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.
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.
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]
23 * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
24 * Copyright 2011 Nexenta Systems, Inc. All rights reserved.
25 * Copyright (c) 2013 by Delphix. All rights reserved.
26 * Copyright (c) 2013, Joyent, Inc. All rights reserved.
29 #include <sys/types.h>
41 #include <sys/zfs_ioctl.h>
44 #include "zfs_namecheck.h"
46 #include "libzfs_impl.h"
47 #include "zfs_comutil.h"
48 #include "zfeature_common.h"
50 static int read_efi_label(nvlist_t *config, diskaddr_t *sb);
52 #define DISK_ROOT "/dev/dsk"
53 #define RDISK_ROOT "/dev/rdsk"
54 #define BACKUP_SLICE "s2"
56 typedef struct prop_flags {
57 int create:1; /* Validate property on creation */
58 int import:1; /* Validate property on import */
62 * ====================================================================
63 * zpool property functions
64 * ====================================================================
68 zpool_get_all_props(zpool_handle_t *zhp)
71 libzfs_handle_t *hdl = zhp->zpool_hdl;
73 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
75 if (zcmd_alloc_dst_nvlist(hdl, &zc, 0) != 0)
78 while (ioctl(hdl->libzfs_fd, ZFS_IOC_POOL_GET_PROPS, &zc) != 0) {
79 if (errno == ENOMEM) {
80 if (zcmd_expand_dst_nvlist(hdl, &zc) != 0) {
81 zcmd_free_nvlists(&zc);
85 zcmd_free_nvlists(&zc);
90 if (zcmd_read_dst_nvlist(hdl, &zc, &zhp->zpool_props) != 0) {
91 zcmd_free_nvlists(&zc);
95 zcmd_free_nvlists(&zc);
101 zpool_props_refresh(zpool_handle_t *zhp)
105 old_props = zhp->zpool_props;
107 if (zpool_get_all_props(zhp) != 0)
110 nvlist_free(old_props);
115 zpool_get_prop_string(zpool_handle_t *zhp, zpool_prop_t prop,
121 zprop_source_t source;
123 nvl = zhp->zpool_props;
124 if (nvlist_lookup_nvlist(nvl, zpool_prop_to_name(prop), &nv) == 0) {
125 verify(nvlist_lookup_uint64(nv, ZPROP_SOURCE, &ival) == 0);
127 verify(nvlist_lookup_string(nv, ZPROP_VALUE, &value) == 0);
129 source = ZPROP_SRC_DEFAULT;
130 if ((value = (char *)zpool_prop_default_string(prop)) == NULL)
141 zpool_get_prop_int(zpool_handle_t *zhp, zpool_prop_t prop, zprop_source_t *src)
145 zprop_source_t source;
147 if (zhp->zpool_props == NULL && zpool_get_all_props(zhp)) {
149 * zpool_get_all_props() has most likely failed because
150 * the pool is faulted, but if all we need is the top level
151 * vdev's guid then get it from the zhp config nvlist.
153 if ((prop == ZPOOL_PROP_GUID) &&
154 (nvlist_lookup_nvlist(zhp->zpool_config,
155 ZPOOL_CONFIG_VDEV_TREE, &nv) == 0) &&
156 (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID, &value)
160 return (zpool_prop_default_numeric(prop));
163 nvl = zhp->zpool_props;
164 if (nvlist_lookup_nvlist(nvl, zpool_prop_to_name(prop), &nv) == 0) {
165 verify(nvlist_lookup_uint64(nv, ZPROP_SOURCE, &value) == 0);
167 verify(nvlist_lookup_uint64(nv, ZPROP_VALUE, &value) == 0);
169 source = ZPROP_SRC_DEFAULT;
170 value = zpool_prop_default_numeric(prop);
180 * Map VDEV STATE to printed strings.
183 zpool_state_to_name(vdev_state_t state, vdev_aux_t aux)
186 case VDEV_STATE_CLOSED:
187 case VDEV_STATE_OFFLINE:
188 return (gettext("OFFLINE"));
189 case VDEV_STATE_REMOVED:
190 return (gettext("REMOVED"));
191 case VDEV_STATE_CANT_OPEN:
192 if (aux == VDEV_AUX_CORRUPT_DATA || aux == VDEV_AUX_BAD_LOG)
193 return (gettext("FAULTED"));
194 else if (aux == VDEV_AUX_SPLIT_POOL)
195 return (gettext("SPLIT"));
197 return (gettext("UNAVAIL"));
198 case VDEV_STATE_FAULTED:
199 return (gettext("FAULTED"));
200 case VDEV_STATE_DEGRADED:
201 return (gettext("DEGRADED"));
202 case VDEV_STATE_HEALTHY:
203 return (gettext("ONLINE"));
206 return (gettext("UNKNOWN"));
210 * Map POOL STATE to printed strings.
213 zpool_pool_state_to_name(pool_state_t state)
216 case POOL_STATE_ACTIVE:
217 return (gettext("ACTIVE"));
218 case POOL_STATE_EXPORTED:
219 return (gettext("EXPORTED"));
220 case POOL_STATE_DESTROYED:
221 return (gettext("DESTROYED"));
222 case POOL_STATE_SPARE:
223 return (gettext("SPARE"));
224 case POOL_STATE_L2CACHE:
225 return (gettext("L2CACHE"));
226 case POOL_STATE_UNINITIALIZED:
227 return (gettext("UNINITIALIZED"));
228 case POOL_STATE_UNAVAIL:
229 return (gettext("UNAVAIL"));
230 case POOL_STATE_POTENTIALLY_ACTIVE:
231 return (gettext("POTENTIALLY_ACTIVE"));
234 return (gettext("UNKNOWN"));
238 * Get a zpool property value for 'prop' and return the value in
239 * a pre-allocated buffer.
242 zpool_get_prop(zpool_handle_t *zhp, zpool_prop_t prop, char *buf, size_t len,
243 zprop_source_t *srctype, boolean_t literal)
247 zprop_source_t src = ZPROP_SRC_NONE;
252 if (zpool_get_state(zhp) == POOL_STATE_UNAVAIL) {
254 case ZPOOL_PROP_NAME:
255 (void) strlcpy(buf, zpool_get_name(zhp), len);
258 case ZPOOL_PROP_HEALTH:
259 (void) strlcpy(buf, "FAULTED", len);
262 case ZPOOL_PROP_GUID:
263 intval = zpool_get_prop_int(zhp, prop, &src);
264 (void) snprintf(buf, len, "%llu", intval);
267 case ZPOOL_PROP_ALTROOT:
268 case ZPOOL_PROP_CACHEFILE:
269 case ZPOOL_PROP_COMMENT:
270 if (zhp->zpool_props != NULL ||
271 zpool_get_all_props(zhp) == 0) {
273 zpool_get_prop_string(zhp, prop, &src),
279 (void) strlcpy(buf, "-", len);
288 if (zhp->zpool_props == NULL && zpool_get_all_props(zhp) &&
289 prop != ZPOOL_PROP_NAME)
292 switch (zpool_prop_get_type(prop)) {
293 case PROP_TYPE_STRING:
294 (void) strlcpy(buf, zpool_get_prop_string(zhp, prop, &src),
298 case PROP_TYPE_NUMBER:
299 intval = zpool_get_prop_int(zhp, prop, &src);
302 case ZPOOL_PROP_SIZE:
303 case ZPOOL_PROP_ALLOCATED:
304 case ZPOOL_PROP_FREE:
305 case ZPOOL_PROP_FREEING:
306 case ZPOOL_PROP_EXPANDSZ:
308 (void) snprintf(buf, len, "%llu",
309 (u_longlong_t)intval);
311 (void) zfs_nicenum(intval, buf, len);
315 case ZPOOL_PROP_CAPACITY:
317 (void) snprintf(buf, len, "%llu",
318 (u_longlong_t)intval);
320 (void) snprintf(buf, len, "%llu%%",
321 (u_longlong_t)intval);
325 case ZPOOL_PROP_DEDUPRATIO:
326 (void) snprintf(buf, len, "%llu.%02llux",
327 (u_longlong_t)(intval / 100),
328 (u_longlong_t)(intval % 100));
331 case ZPOOL_PROP_HEALTH:
332 verify(nvlist_lookup_nvlist(zpool_get_config(zhp, NULL),
333 ZPOOL_CONFIG_VDEV_TREE, &nvroot) == 0);
334 verify(nvlist_lookup_uint64_array(nvroot,
335 ZPOOL_CONFIG_VDEV_STATS, (uint64_t **)&vs, &vsc)
338 (void) strlcpy(buf, zpool_state_to_name(intval,
341 case ZPOOL_PROP_VERSION:
342 if (intval >= SPA_VERSION_FEATURES) {
343 (void) snprintf(buf, len, "-");
348 (void) snprintf(buf, len, "%llu", intval);
352 case PROP_TYPE_INDEX:
353 intval = zpool_get_prop_int(zhp, prop, &src);
354 if (zpool_prop_index_to_string(prop, intval, &strval)
357 (void) strlcpy(buf, strval, len);
371 * Check if the bootfs name has the same pool name as it is set to.
372 * Assuming bootfs is a valid dataset name.
375 bootfs_name_valid(const char *pool, char *bootfs)
377 int len = strlen(pool);
379 if (!zfs_name_valid(bootfs, ZFS_TYPE_FILESYSTEM|ZFS_TYPE_SNAPSHOT))
382 if (strncmp(pool, bootfs, len) == 0 &&
383 (bootfs[len] == '/' || bootfs[len] == '\0'))
390 * Inspect the configuration to determine if any of the devices contain
394 pool_uses_efi(nvlist_t *config)
400 if (nvlist_lookup_nvlist_array(config, ZPOOL_CONFIG_CHILDREN,
401 &child, &children) != 0)
402 return (read_efi_label(config, NULL) >= 0);
404 for (c = 0; c < children; c++) {
405 if (pool_uses_efi(child[c]))
413 zpool_is_bootable(zpool_handle_t *zhp)
415 char bootfs[ZPOOL_MAXNAMELEN];
417 return (zpool_get_prop(zhp, ZPOOL_PROP_BOOTFS, bootfs,
418 sizeof (bootfs), NULL, B_FALSE) == 0 && strncmp(bootfs, "-",
419 sizeof (bootfs)) != 0);
424 * Given an nvlist of zpool properties to be set, validate that they are
425 * correct, and parse any numeric properties (index, boolean, etc) if they are
426 * specified as strings.
429 zpool_valid_proplist(libzfs_handle_t *hdl, const char *poolname,
430 nvlist_t *props, uint64_t version, prop_flags_t flags, char *errbuf)
438 struct stat64 statbuf;
442 if (nvlist_alloc(&retprops, NV_UNIQUE_NAME, 0) != 0) {
443 (void) no_memory(hdl);
448 while ((elem = nvlist_next_nvpair(props, elem)) != NULL) {
449 const char *propname = nvpair_name(elem);
451 prop = zpool_name_to_prop(propname);
452 if (prop == ZPROP_INVAL && zpool_prop_feature(propname)) {
454 char *fname = strchr(propname, '@') + 1;
456 err = zfeature_lookup_name(fname, NULL);
458 ASSERT3U(err, ==, ENOENT);
459 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
460 "invalid feature '%s'"), fname);
461 (void) zfs_error(hdl, EZFS_BADPROP, errbuf);
465 if (nvpair_type(elem) != DATA_TYPE_STRING) {
466 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
467 "'%s' must be a string"), propname);
468 (void) zfs_error(hdl, EZFS_BADPROP, errbuf);
472 (void) nvpair_value_string(elem, &strval);
473 if (strcmp(strval, ZFS_FEATURE_ENABLED) != 0) {
474 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
475 "property '%s' can only be set to "
476 "'enabled'"), propname);
477 (void) zfs_error(hdl, EZFS_BADPROP, errbuf);
481 if (nvlist_add_uint64(retprops, propname, 0) != 0) {
482 (void) no_memory(hdl);
489 * Make sure this property is valid and applies to this type.
491 if (prop == ZPROP_INVAL) {
492 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
493 "invalid property '%s'"), propname);
494 (void) zfs_error(hdl, EZFS_BADPROP, errbuf);
498 if (zpool_prop_readonly(prop)) {
499 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "'%s' "
500 "is readonly"), propname);
501 (void) zfs_error(hdl, EZFS_PROPREADONLY, errbuf);
505 if (zprop_parse_value(hdl, elem, prop, ZFS_TYPE_POOL, retprops,
506 &strval, &intval, errbuf) != 0)
510 * Perform additional checking for specific properties.
513 case ZPOOL_PROP_VERSION:
514 if (intval < version ||
515 !SPA_VERSION_IS_SUPPORTED(intval)) {
516 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
517 "property '%s' number %d is invalid."),
519 (void) zfs_error(hdl, EZFS_BADVERSION, errbuf);
524 case ZPOOL_PROP_BOOTFS:
525 if (flags.create || flags.import) {
526 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
527 "property '%s' cannot be set at creation "
528 "or import time"), propname);
529 (void) zfs_error(hdl, EZFS_BADPROP, errbuf);
533 if (version < SPA_VERSION_BOOTFS) {
534 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
535 "pool must be upgraded to support "
536 "'%s' property"), propname);
537 (void) zfs_error(hdl, EZFS_BADVERSION, errbuf);
542 * bootfs property value has to be a dataset name and
543 * the dataset has to be in the same pool as it sets to.
