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 <sys/types.h>
29#include <sys/stat.h>
30#include <ctype.h>
31#include <errno.h>
32#include <devid.h>
33#include <fcntl.h>
34#include <libintl.h>
35#include <stdio.h>
36#include <stdlib.h>
37#include <strings.h>
38#include <unistd.h>
39#include <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
47static 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
53typedef 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
64static int
65zpool_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
97static int
98zpool_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
111static char *
112zpool_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
137uint64_t
138zpool_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 */
179const char *
180zpool_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 * Map POOL STATE to printed strings.
208 */
209const char *
210zpool_pool_state_to_name(pool_state_t state)
211{
212 switch (state) {
213 case POOL_STATE_ACTIVE:
214 return (gettext("ACTIVE"));
215 case POOL_STATE_EXPORTED:
216 return (gettext("EXPORTED"));
217 case POOL_STATE_DESTROYED:
218 return (gettext("DESTROYED"));
219 case POOL_STATE_SPARE:
220 return (gettext("SPARE"));
221 case POOL_STATE_L2CACHE:
222 return (gettext("L2CACHE"));
223 case POOL_STATE_UNINITIALIZED:
224 return (gettext("UNINITIALIZED"));
225 case POOL_STATE_UNAVAIL:
226 return (gettext("UNAVAIL"));
227 case POOL_STATE_POTENTIALLY_ACTIVE:
228 return (gettext("POTENTIALLY_ACTIVE"));
229 }
230
231 return (gettext("UNKNOWN"));
232}
233
234/*
235 * Get a zpool property value for 'prop' and return the value in
236 * a pre-allocated buffer.
237 */
238int
239zpool_get_prop(zpool_handle_t *zhp, zpool_prop_t prop, char *buf, size_t len,
240 zprop_source_t *srctype)
241{
242 uint64_t intval;
243 const char *strval;
244 zprop_source_t src = ZPROP_SRC_NONE;
245 nvlist_t *nvroot;
246 vdev_stat_t *vs;
247 uint_t vsc;
248
249 if (zpool_get_state(zhp) == POOL_STATE_UNAVAIL) {
250 switch (prop) {
251 case ZPOOL_PROP_NAME:
252 (void) strlcpy(buf, zpool_get_name(zhp), len);
253 break;
254
255 case ZPOOL_PROP_HEALTH:
256 (void) strlcpy(buf, "FAULTED", len);
257 break;
258
259 case ZPOOL_PROP_GUID:
260 intval = zpool_get_prop_int(zhp, prop, &src);
261 (void) snprintf(buf, len, "%llu", intval);
262 break;
263
264 case ZPOOL_PROP_ALTROOT:
265 case ZPOOL_PROP_CACHEFILE:
266 case ZPOOL_PROP_COMMENT:
267 if (zhp->zpool_props != NULL ||
268 zpool_get_all_props(zhp) == 0) {
269 (void) strlcpy(buf,
270 zpool_get_prop_string(zhp, prop, &src),
271 len);
272 if (srctype != NULL)
273 *srctype = src;
274 return (0);
275 }
276 /* FALLTHROUGH */
277 default:
278 (void) strlcpy(buf, "-", len);
279 break;
280 }
281
282 if (srctype != NULL)
283 *srctype = src;
284 return (0);
285 }
286
287 if (zhp->zpool_props == NULL && zpool_get_all_props(zhp) &&
288 prop != ZPOOL_PROP_NAME)
289 return (-1);
290
291 switch (zpool_prop_get_type(prop)) {
292 case PROP_TYPE_STRING:
293 (void) strlcpy(buf, zpool_get_prop_string(zhp, prop, &src),
294 len);
295 break;
296
297 case PROP_TYPE_NUMBER:
298 intval = zpool_get_prop_int(zhp, prop, &src);
299
300 switch (prop) {
301 case ZPOOL_PROP_SIZE:
302 case ZPOOL_PROP_ALLOCATED:
303 case ZPOOL_PROP_FREE:
304 (void) zfs_nicenum(intval, buf, len);
305 break;
306
307 case ZPOOL_PROP_CAPACITY:
308 (void) snprintf(buf, len, "%llu%%",
309 (u_longlong_t)intval);
310 break;
311
312 case ZPOOL_PROP_DEDUPRATIO:
313 (void) snprintf(buf, len, "%llu.%02llux",
314 (u_longlong_t)(intval / 100),
315 (u_longlong_t)(intval % 100));
316 break;
317
318 case ZPOOL_PROP_HEALTH:
319 verify(nvlist_lookup_nvlist(zpool_get_config(zhp, NULL),
320 ZPOOL_CONFIG_VDEV_TREE, &nvroot) == 0);
321 verify(nvlist_lookup_uint64_array(nvroot,
322 ZPOOL_CONFIG_VDEV_STATS, (uint64_t **)&vs, &vsc)
323 == 0);
324
325 (void) strlcpy(buf, zpool_state_to_name(intval,
326 vs->vs_aux), len);
327 break;
328 default:
329 (void) snprintf(buf, len, "%llu", intval);
330 }
331 break;
332
333 case PROP_TYPE_INDEX:
334 intval = zpool_get_prop_int(zhp, prop, &src);
335 if (zpool_prop_index_to_string(prop, intval, &strval)
336 != 0)
337 return (-1);
338 (void) strlcpy(buf, strval, len);
339 break;
340
341 default:
342 abort();
343 }
344
345 if (srctype)
346 *srctype = src;
347
348 return (0);
349}
350
351/*
352 * Check if the bootfs name has the same pool name as it is set to.
353 * Assuming bootfs is a valid dataset name.
354 */
355static boolean_t
356bootfs_name_valid(const char *pool, char *bootfs)
357{
358 int len = strlen(pool);
359
360 if (!zfs_name_valid(bootfs, ZFS_TYPE_FILESYSTEM|ZFS_TYPE_SNAPSHOT))
361 return (B_FALSE);
362
363 if (strncmp(pool, bootfs, len) == 0 &&
364 (bootfs[len] == '/' || bootfs[len] == '\0'))
365 return (B_TRUE);
366
367 return (B_FALSE);
368}
369
370/*
371 * Inspect the configuration to determine if any of the devices contain
372 * an EFI label.
373 */
374static boolean_t
375pool_uses_efi(nvlist_t *config)
376{
377#ifdef sun
378 nvlist_t **child;
379 uint_t c, children;
380
381 if (nvlist_lookup_nvlist_array(config, ZPOOL_CONFIG_CHILDREN,
382 &child, &children) != 0)
383 return (read_efi_label(config, NULL) >= 0);
384
385 for (c = 0; c < children; c++) {
386 if (pool_uses_efi(child[c]))
387 return (B_TRUE);
388 }
389#endif /* sun */
390 return (B_FALSE);
391}
392
393static boolean_t
394pool_is_bootable(zpool_handle_t *zhp)
395{
396 char bootfs[ZPOOL_MAXNAMELEN];
397
398 return (zpool_get_prop(zhp, ZPOOL_PROP_BOOTFS, bootfs,
399 sizeof (bootfs), NULL) == 0 && strncmp(bootfs, "-",
400 sizeof (bootfs)) != 0);
401}
402
403
404/*
405 * Given an nvlist of zpool properties to be set, validate that they are
406 * correct, and parse any numeric properties (index, boolean, etc) if they are
407 * specified as strings.
408 */
409static nvlist_t *
410zpool_valid_proplist(libzfs_handle_t *hdl, const char *poolname,
411 nvlist_t *props, uint64_t version, prop_flags_t flags, char *errbuf)
412{
413 nvpair_t *elem;
414 nvlist_t *retprops;
415 zpool_prop_t prop;
416 char *strval;
417 uint64_t intval;
418 char *slash, *check;
419 struct stat64 statbuf;
420 zpool_handle_t *zhp;
421 nvlist_t *nvroot;
422
423 if (nvlist_alloc(&retprops, NV_UNIQUE_NAME, 0) != 0) {
424 (void) no_memory(hdl);
425 return (NULL);
426 }
427
428 elem = NULL;
429 while ((elem = nvlist_next_nvpair(props, elem)) != NULL) {
430 const char *propname = nvpair_name(elem);
431
432 /*
433 * Make sure this property is valid and applies to this type.
434 */
435 if ((prop = zpool_name_to_prop(propname)) == ZPROP_INVAL) {
436 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
437 "invalid property '%s'"), propname);
438 (void) zfs_error(hdl, EZFS_BADPROP, errbuf);
439 goto error;
440 }
441
442 if (zpool_prop_readonly(prop)) {
443 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "'%s' "
444 "is readonly"), propname);
445 (void) zfs_error(hdl, EZFS_PROPREADONLY, errbuf);
446 goto error;
447 }
448
449 if (zprop_parse_value(hdl, elem, prop, ZFS_TYPE_POOL, retprops,
450 &strval, &intval, errbuf) != 0)
451 goto error;
452
453 /*
454 * Perform additional checking for specific properties.
455 */
456 switch (prop) {
457 case ZPOOL_PROP_VERSION:
458 if (intval < version || intval > SPA_VERSION) {
459 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
460 "property '%s' number %d is invalid."),
461 propname, intval);
462 (void) zfs_error(hdl, EZFS_BADVERSION, errbuf);
463 goto error;
464 }
465 break;
466
467 case ZPOOL_PROP_BOOTFS:
468 if (flags.create || flags.import) {
469 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
470 "property '%s' cannot be set at creation "
471 "or import time"), propname);
472 (void) zfs_error(hdl, EZFS_BADPROP, errbuf);
473 goto error;
474 }
475
476 if (version < SPA_VERSION_BOOTFS) {
477 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
478 "pool must be upgraded to support "
479 "'%s' property"), propname);
480 (void) zfs_error(hdl, EZFS_BADVERSION, errbuf);
481 goto error;
482 }
483
484 /*
485 * bootfs property value has to be a dataset name and
486 * the dataset has to be in the same pool as it sets to.
487 */
488 if (strval[0] != '\0' && !bootfs_name_valid(poolname,
489 strval)) {
490 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "'%s' "
491 "is an invalid name"), strval);
492 (void) zfs_error(hdl, EZFS_INVALIDNAME, errbuf);
493 goto error;
494 }
495
496 if ((zhp = zpool_open_canfail(hdl, poolname)) == NULL) {
497 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
498 "could not open pool '%s'"), poolname);
499 (void) zfs_error(hdl, EZFS_OPENFAILED, errbuf);
500 goto error;
501 }
502 verify(nvlist_lookup_nvlist(zpool_get_config(zhp, NULL),
503 ZPOOL_CONFIG_VDEV_TREE, &nvroot) == 0);
504
505#ifdef sun
506 /*
507 * bootfs property cannot be set on a disk which has
508 * been EFI labeled.
509 */
510 if (pool_uses_efi(nvroot)) {
511 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
512 "property '%s' not supported on "
513 "EFI labeled devices"), propname);
514 (void) zfs_error(hdl, EZFS_POOL_NOTSUP, errbuf);
515 zpool_close(zhp);
516 goto error;
517 }
518#endif /* sun */
519 zpool_close(zhp);
520 break;
521
522 case ZPOOL_PROP_ALTROOT:
523 if (!flags.create && !flags.import) {
524 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
525 "property '%s' can only be set during pool "
526 "creation or import"), propname);
527 (void) zfs_error(hdl, EZFS_BADPROP, errbuf);
528 goto error;
529 }
530
531 if (strval[0] != '/') {
532 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
533 "bad alternate root '%s'"), strval);
534 (void) zfs_error(hdl, EZFS_BADPATH, errbuf);
535 goto error;
536 }
537 break;
538
539 case ZPOOL_PROP_CACHEFILE:
540 if (strval[0] == '\0')
541 break;
542
543 if (strcmp(strval, "none") == 0)
544 break;
545
546 if (strval[0] != '/') {
547 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
548 "property '%s' must be empty, an "
549 "absolute path, or 'none'"), propname);
550 (void) zfs_error(hdl, EZFS_BADPATH, errbuf);
551 goto error;
552 }
553
554 slash = strrchr(strval, '/');
555
556 if (slash[1] == '\0' || strcmp(slash, "/.") == 0 ||
557 strcmp(slash, "/..") == 0) {
558 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
559 "'%s' is not a valid file"), strval);
560 (void) zfs_error(hdl, EZFS_BADPATH, errbuf);
561 goto error;
562 }
563
564 *slash = '\0';
565
566 if (strval[0] != '\0' &&
567 (stat64(strval, &statbuf) != 0 ||
568 !S_ISDIR(statbuf.st_mode))) {
569 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
570 "'%s' is not a valid directory"),
571 strval);
572 (void) zfs_error(hdl, EZFS_BADPATH, errbuf);
573 goto error;
574 }
575
576 *slash = '/';
577 break;
578
579 case ZPOOL_PROP_COMMENT:
580 for (check = strval; *check != '\0'; check++) {
581 if (!isprint(*check)) {
582 zfs_error_aux(hdl,
583 dgettext(TEXT_DOMAIN,
584 "comment may only have printable "
585 "characters"));
586 (void) zfs_error(hdl, EZFS_BADPROP,
587 errbuf);
588 goto error;
589 }
590 }
591 if (strlen(strval) > ZPROP_MAX_COMMENT) {
592 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
593 "comment must not exceed %d characters"),
594 ZPROP_MAX_COMMENT);
595 (void) zfs_error(hdl, EZFS_BADPROP, errbuf);
596 goto error;
597 }
598 break;
599 case ZPOOL_PROP_READONLY:
600 if (!flags.import) {
601 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
602 "property '%s' can only be set at "
603 "import time"), propname);
604 (void) zfs_error(hdl, EZFS_BADPROP, errbuf);
605 goto error;
606 }
607 break;
608 }
609 }
610
611 return (retprops);
612error:
613 nvlist_free(retprops);
614 return (NULL);
615}
616
617/*
618 * Set zpool property : propname=propval.
619 */
620int
621zpool_set_prop(zpool_handle_t *zhp, const char *propname, const char *propval)
622{
623 zfs_cmd_t zc = { 0 };
624 int ret = -1;
625 char errbuf[1024];
626 nvlist_t *nvl = NULL;
627 nvlist_t *realprops;
628 uint64_t version;
629 prop_flags_t flags = { 0 };
630
631 (void) snprintf(errbuf, sizeof (errbuf),
632 dgettext(TEXT_DOMAIN, "cannot set property for '%s'"),
633 zhp->zpool_name);
634
635 if (nvlist_alloc(&nvl, NV_UNIQUE_NAME, 0) != 0)
636 return (no_memory(zhp->zpool_hdl));
637
638 if (nvlist_add_string(nvl, propname, propval) != 0) {
639 nvlist_free(nvl);
640 return (no_memory(zhp->zpool_hdl));
641 }
642
643 version = zpool_get_prop_int(zhp, ZPOOL_PROP_VERSION, NULL);
644 if ((realprops = zpool_valid_proplist(zhp->zpool_hdl,
645 zhp->zpool_name, nvl, version, flags, errbuf)) == NULL) {
646 nvlist_free(nvl);
647 return (-1);
648 }
649
650 nvlist_free(nvl);
651 nvl = realprops;
652
653 /*
654 * Execute the corresponding ioctl() to set this property.
655 */
656 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
657
658 if (zcmd_write_src_nvlist(zhp->zpool_hdl, &zc, nvl) != 0) {
659 nvlist_free(nvl);
660 return (-1);
661 }
662
663 ret = zfs_ioctl(zhp->zpool_hdl, ZFS_IOC_POOL_SET_PROPS, &zc);
664
665 zcmd_free_nvlists(&zc);
666 nvlist_free(nvl);
667
668 if (ret)
669 (void) zpool_standard_error(zhp->zpool_hdl, errno, errbuf);
670 else
671 (void) zpool_props_refresh(zhp);
672
673 return (ret);
674}
675
676int
677zpool_expand_proplist(zpool_handle_t *zhp, zprop_list_t **plp)
678{
679 libzfs_handle_t *hdl = zhp->zpool_hdl;
680 zprop_list_t *entry;
681 char buf[ZFS_MAXPROPLEN];
682
683 if (zprop_expand_list(hdl, plp, ZFS_TYPE_POOL) != 0)
684 return (-1);
685
686 for (entry = *plp; entry != NULL; entry = entry->pl_next) {
687
688 if (entry->pl_fixed)
689 continue;
690
691 if (entry->pl_prop != ZPROP_INVAL &&
692 zpool_get_prop(zhp, entry->pl_prop, buf, sizeof (buf),
693 NULL) == 0) {
694 if (strlen(buf) > entry->pl_width)
695 entry->pl_width = strlen(buf);
696 }
697 }
698
699 return (0);
700}
701
702
703/*
704 * Don't start the slice at the default block of 34; many storage
705 * devices will use a stripe width of 128k, so start there instead.
