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 2015 Nexenta Systems, Inc. All rights reserved.
24 * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
25 * Copyright (c) 2011, 2014 by Delphix. All rights reserved.
40 #include <sys/efi_partition.h>
42 #include <sys/zfs_ioctl.h>
45 #include "zfs_namecheck.h"
47 #include "libzfs_impl.h"
48 #include "zfs_comutil.h"
49 #include "zfeature_common.h"
51 static int read_efi_label(nvlist_t *config, diskaddr_t *sb);
53 typedef struct prop_flags {
54 int create:1; /* Validate property on creation */
55 int import:1; /* Validate property on import */
59 * ====================================================================
60 * zpool property functions
61 * ====================================================================
65 zpool_get_all_props(zpool_handle_t *zhp)
67 zfs_cmd_t zc = {"\0"};
68 libzfs_handle_t *hdl = zhp->zpool_hdl;
70 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
72 if (zcmd_alloc_dst_nvlist(hdl, &zc, 0) != 0)
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);
82 zcmd_free_nvlists(&zc);
87 if (zcmd_read_dst_nvlist(hdl, &zc, &zhp->zpool_props) != 0) {
88 zcmd_free_nvlists(&zc);
92 zcmd_free_nvlists(&zc);
98 zpool_props_refresh(zpool_handle_t *zhp)
102 old_props = zhp->zpool_props;
104 if (zpool_get_all_props(zhp) != 0)
107 nvlist_free(old_props);
112 zpool_get_prop_string(zpool_handle_t *zhp, zpool_prop_t prop,
118 zprop_source_t source;
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);
124 verify(nvlist_lookup_string(nv, ZPROP_VALUE, &value) == 0);
126 source = ZPROP_SRC_DEFAULT;
127 if ((value = (char *)zpool_prop_default_string(prop)) == NULL)
138 zpool_get_prop_int(zpool_handle_t *zhp, zpool_prop_t prop, zprop_source_t *src)
142 zprop_source_t source;
144 if (zhp->zpool_props == NULL && zpool_get_all_props(zhp)) {
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.
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)
157 return (zpool_prop_default_numeric(prop));
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);
164 verify(nvlist_lookup_uint64(nv, ZPROP_VALUE, &value) == 0);
166 source = ZPROP_SRC_DEFAULT;
167 value = zpool_prop_default_numeric(prop);
177 * Map VDEV STATE to printed strings.
180 zpool_state_to_name(vdev_state_t state, vdev_aux_t aux)
185 case VDEV_STATE_CLOSED:
186 case VDEV_STATE_OFFLINE:
187 return (gettext("OFFLINE"));
188 case VDEV_STATE_REMOVED:
189 return (gettext("REMOVED"));
190 case VDEV_STATE_CANT_OPEN:
191 if (aux == VDEV_AUX_CORRUPT_DATA || aux == VDEV_AUX_BAD_LOG)
192 return (gettext("FAULTED"));
193 else if (aux == VDEV_AUX_SPLIT_POOL)
194 return (gettext("SPLIT"));
196 return (gettext("UNAVAIL"));
197 case VDEV_STATE_FAULTED:
198 return (gettext("FAULTED"));
199 case VDEV_STATE_DEGRADED:
200 return (gettext("DEGRADED"));
201 case VDEV_STATE_HEALTHY:
202 return (gettext("ONLINE"));
205 return (gettext("UNKNOWN"));
209 * Map POOL STATE to printed strings.
212 zpool_pool_state_to_name(pool_state_t state)
217 case POOL_STATE_ACTIVE:
218 return (gettext("ACTIVE"));
219 case POOL_STATE_EXPORTED:
220 return (gettext("EXPORTED"));
221 case POOL_STATE_DESTROYED:
222 return (gettext("DESTROYED"));
223 case POOL_STATE_SPARE:
224 return (gettext("SPARE"));
225 case POOL_STATE_L2CACHE:
226 return (gettext("L2CACHE"));
227 case POOL_STATE_UNINITIALIZED:
228 return (gettext("UNINITIALIZED"));
229 case POOL_STATE_UNAVAIL:
230 return (gettext("UNAVAIL"));
231 case POOL_STATE_POTENTIALLY_ACTIVE:
232 return (gettext("POTENTIALLY_ACTIVE"));
235 return (gettext("UNKNOWN"));
239 * Get a zpool property value for 'prop' and return the value in
240 * a pre-allocated buffer.
243 zpool_get_prop(zpool_handle_t *zhp, zpool_prop_t prop, char *buf,
244 size_t len, zprop_source_t *srctype, boolean_t literal)
248 zprop_source_t src = ZPROP_SRC_NONE;
253 if (zpool_get_state(zhp) == POOL_STATE_UNAVAIL) {
255 case ZPOOL_PROP_NAME:
256 (void) strlcpy(buf, zpool_get_name(zhp), len);
259 case ZPOOL_PROP_HEALTH:
260 (void) strlcpy(buf, "FAULTED", len);
263 case ZPOOL_PROP_GUID:
264 intval = zpool_get_prop_int(zhp, prop, &src);
265 (void) snprintf(buf, len, "%llu", (u_longlong_t)intval);
268 case ZPOOL_PROP_ALTROOT:
269 case ZPOOL_PROP_CACHEFILE:
270 case ZPOOL_PROP_COMMENT:
271 if (zhp->zpool_props != NULL ||
272 zpool_get_all_props(zhp) == 0) {
274 zpool_get_prop_string(zhp, prop, &src),
280 (void) strlcpy(buf, "-", len);
289 if (zhp->zpool_props == NULL && zpool_get_all_props(zhp) &&
290 prop != ZPOOL_PROP_NAME)
293 switch (zpool_prop_get_type(prop)) {
294 case PROP_TYPE_STRING:
295 (void) strlcpy(buf, zpool_get_prop_string(zhp, prop, &src),
299 case PROP_TYPE_NUMBER:
300 intval = zpool_get_prop_int(zhp, prop, &src);
303 case ZPOOL_PROP_SIZE:
304 case ZPOOL_PROP_ALLOCATED:
305 case ZPOOL_PROP_FREE:
306 case ZPOOL_PROP_FREEING:
307 case ZPOOL_PROP_LEAKED:
308 case ZPOOL_PROP_ASHIFT:
310 (void) snprintf(buf, len, "%llu",
311 (u_longlong_t)intval);
313 (void) zfs_nicenum(intval, buf, len);
316 case ZPOOL_PROP_EXPANDSZ:
318 (void) strlcpy(buf, "-", len);
319 } else if (literal) {
320 (void) snprintf(buf, len, "%llu",
321 (u_longlong_t)intval);
323 (void) zfs_nicenum(intval, buf, len);
327 case ZPOOL_PROP_CAPACITY:
329 (void) snprintf(buf, len, "%llu",
330 (u_longlong_t)intval);
332 (void) snprintf(buf, len, "%llu%%",
333 (u_longlong_t)intval);
337 case ZPOOL_PROP_FRAGMENTATION:
338 if (intval == UINT64_MAX) {
339 (void) strlcpy(buf, "-", len);
341 (void) snprintf(buf, len, "%llu%%",
342 (u_longlong_t)intval);
346 case ZPOOL_PROP_DEDUPRATIO:
347 (void) snprintf(buf, len, "%llu.%02llux",
348 (u_longlong_t)(intval / 100),
349 (u_longlong_t)(intval % 100));
352 case ZPOOL_PROP_HEALTH:
353 verify(nvlist_lookup_nvlist(zpool_get_config(zhp, NULL),
354 ZPOOL_CONFIG_VDEV_TREE, &nvroot) == 0);
355 verify(nvlist_lookup_uint64_array(nvroot,
356 ZPOOL_CONFIG_VDEV_STATS, (uint64_t **)&vs, &vsc)
359 (void) strlcpy(buf, zpool_state_to_name(intval,
362 case ZPOOL_PROP_VERSION:
363 if (intval >= SPA_VERSION_FEATURES) {
364 (void) snprintf(buf, len, "-");
369 (void) snprintf(buf, len, "%llu", (u_longlong_t)intval);
373 case PROP_TYPE_INDEX:
374 intval = zpool_get_prop_int(zhp, prop, &src);
375 if (zpool_prop_index_to_string(prop, intval, &strval)
378 (void) strlcpy(buf, strval, len);
392 * Check if the bootfs name has the same pool name as it is set to.
393 * Assuming bootfs is a valid dataset name.
396 bootfs_name_valid(const char *pool, char *bootfs)
398 int len = strlen(pool);
400 if (!zfs_name_valid(bootfs, ZFS_TYPE_FILESYSTEM|ZFS_TYPE_SNAPSHOT))
403 if (strncmp(pool, bootfs, len) == 0 &&
404 (bootfs[len] == '/' || bootfs[len] == '\0'))
410 #if defined(__sun__) || defined(__sun)
412 * Inspect the configuration to determine if any of the devices contain
416 pool_uses_efi(nvlist_t *config)
421 if (nvlist_lookup_nvlist_array(config, ZPOOL_CONFIG_CHILDREN,
422 &child, &children) != 0)
423 return (read_efi_label(config, NULL) >= 0);
425 for (c = 0; c < children; c++) {
426 if (pool_uses_efi(child[c]))
434 zpool_is_bootable(zpool_handle_t *zhp)
436 char bootfs[ZPOOL_MAXNAMELEN];
438 return (zpool_get_prop(zhp, ZPOOL_PROP_BOOTFS, bootfs,
439 sizeof (bootfs), NULL, B_FALSE) == 0 && strncmp(bootfs, "-",
440 sizeof (bootfs)) != 0);
445 * Given an nvlist of zpool properties to be set, validate that they are
446 * correct, and parse any numeric properties (index, boolean, etc) if they are
447 * specified as strings.