545 if (strval[0] != '\0' && !bootfs_name_valid(poolname,
547 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "'%s' "
548 "is an invalid name"), strval);
549 (void) zfs_error(hdl, EZFS_INVALIDNAME, errbuf);
553 if ((zhp = zpool_open_canfail(hdl, poolname)) == NULL) {
554 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
555 "could not open pool '%s'"), poolname);
556 (void) zfs_error(hdl, EZFS_OPENFAILED, errbuf);
559 verify(nvlist_lookup_nvlist(zpool_get_config(zhp, NULL),
560 ZPOOL_CONFIG_VDEV_TREE, &nvroot) == 0);
564 * bootfs property cannot be set on a disk which has
567 if (pool_uses_efi(nvroot)) {
568 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
569 "property '%s' not supported on "
570 "EFI labeled devices"), propname);
571 (void) zfs_error(hdl, EZFS_POOL_NOTSUP, errbuf);
579 case ZPOOL_PROP_ALTROOT:
580 if (!flags.create && !flags.import) {
581 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
582 "property '%s' can only be set during pool "
583 "creation or import"), propname);
584 (void) zfs_error(hdl, EZFS_BADPROP, errbuf);
588 if (strval[0] != '/') {
589 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
590 "bad alternate root '%s'"), strval);
591 (void) zfs_error(hdl, EZFS_BADPATH, errbuf);
596 case ZPOOL_PROP_CACHEFILE:
597 if (strval[0] == '\0')
600 if (strcmp(strval, "none") == 0)
603 if (strval[0] != '/') {
604 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
605 "property '%s' must be empty, an "
606 "absolute path, or 'none'"), propname);
607 (void) zfs_error(hdl, EZFS_BADPATH, errbuf);
611 slash = strrchr(strval, '/');
613 if (slash[1] == '\0' || strcmp(slash, "/.") == 0 ||
614 strcmp(slash, "/..") == 0) {
615 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
616 "'%s' is not a valid file"), strval);
617 (void) zfs_error(hdl, EZFS_BADPATH, errbuf);
623 if (strval[0] != '\0' &&
624 (stat64(strval, &statbuf) != 0 ||
625 !S_ISDIR(statbuf.st_mode))) {
626 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
627 "'%s' is not a valid directory"),
629 (void) zfs_error(hdl, EZFS_BADPATH, errbuf);
636 case ZPOOL_PROP_COMMENT:
637 for (check = strval; *check != '\0'; check++) {
638 if (!isprint(*check)) {
640 dgettext(TEXT_DOMAIN,
641 "comment may only have printable "
643 (void) zfs_error(hdl, EZFS_BADPROP,
648 if (strlen(strval) > ZPROP_MAX_COMMENT) {
649 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
650 "comment must not exceed %d characters"),
652 (void) zfs_error(hdl, EZFS_BADPROP, errbuf);
656 case ZPOOL_PROP_READONLY:
658 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
659 "property '%s' can only be set at "
660 "import time"), propname);
661 (void) zfs_error(hdl, EZFS_BADPROP, errbuf);
670 nvlist_free(retprops);
675 * Set zpool property : propname=propval.
678 zpool_set_prop(zpool_handle_t *zhp, const char *propname, const char *propval)
680 zfs_cmd_t zc = { 0 };
683 nvlist_t *nvl = NULL;
686 prop_flags_t flags = { 0 };
688 (void) snprintf(errbuf, sizeof (errbuf),
689 dgettext(TEXT_DOMAIN, "cannot set property for '%s'"),
692 if (nvlist_alloc(&nvl, NV_UNIQUE_NAME, 0) != 0)
693 return (no_memory(zhp->zpool_hdl));
695 if (nvlist_add_string(nvl, propname, propval) != 0) {
697 return (no_memory(zhp->zpool_hdl));
700 version = zpool_get_prop_int(zhp, ZPOOL_PROP_VERSION, NULL);
701 if ((realprops = zpool_valid_proplist(zhp->zpool_hdl,
702 zhp->zpool_name, nvl, version, flags, errbuf)) == NULL) {
711 * Execute the corresponding ioctl() to set this property.
713 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
715 if (zcmd_write_src_nvlist(zhp->zpool_hdl, &zc, nvl) != 0) {
720 ret = zfs_ioctl(zhp->zpool_hdl, ZFS_IOC_POOL_SET_PROPS, &zc);
722 zcmd_free_nvlists(&zc);
726 (void) zpool_standard_error(zhp->zpool_hdl, errno, errbuf);
728 (void) zpool_props_refresh(zhp);
734 zpool_expand_proplist(zpool_handle_t *zhp, zprop_list_t **plp)
736 libzfs_handle_t *hdl = zhp->zpool_hdl;
738 char buf[ZFS_MAXPROPLEN];
739 nvlist_t *features = NULL;
741 boolean_t firstexpand = (NULL == *plp);
743 if (zprop_expand_list(hdl, plp, ZFS_TYPE_POOL) != 0)
747 while (*last != NULL)
748 last = &(*last)->pl_next;
751 features = zpool_get_features(zhp);
753 if ((*plp)->pl_all && firstexpand) {
754 for (int i = 0; i < SPA_FEATURES; i++) {
755 zprop_list_t *entry = zfs_alloc(hdl,
756 sizeof (zprop_list_t));
757 entry->pl_prop = ZPROP_INVAL;
758 entry->pl_user_prop = zfs_asprintf(hdl, "feature@%s",
759 spa_feature_table[i].fi_uname);
760 entry->pl_width = strlen(entry->pl_user_prop);
761 entry->pl_all = B_TRUE;
764 last = &entry->pl_next;
768 /* add any unsupported features */
769 for (nvpair_t *nvp = nvlist_next_nvpair(features, NULL);
770 nvp != NULL; nvp = nvlist_next_nvpair(features, nvp)) {
775 if (zfeature_is_supported(nvpair_name(nvp)))
778 propname = zfs_asprintf(hdl, "unsupported@%s",
782 * Before adding the property to the list make sure that no
783 * other pool already added the same property.
787 while (entry != NULL) {
788 if (entry->pl_user_prop != NULL &&
789 strcmp(propname, entry->pl_user_prop) == 0) {
793 entry = entry->pl_next;
800 entry = zfs_alloc(hdl, sizeof (zprop_list_t));
801 entry->pl_prop = ZPROP_INVAL;
802 entry->pl_user_prop = propname;
803 entry->pl_width = strlen(entry->pl_user_prop);
804 entry->pl_all = B_TRUE;
807 last = &entry->pl_next;
810 for (entry = *plp; entry != NULL; entry = entry->pl_next) {
815 if (entry->pl_prop != ZPROP_INVAL &&
816 zpool_get_prop(zhp, entry->pl_prop, buf, sizeof (buf),
817 NULL, B_FALSE) == 0) {
818 if (strlen(buf) > entry->pl_width)
819 entry->pl_width = strlen(buf);
827 * Get the state for the given feature on the given ZFS pool.
830 zpool_prop_get_feature(zpool_handle_t *zhp, const char *propname, char *buf,
834 boolean_t found = B_FALSE;
835 nvlist_t *features = zpool_get_features(zhp);
837 const char *feature = strchr(propname, '@') + 1;
839 supported = zpool_prop_feature(propname);
840 ASSERT(supported || zpool_prop_unsupported(propname));
843 * Convert from feature name to feature guid. This conversion is
844 * unecessary for unsupported@... properties because they already
851 ret = zfeature_lookup_name(feature, &fid);
853 (void) strlcpy(buf, "-", len);
856 feature = spa_feature_table[fid].fi_guid;
859 if (nvlist_lookup_uint64(features, feature, &refcount) == 0)
864 (void) strlcpy(buf, ZFS_FEATURE_DISABLED, len);
867 (void) strlcpy(buf, ZFS_FEATURE_ENABLED, len);
869 (void) strlcpy(buf, ZFS_FEATURE_ACTIVE, len);
874 (void) strcpy(buf, ZFS_UNSUPPORTED_INACTIVE);
876 (void) strcpy(buf, ZFS_UNSUPPORTED_READONLY);
879 (void) strlcpy(buf, "-", len);
888 * Don't start the slice at the default block of 34; many storage
889 * devices will use a stripe width of 128k, so start there instead.
891 #define NEW_START_BLOCK 256
894 * Validate the given pool name, optionally putting an extended error message in
898 zpool_name_valid(libzfs_handle_t *hdl, boolean_t isopen, const char *pool)
904 ret = pool_namecheck(pool, &why, &what);
907 * The rules for reserved pool names were extended at a later point.
908 * But we need to support users with existing pools that may now be
909 * invalid. So we only check for this expanded set of names during a
910 * create (or import), and only in userland.
912 if (ret == 0 && !isopen &&
913 (strncmp(pool, "mirror", 6) == 0 ||
914 strncmp(pool, "raidz", 5) == 0 ||
915 strncmp(pool, "spare", 5) == 0 ||
916 strcmp(pool, "log") == 0)) {
919 dgettext(TEXT_DOMAIN, "name is reserved"));
927 case NAME_ERR_TOOLONG:
929 dgettext(TEXT_DOMAIN, "name is too long"));
932 case NAME_ERR_INVALCHAR:
934 dgettext(TEXT_DOMAIN, "invalid character "
935 "'%c' in pool name"), what);
938 case NAME_ERR_NOLETTER:
939 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
940 "name must begin with a letter"));
943 case NAME_ERR_RESERVED:
944 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
945 "name is reserved"));
948 case NAME_ERR_DISKLIKE:
949 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
950 "pool name is reserved"));
953 case NAME_ERR_LEADING_SLASH:
954 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
955 "leading slash in name"));
958 case NAME_ERR_EMPTY_COMPONENT:
959 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
960 "empty component in name"));
963 case NAME_ERR_TRAILING_SLASH:
964 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
965 "trailing slash in name"));
968 case NAME_ERR_MULTIPLE_AT:
969 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
970 "multiple '@' delimiters in name"));
982 * Open a handle to the given pool, even if the pool is currently in the FAULTED
986 zpool_open_canfail(libzfs_handle_t *hdl, const char *pool)
992 * Make sure the pool name is valid.
994 if (!zpool_name_valid(hdl, B_TRUE, pool)) {
995 (void) zfs_error_fmt(hdl, EZFS_INVALIDNAME,
996 dgettext(TEXT_DOMAIN, "cannot open '%s'"),
1001 if ((zhp = zfs_alloc(hdl, sizeof (zpool_handle_t))) == NULL)
1004 zhp->zpool_hdl = hdl;
1005 (void) strlcpy(zhp->zpool_name, pool, sizeof (zhp->zpool_name));
1007 if (zpool_refresh_stats(zhp, &missing) != 0) {
1013 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "no such pool"));
1014 (void) zfs_error_fmt(hdl, EZFS_NOENT,
1015 dgettext(TEXT_DOMAIN, "cannot open '%s'"), pool);
1024 * Like the above, but silent on error. Used when iterating over pools (because
1025 * the configuration cache may be out of date).
1028 zpool_open_silent(libzfs_handle_t *hdl, const char *pool, zpool_handle_t **ret)
1030 zpool_handle_t *zhp;
1033 if ((zhp = zfs_alloc(hdl, sizeof (zpool_handle_t))) == NULL)
1036 zhp->zpool_hdl = hdl;
1037 (void) strlcpy(zhp->zpool_name, pool, sizeof (zhp->zpool_name));
1039 if (zpool_refresh_stats(zhp, &missing) != 0) {
1055 * Similar to zpool_open_canfail(), but refuses to open pools in the faulted
1059 zpool_open(libzfs_handle_t *hdl, const char *pool)
1061 zpool_handle_t *zhp;
1063 if ((zhp = zpool_open_canfail(hdl, pool)) == NULL)
1066 if (zhp->zpool_state == POOL_STATE_UNAVAIL) {
1067 (void) zfs_error_fmt(hdl, EZFS_POOLUNAVAIL,
1068 dgettext(TEXT_DOMAIN, "cannot open '%s'"), zhp->zpool_name);
1077 * Close the handle. Simply frees the memory associated with the handle.
1080 zpool_close(zpool_handle_t *zhp)
1082 if (zhp->zpool_config)
1083 nvlist_free(zhp->zpool_config);
1084 if (zhp->zpool_old_config)
1085 nvlist_free(zhp->zpool_old_config);
1086 if (zhp->zpool_props)
1087 nvlist_free(zhp->zpool_props);
1092 * Return the name of the pool.
1095 zpool_get_name(zpool_handle_t *zhp)
1097 return (zhp->zpool_name);
1102 * Return the state of the pool (ACTIVE or UNAVAILABLE)
1105 zpool_get_state(zpool_handle_t *zhp)
1107 return (zhp->zpool_state);
1111 * Create the named pool, using the provided vdev list. It is assumed
1112 * that the consumer has already validated the contents of the nvlist, so we
1113 * don't have to worry about error semantics.
1116 zpool_create(libzfs_handle_t *hdl, const char *pool, nvlist_t *nvroot,
1117 nvlist_t *props, nvlist_t *fsprops)
1119 zfs_cmd_t zc = { 0 };
1120 nvlist_t *zc_fsprops = NULL;
1121 nvlist_t *zc_props = NULL;
1125 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN,
1126 "cannot create '%s'"), pool);
1128 if (!zpool_name_valid(hdl, B_FALSE, pool))
1129 return (zfs_error(hdl, EZFS_INVALIDNAME, msg));
1131 if (zcmd_write_conf_nvlist(hdl, &zc, nvroot) != 0)
1135 prop_flags_t flags = { .create = B_TRUE, .import = B_FALSE };
1137 if ((zc_props = zpool_valid_proplist(hdl, pool, props,
1138 SPA_VERSION_1, flags, msg)) == NULL) {
1147 zoned = ((nvlist_lookup_string(fsprops,
1148 zfs_prop_to_name(ZFS_PROP_ZONED), &zonestr) == 0) &&
1149 strcmp(zonestr, "on") == 0);
1151 if ((zc_fsprops = zfs_valid_proplist(hdl,
1152 ZFS_TYPE_FILESYSTEM, fsprops, zoned, NULL, msg)) == NULL) {
1156 (nvlist_alloc(&zc_props, NV_UNIQUE_NAME, 0) != 0)) {
1159 if (nvlist_add_nvlist(zc_props,
1160 ZPOOL_ROOTFS_PROPS, zc_fsprops) != 0) {
1165 if (zc_props && zcmd_write_src_nvlist(hdl, &zc, zc_props) != 0)
1168 (void) strlcpy(zc.zc_name, pool, sizeof (zc.zc_name));
1170 if ((ret = zfs_ioctl(hdl, ZFS_IOC_POOL_CREATE, &zc)) != 0) {
1172 zcmd_free_nvlists(&zc);
1173 nvlist_free(zc_props);
1174 nvlist_free(zc_fsprops);
1179 * This can happen if the user has specified the same
1180 * device multiple times. We can't reliably detect this
1181 * until we try to add it and see we already have a
1184 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1185 "one or more vdevs refer to the same device"));
1186 return (zfs_error(hdl, EZFS_BADDEV, msg));
1190 * This occurs when one of the devices is below
1191 * SPA_MINDEVSIZE. Unfortunately, we can't detect which
1192 * device was the problem device since there's no
1193 * reliable way to determine device size from userland.
