]> git.proxmox.com Git - mirror_zfs.git/blob - lib/libzfs/libzfs_pool.c
projects / mirror_zfs.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
Rebase master to b105
[mirror_zfs.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 2008 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
46 #include "zfs_namecheck.h"
47 #include "zfs_prop.h"
48 #include "libzfs_impl.h"
49
50 static int read_efi_label(nvlist_t *config, diskaddr_t *sb);
51
52 #if defined(__i386) || defined(__amd64)
53 #define BOOTCMD "installgrub(1M)"
54 #else
55 #define BOOTCMD "installboot(1M)"
56 #endif
57
58 /*
59 * ====================================================================
60 * zpool property functions
61 * ====================================================================
62 */
63
64 static int
65 zpool_get_all_props(zpool_handle_t *zhp)
66 {
67 zfs_cmd_t zc = { 0 };
68 libzfs_handle_t *hdl = zhp->zpool_hdl;
69
70 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
71
72 if (zcmd_alloc_dst_nvlist(hdl, &zc, 0) != 0)
73 return (-1);
74
75 while (ioctl(hdl->libzfs_fd, ZFS_IOC_POOL_GET_PROPS, &zc) != 0) {
76 if (errno == ENOMEM) {
77 if (zcmd_expand_dst_nvlist(hdl, &zc) != 0) {
78 zcmd_free_nvlists(&zc);
79 return (-1);
80 }
81 } else {
82 zcmd_free_nvlists(&zc);
83 return (-1);
84 }
85 }
86
87 if (zcmd_read_dst_nvlist(hdl, &zc, &zhp->zpool_props) != 0) {
88 zcmd_free_nvlists(&zc);
89 return (-1);
90 }
91
92 zcmd_free_nvlists(&zc);
93
94 return (0);
95 }
96
97 static int
98 zpool_props_refresh(zpool_handle_t *zhp)
99 {
100 nvlist_t *old_props;
101
102 old_props = zhp->zpool_props;
103
104 if (zpool_get_all_props(zhp) != 0)
105 return (-1);
106
107 nvlist_free(old_props);
108 return (0);
109 }
110
111 static char *
112 zpool_get_prop_string(zpool_handle_t *zhp, zpool_prop_t prop,
113 zprop_source_t *src)
114 {
115 nvlist_t *nv, *nvl;
116 uint64_t ival;
117 char *value;
118 zprop_source_t source;
119
120 nvl = zhp->zpool_props;
121 if (nvlist_lookup_nvlist(nvl, zpool_prop_to_name(prop), &nv) == 0) {
122 verify(nvlist_lookup_uint64(nv, ZPROP_SOURCE, &ival) == 0);
123 source = ival;
124 verify(nvlist_lookup_string(nv, ZPROP_VALUE, &value) == 0);
125 } else {
126 source = ZPROP_SRC_DEFAULT;
127 if ((value = (char *)zpool_prop_default_string(prop)) == NULL)
128 value = "-";
129 }
130
131 if (src)
132 *src = source;
133
134 return (value);
135 }
136
137 uint64_t
138 zpool_get_prop_int(zpool_handle_t *zhp, zpool_prop_t prop, zprop_source_t *src)
139 {
140 nvlist_t *nv, *nvl;
141 uint64_t value;
142 zprop_source_t source;
143
144 if (zhp->zpool_props == NULL && zpool_get_all_props(zhp)) {
145 /*
146 * zpool_get_all_props() has most likely failed because
147 * the pool is faulted, but if all we need is the top level
148 * vdev's guid then get it from the zhp config nvlist.
149 */
150 if ((prop == ZPOOL_PROP_GUID) &&
151 (nvlist_lookup_nvlist(zhp->zpool_config,
152 ZPOOL_CONFIG_VDEV_TREE, &nv) == 0) &&
153 (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID, &value)
154 == 0)) {
155 return (value);
156 }
157 return (zpool_prop_default_numeric(prop));
158 }
159
160 nvl = zhp->zpool_props;
161 if (nvlist_lookup_nvlist(nvl, zpool_prop_to_name(prop), &nv) == 0) {
162 verify(nvlist_lookup_uint64(nv, ZPROP_SOURCE, &value) == 0);
163 source = value;
164 verify(nvlist_lookup_uint64(nv, ZPROP_VALUE, &value) == 0);
165 } else {
166 source = ZPROP_SRC_DEFAULT;
167 value = zpool_prop_default_numeric(prop);
168 }
169
170 if (src)
171 *src = source;
172
173 return (value);
174 }
175
176 /*
177 * Map VDEV STATE to printed strings.
178 */
179 char *
180 zpool_state_to_name(vdev_state_t state, vdev_aux_t aux)
181 {
182 switch (state) {
183 case VDEV_STATE_CLOSED:
184 case VDEV_STATE_OFFLINE:
185 return (gettext("OFFLINE"));
186 case VDEV_STATE_REMOVED:
187 return (gettext("REMOVED"));
188 case VDEV_STATE_CANT_OPEN:
189 if (aux == VDEV_AUX_CORRUPT_DATA || aux == VDEV_AUX_BAD_LOG)
190 return (gettext("FAULTED"));
191 else
192 return (gettext("UNAVAIL"));
193 case VDEV_STATE_FAULTED:
194 return (gettext("FAULTED"));
195 case VDEV_STATE_DEGRADED:
196 return (gettext("DEGRADED"));
197 case VDEV_STATE_HEALTHY:
198 return (gettext("ONLINE"));
199 }
200
201 return (gettext("UNKNOWN"));
202 }
203
204 /*
205 * Get a zpool property value for 'prop' and return the value in
206 * a pre-allocated buffer.
207 */
208 int
209 zpool_get_prop(zpool_handle_t *zhp, zpool_prop_t prop, char *buf, size_t len,
210 zprop_source_t *srctype)
211 {
212 uint64_t intval;
213 const char *strval;
214 zprop_source_t src = ZPROP_SRC_NONE;
215 nvlist_t *nvroot;
216 vdev_stat_t *vs;
217 uint_t vsc;
218
219 if (zpool_get_state(zhp) == POOL_STATE_UNAVAIL) {
220 if (prop == ZPOOL_PROP_NAME)
221 (void) strlcpy(buf, zpool_get_name(zhp), len);
222 else if (prop == ZPOOL_PROP_HEALTH)
223 (void) strlcpy(buf, "FAULTED", len);
224 else
225 (void) strlcpy(buf, "-", len);
226 return (0);
227 }
228
229 if (zhp->zpool_props == NULL && zpool_get_all_props(zhp) &&
230 prop != ZPOOL_PROP_NAME)
231 return (-1);
232
233 switch (zpool_prop_get_type(prop)) {
234 case PROP_TYPE_STRING:
235 (void) strlcpy(buf, zpool_get_prop_string(zhp, prop, &src),
236 len);
237 break;
238
239 case PROP_TYPE_NUMBER:
240 intval = zpool_get_prop_int(zhp, prop, &src);
241
242 switch (prop) {
243 case ZPOOL_PROP_SIZE:
244 case ZPOOL_PROP_USED:
245 case ZPOOL_PROP_AVAILABLE:
246 (void) zfs_nicenum(intval, buf, len);
247 break;
248
249 case ZPOOL_PROP_CAPACITY:
250 (void) snprintf(buf, len, "%llu%%",
251 (u_longlong_t)intval);
252 break;
253
254 case ZPOOL_PROP_HEALTH:
255 verify(nvlist_lookup_nvlist(zpool_get_config(zhp, NULL),
256 ZPOOL_CONFIG_VDEV_TREE, &nvroot) == 0);
257 verify(nvlist_lookup_uint64_array(nvroot,
258 ZPOOL_CONFIG_STATS, (uint64_t **)&vs, &vsc) == 0);
259
260 (void) strlcpy(buf, zpool_state_to_name(intval,
261 vs->vs_aux), len);
262 break;
263 default:
264 (void) snprintf(buf, len, "%llu", intval);
265 }
266 break;
267
268 case PROP_TYPE_INDEX:
269 intval = zpool_get_prop_int(zhp, prop, &src);
270 if (zpool_prop_index_to_string(prop, intval, &strval)
271 != 0)
272 return (-1);
273 (void) strlcpy(buf, strval, len);
274 break;
275
276 default:
277 abort();
278 }
279
280 if (srctype)
281 *srctype = src;
282
283 return (0);
284 }
285
286 /*
287 * Check if the bootfs name has the same pool name as it is set to.
288 * Assuming bootfs is a valid dataset name.
289 */
290 static boolean_t
291 bootfs_name_valid(const char *pool, char *bootfs)
292 {
293 int len = strlen(pool);
294
295 if (!zfs_name_valid(bootfs, ZFS_TYPE_FILESYSTEM|ZFS_TYPE_SNAPSHOT))
296 return (B_FALSE);
297
298 if (strncmp(pool, bootfs, len) == 0 &&
299 (bootfs[len] == '/' || bootfs[len] == '\0'))
300 return (B_TRUE);
301
302 return (B_FALSE);
303 }
304
305 /*
306 * Inspect the configuration to determine if any of the devices contain
307 * an EFI label.
308 */
309 static boolean_t
310 pool_uses_efi(nvlist_t *config)
311 {
312 nvlist_t **child;
313 uint_t c, children;
314
315 if (nvlist_lookup_nvlist_array(config, ZPOOL_CONFIG_CHILDREN,
316 &child, &children) != 0)
317 return (read_efi_label(config, NULL) >= 0);
318
319 for (c = 0; c < children; c++) {
320 if (pool_uses_efi(child[c]))
321 return (B_TRUE);
322 }
323 return (B_FALSE);
324 }
325
326 static boolean_t
327 pool_is_bootable(zpool_handle_t *zhp)
328 {
329 char bootfs[ZPOOL_MAXNAMELEN];
330
331 return (zpool_get_prop(zhp, ZPOOL_PROP_BOOTFS, bootfs,
332 sizeof (bootfs), NULL) == 0 && strncmp(bootfs, "-",
333 sizeof (bootfs)) != 0);
334 }
335
336
337 /*
338 * Given an nvlist of zpool properties to be set, validate that they are
339 * correct, and parse any numeric properties (index, boolean, etc) if they are
340 * specified as strings.
341 */
342 static nvlist_t *
343 zpool_valid_proplist(libzfs_handle_t *hdl, const char *poolname,
344 nvlist_t *props, uint64_t version, boolean_t create_or_import, char *errbuf)
345 {
346 nvpair_t *elem;
347 nvlist_t *retprops;
348 zpool_prop_t prop;
349 char *strval;
350 uint64_t intval;
351 char *slash;
352 struct stat64 statbuf;
353 zpool_handle_t *zhp;
354 nvlist_t *nvroot;
355
356 if (nvlist_alloc(&retprops, NV_UNIQUE_NAME, 0) != 0) {
357 (void) no_memory(hdl);
358 return (NULL);
359 }
360
361 elem = NULL;
362 while ((elem = nvlist_next_nvpair(props, elem)) != NULL) {
363 const char *propname = nvpair_name(elem);
364
365 /*
366 * Make sure this property is valid and applies to this type.
367 */
368 if ((prop = zpool_name_to_prop(propname)) == ZPROP_INVAL) {
369 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
370 "invalid property '%s'"), propname);
371 (void) zfs_error(hdl, EZFS_BADPROP, errbuf);
372 goto error;
373 }
374
375 if (zpool_prop_readonly(prop)) {
376 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "'%s' "
377 "is readonly"), propname);
378 (void) zfs_error(hdl, EZFS_PROPREADONLY, errbuf);
379 goto error;
380 }
381
382 if (zprop_parse_value(hdl, elem, prop, ZFS_TYPE_POOL, retprops,
383 &strval, &intval, errbuf) != 0)
384 goto error;
385
386 /*
387 * Perform additional checking for specific properties.
388 */
389 switch (prop) {
390 case ZPOOL_PROP_VERSION:
391 if (intval < version || intval > SPA_VERSION) {
392 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
393 "property '%s' number %d is invalid."),
394 propname, intval);
395 (void) zfs_error(hdl, EZFS_BADVERSION, errbuf);
396 goto error;
397 }
398 break;
399
400 case ZPOOL_PROP_BOOTFS:
401 if (create_or_import) {
402 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
403 "property '%s' cannot be set at creation "
404 "or import time"), propname);
405 (void) zfs_error(hdl, EZFS_BADPROP, errbuf);
406 goto error;
407 }
408
409 if (version < SPA_VERSION_BOOTFS) {
410 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
411 "pool must be upgraded to support "
412 "'%s' property"), propname);
413 (void) zfs_error(hdl, EZFS_BADVERSION, errbuf);
414 goto error;
415 }
416
417 /*
418 * bootfs property value has to be a dataset name and
419 * the dataset has to be in the same pool as it sets to.
420 */
421 if (strval[0] != '\0' && !bootfs_name_valid(poolname,
422 strval)) {
423 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "'%s' "
424 "is an invalid name"), strval);
425 (void) zfs_error(hdl, EZFS_INVALIDNAME, errbuf);
426 goto error;
427 }
428
429 if ((zhp = zpool_open_canfail(hdl, poolname)) == NULL) {
430 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
431 "could not open pool '%s'"), poolname);
432 (void) zfs_error(hdl, EZFS_OPENFAILED, errbuf);
433 goto error;
434 }
435 verify(nvlist_lookup_nvlist(zpool_get_config(zhp, NULL),
436 ZPOOL_CONFIG_VDEV_TREE, &nvroot) == 0);
437
438 /*
439 * bootfs property cannot be set on a disk which has
440 * been EFI labeled.