450 zpool_valid_proplist(libzfs_handle_t *hdl, const char *poolname,
451 nvlist_t *props, uint64_t version, prop_flags_t flags, char *errbuf)
459 struct stat64 statbuf;
463 if (nvlist_alloc(&retprops, NV_UNIQUE_NAME, 0) != 0) {
464 (void) no_memory(hdl);
469 while ((elem = nvlist_next_nvpair(props, elem)) != NULL) {
470 const char *propname = nvpair_name(elem);
472 prop = zpool_name_to_prop(propname);
473 if (prop == ZPROP_INVAL && zpool_prop_feature(propname)) {
475 char *fname = strchr(propname, '@') + 1;
477 err = zfeature_lookup_name(fname, NULL);
479 ASSERT3U(err, ==, ENOENT);
480 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
481 "invalid feature '%s'"), fname);
482 (void) zfs_error(hdl, EZFS_BADPROP, errbuf);
486 if (nvpair_type(elem) != DATA_TYPE_STRING) {
487 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
488 "'%s' must be a string"), propname);
489 (void) zfs_error(hdl, EZFS_BADPROP, errbuf);
493 (void) nvpair_value_string(elem, &strval);
494 if (strcmp(strval, ZFS_FEATURE_ENABLED) != 0) {
495 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
496 "property '%s' can only be set to "
497 "'enabled'"), propname);
498 (void) zfs_error(hdl, EZFS_BADPROP, errbuf);
502 if (nvlist_add_uint64(retprops, propname, 0) != 0) {
503 (void) no_memory(hdl);
510 * Make sure this property is valid and applies to this type.
512 if (prop == ZPROP_INVAL) {
513 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
514 "invalid property '%s'"), propname);
515 (void) zfs_error(hdl, EZFS_BADPROP, errbuf);
519 if (zpool_prop_readonly(prop)) {
520 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "'%s' "
521 "is readonly"), propname);
522 (void) zfs_error(hdl, EZFS_PROPREADONLY, errbuf);
526 if (zprop_parse_value(hdl, elem, prop, ZFS_TYPE_POOL, retprops,
527 &strval, &intval, errbuf) != 0)
531 * Perform additional checking for specific properties.
536 case ZPOOL_PROP_VERSION:
537 if (intval < version ||
538 !SPA_VERSION_IS_SUPPORTED(intval)) {
539 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
540 "property '%s' number %d is invalid."),
542 (void) zfs_error(hdl, EZFS_BADVERSION, errbuf);
547 case ZPOOL_PROP_ASHIFT:
549 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
550 "property '%s' can only be set at "
551 "creation time"), propname);
552 (void) zfs_error(hdl, EZFS_BADPROP, errbuf);
556 if (intval != 0 && (intval < 9 || intval > 13)) {
557 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
558 "property '%s' number %d is invalid."),
560 (void) zfs_error(hdl, EZFS_BADPROP, errbuf);
565 case ZPOOL_PROP_BOOTFS:
566 if (flags.create || flags.import) {
567 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
568 "property '%s' cannot be set at creation "
569 "or import time"), propname);
570 (void) zfs_error(hdl, EZFS_BADPROP, errbuf);
574 if (version < SPA_VERSION_BOOTFS) {
575 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
576 "pool must be upgraded to support "
577 "'%s' property"), propname);
578 (void) zfs_error(hdl, EZFS_BADVERSION, errbuf);
583 * bootfs property value has to be a dataset name and
584 * the dataset has to be in the same pool as it sets to.
586 if (strval[0] != '\0' && !bootfs_name_valid(poolname,
588 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "'%s' "
589 "is an invalid name"), strval);
590 (void) zfs_error(hdl, EZFS_INVALIDNAME, errbuf);
594 if ((zhp = zpool_open_canfail(hdl, poolname)) == NULL) {
595 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
596 "could not open pool '%s'"), poolname);
597 (void) zfs_error(hdl, EZFS_OPENFAILED, errbuf);
600 verify(nvlist_lookup_nvlist(zpool_get_config(zhp, NULL),
601 ZPOOL_CONFIG_VDEV_TREE, &nvroot) == 0);
603 #if defined(__sun__) || defined(__sun)
605 * bootfs property cannot be set on a disk which has
608 if (pool_uses_efi(nvroot)) {
609 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
610 "property '%s' not supported on "
611 "EFI labeled devices"), propname);
612 (void) zfs_error(hdl, EZFS_POOL_NOTSUP, errbuf);
620 case ZPOOL_PROP_ALTROOT:
621 if (!flags.create && !flags.import) {
622 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
623 "property '%s' can only be set during pool "
624 "creation or import"), propname);
625 (void) zfs_error(hdl, EZFS_BADPROP, errbuf);
629 if (strval[0] != '/') {
630 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
631 "bad alternate root '%s'"), strval);
632 (void) zfs_error(hdl, EZFS_BADPATH, errbuf);
637 case ZPOOL_PROP_CACHEFILE:
638 if (strval[0] == '\0')
641 if (strcmp(strval, "none") == 0)
644 if (strval[0] != '/') {
645 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
646 "property '%s' must be empty, an "
647 "absolute path, or 'none'"), propname);
648 (void) zfs_error(hdl, EZFS_BADPATH, errbuf);
652 slash = strrchr(strval, '/');
654 if (slash[1] == '\0' || strcmp(slash, "/.") == 0 ||
655 strcmp(slash, "/..") == 0) {
656 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
657 "'%s' is not a valid file"), strval);
658 (void) zfs_error(hdl, EZFS_BADPATH, errbuf);
664 if (strval[0] != '\0' &&
665 (stat64(strval, &statbuf) != 0 ||
666 !S_ISDIR(statbuf.st_mode))) {
667 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
668 "'%s' is not a valid directory"),
670 (void) zfs_error(hdl, EZFS_BADPATH, errbuf);
677 case ZPOOL_PROP_COMMENT:
678 for (check = strval; *check != '\0'; check++) {
679 if (!isprint(*check)) {
681 dgettext(TEXT_DOMAIN,
682 "comment may only have printable "
684 (void) zfs_error(hdl, EZFS_BADPROP,
689 if (strlen(strval) > ZPROP_MAX_COMMENT) {
690 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
691 "comment must not exceed %d characters"),
693 (void) zfs_error(hdl, EZFS_BADPROP, errbuf);
697 case ZPOOL_PROP_READONLY:
699 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
700 "property '%s' can only be set at "
701 "import time"), propname);
702 (void) zfs_error(hdl, EZFS_BADPROP, errbuf);
706 case ZPOOL_PROP_TNAME:
708 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
709 "property '%s' can only be set at "
710 "creation time"), propname);
711 (void) zfs_error(hdl, EZFS_BADPROP, errbuf);
720 nvlist_free(retprops);
725 * Set zpool property : propname=propval.
728 zpool_set_prop(zpool_handle_t *zhp, const char *propname, const char *propval)
730 zfs_cmd_t zc = {"\0"};
733 nvlist_t *nvl = NULL;
736 prop_flags_t flags = { 0 };
738 (void) snprintf(errbuf, sizeof (errbuf),
739 dgettext(TEXT_DOMAIN, "cannot set property for '%s'"),
742 if (nvlist_alloc(&nvl, NV_UNIQUE_NAME, 0) != 0)
743 return (no_memory(zhp->zpool_hdl));
745 if (nvlist_add_string(nvl, propname, propval) != 0) {
747 return (no_memory(zhp->zpool_hdl));
750 version = zpool_get_prop_int(zhp, ZPOOL_PROP_VERSION, NULL);
751 if ((realprops = zpool_valid_proplist(zhp->zpool_hdl,
752 zhp->zpool_name, nvl, version, flags, errbuf)) == NULL) {
761 * Execute the corresponding ioctl() to set this property.
763 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
765 if (zcmd_write_src_nvlist(zhp->zpool_hdl, &zc, nvl) != 0) {
770 ret = zfs_ioctl(zhp->zpool_hdl, ZFS_IOC_POOL_SET_PROPS, &zc);
772 zcmd_free_nvlists(&zc);
776 (void) zpool_standard_error(zhp->zpool_hdl, errno, errbuf);
778 (void) zpool_props_refresh(zhp);
784 zpool_expand_proplist(zpool_handle_t *zhp, zprop_list_t **plp)
786 libzfs_handle_t *hdl = zhp->zpool_hdl;
788 char buf[ZFS_MAXPROPLEN];
789 nvlist_t *features = NULL;
792 boolean_t firstexpand = (NULL == *plp);
795 if (zprop_expand_list(hdl, plp, ZFS_TYPE_POOL) != 0)
799 while (*last != NULL)
800 last = &(*last)->pl_next;
803 features = zpool_get_features(zhp);
805 if ((*plp)->pl_all && firstexpand) {
806 for (i = 0; i < SPA_FEATURES; i++) {
807 zprop_list_t *entry = zfs_alloc(hdl,
808 sizeof (zprop_list_t));
809 entry->pl_prop = ZPROP_INVAL;
810 entry->pl_user_prop = zfs_asprintf(hdl, "feature@%s",
811 spa_feature_table[i].fi_uname);
812 entry->pl_width = strlen(entry->pl_user_prop);
813 entry->pl_all = B_TRUE;
816 last = &entry->pl_next;
820 /* add any unsupported features */
821 for (nvp = nvlist_next_nvpair(features, NULL);
822 nvp != NULL; nvp = nvlist_next_nvpair(features, nvp)) {
827 if (zfeature_is_supported(nvpair_name(nvp)))
830 propname = zfs_asprintf(hdl, "unsupported@%s",
834 * Before adding the property to the list make sure that no
835 * other pool already added the same property.
839 while (entry != NULL) {
840 if (entry->pl_user_prop != NULL &&
841 strcmp(propname, entry->pl_user_prop) == 0) {
845 entry = entry->pl_next;
852 entry = zfs_alloc(hdl, sizeof (zprop_list_t));
853 entry->pl_prop = ZPROP_INVAL;
854 entry->pl_user_prop = propname;
855 entry->pl_width = strlen(entry->pl_user_prop);
856 entry->pl_all = B_TRUE;
859 last = &entry->pl_next;
862 for (entry = *plp; entry != NULL; entry = entry->pl_next) {
867 if (entry->pl_prop != ZPROP_INVAL &&
868 zpool_get_prop(zhp, entry->pl_prop, buf, sizeof (buf),
869 NULL, B_FALSE) == 0) {
870 if (strlen(buf) > entry->pl_width)
871 entry->pl_width = strlen(buf);
879 * Get the state for the given feature on the given ZFS pool.
