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) 2011 by Delphix. All rights reserved.
26 */
27
28 #include <ctype.h>
29 #include <errno.h>
30 #include <devid.h>
31 #include <fcntl.h>
32 #include <libintl.h>
33 #include <stdio.h>
34 #include <stdlib.h>
35 #include <strings.h>
36 #include <unistd.h>
37 #include <sys/efi_partition.h>
38 #include <sys/vtoc.h>
39 #include <sys/zfs_ioctl.h>
40 #include <dlfcn.h>
41
42 #include "zfs_namecheck.h"
43 #include "zfs_prop.h"
44 #include "libzfs_impl.h"
45 #include "zfs_comutil.h"
46
47 static int read_efi_label(nvlist_t *config, diskaddr_t *sb);
48
49 #define DISK_ROOT "/dev/dsk"
50 #define RDISK_ROOT "/dev/rdsk"
51 #define BACKUP_SLICE "s2"
52
53 typedef struct prop_flags {
54 int create:1; /* Validate property on creation */
55 int import:1; /* Validate property on import */
56 } prop_flags_t;
57
58 /*
59 * ====================================================================
60 * zpool property functions
61 * ====================================================================
62 */
63
64 static int
65 zpool_get_all_props(zpool_handle_t *zhp)
66 {
67 zfs_cmd_t zc = { 0 };
68 libzfs_handle_t *hdl = zhp->zpool_hdl;
69
70 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
71
72 if (zcmd_alloc_dst_nvlist(hdl, &zc, 0) != 0)
73 return (-1);
74
75 while (ioctl(hdl->libzfs_fd, ZFS_IOC_POOL_GET_PROPS, &zc) != 0) {
76 if (errno == ENOMEM) {
77 if (zcmd_expand_dst_nvlist(hdl, &zc) != 0) {
78 zcmd_free_nvlists(&zc);
79 return (-1);
80 }
81 } else {
82 zcmd_free_nvlists(&zc);
83 return (-1);
84 }
85 }
86
87 if (zcmd_read_dst_nvlist(hdl, &zc, &zhp->zpool_props) != 0) {
88 zcmd_free_nvlists(&zc);
89 return (-1);
90 }
91
92 zcmd_free_nvlists(&zc);
93
94 return (0);
95 }
96
97 static int
98 zpool_props_refresh(zpool_handle_t *zhp)
99 {
100 nvlist_t *old_props;
101
102 old_props = zhp->zpool_props;
103
104 if (zpool_get_all_props(zhp) != 0)
105 return (-1);
106
107 nvlist_free(old_props);
108 return (0);
109 }
110
111 static char *
112 zpool_get_prop_string(zpool_handle_t *zhp, zpool_prop_t prop,
113 zprop_source_t *src)
114 {
115 nvlist_t *nv, *nvl;
116 uint64_t ival;
117 char *value;
118 zprop_source_t source;
119
120 nvl = zhp->zpool_props;
121 if (nvlist_lookup_nvlist(nvl, zpool_prop_to_name(prop), &nv) == 0) {
122 verify(nvlist_lookup_uint64(nv, ZPROP_SOURCE, &ival) == 0);
123 source = ival;
124 verify(nvlist_lookup_string(nv, ZPROP_VALUE, &value) == 0);
125 } else {
126 source = ZPROP_SRC_DEFAULT;
127 if ((value = (char *)zpool_prop_default_string(prop)) == NULL)
128 value = "-";
129 }
130
131 if (src)
132 *src = source;
133
134 return (value);
135 }
136
137 uint64_t
138 zpool_get_prop_int(zpool_handle_t *zhp, zpool_prop_t prop, zprop_source_t *src)
139 {
140 nvlist_t *nv, *nvl;
141 uint64_t value;
142 zprop_source_t source;
143
144 if (zhp->zpool_props == NULL && zpool_get_all_props(zhp)) {
145 /*
146 * zpool_get_all_props() has most likely failed because
147 * the pool is faulted, but if all we need is the top level
148 * vdev's guid then get it from the zhp config nvlist.
149 */
150 if ((prop == ZPOOL_PROP_GUID) &&
151 (nvlist_lookup_nvlist(zhp->zpool_config,
152 ZPOOL_CONFIG_VDEV_TREE, &nv) == 0) &&
153 (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID, &value)
154 == 0)) {
155 return (value);
156 }
157 return (zpool_prop_default_numeric(prop));
158 }
159
160 nvl = zhp->zpool_props;
161 if (nvlist_lookup_nvlist(nvl, zpool_prop_to_name(prop), &nv) == 0) {
162 verify(nvlist_lookup_uint64(nv, ZPROP_SOURCE, &value) == 0);
163 source = value;
164 verify(nvlist_lookup_uint64(nv, ZPROP_VALUE, &value) == 0);
165 } else {
166 source = ZPROP_SRC_DEFAULT;
167 value = zpool_prop_default_numeric(prop);
168 }
169
170 if (src)
171 *src = source;
172
173 return (value);
174 }
175
176 /*
177 * Map VDEV STATE to printed strings.
178 */
179 char *
180 zpool_state_to_name(vdev_state_t state, vdev_aux_t aux)
181 {
182 switch (state) {
183 case VDEV_STATE_CLOSED:
184 case VDEV_STATE_OFFLINE:
185 return (gettext("OFFLINE"));
186 case VDEV_STATE_REMOVED:
187 return (gettext("REMOVED"));
188 case VDEV_STATE_CANT_OPEN:
189 if (aux == VDEV_AUX_CORRUPT_DATA || aux == VDEV_AUX_BAD_LOG)
190 return (gettext("FAULTED"));
191 else if (aux == VDEV_AUX_SPLIT_POOL)
192 return (gettext("SPLIT"));
193 else
194 return (gettext("UNAVAIL"));
195 case VDEV_STATE_FAULTED:
196 return (gettext("FAULTED"));
197 case VDEV_STATE_DEGRADED:
198 return (gettext("DEGRADED"));
199 case VDEV_STATE_HEALTHY:
200 return (gettext("ONLINE"));
201 }
202
203 return (gettext("UNKNOWN"));
204 }
205
206 /*
207 * Get a zpool property value for 'prop' and return the value in
208 * a pre-allocated buffer.
209 */
210 int
211 zpool_get_prop(zpool_handle_t *zhp, zpool_prop_t prop, char *buf, size_t len,
212 zprop_source_t *srctype)
213 {
214 uint64_t intval;
215 const char *strval;
216 zprop_source_t src = ZPROP_SRC_NONE;
217 nvlist_t *nvroot;
218 vdev_stat_t *vs;
219 uint_t vsc;
220
221 if (zpool_get_state(zhp) == POOL_STATE_UNAVAIL) {
222 switch (prop) {
223 case ZPOOL_PROP_NAME:
224 (void) strlcpy(buf, zpool_get_name(zhp), len);
225 break;
226
227 case ZPOOL_PROP_HEALTH:
228 (void) strlcpy(buf, "FAULTED", len);
229 break;
230
231 case ZPOOL_PROP_GUID:
232 intval = zpool_get_prop_int(zhp, prop, &src);
233 (void) snprintf(buf, len, "%llu", intval);
234 break;
235
236 case ZPOOL_PROP_ALTROOT:
237 case ZPOOL_PROP_CACHEFILE:
238 if (zhp->zpool_props != NULL ||
239 zpool_get_all_props(zhp) == 0) {
240 (void) strlcpy(buf,
241 zpool_get_prop_string(zhp, prop, &src),
242 len);
243 if (srctype != NULL)
244 *srctype = src;
245 return (0);
246 }
247 /* FALLTHROUGH */
248 default:
249 (void) strlcpy(buf, "-", len);
250 break;
251 }
252
253 if (srctype != NULL)
254 *srctype = src;
255 return (0);
256 }
257
258 if (zhp->zpool_props == NULL && zpool_get_all_props(zhp) &&
259 prop != ZPOOL_PROP_NAME)
260 return (-1);
261
262 switch (zpool_prop_get_type(prop)) {
263 case PROP_TYPE_STRING:
264 (void) strlcpy(buf, zpool_get_prop_string(zhp, prop, &src),
265 len);
266 break;
267
268 case PROP_TYPE_NUMBER:
269 intval = zpool_get_prop_int(zhp, prop, &src);
270
271 switch (prop) {
272 case ZPOOL_PROP_SIZE:
273 case ZPOOL_PROP_ALLOCATED:
274 case ZPOOL_PROP_FREE:
275 (void) zfs_nicenum(intval, buf, len);
276 break;
277
278 case ZPOOL_PROP_CAPACITY:
279 (void) snprintf(buf, len, "%llu%%",
280 (u_longlong_t)intval);
281 break;
282
283 case ZPOOL_PROP_DEDUPRATIO:
284 (void) snprintf(buf, len, "%llu.%02llux",
285 (u_longlong_t)(intval / 100),
286 (u_longlong_t)(intval % 100));
287 break;
288
289 case ZPOOL_PROP_HEALTH:
290 verify(nvlist_lookup_nvlist(zpool_get_config(zhp, NULL),
291 ZPOOL_CONFIG_VDEV_TREE, &nvroot) == 0);
292 verify(nvlist_lookup_uint64_array(nvroot,
293 ZPOOL_CONFIG_VDEV_STATS, (uint64_t **)&vs, &vsc)
294 == 0);
295
296 (void) strlcpy(buf, zpool_state_to_name(intval,
297 vs->vs_aux), len);
298 break;
299 default:
300 (void) snprintf(buf, len, "%llu", intval);
301 }
302 break;
303
304 case PROP_TYPE_INDEX:
305 intval = zpool_get_prop_int(zhp, prop, &src);
306 if (zpool_prop_index_to_string(prop, intval, &strval)
307 != 0)
308 return (-1);
309 (void) strlcpy(buf, strval, len);
310 break;
311
312 default:
313 abort();
314 }
315
316 if (srctype)
317 *srctype = src;
318
319 return (0);
320 }
321
322 /*
323 * Check if the bootfs name has the same pool name as it is set to.
324 * Assuming bootfs is a valid dataset name.
325 */
326 static boolean_t
327 bootfs_name_valid(const char *pool, char *bootfs)
328 {
329 int len = strlen(pool);
330
331 if (!zfs_name_valid(bootfs, ZFS_TYPE_FILESYSTEM|ZFS_TYPE_SNAPSHOT))
332 return (B_FALSE);
333
334 if (strncmp(pool, bootfs, len) == 0 &&
335 (bootfs[len] == '/' || bootfs[len] == '\0'))
336 return (B_TRUE);
337
338 return (B_FALSE);
339 }
340
341 /*
342 * Inspect the configuration to determine if any of the devices contain
343 * an EFI label.
344 */
345 static boolean_t
346 pool_uses_efi(nvlist_t *config)
347 {
348 nvlist_t **child;
349 uint_t c, children;
350
351 if (nvlist_lookup_nvlist_array(config, ZPOOL_CONFIG_CHILDREN,
352 &child, &children) != 0)
353 return (read_efi_label(config, NULL) >= 0);
354
355 for (c = 0; c < children; c++) {
356 if (pool_uses_efi(child[c]))
357 return (B_TRUE);
358 }
359 return (B_FALSE);
360 }
361
362 static boolean_t
363 pool_is_bootable(zpool_handle_t *zhp)
364 {
365 char bootfs[ZPOOL_MAXNAMELEN];
366
367 return (zpool_get_prop(zhp, ZPOOL_PROP_BOOTFS, bootfs,
368 sizeof (bootfs), NULL) == 0 && strncmp(bootfs, "-",
369 sizeof (bootfs)) != 0);
370 }
371
372
373 /*
374 * Given an nvlist of zpool properties to be set, validate that they are
375 * correct, and parse any numeric properties (index, boolean, etc) if they are
376 * specified as strings.
377 */
378 static nvlist_t *
379 zpool_valid_proplist(libzfs_handle_t *hdl, const char *poolname,
380 nvlist_t *props, uint64_t version, prop_flags_t flags, char *errbuf)
381 {
382 nvpair_t *elem;
383 nvlist_t *retprops;
384 zpool_prop_t prop;
385 char *strval;
386 uint64_t intval;
387 char *slash;
388 struct stat64 statbuf;
389 zpool_handle_t *zhp;
390 nvlist_t *nvroot;
391
392 if (nvlist_alloc(&retprops, NV_UNIQUE_NAME, 0) != 0) {
393 (void) no_memory(hdl);
394 return (NULL);
395 }
396
397 elem = NULL;
398 while ((elem = nvlist_next_nvpair(props, elem)) != NULL) {
399 const char *propname = nvpair_name(elem);
400
401 /*
402 * Make sure this property is valid and applies to this type.
403 */
404 if ((prop = zpool_name_to_prop(propname)) == ZPROP_INVAL) {
405 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
406 "invalid property '%s'"), propname);
407 (void) zfs_error(hdl, EZFS_BADPROP, errbuf);
408 goto error;
409 }
410
411 if (zpool_prop_readonly(prop)) {
412 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "'%s' "
413 "is readonly"), propname);
414 (void) zfs_error(hdl, EZFS_PROPREADONLY, errbuf);
415 goto error;
416 }
417
418 if (zprop_parse_value(hdl, elem, prop, ZFS_TYPE_POOL, retprops,
419 &strval, &intval, errbuf) != 0)
420 goto error;
421
422 /*
423 * Perform additional checking for specific properties.
424 */
425 switch (prop) {
426 case ZPOOL_PROP_VERSION:
427 if (intval < version || intval > SPA_VERSION) {
428 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
429 "property '%s' number %d is invalid."),
430 propname, intval);
431 (void) zfs_error(hdl, EZFS_BADVERSION, errbuf);
432 goto error;
433 }
434 break;
435
436 case ZPOOL_PROP_BOOTFS:
437 if (flags.create || flags.import) {
438 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
439 "property '%s' cannot be set at creation "
440 "or import time"), propname);
441 (void) zfs_error(hdl, EZFS_BADPROP, errbuf);
442 goto error;
443 }
444
445 if (version < SPA_VERSION_BOOTFS) {
446 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
447 "pool must be upgraded to support "
448 "'%s' property"), propname);
449 (void) zfs_error(hdl, EZFS_BADVERSION, errbuf);
450 goto error;
451 }
452
453 /*
454 * bootfs property value has to be a dataset name and
455 * the dataset has to be in the same pool as it sets to.
456 */
457 if (strval[0] != '\0' && !bootfs_name_valid(poolname,
458 strval)) {
459 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "'%s' "
460 "is an invalid name"), strval);
461 (void) zfs_error(hdl, EZFS_INVALIDNAME, errbuf);
462 goto error;
463 }
464
465 if ((zhp = zpool_open_canfail(hdl, poolname)) == NULL) {
466 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
467 "could not open pool '%s'"), poolname);
468 (void) zfs_error(hdl, EZFS_OPENFAILED, errbuf);
469 goto error;
470 }
471 verify(nvlist_lookup_nvlist(zpool_get_config(zhp, NULL),
472 ZPOOL_CONFIG_VDEV_TREE, &nvroot) == 0);
473
474 /*
475 * bootfs property cannot be set on a disk which has
476 * been EFI labeled.
477 */
478 if (pool_uses_efi(nvroot)) {
479 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
480 "property '%s' not supported on "
481 "EFI labeled devices"), propname);
482 (void) zfs_error(hdl, EZFS_POOL_NOTSUP, errbuf);
483 zpool_close(zhp);
484 goto error;
485 }
486 zpool_close(zhp);
487 break;
488
489 case ZPOOL_PROP_ALTROOT:
490 if (!flags.create && !flags.import) {
491 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
492 "property '%s' can only be set during pool "
493 "creation or import"), propname);
494 (void) zfs_error(hdl, EZFS_BADPROP, errbuf);
495 goto error;
496 }
497
498 if (strval[0] != '/') {
499 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
500 "bad alternate root '%s'"), strval);
501 (void) zfs_error(hdl, EZFS_BADPATH, errbuf);
502 goto error;
503 }
504 break;
505
506 case ZPOOL_PROP_CACHEFILE:
507 if (strval[0] == '\0')
508 break;
509
510 if (strcmp(strval, "none") == 0)
511 break;
512
513 if (strval[0] != '/') {
514 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
515 "property '%s' must be empty, an "
516 "absolute path, or 'none'"), propname);
517 (void) zfs_error(hdl, EZFS_BADPATH, errbuf);
518 goto error;
519 }
520
521 slash = strrchr(strval, '/');
522
523 if (slash[1] == '\0' || strcmp(slash, "/.") == 0 ||
524 strcmp(slash, "/..") == 0) {
525 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
526 "'%s' is not a valid file"), strval);
527 (void) zfs_error(hdl, EZFS_BADPATH, errbuf);
528 goto error;
529 }
530
531 *slash = '\0';
532
533 if (strval[0] != '\0' &&
534 (stat64(strval, &statbuf) != 0 ||
535 !S_ISDIR(statbuf.st_mode))) {
536 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
537 "'%s' is not a valid directory"),
538 strval);
539 (void) zfs_error(hdl, EZFS_BADPATH, errbuf);
540 goto error;
541 }
542
543 *slash = '/';
544 break;
545
546 case ZPOOL_PROP_READONLY:
547 if (!flags.import) {
548 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
549 "property '%s' can only be set at "
550 "import time"), propname);
551 (void) zfs_error(hdl, EZFS_BADPROP, errbuf);
552 goto error;
553 }
554 break;
555 }
556 }
557
558 return (retprops);
559 error:
560 nvlist_free(retprops);
561 return (NULL);
562 }
563
564 /*
565 * Set zpool property : propname=propval.
