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 (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
24 * Copyright 2011 Nexenta Systems, Inc. All rights reserved.
25 * Copyright (c) 2013 by Delphix. All rights reserved.
26 * Copyright (c) 2013, Joyent, Inc. All rights reserved.
27 */
28
29 #include <ctype.h>
30 #include <errno.h>
31 #include <devid.h>
32 #include <fcntl.h>
33 #include <libintl.h>
34 #include <stdio.h>
35 #include <stdlib.h>
36 #include <strings.h>
37 #include <unistd.h>
38 #include <libgen.h>
39 #include <sys/efi_partition.h>
40 #include <sys/vtoc.h>
41 #include <sys/zfs_ioctl.h>
42 #include <dlfcn.h>
43
44 #include "zfs_namecheck.h"
45 #include "zfs_prop.h"
46 #include "libzfs_impl.h"
47 #include "zfs_comutil.h"
48 #include "zfeature_common.h"
49
50 static int read_efi_label(nvlist_t *config, diskaddr_t *sb);
51
52 #define DISK_ROOT "/dev/dsk"
53 #define RDISK_ROOT "/dev/rdsk"
54 #define BACKUP_SLICE "s2"
55
56 typedef struct prop_flags {
57 int create:1; /* Validate property on creation */
58 int import:1; /* Validate property on import */
59 } prop_flags_t;
60
61 /*
62 * ====================================================================
63 * zpool property functions
64 * ====================================================================
65 */
66
67 static int
68 zpool_get_all_props(zpool_handle_t *zhp)
69 {
70 zfs_cmd_t zc = { 0 };
71 libzfs_handle_t *hdl = zhp->zpool_hdl;
72
73 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
74
75 if (zcmd_alloc_dst_nvlist(hdl, &zc, 0) != 0)
76 return (-1);
77
78 while (ioctl(hdl->libzfs_fd, ZFS_IOC_POOL_GET_PROPS, &zc) != 0) {
79 if (errno == ENOMEM) {
80 if (zcmd_expand_dst_nvlist(hdl, &zc) != 0) {
81 zcmd_free_nvlists(&zc);
82 return (-1);
83 }
84 } else {
85 zcmd_free_nvlists(&zc);
86 return (-1);
87 }
88 }
89
90 if (zcmd_read_dst_nvlist(hdl, &zc, &zhp->zpool_props) != 0) {
91 zcmd_free_nvlists(&zc);
92 return (-1);
93 }
94
95 zcmd_free_nvlists(&zc);
96
97 return (0);
98 }
99
100 static int
101 zpool_props_refresh(zpool_handle_t *zhp)
102 {
103 nvlist_t *old_props;
104
105 old_props = zhp->zpool_props;
106
107 if (zpool_get_all_props(zhp) != 0)
108 return (-1);
109
110 nvlist_free(old_props);
111 return (0);
112 }
113
114 static char *
115 zpool_get_prop_string(zpool_handle_t *zhp, zpool_prop_t prop,
116 zprop_source_t *src)
117 {
118 nvlist_t *nv, *nvl;
119 uint64_t ival;
120 char *value;
121 zprop_source_t source;
122
123 nvl = zhp->zpool_props;
124 if (nvlist_lookup_nvlist(nvl, zpool_prop_to_name(prop), &nv) == 0) {
125 verify(nvlist_lookup_uint64(nv, ZPROP_SOURCE, &ival) == 0);
126 source = ival;
127 verify(nvlist_lookup_string(nv, ZPROP_VALUE, &value) == 0);
128 } else {
129 source = ZPROP_SRC_DEFAULT;
130 if ((value = (char *)zpool_prop_default_string(prop)) == NULL)
131 value = "-";
132 }
133
134 if (src)
135 *src = source;
136
137 return (value);
138 }
139
140 uint64_t
141 zpool_get_prop_int(zpool_handle_t *zhp, zpool_prop_t prop, zprop_source_t *src)
142 {
143 nvlist_t *nv, *nvl;
144 uint64_t value;
145 zprop_source_t source;
146
147 if (zhp->zpool_props == NULL && zpool_get_all_props(zhp)) {
148 /*
149 * zpool_get_all_props() has most likely failed because
150 * the pool is faulted, but if all we need is the top level
151 * vdev's guid then get it from the zhp config nvlist.
152 */
153 if ((prop == ZPOOL_PROP_GUID) &&
154 (nvlist_lookup_nvlist(zhp->zpool_config,
155 ZPOOL_CONFIG_VDEV_TREE, &nv) == 0) &&
156 (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID, &value)
157 == 0)) {
158 return (value);
159 }
160 return (zpool_prop_default_numeric(prop));
161 }
162
163 nvl = zhp->zpool_props;
164 if (nvlist_lookup_nvlist(nvl, zpool_prop_to_name(prop), &nv) == 0) {
165 verify(nvlist_lookup_uint64(nv, ZPROP_SOURCE, &value) == 0);
166 source = value;
167 verify(nvlist_lookup_uint64(nv, ZPROP_VALUE, &value) == 0);
168 } else {
169 source = ZPROP_SRC_DEFAULT;
170 value = zpool_prop_default_numeric(prop);
171 }
172
173 if (src)
174 *src = source;
175
176 return (value);
177 }
178
179 /*
180 * Map VDEV STATE to printed strings.
181 */
182 char *
183 zpool_state_to_name(vdev_state_t state, vdev_aux_t aux)
184 {
185 switch (state) {
186 case VDEV_STATE_CLOSED:
187 case VDEV_STATE_OFFLINE:
188 return (gettext("OFFLINE"));
189 case VDEV_STATE_REMOVED:
190 return (gettext("REMOVED"));
191 case VDEV_STATE_CANT_OPEN:
192 if (aux == VDEV_AUX_CORRUPT_DATA || aux == VDEV_AUX_BAD_LOG)
193 return (gettext("FAULTED"));
194 else if (aux == VDEV_AUX_SPLIT_POOL)
195 return (gettext("SPLIT"));
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 case ZPOOL_PROP_COMMENT:
242 if (zhp->zpool_props != NULL ||
243 zpool_get_all_props(zhp) == 0) {
244 (void) strlcpy(buf,
245 zpool_get_prop_string(zhp, prop, &src),
246 len);
247 if (srctype != NULL)
248 *srctype = src;
249 return (0);
250 }
251 /* FALLTHROUGH */
252 default:
253 (void) strlcpy(buf, "-", len);
254 break;
255 }
256
257 if (srctype != NULL)
258 *srctype = src;
259 return (0);
260 }
261
262 if (zhp->zpool_props == NULL && zpool_get_all_props(zhp) &&
263 prop != ZPOOL_PROP_NAME)
264 return (-1);
265
266 switch (zpool_prop_get_type(prop)) {
267 case PROP_TYPE_STRING:
268 (void) strlcpy(buf, zpool_get_prop_string(zhp, prop, &src),
269 len);
270 break;
271
272 case PROP_TYPE_NUMBER:
273 intval = zpool_get_prop_int(zhp, prop, &src);
274
275 switch (prop) {
276 case ZPOOL_PROP_SIZE:
277 case ZPOOL_PROP_ALLOCATED:
278 case ZPOOL_PROP_FREE:
279 case ZPOOL_PROP_FREEING:
280 case ZPOOL_PROP_EXPANDSZ:
281 (void) zfs_nicenum(intval, buf, len);
282 break;
283
284 case ZPOOL_PROP_CAPACITY:
285 (void) snprintf(buf, len, "%llu%%",
286 (u_longlong_t)intval);
287 break;
288
289 case ZPOOL_PROP_DEDUPRATIO:
290 (void) snprintf(buf, len, "%llu.%02llux",
291 (u_longlong_t)(intval / 100),
292 (u_longlong_t)(intval % 100));
293 break;
294
295 case ZPOOL_PROP_HEALTH:
296 verify(nvlist_lookup_nvlist(zpool_get_config(zhp, NULL),
297 ZPOOL_CONFIG_VDEV_TREE, &nvroot) == 0);
298 verify(nvlist_lookup_uint64_array(nvroot,
299 ZPOOL_CONFIG_VDEV_STATS, (uint64_t **)&vs, &vsc)
300 == 0);
301
302 (void) strlcpy(buf, zpool_state_to_name(intval,
303 vs->vs_aux), len);
304 break;
305 case ZPOOL_PROP_VERSION:
306 if (intval >= SPA_VERSION_FEATURES) {
307 (void) snprintf(buf, len, "-");
308 break;
309 }
310 /* FALLTHROUGH */
311 default:
312 (void) snprintf(buf, len, "%llu", intval);
313 }
314 break;
315
316 case PROP_TYPE_INDEX:
317 intval = zpool_get_prop_int(zhp, prop, &src);
318 if (zpool_prop_index_to_string(prop, intval, &strval)
319 != 0)
320 return (-1);
321 (void) strlcpy(buf, strval, len);
322 break;
323
324 default:
325 abort();
326 }
327
328 if (srctype)
329 *srctype = src;
330
331 return (0);
332 }
333
334 /*
335 * Check if the bootfs name has the same pool name as it is set to.
336 * Assuming bootfs is a valid dataset name.
337 */
338 static boolean_t
339 bootfs_name_valid(const char *pool, char *bootfs)
340 {
341 int len = strlen(pool);
342
343 if (!zfs_name_valid(bootfs, ZFS_TYPE_FILESYSTEM|ZFS_TYPE_SNAPSHOT))
344 return (B_FALSE);
345
346 if (strncmp(pool, bootfs, len) == 0 &&
347 (bootfs[len] == '/' || bootfs[len] == '\0'))
348 return (B_TRUE);
349
350 return (B_FALSE);
351 }
352
353 /*
354 * Inspect the configuration to determine if any of the devices contain
355 * an EFI label.
356 */
357 static boolean_t
358 pool_uses_efi(nvlist_t *config)
359 {
360 nvlist_t **child;
361 uint_t c, children;
362
363 if (nvlist_lookup_nvlist_array(config, ZPOOL_CONFIG_CHILDREN,
364 &child, &children) != 0)
365 return (read_efi_label(config, NULL) >= 0);
366
367 for (c = 0; c < children; c++) {
368 if (pool_uses_efi(child[c]))
369 return (B_TRUE);
370 }
371 return (B_FALSE);
372 }
373
374 boolean_t
375 zpool_is_bootable(zpool_handle_t *zhp)
376 {
377 char bootfs[ZPOOL_MAXNAMELEN];
378
379 return (zpool_get_prop(zhp, ZPOOL_PROP_BOOTFS, bootfs,
380 sizeof (bootfs), NULL) == 0 && strncmp(bootfs, "-",
381 sizeof (bootfs)) != 0);
382 }
383
384
385 /*
386 * Given an nvlist of zpool properties to be set, validate that they are
387 * correct, and parse any numeric properties (index, boolean, etc) if they are
388 * specified as strings.
389 */
390 static nvlist_t *
391 zpool_valid_proplist(libzfs_handle_t *hdl, const char *poolname,
392 nvlist_t *props, uint64_t version, prop_flags_t flags, char *errbuf)
393 {
394 nvpair_t *elem;
395 nvlist_t *retprops;
396 zpool_prop_t prop;
397 char *strval;
398 uint64_t intval;
399 char *slash, *check;
400 struct stat64 statbuf;
401 zpool_handle_t *zhp;
402 nvlist_t *nvroot;
403
404 if (nvlist_alloc(&retprops, NV_UNIQUE_NAME, 0) != 0) {
405 (void) no_memory(hdl);
406 return (NULL);
407 }
408
409 elem = NULL;
410 while ((elem = nvlist_next_nvpair(props, elem)) != NULL) {
411 const char *propname = nvpair_name(elem);
412
413 prop = zpool_name_to_prop(propname);
414 if (prop == ZPROP_INVAL && zpool_prop_feature(propname)) {
415 int err;
416 char *fname = strchr(propname, '@') + 1;
417
418 err = zfeature_lookup_name(fname, NULL);
419 if (err != 0) {
420 ASSERT3U(err, ==, ENOENT);
421 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
422 "invalid feature '%s'"), fname);
423 (void) zfs_error(hdl, EZFS_BADPROP, errbuf);
424 goto error;
425 }
426
427 if (nvpair_type(elem) != DATA_TYPE_STRING) {
428 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
429 "'%s' must be a string"), propname);
430 (void) zfs_error(hdl, EZFS_BADPROP, errbuf);
431 goto error;
432 }
433
434 (void) nvpair_value_string(elem, &strval);
435 if (strcmp(strval, ZFS_FEATURE_ENABLED) != 0) {
436 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
437 "property '%s' can only be set to "
438 "'enabled'"), propname);
439 (void) zfs_error(hdl, EZFS_BADPROP, errbuf);
440 goto error;
441 }
442
443 if (nvlist_add_uint64(retprops, propname, 0) != 0) {
444 (void) no_memory(hdl);
445 goto error;
446 }
447 continue;
448 }
449
450 /*
451 * Make sure this property is valid and applies to this type.
452 */
453 if (prop == ZPROP_INVAL) {
454 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
455 "invalid property '%s'"), propname);
456 (void) zfs_error(hdl, EZFS_BADPROP, errbuf);
457 goto error;
458 }
459
460 if (zpool_prop_readonly(prop)) {
461 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "'%s' "
462 "is readonly"), propname);
463 (void) zfs_error(hdl, EZFS_PROPREADONLY, errbuf);
464 goto error;
465 }
466
467 if (zprop_parse_value(hdl, elem, prop, ZFS_TYPE_POOL, retprops,
468 &strval, &intval, errbuf) != 0)
469 goto error;
470
471 /*
472 * Perform additional checking for specific properties.
473 */
474 switch (prop) {
475 case ZPOOL_PROP_VERSION:
476 if (intval < version ||
477 !SPA_VERSION_IS_SUPPORTED(intval)) {
478 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
479 "property '%s' number %d is invalid."),
480 propname, intval);
481 (void) zfs_error(hdl, EZFS_BADVERSION, errbuf);
482 goto error;
483 }
484 break;
485
486 case ZPOOL_PROP_BOOTFS:
487 if (flags.create || flags.import) {
488 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
489 "property '%s' cannot be set at creation "
490 "or import time"), propname);
491 (void) zfs_error(hdl, EZFS_BADPROP, errbuf);
492 goto error;
493 }
494
495 if (version < SPA_VERSION_BOOTFS) {
496 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
497 "pool must be upgraded to support "
498 "'%s' property"), propname);
499 (void) zfs_error(hdl, EZFS_BADVERSION, errbuf);
500 goto error;
501 }
502
503 /*
504 * bootfs property value has to be a dataset name and
505 * the dataset has to be in the same pool as it sets to.
506 */
507 if (strval[0] != '\0' && !bootfs_name_valid(poolname,
508 strval)) {
509 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "'%s' "
510 "is an invalid name"), strval);
511 (void) zfs_error(hdl, EZFS_INVALIDNAME, errbuf);
512 goto error;
513 }
514
515 if ((zhp = zpool_open_canfail(hdl, poolname)) == NULL) {
516 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
517 "could not open pool '%s'"), poolname);
518 (void) zfs_error(hdl, EZFS_OPENFAILED, errbuf);
519 goto error;
520 }
521 verify(nvlist_lookup_nvlist(zpool_get_config(zhp, NULL),
522 ZPOOL_CONFIG_VDEV_TREE, &nvroot) == 0);
523
524 /*
525 * bootfs property cannot be set on a disk which has
526 * been EFI labeled.
527 */
528 if (pool_uses_efi(nvroot)) {
529 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
530 "property '%s' not supported on "
531 "EFI labeled devices"), propname);
532 (void) zfs_error(hdl, EZFS_POOL_NOTSUP, errbuf);
533 zpool_close(zhp);
534 goto error;
535 }
536 zpool_close(zhp);
537 break;
538
539 case ZPOOL_PROP_ALTROOT:
540 if (!flags.create && !flags.import) {
541 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
542 "property '%s' can only be set during pool "
543 "creation or import"), propname);
544 (void) zfs_error(hdl, EZFS_BADPROP, errbuf);
545 goto error;
546 }
547
548 if (strval[0] != '/') {
549 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
550 "bad alternate root '%s'"), strval);
551 (void) zfs_error(hdl, EZFS_BADPATH, errbuf);
552 goto error;
553 }
554 break;
555
556 case ZPOOL_PROP_CACHEFILE:
557 if (strval[0] == '\0')
558 break;
559
560 if (strcmp(strval, "none") == 0)
561 break;
562
563 if (strval[0] != '/') {
564 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
565 "property '%s' must be empty, an "
566 "absolute path, or 'none'"), propname);
567 (void) zfs_error(hdl, EZFS_BADPATH, errbuf);
568 goto error;
569 }
570
571 slash = strrchr(strval, '/');
572
573 if (slash[1] == '\0' || strcmp(slash, "/.") == 0 ||
574 strcmp(slash, "/..") == 0) {
575 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
576 "'%s' is not a valid file"), strval);
577 (void) zfs_error(hdl, EZFS_BADPATH, errbuf);
578 goto error;
579 }
580
581 *slash = '\0';
582
583 if (strval[0] != '\0' &&
584 (stat64(strval, &statbuf) != 0 ||
585 !S_ISDIR(statbuf.st_mode))) {
586 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
587 "'%s' is not a valid directory"),
588 strval);
589 (void) zfs_error(hdl, EZFS_BADPATH, errbuf);
590 goto error;
591 }
592
593 *slash = '/';
594 break;
595
596 case ZPOOL_PROP_COMMENT:
597 for (check = strval; *check != '\0'; check++) {
598 if (!isprint(*check)) {
599 zfs_error_aux(hdl,
600 dgettext(TEXT_DOMAIN,
601 "comment may only have printable "
602 "characters"));
603 (void) zfs_error(hdl, EZFS_BADPROP,
604 errbuf);
605 goto error;
606 }
607 }
608 if (strlen(strval) > ZPROP_MAX_COMMENT) {
609 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
610 "comment must not exceed %d characters"),
611 ZPROP_MAX_COMMENT);
612 (void) zfs_error(hdl, EZFS_BADPROP, errbuf);
613 goto error;
614 }
615 break;
616 case ZPOOL_PROP_READONLY:
617 if (!flags.import) {
618 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
619 "property '%s' can only be set at "
620 "import time"), propname);
621 (void) zfs_error(hdl, EZFS_BADPROP, errbuf);
622 goto error;
623 }
624 break;
625 }
626 }
627
628 return (retprops);
629 error:
630 nvlist_free(retprops);
631 return (NULL);
632 }
633
634 /*
635 * Set zpool property : propname=propval.