882 zpool_prop_get_feature(zpool_handle_t *zhp, const char *propname, char *buf,
886 boolean_t found = B_FALSE;
887 nvlist_t *features = zpool_get_features(zhp);
889 const char *feature = strchr(propname, '@') + 1;
891 supported = zpool_prop_feature(propname);
892 ASSERT(supported || zpool_prop_unsupported(propname));
895 * Convert from feature name to feature guid. This conversion is
896 * unecessary for unsupported@... properties because they already
903 ret = zfeature_lookup_name(feature, &fid);
905 (void) strlcpy(buf, "-", len);
908 feature = spa_feature_table[fid].fi_guid;
911 if (nvlist_lookup_uint64(features, feature, &refcount) == 0)
916 (void) strlcpy(buf, ZFS_FEATURE_DISABLED, len);
919 (void) strlcpy(buf, ZFS_FEATURE_ENABLED, len);
921 (void) strlcpy(buf, ZFS_FEATURE_ACTIVE, len);
926 (void) strcpy(buf, ZFS_UNSUPPORTED_INACTIVE);
928 (void) strcpy(buf, ZFS_UNSUPPORTED_READONLY);
931 (void) strlcpy(buf, "-", len);
940 * Don't start the slice at the default block of 34; many storage
941 * devices will use a stripe width of 128k, other vendors prefer a 1m
942 * alignment. It is best to play it safe and ensure a 1m alignment
943 * given 512B blocks. When the block size is larger by a power of 2
944 * we will still be 1m aligned. Some devices are sensitive to the
945 * partition ending alignment as well.
947 #define NEW_START_BLOCK 2048
948 #define PARTITION_END_ALIGNMENT 2048
951 * Validate the given pool name, optionally putting an extended error message in
955 zpool_name_valid(libzfs_handle_t *hdl, boolean_t isopen, const char *pool)
961 ret = pool_namecheck(pool, &why, &what);
964 * The rules for reserved pool names were extended at a later point.
965 * But we need to support users with existing pools that may now be
966 * invalid. So we only check for this expanded set of names during a
967 * create (or import), and only in userland.
969 if (ret == 0 && !isopen &&
970 (strncmp(pool, "mirror", 6) == 0 ||
971 strncmp(pool, "raidz", 5) == 0 ||
972 strncmp(pool, "spare", 5) == 0 ||
973 strcmp(pool, "log") == 0)) {
976 dgettext(TEXT_DOMAIN, "name is reserved"));
984 case NAME_ERR_TOOLONG:
986 dgettext(TEXT_DOMAIN, "name is too long"));
989 case NAME_ERR_INVALCHAR:
991 dgettext(TEXT_DOMAIN, "invalid character "
992 "'%c' in pool name"), what);
995 case NAME_ERR_NOLETTER:
996 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
997 "name must begin with a letter"));
1000 case NAME_ERR_RESERVED:
1001 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1002 "name is reserved"));
1005 case NAME_ERR_DISKLIKE:
1006 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1007 "pool name is reserved"));
1010 case NAME_ERR_LEADING_SLASH:
1011 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1012 "leading slash in name"));
1015 case NAME_ERR_EMPTY_COMPONENT:
1016 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1017 "empty component in name"));
1020 case NAME_ERR_TRAILING_SLASH:
1021 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1022 "trailing slash in name"));
1025 case NAME_ERR_MULTIPLE_AT:
1026 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1027 "multiple '@' delimiters in name"));
1029 case NAME_ERR_NO_AT:
1030 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1031 "permission set is missing '@'"));
1042 * Open a handle to the given pool, even if the pool is currently in the FAULTED
1046 zpool_open_canfail(libzfs_handle_t *hdl, const char *pool)
1048 zpool_handle_t *zhp;
1052 * Make sure the pool name is valid.
1054 if (!zpool_name_valid(hdl, B_TRUE, pool)) {
1055 (void) zfs_error_fmt(hdl, EZFS_INVALIDNAME,
1056 dgettext(TEXT_DOMAIN, "cannot open '%s'"),
1061 if ((zhp = zfs_alloc(hdl, sizeof (zpool_handle_t))) == NULL)
1064 zhp->zpool_hdl = hdl;
1065 (void) strlcpy(zhp->zpool_name, pool, sizeof (zhp->zpool_name));
1067 if (zpool_refresh_stats(zhp, &missing) != 0) {
1073 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "no such pool"));
1074 (void) zfs_error_fmt(hdl, EZFS_NOENT,
1075 dgettext(TEXT_DOMAIN, "cannot open '%s'"), pool);
1084 * Like the above, but silent on error. Used when iterating over pools (because
1085 * the configuration cache may be out of date).
1088 zpool_open_silent(libzfs_handle_t *hdl, const char *pool, zpool_handle_t **ret)
1090 zpool_handle_t *zhp;
1093 if ((zhp = zfs_alloc(hdl, sizeof (zpool_handle_t))) == NULL)
1096 zhp->zpool_hdl = hdl;
1097 (void) strlcpy(zhp->zpool_name, pool, sizeof (zhp->zpool_name));
1099 if (zpool_refresh_stats(zhp, &missing) != 0) {
1115 * Similar to zpool_open_canfail(), but refuses to open pools in the faulted
1119 zpool_open(libzfs_handle_t *hdl, const char *pool)
1121 zpool_handle_t *zhp;
1123 if ((zhp = zpool_open_canfail(hdl, pool)) == NULL)
1126 if (zhp->zpool_state == POOL_STATE_UNAVAIL) {
1127 (void) zfs_error_fmt(hdl, EZFS_POOLUNAVAIL,
1128 dgettext(TEXT_DOMAIN, "cannot open '%s'"), zhp->zpool_name);
1137 * Close the handle. Simply frees the memory associated with the handle.
1140 zpool_close(zpool_handle_t *zhp)
1142 nvlist_free(zhp->zpool_config);
1143 nvlist_free(zhp->zpool_old_config);
1144 nvlist_free(zhp->zpool_props);
1149 * Return the name of the pool.
1152 zpool_get_name(zpool_handle_t *zhp)
1154 return (zhp->zpool_name);
1159 * Return the state of the pool (ACTIVE or UNAVAILABLE)
1162 zpool_get_state(zpool_handle_t *zhp)
1164 return (zhp->zpool_state);
1168 * Create the named pool, using the provided vdev list. It is assumed
1169 * that the consumer has already validated the contents of the nvlist, so we
1170 * don't have to worry about error semantics.
1173 zpool_create(libzfs_handle_t *hdl, const char *pool, nvlist_t *nvroot,
1174 nvlist_t *props, nvlist_t *fsprops)
1176 zfs_cmd_t zc = {"\0"};
1177 nvlist_t *zc_fsprops = NULL;
1178 nvlist_t *zc_props = NULL;
1182 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN,
1183 "cannot create '%s'"), pool);
1185 if (!zpool_name_valid(hdl, B_FALSE, pool))
1186 return (zfs_error(hdl, EZFS_INVALIDNAME, msg));
1188 if (zcmd_write_conf_nvlist(hdl, &zc, nvroot) != 0)
1192 prop_flags_t flags = { .create = B_TRUE, .import = B_FALSE };
1194 if ((zc_props = zpool_valid_proplist(hdl, pool, props,
1195 SPA_VERSION_1, flags, msg)) == NULL) {
1204 zoned = ((nvlist_lookup_string(fsprops,
1205 zfs_prop_to_name(ZFS_PROP_ZONED), &zonestr) == 0) &&
1206 strcmp(zonestr, "on") == 0);
1208 if ((zc_fsprops = zfs_valid_proplist(hdl, ZFS_TYPE_FILESYSTEM,
1209 fsprops, zoned, NULL, NULL, msg)) == NULL) {
1213 (nvlist_alloc(&zc_props, NV_UNIQUE_NAME, 0) != 0)) {
1216 if (nvlist_add_nvlist(zc_props,
1217 ZPOOL_ROOTFS_PROPS, zc_fsprops) != 0) {
1222 if (zc_props && zcmd_write_src_nvlist(hdl, &zc, zc_props) != 0)
1225 (void) strlcpy(zc.zc_name, pool, sizeof (zc.zc_name));
1227 if ((ret = zfs_ioctl(hdl, ZFS_IOC_POOL_CREATE, &zc)) != 0) {
1229 zcmd_free_nvlists(&zc);
1230 nvlist_free(zc_props);
1231 nvlist_free(zc_fsprops);
1236 * This can happen if the user has specified the same
1237 * device multiple times. We can't reliably detect this
1238 * until we try to add it and see we already have a
1239 * label. This can also happen under if the device is
1240 * part of an active md or lvm device.
1242 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1243 "one or more vdevs refer to the same device, or "
1244 "one of\nthe devices is part of an active md or "
1246 return (zfs_error(hdl, EZFS_BADDEV, msg));
1250 * This happens if the record size is smaller or larger
1251 * than the allowed size range, or not a power of 2.
1253 * NOTE: although zfs_valid_proplist is called earlier,
1254 * this case may have slipped through since the
1255 * pool does not exist yet and it is therefore
1256 * impossible to read properties e.g. max blocksize
1259 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1260 "record size invalid"));
1261 return (zfs_error(hdl, EZFS_BADPROP, msg));
1265 * This occurs when one of the devices is below
1266 * SPA_MINDEVSIZE. Unfortunately, we can't detect which
1267 * device was the problem device since there's no
1268 * reliable way to determine device size from userland.
1273 zfs_nicenum(SPA_MINDEVSIZE, buf, sizeof (buf));
1275 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1276 "one or more devices is less than the "
1277 "minimum size (%s)"), buf);
1279 return (zfs_error(hdl, EZFS_BADDEV, msg));
1282 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1283 "one or more devices is out of space"));
1284 return (zfs_error(hdl, EZFS_BADDEV, msg));
1287 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1288 "cache device must be a disk or disk slice"));
1289 return (zfs_error(hdl, EZFS_BADDEV, msg));
1292 return (zpool_standard_error(hdl, errno, msg));
1297 zcmd_free_nvlists(&zc);
1298 nvlist_free(zc_props);
1299 nvlist_free(zc_fsprops);
1304 * Destroy the given pool. It is up to the caller to ensure that there are no
1305 * datasets left in the pool.
1308 zpool_destroy(zpool_handle_t *zhp, const char *log_str)
1310 zfs_cmd_t zc = {"\0"};
1311 zfs_handle_t *zfp = NULL;
1312 libzfs_handle_t *hdl = zhp->zpool_hdl;
1315 if (zhp->zpool_state == POOL_STATE_ACTIVE &&
1316 (zfp = zfs_open(hdl, zhp->zpool_name, ZFS_TYPE_FILESYSTEM)) == NULL)
1319 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
1320 zc.zc_history = (uint64_t)(uintptr_t)log_str;
1322 if (zfs_ioctl(hdl, ZFS_IOC_POOL_DESTROY, &zc) != 0) {
1323 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN,
1324 "cannot destroy '%s'"), zhp->zpool_name);
1326 if (errno == EROFS) {
1327 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1328 "one or more devices is read only"));
1329 (void) zfs_error(hdl, EZFS_BADDEV, msg);
1331 (void) zpool_standard_error(hdl, errno, msg);
1340 remove_mountpoint(zfp);
1348 * Add the given vdevs to the pool. The caller must have already performed the
1349 * necessary verification to ensure that the vdev specification is well-formed.