566 */
567 int
568 zpool_set_prop(zpool_handle_t *zhp, const char *propname, const char *propval)
569 {
570 zfs_cmd_t zc = { 0 };
571 int ret = -1;
572 char errbuf[1024];
573 nvlist_t *nvl = NULL;
574 nvlist_t *realprops;
575 uint64_t version;
576 prop_flags_t flags = { 0 };
577
578 (void) snprintf(errbuf, sizeof (errbuf),
579 dgettext(TEXT_DOMAIN, "cannot set property for '%s'"),
580 zhp->zpool_name);
581
582 if (nvlist_alloc(&nvl, NV_UNIQUE_NAME, 0) != 0)
583 return (no_memory(zhp->zpool_hdl));
584
585 if (nvlist_add_string(nvl, propname, propval) != 0) {
586 nvlist_free(nvl);
587 return (no_memory(zhp->zpool_hdl));
588 }
589
590 version = zpool_get_prop_int(zhp, ZPOOL_PROP_VERSION, NULL);
591 if ((realprops = zpool_valid_proplist(zhp->zpool_hdl,
592 zhp->zpool_name, nvl, version, flags, errbuf)) == NULL) {
593 nvlist_free(nvl);
594 return (-1);
595 }
596
597 nvlist_free(nvl);
598 nvl = realprops;
599
600 /*
601 * Execute the corresponding ioctl() to set this property.
602 */
603 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
604
605 if (zcmd_write_src_nvlist(zhp->zpool_hdl, &zc, nvl) != 0) {
606 nvlist_free(nvl);
607 return (-1);
608 }
609
610 ret = zfs_ioctl(zhp->zpool_hdl, ZFS_IOC_POOL_SET_PROPS, &zc);
611
612 zcmd_free_nvlists(&zc);
613 nvlist_free(nvl);
614
615 if (ret)
616 (void) zpool_standard_error(zhp->zpool_hdl, errno, errbuf);
617 else
618 (void) zpool_props_refresh(zhp);
619
620 return (ret);
621 }
622
623 int
624 zpool_expand_proplist(zpool_handle_t *zhp, zprop_list_t **plp)
625 {
626 libzfs_handle_t *hdl = zhp->zpool_hdl;
627 zprop_list_t *entry;
628 char buf[ZFS_MAXPROPLEN];
629
630 if (zprop_expand_list(hdl, plp, ZFS_TYPE_POOL) != 0)
631 return (-1);
632
633 for (entry = *plp; entry != NULL; entry = entry->pl_next) {
634
635 if (entry->pl_fixed)
636 continue;
637
638 if (entry->pl_prop != ZPROP_INVAL &&
639 zpool_get_prop(zhp, entry->pl_prop, buf, sizeof (buf),
640 NULL) == 0) {
641 if (strlen(buf) > entry->pl_width)
642 entry->pl_width = strlen(buf);
643 }
644 }
645
646 return (0);
647 }
648
649
650 /*
651 * Don't start the slice at the default block of 34; many storage
652 * devices will use a stripe width of 128k, so start there instead.
653 */
654 #define NEW_START_BLOCK 256
655
656 /*
657 * Validate the given pool name, optionally putting an extended error message in
658 * 'buf'.
659 */
660 boolean_t
661 zpool_name_valid(libzfs_handle_t *hdl, boolean_t isopen, const char *pool)
662 {
663 namecheck_err_t why;
664 char what;
665 int ret;
666
667 ret = pool_namecheck(pool, &why, &what);
668
669 /*
670 * The rules for reserved pool names were extended at a later point.
671 * But we need to support users with existing pools that may now be
672 * invalid. So we only check for this expanded set of names during a
673 * create (or import), and only in userland.
674 */
675 if (ret == 0 && !isopen &&
676 (strncmp(pool, "mirror", 6) == 0 ||
677 strncmp(pool, "raidz", 5) == 0 ||
678 strncmp(pool, "spare", 5) == 0 ||
679 strcmp(pool, "log") == 0)) {
680 if (hdl != NULL)
681 zfs_error_aux(hdl,
682 dgettext(TEXT_DOMAIN, "name is reserved"));
683 return (B_FALSE);
684 }
685
686
687 if (ret != 0) {
688 if (hdl != NULL) {
689 switch (why) {
690 case NAME_ERR_TOOLONG:
691 zfs_error_aux(hdl,
692 dgettext(TEXT_DOMAIN, "name is too long"));
693 break;
694
695 case NAME_ERR_INVALCHAR:
696 zfs_error_aux(hdl,
697 dgettext(TEXT_DOMAIN, "invalid character "
698 "'%c' in pool name"), what);
699 break;
700
701 case NAME_ERR_NOLETTER:
702 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
703 "name must begin with a letter"));
704 break;
705
706 case NAME_ERR_RESERVED:
707 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
708 "name is reserved"));
709 break;
710
711 case NAME_ERR_DISKLIKE:
712 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
713 "pool name is reserved"));
714 break;
715
716 case NAME_ERR_LEADING_SLASH:
717 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
718 "leading slash in name"));
719 break;
720
721 case NAME_ERR_EMPTY_COMPONENT:
722 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
723 "empty component in name"));
724 break;
725
726 case NAME_ERR_TRAILING_SLASH:
727 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
728 "trailing slash in name"));
729 break;
730
731 case NAME_ERR_MULTIPLE_AT:
732 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
733 "multiple '@' delimiters in name"));
734 break;
735
736 }
737 }
738 return (B_FALSE);
739 }
740
741 return (B_TRUE);
742 }
743
744 /*
745 * Open a handle to the given pool, even if the pool is currently in the FAULTED
746 * state.
747 */
748 zpool_handle_t *
749 zpool_open_canfail(libzfs_handle_t *hdl, const char *pool)
750 {
751 zpool_handle_t *zhp;
752 boolean_t missing;
753
754 /*
755 * Make sure the pool name is valid.
756 */
757 if (!zpool_name_valid(hdl, B_TRUE, pool)) {
758 (void) zfs_error_fmt(hdl, EZFS_INVALIDNAME,
759 dgettext(TEXT_DOMAIN, "cannot open '%s'"),
760 pool);
761 return (NULL);
762 }
763
764 if ((zhp = zfs_alloc(hdl, sizeof (zpool_handle_t))) == NULL)
765 return (NULL);
766
767 zhp->zpool_hdl = hdl;
768 (void) strlcpy(zhp->zpool_name, pool, sizeof (zhp->zpool_name));
769
770 if (zpool_refresh_stats(zhp, &missing) != 0) {
771 zpool_close(zhp);
772 return (NULL);
773 }
774
775 if (missing) {
776 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "no such pool"));
777 (void) zfs_error_fmt(hdl, EZFS_NOENT,
778 dgettext(TEXT_DOMAIN, "cannot open '%s'"), pool);
779 zpool_close(zhp);
780 return (NULL);
781 }
782
783 return (zhp);
784 }
785
786 /*
787 * Like the above, but silent on error. Used when iterating over pools (because
788 * the configuration cache may be out of date).
789 */
790 int
791 zpool_open_silent(libzfs_handle_t *hdl, const char *pool, zpool_handle_t **ret)
792 {
793 zpool_handle_t *zhp;
794 boolean_t missing;
795
796 if ((zhp = zfs_alloc(hdl, sizeof (zpool_handle_t))) == NULL)
797 return (-1);
798
799 zhp->zpool_hdl = hdl;
800 (void) strlcpy(zhp->zpool_name, pool, sizeof (zhp->zpool_name));
801
802 if (zpool_refresh_stats(zhp, &missing) != 0) {
803 zpool_close(zhp);
804 return (-1);
805 }
806
807 if (missing) {
808 zpool_close(zhp);
809 *ret = NULL;
810 return (0);
811 }
812
813 *ret = zhp;
814 return (0);
815 }
816
817 /*
818 * Similar to zpool_open_canfail(), but refuses to open pools in the faulted
819 * state.
820 */
821 zpool_handle_t *
822 zpool_open(libzfs_handle_t *hdl, const char *pool)
823 {
824 zpool_handle_t *zhp;
825
826 if ((zhp = zpool_open_canfail(hdl, pool)) == NULL)
827 return (NULL);
828
829 if (zhp->zpool_state == POOL_STATE_UNAVAIL) {
830 (void) zfs_error_fmt(hdl, EZFS_POOLUNAVAIL,
831 dgettext(TEXT_DOMAIN, "cannot open '%s'"), zhp->zpool_name);
832 zpool_close(zhp);
833 return (NULL);
834 }
835
836 return (zhp);
837 }
838
839 /*
840 * Close the handle. Simply frees the memory associated with the handle.
841 */
842 void
843 zpool_close(zpool_handle_t *zhp)
844 {
845 if (zhp->zpool_config)
846 nvlist_free(zhp->zpool_config);
847 if (zhp->zpool_old_config)
848 nvlist_free(zhp->zpool_old_config);
849 if (zhp->zpool_props)
850 nvlist_free(zhp->zpool_props);
851 free(zhp);
852 }
853
854 /*
855 * Return the name of the pool.
856 */
857 const char *
858 zpool_get_name(zpool_handle_t *zhp)
859 {
860 return (zhp->zpool_name);
861 }
862
863
864 /*
865 * Return the state of the pool (ACTIVE or UNAVAILABLE)
866 */
867 int
868 zpool_get_state(zpool_handle_t *zhp)
869 {
870 return (zhp->zpool_state);
871 }
872
873 /*
874 * Create the named pool, using the provided vdev list. It is assumed
875 * that the consumer has already validated the contents of the nvlist, so we
876 * don't have to worry about error semantics.
877 */
878 int
879 zpool_create(libzfs_handle_t *hdl, const char *pool, nvlist_t *nvroot,
880 nvlist_t *props, nvlist_t *fsprops)
881 {
882 zfs_cmd_t zc = { 0 };
883 nvlist_t *zc_fsprops = NULL;
884 nvlist_t *zc_props = NULL;
885 char msg[1024];
886 char *altroot;
887 int ret = -1;
888
889 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN,
890 "cannot create '%s'"), pool);
891
892 if (!zpool_name_valid(hdl, B_FALSE, pool))
893 return (zfs_error(hdl, EZFS_INVALIDNAME, msg));
894
895 if (zcmd_write_conf_nvlist(hdl, &zc, nvroot) != 0)
896 return (-1);
897
898 if (props) {
899 prop_flags_t flags = { .create = B_TRUE, .import = B_FALSE };
900
901 if ((zc_props = zpool_valid_proplist(hdl, pool, props,
902 SPA_VERSION_1, flags, msg)) == NULL) {
903 goto create_failed;
904 }
905 }
906
907 if (fsprops) {
908 uint64_t zoned;
909 char *zonestr;
910
911 zoned = ((nvlist_lookup_string(fsprops,
912 zfs_prop_to_name(ZFS_PROP_ZONED), &zonestr) == 0) &&
913 strcmp(zonestr, "on") == 0);
914
915 if ((zc_fsprops = zfs_valid_proplist(hdl,
916 ZFS_TYPE_FILESYSTEM, fsprops, zoned, NULL, msg)) == NULL) {
917 goto create_failed;
918 }
919 if (!zc_props &&
920 (nvlist_alloc(&zc_props, NV_UNIQUE_NAME, 0) != 0)) {
921 goto create_failed;
922 }
923 if (nvlist_add_nvlist(zc_props,
924 ZPOOL_ROOTFS_PROPS, zc_fsprops) != 0) {
925 goto create_failed;
926 }
927 }
928
929 if (zc_props && zcmd_write_src_nvlist(hdl, &zc, zc_props) != 0)
930 goto create_failed;
931
932 (void) strlcpy(zc.zc_name, pool, sizeof (zc.zc_name));
933
934 if ((ret = zfs_ioctl(hdl, ZFS_IOC_POOL_CREATE, &zc)) != 0) {
935
936 zcmd_free_nvlists(&zc);
937 nvlist_free(zc_props);
938 nvlist_free(zc_fsprops);
939
940 switch (errno) {
941 case EBUSY:
942 /*
943 * This can happen if the user has specified the same
944 * device multiple times. We can't reliably detect this
945 * until we try to add it and see we already have a
946 * label.
947 */
948 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
949 "one or more vdevs refer to the same device"));
950 return (zfs_error(hdl, EZFS_BADDEV, msg));
951
952 case EOVERFLOW:
953 /*
954 * This occurs when one of the devices is below
955 * SPA_MINDEVSIZE. Unfortunately, we can't detect which
956 * device was the problem device since there's no
957 * reliable way to determine device size from userland.
958 */
959 {
960 char buf[64];
961
962 zfs_nicenum(SPA_MINDEVSIZE, buf, sizeof (buf));
963
964 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
965 "one or more devices is less than the "
966 "minimum size (%s)"), buf);
967 }
968 return (zfs_error(hdl, EZFS_BADDEV, msg));
969
970 case ENOSPC:
971 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
972 "one or more devices is out of space"));
973 return (zfs_error(hdl, EZFS_BADDEV, msg));
974
975 case ENOTBLK:
976 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
977 "cache device must be a disk or disk slice"));
978 return (zfs_error(hdl, EZFS_BADDEV, msg));
979
980 default:
981 return (zpool_standard_error(hdl, errno, msg));
982 }
983 }
984
985 /*
986 * If this is an alternate root pool, then we automatically set the
987 * mountpoint of the root dataset to be '/'.
988 */
989 if (nvlist_lookup_string(props, zpool_prop_to_name(ZPOOL_PROP_ALTROOT),
990 &altroot) == 0) {
991 zfs_handle_t *zhp;
992
993 verify((zhp = zfs_open(hdl, pool, ZFS_TYPE_DATASET)) != NULL);
994 verify(zfs_prop_set(zhp, zfs_prop_to_name(ZFS_PROP_MOUNTPOINT),
995 "/") == 0);
996
997 zfs_close(zhp);
998 }
999
1000 create_failed:
1001 zcmd_free_nvlists(&zc);
1002 nvlist_free(zc_props);
1003 nvlist_free(zc_fsprops);
1004 return (ret);
1005 }
1006
1007 /*
1008 * Destroy the given pool. It is up to the caller to ensure that there are no
1009 * datasets left in the pool.
1010 */
1011 int
1012 zpool_destroy(zpool_handle_t *zhp)
1013 {
1014 zfs_cmd_t zc = { 0 };
1015 zfs_handle_t *zfp = NULL;
1016 libzfs_handle_t *hdl = zhp->zpool_hdl;
1017 char msg[1024];
1018
1019 if (zhp->zpool_state == POOL_STATE_ACTIVE &&
1020 (zfp = zfs_open(hdl, zhp->zpool_name, ZFS_TYPE_FILESYSTEM)) == NULL)
1021 return (-1);
1022
1023 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
1024
1025 if (zfs_ioctl(hdl, ZFS_IOC_POOL_DESTROY, &zc) != 0) {
1026 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN,
1027 "cannot destroy '%s'"), zhp->zpool_name);
1028
1029 if (errno == EROFS) {
1030 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1031 "one or more devices is read only"));
1032 (void) zfs_error(hdl, EZFS_BADDEV, msg);
1033 } else {
1034 (void) zpool_standard_error(hdl, errno, msg);
1035 }
1036
1037 if (zfp)
1038 zfs_close(zfp);
1039 return (-1);
1040 }
1041
1042 if (zfp) {
1043 remove_mountpoint(zfp);
1044 zfs_close(zfp);
1045 }
1046
1047 return (0);
1048 }
1049
1050 /*
1051 * Add the given vdevs to the pool. The caller must have already performed the
1052 * necessary verification to ensure that the vdev specification is well-formed.