706 */
707#define NEW_START_BLOCK 256
708
709/*
710 * Validate the given pool name, optionally putting an extended error message in
711 * 'buf'.
712 */
713boolean_t
714zpool_name_valid(libzfs_handle_t *hdl, boolean_t isopen, const char *pool)
715{
716 namecheck_err_t why;
717 char what;
718 int ret;
719
720 ret = pool_namecheck(pool, &why, &what);
721
722 /*
723 * The rules for reserved pool names were extended at a later point.
724 * But we need to support users with existing pools that may now be
725 * invalid. So we only check for this expanded set of names during a
726 * create (or import), and only in userland.
727 */
728 if (ret == 0 && !isopen &&
729 (strncmp(pool, "mirror", 6) == 0 ||
730 strncmp(pool, "raidz", 5) == 0 ||
731 strncmp(pool, "spare", 5) == 0 ||
732 strcmp(pool, "log") == 0)) {
733 if (hdl != NULL)
734 zfs_error_aux(hdl,
735 dgettext(TEXT_DOMAIN, "name is reserved"));
736 return (B_FALSE);
737 }
738
739
740 if (ret != 0) {
741 if (hdl != NULL) {
742 switch (why) {
743 case NAME_ERR_TOOLONG:
744 zfs_error_aux(hdl,
745 dgettext(TEXT_DOMAIN, "name is too long"));
746 break;
747
748 case NAME_ERR_INVALCHAR:
749 zfs_error_aux(hdl,
750 dgettext(TEXT_DOMAIN, "invalid character "
751 "'%c' in pool name"), what);
752 break;
753
754 case NAME_ERR_NOLETTER:
755 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
756 "name must begin with a letter"));
757 break;
758
759 case NAME_ERR_RESERVED:
760 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
761 "name is reserved"));
762 break;
763
764 case NAME_ERR_DISKLIKE:
765 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
766 "pool name is reserved"));
767 break;
768
769 case NAME_ERR_LEADING_SLASH:
770 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
771 "leading slash in name"));
772 break;
773
774 case NAME_ERR_EMPTY_COMPONENT:
775 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
776 "empty component in name"));
777 break;
778
779 case NAME_ERR_TRAILING_SLASH:
780 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
781 "trailing slash in name"));
782 break;
783
784 case NAME_ERR_MULTIPLE_AT:
785 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
786 "multiple '@' delimiters in name"));
787 break;
788
789 }
790 }
791 return (B_FALSE);
792 }
793
794 return (B_TRUE);
795}
796
797/*
798 * Open a handle to the given pool, even if the pool is currently in the FAULTED
799 * state.
800 */
801zpool_handle_t *
802zpool_open_canfail(libzfs_handle_t *hdl, const char *pool)
803{
804 zpool_handle_t *zhp;
805 boolean_t missing;
806
807 /*
808 * Make sure the pool name is valid.
809 */
810 if (!zpool_name_valid(hdl, B_TRUE, pool)) {
811 (void) zfs_error_fmt(hdl, EZFS_INVALIDNAME,
812 dgettext(TEXT_DOMAIN, "cannot open '%s'"),
813 pool);
814 return (NULL);
815 }
816
817 if ((zhp = zfs_alloc(hdl, sizeof (zpool_handle_t))) == NULL)
818 return (NULL);
819
820 zhp->zpool_hdl = hdl;
821 (void) strlcpy(zhp->zpool_name, pool, sizeof (zhp->zpool_name));
822
823 if (zpool_refresh_stats(zhp, &missing) != 0) {
824 zpool_close(zhp);
825 return (NULL);
826 }
827
828 if (missing) {
829 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "no such pool"));
830 (void) zfs_error_fmt(hdl, EZFS_NOENT,
831 dgettext(TEXT_DOMAIN, "cannot open '%s'"), pool);
832 zpool_close(zhp);
833 return (NULL);
834 }
835
836 return (zhp);
837}
838
839/*
840 * Like the above, but silent on error. Used when iterating over pools (because
841 * the configuration cache may be out of date).
842 */
843int
844zpool_open_silent(libzfs_handle_t *hdl, const char *pool, zpool_handle_t **ret)
845{
846 zpool_handle_t *zhp;
847 boolean_t missing;
848
849 if ((zhp = zfs_alloc(hdl, sizeof (zpool_handle_t))) == NULL)
850 return (-1);
851
852 zhp->zpool_hdl = hdl;
853 (void) strlcpy(zhp->zpool_name, pool, sizeof (zhp->zpool_name));
854
855 if (zpool_refresh_stats(zhp, &missing) != 0) {
856 zpool_close(zhp);
857 return (-1);
858 }
859
860 if (missing) {
861 zpool_close(zhp);
862 *ret = NULL;
863 return (0);
864 }
865
866 *ret = zhp;
867 return (0);
868}
869
870/*
871 * Similar to zpool_open_canfail(), but refuses to open pools in the faulted
872 * state.
873 */
874zpool_handle_t *
875zpool_open(libzfs_handle_t *hdl, const char *pool)
876{
877 zpool_handle_t *zhp;
878
879 if ((zhp = zpool_open_canfail(hdl, pool)) == NULL)
880 return (NULL);
881
882 if (zhp->zpool_state == POOL_STATE_UNAVAIL) {
883 (void) zfs_error_fmt(hdl, EZFS_POOLUNAVAIL,
884 dgettext(TEXT_DOMAIN, "cannot open '%s'"), zhp->zpool_name);
885 zpool_close(zhp);
886 return (NULL);
887 }
888
889 return (zhp);
890}
891
892/*
893 * Close the handle. Simply frees the memory associated with the handle.
894 */
895void
896zpool_close(zpool_handle_t *zhp)
897{
898 if (zhp->zpool_config)
899 nvlist_free(zhp->zpool_config);
900 if (zhp->zpool_old_config)
901 nvlist_free(zhp->zpool_old_config);
902 if (zhp->zpool_props)
903 nvlist_free(zhp->zpool_props);
904 free(zhp);
905}
906
907/*
908 * Return the name of the pool.
909 */
910const char *
911zpool_get_name(zpool_handle_t *zhp)
912{
913 return (zhp->zpool_name);
914}
915
916
917/*
918 * Return the state of the pool (ACTIVE or UNAVAILABLE)
919 */
920int
921zpool_get_state(zpool_handle_t *zhp)
922{
923 return (zhp->zpool_state);
924}
925
926/*
927 * Create the named pool, using the provided vdev list. It is assumed
928 * that the consumer has already validated the contents of the nvlist, so we
929 * don't have to worry about error semantics.
930 */
931int
932zpool_create(libzfs_handle_t *hdl, const char *pool, nvlist_t *nvroot,
933 nvlist_t *props, nvlist_t *fsprops)
934{
935 zfs_cmd_t zc = { 0 };
936 nvlist_t *zc_fsprops = NULL;
937 nvlist_t *zc_props = NULL;
938 char msg[1024];
939 char *altroot;
940 int ret = -1;
941
942 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN,
943 "cannot create '%s'"), pool);
944
945 if (!zpool_name_valid(hdl, B_FALSE, pool))
946 return (zfs_error(hdl, EZFS_INVALIDNAME, msg));
947
948 if (zcmd_write_conf_nvlist(hdl, &zc, nvroot) != 0)
949 return (-1);
950
951 if (props) {
952 prop_flags_t flags = { .create = B_TRUE, .import = B_FALSE };
953
954 if ((zc_props = zpool_valid_proplist(hdl, pool, props,
955 SPA_VERSION_1, flags, msg)) == NULL) {
956 goto create_failed;
957 }
958 }
959
960 if (fsprops) {
961 uint64_t zoned;
962 char *zonestr;
963
964 zoned = ((nvlist_lookup_string(fsprops,
965 zfs_prop_to_name(ZFS_PROP_ZONED), &zonestr) == 0) &&
966 strcmp(zonestr, "on") == 0);
967
968 if ((zc_fsprops = zfs_valid_proplist(hdl,
969 ZFS_TYPE_FILESYSTEM, fsprops, zoned, NULL, msg)) == NULL) {
970 goto create_failed;
971 }
972 if (!zc_props &&
973 (nvlist_alloc(&zc_props, NV_UNIQUE_NAME, 0) != 0)) {
974 goto create_failed;
975 }
976 if (nvlist_add_nvlist(zc_props,
977 ZPOOL_ROOTFS_PROPS, zc_fsprops) != 0) {
978 goto create_failed;
979 }
980 }
981
982 if (zc_props && zcmd_write_src_nvlist(hdl, &zc, zc_props) != 0)
983 goto create_failed;
984
985 (void) strlcpy(zc.zc_name, pool, sizeof (zc.zc_name));
986
987 if ((ret = zfs_ioctl(hdl, ZFS_IOC_POOL_CREATE, &zc)) != 0) {
988
989 zcmd_free_nvlists(&zc);
990 nvlist_free(zc_props);
991 nvlist_free(zc_fsprops);
992
993 switch (errno) {
994 case EBUSY:
995 /*
996 * This can happen if the user has specified the same
997 * device multiple times. We can't reliably detect this
998 * until we try to add it and see we already have a
999 * label.
1000 */
1001 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1002 "one or more vdevs refer to the same device"));
1003 return (zfs_error(hdl, EZFS_BADDEV, msg));
1004
1005 case EOVERFLOW:
1006 /*
1007 * This occurs when one of the devices is below
1008 * SPA_MINDEVSIZE. Unfortunately, we can't detect which
1009 * device was the problem device since there's no
1010 * reliable way to determine device size from userland.
1011 */
1012 {
1013 char buf[64];
1014
1015 zfs_nicenum(SPA_MINDEVSIZE, buf, sizeof (buf));
1016
1017 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1018 "one or more devices is less than the "
1019 "minimum size (%s)"), buf);
1020 }
1021 return (zfs_error(hdl, EZFS_BADDEV, msg));
1022
1023 case ENOSPC:
1024 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1025 "one or more devices is out of space"));
1026 return (zfs_error(hdl, EZFS_BADDEV, msg));
1027
1028 case ENOTBLK:
1029 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1030 "cache device must be a disk or disk slice"));
1031 return (zfs_error(hdl, EZFS_BADDEV, msg));
1032
1033 default:
1034 return (zpool_standard_error(hdl, errno, msg));
1035 }
1036 }
1037
1038 /*
1039 * If this is an alternate root pool, then we automatically set the
1040 * mountpoint of the root dataset to be '/'.
1041 */
1042 if (nvlist_lookup_string(props, zpool_prop_to_name(ZPOOL_PROP_ALTROOT),
1043 &altroot) == 0) {
1044 zfs_handle_t *zhp;
1045
1046 verify((zhp = zfs_open(hdl, pool, ZFS_TYPE_DATASET)) != NULL);
1047 verify(zfs_prop_set(zhp, zfs_prop_to_name(ZFS_PROP_MOUNTPOINT),
1048 "/") == 0);
1049
1050 zfs_close(zhp);
1051 }
1052
1053create_failed:
1054 zcmd_free_nvlists(&zc);
1055 nvlist_free(zc_props);
1056 nvlist_free(zc_fsprops);
1057 return (ret);
1058}
1059
1060/*
1061 * Destroy the given pool. It is up to the caller to ensure that there are no
1062 * datasets left in the pool.
1063 */
1064int
1065zpool_destroy(zpool_handle_t *zhp)
1066{
1067 zfs_cmd_t zc = { 0 };
1068 zfs_handle_t *zfp = NULL;
1069 libzfs_handle_t *hdl = zhp->zpool_hdl;
1070 char msg[1024];
1071
1072 if (zhp->zpool_state == POOL_STATE_ACTIVE &&
1073 (zfp = zfs_open(hdl, zhp->zpool_name, ZFS_TYPE_FILESYSTEM)) == NULL)
1074 return (-1);
1075
1076 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
1077
1078 if (zfs_ioctl(hdl, ZFS_IOC_POOL_DESTROY, &zc) != 0) {
1079 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN,
1080 "cannot destroy '%s'"), zhp->zpool_name);
1081
1082 if (errno == EROFS) {
1083 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1084 "one or more devices is read only"));
1085 (void) zfs_error(hdl, EZFS_BADDEV, msg);
1086 } else {
1087 (void) zpool_standard_error(hdl, errno, msg);
1088 }
1089
1090 if (zfp)
1091 zfs_close(zfp);
1092 return (-1);
1093 }
1094
1095 if (zfp) {
1096 remove_mountpoint(zfp);
1097 zfs_close(zfp);
1098 }
1099
1100 return (0);
1101}
1102
1103/*
1104 * Add the given vdevs to the pool. The caller must have already performed the
1105 * necessary verification to ensure that the vdev specification is well-formed.
1106 */
1107int
1108zpool_add(zpool_handle_t *zhp, nvlist_t *nvroot)
1109{
1110 zfs_cmd_t zc = { 0 };
1111 int ret;
1112 libzfs_handle_t *hdl = zhp->zpool_hdl;
1113 char msg[1024];
1114 nvlist_t **spares, **l2cache;
1115 uint_t nspares, nl2cache;
1116
1117 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN,
1118 "cannot add to '%s'"), zhp->zpool_name);
1119
1120 if (zpool_get_prop_int(zhp, ZPOOL_PROP_VERSION, NULL) <
1121 SPA_VERSION_SPARES &&
1122 nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_SPARES,
1123 &spares, &nspares) == 0) {
1124 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "pool must be "
1125 "upgraded to add hot spares"));
1126 return (zfs_error(hdl, EZFS_BADVERSION, msg));
1127 }
1128
1129 if (pool_is_bootable(zhp) && nvlist_lookup_nvlist_array(nvroot,
1130 ZPOOL_CONFIG_SPARES, &spares, &nspares) == 0) {
1131 uint64_t s;
1132
1133 for (s = 0; s < nspares; s++) {
1134 char *path;
1135
1136 if (nvlist_lookup_string(spares[s], ZPOOL_CONFIG_PATH,
1137 &path) == 0 && pool_uses_efi(spares[s])) {
1138 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1139 "device '%s' contains an EFI label and "
1140 "cannot be used on root pools."),
1141 zpool_vdev_name(hdl, NULL, spares[s],
1142 B_FALSE));
1143 return (zfs_error(hdl, EZFS_POOL_NOTSUP, msg));
1144 }
1145 }
1146 }
1147
1148 if (zpool_get_prop_int(zhp, ZPOOL_PROP_VERSION, NULL) <
1149 SPA_VERSION_L2CACHE &&
1150 nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_L2CACHE,
1151 &l2cache, &nl2cache) == 0) {
1152 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "pool must be "
1153 "upgraded to add cache devices"));
1154 return (zfs_error(hdl, EZFS_BADVERSION, msg));
1155 }
1156
1157 if (zcmd_write_conf_nvlist(hdl, &zc, nvroot) != 0)
1158 return (-1);
1159 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
1160
1161 if (zfs_ioctl(hdl, ZFS_IOC_VDEV_ADD, &zc) != 0) {
1162 switch (errno) {
1163 case EBUSY:
1164 /*
1165 * This can happen if the user has specified the same
1166 * device multiple times. We can't reliably detect this
1167 * until we try to add it and see we already have a
1168 * label.
1169 */
1170 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1171 "one or more vdevs refer to the same device"));
1172 (void) zfs_error(hdl, EZFS_BADDEV, msg);
1173 break;
1174
1175 case EOVERFLOW:
1176 /*
1177 * This occurrs when one of the devices is below
1178 * SPA_MINDEVSIZE. Unfortunately, we can't detect which
1179 * device was the problem device since there's no
1180 * reliable way to determine device size from userland.
1181 */
1182 {
1183 char buf[64];
1184
1185 zfs_nicenum(SPA_MINDEVSIZE, buf, sizeof (buf));
1186
1187 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1188 "device is less than the minimum "
1189 "size (%s)"), buf);
1190 }
1191 (void) zfs_error(hdl, EZFS_BADDEV, msg);
1192 break;
1193
1194 case ENOTSUP:
1195 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1196 "pool must be upgraded to add these vdevs"));
1197 (void) zfs_error(hdl, EZFS_BADVERSION, msg);
1198 break;
1199
1200 case EDOM:
1201 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1202 "root pool can not have multiple vdevs"
1203 " or separate logs"));
1204 (void) zfs_error(hdl, EZFS_POOL_NOTSUP, msg);
1205 break;
1206
1207 case ENOTBLK:
1208 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1209 "cache device must be a disk or disk slice"));
1210 (void) zfs_error(hdl, EZFS_BADDEV, msg);
1211 break;
1212
1213 default:
1214 (void) zpool_standard_error(hdl, errno, msg);
1215 }
1216
1217 ret = -1;
1218 } else {
1219 ret = 0;
1220 }
1221
1222 zcmd_free_nvlists(&zc);
1223
1224 return (ret);
1225}
1226
1227/*
1228 * Exports the pool from the system. The caller must ensure that there are no
1229 * mounted datasets in the pool.