1352 zpool_add(zpool_handle_t *zhp, nvlist_t *nvroot)
1354 zfs_cmd_t zc = {"\0"};
1356 libzfs_handle_t *hdl = zhp->zpool_hdl;
1358 nvlist_t **spares, **l2cache;
1359 uint_t nspares, nl2cache;
1361 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN,
1362 "cannot add to '%s'"), zhp->zpool_name);
1364 if (zpool_get_prop_int(zhp, ZPOOL_PROP_VERSION, NULL) <
1365 SPA_VERSION_SPARES &&
1366 nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_SPARES,
1367 &spares, &nspares) == 0) {
1368 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "pool must be "
1369 "upgraded to add hot spares"));
1370 return (zfs_error(hdl, EZFS_BADVERSION, msg));
1373 #if defined(__sun__) || defined(__sun)
1374 if (zpool_is_bootable(zhp) && nvlist_lookup_nvlist_array(nvroot,
1375 ZPOOL_CONFIG_SPARES, &spares, &nspares) == 0) {
1378 for (s = 0; s < nspares; s++) {
1381 if (nvlist_lookup_string(spares[s], ZPOOL_CONFIG_PATH,
1382 &path) == 0 && pool_uses_efi(spares[s])) {
1383 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1384 "device '%s' contains an EFI label and "
1385 "cannot be used on root pools."),
1386 zpool_vdev_name(hdl, NULL, spares[s], 0));
1387 return (zfs_error(hdl, EZFS_POOL_NOTSUP, msg));
1393 if (zpool_get_prop_int(zhp, ZPOOL_PROP_VERSION, NULL) <
1394 SPA_VERSION_L2CACHE &&
1395 nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_L2CACHE,
1396 &l2cache, &nl2cache) == 0) {
1397 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "pool must be "
1398 "upgraded to add cache devices"));
1399 return (zfs_error(hdl, EZFS_BADVERSION, msg));
1402 if (zcmd_write_conf_nvlist(hdl, &zc, nvroot) != 0)
1404 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
1406 if (zfs_ioctl(hdl, ZFS_IOC_VDEV_ADD, &zc) != 0) {
1410 * This can happen if the user has specified the same
1411 * device multiple times. We can't reliably detect this
1412 * until we try to add it and see we already have a
1415 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1416 "one or more vdevs refer to the same device"));
1417 (void) zfs_error(hdl, EZFS_BADDEV, msg);
1422 * This occurrs when one of the devices is below
1423 * SPA_MINDEVSIZE. Unfortunately, we can't detect which
1424 * device was the problem device since there's no
1425 * reliable way to determine device size from userland.
1430 zfs_nicenum(SPA_MINDEVSIZE, buf, sizeof (buf));
1432 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1433 "device is less than the minimum "
1436 (void) zfs_error(hdl, EZFS_BADDEV, msg);
1440 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1441 "pool must be upgraded to add these vdevs"));
1442 (void) zfs_error(hdl, EZFS_BADVERSION, msg);
1446 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1447 "cache device must be a disk or disk slice"));
1448 (void) zfs_error(hdl, EZFS_BADDEV, msg);
1452 (void) zpool_standard_error(hdl, errno, msg);
1460 zcmd_free_nvlists(&zc);
1466 * Exports the pool from the system. The caller must ensure that there are no
1467 * mounted datasets in the pool.
1470 zpool_export_common(zpool_handle_t *zhp, boolean_t force, boolean_t hardforce,
1471 const char *log_str)
1473 zfs_cmd_t zc = {"\0"};
1476 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN,
1477 "cannot export '%s'"), zhp->zpool_name);
1479 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
1480 zc.zc_cookie = force;
1481 zc.zc_guid = hardforce;
1482 zc.zc_history = (uint64_t)(uintptr_t)log_str;
1484 if (zfs_ioctl(zhp->zpool_hdl, ZFS_IOC_POOL_EXPORT, &zc) != 0) {
1487 zfs_error_aux(zhp->zpool_hdl, dgettext(TEXT_DOMAIN,
1488 "use '-f' to override the following errors:\n"
1489 "'%s' has an active shared spare which could be"
1490 " used by other pools once '%s' is exported."),
1491 zhp->zpool_name, zhp->zpool_name);
1492 return (zfs_error(zhp->zpool_hdl, EZFS_ACTIVE_SPARE,
1495 return (zpool_standard_error_fmt(zhp->zpool_hdl, errno,
1504 zpool_export(zpool_handle_t *zhp, boolean_t force, const char *log_str)
1506 return (zpool_export_common(zhp, force, B_FALSE, log_str));
1510 zpool_export_force(zpool_handle_t *zhp, const char *log_str)
1512 return (zpool_export_common(zhp, B_TRUE, B_TRUE, log_str));
1516 zpool_rewind_exclaim(libzfs_handle_t *hdl, const char *name, boolean_t dryrun,
1519 nvlist_t *nv = NULL;
1525 if (!hdl->libzfs_printerr || config == NULL)
1528 if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_LOAD_INFO, &nv) != 0 ||
1529 nvlist_lookup_nvlist(nv, ZPOOL_CONFIG_REWIND_INFO, &nv) != 0) {
1533 if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_LOAD_TIME, &rewindto) != 0)
1535 (void) nvlist_lookup_int64(nv, ZPOOL_CONFIG_REWIND_TIME, &loss);
1537 if (localtime_r((time_t *)&rewindto, &t) != NULL &&
1538 strftime(timestr, 128, "%c", &t) != 0) {
1540 (void) printf(dgettext(TEXT_DOMAIN,
1541 "Would be able to return %s "
1542 "to its state as of %s.\n"),
1545 (void) printf(dgettext(TEXT_DOMAIN,
1546 "Pool %s returned to its state as of %s.\n"),
1550 (void) printf(dgettext(TEXT_DOMAIN,
1551 "%s approximately %lld "),
1552 dryrun ? "Would discard" : "Discarded",
1553 ((longlong_t)loss + 30) / 60);
1554 (void) printf(dgettext(TEXT_DOMAIN,
1555 "minutes of transactions.\n"));
1556 } else if (loss > 0) {
1557 (void) printf(dgettext(TEXT_DOMAIN,
1558 "%s approximately %lld "),
1559 dryrun ? "Would discard" : "Discarded",
1561 (void) printf(dgettext(TEXT_DOMAIN,
1562 "seconds of transactions.\n"));
1568 zpool_explain_recover(libzfs_handle_t *hdl, const char *name, int reason,
1571 nvlist_t *nv = NULL;
1573 uint64_t edata = UINT64_MAX;
1578 if (!hdl->libzfs_printerr)
1582 (void) printf(dgettext(TEXT_DOMAIN, "action: "));
1584 (void) printf(dgettext(TEXT_DOMAIN, "\t"));
1586 /* All attempted rewinds failed if ZPOOL_CONFIG_LOAD_TIME missing */
1587 if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_LOAD_INFO, &nv) != 0 ||
1588 nvlist_lookup_nvlist(nv, ZPOOL_CONFIG_REWIND_INFO, &nv) != 0 ||
1589 nvlist_lookup_uint64(nv, ZPOOL_CONFIG_LOAD_TIME, &rewindto) != 0)
1592 (void) nvlist_lookup_int64(nv, ZPOOL_CONFIG_REWIND_TIME, &loss);
1593 (void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_LOAD_DATA_ERRORS,
1596 (void) printf(dgettext(TEXT_DOMAIN,
1597 "Recovery is possible, but will result in some data loss.\n"));
1599 if (localtime_r((time_t *)&rewindto, &t) != NULL &&
1600 strftime(timestr, 128, "%c", &t) != 0) {
1601 (void) printf(dgettext(TEXT_DOMAIN,
1602 "\tReturning the pool to its state as of %s\n"
1603 "\tshould correct the problem. "),
1606 (void) printf(dgettext(TEXT_DOMAIN,
1607 "\tReverting the pool to an earlier state "
1608 "should correct the problem.\n\t"));
1612 (void) printf(dgettext(TEXT_DOMAIN,
1613 "Approximately %lld minutes of data\n"
1614 "\tmust be discarded, irreversibly. "),
1615 ((longlong_t)loss + 30) / 60);
1616 } else if (loss > 0) {
1617 (void) printf(dgettext(TEXT_DOMAIN,
1618 "Approximately %lld seconds of data\n"
1619 "\tmust be discarded, irreversibly. "),
1622 if (edata != 0 && edata != UINT64_MAX) {
1624 (void) printf(dgettext(TEXT_DOMAIN,
1625 "After rewind, at least\n"
1626 "\tone persistent user-data error will remain. "));
1628 (void) printf(dgettext(TEXT_DOMAIN,
1629 "After rewind, several\n"
1630 "\tpersistent user-data errors will remain. "));
1633 (void) printf(dgettext(TEXT_DOMAIN,
1634 "Recovery can be attempted\n\tby executing 'zpool %s -F %s'. "),
1635 reason >= 0 ? "clear" : "import", name);
1637 (void) printf(dgettext(TEXT_DOMAIN,
1638 "A scrub of the pool\n"
1639 "\tis strongly recommended after recovery.\n"));
1643 (void) printf(dgettext(TEXT_DOMAIN,
1644 "Destroy and re-create the pool from\n\ta backup source.\n"));
1648 * zpool_import() is a contracted interface. Should be kept the same
1651 * Applications should use zpool_import_props() to import a pool with
1652 * new properties value to be set.
