuf);
524 goto error;
525 }
526
527 *slash = '\0';
528
529 if (strval[0] != '\0' &&
530 (stat64(strval, &statbuf) != 0 ||
531 !S_ISDIR(statbuf.st_mode))) {
532 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
533 "'%s' is not a valid directory"),
534 strval);
535 (void) zfs_error(hdl, EZFS_BADPATH, errbuf);
536 goto error;
537 }
538
539 *slash = '/';
540 break;
541 }
542 }
543
544 return (retprops);
545 error:
546 nvlist_free(retprops);
547 return (NULL);
548 }
549
550 /*
551 * Set zpool property : propname=propval.
552 */
553 int
554 zpool_set_prop(zpool_handle_t *zhp, const char *propname, const char *propval)
555 {
556 zfs_cmd_t zc = { 0 };
557 int ret = -1;
558 char errbuf[1024];
559 nvlist_t *nvl = NULL;
560 nvlist_t *realprops;
561 uint64_t version;
562
563 (void) snprintf(errbuf, sizeof (errbuf),
564 dgettext(TEXT_DOMAIN, "cannot set property for '%s'"),
565 zhp->zpool_name);
566
567 if (nvlist_alloc(&nvl, NV_UNIQUE_NAME, 0) != 0)
568 return (no_memory(zhp->zpool_hdl));
569
570 if (nvlist_add_string(nvl, propname, propval) != 0) {
571 nvlist_free(nvl);
572 return (no_memory(zhp->zpool_hdl));
573 }
574
575 version = zpool_get_prop_int(zhp, ZPOOL_PROP_VERSION, NULL);
576 if ((realprops = zpool_valid_proplist(zhp->zpool_hdl,
577 zhp->zpool_name, nvl, version, B_FALSE, errbuf)) == NULL) {
578 nvlist_free(nvl);
579 return (-1);
580 }
581
582 nvlist_free(nvl);
583 nvl = realprops;
584
585 /*
586 * Execute the corresponding ioctl() to set this property.
587 */
588 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
589
590 if (zcmd_write_src_nvlist(zhp->zpool_hdl, &zc, nvl) != 0) {
591 nvlist_free(nvl);
592 return (-1);
593 }
594
595 ret = zfs_ioctl(zhp->zpool_hdl, ZFS_IOC_POOL_SET_PROPS, &zc);
596
597 zcmd_free_nvlists(&zc);
598 nvlist_free(nvl);
599
600 if (ret)
601 (void) zpool_standard_error(zhp->zpool_hdl, errno, errbuf);
602 else
603 (void) zpool_props_refresh(zhp);
604
605 return (ret);
606 }
607
608 int
609 zpool_expand_proplist(zpool_handle_t *zhp, zprop_list_t **plp)
610 {
611 libzfs_handle_t *hdl = zhp->zpool_hdl;
612 zprop_list_t *entry;
613 char buf[ZFS_MAXPROPLEN];
614
615 if (zprop_expand_list(hdl, plp, ZFS_TYPE_POOL) != 0)
616 return (-1);
617
618 for (entry = *plp; entry != NULL; entry = entry->pl_next) {
619
620 if (entry->pl_fixed)
621 continue;
622
623 if (entry->pl_prop != ZPROP_INVAL &&
624 zpool_get_prop(zhp, entry->pl_prop, buf, sizeof (buf),
625 NULL) == 0) {
626 if (strlen(buf) > entry->pl_width)
627 entry->pl_width = strlen(buf);
628 }
629 }
630
631 return (0);
632 }
633
634
635 /*
636 * Don't start the slice at the default block of 34; many storage
637 * devices will use a stripe width of 128k, so start there instead.
638 */
639 #define NEW_START_BLOCK 256
640
641 /*
642 * Validate the given pool name, optionally putting an extended error message in
643 * 'buf'.
644 */
645 boolean_t
646 zpool_name_valid(libzfs_handle_t *hdl, boolean_t isopen, const char *pool)
647 {
648 namecheck_err_t why;
649 char what;
650 int ret;
651
652 ret = pool_namecheck(pool, &why, &what);
653
654 /*
655 * The rules for reserved pool names were extended at a later point.
656 * But we need to support users with existing pools that may now be
657 * invalid. So we only check for this expanded set of names during a
658 * create (or import), and only in userland.
659 */
660 if (ret == 0 && !isopen &&
661 (strncmp(pool, "mirror", 6) == 0 ||
662 strncmp(pool, "raidz", 5) == 0 ||
663 strncmp(pool, "spare", 5) == 0 ||
664 strcmp(pool, "log") == 0)) {
665 if (hdl != NULL)
666 zfs_error_aux(hdl,
667 dgettext(TEXT_DOMAIN, "name is reserved"));
668 return (B_FALSE);
669 }
670
671
672 if (ret != 0) {
673 if (hdl != NULL) {
674 switch (why) {
675 case NAME_ERR_TOOLONG:
676 zfs_error_aux(hdl,
677 dgettext(TEXT_DOMAIN, "name is too long"));
678 break;
679
680 case NAME_ERR_INVALCHAR:
681 zfs_error_aux(hdl,
682 dgettext(TEXT_DOMAIN, "invalid character "
683 "'%c' in pool name"), what);
684 break;
685
686 case NAME_ERR_NOLETTER:
687 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
688 "name must begin with a letter"));
689 break;
690
691 case NAME_ERR_RESERVED:
692 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
693 "name is reserved"));
694 break;
695
696 case NAME_ERR_DISKLIKE:
697 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
698 "pool name is reserved"));
699 break;
700
701 case NAME_ERR_LEADING_SLASH:
702 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
703 "leading slash in name"));
704 break;
705
706 case NAME_ERR_EMPTY_COMPONENT:
707 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
708 "empty component in name"));
709 break;
710
711 case NAME_ERR_TRAILING_SLASH:
712 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
713 "trailing slash in name"));
714 break;
715
716 case NAME_ERR_MULTIPLE_AT:
717 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
718 "multiple '@' delimiters in name"));
719 break;
720
721 }
722 }
723 return (B_FALSE);
724 }
725
726 return (B_TRUE);
727 }
728
729 /*
730 * Open a handle to the given pool, even if the pool is currently in the FAULTED
731 * state.
732 */
733 zpool_handle_t *
734 zpool_open_canfail(libzfs_handle_t *hdl, const char *pool)
735 {
736 zpool_handle_t *zhp;
737 boolean_t missing;
738
739 /*
740 * Make sure the pool name is valid.
741 */
742 if (!zpool_name_valid(hdl, B_TRUE, pool)) {
743 (void) zfs_error_fmt(hdl, EZFS_INVALIDNAME,
744 dgettext(TEXT_DOMAIN, "cannot open '%s'"),
745 pool);
746 return (NULL);
747 }
748
749 if ((zhp = zfs_alloc(hdl, sizeof (zpool_handle_t))) == NULL)
750 return (NULL);
751
752 zhp->zpool_hdl = hdl;
753 (void) strlcpy(zhp->zpool_name, pool, sizeof (zhp->zpool_name));
754
755 if (zpool_refresh_stats(zhp, &missing) != 0) {
756 zpool_close(zhp);
757 return (NULL);
758 }
759
760 if (missing) {
761 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "no such pool"));
762 (void) zfs_error_fmt(hdl, EZFS_NOENT,
763 dgettext(TEXT_DOMAIN, "cannot open '%s'"), pool);
764 zpool_close(zhp);
765 return (NULL);
766 }
767
768 return (zhp);
769 }
770
771 /*
772 * Like the above, but silent on error. Used when iterating over pools (because
773 * the configuration cache may be out of date).
774 */
775 int
776 zpool_open_silent(libzfs_handle_t *hdl, const char *pool, zpool_handle_t **ret)
777 {
778 zpool_handle_t *zhp;
779 boolean_t missing;
780
781 if ((zhp = zfs_alloc(hdl, sizeof (zpool_handle_t))) == NULL)
782 return (-1);
783
784 zhp->zpool_hdl = hdl;
785 (void) strlcpy(zhp->zpool_name, pool, sizeof (zhp->zpool_name));
786
787 if (zpool_refresh_stats(zhp, &missing) != 0) {
788 zpool_close(zhp);
789 return (-1);
790 }
791
792 if (missing) {
793 zpool_close(zhp);
794 *ret = NULL;
795 return (0);
796 }
797
798 *ret = zhp;
799 return (0);
800 }
801
802 /*
803 * Similar to zpool_open_canfail(), but refuses to open pools in the faulted
804 * state.
805 */
806 zpool_handle_t *
807 zpool_open(libzfs_handle_t *hdl, const char *pool)
808 {
809 zpool_handle_t *zhp;
810
811 if ((zhp = zpool_open_canfail(hdl, pool)) == NULL)
812 return (NULL);
813
814 if (zhp->zpool_state == POOL_STATE_UNAVAIL) {
815 (void) zfs_error_fmt(hdl, EZFS_POOLUNAVAIL,
816 dgettext(TEXT_DOMAIN, "cannot open '%s'"), zhp->zpool_name);
817 zpool_close(zhp);
818 return (NULL);
819 }
820
821 return (zhp);
822 }
823
824 /*
825 * Close the handle. Simply frees the memory associated with the handle.
826 */
827 void
828 zpool_close(zpool_handle_t *zhp)
829 {
830 if (zhp->zpool_config)
831 nvlist_free(zhp->zpool_config);
832 if (zhp->zpool_old_config)
833 nvlist_free(zhp->zpool_old_config);
834 if (zhp->zpool_props)
835 nvlist_free(zhp->zpool_props);
836 free(zhp);
837 }
838
839 /*
840 * Return the name of the pool.
841 */
842 const char *
843 zpool_get_name(zpool_handle_t *zhp)
844 {
845 return (zhp->zpool_name);
846 }
847
848
849 /*
850 * Return the state of the pool (ACTIVE or UNAVAILABLE)
851 */
852 int
853 zpool_get_state(zpool_handle_t *zhp)
854 {
855 return (zhp->zpool_state);
856 }
857
858 /*
859 * Create the named pool, using the provided vdev list. It is assumed
860 * that the consumer has already validated the contents of the nvlist, so we
861 * don't have to worry about error semantics.
862 */
863 int
864 zpool_create(libzfs_handle_t *hdl, const char *pool, nvlist_t *nvroot,
865 nvlist_t *props, nvlist_t *fsprops)
866 {
867 zfs_cmd_t zc = { 0 };
868 nvlist_t *zc_fsprops = NULL;
869 nvlist_t *zc_props = NULL;
870 char msg[1024];
871 char *altroot;
872 int ret = -1;
873
874 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN,
875 "cannot create '%s'"), pool);
876
877 if (!zpool_name_valid(hdl, B_FALSE, pool))
878 return (zfs_error(hdl, EZFS_INVALIDNAME, msg));
879
880 if (zcmd_write_conf_nvlist(hdl, &zc, nvroot) != 0)
881 return (-1);
882
883 if (props) {
884 if ((zc_props = zpool_valid_proplist(hdl, pool, props,
885 SPA_VERSION_1, B_TRUE, msg)) == NULL) {
886 goto create_failed;
887 }
888 }
889
890 if (fsprops) {
891 uint64_t zoned;
892 char *zonestr;
893
894 zoned = ((nvlist_lookup_string(fsprops,
895 zfs_prop_to_name(ZFS_PROP_ZONED), &zonestr) == 0) &&
896 strcmp(zonestr, "on") == 0);
897
898 if ((zc_fsprops = zfs_valid_proplist(hdl,
899 ZFS_TYPE_FILESYSTEM, fsprops, zoned, NULL, msg)) == NULL) {
900 goto create_failed;
901 }
902 if (!zc_props &&
903 (nvlist_alloc(&zc_props, NV_UNIQUE_NAME, 0) != 0)) {
904 goto create_failed;
905 }
906 if (nvlist_add_nvlist(zc_props,
907 ZPOOL_ROOTFS_PROPS, zc_fsprops) != 0) {
908 goto create_failed;
909 }
910 }
911
912 if (zc_props && zcmd_write_src_nvlist(hdl, &zc, zc_props) != 0)
913 goto create_failed;
914
915 (void) strlcpy(zc.zc_name, pool, sizeof (zc.zc_name));
916
917 if ((ret = zfs_ioctl(hdl, ZFS_IOC_POOL_CREATE, &zc)) != 0) {
918
919 zcmd_free_nvlists(&zc);
920 nvlist_free(zc_props);
921 nvlist_free(zc_fsprops);
922
923 switch (errno) {
924 case EBUSY:
925 /*
926 * This can happen if the user has specified the same
927 * device multiple times. We can't reliably detect this
928 * until we try to add it and see we already have a
929 * label.