1230 */
1231int
1232zpool_export_common(zpool_handle_t *zhp, boolean_t force, boolean_t hardforce)
1233{
1234 zfs_cmd_t zc = { 0 };
1235 char msg[1024];
1236
1237 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN,
1238 "cannot export '%s'"), zhp->zpool_name);
1239
1240 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
1241 zc.zc_cookie = force;
1242 zc.zc_guid = hardforce;
1243
1244 if (zfs_ioctl(zhp->zpool_hdl, ZFS_IOC_POOL_EXPORT, &zc) != 0) {
1245 switch (errno) {
1246 case EXDEV:
1247 zfs_error_aux(zhp->zpool_hdl, dgettext(TEXT_DOMAIN,
1248 "use '-f' to override the following errors:\n"
1249 "'%s' has an active shared spare which could be"
1250 " used by other pools once '%s' is exported."),
1251 zhp->zpool_name, zhp->zpool_name);
1252 return (zfs_error(zhp->zpool_hdl, EZFS_ACTIVE_SPARE,
1253 msg));
1254 default:
1255 return (zpool_standard_error_fmt(zhp->zpool_hdl, errno,
1256 msg));
1257 }
1258 }
1259
1260 return (0);
1261}
1262
1263int
1264zpool_export(zpool_handle_t *zhp, boolean_t force)
1265{
1266 return (zpool_export_common(zhp, force, B_FALSE));
1267}
1268
1269int
1270zpool_export_force(zpool_handle_t *zhp)
1271{
1272 return (zpool_export_common(zhp, B_TRUE, B_TRUE));
1273}
1274
1275static void
1276zpool_rewind_exclaim(libzfs_handle_t *hdl, const char *name, boolean_t dryrun,
1277 nvlist_t *config)
1278{
1279 nvlist_t *nv = NULL;
1280 uint64_t rewindto;
1281 int64_t loss = -1;
1282 struct tm t;
1283 char timestr[128];
1284
1285 if (!hdl->libzfs_printerr || config == NULL)
1286 return;
1287
1288 if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_LOAD_INFO, &nv) != 0)
1289 return;
1290
1291 if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_LOAD_TIME, &rewindto) != 0)
1292 return;
1293 (void) nvlist_lookup_int64(nv, ZPOOL_CONFIG_REWIND_TIME, &loss);
1294
1295 if (localtime_r((time_t *)&rewindto, &t) != NULL &&
1296 strftime(timestr, 128, 0, &t) != 0) {
1297 if (dryrun) {
1298 (void) printf(dgettext(TEXT_DOMAIN,
1299 "Would be able to return %s "
1300 "to its state as of %s.\n"),
1301 name, timestr);
1302 } else {
1303 (void) printf(dgettext(TEXT_DOMAIN,
1304 "Pool %s returned to its state as of %s.\n"),
1305 name, timestr);
1306 }
1307 if (loss > 120) {
1308 (void) printf(dgettext(TEXT_DOMAIN,
1309 "%s approximately %lld "),
1310 dryrun ? "Would discard" : "Discarded",
1311 (loss + 30) / 60);
1312 (void) printf(dgettext(TEXT_DOMAIN,
1313 "minutes of transactions.\n"));
1314 } else if (loss > 0) {
1315 (void) printf(dgettext(TEXT_DOMAIN,
1316 "%s approximately %lld "),
1317 dryrun ? "Would discard" : "Discarded", loss);
1318 (void) printf(dgettext(TEXT_DOMAIN,
1319 "seconds of transactions.\n"));
1320 }
1321 }
1322}
1323
1324void
1325zpool_explain_recover(libzfs_handle_t *hdl, const char *name, int reason,
1326 nvlist_t *config)
1327{
1328 nvlist_t *nv = NULL;
1329 int64_t loss = -1;
1330 uint64_t edata = UINT64_MAX;
1331 uint64_t rewindto;
1332 struct tm t;
1333 char timestr[128];
1334
1335 if (!hdl->libzfs_printerr)
1336 return;
1337
1338 if (reason >= 0)
1339 (void) printf(dgettext(TEXT_DOMAIN, "action: "));
1340 else
1341 (void) printf(dgettext(TEXT_DOMAIN, "\t"));
1342
1343 /* All attempted rewinds failed if ZPOOL_CONFIG_LOAD_TIME missing */
1344 if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_LOAD_INFO, &nv) != 0 ||
1345 nvlist_lookup_uint64(nv, ZPOOL_CONFIG_LOAD_TIME, &rewindto) != 0)
1346 goto no_info;
1347
1348 (void) nvlist_lookup_int64(nv, ZPOOL_CONFIG_REWIND_TIME, &loss);
1349 (void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_LOAD_DATA_ERRORS,
1350 &edata);
1351
1352 (void) printf(dgettext(TEXT_DOMAIN,
1353 "Recovery is possible, but will result in some data loss.\n"));
1354
1355 if (localtime_r((time_t *)&rewindto, &t) != NULL &&
1356 strftime(timestr, 128, 0, &t) != 0) {
1357 (void) printf(dgettext(TEXT_DOMAIN,
1358 "\tReturning the pool to its state as of %s\n"
1359 "\tshould correct the problem. "),
1360 timestr);
1361 } else {
1362 (void) printf(dgettext(TEXT_DOMAIN,
1363 "\tReverting the pool to an earlier state "
1364 "should correct the problem.\n\t"));
1365 }
1366
1367 if (loss > 120) {
1368 (void) printf(dgettext(TEXT_DOMAIN,
1369 "Approximately %lld minutes of data\n"
1370 "\tmust be discarded, irreversibly. "), (loss + 30) / 60);
1371 } else if (loss > 0) {
1372 (void) printf(dgettext(TEXT_DOMAIN,
1373 "Approximately %lld seconds of data\n"
1374 "\tmust be discarded, irreversibly. "), loss);
1375 }
1376 if (edata != 0 && edata != UINT64_MAX) {
1377 if (edata == 1) {
1378 (void) printf(dgettext(TEXT_DOMAIN,
1379 "After rewind, at least\n"
1380 "\tone persistent user-data error will remain. "));
1381 } else {
1382 (void) printf(dgettext(TEXT_DOMAIN,
1383 "After rewind, several\n"
1384 "\tpersistent user-data errors will remain. "));
1385 }
1386 }
1387 (void) printf(dgettext(TEXT_DOMAIN,
1388 "Recovery can be attempted\n\tby executing 'zpool %s -F %s'. "),
1389 reason >= 0 ? "clear" : "import", name);
1390
1391 (void) printf(dgettext(TEXT_DOMAIN,
1392 "A scrub of the pool\n"
1393 "\tis strongly recommended after recovery.\n"));
1394 return;
1395
1396no_info:
1397 (void) printf(dgettext(TEXT_DOMAIN,
1398 "Destroy and re-create the pool from\n\ta backup source.\n"));
1399}
1400
1401/*
1402 * zpool_import() is a contracted interface. Should be kept the same
1403 * if possible.
1404 *
1405 * Applications should use zpool_import_props() to import a pool with
1406 * new properties value to be set.
1407 */
1408int
1409zpool_import(libzfs_handle_t *hdl, nvlist_t *config, const char *newname,
1410 char *altroot)
1411{
1412 nvlist_t *props = NULL;
1413 int ret;
1414
1415 if (altroot != NULL) {
1416 if (nvlist_alloc(&props, NV_UNIQUE_NAME, 0) != 0) {
1417 return (zfs_error_fmt(hdl, EZFS_NOMEM,
1418 dgettext(TEXT_DOMAIN, "cannot import '%s'"),
1419 newname));
1420 }
1421
1422 if (nvlist_add_string(props,
1423 zpool_prop_to_name(ZPOOL_PROP_ALTROOT), altroot) != 0 ||
1424 nvlist_add_string(props,
1425 zpool_prop_to_name(ZPOOL_PROP_CACHEFILE), "none") != 0) {
1426 nvlist_free(props);
1427 return (zfs_error_fmt(hdl, EZFS_NOMEM,
1428 dgettext(TEXT_DOMAIN, "cannot import '%s'"),
1429 newname));
1430 }
1431 }
1432
1433 ret = zpool_import_props(hdl, config, newname, props,
1434 ZFS_IMPORT_NORMAL);
1435 if (props)
1436 nvlist_free(props);
1437 return (ret);
1438}
1439
1440static void
1441print_vdev_tree(libzfs_handle_t *hdl, const char *name, nvlist_t *nv,
1442 int indent)
1443{
1444 nvlist_t **child;
1445 uint_t c, children;
1446 char *vname;
1447 uint64_t is_log = 0;
1448
1449 (void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_IS_LOG,
1450 &is_log);
1451
1452 if (name != NULL)
1453 (void) printf("\t%*s%s%s\n", indent, "", name,
1454 is_log ? " [log]" : "");
1455
1456 if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN,
1457 &child, &children) != 0)
1458 return;
1459
1460 for (c = 0; c < children; c++) {
1461 vname = zpool_vdev_name(hdl, NULL, child[c], B_TRUE);
1462 print_vdev_tree(hdl, vname, child[c], indent + 2);
1463 free(vname);
1464 }
1465}
1466
1467/*
1468 * Import the given pool using the known configuration and a list of
1469 * properties to be set. The configuration should have come from
1470 * zpool_find_import(). The 'newname' parameters control whether the pool
1471 * is imported with a different name.
1472 */
1473int
1474zpool_import_props(libzfs_handle_t *hdl, nvlist_t *config, const char *newname,
1475 nvlist_t *props, int flags)
1476{
1477 zfs_cmd_t zc = { 0 };
1478 zpool_rewind_policy_t policy;
1479 nvlist_t *nv = NULL;
1480 nvlist_t *nvinfo = NULL;
1481 nvlist_t *missing = NULL;
1482 char *thename;
1483 char *origname;
1484 int ret;
1485 int error = 0;
1486 char errbuf[1024];
1487
1488 verify(nvlist_lookup_string(config, ZPOOL_CONFIG_POOL_NAME,
1489 &origname) == 0);
1490
1491 (void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN,
1492 "cannot import pool '%s'"), origname);
1493
1494 if (newname != NULL) {
1495 if (!zpool_name_valid(hdl, B_FALSE, newname))
1496 return (zfs_error_fmt(hdl, EZFS_INVALIDNAME,
1497 dgettext(TEXT_DOMAIN, "cannot import '%s'"),
1498 newname));
1499 thename = (char *)newname;
1500 } else {
1501 thename = origname;
1502 }
1503
1504 if (props) {
1505 uint64_t version;
1506 prop_flags_t flags = { .create = B_FALSE, .import = B_TRUE };
1507
1508 verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_VERSION,
1509 &version) == 0);
1510
1511 if ((props = zpool_valid_proplist(hdl, origname,
1512 props, version, flags, errbuf)) == NULL) {
1513 return (-1);
1514 } else if (zcmd_write_src_nvlist(hdl, &zc, props) != 0) {
1515 nvlist_free(props);
1516 return (-1);
1517 }
1518 }
1519
1520 (void) strlcpy(zc.zc_name, thename, sizeof (zc.zc_name));
1521
1522 verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID,
1523 &zc.zc_guid) == 0);
1524
1525 if (zcmd_write_conf_nvlist(hdl, &zc, config) != 0) {
1526 nvlist_free(props);
1527 return (-1);
1528 }
1529 if (zcmd_alloc_dst_nvlist(hdl, &zc, zc.zc_nvlist_conf_size * 2) != 0) {
1530 nvlist_free(props);
1531 return (-1);
1532 }
1533
1534 zc.zc_cookie = flags;
1535 while ((ret = zfs_ioctl(hdl, ZFS_IOC_POOL_IMPORT, &zc)) != 0 &&
1536 errno == ENOMEM) {
1537 if (zcmd_expand_dst_nvlist(hdl, &zc) != 0) {
1538 zcmd_free_nvlists(&zc);
1539 return (-1);
1540 }
1541 }
1542 if (ret != 0)
1543 error = errno;
1544
1545 (void) zcmd_read_dst_nvlist(hdl, &zc, &nv);
1546 zpool_get_rewind_policy(config, &policy);
1547
1548 if (error) {
1549 char desc[1024];
1550
1551 /*
1552 * Dry-run failed, but we print out what success
1553 * looks like if we found a best txg
1554 */
1555 if (policy.zrp_request & ZPOOL_TRY_REWIND) {
1556 zpool_rewind_exclaim(hdl, newname ? origname : thename,
1557 B_TRUE, nv);
1558 nvlist_free(nv);
1559 return (-1);
1560 }
1561
1562 if (newname == NULL)
1563 (void) snprintf(desc, sizeof (desc),
1564 dgettext(TEXT_DOMAIN, "cannot import '%s'"),
1565 thename);
1566 else
1567 (void) snprintf(desc, sizeof (desc),
1568 dgettext(TEXT_DOMAIN, "cannot import '%s' as '%s'"),
1569 origname, thename);
1570
1571 switch (error) {
1572 case ENOTSUP:
1573 /*
1574 * Unsupported version.
1575 */
1576 (void) zfs_error(hdl, EZFS_BADVERSION, desc);
1577 break;
1578
1579 case EINVAL:
1580 (void) zfs_error(hdl, EZFS_INVALCONFIG, desc);
1581 break;
1582
1583 case EROFS:
1584 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1585 "one or more devices is read only"));
1586 (void) zfs_error(hdl, EZFS_BADDEV, desc);
1587 break;
1588
1589 case ENXIO:
1590 if (nv && nvlist_lookup_nvlist(nv,
1591 ZPOOL_CONFIG_LOAD_INFO, &nvinfo) == 0 &&
1592 nvlist_lookup_nvlist(nvinfo,
1593 ZPOOL_CONFIG_MISSING_DEVICES, &missing) == 0) {
1594 (void) printf(dgettext(TEXT_DOMAIN,
1595 "The devices below are missing, use "
1596 "'-m' to import the pool anyway:\n"));
1597 print_vdev_tree(hdl, NULL, missing, 2);
1598 (void) printf("\n");
1599 }
1600 (void) zpool_standard_error(hdl, error, desc);
1601 break;
1602
1603 case EEXIST:
1604 (void) zpool_standard_error(hdl, error, desc);
1605 break;
1606
1607 default:
1608 (void) zpool_standard_error(hdl, error, desc);
1609 zpool_explain_recover(hdl,
1610 newname ? origname : thename, -error, nv);
1611 break;
1612 }
1613
1614 nvlist_free(nv);
1615 ret = -1;
1616 } else {
1617 zpool_handle_t *zhp;
1618
1619 /*
1620 * This should never fail, but play it safe anyway.
1621 */
1622 if (zpool_open_silent(hdl, thename, &zhp) != 0)
1623 ret = -1;
1624 else if (zhp != NULL)
1625 zpool_close(zhp);
1626 if (policy.zrp_request &
1627 (ZPOOL_DO_REWIND | ZPOOL_TRY_REWIND)) {
1628 zpool_rewind_exclaim(hdl, newname ? origname : thename,
1629 ((policy.zrp_request & ZPOOL_TRY_REWIND) != 0), nv);
1630 }
1631 nvlist_free(nv);
1632 return (0);
1633 }
1634
1635 zcmd_free_nvlists(&zc);
1636 nvlist_free(props);
1637
1638 return (ret);
1639}
1640
1641/*
1642 * Scan the pool.