1053 */
1054 int
1055 zpool_add(zpool_handle_t *zhp, nvlist_t *nvroot)
1056 {
1057 zfs_cmd_t zc = { 0 };
1058 int ret;
1059 libzfs_handle_t *hdl = zhp->zpool_hdl;
1060 char msg[1024];
1061 nvlist_t **spares, **l2cache;
1062 uint_t nspares, nl2cache;
1063
1064 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN,
1065 "cannot add to '%s'"), zhp->zpool_name);
1066
1067 if (zpool_get_prop_int(zhp, ZPOOL_PROP_VERSION, NULL) <
1068 SPA_VERSION_SPARES &&
1069 nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_SPARES,
1070 &spares, &nspares) == 0) {
1071 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "pool must be "
1072 "upgraded to add hot spares"));
1073 return (zfs_error(hdl, EZFS_BADVERSION, msg));
1074 }
1075
1076 if (pool_is_bootable(zhp) && nvlist_lookup_nvlist_array(nvroot,
1077 ZPOOL_CONFIG_SPARES, &spares, &nspares) == 0) {
1078 uint64_t s;
1079
1080 for (s = 0; s < nspares; s++) {
1081 char *path;
1082
1083 if (nvlist_lookup_string(spares[s], ZPOOL_CONFIG_PATH,
1084 &path) == 0 && pool_uses_efi(spares[s])) {
1085 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1086 "device '%s' contains an EFI label and "
1087 "cannot be used on root pools."),
1088 zpool_vdev_name(hdl, NULL, spares[s],
1089 B_FALSE));
1090 return (zfs_error(hdl, EZFS_POOL_NOTSUP, msg));
1091 }
1092 }
1093 }
1094
1095 if (zpool_get_prop_int(zhp, ZPOOL_PROP_VERSION, NULL) <
1096 SPA_VERSION_L2CACHE &&
1097 nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_L2CACHE,
1098 &l2cache, &nl2cache) == 0) {
1099 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "pool must be "
1100 "upgraded to add cache devices"));
1101 return (zfs_error(hdl, EZFS_BADVERSION, msg));
1102 }
1103
1104 if (zcmd_write_conf_nvlist(hdl, &zc, nvroot) != 0)
1105 return (-1);
1106 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
1107
1108 if (zfs_ioctl(hdl, ZFS_IOC_VDEV_ADD, &zc) != 0) {
1109 switch (errno) {
1110 case EBUSY:
1111 /*
1112 * This can happen if the user has specified the same
1113 * device multiple times. We can't reliably detect this
1114 * until we try to add it and see we already have a
1115 * label.
1116 */
1117 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1118 "one or more vdevs refer to the same device"));
1119 (void) zfs_error(hdl, EZFS_BADDEV, msg);
1120 break;
1121
1122 case EOVERFLOW:
1123 /*
1124 * This occurrs when one of the devices is below
1125 * SPA_MINDEVSIZE. Unfortunately, we can't detect which
1126 * device was the problem device since there's no
1127 * reliable way to determine device size from userland.
1128 */
1129 {
1130 char buf[64];
1131
1132 zfs_nicenum(SPA_MINDEVSIZE, buf, sizeof (buf));
1133
1134 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1135 "device is less than the minimum "
1136 "size (%s)"), buf);
1137 }
1138 (void) zfs_error(hdl, EZFS_BADDEV, msg);
1139 break;
1140
1141 case ENOTSUP:
1142 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1143 "pool must be upgraded to add these vdevs"));
1144 (void) zfs_error(hdl, EZFS_BADVERSION, msg);
1145 break;
1146
1147 case EDOM:
1148 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1149 "root pool can not have multiple vdevs"
1150 " or separate logs"));
1151 (void) zfs_error(hdl, EZFS_POOL_NOTSUP, msg);
1152 break;
1153
1154 case ENOTBLK:
1155 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1156 "cache device must be a disk or disk slice"));
1157 (void) zfs_error(hdl, EZFS_BADDEV, msg);
1158 break;
1159
1160 default:
1161 (void) zpool_standard_error(hdl, errno, msg);
1162 }
1163
1164 ret = -1;
1165 } else {
1166 ret = 0;
1167 }
1168
1169 zcmd_free_nvlists(&zc);
1170
1171 return (ret);
1172 }
1173
1174 /*
1175 * Exports the pool from the system. The caller must ensure that there are no
1176 * mounted datasets in the pool.
1177 */
1178 int
1179 zpool_export_common(zpool_handle_t *zhp, boolean_t force, boolean_t hardforce)
1180 {
1181 zfs_cmd_t zc = { 0 };
1182 char msg[1024];
1183
1184 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN,
1185 "cannot export '%s'"), zhp->zpool_name);
1186
1187 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
1188 zc.zc_cookie = force;
1189 zc.zc_guid = hardforce;
1190
1191 if (zfs_ioctl(zhp->zpool_hdl, ZFS_IOC_POOL_EXPORT, &zc) != 0) {
1192 switch (errno) {
1193 case EXDEV:
1194 zfs_error_aux(zhp->zpool_hdl, dgettext(TEXT_DOMAIN,
1195 "use '-f' to override the following errors:\n"
1196 "'%s' has an active shared spare which could be"
1197 " used by other pools once '%s' is exported."),
1198 zhp->zpool_name, zhp->zpool_name);
1199 return (zfs_error(zhp->zpool_hdl, EZFS_ACTIVE_SPARE,
1200 msg));
1201 default:
1202 return (zpool_standard_error_fmt(zhp->zpool_hdl, errno,
1203 msg));
1204 }
1205 }
1206
1207 return (0);
1208 }
1209
1210 int
1211 zpool_export(zpool_handle_t *zhp, boolean_t force)
1212 {
1213 return (zpool_export_common(zhp, force, B_FALSE));
1214 }
1215
1216 int
1217 zpool_export_force(zpool_handle_t *zhp)
1218 {
1219 return (zpool_export_common(zhp, B_TRUE, B_TRUE));
1220 }
1221
1222 static void
1223 zpool_rewind_exclaim(libzfs_handle_t *hdl, const char *name, boolean_t dryrun,
1224 nvlist_t *config)
1225 {
1226 nvlist_t *nv = NULL;
1227 uint64_t rewindto;
1228 int64_t loss = -1;
1229 struct tm t;
1230 char timestr[128];
1231
1232 if (!hdl->libzfs_printerr || config == NULL)
1233 return;
1234
1235 if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_LOAD_INFO, &nv) != 0)
1236 return;
1237
1238 if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_LOAD_TIME, &rewindto) != 0)
1239 return;
1240 (void) nvlist_lookup_int64(nv, ZPOOL_CONFIG_REWIND_TIME, &loss);
1241
1242 if (localtime_r((time_t *)&rewindto, &t) != NULL &&
1243 strftime(timestr, 128, 0, &t) != 0) {
1244 if (dryrun) {
1245 (void) printf(dgettext(TEXT_DOMAIN,
1246 "Would be able to return %s "
1247 "to its state as of %s.\n"),
1248 name, timestr);
1249 } else {
1250 (void) printf(dgettext(TEXT_DOMAIN,
1251 "Pool %s returned to its state as of %s.\n"),
1252 name, timestr);
1253 }
1254 if (loss > 120) {
1255 (void) printf(dgettext(TEXT_DOMAIN,
1256 "%s approximately %lld "),
1257 dryrun ? "Would discard" : "Discarded",
1258 (loss + 30) / 60);
1259 (void) printf(dgettext(TEXT_DOMAIN,
1260 "minutes of transactions.\n"));
1261 } else if (loss > 0) {
1262 (void) printf(dgettext(TEXT_DOMAIN,
1263 "%s approximately %lld "),
1264 dryrun ? "Would discard" : "Discarded", loss);
1265 (void) printf(dgettext(TEXT_DOMAIN,
1266 "seconds of transactions.\n"));
1267 }
1268 }
1269 }
1270
1271 void
1272 zpool_explain_recover(libzfs_handle_t *hdl, const char *name, int reason,
1273 nvlist_t *config)
1274 {
1275 nvlist_t *nv = NULL;
1276 int64_t loss = -1;
1277 uint64_t edata = UINT64_MAX;
1278 uint64_t rewindto;
1279 struct tm t;
1280 char timestr[128];
1281
1282 if (!hdl->libzfs_printerr)
1283 return;
1284
1285 if (reason >= 0)
1286 (void) printf(dgettext(TEXT_DOMAIN, "action: "));
1287 else
1288 (void) printf(dgettext(TEXT_DOMAIN, "\t"));
1289
1290 /* All attempted rewinds failed if ZPOOL_CONFIG_LOAD_TIME missing */
1291 if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_LOAD_INFO, &nv) != 0 ||
1292 nvlist_lookup_uint64(nv, ZPOOL_CONFIG_LOAD_TIME, &rewindto) != 0)
1293 goto no_info;
1294
1295 (void) nvlist_lookup_int64(nv, ZPOOL_CONFIG_REWIND_TIME, &loss);
1296 (void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_LOAD_DATA_ERRORS,
1297 &edata);
1298
1299 (void) printf(dgettext(TEXT_DOMAIN,
1300 "Recovery is possible, but will result in some data loss.\n"));
1301
1302 if (localtime_r((time_t *)&rewindto, &t) != NULL &&
1303 strftime(timestr, 128, 0, &t) != 0) {
1304 (void) printf(dgettext(TEXT_DOMAIN,
1305 "\tReturning the pool to its state as of %s\n"
1306 "\tshould correct the problem. "),
1307 timestr);
1308 } else {
1309 (void) printf(dgettext(TEXT_DOMAIN,
1310 "\tReverting the pool to an earlier state "
1311 "should correct the problem.\n\t"));
1312 }
1313
1314 if (loss > 120) {
1315 (void) printf(dgettext(TEXT_DOMAIN,
1316 "Approximately %lld minutes of data\n"
1317 "\tmust be discarded, irreversibly. "), (loss + 30) / 60);
1318 } else if (loss > 0) {
1319 (void) printf(dgettext(TEXT_DOMAIN,
1320 "Approximately %lld seconds of data\n"
1321 "\tmust be discarded, irreversibly. "), loss);
1322 }
1323 if (edata != 0 && edata != UINT64_MAX) {
1324 if (edata == 1) {
1325 (void) printf(dgettext(TEXT_DOMAIN,
1326 "After rewind, at least\n"
1327 "\tone persistent user-data error will remain. "));
1328 } else {
1329 (void) printf(dgettext(TEXT_DOMAIN,
1330 "After rewind, several\n"
1331 "\tpersistent user-data errors will remain. "));
1332 }
1333 }
1334 (void) printf(dgettext(TEXT_DOMAIN,
1335 "Recovery can be attempted\n\tby executing 'zpool %s -F %s'. "),
1336 reason >= 0 ? "clear" : "import", name);
1337
1338 (void) printf(dgettext(TEXT_DOMAIN,
1339 "A scrub of the pool\n"
1340 "\tis strongly recommended after recovery.\n"));
1341 return;
1342
1343 no_info:
1344 (void) printf(dgettext(TEXT_DOMAIN,
1345 "Destroy and re-create the pool from\n\ta backup source.\n"));
1346 }
1347
1348 /*
1349 * zpool_import() is a contracted interface. Should be kept the same
1350 * if possible.
1351 *
1352 * Applications should use zpool_import_props() to import a pool with
1353 * new properties value to be set.
1354 */
1355 int
1356 zpool_import(libzfs_handle_t *hdl, nvlist_t *config, const char *newname,
1357 char *altroot)
1358 {
1359 nvlist_t *props = NULL;
1360 int ret;
1361
1362 if (altroot != NULL) {
1363 if (nvlist_alloc(&props, NV_UNIQUE_NAME, 0) != 0) {
1364 return (zfs_error_fmt(hdl, EZFS_NOMEM,
1365 dgettext(TEXT_DOMAIN, "cannot import '%s'"),
1366 newname));
1367 }
1368
1369 if (nvlist_add_string(props,
1370 zpool_prop_to_name(ZPOOL_PROP_ALTROOT), altroot) != 0 ||
1371 nvlist_add_string(props,
1372 zpool_prop_to_name(ZPOOL_PROP_CACHEFILE), "none") != 0) {
1373 nvlist_free(props);
1374 return (zfs_error_fmt(hdl, EZFS_NOMEM,
1375 dgettext(TEXT_DOMAIN, "cannot import '%s'"),
1376 newname));
1377 }
1378 }
1379
1380 ret = zpool_import_props(hdl, config, newname, props,
1381 ZFS_IMPORT_NORMAL);
1382 if (props)
1383 nvlist_free(props);
1384 return (ret);
1385 }
1386
1387 static void
1388 print_vdev_tree(libzfs_handle_t *hdl, const char *name, nvlist_t *nv,
1389 int indent)
1390 {
1391 nvlist_t **child;
1392 uint_t c, children;
1393 char *vname;
1394 uint64_t is_log = 0;
1395
1396 (void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_IS_LOG,
1397 &is_log);
1398
1399 if (name != NULL)
1400 (void) printf("\t%*s%s%s\n", indent, "", name,
1401 is_log ? " [log]" : "");
1402
1403 if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN,
1404 &child, &children) != 0)
1405 return;
1406
1407 for (c = 0; c < children; c++) {
1408 vname = zpool_vdev_name(hdl, NULL, child[c], B_TRUE);
1409 print_vdev_tree(hdl, vname, child[c], indent + 2);
1410 free(vname);
1411 }
1412 }
1413
1414 /*
1415 * Import the given pool using the known configuration and a list of
1416 * properties to be set. The configuration should have come from
1417 * zpool_find_import(). The 'newname' parameters control whether the pool
1418 * is imported with a different name.
1419 */
1420 int
1421 zpool_import_props(libzfs_handle_t *hdl, nvlist_t *config, const char *newname,
1422 nvlist_t *props, int flags)
1423 {
1424 zfs_cmd_t zc = { 0 };
1425 zpool_rewind_policy_t policy;
1426 nvlist_t *nv = NULL;
1427 nvlist_t *nvinfo = NULL;
1428 nvlist_t *missing = NULL;
1429 char *thename;
1430 char *origname;
1431 int ret;
1432 int error = 0;
1433 char errbuf[1024];
1434
1435 verify(nvlist_lookup_string(config, ZPOOL_CONFIG_POOL_NAME,
1436 &origname) == 0);
1437
1438 (void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN,
1439 "cannot import pool '%s'"), origname);
1440
1441 if (newname != NULL) {
1442 if (!zpool_name_valid(hdl, B_FALSE, newname))
1443 return (zfs_error_fmt(hdl, EZFS_INVALIDNAME,
1444 dgettext(TEXT_DOMAIN, "cannot import '%s'"),
1445 newname));
1446 thename = (char *)newname;
1447 } else {
1448 thename = origname;
1449 }
1450
1451 if (props) {
1452 uint64_t version;
1453 prop_flags_t flags = { .create = B_FALSE, .import = B_TRUE };
1454
1455 verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_VERSION,
1456 &version) == 0);
1457
1458 if ((props = zpool_valid_proplist(hdl, origname,
1459 props, version, flags, errbuf)) == NULL) {
1460 return (-1);
1461 } else if (zcmd_write_src_nvlist(hdl, &zc, props) != 0) {
1462 nvlist_free(props);
1463 return (-1);
1464 }
1465 }
1466
1467 (void) strlcpy(zc.zc_name, thename, sizeof (zc.zc_name));
1468
1469 verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID,
1470 &zc.zc_guid) == 0);
1471
1472 if (zcmd_write_conf_nvlist(hdl, &zc, config) != 0) {
1473 nvlist_free(props);
1474 return (-1);
1475 }
1476 if (zcmd_alloc_dst_nvlist(hdl, &zc, zc.zc_nvlist_conf_size * 2) != 0) {
1477 nvlist_free(props);
1478 return (-1);
1479 }
1480
1481 zc.zc_cookie = flags;
1482 while ((ret = zfs_ioctl(hdl, ZFS_IOC_POOL_IMPORT, &zc)) != 0 &&
1483 errno == ENOMEM) {
1484 if (zcmd_expand_dst_nvlist(hdl, &zc) != 0) {
1485 zcmd_free_nvlists(&zc);
1486 return (-1);
1487 }
1488 }
1489 if (ret != 0)
1490 error = errno;
1491
1492 (void) zcmd_read_dst_nvlist(hdl, &zc, &nv);
1493 zpool_get_rewind_policy(config, &policy);
1494
1495 if (error) {
1496 char desc[1024];
1497
1498 /*
1499 * Dry-run failed, but we print out what success
1500 * looks like if we found a best txg
1501 */
1502 if (policy.zrp_request & ZPOOL_TRY_REWIND) {
1503 zpool_rewind_exclaim(hdl, newname ? origname : thename,
1504 B_TRUE, nv);
1505 nvlist_free(nv);
1506 return (-1);
1507 }
1508
1509 if (newname == NULL)
1510 (void) snprintf(desc, sizeof (desc),
1511 dgettext(TEXT_DOMAIN, "cannot import '%s'"),
1512 thename);
1513 else
1514 (void) snprintf(desc, sizeof (desc),
1515 dgettext(TEXT_DOMAIN, "cannot import '%s' as '%s'"),
1516 origname, thename);
1517
1518 switch (error) {
1519 case ENOTSUP:
1520 /*
1521 * Unsupported version.