636 */
637 int
638 zpool_set_prop(zpool_handle_t *zhp, const char *propname, const char *propval)
639 {
640 zfs_cmd_t zc = { 0 };
641 int ret = -1;
642 char errbuf[1024];
643 nvlist_t *nvl = NULL;
644 nvlist_t *realprops;
645 uint64_t version;
646 prop_flags_t flags = { 0 };
647
648 (void) snprintf(errbuf, sizeof (errbuf),
649 dgettext(TEXT_DOMAIN, "cannot set property for '%s'"),
650 zhp->zpool_name);
651
652 if (nvlist_alloc(&nvl, NV_UNIQUE_NAME, 0) != 0)
653 return (no_memory(zhp->zpool_hdl));
654
655 if (nvlist_add_string(nvl, propname, propval) != 0) {
656 nvlist_free(nvl);
657 return (no_memory(zhp->zpool_hdl));
658 }
659
660 version = zpool_get_prop_int(zhp, ZPOOL_PROP_VERSION, NULL);
661 if ((realprops = zpool_valid_proplist(zhp->zpool_hdl,
662 zhp->zpool_name, nvl, version, flags, errbuf)) == NULL) {
663 nvlist_free(nvl);
664 return (-1);
665 }
666
667 nvlist_free(nvl);
668 nvl = realprops;
669
670 /*
671 * Execute the corresponding ioctl() to set this property.
672 */
673 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
674
675 if (zcmd_write_src_nvlist(zhp->zpool_hdl, &zc, nvl) != 0) {
676 nvlist_free(nvl);
677 return (-1);
678 }
679
680 ret = zfs_ioctl(zhp->zpool_hdl, ZFS_IOC_POOL_SET_PROPS, &zc);
681
682 zcmd_free_nvlists(&zc);
683 nvlist_free(nvl);
684
685 if (ret)
686 (void) zpool_standard_error(zhp->zpool_hdl, errno, errbuf);
687 else
688 (void) zpool_props_refresh(zhp);
689
690 return (ret);
691 }
692
693 int
694 zpool_expand_proplist(zpool_handle_t *zhp, zprop_list_t **plp)
695 {
696 libzfs_handle_t *hdl = zhp->zpool_hdl;
697 zprop_list_t *entry;
698 char buf[ZFS_MAXPROPLEN];
699 nvlist_t *features = NULL;
700 zprop_list_t **last;
701 boolean_t firstexpand = (NULL == *plp);
702
703 if (zprop_expand_list(hdl, plp, ZFS_TYPE_POOL) != 0)
704 return (-1);
705
706 last = plp;
707 while (*last != NULL)
708 last = &(*last)->pl_next;
709
710 if ((*plp)->pl_all)
711 features = zpool_get_features(zhp);
712
713 if ((*plp)->pl_all && firstexpand) {
714 for (int i = 0; i < SPA_FEATURES; i++) {
715 zprop_list_t *entry = zfs_alloc(hdl,
716 sizeof (zprop_list_t));
717 entry->pl_prop = ZPROP_INVAL;
718 entry->pl_user_prop = zfs_asprintf(hdl, "feature@%s",
719 spa_feature_table[i].fi_uname);
720 entry->pl_width = strlen(entry->pl_user_prop);
721 entry->pl_all = B_TRUE;
722
723 *last = entry;
724 last = &entry->pl_next;
725 }
726 }
727
728 /* add any unsupported features */
729 for (nvpair_t *nvp = nvlist_next_nvpair(features, NULL);
730 nvp != NULL; nvp = nvlist_next_nvpair(features, nvp)) {
731 char *propname;
732 boolean_t found;
733 zprop_list_t *entry;
734
735 if (zfeature_is_supported(nvpair_name(nvp)))
736 continue;
737
738 propname = zfs_asprintf(hdl, "unsupported@%s",
739 nvpair_name(nvp));
740
741 /*
742 * Before adding the property to the list make sure that no
743 * other pool already added the same property.
744 */
745 found = B_FALSE;
746 entry = *plp;
747 while (entry != NULL) {
748 if (entry->pl_user_prop != NULL &&
749 strcmp(propname, entry->pl_user_prop) == 0) {
750 found = B_TRUE;
751 break;
752 }
753 entry = entry->pl_next;
754 }
755 if (found) {
756 free(propname);
757 continue;
758 }
759
760 entry = zfs_alloc(hdl, sizeof (zprop_list_t));
761 entry->pl_prop = ZPROP_INVAL;
762 entry->pl_user_prop = propname;
763 entry->pl_width = strlen(entry->pl_user_prop);
764 entry->pl_all = B_TRUE;
765
766 *last = entry;
767 last = &entry->pl_next;
768 }
769
770 for (entry = *plp; entry != NULL; entry = entry->pl_next) {
771
772 if (entry->pl_fixed)
773 continue;
774
775 if (entry->pl_prop != ZPROP_INVAL &&
776 zpool_get_prop(zhp, entry->pl_prop, buf, sizeof (buf),
777 NULL) == 0) {
778 if (strlen(buf) > entry->pl_width)
779 entry->pl_width = strlen(buf);
780 }
781 }
782
783 return (0);
784 }
785
786 /*
787 * Get the state for the given feature on the given ZFS pool.
788 */
789 int
790 zpool_prop_get_feature(zpool_handle_t *zhp, const char *propname, char *buf,
791 size_t len)
792 {
793 uint64_t refcount;
794 boolean_t found = B_FALSE;
795 nvlist_t *features = zpool_get_features(zhp);
796 boolean_t supported;
797 const char *feature = strchr(propname, '@') + 1;
798
799 supported = zpool_prop_feature(propname);
800 ASSERT(supported || zfs_prop_unsupported(propname));
801
802 /*
803 * Convert from feature name to feature guid. This conversion is
804 * unecessary for unsupported@... properties because they already
805 * use guids.
806 */
807 if (supported) {
808 int ret;
809 spa_feature_t fid;
810
811 ret = zfeature_lookup_name(feature, &fid);
812 if (ret != 0) {
813 (void) strlcpy(buf, "-", len);
814 return (ENOTSUP);
815 }
816 feature = spa_feature_table[fid].fi_guid;
817 }
818
819 if (nvlist_lookup_uint64(features, feature, &refcount) == 0)
820 found = B_TRUE;
821
822 if (supported) {
823 if (!found) {
824 (void) strlcpy(buf, ZFS_FEATURE_DISABLED, len);
825 } else {
826 if (refcount == 0)
827 (void) strlcpy(buf, ZFS_FEATURE_ENABLED, len);
828 else
829 (void) strlcpy(buf, ZFS_FEATURE_ACTIVE, len);
830 }
831 } else {
832 if (found) {
833 if (refcount == 0) {
834 (void) strcpy(buf, ZFS_UNSUPPORTED_INACTIVE);
835 } else {
836 (void) strcpy(buf, ZFS_UNSUPPORTED_READONLY);
837 }
838 } else {
839 (void) strlcpy(buf, "-", len);
840 return (ENOTSUP);
841 }
842 }
843
844 return (0);
845 }
846
847 /*
848 * Don't start the slice at the default block of 34; many storage
849 * devices will use a stripe width of 128k, so start there instead.
850 */
851 #define NEW_START_BLOCK 256
852
853 /*
854 * Validate the given pool name, optionally putting an extended error message in
855 * 'buf'.
856 */
857 boolean_t
858 zpool_name_valid(libzfs_handle_t *hdl, boolean_t isopen, const char *pool)
859 {
860 namecheck_err_t why;
861 char what;
862 int ret;
863
864 ret = pool_namecheck(pool, &why, &what);
865
866 /*
867 * The rules for reserved pool names were extended at a later point.
868 * But we need to support users with existing pools that may now be
869 * invalid. So we only check for this expanded set of names during a
870 * create (or import), and only in userland.
871 */
872 if (ret == 0 && !isopen &&
873 (strncmp(pool, "mirror", 6) == 0 ||
874 strncmp(pool, "raidz", 5) == 0 ||
875 strncmp(pool, "spare", 5) == 0 ||
876 strcmp(pool, "log") == 0)) {
877 if (hdl != NULL)
878 zfs_error_aux(hdl,
879 dgettext(TEXT_DOMAIN, "name is reserved"));
880 return (B_FALSE);
881 }
882
883
884 if (ret != 0) {
885 if (hdl != NULL) {
886 switch (why) {
887 case NAME_ERR_TOOLONG:
888 zfs_error_aux(hdl,
889 dgettext(TEXT_DOMAIN, "name is too long"));
890 break;
891
892 case NAME_ERR_INVALCHAR:
893 zfs_error_aux(hdl,
894 dgettext(TEXT_DOMAIN, "invalid character "
895 "'%c' in pool name"), what);
896 break;
897
898 case NAME_ERR_NOLETTER:
899 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
900 "name must begin with a letter"));
901 break;
902
903 case NAME_ERR_RESERVED:
904 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
905 "name is reserved"));
906 break;
907
908 case NAME_ERR_DISKLIKE:
909 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
910 "pool name is reserved"));
911 break;
912
913 case NAME_ERR_LEADING_SLASH:
914 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
915 "leading slash in name"));
916 break;
917
918 case NAME_ERR_EMPTY_COMPONENT:
919 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
920 "empty component in name"));
921 break;
922
923 case NAME_ERR_TRAILING_SLASH:
924 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
925 "trailing slash in name"));
926 break;
927
928 case NAME_ERR_MULTIPLE_AT:
929 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
930 "multiple '@' delimiters in name"));
931 break;
932
933 }
934 }
935 return (B_FALSE);
936 }
937
938 return (B_TRUE);
939 }
940
941 /*
942 * Open a handle to the given pool, even if the pool is currently in the FAULTED
943 * state.
944 */
945 zpool_handle_t *
946 zpool_open_canfail(libzfs_handle_t *hdl, const char *pool)
947 {
948 zpool_handle_t *zhp;
949 boolean_t missing;
950
951 /*
952 * Make sure the pool name is valid.
953 */
954 if (!zpool_name_valid(hdl, B_TRUE, pool)) {
955 (void) zfs_error_fmt(hdl, EZFS_INVALIDNAME,
956 dgettext(TEXT_DOMAIN, "cannot open '%s'"),
957 pool);
958 return (NULL);
959 }
960
961 if ((zhp = zfs_alloc(hdl, sizeof (zpool_handle_t))) == NULL)
962 return (NULL);
963
964 zhp->zpool_hdl = hdl;
965 (void) strlcpy(zhp->zpool_name, pool, sizeof (zhp->zpool_name));
966
967 if (zpool_refresh_stats(zhp, &missing) != 0) {
968 zpool_close(zhp);
969 return (NULL);
970 }
971
972 if (missing) {
973 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "no such pool"));
974 (void) zfs_error_fmt(hdl, EZFS_NOENT,
975 dgettext(TEXT_DOMAIN, "cannot open '%s'"), pool);
976 zpool_close(zhp);
977 return (NULL);
978 }
979
980 return (zhp);
981 }
982
983 /*
984 * Like the above, but silent on error. Used when iterating over pools (because
985 * the configuration cache may be out of date).
986 */
987 int
988 zpool_open_silent(libzfs_handle_t *hdl, const char *pool, zpool_handle_t **ret)
989 {
990 zpool_handle_t *zhp;
991 boolean_t missing;
992
993 if ((zhp = zfs_alloc(hdl, sizeof (zpool_handle_t))) == NULL)
994 return (-1);
995
996 zhp->zpool_hdl = hdl;
997 (void) strlcpy(zhp->zpool_name, pool, sizeof (zhp->zpool_name));
998
999 if (zpool_refresh_stats(zhp, &missing) != 0) {
1000 zpool_close(zhp);
1001 return (-1);
1002 }
1003
1004 if (missing) {
1005 zpool_close(zhp);
1006 *ret = NULL;
1007 return (0);
1008 }
1009
1010 *ret = zhp;
1011 return (0);
1012 }
1013
1014 /*
1015 * Similar to zpool_open_canfail(), but refuses to open pools in the faulted
1016 * state.
1017 */
1018 zpool_handle_t *
1019 zpool_open(libzfs_handle_t *hdl, const char *pool)
1020 {
1021 zpool_handle_t *zhp;
1022
1023 if ((zhp = zpool_open_canfail(hdl, pool)) == NULL)
1024 return (NULL);
1025
1026 if (zhp->zpool_state == POOL_STATE_UNAVAIL) {
1027 (void) zfs_error_fmt(hdl, EZFS_POOLUNAVAIL,
1028 dgettext(TEXT_DOMAIN, "cannot open '%s'"), zhp->zpool_name);
1029 zpool_close(zhp);
1030 return (NULL);
1031 }
1032
1033 return (zhp);
1034 }
1035
1036 /*
1037 * Close the handle. Simply frees the memory associated with the handle.
1038 */
1039 void
1040 zpool_close(zpool_handle_t *zhp)
1041 {
1042 if (zhp->zpool_config)
1043 nvlist_free(zhp->zpool_config);
1044 if (zhp->zpool_old_config)
1045 nvlist_free(zhp->zpool_old_config);
1046 if (zhp->zpool_props)
1047 nvlist_free(zhp->zpool_props);
1048 free(zhp);
1049 }
1050
1051 /*
1052 * Return the name of the pool.
1053 */
1054 const char *
1055 zpool_get_name(zpool_handle_t *zhp)
1056 {
1057 return (zhp->zpool_name);
1058 }
1059
1060
1061 /*
1062 * Return the state of the pool (ACTIVE or UNAVAILABLE)
1063 */
1064 int
1065 zpool_get_state(zpool_handle_t *zhp)
1066 {
1067 return (zhp->zpool_state);
1068 }
1069
1070 /*
1071 * Create the named pool, using the provided vdev list. It is assumed
1072 * that the consumer has already validated the contents of the nvlist, so we
1073 * don't have to worry about error semantics.
1074 */
1075 int
1076 zpool_create(libzfs_handle_t *hdl, const char *pool, nvlist_t *nvroot,
1077 nvlist_t *props, nvlist_t *fsprops)
1078 {
1079 zfs_cmd_t zc = { 0 };
1080 nvlist_t *zc_fsprops = NULL;
1081 nvlist_t *zc_props = NULL;
1082 char msg[1024];
1083 int ret = -1;
1084
1085 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN,
1086 "cannot create '%s'"), pool);
1087
1088 if (!zpool_name_valid(hdl, B_FALSE, pool))
1089 return (zfs_error(hdl, EZFS_INVALIDNAME, msg));
1090
1091 if (zcmd_write_conf_nvlist(hdl, &zc, nvroot) != 0)
1092 return (-1);
1093
1094 if (props) {
1095 prop_flags_t flags = { .create = B_TRUE, .import = B_FALSE };
1096
1097 if ((zc_props = zpool_valid_proplist(hdl, pool, props,
1098 SPA_VERSION_1, flags, msg)) == NULL) {
1099 goto create_failed;
1100 }
1101 }
1102
1103 if (fsprops) {
1104 uint64_t zoned;
1105 char *zonestr;
1106
1107 zoned = ((nvlist_lookup_string(fsprops,
1108 zfs_prop_to_name(ZFS_PROP_ZONED), &zonestr) == 0) &&
1109 strcmp(zonestr, "on") == 0);
1110
1111 if ((zc_fsprops = zfs_valid_proplist(hdl,
1112 ZFS_TYPE_FILESYSTEM, fsprops, zoned, NULL, msg)) == NULL) {
1113 goto create_failed;
1114 }
1115 if (!zc_props &&
1116 (nvlist_alloc(&zc_props, NV_UNIQUE_NAME, 0) != 0)) {
1117 goto create_failed;
1118 }
1119 if (nvlist_add_nvlist(zc_props,
1120 ZPOOL_ROOTFS_PROPS, zc_fsprops) != 0) {
1121 goto create_failed;
1122 }
1123 }
1124
1125 if (zc_props && zcmd_write_src_nvlist(hdl, &zc, zc_props) != 0)
1126 goto create_failed;
1127
1128 (void) strlcpy(zc.zc_name, pool, sizeof (zc.zc_name));
1129
1130 if ((ret = zfs_ioctl(hdl, ZFS_IOC_POOL_CREATE, &zc)) != 0) {
1131
1132 zcmd_free_nvlists(&zc);
1133 nvlist_free(zc_props);
1134 nvlist_free(zc_fsprops);
1135
1136 switch (errno) {
1137 case EBUSY:
1138 /*
1139 * This can happen if the user has specified the same
1140 * device multiple times. We can't reliably detect this
1141 * until we try to add it and see we already have a
1142 * label.
1143 */
1144 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1145 "one or more vdevs refer to the same device"));
1146 return (zfs_error(hdl, EZFS_BADDEV, msg));
1147
1148 case EOVERFLOW:
1149 /*
1150 * This occurs when one of the devices is below
1151 * SPA_MINDEVSIZE. Unfortunately, we can't detect which
1152 * device was the problem device since there's no
1153 * reliable way to determine device size from userland.