930 */
931 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
932 "one or more vdevs refer to the same device"));
933 return (zfs_error(hdl, EZFS_BADDEV, msg));
934
935 case EOVERFLOW:
936 /*
937 * This occurs when one of the devices is below
938 * SPA_MINDEVSIZE. Unfortunately, we can't detect which
939 * device was the problem device since there's no
940 * reliable way to determine device size from userland.
941 */
942 {
943 char buf[64];
944
945 zfs_nicenum(SPA_MINDEVSIZE, buf, sizeof (buf));
946
947 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
948 "one or more devices is less than the "
949 "minimum size (%s)"), buf);
950 }
951 return (zfs_error(hdl, EZFS_BADDEV, msg));
952
953 case ENOSPC:
954 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
955 "one or more devices is out of space"));
956 return (zfs_error(hdl, EZFS_BADDEV, msg));
957
958 case ENOTBLK:
959 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
960 "cache device must be a disk or disk slice"));
961 return (zfs_error(hdl, EZFS_BADDEV, msg));
962
963 default:
964 return (zpool_standard_error(hdl, errno, msg));
965 }
966 }
967
968 /*
969 * If this is an alternate root pool, then we automatically set the
970 * mountpoint of the root dataset to be '/'.
971 */
972 if (nvlist_lookup_string(props, zpool_prop_to_name(ZPOOL_PROP_ALTROOT),
973 &altroot) == 0) {
974 zfs_handle_t *zhp;
975
976 verify((zhp = zfs_open(hdl, pool, ZFS_TYPE_DATASET)) != NULL);
977 verify(zfs_prop_set(zhp, zfs_prop_to_name(ZFS_PROP_MOUNTPOINT),
978 "/") == 0);
979
980 zfs_close(zhp);
981 }
982
983 create_failed:
984 zcmd_free_nvlists(&zc);
985 nvlist_free(zc_props);
986 nvlist_free(zc_fsprops);
987 return (ret);
988 }
989
990 /*
991 * Destroy the given pool. It is up to the caller to ensure that there are no
992 * datasets left in the pool.
993 */
994 int
995 zpool_destroy(zpool_handle_t *zhp)
996 {
997 zfs_cmd_t zc = { 0 };
998 zfs_handle_t *zfp = NULL;
999 libzfs_handle_t *hdl = zhp->zpool_hdl;
1000 char msg[1024];
1001
1002 if (zhp->zpool_state == POOL_STATE_ACTIVE &&
1003 (zfp = zfs_open(zhp->zpool_hdl, zhp->zpool_name,
1004 ZFS_TYPE_FILESYSTEM)) == NULL)
1005 return (-1);
1006
1007 if (zpool_remove_zvol_links(zhp) != 0)
1008 return (-1);
1009
1010 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
1011
1012 if (zfs_ioctl(zhp->zpool_hdl, ZFS_IOC_POOL_DESTROY, &zc) != 0) {
1013 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN,
1014 "cannot destroy '%s'"), zhp->zpool_name);
1015
1016 if (errno == EROFS) {
1017 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1018 "one or more devices is read only"));
1019 (void) zfs_error(hdl, EZFS_BADDEV, msg);
1020 } else {
1021 (void) zpool_standard_error(hdl, errno, msg);
1022 }
1023
1024 if (zfp)
1025 zfs_close(zfp);
1026 return (-1);
1027 }
1028
1029 if (zfp) {
1030 remove_mountpoint(zfp);
1031 zfs_close(zfp);
1032 }
1033
1034 return (0);
1035 }
1036
1037 /*
1038 * Add the given vdevs to the pool. The caller must have already performed the
1039 * necessary verification to ensure that the vdev specification is well-formed.
1040 */
1041 int
1042 zpool_add(zpool_handle_t *zhp, nvlist_t *nvroot)
1043 {
1044 zfs_cmd_t zc = { 0 };
1045 int ret;
1046 libzfs_handle_t *hdl = zhp->zpool_hdl;
1047 char msg[1024];
1048 nvlist_t **spares, **l2cache;
1049 uint_t nspares, nl2cache;
1050
1051 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN,
1052 "cannot add to '%s'"), zhp->zpool_name);
1053
1054 if (zpool_get_prop_int(zhp, ZPOOL_PROP_VERSION, NULL) <
1055 SPA_VERSION_SPARES &&
1056 nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_SPARES,
1057 &spares, &nspares) == 0) {
1058 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "pool must be "
1059 "upgraded to add hot spares"));
1060 return (zfs_error(hdl, EZFS_BADVERSION, msg));
1061 }
1062
1063 if (pool_is_bootable(zhp) && nvlist_lookup_nvlist_array(nvroot,
1064 ZPOOL_CONFIG_SPARES, &spares, &nspares) == 0) {
1065 uint64_t s;
1066
1067 for (s = 0; s < nspares; s++) {
1068 char *path;
1069
1070 if (nvlist_lookup_string(spares[s], ZPOOL_CONFIG_PATH,
1071 &path) == 0 && pool_uses_efi(spares[s])) {
1072 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1073 "device '%s' contains an EFI label and "
1074 "cannot be used on root pools."),
1075 zpool_vdev_name(hdl, NULL, spares[s]));
1076 return (zfs_error(hdl, EZFS_POOL_NOTSUP, msg));
1077 }
1078 }
1079 }
1080
1081 if (zpool_get_prop_int(zhp, ZPOOL_PROP_VERSION, NULL) <
1082 SPA_VERSION_L2CACHE &&
1083 nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_L2CACHE,
1084 &l2cache, &nl2cache) == 0) {
1085 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "pool must be "
1086 "upgraded to add cache devices"));
1087 return (zfs_error(hdl, EZFS_BADVERSION, msg));
1088 }
1089
1090 if (zcmd_write_conf_nvlist(hdl, &zc, nvroot) != 0)
1091 return (-1);
1092 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
1093
1094 if (zfs_ioctl(zhp->zpool_hdl, ZFS_IOC_VDEV_ADD, &zc) != 0) {
1095 switch (errno) {
1096 case EBUSY:
1097 /*
1098 * This can happen if the user has specified the same
1099 * device multiple times. We can't reliably detect this
1100 * until we try to add it and see we already have a
1101 * label.
1102 */
1103 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1104 "one or more vdevs refer to the same device"));
1105 (void) zfs_error(hdl, EZFS_BADDEV, msg);
1106 break;
1107
1108 case EOVERFLOW:
1109 /*
1110 * This occurrs when one of the devices is below
1111 * SPA_MINDEVSIZE. Unfortunately, we can't detect which
1112 * device was the problem device since there's no
1113 * reliable way to determine device size from userland.
1114 */
1115 {
1116 char buf[64];
1117
1118 zfs_nicenum(SPA_MINDEVSIZE, buf, sizeof (buf));
1119
1120 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1121 "device is less than the minimum "
1122 "size (%s)"), buf);
1123 }
1124 (void) zfs_error(hdl, EZFS_BADDEV, msg);
1125 break;
1126
1127 case ENOTSUP:
1128 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1129 "pool must be upgraded to add these vdevs"));
1130 (void) zfs_error(hdl, EZFS_BADVERSION, msg);
1131 break;
1132
1133 case EDOM:
1134 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1135 "root pool can not have multiple vdevs"
1136 " or separate logs"));
1137 (void) zfs_error(hdl, EZFS_POOL_NOTSUP, msg);
1138 break;
1139
1140 case ENOTBLK:
1141 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1142 "cache device must be a disk or disk slice"));
1143 (void) zfs_error(hdl, EZFS_BADDEV, msg);
1144 break;
1145
1146 default:
1147 (void) zpool_standard_error(hdl, errno, msg);
1148 }
1149
1150 ret = -1;
1151 } else {
1152 ret = 0;
1153 }
1154
1155 zcmd_free_nvlists(&zc);
1156
1157 return (ret);
1158 }
1159
1160 /*
1161 * Exports the pool from the system. The caller must ensure that there are no
1162 * mounted datasets in the pool.
1163 */
1164 int
1165 zpool_export_common(zpool_handle_t *zhp, boolean_t force, boolean_t hardforce)
1166 {
1167 zfs_cmd_t zc = { 0 };
1168 char msg[1024];
1169
1170 if (zpool_remove_zvol_links(zhp) != 0)
1171 return (-1);
1172
1173 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN,
1174 "cannot export '%s'"), zhp->zpool_name);
1175
1176 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
1177 zc.zc_cookie = force;
1178 zc.zc_guid = hardforce;
1179
1180 if (zfs_ioctl(zhp->zpool_hdl, ZFS_IOC_POOL_EXPORT, &zc) != 0) {
1181 switch (errno) {
1182 case EXDEV:
1183 zfs_error_aux(zhp->zpool_hdl, dgettext(TEXT_DOMAIN,
1184 "use '-f' to override the following errors:\n"
1185 "'%s' has an active shared spare which could be"
1186 " used by other pools once '%s' is exported."),
1187 zhp->zpool_name, zhp->zpool_name);
1188 return (zfs_error(zhp->zpool_hdl, EZFS_ACTIVE_SPARE,
1189 msg));
1190 default:
1191 return (zpool_standard_error_fmt(zhp->zpool_hdl, errno,
1192 msg));
1193 }
1194 }
1195
1196 return (0);
1197 }
1198
1199 int
1200 zpool_export(zpool_handle_t *zhp, boolean_t force)
1201 {
1202 return (zpool_export_common(zhp, force, B_FALSE));
1203 }
1204
1205 int
1206 zpool_export_force(zpool_handle_t *zhp)
1207 {
1208 return (zpool_export_common(zhp, B_TRUE, B_TRUE));
1209 }
1210
1211 /*
1212 * zpool_import() is a contracted interface. Should be kept the same
1213 * if possible.
1214 *
1215 * Applications should use zpool_import_props() to import a pool with
1216 * new properties value to be set.
1217 */
1218 int
1219 zpool_import(libzfs_handle_t *hdl, nvlist_t *config, const char *newname,
1220 char *altroot)
1221 {
1222 nvlist_t *props = NULL;
1223 int ret;
1224
1225 if (altroot != NULL) {
1226 if (nvlist_alloc(&props, NV_UNIQUE_NAME, 0) != 0) {
1227 return (zfs_error_fmt(hdl, EZFS_NOMEM,
1228 dgettext(TEXT_DOMAIN, "cannot import '%s'"),
1229 newname));
1230 }
1231
1232 if (nvlist_add_string(props,
1233 zpool_prop_to_name(ZPOOL_PROP_ALTROOT), altroot) != 0 ||
1234 nvlist_add_string(props,
1235 zpool_prop_to_name(ZPOOL_PROP_CACHEFILE), "none") != 0) {
1236 nvlist_free(props);
1237 return (zfs_error_fmt(hdl, EZFS_NOMEM,
1238 dgettext(TEXT_DOMAIN, "cannot import '%s'"),
1239 newname));
1240 }
1241 }
1242
1243 ret = zpool_import_props(hdl, config, newname, props, B_FALSE);
1244 if (props)
1245 nvlist_free(props);
1246 return (ret);
1247 }
1248
1249 /*
1250 * Import the given pool using the known configuration and a list of
1251 * properties to be set. The configuration should have come from
1252 * zpool_find_import(). The 'newname' parameters control whether the pool
1253 * is imported with a different name.