441 */
442 if (pool_uses_efi(nvroot)) {
443 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
444 "property '%s' not supported on "
445 "EFI labeled devices"), propname);
446 (void) zfs_error(hdl, EZFS_POOL_NOTSUP, errbuf);
447 zpool_close(zhp);
448 goto error;
449 }
450 zpool_close(zhp);
451 break;
452
453 case ZPOOL_PROP_ALTROOT:
454 if (!create_or_import) {
455 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
456 "property '%s' can only be set during pool "
457 "creation or import"), propname);
458 (void) zfs_error(hdl, EZFS_BADPROP, errbuf);
459 goto error;
460 }
461
462 if (strval[0] != '/') {
463 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
464 "bad alternate root '%s'"), strval);
465 (void) zfs_error(hdl, EZFS_BADPATH, errbuf);
466 goto error;
467 }
468 break;
469
470 case ZPOOL_PROP_CACHEFILE:
471 if (strval[0] == '\0')
472 break;
473
474 if (strcmp(strval, "none") == 0)
475 break;
476
477 if (strval[0] != '/') {
478 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
479 "property '%s' must be empty, an "
480 "absolute path, or 'none'"), propname);
481 (void) zfs_error(hdl, EZFS_BADPATH, errbuf);
482 goto error;
483 }
484
485 slash = strrchr(strval, '/');
486
487 if (slash[1] == '\0' || strcmp(slash, "/.") == 0 ||
488 strcmp(slash, "/..") == 0) {
489 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
490 "'%s' is not a valid file"), strval);
491 (void) zfs_error(hdl, EZFS_BADPATH, errbuf);
492 goto error;
493 }
494
495 *slash = '\0';
496
497 if (strval[0] != '\0' &&
498 (stat64(strval, &statbuf) != 0 ||
499 !S_ISDIR(statbuf.st_mode))) {
500 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
501 "'%s' is not a valid directory"),
502 strval);
503 (void) zfs_error(hdl, EZFS_BADPATH, errbuf);
504 goto error;
505 }
506
507 *slash = '/';
508 break;
509 }
510 }
511
512 return (retprops);
513 error:
514 nvlist_free(retprops);
515 return (NULL);
516 }
517
518 /*
519 * Set zpool property : propname=propval.
520 */
521 int
522 zpool_set_prop(zpool_handle_t *zhp, const char *propname, const char *propval)
523 {
524 zfs_cmd_t zc = { 0 };
525 int ret = -1;
526 char errbuf[1024];
527 nvlist_t *nvl = NULL;
528 nvlist_t *realprops;
529 uint64_t version;
530
531 (void) snprintf(errbuf, sizeof (errbuf),
532 dgettext(TEXT_DOMAIN, "cannot set property for '%s'"),
533 zhp->zpool_name);
534
535 if (zhp->zpool_props == NULL && zpool_get_all_props(zhp))
536 return (zfs_error(zhp->zpool_hdl, EZFS_POOLPROPS, errbuf));
537
538 if (nvlist_alloc(&nvl, NV_UNIQUE_NAME, 0) != 0)
539 return (no_memory(zhp->zpool_hdl));
540
541 if (nvlist_add_string(nvl, propname, propval) != 0) {
542 nvlist_free(nvl);
543 return (no_memory(zhp->zpool_hdl));
544 }
545
546 version = zpool_get_prop_int(zhp, ZPOOL_PROP_VERSION, NULL);
547 if ((realprops = zpool_valid_proplist(zhp->zpool_hdl,
548 zhp->zpool_name, nvl, version, B_FALSE, errbuf)) == NULL) {
549 nvlist_free(nvl);
550 return (-1);
551 }
552
553 nvlist_free(nvl);
554 nvl = realprops;
555
556 /*
557 * Execute the corresponding ioctl() to set this property.
558 */
559 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
560
561 if (zcmd_write_src_nvlist(zhp->zpool_hdl, &zc, nvl) != 0) {
562 nvlist_free(nvl);
563 return (-1);
564 }
565
566 ret = zfs_ioctl(zhp->zpool_hdl, ZFS_IOC_POOL_SET_PROPS, &zc);
567
568 zcmd_free_nvlists(&zc);
569 nvlist_free(nvl);
570
571 if (ret)
572 (void) zpool_standard_error(zhp->zpool_hdl, errno, errbuf);
573 else
574 (void) zpool_props_refresh(zhp);
575
576 return (ret);
577 }
578
579 int
580 zpool_expand_proplist(zpool_handle_t *zhp, zprop_list_t **plp)
581 {
582 libzfs_handle_t *hdl = zhp->zpool_hdl;
583 zprop_list_t *entry;
584 char buf[ZFS_MAXPROPLEN];
585
586 if (zprop_expand_list(hdl, plp, ZFS_TYPE_POOL) != 0)
587 return (-1);
588
589 for (entry = *plp; entry != NULL; entry = entry->pl_next) {
590
591 if (entry->pl_fixed)
592 continue;
593
594 if (entry->pl_prop != ZPROP_INVAL &&
595 zpool_get_prop(zhp, entry->pl_prop, buf, sizeof (buf),
596 NULL) == 0) {
597 if (strlen(buf) > entry->pl_width)
598 entry->pl_width = strlen(buf);
599 }
600 }
601
602 return (0);
603 }
604
605
606 /*
607 * Validate the given pool name, optionally putting an extended error message in
608 * 'buf'.
609 */
610 boolean_t
611 zpool_name_valid(libzfs_handle_t *hdl, boolean_t isopen, const char *pool)
612 {
613 namecheck_err_t why;
614 char what;
615 int ret;
616
617 ret = pool_namecheck(pool, &why, &what);
618
619 /*
620 * The rules for reserved pool names were extended at a later point.
621 * But we need to support users with existing pools that may now be
622 * invalid. So we only check for this expanded set of names during a
623 * create (or import), and only in userland.
624 */
625 if (ret == 0 && !isopen &&
626 (strncmp(pool, "mirror", 6) == 0 ||
627 strncmp(pool, "raidz", 5) == 0 ||
628 strncmp(pool, "spare", 5) == 0 ||
629 strcmp(pool, "log") == 0)) {
630 if (hdl != NULL)
631 zfs_error_aux(hdl,
632 dgettext(TEXT_DOMAIN, "name is reserved"));
633 return (B_FALSE);
634 }
635
636
637 if (ret != 0) {
638 if (hdl != NULL) {
639 switch (why) {
640 case NAME_ERR_TOOLONG:
641 zfs_error_aux(hdl,
642 dgettext(TEXT_DOMAIN, "name is too long"));
643 break;
644
645 case NAME_ERR_INVALCHAR:
646 zfs_error_aux(hdl,
647 dgettext(TEXT_DOMAIN, "invalid character "
648 "'%c' in pool name"), what);
649 break;
650
651 case NAME_ERR_NOLETTER:
652 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
653 "name must begin with a letter"));
654 break;
655
656 case NAME_ERR_RESERVED:
657 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
658 "name is reserved"));
659 break;
660
661 case NAME_ERR_DISKLIKE:
662 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
663 "pool name is reserved"));
664 break;
665
666 case NAME_ERR_LEADING_SLASH:
667 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
668 "leading slash in name"));
669 break;
670
671 case NAME_ERR_EMPTY_COMPONENT:
672 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
673 "empty component in name"));
674 break;
675
676 case NAME_ERR_TRAILING_SLASH:
677 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
678 "trailing slash in name"));
679 break;
680
681 case NAME_ERR_MULTIPLE_AT:
682 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
683 "multiple '@' delimiters in name"));
684 break;
685
686 }
687 }
688 return (B_FALSE);
689 }
690
691 return (B_TRUE);
692 }
693
694 /*
695 * Open a handle to the given pool, even if the pool is currently in the FAULTED
696 * state.
697 */
698 zpool_handle_t *
699 zpool_open_canfail(libzfs_handle_t *hdl, const char *pool)
700 {
701 zpool_handle_t *zhp;
702 boolean_t missing;
703
704 /*
705 * Make sure the pool name is valid.
706 */
707 if (!zpool_name_valid(hdl, B_TRUE, pool)) {
708 (void) zfs_error_fmt(hdl, EZFS_INVALIDNAME,
709 dgettext(TEXT_DOMAIN, "cannot open '%s'"),
710 pool);
711 return (NULL);
712 }
713
714 if ((zhp = zfs_alloc(hdl, sizeof (zpool_handle_t))) == NULL)
715 return (NULL);
716
717 zhp->zpool_hdl = hdl;
718 (void) strlcpy(zhp->zpool_name, pool, sizeof (zhp->zpool_name));
719
720 if (zpool_refresh_stats(zhp, &missing) != 0) {
721 zpool_close(zhp);
722 return (NULL);
723 }
724
725 if (missing) {
726 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "no such pool"));
727 (void) zfs_error_fmt(hdl, EZFS_NOENT,
728 dgettext(TEXT_DOMAIN, "cannot open '%s'"), pool);
729 zpool_close(zhp);
730 return (NULL);
731 }
732
733 return (zhp);
734 }
735
736 /*
737 * Like the above, but silent on error. Used when iterating over pools (because
738 * the configuration cache may be out of date).
739 */
740 int
741 zpool_open_silent(libzfs_handle_t *hdl, const char *pool, zpool_handle_t **ret)
742 {
743 zpool_handle_t *zhp;
744 boolean_t missing;
745
746 if ((zhp = zfs_alloc(hdl, sizeof (zpool_handle_t))) == NULL)
747 return (-1);
748
749 zhp->zpool_hdl = hdl;
750 (void) strlcpy(zhp->zpool_name, pool, sizeof (zhp->zpool_name));
751
752 if (zpool_refresh_stats(zhp, &missing) != 0) {
753 zpool_close(zhp);
754 return (-1);
755 }
756
757 if (missing) {
758 zpool_close(zhp);
759 *ret = NULL;
760 return (0);
761 }
762
763 *ret = zhp;
764 return (0);
765 }
766
767 /*
768 * Similar to zpool_open_canfail(), but refuses to open pools in the faulted
769 * state.
770 */
771 zpool_handle_t *
772 zpool_open(libzfs_handle_t *hdl, const char *pool)
773 {
774 zpool_handle_t *zhp;
775
776 if ((zhp = zpool_open_canfail(hdl, pool)) == NULL)
777 return (NULL);
778
779 if (zhp->zpool_state == POOL_STATE_UNAVAIL) {
780 (void) zfs_error_fmt(hdl, EZFS_POOLUNAVAIL,
781 dgettext(TEXT_DOMAIN, "cannot open '%s'"), zhp->zpool_name);
782 zpool_close(zhp);
783 return (NULL);
784 }
785
786 return (zhp);
787 }
788
789 /*
790 * Close the handle. Simply frees the memory associated with the handle.
791 */
792 void
793 zpool_close(zpool_handle_t *zhp)
794 {
795 if (zhp->zpool_config)
796 nvlist_free(zhp->zpool_config);
797 if (zhp->zpool_old_config)
798 nvlist_free(zhp->zpool_old_config);
799 if (zhp->zpool_props)
800 nvlist_free(zhp->zpool_props);
801 free(zhp);
802 }
803
804 /*
805 * Return the name of the pool.
806 */
807 const char *
808 zpool_get_name(zpool_handle_t *zhp)
809 {
810 return (zhp->zpool_name);
811 }
812
813
814 /*
815 * Return the state of the pool (ACTIVE or UNAVAILABLE)
816 */
817 int
818 zpool_get_state(zpool_handle_t *zhp)
819 {
820 return (zhp->zpool_state);
821 }
822
823 /*
824 * Create the named pool, using the provided vdev list. It is assumed
825 * that the consumer has already validated the contents of the nvlist, so we
826 * don't have to worry about error semantics.
827 */
828 int
829 zpool_create(libzfs_handle_t *hdl, const char *pool, nvlist_t *nvroot,
830 nvlist_t *props, nvlist_t *fsprops)
831 {
832 zfs_cmd_t zc = { 0 };
833 nvlist_t *zc_fsprops = NULL;
834 nvlist_t *zc_props = NULL;
835 char msg[1024];
836 char *altroot;
837 int ret = -1;
838
839 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN,
840 "cannot create '%s'"), pool);
841
842 if (!zpool_name_valid(hdl, B_FALSE, pool))
843 return (zfs_error(hdl, EZFS_INVALIDNAME, msg));
844
845 if (zcmd_write_conf_nvlist(hdl, &zc, nvroot) != 0)
846 return (-1);
847
848 if (props) {
849 if ((zc_props = zpool_valid_proplist(hdl, pool, props,
850 SPA_VERSION_1, B_TRUE, msg)) == NULL) {
851 goto create_failed;
852 }
853 }
854
855 if (fsprops) {
856 uint64_t zoned;
857 char *zonestr;
858
859 zoned = ((nvlist_lookup_string(fsprops,
860 zfs_prop_to_name(ZFS_PROP_ZONED), &zonestr) == 0) &&
861 strcmp(zonestr, "on") == 0);
862
863 if ((zc_fsprops = zfs_valid_proplist(hdl,
864 ZFS_TYPE_FILESYSTEM, fsprops, zoned, NULL, msg)) == NULL) {
865 goto create_failed;
866 }
867 if (!zc_props &&
868 (nvlist_alloc(&zc_props, NV_UNIQUE_NAME, 0) != 0)) {
869 goto create_failed;
870 }
871 if (nvlist_add_nvlist(zc_props,
872 ZPOOL_ROOTFS_PROPS, zc_fsprops) != 0) {
873 goto create_failed;
874 }
875 }
876
877 if (zc_props && zcmd_write_src_nvlist(hdl, &zc, zc_props) != 0)
878 goto create_failed;
879
880 (void) strlcpy(zc.zc_name, pool, sizeof (zc.zc_name));
881
882 if ((ret = zfs_ioctl(hdl, ZFS_IOC_POOL_CREATE, &zc)) != 0) {
883
884 zcmd_free_nvlists(&zc);
885 nvlist_free(zc_props);
886 nvlist_free(zc_fsprops);
887
888 switch (errno) {
889 case EBUSY:
890 /*
891 * This can happen if the user has specified the same
892 * device multiple times. We can't reliably detect this
893 * until we try to add it and see we already have a
894 * label.