1154 */
1155 {
1156 char buf[64];
1157
1158 zfs_nicenum(SPA_MINDEVSIZE, buf, sizeof (buf));
1159
1160 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1161 "one or more devices is less than the "
1162 "minimum size (%s)"), buf);
1163 }
1164 return (zfs_error(hdl, EZFS_BADDEV, msg));
1165
1166 case ENOSPC:
1167 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1168 "one or more devices is out of space"));
1169 return (zfs_error(hdl, EZFS_BADDEV, msg));
1170
1171 case ENOTBLK:
1172 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1173 "cache device must be a disk or disk slice"));
1174 return (zfs_error(hdl, EZFS_BADDEV, msg));
1175
1176 default:
1177 return (zpool_standard_error(hdl, errno, msg));
1178 }
1179 }
1180
1181 create_failed:
1182 zcmd_free_nvlists(&zc);
1183 nvlist_free(zc_props);
1184 nvlist_free(zc_fsprops);
1185 return (ret);
1186 }
1187
1188 /*
1189 * Destroy the given pool. It is up to the caller to ensure that there are no
1190 * datasets left in the pool.
1191 */
1192 int
1193 zpool_destroy(zpool_handle_t *zhp, const char *log_str)
1194 {
1195 zfs_cmd_t zc = { 0 };
1196 zfs_handle_t *zfp = NULL;
1197 libzfs_handle_t *hdl = zhp->zpool_hdl;
1198 char msg[1024];
1199
1200 if (zhp->zpool_state == POOL_STATE_ACTIVE &&
1201 (zfp = zfs_open(hdl, zhp->zpool_name, ZFS_TYPE_FILESYSTEM)) == NULL)
1202 return (-1);
1203
1204 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
1205 zc.zc_history = (uint64_t)(uintptr_t)log_str;
1206
1207 if (zfs_ioctl(hdl, ZFS_IOC_POOL_DESTROY, &zc) != 0) {
1208 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN,
1209 "cannot destroy '%s'"), zhp->zpool_name);
1210
1211 if (errno == EROFS) {
1212 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1213 "one or more devices is read only"));
1214 (void) zfs_error(hdl, EZFS_BADDEV, msg);
1215 } else {
1216 (void) zpool_standard_error(hdl, errno, msg);
1217 }
1218
1219 if (zfp)
1220 zfs_close(zfp);
1221 return (-1);
1222 }
1223
1224 if (zfp) {
1225 remove_mountpoint(zfp);
1226 zfs_close(zfp);
1227 }
1228
1229 return (0);
1230 }
1231
1232 /*
1233 * Add the given vdevs to the pool. The caller must have already performed the
1234 * necessary verification to ensure that the vdev specification is well-formed.
1235 */
1236 int
1237 zpool_add(zpool_handle_t *zhp, nvlist_t *nvroot)
1238 {
1239 zfs_cmd_t zc = { 0 };
1240 int ret;
1241 libzfs_handle_t *hdl = zhp->zpool_hdl;
1242 char msg[1024];
1243 nvlist_t **spares, **l2cache;
1244 uint_t nspares, nl2cache;
1245
1246 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN,
1247 "cannot add to '%s'"), zhp->zpool_name);
1248
1249 if (zpool_get_prop_int(zhp, ZPOOL_PROP_VERSION, NULL) <
1250 SPA_VERSION_SPARES &&
1251 nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_SPARES,
1252 &spares, &nspares) == 0) {
1253 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "pool must be "
1254 "upgraded to add hot spares"));
1255 return (zfs_error(hdl, EZFS_BADVERSION, msg));
1256 }
1257
1258 if (zpool_is_bootable(zhp) && nvlist_lookup_nvlist_array(nvroot,
1259 ZPOOL_CONFIG_SPARES, &spares, &nspares) == 0) {
1260 uint64_t s;
1261
1262 for (s = 0; s < nspares; s++) {
1263 char *path;
1264
1265 if (nvlist_lookup_string(spares[s], ZPOOL_CONFIG_PATH,
1266 &path) == 0 && pool_uses_efi(spares[s])) {
1267 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1268 "device '%s' contains an EFI label and "
1269 "cannot be used on root pools."),
1270 zpool_vdev_name(hdl, NULL, spares[s],
1271 B_FALSE));
1272 return (zfs_error(hdl, EZFS_POOL_NOTSUP, msg));
1273 }
1274 }
1275 }
1276
1277 if (zpool_get_prop_int(zhp, ZPOOL_PROP_VERSION, NULL) <
1278 SPA_VERSION_L2CACHE &&
1279 nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_L2CACHE,
1280 &l2cache, &nl2cache) == 0) {
1281 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "pool must be "
1282 "upgraded to add cache devices"));
1283 return (zfs_error(hdl, EZFS_BADVERSION, msg));
1284 }
1285
1286 if (zcmd_write_conf_nvlist(hdl, &zc, nvroot) != 0)
1287 return (-1);
1288 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
1289
1290 if (zfs_ioctl(hdl, ZFS_IOC_VDEV_ADD, &zc) != 0) {
1291 switch (errno) {
1292 case EBUSY:
1293 /*
1294 * This can happen if the user has specified the same
1295 * device multiple times. We can't reliably detect this
1296 * until we try to add it and see we already have a
1297 * label.
1298 */
1299 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1300 "one or more vdevs refer to the same device"));
1301 (void) zfs_error(hdl, EZFS_BADDEV, msg);
1302 break;
1303
1304 case EOVERFLOW:
1305 /*
1306 * This occurrs when one of the devices is below
1307 * SPA_MINDEVSIZE. Unfortunately, we can't detect which
1308 * device was the problem device since there's no
1309 * reliable way to determine device size from userland.
1310 */
1311 {
1312 char buf[64];
1313
1314 zfs_nicenum(SPA_MINDEVSIZE, buf, sizeof (buf));
1315
1316 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1317 "device is less than the minimum "
1318 "size (%s)"), buf);
1319 }
1320 (void) zfs_error(hdl, EZFS_BADDEV, msg);
1321 break;
1322
1323 case ENOTSUP:
1324 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1325 "pool must be upgraded to add these vdevs"));
1326 (void) zfs_error(hdl, EZFS_BADVERSION, msg);
1327 break;
1328
1329 case EDOM:
1330 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1331 "root pool can not have multiple vdevs"
1332 " or separate logs"));
1333 (void) zfs_error(hdl, EZFS_POOL_NOTSUP, msg);
1334 break;
1335
1336 case ENOTBLK:
1337 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1338 "cache device must be a disk or disk slice"));
1339 (void) zfs_error(hdl, EZFS_BADDEV, msg);
1340 break;
1341
1342 default:
1343 (void) zpool_standard_error(hdl, errno, msg);
1344 }
1345
1346 ret = -1;
1347 } else {
1348 ret = 0;
1349 }
1350
1351 zcmd_free_nvlists(&zc);
1352
1353 return (ret);
1354 }
1355
1356 /*
1357 * Exports the pool from the system. The caller must ensure that there are no
1358 * mounted datasets in the pool.
1359 */
1360 static int
1361 zpool_export_common(zpool_handle_t *zhp, boolean_t force, boolean_t hardforce,
1362 const char *log_str)
1363 {
1364 zfs_cmd_t zc = { 0 };
1365 char msg[1024];
1366
1367 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN,
1368 "cannot export '%s'"), zhp->zpool_name);
1369
1370 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
1371 zc.zc_cookie = force;
1372 zc.zc_guid = hardforce;
1373 zc.zc_history = (uint64_t)(uintptr_t)log_str;
1374
1375 if (zfs_ioctl(zhp->zpool_hdl, ZFS_IOC_POOL_EXPORT, &zc) != 0) {
1376 switch (errno) {
1377 case EXDEV:
1378 zfs_error_aux(zhp->zpool_hdl, dgettext(TEXT_DOMAIN,
1379 "use '-f' to override the following errors:\n"
1380 "'%s' has an active shared spare which could be"
1381 " used by other pools once '%s' is exported."),
1382 zhp->zpool_name, zhp->zpool_name);
1383 return (zfs_error(zhp->zpool_hdl, EZFS_ACTIVE_SPARE,
1384 msg));
1385 default:
1386 return (zpool_standard_error_fmt(zhp->zpool_hdl, errno,
1387 msg));
1388 }
1389 }
1390
1391 return (0);
1392 }
1393
1394 int
1395 zpool_export(zpool_handle_t *zhp, boolean_t force, const char *log_str)
1396 {
1397 return (zpool_export_common(zhp, force, B_FALSE, log_str));
1398 }
1399
1400 int
1401 zpool_export_force(zpool_handle_t *zhp, const char *log_str)
1402 {
1403 return (zpool_export_common(zhp, B_TRUE, B_TRUE, log_str));
1404 }
1405
1406 static void
1407 zpool_rewind_exclaim(libzfs_handle_t *hdl, const char *name, boolean_t dryrun,
1408 nvlist_t *config)
1409 {
1410 nvlist_t *nv = NULL;
1411 uint64_t rewindto;
1412 int64_t loss = -1;
1413 struct tm t;
1414 char timestr[128];
1415
1416 if (!hdl->libzfs_printerr || config == NULL)
1417 return;
1418
1419 if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_LOAD_INFO, &nv) != 0 ||
1420 nvlist_lookup_nvlist(nv, ZPOOL_CONFIG_REWIND_INFO, &nv) != 0) {
1421 return;
1422 }
1423
1424 if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_LOAD_TIME, &rewindto) != 0)
1425 return;
1426 (void) nvlist_lookup_int64(nv, ZPOOL_CONFIG_REWIND_TIME, &loss);
1427
1428 if (localtime_r((time_t *)&rewindto, &t) != NULL &&
1429 strftime(timestr, 128, 0, &t) != 0) {
1430 if (dryrun) {
1431 (void) printf(dgettext(TEXT_DOMAIN,
1432 "Would be able to return %s "
1433 "to its state as of %s.\n"),
1434 name, timestr);
1435 } else {
1436 (void) printf(dgettext(TEXT_DOMAIN,
1437 "Pool %s returned to its state as of %s.\n"),
1438 name, timestr);
1439 }
1440 if (loss > 120) {
1441 (void) printf(dgettext(TEXT_DOMAIN,
1442 "%s approximately %lld "),
1443 dryrun ? "Would discard" : "Discarded",
1444 (loss + 30) / 60);
1445 (void) printf(dgettext(TEXT_DOMAIN,
1446 "minutes of transactions.\n"));
1447 } else if (loss > 0) {
1448 (void) printf(dgettext(TEXT_DOMAIN,
1449 "%s approximately %lld "),
1450 dryrun ? "Would discard" : "Discarded", loss);
1451 (void) printf(dgettext(TEXT_DOMAIN,
1452 "seconds of transactions.\n"));
1453 }
1454 }
1455 }
1456
1457 void
1458 zpool_explain_recover(libzfs_handle_t *hdl, const char *name, int reason,
1459 nvlist_t *config)
1460 {
1461 nvlist_t *nv = NULL;
1462 int64_t loss = -1;
1463 uint64_t edata = UINT64_MAX;
1464 uint64_t rewindto;
1465 struct tm t;
1466 char timestr[128];
1467
1468 if (!hdl->libzfs_printerr)
1469 return;
1470
1471 if (reason >= 0)
1472 (void) printf(dgettext(TEXT_DOMAIN, "action: "));
1473 else
1474 (void) printf(dgettext(TEXT_DOMAIN, "\t"));
1475
1476 /* All attempted rewinds failed if ZPOOL_CONFIG_LOAD_TIME missing */
1477 if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_LOAD_INFO, &nv) != 0 ||
1478 nvlist_lookup_nvlist(nv, ZPOOL_CONFIG_REWIND_INFO, &nv) != 0 ||
1479 nvlist_lookup_uint64(nv, ZPOOL_CONFIG_LOAD_TIME, &rewindto) != 0)
1480 goto no_info;
1481
1482 (void) nvlist_lookup_int64(nv, ZPOOL_CONFIG_REWIND_TIME, &loss);
1483 (void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_LOAD_DATA_ERRORS,
1484 &edata);
1485
1486 (void) printf(dgettext(TEXT_DOMAIN,
1487 "Recovery is possible, but will result in some data loss.\n"));
1488
1489 if (localtime_r((time_t *)&rewindto, &t) != NULL &&
1490 strftime(timestr, 128, 0, &t) != 0) {
1491 (void) printf(dgettext(TEXT_DOMAIN,
1492 "\tReturning the pool to its state as of %s\n"
1493 "\tshould correct the problem. "),
1494 timestr);
1495 } else {
1496 (void) printf(dgettext(TEXT_DOMAIN,
1497 "\tReverting the pool to an earlier state "
1498 "should correct the problem.\n\t"));
1499 }
1500
1501 if (loss > 120) {
1502 (void) printf(dgettext(TEXT_DOMAIN,
1503 "Approximately %lld minutes of data\n"
1504 "\tmust be discarded, irreversibly. "), (loss + 30) / 60);
1505 } else if (loss > 0) {
1506 (void) printf(dgettext(TEXT_DOMAIN,
1507 "Approximately %lld seconds of data\n"
1508 "\tmust be discarded, irreversibly. "), loss);
1509 }
1510 if (edata != 0 && edata != UINT64_MAX) {
1511 if (edata == 1) {
1512 (void) printf(dgettext(TEXT_DOMAIN,
1513 "After rewind, at least\n"
1514 "\tone persistent user-data error will remain. "));
1515 } else {
1516 (void) printf(dgettext(TEXT_DOMAIN,
1517 "After rewind, several\n"
1518 "\tpersistent user-data errors will remain. "));
1519 }
1520 }
1521 (void) printf(dgettext(TEXT_DOMAIN,
1522 "Recovery can be attempted\n\tby executing 'zpool %s -F %s'. "),
1523 reason >= 0 ? "clear" : "import", name);
1524
1525 (void) printf(dgettext(TEXT_DOMAIN,
1526 "A scrub of the pool\n"
1527 "\tis strongly recommended after recovery.\n"));
1528 return;
1529
1530 no_info:
1531 (void) printf(dgettext(TEXT_DOMAIN,
1532 "Destroy and re-create the pool from\n\ta backup source.\n"));
1533 }
1534
1535 /*
1536 * zpool_import() is a contracted interface. Should be kept the same
1537 * if possible.
1538 *
1539 * Applications should use zpool_import_props() to import a pool with
1540 * new properties value to be set.
1541 */
1542 int
1543 zpool_import(libzfs_handle_t *hdl, nvlist_t *config, const char *newname,
1544 char *altroot)
1545 {
1546 nvlist_t *props = NULL;
1547 int ret;
1548
1549 if (altroot != NULL) {
1550 if (nvlist_alloc(&props, NV_UNIQUE_NAME, 0) != 0) {
1551 return (zfs_error_fmt(hdl, EZFS_NOMEM,
1552 dgettext(TEXT_DOMAIN, "cannot import '%s'"),
1553 newname));
1554 }
1555
1556 if (nvlist_add_string(props,
1557 zpool_prop_to_name(ZPOOL_PROP_ALTROOT), altroot) != 0 ||
1558 nvlist_add_string(props,
1559 zpool_prop_to_name(ZPOOL_PROP_CACHEFILE), "none") != 0) {
1560 nvlist_free(props);
1561 return (zfs_error_fmt(hdl, EZFS_NOMEM,
1562 dgettext(TEXT_DOMAIN, "cannot import '%s'"),
1563 newname));
1564 }
1565 }
1566
1567 ret = zpool_import_props(hdl, config, newname, props,
1568 ZFS_IMPORT_NORMAL);
1569 if (props)
1570 nvlist_free(props);
1571 return (ret);
1572 }
1573
1574 static void
1575 print_vdev_tree(libzfs_handle_t *hdl, const char *name, nvlist_t *nv,
1576 int indent)
1577 {
1578 nvlist_t **child;
1579 uint_t c, children;
1580 char *vname;
1581 uint64_t is_log = 0;
1582
1583 (void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_IS_LOG,
1584 &is_log);
1585
1586 if (name != NULL)
1587 (void) printf("\t%*s%s%s\n", indent, "", name,
1588 is_log ? " [log]" : "");
1589
1590 if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN,
1591 &child, &children) != 0)
1592 return;
1593
1594 for (c = 0; c < children; c++) {
1595 vname = zpool_vdev_name(hdl, NULL, child[c], B_TRUE);
1596 print_vdev_tree(hdl, vname, child[c], indent + 2);
1597 free(vname);
1598 }
1599 }
1600
1601 void
1602 zpool_print_unsup_feat(nvlist_t *config)
1603 {
1604 nvlist_t *nvinfo, *unsup_feat;
1605
1606 verify(nvlist_lookup_nvlist(config, ZPOOL_CONFIG_LOAD_INFO, &nvinfo) ==
1607 0);
1608 verify(nvlist_lookup_nvlist(nvinfo, ZPOOL_CONFIG_UNSUP_FEAT,
1609 &unsup_feat) == 0);
1610
1611 for (nvpair_t *nvp = nvlist_next_nvpair(unsup_feat, NULL); nvp != NULL;
1612 nvp = nvlist_next_nvpair(unsup_feat, nvp)) {
1613 char *desc;
1614
1615 verify(nvpair_type(nvp) == DATA_TYPE_STRING);
1616 verify(nvpair_value_string(nvp, &desc) == 0);
1617
1618 if (strlen(desc) > 0)
1619 (void) printf("\t%s (%s)\n", nvpair_name(nvp), desc);
1620 else
1621 (void) printf("\t%s\n", nvpair_name(nvp));
1622 }
1623 }
1624
1625 /*
1626 * Import the given pool using the known configuration and a list of
1627 * properties to be set. The configuration should have come from
1628 * zpool_find_import(). The 'newname' parameters control whether the pool
1629 * is imported with a different name.