1643 */
1644int
1645zpool_scan(zpool_handle_t *zhp, pool_scan_func_t func)
1646{
1647 zfs_cmd_t zc = { 0 };
1648 char msg[1024];
1649 libzfs_handle_t *hdl = zhp->zpool_hdl;
1650
1651 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
1652 zc.zc_cookie = func;
1653
1654 if (zfs_ioctl(hdl, ZFS_IOC_POOL_SCAN, &zc) == 0 ||
1655 (errno == ENOENT && func != POOL_SCAN_NONE))
1656 return (0);
1657
1658 if (func == POOL_SCAN_SCRUB) {
1659 (void) snprintf(msg, sizeof (msg),
1660 dgettext(TEXT_DOMAIN, "cannot scrub %s"), zc.zc_name);
1661 } else if (func == POOL_SCAN_NONE) {
1662 (void) snprintf(msg, sizeof (msg),
1663 dgettext(TEXT_DOMAIN, "cannot cancel scrubbing %s"),
1664 zc.zc_name);
1665 } else {
1666 assert(!"unexpected result");
1667 }
1668
1669 if (errno == EBUSY) {
1670 nvlist_t *nvroot;
1671 pool_scan_stat_t *ps = NULL;
1672 uint_t psc;
1673
1674 verify(nvlist_lookup_nvlist(zhp->zpool_config,
1675 ZPOOL_CONFIG_VDEV_TREE, &nvroot) == 0);
1676 (void) nvlist_lookup_uint64_array(nvroot,
1677 ZPOOL_CONFIG_SCAN_STATS, (uint64_t **)&ps, &psc);
1678 if (ps && ps->pss_func == POOL_SCAN_SCRUB)
1679 return (zfs_error(hdl, EZFS_SCRUBBING, msg));
1680 else
1681 return (zfs_error(hdl, EZFS_RESILVERING, msg));
1682 } else if (errno == ENOENT) {
1683 return (zfs_error(hdl, EZFS_NO_SCRUB, msg));
1684 } else {
1685 return (zpool_standard_error(hdl, errno, msg));
1686 }
1687}
1688
1689/*
1690 * This provides a very minimal check whether a given string is likely a
1691 * c#t#d# style string. Users of this are expected to do their own
1692 * verification of the s# part.
1693 */
1694#define CTD_CHECK(str) (str && str[0] == 'c' && isdigit(str[1]))
1695
1696/*
1697 * More elaborate version for ones which may start with "/dev/dsk/"
1698 * and the like.
1699 */
1700static int
1701ctd_check_path(char *str) {
1702 /*
1703 * If it starts with a slash, check the last component.
1704 */
1705 if (str && str[0] == '/') {
1706 char *tmp = strrchr(str, '/');
1707
1708 /*
1709 * If it ends in "/old", check the second-to-last
1710 * component of the string instead.
1711 */
1712 if (tmp != str && strcmp(tmp, "/old") == 0) {
1713 for (tmp--; *tmp != '/'; tmp--)
1714 ;
1715 }
1716 str = tmp + 1;
1717 }
1718 return (CTD_CHECK(str));
1719}
1720
1721/*
1722 * Find a vdev that matches the search criteria specified. We use the
1723 * the nvpair name to determine how we should look for the device.
1724 * 'avail_spare' is set to TRUE if the provided guid refers to an AVAIL
1725 * spare; but FALSE if its an INUSE spare.
1726 */
1727static nvlist_t *
1728vdev_to_nvlist_iter(nvlist_t *nv, nvlist_t *search, boolean_t *avail_spare,
1729 boolean_t *l2cache, boolean_t *log)
1730{
1731 uint_t c, children;
1732 nvlist_t **child;
1733 nvlist_t *ret;
1734 uint64_t is_log;
1735 char *srchkey;
1736 nvpair_t *pair = nvlist_next_nvpair(search, NULL);
1737
1738 /* Nothing to look for */
1739 if (search == NULL || pair == NULL)
1740 return (NULL);
1741
1742 /* Obtain the key we will use to search */
1743 srchkey = nvpair_name(pair);
1744
1745 switch (nvpair_type(pair)) {
1746 case DATA_TYPE_UINT64:
1747 if (strcmp(srchkey, ZPOOL_CONFIG_GUID) == 0) {
1748 uint64_t srchval, theguid;
1749
1750 verify(nvpair_value_uint64(pair, &srchval) == 0);
1751 verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID,
1752 &theguid) == 0);
1753 if (theguid == srchval)
1754 return (nv);
1755 }
1756 break;
1757
1758 case DATA_TYPE_STRING: {
1759 char *srchval, *val;
1760
1761 verify(nvpair_value_string(pair, &srchval) == 0);
1762 if (nvlist_lookup_string(nv, srchkey, &val) != 0)
1763 break;
1764
1765 /*
1766 * Search for the requested value. Special cases:
1767 *
1768 * - ZPOOL_CONFIG_PATH for whole disk entries. These end in
1769 * "s0" or "s0/old". The "s0" part is hidden from the user,
1770 * but included in the string, so this matches around it.
1771 * - looking for a top-level vdev name (i.e. ZPOOL_CONFIG_TYPE).
1772 *
1773 * Otherwise, all other searches are simple string compares.
1774 */
1775 if (strcmp(srchkey, ZPOOL_CONFIG_PATH) == 0 &&
1776 ctd_check_path(val)) {
1777 uint64_t wholedisk = 0;
1778
1779 (void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_WHOLE_DISK,
1780 &wholedisk);
1781 if (wholedisk) {
1782 int slen = strlen(srchval);
1783 int vlen = strlen(val);
1784
1785 if (slen != vlen - 2)
1786 break;
1787
1788 /*
1789 * make_leaf_vdev() should only set
1790 * wholedisk for ZPOOL_CONFIG_PATHs which
1791 * will include "/dev/dsk/", giving plenty of
1792 * room for the indices used next.
1793 */
1794 ASSERT(vlen >= 6);
1795
1796 /*
1797 * strings identical except trailing "s0"
1798 */
1799 if (strcmp(&val[vlen - 2], "s0") == 0 &&
1800 strncmp(srchval, val, slen) == 0)
1801 return (nv);
1802
1803 /*
1804 * strings identical except trailing "s0/old"
1805 */
1806 if (strcmp(&val[vlen - 6], "s0/old") == 0 &&
1807 strcmp(&srchval[slen - 4], "/old") == 0 &&
1808 strncmp(srchval, val, slen - 4) == 0)
1809 return (nv);
1810
1811 break;
1812 }
1813 } else if (strcmp(srchkey, ZPOOL_CONFIG_TYPE) == 0 && val) {
1814 char *type, *idx, *end, *p;
1815 uint64_t id, vdev_id;
1816
1817 /*
1818 * Determine our vdev type, keeping in mind
1819 * that the srchval is composed of a type and
1820 * vdev id pair (i.e. mirror-4).
1821 */
1822 if ((type = strdup(srchval)) == NULL)
1823 return (NULL);
1824
1825 if ((p = strrchr(type, '-')) == NULL) {
1826 free(type);
1827 break;
1828 }
1829 idx = p + 1;
1830 *p = '\0';
1831
1832 /*
1833 * If the types don't match then keep looking.
1834 */
1835 if (strncmp(val, type, strlen(val)) != 0) {
1836 free(type);
1837 break;
1838 }
1839
1840 verify(strncmp(type, VDEV_TYPE_RAIDZ,
1841 strlen(VDEV_TYPE_RAIDZ)) == 0 ||
1842 strncmp(type, VDEV_TYPE_MIRROR,
1843 strlen(VDEV_TYPE_MIRROR)) == 0);
1844 verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_ID,
1845 &id) == 0);
1846
1847 errno = 0;
1848 vdev_id = strtoull(idx, &end, 10);
1849
1850 free(type);
1851 if (errno != 0)
1852 return (NULL);
1853
1854 /*
1855 * Now verify that we have the correct vdev id.
1856 */
1857 if (vdev_id == id)
1858 return (nv);
1859 }
1860
1861 /*
1862 * Common case
1863 */
1864 if (strcmp(srchval, val) == 0)
1865 return (nv);
1866 break;
1867 }
1868
1869 default:
1870 break;
1871 }
1872
1873 if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN,
1874 &child, &children) != 0)
1875 return (NULL);
1876
1877 for (c = 0; c < children; c++) {
1878 if ((ret = vdev_to_nvlist_iter(child[c], search,
1879 avail_spare, l2cache, NULL)) != NULL) {
1880 /*
1881 * The 'is_log' value is only set for the toplevel
1882 * vdev, not the leaf vdevs. So we always lookup the
1883 * log device from the root of the vdev tree (where
1884 * 'log' is non-NULL).
1885 */
1886 if (log != NULL &&
1887 nvlist_lookup_uint64(child[c],
1888 ZPOOL_CONFIG_IS_LOG, &is_log) == 0 &&
1889 is_log) {
1890 *log = B_TRUE;
1891 }
1892 return (ret);
1893 }
1894 }
1895
1896 if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_SPARES,
1897 &child, &children) == 0) {
1898 for (c = 0; c < children; c++) {
1899 if ((ret = vdev_to_nvlist_iter(child[c], search,
1900 avail_spare, l2cache, NULL)) != NULL) {
1901 *avail_spare = B_TRUE;
1902 return (ret);
1903 }
1904 }
1905 }
1906
1907 if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_L2CACHE,
1908 &child, &children) == 0) {
1909 for (c = 0; c < children; c++) {
1910 if ((ret = vdev_to_nvlist_iter(child[c], search,
1911 avail_spare, l2cache, NULL)) != NULL) {
1912 *l2cache = B_TRUE;
1913 return (ret);
1914 }
1915 }
1916 }
1917
1918 return (NULL);
1919}
1920
1921/*
1922 * Given a physical path (minus the "/devices" prefix), find the
1923 * associated vdev.
1924 */
1925nvlist_t *
1926zpool_find_vdev_by_physpath(zpool_handle_t *zhp, const char *ppath,
1927 boolean_t *avail_spare, boolean_t *l2cache, boolean_t *log)
1928{
1929 nvlist_t *search, *nvroot, *ret;
1930
1931 verify(nvlist_alloc(&search, NV_UNIQUE_NAME, KM_SLEEP) == 0);
1932 verify(nvlist_add_string(search, ZPOOL_CONFIG_PHYS_PATH, ppath) == 0);
1933
1934 verify(nvlist_lookup_nvlist(zhp->zpool_config, ZPOOL_CONFIG_VDEV_TREE,
1935 &nvroot) == 0);
1936
1937 *avail_spare = B_FALSE;
1938 *l2cache = B_FALSE;
1939 if (log != NULL)
1940 *log = B_FALSE;
1941 ret = vdev_to_nvlist_iter(nvroot, search, avail_spare, l2cache, log);
1942 nvlist_free(search);
1943
1944 return (ret);
1945}
1946
1947/*
1948 * Determine if we have an "interior" top-level vdev (i.e mirror/raidz).
1949 */
1950boolean_t
1951zpool_vdev_is_interior(const char *name)
1952{
1953 if (strncmp(name, VDEV_TYPE_RAIDZ, strlen(VDEV_TYPE_RAIDZ)) == 0 ||
1954 strncmp(name, VDEV_TYPE_MIRROR, strlen(VDEV_TYPE_MIRROR)) == 0)
1955 return (B_TRUE);
1956 return (B_FALSE);
1957}
1958
1959nvlist_t *
1960zpool_find_vdev(zpool_handle_t *zhp, const char *path, boolean_t *avail_spare,
1961 boolean_t *l2cache, boolean_t *log)
1962{
1963 char buf[MAXPATHLEN];
1964 char *end;
1965 nvlist_t *nvroot, *search, *ret;
1966 uint64_t guid;
1967
1968 verify(nvlist_alloc(&search, NV_UNIQUE_NAME, KM_SLEEP) == 0);
1969
1970 guid = strtoull(path, &end, 10);
1971 if (guid != 0 && *end == '\0') {
1972 verify(nvlist_add_uint64(search, ZPOOL_CONFIG_GUID, guid) == 0);
1973 } else if (zpool_vdev_is_interior(path)) {
1974 verify(nvlist_add_string(search, ZPOOL_CONFIG_TYPE, path) == 0);
1975 } else if (path[0] != '/') {
1976 (void) snprintf(buf, sizeof (buf), "%s%s", _PATH_DEV, path);
1977 verify(nvlist_add_string(search, ZPOOL_CONFIG_PATH, buf) == 0);
1978 } else {
1979 verify(nvlist_add_string(search, ZPOOL_CONFIG_PATH, path) == 0);
1980 }
1981
1982 verify(nvlist_lookup_nvlist(zhp->zpool_config, ZPOOL_CONFIG_VDEV_TREE,
1983 &nvroot) == 0);
1984
1985 *avail_spare = B_FALSE;
1986 *l2cache = B_FALSE;
1987 if (log != NULL)
1988 *log = B_FALSE;
1989 ret = vdev_to_nvlist_iter(nvroot, search, avail_spare, l2cache, log);
1990 nvlist_free(search);
1991
1992 return (ret);
1993}
1994
1995static int
1996vdev_online(nvlist_t *nv)
1997{
1998 uint64_t ival;
1999
2000 if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_OFFLINE, &ival) == 0 ||
2001 nvlist_lookup_uint64(nv, ZPOOL_CONFIG_FAULTED, &ival) == 0 ||
2002 nvlist_lookup_uint64(nv, ZPOOL_CONFIG_REMOVED, &ival) == 0)
2003 return (0);
2004
2005 return (1);
2006}
2007
2008/*
2009 * Helper function for zpool_get_physpaths().
2010 */
2011static int
2012vdev_get_one_physpath(nvlist_t *config, char *physpath, size_t physpath_size,
2013 size_t *bytes_written)
2014{
2015 size_t bytes_left, pos, rsz;
2016 char *tmppath;
2017 const char *format;
2018
2019 if (nvlist_lookup_string(config, ZPOOL_CONFIG_PHYS_PATH,
2020 &tmppath) != 0)
2021 return (EZFS_NODEVICE);
2022
2023 pos = *bytes_written;
2024 bytes_left = physpath_size - pos;
2025 format = (pos == 0) ? "%s" : " %s";
2026
2027 rsz = snprintf(physpath + pos, bytes_left, format, tmppath);
2028 *bytes_written += rsz;
2029
2030 if (rsz >= bytes_left) {
2031 /* if physpath was not copied properly, clear it */
2032 if (bytes_left != 0) {
2033 physpath[pos] = 0;
2034 }
2035 return (EZFS_NOSPC);
2036 }
2037 return (0);
2038}
2039
2040static int
2041vdev_get_physpaths(nvlist_t *nv, char *physpath, size_t phypath_size,
2042 size_t *rsz, boolean_t is_spare)
2043{
2044 char *type;
2045 int ret;
2046
2047 if (nvlist_lookup_string(nv, ZPOOL_CONFIG_TYPE, &type) != 0)
2048 return (EZFS_INVALCONFIG);
2049
2050 if (strcmp(type, VDEV_TYPE_DISK) == 0) {
2051 /*
2052 * An active spare device has ZPOOL_CONFIG_IS_SPARE set.
2053 * For a spare vdev, we only want to boot from the active
2054 * spare device.
2055 */
2056 if (is_spare) {
2057 uint64_t spare = 0;
2058 (void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_IS_SPARE,
2059 &spare);
2060 if (!spare)
2061 return (EZFS_INVALCONFIG);
2062 }
2063
2064 if (vdev_online(nv)) {
2065 if ((ret = vdev_get_one_physpath(nv, physpath,
2066 phypath_size, rsz)) != 0)
2067 return (ret);
2068 }
2069 } else if (strcmp(type, VDEV_TYPE_MIRROR) == 0 ||
2070 strcmp(type, VDEV_TYPE_REPLACING) == 0 ||
2071 (is_spare = (strcmp(type, VDEV_TYPE_SPARE) == 0))) {
2072 nvlist_t **child;
2073 uint_t count;
2074 int i, ret;
2075
2076 if (nvlist_lookup_nvlist_array(nv,
2077 ZPOOL_CONFIG_CHILDREN, &child, &count) != 0)
2078 return (EZFS_INVALCONFIG);
2079
2080 for (i = 0; i < count; i++) {
2081 ret = vdev_get_physpaths(child[i], physpath,
2082 phypath_size, rsz, is_spare);
2083 if (ret == EZFS_NOSPC)
2084 return (ret);
2085 }
2086 }
2087
2088 return (EZFS_POOL_INVALARG);
2089}
2090
2091/*
2092 * Get phys_path for a root pool config.
2093 * Return 0 on success; non-zero on failure.
2094 */
2095static int
2096zpool_get_config_physpath(nvlist_t *config, char *physpath, size_t phypath_size)
2097{
2098 size_t rsz;
2099 nvlist_t *vdev_root;
2100 nvlist_t **child;
2101 uint_t count;
2102 char *type;
2103
2104 rsz = 0;
2105
2106 if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE,
2107 &vdev_root) != 0)
2108 return (EZFS_INVALCONFIG);
2109
2110 if (nvlist_lookup_string(vdev_root, ZPOOL_CONFIG_TYPE, &type) != 0 ||
2111 nvlist_lookup_nvlist_array(vdev_root, ZPOOL_CONFIG_CHILDREN,
2112 &child, &count) != 0)
2113 return (EZFS_INVALCONFIG);
2114
2115 /*
2116 * root pool can not have EFI labeled disks and can only have
2117 * a single top-level vdev.