895 */
896 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
897 "one or more vdevs refer to the same device"));
898 return (zfs_error(hdl, EZFS_BADDEV, msg));
899
900 case EOVERFLOW:
901 /*
902 * This occurs when one of the devices is below
903 * SPA_MINDEVSIZE. Unfortunately, we can't detect which
904 * device was the problem device since there's no
905 * reliable way to determine device size from userland.
906 */
907 {
908 char buf[64];
909
910 zfs_nicenum(SPA_MINDEVSIZE, buf, sizeof (buf));
911
912 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
913 "one or more devices is less than the "
914 "minimum size (%s)"), buf);
915 }
916 return (zfs_error(hdl, EZFS_BADDEV, msg));
917
918 case ENOSPC:
919 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
920 "one or more devices is out of space"));
921 return (zfs_error(hdl, EZFS_BADDEV, msg));
922
923 case ENOTBLK:
924 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
925 "cache device must be a disk or disk slice"));
926 return (zfs_error(hdl, EZFS_BADDEV, msg));
927
928 default:
929 return (zpool_standard_error(hdl, errno, msg));
930 }
931 }
932
933 /*
934 * If this is an alternate root pool, then we automatically set the
935 * mountpoint of the root dataset to be '/'.
936 */
937 if (nvlist_lookup_string(props, zpool_prop_to_name(ZPOOL_PROP_ALTROOT),
938 &altroot) == 0) {
939 zfs_handle_t *zhp;
940
941 verify((zhp = zfs_open(hdl, pool, ZFS_TYPE_DATASET)) != NULL);
942 verify(zfs_prop_set(zhp, zfs_prop_to_name(ZFS_PROP_MOUNTPOINT),
943 "/") == 0);
944
945 zfs_close(zhp);
946 }
947
948 create_failed:
949 zcmd_free_nvlists(&zc);
950 nvlist_free(zc_props);
951 nvlist_free(zc_fsprops);
952 return (ret);
953 }
954
955 /*
956 * Destroy the given pool. It is up to the caller to ensure that there are no
957 * datasets left in the pool.
958 */
959 int
960 zpool_destroy(zpool_handle_t *zhp)
961 {
962 zfs_cmd_t zc = { 0 };
963 zfs_handle_t *zfp = NULL;
964 libzfs_handle_t *hdl = zhp->zpool_hdl;
965 char msg[1024];
966
967 if (zhp->zpool_state == POOL_STATE_ACTIVE &&
968 (zfp = zfs_open(zhp->zpool_hdl, zhp->zpool_name,
969 ZFS_TYPE_FILESYSTEM)) == NULL)
970 return (-1);
971
972 if (zpool_remove_zvol_links(zhp) != 0)
973 return (-1);
974
975 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
976
977 if (zfs_ioctl(zhp->zpool_hdl, ZFS_IOC_POOL_DESTROY, &zc) != 0) {
978 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN,
979 "cannot destroy '%s'"), zhp->zpool_name);
980
981 if (errno == EROFS) {
982 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
983 "one or more devices is read only"));
984 (void) zfs_error(hdl, EZFS_BADDEV, msg);
985 } else {
986 (void) zpool_standard_error(hdl, errno, msg);
987 }
988
989 if (zfp)
990 zfs_close(zfp);
991 return (-1);
992 }
993
994 if (zfp) {
995 remove_mountpoint(zfp);
996 zfs_close(zfp);
997 }
998
999 return (0);
1000 }
1001
1002 /*
1003 * Add the given vdevs to the pool. The caller must have already performed the
1004 * necessary verification to ensure that the vdev specification is well-formed.
1005 */
1006 int
1007 zpool_add(zpool_handle_t *zhp, nvlist_t *nvroot)
1008 {
1009 zfs_cmd_t zc = { 0 };
1010 int ret;
1011 libzfs_handle_t *hdl = zhp->zpool_hdl;
1012 char msg[1024];
1013 nvlist_t **spares, **l2cache;
1014 uint_t nspares, nl2cache;
1015
1016 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN,
1017 "cannot add to '%s'"), zhp->zpool_name);
1018
1019 if (zpool_get_prop_int(zhp, ZPOOL_PROP_VERSION, NULL) <
1020 SPA_VERSION_SPARES &&
1021 nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_SPARES,
1022 &spares, &nspares) == 0) {
1023 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "pool must be "
1024 "upgraded to add hot spares"));
1025 return (zfs_error(hdl, EZFS_BADVERSION, msg));
1026 }
1027
1028 if (pool_is_bootable(zhp) && nvlist_lookup_nvlist_array(nvroot,
1029 ZPOOL_CONFIG_SPARES, &spares, &nspares) == 0) {
1030 uint64_t s;
1031
1032 for (s = 0; s < nspares; s++) {
1033 char *path;
1034
1035 if (nvlist_lookup_string(spares[s], ZPOOL_CONFIG_PATH,
1036 &path) == 0 && pool_uses_efi(spares[s])) {
1037 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1038 "device '%s' contains an EFI label and "
1039 "cannot be used on root pools."),
1040 zpool_vdev_name(hdl, NULL, spares[s]));
1041 return (zfs_error(hdl, EZFS_POOL_NOTSUP, msg));
1042 }
1043 }
1044 }
1045
1046 if (zpool_get_prop_int(zhp, ZPOOL_PROP_VERSION, NULL) <
1047 SPA_VERSION_L2CACHE &&
1048 nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_L2CACHE,
1049 &l2cache, &nl2cache) == 0) {
1050 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "pool must be "
1051 "upgraded to add cache devices"));
1052 return (zfs_error(hdl, EZFS_BADVERSION, msg));
1053 }
1054
1055 if (zcmd_write_conf_nvlist(hdl, &zc, nvroot) != 0)
1056 return (-1);
1057 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
1058
1059 if (zfs_ioctl(zhp->zpool_hdl, ZFS_IOC_VDEV_ADD, &zc) != 0) {
1060 switch (errno) {
1061 case EBUSY:
1062 /*
1063 * This can happen if the user has specified the same
1064 * device multiple times. We can't reliably detect this
1065 * until we try to add it and see we already have a
1066 * label.
1067 */
1068 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1069 "one or more vdevs refer to the same device"));
1070 (void) zfs_error(hdl, EZFS_BADDEV, msg);
1071 break;
1072
1073 case EOVERFLOW:
1074 /*
1075 * This occurrs when one of the devices is below
1076 * SPA_MINDEVSIZE. Unfortunately, we can't detect which
1077 * device was the problem device since there's no
1078 * reliable way to determine device size from userland.
1079 */
1080 {
1081 char buf[64];
1082
1083 zfs_nicenum(SPA_MINDEVSIZE, buf, sizeof (buf));
1084
1085 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1086 "device is less than the minimum "
1087 "size (%s)"), buf);
1088 }
1089 (void) zfs_error(hdl, EZFS_BADDEV, msg);
1090 break;
1091
1092 case ENOTSUP:
1093 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1094 "pool must be upgraded to add these vdevs"));
1095 (void) zfs_error(hdl, EZFS_BADVERSION, msg);
1096 break;
1097
1098 case EDOM:
1099 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1100 "root pool can not have multiple vdevs"
1101 " or separate logs"));
1102 (void) zfs_error(hdl, EZFS_POOL_NOTSUP, msg);
1103 break;
1104
1105 case ENOTBLK:
1106 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1107 "cache device must be a disk or disk slice"));
1108 (void) zfs_error(hdl, EZFS_BADDEV, msg);
1109 break;
1110
1111 default:
1112 (void) zpool_standard_error(hdl, errno, msg);
1113 }
1114
1115 ret = -1;
1116 } else {
1117 ret = 0;
1118 }
1119
1120 zcmd_free_nvlists(&zc);
1121
1122 return (ret);
1123 }
1124
1125 /*
1126 * Exports the pool from the system. The caller must ensure that there are no
1127 * mounted datasets in the pool.
1128 */
1129 int
1130 zpool_export_common(zpool_handle_t *zhp, boolean_t force, boolean_t hardforce)
1131 {
1132 zfs_cmd_t zc = { 0 };
1133 char msg[1024];
1134
1135 if (zpool_remove_zvol_links(zhp) != 0)
1136 return (-1);
1137
1138 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN,
1139 "cannot export '%s'"), zhp->zpool_name);
1140
1141 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
1142 zc.zc_cookie = force;
1143 zc.zc_guid = hardforce;
1144
1145 if (zfs_ioctl(zhp->zpool_hdl, ZFS_IOC_POOL_EXPORT, &zc) != 0) {
1146 switch (errno) {
1147 case EXDEV:
1148 zfs_error_aux(zhp->zpool_hdl, dgettext(TEXT_DOMAIN,
1149 "use '-f' to override the following errors:\n"
1150 "'%s' has an active shared spare which could be"
1151 " used by other pools once '%s' is exported."),
1152 zhp->zpool_name, zhp->zpool_name);
1153 return (zfs_error(zhp->zpool_hdl, EZFS_ACTIVE_SPARE,
1154 msg));
1155 default:
1156 return (zpool_standard_error_fmt(zhp->zpool_hdl, errno,
1157 msg));
1158 }
1159 }
1160
1161 return (0);
1162 }
1163
1164 int
1165 zpool_export(zpool_handle_t *zhp, boolean_t force)
1166 {
1167 return (zpool_export_common(zhp, force, B_FALSE));
1168 }
1169
1170 int
1171 zpool_export_force(zpool_handle_t *zhp)
1172 {
1173 return (zpool_export_common(zhp, B_TRUE, B_TRUE));
1174 }
1175
1176 /*
1177 * zpool_import() is a contracted interface. Should be kept the same
1178 * if possible.
1179 *
1180 * Applications should use zpool_import_props() to import a pool with
1181 * new properties value to be set.
1182 */
1183 int
1184 zpool_import(libzfs_handle_t *hdl, nvlist_t *config, const char *newname,
1185 char *altroot)
1186 {
1187 nvlist_t *props = NULL;
1188 int ret;
1189
1190 if (altroot != NULL) {
1191 if (nvlist_alloc(&props, NV_UNIQUE_NAME, 0) != 0) {
1192 return (zfs_error_fmt(hdl, EZFS_NOMEM,
1193 dgettext(TEXT_DOMAIN, "cannot import '%s'"),
1194 newname));
1195 }
1196
1197 if (nvlist_add_string(props,
1198 zpool_prop_to_name(ZPOOL_PROP_ALTROOT), altroot) != 0 ||
1199 nvlist_add_string(props,
1200 zpool_prop_to_name(ZPOOL_PROP_CACHEFILE), "none") != 0) {
1201 nvlist_free(props);
1202 return (zfs_error_fmt(hdl, EZFS_NOMEM,
1203 dgettext(TEXT_DOMAIN, "cannot import '%s'"),
1204 newname));
1205 }
1206 }
1207
1208 ret = zpool_import_props(hdl, config, newname, props, B_FALSE);
1209 if (props)
1210 nvlist_free(props);
1211 return (ret);
1212 }
1213
1214 /*
1215 * Import the given pool using the known configuration and a list of
1216 * properties to be set. The configuration should have come from
1217 * zpool_find_import(). The 'newname' parameters control whether the pool
1218 * is imported with a different name.