1655 zpool_import(libzfs_handle_t *hdl, nvlist_t *config, const char *newname,
1658 nvlist_t *props = NULL;
1661 if (altroot != NULL) {
1662 if (nvlist_alloc(&props, NV_UNIQUE_NAME, 0) != 0) {
1663 return (zfs_error_fmt(hdl, EZFS_NOMEM,
1664 dgettext(TEXT_DOMAIN, "cannot import '%s'"),
1668 if (nvlist_add_string(props,
1669 zpool_prop_to_name(ZPOOL_PROP_ALTROOT), altroot) != 0 ||
1670 nvlist_add_string(props,
1671 zpool_prop_to_name(ZPOOL_PROP_CACHEFILE), "none") != 0) {
1673 return (zfs_error_fmt(hdl, EZFS_NOMEM,
1674 dgettext(TEXT_DOMAIN, "cannot import '%s'"),
1679 ret = zpool_import_props(hdl, config, newname, props,
1686 print_vdev_tree(libzfs_handle_t *hdl, const char *name, nvlist_t *nv,
1692 uint64_t is_log = 0;
1694 (void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_IS_LOG,
1698 (void) printf("\t%*s%s%s\n", indent, "", name,
1699 is_log ? " [log]" : "");
1701 if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN,
1702 &child, &children) != 0)
1705 for (c = 0; c < children; c++) {
1706 vname = zpool_vdev_name(hdl, NULL, child[c], VDEV_NAME_TYPE_ID);
1707 print_vdev_tree(hdl, vname, child[c], indent + 2);
1713 zpool_print_unsup_feat(nvlist_t *config)
1715 nvlist_t *nvinfo, *unsup_feat;
1718 verify(nvlist_lookup_nvlist(config, ZPOOL_CONFIG_LOAD_INFO, &nvinfo) ==
1720 verify(nvlist_lookup_nvlist(nvinfo, ZPOOL_CONFIG_UNSUP_FEAT,
1723 for (nvp = nvlist_next_nvpair(unsup_feat, NULL); nvp != NULL;
1724 nvp = nvlist_next_nvpair(unsup_feat, nvp)) {
1727 verify(nvpair_type(nvp) == DATA_TYPE_STRING);
1728 verify(nvpair_value_string(nvp, &desc) == 0);
1730 if (strlen(desc) > 0)
1731 (void) printf("\t%s (%s)\n", nvpair_name(nvp), desc);
1733 (void) printf("\t%s\n", nvpair_name(nvp));
1738 * Import the given pool using the known configuration and a list of
1739 * properties to be set. The configuration should have come from
1740 * zpool_find_import(). The 'newname' parameters control whether the pool
1741 * is imported with a different name.
1744 zpool_import_props(libzfs_handle_t *hdl, nvlist_t *config, const char *newname,
1745 nvlist_t *props, int flags)
1747 zfs_cmd_t zc = {"\0"};
1748 zpool_rewind_policy_t policy;
1749 nvlist_t *nv = NULL;
1750 nvlist_t *nvinfo = NULL;
1751 nvlist_t *missing = NULL;
1758 verify(nvlist_lookup_string(config, ZPOOL_CONFIG_POOL_NAME,
1761 (void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN,
1762 "cannot import pool '%s'"), origname);
1764 if (newname != NULL) {
1765 if (!zpool_name_valid(hdl, B_FALSE, newname))
1766 return (zfs_error_fmt(hdl, EZFS_INVALIDNAME,
1767 dgettext(TEXT_DOMAIN, "cannot import '%s'"),
1769 thename = (char *)newname;
1774 if (props != NULL) {
1776 prop_flags_t flags = { .create = B_FALSE, .import = B_TRUE };
1778 verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_VERSION,
1781 if ((props = zpool_valid_proplist(hdl, origname,
1782 props, version, flags, errbuf)) == NULL)
1784 if (zcmd_write_src_nvlist(hdl, &zc, props) != 0) {
1791 (void) strlcpy(zc.zc_name, thename, sizeof (zc.zc_name));
1793 verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID,
1796 if (zcmd_write_conf_nvlist(hdl, &zc, config) != 0) {
1797 zcmd_free_nvlists(&zc);
1800 if (zcmd_alloc_dst_nvlist(hdl, &zc, zc.zc_nvlist_conf_size * 2) != 0) {
1801 zcmd_free_nvlists(&zc);
1805 zc.zc_cookie = flags;
1806 while ((ret = zfs_ioctl(hdl, ZFS_IOC_POOL_IMPORT, &zc)) != 0 &&
1808 if (zcmd_expand_dst_nvlist(hdl, &zc) != 0) {
1809 zcmd_free_nvlists(&zc);
1816 (void) zcmd_read_dst_nvlist(hdl, &zc, &nv);
1818 zcmd_free_nvlists(&zc);
1820 zpool_get_rewind_policy(config, &policy);
1826 * Dry-run failed, but we print out what success
1827 * looks like if we found a best txg
1829 if (policy.zrp_request & ZPOOL_TRY_REWIND) {
1830 zpool_rewind_exclaim(hdl, newname ? origname : thename,
1836 if (newname == NULL)
1837 (void) snprintf(desc, sizeof (desc),
1838 dgettext(TEXT_DOMAIN, "cannot import '%s'"),
1841 (void) snprintf(desc, sizeof (desc),
1842 dgettext(TEXT_DOMAIN, "cannot import '%s' as '%s'"),
1847 if (nv != NULL && nvlist_lookup_nvlist(nv,
1848 ZPOOL_CONFIG_LOAD_INFO, &nvinfo) == 0 &&
1849 nvlist_exists(nvinfo, ZPOOL_CONFIG_UNSUP_FEAT)) {
1850 (void) printf(dgettext(TEXT_DOMAIN, "This "
1851 "pool uses the following feature(s) not "
1852 "supported by this system:\n"));
1853 zpool_print_unsup_feat(nv);
1854 if (nvlist_exists(nvinfo,
1855 ZPOOL_CONFIG_CAN_RDONLY)) {
1856 (void) printf(dgettext(TEXT_DOMAIN,
1857 "All unsupported features are only "
1858 "required for writing to the pool."
1859 "\nThe pool can be imported using "
1860 "'-o readonly=on'.\n"));
1864 * Unsupported version.
1866 (void) zfs_error(hdl, EZFS_BADVERSION, desc);
1870 (void) zfs_error(hdl, EZFS_INVALCONFIG, desc);
1874 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1875 "one or more devices is read only"));
1876 (void) zfs_error(hdl, EZFS_BADDEV, desc);
1880 if (nv && nvlist_lookup_nvlist(nv,
1881 ZPOOL_CONFIG_LOAD_INFO, &nvinfo) == 0 &&
1882 nvlist_lookup_nvlist(nvinfo,
1883 ZPOOL_CONFIG_MISSING_DEVICES, &missing) == 0) {
1884 (void) printf(dgettext(TEXT_DOMAIN,
1885 "The devices below are missing, use "
1886 "'-m' to import the pool anyway:\n"));
1887 print_vdev_tree(hdl, NULL, missing, 2);
1888 (void) printf("\n");
1890 (void) zpool_standard_error(hdl, error, desc);
1894 (void) zpool_standard_error(hdl, error, desc);
1898 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1899 "one or more devices are already in use\n"));
1900 (void) zfs_error(hdl, EZFS_BADDEV, desc);
1904 (void) zpool_standard_error(hdl, error, desc);
1905 zpool_explain_recover(hdl,
1906 newname ? origname : thename, -error, nv);
1913 zpool_handle_t *zhp;
1916 * This should never fail, but play it safe anyway.
1918 if (zpool_open_silent(hdl, thename, &zhp) != 0)
1920 else if (zhp != NULL)
1922 if (policy.zrp_request &
1923 (ZPOOL_DO_REWIND | ZPOOL_TRY_REWIND)) {
1924 zpool_rewind_exclaim(hdl, newname ? origname : thename,
1925 ((policy.zrp_request & ZPOOL_TRY_REWIND) != 0), nv);
1938 zpool_scan(zpool_handle_t *zhp, pool_scan_func_t func)
1940 zfs_cmd_t zc = {"\0"};
1942 libzfs_handle_t *hdl = zhp->zpool_hdl;
1944 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
1945 zc.zc_cookie = func;
1947 if (zfs_ioctl(hdl, ZFS_IOC_POOL_SCAN, &zc) == 0 ||
1948 (errno == ENOENT && func != POOL_SCAN_NONE))
1951 if (func == POOL_SCAN_SCRUB) {
1952 (void) snprintf(msg, sizeof (msg),
1953 dgettext(TEXT_DOMAIN, "cannot scrub %s"), zc.zc_name);
1954 } else if (func == POOL_SCAN_NONE) {
1955 (void) snprintf(msg, sizeof (msg),
1956 dgettext(TEXT_DOMAIN, "cannot cancel scrubbing %s"),
1959 assert(!"unexpected result");
1962 if (errno == EBUSY) {
1964 pool_scan_stat_t *ps = NULL;
1967 verify(nvlist_lookup_nvlist(zhp->zpool_config,
1968 ZPOOL_CONFIG_VDEV_TREE, &nvroot) == 0);
1969 (void) nvlist_lookup_uint64_array(nvroot,
1970 ZPOOL_CONFIG_SCAN_STATS, (uint64_t **)&ps, &psc);
1971 if (ps && ps->pss_func == POOL_SCAN_SCRUB)
1972 return (zfs_error(hdl, EZFS_SCRUBBING, msg));
1974 return (zfs_error(hdl, EZFS_RESILVERING, msg));
1975 } else if (errno == ENOENT) {
1976 return (zfs_error(hdl, EZFS_NO_SCRUB, msg));
1978 return (zpool_standard_error(hdl, errno, msg));
1983 * Find a vdev that matches the search criteria specified. We use the
1984 * the nvpair name to determine how we should look for the device.
1985 * 'avail_spare' is set to TRUE if the provided guid refers to an AVAIL
1986 * spare; but FALSE if its an INUSE spare.
1989 vdev_to_nvlist_iter(nvlist_t *nv, nvlist_t *search, boolean_t *avail_spare,
1990 boolean_t *l2cache, boolean_t *log)
1997 nvpair_t *pair = nvlist_next_nvpair(search, NULL);
1999 /* Nothing to look for */
2000 if (search == NULL || pair == NULL)
2003 /* Obtain the key we will use to search */
2004 srchkey = nvpair_name(pair);
2006 switch (nvpair_type(pair)) {
2007 case DATA_TYPE_UINT64:
2008 if (strcmp(srchkey, ZPOOL_CONFIG_GUID) == 0) {
2009 uint64_t srchval, theguid;
2011 verify(nvpair_value_uint64(pair, &srchval) == 0);
2012 verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID,
2014 if (theguid == srchval)
2019 case DATA_TYPE_STRING: {
2020 char *srchval, *val;
2022 verify(nvpair_value_string(pair, &srchval) == 0);
2023 if (nvlist_lookup_string(nv, srchkey, &val) != 0)
2027 * Search for the requested value. Special cases:
2029 * - ZPOOL_CONFIG_PATH for whole disk entries. These end in
2030 * "-part1", or "p1". The suffix is hidden from the user,
2031 * but included in the string, so this matches around it.
2032 * - ZPOOL_CONFIG_PATH for short names zfs_strcmp_shortname()
2033 * is used to check all possible expanded paths.
2034 * - looking for a top-level vdev name (i.e. ZPOOL_CONFIG_TYPE).