1254 */
1255 int
1256 zpool_import_props(libzfs_handle_t *hdl, nvlist_t *config, const char *newname,
1257 nvlist_t *props, boolean_t importfaulted)
1258 {
1259 zfs_cmd_t zc = { 0 };
1260 char *thename;
1261 char *origname;
1262 int ret;
1263 char errbuf[1024];
1264
1265 verify(nvlist_lookup_string(config, ZPOOL_CONFIG_POOL_NAME,
1266 &origname) == 0);
1267
1268 (void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN,
1269 "cannot import pool '%s'"), origname);
1270
1271 if (newname != NULL) {
1272 if (!zpool_name_valid(hdl, B_FALSE, newname))
1273 return (zfs_error_fmt(hdl, EZFS_INVALIDNAME,
1274 dgettext(TEXT_DOMAIN, "cannot import '%s'"),
1275 newname));
1276 thename = (char *)newname;
1277 } else {
1278 thename = origname;
1279 }
1280
1281 if (props) {
1282 uint64_t version;
1283
1284 verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_VERSION,
1285 &version) == 0);
1286
1287 if ((props = zpool_valid_proplist(hdl, origname,
1288 props, version, B_TRUE, errbuf)) == NULL) {
1289 return (-1);
1290 } else if (zcmd_write_src_nvlist(hdl, &zc, props) != 0) {
1291 nvlist_free(props);
1292 return (-1);
1293 }
1294 }
1295
1296 (void) strlcpy(zc.zc_name, thename, sizeof (zc.zc_name));
1297
1298 verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID,
1299 &zc.zc_guid) == 0);
1300
1301 if (zcmd_write_conf_nvlist(hdl, &zc, config) != 0) {
1302 nvlist_free(props);
1303 return (-1);
1304 }
1305
1306 zc.zc_cookie = (uint64_t)importfaulted;
1307 ret = 0;
1308 if (zfs_ioctl(hdl, ZFS_IOC_POOL_IMPORT, &zc) != 0) {
1309 char desc[1024];
1310 if (newname == NULL)
1311 (void) snprintf(desc, sizeof (desc),
1312 dgettext(TEXT_DOMAIN, "cannot import '%s'"),
1313 thename);
1314 else
1315 (void) snprintf(desc, sizeof (desc),
1316 dgettext(TEXT_DOMAIN, "cannot import '%s' as '%s'"),
1317 origname, thename);
1318
1319 switch (errno) {
1320 case ENOTSUP:
1321 /*
1322 * Unsupported version.
1323 */
1324 (void) zfs_error(hdl, EZFS_BADVERSION, desc);
1325 break;
1326
1327 case EINVAL:
1328 (void) zfs_error(hdl, EZFS_INVALCONFIG, desc);
1329 break;
1330
1331 default:
1332 (void) zpool_standard_error(hdl, errno, desc);
1333 }
1334
1335 ret = -1;
1336 } else {
1337 zpool_handle_t *zhp;
1338
1339 /*
1340 * This should never fail, but play it safe anyway.
1341 */
1342 if (zpool_open_silent(hdl, thename, &zhp) != 0) {
1343 ret = -1;
1344 } else if (zhp != NULL) {
1345 ret = zpool_create_zvol_links(zhp);
1346 zpool_close(zhp);
1347 }
1348
1349 }
1350
1351 zcmd_free_nvlists(&zc);
1352 nvlist_free(props);
1353
1354 return (ret);
1355 }
1356
1357 /*
1358 * Scrub the pool.
1359 */
1360 int
1361 zpool_scrub(zpool_handle_t *zhp, pool_scrub_type_t type)
1362 {
1363 zfs_cmd_t zc = { 0 };
1364 char msg[1024];
1365 libzfs_handle_t *hdl = zhp->zpool_hdl;
1366
1367 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
1368 zc.zc_cookie = type;
1369
1370 if (zfs_ioctl(zhp->zpool_hdl, ZFS_IOC_POOL_SCRUB, &zc) == 0)
1371 return (0);
1372
1373 (void) snprintf(msg, sizeof (msg),
1374 dgettext(TEXT_DOMAIN, "cannot scrub %s"), zc.zc_name);
1375
1376 if (errno == EBUSY)
1377 return (zfs_error(hdl, EZFS_RESILVERING, msg));
1378 else
1379 return (zpool_standard_error(hdl, errno, msg));
1380 }
1381
1382 /*
1383 * Find a vdev that matches the search criteria specified. We use the
1384 * the nvpair name to determine how we should look for the device.
1385 * 'avail_spare' is set to TRUE if the provided guid refers to an AVAIL
1386 * spare; but FALSE if its an INUSE spare.
1387 */
1388 static nvlist_t *
1389 vdev_to_nvlist_iter(nvlist_t *nv, nvlist_t *search, boolean_t *avail_spare,
1390 boolean_t *l2cache, boolean_t *log)
1391 {
1392 uint_t c, children;
1393 nvlist_t **child;
1394 nvlist_t *ret;
1395 uint64_t is_log;
1396 char *srchkey;
1397 nvpair_t *pair = nvlist_next_nvpair(search, NULL);
1398
1399 /* Nothing to look for */
1400 if (search == NULL || pair == NULL)
1401 return (NULL);
1402
1403 /* Obtain the key we will use to search */
1404 srchkey = nvpair_name(pair);
1405
1406 switch (nvpair_type(pair)) {
1407 case DATA_TYPE_UINT64: {
1408 uint64_t srchval, theguid, present;
1409
1410 verify(nvpair_value_uint64(pair, &srchval) == 0);
1411 if (strcmp(srchkey, ZPOOL_CONFIG_GUID) == 0) {
1412 if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_NOT_PRESENT,
1413 &present) == 0) {
1414 /*
1415 * If the device has never been present since
1416 * import, the only reliable way to match the
1417 * vdev is by GUID.
1418 */
1419 verify(nvlist_lookup_uint64(nv,
1420 ZPOOL_CONFIG_GUID, &theguid) == 0);
1421 if (theguid == srchval)
1422 return (nv);
1423 }
1424 }
1425 break;
1426 }
1427
1428 case DATA_TYPE_STRING: {
1429 char *srchval, *val;
1430
1431 verify(nvpair_value_string(pair, &srchval) == 0);
1432 if (nvlist_lookup_string(nv, srchkey, &val) != 0)
1433 break;
1434
1435 /*
1436 * Search for the requested value. We special case the search
1437 * for ZPOOL_CONFIG_PATH when it's a wholedisk. Otherwise,
1438 * all other searches are simple string compares.
1439 */
1440 if (strcmp(srchkey, ZPOOL_CONFIG_PATH) == 0 && val) {
1441 uint64_t wholedisk = 0;
1442
1443 (void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_WHOLE_DISK,
1444 &wholedisk);
1445 if (wholedisk) {
1446 /*
1447 * For whole disks, the internal path has 's0',
1448 * but the path passed in by the user doesn't.
1449 */
1450 if (strlen(srchval) == strlen(val) - 2 &&
1451 strncmp(srchval, val, strlen(srchval)) == 0)
1452 return (nv);
1453 break;
1454 }
1455 }
1456
1457 /*
1458 * Common case
1459 */
1460 if (strcmp(srchval, val) == 0)
1461 return (nv);
1462 break;
1463 }
1464
1465 default:
1466 break;
1467 }
1468
1469 if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN,
1470 &child, &children) != 0)
1471 return (NULL);
1472
1473 for (c = 0; c < children; c++) {
1474 if ((ret = vdev_to_nvlist_iter(child[c], search,
1475 avail_spare, l2cache, NULL)) != NULL) {
1476 /*
1477 * The 'is_log' value is only set for the toplevel
1478 * vdev, not the leaf vdevs. So we always lookup the
1479 * log device from the root of the vdev tree (where
1480 * 'log' is non-NULL).
1481 */
1482 if (log != NULL &&
1483 nvlist_lookup_uint64(child[c],
1484 ZPOOL_CONFIG_IS_LOG, &is_log) == 0 &&
1485 is_log) {
1486 *log = B_TRUE;
1487 }
1488 return (ret);
1489 }
1490 }
1491
1492 if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_SPARES,
1493 &child, &children) == 0) {
1494 for (c = 0; c < children; c++) {
1495 if ((ret = vdev_to_nvlist_iter(child[c], search,
1496 avail_spare, l2cache, NULL)) != NULL) {
1497 *avail_spare = B_TRUE;
1498 return (ret);
1499 }
1500 }
1501 }
1502
1503 if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_L2CACHE,
1504 &child, &children) == 0) {
1505 for (c = 0; c < children; c++) {
1506 if ((ret = vdev_to_nvlist_iter(child[c], search,
1507 avail_spare, l2cache, NULL)) != NULL) {
1508 *l2cache = B_TRUE;
1509 return (ret);
1510 }
1511 }
1512 }
1513
1514 return (NULL);
1515 }
1516
1517 /*
1518 * Given a physical path (minus the "/devices" prefix), find the
1519 * associated vdev.
1520 */
1521 nvlist_t *
1522 zpool_find_vdev_by_physpath(zpool_handle_t *zhp, const char *ppath,
1523 boolean_t *avail_spare, boolean_t *l2cache, boolean_t *log)
1524 {
1525 nvlist_t *search, *nvroot, *ret;
1526
1527 verify(nvlist_alloc(&search, NV_UNIQUE_NAME, KM_SLEEP) == 0);
1528 verify(nvlist_add_string(search, ZPOOL_CONFIG_PHYS_PATH, ppath) == 0);
1529
1530 verify(nvlist_lookup_nvlist(zhp->zpool_config, ZPOOL_CONFIG_VDEV_TREE,
1531 &nvroot) == 0);
1532
1533 *avail_spare = B_FALSE;
1534 ret = vdev_to_nvlist_iter(nvroot, search, avail_spare, l2cache, log);
1535 nvlist_free(search);
1536
1537 return (ret);
1538 }
1539
1540 nvlist_t *
1541 zpool_find_vdev(zpool_handle_t *zhp, const char *path, boolean_t *avail_spare,
1542 boolean_t *l2cache, boolean_t *log)
1543 {
1544 char buf[MAXPATHLEN];
1545 char *end;
1546 nvlist_t *nvroot, *search, *ret;
1547 uint64_t guid;
1548
1549 verify(nvlist_alloc(&search, NV_UNIQUE_NAME, KM_SLEEP) == 0);
1550
1551 guid = strtoull(path, &end, 10);
1552 if (guid != 0 && *end == '\0') {
1553 verify(nvlist_add_uint64(search, ZPOOL_CONFIG_GUID, guid) == 0);
1554 } else if (path[0] != '/') {
1555 (void) snprintf(buf, sizeof (buf), "%s%s", "/dev/dsk/", path);
1556 verify(nvlist_add_string(search, ZPOOL_CONFIG_PATH, buf) == 0);
1557 } else {
1558 verify(nvlist_add_string(search, ZPOOL_CONFIG_PATH, path) == 0);
1559 }
1560
1561 verify(nvlist_lookup_nvlist(zhp->zpool_config, ZPOOL_CONFIG_VDEV_TREE,
1562 &nvroot) == 0);
1563
1564 *avail_spare = B_FALSE;
1565 *l2cache = B_FALSE;
1566 if (log != NULL)
1567 *log = B_FALSE;
1568 ret = vdev_to_nvlist_iter(nvroot, search, avail_spare, l2cache, log);
1569 nvlist_free(search);
1570
1571 return (ret);
1572 }
1573
1574 static int
1575 vdev_online(nvlist_t *nv)
1576 {
1577 uint64_t ival;
1578
1579 if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_OFFLINE, &ival) == 0 ||
1580 nvlist_lookup_uint64(nv, ZPOOL_CONFIG_FAULTED, &ival) == 0 ||
1581 nvlist_lookup_uint64(nv, ZPOOL_CONFIG_REMOVED, &ival) == 0)
1582 return (0);
1583
1584 return (1);
1585 }
1586
1587 /*
1588 * Helper function for zpool_get_physpaths().
1589 */
1590 static int
1591 vdev_get_one_physpath(nvlist_t *config, char *physpath, size_t physpath_size,
1592 size_t *bytes_written)
1593 {
1594 size_t bytes_left, pos, rsz;
1595 char *tmppath;
1596 const char *format;
1597
1598 if (nvlist_lookup_string(config, ZPOOL_CONFIG_PHYS_PATH,
1599 &tmppath) != 0)
1600 return (EZFS_NODEVICE);
1601
1602 pos = *bytes_written;
1603 bytes_left = physpath_size - pos;
1604 format = (pos == 0) ? "%s" : " %s";
1605
1606 rsz = snprintf(physpath + pos, bytes_left, format, tmppath);
1607 *bytes_written += rsz;
1608
1609 if (rsz >= bytes_left) {
1610 /* if physpath was not copied properly, clear it */
1611 if (bytes_left != 0) {
1612 physpath[pos] = 0;
1613 }
1614 return (EZFS_NOSPC);
1615 }
1616 return (0);
1617 }
1618
1619 static int
1620 vdev_get_physpaths(nvlist_t *nv, char *physpath, size_t phypath_size,
1621 size_t *rsz, boolean_t is_spare)
1622 {
1623 char *type;
1624 int ret;
1625
1626 if (nvlist_lookup_string(nv, ZPOOL_CONFIG_TYPE, &type) != 0)
1627 return (EZFS_INVALCONFIG);
1628
1629 if (strcmp(type, VDEV_TYPE_DISK) == 0) {
1630 /*
1631 * An active spare device has ZPOOL_CONFIG_IS_SPARE set.