1219 */
1220 int
1221 zpool_import_props(libzfs_handle_t *hdl, nvlist_t *config, const char *newname,
1222 nvlist_t *props, boolean_t importfaulted)
1223 {
1224 zfs_cmd_t zc = { 0 };
1225 char *thename;
1226 char *origname;
1227 int ret;
1228 char errbuf[1024];
1229
1230 verify(nvlist_lookup_string(config, ZPOOL_CONFIG_POOL_NAME,
1231 &origname) == 0);
1232
1233 (void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN,
1234 "cannot import pool '%s'"), origname);
1235
1236 if (newname != NULL) {
1237 if (!zpool_name_valid(hdl, B_FALSE, newname))
1238 return (zfs_error_fmt(hdl, EZFS_INVALIDNAME,
1239 dgettext(TEXT_DOMAIN, "cannot import '%s'"),
1240 newname));
1241 thename = (char *)newname;
1242 } else {
1243 thename = origname;
1244 }
1245
1246 if (props) {
1247 uint64_t version;
1248
1249 verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_VERSION,
1250 &version) == 0);
1251
1252 if ((props = zpool_valid_proplist(hdl, origname,
1253 props, version, B_TRUE, errbuf)) == NULL) {
1254 return (-1);
1255 } else if (zcmd_write_src_nvlist(hdl, &zc, props) != 0) {
1256 nvlist_free(props);
1257 return (-1);
1258 }
1259 }
1260
1261 (void) strlcpy(zc.zc_name, thename, sizeof (zc.zc_name));
1262
1263 verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID,
1264 &zc.zc_guid) == 0);
1265
1266 if (zcmd_write_conf_nvlist(hdl, &zc, config) != 0) {
1267 nvlist_free(props);
1268 return (-1);
1269 }
1270
1271 zc.zc_cookie = (uint64_t)importfaulted;
1272 ret = 0;
1273 if (zfs_ioctl(hdl, ZFS_IOC_POOL_IMPORT, &zc) != 0) {
1274 char desc[1024];
1275 if (newname == NULL)
1276 (void) snprintf(desc, sizeof (desc),
1277 dgettext(TEXT_DOMAIN, "cannot import '%s'"),
1278 thename);
1279 else
1280 (void) snprintf(desc, sizeof (desc),
1281 dgettext(TEXT_DOMAIN, "cannot import '%s' as '%s'"),
1282 origname, thename);
1283
1284 switch (errno) {
1285 case ENOTSUP:
1286 /*
1287 * Unsupported version.
1288 */
1289 (void) zfs_error(hdl, EZFS_BADVERSION, desc);
1290 break;
1291
1292 case EINVAL:
1293 (void) zfs_error(hdl, EZFS_INVALCONFIG, desc);
1294 break;
1295
1296 default:
1297 (void) zpool_standard_error(hdl, errno, desc);
1298 }
1299
1300 ret = -1;
1301 } else {
1302 zpool_handle_t *zhp;
1303
1304 /*
1305 * This should never fail, but play it safe anyway.
1306 */
1307 if (zpool_open_silent(hdl, thename, &zhp) != 0) {
1308 ret = -1;
1309 } else if (zhp != NULL) {
1310 ret = zpool_create_zvol_links(zhp);
1311 zpool_close(zhp);
1312 }
1313
1314 }
1315
1316 zcmd_free_nvlists(&zc);
1317 nvlist_free(props);
1318
1319 return (ret);
1320 }
1321
1322 /*
1323 * Scrub the pool.
1324 */
1325 int
1326 zpool_scrub(zpool_handle_t *zhp, pool_scrub_type_t type)
1327 {
1328 zfs_cmd_t zc = { 0 };
1329 char msg[1024];
1330 libzfs_handle_t *hdl = zhp->zpool_hdl;
1331
1332 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
1333 zc.zc_cookie = type;
1334
1335 if (zfs_ioctl(zhp->zpool_hdl, ZFS_IOC_POOL_SCRUB, &zc) == 0)
1336 return (0);
1337
1338 (void) snprintf(msg, sizeof (msg),
1339 dgettext(TEXT_DOMAIN, "cannot scrub %s"), zc.zc_name);
1340
1341 if (errno == EBUSY)
1342 return (zfs_error(hdl, EZFS_RESILVERING, msg));
1343 else
1344 return (zpool_standard_error(hdl, errno, msg));
1345 }
1346
1347 /*
1348 * 'avail_spare' is set to TRUE if the provided guid refers to an AVAIL
1349 * spare; but FALSE if its an INUSE spare.
1350 */
1351 static nvlist_t *
1352 vdev_to_nvlist_iter(nvlist_t *nv, const char *search, uint64_t guid,
1353 boolean_t *avail_spare, boolean_t *l2cache, boolean_t *log)
1354 {
1355 uint_t c, children;
1356 nvlist_t **child;
1357 uint64_t theguid, present;
1358 char *path;
1359 uint64_t wholedisk = 0;
1360 nvlist_t *ret;
1361 uint64_t is_log;
1362
1363 verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID, &theguid) == 0);
1364
1365 if (search == NULL &&
1366 nvlist_lookup_uint64(nv, ZPOOL_CONFIG_NOT_PRESENT, &present) == 0) {
1367 /*
1368 * If the device has never been present since import, the only
1369 * reliable way to match the vdev is by GUID.
1370 */
1371 if (theguid == guid)
1372 return (nv);
1373 } else if (search != NULL &&
1374 nvlist_lookup_string(nv, ZPOOL_CONFIG_PATH, &path) == 0) {
1375 (void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_WHOLE_DISK,
1376 &wholedisk);
1377 if (wholedisk) {
1378 /*
1379 * For whole disks, the internal path has 's0', but the
1380 * path passed in by the user doesn't.
1381 */
1382 if (strlen(search) == strlen(path) - 2 &&
1383 strncmp(search, path, strlen(search)) == 0)
1384 return (nv);
1385 } else if (strcmp(search, path) == 0) {
1386 return (nv);
1387 }
1388 }
1389
1390 if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN,
1391 &child, &children) != 0)
1392 return (NULL);
1393
1394 for (c = 0; c < children; c++) {
1395 if ((ret = vdev_to_nvlist_iter(child[c], search, guid,
1396 avail_spare, l2cache, NULL)) != NULL) {
1397 /*
1398 * The 'is_log' value is only set for the toplevel
1399 * vdev, not the leaf vdevs. So we always lookup the
1400 * log device from the root of the vdev tree (where
1401 * 'log' is non-NULL).
1402 */
1403 if (log != NULL &&
1404 nvlist_lookup_uint64(child[c],
1405 ZPOOL_CONFIG_IS_LOG, &is_log) == 0 &&
1406 is_log) {
1407 *log = B_TRUE;
1408 }
1409 return (ret);
1410 }
1411 }
1412
1413 if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_SPARES,
1414 &child, &children) == 0) {
1415 for (c = 0; c < children; c++) {
1416 if ((ret = vdev_to_nvlist_iter(child[c], search, guid,
1417 avail_spare, l2cache, NULL)) != NULL) {
1418 *avail_spare = B_TRUE;
1419 return (ret);
1420 }
1421 }
1422 }
1423
1424 if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_L2CACHE,
1425 &child, &children) == 0) {
1426 for (c = 0; c < children; c++) {
1427 if ((ret = vdev_to_nvlist_iter(child[c], search, guid,
1428 avail_spare, l2cache, NULL)) != NULL) {
1429 *l2cache = B_TRUE;
1430 return (ret);
1431 }
1432 }
1433 }
1434
1435 return (NULL);
1436 }
1437
1438 nvlist_t *
1439 zpool_find_vdev(zpool_handle_t *zhp, const char *path, boolean_t *avail_spare,
1440 boolean_t *l2cache, boolean_t *log)
1441 {
1442 char buf[MAXPATHLEN];
1443 const char *search;
1444 char *end;
1445 nvlist_t *nvroot;
1446 uint64_t guid;
1447
1448 guid = strtoull(path, &end, 10);
1449 if (guid != 0 && *end == '\0') {
1450 search = NULL;
1451 } else if (path[0] != '/') {
1452 (void) snprintf(buf, sizeof (buf), "%s%s", "/dev/dsk/", path);
1453 search = buf;
1454 } else {
1455 search = path;
1456 }
1457
1458 verify(nvlist_lookup_nvlist(zhp->zpool_config, ZPOOL_CONFIG_VDEV_TREE,
1459 &nvroot) == 0);
1460
1461 *avail_spare = B_FALSE;
1462 *l2cache = B_FALSE;
1463 if (log != NULL)
1464 *log = B_FALSE;
1465 return (vdev_to_nvlist_iter(nvroot, search, guid, avail_spare,
1466 l2cache, log));
1467 }
1468
1469 static int
1470 vdev_online(nvlist_t *nv)
1471 {
1472 uint64_t ival;
1473
1474 if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_OFFLINE, &ival) == 0 ||
1475 nvlist_lookup_uint64(nv, ZPOOL_CONFIG_FAULTED, &ival) == 0 ||
1476 nvlist_lookup_uint64(nv, ZPOOL_CONFIG_REMOVED, &ival) == 0)
1477 return (0);
1478
1479 return (1);
1480 }
1481
1482 /*
1483 * Get phys_path for a root pool
1484 * Return 0 on success; non-zeron on failure.
1485 */
1486 int
1487 zpool_get_physpath(zpool_handle_t *zhp, char *physpath)
1488 {
1489 nvlist_t *vdev_root;
1490 nvlist_t **child;
1491 uint_t count;
1492 int i;
1493
1494 /*
1495 * Make sure this is a root pool, as phys_path doesn't mean
1496 * anything to a non-root pool.
1497 */
1498 if (!pool_is_bootable(zhp))
1499 return (-1);
1500
1501 verify(nvlist_lookup_nvlist(zhp->zpool_config,
1502 ZPOOL_CONFIG_VDEV_TREE, &vdev_root) == 0);
1503
1504 if (nvlist_lookup_nvlist_array(vdev_root, ZPOOL_CONFIG_CHILDREN,
1505 &child, &count) != 0)
1506 return (-2);
1507
1508 for (i = 0; i < count; i++) {
1509 nvlist_t **child2;
1510 uint_t count2;
1511 char *type;
1512 char *tmppath;
1513 int j;
1514
1515 if (nvlist_lookup_string(child[i], ZPOOL_CONFIG_TYPE, &type)
1516 != 0)
1517 return (-3);
1518
1519 if (strcmp(type, VDEV_TYPE_DISK) == 0) {
1520 if (!vdev_online(child[i]))
1521 return (-8);
1522 verify(nvlist_lookup_string(child[i],
1523 ZPOOL_CONFIG_PHYS_PATH, &tmppath) == 0);
1524 (void) strncpy(physpath, tmppath, strlen(tmppath));
1525 } else if (strcmp(type, VDEV_TYPE_MIRROR) == 0) {
1526 if (nvlist_lookup_nvlist_array(child[i],
1527 ZPOOL_CONFIG_CHILDREN, &child2, &count2) != 0)
1528 return (-4);
1529
1530 for (j = 0; j < count2; j++) {
1531 if (!vdev_online(child2[j]))
1532 return (-8);
1533 if (nvlist_lookup_string(child2[j],
1534 ZPOOL_CONFIG_PHYS_PATH, &tmppath) != 0)
1535 return (-5);
1536
1537 if ((strlen(physpath) + strlen(tmppath)) >
1538 MAXNAMELEN)
1539 return (-6);
1540
1541 if (strlen(physpath) == 0) {
1542 (void) strncpy(physpath, tmppath,
1543 strlen(tmppath));
1544 } else {
1545 (void) strcat(physpath, " ");
1546 (void) strcat(physpath, tmppath);
1547 }
1548 }
1549 } else {
1550 return (-7);
1551 }
1552 }
1553
1554 return (0);
1555 }
1556
1557 /*
1558 * Returns TRUE if the given guid corresponds to the given type.
1559 * This is used to check for hot spares (INUSE or not), and level 2 cache
1560 * devices.
1561 */
1562 static boolean_t
1563 is_guid_type(zpool_handle_t *zhp, uint64_t guid, const char *type)
1564 {
1565 uint64_t target_guid;
1566 nvlist_t *nvroot;
1567 nvlist_t **list;
1568 uint_t count;
1569 int i;
1570
1571 verify(nvlist_lookup_nvlist(zhp->zpool_config, ZPOOL_CONFIG_VDEV_TREE,
1572 &nvroot) == 0);
1573 if (nvlist_lookup_nvlist_array(nvroot, type, &list, &count) == 0) {
1574 for (i = 0; i < count; i++) {
1575 verify(nvlist_lookup_uint64(list[i], ZPOOL_CONFIG_GUID,
1576 &target_guid) == 0);
1577 if (guid == target_guid)
1578 return (B_TRUE);
1579 }
1580 }
1581
1582 return (B_FALSE);
1583 }
1584
1585 /*
1586 * Bring the specified vdev online. The 'flags' parameter is a set of the
1587 * ZFS_ONLINE_* flags.