1630 */
1631 int
1632 zpool_import_props(libzfs_handle_t *hdl, nvlist_t *config, const char *newname,
1633 nvlist_t *props, int flags)
1634 {
1635 zfs_cmd_t zc = { 0 };
1636 zpool_rewind_policy_t policy;
1637 nvlist_t *nv = NULL;
1638 nvlist_t *nvinfo = NULL;
1639 nvlist_t *missing = NULL;
1640 char *thename;
1641 char *origname;
1642 int ret;
1643 int error = 0;
1644 char errbuf[1024];
1645
1646 verify(nvlist_lookup_string(config, ZPOOL_CONFIG_POOL_NAME,
1647 &origname) == 0);
1648
1649 (void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN,
1650 "cannot import pool '%s'"), origname);
1651
1652 if (newname != NULL) {
1653 if (!zpool_name_valid(hdl, B_FALSE, newname))
1654 return (zfs_error_fmt(hdl, EZFS_INVALIDNAME,
1655 dgettext(TEXT_DOMAIN, "cannot import '%s'"),
1656 newname));
1657 thename = (char *)newname;
1658 } else {
1659 thename = origname;
1660 }
1661
1662 if (props) {
1663 uint64_t version;
1664 prop_flags_t flags = { .create = B_FALSE, .import = B_TRUE };
1665
1666 verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_VERSION,
1667 &version) == 0);
1668
1669 if ((props = zpool_valid_proplist(hdl, origname,
1670 props, version, flags, errbuf)) == NULL) {
1671 return (-1);
1672 } else if (zcmd_write_src_nvlist(hdl, &zc, props) != 0) {
1673 nvlist_free(props);
1674 return (-1);
1675 }
1676 }
1677
1678 (void) strlcpy(zc.zc_name, thename, sizeof (zc.zc_name));
1679
1680 verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID,
1681 &zc.zc_guid) == 0);
1682
1683 if (zcmd_write_conf_nvlist(hdl, &zc, config) != 0) {
1684 nvlist_free(props);
1685 return (-1);
1686 }
1687 if (zcmd_alloc_dst_nvlist(hdl, &zc, zc.zc_nvlist_conf_size * 2) != 0) {
1688 nvlist_free(props);
1689 return (-1);
1690 }
1691
1692 zc.zc_cookie = flags;
1693 while ((ret = zfs_ioctl(hdl, ZFS_IOC_POOL_IMPORT, &zc)) != 0 &&
1694 errno == ENOMEM) {
1695 if (zcmd_expand_dst_nvlist(hdl, &zc) != 0) {
1696 zcmd_free_nvlists(&zc);
1697 return (-1);
1698 }
1699 }
1700 if (ret != 0)
1701 error = errno;
1702
1703 (void) zcmd_read_dst_nvlist(hdl, &zc, &nv);
1704 zpool_get_rewind_policy(config, &policy);
1705
1706 if (error) {
1707 char desc[1024];
1708
1709 /*
1710 * Dry-run failed, but we print out what success
1711 * looks like if we found a best txg
1712 */
1713 if (policy.zrp_request & ZPOOL_TRY_REWIND) {
1714 zpool_rewind_exclaim(hdl, newname ? origname : thename,
1715 B_TRUE, nv);
1716 nvlist_free(nv);
1717 return (-1);
1718 }
1719
1720 if (newname == NULL)
1721 (void) snprintf(desc, sizeof (desc),
1722 dgettext(TEXT_DOMAIN, "cannot import '%s'"),
1723 thename);
1724 else
1725 (void) snprintf(desc, sizeof (desc),
1726 dgettext(TEXT_DOMAIN, "cannot import '%s' as '%s'"),
1727 origname, thename);
1728
1729 switch (error) {
1730 case ENOTSUP:
1731 if (nv != NULL && nvlist_lookup_nvlist(nv,
1732 ZPOOL_CONFIG_LOAD_INFO, &nvinfo) == 0 &&
1733 nvlist_exists(nvinfo, ZPOOL_CONFIG_UNSUP_FEAT)) {
1734 (void) printf(dgettext(TEXT_DOMAIN, "This "
1735 "pool uses the following feature(s) not "
1736 "supported by this system:\n"));
1737 zpool_print_unsup_feat(nv);
1738 if (nvlist_exists(nvinfo,
1739 ZPOOL_CONFIG_CAN_RDONLY)) {
1740 (void) printf(dgettext(TEXT_DOMAIN,
1741 "All unsupported features are only "
1742 "required for writing to the pool."
1743 "\nThe pool can be imported using "
1744 "'-o readonly=on'.\n"));
1745 }
1746 }
1747 /*
1748 * Unsupported version.
1749 */
1750 (void) zfs_error(hdl, EZFS_BADVERSION, desc);
1751 break;
1752
1753 case EINVAL:
1754 (void) zfs_error(hdl, EZFS_INVALCONFIG, desc);
1755 break;
1756
1757 case EROFS:
1758 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1759 "one or more devices is read only"));
1760 (void) zfs_error(hdl, EZFS_BADDEV, desc);
1761 break;
1762
1763 case ENXIO:
1764 if (nv && nvlist_lookup_nvlist(nv,
1765 ZPOOL_CONFIG_LOAD_INFO, &nvinfo) == 0 &&
1766 nvlist_lookup_nvlist(nvinfo,
1767 ZPOOL_CONFIG_MISSING_DEVICES, &missing) == 0) {
1768 (void) printf(dgettext(TEXT_DOMAIN,
1769 "The devices below are missing, use "
1770 "'-m' to import the pool anyway:\n"));
1771 print_vdev_tree(hdl, NULL, missing, 2);
1772 (void) printf("\n");
1773 }
1774 (void) zpool_standard_error(hdl, error, desc);
1775 break;
1776
1777 case EEXIST:
1778 (void) zpool_standard_error(hdl, error, desc);
1779 break;
1780
1781 default:
1782 (void) zpool_standard_error(hdl, error, desc);
1783 zpool_explain_recover(hdl,
1784 newname ? origname : thename, -error, nv);
1785 break;
1786 }
1787
1788 nvlist_free(nv);
1789 ret = -1;
1790 } else {
1791 zpool_handle_t *zhp;
1792
1793 /*
1794 * This should never fail, but play it safe anyway.
1795 */
1796 if (zpool_open_silent(hdl, thename, &zhp) != 0)
1797 ret = -1;
1798 else if (zhp != NULL)
1799 zpool_close(zhp);
1800 if (policy.zrp_request &
1801 (ZPOOL_DO_REWIND | ZPOOL_TRY_REWIND)) {
1802 zpool_rewind_exclaim(hdl, newname ? origname : thename,
1803 ((policy.zrp_request & ZPOOL_TRY_REWIND) != 0), nv);
1804 }
1805 nvlist_free(nv);
1806 return (0);
1807 }
1808
1809 zcmd_free_nvlists(&zc);
1810 nvlist_free(props);
1811
1812 return (ret);
1813 }
1814
1815 /*
1816 * Scan the pool.
1817 */
1818 int
1819 zpool_scan(zpool_handle_t *zhp, pool_scan_func_t func)
1820 {
1821 zfs_cmd_t zc = { 0 };
1822 char msg[1024];
1823 libzfs_handle_t *hdl = zhp->zpool_hdl;
1824
1825 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
1826 zc.zc_cookie = func;
1827
1828 if (zfs_ioctl(hdl, ZFS_IOC_POOL_SCAN, &zc) == 0 ||
1829 (errno == ENOENT && func != POOL_SCAN_NONE))
1830 return (0);
1831
1832 if (func == POOL_SCAN_SCRUB) {
1833 (void) snprintf(msg, sizeof (msg),
1834 dgettext(TEXT_DOMAIN, "cannot scrub %s"), zc.zc_name);
1835 } else if (func == POOL_SCAN_NONE) {
1836 (void) snprintf(msg, sizeof (msg),
1837 dgettext(TEXT_DOMAIN, "cannot cancel scrubbing %s"),
1838 zc.zc_name);
1839 } else {
1840 assert(!"unexpected result");
1841 }
1842
1843 if (errno == EBUSY) {
1844 nvlist_t *nvroot;
1845 pool_scan_stat_t *ps = NULL;
1846 uint_t psc;
1847
1848 verify(nvlist_lookup_nvlist(zhp->zpool_config,
1849 ZPOOL_CONFIG_VDEV_TREE, &nvroot) == 0);
1850 (void) nvlist_lookup_uint64_array(nvroot,
1851 ZPOOL_CONFIG_SCAN_STATS, (uint64_t **)&ps, &psc);
1852 if (ps && ps->pss_func == POOL_SCAN_SCRUB)
1853 return (zfs_error(hdl, EZFS_SCRUBBING, msg));
1854 else
1855 return (zfs_error(hdl, EZFS_RESILVERING, msg));
1856 } else if (errno == ENOENT) {
1857 return (zfs_error(hdl, EZFS_NO_SCRUB, msg));
1858 } else {
1859 return (zpool_standard_error(hdl, errno, msg));
1860 }
1861 }
1862
1863 /*
1864 * This provides a very minimal check whether a given string is likely a
1865 * c#t#d# style string. Users of this are expected to do their own
1866 * verification of the s# part.
1867 */
1868 #define CTD_CHECK(str) (str && str[0] == 'c' && isdigit(str[1]))
1869
1870 /*
1871 * More elaborate version for ones which may start with "/dev/dsk/"
1872 * and the like.
1873 */
1874 static int
1875 ctd_check_path(char *str) {
1876 /*
1877 * If it starts with a slash, check the last component.
1878 */
1879 if (str && str[0] == '/') {
1880 char *tmp = strrchr(str, '/');
1881
1882 /*
1883 * If it ends in "/old", check the second-to-last
1884 * component of the string instead.
1885 */
1886 if (tmp != str && strcmp(tmp, "/old") == 0) {
1887 for (tmp--; *tmp != '/'; tmp--)
1888 ;
1889 }
1890 str = tmp + 1;
1891 }
1892 return (CTD_CHECK(str));
1893 }
1894
1895 /*
1896 * Find a vdev that matches the search criteria specified. We use the
1897 * the nvpair name to determine how we should look for the device.
1898 * 'avail_spare' is set to TRUE if the provided guid refers to an AVAIL
1899 * spare; but FALSE if its an INUSE spare.
1900 */
1901 static nvlist_t *
1902 vdev_to_nvlist_iter(nvlist_t *nv, nvlist_t *search, boolean_t *avail_spare,
1903 boolean_t *l2cache, boolean_t *log)
1904 {
1905 uint_t c, children;
1906 nvlist_t **child;
1907 nvlist_t *ret;
1908 uint64_t is_log;
1909 char *srchkey;
1910 nvpair_t *pair = nvlist_next_nvpair(search, NULL);
1911
1912 /* Nothing to look for */
1913 if (search == NULL || pair == NULL)
1914 return (NULL);
1915
1916 /* Obtain the key we will use to search */
1917 srchkey = nvpair_name(pair);
1918
1919 switch (nvpair_type(pair)) {
1920 case DATA_TYPE_UINT64:
1921 if (strcmp(srchkey, ZPOOL_CONFIG_GUID) == 0) {
1922 uint64_t srchval, theguid;
1923
1924 verify(nvpair_value_uint64(pair, &srchval) == 0);
1925 verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID,
1926 &theguid) == 0);
1927 if (theguid == srchval)
1928 return (nv);
1929 }
1930 break;
1931
1932 case DATA_TYPE_STRING: {
1933 char *srchval, *val;
1934
1935 verify(nvpair_value_string(pair, &srchval) == 0);
1936 if (nvlist_lookup_string(nv, srchkey, &val) != 0)
1937 break;
1938
1939 /*
1940 * Search for the requested value. Special cases:
1941 *
1942 * - ZPOOL_CONFIG_PATH for whole disk entries. These end in
1943 * "s0" or "s0/old". The "s0" part is hidden from the user,
1944 * but included in the string, so this matches around it.
1945 * - looking for a top-level vdev name (i.e. ZPOOL_CONFIG_TYPE).
1946 *
1947 * Otherwise, all other searches are simple string compares.
1948 */
1949 if (strcmp(srchkey, ZPOOL_CONFIG_PATH) == 0 &&
1950 ctd_check_path(val)) {
1951 uint64_t wholedisk = 0;
1952
1953 (void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_WHOLE_DISK,
1954 &wholedisk);
1955 if (wholedisk) {
1956 int slen = strlen(srchval);
1957 int vlen = strlen(val);
1958
1959 if (slen != vlen - 2)
1960 break;
1961
1962 /*
1963 * make_leaf_vdev() should only set
1964 * wholedisk for ZPOOL_CONFIG_PATHs which
1965 * will include "/dev/dsk/", giving plenty of
1966 * room for the indices used next.
1967 */
1968 ASSERT(vlen >= 6);
1969
1970 /*
1971 * strings identical except trailing "s0"
1972 */
1973 if (strcmp(&val[vlen - 2], "s0") == 0 &&
1974 strncmp(srchval, val, slen) == 0)
1975 return (nv);
1976
1977 /*
1978 * strings identical except trailing "s0/old"
1979 */
1980 if (strcmp(&val[vlen - 6], "s0/old") == 0 &&
1981 strcmp(&srchval[slen - 4], "/old") == 0 &&
1982 strncmp(srchval, val, slen - 4) == 0)
1983 return (nv);
1984
1985 break;
1986 }
1987 } else if (strcmp(srchkey, ZPOOL_CONFIG_TYPE) == 0 && val) {
1988 char *type, *idx, *end, *p;
1989 uint64_t id, vdev_id;
1990
1991 /*
1992 * Determine our vdev type, keeping in mind
1993 * that the srchval is composed of a type and
1994 * vdev id pair (i.e. mirror-4).
1995 */
1996 if ((type = strdup(srchval)) == NULL)
1997 return (NULL);
1998
1999 if ((p = strrchr(type, '-')) == NULL) {
2000 free(type);
2001 break;
2002 }
2003 idx = p + 1;
2004 *p = '\0';
2005
2006 /*
2007 * If the types don't match then keep looking.
2008 */
2009 if (strncmp(val, type, strlen(val)) != 0) {
2010 free(type);
2011 break;
2012 }
2013
2014 verify(strncmp(type, VDEV_TYPE_RAIDZ,
2015 strlen(VDEV_TYPE_RAIDZ)) == 0 ||
2016 strncmp(type, VDEV_TYPE_MIRROR,
2017 strlen(VDEV_TYPE_MIRROR)) == 0);
2018 verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_ID,
2019 &id) == 0);
2020
2021 errno = 0;
2022 vdev_id = strtoull(idx, &end, 10);
2023
2024 free(type);
2025 if (errno != 0)
2026 return (NULL);
2027
2028 /*
2029 * Now verify that we have the correct vdev id.
2030 */
2031 if (vdev_id == id)
2032 return (nv);
2033 }
2034
2035 /*
2036 * Common case
2037 */
2038 if (strcmp(srchval, val) == 0)
2039 return (nv);
2040 break;
2041 }
2042
2043 default:
2044 break;
2045 }
2046
2047 if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN,
2048 &child, &children) != 0)
2049 return (NULL);
2050
2051 for (c = 0; c < children; c++) {
2052 if ((ret = vdev_to_nvlist_iter(child[c], search,
2053 avail_spare, l2cache, NULL)) != NULL) {
2054 /*
2055 * The 'is_log' value is only set for the toplevel
2056 * vdev, not the leaf vdevs. So we always lookup the
2057 * log device from the root of the vdev tree (where
2058 * 'log' is non-NULL).
2059 */
2060 if (log != NULL &&
2061 nvlist_lookup_uint64(child[c],
2062 ZPOOL_CONFIG_IS_LOG, &is_log) == 0 &&
2063 is_log) {
2064 *log = B_TRUE;
2065 }
2066 return (ret);
2067 }
2068 }
2069
2070 if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_SPARES,
2071 &child, &children) == 0) {
2072 for (c = 0; c < children; c++) {
2073 if ((ret = vdev_to_nvlist_iter(child[c], search,
2074 avail_spare, l2cache, NULL)) != NULL) {
2075 *avail_spare = B_TRUE;
2076 return (ret);
2077 }
2078 }
2079 }
2080
2081 if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_L2CACHE,
2082 &child, &children) == 0) {
2083 for (c = 0; c < children; c++) {
2084 if ((ret = vdev_to_nvlist_iter(child[c], search,
2085 avail_spare, l2cache, NULL)) != NULL) {
2086 *l2cache = B_TRUE;
2087 return (ret);
2088 }
2089 }
2090 }
2091
2092 return (NULL);
2093 }
2094
2095 /*
2096 * Given a physical path (minus the "/devices" prefix), find the
2097 * associated vdev.
2098 */
2099 nvlist_t *
2100 zpool_find_vdev_by_physpath(zpool_handle_t *zhp, const char *ppath,
2101 boolean_t *avail_spare, boolean_t *l2cache, boolean_t *log)
2102 {
2103 nvlist_t *search, *nvroot, *ret;
2104
2105 verify(nvlist_alloc(&search, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2106 verify(nvlist_add_string(search, ZPOOL_CONFIG_PHYS_PATH, ppath) == 0);
2107
2108 verify(nvlist_lookup_nvlist(zhp->zpool_config, ZPOOL_CONFIG_VDEV_TREE,
2109 &nvroot) == 0);
2110
2111 *avail_spare = B_FALSE;
2112 *l2cache = B_FALSE;
2113 if (log != NULL)
2114 *log = B_FALSE;
2115 ret = vdev_to_nvlist_iter(nvroot, search, avail_spare, l2cache, log);
2116 nvlist_free(search);
2117
2118 return (ret);
2119 }
2120
2121 /*
2122 * Determine if we have an "interior" top-level vdev (i.e mirror/raidz).