1522 */
1523 (void) zfs_error(hdl, EZFS_BADVERSION, desc);
1524 break;
1525
1526 case EINVAL:
1527 (void) zfs_error(hdl, EZFS_INVALCONFIG, desc);
1528 break;
1529
1530 case EROFS:
1531 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1532 "one or more devices is read only"));
1533 (void) zfs_error(hdl, EZFS_BADDEV, desc);
1534 break;
1535
1536 case ENXIO:
1537 if (nv && nvlist_lookup_nvlist(nv,
1538 ZPOOL_CONFIG_LOAD_INFO, &nvinfo) == 0 &&
1539 nvlist_lookup_nvlist(nvinfo,
1540 ZPOOL_CONFIG_MISSING_DEVICES, &missing) == 0) {
1541 (void) printf(dgettext(TEXT_DOMAIN,
1542 "The devices below are missing, use "
1543 "'-m' to import the pool anyway:\n"));
1544 print_vdev_tree(hdl, NULL, missing, 2);
1545 (void) printf("\n");
1546 }
1547 (void) zpool_standard_error(hdl, error, desc);
1548 break;
1549
1550 case EEXIST:
1551 (void) zpool_standard_error(hdl, error, desc);
1552 break;
1553
1554 default:
1555 (void) zpool_standard_error(hdl, error, desc);
1556 zpool_explain_recover(hdl,
1557 newname ? origname : thename, -error, nv);
1558 break;
1559 }
1560
1561 nvlist_free(nv);
1562 ret = -1;
1563 } else {
1564 zpool_handle_t *zhp;
1565
1566 /*
1567 * This should never fail, but play it safe anyway.
1568 */
1569 if (zpool_open_silent(hdl, thename, &zhp) != 0)
1570 ret = -1;
1571 else if (zhp != NULL)
1572 zpool_close(zhp);
1573 if (policy.zrp_request &
1574 (ZPOOL_DO_REWIND | ZPOOL_TRY_REWIND)) {
1575 zpool_rewind_exclaim(hdl, newname ? origname : thename,
1576 ((policy.zrp_request & ZPOOL_TRY_REWIND) != 0), nv);
1577 }
1578 nvlist_free(nv);
1579 return (0);
1580 }
1581
1582 zcmd_free_nvlists(&zc);
1583 nvlist_free(props);
1584
1585 return (ret);
1586 }
1587
1588 /*
1589 * Scan the pool.
1590 */
1591 int
1592 zpool_scan(zpool_handle_t *zhp, pool_scan_func_t func)
1593 {
1594 zfs_cmd_t zc = { 0 };
1595 char msg[1024];
1596 libzfs_handle_t *hdl = zhp->zpool_hdl;
1597
1598 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
1599 zc.zc_cookie = func;
1600
1601 if (zfs_ioctl(hdl, ZFS_IOC_POOL_SCAN, &zc) == 0 ||
1602 (errno == ENOENT && func != POOL_SCAN_NONE))
1603 return (0);
1604
1605 if (func == POOL_SCAN_SCRUB) {
1606 (void) snprintf(msg, sizeof (msg),
1607 dgettext(TEXT_DOMAIN, "cannot scrub %s"), zc.zc_name);
1608 } else if (func == POOL_SCAN_NONE) {
1609 (void) snprintf(msg, sizeof (msg),
1610 dgettext(TEXT_DOMAIN, "cannot cancel scrubbing %s"),
1611 zc.zc_name);
1612 } else {
1613 assert(!"unexpected result");
1614 }
1615
1616 if (errno == EBUSY) {
1617 nvlist_t *nvroot;
1618 pool_scan_stat_t *ps = NULL;
1619 uint_t psc;
1620
1621 verify(nvlist_lookup_nvlist(zhp->zpool_config,
1622 ZPOOL_CONFIG_VDEV_TREE, &nvroot) == 0);
1623 (void) nvlist_lookup_uint64_array(nvroot,
1624 ZPOOL_CONFIG_SCAN_STATS, (uint64_t **)&ps, &psc);
1625 if (ps && ps->pss_func == POOL_SCAN_SCRUB)
1626 return (zfs_error(hdl, EZFS_SCRUBBING, msg));
1627 else
1628 return (zfs_error(hdl, EZFS_RESILVERING, msg));
1629 } else if (errno == ENOENT) {
1630 return (zfs_error(hdl, EZFS_NO_SCRUB, msg));
1631 } else {
1632 return (zpool_standard_error(hdl, errno, msg));
1633 }
1634 }
1635
1636 /*
1637 * This provides a very minimal check whether a given string is likely a
1638 * c#t#d# style string. Users of this are expected to do their own
1639 * verification of the s# part.
1640 */
1641 #define CTD_CHECK(str) (str && str[0] == 'c' && isdigit(str[1]))
1642
1643 /*
1644 * More elaborate version for ones which may start with "/dev/dsk/"
1645 * and the like.
1646 */
1647 static int
1648 ctd_check_path(char *str) {
1649 /*
1650 * If it starts with a slash, check the last component.
1651 */
1652 if (str && str[0] == '/') {
1653 char *tmp = strrchr(str, '/');
1654
1655 /*
1656 * If it ends in "/old", check the second-to-last
1657 * component of the string instead.
1658 */
1659 if (tmp != str && strcmp(tmp, "/old") == 0) {
1660 for (tmp--; *tmp != '/'; tmp--)
1661 ;
1662 }
1663 str = tmp + 1;
1664 }
1665 return (CTD_CHECK(str));
1666 }
1667
1668 /*
1669 * Find a vdev that matches the search criteria specified. We use the
1670 * the nvpair name to determine how we should look for the device.
1671 * 'avail_spare' is set to TRUE if the provided guid refers to an AVAIL
1672 * spare; but FALSE if its an INUSE spare.
1673 */
1674 static nvlist_t *
1675 vdev_to_nvlist_iter(nvlist_t *nv, nvlist_t *search, boolean_t *avail_spare,
1676 boolean_t *l2cache, boolean_t *log)
1677 {
1678 uint_t c, children;
1679 nvlist_t **child;
1680 nvlist_t *ret;
1681 uint64_t is_log;
1682 char *srchkey;
1683 nvpair_t *pair = nvlist_next_nvpair(search, NULL);
1684
1685 /* Nothing to look for */
1686 if (search == NULL || pair == NULL)
1687 return (NULL);
1688
1689 /* Obtain the key we will use to search */
1690 srchkey = nvpair_name(pair);
1691
1692 switch (nvpair_type(pair)) {
1693 case DATA_TYPE_UINT64:
1694 if (strcmp(srchkey, ZPOOL_CONFIG_GUID) == 0) {
1695 uint64_t srchval, theguid;
1696
1697 verify(nvpair_value_uint64(pair, &srchval) == 0);
1698 verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID,
1699 &theguid) == 0);
1700 if (theguid == srchval)
1701 return (nv);
1702 }
1703 break;
1704
1705 case DATA_TYPE_STRING: {
1706 char *srchval, *val;
1707
1708 verify(nvpair_value_string(pair, &srchval) == 0);
1709 if (nvlist_lookup_string(nv, srchkey, &val) != 0)
1710 break;
1711
1712 /*
1713 * Search for the requested value. Special cases:
1714 *
1715 * - ZPOOL_CONFIG_PATH for whole disk entries. These end in
1716 * "s0" or "s0/old". The "s0" part is hidden from the user,
1717 * but included in the string, so this matches around it.
1718 * - looking for a top-level vdev name (i.e. ZPOOL_CONFIG_TYPE).
1719 *
1720 * Otherwise, all other searches are simple string compares.
1721 */
1722 if (strcmp(srchkey, ZPOOL_CONFIG_PATH) == 0 &&
1723 ctd_check_path(val)) {
1724 uint64_t wholedisk = 0;
1725
1726 (void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_WHOLE_DISK,
1727 &wholedisk);
1728 if (wholedisk) {
1729 int slen = strlen(srchval);
1730 int vlen = strlen(val);
1731
1732 if (slen != vlen - 2)
1733 break;
1734
1735 /*
1736 * make_leaf_vdev() should only set
1737 * wholedisk for ZPOOL_CONFIG_PATHs which
1738 * will include "/dev/dsk/", giving plenty of
1739 * room for the indices used next.
1740 */
1741 ASSERT(vlen >= 6);
1742
1743 /*
1744 * strings identical except trailing "s0"
1745 */
1746 if (strcmp(&val[vlen - 2], "s0") == 0 &&
1747 strncmp(srchval, val, slen) == 0)
1748 return (nv);
1749
1750 /*
1751 * strings identical except trailing "s0/old"
1752 */
1753 if (strcmp(&val[vlen - 6], "s0/old") == 0 &&
1754 strcmp(&srchval[slen - 4], "/old") == 0 &&
1755 strncmp(srchval, val, slen - 4) == 0)
1756 return (nv);
1757
1758 break;
1759 }
1760 } else if (strcmp(srchkey, ZPOOL_CONFIG_TYPE) == 0 && val) {
1761 char *type, *idx, *end, *p;
1762 uint64_t id, vdev_id;
1763
1764 /*
1765 * Determine our vdev type, keeping in mind
1766 * that the srchval is composed of a type and
1767 * vdev id pair (i.e. mirror-4).
1768 */
1769 if ((type = strdup(srchval)) == NULL)
1770 return (NULL);
1771
1772 if ((p = strrchr(type, '-')) == NULL) {
1773 free(type);
1774 break;
1775 }
1776 idx = p + 1;
1777 *p = '\0';
1778
1779 /*
1780 * If the types don't match then keep looking.
1781 */
1782 if (strncmp(val, type, strlen(val)) != 0) {
1783 free(type);
1784 break;
1785 }
1786
1787 verify(strncmp(type, VDEV_TYPE_RAIDZ,
1788 strlen(VDEV_TYPE_RAIDZ)) == 0 ||
1789 strncmp(type, VDEV_TYPE_MIRROR,
1790 strlen(VDEV_TYPE_MIRROR)) == 0);
1791 verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_ID,
1792 &id) == 0);
1793
1794 errno = 0;
1795 vdev_id = strtoull(idx, &end, 10);
1796
1797 free(type);
1798 if (errno != 0)
1799 return (NULL);
1800
1801 /*
1802 * Now verify that we have the correct vdev id.
1803 */
1804 if (vdev_id == id)
1805 return (nv);
1806 }
1807
1808 /*
1809 * Common case
1810 */
1811 if (strcmp(srchval, val) == 0)
1812 return (nv);
1813 break;
1814 }
1815
1816 default:
1817 break;
1818 }
1819
1820 if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN,
1821 &child, &children) != 0)
1822 return (NULL);
1823
1824 for (c = 0; c < children; c++) {
1825 if ((ret = vdev_to_nvlist_iter(child[c], search,
1826 avail_spare, l2cache, NULL)) != NULL) {
1827 /*
1828 * The 'is_log' value is only set for the toplevel
1829 * vdev, not the leaf vdevs. So we always lookup the
1830 * log device from the root of the vdev tree (where
1831 * 'log' is non-NULL).
1832 */
1833 if (log != NULL &&
1834 nvlist_lookup_uint64(child[c],
1835 ZPOOL_CONFIG_IS_LOG, &is_log) == 0 &&
1836 is_log) {
1837 *log = B_TRUE;
1838 }
1839 return (ret);
1840 }
1841 }
1842
1843 if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_SPARES,
1844 &child, &children) == 0) {
1845 for (c = 0; c < children; c++) {
1846 if ((ret = vdev_to_nvlist_iter(child[c], search,
1847 avail_spare, l2cache, NULL)) != NULL) {
1848 *avail_spare = B_TRUE;
1849 return (ret);
1850 }
1851 }
1852 }
1853
1854 if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_L2CACHE,
1855 &child, &children) == 0) {
1856 for (c = 0; c < children; c++) {
1857 if ((ret = vdev_to_nvlist_iter(child[c], search,
1858 avail_spare, l2cache, NULL)) != NULL) {
1859 *l2cache = B_TRUE;
1860 return (ret);
1861 }
1862 }
1863 }
1864
1865 return (NULL);
1866 }
1867
1868 /*
1869 * Given a physical path (minus the "/devices" prefix), find the
1870 * associated vdev.
1871 */
1872 nvlist_t *
1873 zpool_find_vdev_by_physpath(zpool_handle_t *zhp, const char *ppath,
1874 boolean_t *avail_spare, boolean_t *l2cache, boolean_t *log)
1875 {
1876 nvlist_t *search, *nvroot, *ret;
1877
1878 verify(nvlist_alloc(&search, NV_UNIQUE_NAME, KM_SLEEP) == 0);
1879 verify(nvlist_add_string(search, ZPOOL_CONFIG_PHYS_PATH, ppath) == 0);
1880
1881 verify(nvlist_lookup_nvlist(zhp->zpool_config, ZPOOL_CONFIG_VDEV_TREE,
1882 &nvroot) == 0);
1883
1884 *avail_spare = B_FALSE;
1885 *l2cache = B_FALSE;
1886 if (log != NULL)
1887 *log = B_FALSE;
1888 ret = vdev_to_nvlist_iter(nvroot, search, avail_spare, l2cache, log);
1889 nvlist_free(search);
1890
1891 return (ret);
1892 }
1893
1894 /*
1895 * Determine if we have an "interior" top-level vdev (i.e mirror/raidz).
1896 */
1897 boolean_t
1898 zpool_vdev_is_interior(const char *name)
1899 {
1900 if (strncmp(name, VDEV_TYPE_RAIDZ, strlen(VDEV_TYPE_RAIDZ)) == 0 ||
1901 strncmp(name, VDEV_TYPE_MIRROR, strlen(VDEV_TYPE_MIRROR)) == 0)
1902 return (B_TRUE);
1903 return (B_FALSE);
1904 }
1905
1906 nvlist_t *
1907 zpool_find_vdev(zpool_handle_t *zhp, const char *path, boolean_t *avail_spare,
1908 boolean_t *l2cache, boolean_t *log)
1909 {
1910 char buf[MAXPATHLEN];
1911 char *end;
1912 nvlist_t *nvroot, *search, *ret;
1913 uint64_t guid;
1914
1915 verify(nvlist_alloc(&search, NV_UNIQUE_NAME, KM_SLEEP) == 0);
1916
1917 guid = strtoull(path, &end, 10);
1918 if (guid != 0 && *end == '\0') {
1919 verify(nvlist_add_uint64(search, ZPOOL_CONFIG_GUID, guid) == 0);
1920 } else if (zpool_vdev_is_interior(path)) {
1921 verify(nvlist_add_string(search, ZPOOL_CONFIG_TYPE, path) == 0);
1922 } else if (path[0] != '/') {
1923 (void) snprintf(buf, sizeof (buf), "%s%s", "/dev/dsk/", path);
1924 verify(nvlist_add_string(search, ZPOOL_CONFIG_PATH, buf) == 0);
1925 } else {
1926 verify(nvlist_add_string(search, ZPOOL_CONFIG_PATH, path) == 0);
1927 }
1928
1929 verify(nvlist_lookup_nvlist(zhp->zpool_config, ZPOOL_CONFIG_VDEV_TREE,
1930 &nvroot) == 0);
1931
1932 *avail_spare = B_FALSE;
1933 *l2cache = B_FALSE;
1934 if (log != NULL)
1935 *log = B_FALSE;
1936 ret = vdev_to_nvlist_iter(nvroot, search, avail_spare, l2cache, log);
1937 nvlist_free(search);
1938
1939 return (ret);
1940 }
1941
1942 static int
1943 vdev_online(nvlist_t *nv)
1944 {
1945 uint64_t ival;
1946
1947 if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_OFFLINE, &ival) == 0 ||
1948 nvlist_lookup_uint64(nv, ZPOOL_CONFIG_FAULTED, &ival) == 0 ||
1949 nvlist_lookup_uint64(nv, ZPOOL_CONFIG_REMOVED, &ival) == 0)
1950 return (0);
1951
1952 return (1);
1953 }
1954
1955 /*
1956 * Helper function for zpool_get_physpaths().
1957 */
1958 static int
1959 vdev_get_one_physpath(nvlist_t *config, char *physpath, size_t physpath_size,
1960 size_t *bytes_written)
1961 {
1962 size_t bytes_left, pos, rsz;
1963 char *tmppath;
1964 const char *format;
1965
1966 if (nvlist_lookup_string(config, ZPOOL_CONFIG_PHYS_PATH,
1967 &tmppath) != 0)
1968 return (EZFS_NODEVICE);
1969
1970 pos = *bytes_written;
1971 bytes_left = physpath_size - pos;
1972 format = (pos == 0) ? "%s" : " %s";
1973
1974 rsz = snprintf(physpath + pos, bytes_left, format, tmppath);
1975 *bytes_written += rsz;
1976
1977 if (rsz >= bytes_left) {
1978 /* if physpath was not copied properly, clear it */
1979 if (bytes_left != 0) {
1980 physpath[pos] = 0;
1981 }
1982 return (EZFS_NOSPC);
1983 }
1984 return (0);
1985 }
1986
1987 static int
1988 vdev_get_physpaths(nvlist_t *nv, char *physpath, size_t phypath_size,
1989 size_t *rsz, boolean_t is_spare)
1990 {
1991 char *type;
1992 int ret;
1993
1994 if (nvlist_lookup_string(nv, ZPOOL_CONFIG_TYPE, &type) != 0)
1995 return (EZFS_INVALCONFIG);
1996
1997 if (strcmp(type, VDEV_TYPE_DISK) == 0) {
1998 /*
1999 * An active spare device has ZPOOL_CONFIG_IS_SPARE set.