1588 */
1589 int
1590 zpool_vdev_online(zpool_handle_t *zhp, const char *path, int flags,
1591 vdev_state_t *newstate)
1592 {
1593 zfs_cmd_t zc = { 0 };
1594 char msg[1024];
1595 nvlist_t *tgt;
1596 boolean_t avail_spare, l2cache;
1597 libzfs_handle_t *hdl = zhp->zpool_hdl;
1598
1599 (void) snprintf(msg, sizeof (msg),
1600 dgettext(TEXT_DOMAIN, "cannot online %s"), path);
1601
1602 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
1603 if ((tgt = zpool_find_vdev(zhp, path, &avail_spare, &l2cache,
1604 NULL)) == NULL)
1605 return (zfs_error(hdl, EZFS_NODEVICE, msg));
1606
1607 verify(nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID, &zc.zc_guid) == 0);
1608
1609 if (avail_spare ||
1610 is_guid_type(zhp, zc.zc_guid, ZPOOL_CONFIG_SPARES) == B_TRUE)
1611 return (zfs_error(hdl, EZFS_ISSPARE, msg));
1612
1613 zc.zc_cookie = VDEV_STATE_ONLINE;
1614 zc.zc_obj = flags;
1615
1616 if (zfs_ioctl(zhp->zpool_hdl, ZFS_IOC_VDEV_SET_STATE, &zc) != 0)
1617 return (zpool_standard_error(hdl, errno, msg));
1618
1619 *newstate = zc.zc_cookie;
1620 return (0);
1621 }
1622
1623 /*
1624 * Take the specified vdev offline
1625 */
1626 int
1627 zpool_vdev_offline(zpool_handle_t *zhp, const char *path, boolean_t istmp)
1628 {
1629 zfs_cmd_t zc = { 0 };
1630 char msg[1024];
1631 nvlist_t *tgt;
1632 boolean_t avail_spare, l2cache;
1633 libzfs_handle_t *hdl = zhp->zpool_hdl;
1634
1635 (void) snprintf(msg, sizeof (msg),
1636 dgettext(TEXT_DOMAIN, "cannot offline %s"), path);
1637
1638 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
1639 if ((tgt = zpool_find_vdev(zhp, path, &avail_spare, &l2cache,
1640 NULL)) == NULL)
1641 return (zfs_error(hdl, EZFS_NODEVICE, msg));
1642
1643 verify(nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID, &zc.zc_guid) == 0);
1644
1645 if (avail_spare ||
1646 is_guid_type(zhp, zc.zc_guid, ZPOOL_CONFIG_SPARES) == B_TRUE)
1647 return (zfs_error(hdl, EZFS_ISSPARE, msg));
1648
1649 zc.zc_cookie = VDEV_STATE_OFFLINE;
1650 zc.zc_obj = istmp ? ZFS_OFFLINE_TEMPORARY : 0;
1651
1652 if (zfs_ioctl(zhp->zpool_hdl, ZFS_IOC_VDEV_SET_STATE, &zc) == 0)
1653 return (0);
1654
1655 switch (errno) {
1656 case EBUSY:
1657
1658 /*
1659 * There are no other replicas of this device.
1660 */
1661 return (zfs_error(hdl, EZFS_NOREPLICAS, msg));
1662
1663 default:
1664 return (zpool_standard_error(hdl, errno, msg));
1665 }
1666 }
1667
1668 /*
1669 * Mark the given vdev faulted.
1670 */
1671 int
1672 zpool_vdev_fault(zpool_handle_t *zhp, uint64_t guid)
1673 {
1674 zfs_cmd_t zc = { 0 };
1675 char msg[1024];
1676 libzfs_handle_t *hdl = zhp->zpool_hdl;
1677
1678 (void) snprintf(msg, sizeof (msg),
1679 dgettext(TEXT_DOMAIN, "cannot fault %llu"), guid);
1680
1681 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
1682 zc.zc_guid = guid;
1683 zc.zc_cookie = VDEV_STATE_FAULTED;
1684
1685 if (ioctl(zhp->zpool_hdl->libzfs_fd, ZFS_IOC_VDEV_SET_STATE, &zc) == 0)
1686 return (0);
1687
1688 switch (errno) {
1689 case EBUSY:
1690
1691 /*
1692 * There are no other replicas of this device.
1693 */
1694 return (zfs_error(hdl, EZFS_NOREPLICAS, msg));
1695
1696 default:
1697 return (zpool_standard_error(hdl, errno, msg));
1698 }
1699
1700 }
1701
1702 /*
1703 * Mark the given vdev degraded.
1704 */
1705 int
1706 zpool_vdev_degrade(zpool_handle_t *zhp, uint64_t guid)
1707 {
1708 zfs_cmd_t zc = { 0 };
1709 char msg[1024];
1710 libzfs_handle_t *hdl = zhp->zpool_hdl;
1711
1712 (void) snprintf(msg, sizeof (msg),
1713 dgettext(TEXT_DOMAIN, "cannot degrade %llu"), guid);
1714
1715 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
1716 zc.zc_guid = guid;
1717 zc.zc_cookie = VDEV_STATE_DEGRADED;
1718
1719 if (ioctl(zhp->zpool_hdl->libzfs_fd, ZFS_IOC_VDEV_SET_STATE, &zc) == 0)
1720 return (0);
1721
1722 return (zpool_standard_error(hdl, errno, msg));
1723 }
1724
1725 /*
1726 * Returns TRUE if the given nvlist is a vdev that was originally swapped in as
1727 * a hot spare.
1728 */
1729 static boolean_t
1730 is_replacing_spare(nvlist_t *search, nvlist_t *tgt, int which)
1731 {
1732 nvlist_t **child;
1733 uint_t c, children;
1734 char *type;
1735
1736 if (nvlist_lookup_nvlist_array(search, ZPOOL_CONFIG_CHILDREN, &child,
1737 &children) == 0) {
1738 verify(nvlist_lookup_string(search, ZPOOL_CONFIG_TYPE,
1739 &type) == 0);
1740
1741 if (strcmp(type, VDEV_TYPE_SPARE) == 0 &&
1742 children == 2 && child[which] == tgt)
1743 return (B_TRUE);
1744
1745 for (c = 0; c < children; c++)
1746 if (is_replacing_spare(child[c], tgt, which))
1747 return (B_TRUE);
1748 }
1749
1750 return (B_FALSE);
1751 }
1752
1753 /*
1754 * Attach new_disk (fully described by nvroot) to old_disk.
1755 * If 'replacing' is specified, the new disk will replace the old one.
1756 */
1757 int
1758 zpool_vdev_attach(zpool_handle_t *zhp,
1759 const char *old_disk, const char *new_disk, nvlist_t *nvroot, int replacing)
1760 {
1761 zfs_cmd_t zc = { 0 };
1762 char msg[1024];
1763 int ret;
1764 nvlist_t *tgt;
1765 boolean_t avail_spare, l2cache, islog;
1766 uint64_t val;
1767 char *path, *newname;
1768 nvlist_t **child;
1769 uint_t children;
1770 nvlist_t *config_root;
1771 libzfs_handle_t *hdl = zhp->zpool_hdl;
1772 boolean_t rootpool = pool_is_bootable(zhp);
1773
1774 if (replacing)
1775 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN,
1776 "cannot replace %s with %s"), old_disk, new_disk);
1777 else
1778 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN,
1779 "cannot attach %s to %s"), new_disk, old_disk);
1780
1781 /*
1782 * If this is a root pool, make sure that we're not attaching an
1783 * EFI labeled device.
1784 */
1785 if (rootpool && pool_uses_efi(nvroot)) {
1786 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1787 "EFI labeled devices are not supported on root pools."));
1788 return (zfs_error(hdl, EZFS_POOL_NOTSUP, msg));
1789 }
1790
1791 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
1792 if ((tgt = zpool_find_vdev(zhp, old_disk, &avail_spare, &l2cache,
1793 &islog)) == 0)
1794 return (zfs_error(hdl, EZFS_NODEVICE, msg));
1795
1796 if (avail_spare)
1797 return (zfs_error(hdl, EZFS_ISSPARE, msg));
1798
1799 if (l2cache)
1800 return (zfs_error(hdl, EZFS_ISL2CACHE, msg));
1801
1802 verify(nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID, &zc.zc_guid) == 0);
1803 zc.zc_cookie = replacing;
1804
1805 if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_CHILDREN,
1806 &child, &children) != 0 || children != 1) {
1807 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1808 "new device must be a single disk"));
1809 return (zfs_error(hdl, EZFS_INVALCONFIG, msg));
1810 }
1811
1812 verify(nvlist_lookup_nvlist(zpool_get_config(zhp, NULL),
1813 ZPOOL_CONFIG_VDEV_TREE, &config_root) == 0);
1814
1815 if ((newname = zpool_vdev_name(NULL, NULL, child[0])) == NULL)
1816 return (-1);
1817
1818 /*
1819 * If the target is a hot spare that has been swapped in, we can only
1820 * replace it with another hot spare.
1821 */
1822 if (replacing &&
1823 nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_IS_SPARE, &val) == 0 &&
1824 (zpool_find_vdev(zhp, newname, &avail_spare, &l2cache,
1825 NULL) == NULL || !avail_spare) &&
1826 is_replacing_spare(config_root, tgt, 1)) {
1827 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1828 "can only be replaced by another hot spare"));
1829 free(newname);
1830 return (zfs_error(hdl, EZFS_BADTARGET, msg));
1831 }
1832
1833 /*
1834 * If we are attempting to replace a spare, it canot be applied to an
1835 * already spared device.
1836 */
1837 if (replacing &&
1838 nvlist_lookup_string(child[0], ZPOOL_CONFIG_PATH, &path) == 0 &&
1839 zpool_find_vdev(zhp, newname, &avail_spare,
1840 &l2cache, NULL) != NULL && avail_spare &&
1841 is_replacing_spare(config_root, tgt, 0)) {
1842 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1843 "device has already been replaced with a spare"));
1844 free(newname);
1845 return (zfs_error(hdl, EZFS_BADTARGET, msg));
1846 }
1847
1848 free(newname);
1849
1850 if (zcmd_write_conf_nvlist(hdl, &zc, nvroot) != 0)
1851 return (-1);
1852
1853 ret = zfs_ioctl(zhp->zpool_hdl, ZFS_IOC_VDEV_ATTACH, &zc);
1854
1855 zcmd_free_nvlists(&zc);
1856
1857 if (ret == 0) {
1858 if (rootpool) {
1859 /*
1860 * XXX - This should be removed once we can
1861 * automatically install the bootblocks on the
1862 * newly attached disk.
1863 */
1864 (void) fprintf(stderr, dgettext(TEXT_DOMAIN, "Please "
1865 "be sure to invoke %s to make '%s' bootable.\n"),
1866 BOOTCMD, new_disk);
1867 }
1868 return (0);
1869 }
1870
1871 switch (errno) {
1872 case ENOTSUP:
1873 /*
1874 * Can't attach to or replace this type of vdev.
1875 */
1876 if (replacing) {
1877 if (islog)
1878 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1879 "cannot replace a log with a spare"));
1880 else
1881 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1882 "cannot replace a replacing device"));
1883 } else {
1884 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1885 "can only attach to mirrors and top-level "
1886 "disks"));
1887 }
1888 (void) zfs_error(hdl, EZFS_BADTARGET, msg);
1889 break;
1890
1891 case EINVAL:
1892 /*
1893 * The new device must be a single disk.
1894 */
1895 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1896 "new device must be a single disk"));
1897 (void) zfs_error(hdl, EZFS_INVALCONFIG, msg);
1898 break;
1899
1900 case EBUSY:
1901 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "%s is busy"),
1902 new_disk);
1903 (void) zfs_error(hdl, EZFS_BADDEV, msg);
1904 break;
1905
1906 case EOVERFLOW:
1907 /*
1908 * The new device is too small.
1909 */
1910 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1911 "device is too small"));
1912 (void) zfs_error(hdl, EZFS_BADDEV, msg);
1913 break;
1914
1915 case EDOM:
1916 /*
1917 * The new device has a different alignment requirement.
1918 */
1919 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1920 "devices have different sector alignment"));
1921 (void) zfs_error(hdl, EZFS_BADDEV, msg);
1922 break;
1923
1924 case ENAMETOOLONG:
1925 /*
1926 * The resulting top-level vdev spec won't fit in the label.
1927 */
1928 (void) zfs_error(hdl, EZFS_DEVOVERFLOW, msg);
1929 break;
1930
1931 default:
1932 (void) zpool_standard_error(hdl, errno, msg);
1933 }
1934
1935 return (-1);
1936 }
1937
1938 /*
1939 * Detach the specified device.
1940 */
1941 int
1942 zpool_vdev_detach(zpool_handle_t *zhp, const char *path)
1943 {
1944 zfs_cmd_t zc = { 0 };
1945 char msg[1024];
1946 nvlist_t *tgt;
1947 boolean_t avail_spare, l2cache;
1948 libzfs_handle_t *hdl = zhp->zpool_hdl;
1949
1950 (void) snprintf(msg, sizeof (msg),
1951 dgettext(TEXT_DOMAIN, "cannot detach %s"), path);
1952
1953 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
1954 if ((tgt = zpool_find_vdev(zhp, path, &avail_spare, &l2cache,
1955 NULL)) == 0)
1956 return (zfs_error(hdl, EZFS_NODEVICE, msg));
1957
1958 if (avail_spare)
1959 return (zfs_error(hdl, EZFS_ISSPARE, msg));
1960
1961 if (l2cache)
1962 return (zfs_error(hdl, EZFS_ISL2CACHE, msg));
1963
1964 verify(nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID, &zc.zc_guid) == 0);
1965
1966 if (zfs_ioctl(hdl, ZFS_IOC_VDEV_DETACH, &zc) == 0)
1967 return (0);
1968
1969 switch (errno) {
1970
1971 case ENOTSUP:
1972 /*
1973 * Can't detach from this type of vdev.
1974 */
1975 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "only "
1976 "applicable to mirror and replacing vdevs"));
1977 (void) zfs_error(zhp->zpool_hdl, EZFS_BADTARGET, msg);
1978 break;
1979
1980 case EBUSY:
1981 /*
1982 * There are no other replicas of this device.
1983 */
1984 (void) zfs_error(hdl, EZFS_NOREPLICAS, msg);
1985 break;
1986
1987 default:
1988 (void) zpool_standard_error(hdl, errno, msg);
1989 }
1990
1991 return (-1);
1992 }
1993
1994 /*
1995 * Remove the given device. Currently, this is supported only for hot spares
1996 * and level 2 cache devices.