2118 */
2119 if (strcmp(type, VDEV_TYPE_ROOT) != 0 || count != 1 ||
2120 pool_uses_efi(vdev_root))
2121 return (EZFS_POOL_INVALARG);
2122
2123 (void) vdev_get_physpaths(child[0], physpath, phypath_size, &rsz,
2124 B_FALSE);
2125
2126 /* No online devices */
2127 if (rsz == 0)
2128 return (EZFS_NODEVICE);
2129
2130 return (0);
2131}
2132
2133/*
2134 * Get phys_path for a root pool
2135 * Return 0 on success; non-zero on failure.
2136 */
2137int
2138zpool_get_physpath(zpool_handle_t *zhp, char *physpath, size_t phypath_size)
2139{
2140 return (zpool_get_config_physpath(zhp->zpool_config, physpath,
2141 phypath_size));
2142}
2143
2144/*
2145 * If the device has being dynamically expanded then we need to relabel
2146 * the disk to use the new unallocated space.
2147 */
2148static int
2149zpool_relabel_disk(libzfs_handle_t *hdl, const char *name)
2150{
2151#ifdef sun
2152 char path[MAXPATHLEN];
2153 char errbuf[1024];
2154 int fd, error;
2155 int (*_efi_use_whole_disk)(int);
2156
2157 if ((_efi_use_whole_disk = (int (*)(int))dlsym(RTLD_DEFAULT,
2158 "efi_use_whole_disk")) == NULL)
2159 return (-1);
2160
2161 (void) snprintf(path, sizeof (path), "%s/%s", RDISK_ROOT, name);
2162
2163 if ((fd = open(path, O_RDWR | O_NDELAY)) < 0) {
2164 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "cannot "
2165 "relabel '%s': unable to open device"), name);
2166 return (zfs_error(hdl, EZFS_OPENFAILED, errbuf));
2167 }
2168
2169 /*
2170 * It's possible that we might encounter an error if the device
2171 * does not have any unallocated space left. If so, we simply
2172 * ignore that error and continue on.
2173 */
2174 error = _efi_use_whole_disk(fd);
2175 (void) close(fd);
2176 if (error && error != VT_ENOSPC) {
2177 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "cannot "
2178 "relabel '%s': unable to read disk capacity"), name);
2179 return (zfs_error(hdl, EZFS_NOCAP, errbuf));
2180 }
2181#endif /* sun */
2182 return (0);
2183}
2184
2185/*
2186 * Bring the specified vdev online. The 'flags' parameter is a set of the
2187 * ZFS_ONLINE_* flags.
2188 */
2189int
2190zpool_vdev_online(zpool_handle_t *zhp, const char *path, int flags,
2191 vdev_state_t *newstate)
2192{
2193 zfs_cmd_t zc = { 0 };
2194 char msg[1024];
2195 nvlist_t *tgt;
2196 boolean_t avail_spare, l2cache, islog;
2197 libzfs_handle_t *hdl = zhp->zpool_hdl;
2198
2199 if (flags & ZFS_ONLINE_EXPAND) {
2200 (void) snprintf(msg, sizeof (msg),
2201 dgettext(TEXT_DOMAIN, "cannot expand %s"), path);
2202 } else {
2203 (void) snprintf(msg, sizeof (msg),
2204 dgettext(TEXT_DOMAIN, "cannot online %s"), path);
2205 }
2206
2207 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
2208 if ((tgt = zpool_find_vdev(zhp, path, &avail_spare, &l2cache,
2209 &islog)) == NULL)
2210 return (zfs_error(hdl, EZFS_NODEVICE, msg));
2211
2212 verify(nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID, &zc.zc_guid) == 0);
2213
2214 if (avail_spare)
2215 return (zfs_error(hdl, EZFS_ISSPARE, msg));
2216
2217 if (flags & ZFS_ONLINE_EXPAND ||
2218 zpool_get_prop_int(zhp, ZPOOL_PROP_AUTOEXPAND, NULL)) {
2219 char *pathname = NULL;
2220 uint64_t wholedisk = 0;
2221
2222 (void) nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_WHOLE_DISK,
2223 &wholedisk);
2224 verify(nvlist_lookup_string(tgt, ZPOOL_CONFIG_PATH,
2225 &pathname) == 0);
2226
2227 /*
2228 * XXX - L2ARC 1.0 devices can't support expansion.
2229 */
2230 if (l2cache) {
2231 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2232 "cannot expand cache devices"));
2233 return (zfs_error(hdl, EZFS_VDEVNOTSUP, msg));
2234 }
2235
2236 if (wholedisk) {
2237 pathname += strlen(DISK_ROOT) + 1;
2238 (void) zpool_relabel_disk(hdl, pathname);
2239 }
2240 }
2241
2242 zc.zc_cookie = VDEV_STATE_ONLINE;
2243 zc.zc_obj = flags;
2244
2245 if (zfs_ioctl(hdl, ZFS_IOC_VDEV_SET_STATE, &zc) != 0) {
2246 if (errno == EINVAL) {
2247 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "was split "
2248 "from this pool into a new one. Use '%s' "
2249 "instead"), "zpool detach");
2250 return (zfs_error(hdl, EZFS_POSTSPLIT_ONLINE, msg));
2251 }
2252 return (zpool_standard_error(hdl, errno, msg));
2253 }
2254
2255 *newstate = zc.zc_cookie;
2256 return (0);
2257}
2258
2259/*
2260 * Take the specified vdev offline
2261 */
2262int
2263zpool_vdev_offline(zpool_handle_t *zhp, const char *path, boolean_t istmp)
2264{
2265 zfs_cmd_t zc = { 0 };
2266 char msg[1024];
2267 nvlist_t *tgt;
2268 boolean_t avail_spare, l2cache;
2269 libzfs_handle_t *hdl = zhp->zpool_hdl;
2270
2271 (void) snprintf(msg, sizeof (msg),
2272 dgettext(TEXT_DOMAIN, "cannot offline %s"), path);
2273
2274 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
2275 if ((tgt = zpool_find_vdev(zhp, path, &avail_spare, &l2cache,
2276 NULL)) == NULL)
2277 return (zfs_error(hdl, EZFS_NODEVICE, msg));
2278
2279 verify(nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID, &zc.zc_guid) == 0);
2280
2281 if (avail_spare)
2282 return (zfs_error(hdl, EZFS_ISSPARE, msg));
2283
2284 zc.zc_cookie = VDEV_STATE_OFFLINE;
2285 zc.zc_obj = istmp ? ZFS_OFFLINE_TEMPORARY : 0;
2286
2287 if (zfs_ioctl(hdl, ZFS_IOC_VDEV_SET_STATE, &zc) == 0)
2288 return (0);
2289
2290 switch (errno) {
2291 case EBUSY:
2292
2293 /*
2294 * There are no other replicas of this device.
2295 */
2296 return (zfs_error(hdl, EZFS_NOREPLICAS, msg));
2297
2298 case EEXIST:
2299 /*
2300 * The log device has unplayed logs
2301 */
2302 return (zfs_error(hdl, EZFS_UNPLAYED_LOGS, msg));
2303
2304 default:
2305 return (zpool_standard_error(hdl, errno, msg));
2306 }
2307}
2308
2309/*
2310 * Mark the given vdev faulted.
2311 */
2312int
2313zpool_vdev_fault(zpool_handle_t *zhp, uint64_t guid, vdev_aux_t aux)
2314{
2315 zfs_cmd_t zc = { 0 };
2316 char msg[1024];
2317 libzfs_handle_t *hdl = zhp->zpool_hdl;
2318
2319 (void) snprintf(msg, sizeof (msg),
2320 dgettext(TEXT_DOMAIN, "cannot fault %llu"), guid);
2321
2322 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
2323 zc.zc_guid = guid;
2324 zc.zc_cookie = VDEV_STATE_FAULTED;
2325 zc.zc_obj = aux;
2326
2327 if (ioctl(hdl->libzfs_fd, ZFS_IOC_VDEV_SET_STATE, &zc) == 0)
2328 return (0);
2329
2330 switch (errno) {
2331 case EBUSY:
2332
2333 /*
2334 * There are no other replicas of this device.
2335 */
2336 return (zfs_error(hdl, EZFS_NOREPLICAS, msg));
2337
2338 default:
2339 return (zpool_standard_error(hdl, errno, msg));
2340 }
2341
2342}
2343
2344/*
2345 * Mark the given vdev degraded.
2346 */
2347int
2348zpool_vdev_degrade(zpool_handle_t *zhp, uint64_t guid, vdev_aux_t aux)
2349{
2350 zfs_cmd_t zc = { 0 };
2351 char msg[1024];
2352 libzfs_handle_t *hdl = zhp->zpool_hdl;
2353
2354 (void) snprintf(msg, sizeof (msg),
2355 dgettext(TEXT_DOMAIN, "cannot degrade %llu"), guid);
2356
2357 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
2358 zc.zc_guid = guid;
2359 zc.zc_cookie = VDEV_STATE_DEGRADED;
2360 zc.zc_obj = aux;
2361
2362 if (ioctl(hdl->libzfs_fd, ZFS_IOC_VDEV_SET_STATE, &zc) == 0)
2363 return (0);
2364
2365 return (zpool_standard_error(hdl, errno, msg));
2366}
2367
2368/*
2369 * Returns TRUE if the given nvlist is a vdev that was originally swapped in as
2370 * a hot spare.
2371 */
2372static boolean_t
2373is_replacing_spare(nvlist_t *search, nvlist_t *tgt, int which)
2374{
2375 nvlist_t **child;
2376 uint_t c, children;
2377 char *type;
2378
2379 if (nvlist_lookup_nvlist_array(search, ZPOOL_CONFIG_CHILDREN, &child,
2380 &children) == 0) {
2381 verify(nvlist_lookup_string(search, ZPOOL_CONFIG_TYPE,
2382 &type) == 0);
2383
2384 if (strcmp(type, VDEV_TYPE_SPARE) == 0 &&
2385 children == 2 && child[which] == tgt)
2386 return (B_TRUE);
2387
2388 for (c = 0; c < children; c++)
2389 if (is_replacing_spare(child[c], tgt, which))
2390 return (B_TRUE);
2391 }
2392
2393 return (B_FALSE);
2394}
2395
2396/*
2397 * Attach new_disk (fully described by nvroot) to old_disk.
2398 * If 'replacing' is specified, the new disk will replace the old one.
2399 */
2400int
2401zpool_vdev_attach(zpool_handle_t *zhp,
2402 const char *old_disk, const char *new_disk, nvlist_t *nvroot, int replacing)
2403{
2404 zfs_cmd_t zc = { 0 };
2405 char msg[1024];
2406 int ret;
2407 nvlist_t *tgt;
2408 boolean_t avail_spare, l2cache, islog;
2409 uint64_t val;
2410 char *newname;
2411 nvlist_t **child;
2412 uint_t children;
2413 nvlist_t *config_root;
2414 libzfs_handle_t *hdl = zhp->zpool_hdl;
2415 boolean_t rootpool = pool_is_bootable(zhp);
2416
2417 if (replacing)
2418 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN,
2419 "cannot replace %s with %s"), old_disk, new_disk);
2420 else
2421 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN,
2422 "cannot attach %s to %s"), new_disk, old_disk);
2423
2424 /*
2425 * If this is a root pool, make sure that we're not attaching an
2426 * EFI labeled device.
2427 */
2428 if (rootpool && pool_uses_efi(nvroot)) {
2429 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2430 "EFI labeled devices are not supported on root pools."));
2431 return (zfs_error(hdl, EZFS_POOL_NOTSUP, msg));
2432 }
2433
2434 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
2435 if ((tgt = zpool_find_vdev(zhp, old_disk, &avail_spare, &l2cache,
2436 &islog)) == 0)
2437 return (zfs_error(hdl, EZFS_NODEVICE, msg));
2438
2439 if (avail_spare)
2440 return (zfs_error(hdl, EZFS_ISSPARE, msg));
2441
2442 if (l2cache)
2443 return (zfs_error(hdl, EZFS_ISL2CACHE, msg));
2444
2445 verify(nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID, &zc.zc_guid) == 0);
2446 zc.zc_cookie = replacing;
2447
2448 if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_CHILDREN,
2449 &child, &children) != 0 || children != 1) {
2450 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2451 "new device must be a single disk"));
2452 return (zfs_error(hdl, EZFS_INVALCONFIG, msg));
2453 }
2454
2455 verify(nvlist_lookup_nvlist(zpool_get_config(zhp, NULL),
2456 ZPOOL_CONFIG_VDEV_TREE, &config_root) == 0);
2457
2458 if ((newname = zpool_vdev_name(NULL, NULL, child[0], B_FALSE)) == NULL)
2459 return (-1);
2460
2461 /*
2462 * If the target is a hot spare that has been swapped in, we can only
2463 * replace it with another hot spare.
2464 */
2465 if (replacing &&
2466 nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_IS_SPARE, &val) == 0 &&
2467 (zpool_find_vdev(zhp, newname, &avail_spare, &l2cache,
2468 NULL) == NULL || !avail_spare) &&
2469 is_replacing_spare(config_root, tgt, 1)) {
2470 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2471 "can only be replaced by another hot spare"));
2472 free(newname);
2473 return (zfs_error(hdl, EZFS_BADTARGET, msg));
2474 }
2475
2476 free(newname);
2477
2478 if (zcmd_write_conf_nvlist(hdl, &zc, nvroot) != 0)
2479 return (-1);
2480
2481 ret = zfs_ioctl(hdl, ZFS_IOC_VDEV_ATTACH, &zc);
2482
2483 zcmd_free_nvlists(&zc);
2484
2485 if (ret == 0) {
2486 if (rootpool) {
2487 /*
2488 * XXX need a better way to prevent user from
2489 * booting up a half-baked vdev.
2490 */
2491 (void) fprintf(stderr, dgettext(TEXT_DOMAIN, "Make "
2492 "sure to wait until resilver is done "
2493 "before rebooting.\n"));
2494 (void) fprintf(stderr, "\n");
2495 (void) fprintf(stderr, dgettext(TEXT_DOMAIN, "If "
2496 "you boot from pool '%s', you may need to update\n"
2497 "boot code on newly attached disk '%s'.\n\n"
2498 "Assuming you use GPT partitioning and 'da0' is "
2499 "your new boot disk\n"
2500 "you may use the following command:\n\n"
2501 "\tgpart bootcode -b /boot/pmbr -p "
2502 "/boot/gptzfsboot -i 1 da0\n\n"),
2503 zhp->zpool_name, new_disk);
2504 }
2505 return (0);
2506 }
2507
2508 switch (errno) {
2509 case ENOTSUP:
2510 /*
2511 * Can't attach to or replace this type of vdev.
2512 */
2513 if (replacing) {
2514 uint64_t version = zpool_get_prop_int(zhp,
2515 ZPOOL_PROP_VERSION, NULL);
2516
2517 if (islog)
2518 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2519 "cannot replace a log with a spare"));
2520 else if (version >= SPA_VERSION_MULTI_REPLACE)
2521 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2522 "already in replacing/spare config; wait "
2523 "for completion or use 'zpool detach'"));
2524 else
2525 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2526 "cannot replace a replacing device"));
2527 } else {
2528 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2529 "can only attach to mirrors and top-level "
2530 "disks"));
2531 }
2532 (void) zfs_error(hdl, EZFS_BADTARGET, msg);
2533 break;
2534
2535 case EINVAL:
2536 /*
2537 * The new device must be a single disk.
2538 */
2539 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2540 "new device must be a single disk"));
2541 (void) zfs_error(hdl, EZFS_INVALCONFIG, msg);
2542 break;
2543
2544 case EBUSY:
2545 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "%s is busy"),
2546 new_disk);
2547 (void) zfs_error(hdl, EZFS_BADDEV, msg);
2548 break;
2549
2550 case EOVERFLOW:
2551 /*
2552 * The new device is too small.
2553 */
2554 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2555 "device is too small"));
2556 (void) zfs_error(hdl, EZFS_BADDEV, msg);
2557 break;
2558
2559 case EDOM:
2560 /*
2561 * The new device has a different alignment requirement.
2562 */
2563 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2564 "devices have different sector alignment"));
2565 (void) zfs_error(hdl, EZFS_BADDEV, msg);
2566 break;
2567
2568 case ENAMETOOLONG:
2569 /*
2570 * The resulting top-level vdev spec won't fit in the label.
2571 */
2572 (void) zfs_error(hdl, EZFS_DEVOVERFLOW, msg);
2573 break;
2574
2575 default:
2576 (void) zpool_standard_error(hdl, errno, msg);
2577 }
2578
2579 return (-1);
2580}
2581
2582/*
2583 * Detach the specified device.