2123 */
2124 boolean_t
2125 zpool_vdev_is_interior(const char *name)
2126 {
2127 if (strncmp(name, VDEV_TYPE_RAIDZ, strlen(VDEV_TYPE_RAIDZ)) == 0 ||
2128 strncmp(name, VDEV_TYPE_MIRROR, strlen(VDEV_TYPE_MIRROR)) == 0)
2129 return (B_TRUE);
2130 return (B_FALSE);
2131 }
2132
2133 nvlist_t *
2134 zpool_find_vdev(zpool_handle_t *zhp, const char *path, boolean_t *avail_spare,
2135 boolean_t *l2cache, boolean_t *log)
2136 {
2137 char buf[MAXPATHLEN];
2138 char *end;
2139 nvlist_t *nvroot, *search, *ret;
2140 uint64_t guid;
2141
2142 verify(nvlist_alloc(&search, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2143
2144 guid = strtoull(path, &end, 10);
2145 if (guid != 0 && *end == '\0') {
2146 verify(nvlist_add_uint64(search, ZPOOL_CONFIG_GUID, guid) == 0);
2147 } else if (zpool_vdev_is_interior(path)) {
2148 verify(nvlist_add_string(search, ZPOOL_CONFIG_TYPE, path) == 0);
2149 } else if (path[0] != '/') {
2150 (void) snprintf(buf, sizeof (buf), "%s%s", "/dev/dsk/", path);
2151 verify(nvlist_add_string(search, ZPOOL_CONFIG_PATH, buf) == 0);
2152 } else {
2153 verify(nvlist_add_string(search, ZPOOL_CONFIG_PATH, path) == 0);
2154 }
2155
2156 verify(nvlist_lookup_nvlist(zhp->zpool_config, ZPOOL_CONFIG_VDEV_TREE,
2157 &nvroot) == 0);
2158
2159 *avail_spare = B_FALSE;
2160 *l2cache = B_FALSE;
2161 if (log != NULL)
2162 *log = B_FALSE;
2163 ret = vdev_to_nvlist_iter(nvroot, search, avail_spare, l2cache, log);
2164 nvlist_free(search);
2165
2166 return (ret);
2167 }
2168
2169 static int
2170 vdev_online(nvlist_t *nv)
2171 {
2172 uint64_t ival;
2173
2174 if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_OFFLINE, &ival) == 0 ||
2175 nvlist_lookup_uint64(nv, ZPOOL_CONFIG_FAULTED, &ival) == 0 ||
2176 nvlist_lookup_uint64(nv, ZPOOL_CONFIG_REMOVED, &ival) == 0)
2177 return (0);
2178
2179 return (1);
2180 }
2181
2182 /*
2183 * Helper function for zpool_get_physpaths().
2184 */
2185 static int
2186 vdev_get_one_physpath(nvlist_t *config, char *physpath, size_t physpath_size,
2187 size_t *bytes_written)
2188 {
2189 size_t bytes_left, pos, rsz;
2190 char *tmppath;
2191 const char *format;
2192
2193 if (nvlist_lookup_string(config, ZPOOL_CONFIG_PHYS_PATH,
2194 &tmppath) != 0)
2195 return (EZFS_NODEVICE);
2196
2197 pos = *bytes_written;
2198 bytes_left = physpath_size - pos;
2199 format = (pos == 0) ? "%s" : " %s";
2200
2201 rsz = snprintf(physpath + pos, bytes_left, format, tmppath);
2202 *bytes_written += rsz;
2203
2204 if (rsz >= bytes_left) {
2205 /* if physpath was not copied properly, clear it */
2206 if (bytes_left != 0) {
2207 physpath[pos] = 0;
2208 }
2209 return (EZFS_NOSPC);
2210 }
2211 return (0);
2212 }
2213
2214 static int
2215 vdev_get_physpaths(nvlist_t *nv, char *physpath, size_t phypath_size,
2216 size_t *rsz, boolean_t is_spare)
2217 {
2218 char *type;
2219 int ret;
2220
2221 if (nvlist_lookup_string(nv, ZPOOL_CONFIG_TYPE, &type) != 0)
2222 return (EZFS_INVALCONFIG);
2223
2224 if (strcmp(type, VDEV_TYPE_DISK) == 0) {
2225 /*
2226 * An active spare device has ZPOOL_CONFIG_IS_SPARE set.
2227 * For a spare vdev, we only want to boot from the active
2228 * spare device.
2229 */
2230 if (is_spare) {
2231 uint64_t spare = 0;
2232 (void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_IS_SPARE,
2233 &spare);
2234 if (!spare)
2235 return (EZFS_INVALCONFIG);
2236 }
2237
2238 if (vdev_online(nv)) {
2239 if ((ret = vdev_get_one_physpath(nv, physpath,
2240 phypath_size, rsz)) != 0)
2241 return (ret);
2242 }
2243 } else if (strcmp(type, VDEV_TYPE_MIRROR) == 0 ||
2244 strcmp(type, VDEV_TYPE_REPLACING) == 0 ||
2245 (is_spare = (strcmp(type, VDEV_TYPE_SPARE) == 0))) {
2246 nvlist_t **child;
2247 uint_t count;
2248 int i, ret;
2249
2250 if (nvlist_lookup_nvlist_array(nv,
2251 ZPOOL_CONFIG_CHILDREN, &child, &count) != 0)
2252 return (EZFS_INVALCONFIG);
2253
2254 for (i = 0; i < count; i++) {
2255 ret = vdev_get_physpaths(child[i], physpath,
2256 phypath_size, rsz, is_spare);
2257 if (ret == EZFS_NOSPC)
2258 return (ret);
2259 }
2260 }
2261
2262 return (EZFS_POOL_INVALARG);
2263 }
2264
2265 /*
2266 * Get phys_path for a root pool config.
2267 * Return 0 on success; non-zero on failure.
2268 */
2269 static int
2270 zpool_get_config_physpath(nvlist_t *config, char *physpath, size_t phypath_size)
2271 {
2272 size_t rsz;
2273 nvlist_t *vdev_root;
2274 nvlist_t **child;
2275 uint_t count;
2276 char *type;
2277
2278 rsz = 0;
2279
2280 if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE,
2281 &vdev_root) != 0)
2282 return (EZFS_INVALCONFIG);
2283
2284 if (nvlist_lookup_string(vdev_root, ZPOOL_CONFIG_TYPE, &type) != 0 ||
2285 nvlist_lookup_nvlist_array(vdev_root, ZPOOL_CONFIG_CHILDREN,
2286 &child, &count) != 0)
2287 return (EZFS_INVALCONFIG);
2288
2289 /*
2290 * root pool can not have EFI labeled disks and can only have
2291 * a single top-level vdev.
2292 */
2293 if (strcmp(type, VDEV_TYPE_ROOT) != 0 || count != 1 ||
2294 pool_uses_efi(vdev_root))
2295 return (EZFS_POOL_INVALARG);
2296
2297 (void) vdev_get_physpaths(child[0], physpath, phypath_size, &rsz,
2298 B_FALSE);
2299
2300 /* No online devices */
2301 if (rsz == 0)
2302 return (EZFS_NODEVICE);
2303
2304 return (0);
2305 }
2306
2307 /*
2308 * Get phys_path for a root pool
2309 * Return 0 on success; non-zero on failure.
2310 */
2311 int
2312 zpool_get_physpath(zpool_handle_t *zhp, char *physpath, size_t phypath_size)
2313 {
2314 return (zpool_get_config_physpath(zhp->zpool_config, physpath,
2315 phypath_size));
2316 }
2317
2318 /*
2319 * If the device has being dynamically expanded then we need to relabel
2320 * the disk to use the new unallocated space.
2321 */
2322 static int
2323 zpool_relabel_disk(libzfs_handle_t *hdl, const char *name)
2324 {
2325 char path[MAXPATHLEN];
2326 char errbuf[1024];
2327 int fd, error;
2328 int (*_efi_use_whole_disk)(int);
2329
2330 if ((_efi_use_whole_disk = (int (*)(int))dlsym(RTLD_DEFAULT,
2331 "efi_use_whole_disk")) == NULL)
2332 return (-1);
2333
2334 (void) snprintf(path, sizeof (path), "%s/%s", RDISK_ROOT, name);
2335
2336 if ((fd = open(path, O_RDWR | O_NDELAY)) < 0) {
2337 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "cannot "
2338 "relabel '%s': unable to open device"), name);
2339 return (zfs_error(hdl, EZFS_OPENFAILED, errbuf));
2340 }
2341
2342 /*
2343 * It's possible that we might encounter an error if the device
2344 * does not have any unallocated space left. If so, we simply
2345 * ignore that error and continue on.
2346 */
2347 error = _efi_use_whole_disk(fd);
2348 (void) close(fd);
2349 if (error && error != VT_ENOSPC) {
2350 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "cannot "
2351 "relabel '%s': unable to read disk capacity"), name);
2352 return (zfs_error(hdl, EZFS_NOCAP, errbuf));
2353 }
2354 return (0);
2355 }
2356
2357 /*
2358 * Bring the specified vdev online. The 'flags' parameter is a set of the
2359 * ZFS_ONLINE_* flags.
2360 */
2361 int
2362 zpool_vdev_online(zpool_handle_t *zhp, const char *path, int flags,
2363 vdev_state_t *newstate)
2364 {
2365 zfs_cmd_t zc = { 0 };
2366 char msg[1024];
2367 nvlist_t *tgt;
2368 boolean_t avail_spare, l2cache, islog;
2369 libzfs_handle_t *hdl = zhp->zpool_hdl;
2370
2371 if (flags & ZFS_ONLINE_EXPAND) {
2372 (void) snprintf(msg, sizeof (msg),
2373 dgettext(TEXT_DOMAIN, "cannot expand %s"), path);
2374 } else {
2375 (void) snprintf(msg, sizeof (msg),
2376 dgettext(TEXT_DOMAIN, "cannot online %s"), path);
2377 }
2378
2379 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
2380 if ((tgt = zpool_find_vdev(zhp, path, &avail_spare, &l2cache,
2381 &islog)) == NULL)
2382 return (zfs_error(hdl, EZFS_NODEVICE, msg));
2383
2384 verify(nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID, &zc.zc_guid) == 0);
2385
2386 if (avail_spare)
2387 return (zfs_error(hdl, EZFS_ISSPARE, msg));
2388
2389 if (flags & ZFS_ONLINE_EXPAND ||
2390 zpool_get_prop_int(zhp, ZPOOL_PROP_AUTOEXPAND, NULL)) {
2391 char *pathname = NULL;
2392 uint64_t wholedisk = 0;
2393
2394 (void) nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_WHOLE_DISK,
2395 &wholedisk);
2396 verify(nvlist_lookup_string(tgt, ZPOOL_CONFIG_PATH,
2397 &pathname) == 0);
2398
2399 /*
2400 * XXX - L2ARC 1.0 devices can't support expansion.
2401 */
2402 if (l2cache) {
2403 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2404 "cannot expand cache devices"));
2405 return (zfs_error(hdl, EZFS_VDEVNOTSUP, msg));
2406 }
2407
2408 if (wholedisk) {
2409 pathname += strlen(DISK_ROOT) + 1;
2410 (void) zpool_relabel_disk(hdl, pathname);
2411 }
2412 }
2413
2414 zc.zc_cookie = VDEV_STATE_ONLINE;
2415 zc.zc_obj = flags;
2416
2417 if (zfs_ioctl(hdl, ZFS_IOC_VDEV_SET_STATE, &zc) != 0) {
2418 if (errno == EINVAL) {
2419 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "was split "
2420 "from this pool into a new one. Use '%s' "
2421 "instead"), "zpool detach");
2422 return (zfs_error(hdl, EZFS_POSTSPLIT_ONLINE, msg));
2423 }
2424 return (zpool_standard_error(hdl, errno, msg));
2425 }
2426
2427 *newstate = zc.zc_cookie;
2428 return (0);
2429 }
2430
2431 /*
2432 * Take the specified vdev offline
2433 */
2434 int
2435 zpool_vdev_offline(zpool_handle_t *zhp, const char *path, boolean_t istmp)
2436 {
2437 zfs_cmd_t zc = { 0 };
2438 char msg[1024];
2439 nvlist_t *tgt;
2440 boolean_t avail_spare, l2cache;
2441 libzfs_handle_t *hdl = zhp->zpool_hdl;
2442
2443 (void) snprintf(msg, sizeof (msg),
2444 dgettext(TEXT_DOMAIN, "cannot offline %s"), path);
2445
2446 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
2447 if ((tgt = zpool_find_vdev(zhp, path, &avail_spare, &l2cache,
2448 NULL)) == NULL)
2449 return (zfs_error(hdl, EZFS_NODEVICE, msg));
2450
2451 verify(nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID, &zc.zc_guid) == 0);
2452
2453 if (avail_spare)
2454 return (zfs_error(hdl, EZFS_ISSPARE, msg));
2455
2456 zc.zc_cookie = VDEV_STATE_OFFLINE;
2457 zc.zc_obj = istmp ? ZFS_OFFLINE_TEMPORARY : 0;
2458
2459 if (zfs_ioctl(hdl, ZFS_IOC_VDEV_SET_STATE, &zc) == 0)
2460 return (0);
2461
2462 switch (errno) {
2463 case EBUSY:
2464
2465 /*
2466 * There are no other replicas of this device.
2467 */
2468 return (zfs_error(hdl, EZFS_NOREPLICAS, msg));
2469
2470 case EEXIST:
2471 /*
2472 * The log device has unplayed logs
2473 */
2474 return (zfs_error(hdl, EZFS_UNPLAYED_LOGS, msg));
2475
2476 default:
2477 return (zpool_standard_error(hdl, errno, msg));
2478 }
2479 }
2480
2481 /*
2482 * Mark the given vdev faulted.
2483 */
2484 int
2485 zpool_vdev_fault(zpool_handle_t *zhp, uint64_t guid, vdev_aux_t aux)
2486 {
2487 zfs_cmd_t zc = { 0 };
2488 char msg[1024];
2489 libzfs_handle_t *hdl = zhp->zpool_hdl;
2490
2491 (void) snprintf(msg, sizeof (msg),
2492 dgettext(TEXT_DOMAIN, "cannot fault %llu"), guid);
2493
2494 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
2495 zc.zc_guid = guid;
2496 zc.zc_cookie = VDEV_STATE_FAULTED;
2497 zc.zc_obj = aux;
2498
2499 if (ioctl(hdl->libzfs_fd, ZFS_IOC_VDEV_SET_STATE, &zc) == 0)
2500 return (0);
2501
2502 switch (errno) {
2503 case EBUSY:
2504
2505 /*
2506 * There are no other replicas of this device.
2507 */
2508 return (zfs_error(hdl, EZFS_NOREPLICAS, msg));
2509
2510 default:
2511 return (zpool_standard_error(hdl, errno, msg));
2512 }
2513
2514 }
2515
2516 /*
2517 * Mark the given vdev degraded.
2518 */
2519 int
2520 zpool_vdev_degrade(zpool_handle_t *zhp, uint64_t guid, vdev_aux_t aux)
2521 {
2522 zfs_cmd_t zc = { 0 };
2523 char msg[1024];
2524 libzfs_handle_t *hdl = zhp->zpool_hdl;
2525
2526 (void) snprintf(msg, sizeof (msg),
2527 dgettext(TEXT_DOMAIN, "cannot degrade %llu"), guid);
2528
2529 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
2530 zc.zc_guid = guid;
2531 zc.zc_cookie = VDEV_STATE_DEGRADED;
2532 zc.zc_obj = aux;
2533
2534 if (ioctl(hdl->libzfs_fd, ZFS_IOC_VDEV_SET_STATE, &zc) == 0)
2535 return (0);
2536
2537 return (zpool_standard_error(hdl, errno, msg));
2538 }
2539
2540 /*
2541 * Returns TRUE if the given nvlist is a vdev that was originally swapped in as
2542 * a hot spare.
2543 */
2544 static boolean_t
2545 is_replacing_spare(nvlist_t *search, nvlist_t *tgt, int which)
2546 {
2547 nvlist_t **child;
2548 uint_t c, children;
2549 char *type;
2550
2551 if (nvlist_lookup_nvlist_array(search, ZPOOL_CONFIG_CHILDREN, &child,
2552 &children) == 0) {
2553 verify(nvlist_lookup_string(search, ZPOOL_CONFIG_TYPE,
2554 &type) == 0);
2555
2556 if (strcmp(type, VDEV_TYPE_SPARE) == 0 &&
2557 children == 2 && child[which] == tgt)
2558 return (B_TRUE);
2559
2560 for (c = 0; c < children; c++)
2561 if (is_replacing_spare(child[c], tgt, which))
2562 return (B_TRUE);
2563 }
2564
2565 return (B_FALSE);
2566 }
2567
2568 /*
2569 * Attach new_disk (fully described by nvroot) to old_disk.
2570 * If 'replacing' is specified, the new disk will replace the old one.
2571 */
2572 int
2573 zpool_vdev_attach(zpool_handle_t *zhp,
2574 const char *old_disk, const char *new_disk, nvlist_t *nvroot, int replacing)
2575 {
2576 zfs_cmd_t zc = { 0 };
2577 char msg[1024];
2578 int ret;
2579 nvlist_t *tgt;
2580 boolean_t avail_spare, l2cache, islog;
2581 uint64_t val;
2582 char *newname;
2583 nvlist_t **child;
2584 uint_t children;
2585 nvlist_t *config_root;
2586 libzfs_handle_t *hdl = zhp->zpool_hdl;
2587 boolean_t rootpool = zpool_is_bootable(zhp);
2588
2589 if (replacing)
2590 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN,
2591 "cannot replace %s with %s"), old_disk, new_disk);
2592 else
2593 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN,
2594 "cannot attach %s to %s"), new_disk, old_disk);
2595
2596 /*
2597 * If this is a root pool, make sure that we're not attaching an
2598 * EFI labeled device.
2599 */
2600 if (rootpool && pool_uses_efi(nvroot)) {
2601 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2602 "EFI labeled devices are not supported on root pools."));
2603 return (zfs_error(hdl, EZFS_POOL_NOTSUP, msg));
2604 }
2605
2606 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
2607 if ((tgt = zpool_find_vdev(zhp, old_disk, &avail_spare, &l2cache,
2608 &islog)) == 0)
2609 return (zfs_error(hdl, EZFS_NODEVICE, msg));
2610
2611 if (avail_spare)
2612 return (zfs_error(hdl, EZFS_ISSPARE, msg));
2613
2614 if (l2cache)
2615 return (zfs_error(hdl, EZFS_ISL2CACHE, msg));
2616
2617 verify(nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID, &zc.zc_guid) == 0);
2618 zc.zc_cookie = replacing;
2619
2620 if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_CHILDREN,
2621 &child, &children) != 0 || children != 1) {
2622 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2623 "new device must be a single disk"));
2624 return (zfs_error(hdl, EZFS_INVALCONFIG, msg));
2625 }
2626
2627 verify(nvlist_lookup_nvlist(zpool_get_config(zhp, NULL),
2628 ZPOOL_CONFIG_VDEV_TREE, &config_root) == 0);
2629
2630 if ((newname = zpool_vdev_name(NULL, NULL, child[0], B_FALSE)) == NULL)
2631 return (-1);
2632
2633 /*
2634 * If the target is a hot spare that has been swapped in, we can only
2635 * replace it with another hot spare.