1997 */
1998 int
1999 zpool_vdev_remove(zpool_handle_t *zhp, const char *path)
2000 {
2001 zfs_cmd_t zc = { 0 };
2002 char msg[1024];
2003 nvlist_t *tgt;
2004 boolean_t avail_spare, l2cache;
2005 libzfs_handle_t *hdl = zhp->zpool_hdl;
2006
2007 (void) snprintf(msg, sizeof (msg),
2008 dgettext(TEXT_DOMAIN, "cannot remove %s"), path);
2009
2010 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
2011 if ((tgt = zpool_find_vdev(zhp, path, &avail_spare, &l2cache,
2012 NULL)) == 0)
2013 return (zfs_error(hdl, EZFS_NODEVICE, msg));
2014
2015 if (!avail_spare && !l2cache) {
2016 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2017 "only inactive hot spares or cache devices "
2018 "can be removed"));
2019 return (zfs_error(hdl, EZFS_NODEVICE, msg));
2020 }
2021
2022 verify(nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID, &zc.zc_guid) == 0);
2023
2024 if (zfs_ioctl(hdl, ZFS_IOC_VDEV_REMOVE, &zc) == 0)
2025 return (0);
2026
2027 return (zpool_standard_error(hdl, errno, msg));
2028 }
2029
2030 /*
2031 * Clear the errors for the pool, or the particular device if specified.
2032 */
2033 int
2034 zpool_clear(zpool_handle_t *zhp, const char *path)
2035 {
2036 zfs_cmd_t zc = { 0 };
2037 char msg[1024];
2038 nvlist_t *tgt;
2039 boolean_t avail_spare, l2cache;
2040 libzfs_handle_t *hdl = zhp->zpool_hdl;
2041
2042 if (path)
2043 (void) snprintf(msg, sizeof (msg),
2044 dgettext(TEXT_DOMAIN, "cannot clear errors for %s"),
2045 path);
2046 else
2047 (void) snprintf(msg, sizeof (msg),
2048 dgettext(TEXT_DOMAIN, "cannot clear errors for %s"),
2049 zhp->zpool_name);
2050
2051 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
2052 if (path) {
2053 if ((tgt = zpool_find_vdev(zhp, path, &avail_spare,
2054 &l2cache, NULL)) == 0)
2055 return (zfs_error(hdl, EZFS_NODEVICE, msg));
2056
2057 /*
2058 * Don't allow error clearing for hot spares. Do allow
2059 * error clearing for l2cache devices.
2060 */
2061 if (avail_spare)
2062 return (zfs_error(hdl, EZFS_ISSPARE, msg));
2063
2064 verify(nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID,
2065 &zc.zc_guid) == 0);
2066 }
2067
2068 if (zfs_ioctl(hdl, ZFS_IOC_CLEAR, &zc) == 0)
2069 return (0);
2070
2071 return (zpool_standard_error(hdl, errno, msg));
2072 }
2073
2074 /*
2075 * Similar to zpool_clear(), but takes a GUID (used by fmd).
2076 */
2077 int
2078 zpool_vdev_clear(zpool_handle_t *zhp, uint64_t guid)
2079 {
2080 zfs_cmd_t zc = { 0 };
2081 char msg[1024];
2082 libzfs_handle_t *hdl = zhp->zpool_hdl;
2083
2084 (void) snprintf(msg, sizeof (msg),
2085 dgettext(TEXT_DOMAIN, "cannot clear errors for %llx"),
2086 guid);
2087
2088 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
2089 zc.zc_guid = guid;
2090
2091 if (ioctl(hdl->libzfs_fd, ZFS_IOC_CLEAR, &zc) == 0)
2092 return (0);
2093
2094 return (zpool_standard_error(hdl, errno, msg));
2095 }
2096
2097 /*
2098 * Iterate over all zvols in a given pool by walking the /dev/zvol/dsk/<pool>
2099 * hierarchy.
2100 */
2101 int
2102 zpool_iter_zvol(zpool_handle_t *zhp, int (*cb)(const char *, void *),
2103 void *data)
2104 {
2105 libzfs_handle_t *hdl = zhp->zpool_hdl;
2106 char (*paths)[MAXPATHLEN];
2107 size_t size = 4;
2108 int curr, fd, base, ret = 0;
2109 DIR *dirp;
2110 struct dirent *dp;
2111 struct stat st;
2112
2113 if ((base = open("/dev/zvol/dsk", O_RDONLY)) < 0)
2114 return (errno == ENOENT ? 0 : -1);
2115
2116 if (fstatat(base, zhp->zpool_name, &st, 0) != 0) {
2117 int err = errno;
2118 (void) close(base);
2119 return (err == ENOENT ? 0 : -1);
2120 }
2121
2122 /*
2123 * Oddly this wasn't a directory -- ignore that failure since we
2124 * know there are no links lower in the (non-existant) hierarchy.
2125 */
2126 if (!S_ISDIR(st.st_mode)) {
2127 (void) close(base);
2128 return (0);
2129 }
2130
2131 if ((paths = zfs_alloc(hdl, size * sizeof (paths[0]))) == NULL) {
2132 (void) close(base);
2133 return (-1);
2134 }
2135
2136 (void) strlcpy(paths[0], zhp->zpool_name, sizeof (paths[0]));
2137 curr = 0;
2138
2139 while (curr >= 0) {
2140 if (fstatat(base, paths[curr], &st, AT_SYMLINK_NOFOLLOW) != 0)
2141 goto err;
2142
2143 if (S_ISDIR(st.st_mode)) {
2144 if ((fd = openat(base, paths[curr], O_RDONLY)) < 0)
2145 goto err;
2146
2147 if ((dirp = fdopendir(fd)) == NULL) {
2148 (void) close(fd);
2149 goto err;
2150 }
2151
2152 while ((dp = readdir(dirp)) != NULL) {
2153 if (dp->d_name[0] == '.')
2154 continue;
2155
2156 if (curr + 1 == size) {
2157 paths = zfs_realloc(hdl, paths,
2158 size * sizeof (paths[0]),
2159 size * 2 * sizeof (paths[0]));
2160 if (paths == NULL) {
2161 (void) closedir(dirp);
2162 (void) close(fd);
2163 goto err;
2164 }
2165
2166 size *= 2;
2167 }
2168
2169 (void) strlcpy(paths[curr + 1], paths[curr],
2170 sizeof (paths[curr + 1]));
2171 (void) strlcat(paths[curr], "/",
2172 sizeof (paths[curr]));
2173 (void) strlcat(paths[curr], dp->d_name,
2174 sizeof (paths[curr]));
2175 curr++;
2176 }
2177
2178 (void) closedir(dirp);
2179
2180 } else {
2181 if ((ret = cb(paths[curr], data)) != 0)
2182 break;
2183 }
2184
2185 curr--;
2186 }
2187
2188 free(paths);
2189 (void) close(base);
2190
2191 return (ret);
2192
2193 err:
2194 free(paths);
2195 (void) close(base);
2196 return (-1);
2197 }
2198
2199 typedef struct zvol_cb {
2200 zpool_handle_t *zcb_pool;
2201 boolean_t zcb_create;
2202 } zvol_cb_t;
2203
2204 /*ARGSUSED*/
2205 static int
2206 do_zvol_create(zfs_handle_t *zhp, void *data)
2207 {
2208 int ret = 0;
2209
2210 if (ZFS_IS_VOLUME(zhp)) {
2211 (void) zvol_create_link(zhp->zfs_hdl, zhp->zfs_name);
2212 ret = zfs_iter_snapshots(zhp, do_zvol_create, NULL);
2213 }
2214
2215 if (ret == 0)
2216 ret = zfs_iter_filesystems(zhp, do_zvol_create, NULL);
2217
2218 zfs_close(zhp);
2219
2220 return (ret);
2221 }
2222
2223 /*
2224 * Iterate over all zvols in the pool and make any necessary minor nodes.
2225 */
2226 int
2227 zpool_create_zvol_links(zpool_handle_t *zhp)
2228 {
2229 zfs_handle_t *zfp;
2230 int ret;
2231
2232 /*
2233 * If the pool is unavailable, just return success.
2234 */
2235 if ((zfp = make_dataset_handle(zhp->zpool_hdl,
2236 zhp->zpool_name)) == NULL)
2237 return (0);
2238
2239 ret = zfs_iter_filesystems(zfp, do_zvol_create, NULL);
2240
2241 zfs_close(zfp);
2242 return (ret);
2243 }
2244
2245 static int
2246 do_zvol_remove(const char *dataset, void *data)
2247 {
2248 zpool_handle_t *zhp = data;
2249
2250 return (zvol_remove_link(zhp->zpool_hdl, dataset));
2251 }
2252
2253 /*
2254 * Iterate over all zvols in the pool and remove any minor nodes. We iterate
2255 * by examining the /dev links so that a corrupted pool doesn't impede this
2256 * operation.
2257 */
2258 int
2259 zpool_remove_zvol_links(zpool_handle_t *zhp)
2260 {
2261 return (zpool_iter_zvol(zhp, do_zvol_remove, zhp));
2262 }
2263
2264 /*
2265 * Convert from a devid string to a path.
2266 */
2267 static char *
2268 devid_to_path(char *devid_str)
2269 {
2270 ddi_devid_t devid;
2271 char *minor;
2272 char *path;
2273 devid_nmlist_t *list = NULL;
2274 int ret;
2275
2276 if (devid_str_decode(devid_str, &devid, &minor) != 0)
2277 return (NULL);
2278
2279 ret = devid_deviceid_to_nmlist("/dev", devid, minor, &list);
2280
2281 devid_str_free(minor);
2282 devid_free(devid);
2283
2284 if (ret != 0)
2285 return (NULL);
2286
2287 if ((path = strdup(list[0].devname)) == NULL)
2288 return (NULL);
2289
2290 devid_free_nmlist(list);
2291
2292 return (path);
2293 }
2294
2295 /*
2296 * Convert from a path to a devid string.
2297 */
2298 static char *
2299 path_to_devid(const char *path)
2300 {
2301 int fd;
2302 ddi_devid_t devid;
2303 char *minor, *ret;
2304
2305 if ((fd = open(path, O_RDONLY)) < 0)
2306 return (NULL);
2307
2308 minor = NULL;
2309 ret = NULL;
2310 if (devid_get(fd, &devid) == 0) {
2311 if (devid_get_minor_name(fd, &minor) == 0)
2312 ret = devid_str_encode(devid, minor);
2313 if (minor != NULL)
2314 devid_str_free(minor);
2315 devid_free(devid);
2316 }
2317 (void) close(fd);
2318
2319 return (ret);
2320 }
2321
2322 /*
2323 * Issue the necessary ioctl() to update the stored path value for the vdev. We
2324 * ignore any failure here, since a common case is for an unprivileged user to
2325 * type 'zpool status', and we'll display the correct information anyway.
2326 */
2327 static void
2328 set_path(zpool_handle_t *zhp, nvlist_t *nv, const char *path)
2329 {
2330 zfs_cmd_t zc = { 0 };
2331
2332 (void) strncpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
2333 (void) strncpy(zc.zc_value, path, sizeof (zc.zc_value));
2334 verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID,
2335 &zc.zc_guid) == 0);
2336
2337 (void) ioctl(zhp->zpool_hdl->libzfs_fd, ZFS_IOC_VDEV_SETPATH, &zc);
2338 }
2339
2340 /*
2341 * Given a vdev, return the name to display in iostat. If the vdev has a path,
2342 * we use that, stripping off any leading "/dev/dsk/"; if not, we use the type.
2343 * We also check if this is a whole disk, in which case we strip off the
2344 * trailing 's0' slice name.
2345 *
2346 * This routine is also responsible for identifying when disks have been
2347 * reconfigured in a new location. The kernel will have opened the device by
2348 * devid, but the path will still refer to the old location. To catch this, we
2349 * first do a path -> devid translation (which is fast for the common case). If
2350 * the devid matches, we're done. If not, we do a reverse devid -> path
2351 * translation and issue the appropriate ioctl() to update the path of the vdev.
2352 * If 'zhp' is NULL, then this is an exported pool, and we don't need to do any
2353 * of these checks.
2354 */
2355 char *
2356 zpool_vdev_name(libzfs_handle_t *hdl, zpool_handle_t *zhp, nvlist_t *nv)
2357 {
2358 char *path, *devid;
2359 uint64_t value;
2360 char buf[64];
2361 vdev_stat_t *vs;
2362 uint_t vsc;
2363
2364 if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_NOT_PRESENT,
2365 &value) == 0) {
2366 verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID,
2367 &value) == 0);
2368 (void) snprintf(buf, sizeof (buf), "%llu",
2369 (u_longlong_t)value);
2370 path = buf;
2371 } else if (nvlist_lookup_string(nv, ZPOOL_CONFIG_PATH, &path) == 0) {
2372
2373 /*
2374 * If the device is dead (faulted, offline, etc) then don't
2375 * bother opening it. Otherwise we may be forcing the user to
2376 * open a misbehaving device, which can have undesirable
2377 * effects.
2378 */
2379 if ((nvlist_lookup_uint64_array(nv, ZPOOL_CONFIG_STATS,
2380 (uint64_t **)&vs, &vsc) != 0 ||
2381 vs->vs_state >= VDEV_STATE_DEGRADED) &&
2382 zhp != NULL &&
2383 nvlist_lookup_string(nv, ZPOOL_CONFIG_DEVID, &devid) == 0) {
2384 /*
2385 * Determine if the current path is correct.