1198 zfs_nicenum(SPA_MINDEVSIZE, buf, sizeof (buf));
1200 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1201 "one or more devices is less than the "
1202 "minimum size (%s)"), buf);
1204 return (zfs_error(hdl, EZFS_BADDEV, msg));
1207 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1208 "one or more devices is out of space"));
1209 return (zfs_error(hdl, EZFS_BADDEV, msg));
1212 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1213 "cache device must be a disk or disk slice"));
1214 return (zfs_error(hdl, EZFS_BADDEV, msg));
1217 return (zpool_standard_error(hdl, errno, msg));
1222 zcmd_free_nvlists(&zc);
1223 nvlist_free(zc_props);
1224 nvlist_free(zc_fsprops);
1229 * Destroy the given pool. It is up to the caller to ensure that there are no
1230 * datasets left in the pool.
1233 zpool_destroy(zpool_handle_t *zhp, const char *log_str)
1235 zfs_cmd_t zc = { 0 };
1236 zfs_handle_t *zfp = NULL;
1237 libzfs_handle_t *hdl = zhp->zpool_hdl;
1240 if (zhp->zpool_state == POOL_STATE_ACTIVE &&
1241 (zfp = zfs_open(hdl, zhp->zpool_name, ZFS_TYPE_FILESYSTEM)) == NULL)
1244 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
1245 zc.zc_history = (uint64_t)(uintptr_t)log_str;
1247 if (zfs_ioctl(hdl, ZFS_IOC_POOL_DESTROY, &zc) != 0) {
1248 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN,
1249 "cannot destroy '%s'"), zhp->zpool_name);
1251 if (errno == EROFS) {
1252 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1253 "one or more devices is read only"));
1254 (void) zfs_error(hdl, EZFS_BADDEV, msg);
1256 (void) zpool_standard_error(hdl, errno, msg);
1265 remove_mountpoint(zfp);
1273 * Add the given vdevs to the pool. The caller must have already performed the
1274 * necessary verification to ensure that the vdev specification is well-formed.
1277 zpool_add(zpool_handle_t *zhp, nvlist_t *nvroot)
1279 zfs_cmd_t zc = { 0 };
1281 libzfs_handle_t *hdl = zhp->zpool_hdl;
1283 nvlist_t **spares, **l2cache;
1284 uint_t nspares, nl2cache;
1286 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN,
1287 "cannot add to '%s'"), zhp->zpool_name);
1289 if (zpool_get_prop_int(zhp, ZPOOL_PROP_VERSION, NULL) <
1290 SPA_VERSION_SPARES &&
1291 nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_SPARES,
1292 &spares, &nspares) == 0) {
1293 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "pool must be "
1294 "upgraded to add hot spares"));
1295 return (zfs_error(hdl, EZFS_BADVERSION, msg));
1298 if (zpool_is_bootable(zhp) && nvlist_lookup_nvlist_array(nvroot,
1299 ZPOOL_CONFIG_SPARES, &spares, &nspares) == 0) {
1302 for (s = 0; s < nspares; s++) {
1305 if (nvlist_lookup_string(spares[s], ZPOOL_CONFIG_PATH,
1306 &path) == 0 && pool_uses_efi(spares[s])) {
1307 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1308 "device '%s' contains an EFI label and "
1309 "cannot be used on root pools."),
1310 zpool_vdev_name(hdl, NULL, spares[s],
1312 return (zfs_error(hdl, EZFS_POOL_NOTSUP, msg));
1317 if (zpool_get_prop_int(zhp, ZPOOL_PROP_VERSION, NULL) <
1318 SPA_VERSION_L2CACHE &&
1319 nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_L2CACHE,
1320 &l2cache, &nl2cache) == 0) {
1321 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "pool must be "
1322 "upgraded to add cache devices"));
1323 return (zfs_error(hdl, EZFS_BADVERSION, msg));
1326 if (zcmd_write_conf_nvlist(hdl, &zc, nvroot) != 0)
1328 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
1330 if (zfs_ioctl(hdl, ZFS_IOC_VDEV_ADD, &zc) != 0) {
1334 * This can happen if the user has specified the same
1335 * device multiple times. We can't reliably detect this
1336 * until we try to add it and see we already have a
1339 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1340 "one or more vdevs refer to the same device"));
1341 (void) zfs_error(hdl, EZFS_BADDEV, msg);
1346 * This occurrs when one of the devices is below
1347 * SPA_MINDEVSIZE. Unfortunately, we can't detect which
1348 * device was the problem device since there's no
1349 * reliable way to determine device size from userland.
1354 zfs_nicenum(SPA_MINDEVSIZE, buf, sizeof (buf));
1356 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1357 "device is less than the minimum "
1360 (void) zfs_error(hdl, EZFS_BADDEV, msg);
1364 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1365 "pool must be upgraded to add these vdevs"));
1366 (void) zfs_error(hdl, EZFS_BADVERSION, msg);
1370 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1371 "root pool can not have multiple vdevs"
1372 " or separate logs"));
1373 (void) zfs_error(hdl, EZFS_POOL_NOTSUP, msg);
1377 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1378 "cache device must be a disk or disk slice"));
1379 (void) zfs_error(hdl, EZFS_BADDEV, msg);
1383 (void) zpool_standard_error(hdl, errno, msg);
1391 zcmd_free_nvlists(&zc);
1397 * Exports the pool from the system. The caller must ensure that there are no
1398 * mounted datasets in the pool.
1401 zpool_export_common(zpool_handle_t *zhp, boolean_t force, boolean_t hardforce,
1402 const char *log_str)
1404 zfs_cmd_t zc = { 0 };
1407 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN,
1408 "cannot export '%s'"), zhp->zpool_name);
1410 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
1411 zc.zc_cookie = force;
1412 zc.zc_guid = hardforce;
1413 zc.zc_history = (uint64_t)(uintptr_t)log_str;
1415 if (zfs_ioctl(zhp->zpool_hdl, ZFS_IOC_POOL_EXPORT, &zc) != 0) {
1418 zfs_error_aux(zhp->zpool_hdl, dgettext(TEXT_DOMAIN,
1419 "use '-f' to override the following errors:\n"
1420 "'%s' has an active shared spare which could be"
1421 " used by other pools once '%s' is exported."),
1422 zhp->zpool_name, zhp->zpool_name);
1423 return (zfs_error(zhp->zpool_hdl, EZFS_ACTIVE_SPARE,
1426 return (zpool_standard_error_fmt(zhp->zpool_hdl, errno,
1435 zpool_export(zpool_handle_t *zhp, boolean_t force, const char *log_str)
1437 return (zpool_export_common(zhp, force, B_FALSE, log_str));
1441 zpool_export_force(zpool_handle_t *zhp, const char *log_str)
1443 return (zpool_export_common(zhp, B_TRUE, B_TRUE, log_str));
1447 zpool_rewind_exclaim(libzfs_handle_t *hdl, const char *name, boolean_t dryrun,
1450 nvlist_t *nv = NULL;
1456 if (!hdl->libzfs_printerr || config == NULL)
1459 if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_LOAD_INFO, &nv) != 0 ||
1460 nvlist_lookup_nvlist(nv, ZPOOL_CONFIG_REWIND_INFO, &nv) != 0) {
1464 if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_LOAD_TIME, &rewindto) != 0)
1466 (void) nvlist_lookup_int64(nv, ZPOOL_CONFIG_REWIND_TIME, &loss);
1468 if (localtime_r((time_t *)&rewindto, &t) != NULL &&
1469 strftime(timestr, 128, 0, &t) != 0) {
1471 (void) printf(dgettext(TEXT_DOMAIN,
1472 "Would be able to return %s "
1473 "to its state as of %s.\n"),
1476 (void) printf(dgettext(TEXT_DOMAIN,
1477 "Pool %s returned to its state as of %s.\n"),
1481 (void) printf(dgettext(TEXT_DOMAIN,
1482 "%s approximately %lld "),
1483 dryrun ? "Would discard" : "Discarded",
1485 (void) printf(dgettext(TEXT_DOMAIN,
1486 "minutes of transactions.\n"));
1487 } else if (loss > 0) {
1488 (void) printf(dgettext(TEXT_DOMAIN,
1489 "%s approximately %lld "),
1490 dryrun ? "Would discard" : "Discarded", loss);
1491 (void) printf(dgettext(TEXT_DOMAIN,
1492 "seconds of transactions.\n"));
1498 zpool_explain_recover(libzfs_handle_t *hdl, const char *name, int reason,
1501 nvlist_t *nv = NULL;
1503 uint64_t edata = UINT64_MAX;
1508 if (!hdl->libzfs_printerr)
1512 (void) printf(dgettext(TEXT_DOMAIN, "action: "));
1514 (void) printf(dgettext(TEXT_DOMAIN, "\t"));
1516 /* All attempted rewinds failed if ZPOOL_CONFIG_LOAD_TIME missing */
1517 if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_LOAD_INFO, &nv) != 0 ||
1518 nvlist_lookup_nvlist(nv, ZPOOL_CONFIG_REWIND_INFO, &nv) != 0 ||
1519 nvlist_lookup_uint64(nv, ZPOOL_CONFIG_LOAD_TIME, &rewindto) != 0)
1522 (void) nvlist_lookup_int64(nv, ZPOOL_CONFIG_REWIND_TIME, &loss);
1523 (void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_LOAD_DATA_ERRORS,
1526 (void) printf(dgettext(TEXT_DOMAIN,
1527 "Recovery is possible, but will result in some data loss.\n"));
1529 if (localtime_r((time_t *)&rewindto, &t) != NULL &&
1530 strftime(timestr, 128, 0, &t) != 0) {
1531 (void) printf(dgettext(TEXT_DOMAIN,
1532 "\tReturning the pool to its state as of %s\n"
1533 "\tshould correct the problem. "),
1536 (void) printf(dgettext(TEXT_DOMAIN,
1537 "\tReverting the pool to an earlier state "
1538 "should correct the problem.\n\t"));
1542 (void) printf(dgettext(TEXT_DOMAIN,
1543 "Approximately %lld minutes of data\n"
1544 "\tmust be discarded, irreversibly. "), (loss + 30) / 60);
1545 } else if (loss > 0) {
1546 (void) printf(dgettext(TEXT_DOMAIN,
1547 "Approximately %lld seconds of data\n"
1548 "\tmust be discarded, irreversibly. "), loss);
1550 if (edata != 0 && edata != UINT64_MAX) {
1552 (void) printf(dgettext(TEXT_DOMAIN,
1553 "After rewind, at least\n"
1554 "\tone persistent user-data error will remain. "));
1556 (void) printf(dgettext(TEXT_DOMAIN,
1557 "After rewind, several\n"
1558 "\tpersistent user-data errors will remain. "));
1561 (void) printf(dgettext(TEXT_DOMAIN,
1562 "Recovery can be attempted\n\tby executing 'zpool %s -F %s'. "),
1563 reason >= 0 ? "clear" : "import", name);
1565 (void) printf(dgettext(TEXT_DOMAIN,
1566 "A scrub of the pool\n"
1567 "\tis strongly recommended after recovery.\n"));
1571 (void) printf(dgettext(TEXT_DOMAIN,
1572 "Destroy and re-create the pool from\n\ta backup source.\n"));
1576 * zpool_import() is a contracted interface. Should be kept the same
1579 * Applications should use zpool_import_props() to import a pool with
1580 * new properties value to be set.
1583 zpool_import(libzfs_handle_t *hdl, nvlist_t *config, const char *newname,
1586 nvlist_t *props = NULL;
1589 if (altroot != NULL) {
1590 if (nvlist_alloc(&props, NV_UNIQUE_NAME, 0) != 0) {
1591 return (zfs_error_fmt(hdl, EZFS_NOMEM,
1592 dgettext(TEXT_DOMAIN, "cannot import '%s'"),
1596 if (nvlist_add_string(props,
1597 zpool_prop_to_name(ZPOOL_PROP_ALTROOT), altroot) != 0 ||
1598 nvlist_add_string(props,
1599 zpool_prop_to_name(ZPOOL_PROP_CACHEFILE), "none") != 0) {
1601 return (zfs_error_fmt(hdl, EZFS_NOMEM,
1602 dgettext(TEXT_DOMAIN, "cannot import '%s'"),
1607 ret = zpool_import_props(hdl, config, newname, props,
1615 print_vdev_tree(libzfs_handle_t *hdl, const char *name, nvlist_t *nv,
1621 uint64_t is_log = 0;
1623 (void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_IS_LOG,
1627 (void) printf("\t%*s%s%s\n", indent, "", name,
1628 is_log ? " [log]" : "");
1630 if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN,
1631 &child, &children) != 0)
1634 for (c = 0; c < children; c++) {
1635 vname = zpool_vdev_name(hdl, NULL, child[c], B_TRUE);
1636 print_vdev_tree(hdl, vname, child[c], indent + 2);
1642 zpool_print_unsup_feat(nvlist_t *config)
1644 nvlist_t *nvinfo, *unsup_feat;
1646 verify(nvlist_lookup_nvlist(config, ZPOOL_CONFIG_LOAD_INFO, &nvinfo) ==
1648 verify(nvlist_lookup_nvlist(nvinfo, ZPOOL_CONFIG_UNSUP_FEAT,
1651 for (nvpair_t *nvp = nvlist_next_nvpair(unsup_feat, NULL); nvp != NULL;
1652 nvp = nvlist_next_nvpair(unsup_feat, nvp)) {
1655 verify(nvpair_type(nvp) == DATA_TYPE_STRING);
1656 verify(nvpair_value_string(nvp, &desc) == 0);
1658 if (strlen(desc) > 0)
1659 (void) printf("\t%s (%s)\n", nvpair_name(nvp), desc);
1661 (void) printf("\t%s\n", nvpair_name(nvp));
1666 * Import the given pool using the known configuration and a list of
1667 * properties to be set. The configuration should have come from
1668 * zpool_find_import(). The 'newname' parameters control whether the pool
1669 * is imported with a different name.