2036 * Otherwise, all other searches are simple string compares.
2038 if (strcmp(srchkey, ZPOOL_CONFIG_PATH) == 0) {
2039 uint64_t wholedisk = 0;
2041 (void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_WHOLE_DISK,
2043 if (zfs_strcmp_pathname(srchval, val, wholedisk) == 0)
2046 } else if (strcmp(srchkey, ZPOOL_CONFIG_TYPE) == 0 && val) {
2047 char *type, *idx, *end, *p;
2048 uint64_t id, vdev_id;
2051 * Determine our vdev type, keeping in mind
2052 * that the srchval is composed of a type and
2053 * vdev id pair (i.e. mirror-4).
2055 if ((type = strdup(srchval)) == NULL)
2058 if ((p = strrchr(type, '-')) == NULL) {
2066 * If the types don't match then keep looking.
2068 if (strncmp(val, type, strlen(val)) != 0) {
2073 verify(strncmp(type, VDEV_TYPE_RAIDZ,
2074 strlen(VDEV_TYPE_RAIDZ)) == 0 ||
2075 strncmp(type, VDEV_TYPE_MIRROR,
2076 strlen(VDEV_TYPE_MIRROR)) == 0);
2077 verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_ID,
2081 vdev_id = strtoull(idx, &end, 10);
2088 * Now verify that we have the correct vdev id.
2097 if (strcmp(srchval, val) == 0)
2106 if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN,
2107 &child, &children) != 0)
2110 for (c = 0; c < children; c++) {
2111 if ((ret = vdev_to_nvlist_iter(child[c], search,
2112 avail_spare, l2cache, NULL)) != NULL) {
2114 * The 'is_log' value is only set for the toplevel
2115 * vdev, not the leaf vdevs. So we always lookup the
2116 * log device from the root of the vdev tree (where
2117 * 'log' is non-NULL).
2120 nvlist_lookup_uint64(child[c],
2121 ZPOOL_CONFIG_IS_LOG, &is_log) == 0 &&
2129 if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_SPARES,
2130 &child, &children) == 0) {
2131 for (c = 0; c < children; c++) {
2132 if ((ret = vdev_to_nvlist_iter(child[c], search,
2133 avail_spare, l2cache, NULL)) != NULL) {
2134 *avail_spare = B_TRUE;
2140 if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_L2CACHE,
2141 &child, &children) == 0) {
2142 for (c = 0; c < children; c++) {
2143 if ((ret = vdev_to_nvlist_iter(child[c], search,
2144 avail_spare, l2cache, NULL)) != NULL) {
2155 * Given a physical path (minus the "/devices" prefix), find the
2159 zpool_find_vdev_by_physpath(zpool_handle_t *zhp, const char *ppath,
2160 boolean_t *avail_spare, boolean_t *l2cache, boolean_t *log)
2162 nvlist_t *search, *nvroot, *ret;
2164 verify(nvlist_alloc(&search, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2165 verify(nvlist_add_string(search, ZPOOL_CONFIG_PHYS_PATH, ppath) == 0);
2167 verify(nvlist_lookup_nvlist(zhp->zpool_config, ZPOOL_CONFIG_VDEV_TREE,
2170 *avail_spare = B_FALSE;
2174 ret = vdev_to_nvlist_iter(nvroot, search, avail_spare, l2cache, log);
2175 nvlist_free(search);
2181 * Determine if we have an "interior" top-level vdev (i.e mirror/raidz).
2184 zpool_vdev_is_interior(const char *name)
2186 if (strncmp(name, VDEV_TYPE_RAIDZ, strlen(VDEV_TYPE_RAIDZ)) == 0 ||
2187 strncmp(name, VDEV_TYPE_MIRROR, strlen(VDEV_TYPE_MIRROR)) == 0)
2193 zpool_find_vdev(zpool_handle_t *zhp, const char *path, boolean_t *avail_spare,
2194 boolean_t *l2cache, boolean_t *log)
2197 nvlist_t *nvroot, *search, *ret;
2200 verify(nvlist_alloc(&search, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2202 guid = strtoull(path, &end, 0);
2203 if (guid != 0 && *end == '\0') {
2204 verify(nvlist_add_uint64(search, ZPOOL_CONFIG_GUID, guid) == 0);
2205 } else if (zpool_vdev_is_interior(path)) {
2206 verify(nvlist_add_string(search, ZPOOL_CONFIG_TYPE, path) == 0);
2208 verify(nvlist_add_string(search, ZPOOL_CONFIG_PATH, path) == 0);
2211 verify(nvlist_lookup_nvlist(zhp->zpool_config, ZPOOL_CONFIG_VDEV_TREE,
2214 *avail_spare = B_FALSE;
2218 ret = vdev_to_nvlist_iter(nvroot, search, avail_spare, l2cache, log);
2219 nvlist_free(search);
2225 vdev_online(nvlist_t *nv)
2229 if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_OFFLINE, &ival) == 0 ||
2230 nvlist_lookup_uint64(nv, ZPOOL_CONFIG_FAULTED, &ival) == 0 ||
2231 nvlist_lookup_uint64(nv, ZPOOL_CONFIG_REMOVED, &ival) == 0)
2238 * Helper function for zpool_get_physpaths().
2241 vdev_get_one_physpath(nvlist_t *config, char *physpath, size_t physpath_size,
2242 size_t *bytes_written)
2244 size_t bytes_left, pos, rsz;
2248 if (nvlist_lookup_string(config, ZPOOL_CONFIG_PHYS_PATH,
2250 return (EZFS_NODEVICE);
2252 pos = *bytes_written;
2253 bytes_left = physpath_size - pos;
2254 format = (pos == 0) ? "%s" : " %s";
2256 rsz = snprintf(physpath + pos, bytes_left, format, tmppath);
2257 *bytes_written += rsz;
2259 if (rsz >= bytes_left) {
2260 /* if physpath was not copied properly, clear it */
2261 if (bytes_left != 0) {
2264 return (EZFS_NOSPC);
2270 vdev_get_physpaths(nvlist_t *nv, char *physpath, size_t phypath_size,
2271 size_t *rsz, boolean_t is_spare)
2276 if (nvlist_lookup_string(nv, ZPOOL_CONFIG_TYPE, &type) != 0)
2277 return (EZFS_INVALCONFIG);
2279 if (strcmp(type, VDEV_TYPE_DISK) == 0) {
2281 * An active spare device has ZPOOL_CONFIG_IS_SPARE set.
2282 * For a spare vdev, we only want to boot from the active
2287 (void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_IS_SPARE,
2290 return (EZFS_INVALCONFIG);
2293 if (vdev_online(nv)) {
2294 if ((ret = vdev_get_one_physpath(nv, physpath,
2295 phypath_size, rsz)) != 0)
2298 } else if (strcmp(type, VDEV_TYPE_MIRROR) == 0 ||
2299 strcmp(type, VDEV_TYPE_REPLACING) == 0 ||
2300 (is_spare = (strcmp(type, VDEV_TYPE_SPARE) == 0))) {
2305 if (nvlist_lookup_nvlist_array(nv,
2306 ZPOOL_CONFIG_CHILDREN, &child, &count) != 0)
2307 return (EZFS_INVALCONFIG);
2309 for (i = 0; i < count; i++) {
2310 ret = vdev_get_physpaths(child[i], physpath,
2311 phypath_size, rsz, is_spare);
2312 if (ret == EZFS_NOSPC)
2317 return (EZFS_POOL_INVALARG);
2321 * Get phys_path for a root pool config.
2322 * Return 0 on success; non-zero on failure.
2325 zpool_get_config_physpath(nvlist_t *config, char *physpath, size_t phypath_size)
2328 nvlist_t *vdev_root;
2335 if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE,
2337 return (EZFS_INVALCONFIG);
2339 if (nvlist_lookup_string(vdev_root, ZPOOL_CONFIG_TYPE, &type) != 0 ||
2340 nvlist_lookup_nvlist_array(vdev_root, ZPOOL_CONFIG_CHILDREN,
2341 &child, &count) != 0)
2342 return (EZFS_INVALCONFIG);
2344 #if defined(__sun__) || defined(__sun)
2346 * root pool can not have EFI labeled disks and can only have
2347 * a single top-level vdev.
2349 if (strcmp(type, VDEV_TYPE_ROOT) != 0 || count != 1 ||
2350 pool_uses_efi(vdev_root))
2351 return (EZFS_POOL_INVALARG);
2354 (void) vdev_get_physpaths(child[0], physpath, phypath_size, &rsz,
2357 /* No online devices */
2359 return (EZFS_NODEVICE);
2365 * Get phys_path for a root pool
2366 * Return 0 on success; non-zero on failure.
2369 zpool_get_physpath(zpool_handle_t *zhp, char *physpath, size_t phypath_size)
2371 return (zpool_get_config_physpath(zhp->zpool_config, physpath,
2376 * If the device has being dynamically expanded then we need to relabel
2377 * the disk to use the new unallocated space.
2380 zpool_relabel_disk(libzfs_handle_t *hdl, const char *path, const char *msg)
2384 if ((fd = open(path, O_RDWR|O_DIRECT)) < 0) {
2385 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "cannot "
2386 "relabel '%s': unable to open device: %d"), path, errno);
2387 return (zfs_error(hdl, EZFS_OPENFAILED, msg));
2391 * It's possible that we might encounter an error if the device
2392 * does not have any unallocated space left. If so, we simply
2393 * ignore that error and continue on.
2395 * Also, we don't call efi_rescan() - that would just return EBUSY.
2396 * The module will do it for us in vdev_disk_open().
2398 error = efi_use_whole_disk(fd);
2400 if (error && error != VT_ENOSPC) {
2401 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "cannot "
2402 "relabel '%s': unable to read disk capacity"), path);
2403 return (zfs_error(hdl, EZFS_NOCAP, msg));
2409 * Bring the specified vdev online. The 'flags' parameter is a set of the
2410 * ZFS_ONLINE_* flags.