2386 */
2387 char *newdevid = path_to_devid(path);
2388
2389 if (newdevid == NULL ||
2390 strcmp(devid, newdevid) != 0) {
2391 char *newpath;
2392
2393 if ((newpath = devid_to_path(devid)) != NULL) {
2394 /*
2395 * Update the path appropriately.
2396 */
2397 set_path(zhp, nv, newpath);
2398 if (nvlist_add_string(nv,
2399 ZPOOL_CONFIG_PATH, newpath) == 0)
2400 verify(nvlist_lookup_string(nv,
2401 ZPOOL_CONFIG_PATH,
2402 &path) == 0);
2403 free(newpath);
2404 }
2405 }
2406
2407 if (newdevid)
2408 devid_str_free(newdevid);
2409 }
2410
2411 if (strncmp(path, "/dev/dsk/", 9) == 0)
2412 path += 9;
2413
2414 if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_WHOLE_DISK,
2415 &value) == 0 && value) {
2416 char *tmp = zfs_strdup(hdl, path);
2417 if (tmp == NULL)
2418 return (NULL);
2419 tmp[strlen(path) - 2] = '\0';
2420 return (tmp);
2421 }
2422 } else {
2423 verify(nvlist_lookup_string(nv, ZPOOL_CONFIG_TYPE, &path) == 0);
2424
2425 /*
2426 * If it's a raidz device, we need to stick in the parity level.
2427 */
2428 if (strcmp(path, VDEV_TYPE_RAIDZ) == 0) {
2429 verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_NPARITY,
2430 &value) == 0);
2431 (void) snprintf(buf, sizeof (buf), "%s%llu", path,
2432 (u_longlong_t)value);
2433 path = buf;
2434 }
2435 }
2436
2437 return (zfs_strdup(hdl, path));
2438 }
2439
2440 static int
2441 zbookmark_compare(const void *a, const void *b)
2442 {
2443 return (memcmp(a, b, sizeof (zbookmark_t)));
2444 }
2445
2446 /*
2447 * Retrieve the persistent error log, uniquify the members, and return to the
2448 * caller.
2449 */
2450 int
2451 zpool_get_errlog(zpool_handle_t *zhp, nvlist_t **nverrlistp)
2452 {
2453 zfs_cmd_t zc = { 0 };
2454 uint64_t count;
2455 zbookmark_t *zb = NULL;
2456 int i;
2457
2458 /*
2459 * Retrieve the raw error list from the kernel. If the number of errors
2460 * has increased, allocate more space and continue until we get the
2461 * entire list.
2462 */
2463 verify(nvlist_lookup_uint64(zhp->zpool_config, ZPOOL_CONFIG_ERRCOUNT,
2464 &count) == 0);
2465 if (count == 0)
2466 return (0);
2467 if ((zc.zc_nvlist_dst = (uintptr_t)zfs_alloc(zhp->zpool_hdl,
2468 count * sizeof (zbookmark_t))) == (uintptr_t)NULL)
2469 return (-1);
2470 zc.zc_nvlist_dst_size = count;
2471 (void) strcpy(zc.zc_name, zhp->zpool_name);
2472 for (;;) {
2473 if (ioctl(zhp->zpool_hdl->libzfs_fd, ZFS_IOC_ERROR_LOG,
2474 &zc) != 0) {
2475 free((void *)(uintptr_t)zc.zc_nvlist_dst);
2476 if (errno == ENOMEM) {
2477 count = zc.zc_nvlist_dst_size;
2478 if ((zc.zc_nvlist_dst = (uintptr_t)
2479 zfs_alloc(zhp->zpool_hdl, count *
2480 sizeof (zbookmark_t))) == (uintptr_t)NULL)
2481 return (-1);
2482 } else {
2483 return (-1);
2484 }
2485 } else {
2486 break;
2487 }
2488 }
2489
2490 /*
2491 * Sort the resulting bookmarks. This is a little confusing due to the
2492 * implementation of ZFS_IOC_ERROR_LOG. The bookmarks are copied last
2493 * to first, and 'zc_nvlist_dst_size' indicates the number of boomarks
2494 * _not_ copied as part of the process. So we point the start of our
2495 * array appropriate and decrement the total number of elements.
2496 */
2497 zb = ((zbookmark_t *)(uintptr_t)zc.zc_nvlist_dst) +
2498 zc.zc_nvlist_dst_size;
2499 count -= zc.zc_nvlist_dst_size;
2500
2501 qsort(zb, count, sizeof (zbookmark_t), zbookmark_compare);
2502
2503 verify(nvlist_alloc(nverrlistp, 0, KM_SLEEP) == 0);
2504
2505 /*
2506 * Fill in the nverrlistp with nvlist's of dataset and object numbers.
2507 */
2508 for (i = 0; i < count; i++) {
2509 nvlist_t *nv;
2510
2511 /* ignoring zb_blkid and zb_level for now */
2512 if (i > 0 && zb[i-1].zb_objset == zb[i].zb_objset &&
2513 zb[i-1].zb_object == zb[i].zb_object)
2514 continue;
2515
2516 if (nvlist_alloc(&nv, NV_UNIQUE_NAME, KM_SLEEP) != 0)
2517 goto nomem;
2518 if (nvlist_add_uint64(nv, ZPOOL_ERR_DATASET,
2519 zb[i].zb_objset) != 0) {
2520 nvlist_free(nv);
2521 goto nomem;
2522 }
2523 if (nvlist_add_uint64(nv, ZPOOL_ERR_OBJECT,
2524 zb[i].zb_object) != 0) {
2525 nvlist_free(nv);
2526 goto nomem;
2527 }
2528 if (nvlist_add_nvlist(*nverrlistp, "ejk", nv) != 0) {
2529 nvlist_free(nv);
2530 goto nomem;
2531 }
2532 nvlist_free(nv);
2533 }
2534
2535 free((void *)(uintptr_t)zc.zc_nvlist_dst);
2536 return (0);
2537
2538 nomem:
2539 free((void *)(uintptr_t)zc.zc_nvlist_dst);
2540 return (no_memory(zhp->zpool_hdl));
2541 }
2542
2543 /*
2544 * Upgrade a ZFS pool to the latest on-disk version.
2545 */
2546 int
2547 zpool_upgrade(zpool_handle_t *zhp, uint64_t new_version)
2548 {
2549 zfs_cmd_t zc = { 0 };
2550 libzfs_handle_t *hdl = zhp->zpool_hdl;
2551
2552 (void) strcpy(zc.zc_name, zhp->zpool_name);
2553 zc.zc_cookie = new_version;
2554
2555 if (zfs_ioctl(hdl, ZFS_IOC_POOL_UPGRADE, &zc) != 0)
2556 return (zpool_standard_error_fmt(hdl, errno,
2557 dgettext(TEXT_DOMAIN, "cannot upgrade '%s'"),
2558 zhp->zpool_name));
2559 return (0);
2560 }
2561
2562 void
2563 zpool_set_history_str(const char *subcommand, int argc, char **argv,
2564 char *history_str)
2565 {
2566 int i;
2567
2568 (void) strlcpy(history_str, subcommand, HIS_MAX_RECORD_LEN);
2569 for (i = 1; i < argc; i++) {
2570 if (strlen(history_str) + 1 + strlen(argv[i]) >
2571 HIS_MAX_RECORD_LEN)
2572 break;
2573 (void) strlcat(history_str, " ", HIS_MAX_RECORD_LEN);
2574 (void) strlcat(history_str, argv[i], HIS_MAX_RECORD_LEN);
2575 }
2576 }
2577
2578 /*
2579 * Stage command history for logging.
2580 */
2581 int
2582 zpool_stage_history(libzfs_handle_t *hdl, const char *history_str)
2583 {
2584 if (history_str == NULL)
2585 return (EINVAL);
2586
2587 if (strlen(history_str) > HIS_MAX_RECORD_LEN)
2588 return (EINVAL);
2589
2590 if (hdl->libzfs_log_str != NULL)
2591 free(hdl->libzfs_log_str);
2592
2593 if ((hdl->libzfs_log_str = strdup(history_str)) == NULL)
2594 return (no_memory(hdl));
2595
2596 return (0);
2597 }
2598
2599 /*
2600 * Perform ioctl to get some command history of a pool.
2601 *
2602 * 'buf' is the buffer to fill up to 'len' bytes. 'off' is the
2603 * logical offset of the history buffer to start reading from.
2604 *
2605 * Upon return, 'off' is the next logical offset to read from and
2606 * 'len' is the actual amount of bytes read into 'buf'.
2607 */
2608 static int
2609 get_history(zpool_handle_t *zhp, char *buf, uint64_t *off, uint64_t *len)
2610 {
2611 zfs_cmd_t zc = { 0 };
2612 libzfs_handle_t *hdl = zhp->zpool_hdl;
2613
2614 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
2615
2616 zc.zc_history = (uint64_t)(uintptr_t)buf;
2617 zc.zc_history_len = *len;
2618 zc.zc_history_offset = *off;
2619
2620 if (ioctl(hdl->libzfs_fd, ZFS_IOC_POOL_GET_HISTORY, &zc) != 0) {
2621 switch (errno) {
2622 case EPERM:
2623 return (zfs_error_fmt(hdl, EZFS_PERM,
2624 dgettext(TEXT_DOMAIN,
2625 "cannot show history for pool '%s'"),
2626 zhp->zpool_name));
2627 case ENOENT:
2628 return (zfs_error_fmt(hdl, EZFS_NOHISTORY,
2629 dgettext(TEXT_DOMAIN, "cannot get history for pool "
2630 "'%s'"), zhp->zpool_name));
2631 case ENOTSUP:
2632 return (zfs_error_fmt(hdl, EZFS_BADVERSION,
2633 dgettext(TEXT_DOMAIN, "cannot get history for pool "
2634 "'%s', pool must be upgraded"), zhp->zpool_name));
2635 default:
2636 return (zpool_standard_error_fmt(hdl, errno,
2637 dgettext(TEXT_DOMAIN,
2638 "cannot get history for '%s'"), zhp->zpool_name));
2639 }
2640 }
2641
2642 *len = zc.zc_history_len;
2643 *off = zc.zc_history_offset;
2644
2645 return (0);
2646 }
2647
2648 /*
2649 * Process the buffer of nvlists, unpacking and storing each nvlist record
2650 * into 'records'. 'leftover' is set to the number of bytes that weren't
2651 * processed as there wasn't a complete record.
2652 */
2653 static int
2654 zpool_history_unpack(char *buf, uint64_t bytes_read, uint64_t *leftover,
2655 nvlist_t ***records, uint_t *numrecords)
2656 {
2657 uint64_t reclen;
2658 nvlist_t *nv;
2659 int i;
2660
2661 while (bytes_read > sizeof (reclen)) {
2662
2663 /* get length of packed record (stored as little endian) */
2664 for (i = 0, reclen = 0; i < sizeof (reclen); i++)
2665 reclen += (uint64_t)(((uchar_t *)buf)[i]) << (8*i);
2666
2667 if (bytes_read < sizeof (reclen) + reclen)
2668 break;
2669
2670 /* unpack record */
2671 if (nvlist_unpack(buf + sizeof (reclen), reclen, &nv, 0) != 0)
2672 return (ENOMEM);
2673 bytes_read -= sizeof (reclen) + reclen;
2674 buf += sizeof (reclen) + reclen;
2675
2676 /* add record to nvlist array */
2677 (*numrecords)++;
2678 if (ISP2(*numrecords + 1)) {
2679 *records = realloc(*records,
2680 *numrecords * 2 * sizeof (nvlist_t *));
2681 }
2682 (*records)[*numrecords - 1] = nv;
2683 }
2684
2685 *leftover = bytes_read;
2686 return (0);
2687 }
2688
2689 #define HIS_BUF_LEN (128*1024)
2690
2691 /*
2692 * Retrieve the command history of a pool.