2584 */
2585int
2586zpool_vdev_detach(zpool_handle_t *zhp, const char *path)
2587{
2588 zfs_cmd_t zc = { 0 };
2589 char msg[1024];
2590 nvlist_t *tgt;
2591 boolean_t avail_spare, l2cache;
2592 libzfs_handle_t *hdl = zhp->zpool_hdl;
2593
2594 (void) snprintf(msg, sizeof (msg),
2595 dgettext(TEXT_DOMAIN, "cannot detach %s"), path);
2596
2597 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
2598 if ((tgt = zpool_find_vdev(zhp, path, &avail_spare, &l2cache,
2599 NULL)) == 0)
2600 return (zfs_error(hdl, EZFS_NODEVICE, msg));
2601
2602 if (avail_spare)
2603 return (zfs_error(hdl, EZFS_ISSPARE, msg));
2604
2605 if (l2cache)
2606 return (zfs_error(hdl, EZFS_ISL2CACHE, msg));
2607
2608 verify(nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID, &zc.zc_guid) == 0);
2609
2610 if (zfs_ioctl(hdl, ZFS_IOC_VDEV_DETACH, &zc) == 0)
2611 return (0);
2612
2613 switch (errno) {
2614
2615 case ENOTSUP:
2616 /*
2617 * Can't detach from this type of vdev.
2618 */
2619 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "only "
2620 "applicable to mirror and replacing vdevs"));
2621 (void) zfs_error(hdl, EZFS_BADTARGET, msg);
2622 break;
2623
2624 case EBUSY:
2625 /*
2626 * There are no other replicas of this device.
2627 */
2628 (void) zfs_error(hdl, EZFS_NOREPLICAS, msg);
2629 break;
2630
2631 default:
2632 (void) zpool_standard_error(hdl, errno, msg);
2633 }
2634
2635 return (-1);
2636}
2637
2638/*
2639 * Find a mirror vdev in the source nvlist.
2640 *
2641 * The mchild array contains a list of disks in one of the top-level mirrors
2642 * of the source pool. The schild array contains a list of disks that the
2643 * user specified on the command line. We loop over the mchild array to
2644 * see if any entry in the schild array matches.
2645 *
2646 * If a disk in the mchild array is found in the schild array, we return
2647 * the index of that entry. Otherwise we return -1.
2648 */
2649static int
2650find_vdev_entry(zpool_handle_t *zhp, nvlist_t **mchild, uint_t mchildren,
2651 nvlist_t **schild, uint_t schildren)
2652{
2653 uint_t mc;
2654
2655 for (mc = 0; mc < mchildren; mc++) {
2656 uint_t sc;
2657 char *mpath = zpool_vdev_name(zhp->zpool_hdl, zhp,
2658 mchild[mc], B_FALSE);
2659
2660 for (sc = 0; sc < schildren; sc++) {
2661 char *spath = zpool_vdev_name(zhp->zpool_hdl, zhp,
2662 schild[sc], B_FALSE);
2663 boolean_t result = (strcmp(mpath, spath) == 0);
2664
2665 free(spath);
2666 if (result) {
2667 free(mpath);
2668 return (mc);
2669 }
2670 }
2671
2672 free(mpath);
2673 }
2674
2675 return (-1);
2676}
2677
2678/*
2679 * Split a mirror pool. If newroot points to null, then a new nvlist
2680 * is generated and it is the responsibility of the caller to free it.
2681 */
2682int
2683zpool_vdev_split(zpool_handle_t *zhp, char *newname, nvlist_t **newroot,
2684 nvlist_t *props, splitflags_t flags)
2685{
2686 zfs_cmd_t zc = { 0 };
2687 char msg[1024];
2688 nvlist_t *tree, *config, **child, **newchild, *newconfig = NULL;
2689 nvlist_t **varray = NULL, *zc_props = NULL;
2690 uint_t c, children, newchildren, lastlog = 0, vcount, found = 0;
2691 libzfs_handle_t *hdl = zhp->zpool_hdl;
2692 uint64_t vers;
2693 boolean_t freelist = B_FALSE, memory_err = B_TRUE;
2694 int retval = 0;
2695
2696 (void) snprintf(msg, sizeof (msg),
2697 dgettext(TEXT_DOMAIN, "Unable to split %s"), zhp->zpool_name);
2698
2699 if (!zpool_name_valid(hdl, B_FALSE, newname))
2700 return (zfs_error(hdl, EZFS_INVALIDNAME, msg));
2701
2702 if ((config = zpool_get_config(zhp, NULL)) == NULL) {
2703 (void) fprintf(stderr, gettext("Internal error: unable to "
2704 "retrieve pool configuration\n"));
2705 return (-1);
2706 }
2707
2708 verify(nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE, &tree)
2709 == 0);
2710 verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_VERSION, &vers) == 0);
2711
2712 if (props) {
2713 prop_flags_t flags = { .create = B_FALSE, .import = B_TRUE };
2714 if ((zc_props = zpool_valid_proplist(hdl, zhp->zpool_name,
2715 props, vers, flags, msg)) == NULL)
2716 return (-1);
2717 }
2718
2719 if (nvlist_lookup_nvlist_array(tree, ZPOOL_CONFIG_CHILDREN, &child,
2720 &children) != 0) {
2721 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2722 "Source pool is missing vdev tree"));
2723 if (zc_props)
2724 nvlist_free(zc_props);
2725 return (-1);
2726 }
2727
2728 varray = zfs_alloc(hdl, children * sizeof (nvlist_t *));
2729 vcount = 0;
2730
2731 if (*newroot == NULL ||
2732 nvlist_lookup_nvlist_array(*newroot, ZPOOL_CONFIG_CHILDREN,
2733 &newchild, &newchildren) != 0)
2734 newchildren = 0;
2735
2736 for (c = 0; c < children; c++) {
2737 uint64_t is_log = B_FALSE, is_hole = B_FALSE;
2738 char *type;
2739 nvlist_t **mchild, *vdev;
2740 uint_t mchildren;
2741 int entry;
2742
2743 /*
2744 * Unlike cache & spares, slogs are stored in the
2745 * ZPOOL_CONFIG_CHILDREN array. We filter them out here.
2746 */
2747 (void) nvlist_lookup_uint64(child[c], ZPOOL_CONFIG_IS_LOG,
2748 &is_log);
2749 (void) nvlist_lookup_uint64(child[c], ZPOOL_CONFIG_IS_HOLE,
2750 &is_hole);
2751 if (is_log || is_hole) {
2752 /*
2753 * Create a hole vdev and put it in the config.
2754 */
2755 if (nvlist_alloc(&vdev, NV_UNIQUE_NAME, 0) != 0)
2756 goto out;
2757 if (nvlist_add_string(vdev, ZPOOL_CONFIG_TYPE,
2758 VDEV_TYPE_HOLE) != 0)
2759 goto out;
2760 if (nvlist_add_uint64(vdev, ZPOOL_CONFIG_IS_HOLE,
2761 1) != 0)
2762 goto out;
2763 if (lastlog == 0)
2764 lastlog = vcount;
2765 varray[vcount++] = vdev;
2766 continue;
2767 }
2768 lastlog = 0;
2769 verify(nvlist_lookup_string(child[c], ZPOOL_CONFIG_TYPE, &type)
2770 == 0);
2771 if (strcmp(type, VDEV_TYPE_MIRROR) != 0) {
2772 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2773 "Source pool must be composed only of mirrors\n"));
2774 retval = zfs_error(hdl, EZFS_INVALCONFIG, msg);
2775 goto out;
2776 }
2777
2778 verify(nvlist_lookup_nvlist_array(child[c],
2779 ZPOOL_CONFIG_CHILDREN, &mchild, &mchildren) == 0);
2780
2781 /* find or add an entry for this top-level vdev */
2782 if (newchildren > 0 &&
2783 (entry = find_vdev_entry(zhp, mchild, mchildren,
2784 newchild, newchildren)) >= 0) {
2785 /* We found a disk that the user specified. */
2786 vdev = mchild[entry];
2787 ++found;
2788 } else {
2789 /* User didn't specify a disk for this vdev. */
2790 vdev = mchild[mchildren - 1];
2791 }
2792
2793 if (nvlist_dup(vdev, &varray[vcount++], 0) != 0)
2794 goto out;
2795 }
2796
2797 /* did we find every disk the user specified? */
2798 if (found != newchildren) {
2799 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "Device list must "
2800 "include at most one disk from each mirror"));
2801 retval = zfs_error(hdl, EZFS_INVALCONFIG, msg);
2802 goto out;
2803 }
2804
2805 /* Prepare the nvlist for populating. */
2806 if (*newroot == NULL) {
2807 if (nvlist_alloc(newroot, NV_UNIQUE_NAME, 0) != 0)
2808 goto out;
2809 freelist = B_TRUE;
2810 if (nvlist_add_string(*newroot, ZPOOL_CONFIG_TYPE,
2811 VDEV_TYPE_ROOT) != 0)
2812 goto out;
2813 } else {
2814 verify(nvlist_remove_all(*newroot, ZPOOL_CONFIG_CHILDREN) == 0);
2815 }
2816
2817 /* Add all the children we found */
2818 if (nvlist_add_nvlist_array(*newroot, ZPOOL_CONFIG_CHILDREN, varray,
2819 lastlog == 0 ? vcount : lastlog) != 0)
2820 goto out;
2821
2822 /*
2823 * If we're just doing a dry run, exit now with success.
2824 */
2825 if (flags.dryrun) {
2826 memory_err = B_FALSE;
2827 freelist = B_FALSE;
2828 goto out;
2829 }
2830
2831 /* now build up the config list & call the ioctl */
2832 if (nvlist_alloc(&newconfig, NV_UNIQUE_NAME, 0) != 0)
2833 goto out;
2834
2835 if (nvlist_add_nvlist(newconfig,
2836 ZPOOL_CONFIG_VDEV_TREE, *newroot) != 0 ||
2837 nvlist_add_string(newconfig,
2838 ZPOOL_CONFIG_POOL_NAME, newname) != 0 ||
2839 nvlist_add_uint64(newconfig, ZPOOL_CONFIG_VERSION, vers) != 0)
2840 goto out;
2841
2842 /*
2843 * The new pool is automatically part of the namespace unless we
2844 * explicitly export it.
2845 */
2846 if (!flags.import)
2847 zc.zc_cookie = ZPOOL_EXPORT_AFTER_SPLIT;
2848 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
2849 (void) strlcpy(zc.zc_string, newname, sizeof (zc.zc_string));
2850 if (zcmd_write_conf_nvlist(hdl, &zc, newconfig) != 0)
2851 goto out;
2852 if (zc_props != NULL && zcmd_write_src_nvlist(hdl, &zc, zc_props) != 0)
2853 goto out;
2854
2855 if (zfs_ioctl(hdl, ZFS_IOC_VDEV_SPLIT, &zc) != 0) {
2856 retval = zpool_standard_error(hdl, errno, msg);
2857 goto out;
2858 }
2859
2860 freelist = B_FALSE;
2861 memory_err = B_FALSE;
2862
2863out:
2864 if (varray != NULL) {
2865 int v;
2866
2867 for (v = 0; v < vcount; v++)
2868 nvlist_free(varray[v]);
2869 free(varray);
2870 }
2871 zcmd_free_nvlists(&zc);
2872 if (zc_props)
2873 nvlist_free(zc_props);
2874 if (newconfig)
2875 nvlist_free(newconfig);
2876 if (freelist) {
2877 nvlist_free(*newroot);
2878 *newroot = NULL;
2879 }
2880
2881 if (retval != 0)
2882 return (retval);
2883
2884 if (memory_err)
2885 return (no_memory(hdl));
2886
2887 return (0);
2888}
2889
2890/*
2891 * Remove the given device. Currently, this is supported only for hot spares
2892 * and level 2 cache devices.
2893 */
2894int
2895zpool_vdev_remove(zpool_handle_t *zhp, const char *path)
2896{
2897 zfs_cmd_t zc = { 0 };
2898 char msg[1024];
2899 nvlist_t *tgt;
2900 boolean_t avail_spare, l2cache, islog;
2901 libzfs_handle_t *hdl = zhp->zpool_hdl;
2902 uint64_t version;
2903
2904 (void) snprintf(msg, sizeof (msg),
2905 dgettext(TEXT_DOMAIN, "cannot remove %s"), path);
2906
2907 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
2908 if ((tgt = zpool_find_vdev(zhp, path, &avail_spare, &l2cache,
2909 &islog)) == 0)
2910 return (zfs_error(hdl, EZFS_NODEVICE, msg));
2911 /*
2912 * XXX - this should just go away.
2913 */
2914 if (!avail_spare && !l2cache && !islog) {
2915 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2916 "only inactive hot spares, cache, top-level, "
2917 "or log devices can be removed"));
2918 return (zfs_error(hdl, EZFS_NODEVICE, msg));
2919 }
2920
2921 version = zpool_get_prop_int(zhp, ZPOOL_PROP_VERSION, NULL);
2922 if (islog && version < SPA_VERSION_HOLES) {
2923 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2924 "pool must be upgrade to support log removal"));
2925 return (zfs_error(hdl, EZFS_BADVERSION, msg));
2926 }
2927
2928 verify(nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID, &zc.zc_guid) == 0);
2929
2930 if (zfs_ioctl(hdl, ZFS_IOC_VDEV_REMOVE, &zc) == 0)
2931 return (0);
2932
2933 return (zpool_standard_error(hdl, errno, msg));
2934}
2935
2936/*
2937 * Clear the errors for the pool, or the particular device if specified.
2938 */
2939int
2940zpool_clear(zpool_handle_t *zhp, const char *path, nvlist_t *rewindnvl)
2941{
2942 zfs_cmd_t zc = { 0 };
2943 char msg[1024];
2944 nvlist_t *tgt;
2945 zpool_rewind_policy_t policy;
2946 boolean_t avail_spare, l2cache;
2947 libzfs_handle_t *hdl = zhp->zpool_hdl;
2948 nvlist_t *nvi = NULL;
2949 int error;
2950
2951 if (path)
2952 (void) snprintf(msg, sizeof (msg),
2953 dgettext(TEXT_DOMAIN, "cannot clear errors for %s"),
2954 path);
2955 else
2956 (void) snprintf(msg, sizeof (msg),
2957 dgettext(TEXT_DOMAIN, "cannot clear errors for %s"),
2958 zhp->zpool_name);
2959
2960 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
2961 if (path) {
2962 if ((tgt = zpool_find_vdev(zhp, path, &avail_spare,
2963 &l2cache, NULL)) == 0)
2964 return (zfs_error(hdl, EZFS_NODEVICE, msg));
2965
2966 /*
2967 * Don't allow error clearing for hot spares. Do allow
2968 * error clearing for l2cache devices.
2969 */
2970 if (avail_spare)
2971 return (zfs_error(hdl, EZFS_ISSPARE, msg));
2972
2973 verify(nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID,
2974 &zc.zc_guid) == 0);
2975 }
2976
2977 zpool_get_rewind_policy(rewindnvl, &policy);
2978 zc.zc_cookie = policy.zrp_request;
2979
2980 if (zcmd_alloc_dst_nvlist(hdl, &zc, zhp->zpool_config_size * 2) != 0)
2981 return (-1);
2982
2983 if (zcmd_write_src_nvlist(hdl, &zc, rewindnvl) != 0)
2984 return (-1);
2985
2986 while ((error = zfs_ioctl(hdl, ZFS_IOC_CLEAR, &zc)) != 0 &&
2987 errno == ENOMEM) {
2988 if (zcmd_expand_dst_nvlist(hdl, &zc) != 0) {
2989 zcmd_free_nvlists(&zc);
2990 return (-1);
2991 }
2992 }
2993
2994 if (!error || ((policy.zrp_request & ZPOOL_TRY_REWIND) &&
2995 errno != EPERM && errno != EACCES)) {
2996 if (policy.zrp_request &
2997 (ZPOOL_DO_REWIND | ZPOOL_TRY_REWIND)) {
2998 (void) zcmd_read_dst_nvlist(hdl, &zc, &nvi);
2999 zpool_rewind_exclaim(hdl, zc.zc_name,
3000 ((policy.zrp_request & ZPOOL_TRY_REWIND) != 0),
3001 nvi);
3002 nvlist_free(nvi);
3003 }
3004 zcmd_free_nvlists(&zc);
3005 return (0);
3006 }
3007
3008 zcmd_free_nvlists(&zc);
3009 return (zpool_standard_error(hdl, errno, msg));
3010}
3011
3012/*
3013 * Similar to zpool_clear(), but takes a GUID (used by fmd).