1632 * For a spare vdev, we only want to boot from the active
1633 * spare device.
1634 */
1635 if (is_spare) {
1636 uint64_t spare = 0;
1637 (void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_IS_SPARE,
1638 &spare);
1639 if (!spare)
1640 return (EZFS_INVALCONFIG);
1641 }
1642
1643 if (vdev_online(nv)) {
1644 if ((ret = vdev_get_one_physpath(nv, physpath,
1645 phypath_size, rsz)) != 0)
1646 return (ret);
1647 }
1648 } else if (strcmp(type, VDEV_TYPE_MIRROR) == 0 ||
1649 strcmp(type, VDEV_TYPE_REPLACING) == 0 ||
1650 (is_spare = (strcmp(type, VDEV_TYPE_SPARE) == 0))) {
1651 nvlist_t **child;
1652 uint_t count;
1653 int i, ret;
1654
1655 if (nvlist_lookup_nvlist_array(nv,
1656 ZPOOL_CONFIG_CHILDREN, &child, &count) != 0)
1657 return (EZFS_INVALCONFIG);
1658
1659 for (i = 0; i < count; i++) {
1660 ret = vdev_get_physpaths(child[i], physpath,
1661 phypath_size, rsz, is_spare);
1662 if (ret == EZFS_NOSPC)
1663 return (ret);
1664 }
1665 }
1666
1667 return (EZFS_POOL_INVALARG);
1668 }
1669
1670 /*
1671 * Get phys_path for a root pool config.
1672 * Return 0 on success; non-zero on failure.
1673 */
1674 static int
1675 zpool_get_config_physpath(nvlist_t *config, char *physpath, size_t phypath_size)
1676 {
1677 size_t rsz;
1678 nvlist_t *vdev_root;
1679 nvlist_t **child;
1680 uint_t count;
1681 char *type;
1682
1683 rsz = 0;
1684
1685 if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE,
1686 &vdev_root) != 0)
1687 return (EZFS_INVALCONFIG);
1688
1689 if (nvlist_lookup_string(vdev_root, ZPOOL_CONFIG_TYPE, &type) != 0 ||
1690 nvlist_lookup_nvlist_array(vdev_root, ZPOOL_CONFIG_CHILDREN,
1691 &child, &count) != 0)
1692 return (EZFS_INVALCONFIG);
1693
1694 /*
1695 * root pool can not have EFI labeled disks and can only have
1696 * a single top-level vdev.
1697 */
1698 if (strcmp(type, VDEV_TYPE_ROOT) != 0 || count != 1 ||
1699 pool_uses_efi(vdev_root))
1700 return (EZFS_POOL_INVALARG);
1701
1702 (void) vdev_get_physpaths(child[0], physpath, phypath_size, &rsz,
1703 B_FALSE);
1704
1705 /* No online devices */
1706 if (rsz == 0)
1707 return (EZFS_NODEVICE);
1708
1709 return (0);
1710 }
1711
1712 /*
1713 * Get phys_path for a root pool
1714 * Return 0 on success; non-zero on failure.
1715 */
1716 int
1717 zpool_get_physpath(zpool_handle_t *zhp, char *physpath, size_t phypath_size)
1718 {
1719 return (zpool_get_config_physpath(zhp->zpool_config, physpath,
1720 phypath_size));
1721 }
1722
1723 /*
1724 * Returns TRUE if the given guid corresponds to the given type.
1725 * This is used to check for hot spares (INUSE or not), and level 2 cache
1726 * devices.
1727 */
1728 static boolean_t
1729 is_guid_type(zpool_handle_t *zhp, uint64_t guid, const char *type)
1730 {
1731 uint64_t target_guid;
1732 nvlist_t *nvroot;
1733 nvlist_t **list;
1734 uint_t count;
1735 int i;
1736
1737 verify(nvlist_lookup_nvlist(zhp->zpool_config, ZPOOL_CONFIG_VDEV_TREE,
1738 &nvroot) == 0);
1739 if (nvlist_lookup_nvlist_array(nvroot, type, &list, &count) == 0) {
1740 for (i = 0; i < count; i++) {
1741 verify(nvlist_lookup_uint64(list[i], ZPOOL_CONFIG_GUID,
1742 &target_guid) == 0);
1743 if (guid == target_guid)
1744 return (B_TRUE);
1745 }
1746 }
1747
1748 return (B_FALSE);
1749 }
1750
1751 /*
1752 * If the device has being dynamically expanded then we need to relabel
1753 * the disk to use the new unallocated space.
1754 */
1755 static int
1756 zpool_relabel_disk(libzfs_handle_t *hdl, const char *name)
1757 {
1758 char path[MAXPATHLEN];
1759 char errbuf[1024];
1760 int fd, error;
1761 int (*_efi_use_whole_disk)(int);
1762
1763 if ((_efi_use_whole_disk = (int (*)(int))dlsym(RTLD_DEFAULT,
1764 "efi_use_whole_disk")) == NULL)
1765 return (-1);
1766
1767 (void) snprintf(path, sizeof (path), "%s/%s", RDISK_ROOT, name);
1768
1769 if ((fd = open(path, O_RDWR | O_NDELAY)) < 0) {
1770 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "cannot "
1771 "relabel '%s': unable to open device"), name);
1772 return (zfs_error(hdl, EZFS_OPENFAILED, errbuf));
1773 }
1774
1775 /*
1776 * It's possible that we might encounter an error if the device
1777 * does not have any unallocated space left. If so, we simply
1778 * ignore that error and continue on.
1779 */
1780 error = _efi_use_whole_disk(fd);
1781 (void) close(fd);
1782 if (error && error != VT_ENOSPC) {
1783 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "cannot "
1784 "relabel '%s': unable to read disk capacity"), name);
1785 return (zfs_error(hdl, EZFS_NOCAP, errbuf));
1786 }
1787 return (0);
1788 }
1789
1790 /*
1791 * Bring the specified vdev online. The 'flags' parameter is a set of the
1792 * ZFS_ONLINE_* flags.
1793 */
1794 int
1795 zpool_vdev_online(zpool_handle_t *zhp, const char *path, int flags,
1796 vdev_state_t *newstate)
1797 {
1798 zfs_cmd_t zc = { 0 };
1799 char msg[1024];
1800 nvlist_t *tgt;
1801 boolean_t avail_spare, l2cache, islog;
1802 libzfs_handle_t *hdl = zhp->zpool_hdl;
1803
1804 if (flags & ZFS_ONLINE_EXPAND) {
1805 (void) snprintf(msg, sizeof (msg),
1806 dgettext(TEXT_DOMAIN, "cannot expand %s"), path);
1807 } else {
1808 (void) snprintf(msg, sizeof (msg),
1809 dgettext(TEXT_DOMAIN, "cannot online %s"), path);
1810 }
1811
1812 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
1813 if ((tgt = zpool_find_vdev(zhp, path, &avail_spare, &l2cache,
1814 &islog)) == NULL)
1815 return (zfs_error(hdl, EZFS_NODEVICE, msg));
1816
1817 verify(nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID, &zc.zc_guid) == 0);
1818
1819 if (avail_spare ||
1820 is_guid_type(zhp, zc.zc_guid, ZPOOL_CONFIG_SPARES) == B_TRUE)
1821 return (zfs_error(hdl, EZFS_ISSPARE, msg));
1822
1823 if (flags & ZFS_ONLINE_EXPAND ||
1824 zpool_get_prop_int(zhp, ZPOOL_PROP_AUTOEXPAND, NULL)) {
1825 char *pathname = NULL;
1826 uint64_t wholedisk = 0;
1827
1828 (void) nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_WHOLE_DISK,
1829 &wholedisk);
1830 verify(nvlist_lookup_string(tgt, ZPOOL_CONFIG_PATH,
1831 &pathname) == 0);
1832
1833 /*
1834 * XXX - L2ARC 1.0 devices can't support expansion.
1835 */
1836 if (l2cache) {
1837 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1838 "cannot expand cache devices"));
1839 return (zfs_error(hdl, EZFS_VDEVNOTSUP, msg));
1840 }
1841
1842 if (wholedisk) {
1843 pathname += strlen(DISK_ROOT) + 1;
1844 (void) zpool_relabel_disk(zhp->zpool_hdl, pathname);
1845 }
1846 }
1847
1848 zc.zc_cookie = VDEV_STATE_ONLINE;
1849 zc.zc_obj = flags;
1850
1851 if (zfs_ioctl(zhp->zpool_hdl, ZFS_IOC_VDEV_SET_STATE, &zc) != 0)
1852 return (zpool_standard_error(hdl, errno, msg));
1853
1854 *newstate = zc.zc_cookie;
1855 return (0);
1856 }
1857
1858 /*
1859 * Take the specified vdev offline
1860 */
1861 int
1862 zpool_vdev_offline(zpool_handle_t *zhp, const char *path, boolean_t istmp)
1863 {
1864 zfs_cmd_t zc = { 0 };
1865 char msg[1024];
1866 nvlist_t *tgt;
1867 boolean_t avail_spare, l2cache;
1868 libzfs_handle_t *hdl = zhp->zpool_hdl;
1869
1870 (void) snprintf(msg, sizeof (msg),
1871 dgettext(TEXT_DOMAIN, "cannot offline %s"), path);
1872
1873 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
1874 if ((tgt = zpool_find_vdev(zhp, path, &avail_spare, &l2cache,
1875 NULL)) == NULL)
1876 return (zfs_error(hdl, EZFS_NODEVICE, msg));
1877
1878 verify(nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID, &zc.zc_guid) == 0);
1879
1880 if (avail_spare ||
1881 is_guid_type(zhp, zc.zc_guid, ZPOOL_CONFIG_SPARES) == B_TRUE)
1882 return (zfs_error(hdl, EZFS_ISSPARE, msg));
1883
1884 zc.zc_cookie = VDEV_STATE_OFFLINE;
1885 zc.zc_obj = istmp ? ZFS_OFFLINE_TEMPORARY : 0;
1886
1887 if (zfs_ioctl(zhp->zpool_hdl, ZFS_IOC_VDEV_SET_STATE, &zc) == 0)
1888 return (0);
1889
1890 switch (errno) {
1891 case EBUSY:
1892
1893 /*
1894 * There are no other replicas of this device.
1895 */
1896 return (zfs_error(hdl, EZFS_NOREPLICAS, msg));
1897
1898 case EEXIST:
1899 /*
1900 * The log device has unplayed logs
1901 */
1902 return (zfs_error(hdl, EZFS_UNPLAYED_LOGS, msg));
1903
1904 default:
1905 return (zpool_standard_error(hdl, errno, msg));
1906 }
1907 }
1908
1909 /*
1910 * Mark the given vdev faulted.
1911 */
1912 int
1913 zpool_vdev_fault(zpool_handle_t *zhp, uint64_t guid)
1914 {
1915 zfs_cmd_t zc = { 0 };
1916 char msg[1024];
1917 libzfs_handle_t *hdl = zhp->zpool_hdl;
1918
1919 (void) snprintf(msg, sizeof (msg),
1920 dgettext(TEXT_DOMAIN, "cannot fault %llu"), guid);
1921
1922 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
1923 zc.zc_guid = guid;
1924 zc.zc_cookie = VDEV_STATE_FAULTED;
1925
1926 if (ioctl(zhp->zpool_hdl->libzfs_fd, ZFS_IOC_VDEV_SET_STATE, &zc) == 0)
1927 return (0);
1928
1929 switch (errno) {
1930 case EBUSY:
1931
1932 /*
1933 * There are no other replicas of this device.
1934 */
1935 return (zfs_error(hdl, EZFS_NOREPLICAS, msg));
1936
1937 default:
1938 return (zpool_standard_error(hdl, errno, msg));
1939 }
1940
1941 }
1942
1943 /*
1944 * Mark the given vdev degraded.