2413 zpool_vdev_online(zpool_handle_t *zhp, const char *path, int flags,
2414 vdev_state_t *newstate)
2416 zfs_cmd_t zc = {"\0"};
2419 boolean_t avail_spare, l2cache, islog;
2420 libzfs_handle_t *hdl = zhp->zpool_hdl;
2423 if (flags & ZFS_ONLINE_EXPAND) {
2424 (void) snprintf(msg, sizeof (msg),
2425 dgettext(TEXT_DOMAIN, "cannot expand %s"), path);
2427 (void) snprintf(msg, sizeof (msg),
2428 dgettext(TEXT_DOMAIN, "cannot online %s"), path);
2431 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
2432 if ((tgt = zpool_find_vdev(zhp, path, &avail_spare, &l2cache,
2434 return (zfs_error(hdl, EZFS_NODEVICE, msg));
2436 verify(nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID, &zc.zc_guid) == 0);
2439 return (zfs_error(hdl, EZFS_ISSPARE, msg));
2441 if (flags & ZFS_ONLINE_EXPAND ||
2442 zpool_get_prop_int(zhp, ZPOOL_PROP_AUTOEXPAND, NULL)) {
2443 uint64_t wholedisk = 0;
2445 (void) nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_WHOLE_DISK,
2449 * XXX - L2ARC 1.0 devices can't support expansion.
2452 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2453 "cannot expand cache devices"));
2454 return (zfs_error(hdl, EZFS_VDEVNOTSUP, msg));
2458 const char *fullpath = path;
2459 char buf[MAXPATHLEN];
2461 if (path[0] != '/') {
2462 error = zfs_resolve_shortname(path, buf,
2465 return (zfs_error(hdl, EZFS_NODEVICE,
2471 error = zpool_relabel_disk(hdl, fullpath, msg);
2477 zc.zc_cookie = VDEV_STATE_ONLINE;
2480 if (zfs_ioctl(hdl, ZFS_IOC_VDEV_SET_STATE, &zc) != 0) {
2481 if (errno == EINVAL) {
2482 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "was split "
2483 "from this pool into a new one. Use '%s' "
2484 "instead"), "zpool detach");
2485 return (zfs_error(hdl, EZFS_POSTSPLIT_ONLINE, msg));
2487 return (zpool_standard_error(hdl, errno, msg));
2490 *newstate = zc.zc_cookie;
2495 * Take the specified vdev offline
2498 zpool_vdev_offline(zpool_handle_t *zhp, const char *path, boolean_t istmp)
2500 zfs_cmd_t zc = {"\0"};
2503 boolean_t avail_spare, l2cache;
2504 libzfs_handle_t *hdl = zhp->zpool_hdl;
2506 (void) snprintf(msg, sizeof (msg),
2507 dgettext(TEXT_DOMAIN, "cannot offline %s"), path);
2509 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
2510 if ((tgt = zpool_find_vdev(zhp, path, &avail_spare, &l2cache,
2512 return (zfs_error(hdl, EZFS_NODEVICE, msg));
2514 verify(nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID, &zc.zc_guid) == 0);
2517 return (zfs_error(hdl, EZFS_ISSPARE, msg));
2519 zc.zc_cookie = VDEV_STATE_OFFLINE;
2520 zc.zc_obj = istmp ? ZFS_OFFLINE_TEMPORARY : 0;
2522 if (zfs_ioctl(hdl, ZFS_IOC_VDEV_SET_STATE, &zc) == 0)
2529 * There are no other replicas of this device.
2531 return (zfs_error(hdl, EZFS_NOREPLICAS, msg));
2535 * The log device has unplayed logs
2537 return (zfs_error(hdl, EZFS_UNPLAYED_LOGS, msg));
2540 return (zpool_standard_error(hdl, errno, msg));
2545 * Mark the given vdev faulted.
2548 zpool_vdev_fault(zpool_handle_t *zhp, uint64_t guid, vdev_aux_t aux)
2550 zfs_cmd_t zc = {"\0"};
2552 libzfs_handle_t *hdl = zhp->zpool_hdl;
2554 (void) snprintf(msg, sizeof (msg),
2555 dgettext(TEXT_DOMAIN, "cannot fault %llu"), (u_longlong_t)guid);
2557 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
2559 zc.zc_cookie = VDEV_STATE_FAULTED;
2562 if (ioctl(hdl->libzfs_fd, ZFS_IOC_VDEV_SET_STATE, &zc) == 0)
2569 * There are no other replicas of this device.
2571 return (zfs_error(hdl, EZFS_NOREPLICAS, msg));
2574 return (zpool_standard_error(hdl, errno, msg));
2580 * Mark the given vdev degraded.
2583 zpool_vdev_degrade(zpool_handle_t *zhp, uint64_t guid, vdev_aux_t aux)
2585 zfs_cmd_t zc = {"\0"};
2587 libzfs_handle_t *hdl = zhp->zpool_hdl;
2589 (void) snprintf(msg, sizeof (msg),
2590 dgettext(TEXT_DOMAIN, "cannot degrade %llu"), (u_longlong_t)guid);
2592 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
2594 zc.zc_cookie = VDEV_STATE_DEGRADED;
2597 if (ioctl(hdl->libzfs_fd, ZFS_IOC_VDEV_SET_STATE, &zc) == 0)
2600 return (zpool_standard_error(hdl, errno, msg));
2604 * Returns TRUE if the given nvlist is a vdev that was originally swapped in as
2608 is_replacing_spare(nvlist_t *search, nvlist_t *tgt, int which)
2614 if (nvlist_lookup_nvlist_array(search, ZPOOL_CONFIG_CHILDREN, &child,
2616 verify(nvlist_lookup_string(search, ZPOOL_CONFIG_TYPE,
2619 if (strcmp(type, VDEV_TYPE_SPARE) == 0 &&
2620 children == 2 && child[which] == tgt)
2623 for (c = 0; c < children; c++)
2624 if (is_replacing_spare(child[c], tgt, which))
2632 * Attach new_disk (fully described by nvroot) to old_disk.
2633 * If 'replacing' is specified, the new disk will replace the old one.
2636 zpool_vdev_attach(zpool_handle_t *zhp,
2637 const char *old_disk, const char *new_disk, nvlist_t *nvroot, int replacing)
2639 zfs_cmd_t zc = {"\0"};
2643 boolean_t avail_spare, l2cache, islog;
2648 nvlist_t *config_root;
2649 libzfs_handle_t *hdl = zhp->zpool_hdl;
2650 boolean_t rootpool = zpool_is_bootable(zhp);
2653 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN,
2654 "cannot replace %s with %s"), old_disk, new_disk);
2656 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN,
2657 "cannot attach %s to %s"), new_disk, old_disk);
2659 #if defined(__sun__) || defined(__sun)
2661 * If this is a root pool, make sure that we're not attaching an
2662 * EFI labeled device.
2664 if (rootpool && pool_uses_efi(nvroot)) {
2665 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2666 "EFI labeled devices are not supported on root pools."));
2667 return (zfs_error(hdl, EZFS_POOL_NOTSUP, msg));
2671 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
2672 if ((tgt = zpool_find_vdev(zhp, old_disk, &avail_spare, &l2cache,
2674 return (zfs_error(hdl, EZFS_NODEVICE, msg));
2677 return (zfs_error(hdl, EZFS_ISSPARE, msg));
2680 return (zfs_error(hdl, EZFS_ISL2CACHE, msg));
2682 verify(nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID, &zc.zc_guid) == 0);
2683 zc.zc_cookie = replacing;
2685 if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_CHILDREN,
2686 &child, &children) != 0 || children != 1) {
2687 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2688 "new device must be a single disk"));
2689 return (zfs_error(hdl, EZFS_INVALCONFIG, msg));
2692 verify(nvlist_lookup_nvlist(zpool_get_config(zhp, NULL),
2693 ZPOOL_CONFIG_VDEV_TREE, &config_root) == 0);
2695 if ((newname = zpool_vdev_name(NULL, NULL, child[0], 0)) == NULL)
2699 * If the target is a hot spare that has been swapped in, we can only
2700 * replace it with another hot spare.
2703 nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_IS_SPARE, &val) == 0 &&
2704 (zpool_find_vdev(zhp, newname, &avail_spare, &l2cache,
2705 NULL) == NULL || !avail_spare) &&
2706 is_replacing_spare(config_root, tgt, 1)) {
2707 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2708 "can only be replaced by another hot spare"));
2710 return (zfs_error(hdl, EZFS_BADTARGET, msg));
2715 if (zcmd_write_conf_nvlist(hdl, &zc, nvroot) != 0)
2718 ret = zfs_ioctl(hdl, ZFS_IOC_VDEV_ATTACH, &zc);
2720 zcmd_free_nvlists(&zc);
2725 * XXX need a better way to prevent user from
2726 * booting up a half-baked vdev.
2728 (void) fprintf(stderr, dgettext(TEXT_DOMAIN, "Make "
2729 "sure to wait until resilver is done "
2730 "before rebooting.\n"));
2738 * Can't attach to or replace this type of vdev.
2741 uint64_t version = zpool_get_prop_int(zhp,
2742 ZPOOL_PROP_VERSION, NULL);
2745 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2746 "cannot replace a log with a spare"));
2747 else if (version >= SPA_VERSION_MULTI_REPLACE)
2748 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2749 "already in replacing/spare config; wait "
2750 "for completion or use 'zpool detach'"));
2752 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2753 "cannot replace a replacing device"));
2755 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2756 "can only attach to mirrors and top-level "
2759 (void) zfs_error(hdl, EZFS_BADTARGET, msg);
2764 * The new device must be a single disk.
2766 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2767 "new device must be a single disk"));
2768 (void) zfs_error(hdl, EZFS_INVALCONFIG, msg);
2772 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "%s is busy"),
2774 (void) zfs_error(hdl, EZFS_BADDEV, msg);
2779 * The new device is too small.
2781 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2782 "device is too small"));
2783 (void) zfs_error(hdl, EZFS_BADDEV, msg);
2788 * The new device has a different optimal sector size.
2790 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2791 "new device has a different optimal sector size; use the "
2792 "option '-o ashift=N' to override the optimal size"));
2793 (void) zfs_error(hdl, EZFS_BADDEV, msg);
2798 * The resulting top-level vdev spec won't fit in the label.
2800 (void) zfs_error(hdl, EZFS_DEVOVERFLOW, msg);
2804 (void) zpool_standard_error(hdl, errno, msg);
2811 * Detach the specified device.
2814 zpool_vdev_detach(zpool_handle_t *zhp, const char *path)
2816 zfs_cmd_t zc = {"\0"};
2819 boolean_t avail_spare, l2cache;
2820 libzfs_handle_t *hdl = zhp->zpool_hdl;
2822 (void) snprintf(msg, sizeof (msg),
2823 dgettext(TEXT_DOMAIN, "cannot detach %s"), path);
2825 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
2826 if ((tgt = zpool_find_vdev(zhp, path, &avail_spare, &l2cache,
2828 return (zfs_error(hdl, EZFS_NODEVICE, msg));
2831 return (zfs_error(hdl, EZFS_ISSPARE, msg));
2834 return (zfs_error(hdl, EZFS_ISL2CACHE, msg));
2836 verify(nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID, &zc.zc_guid) == 0);
2838 if (zfs_ioctl(hdl, ZFS_IOC_VDEV_DETACH, &zc) == 0)
2845 * Can't detach from this type of vdev.