1672 zpool_import_props(libzfs_handle_t *hdl, nvlist_t *config, const char *newname,
1673 nvlist_t *props, int flags)
1675 zfs_cmd_t zc = { 0 };
1676 zpool_rewind_policy_t policy;
1677 nvlist_t *nv = NULL;
1678 nvlist_t *nvinfo = NULL;
1679 nvlist_t *missing = NULL;
1686 verify(nvlist_lookup_string(config, ZPOOL_CONFIG_POOL_NAME,
1689 (void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN,
1690 "cannot import pool '%s'"), origname);
1692 if (newname != NULL) {
1693 if (!zpool_name_valid(hdl, B_FALSE, newname))
1694 return (zfs_error_fmt(hdl, EZFS_INVALIDNAME,
1695 dgettext(TEXT_DOMAIN, "cannot import '%s'"),
1697 thename = (char *)newname;
1704 prop_flags_t flags = { .create = B_FALSE, .import = B_TRUE };
1706 verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_VERSION,
1709 if ((props = zpool_valid_proplist(hdl, origname,
1710 props, version, flags, errbuf)) == NULL) {
1712 } else if (zcmd_write_src_nvlist(hdl, &zc, props) != 0) {
1718 (void) strlcpy(zc.zc_name, thename, sizeof (zc.zc_name));
1720 verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID,
1723 if (zcmd_write_conf_nvlist(hdl, &zc, config) != 0) {
1727 if (zcmd_alloc_dst_nvlist(hdl, &zc, zc.zc_nvlist_conf_size * 2) != 0) {
1732 zc.zc_cookie = flags;
1733 while ((ret = zfs_ioctl(hdl, ZFS_IOC_POOL_IMPORT, &zc)) != 0 &&
1735 if (zcmd_expand_dst_nvlist(hdl, &zc) != 0) {
1736 zcmd_free_nvlists(&zc);
1743 (void) zcmd_read_dst_nvlist(hdl, &zc, &nv);
1744 zpool_get_rewind_policy(config, &policy);
1750 * Dry-run failed, but we print out what success
1751 * looks like if we found a best txg
1753 if (policy.zrp_request & ZPOOL_TRY_REWIND) {
1754 zpool_rewind_exclaim(hdl, newname ? origname : thename,
1760 if (newname == NULL)
1761 (void) snprintf(desc, sizeof (desc),
1762 dgettext(TEXT_DOMAIN, "cannot import '%s'"),
1765 (void) snprintf(desc, sizeof (desc),
1766 dgettext(TEXT_DOMAIN, "cannot import '%s' as '%s'"),
1771 if (nv != NULL && nvlist_lookup_nvlist(nv,
1772 ZPOOL_CONFIG_LOAD_INFO, &nvinfo) == 0 &&
1773 nvlist_exists(nvinfo, ZPOOL_CONFIG_UNSUP_FEAT)) {
1774 (void) printf(dgettext(TEXT_DOMAIN, "This "
1775 "pool uses the following feature(s) not "
1776 "supported by this system:\n"));
1777 zpool_print_unsup_feat(nv);
1778 if (nvlist_exists(nvinfo,
1779 ZPOOL_CONFIG_CAN_RDONLY)) {
1780 (void) printf(dgettext(TEXT_DOMAIN,
1781 "All unsupported features are only "
1782 "required for writing to the pool."
1783 "\nThe pool can be imported using "
1784 "'-o readonly=on'.\n"));
1788 * Unsupported version.
1790 (void) zfs_error(hdl, EZFS_BADVERSION, desc);
1794 (void) zfs_error(hdl, EZFS_INVALCONFIG, desc);
1798 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1799 "one or more devices is read only"));
1800 (void) zfs_error(hdl, EZFS_BADDEV, desc);
1804 if (nv && nvlist_lookup_nvlist(nv,
1805 ZPOOL_CONFIG_LOAD_INFO, &nvinfo) == 0 &&
1806 nvlist_lookup_nvlist(nvinfo,
1807 ZPOOL_CONFIG_MISSING_DEVICES, &missing) == 0) {
1808 (void) printf(dgettext(TEXT_DOMAIN,
1809 "The devices below are missing, use "
1810 "'-m' to import the pool anyway:\n"));
1811 print_vdev_tree(hdl, NULL, missing, 2);
1812 (void) printf("\n");
1814 (void) zpool_standard_error(hdl, error, desc);
1818 (void) zpool_standard_error(hdl, error, desc);
1822 (void) zpool_standard_error(hdl, error, desc);
1823 zpool_explain_recover(hdl,
1824 newname ? origname : thename, -error, nv);
1831 zpool_handle_t *zhp;
1834 * This should never fail, but play it safe anyway.
1836 if (zpool_open_silent(hdl, thename, &zhp) != 0)
1838 else if (zhp != NULL)
1840 if (policy.zrp_request &
1841 (ZPOOL_DO_REWIND | ZPOOL_TRY_REWIND)) {
1842 zpool_rewind_exclaim(hdl, newname ? origname : thename,
1843 ((policy.zrp_request & ZPOOL_TRY_REWIND) != 0), nv);
1849 zcmd_free_nvlists(&zc);
1859 zpool_scan(zpool_handle_t *zhp, pool_scan_func_t func)
1861 zfs_cmd_t zc = { 0 };
1863 libzfs_handle_t *hdl = zhp->zpool_hdl;
1865 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
1866 zc.zc_cookie = func;
1868 if (zfs_ioctl(hdl, ZFS_IOC_POOL_SCAN, &zc) == 0 ||
1869 (errno == ENOENT && func != POOL_SCAN_NONE))
1872 if (func == POOL_SCAN_SCRUB) {
1873 (void) snprintf(msg, sizeof (msg),
1874 dgettext(TEXT_DOMAIN, "cannot scrub %s"), zc.zc_name);
1875 } else if (func == POOL_SCAN_NONE) {
1876 (void) snprintf(msg, sizeof (msg),
1877 dgettext(TEXT_DOMAIN, "cannot cancel scrubbing %s"),
1880 assert(!"unexpected result");
1883 if (errno == EBUSY) {
1885 pool_scan_stat_t *ps = NULL;
1888 verify(nvlist_lookup_nvlist(zhp->zpool_config,
1889 ZPOOL_CONFIG_VDEV_TREE, &nvroot) == 0);
1890 (void) nvlist_lookup_uint64_array(nvroot,
1891 ZPOOL_CONFIG_SCAN_STATS, (uint64_t **)&ps, &psc);
1892 if (ps && ps->pss_func == POOL_SCAN_SCRUB)
1893 return (zfs_error(hdl, EZFS_SCRUBBING, msg));
1895 return (zfs_error(hdl, EZFS_RESILVERING, msg));
1896 } else if (errno == ENOENT) {
1897 return (zfs_error(hdl, EZFS_NO_SCRUB, msg));
1899 return (zpool_standard_error(hdl, errno, msg));
1904 * This provides a very minimal check whether a given string is likely a
1905 * c#t#d# style string. Users of this are expected to do their own
1906 * verification of the s# part.
1908 #define CTD_CHECK(str) (str && str[0] == 'c' && isdigit(str[1]))
1911 * More elaborate version for ones which may start with "/dev/dsk/"
1915 ctd_check_path(char *str) {
1917 * If it starts with a slash, check the last component.
1919 if (str && str[0] == '/') {
1920 char *tmp = strrchr(str, '/');
1923 * If it ends in "/old", check the second-to-last
1924 * component of the string instead.
1926 if (tmp != str && strcmp(tmp, "/old") == 0) {
1927 for (tmp--; *tmp != '/'; tmp--)
1932 return (CTD_CHECK(str));
1936 * Find a vdev that matches the search criteria specified. We use the
1937 * the nvpair name to determine how we should look for the device.
1938 * 'avail_spare' is set to TRUE if the provided guid refers to an AVAIL
1939 * spare; but FALSE if its an INUSE spare.
1942 vdev_to_nvlist_iter(nvlist_t *nv, nvlist_t *search, boolean_t *avail_spare,
1943 boolean_t *l2cache, boolean_t *log)
1950 nvpair_t *pair = nvlist_next_nvpair(search, NULL);
1952 /* Nothing to look for */
1953 if (search == NULL || pair == NULL)
1956 /* Obtain the key we will use to search */
1957 srchkey = nvpair_name(pair);
1959 switch (nvpair_type(pair)) {
1960 case DATA_TYPE_UINT64:
1961 if (strcmp(srchkey, ZPOOL_CONFIG_GUID) == 0) {
1962 uint64_t srchval, theguid;
1964 verify(nvpair_value_uint64(pair, &srchval) == 0);
1965 verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID,
1967 if (theguid == srchval)
1972 case DATA_TYPE_STRING: {
1973 char *srchval, *val;
1975 verify(nvpair_value_string(pair, &srchval) == 0);
1976 if (nvlist_lookup_string(nv, srchkey, &val) != 0)
1980 * Search for the requested value. Special cases:
1982 * - ZPOOL_CONFIG_PATH for whole disk entries. These end in
1983 * "s0" or "s0/old". The "s0" part is hidden from the user,
1984 * but included in the string, so this matches around it.
1985 * - looking for a top-level vdev name (i.e. ZPOOL_CONFIG_TYPE).
1987 * Otherwise, all other searches are simple string compares.
1989 if (strcmp(srchkey, ZPOOL_CONFIG_PATH) == 0 &&
1990 ctd_check_path(val)) {
1991 uint64_t wholedisk = 0;
1993 (void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_WHOLE_DISK,
1996 int slen = strlen(srchval);
1997 int vlen = strlen(val);
1999 if (slen != vlen - 2)
2003 * make_leaf_vdev() should only set
2004 * wholedisk for ZPOOL_CONFIG_PATHs which
2005 * will include "/dev/dsk/", giving plenty of
2006 * room for the indices used next.
2011 * strings identical except trailing "s0"
2013 if (strcmp(&val[vlen - 2], "s0") == 0 &&
2014 strncmp(srchval, val, slen) == 0)
2018 * strings identical except trailing "s0/old"
2020 if (strcmp(&val[vlen - 6], "s0/old") == 0 &&
2021 strcmp(&srchval[slen - 4], "/old") == 0 &&
2022 strncmp(srchval, val, slen - 4) == 0)
2027 } else if (strcmp(srchkey, ZPOOL_CONFIG_TYPE) == 0 && val) {
2028 char *type, *idx, *end, *p;
2029 uint64_t id, vdev_id;
2032 * Determine our vdev type, keeping in mind
2033 * that the srchval is composed of a type and
2034 * vdev id pair (i.e. mirror-4).
2036 if ((type = strdup(srchval)) == NULL)
2039 if ((p = strrchr(type, '-')) == NULL) {
2047 * If the types don't match then keep looking.
2049 if (strncmp(val, type, strlen(val)) != 0) {
2054 verify(strncmp(type, VDEV_TYPE_RAIDZ,
2055 strlen(VDEV_TYPE_RAIDZ)) == 0 ||
2056 strncmp(type, VDEV_TYPE_MIRROR,
2057 strlen(VDEV_TYPE_MIRROR)) == 0);
2058 verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_ID,
2062 vdev_id = strtoull(idx, &end, 10);
2069 * Now verify that we have the correct vdev id.
2078 if (strcmp(srchval, val) == 0)
2087 if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN,
2088 &child, &children) != 0)
2091 for (c = 0; c < children; c++) {
2092 if ((ret = vdev_to_nvlist_iter(child[c], search,
2093 avail_spare, l2cache, NULL)) != NULL) {
2095 * The 'is_log' value is only set for the toplevel
2096 * vdev, not the leaf vdevs. So we always lookup the
2097 * log device from the root of the vdev tree (where
2098 * 'log' is non-NULL).
2101 nvlist_lookup_uint64(child[c],
2102 ZPOOL_CONFIG_IS_LOG, &is_log) == 0 &&
2110 if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_SPARES,
2111 &child, &children) == 0) {
2112 for (c = 0; c < children; c++) {
2113 if ((ret = vdev_to_nvlist_iter(child[c], search,
2114 avail_spare, l2cache, NULL)) != NULL) {
2115 *avail_spare = B_TRUE;
2121 if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_L2CACHE,
2122 &child, &children) == 0) {
2123 for (c = 0; c < children; c++) {
2124 if ((ret = vdev_to_nvlist_iter(child[c], search,
2125 avail_spare, l2cache, NULL)) != NULL) {
2136 * Given a physical path (minus the "/devices" prefix), find the
2140 zpool_find_vdev_by_physpath(zpool_handle_t *zhp, const char *ppath,
2141 boolean_t *avail_spare, boolean_t *l2cache, boolean_t *log)
2143 nvlist_t *search, *nvroot, *ret;
2145 verify(nvlist_alloc(&search, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2146 verify(nvlist_add_string(search, ZPOOL_CONFIG_PHYS_PATH, ppath) == 0);
2148 verify(nvlist_lookup_nvlist(zhp->zpool_config, ZPOOL_CONFIG_VDEV_TREE,
2151 *avail_spare = B_FALSE;
2155 ret = vdev_to_nvlist_iter(nvroot, search, avail_spare, l2cache, log);
2156 nvlist_free(search);
2162 * Determine if we have an "interior" top-level vdev (i.e mirror/raidz).