1945 */
1946 int
1947 zpool_vdev_degrade(zpool_handle_t *zhp, uint64_t guid)
1948 {
1949 zfs_cmd_t zc = { 0 };
1950 char msg[1024];
1951 libzfs_handle_t *hdl = zhp->zpool_hdl;
1952
1953 (void) snprintf(msg, sizeof (msg),
1954 dgettext(TEXT_DOMAIN, "cannot degrade %llu"), guid);
1955
1956 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
1957 zc.zc_guid = guid;
1958 zc.zc_cookie = VDEV_STATE_DEGRADED;
1959
1960 if (ioctl(zhp->zpool_hdl->libzfs_fd, ZFS_IOC_VDEV_SET_STATE, &zc) == 0)
1961 return (0);
1962
1963 return (zpool_standard_error(hdl, errno, msg));
1964 }
1965
1966 /*
1967 * Returns TRUE if the given nvlist is a vdev that was originally swapped in as
1968 * a hot spare.
1969 */
1970 static boolean_t
1971 is_replacing_spare(nvlist_t *search, nvlist_t *tgt, int which)
1972 {
1973 nvlist_t **child;
1974 uint_t c, children;
1975 char *type;
1976
1977 if (nvlist_lookup_nvlist_array(search, ZPOOL_CONFIG_CHILDREN, &child,
1978 &children) == 0) {
1979 verify(nvlist_lookup_string(search, ZPOOL_CONFIG_TYPE,
1980 &type) == 0);
1981
1982 if (strcmp(type, VDEV_TYPE_SPARE) == 0 &&
1983 children == 2 && child[which] == tgt)
1984 return (B_TRUE);
1985
1986 for (c = 0; c < children; c++)
1987 if (is_replacing_spare(child[c], tgt, which))
1988 return (B_TRUE);
1989 }
1990
1991 return (B_FALSE);
1992 }
1993
1994 /*
1995 * Attach new_disk (fully described by nvroot) to old_disk.
1996 * If 'replacing' is specified, the new disk will replace the old one.
1997 */
1998 int
1999 zpool_vdev_attach(zpool_handle_t *zhp,
2000 const char *old_disk, const char *new_disk, nvlist_t *nvroot, int replacing)
2001 {
2002 zfs_cmd_t zc = { 0 };
2003 char msg[1024];
2004 int ret;
2005 nvlist_t *tgt;
2006 boolean_t avail_spare, l2cache, islog;
2007 uint64_t val;
2008 char *path, *newname;
2009 nvlist_t **child;
2010 uint_t children;
2011 nvlist_t *config_root;
2012 libzfs_handle_t *hdl = zhp->zpool_hdl;
2013 boolean_t rootpool = pool_is_bootable(zhp);
2014
2015 if (replacing)
2016 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN,
2017 "cannot replace %s with %s"), old_disk, new_disk);
2018 else
2019 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN,
2020 "cannot attach %s to %s"), new_disk, old_disk);
2021
2022 /*
2023 * If this is a root pool, make sure that we're not attaching an
2024 * EFI labeled device.
2025 */
2026 if (rootpool && pool_uses_efi(nvroot)) {
2027 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2028 "EFI labeled devices are not supported on root pools."));
2029 return (zfs_error(hdl, EZFS_POOL_NOTSUP, msg));
2030 }
2031
2032 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
2033 if ((tgt = zpool_find_vdev(zhp, old_disk, &avail_spare, &l2cache,
2034 &islog)) == 0)
2035 return (zfs_error(hdl, EZFS_NODEVICE, msg));
2036
2037 if (avail_spare)
2038 return (zfs_error(hdl, EZFS_ISSPARE, msg));
2039
2040 if (l2cache)
2041 return (zfs_error(hdl, EZFS_ISL2CACHE, msg));
2042
2043 verify(nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID, &zc.zc_guid) == 0);
2044 zc.zc_cookie = replacing;
2045
2046 if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_CHILDREN,
2047 &child, &children) != 0 || children != 1) {
2048 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2049 "new device must be a single disk"));
2050 return (zfs_error(hdl, EZFS_INVALCONFIG, msg));
2051 }
2052
2053 verify(nvlist_lookup_nvlist(zpool_get_config(zhp, NULL),
2054 ZPOOL_CONFIG_VDEV_TREE, &config_root) == 0);
2055
2056 if ((newname = zpool_vdev_name(NULL, NULL, child[0])) == NULL)
2057 return (-1);
2058
2059 /*
2060 * If the target is a hot spare that has been swapped in, we can only
2061 * replace it with another hot spare.
2062 */
2063 if (replacing &&
2064 nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_IS_SPARE, &val) == 0 &&
2065 (zpool_find_vdev(zhp, newname, &avail_spare, &l2cache,
2066 NULL) == NULL || !avail_spare) &&
2067 is_replacing_spare(config_root, tgt, 1)) {
2068 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2069 "can only be replaced by another hot spare"));
2070 free(newname);
2071 return (zfs_error(hdl, EZFS_BADTARGET, msg));
2072 }
2073
2074 /*
2075 * If we are attempting to replace a spare, it canot be applied to an
2076 * already spared device.
2077 */
2078 if (replacing &&
2079 nvlist_lookup_string(child[0], ZPOOL_CONFIG_PATH, &path) == 0 &&
2080 zpool_find_vdev(zhp, newname, &avail_spare,
2081 &l2cache, NULL) != NULL && avail_spare &&
2082 is_replacing_spare(config_root, tgt, 0)) {
2083 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2084 "device has already been replaced with a spare"));
2085 free(newname);
2086 return (zfs_error(hdl, EZFS_BADTARGET, msg));
2087 }
2088
2089 free(newname);
2090
2091 if (zcmd_write_conf_nvlist(hdl, &zc, nvroot) != 0)
2092 return (-1);
2093
2094 ret = zfs_ioctl(zhp->zpool_hdl, ZFS_IOC_VDEV_ATTACH, &zc);
2095
2096 zcmd_free_nvlists(&zc);
2097
2098 if (ret == 0) {
2099 if (rootpool) {
2100 /*
2101 * XXX - This should be removed once we can
2102 * automatically install the bootblocks on the
2103 * newly attached disk.
2104 */
2105 (void) fprintf(stderr, dgettext(TEXT_DOMAIN, "Please "
2106 "be sure to invoke %s to make '%s' bootable.\n"),
2107 BOOTCMD, new_disk);
2108
2109 /*
2110 * XXX need a better way to prevent user from
2111 * booting up a half-baked vdev.
2112 */
2113 (void) fprintf(stderr, dgettext(TEXT_DOMAIN, "Make "
2114 "sure to wait until resilver is done "
2115 "before rebooting.\n"));
2116 }
2117 return (0);
2118 }
2119
2120 switch (errno) {
2121 case ENOTSUP:
2122 /*
2123 * Can't attach to or replace this type of vdev.
2124 */
2125 if (replacing) {
2126 if (islog)
2127 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2128 "cannot replace a log with a spare"));
2129 else
2130 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2131 "cannot replace a replacing device"));
2132 } else {
2133 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2134 "can only attach to mirrors and top-level "
2135 "disks"));
2136 }
2137 (void) zfs_error(hdl, EZFS_BADTARGET, msg);
2138 break;
2139
2140 case EINVAL:
2141 /*
2142 * The new device must be a single disk.
2143 */
2144 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2145 "new device must be a single disk"));
2146 (void) zfs_error(hdl, EZFS_INVALCONFIG, msg);
2147 break;
2148
2149 case EBUSY:
2150 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "%s is busy"),
2151 new_disk);
2152 (void) zfs_error(hdl, EZFS_BADDEV, msg);
2153 break;
2154
2155 case EOVERFLOW:
2156 /*
2157 * The new device is too small.
2158 */
2159 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2160 "device is too small"));
2161 (void) zfs_error(hdl, EZFS_BADDEV, msg);
2162 break;
2163
2164 case EDOM:
2165 /*
2166 * The new device has a different alignment requirement.
2167 */
2168 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2169 "devices have different sector alignment"));
2170 (void) zfs_error(hdl, EZFS_BADDEV, msg);
2171 break;
2172
2173 case ENAMETOOLONG:
2174 /*
2175 * The resulting top-level vdev spec won't fit in the label.
2176 */
2177 (void) zfs_error(hdl, EZFS_DEVOVERFLOW, msg);
2178 break;
2179
2180 default:
2181 (void) zpool_standard_error(hdl, errno, msg);
2182 }
2183
2184 return (-1);
2185 }
2186
2187 /*
2188 * Detach the specified device.
2189 */
2190 int
2191 zpool_vdev_detach(zpool_handle_t *zhp, const char *path)
2192 {
2193 zfs_cmd_t zc = { 0 };
2194 char msg[1024];
2195 nvlist_t *tgt;
2196 boolean_t avail_spare, l2cache;
2197 libzfs_handle_t *hdl = zhp->zpool_hdl;
2198
2199 (void) snprintf(msg, sizeof (msg),
2200 dgettext(TEXT_DOMAIN, "cannot detach %s"), path);
2201
2202 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
2203 if ((tgt = zpool_find_vdev(zhp, path, &avail_spare, &l2cache,
2204 NULL)) == 0)
2205 return (zfs_error(hdl, EZFS_NODEVICE, msg));
2206
2207 if (avail_spare)
2208 return (zfs_error(hdl, EZFS_ISSPARE, msg));
2209
2210 if (l2cache)
2211 return (zfs_error(hdl, EZFS_ISL2CACHE, msg));
2212
2213 verify(nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID, &zc.zc_guid) == 0);
2214
2215 if (zfs_ioctl(hdl, ZFS_IOC_VDEV_DETACH, &zc) == 0)
2216 return (0);
2217
2218 switch (errno) {
2219
2220 case ENOTSUP:
2221 /*
2222 * Can't detach from this type of vdev.
2223 */
2224 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "only "
2225 "applicable to mirror and replacing vdevs"));
2226 (void) zfs_error(zhp->zpool_hdl, EZFS_BADTARGET, msg);
2227 break;
2228
2229 case EBUSY:
2230 /*
2231 * There are no other replicas of this device.
2232 */
2233 (void) zfs_error(hdl, EZFS_NOREPLICAS, msg);
2234 break;
2235
2236 default:
2237 (void) zpool_standard_error(hdl, errno, msg);
2238 }
2239
2240 return (-1);
2241 }
2242
2243 /*
2244 * Remove the given device. Currently, this is supported only for hot spares
2245 * and level 2 cache devices.
2246 */
2247 int
2248 zpool_vdev_remove(zpool_handle_t *zhp, const char *path)
2249 {
2250 zfs_cmd_t zc = { 0 };
2251 char msg[1024];
2252 nvlist_t *tgt;
2253 boolean_t avail_spare, l2cache;
2254 libzfs_handle_t *hdl = zhp->zpool_hdl;
2255
2256 (void) snprintf(msg, sizeof (msg),
2257 dgettext(TEXT_DOMAIN, "cannot remove %s"), path);
2258
2259 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
2260 if ((tgt = zpool_find_vdev(zhp, path, &avail_spare, &l2cache,
2261 NULL)) == 0)
2262 return (zfs_error(hdl, EZFS_NODEVICE, msg));
2263
2264 if (!avail_spare && !l2cache) {
2265 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2266 "only inactive hot spares or cache devices "
2267 "can be removed"));
2268 return (zfs_error(hdl, EZFS_NODEVICE, msg));
2269 }
2270
2271 verify(nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID, &zc.zc_guid) == 0);
2272
2273 if (zfs_ioctl(hdl, ZFS_IOC_VDEV_REMOVE, &zc) == 0)
2274 return (0);
2275
2276 return (zpool_standard_error(hdl, errno, msg));
2277 }
2278
2279 /*
2280 * Clear the errors for the pool, or the particular device if specified.
2281 */
2282 int
2283 zpool_clear(zpool_handle_t *zhp, const char *path)
2284 {
2285 zfs_cmd_t zc = { 0 };
2286 char msg[1024];
2287 nvlist_t *tgt;
2288 boolean_t avail_spare, l2cache;
2289 libzfs_handle_t *hdl = zhp->zpool_hdl;
2290
2291 if (path)
2292 (void) snprintf(msg, sizeof (msg),
2293 dgettext(TEXT_DOMAIN, "cannot clear errors for %s"),
2294 path);
2295 else
2296 (void) snprintf(msg, sizeof (msg),
2297 dgettext(TEXT_DOMAIN, "cannot clear errors for %s"),
2298 zhp->zpool_name);
2299
2300 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
2301 if (path) {
2302 if ((tgt = zpool_find_vdev(zhp, path, &avail_spare,
2303 &l2cache, NULL)) == 0)
2304 return (zfs_error(hdl, EZFS_NODEVICE, msg));
2305
2306 /*
2307 * Don't allow error clearing for hot spares. Do allow
2308 * error clearing for l2cache devices.