3014 */
3015int
3016zpool_vdev_clear(zpool_handle_t *zhp, uint64_t guid)
3017{
3018 zfs_cmd_t zc = { 0 };
3019 char msg[1024];
3020 libzfs_handle_t *hdl = zhp->zpool_hdl;
3021
3022 (void) snprintf(msg, sizeof (msg),
3023 dgettext(TEXT_DOMAIN, "cannot clear errors for %llx"),
3024 guid);
3025
3026 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
3027 zc.zc_guid = guid;
3028 zc.zc_cookie = ZPOOL_NO_REWIND;
3029
3030 if (ioctl(hdl->libzfs_fd, ZFS_IOC_CLEAR, &zc) == 0)
3031 return (0);
3032
3033 return (zpool_standard_error(hdl, errno, msg));
3034}
3035
3036/*
3037 * Change the GUID for a pool.
3038 */
3039int
3040zpool_reguid(zpool_handle_t *zhp)
3041{
3042 char msg[1024];
3043 libzfs_handle_t *hdl = zhp->zpool_hdl;
3044 zfs_cmd_t zc = { 0 };
3045
3046 (void) snprintf(msg, sizeof (msg),
3047 dgettext(TEXT_DOMAIN, "cannot reguid '%s'"), zhp->zpool_name);
3048
3049 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
3050 if (zfs_ioctl(hdl, ZFS_IOC_POOL_REGUID, &zc) == 0)
3051 return (0);
3052
3053 return (zpool_standard_error(hdl, errno, msg));
3054}
3055
3056/*
3057 * Convert from a devid string to a path.
3058 */
3059static char *
3060devid_to_path(char *devid_str)
3061{
3062 ddi_devid_t devid;
3063 char *minor;
3064 char *path;
3065 devid_nmlist_t *list = NULL;
3066 int ret;
3067
3068 if (devid_str_decode(devid_str, &devid, &minor) != 0)
3069 return (NULL);
3070
3071 ret = devid_deviceid_to_nmlist("/dev", devid, minor, &list);
3072
3073 devid_str_free(minor);
3074 devid_free(devid);
3075
3076 if (ret != 0)
3077 return (NULL);
3078
3079 if ((path = strdup(list[0].devname)) == NULL)
3080 return (NULL);
3081
3082 devid_free_nmlist(list);
3083
3084 return (path);
3085}
3086
3087/*
3088 * Convert from a path to a devid string.
3089 */
3090static char *
3091path_to_devid(const char *path)
3092{
3093 int fd;
3094 ddi_devid_t devid;
3095 char *minor, *ret;
3096
3097 if ((fd = open(path, O_RDONLY)) < 0)
3098 return (NULL);
3099
3100 minor = NULL;
3101 ret = NULL;
3102 if (devid_get(fd, &devid) == 0) {
3103 if (devid_get_minor_name(fd, &minor) == 0)
3104 ret = devid_str_encode(devid, minor);
3105 if (minor != NULL)
3106 devid_str_free(minor);
3107 devid_free(devid);
3108 }
3109 (void) close(fd);
3110
3111 return (ret);
3112}
3113
3114/*
3115 * Issue the necessary ioctl() to update the stored path value for the vdev. We
3116 * ignore any failure here, since a common case is for an unprivileged user to
3117 * type 'zpool status', and we'll display the correct information anyway.
3118 */
3119static void
3120set_path(zpool_handle_t *zhp, nvlist_t *nv, const char *path)
3121{
3122 zfs_cmd_t zc = { 0 };
3123
3124 (void) strncpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
3125 (void) strncpy(zc.zc_value, path, sizeof (zc.zc_value));
3126 verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID,
3127 &zc.zc_guid) == 0);
3128
3129 (void) ioctl(zhp->zpool_hdl->libzfs_fd, ZFS_IOC_VDEV_SETPATH, &zc);
3130}
3131
3132/*
3133 * Given a vdev, return the name to display in iostat. If the vdev has a path,
3134 * we use that, stripping off any leading "/dev/dsk/"; if not, we use the type.
3135 * We also check if this is a whole disk, in which case we strip off the
3136 * trailing 's0' slice name.
3137 *
3138 * This routine is also responsible for identifying when disks have been
3139 * reconfigured in a new location. The kernel will have opened the device by
3140 * devid, but the path will still refer to the old location. To catch this, we
3141 * first do a path -> devid translation (which is fast for the common case). If
3142 * the devid matches, we're done. If not, we do a reverse devid -> path
3143 * translation and issue the appropriate ioctl() to update the path of the vdev.
3144 * If 'zhp' is NULL, then this is an exported pool, and we don't need to do any
3145 * of these checks.
3146 */
3147char *
3148zpool_vdev_name(libzfs_handle_t *hdl, zpool_handle_t *zhp, nvlist_t *nv,
3149 boolean_t verbose)
3150{
3151 char *path, *devid;
3152 uint64_t value;
3153 char buf[64];
3154 vdev_stat_t *vs;
3155 uint_t vsc;
3156 int have_stats;
3157 int have_path;
3158
3159 have_stats = nvlist_lookup_uint64_array(nv, ZPOOL_CONFIG_VDEV_STATS,
3160 (uint64_t **)&vs, &vsc) == 0;
3161 have_path = nvlist_lookup_string(nv, ZPOOL_CONFIG_PATH, &path) == 0;
3162
3163 /*
3164 * If the device is not currently present, assume it will not
3165 * come back at the same device path. Display the device by GUID.
3166 */
3167 if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_NOT_PRESENT, &value) == 0 ||
3168 have_path && have_stats && vs->vs_state <= VDEV_STATE_CANT_OPEN) {
3169 verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID,
3170 &value) == 0);
3171 (void) snprintf(buf, sizeof (buf), "%llu",
3172 (u_longlong_t)value);
3173 path = buf;
3174 } else if (have_path) {
3175
3176 /*
3177 * If the device is dead (faulted, offline, etc) then don't
3178 * bother opening it. Otherwise we may be forcing the user to
3179 * open a misbehaving device, which can have undesirable
3180 * effects.
3181 */
3182 if ((have_stats == 0 ||
3183 vs->vs_state >= VDEV_STATE_DEGRADED) &&
3184 zhp != NULL &&
3185 nvlist_lookup_string(nv, ZPOOL_CONFIG_DEVID, &devid) == 0) {
3186 /*
3187 * Determine if the current path is correct.
3188 */
3189 char *newdevid = path_to_devid(path);
3190
3191 if (newdevid == NULL ||
3192 strcmp(devid, newdevid) != 0) {
3193 char *newpath;
3194
3195 if ((newpath = devid_to_path(devid)) != NULL) {
3196 /*
3197 * Update the path appropriately.
3198 */
3199 set_path(zhp, nv, newpath);
3200 if (nvlist_add_string(nv,
3201 ZPOOL_CONFIG_PATH, newpath) == 0)
3202 verify(nvlist_lookup_string(nv,
3203 ZPOOL_CONFIG_PATH,
3204 &path) == 0);
3205 free(newpath);
3206 }
3207 }
3208
3209 if (newdevid)
3210 devid_str_free(newdevid);
3211 }
3212
3213#ifdef sun
3214 if (strncmp(path, "/dev/dsk/", 9) == 0)
3215 path += 9;
3216
3217 if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_WHOLE_DISK,
3218 &value) == 0 && value) {
3219 int pathlen = strlen(path);
3220 char *tmp = zfs_strdup(hdl, path);
3221
3222 /*
3223 * If it starts with c#, and ends with "s0", chop
3224 * the "s0" off, or if it ends with "s0/old", remove
3225 * the "s0" from the middle.
3226 */
3227 if (CTD_CHECK(tmp)) {
3228 if (strcmp(&tmp[pathlen - 2], "s0") == 0) {
3229 tmp[pathlen - 2] = '\0';
3230 } else if (pathlen > 6 &&
3231 strcmp(&tmp[pathlen - 6], "s0/old") == 0) {
3232 (void) strcpy(&tmp[pathlen - 6],
3233 "/old");
3234 }
3235 }
3236 return (tmp);
3237 }
3238#else /* !sun */
3239 if (strncmp(path, _PATH_DEV, sizeof(_PATH_DEV) - 1) == 0)
3240 path += sizeof(_PATH_DEV) - 1;
3241#endif /* !sun */
3242 } else {
3243 verify(nvlist_lookup_string(nv, ZPOOL_CONFIG_TYPE, &path) == 0);
3244
3245 /*
3246 * If it's a raidz device, we need to stick in the parity level.
3247 */
3248 if (strcmp(path, VDEV_TYPE_RAIDZ) == 0) {
3249 verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_NPARITY,
3250 &value) == 0);
3251 (void) snprintf(buf, sizeof (buf), "%s%llu", path,
3252 (u_longlong_t)value);
3253 path = buf;
3254 }
3255
3256 /*
3257 * We identify each top-level vdev by using a <type-id>
3258 * naming convention.
3259 */
3260 if (verbose) {
3261 uint64_t id;
3262
3263 verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_ID,
3264 &id) == 0);
3265 (void) snprintf(buf, sizeof (buf), "%s-%llu", path,
3266 (u_longlong_t)id);
3267 path = buf;
3268 }
3269 }
3270
3271 return (zfs_strdup(hdl, path));
3272}
3273
3274static int
3275zbookmark_compare(const void *a, const void *b)
3276{
3277 return (memcmp(a, b, sizeof (zbookmark_t)));
3278}
3279
3280/*
3281 * Retrieve the persistent error log, uniquify the members, and return to the
3282 * caller.
3283 */
3284int
3285zpool_get_errlog(zpool_handle_t *zhp, nvlist_t **nverrlistp)
3286{
3287 zfs_cmd_t zc = { 0 };
3288 uint64_t count;
3289 zbookmark_t *zb = NULL;
3290 int i;
3291
3292 /*
3293 * Retrieve the raw error list from the kernel. If the number of errors
3294 * has increased, allocate more space and continue until we get the
3295 * entire list.
3296 */
3297 verify(nvlist_lookup_uint64(zhp->zpool_config, ZPOOL_CONFIG_ERRCOUNT,
3298 &count) == 0);
3299 if (count == 0)
3300 return (0);
3301 if ((zc.zc_nvlist_dst = (uintptr_t)zfs_alloc(zhp->zpool_hdl,
3302 count * sizeof (zbookmark_t))) == (uintptr_t)NULL)
3303 return (-1);
3304 zc.zc_nvlist_dst_size = count;
3305 (void) strcpy(zc.zc_name, zhp->zpool_name);
3306 for (;;) {
3307 if (ioctl(zhp->zpool_hdl->libzfs_fd, ZFS_IOC_ERROR_LOG,
3308 &zc) != 0) {
3309 free((void *)(uintptr_t)zc.zc_nvlist_dst);
3310 if (errno == ENOMEM) {
3311 count = zc.zc_nvlist_dst_size;
3312 if ((zc.zc_nvlist_dst = (uintptr_t)
3313 zfs_alloc(zhp->zpool_hdl, count *
3314 sizeof (zbookmark_t))) == (uintptr_t)NULL)
3315 return (-1);
3316 } else {
3317 return (-1);
3318 }
3319 } else {
3320 break;
3321 }
3322 }
3323
3324 /*
3325 * Sort the resulting bookmarks. This is a little confusing due to the
3326 * implementation of ZFS_IOC_ERROR_LOG. The bookmarks are copied last
3327 * to first, and 'zc_nvlist_dst_size' indicates the number of boomarks
3328 * _not_ copied as part of the process. So we point the start of our
3329 * array appropriate and decrement the total number of elements.
3330 */
3331 zb = ((zbookmark_t *)(uintptr_t)zc.zc_nvlist_dst) +
3332 zc.zc_nvlist_dst_size;
3333 count -= zc.zc_nvlist_dst_size;
3334
3335 qsort(zb, count, sizeof (zbookmark_t), zbookmark_compare);
3336
3337 verify(nvlist_alloc(nverrlistp, 0, KM_SLEEP) == 0);
3338
3339 /*
3340 * Fill in the nverrlistp with nvlist's of dataset and object numbers.
3341 */
3342 for (i = 0; i < count; i++) {
3343 nvlist_t *nv;
3344
3345 /* ignoring zb_blkid and zb_level for now */
3346 if (i > 0 && zb[i-1].zb_objset == zb[i].zb_objset &&
3347 zb[i-1].zb_object == zb[i].zb_object)
3348 continue;
3349
3350 if (nvlist_alloc(&nv, NV_UNIQUE_NAME, KM_SLEEP) != 0)
3351 goto nomem;
3352 if (nvlist_add_uint64(nv, ZPOOL_ERR_DATASET,
3353 zb[i].zb_objset) != 0) {
3354 nvlist_free(nv);
3355 goto nomem;
3356 }
3357 if (nvlist_add_uint64(nv, ZPOOL_ERR_OBJECT,
3358 zb[i].zb_object) != 0) {
3359 nvlist_free(nv);
3360 goto nomem;
3361 }
3362 if (nvlist_add_nvlist(*nverrlistp, "ejk", nv) != 0) {
3363 nvlist_free(nv);
3364 goto nomem;
3365 }
3366 nvlist_free(nv);
3367 }
3368
3369 free((void *)(uintptr_t)zc.zc_nvlist_dst);
3370 return (0);
3371
3372nomem:
3373 free((void *)(uintptr_t)zc.zc_nvlist_dst);
3374 return (no_memory(zhp->zpool_hdl));
3375}
3376
3377/*
3378 * Upgrade a ZFS pool to the latest on-disk version.
3379 */
3380int
3381zpool_upgrade(zpool_handle_t *zhp, uint64_t new_version)
3382{
3383 zfs_cmd_t zc = { 0 };
3384 libzfs_handle_t *hdl = zhp->zpool_hdl;
3385
3386 (void) strcpy(zc.zc_name, zhp->zpool_name);
3387 zc.zc_cookie = new_version;
3388
3389 if (zfs_ioctl(hdl, ZFS_IOC_POOL_UPGRADE, &zc) != 0)
3390 return (zpool_standard_error_fmt(hdl, errno,
3391 dgettext(TEXT_DOMAIN, "cannot upgrade '%s'"),
3392 zhp->zpool_name));
3393 return (0);
3394}
3395
3396void
3397zpool_set_history_str(const char *subcommand, int argc, char **argv,
3398 char *history_str)
3399{
3400 int i;
3401
3402 (void) strlcpy(history_str, subcommand, HIS_MAX_RECORD_LEN);
3403 for (i = 1; i < argc; i++) {
3404 if (strlen(history_str) + 1 + strlen(argv[i]) >
3405 HIS_MAX_RECORD_LEN)
3406 break;
3407 (void) strlcat(history_str, " ", HIS_MAX_RECORD_LEN);
3408 (void) strlcat(history_str, argv[i], HIS_MAX_RECORD_LEN);
3409 }
3410}
3411
3412/*
3413 * Stage command history for logging.
3414 */
3415int
3416zpool_stage_history(libzfs_handle_t *hdl, const char *history_str)
3417{
3418 if (history_str == NULL)
3419 return (EINVAL);
3420
3421 if (strlen(history_str) > HIS_MAX_RECORD_LEN)
3422 return (EINVAL);
3423
3424 if (hdl->libzfs_log_str != NULL)
3425 free(hdl->libzfs_log_str);
3426
3427 if ((hdl->libzfs_log_str = strdup(history_str)) == NULL)
3428 return (no_memory(hdl));
3429
3430 return (0);
3431}
3432
3433/*
3434 * Perform ioctl to get some command history of a pool.
3435 *
3436 * 'buf' is the buffer to fill up to 'len' bytes. 'off' is the
3437 * logical offset of the history buffer to start reading from.
3438 *
3439 * Upon return, 'off' is the next logical offset to read from and
3440 * 'len' is the actual amount of bytes read into 'buf'.
3441 */
3442static int
3443get_history(zpool_handle_t *zhp, char *buf, uint64_t *off, uint64_t *len)
3444{
3445 zfs_cmd_t zc = { 0 };
3446 libzfs_handle_t *hdl = zhp->zpool_hdl;
3447
3448 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
3449
3450 zc.zc_history = (uint64_t)(uintptr_t)buf;
3451 zc.zc_history_len = *len;
3452 zc.zc_history_offset = *off;
3453
3454 if (ioctl(hdl->libzfs_fd, ZFS_IOC_POOL_GET_HISTORY, &zc) != 0) {
3455 switch (errno) {
3456 case EPERM:
3457 return (zfs_error_fmt(hdl, EZFS_PERM,
3458 dgettext(TEXT_DOMAIN,
3459 "cannot show history for pool '%s'"),
3460 zhp->zpool_name));
3461 case ENOENT:
3462 return (zfs_error_fmt(hdl, EZFS_NOHISTORY,
3463 dgettext(TEXT_DOMAIN, "cannot get history for pool "
3464 "'%s'"), zhp->zpool_name));
3465 case ENOTSUP:
3466 return (zfs_error_fmt(hdl, EZFS_BADVERSION,
3467 dgettext(TEXT_DOMAIN, "cannot get history for pool "
3468 "'%s', pool must be upgraded"), zhp->zpool_name));
3469 default:
3470 return (zpool_standard_error_fmt(hdl, errno,
3471 dgettext(TEXT_DOMAIN,
3472 "cannot get history for '%s'"), zhp->zpool_name));
3473 }
3474 }
3475
3476 *len = zc.zc_history_len;
3477 *off = zc.zc_history_offset;
3478
3479 return (0);
3480}
3481
3482/*
3483 * Process the buffer of nvlists, unpacking and storing each nvlist record
3484 * into 'records'. 'leftover' is set to the number of bytes that weren't
3485 * processed as there wasn't a complete record.