2165 zpool_vdev_is_interior(const char *name)
2167 if (strncmp(name, VDEV_TYPE_RAIDZ, strlen(VDEV_TYPE_RAIDZ)) == 0 ||
2168 strncmp(name, VDEV_TYPE_MIRROR, strlen(VDEV_TYPE_MIRROR)) == 0)
2174 zpool_find_vdev(zpool_handle_t *zhp, const char *path, boolean_t *avail_spare,
2175 boolean_t *l2cache, boolean_t *log)
2177 char buf[MAXPATHLEN];
2179 nvlist_t *nvroot, *search, *ret;
2182 verify(nvlist_alloc(&search, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2184 guid = strtoull(path, &end, 10);
2185 if (guid != 0 && *end == '\0') {
2186 verify(nvlist_add_uint64(search, ZPOOL_CONFIG_GUID, guid) == 0);
2187 } else if (zpool_vdev_is_interior(path)) {
2188 verify(nvlist_add_string(search, ZPOOL_CONFIG_TYPE, path) == 0);
2189 } else if (path[0] != '/') {
2190 (void) snprintf(buf, sizeof (buf), "%s%s", _PATH_DEV, path);
2191 verify(nvlist_add_string(search, ZPOOL_CONFIG_PATH, buf) == 0);
2193 verify(nvlist_add_string(search, ZPOOL_CONFIG_PATH, path) == 0);
2196 verify(nvlist_lookup_nvlist(zhp->zpool_config, ZPOOL_CONFIG_VDEV_TREE,
2199 *avail_spare = B_FALSE;
2203 ret = vdev_to_nvlist_iter(nvroot, search, avail_spare, l2cache, log);
2204 nvlist_free(search);
2210 vdev_online(nvlist_t *nv)
2214 if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_OFFLINE, &ival) == 0 ||
2215 nvlist_lookup_uint64(nv, ZPOOL_CONFIG_FAULTED, &ival) == 0 ||
2216 nvlist_lookup_uint64(nv, ZPOOL_CONFIG_REMOVED, &ival) == 0)
2223 * Helper function for zpool_get_physpaths().
2226 vdev_get_one_physpath(nvlist_t *config, char *physpath, size_t physpath_size,
2227 size_t *bytes_written)
2229 size_t bytes_left, pos, rsz;
2233 if (nvlist_lookup_string(config, ZPOOL_CONFIG_PHYS_PATH,
2235 return (EZFS_NODEVICE);
2237 pos = *bytes_written;
2238 bytes_left = physpath_size - pos;
2239 format = (pos == 0) ? "%s" : " %s";
2241 rsz = snprintf(physpath + pos, bytes_left, format, tmppath);
2242 *bytes_written += rsz;
2244 if (rsz >= bytes_left) {
2245 /* if physpath was not copied properly, clear it */
2246 if (bytes_left != 0) {
2249 return (EZFS_NOSPC);
2255 vdev_get_physpaths(nvlist_t *nv, char *physpath, size_t phypath_size,
2256 size_t *rsz, boolean_t is_spare)
2261 if (nvlist_lookup_string(nv, ZPOOL_CONFIG_TYPE, &type) != 0)
2262 return (EZFS_INVALCONFIG);
2264 if (strcmp(type, VDEV_TYPE_DISK) == 0) {
2266 * An active spare device has ZPOOL_CONFIG_IS_SPARE set.
2267 * For a spare vdev, we only want to boot from the active
2272 (void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_IS_SPARE,
2275 return (EZFS_INVALCONFIG);
2278 if (vdev_online(nv)) {
2279 if ((ret = vdev_get_one_physpath(nv, physpath,
2280 phypath_size, rsz)) != 0)
2283 } else if (strcmp(type, VDEV_TYPE_MIRROR) == 0 ||
2284 strcmp(type, VDEV_TYPE_REPLACING) == 0 ||
2285 (is_spare = (strcmp(type, VDEV_TYPE_SPARE) == 0))) {
2290 if (nvlist_lookup_nvlist_array(nv,
2291 ZPOOL_CONFIG_CHILDREN, &child, &count) != 0)
2292 return (EZFS_INVALCONFIG);
2294 for (i = 0; i < count; i++) {
2295 ret = vdev_get_physpaths(child[i], physpath,
2296 phypath_size, rsz, is_spare);
2297 if (ret == EZFS_NOSPC)
2302 return (EZFS_POOL_INVALARG);
2306 * Get phys_path for a root pool config.
2307 * Return 0 on success; non-zero on failure.
2310 zpool_get_config_physpath(nvlist_t *config, char *physpath, size_t phypath_size)
2313 nvlist_t *vdev_root;
2320 if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE,
2322 return (EZFS_INVALCONFIG);
2324 if (nvlist_lookup_string(vdev_root, ZPOOL_CONFIG_TYPE, &type) != 0 ||
2325 nvlist_lookup_nvlist_array(vdev_root, ZPOOL_CONFIG_CHILDREN,
2326 &child, &count) != 0)
2327 return (EZFS_INVALCONFIG);
2330 * root pool can not have EFI labeled disks and can only have
2331 * a single top-level vdev.
2333 if (strcmp(type, VDEV_TYPE_ROOT) != 0 || count != 1 ||
2334 pool_uses_efi(vdev_root))
2335 return (EZFS_POOL_INVALARG);
2337 (void) vdev_get_physpaths(child[0], physpath, phypath_size, &rsz,
2340 /* No online devices */
2342 return (EZFS_NODEVICE);
2348 * Get phys_path for a root pool
2349 * Return 0 on success; non-zero on failure.
2352 zpool_get_physpath(zpool_handle_t *zhp, char *physpath, size_t phypath_size)
2354 return (zpool_get_config_physpath(zhp->zpool_config, physpath,
2359 * If the device has being dynamically expanded then we need to relabel
2360 * the disk to use the new unallocated space.
2363 zpool_relabel_disk(libzfs_handle_t *hdl, const char *name)
2366 char path[MAXPATHLEN];
2369 int (*_efi_use_whole_disk)(int);
2371 if ((_efi_use_whole_disk = (int (*)(int))dlsym(RTLD_DEFAULT,
2372 "efi_use_whole_disk")) == NULL)
2375 (void) snprintf(path, sizeof (path), "%s/%s", RDISK_ROOT, name);
2377 if ((fd = open(path, O_RDWR | O_NDELAY)) < 0) {
2378 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "cannot "
2379 "relabel '%s': unable to open device"), name);
2380 return (zfs_error(hdl, EZFS_OPENFAILED, errbuf));
2384 * It's possible that we might encounter an error if the device
2385 * does not have any unallocated space left. If so, we simply
2386 * ignore that error and continue on.
2388 error = _efi_use_whole_disk(fd);
2390 if (error && error != VT_ENOSPC) {
2391 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "cannot "
2392 "relabel '%s': unable to read disk capacity"), name);
2393 return (zfs_error(hdl, EZFS_NOCAP, errbuf));
2400 * Bring the specified vdev online. The 'flags' parameter is a set of the
2401 * ZFS_ONLINE_* flags.
2404 zpool_vdev_online(zpool_handle_t *zhp, const char *path, int flags,
2405 vdev_state_t *newstate)
2407 zfs_cmd_t zc = { 0 };
2410 boolean_t avail_spare, l2cache, islog;
2411 libzfs_handle_t *hdl = zhp->zpool_hdl;
2413 if (flags & ZFS_ONLINE_EXPAND) {
2414 (void) snprintf(msg, sizeof (msg),
2415 dgettext(TEXT_DOMAIN, "cannot expand %s"), path);
2417 (void) snprintf(msg, sizeof (msg),
2418 dgettext(TEXT_DOMAIN, "cannot online %s"), path);
2421 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
2422 if ((tgt = zpool_find_vdev(zhp, path, &avail_spare, &l2cache,
2424 return (zfs_error(hdl, EZFS_NODEVICE, msg));
2426 verify(nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID, &zc.zc_guid) == 0);
2429 return (zfs_error(hdl, EZFS_ISSPARE, msg));
2431 if (flags & ZFS_ONLINE_EXPAND ||
2432 zpool_get_prop_int(zhp, ZPOOL_PROP_AUTOEXPAND, NULL)) {
2433 char *pathname = NULL;
2434 uint64_t wholedisk = 0;
2436 (void) nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_WHOLE_DISK,
2438 verify(nvlist_lookup_string(tgt, ZPOOL_CONFIG_PATH,
2442 * XXX - L2ARC 1.0 devices can't support expansion.
2445 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2446 "cannot expand cache devices"));
2447 return (zfs_error(hdl, EZFS_VDEVNOTSUP, msg));
2451 pathname += strlen(DISK_ROOT) + 1;
2452 (void) zpool_relabel_disk(hdl, pathname);
2456 zc.zc_cookie = VDEV_STATE_ONLINE;
2459 if (zfs_ioctl(hdl, ZFS_IOC_VDEV_SET_STATE, &zc) != 0) {
2460 if (errno == EINVAL) {
2461 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "was split "
2462 "from this pool into a new one. Use '%s' "
2463 "instead"), "zpool detach");
2464 return (zfs_error(hdl, EZFS_POSTSPLIT_ONLINE, msg));
2466 return (zpool_standard_error(hdl, errno, msg));
2469 *newstate = zc.zc_cookie;
2474 * Take the specified vdev offline
2477 zpool_vdev_offline(zpool_handle_t *zhp, const char *path, boolean_t istmp)
2479 zfs_cmd_t zc = { 0 };
2482 boolean_t avail_spare, l2cache;
2483 libzfs_handle_t *hdl = zhp->zpool_hdl;
2485 (void) snprintf(msg, sizeof (msg),
2486 dgettext(TEXT_DOMAIN, "cannot offline %s"), path);
2488 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
2489 if ((tgt = zpool_find_vdev(zhp, path, &avail_spare, &l2cache,
2491 return (zfs_error(hdl, EZFS_NODEVICE, msg));
2493 verify(nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID, &zc.zc_guid) == 0);
2496 return (zfs_error(hdl, EZFS_ISSPARE, msg));
2498 zc.zc_cookie = VDEV_STATE_OFFLINE;
2499 zc.zc_obj = istmp ? ZFS_OFFLINE_TEMPORARY : 0;
2501 if (zfs_ioctl(hdl, ZFS_IOC_VDEV_SET_STATE, &zc) == 0)
2508 * There are no other replicas of this device.
2510 return (zfs_error(hdl, EZFS_NOREPLICAS, msg));
2514 * The log device has unplayed logs
2516 return (zfs_error(hdl, EZFS_UNPLAYED_LOGS, msg));
2519 return (zpool_standard_error(hdl, errno, msg));
2524 * Mark the given vdev faulted.
2527 zpool_vdev_fault(zpool_handle_t *zhp, uint64_t guid, vdev_aux_t aux)
2529 zfs_cmd_t zc = { 0 };
2531 libzfs_handle_t *hdl = zhp->zpool_hdl;
2533 (void) snprintf(msg, sizeof (msg),
2534 dgettext(TEXT_DOMAIN, "cannot fault %llu"), guid);
2536 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
2538 zc.zc_cookie = VDEV_STATE_FAULTED;
2541 if (ioctl(hdl->libzfs_fd, ZFS_IOC_VDEV_SET_STATE, &zc) == 0)
2548 * There are no other replicas of this device.
2550 return (zfs_error(hdl, EZFS_NOREPLICAS, msg));
2553 return (zpool_standard_error(hdl, errno, msg));
2559 * Mark the given vdev degraded.
2562 zpool_vdev_degrade(zpool_handle_t *zhp, uint64_t guid, vdev_aux_t aux)
2564 zfs_cmd_t zc = { 0 };
2566 libzfs_handle_t *hdl = zhp->zpool_hdl;
2568 (void) snprintf(msg, sizeof (msg),
2569 dgettext(TEXT_DOMAIN, "cannot degrade %llu"), guid);
2571 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
2573 zc.zc_cookie = VDEV_STATE_DEGRADED;
2576 if (ioctl(hdl->libzfs_fd, ZFS_IOC_VDEV_SET_STATE, &zc) == 0)
2579 return (zpool_standard_error(hdl, errno, msg));
2583 * Returns TRUE if the given nvlist is a vdev that was originally swapped in as
2587 is_replacing_spare(nvlist_t *search, nvlist_t *tgt, int which)
2593 if (nvlist_lookup_nvlist_array(search, ZPOOL_CONFIG_CHILDREN, &child,
2595 verify(nvlist_lookup_string(search, ZPOOL_CONFIG_TYPE,
2598 if (strcmp(type, VDEV_TYPE_SPARE) == 0 &&
2599 children == 2 && child[which] == tgt)
2602 for (c = 0; c < children; c++)
2603 if (is_replacing_spare(child[c], tgt, which))
2611 * Attach new_disk (fully described by nvroot) to old_disk.
2612 * If 'replacing' is specified, the new disk will replace the old one.
2615 zpool_vdev_attach(zpool_handle_t *zhp,
2616 const char *old_disk, const char *new_disk, nvlist_t *nvroot, int replacing)
2618 zfs_cmd_t zc = { 0 };
2622 boolean_t avail_spare, l2cache, islog;
2627 nvlist_t *config_root;
2628 libzfs_handle_t *hdl = zhp->zpool_hdl;
2629 boolean_t rootpool = zpool_is_bootable(zhp);
2632 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN,
2633 "cannot replace %s with %s"), old_disk, new_disk);
2635 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN,
2636 "cannot attach %s to %s"), new_disk, old_disk);
2639 * If this is a root pool, make sure that we're not attaching an
2640 * EFI labeled device.
2642 if (rootpool && pool_uses_efi(nvroot)) {
2643 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2644 "EFI labeled devices are not supported on root pools."));
2645 return (zfs_error(hdl, EZFS_POOL_NOTSUP, msg));
2648 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
2649 if ((tgt = zpool_find_vdev(zhp, old_disk, &avail_spare, &l2cache,
2651 return (zfs_error(hdl, EZFS_NODEVICE, msg));
2654 return (zfs_error(hdl, EZFS_ISSPARE, msg));
2657 return (zfs_error(hdl, EZFS_ISL2CACHE, msg));
2659 verify(nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID, &zc.zc_guid) == 0);
2660 zc.zc_cookie = replacing;
2662 if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_CHILDREN,
2663 &child, &children) != 0 || children != 1) {
2664 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2665 "new device must be a single disk"));
2666 return (zfs_error(hdl, EZFS_INVALCONFIG, msg));
2669 verify(nvlist_lookup_nvlist(zpool_get_config(zhp, NULL),
2670 ZPOOL_CONFIG_VDEV_TREE, &config_root) == 0);
2672 if ((newname = zpool_vdev_name(NULL, NULL, child[0], B_FALSE)) == NULL)
2676 * If the target is a hot spare that has been swapped in, we can only
2677 * replace it with another hot spare.