2636 */
2637 if (replacing &&
2638 nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_IS_SPARE, &val) == 0 &&
2639 (zpool_find_vdev(zhp, newname, &avail_spare, &l2cache,
2640 NULL) == NULL || !avail_spare) &&
2641 is_replacing_spare(config_root, tgt, 1)) {
2642 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2643 "can only be replaced by another hot spare"));
2644 free(newname);
2645 return (zfs_error(hdl, EZFS_BADTARGET, msg));
2646 }
2647
2648 free(newname);
2649
2650 if (zcmd_write_conf_nvlist(hdl, &zc, nvroot) != 0)
2651 return (-1);
2652
2653 ret = zfs_ioctl(hdl, ZFS_IOC_VDEV_ATTACH, &zc);
2654
2655 zcmd_free_nvlists(&zc);
2656
2657 if (ret == 0) {
2658 if (rootpool) {
2659 /*
2660 * XXX need a better way to prevent user from
2661 * booting up a half-baked vdev.
2662 */
2663 (void) fprintf(stderr, dgettext(TEXT_DOMAIN, "Make "
2664 "sure to wait until resilver is done "
2665 "before rebooting.\n"));
2666 }
2667 return (0);
2668 }
2669
2670 switch (errno) {
2671 case ENOTSUP:
2672 /*
2673 * Can't attach to or replace this type of vdev.
2674 */
2675 if (replacing) {
2676 uint64_t version = zpool_get_prop_int(zhp,
2677 ZPOOL_PROP_VERSION, NULL);
2678
2679 if (islog)
2680 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2681 "cannot replace a log with a spare"));
2682 else if (version >= SPA_VERSION_MULTI_REPLACE)
2683 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2684 "already in replacing/spare config; wait "
2685 "for completion or use 'zpool detach'"));
2686 else
2687 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2688 "cannot replace a replacing device"));
2689 } else {
2690 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2691 "can only attach to mirrors and top-level "
2692 "disks"));
2693 }
2694 (void) zfs_error(hdl, EZFS_BADTARGET, msg);
2695 break;
2696
2697 case EINVAL:
2698 /*
2699 * The new device must be a single disk.
2700 */
2701 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2702 "new device must be a single disk"));
2703 (void) zfs_error(hdl, EZFS_INVALCONFIG, msg);
2704 break;
2705
2706 case EBUSY:
2707 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "%s is busy"),
2708 new_disk);
2709 (void) zfs_error(hdl, EZFS_BADDEV, msg);
2710 break;
2711
2712 case EOVERFLOW:
2713 /*
2714 * The new device is too small.
2715 */
2716 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2717 "device is too small"));
2718 (void) zfs_error(hdl, EZFS_BADDEV, msg);
2719 break;
2720
2721 case EDOM:
2722 /*
2723 * The new device has a different alignment requirement.
2724 */
2725 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2726 "devices have different sector alignment"));
2727 (void) zfs_error(hdl, EZFS_BADDEV, msg);
2728 break;
2729
2730 case ENAMETOOLONG:
2731 /*
2732 * The resulting top-level vdev spec won't fit in the label.
2733 */
2734 (void) zfs_error(hdl, EZFS_DEVOVERFLOW, msg);
2735 break;
2736
2737 default:
2738 (void) zpool_standard_error(hdl, errno, msg);
2739 }
2740
2741 return (-1);
2742 }
2743
2744 /*
2745 * Detach the specified device.
2746 */
2747 int
2748 zpool_vdev_detach(zpool_handle_t *zhp, const char *path)
2749 {
2750 zfs_cmd_t zc = { 0 };
2751 char msg[1024];
2752 nvlist_t *tgt;
2753 boolean_t avail_spare, l2cache;
2754 libzfs_handle_t *hdl = zhp->zpool_hdl;
2755
2756 (void) snprintf(msg, sizeof (msg),
2757 dgettext(TEXT_DOMAIN, "cannot detach %s"), path);
2758
2759 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
2760 if ((tgt = zpool_find_vdev(zhp, path, &avail_spare, &l2cache,
2761 NULL)) == 0)
2762 return (zfs_error(hdl, EZFS_NODEVICE, msg));
2763
2764 if (avail_spare)
2765 return (zfs_error(hdl, EZFS_ISSPARE, msg));
2766
2767 if (l2cache)
2768 return (zfs_error(hdl, EZFS_ISL2CACHE, msg));
2769
2770 verify(nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID, &zc.zc_guid) == 0);
2771
2772 if (zfs_ioctl(hdl, ZFS_IOC_VDEV_DETACH, &zc) == 0)
2773 return (0);
2774
2775 switch (errno) {
2776
2777 case ENOTSUP:
2778 /*
2779 * Can't detach from this type of vdev.
2780 */
2781 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "only "
2782 "applicable to mirror and replacing vdevs"));
2783 (void) zfs_error(hdl, EZFS_BADTARGET, msg);
2784 break;
2785
2786 case EBUSY:
2787 /*
2788 * There are no other replicas of this device.
2789 */
2790 (void) zfs_error(hdl, EZFS_NOREPLICAS, msg);
2791 break;
2792
2793 default:
2794 (void) zpool_standard_error(hdl, errno, msg);
2795 }
2796
2797 return (-1);
2798 }
2799
2800 /*
2801 * Find a mirror vdev in the source nvlist.
2802 *
2803 * The mchild array contains a list of disks in one of the top-level mirrors
2804 * of the source pool. The schild array contains a list of disks that the
2805 * user specified on the command line. We loop over the mchild array to
2806 * see if any entry in the schild array matches.
2807 *
2808 * If a disk in the mchild array is found in the schild array, we return
2809 * the index of that entry. Otherwise we return -1.
2810 */
2811 static int
2812 find_vdev_entry(zpool_handle_t *zhp, nvlist_t **mchild, uint_t mchildren,
2813 nvlist_t **schild, uint_t schildren)
2814 {
2815 uint_t mc;
2816
2817 for (mc = 0; mc < mchildren; mc++) {
2818 uint_t sc;
2819 char *mpath = zpool_vdev_name(zhp->zpool_hdl, zhp,
2820 mchild[mc], B_FALSE);
2821
2822 for (sc = 0; sc < schildren; sc++) {
2823 char *spath = zpool_vdev_name(zhp->zpool_hdl, zhp,
2824 schild[sc], B_FALSE);
2825 boolean_t result = (strcmp(mpath, spath) == 0);
2826
2827 free(spath);
2828 if (result) {
2829 free(mpath);
2830 return (mc);
2831 }
2832 }
2833
2834 free(mpath);
2835 }
2836
2837 return (-1);
2838 }
2839
2840 /*
2841 * Split a mirror pool. If newroot points to null, then a new nvlist
2842 * is generated and it is the responsibility of the caller to free it.
2843 */
2844 int
2845 zpool_vdev_split(zpool_handle_t *zhp, char *newname, nvlist_t **newroot,
2846 nvlist_t *props, splitflags_t flags)
2847 {
2848 zfs_cmd_t zc = { 0 };
2849 char msg[1024];
2850 nvlist_t *tree, *config, **child, **newchild, *newconfig = NULL;
2851 nvlist_t **varray = NULL, *zc_props = NULL;
2852 uint_t c, children, newchildren, lastlog = 0, vcount, found = 0;
2853 libzfs_handle_t *hdl = zhp->zpool_hdl;
2854 uint64_t vers;
2855 boolean_t freelist = B_FALSE, memory_err = B_TRUE;
2856 int retval = 0;
2857
2858 (void) snprintf(msg, sizeof (msg),
2859 dgettext(TEXT_DOMAIN, "Unable to split %s"), zhp->zpool_name);
2860
2861 if (!zpool_name_valid(hdl, B_FALSE, newname))
2862 return (zfs_error(hdl, EZFS_INVALIDNAME, msg));
2863
2864 if ((config = zpool_get_config(zhp, NULL)) == NULL) {
2865 (void) fprintf(stderr, gettext("Internal error: unable to "
2866 "retrieve pool configuration\n"));
2867 return (-1);
2868 }
2869
2870 verify(nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE, &tree)
2871 == 0);
2872 verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_VERSION, &vers) == 0);
2873
2874 if (props) {
2875 prop_flags_t flags = { .create = B_FALSE, .import = B_TRUE };
2876 if ((zc_props = zpool_valid_proplist(hdl, zhp->zpool_name,
2877 props, vers, flags, msg)) == NULL)
2878 return (-1);
2879 }
2880
2881 if (nvlist_lookup_nvlist_array(tree, ZPOOL_CONFIG_CHILDREN, &child,
2882 &children) != 0) {
2883 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2884 "Source pool is missing vdev tree"));
2885 if (zc_props)
2886 nvlist_free(zc_props);
2887 return (-1);
2888 }
2889
2890 varray = zfs_alloc(hdl, children * sizeof (nvlist_t *));
2891 vcount = 0;
2892
2893 if (*newroot == NULL ||
2894 nvlist_lookup_nvlist_array(*newroot, ZPOOL_CONFIG_CHILDREN,
2895 &newchild, &newchildren) != 0)
2896 newchildren = 0;
2897
2898 for (c = 0; c < children; c++) {
2899 uint64_t is_log = B_FALSE, is_hole = B_FALSE;
2900 char *type;
2901 nvlist_t **mchild, *vdev;
2902 uint_t mchildren;
2903 int entry;
2904
2905 /*
2906 * Unlike cache & spares, slogs are stored in the
2907 * ZPOOL_CONFIG_CHILDREN array. We filter them out here.
2908 */
2909 (void) nvlist_lookup_uint64(child[c], ZPOOL_CONFIG_IS_LOG,
2910 &is_log);
2911 (void) nvlist_lookup_uint64(child[c], ZPOOL_CONFIG_IS_HOLE,
2912 &is_hole);
2913 if (is_log || is_hole) {
2914 /*
2915 * Create a hole vdev and put it in the config.
2916 */
2917 if (nvlist_alloc(&vdev, NV_UNIQUE_NAME, 0) != 0)
2918 goto out;
2919 if (nvlist_add_string(vdev, ZPOOL_CONFIG_TYPE,
2920 VDEV_TYPE_HOLE) != 0)
2921 goto out;
2922 if (nvlist_add_uint64(vdev, ZPOOL_CONFIG_IS_HOLE,
2923 1) != 0)
2924 goto out;
2925 if (lastlog == 0)
2926 lastlog = vcount;
2927 varray[vcount++] = vdev;
2928 continue;
2929 }
2930 lastlog = 0;
2931 verify(nvlist_lookup_string(child[c], ZPOOL_CONFIG_TYPE, &type)
2932 == 0);
2933 if (strcmp(type, VDEV_TYPE_MIRROR) != 0) {
2934 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2935 "Source pool must be composed only of mirrors\n"));
2936 retval = zfs_error(hdl, EZFS_INVALCONFIG, msg);
2937 goto out;
2938 }
2939
2940 verify(nvlist_lookup_nvlist_array(child[c],
2941 ZPOOL_CONFIG_CHILDREN, &mchild, &mchildren) == 0);
2942
2943 /* find or add an entry for this top-level vdev */
2944 if (newchildren > 0 &&
2945 (entry = find_vdev_entry(zhp, mchild, mchildren,
2946 newchild, newchildren)) >= 0) {
2947 /* We found a disk that the user specified. */
2948 vdev = mchild[entry];
2949 ++found;
2950 } else {
2951 /* User didn't specify a disk for this vdev. */
2952 vdev = mchild[mchildren - 1];
2953 }
2954
2955 if (nvlist_dup(vdev, &varray[vcount++], 0) != 0)
2956 goto out;
2957 }
2958
2959 /* did we find every disk the user specified? */
2960 if (found != newchildren) {
2961 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "Device list must "
2962 "include at most one disk from each mirror"));
2963 retval = zfs_error(hdl, EZFS_INVALCONFIG, msg);
2964 goto out;
2965 }
2966
2967 /* Prepare the nvlist for populating. */
2968 if (*newroot == NULL) {
2969 if (nvlist_alloc(newroot, NV_UNIQUE_NAME, 0) != 0)
2970 goto out;
2971 freelist = B_TRUE;
2972 if (nvlist_add_string(*newroot, ZPOOL_CONFIG_TYPE,
2973 VDEV_TYPE_ROOT) != 0)
2974 goto out;
2975 } else {
2976 verify(nvlist_remove_all(*newroot, ZPOOL_CONFIG_CHILDREN) == 0);
2977 }
2978
2979 /* Add all the children we found */
2980 if (nvlist_add_nvlist_array(*newroot, ZPOOL_CONFIG_CHILDREN, varray,
2981 lastlog == 0 ? vcount : lastlog) != 0)
2982 goto out;
2983
2984 /*
2985 * If we're just doing a dry run, exit now with success.
2986 */
2987 if (flags.dryrun) {
2988 memory_err = B_FALSE;
2989 freelist = B_FALSE;
2990 goto out;
2991 }
2992
2993 /* now build up the config list & call the ioctl */
2994 if (nvlist_alloc(&newconfig, NV_UNIQUE_NAME, 0) != 0)
2995 goto out;
2996
2997 if (nvlist_add_nvlist(newconfig,
2998 ZPOOL_CONFIG_VDEV_TREE, *newroot) != 0 ||
2999 nvlist_add_string(newconfig,
3000 ZPOOL_CONFIG_POOL_NAME, newname) != 0 ||
3001 nvlist_add_uint64(newconfig, ZPOOL_CONFIG_VERSION, vers) != 0)
3002 goto out;
3003
3004 /*
3005 * The new pool is automatically part of the namespace unless we
3006 * explicitly export it.
3007 */
3008 if (!flags.import)
3009 zc.zc_cookie = ZPOOL_EXPORT_AFTER_SPLIT;
3010 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
3011 (void) strlcpy(zc.zc_string, newname, sizeof (zc.zc_string));
3012 if (zcmd_write_conf_nvlist(hdl, &zc, newconfig) != 0)
3013 goto out;
3014 if (zc_props != NULL && zcmd_write_src_nvlist(hdl, &zc, zc_props) != 0)
3015 goto out;
3016
3017 if (zfs_ioctl(hdl, ZFS_IOC_VDEV_SPLIT, &zc) != 0) {
3018 retval = zpool_standard_error(hdl, errno, msg);
3019 goto out;
3020 }
3021
3022 freelist = B_FALSE;
3023 memory_err = B_FALSE;
3024
3025 out:
3026 if (varray != NULL) {
3027 int v;
3028
3029 for (v = 0; v < vcount; v++)
3030 nvlist_free(varray[v]);
3031 free(varray);
3032 }
3033 zcmd_free_nvlists(&zc);
3034 if (zc_props)
3035 nvlist_free(zc_props);
3036 if (newconfig)
3037 nvlist_free(newconfig);
3038 if (freelist) {
3039 nvlist_free(*newroot);
3040 *newroot = NULL;
3041 }
3042
3043 if (retval != 0)
3044 return (retval);
3045
3046 if (memory_err)
3047 return (no_memory(hdl));
3048
3049 return (0);
3050 }
3051
3052 /*
3053 * Remove the given device. Currently, this is supported only for hot spares
3054 * and level 2 cache devices.
3055 */
3056 int
3057 zpool_vdev_remove(zpool_handle_t *zhp, const char *path)
3058 {
3059 zfs_cmd_t zc = { 0 };
3060 char msg[1024];
3061 nvlist_t *tgt;
3062 boolean_t avail_spare, l2cache, islog;
3063 libzfs_handle_t *hdl = zhp->zpool_hdl;
3064 uint64_t version;
3065
3066 (void) snprintf(msg, sizeof (msg),
3067 dgettext(TEXT_DOMAIN, "cannot remove %s"), path);
3068
3069 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
3070 if ((tgt = zpool_find_vdev(zhp, path, &avail_spare, &l2cache,
3071 &islog)) == 0)
3072 return (zfs_error(hdl, EZFS_NODEVICE, msg));
3073 /*
3074 * XXX - this should just go away.
3075 */
3076 if (!avail_spare && !l2cache && !islog) {
3077 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3078 "only inactive hot spares, cache, top-level, "
3079 "or log devices can be removed"));
3080 return (zfs_error(hdl, EZFS_NODEVICE, msg));
3081 }
3082
3083 version = zpool_get_prop_int(zhp, ZPOOL_PROP_VERSION, NULL);
3084 if (islog && version < SPA_VERSION_HOLES) {
3085 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3086 "pool must be upgrade to support log removal"));
3087 return (zfs_error(hdl, EZFS_BADVERSION, msg));
3088 }
3089
3090 verify(nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID, &zc.zc_guid) == 0);
3091
3092 if (zfs_ioctl(hdl, ZFS_IOC_VDEV_REMOVE, &zc) == 0)
3093 return (0);
3094
3095 return (zpool_standard_error(hdl, errno, msg));
3096 }
3097
3098 /*
3099 * Clear the errors for the pool, or the particular device if specified.