2309 */
2310 if (avail_spare)
2311 return (zfs_error(hdl, EZFS_ISSPARE, msg));
2312
2313 verify(nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID,
2314 &zc.zc_guid) == 0);
2315 }
2316
2317 if (zfs_ioctl(hdl, ZFS_IOC_CLEAR, &zc) == 0)
2318 return (0);
2319
2320 return (zpool_standard_error(hdl, errno, msg));
2321 }
2322
2323 /*
2324 * Similar to zpool_clear(), but takes a GUID (used by fmd).
2325 */
2326 int
2327 zpool_vdev_clear(zpool_handle_t *zhp, uint64_t guid)
2328 {
2329 zfs_cmd_t zc = { 0 };
2330 char msg[1024];
2331 libzfs_handle_t *hdl = zhp->zpool_hdl;
2332
2333 (void) snprintf(msg, sizeof (msg),
2334 dgettext(TEXT_DOMAIN, "cannot clear errors for %llx"),
2335 guid);
2336
2337 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
2338 zc.zc_guid = guid;
2339
2340 if (ioctl(hdl->libzfs_fd, ZFS_IOC_CLEAR, &zc) == 0)
2341 return (0);
2342
2343 return (zpool_standard_error(hdl, errno, msg));
2344 }
2345
2346 /*
2347 * Iterate over all zvols in a given pool by walking the /dev/zvol/dsk/<pool>
2348 * hierarchy.
2349 */
2350 int
2351 zpool_iter_zvol(zpool_handle_t *zhp, int (*cb)(const char *, void *),
2352 void *data)
2353 {
2354 libzfs_handle_t *hdl = zhp->zpool_hdl;
2355 char (*paths)[MAXPATHLEN];
2356 size_t size = 4;
2357 int curr, fd, base, ret = 0;
2358 DIR *dirp;
2359 struct dirent *dp;
2360 struct stat st;
2361
2362 if ((base = open("/dev/zvol/dsk", O_RDONLY)) < 0)
2363 return (errno == ENOENT ? 0 : -1);
2364
2365 if (fstatat(base, zhp->zpool_name, &st, 0) != 0) {
2366 int err = errno;
2367 (void) close(base);
2368 return (err == ENOENT ? 0 : -1);
2369 }
2370
2371 /*
2372 * Oddly this wasn't a directory -- ignore that failure since we
2373 * know there are no links lower in the (non-existant) hierarchy.
2374 */
2375 if (!S_ISDIR(st.st_mode)) {
2376 (void) close(base);
2377 return (0);
2378 }
2379
2380 if ((paths = zfs_alloc(hdl, size * sizeof (paths[0]))) == NULL) {
2381 (void) close(base);
2382 return (-1);
2383 }
2384
2385 (void) strlcpy(paths[0], zhp->zpool_name, sizeof (paths[0]));
2386 curr = 0;
2387
2388 while (curr >= 0) {
2389 if (fstatat(base, paths[curr], &st, AT_SYMLINK_NOFOLLOW) != 0)
2390 goto err;
2391
2392 if (S_ISDIR(st.st_mode)) {
2393 if ((fd = openat(base, paths[curr], O_RDONLY)) < 0)
2394 goto err;
2395
2396 if ((dirp = fdopendir(fd)) == NULL) {
2397 (void) close(fd);
2398 goto err;
2399 }
2400
2401 while ((dp = readdir(dirp)) != NULL) {
2402 if (dp->d_name[0] == '.')
2403 continue;
2404
2405 if (curr + 1 == size) {
2406 paths = zfs_realloc(hdl, paths,
2407 size * sizeof (paths[0]),
2408 size * 2 * sizeof (paths[0]));
2409 if (paths == NULL) {
2410 (void) closedir(dirp);
2411 (void) close(fd);
2412 goto err;
2413 }
2414
2415 size *= 2;
2416 }
2417
2418 (void) strlcpy(paths[curr + 1], paths[curr],
2419 sizeof (paths[curr + 1]));
2420 (void) strlcat(paths[curr], "/",
2421 sizeof (paths[curr]));
2422 (void) strlcat(paths[curr], dp->d_name,
2423 sizeof (paths[curr]));
2424 curr++;
2425 }
2426
2427 (void) closedir(dirp);
2428
2429 } else {
2430 if ((ret = cb(paths[curr], data)) != 0)
2431 break;
2432 }
2433
2434 curr--;
2435 }
2436
2437 free(paths);
2438 (void) close(base);
2439
2440 return (ret);
2441
2442 err:
2443 free(paths);
2444 (void) close(base);
2445 return (-1);
2446 }
2447
2448 typedef struct zvol_cb {
2449 zpool_handle_t *zcb_pool;
2450 boolean_t zcb_create;
2451 } zvol_cb_t;
2452
2453 /*ARGSUSED*/
2454 static int
2455 do_zvol_create(zfs_handle_t *zhp, void *data)
2456 {
2457 int ret = 0;
2458
2459 if (ZFS_IS_VOLUME(zhp)) {
2460 (void) zvol_create_link(zhp->zfs_hdl, zhp->zfs_name);
2461 ret = zfs_iter_snapshots(zhp, do_zvol_create, NULL);
2462 }
2463
2464 if (ret == 0)
2465 ret = zfs_iter_filesystems(zhp, do_zvol_create, NULL);
2466
2467 zfs_close(zhp);
2468
2469 return (ret);
2470 }
2471
2472 /*
2473 * Iterate over all zvols in the pool and make any necessary minor nodes.
2474 */
2475 int
2476 zpool_create_zvol_links(zpool_handle_t *zhp)
2477 {
2478 zfs_handle_t *zfp;
2479 int ret;
2480
2481 /*
2482 * If the pool is unavailable, just return success.
2483 */
2484 if ((zfp = make_dataset_handle(zhp->zpool_hdl,
2485 zhp->zpool_name)) == NULL)
2486 return (0);
2487
2488 ret = zfs_iter_filesystems(zfp, do_zvol_create, NULL);
2489
2490 zfs_close(zfp);
2491 return (ret);
2492 }
2493
2494 static int
2495 do_zvol_remove(const char *dataset, void *data)
2496 {
2497 zpool_handle_t *zhp = data;
2498
2499 return (zvol_remove_link(zhp->zpool_hdl, dataset));
2500 }
2501
2502 /*
2503 * Iterate over all zvols in the pool and remove any minor nodes. We iterate
2504 * by examining the /dev links so that a corrupted pool doesn't impede this
2505 * operation.
2506 */
2507 int
2508 zpool_remove_zvol_links(zpool_handle_t *zhp)
2509 {
2510 return (zpool_iter_zvol(zhp, do_zvol_remove, zhp));
2511 }
2512
2513 /*
2514 * Convert from a devid string to a path.
2515 */
2516 static char *
2517 devid_to_path(char *devid_str)
2518 {
2519 ddi_devid_t devid;
2520 char *minor;
2521 char *path;
2522 devid_nmlist_t *list = NULL;
2523 int ret;
2524
2525 if (devid_str_decode(devid_str, &devid, &minor) != 0)
2526 return (NULL);
2527
2528 ret = devid_deviceid_to_nmlist("/dev", devid, minor, &list);
2529
2530 devid_str_free(minor);
2531 devid_free(devid);
2532
2533 if (ret != 0)
2534 return (NULL);
2535
2536 if ((path = strdup(list[0].devname)) == NULL)
2537 return (NULL);
2538
2539 devid_free_nmlist(list);
2540
2541 return (path);
2542 }
2543
2544 /*
2545 * Convert from a path to a devid string.
2546 */
2547 static char *
2548 path_to_devid(const char *path)
2549 {
2550 int fd;
2551 ddi_devid_t devid;
2552 char *minor, *ret;
2553
2554 if ((fd = open(path, O_RDONLY)) < 0)
2555 return (NULL);
2556
2557 minor = NULL;
2558 ret = NULL;
2559 if (devid_get(fd, &devid) == 0) {
2560 if (devid_get_minor_name(fd, &minor) == 0)
2561 ret = devid_str_encode(devid, minor);
2562 if (minor != NULL)
2563 devid_str_free(minor);
2564 devid_free(devid);
2565 }
2566 (void) close(fd);
2567
2568 return (ret);
2569 }
2570
2571 /*
2572 * Issue the necessary ioctl() to update the stored path value for the vdev. We
2573 * ignore any failure here, since a common case is for an unprivileged user to
2574 * type 'zpool status', and we'll display the correct information anyway.
2575 */
2576 static void
2577 set_path(zpool_handle_t *zhp, nvlist_t *nv, const char *path)
2578 {
2579 zfs_cmd_t zc = { 0 };
2580
2581 (void) strncpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
2582 (void) strncpy(zc.zc_value, path, sizeof (zc.zc_value));
2583 verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID,
2584 &zc.zc_guid) == 0);
2585
2586 (void) ioctl(zhp->zpool_hdl->libzfs_fd, ZFS_IOC_VDEV_SETPATH, &zc);
2587 }
2588
2589 /*
2590 * Given a vdev, return the name to display in iostat. If the vdev has a path,
2591 * we use that, stripping off any leading "/dev/dsk/"; if not, we use the type.
2592 * We also check if this is a whole disk, in which case we strip off the
2593 * trailing 's0' slice name.
2594 *
2595 * This routine is also responsible for identifying when disks have been
2596 * reconfigured in a new location. The kernel will have opened the device by
2597 * devid, but the path will still refer to the old location. To catch this, we
2598 * first do a path -> devid translation (which is fast for the common case). If
2599 * the devid matches, we're done. If not, we do a reverse devid -> path
2600 * translation and issue the appropriate ioctl() to update the path of the vdev.
2601 * If 'zhp' is NULL, then this is an exported pool, and we don't need to do any
2602 * of these checks.
2603 */
2604 char *
2605 zpool_vdev_name(libzfs_handle_t *hdl, zpool_handle_t *zhp, nvlist_t *nv)
2606 {
2607 char *path, *devid;
2608 uint64_t value;
2609 char buf[64];
2610 vdev_stat_t *vs;
2611 uint_t vsc;
2612
2613 if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_NOT_PRESENT,
2614 &value) == 0) {
2615 verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID,
2616 &value) == 0);
2617 (void) snprintf(buf, sizeof (buf), "%llu",
2618 (u_longlong_t)value);
2619 path = buf;
2620 } else if (nvlist_lookup_string(nv, ZPOOL_CONFIG_PATH, &path) == 0) {
2621
2622 /*
2623 * If the device is dead (faulted, offline, etc) then don't
2624 * bother opening it. Otherwise we may be forcing the user to
2625 * open a misbehaving device, which can have undesirable
2626 * effects.
2627 */
2628 if ((nvlist_lookup_uint64_array(nv, ZPOOL_CONFIG_STATS,
2629 (uint64_t **)&vs, &vsc) != 0 ||
2630 vs->vs_state >= VDEV_STATE_DEGRADED) &&
2631 zhp != NULL &&
2632 nvlist_lookup_string(nv, ZPOOL_CONFIG_DEVID, &devid) == 0) {
2633 /*
2634 * Determine if the current path is correct.
2635 */
2636 char *newdevid = path_to_devid(path);
2637
2638 if (newdevid == NULL ||
2639 strcmp(devid, newdevid) != 0) {
2640 char *newpath;
2641
2642 if ((newpath = devid_to_path(devid)) != NULL) {
2643 /*
2644 * Update the path appropriately.
2645 */
2646 set_path(zhp, nv, newpath);
2647 if (nvlist_add_string(nv,
2648 ZPOOL_CONFIG_PATH, newpath) == 0)
2649 verify(nvlist_lookup_string(nv,
2650 ZPOOL_CONFIG_PATH,
2651 &path) == 0);
2652 free(newpath);
2653 }
2654 }
2655
2656 if (newdevid)
2657 devid_str_free(newdevid);
2658 }
2659
2660 if (strncmp(path, "/dev/dsk/", 9) == 0)
2661 path += 9;
2662
2663 if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_WHOLE_DISK,
2664 &value) == 0 && value) {
2665 char *tmp = zfs_strdup(hdl, path);
2666 if (tmp == NULL)
2667 return (NULL);
2668 tmp[strlen(path) - 2] = '\0';
2669 return (tmp);
2670 }
2671 } else {
2672 verify(nvlist_lookup_string(nv, ZPOOL_CONFIG_TYPE, &path) == 0);
2673
2674 /*
2675 * If it's a raidz device, we need to stick in the parity level.