2680 nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_IS_SPARE, &val) == 0 &&
2681 (zpool_find_vdev(zhp, newname, &avail_spare, &l2cache,
2682 NULL) == NULL || !avail_spare) &&
2683 is_replacing_spare(config_root, tgt, 1)) {
2684 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2685 "can only be replaced by another hot spare"));
2687 return (zfs_error(hdl, EZFS_BADTARGET, msg));
2692 if (zcmd_write_conf_nvlist(hdl, &zc, nvroot) != 0)
2695 ret = zfs_ioctl(hdl, ZFS_IOC_VDEV_ATTACH, &zc);
2697 zcmd_free_nvlists(&zc);
2702 * XXX need a better way to prevent user from
2703 * booting up a half-baked vdev.
2705 (void) fprintf(stderr, dgettext(TEXT_DOMAIN, "Make "
2706 "sure to wait until resilver is done "
2707 "before rebooting.\n"));
2708 (void) fprintf(stderr, "\n");
2709 (void) fprintf(stderr, dgettext(TEXT_DOMAIN, "If "
2710 "you boot from pool '%s', you may need to update\n"
2711 "boot code on newly attached disk '%s'.\n\n"
2712 "Assuming you use GPT partitioning and 'da0' is "
2713 "your new boot disk\n"
2714 "you may use the following command:\n\n"
2715 "\tgpart bootcode -b /boot/pmbr -p "
2716 "/boot/gptzfsboot -i 1 da0\n\n"),
2717 zhp->zpool_name, new_disk);
2725 * Can't attach to or replace this type of vdev.
2728 uint64_t version = zpool_get_prop_int(zhp,
2729 ZPOOL_PROP_VERSION, NULL);
2732 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2733 "cannot replace a log with a spare"));
2734 else if (version >= SPA_VERSION_MULTI_REPLACE)
2735 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2736 "already in replacing/spare config; wait "
2737 "for completion or use 'zpool detach'"));
2739 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2740 "cannot replace a replacing device"));
2742 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2743 "can only attach to mirrors and top-level "
2746 (void) zfs_error(hdl, EZFS_BADTARGET, msg);
2751 * The new device must be a single disk.
2753 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2754 "new device must be a single disk"));
2755 (void) zfs_error(hdl, EZFS_INVALCONFIG, msg);
2759 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "%s is busy"),
2761 (void) zfs_error(hdl, EZFS_BADDEV, msg);
2766 * The new device is too small.
2768 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2769 "device is too small"));
2770 (void) zfs_error(hdl, EZFS_BADDEV, msg);
2775 * The new device has a different alignment requirement.
2777 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2778 "devices have different sector alignment"));
2779 (void) zfs_error(hdl, EZFS_BADDEV, msg);
2784 * The resulting top-level vdev spec won't fit in the label.
2786 (void) zfs_error(hdl, EZFS_DEVOVERFLOW, msg);
2790 (void) zpool_standard_error(hdl, errno, msg);
2797 * Detach the specified device.
2800 zpool_vdev_detach(zpool_handle_t *zhp, const char *path)
2802 zfs_cmd_t zc = { 0 };
2805 boolean_t avail_spare, l2cache;
2806 libzfs_handle_t *hdl = zhp->zpool_hdl;
2808 (void) snprintf(msg, sizeof (msg),
2809 dgettext(TEXT_DOMAIN, "cannot detach %s"), path);
2811 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
2812 if ((tgt = zpool_find_vdev(zhp, path, &avail_spare, &l2cache,
2814 return (zfs_error(hdl, EZFS_NODEVICE, msg));
2817 return (zfs_error(hdl, EZFS_ISSPARE, msg));
2820 return (zfs_error(hdl, EZFS_ISL2CACHE, msg));
2822 verify(nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID, &zc.zc_guid) == 0);
2824 if (zfs_ioctl(hdl, ZFS_IOC_VDEV_DETACH, &zc) == 0)
2831 * Can't detach from this type of vdev.
2833 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "only "
2834 "applicable to mirror and replacing vdevs"));
2835 (void) zfs_error(hdl, EZFS_BADTARGET, msg);
2840 * There are no other replicas of this device.
2842 (void) zfs_error(hdl, EZFS_NOREPLICAS, msg);
2846 (void) zpool_standard_error(hdl, errno, msg);
2853 * Find a mirror vdev in the source nvlist.
2855 * The mchild array contains a list of disks in one of the top-level mirrors
2856 * of the source pool. The schild array contains a list of disks that the
2857 * user specified on the command line. We loop over the mchild array to
2858 * see if any entry in the schild array matches.
2860 * If a disk in the mchild array is found in the schild array, we return
2861 * the index of that entry. Otherwise we return -1.
2864 find_vdev_entry(zpool_handle_t *zhp, nvlist_t **mchild, uint_t mchildren,
2865 nvlist_t **schild, uint_t schildren)
2869 for (mc = 0; mc < mchildren; mc++) {
2871 char *mpath = zpool_vdev_name(zhp->zpool_hdl, zhp,
2872 mchild[mc], B_FALSE);
2874 for (sc = 0; sc < schildren; sc++) {
2875 char *spath = zpool_vdev_name(zhp->zpool_hdl, zhp,
2876 schild[sc], B_FALSE);
2877 boolean_t result = (strcmp(mpath, spath) == 0);
2893 * Split a mirror pool. If newroot points to null, then a new nvlist
2894 * is generated and it is the responsibility of the caller to free it.
2897 zpool_vdev_split(zpool_handle_t *zhp, char *newname, nvlist_t **newroot,
2898 nvlist_t *props, splitflags_t flags)
2900 zfs_cmd_t zc = { 0 };
2902 nvlist_t *tree, *config, **child, **newchild, *newconfig = NULL;
2903 nvlist_t **varray = NULL, *zc_props = NULL;
2904 uint_t c, children, newchildren, lastlog = 0, vcount, found = 0;
2905 libzfs_handle_t *hdl = zhp->zpool_hdl;
2907 boolean_t freelist = B_FALSE, memory_err = B_TRUE;
2910 (void) snprintf(msg, sizeof (msg),
2911 dgettext(TEXT_DOMAIN, "Unable to split %s"), zhp->zpool_name);
2913 if (!zpool_name_valid(hdl, B_FALSE, newname))
2914 return (zfs_error(hdl, EZFS_INVALIDNAME, msg));
2916 if ((config = zpool_get_config(zhp, NULL)) == NULL) {
2917 (void) fprintf(stderr, gettext("Internal error: unable to "
2918 "retrieve pool configuration\n"));
2922 verify(nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE, &tree)
2924 verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_VERSION, &vers) == 0);
2927 prop_flags_t flags = { .create = B_FALSE, .import = B_TRUE };
2928 if ((zc_props = zpool_valid_proplist(hdl, zhp->zpool_name,
2929 props, vers, flags, msg)) == NULL)
2933 if (nvlist_lookup_nvlist_array(tree, ZPOOL_CONFIG_CHILDREN, &child,
2935 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2936 "Source pool is missing vdev tree"));
2938 nvlist_free(zc_props);
2942 varray = zfs_alloc(hdl, children * sizeof (nvlist_t *));
2945 if (*newroot == NULL ||
2946 nvlist_lookup_nvlist_array(*newroot, ZPOOL_CONFIG_CHILDREN,
2947 &newchild, &newchildren) != 0)
2950 for (c = 0; c < children; c++) {
2951 uint64_t is_log = B_FALSE, is_hole = B_FALSE;
2953 nvlist_t **mchild, *vdev;
2958 * Unlike cache & spares, slogs are stored in the
2959 * ZPOOL_CONFIG_CHILDREN array. We filter them out here.
2961 (void) nvlist_lookup_uint64(child[c], ZPOOL_CONFIG_IS_LOG,
2963 (void) nvlist_lookup_uint64(child[c], ZPOOL_CONFIG_IS_HOLE,
2965 if (is_log || is_hole) {
2967 * Create a hole vdev and put it in the config.
2969 if (nvlist_alloc(&vdev, NV_UNIQUE_NAME, 0) != 0)
2971 if (nvlist_add_string(vdev, ZPOOL_CONFIG_TYPE,
2972 VDEV_TYPE_HOLE) != 0)
2974 if (nvlist_add_uint64(vdev, ZPOOL_CONFIG_IS_HOLE,
2979 varray[vcount++] = vdev;
2983 verify(nvlist_lookup_string(child[c], ZPOOL_CONFIG_TYPE, &type)
2985 if (strcmp(type, VDEV_TYPE_MIRROR) != 0) {
2986 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2987 "Source pool must be composed only of mirrors\n"));
2988 retval = zfs_error(hdl, EZFS_INVALCONFIG, msg);
2992 verify(nvlist_lookup_nvlist_array(child[c],
2993 ZPOOL_CONFIG_CHILDREN, &mchild, &mchildren) == 0);
2995 /* find or add an entry for this top-level vdev */
2996 if (newchildren > 0 &&
2997 (entry = find_vdev_entry(zhp, mchild, mchildren,
2998 newchild, newchildren)) >= 0) {
2999 /* We found a disk that the user specified. */
3000 vdev = mchild[entry];
3003 /* User didn't specify a disk for this vdev. */
3004 vdev = mchild[mchildren - 1];
3007 if (nvlist_dup(vdev, &varray[vcount++], 0) != 0)
3011 /* did we find every disk the user specified? */
3012 if (found != newchildren) {
3013 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "Device list must "
3014 "include at most one disk from each mirror"));
3015 retval = zfs_error(hdl, EZFS_INVALCONFIG, msg);
3019 /* Prepare the nvlist for populating. */
3020 if (*newroot == NULL) {
3021 if (nvlist_alloc(newroot, NV_UNIQUE_NAME, 0) != 0)
3024 if (nvlist_add_string(*newroot, ZPOOL_CONFIG_TYPE,
3025 VDEV_TYPE_ROOT) != 0)
3028 verify(nvlist_remove_all(*newroot, ZPOOL_CONFIG_CHILDREN) == 0);
3031 /* Add all the children we found */
3032 if (nvlist_add_nvlist_array(*newroot, ZPOOL_CONFIG_CHILDREN, varray,
3033 lastlog == 0 ? vcount : lastlog) != 0)
3037 * If we're just doing a dry run, exit now with success.
3040 memory_err = B_FALSE;
3045 /* now build up the config list & call the ioctl */
3046 if (nvlist_alloc(&newconfig, NV_UNIQUE_NAME, 0) != 0)
3049 if (nvlist_add_nvlist(newconfig,
3050 ZPOOL_CONFIG_VDEV_TREE, *newroot) != 0 ||
3051 nvlist_add_string(newconfig,
3052 ZPOOL_CONFIG_POOL_NAME, newname) != 0 ||
3053 nvlist_add_uint64(newconfig, ZPOOL_CONFIG_VERSION, vers) != 0)
3057 * The new pool is automatically part of the namespace unless we
3058 * explicitly export it.
3061 zc.zc_cookie = ZPOOL_EXPORT_AFTER_SPLIT;
3062 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
3063 (void) strlcpy(zc.zc_string, newname, sizeof (zc.zc_string));
3064 if (zcmd_write_conf_nvlist(hdl, &zc, newconfig) != 0)
3066 if (zc_props != NULL && zcmd_write_src_nvlist(hdl, &zc, zc_props) != 0)
3069 if (zfs_ioctl(hdl, ZFS_IOC_VDEV_SPLIT, &zc) != 0) {
3070 retval = zpool_standard_error(hdl, errno, msg);
3075 memory_err = B_FALSE;
3078 if (varray != NULL) {
3081 for (v = 0; v < vcount; v++)
3082 nvlist_free(varray[v]);
3085 zcmd_free_nvlists(&zc);
3087 nvlist_free(zc_props);
3089 nvlist_free(newconfig);
3091 nvlist_free(*newroot);
3099 return (no_memory(hdl));
3105 * Remove the given device. Currently, this is supported only for hot spares
3106 * and level 2 cache devices.
3109 zpool_vdev_remove(zpool_handle_t *zhp, const char *path)
3111 zfs_cmd_t zc = { 0 };
3114 boolean_t avail_spare, l2cache, islog;
3115 libzfs_handle_t *hdl = zhp->zpool_hdl;
3118 (void) snprintf(msg, sizeof (msg),
3119 dgettext(TEXT_DOMAIN, "cannot remove %s"), path);
3121 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
3122 if ((tgt = zpool_find_vdev(zhp, path, &avail_spare, &l2cache,
3124 return (zfs_error(hdl, EZFS_NODEVICE, msg));
3126 * XXX - this should just go away.
3128 if (!avail_spare && !l2cache && !islog) {
3129 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3130 "only inactive hot spares, cache, top-level, "
3131 "or log devices can be removed"));
3132 return (zfs_error(hdl, EZFS_NODEVICE, msg));
3135 version = zpool_get_prop_int(zhp, ZPOOL_PROP_VERSION, NULL);
3136 if (islog && version < SPA_VERSION_HOLES) {
3137 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3138 "pool must be upgrade to support log removal"));
3139 return (zfs_error(hdl, EZFS_BADVERSION, msg));
3142 verify(nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID, &zc.zc_guid) == 0);
3144 if (zfs_ioctl(hdl, ZFS_IOC_VDEV_REMOVE, &zc) == 0)
3147 return (zpool_standard_error(hdl, errno, msg));
3151 * Clear the errors for the pool, or the particular device if specified.