3100 */
3101 int
3102 zpool_clear(zpool_handle_t *zhp, const char *path, nvlist_t *rewindnvl)
3103 {
3104 zfs_cmd_t zc = { 0 };
3105 char msg[1024];
3106 nvlist_t *tgt;
3107 zpool_rewind_policy_t policy;
3108 boolean_t avail_spare, l2cache;
3109 libzfs_handle_t *hdl = zhp->zpool_hdl;
3110 nvlist_t *nvi = NULL;
3111 int error;
3112
3113 if (path)
3114 (void) snprintf(msg, sizeof (msg),
3115 dgettext(TEXT_DOMAIN, "cannot clear errors for %s"),
3116 path);
3117 else
3118 (void) snprintf(msg, sizeof (msg),
3119 dgettext(TEXT_DOMAIN, "cannot clear errors for %s"),
3120 zhp->zpool_name);
3121
3122 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
3123 if (path) {
3124 if ((tgt = zpool_find_vdev(zhp, path, &avail_spare,
3125 &l2cache, NULL)) == 0)
3126 return (zfs_error(hdl, EZFS_NODEVICE, msg));
3127
3128 /*
3129 * Don't allow error clearing for hot spares. Do allow
3130 * error clearing for l2cache devices.
3131 */
3132 if (avail_spare)
3133 return (zfs_error(hdl, EZFS_ISSPARE, msg));
3134
3135 verify(nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID,
3136 &zc.zc_guid) == 0);
3137 }
3138
3139 zpool_get_rewind_policy(rewindnvl, &policy);
3140 zc.zc_cookie = policy.zrp_request;
3141
3142 if (zcmd_alloc_dst_nvlist(hdl, &zc, zhp->zpool_config_size * 2) != 0)
3143 return (-1);
3144
3145 if (zcmd_write_src_nvlist(hdl, &zc, rewindnvl) != 0)
3146 return (-1);
3147
3148 while ((error = zfs_ioctl(hdl, ZFS_IOC_CLEAR, &zc)) != 0 &&
3149 errno == ENOMEM) {
3150 if (zcmd_expand_dst_nvlist(hdl, &zc) != 0) {
3151 zcmd_free_nvlists(&zc);
3152 return (-1);
3153 }
3154 }
3155
3156 if (!error || ((policy.zrp_request & ZPOOL_TRY_REWIND) &&
3157 errno != EPERM && errno != EACCES)) {
3158 if (policy.zrp_request &
3159 (ZPOOL_DO_REWIND | ZPOOL_TRY_REWIND)) {
3160 (void) zcmd_read_dst_nvlist(hdl, &zc, &nvi);
3161 zpool_rewind_exclaim(hdl, zc.zc_name,
3162 ((policy.zrp_request & ZPOOL_TRY_REWIND) != 0),
3163 nvi);
3164 nvlist_free(nvi);
3165 }
3166 zcmd_free_nvlists(&zc);
3167 return (0);
3168 }
3169
3170 zcmd_free_nvlists(&zc);
3171 return (zpool_standard_error(hdl, errno, msg));
3172 }
3173
3174 /*
3175 * Similar to zpool_clear(), but takes a GUID (used by fmd).
3176 */
3177 int
3178 zpool_vdev_clear(zpool_handle_t *zhp, uint64_t guid)
3179 {
3180 zfs_cmd_t zc = { 0 };
3181 char msg[1024];
3182 libzfs_handle_t *hdl = zhp->zpool_hdl;
3183
3184 (void) snprintf(msg, sizeof (msg),
3185 dgettext(TEXT_DOMAIN, "cannot clear errors for %llx"),
3186 guid);
3187
3188 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
3189 zc.zc_guid = guid;
3190 zc.zc_cookie = ZPOOL_NO_REWIND;
3191
3192 if (ioctl(hdl->libzfs_fd, ZFS_IOC_CLEAR, &zc) == 0)
3193 return (0);
3194
3195 return (zpool_standard_error(hdl, errno, msg));
3196 }
3197
3198 /*
3199 * Change the GUID for a pool.
3200 */
3201 int
3202 zpool_reguid(zpool_handle_t *zhp)
3203 {
3204 char msg[1024];
3205 libzfs_handle_t *hdl = zhp->zpool_hdl;
3206 zfs_cmd_t zc = { 0 };
3207
3208 (void) snprintf(msg, sizeof (msg),
3209 dgettext(TEXT_DOMAIN, "cannot reguid '%s'"), zhp->zpool_name);
3210
3211 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
3212 if (zfs_ioctl(hdl, ZFS_IOC_POOL_REGUID, &zc) == 0)
3213 return (0);
3214
3215 return (zpool_standard_error(hdl, errno, msg));
3216 }
3217
3218 /*
3219 * Reopen the pool.
3220 */
3221 int
3222 zpool_reopen(zpool_handle_t *zhp)
3223 {
3224 zfs_cmd_t zc = { 0 };
3225 char msg[1024];
3226 libzfs_handle_t *hdl = zhp->zpool_hdl;
3227
3228 (void) snprintf(msg, sizeof (msg),
3229 dgettext(TEXT_DOMAIN, "cannot reopen '%s'"),
3230 zhp->zpool_name);
3231
3232 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
3233 if (zfs_ioctl(hdl, ZFS_IOC_POOL_REOPEN, &zc) == 0)
3234 return (0);
3235 return (zpool_standard_error(hdl, errno, msg));
3236 }
3237
3238 /*
3239 * Convert from a devid string to a path.
3240 */
3241 static char *
3242 devid_to_path(char *devid_str)
3243 {
3244 ddi_devid_t devid;
3245 char *minor;
3246 char *path;
3247 devid_nmlist_t *list = NULL;
3248 int ret;
3249
3250 if (devid_str_decode(devid_str, &devid, &minor) != 0)
3251 return (NULL);
3252
3253 ret = devid_deviceid_to_nmlist("/dev", devid, minor, &list);
3254
3255 devid_str_free(minor);
3256 devid_free(devid);
3257
3258 if (ret != 0)
3259 return (NULL);
3260
3261 if ((path = strdup(list[0].devname)) == NULL)
3262 return (NULL);
3263
3264 devid_free_nmlist(list);
3265
3266 return (path);
3267 }
3268
3269 /*
3270 * Convert from a path to a devid string.
3271 */
3272 static char *
3273 path_to_devid(const char *path)
3274 {
3275 int fd;
3276 ddi_devid_t devid;
3277 char *minor, *ret;
3278
3279 if ((fd = open(path, O_RDONLY)) < 0)
3280 return (NULL);
3281
3282 minor = NULL;
3283 ret = NULL;
3284 if (devid_get(fd, &devid) == 0) {
3285 if (devid_get_minor_name(fd, &minor) == 0)
3286 ret = devid_str_encode(devid, minor);
3287 if (minor != NULL)
3288 devid_str_free(minor);
3289 devid_free(devid);
3290 }
3291 (void) close(fd);
3292
3293 return (ret);
3294 }
3295
3296 /*
3297 * Issue the necessary ioctl() to update the stored path value for the vdev. We
3298 * ignore any failure here, since a common case is for an unprivileged user to
3299 * type 'zpool status', and we'll display the correct information anyway.
3300 */
3301 static void
3302 set_path(zpool_handle_t *zhp, nvlist_t *nv, const char *path)
3303 {
3304 zfs_cmd_t zc = { 0 };
3305
3306 (void) strncpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
3307 (void) strncpy(zc.zc_value, path, sizeof (zc.zc_value));
3308 verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID,
3309 &zc.zc_guid) == 0);
3310
3311 (void) ioctl(zhp->zpool_hdl->libzfs_fd, ZFS_IOC_VDEV_SETPATH, &zc);
3312 }
3313
3314 /*
3315 * Given a vdev, return the name to display in iostat. If the vdev has a path,
3316 * we use that, stripping off any leading "/dev/dsk/"; if not, we use the type.
3317 * We also check if this is a whole disk, in which case we strip off the
3318 * trailing 's0' slice name.
3319 *
3320 * This routine is also responsible for identifying when disks have been
3321 * reconfigured in a new location. The kernel will have opened the device by
3322 * devid, but the path will still refer to the old location. To catch this, we
3323 * first do a path -> devid translation (which is fast for the common case). If
3324 * the devid matches, we're done. If not, we do a reverse devid -> path
3325 * translation and issue the appropriate ioctl() to update the path of the vdev.
3326 * If 'zhp' is NULL, then this is an exported pool, and we don't need to do any
3327 * of these checks.
3328 */
3329 char *
3330 zpool_vdev_name(libzfs_handle_t *hdl, zpool_handle_t *zhp, nvlist_t *nv,
3331 boolean_t verbose)
3332 {
3333 char *path, *devid;
3334 uint64_t value;
3335 char buf[64];
3336 vdev_stat_t *vs;
3337 uint_t vsc;
3338
3339 if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_NOT_PRESENT,
3340 &value) == 0) {
3341 verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID,
3342 &value) == 0);
3343 (void) snprintf(buf, sizeof (buf), "%llu",
3344 (u_longlong_t)value);
3345 path = buf;
3346 } else if (nvlist_lookup_string(nv, ZPOOL_CONFIG_PATH, &path) == 0) {
3347
3348 /*
3349 * If the device is dead (faulted, offline, etc) then don't
3350 * bother opening it. Otherwise we may be forcing the user to
3351 * open a misbehaving device, which can have undesirable
3352 * effects.
3353 */
3354 if ((nvlist_lookup_uint64_array(nv, ZPOOL_CONFIG_VDEV_STATS,
3355 (uint64_t **)&vs, &vsc) != 0 ||
3356 vs->vs_state >= VDEV_STATE_DEGRADED) &&
3357 zhp != NULL &&
3358 nvlist_lookup_string(nv, ZPOOL_CONFIG_DEVID, &devid) == 0) {
3359 /*
3360 * Determine if the current path is correct.
3361 */
3362 char *newdevid = path_to_devid(path);
3363
3364 if (newdevid == NULL ||
3365 strcmp(devid, newdevid) != 0) {
3366 char *newpath;
3367
3368 if ((newpath = devid_to_path(devid)) != NULL) {
3369 /*
3370 * Update the path appropriately.
3371 */
3372 set_path(zhp, nv, newpath);
3373 if (nvlist_add_string(nv,
3374 ZPOOL_CONFIG_PATH, newpath) == 0)
3375 verify(nvlist_lookup_string(nv,
3376 ZPOOL_CONFIG_PATH,
3377 &path) == 0);
3378 free(newpath);
3379 }
3380 }
3381
3382 if (newdevid)
3383 devid_str_free(newdevid);
3384 }
3385
3386 if (strncmp(path, "/dev/dsk/", 9) == 0)
3387 path += 9;
3388
3389 if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_WHOLE_DISK,
3390 &value) == 0 && value) {
3391 int pathlen = strlen(path);
3392 char *tmp = zfs_strdup(hdl, path);
3393
3394 /*
3395 * If it starts with c#, and ends with "s0", chop
3396 * the "s0" off, or if it ends with "s0/old", remove
3397 * the "s0" from the middle.
3398 */
3399 if (CTD_CHECK(tmp)) {
3400 if (strcmp(&tmp[pathlen - 2], "s0") == 0) {
3401 tmp[pathlen - 2] = '\0';
3402 } else if (pathlen > 6 &&
3403 strcmp(&tmp[pathlen - 6], "s0/old") == 0) {
3404 (void) strcpy(&tmp[pathlen - 6],
3405 "/old");
3406 }
3407 }
3408 return (tmp);
3409 }
3410 } else {
3411 verify(nvlist_lookup_string(nv, ZPOOL_CONFIG_TYPE, &path) == 0);
3412
3413 /*
3414 * If it's a raidz device, we need to stick in the parity level.
3415 */
3416 if (strcmp(path, VDEV_TYPE_RAIDZ) == 0) {
3417 verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_NPARITY,
3418 &value) == 0);
3419 (void) snprintf(buf, sizeof (buf), "%s%llu", path,
3420 (u_longlong_t)value);
3421 path = buf;
3422 }
3423
3424 /*
3425 * We identify each top-level vdev by using a <type-id>
3426 * naming convention.
3427 */
3428 if (verbose) {
3429 uint64_t id;
3430
3431 verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_ID,
3432 &id) == 0);
3433 (void) snprintf(buf, sizeof (buf), "%s-%llu", path,
3434 (u_longlong_t)id);
3435 path = buf;
3436 }
3437 }
3438
3439 return (zfs_strdup(hdl, path));
3440 }
3441
3442 static int
3443 zbookmark_compare(const void *a, const void *b)
3444 {
3445 return (memcmp(a, b, sizeof (zbookmark_t)));
3446 }
3447
3448 /*
3449 * Retrieve the persistent error log, uniquify the members, and return to the
3450 * caller.
3451 */
3452 int
3453 zpool_get_errlog(zpool_handle_t *zhp, nvlist_t **nverrlistp)
3454 {
3455 zfs_cmd_t zc = { 0 };
3456 uint64_t count;
3457 zbookmark_t *zb = NULL;
3458 int i;
3459
3460 /*
3461 * Retrieve the raw error list from the kernel. If the number of errors
3462 * has increased, allocate more space and continue until we get the
3463 * entire list.
3464 */
3465 verify(nvlist_lookup_uint64(zhp->zpool_config, ZPOOL_CONFIG_ERRCOUNT,
3466 &count) == 0);
3467 if (count == 0)
3468 return (0);
3469 if ((zc.zc_nvlist_dst = (uintptr_t)zfs_alloc(zhp->zpool_hdl,
3470 count * sizeof (zbookmark_t))) == (uintptr_t)NULL)
3471 return (-1);
3472 zc.zc_nvlist_dst_size = count;
3473 (void) strcpy(zc.zc_name, zhp->zpool_name);
3474 for (;;) {
3475 if (ioctl(zhp->zpool_hdl->libzfs_fd, ZFS_IOC_ERROR_LOG,
3476 &zc) != 0) {
3477 free((void *)(uintptr_t)zc.zc_nvlist_dst);
3478 if (errno == ENOMEM) {
3479 count = zc.zc_nvlist_dst_size;
3480 if ((zc.zc_nvlist_dst = (uintptr_t)
3481 zfs_alloc(zhp->zpool_hdl, count *
3482 sizeof (zbookmark_t))) == (uintptr_t)NULL)
3483 return (-1);
3484 } else {
3485 return (-1);
3486 }
3487 } else {
3488 break;
3489 }
3490 }
3491
3492 /*
3493 * Sort the resulting bookmarks. This is a little confusing due to the
3494 * implementation of ZFS_IOC_ERROR_LOG. The bookmarks are copied last
3495 * to first, and 'zc_nvlist_dst_size' indicates the number of boomarks
3496 * _not_ copied as part of the process. So we point the start of our
3497 * array appropriate and decrement the total number of elements.
3498 */
3499 zb = ((zbookmark_t *)(uintptr_t)zc.zc_nvlist_dst) +
3500 zc.zc_nvlist_dst_size;
3501 count -= zc.zc_nvlist_dst_size;
3502
3503 qsort(zb, count, sizeof (zbookmark_t), zbookmark_compare);
3504
3505 verify(nvlist_alloc(nverrlistp, 0, KM_SLEEP) == 0);
3506
3507 /*
3508 * Fill in the nverrlistp with nvlist's of dataset and object numbers.
3509 */
3510 for (i = 0; i < count; i++) {
3511 nvlist_t *nv;
3512
3513 /* ignoring zb_blkid and zb_level for now */
3514 if (i > 0 && zb[i-1].zb_objset == zb[i].zb_objset &&
3515 zb[i-1].zb_object == zb[i].zb_object)
3516 continue;
3517
3518 if (nvlist_alloc(&nv, NV_UNIQUE_NAME, KM_SLEEP) != 0)
3519 goto nomem;
3520 if (nvlist_add_uint64(nv, ZPOOL_ERR_DATASET,
3521 zb[i].zb_objset) != 0) {
3522 nvlist_free(nv);
3523 goto nomem;
3524 }
3525 if (nvlist_add_uint64(nv, ZPOOL_ERR_OBJECT,
3526 zb[i].zb_object) != 0) {
3527 nvlist_free(nv);
3528 goto nomem;
3529 }
3530 if (nvlist_add_nvlist(*nverrlistp, "ejk", nv) != 0) {
3531 nvlist_free(nv);
3532 goto nomem;
3533 }
3534 nvlist_free(nv);
3535 }
3536
3537 free((void *)(uintptr_t)zc.zc_nvlist_dst);
3538 return (0);
3539
3540 nomem:
3541 free((void *)(uintptr_t)zc.zc_nvlist_dst);
3542 return (no_memory(zhp->zpool_hdl));
3543 }
3544
3545 /*
3546 * Upgrade a ZFS pool to the latest on-disk version.
3547 */
3548 int
3549 zpool_upgrade(zpool_handle_t *zhp, uint64_t new_version)
3550 {
3551 zfs_cmd_t zc = { 0 };
3552 libzfs_handle_t *hdl = zhp->zpool_hdl;
3553
3554 (void) strcpy(zc.zc_name, zhp->zpool_name);
3555 zc.zc_cookie = new_version;
3556
3557 if (zfs_ioctl(hdl, ZFS_IOC_POOL_UPGRADE, &zc) != 0)
3558 return (zpool_standard_error_fmt(hdl, errno,
3559 dgettext(TEXT_DOMAIN, "cannot upgrade '%s'"),
3560 zhp->zpool_name));
3561 return (0);
3562 }
3563
3564 void
3565 zfs_save_arguments(int argc, char **argv, char *string, int len)
3566 {
3567 (void) strlcpy(string, basename(argv[0]), len);
3568 for (int i = 1; i < argc; i++) {
3569 (void) strlcat(string, " ", len);
3570 (void) strlcat(string, argv[i], len);
3571 }
3572 }
3573
3574 int
3575 zpool_log_history(libzfs_handle_t *hdl, const char *message)
3576 {
3577 zfs_cmd_t zc = { 0 };
3578 nvlist_t *args;
3579 int err;
3580
3581 args = fnvlist_alloc();
3582 fnvlist_add_string(args, "message", message);
3583 err = zcmd_write_src_nvlist(hdl, &zc, args);
3584 if (err == 0)
3585 err = ioctl(hdl->libzfs_fd, ZFS_IOC_LOG_HISTORY, &zc);
3586 nvlist_free(args);
3587 zcmd_free_nvlists(&zc);
3588 return (err);
3589 }
3590
3591 /*
3592 * Perform ioctl to get some command history of a pool.