2000 * For a spare vdev, we only want to boot from the active
2001 * spare device.
2002 */
2003 if (is_spare) {
2004 uint64_t spare = 0;
2005 (void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_IS_SPARE,
2006 &spare);
2007 if (!spare)
2008 return (EZFS_INVALCONFIG);
2009 }
2010
2011 if (vdev_online(nv)) {
2012 if ((ret = vdev_get_one_physpath(nv, physpath,
2013 phypath_size, rsz)) != 0)
2014 return (ret);
2015 }
2016 } else if (strcmp(type, VDEV_TYPE_MIRROR) == 0 ||
2017 strcmp(type, VDEV_TYPE_REPLACING) == 0 ||
2018 (is_spare = (strcmp(type, VDEV_TYPE_SPARE) == 0))) {
2019 nvlist_t **child;
2020 uint_t count;
2021 int i, ret;
2022
2023 if (nvlist_lookup_nvlist_array(nv,
2024 ZPOOL_CONFIG_CHILDREN, &child, &count) != 0)
2025 return (EZFS_INVALCONFIG);
2026
2027 for (i = 0; i < count; i++) {
2028 ret = vdev_get_physpaths(child[i], physpath,
2029 phypath_size, rsz, is_spare);
2030 if (ret == EZFS_NOSPC)
2031 return (ret);
2032 }
2033 }
2034
2035 return (EZFS_POOL_INVALARG);
2036 }
2037
2038 /*
2039 * Get phys_path for a root pool config.
2040 * Return 0 on success; non-zero on failure.
2041 */
2042 static int
2043 zpool_get_config_physpath(nvlist_t *config, char *physpath, size_t phypath_size)
2044 {
2045 size_t rsz;
2046 nvlist_t *vdev_root;
2047 nvlist_t **child;
2048 uint_t count;
2049 char *type;
2050
2051 rsz = 0;
2052
2053 if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE,
2054 &vdev_root) != 0)
2055 return (EZFS_INVALCONFIG);
2056
2057 if (nvlist_lookup_string(vdev_root, ZPOOL_CONFIG_TYPE, &type) != 0 ||
2058 nvlist_lookup_nvlist_array(vdev_root, ZPOOL_CONFIG_CHILDREN,
2059 &child, &count) != 0)
2060 return (EZFS_INVALCONFIG);
2061
2062 /*
2063 * root pool can not have EFI labeled disks and can only have
2064 * a single top-level vdev.
2065 */
2066 if (strcmp(type, VDEV_TYPE_ROOT) != 0 || count != 1 ||
2067 pool_uses_efi(vdev_root))
2068 return (EZFS_POOL_INVALARG);
2069
2070 (void) vdev_get_physpaths(child[0], physpath, phypath_size, &rsz,
2071 B_FALSE);
2072
2073 /* No online devices */
2074 if (rsz == 0)
2075 return (EZFS_NODEVICE);
2076
2077 return (0);
2078 }
2079
2080 /*
2081 * Get phys_path for a root pool
2082 * Return 0 on success; non-zero on failure.
2083 */
2084 int
2085 zpool_get_physpath(zpool_handle_t *zhp, char *physpath, size_t phypath_size)
2086 {
2087 return (zpool_get_config_physpath(zhp->zpool_config, physpath,
2088 phypath_size));
2089 }
2090
2091 /*
2092 * If the device has being dynamically expanded then we need to relabel
2093 * the disk to use the new unallocated space.
2094 */
2095 static int
2096 zpool_relabel_disk(libzfs_handle_t *hdl, const char *name)
2097 {
2098 char path[MAXPATHLEN];
2099 char errbuf[1024];
2100 int fd, error;
2101 int (*_efi_use_whole_disk)(int);
2102
2103 if ((_efi_use_whole_disk = (int (*)(int))dlsym(RTLD_DEFAULT,
2104 "efi_use_whole_disk")) == NULL)
2105 return (-1);
2106
2107 (void) snprintf(path, sizeof (path), "%s/%s", RDISK_ROOT, name);
2108
2109 if ((fd = open(path, O_RDWR | O_NDELAY)) < 0) {
2110 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "cannot "
2111 "relabel '%s': unable to open device"), name);
2112 return (zfs_error(hdl, EZFS_OPENFAILED, errbuf));
2113 }
2114
2115 /*
2116 * It's possible that we might encounter an error if the device
2117 * does not have any unallocated space left. If so, we simply
2118 * ignore that error and continue on.
2119 */
2120 error = _efi_use_whole_disk(fd);
2121 (void) close(fd);
2122 if (error && error != VT_ENOSPC) {
2123 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "cannot "
2124 "relabel '%s': unable to read disk capacity"), name);
2125 return (zfs_error(hdl, EZFS_NOCAP, errbuf));
2126 }
2127 return (0);
2128 }
2129
2130 /*
2131 * Bring the specified vdev online. The 'flags' parameter is a set of the
2132 * ZFS_ONLINE_* flags.
2133 */
2134 int
2135 zpool_vdev_online(zpool_handle_t *zhp, const char *path, int flags,
2136 vdev_state_t *newstate)
2137 {
2138 zfs_cmd_t zc = { 0 };
2139 char msg[1024];
2140 nvlist_t *tgt;
2141 boolean_t avail_spare, l2cache, islog;
2142 libzfs_handle_t *hdl = zhp->zpool_hdl;
2143
2144 if (flags & ZFS_ONLINE_EXPAND) {
2145 (void) snprintf(msg, sizeof (msg),
2146 dgettext(TEXT_DOMAIN, "cannot expand %s"), path);
2147 } else {
2148 (void) snprintf(msg, sizeof (msg),
2149 dgettext(TEXT_DOMAIN, "cannot online %s"), path);
2150 }
2151
2152 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
2153 if ((tgt = zpool_find_vdev(zhp, path, &avail_spare, &l2cache,
2154 &islog)) == NULL)
2155 return (zfs_error(hdl, EZFS_NODEVICE, msg));
2156
2157 verify(nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID, &zc.zc_guid) == 0);
2158
2159 if (avail_spare)
2160 return (zfs_error(hdl, EZFS_ISSPARE, msg));
2161
2162 if (flags & ZFS_ONLINE_EXPAND ||
2163 zpool_get_prop_int(zhp, ZPOOL_PROP_AUTOEXPAND, NULL)) {
2164 char *pathname = NULL;
2165 uint64_t wholedisk = 0;
2166
2167 (void) nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_WHOLE_DISK,
2168 &wholedisk);
2169 verify(nvlist_lookup_string(tgt, ZPOOL_CONFIG_PATH,
2170 &pathname) == 0);
2171
2172 /*
2173 * XXX - L2ARC 1.0 devices can't support expansion.
2174 */
2175 if (l2cache) {
2176 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2177 "cannot expand cache devices"));
2178 return (zfs_error(hdl, EZFS_VDEVNOTSUP, msg));
2179 }
2180
2181 if (wholedisk) {
2182 pathname += strlen(DISK_ROOT) + 1;
2183 (void) zpool_relabel_disk(hdl, pathname);
2184 }
2185 }
2186
2187 zc.zc_cookie = VDEV_STATE_ONLINE;
2188 zc.zc_obj = flags;
2189
2190 if (zfs_ioctl(hdl, ZFS_IOC_VDEV_SET_STATE, &zc) != 0) {
2191 if (errno == EINVAL) {
2192 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "was split "
2193 "from this pool into a new one. Use '%s' "
2194 "instead"), "zpool detach");
2195 return (zfs_error(hdl, EZFS_POSTSPLIT_ONLINE, msg));
2196 }
2197 return (zpool_standard_error(hdl, errno, msg));
2198 }
2199
2200 *newstate = zc.zc_cookie;
2201 return (0);
2202 }
2203
2204 /*
2205 * Take the specified vdev offline
2206 */
2207 int
2208 zpool_vdev_offline(zpool_handle_t *zhp, const char *path, boolean_t istmp)
2209 {
2210 zfs_cmd_t zc = { 0 };
2211 char msg[1024];
2212 nvlist_t *tgt;
2213 boolean_t avail_spare, l2cache;
2214 libzfs_handle_t *hdl = zhp->zpool_hdl;
2215
2216 (void) snprintf(msg, sizeof (msg),
2217 dgettext(TEXT_DOMAIN, "cannot offline %s"), path);
2218
2219 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
2220 if ((tgt = zpool_find_vdev(zhp, path, &avail_spare, &l2cache,
2221 NULL)) == NULL)
2222 return (zfs_error(hdl, EZFS_NODEVICE, msg));
2223
2224 verify(nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID, &zc.zc_guid) == 0);
2225
2226 if (avail_spare)
2227 return (zfs_error(hdl, EZFS_ISSPARE, msg));
2228
2229 zc.zc_cookie = VDEV_STATE_OFFLINE;
2230 zc.zc_obj = istmp ? ZFS_OFFLINE_TEMPORARY : 0;
2231
2232 if (zfs_ioctl(hdl, ZFS_IOC_VDEV_SET_STATE, &zc) == 0)
2233 return (0);
2234
2235 switch (errno) {
2236 case EBUSY:
2237
2238 /*
2239 * There are no other replicas of this device.
2240 */
2241 return (zfs_error(hdl, EZFS_NOREPLICAS, msg));
2242
2243 case EEXIST:
2244 /*
2245 * The log device has unplayed logs
2246 */
2247 return (zfs_error(hdl, EZFS_UNPLAYED_LOGS, msg));
2248
2249 default:
2250 return (zpool_standard_error(hdl, errno, msg));
2251 }
2252 }
2253
2254 /*
2255 * Mark the given vdev faulted.
2256 */
2257 int
2258 zpool_vdev_fault(zpool_handle_t *zhp, uint64_t guid, vdev_aux_t aux)
2259 {
2260 zfs_cmd_t zc = { 0 };
2261 char msg[1024];
2262 libzfs_handle_t *hdl = zhp->zpool_hdl;
2263
2264 (void) snprintf(msg, sizeof (msg),
2265 dgettext(TEXT_DOMAIN, "cannot fault %llu"), guid);
2266
2267 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
2268 zc.zc_guid = guid;
2269 zc.zc_cookie = VDEV_STATE_FAULTED;
2270 zc.zc_obj = aux;
2271
2272 if (ioctl(hdl->libzfs_fd, ZFS_IOC_VDEV_SET_STATE, &zc) == 0)
2273 return (0);
2274
2275 switch (errno) {
2276 case EBUSY:
2277
2278 /*
2279 * There are no other replicas of this device.
2280 */
2281 return (zfs_error(hdl, EZFS_NOREPLICAS, msg));
2282
2283 default:
2284 return (zpool_standard_error(hdl, errno, msg));
2285 }
2286
2287 }
2288
2289 /*
2290 * Mark the given vdev degraded.
2291 */
2292 int
2293 zpool_vdev_degrade(zpool_handle_t *zhp, uint64_t guid, vdev_aux_t aux)
2294 {
2295 zfs_cmd_t zc = { 0 };
2296 char msg[1024];
2297 libzfs_handle_t *hdl = zhp->zpool_hdl;
2298
2299 (void) snprintf(msg, sizeof (msg),
2300 dgettext(TEXT_DOMAIN, "cannot degrade %llu"), guid);
2301
2302 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
2303 zc.zc_guid = guid;
2304 zc.zc_cookie = VDEV_STATE_DEGRADED;
2305 zc.zc_obj = aux;
2306
2307 if (ioctl(hdl->libzfs_fd, ZFS_IOC_VDEV_SET_STATE, &zc) == 0)
2308 return (0);
2309
2310 return (zpool_standard_error(hdl, errno, msg));
2311 }
2312
2313 /*
2314 * Returns TRUE if the given nvlist is a vdev that was originally swapped in as
2315 * a hot spare.
2316 */
2317 static boolean_t
2318 is_replacing_spare(nvlist_t *search, nvlist_t *tgt, int which)
2319 {
2320 nvlist_t **child;
2321 uint_t c, children;
2322 char *type;
2323
2324 if (nvlist_lookup_nvlist_array(search, ZPOOL_CONFIG_CHILDREN, &child,
2325 &children) == 0) {
2326 verify(nvlist_lookup_string(search, ZPOOL_CONFIG_TYPE,
2327 &type) == 0);
2328
2329 if (strcmp(type, VDEV_TYPE_SPARE) == 0 &&
2330 children == 2 && child[which] == tgt)
2331 return (B_TRUE);
2332
2333 for (c = 0; c < children; c++)
2334 if (is_replacing_spare(child[c], tgt, which))
2335 return (B_TRUE);
2336 }
2337
2338 return (B_FALSE);
2339 }
2340
2341 /*
2342 * Attach new_disk (fully described by nvroot) to old_disk.
2343 * If 'replacing' is specified, the new disk will replace the old one.
2344 */
2345 int
2346 zpool_vdev_attach(zpool_handle_t *zhp,
2347 const char *old_disk, const char *new_disk, nvlist_t *nvroot, int replacing)
2348 {
2349 zfs_cmd_t zc = { 0 };
2350 char msg[1024];
2351 int ret;
2352 nvlist_t *tgt;
2353 boolean_t avail_spare, l2cache, islog;
2354 uint64_t val;
2355 char *newname;
2356 nvlist_t **child;
2357 uint_t children;
2358 nvlist_t *config_root;
2359 libzfs_handle_t *hdl = zhp->zpool_hdl;
2360 boolean_t rootpool = pool_is_bootable(zhp);
2361
2362 if (replacing)
2363 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN,
2364 "cannot replace %s with %s"), old_disk, new_disk);
2365 else
2366 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN,
2367 "cannot attach %s to %s"), new_disk, old_disk);
2368
2369 /*
2370 * If this is a root pool, make sure that we're not attaching an
2371 * EFI labeled device.
2372 */
2373 if (rootpool && pool_uses_efi(nvroot)) {
2374 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2375 "EFI labeled devices are not supported on root pools."));
2376 return (zfs_error(hdl, EZFS_POOL_NOTSUP, msg));
2377 }
2378
2379 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
2380 if ((tgt = zpool_find_vdev(zhp, old_disk, &avail_spare, &l2cache,
2381 &islog)) == 0)
2382 return (zfs_error(hdl, EZFS_NODEVICE, msg));
2383
2384 if (avail_spare)
2385 return (zfs_error(hdl, EZFS_ISSPARE, msg));
2386
2387 if (l2cache)
2388 return (zfs_error(hdl, EZFS_ISL2CACHE, msg));
2389
2390 verify(nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID, &zc.zc_guid) == 0);
2391 zc.zc_cookie = replacing;
2392
2393 if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_CHILDREN,
2394 &child, &children) != 0 || children != 1) {
2395 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2396 "new device must be a single disk"));
2397 return (zfs_error(hdl, EZFS_INVALCONFIG, msg));
2398 }
2399
2400 verify(nvlist_lookup_nvlist(zpool_get_config(zhp, NULL),
2401 ZPOOL_CONFIG_VDEV_TREE, &config_root) == 0);
2402
2403 if ((newname = zpool_vdev_name(NULL, NULL, child[0], B_FALSE)) == NULL)
2404 return (-1);
2405
2406 /*
2407 * If the target is a hot spare that has been swapped in, we can only
2408 * replace it with another hot spare.
2409 */
2410 if (replacing &&
2411 nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_IS_SPARE, &val) == 0 &&
2412 (zpool_find_vdev(zhp, newname, &avail_spare, &l2cache,
2413 NULL) == NULL || !avail_spare) &&
2414 is_replacing_spare(config_root, tgt, 1)) {
2415 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2416 "can only be replaced by another hot spare"));
2417 free(newname);
2418 return (zfs_error(hdl, EZFS_BADTARGET, msg));
2419 }
2420
2421 free(newname);
2422
2423 if (zcmd_write_conf_nvlist(hdl, &zc, nvroot) != 0)
2424 return (-1);
2425
2426 ret = zfs_ioctl(hdl, ZFS_IOC_VDEV_ATTACH, &zc);
2427
2428 zcmd_free_nvlists(&zc);
2429
2430 if (ret == 0) {
2431 if (rootpool) {
2432 /*
2433 * XXX need a better way to prevent user from
2434 * booting up a half-baked vdev.
2435 */
2436 (void) fprintf(stderr, dgettext(TEXT_DOMAIN, "Make "
2437 "sure to wait until resilver is done "
2438 "before rebooting.\n"));
2439 }
2440 return (0);
2441 }
2442
2443 switch (errno) {
2444 case ENOTSUP:
2445 /*
2446 * Can't attach to or replace this type of vdev.
2447 */
2448 if (replacing) {
2449 uint64_t version = zpool_get_prop_int(zhp,
2450 ZPOOL_PROP_VERSION, NULL);
2451
2452 if (islog)
2453 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2454 "cannot replace a log with a spare"));
2455 else if (version >= SPA_VERSION_MULTI_REPLACE)
2456 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2457 "already in replacing/spare config; wait "
2458 "for completion or use 'zpool detach'"));
2459 else
2460 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2461 "cannot replace a replacing device"));
2462 } else {
2463 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2464 "can only attach to mirrors and top-level "
2465 "disks"));
2466 }
2467 (void) zfs_error(hdl, EZFS_BADTARGET, msg);
2468 break;
2469
2470 case EINVAL:
2471 /*
2472 * The new device must be a single disk.