3154 zpool_clear(zpool_handle_t *zhp, const char *path, nvlist_t *rewindnvl)
3156 zfs_cmd_t zc = { 0 };
3159 zpool_rewind_policy_t policy;
3160 boolean_t avail_spare, l2cache;
3161 libzfs_handle_t *hdl = zhp->zpool_hdl;
3162 nvlist_t *nvi = NULL;
3166 (void) snprintf(msg, sizeof (msg),
3167 dgettext(TEXT_DOMAIN, "cannot clear errors for %s"),
3170 (void) snprintf(msg, sizeof (msg),
3171 dgettext(TEXT_DOMAIN, "cannot clear errors for %s"),
3174 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
3176 if ((tgt = zpool_find_vdev(zhp, path, &avail_spare,
3177 &l2cache, NULL)) == 0)
3178 return (zfs_error(hdl, EZFS_NODEVICE, msg));
3181 * Don't allow error clearing for hot spares. Do allow
3182 * error clearing for l2cache devices.
3185 return (zfs_error(hdl, EZFS_ISSPARE, msg));
3187 verify(nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID,
3191 zpool_get_rewind_policy(rewindnvl, &policy);
3192 zc.zc_cookie = policy.zrp_request;
3194 if (zcmd_alloc_dst_nvlist(hdl, &zc, zhp->zpool_config_size * 2) != 0)
3197 if (zcmd_write_src_nvlist(hdl, &zc, rewindnvl) != 0)
3200 while ((error = zfs_ioctl(hdl, ZFS_IOC_CLEAR, &zc)) != 0 &&
3202 if (zcmd_expand_dst_nvlist(hdl, &zc) != 0) {
3203 zcmd_free_nvlists(&zc);
3208 if (!error || ((policy.zrp_request & ZPOOL_TRY_REWIND) &&
3209 errno != EPERM && errno != EACCES)) {
3210 if (policy.zrp_request &
3211 (ZPOOL_DO_REWIND | ZPOOL_TRY_REWIND)) {
3212 (void) zcmd_read_dst_nvlist(hdl, &zc, &nvi);
3213 zpool_rewind_exclaim(hdl, zc.zc_name,
3214 ((policy.zrp_request & ZPOOL_TRY_REWIND) != 0),
3218 zcmd_free_nvlists(&zc);
3222 zcmd_free_nvlists(&zc);
3223 return (zpool_standard_error(hdl, errno, msg));
3227 * Similar to zpool_clear(), but takes a GUID (used by fmd).
3230 zpool_vdev_clear(zpool_handle_t *zhp, uint64_t guid)
3232 zfs_cmd_t zc = { 0 };
3234 libzfs_handle_t *hdl = zhp->zpool_hdl;
3236 (void) snprintf(msg, sizeof (msg),
3237 dgettext(TEXT_DOMAIN, "cannot clear errors for %llx"),
3240 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
3242 zc.zc_cookie = ZPOOL_NO_REWIND;
3244 if (ioctl(hdl->libzfs_fd, ZFS_IOC_CLEAR, &zc) == 0)
3247 return (zpool_standard_error(hdl, errno, msg));
3251 * Change the GUID for a pool.
3254 zpool_reguid(zpool_handle_t *zhp)
3257 libzfs_handle_t *hdl = zhp->zpool_hdl;
3258 zfs_cmd_t zc = { 0 };
3260 (void) snprintf(msg, sizeof (msg),
3261 dgettext(TEXT_DOMAIN, "cannot reguid '%s'"), zhp->zpool_name);
3263 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
3264 if (zfs_ioctl(hdl, ZFS_IOC_POOL_REGUID, &zc) == 0)
3267 return (zpool_standard_error(hdl, errno, msg));
3274 zpool_reopen(zpool_handle_t *zhp)
3276 zfs_cmd_t zc = { 0 };
3278 libzfs_handle_t *hdl = zhp->zpool_hdl;
3280 (void) snprintf(msg, sizeof (msg),
3281 dgettext(TEXT_DOMAIN, "cannot reopen '%s'"),
3284 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
3285 if (zfs_ioctl(hdl, ZFS_IOC_POOL_REOPEN, &zc) == 0)
3287 return (zpool_standard_error(hdl, errno, msg));
3291 * Convert from a devid string to a path.
3294 devid_to_path(char *devid_str)
3299 devid_nmlist_t *list = NULL;
3302 if (devid_str_decode(devid_str, &devid, &minor) != 0)
3305 ret = devid_deviceid_to_nmlist("/dev", devid, minor, &list);
3307 devid_str_free(minor);
3313 if ((path = strdup(list[0].devname)) == NULL)
3316 devid_free_nmlist(list);
3322 * Convert from a path to a devid string.
3325 path_to_devid(const char *path)
3331 if ((fd = open(path, O_RDONLY)) < 0)
3336 if (devid_get(fd, &devid) == 0) {
3337 if (devid_get_minor_name(fd, &minor) == 0)
3338 ret = devid_str_encode(devid, minor);
3340 devid_str_free(minor);
3349 * Issue the necessary ioctl() to update the stored path value for the vdev. We
3350 * ignore any failure here, since a common case is for an unprivileged user to
3351 * type 'zpool status', and we'll display the correct information anyway.
3354 set_path(zpool_handle_t *zhp, nvlist_t *nv, const char *path)
3356 zfs_cmd_t zc = { 0 };
3358 (void) strncpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
3359 (void) strncpy(zc.zc_value, path, sizeof (zc.zc_value));
3360 verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID,
3363 (void) ioctl(zhp->zpool_hdl->libzfs_fd, ZFS_IOC_VDEV_SETPATH, &zc);
3367 * Given a vdev, return the name to display in iostat. If the vdev has a path,
3368 * we use that, stripping off any leading "/dev/dsk/"; if not, we use the type.
3369 * We also check if this is a whole disk, in which case we strip off the
3370 * trailing 's0' slice name.
3372 * This routine is also responsible for identifying when disks have been
3373 * reconfigured in a new location. The kernel will have opened the device by
3374 * devid, but the path will still refer to the old location. To catch this, we
3375 * first do a path -> devid translation (which is fast for the common case). If
3376 * the devid matches, we're done. If not, we do a reverse devid -> path
3377 * translation and issue the appropriate ioctl() to update the path of the vdev.
3378 * If 'zhp' is NULL, then this is an exported pool, and we don't need to do any
3382 zpool_vdev_name(libzfs_handle_t *hdl, zpool_handle_t *zhp, nvlist_t *nv,
3393 have_stats = nvlist_lookup_uint64_array(nv, ZPOOL_CONFIG_VDEV_STATS,
3394 (uint64_t **)&vs, &vsc) == 0;
3395 have_path = nvlist_lookup_string(nv, ZPOOL_CONFIG_PATH, &path) == 0;
3398 * If the device is not currently present, assume it will not
3399 * come back at the same device path. Display the device by GUID.
3401 if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_NOT_PRESENT, &value) == 0 ||
3402 have_path && have_stats && vs->vs_state <= VDEV_STATE_CANT_OPEN) {
3403 verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID,
3405 (void) snprintf(buf, sizeof (buf), "%llu",
3406 (u_longlong_t)value);
3408 } else if (have_path) {
3411 * If the device is dead (faulted, offline, etc) then don't
3412 * bother opening it. Otherwise we may be forcing the user to
3413 * open a misbehaving device, which can have undesirable
3416 if ((have_stats == 0 ||
3417 vs->vs_state >= VDEV_STATE_DEGRADED) &&
3419 nvlist_lookup_string(nv, ZPOOL_CONFIG_DEVID, &devid) == 0) {
3421 * Determine if the current path is correct.
3423 char *newdevid = path_to_devid(path);
3425 if (newdevid == NULL ||
3426 strcmp(devid, newdevid) != 0) {
3429 if ((newpath = devid_to_path(devid)) != NULL) {
3431 * Update the path appropriately.
3433 set_path(zhp, nv, newpath);
3434 if (nvlist_add_string(nv,
3435 ZPOOL_CONFIG_PATH, newpath) == 0)
3436 verify(nvlist_lookup_string(nv,
3444 devid_str_free(newdevid);
3448 if (strncmp(path, "/dev/dsk/", 9) == 0)
3451 if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_WHOLE_DISK,
3452 &value) == 0 && value) {
3453 int pathlen = strlen(path);
3454 char *tmp = zfs_strdup(hdl, path);
3457 * If it starts with c#, and ends with "s0", chop
3458 * the "s0" off, or if it ends with "s0/old", remove
3459 * the "s0" from the middle.
3461 if (CTD_CHECK(tmp)) {
3462 if (strcmp(&tmp[pathlen - 2], "s0") == 0) {
3463 tmp[pathlen - 2] = '\0';
3464 } else if (pathlen > 6 &&
3465 strcmp(&tmp[pathlen - 6], "s0/old") == 0) {
3466 (void) strcpy(&tmp[pathlen - 6],
3473 if (strncmp(path, _PATH_DEV, sizeof(_PATH_DEV) - 1) == 0)
3474 path += sizeof(_PATH_DEV) - 1;
3477 verify(nvlist_lookup_string(nv, ZPOOL_CONFIG_TYPE, &path) == 0);
3480 * If it's a raidz device, we need to stick in the parity level.
3482 if (strcmp(path, VDEV_TYPE_RAIDZ) == 0) {
3483 verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_NPARITY,
3485 (void) snprintf(buf, sizeof (buf), "%s%llu", path,
3486 (u_longlong_t)value);
3491 * We identify each top-level vdev by using a <type-id>
3492 * naming convention.
3497 verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_ID,
3499 (void) snprintf(buf, sizeof (buf), "%s-%llu", path,
3505 return (zfs_strdup(hdl, path));
3509 zbookmark_compare(const void *a, const void *b)
3511 return (memcmp(a, b, sizeof (zbookmark_t)));
3515 * Retrieve the persistent error log, uniquify the members, and return to the
3519 zpool_get_errlog(zpool_handle_t *zhp, nvlist_t **nverrlistp)
3521 zfs_cmd_t zc = { 0 };
3523 zbookmark_t *zb = NULL;
3527 * Retrieve the raw error list from the kernel. If the number of errors
3528 * has increased, allocate more space and continue until we get the
3531 verify(nvlist_lookup_uint64(zhp->zpool_config, ZPOOL_CONFIG_ERRCOUNT,
3535 if ((zc.zc_nvlist_dst = (uintptr_t)zfs_alloc(zhp->zpool_hdl,
3536 count * sizeof (zbookmark_t))) == (uintptr_t)NULL)
3538 zc.zc_nvlist_dst_size = count;
3539 (void) strcpy(zc.zc_name, zhp->zpool_name);
3541 if (ioctl(zhp->zpool_hdl->libzfs_fd, ZFS_IOC_ERROR_LOG,
3543 free((void *)(uintptr_t)zc.zc_nvlist_dst);
3544 if (errno == ENOMEM) {
3545 count = zc.zc_nvlist_dst_size;
3546 if ((zc.zc_nvlist_dst = (uintptr_t)
3547 zfs_alloc(zhp->zpool_hdl, count *
3548 sizeof (zbookmark_t))) == (uintptr_t)NULL)
3559 * Sort the resulting bookmarks. This is a little confusing due to the
3560 * implementation of ZFS_IOC_ERROR_LOG. The bookmarks are copied last
3561 * to first, and 'zc_nvlist_dst_size' indicates the number of boomarks
3562 * _not_ copied as part of the process. So we point the start of our
3563 * array appropriate and decrement the total number of elements.
3565 zb = ((zbookmark_t *)(uintptr_t)zc.zc_nvlist_dst) +
3566 zc.zc_nvlist_dst_size;
3567 count -= zc.zc_nvlist_dst_size;
3569 qsort(zb, count, sizeof (zbookmark_t), zbookmark_compare);
3571 verify(nvlist_alloc(nverrlistp, 0, KM_SLEEP) == 0);
3574 * Fill in the nverrlistp with nvlist's of dataset and object numbers.
3576 for (i = 0; i < count; i++) {
3579 /* ignoring zb_blkid and zb_level for now */
3580 if (i > 0 && zb[i-1].zb_objset == zb[i].zb_objset &&
3581 zb[i-1].zb_object == zb[i].zb_object)
3584 if (nvlist_alloc(&nv, NV_UNIQUE_NAME, KM_SLEEP) != 0)
3586 if (nvlist_add_uint64(nv, ZPOOL_ERR_DATASET,
3587 zb[i].zb_objset) != 0) {
3591 if (nvlist_add_uint64(nv, ZPOOL_ERR_OBJECT,
3592 zb[i].zb_object) != 0) {
3596 if (nvlist_add_nvlist(*nverrlistp, "ejk", nv) != 0) {
3603 free((void *)(uintptr_t)zc.zc_nvlist_dst);
3607 free((void *)(uintptr_t)zc.zc_nvlist_dst);
3608 return (no_memory(zhp->zpool_hdl));
3612 * Upgrade a ZFS pool to the latest on-disk version.
3615 zpool_upgrade(zpool_handle_t *zhp, uint64_t new_version)
3617 zfs_cmd_t zc = { 0 };
3618 libzfs_handle_t *hdl = zhp->zpool_hdl;
3620 (void) strcpy(zc.zc_name, zhp->zpool_name);
3621 zc.zc_cookie = new_version;
3623 if (zfs_ioctl(hdl, ZFS_IOC_POOL_UPGRADE, &zc) != 0)
3624 return (zpool_standard_error_fmt(hdl, errno,
3625 dgettext(TEXT_DOMAIN, "cannot upgrade '%s'"),
3631 zfs_save_arguments(int argc, char **argv, char *string, int len)
3633 (void) strlcpy(string, basename(argv[0]), len);
3634 for (int i = 1; i < argc; i++) {
3635 (void) strlcat(string, " ", len);
3636 (void) strlcat(string, argv[i], len);
3641 zpool_log_history(libzfs_handle_t *hdl, const char *message)
3643 zfs_cmd_t zc = { 0 };
3647 args = fnvlist_alloc();
3648 fnvlist_add_string(args, "message", message);
3649 err = zcmd_write_src_nvlist(hdl, &zc, args);
3651 err = ioctl(hdl->libzfs_fd, ZFS_IOC_LOG_HISTORY, &zc);
3653 zcmd_free_nvlists(&zc);
3658 * Perform ioctl to get some command history of a pool.
3660 * 'buf' is the buffer to fill up to 'len' bytes. 'off' is the
3661 * logical offset of the history buffer to start reading from.