3593 *
3594 * 'buf' is the buffer to fill up to 'len' bytes. 'off' is the
3595 * logical offset of the history buffer to start reading from.
3596 *
3597 * Upon return, 'off' is the next logical offset to read from and
3598 * 'len' is the actual amount of bytes read into 'buf'.
3599 */
3600 static int
3601 get_history(zpool_handle_t *zhp, char *buf, uint64_t *off, uint64_t *len)
3602 {
3603 zfs_cmd_t zc = { 0 };
3604 libzfs_handle_t *hdl = zhp->zpool_hdl;
3605
3606 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
3607
3608 zc.zc_history = (uint64_t)(uintptr_t)buf;
3609 zc.zc_history_len = *len;
3610 zc.zc_history_offset = *off;
3611
3612 if (ioctl(hdl->libzfs_fd, ZFS_IOC_POOL_GET_HISTORY, &zc) != 0) {
3613 switch (errno) {
3614 case EPERM:
3615 return (zfs_error_fmt(hdl, EZFS_PERM,
3616 dgettext(TEXT_DOMAIN,
3617 "cannot show history for pool '%s'"),
3618 zhp->zpool_name));
3619 case ENOENT:
3620 return (zfs_error_fmt(hdl, EZFS_NOHISTORY,
3621 dgettext(TEXT_DOMAIN, "cannot get history for pool "
3622 "'%s'"), zhp->zpool_name));
3623 case ENOTSUP:
3624 return (zfs_error_fmt(hdl, EZFS_BADVERSION,
3625 dgettext(TEXT_DOMAIN, "cannot get history for pool "
3626 "'%s', pool must be upgraded"), zhp->zpool_name));
3627 default:
3628 return (zpool_standard_error_fmt(hdl, errno,
3629 dgettext(TEXT_DOMAIN,
3630 "cannot get history for '%s'"), zhp->zpool_name));
3631 }
3632 }
3633
3634 *len = zc.zc_history_len;
3635 *off = zc.zc_history_offset;
3636
3637 return (0);
3638 }
3639
3640 /*
3641 * Process the buffer of nvlists, unpacking and storing each nvlist record
3642 * into 'records'. 'leftover' is set to the number of bytes that weren't
3643 * processed as there wasn't a complete record.
3644 */
3645 int
3646 zpool_history_unpack(char *buf, uint64_t bytes_read, uint64_t *leftover,
3647 nvlist_t ***records, uint_t *numrecords)
3648 {
3649 uint64_t reclen;
3650 nvlist_t *nv;
3651 int i;
3652
3653 while (bytes_read > sizeof (reclen)) {
3654
3655 /* get length of packed record (stored as little endian) */
3656 for (i = 0, reclen = 0; i < sizeof (reclen); i++)
3657 reclen += (uint64_t)(((uchar_t *)buf)[i]) << (8*i);
3658
3659 if (bytes_read < sizeof (reclen) + reclen)
3660 break;
3661
3662 /* unpack record */
3663 if (nvlist_unpack(buf + sizeof (reclen), reclen, &nv, 0) != 0)
3664 return (ENOMEM);
3665 bytes_read -= sizeof (reclen) + reclen;
3666 buf += sizeof (reclen) + reclen;
3667
3668 /* add record to nvlist array */
3669 (*numrecords)++;
3670 if (ISP2(*numrecords + 1)) {
3671 *records = realloc(*records,
3672 *numrecords * 2 * sizeof (nvlist_t *));
3673 }
3674 (*records)[*numrecords - 1] = nv;
3675 }
3676
3677 *leftover = bytes_read;
3678 return (0);
3679 }
3680
3681 #define HIS_BUF_LEN (128*1024)
3682
3683 /*
3684 * Retrieve the command history of a pool.
3685 */
3686 int
3687 zpool_get_history(zpool_handle_t *zhp, nvlist_t **nvhisp)
3688 {
3689 char buf[HIS_BUF_LEN];
3690 uint64_t off = 0;
3691 nvlist_t **records = NULL;
3692 uint_t numrecords = 0;
3693 int err, i;
3694
3695 do {
3696 uint64_t bytes_read = sizeof (buf);
3697 uint64_t leftover;
3698
3699 if ((err = get_history(zhp, buf, &off, &bytes_read)) != 0)
3700 break;
3701
3702 /* if nothing else was read in, we're at EOF, just return */
3703 if (!bytes_read)
3704 break;
3705
3706 if ((err = zpool_history_unpack(buf, bytes_read,
3707 &leftover, &records, &numrecords)) != 0)
3708 break;
3709 off -= leftover;
3710
3711 /* CONSTCOND */
3712 } while (1);
3713
3714 if (!err) {
3715 verify(nvlist_alloc(nvhisp, NV_UNIQUE_NAME, 0) == 0);
3716 verify(nvlist_add_nvlist_array(*nvhisp, ZPOOL_HIST_RECORD,
3717 records, numrecords) == 0);
3718 }
3719 for (i = 0; i < numrecords; i++)
3720 nvlist_free(records[i]);
3721 free(records);
3722
3723 return (err);
3724 }
3725
3726 void
3727 zpool_obj_to_path(zpool_handle_t *zhp, uint64_t dsobj, uint64_t obj,
3728 char *pathname, size_t len)
3729 {
3730 zfs_cmd_t zc = { 0 };
3731 boolean_t mounted = B_FALSE;
3732 char *mntpnt = NULL;
3733 char dsname[MAXNAMELEN];
3734
3735 if (dsobj == 0) {
3736 /* special case for the MOS */
3737 (void) snprintf(pathname, len, "<metadata>:<0x%llx>", obj);
3738 return;
3739 }
3740
3741 /* get the dataset's name */
3742 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
3743 zc.zc_obj = dsobj;
3744 if (ioctl(zhp->zpool_hdl->libzfs_fd,
3745 ZFS_IOC_DSOBJ_TO_DSNAME, &zc) != 0) {
3746 /* just write out a path of two object numbers */
3747 (void) snprintf(pathname, len, "<0x%llx>:<0x%llx>",
3748 dsobj, obj);
3749 return;
3750 }
3751 (void) strlcpy(dsname, zc.zc_value, sizeof (dsname));
3752
3753 /* find out if the dataset is mounted */
3754 mounted = is_mounted(zhp->zpool_hdl, dsname, &mntpnt);
3755
3756 /* get the corrupted object's path */
3757 (void) strlcpy(zc.zc_name, dsname, sizeof (zc.zc_name));
3758 zc.zc_obj = obj;
3759 if (ioctl(zhp->zpool_hdl->libzfs_fd, ZFS_IOC_OBJ_TO_PATH,
3760 &zc) == 0) {
3761 if (mounted) {
3762 (void) snprintf(pathname, len, "%s%s", mntpnt,
3763 zc.zc_value);
3764 } else {
3765 (void) snprintf(pathname, len, "%s:%s",
3766 dsname, zc.zc_value);
3767 }
3768 } else {
3769 (void) snprintf(pathname, len, "%s:<0x%llx>", dsname, obj);
3770 }
3771 free(mntpnt);
3772 }
3773
3774 /*
3775 * Read the EFI label from the config, if a label does not exist then
3776 * pass back the error to the caller. If the caller has passed a non-NULL
3777 * diskaddr argument then we set it to the starting address of the EFI
3778 * partition.
3779 */
3780 static int
3781 read_efi_label(nvlist_t *config, diskaddr_t *sb)
3782 {
3783 char *path;
3784 int fd;
3785 char diskname[MAXPATHLEN];
3786 int err = -1;
3787
3788 if (nvlist_lookup_string(config, ZPOOL_CONFIG_PATH, &path) != 0)
3789 return (err);
3790
3791 (void) snprintf(diskname, sizeof (diskname), "%s%s", RDISK_ROOT,
3792 strrchr(path, '/'));
3793 if ((fd = open(diskname, O_RDONLY|O_NDELAY)) >= 0) {
3794 struct dk_gpt *vtoc;
3795
3796 if ((err = efi_alloc_and_read(fd, &vtoc)) >= 0) {
3797 if (sb != NULL)
3798 *sb = vtoc->efi_parts[0].p_start;
3799 efi_free(vtoc);
3800 }
3801 (void) close(fd);
3802 }
3803 return (err);
3804 }
3805
3806 /*
3807 * determine where a partition starts on a disk in the current
3808 * configuration
3809 */
3810 static diskaddr_t
3811 find_start_block(nvlist_t *config)
3812 {
3813 nvlist_t **child;
3814 uint_t c, children;
3815 diskaddr_t sb = MAXOFFSET_T;
3816 uint64_t wholedisk;
3817
3818 if (nvlist_lookup_nvlist_array(config,
3819 ZPOOL_CONFIG_CHILDREN, &child, &children) != 0) {
3820 if (nvlist_lookup_uint64(config,
3821 ZPOOL_CONFIG_WHOLE_DISK,
3822 &wholedisk) != 0 || !wholedisk) {
3823 return (MAXOFFSET_T);
3824 }
3825 if (read_efi_label(config, &sb) < 0)
3826 sb = MAXOFFSET_T;
3827 return (sb);
3828 }
3829
3830 for (c = 0; c < children; c++) {
3831 sb = find_start_block(child[c]);
3832 if (sb != MAXOFFSET_T) {
3833 return (sb);
3834 }
3835 }
3836 return (MAXOFFSET_T);
3837 }
3838
3839 /*
3840 * Label an individual disk. The name provided is the short name,
3841 * stripped of any leading /dev path.
3842 */
3843 int
3844 zpool_label_disk(libzfs_handle_t *hdl, zpool_handle_t *zhp, char *name)
3845 {
3846 char path[MAXPATHLEN];
3847 struct dk_gpt *vtoc;
3848 int fd;
3849 size_t resv = EFI_MIN_RESV_SIZE;
3850 uint64_t slice_size;
3851 diskaddr_t start_block;
3852 char errbuf[1024];
3853
3854 /* prepare an error message just in case */
3855 (void) snprintf(errbuf, sizeof (errbuf),
3856 dgettext(TEXT_DOMAIN, "cannot label '%s'"), name);
3857
3858 if (zhp) {
3859 nvlist_t *nvroot;
3860
3861 if (zpool_is_bootable(zhp)) {
3862 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3863 "EFI labeled devices are not supported on root "
3864 "pools."));
3865 return (zfs_error(hdl, EZFS_POOL_NOTSUP, errbuf));
3866 }
3867
3868 verify(nvlist_lookup_nvlist(zhp->zpool_config,
3869 ZPOOL_CONFIG_VDEV_TREE, &nvroot) == 0);
3870
3871 if (zhp->zpool_start_block == 0)
3872 start_block = find_start_block(nvroot);
3873 else
3874 start_block = zhp->zpool_start_block;
3875 zhp->zpool_start_block = start_block;
3876 } else {
3877 /* new pool */
3878 start_block = NEW_START_BLOCK;
3879 }
3880
3881 (void) snprintf(path, sizeof (path), "%s/%s%s", RDISK_ROOT, name,
3882 BACKUP_SLICE);
3883
3884 if ((fd = open(path, O_RDWR | O_NDELAY)) < 0) {
3885 /*
3886 * This shouldn't happen. We've long since verified that this
3887 * is a valid device.
3888 */
3889 zfs_error_aux(hdl,
3890 dgettext(TEXT_DOMAIN, "unable to open device"));
3891 return (zfs_error(hdl, EZFS_OPENFAILED, errbuf));
3892 }
3893
3894 if (efi_alloc_and_init(fd, EFI_NUMPAR, &vtoc) != 0) {
3895 /*
3896 * The only way this can fail is if we run out of memory, or we
3897 * were unable to read the disk's capacity
3898 */
3899 if (errno == ENOMEM)
3900 (void) no_memory(hdl);
3901
3902 (void) close(fd);
3903 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3904 "unable to read disk capacity"), name);
3905
3906 return (zfs_error(hdl, EZFS_NOCAP, errbuf));
3907 }
3908
3909 slice_size = vtoc->efi_last_u_lba + 1;
3910 slice_size -= EFI_MIN_RESV_SIZE;
3911 if (start_block == MAXOFFSET_T)
3912 start_block = NEW_START_BLOCK;
3913 slice_size -= start_block;
3914
3915 vtoc->efi_parts[0].p_start = start_block;
3916 vtoc->efi_parts[0].p_size = slice_size;
3917
3918 /*
3919 * Why we use V_USR: V_BACKUP confuses users, and is considered
3920 * disposable by some EFI utilities (since EFI doesn't have a backup
3921 * slice). V_UNASSIGNED is supposed to be used only for zero size
3922 * partitions, and efi_write() will fail if we use it. V_ROOT, V_BOOT,
3923 * etc. were all pretty specific. V_USR is as close to reality as we
3924 * can get, in the absence of V_OTHER.
3925 */
3926 vtoc->efi_parts[0].p_tag = V_USR;
3927 (void) strcpy(vtoc->efi_parts[0].p_name, "zfs");
3928
3929 vtoc->efi_parts[8].p_start = slice_size + start_block;
3930 vtoc->efi_parts[8].p_size = resv;
3931 vtoc->efi_parts[8].p_tag = V_RESERVED;
3932
3933 if (efi_write(fd, vtoc) != 0) {
3934 /*
3935 * Some block drivers (like pcata) may not support EFI
3936 * GPT labels. Print out a helpful error message dir-
3937 * ecting the user to manually label the disk and give
3938 * a specific slice.
3939 */
3940 (void) close(fd);
3941 efi_free(vtoc);
3942
3943 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3944 "try using fdisk(1M) and then provide a specific slice"));
3945 return (zfs_error(hdl, EZFS_LABELFAILED, errbuf));
3946 }
3947
3948 (void) close(fd);
3949 efi_free(vtoc);
3950 return (0);
3951 }
3952
3953 static boolean_t
3954 supported_dump_vdev_type(libzfs_handle_t *hdl, nvlist_t *config, char *errbuf)
3955 {
3956 char *type;
3957 nvlist_t **child;
3958 uint_t children, c;
3959
3960 verify(nvlist_lookup_string(config, ZPOOL_CONFIG_TYPE, &type) == 0);
3961 if (strcmp(type, VDEV_TYPE_FILE) == 0 ||
3962 strcmp(type, VDEV_TYPE_HOLE) == 0 ||
3963 strcmp(type, VDEV_TYPE_MISSING) == 0) {
3964 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3965 "vdev type '%s' is not supported"), type);
3966 (void) zfs_error(hdl, EZFS_VDEVNOTSUP, errbuf);
3967 return (B_FALSE);
3968 }
3969 if (nvlist_lookup_nvlist_array(config, ZPOOL_CONFIG_CHILDREN,
3970 &child, &children) == 0) {
3971 for (c = 0; c < children; c++) {
3972 if (!supported_dump_vdev_type(hdl, child[c], errbuf))
3973 return (B_FALSE);
3974 }
3975 }
3976 return (B_TRUE);
3977 }
3978
3979 /*
3980 * Check if this zvol is allowable for use as a dump device; zero if
3981 * it is, > 0 if it isn't, < 0 if it isn't a zvol.
3982 *
3983 * Allowable storage configurations include mirrors, all raidz variants, and
3984 * pools with log, cache, and spare devices. Pools which are backed by files or
3985 * have missing/hole vdevs are not suitable.
3986 */
3987 int
3988 zvol_check_dump_config(char *arg)
3989 {
3990 zpool_handle_t *zhp = NULL;
3991 nvlist_t *config, *nvroot;
3992 char *p, *volname;
3993 nvlist_t **top;
3994 uint_t toplevels;
3995 libzfs_handle_t *hdl;
3996 char errbuf[1024];
3997 char poolname[ZPOOL_MAXNAMELEN];
3998 int pathlen = strlen(ZVOL_FULL_DEV_DIR);
3999 int ret = 1;
4000
4001 if (strncmp(arg, ZVOL_FULL_DEV_DIR, pathlen)) {
4002 return (-1);
4003 }
4004
4005 (void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN,
4006 "dump is not supported on device '%s'"), arg);
4007
4008 if ((hdl = libzfs_init()) == NULL)
4009 return (1);
4010 libzfs_print_on_error(hdl, B_TRUE);
4011
4012 volname = arg + pathlen;
4013
4014 /* check the configuration of the pool */
4015 if ((p = strchr(volname, '/')) == NULL) {
4016 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
4017 "malformed dataset name"));
4018 (void) zfs_error(hdl, EZFS_INVALIDNAME, errbuf);
4019 return (1);
4020 } else if (p - volname >= ZFS_MAXNAMELEN) {
4021 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
4022 "dataset name is too long"));
4023 (void) zfs_error(hdl, EZFS_NAMETOOLONG, errbuf);
4024 return (1);
4025 } else {
4026 (void) strncpy(poolname, volname, p - volname);
4027 poolname[p - volname] = '\0';
4028 }
4029
4030 if ((zhp = zpool_open(hdl, poolname)) == NULL) {
4031 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
4032 "could not open pool '%s'"), poolname);
4033 (void) zfs_error(hdl, EZFS_OPENFAILED, errbuf);
4034 goto out;
4035 }
4036 config = zpool_get_config(zhp, NULL);
4037 if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE,
4038 &nvroot) != 0) {
4039 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
4040 "could not obtain vdev configuration for '%s'"), poolname);
4041 (void) zfs_error(hdl, EZFS_INVALCONFIG, errbuf);
4042 goto out;
4043 }
4044
4045 verify(nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_CHILDREN,
4046 &top, &toplevels) == 0);
4047
4048 if (!supported_dump_vdev_type(hdl, top[0], errbuf)) {
4049 goto out;
4050 }
4051 ret = 0;
4052
4053 out:
4054 if (zhp)
4055 zpool_close(zhp);
4056 libzfs_fini(hdl);
4057 return (ret);
4058 }