2676 */
2677 if (strcmp(path, VDEV_TYPE_RAIDZ) == 0) {
2678 verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_NPARITY,
2679 &value) == 0);
2680 (void) snprintf(buf, sizeof (buf), "%s%llu", path,
2681 (u_longlong_t)value);
2682 path = buf;
2683 }
2684 }
2685
2686 return (zfs_strdup(hdl, path));
2687 }
2688
2689 static int
2690 zbookmark_compare(const void *a, const void *b)
2691 {
2692 return (memcmp(a, b, sizeof (zbookmark_t)));
2693 }
2694
2695 /*
2696 * Retrieve the persistent error log, uniquify the members, and return to the
2697 * caller.
2698 */
2699 int
2700 zpool_get_errlog(zpool_handle_t *zhp, nvlist_t **nverrlistp)
2701 {
2702 zfs_cmd_t zc = { 0 };
2703 uint64_t count;
2704 zbookmark_t *zb = NULL;
2705 int i;
2706
2707 /*
2708 * Retrieve the raw error list from the kernel. If the number of errors
2709 * has increased, allocate more space and continue until we get the
2710 * entire list.
2711 */
2712 verify(nvlist_lookup_uint64(zhp->zpool_config, ZPOOL_CONFIG_ERRCOUNT,
2713 &count) == 0);
2714 if (count == 0)
2715 return (0);
2716 if ((zc.zc_nvlist_dst = (uintptr_t)zfs_alloc(zhp->zpool_hdl,
2717 count * sizeof (zbookmark_t))) == (uintptr_t)NULL)
2718 return (-1);
2719 zc.zc_nvlist_dst_size = count;
2720 (void) strcpy(zc.zc_name, zhp->zpool_name);
2721 for (;;) {
2722 if (ioctl(zhp->zpool_hdl->libzfs_fd, ZFS_IOC_ERROR_LOG,
2723 &zc) != 0) {
2724 free((void *)(uintptr_t)zc.zc_nvlist_dst);
2725 if (errno == ENOMEM) {
2726 count = zc.zc_nvlist_dst_size;
2727 if ((zc.zc_nvlist_dst = (uintptr_t)
2728 zfs_alloc(zhp->zpool_hdl, count *
2729 sizeof (zbookmark_t))) == (uintptr_t)NULL)
2730 return (-1);
2731 } else {
2732 return (-1);
2733 }
2734 } else {
2735 break;
2736 }
2737 }
2738
2739 /*
2740 * Sort the resulting bookmarks. This is a little confusing due to the
2741 * implementation of ZFS_IOC_ERROR_LOG. The bookmarks are copied last
2742 * to first, and 'zc_nvlist_dst_size' indicates the number of boomarks
2743 * _not_ copied as part of the process. So we point the start of our
2744 * array appropriate and decrement the total number of elements.
2745 */
2746 zb = ((zbookmark_t *)(uintptr_t)zc.zc_nvlist_dst) +
2747 zc.zc_nvlist_dst_size;
2748 count -= zc.zc_nvlist_dst_size;
2749
2750 qsort(zb, count, sizeof (zbookmark_t), zbookmark_compare);
2751
2752 verify(nvlist_alloc(nverrlistp, 0, KM_SLEEP) == 0);
2753
2754 /*
2755 * Fill in the nverrlistp with nvlist's of dataset and object numbers.
2756 */
2757 for (i = 0; i < count; i++) {
2758 nvlist_t *nv;
2759
2760 /* ignoring zb_blkid and zb_level for now */
2761 if (i > 0 && zb[i-1].zb_objset == zb[i].zb_objset &&
2762 zb[i-1].zb_object == zb[i].zb_object)
2763 continue;
2764
2765 if (nvlist_alloc(&nv, NV_UNIQUE_NAME, KM_SLEEP) != 0)
2766 goto nomem;
2767 if (nvlist_add_uint64(nv, ZPOOL_ERR_DATASET,
2768 zb[i].zb_objset) != 0) {
2769 nvlist_free(nv);
2770 goto nomem;
2771 }
2772 if (nvlist_add_uint64(nv, ZPOOL_ERR_OBJECT,
2773 zb[i].zb_object) != 0) {
2774 nvlist_free(nv);
2775 goto nomem;
2776 }
2777 if (nvlist_add_nvlist(*nverrlistp, "ejk", nv) != 0) {
2778 nvlist_free(nv);
2779 goto nomem;
2780 }
2781 nvlist_free(nv);
2782 }
2783
2784 free((void *)(uintptr_t)zc.zc_nvlist_dst);
2785 return (0);
2786
2787 nomem:
2788 free((void *)(uintptr_t)zc.zc_nvlist_dst);
2789 return (no_memory(zhp->zpool_hdl));
2790 }
2791
2792 /*
2793 * Upgrade a ZFS pool to the latest on-disk version.
2794 */
2795 int
2796 zpool_upgrade(zpool_handle_t *zhp, uint64_t new_version)
2797 {
2798 zfs_cmd_t zc = { 0 };
2799 libzfs_handle_t *hdl = zhp->zpool_hdl;
2800
2801 (void) strcpy(zc.zc_name, zhp->zpool_name);
2802 zc.zc_cookie = new_version;
2803
2804 if (zfs_ioctl(hdl, ZFS_IOC_POOL_UPGRADE, &zc) != 0)
2805 return (zpool_standard_error_fmt(hdl, errno,
2806 dgettext(TEXT_DOMAIN, "cannot upgrade '%s'"),
2807 zhp->zpool_name));
2808 return (0);
2809 }
2810
2811 void
2812 zpool_set_history_str(const char *subcommand, int argc, char **argv,
2813 char *history_str)
2814 {
2815 int i;
2816
2817 (void) strlcpy(history_str, subcommand, HIS_MAX_RECORD_LEN);
2818 for (i = 1; i < argc; i++) {
2819 if (strlen(history_str) + 1 + strlen(argv[i]) >
2820 HIS_MAX_RECORD_LEN)
2821 break;
2822 (void) strlcat(history_str, " ", HIS_MAX_RECORD_LEN);
2823 (void) strlcat(history_str, argv[i], HIS_MAX_RECORD_LEN);
2824 }
2825 }
2826
2827 /*
2828 * Stage command history for logging.
2829 */
2830 int
2831 zpool_stage_history(libzfs_handle_t *hdl, const char *history_str)
2832 {
2833 if (history_str == NULL)
2834 return (EINVAL);
2835
2836 if (strlen(history_str) > HIS_MAX_RECORD_LEN)
2837 return (EINVAL);
2838
2839 if (hdl->libzfs_log_str != NULL)
2840 free(hdl->libzfs_log_str);
2841
2842 if ((hdl->libzfs_log_str = strdup(history_str)) == NULL)
2843 return (no_memory(hdl));
2844
2845 return (0);
2846 }
2847
2848 /*
2849 * Perform ioctl to get some command history of a pool.
2850 *
2851 * 'buf' is the buffer to fill up to 'len' bytes. 'off' is the
2852 * logical offset of the history buffer to start reading from.
2853 *
2854 * Upon return, 'off' is the next logical offset to read from and
2855 * 'len' is the actual amount of bytes read into 'buf'.
2856 */
2857 static int
2858 get_history(zpool_handle_t *zhp, char *buf, uint64_t *off, uint64_t *len)
2859 {
2860 zfs_cmd_t zc = { 0 };
2861 libzfs_handle_t *hdl = zhp->zpool_hdl;
2862
2863 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
2864
2865 zc.zc_history = (uint64_t)(uintptr_t)buf;
2866 zc.zc_history_len = *len;
2867 zc.zc_history_offset = *off;
2868
2869 if (ioctl(hdl->libzfs_fd, ZFS_IOC_POOL_GET_HISTORY, &zc) != 0) {
2870 switch (errno) {
2871 case EPERM:
2872 return (zfs_error_fmt(hdl, EZFS_PERM,
2873 dgettext(TEXT_DOMAIN,
2874 "cannot show history for pool '%s'"),
2875 zhp->zpool_name));
2876 case ENOENT:
2877 return (zfs_error_fmt(hdl, EZFS_NOHISTORY,
2878 dgettext(TEXT_DOMAIN, "cannot get history for pool "
2879 "'%s'"), zhp->zpool_name));
2880 case ENOTSUP:
2881 return (zfs_error_fmt(hdl, EZFS_BADVERSION,
2882 dgettext(TEXT_DOMAIN, "cannot get history for pool "
2883 "'%s', pool must be upgraded"), zhp->zpool_name));
2884 default:
2885 return (zpool_standard_error_fmt(hdl, errno,
2886 dgettext(TEXT_DOMAIN,
2887 "cannot get history for '%s'"), zhp->zpool_name));
2888 }
2889 }
2890
2891 *len = zc.zc_history_len;
2892 *off = zc.zc_history_offset;
2893
2894 return (0);
2895 }
2896
2897 /*
2898 * Process the buffer of nvlists, unpacking and storing each nvlist record
2899 * into 'records'. 'leftover' is set to the number of bytes that weren't
2900 * processed as there wasn't a complete record.
2901 */
2902 static int
2903 zpool_history_unpack(char *buf, uint64_t bytes_read, uint64_t *leftover,
2904 nvlist_t ***records, uint_t *numrecords)
2905 {
2906 uint64_t reclen;
2907 nvlist_t *nv;
2908 int i;
2909
2910 while (bytes_read > sizeof (reclen)) {
2911
2912 /* get length of packed record (stored as little endian) */
2913 for (i = 0, reclen = 0; i < sizeof (reclen); i++)
2914 reclen += (uint64_t)(((uchar_t *)buf)[i]) << (8*i);
2915
2916 if (bytes_read < sizeof (reclen) + reclen)
2917 break;
2918
2919 /* unpack record */
2920 if (nvlist_unpack(buf + sizeof (reclen), reclen, &nv, 0) != 0)
2921 return (ENOMEM);
2922 bytes_read -= sizeof (reclen) + reclen;
2923 buf += sizeof (reclen) + reclen;
2924
2925 /* add record to nvlist array */
2926 (*numrecords)++;
2927 if (ISP2(*numrecords + 1)) {
2928 *records = realloc(*records,
2929 *numrecords * 2 * sizeof (nvlist_t *));
2930 }
2931 (*records)[*numrecords - 1] = nv;
2932 }
2933
2934 *leftover = bytes_read;
2935 return (0);
2936 }
2937
2938 #define HIS_BUF_LEN (128*1024)
2939
2940 /*
2941 * Retrieve the command history of a pool.
2942 */
2943 int
2944 zpool_get_history(zpool_handle_t *zhp, nvlist_t **nvhisp)
2945 {
2946 char buf[HIS_BUF_LEN];
2947 uint64_t off = 0;
2948 nvlist_t **records = NULL;
2949 uint_t numrecords = 0;
2950 int err, i;
2951
2952 do {
2953 uint64_t bytes_read = sizeof (buf);
2954 uint64_t leftover;
2955
2956 if ((err = get_history(zhp, buf, &off, &bytes_read)) != 0)
2957 break;
2958
2959 /* if nothing else was read in, we're at EOF, just return */
2960 if (!bytes_read)
2961 break;
2962
2963 if ((err = zpool_history_unpack(buf, bytes_read,
2964 &leftover, &records, &numrecords)) != 0)
2965 break;
2966 off -= leftover;
2967
2968 /* CONSTCOND */
2969 } while (1);
2970
2971 if (!err) {
2972 verify(nvlist_alloc(nvhisp, NV_UNIQUE_NAME, 0) == 0);
2973 verify(nvlist_add_nvlist_array(*nvhisp, ZPOOL_HIST_RECORD,
2974 records, numrecords) == 0);
2975 }
2976 for (i = 0; i < numrecords; i++)
2977 nvlist_free(records[i]);
2978 free(records);
2979
2980 return (err);
2981 }
2982
2983 void
2984 zpool_obj_to_path(zpool_handle_t *zhp, uint64_t dsobj, uint64_t obj,
2985 char *pathname, size_t len)
2986 {
2987 zfs_cmd_t zc = { 0 };
2988 boolean_t mounted = B_FALSE;
2989 char *mntpnt = NULL;
2990 char dsname[MAXNAMELEN];
2991
2992 if (dsobj == 0) {
2993 /* special case for the MOS */
2994 (void) snprintf(pathname, len, "<metadata>:<0x%llx>", obj);
2995 return;
2996 }
2997
2998 /* get the dataset's name */
2999 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
3000 zc.zc_obj = dsobj;
3001 if (ioctl(zhp->zpool_hdl->libzfs_fd,
3002 ZFS_IOC_DSOBJ_TO_DSNAME, &zc) != 0) {
3003 /* just write out a path of two object numbers */
3004 (void) snprintf(pathname, len, "<0x%llx>:<0x%llx>",
3005 dsobj, obj);
3006 return;
3007 }
3008 (void) strlcpy(dsname, zc.zc_value, sizeof (dsname));
3009
3010 /* find out if the dataset is mounted */
3011 mounted = is_mounted(zhp->zpool_hdl, dsname, &mntpnt);
3012
3013 /* get the corrupted object's path */
3014 (void) strlcpy(zc.zc_name, dsname, sizeof (zc.zc_name));
3015 zc.zc_obj = obj;
3016 if (ioctl(zhp->zpool_hdl->libzfs_fd, ZFS_IOC_OBJ_TO_PATH,
3017 &zc) == 0) {
3018 if (mounted) {
3019 (void) snprintf(pathname, len, "%s%s", mntpnt,
3020 zc.zc_value);
3021 } else {
3022 (void) snprintf(pathname, len, "%s:%s",
3023 dsname, zc.zc_value);
3024 }
3025 } else {
3026 (void) snprintf(pathname, len, "%s:<0x%llx>", dsname, obj);
3027 }
3028 free(mntpnt);
3029 }
3030
3031 /*
3032 * Read the EFI label from the config, if a label does not exist then
3033 * pass back the error to the caller. If the caller has passed a non-NULL
3034 * diskaddr argument then we set it to the starting address of the EFI
3035 * partition.