2473 */
2474 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2475 "new device must be a single disk"));
2476 (void) zfs_error(hdl, EZFS_INVALCONFIG, msg);
2477 break;
2478
2479 case EBUSY:
2480 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "%s is busy"),
2481 new_disk);
2482 (void) zfs_error(hdl, EZFS_BADDEV, msg);
2483 break;
2484
2485 case EOVERFLOW:
2486 /*
2487 * The new device is too small.
2488 */
2489 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2490 "device is too small"));
2491 (void) zfs_error(hdl, EZFS_BADDEV, msg);
2492 break;
2493
2494 case EDOM:
2495 /*
2496 * The new device has a different alignment requirement.
2497 */
2498 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2499 "devices have different sector alignment"));
2500 (void) zfs_error(hdl, EZFS_BADDEV, msg);
2501 break;
2502
2503 case ENAMETOOLONG:
2504 /*
2505 * The resulting top-level vdev spec won't fit in the label.
2506 */
2507 (void) zfs_error(hdl, EZFS_DEVOVERFLOW, msg);
2508 break;
2509
2510 default:
2511 (void) zpool_standard_error(hdl, errno, msg);
2512 }
2513
2514 return (-1);
2515 }
2516
2517 /*
2518 * Detach the specified device.
2519 */
2520 int
2521 zpool_vdev_detach(zpool_handle_t *zhp, const char *path)
2522 {
2523 zfs_cmd_t zc = { 0 };
2524 char msg[1024];
2525 nvlist_t *tgt;
2526 boolean_t avail_spare, l2cache;
2527 libzfs_handle_t *hdl = zhp->zpool_hdl;
2528
2529 (void) snprintf(msg, sizeof (msg),
2530 dgettext(TEXT_DOMAIN, "cannot detach %s"), path);
2531
2532 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
2533 if ((tgt = zpool_find_vdev(zhp, path, &avail_spare, &l2cache,
2534 NULL)) == 0)
2535 return (zfs_error(hdl, EZFS_NODEVICE, msg));
2536
2537 if (avail_spare)
2538 return (zfs_error(hdl, EZFS_ISSPARE, msg));
2539
2540 if (l2cache)
2541 return (zfs_error(hdl, EZFS_ISL2CACHE, msg));
2542
2543 verify(nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID, &zc.zc_guid) == 0);
2544
2545 if (zfs_ioctl(hdl, ZFS_IOC_VDEV_DETACH, &zc) == 0)
2546 return (0);
2547
2548 switch (errno) {
2549
2550 case ENOTSUP:
2551 /*
2552 * Can't detach from this type of vdev.
2553 */
2554 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "only "
2555 "applicable to mirror and replacing vdevs"));
2556 (void) zfs_error(hdl, EZFS_BADTARGET, msg);
2557 break;
2558
2559 case EBUSY:
2560 /*
2561 * There are no other replicas of this device.
2562 */
2563 (void) zfs_error(hdl, EZFS_NOREPLICAS, msg);
2564 break;
2565
2566 default:
2567 (void) zpool_standard_error(hdl, errno, msg);
2568 }
2569
2570 return (-1);
2571 }
2572
2573 /*
2574 * Find a mirror vdev in the source nvlist.
2575 *
2576 * The mchild array contains a list of disks in one of the top-level mirrors
2577 * of the source pool. The schild array contains a list of disks that the
2578 * user specified on the command line. We loop over the mchild array to
2579 * see if any entry in the schild array matches.
2580 *
2581 * If a disk in the mchild array is found in the schild array, we return
2582 * the index of that entry. Otherwise we return -1.
2583 */
2584 static int
2585 find_vdev_entry(zpool_handle_t *zhp, nvlist_t **mchild, uint_t mchildren,
2586 nvlist_t **schild, uint_t schildren)
2587 {
2588 uint_t mc;
2589
2590 for (mc = 0; mc < mchildren; mc++) {
2591 uint_t sc;
2592 char *mpath = zpool_vdev_name(zhp->zpool_hdl, zhp,
2593 mchild[mc], B_FALSE);
2594
2595 for (sc = 0; sc < schildren; sc++) {
2596 char *spath = zpool_vdev_name(zhp->zpool_hdl, zhp,
2597 schild[sc], B_FALSE);
2598 boolean_t result = (strcmp(mpath, spath) == 0);
2599
2600 free(spath);
2601 if (result) {
2602 free(mpath);
2603 return (mc);
2604 }
2605 }
2606
2607 free(mpath);
2608 }
2609
2610 return (-1);
2611 }
2612
2613 /*
2614 * Split a mirror pool. If newroot points to null, then a new nvlist
2615 * is generated and it is the responsibility of the caller to free it.
2616 */
2617 int
2618 zpool_vdev_split(zpool_handle_t *zhp, char *newname, nvlist_t **newroot,
2619 nvlist_t *props, splitflags_t flags)
2620 {
2621 zfs_cmd_t zc = { 0 };
2622 char msg[1024];
2623 nvlist_t *tree, *config, **child, **newchild, *newconfig = NULL;
2624 nvlist_t **varray = NULL, *zc_props = NULL;
2625 uint_t c, children, newchildren, lastlog = 0, vcount, found = 0;
2626 libzfs_handle_t *hdl = zhp->zpool_hdl;
2627 uint64_t vers;
2628 boolean_t freelist = B_FALSE, memory_err = B_TRUE;
2629 int retval = 0;
2630
2631 (void) snprintf(msg, sizeof (msg),
2632 dgettext(TEXT_DOMAIN, "Unable to split %s"), zhp->zpool_name);
2633
2634 if (!zpool_name_valid(hdl, B_FALSE, newname))
2635 return (zfs_error(hdl, EZFS_INVALIDNAME, msg));
2636
2637 if ((config = zpool_get_config(zhp, NULL)) == NULL) {
2638 (void) fprintf(stderr, gettext("Internal error: unable to "
2639 "retrieve pool configuration\n"));
2640 return (-1);
2641 }
2642
2643 verify(nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE, &tree)
2644 == 0);
2645 verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_VERSION, &vers) == 0);
2646
2647 if (props) {
2648 prop_flags_t flags = { .create = B_FALSE, .import = B_TRUE };
2649 if ((zc_props = zpool_valid_proplist(hdl, zhp->zpool_name,
2650 props, vers, flags, msg)) == NULL)
2651 return (-1);
2652 }
2653
2654 if (nvlist_lookup_nvlist_array(tree, ZPOOL_CONFIG_CHILDREN, &child,
2655 &children) != 0) {
2656 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2657 "Source pool is missing vdev tree"));
2658 if (zc_props)
2659 nvlist_free(zc_props);
2660 return (-1);
2661 }
2662
2663 varray = zfs_alloc(hdl, children * sizeof (nvlist_t *));
2664 vcount = 0;
2665
2666 if (*newroot == NULL ||
2667 nvlist_lookup_nvlist_array(*newroot, ZPOOL_CONFIG_CHILDREN,
2668 &newchild, &newchildren) != 0)
2669 newchildren = 0;
2670
2671 for (c = 0; c < children; c++) {
2672 uint64_t is_log = B_FALSE, is_hole = B_FALSE;
2673 char *type;
2674 nvlist_t **mchild, *vdev;
2675 uint_t mchildren;
2676 int entry;
2677
2678 /*
2679 * Unlike cache & spares, slogs are stored in the
2680 * ZPOOL_CONFIG_CHILDREN array. We filter them out here.
2681 */
2682 (void) nvlist_lookup_uint64(child[c], ZPOOL_CONFIG_IS_LOG,
2683 &is_log);
2684 (void) nvlist_lookup_uint64(child[c], ZPOOL_CONFIG_IS_HOLE,
2685 &is_hole);
2686 if (is_log || is_hole) {
2687 /*
2688 * Create a hole vdev and put it in the config.
2689 */
2690 if (nvlist_alloc(&vdev, NV_UNIQUE_NAME, 0) != 0)
2691 goto out;
2692 if (nvlist_add_string(vdev, ZPOOL_CONFIG_TYPE,
2693 VDEV_TYPE_HOLE) != 0)
2694 goto out;
2695 if (nvlist_add_uint64(vdev, ZPOOL_CONFIG_IS_HOLE,
2696 1) != 0)
2697 goto out;
2698 if (lastlog == 0)
2699 lastlog = vcount;
2700 varray[vcount++] = vdev;
2701 continue;
2702 }
2703 lastlog = 0;
2704 verify(nvlist_lookup_string(child[c], ZPOOL_CONFIG_TYPE, &type)
2705 == 0);
2706 if (strcmp(type, VDEV_TYPE_MIRROR) != 0) {
2707 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2708 "Source pool must be composed only of mirrors\n"));
2709 retval = zfs_error(hdl, EZFS_INVALCONFIG, msg);
2710 goto out;
2711 }
2712
2713 verify(nvlist_lookup_nvlist_array(child[c],
2714 ZPOOL_CONFIG_CHILDREN, &mchild, &mchildren) == 0);
2715
2716 /* find or add an entry for this top-level vdev */
2717 if (newchildren > 0 &&
2718 (entry = find_vdev_entry(zhp, mchild, mchildren,
2719 newchild, newchildren)) >= 0) {
2720 /* We found a disk that the user specified. */
2721 vdev = mchild[entry];
2722 ++found;
2723 } else {
2724 /* User didn't specify a disk for this vdev. */
2725 vdev = mchild[mchildren - 1];
2726 }
2727
2728 if (nvlist_dup(vdev, &varray[vcount++], 0) != 0)
2729 goto out;
2730 }
2731
2732 /* did we find every disk the user specified? */
2733 if (found != newchildren) {
2734 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "Device list must "
2735 "include at most one disk from each mirror"));
2736 retval = zfs_error(hdl, EZFS_INVALCONFIG, msg);
2737 goto out;
2738 }
2739
2740 /* Prepare the nvlist for populating. */
2741 if (*newroot == NULL) {
2742 if (nvlist_alloc(newroot, NV_UNIQUE_NAME, 0) != 0)
2743 goto out;
2744 freelist = B_TRUE;
2745 if (nvlist_add_string(*newroot, ZPOOL_CONFIG_TYPE,
2746 VDEV_TYPE_ROOT) != 0)
2747 goto out;
2748 } else {
2749 verify(nvlist_remove_all(*newroot, ZPOOL_CONFIG_CHILDREN) == 0);
2750 }
2751
2752 /* Add all the children we found */
2753 if (nvlist_add_nvlist_array(*newroot, ZPOOL_CONFIG_CHILDREN, varray,
2754 lastlog == 0 ? vcount : lastlog) != 0)
2755 goto out;
2756
2757 /*
2758 * If we're just doing a dry run, exit now with success.
2759 */
2760 if (flags.dryrun) {
2761 memory_err = B_FALSE;
2762 freelist = B_FALSE;
2763 goto out;
2764 }
2765
2766 /* now build up the config list & call the ioctl */
2767 if (nvlist_alloc(&newconfig, NV_UNIQUE_NAME, 0) != 0)
2768 goto out;
2769
2770 if (nvlist_add_nvlist(newconfig,
2771 ZPOOL_CONFIG_VDEV_TREE, *newroot) != 0 ||
2772 nvlist_add_string(newconfig,
2773 ZPOOL_CONFIG_POOL_NAME, newname) != 0 ||
2774 nvlist_add_uint64(newconfig, ZPOOL_CONFIG_VERSION, vers) != 0)
2775 goto out;
2776
2777 /*
2778 * The new pool is automatically part of the namespace unless we
2779 * explicitly export it.
2780 */
2781 if (!flags.import)
2782 zc.zc_cookie = ZPOOL_EXPORT_AFTER_SPLIT;
2783 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
2784 (void) strlcpy(zc.zc_string, newname, sizeof (zc.zc_string));
2785 if (zcmd_write_conf_nvlist(hdl, &zc, newconfig) != 0)
2786 goto out;
2787 if (zc_props != NULL && zcmd_write_src_nvlist(hdl, &zc, zc_props) != 0)
2788 goto out;
2789
2790 if (zfs_ioctl(hdl, ZFS_IOC_VDEV_SPLIT, &zc) != 0) {
2791 retval = zpool_standard_error(hdl, errno, msg);
2792 goto out;
2793 }
2794
2795 freelist = B_FALSE;
2796 memory_err = B_FALSE;
2797
2798 out:
2799 if (varray != NULL) {
2800 int v;
2801
2802 for (v = 0; v < vcount; v++)
2803 nvlist_free(varray[v]);
2804 free(varray);
2805 }
2806 zcmd_free_nvlists(&zc);
2807 if (zc_props)
2808 nvlist_free(zc_props);
2809 if (newconfig)
2810 nvlist_free(newconfig);
2811 if (freelist) {
2812 nvlist_free(*newroot);
2813 *newroot = NULL;
2814 }
2815
2816 if (retval != 0)
2817 return (retval);
2818
2819 if (memory_err)
2820 return (no_memory(hdl));
2821
2822 return (0);
2823 }
2824
2825 /*
2826 * Remove the given device. Currently, this is supported only for hot spares
2827 * and level 2 cache devices.
2828 */
2829 int
2830 zpool_vdev_remove(zpool_handle_t *zhp, const char *path)
2831 {
2832 zfs_cmd_t zc = { 0 };
2833 char msg[1024];
2834 nvlist_t *tgt;
2835 boolean_t avail_spare, l2cache, islog;
2836 libzfs_handle_t *hdl = zhp->zpool_hdl;
2837 uint64_t version;
2838
2839 (void) snprintf(msg, sizeof (msg),
2840 dgettext(TEXT_DOMAIN, "cannot remove %s"), path);
2841
2842 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
2843 if ((tgt = zpool_find_vdev(zhp, path, &avail_spare, &l2cache,
2844 &islog)) == 0)
2845 return (zfs_error(hdl, EZFS_NODEVICE, msg));
2846 /*
2847 * XXX - this should just go away.
2848 */
2849 if (!avail_spare && !l2cache && !islog) {
2850 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2851 "only inactive hot spares, cache, top-level, "
2852 "or log devices can be removed"));
2853 return (zfs_error(hdl, EZFS_NODEVICE, msg));
2854 }
2855
2856 version = zpool_get_prop_int(zhp, ZPOOL_PROP_VERSION, NULL);
2857 if (islog && version < SPA_VERSION_HOLES) {
2858 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2859 "pool must be upgrade to support log removal"));
2860 return (zfs_error(hdl, EZFS_BADVERSION, msg));
2861 }
2862
2863 verify(nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID, &zc.zc_guid) == 0);
2864
2865 if (zfs_ioctl(hdl, ZFS_IOC_VDEV_REMOVE, &zc) == 0)
2866 return (0);
2867
2868 return (zpool_standard_error(hdl, errno, msg));
2869 }
2870
2871 /*
2872 * Clear the errors for the pool, or the particular device if specified.
2873 */
2874 int
2875 zpool_clear(zpool_handle_t *zhp, const char *path, nvlist_t *rewindnvl)
2876 {
2877 zfs_cmd_t zc = { 0 };
2878 char msg[1024];
2879 nvlist_t *tgt;
2880 zpool_rewind_policy_t policy;
2881 boolean_t avail_spare, l2cache;
2882 libzfs_handle_t *hdl = zhp->zpool_hdl;
2883 nvlist_t *nvi = NULL;
2884 int error;
2885
2886 if (path)
2887 (void) snprintf(msg, sizeof (msg),
2888 dgettext(TEXT_DOMAIN, "cannot clear errors for %s"),
2889 path);
2890 else
2891 (void) snprintf(msg, sizeof (msg),
2892 dgettext(TEXT_DOMAIN, "cannot clear errors for %s"),
2893 zhp->zpool_name);
2894
2895 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
2896 if (path) {
2897 if ((tgt = zpool_find_vdev(zhp, path, &avail_spare,
2898 &l2cache, NULL)) == 0)
2899 return (zfs_error(hdl, EZFS_NODEVICE, msg));
2900
2901 /*
2902 * Don't allow error clearing for hot spares. Do allow
2903 * error clearing for l2cache devices.