2693 */
2694 int
2695 zpool_get_history(zpool_handle_t *zhp, nvlist_t **nvhisp)
2696 {
2697 char buf[HIS_BUF_LEN];
2698 uint64_t off = 0;
2699 nvlist_t **records = NULL;
2700 uint_t numrecords = 0;
2701 int err, i;
2702
2703 do {
2704 uint64_t bytes_read = sizeof (buf);
2705 uint64_t leftover;
2706
2707 if ((err = get_history(zhp, buf, &off, &bytes_read)) != 0)
2708 break;
2709
2710 /* if nothing else was read in, we're at EOF, just return */
2711 if (!bytes_read)
2712 break;
2713
2714 if ((err = zpool_history_unpack(buf, bytes_read,
2715 &leftover, &records, &numrecords)) != 0)
2716 break;
2717 off -= leftover;
2718
2719 /* CONSTCOND */
2720 } while (1);
2721
2722 if (!err) {
2723 verify(nvlist_alloc(nvhisp, NV_UNIQUE_NAME, 0) == 0);
2724 verify(nvlist_add_nvlist_array(*nvhisp, ZPOOL_HIST_RECORD,
2725 records, numrecords) == 0);
2726 }
2727 for (i = 0; i < numrecords; i++)
2728 nvlist_free(records[i]);
2729 free(records);
2730
2731 return (err);
2732 }
2733
2734 void
2735 zpool_obj_to_path(zpool_handle_t *zhp, uint64_t dsobj, uint64_t obj,
2736 char *pathname, size_t len)
2737 {
2738 zfs_cmd_t zc = { 0 };
2739 boolean_t mounted = B_FALSE;
2740 char *mntpnt = NULL;
2741 char dsname[MAXNAMELEN];
2742
2743 if (dsobj == 0) {
2744 /* special case for the MOS */
2745 (void) snprintf(pathname, len, "<metadata>:<0x%llx>", obj);
2746 return;
2747 }
2748
2749 /* get the dataset's name */
2750 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
2751 zc.zc_obj = dsobj;
2752 if (ioctl(zhp->zpool_hdl->libzfs_fd,
2753 ZFS_IOC_DSOBJ_TO_DSNAME, &zc) != 0) {
2754 /* just write out a path of two object numbers */
2755 (void) snprintf(pathname, len, "<0x%llx>:<0x%llx>",
2756 dsobj, obj);
2757 return;
2758 }
2759 (void) strlcpy(dsname, zc.zc_value, sizeof (dsname));
2760
2761 /* find out if the dataset is mounted */
2762 mounted = is_mounted(zhp->zpool_hdl, dsname, &mntpnt);
2763
2764 /* get the corrupted object's path */
2765 (void) strlcpy(zc.zc_name, dsname, sizeof (zc.zc_name));
2766 zc.zc_obj = obj;
2767 if (ioctl(zhp->zpool_hdl->libzfs_fd, ZFS_IOC_OBJ_TO_PATH,
2768 &zc) == 0) {
2769 if (mounted) {
2770 (void) snprintf(pathname, len, "%s%s", mntpnt,
2771 zc.zc_value);
2772 } else {
2773 (void) snprintf(pathname, len, "%s:%s",
2774 dsname, zc.zc_value);
2775 }
2776 } else {
2777 (void) snprintf(pathname, len, "%s:<0x%llx>", dsname, obj);
2778 }
2779 free(mntpnt);
2780 }
2781
2782 #define RDISK_ROOT "/dev/rdsk"
2783 #define BACKUP_SLICE "s2"
2784 /*
2785 * Don't start the slice at the default block of 34; many storage
2786 * devices will use a stripe width of 128k, so start there instead.
2787 */
2788 #define NEW_START_BLOCK 256
2789
2790 /*
2791 * Read the EFI label from the config, if a label does not exist then
2792 * pass back the error to the caller. If the caller has passed a non-NULL
2793 * diskaddr argument then we set it to the starting address of the EFI
2794 * partition.
2795 */
2796 static int
2797 read_efi_label(nvlist_t *config, diskaddr_t *sb)
2798 {
2799 char *path;
2800 int fd;
2801 char diskname[MAXPATHLEN];
2802 int err = -1;
2803
2804 if (nvlist_lookup_string(config, ZPOOL_CONFIG_PATH, &path) != 0)
2805 return (err);
2806
2807 (void) snprintf(diskname, sizeof (diskname), "%s%s", RDISK_ROOT,
2808 strrchr(path, '/'));
2809 if ((fd = open(diskname, O_RDONLY|O_NDELAY)) >= 0) {
2810 struct dk_gpt *vtoc;
2811
2812 if ((err = efi_alloc_and_read(fd, &vtoc)) >= 0) {
2813 if (sb != NULL)
2814 *sb = vtoc->efi_parts[0].p_start;
2815 efi_free(vtoc);
2816 }
2817 (void) close(fd);
2818 }
2819 return (err);
2820 }
2821
2822 /*
2823 * determine where a partition starts on a disk in the current
2824 * configuration
2825 */
2826 static diskaddr_t
2827 find_start_block(nvlist_t *config)
2828 {
2829 nvlist_t **child;
2830 uint_t c, children;
2831 diskaddr_t sb = MAXOFFSET_T;
2832 uint64_t wholedisk;
2833
2834 if (nvlist_lookup_nvlist_array(config,
2835 ZPOOL_CONFIG_CHILDREN, &child, &children) != 0) {
2836 if (nvlist_lookup_uint64(config,
2837 ZPOOL_CONFIG_WHOLE_DISK,
2838 &wholedisk) != 0 || !wholedisk) {
2839 return (MAXOFFSET_T);
2840 }
2841 if (read_efi_label(config, &sb) < 0)
2842 sb = MAXOFFSET_T;
2843 return (sb);
2844 }
2845
2846 for (c = 0; c < children; c++) {
2847 sb = find_start_block(child[c]);
2848 if (sb != MAXOFFSET_T) {
2849 return (sb);
2850 }
2851 }
2852 return (MAXOFFSET_T);
2853 }
2854
2855 /*
2856 * Label an individual disk. The name provided is the short name,
2857 * stripped of any leading /dev path.
2858 */
2859 int
2860 zpool_label_disk(libzfs_handle_t *hdl, zpool_handle_t *zhp, char *name)
2861 {
2862 char path[MAXPATHLEN];
2863 struct dk_gpt *vtoc;
2864 int fd;
2865 size_t resv = EFI_MIN_RESV_SIZE;
2866 uint64_t slice_size;
2867 diskaddr_t start_block;
2868 char errbuf[1024];
2869
2870 /* prepare an error message just in case */
2871 (void) snprintf(errbuf, sizeof (errbuf),
2872 dgettext(TEXT_DOMAIN, "cannot label '%s'"), name);
2873
2874 if (zhp) {
2875 nvlist_t *nvroot;
2876
2877 if (pool_is_bootable(zhp)) {
2878 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2879 "EFI labeled devices are not supported on root "
2880 "pools."));
2881 return (zfs_error(hdl, EZFS_POOL_NOTSUP, errbuf));
2882 }
2883
2884 verify(nvlist_lookup_nvlist(zhp->zpool_config,
2885 ZPOOL_CONFIG_VDEV_TREE, &nvroot) == 0);
2886
2887 if (zhp->zpool_start_block == 0)
2888 start_block = find_start_block(nvroot);
2889 else
2890 start_block = zhp->zpool_start_block;
2891 zhp->zpool_start_block = start_block;
2892 } else {
2893 /* new pool */
2894 start_block = NEW_START_BLOCK;
2895 }
2896
2897 (void) snprintf(path, sizeof (path), "%s/%s%s", RDISK_ROOT, name,
2898 BACKUP_SLICE);
2899
2900 if ((fd = open(path, O_RDWR | O_NDELAY)) < 0) {
2901 /*
2902 * This shouldn't happen. We've long since verified that this
2903 * is a valid device.
2904 */
2905 zfs_error_aux(hdl,
2906 dgettext(TEXT_DOMAIN, "unable to open device"));
2907 return (zfs_error(hdl, EZFS_OPENFAILED, errbuf));
2908 }
2909
2910 if (efi_alloc_and_init(fd, EFI_NUMPAR, &vtoc) != 0) {
2911 /*
2912 * The only way this can fail is if we run out of memory, or we
2913 * were unable to read the disk's capacity
2914 */
2915 if (errno == ENOMEM)
2916 (void) no_memory(hdl);
2917
2918 (void) close(fd);
2919 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2920 "unable to read disk capacity"), name);
2921
2922 return (zfs_error(hdl, EZFS_NOCAP, errbuf));
2923 }
2924
2925 slice_size = vtoc->efi_last_u_lba + 1;
2926 slice_size -= EFI_MIN_RESV_SIZE;
2927 if (start_block == MAXOFFSET_T)
2928 start_block = NEW_START_BLOCK;
2929 slice_size -= start_block;
2930
2931 vtoc->efi_parts[0].p_start = start_block;
2932 vtoc->efi_parts[0].p_size = slice_size;
2933
2934 /*
2935 * Why we use V_USR: V_BACKUP confuses users, and is considered
2936 * disposable by some EFI utilities (since EFI doesn't have a backup
2937 * slice). V_UNASSIGNED is supposed to be used only for zero size
2938 * partitions, and efi_write() will fail if we use it. V_ROOT, V_BOOT,
2939 * etc. were all pretty specific. V_USR is as close to reality as we
2940 * can get, in the absence of V_OTHER.
2941 */
2942 vtoc->efi_parts[0].p_tag = V_USR;
2943 (void) strcpy(vtoc->efi_parts[0].p_name, "zfs");
2944
2945 vtoc->efi_parts[8].p_start = slice_size + start_block;
2946 vtoc->efi_parts[8].p_size = resv;
2947 vtoc->efi_parts[8].p_tag = V_RESERVED;
2948
2949 if (efi_write(fd, vtoc) != 0) {
2950 /*
2951 * Some block drivers (like pcata) may not support EFI
2952 * GPT labels. Print out a helpful error message dir-
2953 * ecting the user to manually label the disk and give
2954 * a specific slice.
2955 */
2956 (void) close(fd);
2957 efi_free(vtoc);
2958
2959 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2960 "try using fdisk(1M) and then provide a specific slice"));
2961 return (zfs_error(hdl, EZFS_LABELFAILED, errbuf));
2962 }
2963
2964 (void) close(fd);
2965 efi_free(vtoc);
2966 return (0);
2967 }
2968
2969 static boolean_t
2970 supported_dump_vdev_type(libzfs_handle_t *hdl, nvlist_t *config, char *errbuf)
2971 {
2972 char *type;
2973 nvlist_t **child;
2974 uint_t children, c;
2975
2976 verify(nvlist_lookup_string(config, ZPOOL_CONFIG_TYPE, &type) == 0);
2977 if (strcmp(type, VDEV_TYPE_RAIDZ) == 0 ||
2978 strcmp(type, VDEV_TYPE_FILE) == 0 ||
2979 strcmp(type, VDEV_TYPE_LOG) == 0 ||
2980 strcmp(type, VDEV_TYPE_MISSING) == 0) {
2981 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2982 "vdev type '%s' is not supported"), type);
2983 (void) zfs_error(hdl, EZFS_VDEVNOTSUP, errbuf);
2984 return (B_FALSE);
2985 }
2986 if (nvlist_lookup_nvlist_array(config, ZPOOL_CONFIG_CHILDREN,
2987 &child, &children) == 0) {
2988 for (c = 0; c < children; c++) {
2989 if (!supported_dump_vdev_type(hdl, child[c], errbuf))
2990 return (B_FALSE);
2991 }
2992 }
2993 return (B_TRUE);
2994 }
2995
2996 /*
2997 * check if this zvol is allowable for use as a dump device; zero if
2998 * it is, > 0 if it isn't, < 0 if it isn't a zvol
2999 */
3000 int
3001 zvol_check_dump_config(char *arg)
3002 {
3003 zpool_handle_t *zhp = NULL;
3004 nvlist_t *config, *nvroot;
3005 char *p, *volname;
3006 nvlist_t **top;
3007 uint_t toplevels;
3008 libzfs_handle_t *hdl;
3009 char errbuf[1024];
3010 char poolname[ZPOOL_MAXNAMELEN];
3011 int pathlen = strlen(ZVOL_FULL_DEV_DIR);
3012 int ret = 1;
3013
3014 if (strncmp(arg, ZVOL_FULL_DEV_DIR, pathlen)) {
3015 return (-1);
3016 }
3017
3018 (void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN,
3019 "dump is not supported on device '%s'"), arg);
3020
3021 if ((hdl = libzfs_init()) == NULL)
3022 return (1);
3023 libzfs_print_on_error(hdl, B_TRUE);
3024
3025 volname = arg + pathlen;
3026
3027 /* check the configuration of the pool */
3028 if ((p = strchr(volname, '/')) == NULL) {
3029 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3030 "malformed dataset name"));
3031 (void) zfs_error(hdl, EZFS_INVALIDNAME, errbuf);
3032 return (1);
3033 } else if (p - volname >= ZFS_MAXNAMELEN) {
3034 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3035 "dataset name is too long"));
3036 (void) zfs_error(hdl, EZFS_NAMETOOLONG, errbuf);
3037 return (1);
3038 } else {
3039 (void) strncpy(poolname, volname, p - volname);
3040 poolname[p - volname] = '\0';
3041 }
3042
3043 if ((zhp = zpool_open(hdl, poolname)) == NULL) {
3044 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3045 "could not open pool '%s'"), poolname);
3046 (void) zfs_error(hdl, EZFS_OPENFAILED, errbuf);
3047 goto out;
3048 }
3049 config = zpool_get_config(zhp, NULL);
3050 if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE,
3051 &nvroot) != 0) {
3052 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3053 "could not obtain vdev configuration for '%s'"), poolname);
3054 (void) zfs_error(hdl, EZFS_INVALCONFIG, errbuf);
3055 goto out;
3056 }
3057
3058 verify(nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_CHILDREN,
3059 &top, &toplevels) == 0);
3060 if (toplevels != 1) {
3061 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3062 "'%s' has multiple top level vdevs"), poolname);
3063 (void) zfs_error(hdl, EZFS_DEVOVERFLOW, errbuf);
3064 goto out;
3065 }
3066
3067 if (!supported_dump_vdev_type(hdl, top[0], errbuf)) {
3068 goto out;
3069 }
3070 ret = 0;
3071
3072 out:
3073 if (zhp)
3074 zpool_close(zhp);
3075 libzfs_fini(hdl);
3076 return (ret);
3077 }
ZFS On Linux mirror for PVE