3036 */
3037 static int
3038 read_efi_label(nvlist_t *config, diskaddr_t *sb)
3039 {
3040 char *path;
3041 int fd;
3042 char diskname[MAXPATHLEN];
3043 int err = -1;
3044
3045 if (nvlist_lookup_string(config, ZPOOL_CONFIG_PATH, &path) != 0)
3046 return (err);
3047
3048 (void) snprintf(diskname, sizeof (diskname), "%s%s", RDISK_ROOT,
3049 strrchr(path, '/'));
3050 if ((fd = open(diskname, O_RDONLY|O_NDELAY)) >= 0) {
3051 struct dk_gpt *vtoc;
3052
3053 if ((err = efi_alloc_and_read(fd, &vtoc)) >= 0) {
3054 if (sb != NULL)
3055 *sb = vtoc->efi_parts[0].p_start;
3056 efi_free(vtoc);
3057 }
3058 (void) close(fd);
3059 }
3060 return (err);
3061 }
3062
3063 /*
3064 * determine where a partition starts on a disk in the current
3065 * configuration
3066 */
3067 static diskaddr_t
3068 find_start_block(nvlist_t *config)
3069 {
3070 nvlist_t **child;
3071 uint_t c, children;
3072 diskaddr_t sb = MAXOFFSET_T;
3073 uint64_t wholedisk;
3074
3075 if (nvlist_lookup_nvlist_array(config,
3076 ZPOOL_CONFIG_CHILDREN, &child, &children) != 0) {
3077 if (nvlist_lookup_uint64(config,
3078 ZPOOL_CONFIG_WHOLE_DISK,
3079 &wholedisk) != 0 || !wholedisk) {
3080 return (MAXOFFSET_T);
3081 }
3082 if (read_efi_label(config, &sb) < 0)
3083 sb = MAXOFFSET_T;
3084 return (sb);
3085 }
3086
3087 for (c = 0; c < children; c++) {
3088 sb = find_start_block(child[c]);
3089 if (sb != MAXOFFSET_T) {
3090 return (sb);
3091 }
3092 }
3093 return (MAXOFFSET_T);
3094 }
3095
3096 /*
3097 * Label an individual disk. The name provided is the short name,
3098 * stripped of any leading /dev path.
3099 */
3100 int
3101 zpool_label_disk(libzfs_handle_t *hdl, zpool_handle_t *zhp, char *name)
3102 {
3103 char path[MAXPATHLEN];
3104 struct dk_gpt *vtoc;
3105 int fd;
3106 size_t resv = EFI_MIN_RESV_SIZE;
3107 uint64_t slice_size;
3108 diskaddr_t start_block;
3109 char errbuf[1024];
3110
3111 /* prepare an error message just in case */
3112 (void) snprintf(errbuf, sizeof (errbuf),
3113 dgettext(TEXT_DOMAIN, "cannot label '%s'"), name);
3114
3115 if (zhp) {
3116 nvlist_t *nvroot;
3117
3118 if (pool_is_bootable(zhp)) {
3119 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3120 "EFI labeled devices are not supported on root "
3121 "pools."));
3122 return (zfs_error(hdl, EZFS_POOL_NOTSUP, errbuf));
3123 }
3124
3125 verify(nvlist_lookup_nvlist(zhp->zpool_config,
3126 ZPOOL_CONFIG_VDEV_TREE, &nvroot) == 0);
3127
3128 if (zhp->zpool_start_block == 0)
3129 start_block = find_start_block(nvroot);
3130 else
3131 start_block = zhp->zpool_start_block;
3132 zhp->zpool_start_block = start_block;
3133 } else {
3134 /* new pool */
3135 start_block = NEW_START_BLOCK;
3136 }
3137
3138 (void) snprintf(path, sizeof (path), "%s/%s%s", RDISK_ROOT, name,
3139 BACKUP_SLICE);
3140
3141 if ((fd = open(path, O_RDWR | O_NDELAY)) < 0) {
3142 /*
3143 * This shouldn't happen. We've long since verified that this
3144 * is a valid device.
3145 */
3146 zfs_error_aux(hdl,
3147 dgettext(TEXT_DOMAIN, "unable to open device"));
3148 return (zfs_error(hdl, EZFS_OPENFAILED, errbuf));
3149 }
3150
3151 if (efi_alloc_and_init(fd, EFI_NUMPAR, &vtoc) != 0) {
3152 /*
3153 * The only way this can fail is if we run out of memory, or we
3154 * were unable to read the disk's capacity
3155 */
3156 if (errno == ENOMEM)
3157 (void) no_memory(hdl);
3158
3159 (void) close(fd);
3160 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3161 "unable to read disk capacity"), name);
3162
3163 return (zfs_error(hdl, EZFS_NOCAP, errbuf));
3164 }
3165
3166 slice_size = vtoc->efi_last_u_lba + 1;
3167 slice_size -= EFI_MIN_RESV_SIZE;
3168 if (start_block == MAXOFFSET_T)
3169 start_block = NEW_START_BLOCK;
3170 slice_size -= start_block;
3171
3172 vtoc->efi_parts[0].p_start = start_block;
3173 vtoc->efi_parts[0].p_size = slice_size;
3174
3175 /*
3176 * Why we use V_USR: V_BACKUP confuses users, and is considered
3177 * disposable by some EFI utilities (since EFI doesn't have a backup
3178 * slice). V_UNASSIGNED is supposed to be used only for zero size
3179 * partitions, and efi_write() will fail if we use it. V_ROOT, V_BOOT,
3180 * etc. were all pretty specific. V_USR is as close to reality as we
3181 * can get, in the absence of V_OTHER.
3182 */
3183 vtoc->efi_parts[0].p_tag = V_USR;
3184 (void) strcpy(vtoc->efi_parts[0].p_name, "zfs");
3185
3186 vtoc->efi_parts[8].p_start = slice_size + start_block;
3187 vtoc->efi_parts[8].p_size = resv;
3188 vtoc->efi_parts[8].p_tag = V_RESERVED;
3189
3190 if (efi_write(fd, vtoc) != 0) {
3191 /*
3192 * Some block drivers (like pcata) may not support EFI
3193 * GPT labels. Print out a helpful error message dir-
3194 * ecting the user to manually label the disk and give
3195 * a specific slice.
3196 */
3197 (void) close(fd);
3198 efi_free(vtoc);
3199
3200 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3201 "try using fdisk(1M) and then provide a specific slice"));
3202 return (zfs_error(hdl, EZFS_LABELFAILED, errbuf));
3203 }
3204
3205 (void) close(fd);
3206 efi_free(vtoc);
3207 return (0);
3208 }
3209
3210 static boolean_t
3211 supported_dump_vdev_type(libzfs_handle_t *hdl, nvlist_t *config, char *errbuf)
3212 {
3213 char *type;
3214 nvlist_t **child;
3215 uint_t children, c;
3216
3217 verify(nvlist_lookup_string(config, ZPOOL_CONFIG_TYPE, &type) == 0);
3218 if (strcmp(type, VDEV_TYPE_RAIDZ) == 0 ||
3219 strcmp(type, VDEV_TYPE_FILE) == 0 ||
3220 strcmp(type, VDEV_TYPE_LOG) == 0 ||
3221 strcmp(type, VDEV_TYPE_MISSING) == 0) {
3222 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3223 "vdev type '%s' is not supported"), type);
3224 (void) zfs_error(hdl, EZFS_VDEVNOTSUP, errbuf);
3225 return (B_FALSE);
3226 }
3227 if (nvlist_lookup_nvlist_array(config, ZPOOL_CONFIG_CHILDREN,
3228 &child, &children) == 0) {
3229 for (c = 0; c < children; c++) {
3230 if (!supported_dump_vdev_type(hdl, child[c], errbuf))
3231 return (B_FALSE);
3232 }
3233 }
3234 return (B_TRUE);
3235 }
3236
3237 /*
3238 * check if this zvol is allowable for use as a dump device; zero if
3239 * it is, > 0 if it isn't, < 0 if it isn't a zvol
3240 */
3241 int
3242 zvol_check_dump_config(char *arg)
3243 {
3244 zpool_handle_t *zhp = NULL;
3245 nvlist_t *config, *nvroot;
3246 char *p, *volname;
3247 nvlist_t **top;
3248 uint_t toplevels;
3249 libzfs_handle_t *hdl;
3250 char errbuf[1024];
3251 char poolname[ZPOOL_MAXNAMELEN];
3252 int pathlen = strlen(ZVOL_FULL_DEV_DIR);
3253 int ret = 1;
3254
3255 if (strncmp(arg, ZVOL_FULL_DEV_DIR, pathlen)) {
3256 return (-1);
3257 }
3258
3259 (void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN,
3260 "dump is not supported on device '%s'"), arg);
3261
3262 if ((hdl = libzfs_init()) == NULL)
3263 return (1);
3264 libzfs_print_on_error(hdl, B_TRUE);
3265
3266 volname = arg + pathlen;
3267
3268 /* check the configuration of the pool */
3269 if ((p = strchr(volname, '/')) == NULL) {
3270 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3271 "malformed dataset name"));
3272 (void) zfs_error(hdl, EZFS_INVALIDNAME, errbuf);
3273 return (1);
3274 } else if (p - volname >= ZFS_MAXNAMELEN) {
3275 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3276 "dataset name is too long"));
3277 (void) zfs_error(hdl, EZFS_NAMETOOLONG, errbuf);
3278 return (1);
3279 } else {
3280 (void) strncpy(poolname, volname, p - volname);
3281 poolname[p - volname] = '\0';
3282 }
3283
3284 if ((zhp = zpool_open(hdl, poolname)) == NULL) {
3285 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3286 "could not open pool '%s'"), poolname);
3287 (void) zfs_error(hdl, EZFS_OPENFAILED, errbuf);
3288 goto out;
3289 }
3290 config = zpool_get_config(zhp, NULL);
3291 if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE,
3292 &nvroot) != 0) {
3293 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3294 "could not obtain vdev configuration for '%s'"), poolname);
3295 (void) zfs_error(hdl, EZFS_INVALCONFIG, errbuf);
3296 goto out;
3297 }
3298
3299 verify(nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_CHILDREN,
3300 &top, &toplevels) == 0);
3301 if (toplevels != 1) {
3302 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3303 "'%s' has multiple top level vdevs"), poolname);
3304 (void) zfs_error(hdl, EZFS_DEVOVERFLOW, errbuf);
3305 goto out;
3306 }
3307
3308 if (!supported_dump_vdev_type(hdl, top[0], errbuf)) {
3309 goto out;
3310 }
3311 ret = 0;
3312
3313 out:
3314 if (zhp)
3315 zpool_close(zhp);
3316 libzfs_fini(hdl);
3317 return (ret);
3318 }