2904 */
2905 if (avail_spare)
2906 return (zfs_error(hdl, EZFS_ISSPARE, msg));
2907
2908 verify(nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID,
2909 &zc.zc_guid) == 0);
2910 }
2911
2912 zpool_get_rewind_policy(rewindnvl, &policy);
2913 zc.zc_cookie = policy.zrp_request;
2914
2915 if (zcmd_alloc_dst_nvlist(hdl, &zc, zhp->zpool_config_size * 2) != 0)
2916 return (-1);
2917
2918 if (zcmd_write_src_nvlist(hdl, &zc, rewindnvl) != 0)
2919 return (-1);
2920
2921 while ((error = zfs_ioctl(hdl, ZFS_IOC_CLEAR, &zc)) != 0 &&
2922 errno == ENOMEM) {
2923 if (zcmd_expand_dst_nvlist(hdl, &zc) != 0) {
2924 zcmd_free_nvlists(&zc);
2925 return (-1);
2926 }
2927 }
2928
2929 if (!error || ((policy.zrp_request & ZPOOL_TRY_REWIND) &&
2930 errno != EPERM && errno != EACCES)) {
2931 if (policy.zrp_request &
2932 (ZPOOL_DO_REWIND | ZPOOL_TRY_REWIND)) {
2933 (void) zcmd_read_dst_nvlist(hdl, &zc, &nvi);
2934 zpool_rewind_exclaim(hdl, zc.zc_name,
2935 ((policy.zrp_request & ZPOOL_TRY_REWIND) != 0),
2936 nvi);
2937 nvlist_free(nvi);
2938 }
2939 zcmd_free_nvlists(&zc);
2940 return (0);
2941 }
2942
2943 zcmd_free_nvlists(&zc);
2944 return (zpool_standard_error(hdl, errno, msg));
2945 }
2946
2947 /*
2948 * Similar to zpool_clear(), but takes a GUID (used by fmd).
2949 */
2950 int
2951 zpool_vdev_clear(zpool_handle_t *zhp, uint64_t guid)
2952 {
2953 zfs_cmd_t zc = { 0 };
2954 char msg[1024];
2955 libzfs_handle_t *hdl = zhp->zpool_hdl;
2956
2957 (void) snprintf(msg, sizeof (msg),
2958 dgettext(TEXT_DOMAIN, "cannot clear errors for %llx"),
2959 guid);
2960
2961 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
2962 zc.zc_guid = guid;
2963 zc.zc_cookie = ZPOOL_NO_REWIND;
2964
2965 if (ioctl(hdl->libzfs_fd, ZFS_IOC_CLEAR, &zc) == 0)
2966 return (0);
2967
2968 return (zpool_standard_error(hdl, errno, msg));
2969 }
2970
2971 /*
2972 * Change the GUID for a pool.
2973 */
2974 int
2975 zpool_reguid(zpool_handle_t *zhp)
2976 {
2977 char msg[1024];
2978 libzfs_handle_t *hdl = zhp->zpool_hdl;
2979 zfs_cmd_t zc = { 0 };
2980
2981 (void) snprintf(msg, sizeof (msg),
2982 dgettext(TEXT_DOMAIN, "cannot reguid '%s'"), zhp->zpool_name);
2983
2984 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
2985 if (zfs_ioctl(hdl, ZFS_IOC_POOL_REGUID, &zc) == 0)
2986 return (0);
2987
2988 return (zpool_standard_error(hdl, errno, msg));
2989 }
2990
2991 /*
2992 * Convert from a devid string to a path.
2993 */
2994 static char *
2995 devid_to_path(char *devid_str)
2996 {
2997 ddi_devid_t devid;
2998 char *minor;
2999 char *path;
3000 devid_nmlist_t *list = NULL;
3001 int ret;
3002
3003 if (devid_str_decode(devid_str, &devid, &minor) != 0)
3004 return (NULL);
3005
3006 ret = devid_deviceid_to_nmlist("/dev", devid, minor, &list);
3007
3008 devid_str_free(minor);
3009 devid_free(devid);
3010
3011 if (ret != 0)
3012 return (NULL);
3013
3014 if ((path = strdup(list[0].devname)) == NULL)
3015 return (NULL);
3016
3017 devid_free_nmlist(list);
3018
3019 return (path);
3020 }
3021
3022 /*
3023 * Convert from a path to a devid string.
3024 */
3025 static char *
3026 path_to_devid(const char *path)
3027 {
3028 int fd;
3029 ddi_devid_t devid;
3030 char *minor, *ret;
3031
3032 if ((fd = open(path, O_RDONLY)) < 0)
3033 return (NULL);
3034
3035 minor = NULL;
3036 ret = NULL;
3037 if (devid_get(fd, &devid) == 0) {
3038 if (devid_get_minor_name(fd, &minor) == 0)
3039 ret = devid_str_encode(devid, minor);
3040 if (minor != NULL)
3041 devid_str_free(minor);
3042 devid_free(devid);
3043 }
3044 (void) close(fd);
3045
3046 return (ret);
3047 }
3048
3049 /*
3050 * Issue the necessary ioctl() to update the stored path value for the vdev. We
3051 * ignore any failure here, since a common case is for an unprivileged user to
3052 * type 'zpool status', and we'll display the correct information anyway.
3053 */
3054 static void
3055 set_path(zpool_handle_t *zhp, nvlist_t *nv, const char *path)
3056 {
3057 zfs_cmd_t zc = { 0 };
3058
3059 (void) strncpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
3060 (void) strncpy(zc.zc_value, path, sizeof (zc.zc_value));
3061 verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID,
3062 &zc.zc_guid) == 0);
3063
3064 (void) ioctl(zhp->zpool_hdl->libzfs_fd, ZFS_IOC_VDEV_SETPATH, &zc);
3065 }
3066
3067 /*
3068 * Given a vdev, return the name to display in iostat. If the vdev has a path,
3069 * we use that, stripping off any leading "/dev/dsk/"; if not, we use the type.
3070 * We also check if this is a whole disk, in which case we strip off the
3071 * trailing 's0' slice name.
3072 *
3073 * This routine is also responsible for identifying when disks have been
3074 * reconfigured in a new location. The kernel will have opened the device by
3075 * devid, but the path will still refer to the old location. To catch this, we
3076 * first do a path -> devid translation (which is fast for the common case). If
3077 * the devid matches, we're done. If not, we do a reverse devid -> path
3078 * translation and issue the appropriate ioctl() to update the path of the vdev.
3079 * If 'zhp' is NULL, then this is an exported pool, and we don't need to do any
3080 * of these checks.
3081 */
3082 char *
3083 zpool_vdev_name(libzfs_handle_t *hdl, zpool_handle_t *zhp, nvlist_t *nv,
3084 boolean_t verbose)
3085 {
3086 char *path, *devid;
3087 uint64_t value;
3088 char buf[64];
3089 vdev_stat_t *vs;
3090 uint_t vsc;
3091
3092 if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_NOT_PRESENT,
3093 &value) == 0) {
3094 verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID,
3095 &value) == 0);
3096 (void) snprintf(buf, sizeof (buf), "%llu",
3097 (u_longlong_t)value);
3098 path = buf;
3099 } else if (nvlist_lookup_string(nv, ZPOOL_CONFIG_PATH, &path) == 0) {
3100
3101 /*
3102 * If the device is dead (faulted, offline, etc) then don't
3103 * bother opening it. Otherwise we may be forcing the user to
3104 * open a misbehaving device, which can have undesirable
3105 * effects.
3106 */
3107 if ((nvlist_lookup_uint64_array(nv, ZPOOL_CONFIG_VDEV_STATS,
3108 (uint64_t **)&vs, &vsc) != 0 ||
3109 vs->vs_state >= VDEV_STATE_DEGRADED) &&
3110 zhp != NULL &&
3111 nvlist_lookup_string(nv, ZPOOL_CONFIG_DEVID, &devid) == 0) {
3112 /*
3113 * Determine if the current path is correct.
3114 */
3115 char *newdevid = path_to_devid(path);
3116
3117 if (newdevid == NULL ||
3118 strcmp(devid, newdevid) != 0) {
3119 char *newpath;
3120
3121 if ((newpath = devid_to_path(devid)) != NULL) {
3122 /*
3123 * Update the path appropriately.
3124 */
3125 set_path(zhp, nv, newpath);
3126 if (nvlist_add_string(nv,
3127 ZPOOL_CONFIG_PATH, newpath) == 0)
3128 verify(nvlist_lookup_string(nv,
3129 ZPOOL_CONFIG_PATH,
3130 &path) == 0);
3131 free(newpath);
3132 }
3133 }
3134
3135 if (newdevid)
3136 devid_str_free(newdevid);
3137 }
3138
3139 if (strncmp(path, "/dev/dsk/", 9) == 0)
3140 path += 9;
3141
3142 if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_WHOLE_DISK,
3143 &value) == 0 && value) {
3144 int pathlen = strlen(path);
3145 char *tmp = zfs_strdup(hdl, path);
3146
3147 /*
3148 * If it starts with c#, and ends with "s0", chop
3149 * the "s0" off, or if it ends with "s0/old", remove
3150 * the "s0" from the middle.
3151 */
3152 if (CTD_CHECK(tmp)) {
3153 if (strcmp(&tmp[pathlen - 2], "s0") == 0) {
3154 tmp[pathlen - 2] = '\0';
3155 } else if (pathlen > 6 &&
3156 strcmp(&tmp[pathlen - 6], "s0/old") == 0) {
3157 (void) strcpy(&tmp[pathlen - 6],
3158 "/old");
3159 }
3160 }
3161 return (tmp);
3162 }
3163 } else {
3164 verify(nvlist_lookup_string(nv, ZPOOL_CONFIG_TYPE, &path) == 0);
3165
3166 /*
3167 * If it's a raidz device, we need to stick in the parity level.
3168 */
3169 if (strcmp(path, VDEV_TYPE_RAIDZ) == 0) {
3170 verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_NPARITY,
3171 &value) == 0);
3172 (void) snprintf(buf, sizeof (buf), "%s%llu", path,
3173 (u_longlong_t)value);
3174 path = buf;
3175 }
3176
3177 /*
3178 * We identify each top-level vdev by using a <type-id>
3179 * naming convention.
3180 */
3181 if (verbose) {
3182 uint64_t id;
3183
3184 verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_ID,
3185 &id) == 0);
3186 (void) snprintf(buf, sizeof (buf), "%s-%llu", path,
3187 (u_longlong_t)id);
3188 path = buf;
3189 }
3190 }
3191
3192 return (zfs_strdup(hdl, path));
3193 }
3194
3195 static int
3196 zbookmark_compare(const void *a, const void *b)
3197 {
3198 return (memcmp(a, b, sizeof (zbookmark_t)));
3199 }
3200
3201 /*
3202 * Retrieve the persistent error log, uniquify the members, and return to the
3203 * caller.
3204 */
3205 int
3206 zpool_get_errlog(zpool_handle_t *zhp, nvlist_t **nverrlistp)
3207 {
3208 zfs_cmd_t zc = { 0 };
3209 uint64_t count;
3210 zbookmark_t *zb = NULL;
3211 int i;
3212
3213 /*
3214 * Retrieve the raw error list from the kernel. If the number of errors
3215 * has increased, allocate more space and continue until we get the
3216 * entire list.
3217 */
3218 verify(nvlist_lookup_uint64(zhp->zpool_config, ZPOOL_CONFIG_ERRCOUNT,
3219 &count) == 0);
3220 if (count == 0)
3221 return (0);
3222 if ((zc.zc_nvlist_dst = (uintptr_t)zfs_alloc(zhp->zpool_hdl,
3223 count * sizeof (zbookmark_t))) == (uintptr_t)NULL)
3224 return (-1);
3225 zc.zc_nvlist_dst_size = count;
3226 (void) strcpy(zc.zc_name, zhp->zpool_name);
3227 for (;;) {
3228 if (ioctl(zhp->zpool_hdl->libzfs_fd, ZFS_IOC_ERROR_LOG,
3229 &zc) != 0) {
3230 free((void *)(uintptr_t)zc.zc_nvlist_dst);
3231 if (errno == ENOMEM) {
3232 count = zc.zc_nvlist_dst_size;
3233 if ((zc.zc_nvlist_dst = (uintptr_t)
3234 zfs_alloc(zhp->zpool_hdl, count *
3235 sizeof (zbookmark_t))) == (uintptr_t)NULL)
3236 return (-1);
3237 } else {
3238 return (-1);
3239 }
3240 } else {
3241 break;
3242 }
3243 }
3244
3245 /*
3246 * Sort the resulting bookmarks. This is a little confusing due to the
3247 * implementation of ZFS_IOC_ERROR_LOG. The bookmarks are copied last
3248 * to first, and 'zc_nvlist_dst_size' indicates the number of boomarks
3249 * _not_ copied as part of the process. So we point the start of our
3250 * array appropriate and decrement the total number of elements.
3251 */
3252 zb = ((zbookmark_t *)(uintptr_t)zc.zc_nvlist_dst) +
3253 zc.zc_nvlist_dst_size;
3254 count -= zc.zc_nvlist_dst_size;
3255
3256 qsort(zb, count, sizeof (zbookmark_t), zbookmark_compare);
3257
3258 verify(nvlist_alloc(nverrlistp, 0, KM_SLEEP) == 0);
3259
3260 /*
3261 * Fill in the nverrlistp with nvlist's of dataset and object numbers.
3262 */
3263 for (i = 0; i < count; i++) {
3264 nvlist_t *nv;
3265
3266 /* ignoring zb_blkid and zb_level for now */
3267 if (i > 0 && zb[i-1].zb_objset == zb[i].zb_objset &&
3268 zb[i-1].zb_object == zb[i].zb_object)
3269 continue;
3270
3271 if (nvlist_alloc(&nv, NV_UNIQUE_NAME, KM_SLEEP) != 0)
3272 goto nomem;
3273 if (nvlist_add_uint64(nv, ZPOOL_ERR_DATASET,
3274 zb[i].zb_objset) != 0) {
3275 nvlist_free(nv);
3276 goto nomem;
3277 }
3278 if (nvlist_add_uint64(nv, ZPOOL_ERR_OBJECT,
3279 zb[i].zb_object) != 0) {
3280 nvlist_free(nv);
3281 goto nomem;
3282 }
3283 if (nvlist_add_nvlist(*nverrlistp, "ejk", nv) != 0) {
3284 nvlist_free(nv);
3285 goto nomem;
3286 }
3287 nvlist_free(nv);
3288 }
3289
3290 free((void *)(uintptr_t)zc.zc_nvlist_dst);
3291 return (0);
3292
3293 nomem:
3294 free((void *)(uintptr_t)zc.zc_nvlist_dst);
3295 return (no_memory(zhp->zpool_hdl));
3296 }
3297
3298 /*
3299 * Upgrade a ZFS pool to the latest on-disk version.
3300 */
3301 int
3302 zpool_upgrade(zpool_handle_t *zhp, uint64_t new_version)
3303 {
3304 zfs_cmd_t zc = { 0 };
3305 libzfs_handle_t *hdl = zhp->zpool_hdl;
3306
3307 (void) strcpy(zc.zc_name, zhp->zpool_name);
3308 zc.zc_cookie = new_version;
3309
3310 if (zfs_ioctl(hdl, ZFS_IOC_POOL_UPGRADE, &zc) != 0)
3311 return (zpool_standard_error_fmt(hdl, errno,
3312 dgettext(TEXT_DOMAIN, "cannot upgrade '%s'"),
3313 zhp->zpool_name));
3314 return (0);
3315 }
3316
3317 void
3318 zpool_set_history_str(const char *subcommand, int argc, char **argv,
3319 char *history_str)
3320 {
3321 int i;
3322
3323 (void) strlcpy(history_str, subcommand, HIS_MAX_RECORD_LEN);
3324 for (i = 1; i < argc; i++) {
3325 if (strlen(history_str) + 1 + strlen(argv[i]) >
3326 HIS_MAX_RECORD_LEN)
3327 break;
3328 (void) strlcat(history_str, " ", HIS_MAX_RECORD_LEN);
3329 (void) strlcat(history_str, argv[i], HIS_MAX_RECORD_LEN);
3330 }
3331 }
3332
3333 /*
3334 * Stage command history for logging.
3335 */
3336 int
3337 zpool_stage_history(libzfs_handle_t *hdl, const char *history_str)
3338 {
3339 if (history_str == NULL)
3340 return (EINVAL);
3341
3342 if (strlen(history_str) > HIS_MAX_RECORD_LEN)
3343 return (EINVAL);
3344
3345 if (hdl->libzfs_log_str != NULL)
3346 free(hdl->libzfs_log_str);
3347
3348 if ((hdl->libzfs_log_str = strdup(history_str)) == NULL)
3349 return (no_memory(hdl));
3350
3351 return (0);
3352 }
3353
3354 /*
3355 * Perform ioctl to get some command history of a pool.
3356 *
3357 * 'buf' is the buffer to fill up to 'len' bytes. 'off' is the
3358 * logical offset of the history buffer to start reading from.
3359 *
3360 * Upon return, 'off' is the next logical offset to read from and
3361 * 'len' is the actual amount of bytes read into 'buf'.