3486 */
3487int
3488zpool_history_unpack(char *buf, uint64_t bytes_read, uint64_t *leftover,
3489 nvlist_t ***records, uint_t *numrecords)
3490{
3491 uint64_t reclen;
3492 nvlist_t *nv;
3493 int i;
3494
3495 while (bytes_read > sizeof (reclen)) {
3496
3497 /* get length of packed record (stored as little endian) */
3498 for (i = 0, reclen = 0; i < sizeof (reclen); i++)
3499 reclen += (uint64_t)(((uchar_t *)buf)[i]) << (8*i);
3500
3501 if (bytes_read < sizeof (reclen) + reclen)
3502 break;
3503
3504 /* unpack record */
3505 if (nvlist_unpack(buf + sizeof (reclen), reclen, &nv, 0) != 0)
3506 return (ENOMEM);
3507 bytes_read -= sizeof (reclen) + reclen;
3508 buf += sizeof (reclen) + reclen;
3509
3510 /* add record to nvlist array */
3511 (*numrecords)++;
3512 if (ISP2(*numrecords + 1)) {
3513 *records = realloc(*records,
3514 *numrecords * 2 * sizeof (nvlist_t *));
3515 }
3516 (*records)[*numrecords - 1] = nv;
3517 }
3518
3519 *leftover = bytes_read;
3520 return (0);
3521}
3522
3523#define HIS_BUF_LEN (128*1024)
3524
3525/*
3526 * Retrieve the command history of a pool.
3527 */
3528int
3529zpool_get_history(zpool_handle_t *zhp, nvlist_t **nvhisp)
3530{
3531 char buf[HIS_BUF_LEN];
3532 uint64_t off = 0;
3533 nvlist_t **records = NULL;
3534 uint_t numrecords = 0;
3535 int err, i;
3536
3537 do {
3538 uint64_t bytes_read = sizeof (buf);
3539 uint64_t leftover;
3540
3541 if ((err = get_history(zhp, buf, &off, &bytes_read)) != 0)
3542 break;
3543
3544 /* if nothing else was read in, we're at EOF, just return */
3545 if (!bytes_read)
3546 break;
3547
3548 if ((err = zpool_history_unpack(buf, bytes_read,
3549 &leftover, &records, &numrecords)) != 0)
3550 break;
3551 off -= leftover;
3552
3553 /* CONSTCOND */
3554 } while (1);
3555
3556 if (!err) {
3557 verify(nvlist_alloc(nvhisp, NV_UNIQUE_NAME, 0) == 0);
3558 verify(nvlist_add_nvlist_array(*nvhisp, ZPOOL_HIST_RECORD,
3559 records, numrecords) == 0);
3560 }
3561 for (i = 0; i < numrecords; i++)
3562 nvlist_free(records[i]);
3563 free(records);
3564
3565 return (err);
3566}
3567
3568void
3569zpool_obj_to_path(zpool_handle_t *zhp, uint64_t dsobj, uint64_t obj,
3570 char *pathname, size_t len)
3571{
3572 zfs_cmd_t zc = { 0 };
3573 boolean_t mounted = B_FALSE;
3574 char *mntpnt = NULL;
3575 char dsname[MAXNAMELEN];
3576
3577 if (dsobj == 0) {
3578 /* special case for the MOS */
3579 (void) snprintf(pathname, len, "<metadata>:<0x%llx>", obj);
3580 return;
3581 }
3582
3583 /* get the dataset's name */
3584 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
3585 zc.zc_obj = dsobj;
3586 if (ioctl(zhp->zpool_hdl->libzfs_fd,
3587 ZFS_IOC_DSOBJ_TO_DSNAME, &zc) != 0) {
3588 /* just write out a path of two object numbers */
3589 (void) snprintf(pathname, len, "<0x%llx>:<0x%llx>",
3590 dsobj, obj);
3591 return;
3592 }
3593 (void) strlcpy(dsname, zc.zc_value, sizeof (dsname));
3594
3595 /* find out if the dataset is mounted */
3596 mounted = is_mounted(zhp->zpool_hdl, dsname, &mntpnt);
3597
3598 /* get the corrupted object's path */
3599 (void) strlcpy(zc.zc_name, dsname, sizeof (zc.zc_name));
3600 zc.zc_obj = obj;
3601 if (ioctl(zhp->zpool_hdl->libzfs_fd, ZFS_IOC_OBJ_TO_PATH,
3602 &zc) == 0) {
3603 if (mounted) {
3604 (void) snprintf(pathname, len, "%s%s", mntpnt,
3605 zc.zc_value);
3606 } else {
3607 (void) snprintf(pathname, len, "%s:%s",
3608 dsname, zc.zc_value);
3609 }
3610 } else {
3611 (void) snprintf(pathname, len, "%s:<0x%llx>", dsname, obj);
3612 }
3613 free(mntpnt);
3614}
3615
3616#ifdef sun
3617/*
3618 * Read the EFI label from the config, if a label does not exist then
3619 * pass back the error to the caller. If the caller has passed a non-NULL
3620 * diskaddr argument then we set it to the starting address of the EFI
3621 * partition.
3622 */
3623static int
3624read_efi_label(nvlist_t *config, diskaddr_t *sb)
3625{
3626 char *path;
3627 int fd;
3628 char diskname[MAXPATHLEN];
3629 int err = -1;
3630
3631 if (nvlist_lookup_string(config, ZPOOL_CONFIG_PATH, &path) != 0)
3632 return (err);
3633
3634 (void) snprintf(diskname, sizeof (diskname), "%s%s", RDISK_ROOT,
3635 strrchr(path, '/'));
3636 if ((fd = open(diskname, O_RDONLY|O_NDELAY)) >= 0) {
3637 struct dk_gpt *vtoc;
3638
3639 if ((err = efi_alloc_and_read(fd, &vtoc)) >= 0) {
3640 if (sb != NULL)
3641 *sb = vtoc->efi_parts[0].p_start;
3642 efi_free(vtoc);
3643 }
3644 (void) close(fd);
3645 }
3646 return (err);
3647}
3648
3649/*
3650 * determine where a partition starts on a disk in the current
3651 * configuration
3652 */
3653static diskaddr_t
3654find_start_block(nvlist_t *config)
3655{
3656 nvlist_t **child;
3657 uint_t c, children;
3658 diskaddr_t sb = MAXOFFSET_T;
3659 uint64_t wholedisk;
3660
3661 if (nvlist_lookup_nvlist_array(config,
3662 ZPOOL_CONFIG_CHILDREN, &child, &children) != 0) {
3663 if (nvlist_lookup_uint64(config,
3664 ZPOOL_CONFIG_WHOLE_DISK,
3665 &wholedisk) != 0 || !wholedisk) {
3666 return (MAXOFFSET_T);
3667 }
3668 if (read_efi_label(config, &sb) < 0)
3669 sb = MAXOFFSET_T;
3670 return (sb);
3671 }
3672
3673 for (c = 0; c < children; c++) {
3674 sb = find_start_block(child[c]);
3675 if (sb != MAXOFFSET_T) {
3676 return (sb);
3677 }
3678 }
3679 return (MAXOFFSET_T);
3680}
3681#endif /* sun */
3682
3683/*
3684 * Label an individual disk. The name provided is the short name,
3685 * stripped of any leading /dev path.
3686 */
3687int
3688zpool_label_disk(libzfs_handle_t *hdl, zpool_handle_t *zhp, const char *name)
3689{
3690#ifdef sun
3691 char path[MAXPATHLEN];
3692 struct dk_gpt *vtoc;
3693 int fd;
3694 size_t resv = EFI_MIN_RESV_SIZE;
3695 uint64_t slice_size;
3696 diskaddr_t start_block;
3697 char errbuf[1024];
3698
3699 /* prepare an error message just in case */
3700 (void) snprintf(errbuf, sizeof (errbuf),
3701 dgettext(TEXT_DOMAIN, "cannot label '%s'"), name);
3702
3703 if (zhp) {
3704 nvlist_t *nvroot;
3705
3706 if (pool_is_bootable(zhp)) {
3707 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3708 "EFI labeled devices are not supported on root "
3709 "pools."));
3710 return (zfs_error(hdl, EZFS_POOL_NOTSUP, errbuf));
3711 }
3712
3713 verify(nvlist_lookup_nvlist(zhp->zpool_config,
3714 ZPOOL_CONFIG_VDEV_TREE, &nvroot) == 0);
3715
3716 if (zhp->zpool_start_block == 0)
3717 start_block = find_start_block(nvroot);
3718 else
3719 start_block = zhp->zpool_start_block;
3720 zhp->zpool_start_block = start_block;
3721 } else {
3722 /* new pool */
3723 start_block = NEW_START_BLOCK;
3724 }
3725
3726 (void) snprintf(path, sizeof (path), "%s/%s%s", RDISK_ROOT, name,
3727 BACKUP_SLICE);
3728
3729 if ((fd = open(path, O_RDWR | O_NDELAY)) < 0) {
3730 /*
3731 * This shouldn't happen. We've long since verified that this
3732 * is a valid device.
3733 */
3734 zfs_error_aux(hdl,
3735 dgettext(TEXT_DOMAIN, "unable to open device"));
3736 return (zfs_error(hdl, EZFS_OPENFAILED, errbuf));
3737 }
3738
3739 if (efi_alloc_and_init(fd, EFI_NUMPAR, &vtoc) != 0) {
3740 /*
3741 * The only way this can fail is if we run out of memory, or we
3742 * were unable to read the disk's capacity
3743 */
3744 if (errno == ENOMEM)
3745 (void) no_memory(hdl);
3746
3747 (void) close(fd);
3748 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3749 "unable to read disk capacity"), name);
3750
3751 return (zfs_error(hdl, EZFS_NOCAP, errbuf));
3752 }
3753
3754 slice_size = vtoc->efi_last_u_lba + 1;
3755 slice_size -= EFI_MIN_RESV_SIZE;
3756 if (start_block == MAXOFFSET_T)
3757 start_block = NEW_START_BLOCK;
3758 slice_size -= start_block;
3759
3760 vtoc->efi_parts[0].p_start = start_block;
3761 vtoc->efi_parts[0].p_size = slice_size;
3762
3763 /*
3764 * Why we use V_USR: V_BACKUP confuses users, and is considered
3765 * disposable by some EFI utilities (since EFI doesn't have a backup
3766 * slice). V_UNASSIGNED is supposed to be used only for zero size
3767 * partitions, and efi_write() will fail if we use it. V_ROOT, V_BOOT,
3768 * etc. were all pretty specific. V_USR is as close to reality as we
3769 * can get, in the absence of V_OTHER.
3770 */
3771 vtoc->efi_parts[0].p_tag = V_USR;
3772 (void) strcpy(vtoc->efi_parts[0].p_name, "zfs");
3773
3774 vtoc->efi_parts[8].p_start = slice_size + start_block;
3775 vtoc->efi_parts[8].p_size = resv;
3776 vtoc->efi_parts[8].p_tag = V_RESERVED;
3777
3778 if (efi_write(fd, vtoc) != 0) {
3779 /*
3780 * Some block drivers (like pcata) may not support EFI
3781 * GPT labels. Print out a helpful error message dir-
3782 * ecting the user to manually label the disk and give
3783 * a specific slice.
3784 */
3785 (void) close(fd);
3786 efi_free(vtoc);
3787
3788 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3789 "try using fdisk(1M) and then provide a specific slice"));
3790 return (zfs_error(hdl, EZFS_LABELFAILED, errbuf));
3791 }
3792
3793 (void) close(fd);
3794 efi_free(vtoc);
3795#endif /* sun */
3796 return (0);
3797}
3798
3799static boolean_t
3800supported_dump_vdev_type(libzfs_handle_t *hdl, nvlist_t *config, char *errbuf)
3801{
3802 char *type;
3803 nvlist_t **child;
3804 uint_t children, c;
3805
3806 verify(nvlist_lookup_string(config, ZPOOL_CONFIG_TYPE, &type) == 0);
3807 if (strcmp(type, VDEV_TYPE_RAIDZ) == 0 ||
3808 strcmp(type, VDEV_TYPE_FILE) == 0 ||
3809 strcmp(type, VDEV_TYPE_LOG) == 0 ||
3810 strcmp(type, VDEV_TYPE_HOLE) == 0 ||
3811 strcmp(type, VDEV_TYPE_MISSING) == 0) {
3812 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3813 "vdev type '%s' is not supported"), type);
3814 (void) zfs_error(hdl, EZFS_VDEVNOTSUP, errbuf);
3815 return (B_FALSE);
3816 }
3817 if (nvlist_lookup_nvlist_array(config, ZPOOL_CONFIG_CHILDREN,
3818 &child, &children) == 0) {
3819 for (c = 0; c < children; c++) {
3820 if (!supported_dump_vdev_type(hdl, child[c], errbuf))
3821 return (B_FALSE);
3822 }
3823 }
3824 return (B_TRUE);
3825}
3826
3827/*
3828 * check if this zvol is allowable for use as a dump device; zero if
3829 * it is, > 0 if it isn't, < 0 if it isn't a zvol
3830 */
3831int
3832zvol_check_dump_config(char *arg)
3833{
3834 zpool_handle_t *zhp = NULL;
3835 nvlist_t *config, *nvroot;
3836 char *p, *volname;
3837 nvlist_t **top;
3838 uint_t toplevels;
3839 libzfs_handle_t *hdl;
3840 char errbuf[1024];
3841 char poolname[ZPOOL_MAXNAMELEN];
3842 int pathlen = strlen(ZVOL_FULL_DEV_DIR);
3843 int ret = 1;
3844
3845 if (strncmp(arg, ZVOL_FULL_DEV_DIR, pathlen)) {
3846 return (-1);
3847 }
3848
3849 (void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN,
3850 "dump is not supported on device '%s'"), arg);
3851
3852 if ((hdl = libzfs_init()) == NULL)
3853 return (1);
3854 libzfs_print_on_error(hdl, B_TRUE);
3855
3856 volname = arg + pathlen;
3857
3858 /* check the configuration of the pool */
3859 if ((p = strchr(volname, '/')) == NULL) {
3860 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3861 "malformed dataset name"));
3862 (void) zfs_error(hdl, EZFS_INVALIDNAME, errbuf);
3863 return (1);
3864 } else if (p - volname >= ZFS_MAXNAMELEN) {
3865 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3866 "dataset name is too long"));
3867 (void) zfs_error(hdl, EZFS_NAMETOOLONG, errbuf);
3868 return (1);
3869 } else {
3870 (void) strncpy(poolname, volname, p - volname);
3871 poolname[p - volname] = '\0';
3872 }
3873
3874 if ((zhp = zpool_open(hdl, poolname)) == NULL) {
3875 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3876 "could not open pool '%s'"), poolname);
3877 (void) zfs_error(hdl, EZFS_OPENFAILED, errbuf);
3878 goto out;
3879 }
3880 config = zpool_get_config(zhp, NULL);
3881 if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE,
3882 &nvroot) != 0) {
3883 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3884 "could not obtain vdev configuration for '%s'"), poolname);
3885 (void) zfs_error(hdl, EZFS_INVALCONFIG, errbuf);
3886 goto out;
3887 }
3888
3889 verify(nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_CHILDREN,
3890 &top, &toplevels) == 0);
3891 if (toplevels != 1) {
3892 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3893 "'%s' has multiple top level vdevs"), poolname);
3894 (void) zfs_error(hdl, EZFS_DEVOVERFLOW, errbuf);
3895 goto out;
3896 }
3897
3898 if (!supported_dump_vdev_type(hdl, top[0], errbuf)) {
3899 goto out;
3900 }
3901 ret = 0;
3902
3903out:
3904 if (zhp)
3905 zpool_close(zhp);
3906 libzfs_fini(hdl);
3907 return (ret);
3908}