3663 * Upon return, 'off' is the next logical offset to read from and
3664 * 'len' is the actual amount of bytes read into 'buf'.
3667 get_history(zpool_handle_t *zhp, char *buf, uint64_t *off, uint64_t *len)
3669 zfs_cmd_t zc = { 0 };
3670 libzfs_handle_t *hdl = zhp->zpool_hdl;
3672 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
3674 zc.zc_history = (uint64_t)(uintptr_t)buf;
3675 zc.zc_history_len = *len;
3676 zc.zc_history_offset = *off;
3678 if (ioctl(hdl->libzfs_fd, ZFS_IOC_POOL_GET_HISTORY, &zc) != 0) {
3681 return (zfs_error_fmt(hdl, EZFS_PERM,
3682 dgettext(TEXT_DOMAIN,
3683 "cannot show history for pool '%s'"),
3686 return (zfs_error_fmt(hdl, EZFS_NOHISTORY,
3687 dgettext(TEXT_DOMAIN, "cannot get history for pool "
3688 "'%s'"), zhp->zpool_name));
3690 return (zfs_error_fmt(hdl, EZFS_BADVERSION,
3691 dgettext(TEXT_DOMAIN, "cannot get history for pool "
3692 "'%s', pool must be upgraded"), zhp->zpool_name));
3694 return (zpool_standard_error_fmt(hdl, errno,
3695 dgettext(TEXT_DOMAIN,
3696 "cannot get history for '%s'"), zhp->zpool_name));
3700 *len = zc.zc_history_len;
3701 *off = zc.zc_history_offset;
3707 * Process the buffer of nvlists, unpacking and storing each nvlist record
3708 * into 'records'. 'leftover' is set to the number of bytes that weren't
3709 * processed as there wasn't a complete record.
3712 zpool_history_unpack(char *buf, uint64_t bytes_read, uint64_t *leftover,
3713 nvlist_t ***records, uint_t *numrecords)
3719 while (bytes_read > sizeof (reclen)) {
3721 /* get length of packed record (stored as little endian) */
3722 for (i = 0, reclen = 0; i < sizeof (reclen); i++)
3723 reclen += (uint64_t)(((uchar_t *)buf)[i]) << (8*i);
3725 if (bytes_read < sizeof (reclen) + reclen)
3729 if (nvlist_unpack(buf + sizeof (reclen), reclen, &nv, 0) != 0)
3731 bytes_read -= sizeof (reclen) + reclen;
3732 buf += sizeof (reclen) + reclen;
3734 /* add record to nvlist array */
3736 if (ISP2(*numrecords + 1)) {
3737 *records = realloc(*records,
3738 *numrecords * 2 * sizeof (nvlist_t *));
3740 (*records)[*numrecords - 1] = nv;
3743 *leftover = bytes_read;
3747 /* from spa_history.c: spa_history_create_obj() */
3748 #define HIS_BUF_LEN_DEF (128 << 10)
3749 #define HIS_BUF_LEN_MAX (1 << 30)
3752 * Retrieve the command history of a pool.
3755 zpool_get_history(zpool_handle_t *zhp, nvlist_t **nvhisp)
3758 uint64_t bufsize = HIS_BUF_LEN_DEF;
3760 nvlist_t **records = NULL;
3761 uint_t numrecords = 0;
3764 if ((buf = malloc(bufsize)) == NULL)
3767 uint64_t bytes_read = bufsize;
3770 if ((err = get_history(zhp, buf, &off, &bytes_read)) != 0)
3773 /* if nothing else was read in, we're at EOF, just return */
3774 if (bytes_read == 0)
3777 if ((err = zpool_history_unpack(buf, bytes_read,
3778 &leftover, &records, &numrecords)) != 0)
3783 * If the history block is too big, double the buffer
3784 * size and try again.
3786 if (leftover == bytes_read) {
3791 if ((bufsize >= HIS_BUF_LEN_MAX) ||
3792 ((buf = malloc(bufsize)) == NULL)) {
3803 verify(nvlist_alloc(nvhisp, NV_UNIQUE_NAME, 0) == 0);
3804 verify(nvlist_add_nvlist_array(*nvhisp, ZPOOL_HIST_RECORD,
3805 records, numrecords) == 0);
3807 for (i = 0; i < numrecords; i++)
3808 nvlist_free(records[i]);
3815 zpool_obj_to_path(zpool_handle_t *zhp, uint64_t dsobj, uint64_t obj,
3816 char *pathname, size_t len)
3818 zfs_cmd_t zc = { 0 };
3819 boolean_t mounted = B_FALSE;
3820 char *mntpnt = NULL;
3821 char dsname[MAXNAMELEN];
3824 /* special case for the MOS */
3825 (void) snprintf(pathname, len, "<metadata>:<0x%llx>", obj);
3829 /* get the dataset's name */
3830 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
3832 if (ioctl(zhp->zpool_hdl->libzfs_fd,
3833 ZFS_IOC_DSOBJ_TO_DSNAME, &zc) != 0) {
3834 /* just write out a path of two object numbers */
3835 (void) snprintf(pathname, len, "<0x%llx>:<0x%llx>",
3839 (void) strlcpy(dsname, zc.zc_value, sizeof (dsname));
3841 /* find out if the dataset is mounted */
3842 mounted = is_mounted(zhp->zpool_hdl, dsname, &mntpnt);
3844 /* get the corrupted object's path */
3845 (void) strlcpy(zc.zc_name, dsname, sizeof (zc.zc_name));
3847 if (ioctl(zhp->zpool_hdl->libzfs_fd, ZFS_IOC_OBJ_TO_PATH,
3850 (void) snprintf(pathname, len, "%s%s", mntpnt,
3853 (void) snprintf(pathname, len, "%s:%s",
3854 dsname, zc.zc_value);
3857 (void) snprintf(pathname, len, "%s:<0x%llx>", dsname, obj);
3864 * Read the EFI label from the config, if a label does not exist then
3865 * pass back the error to the caller. If the caller has passed a non-NULL
3866 * diskaddr argument then we set it to the starting address of the EFI
3870 read_efi_label(nvlist_t *config, diskaddr_t *sb)
3874 char diskname[MAXPATHLEN];
3877 if (nvlist_lookup_string(config, ZPOOL_CONFIG_PATH, &path) != 0)
3880 (void) snprintf(diskname, sizeof (diskname), "%s%s", RDISK_ROOT,
3881 strrchr(path, '/'));
3882 if ((fd = open(diskname, O_RDONLY|O_NDELAY)) >= 0) {
3883 struct dk_gpt *vtoc;
3885 if ((err = efi_alloc_and_read(fd, &vtoc)) >= 0) {
3887 *sb = vtoc->efi_parts[0].p_start;
3896 * determine where a partition starts on a disk in the current
3900 find_start_block(nvlist_t *config)
3904 diskaddr_t sb = MAXOFFSET_T;
3907 if (nvlist_lookup_nvlist_array(config,
3908 ZPOOL_CONFIG_CHILDREN, &child, &children) != 0) {
3909 if (nvlist_lookup_uint64(config,
3910 ZPOOL_CONFIG_WHOLE_DISK,
3911 &wholedisk) != 0 || !wholedisk) {
3912 return (MAXOFFSET_T);
3914 if (read_efi_label(config, &sb) < 0)
3919 for (c = 0; c < children; c++) {
3920 sb = find_start_block(child[c]);
3921 if (sb != MAXOFFSET_T) {
3925 return (MAXOFFSET_T);
3930 * Label an individual disk. The name provided is the short name,
3931 * stripped of any leading /dev path.
3934 zpool_label_disk(libzfs_handle_t *hdl, zpool_handle_t *zhp, const char *name)
3937 char path[MAXPATHLEN];
3938 struct dk_gpt *vtoc;
3940 size_t resv = EFI_MIN_RESV_SIZE;
3941 uint64_t slice_size;
3942 diskaddr_t start_block;
3945 /* prepare an error message just in case */
3946 (void) snprintf(errbuf, sizeof (errbuf),
3947 dgettext(TEXT_DOMAIN, "cannot label '%s'"), name);
3952 if (zpool_is_bootable(zhp)) {
3953 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3954 "EFI labeled devices are not supported on root "
3956 return (zfs_error(hdl, EZFS_POOL_NOTSUP, errbuf));
3959 verify(nvlist_lookup_nvlist(zhp->zpool_config,
3960 ZPOOL_CONFIG_VDEV_TREE, &nvroot) == 0);
3962 if (zhp->zpool_start_block == 0)
3963 start_block = find_start_block(nvroot);
3965 start_block = zhp->zpool_start_block;
3966 zhp->zpool_start_block = start_block;
3969 start_block = NEW_START_BLOCK;
3972 (void) snprintf(path, sizeof (path), "%s/%s%s", RDISK_ROOT, name,
3975 if ((fd = open(path, O_RDWR | O_NDELAY)) < 0) {
3977 * This shouldn't happen. We've long since verified that this
3978 * is a valid device.
3981 dgettext(TEXT_DOMAIN, "unable to open device"));
3982 return (zfs_error(hdl, EZFS_OPENFAILED, errbuf));
3985 if (efi_alloc_and_init(fd, EFI_NUMPAR, &vtoc) != 0) {
3987 * The only way this can fail is if we run out of memory, or we
3988 * were unable to read the disk's capacity
3990 if (errno == ENOMEM)
3991 (void) no_memory(hdl);
3994 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3995 "unable to read disk capacity"), name);
3997 return (zfs_error(hdl, EZFS_NOCAP, errbuf));
4000 slice_size = vtoc->efi_last_u_lba + 1;
4001 slice_size -= EFI_MIN_RESV_SIZE;
4002 if (start_block == MAXOFFSET_T)
4003 start_block = NEW_START_BLOCK;
4004 slice_size -= start_block;
4006 vtoc->efi_parts[0].p_start = start_block;
4007 vtoc->efi_parts[0].p_size = slice_size;
4010 * Why we use V_USR: V_BACKUP confuses users, and is considered
4011 * disposable by some EFI utilities (since EFI doesn't have a backup
4012 * slice). V_UNASSIGNED is supposed to be used only for zero size
4013 * partitions, and efi_write() will fail if we use it. V_ROOT, V_BOOT,
4014 * etc. were all pretty specific. V_USR is as close to reality as we
4015 * can get, in the absence of V_OTHER.
4017 vtoc->efi_parts[0].p_tag = V_USR;
4018 (void) strcpy(vtoc->efi_parts[0].p_name, "zfs");
4020 vtoc->efi_parts[8].p_start = slice_size + start_block;
4021 vtoc->efi_parts[8].p_size = resv;
4022 vtoc->efi_parts[8].p_tag = V_RESERVED;
4024 if (efi_write(fd, vtoc) != 0) {
4026 * Some block drivers (like pcata) may not support EFI
4027 * GPT labels. Print out a helpful error message dir-
4028 * ecting the user to manually label the disk and give
4034 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
4035 "try using fdisk(1M) and then provide a specific slice"));
4036 return (zfs_error(hdl, EZFS_LABELFAILED, errbuf));
4046 supported_dump_vdev_type(libzfs_handle_t *hdl, nvlist_t *config, char *errbuf)
4052 verify(nvlist_lookup_string(config, ZPOOL_CONFIG_TYPE, &type) == 0);
4053 if (strcmp(type, VDEV_TYPE_FILE) == 0 ||
4054 strcmp(type, VDEV_TYPE_HOLE) == 0 ||
4055 strcmp(type, VDEV_TYPE_MISSING) == 0) {
4056 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
4057 "vdev type '%s' is not supported"), type);
4058 (void) zfs_error(hdl, EZFS_VDEVNOTSUP, errbuf);
4061 if (nvlist_lookup_nvlist_array(config, ZPOOL_CONFIG_CHILDREN,
4062 &child, &children) == 0) {
4063 for (c = 0; c < children; c++) {
4064 if (!supported_dump_vdev_type(hdl, child[c], errbuf))
4072 * Check if this zvol is allowable for use as a dump device; zero if
4073 * it is, > 0 if it isn't, < 0 if it isn't a zvol.
4075 * Allowable storage configurations include mirrors, all raidz variants, and
4076 * pools with log, cache, and spare devices. Pools which are backed by files or
4077 * have missing/hole vdevs are not suitable.
4080 zvol_check_dump_config(char *arg)
4082 zpool_handle_t *zhp = NULL;
4083 nvlist_t *config, *nvroot;
4087 libzfs_handle_t *hdl;
4089 char poolname[ZPOOL_MAXNAMELEN];
4090 int pathlen = strlen(ZVOL_FULL_DEV_DIR);
4093 if (strncmp(arg, ZVOL_FULL_DEV_DIR, pathlen)) {
4097 (void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN,
4098 "dump is not supported on device '%s'"), arg);
4100 if ((hdl = libzfs_init()) == NULL)
4102 libzfs_print_on_error(hdl, B_TRUE);
4104 volname = arg + pathlen;
4106 /* check the configuration of the pool */
4107 if ((p = strchr(volname, '/')) == NULL) {
4108 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
4109 "malformed dataset name"));
4110 (void) zfs_error(hdl, EZFS_INVALIDNAME, errbuf);
4112 } else if (p - volname >= ZFS_MAXNAMELEN) {
4113 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
4114 "dataset name is too long"));
4115 (void) zfs_error(hdl, EZFS_NAMETOOLONG, errbuf);
4118 (void) strncpy(poolname, volname, p - volname);
4119 poolname[p - volname] = '\0';
4122 if ((zhp = zpool_open(hdl, poolname)) == NULL) {
4123 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
4124 "could not open pool '%s'"), poolname);
4125 (void) zfs_error(hdl, EZFS_OPENFAILED, errbuf);
4128 config = zpool_get_config(zhp, NULL);
4129 if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE,
4131 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
4132 "could not obtain vdev configuration for '%s'"), poolname);
4133 (void) zfs_error(hdl, EZFS_INVALCONFIG, errbuf);
4137 verify(nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_CHILDREN,
4138 &top, &toplevels) == 0);
4140 if (!supported_dump_vdev_type(hdl, top[0], errbuf)) {