3362 */
3363 static int
3364 get_history(zpool_handle_t *zhp, char *buf, uint64_t *off, uint64_t *len)
3365 {
3366 zfs_cmd_t zc = { 0 };
3367 libzfs_handle_t *hdl = zhp->zpool_hdl;
3368
3369 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
3370
3371 zc.zc_history = (uint64_t)(uintptr_t)buf;
3372 zc.zc_history_len = *len;
3373 zc.zc_history_offset = *off;
3374
3375 if (ioctl(hdl->libzfs_fd, ZFS_IOC_POOL_GET_HISTORY, &zc) != 0) {
3376 switch (errno) {
3377 case EPERM:
3378 return (zfs_error_fmt(hdl, EZFS_PERM,
3379 dgettext(TEXT_DOMAIN,
3380 "cannot show history for pool '%s'"),
3381 zhp->zpool_name));
3382 case ENOENT:
3383 return (zfs_error_fmt(hdl, EZFS_NOHISTORY,
3384 dgettext(TEXT_DOMAIN, "cannot get history for pool "
3385 "'%s'"), zhp->zpool_name));
3386 case ENOTSUP:
3387 return (zfs_error_fmt(hdl, EZFS_BADVERSION,
3388 dgettext(TEXT_DOMAIN, "cannot get history for pool "
3389 "'%s', pool must be upgraded"), zhp->zpool_name));
3390 default:
3391 return (zpool_standard_error_fmt(hdl, errno,
3392 dgettext(TEXT_DOMAIN,
3393 "cannot get history for '%s'"), zhp->zpool_name));
3394 }
3395 }
3396
3397 *len = zc.zc_history_len;
3398 *off = zc.zc_history_offset;
3399
3400 return (0);
3401 }
3402
3403 /*
3404 * Process the buffer of nvlists, unpacking and storing each nvlist record
3405 * into 'records'. 'leftover' is set to the number of bytes that weren't
3406 * processed as there wasn't a complete record.
3407 */
3408 int
3409 zpool_history_unpack(char *buf, uint64_t bytes_read, uint64_t *leftover,
3410 nvlist_t ***records, uint_t *numrecords)
3411 {
3412 uint64_t reclen;
3413 nvlist_t *nv;
3414 int i;
3415
3416 while (bytes_read > sizeof (reclen)) {
3417
3418 /* get length of packed record (stored as little endian) */
3419 for (i = 0, reclen = 0; i < sizeof (reclen); i++)
3420 reclen += (uint64_t)(((uchar_t *)buf)[i]) << (8*i);
3421
3422 if (bytes_read < sizeof (reclen) + reclen)
3423 break;
3424
3425 /* unpack record */
3426 if (nvlist_unpack(buf + sizeof (reclen), reclen, &nv, 0) != 0)
3427 return (ENOMEM);
3428 bytes_read -= sizeof (reclen) + reclen;
3429 buf += sizeof (reclen) + reclen;
3430
3431 /* add record to nvlist array */
3432 (*numrecords)++;
3433 if (ISP2(*numrecords + 1)) {
3434 *records = realloc(*records,
3435 *numrecords * 2 * sizeof (nvlist_t *));
3436 }
3437 (*records)[*numrecords - 1] = nv;
3438 }
3439
3440 *leftover = bytes_read;
3441 return (0);
3442 }
3443
3444 #define HIS_BUF_LEN (128*1024)
3445
3446 /*
3447 * Retrieve the command history of a pool.
3448 */
3449 int
3450 zpool_get_history(zpool_handle_t *zhp, nvlist_t **nvhisp)
3451 {
3452 char buf[HIS_BUF_LEN];
3453 uint64_t off = 0;
3454 nvlist_t **records = NULL;
3455 uint_t numrecords = 0;
3456 int err, i;
3457
3458 do {
3459 uint64_t bytes_read = sizeof (buf);
3460 uint64_t leftover;
3461
3462 if ((err = get_history(zhp, buf, &off, &bytes_read)) != 0)
3463 break;
3464
3465 /* if nothing else was read in, we're at EOF, just return */
3466 if (!bytes_read)
3467 break;
3468
3469 if ((err = zpool_history_unpack(buf, bytes_read,
3470 &leftover, &records, &numrecords)) != 0)
3471 break;
3472 off -= leftover;
3473
3474 /* CONSTCOND */
3475 } while (1);
3476
3477 if (!err) {
3478 verify(nvlist_alloc(nvhisp, NV_UNIQUE_NAME, 0) == 0);
3479 verify(nvlist_add_nvlist_array(*nvhisp, ZPOOL_HIST_RECORD,
3480 records, numrecords) == 0);
3481 }
3482 for (i = 0; i < numrecords; i++)
3483 nvlist_free(records[i]);
3484 free(records);
3485
3486 return (err);
3487 }
3488
3489 void
3490 zpool_obj_to_path(zpool_handle_t *zhp, uint64_t dsobj, uint64_t obj,
3491 char *pathname, size_t len)
3492 {
3493 zfs_cmd_t zc = { 0 };
3494 boolean_t mounted = B_FALSE;
3495 char *mntpnt = NULL;
3496 char dsname[MAXNAMELEN];
3497
3498 if (dsobj == 0) {
3499 /* special case for the MOS */
3500 (void) snprintf(pathname, len, "<metadata>:<0x%llx>", obj);
3501 return;
3502 }
3503
3504 /* get the dataset's name */
3505 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
3506 zc.zc_obj = dsobj;
3507 if (ioctl(zhp->zpool_hdl->libzfs_fd,
3508 ZFS_IOC_DSOBJ_TO_DSNAME, &zc) != 0) {
3509 /* just write out a path of two object numbers */
3510 (void) snprintf(pathname, len, "<0x%llx>:<0x%llx>",
3511 dsobj, obj);
3512 return;
3513 }
3514 (void) strlcpy(dsname, zc.zc_value, sizeof (dsname));
3515
3516 /* find out if the dataset is mounted */
3517 mounted = is_mounted(zhp->zpool_hdl, dsname, &mntpnt);
3518
3519 /* get the corrupted object's path */
3520 (void) strlcpy(zc.zc_name, dsname, sizeof (zc.zc_name));
3521 zc.zc_obj = obj;
3522 if (ioctl(zhp->zpool_hdl->libzfs_fd, ZFS_IOC_OBJ_TO_PATH,
3523 &zc) == 0) {
3524 if (mounted) {
3525 (void) snprintf(pathname, len, "%s%s", mntpnt,
3526 zc.zc_value);
3527 } else {
3528 (void) snprintf(pathname, len, "%s:%s",
3529 dsname, zc.zc_value);
3530 }
3531 } else {
3532 (void) snprintf(pathname, len, "%s:<0x%llx>", dsname, obj);
3533 }
3534 free(mntpnt);
3535 }
3536
3537 /*
3538 * Read the EFI label from the config, if a label does not exist then
3539 * pass back the error to the caller. If the caller has passed a non-NULL
3540 * diskaddr argument then we set it to the starting address of the EFI
3541 * partition.
3542 */
3543 static int
3544 read_efi_label(nvlist_t *config, diskaddr_t *sb)
3545 {
3546 char *path;
3547 int fd;
3548 char diskname[MAXPATHLEN];
3549 int err = -1;
3550
3551 if (nvlist_lookup_string(config, ZPOOL_CONFIG_PATH, &path) != 0)
3552 return (err);
3553
3554 (void) snprintf(diskname, sizeof (diskname), "%s%s", RDISK_ROOT,
3555 strrchr(path, '/'));
3556 if ((fd = open(diskname, O_RDONLY|O_NDELAY)) >= 0) {
3557 struct dk_gpt *vtoc;
3558
3559 if ((err = efi_alloc_and_read(fd, &vtoc)) >= 0) {
3560 if (sb != NULL)
3561 *sb = vtoc->efi_parts[0].p_start;
3562 efi_free(vtoc);
3563 }
3564 (void) close(fd);
3565 }
3566 return (err);
3567 }
3568
3569 /*
3570 * determine where a partition starts on a disk in the current
3571 * configuration
3572 */
3573 static diskaddr_t
3574 find_start_block(nvlist_t *config)
3575 {
3576 nvlist_t **child;
3577 uint_t c, children;
3578 diskaddr_t sb = MAXOFFSET_T;
3579 uint64_t wholedisk;
3580
3581 if (nvlist_lookup_nvlist_array(config,
3582 ZPOOL_CONFIG_CHILDREN, &child, &children) != 0) {
3583 if (nvlist_lookup_uint64(config,
3584 ZPOOL_CONFIG_WHOLE_DISK,
3585 &wholedisk) != 0 || !wholedisk) {
3586 return (MAXOFFSET_T);
3587 }
3588 if (read_efi_label(config, &sb) < 0)
3589 sb = MAXOFFSET_T;
3590 return (sb);
3591 }
3592
3593 for (c = 0; c < children; c++) {
3594 sb = find_start_block(child[c]);
3595 if (sb != MAXOFFSET_T) {
3596 return (sb);
3597 }
3598 }
3599 return (MAXOFFSET_T);
3600 }
3601
3602 /*
3603 * Label an individual disk. The name provided is the short name,
3604 * stripped of any leading /dev path.
3605 */
3606 int
3607 zpool_label_disk(libzfs_handle_t *hdl, zpool_handle_t *zhp, char *name)
3608 {
3609 char path[MAXPATHLEN];
3610 struct dk_gpt *vtoc;
3611 int fd;
3612 size_t resv = EFI_MIN_RESV_SIZE;
3613 uint64_t slice_size;
3614 diskaddr_t start_block;
3615 char errbuf[1024];
3616
3617 /* prepare an error message just in case */
3618 (void) snprintf(errbuf, sizeof (errbuf),
3619 dgettext(TEXT_DOMAIN, "cannot label '%s'"), name);
3620
3621 if (zhp) {
3622 nvlist_t *nvroot;
3623
3624 if (pool_is_bootable(zhp)) {
3625 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3626 "EFI labeled devices are not supported on root "
3627 "pools."));
3628 return (zfs_error(hdl, EZFS_POOL_NOTSUP, errbuf));
3629 }
3630
3631 verify(nvlist_lookup_nvlist(zhp->zpool_config,
3632 ZPOOL_CONFIG_VDEV_TREE, &nvroot) == 0);
3633
3634 if (zhp->zpool_start_block == 0)
3635 start_block = find_start_block(nvroot);
3636 else
3637 start_block = zhp->zpool_start_block;
3638 zhp->zpool_start_block = start_block;
3639 } else {
3640 /* new pool */
3641 start_block = NEW_START_BLOCK;
3642 }
3643
3644 (void) snprintf(path, sizeof (path), "%s/%s%s", RDISK_ROOT, name,
3645 BACKUP_SLICE);
3646
3647 if ((fd = open(path, O_RDWR | O_NDELAY)) < 0) {
3648 /*
3649 * This shouldn't happen. We've long since verified that this
3650 * is a valid device.
3651 */
3652 zfs_error_aux(hdl,
3653 dgettext(TEXT_DOMAIN, "unable to open device"));
3654 return (zfs_error(hdl, EZFS_OPENFAILED, errbuf));
3655 }
3656
3657 if (efi_alloc_and_init(fd, EFI_NUMPAR, &vtoc) != 0) {
3658 /*
3659 * The only way this can fail is if we run out of memory, or we
3660 * were unable to read the disk's capacity
3661 */
3662 if (errno == ENOMEM)
3663 (void) no_memory(hdl);
3664
3665 (void) close(fd);
3666 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3667 "unable to read disk capacity"), name);
3668
3669 return (zfs_error(hdl, EZFS_NOCAP, errbuf));
3670 }
3671
3672 slice_size = vtoc->efi_last_u_lba + 1;
3673 slice_size -= EFI_MIN_RESV_SIZE;
3674 if (start_block == MAXOFFSET_T)
3675 start_block = NEW_START_BLOCK;
3676 slice_size -= start_block;
3677
3678 vtoc->efi_parts[0].p_start = start_block;
3679 vtoc->efi_parts[0].p_size = slice_size;
3680
3681 /*
3682 * Why we use V_USR: V_BACKUP confuses users, and is considered
3683 * disposable by some EFI utilities (since EFI doesn't have a backup
3684 * slice). V_UNASSIGNED is supposed to be used only for zero size
3685 * partitions, and efi_write() will fail if we use it. V_ROOT, V_BOOT,
3686 * etc. were all pretty specific. V_USR is as close to reality as we
3687 * can get, in the absence of V_OTHER.
3688 */
3689 vtoc->efi_parts[0].p_tag = V_USR;
3690 (void) strcpy(vtoc->efi_parts[0].p_name, "zfs");
3691
3692 vtoc->efi_parts[8].p_start = slice_size + start_block;
3693 vtoc->efi_parts[8].p_size = resv;
3694 vtoc->efi_parts[8].p_tag = V_RESERVED;
3695
3696 if (efi_write(fd, vtoc) != 0) {
3697 /*
3698 * Some block drivers (like pcata) may not support EFI
3699 * GPT labels. Print out a helpful error message dir-
3700 * ecting the user to manually label the disk and give
3701 * a specific slice.
3702 */
3703 (void) close(fd);
3704 efi_free(vtoc);
3705
3706 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3707 "try using fdisk(1M) and then provide a specific slice"));
3708 return (zfs_error(hdl, EZFS_LABELFAILED, errbuf));
3709 }
3710
3711 (void) close(fd);
3712 efi_free(vtoc);
3713 return (0);
3714 }
3715
3716 static boolean_t
3717 supported_dump_vdev_type(libzfs_handle_t *hdl, nvlist_t *config, char *errbuf)
3718 {
3719 char *type;
3720 nvlist_t **child;
3721 uint_t children, c;
3722
3723 verify(nvlist_lookup_string(config, ZPOOL_CONFIG_TYPE, &type) == 0);
3724 if (strcmp(type, VDEV_TYPE_RAIDZ) == 0 ||
3725 strcmp(type, VDEV_TYPE_FILE) == 0 ||
3726 strcmp(type, VDEV_TYPE_LOG) == 0 ||
3727 strcmp(type, VDEV_TYPE_HOLE) == 0 ||
3728 strcmp(type, VDEV_TYPE_MISSING) == 0) {
3729 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3730 "vdev type '%s' is not supported"), type);
3731 (void) zfs_error(hdl, EZFS_VDEVNOTSUP, errbuf);
3732 return (B_FALSE);
3733 }
3734 if (nvlist_lookup_nvlist_array(config, ZPOOL_CONFIG_CHILDREN,
3735 &child, &children) == 0) {
3736 for (c = 0; c < children; c++) {
3737 if (!supported_dump_vdev_type(hdl, child[c], errbuf))
3738 return (B_FALSE);
3739 }
3740 }
3741 return (B_TRUE);
3742 }
3743
3744 /*
3745 * check if this zvol is allowable for use as a dump device; zero if
3746 * it is, > 0 if it isn't, < 0 if it isn't a zvol
3747 */
3748 int
3749 zvol_check_dump_config(char *arg)
3750 {
3751 zpool_handle_t *zhp = NULL;
3752 nvlist_t *config, *nvroot;
3753 char *p, *volname;
3754 nvlist_t **top;
3755 uint_t toplevels;
3756 libzfs_handle_t *hdl;
3757 char errbuf[1024];
3758 char poolname[ZPOOL_MAXNAMELEN];
3759 int pathlen = strlen(ZVOL_FULL_DEV_DIR);
3760 int ret = 1;
3761
3762 if (strncmp(arg, ZVOL_FULL_DEV_DIR, pathlen)) {
3763 return (-1);
3764 }
3765
3766 (void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN,
3767 "dump is not supported on device '%s'"), arg);
3768
3769 if ((hdl = libzfs_init()) == NULL)
3770 return (1);
3771 libzfs_print_on_error(hdl, B_TRUE);
3772
3773 volname = arg + pathlen;
3774
3775 /* check the configuration of the pool */
3776 if ((p = strchr(volname, '/')) == NULL) {
3777 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3778 "malformed dataset name"));
3779 (void) zfs_error(hdl, EZFS_INVALIDNAME, errbuf);
3780 return (1);
3781 } else if (p - volname >= ZFS_MAXNAMELEN) {
3782 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3783 "dataset name is too long"));
3784 (void) zfs_error(hdl, EZFS_NAMETOOLONG, errbuf);
3785 return (1);
3786 } else {
3787 (void) strncpy(poolname, volname, p - volname);
3788 poolname[p - volname] = '\0';
3789 }
3790
3791 if ((zhp = zpool_open(hdl, poolname)) == NULL) {
3792 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3793 "could not open pool '%s'"), poolname);
3794 (void) zfs_error(hdl, EZFS_OPENFAILED, errbuf);
3795 goto out;
3796 }
3797 config = zpool_get_config(zhp, NULL);
3798 if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE,
3799 &nvroot) != 0) {
3800 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3801 "could not obtain vdev configuration for '%s'"), poolname);
3802 (void) zfs_error(hdl, EZFS_INVALCONFIG, errbuf);
3803 goto out;
3804 }
3805
3806 verify(nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_CHILDREN,
3807 &top, &toplevels) == 0);
3808 if (toplevels != 1) {
3809 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3810 "'%s' has multiple top level vdevs"), poolname);
3811 (void) zfs_error(hdl, EZFS_DEVOVERFLOW, errbuf);
3812 goto out;
3813 }
3814
3815 if (!supported_dump_vdev_type(hdl, top[0], errbuf)) {
3816 goto out;
3817 }
3818 ret = 0;
3819
3820 out:
3821 if (zhp)
3822 zpool_close(zhp);
3823 libzfs_fini(hdl);
3824 return (ret);
3825 }