2847 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "only "
2848 "applicable to mirror and replacing vdevs"));
2849 (void) zfs_error(hdl, EZFS_BADTARGET, msg);
2854 * There are no other replicas of this device.
2856 (void) zfs_error(hdl, EZFS_NOREPLICAS, msg);
2860 (void) zpool_standard_error(hdl, errno, msg);
2867 * Find a mirror vdev in the source nvlist.
2869 * The mchild array contains a list of disks in one of the top-level mirrors
2870 * of the source pool. The schild array contains a list of disks that the
2871 * user specified on the command line. We loop over the mchild array to
2872 * see if any entry in the schild array matches.
2874 * If a disk in the mchild array is found in the schild array, we return
2875 * the index of that entry. Otherwise we return -1.
2878 find_vdev_entry(zpool_handle_t *zhp, nvlist_t **mchild, uint_t mchildren,
2879 nvlist_t **schild, uint_t schildren)
2883 for (mc = 0; mc < mchildren; mc++) {
2885 char *mpath = zpool_vdev_name(zhp->zpool_hdl, zhp,
2888 for (sc = 0; sc < schildren; sc++) {
2889 char *spath = zpool_vdev_name(zhp->zpool_hdl, zhp,
2891 boolean_t result = (strcmp(mpath, spath) == 0);
2907 * Split a mirror pool. If newroot points to null, then a new nvlist
2908 * is generated and it is the responsibility of the caller to free it.
2911 zpool_vdev_split(zpool_handle_t *zhp, char *newname, nvlist_t **newroot,
2912 nvlist_t *props, splitflags_t flags)
2914 zfs_cmd_t zc = {"\0"};
2916 nvlist_t *tree, *config, **child, **newchild, *newconfig = NULL;
2917 nvlist_t **varray = NULL, *zc_props = NULL;
2918 uint_t c, children, newchildren, lastlog = 0, vcount, found = 0;
2919 libzfs_handle_t *hdl = zhp->zpool_hdl;
2921 boolean_t freelist = B_FALSE, memory_err = B_TRUE;
2924 (void) snprintf(msg, sizeof (msg),
2925 dgettext(TEXT_DOMAIN, "Unable to split %s"), zhp->zpool_name);
2927 if (!zpool_name_valid(hdl, B_FALSE, newname))
2928 return (zfs_error(hdl, EZFS_INVALIDNAME, msg));
2930 if ((config = zpool_get_config(zhp, NULL)) == NULL) {
2931 (void) fprintf(stderr, gettext("Internal error: unable to "
2932 "retrieve pool configuration\n"));
2936 verify(nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE, &tree)
2938 verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_VERSION, &vers) == 0);
2941 prop_flags_t flags = { .create = B_FALSE, .import = B_TRUE };
2942 if ((zc_props = zpool_valid_proplist(hdl, zhp->zpool_name,
2943 props, vers, flags, msg)) == NULL)
2947 if (nvlist_lookup_nvlist_array(tree, ZPOOL_CONFIG_CHILDREN, &child,
2949 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2950 "Source pool is missing vdev tree"));
2951 nvlist_free(zc_props);
2955 varray = zfs_alloc(hdl, children * sizeof (nvlist_t *));
2958 if (*newroot == NULL ||
2959 nvlist_lookup_nvlist_array(*newroot, ZPOOL_CONFIG_CHILDREN,
2960 &newchild, &newchildren) != 0)
2963 for (c = 0; c < children; c++) {
2964 uint64_t is_log = B_FALSE, is_hole = B_FALSE;
2966 nvlist_t **mchild, *vdev;
2971 * Unlike cache & spares, slogs are stored in the
2972 * ZPOOL_CONFIG_CHILDREN array. We filter them out here.
2974 (void) nvlist_lookup_uint64(child[c], ZPOOL_CONFIG_IS_LOG,
2976 (void) nvlist_lookup_uint64(child[c], ZPOOL_CONFIG_IS_HOLE,
2978 if (is_log || is_hole) {
2980 * Create a hole vdev and put it in the config.
2982 if (nvlist_alloc(&vdev, NV_UNIQUE_NAME, 0) != 0)
2984 if (nvlist_add_string(vdev, ZPOOL_CONFIG_TYPE,
2985 VDEV_TYPE_HOLE) != 0)
2987 if (nvlist_add_uint64(vdev, ZPOOL_CONFIG_IS_HOLE,
2992 varray[vcount++] = vdev;
2996 verify(nvlist_lookup_string(child[c], ZPOOL_CONFIG_TYPE, &type)
2998 if (strcmp(type, VDEV_TYPE_MIRROR) != 0) {
2999 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3000 "Source pool must be composed only of mirrors\n"));
3001 retval = zfs_error(hdl, EZFS_INVALCONFIG, msg);
3005 verify(nvlist_lookup_nvlist_array(child[c],
3006 ZPOOL_CONFIG_CHILDREN, &mchild, &mchildren) == 0);
3008 /* find or add an entry for this top-level vdev */
3009 if (newchildren > 0 &&
3010 (entry = find_vdev_entry(zhp, mchild, mchildren,
3011 newchild, newchildren)) >= 0) {
3012 /* We found a disk that the user specified. */
3013 vdev = mchild[entry];
3016 /* User didn't specify a disk for this vdev. */
3017 vdev = mchild[mchildren - 1];
3020 if (nvlist_dup(vdev, &varray[vcount++], 0) != 0)
3024 /* did we find every disk the user specified? */
3025 if (found != newchildren) {
3026 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "Device list must "
3027 "include at most one disk from each mirror"));
3028 retval = zfs_error(hdl, EZFS_INVALCONFIG, msg);
3032 /* Prepare the nvlist for populating. */
3033 if (*newroot == NULL) {
3034 if (nvlist_alloc(newroot, NV_UNIQUE_NAME, 0) != 0)
3037 if (nvlist_add_string(*newroot, ZPOOL_CONFIG_TYPE,
3038 VDEV_TYPE_ROOT) != 0)
3041 verify(nvlist_remove_all(*newroot, ZPOOL_CONFIG_CHILDREN) == 0);
3044 /* Add all the children we found */
3045 if (nvlist_add_nvlist_array(*newroot, ZPOOL_CONFIG_CHILDREN, varray,
3046 lastlog == 0 ? vcount : lastlog) != 0)
3050 * If we're just doing a dry run, exit now with success.
3053 memory_err = B_FALSE;
3058 /* now build up the config list & call the ioctl */
3059 if (nvlist_alloc(&newconfig, NV_UNIQUE_NAME, 0) != 0)
3062 if (nvlist_add_nvlist(newconfig,
3063 ZPOOL_CONFIG_VDEV_TREE, *newroot) != 0 ||
3064 nvlist_add_string(newconfig,
3065 ZPOOL_CONFIG_POOL_NAME, newname) != 0 ||
3066 nvlist_add_uint64(newconfig, ZPOOL_CONFIG_VERSION, vers) != 0)
3070 * The new pool is automatically part of the namespace unless we
3071 * explicitly export it.
3074 zc.zc_cookie = ZPOOL_EXPORT_AFTER_SPLIT;
3075 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
3076 (void) strlcpy(zc.zc_string, newname, sizeof (zc.zc_string));
3077 if (zcmd_write_conf_nvlist(hdl, &zc, newconfig) != 0)
3079 if (zc_props != NULL && zcmd_write_src_nvlist(hdl, &zc, zc_props) != 0)
3082 if (zfs_ioctl(hdl, ZFS_IOC_VDEV_SPLIT, &zc) != 0) {
3083 retval = zpool_standard_error(hdl, errno, msg);
3088 memory_err = B_FALSE;
3091 if (varray != NULL) {
3094 for (v = 0; v < vcount; v++)
3095 nvlist_free(varray[v]);
3098 zcmd_free_nvlists(&zc);
3099 nvlist_free(zc_props);
3100 nvlist_free(newconfig);
3102 nvlist_free(*newroot);
3110 return (no_memory(hdl));
3116 * Remove the given device. Currently, this is supported only for hot spares
3117 * and level 2 cache devices.
3120 zpool_vdev_remove(zpool_handle_t *zhp, const char *path)
3122 zfs_cmd_t zc = {"\0"};
3125 boolean_t avail_spare, l2cache, islog;
3126 libzfs_handle_t *hdl = zhp->zpool_hdl;
3129 (void) snprintf(msg, sizeof (msg),
3130 dgettext(TEXT_DOMAIN, "cannot remove %s"), path);
3132 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
3133 if ((tgt = zpool_find_vdev(zhp, path, &avail_spare, &l2cache,
3135 return (zfs_error(hdl, EZFS_NODEVICE, msg));
3137 * XXX - this should just go away.
3139 if (!avail_spare && !l2cache && !islog) {
3140 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3141 "only inactive hot spares, cache, top-level, "
3142 "or log devices can be removed"));
3143 return (zfs_error(hdl, EZFS_NODEVICE, msg));
3146 version = zpool_get_prop_int(zhp, ZPOOL_PROP_VERSION, NULL);
3147 if (islog && version < SPA_VERSION_HOLES) {
3148 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3149 "pool must be upgrade to support log removal"));
3150 return (zfs_error(hdl, EZFS_BADVERSION, msg));
3153 verify(nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID, &zc.zc_guid) == 0);
3155 if (zfs_ioctl(hdl, ZFS_IOC_VDEV_REMOVE, &zc) == 0)
3158 return (zpool_standard_error(hdl, errno, msg));
3162 * Clear the errors for the pool, or the particular device if specified.
3165 zpool_clear(zpool_handle_t *zhp, const char *path, nvlist_t *rewindnvl)
3167 zfs_cmd_t zc = {"\0"};
3170 zpool_rewind_policy_t policy;
3171 boolean_t avail_spare, l2cache;
3172 libzfs_handle_t *hdl = zhp->zpool_hdl;
3173 nvlist_t *nvi = NULL;
3177 (void) snprintf(msg, sizeof (msg),
3178 dgettext(TEXT_DOMAIN, "cannot clear errors for %s"),
3181 (void) snprintf(msg, sizeof (msg),
3182 dgettext(TEXT_DOMAIN, "cannot clear errors for %s"),
3185 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
3187 if ((tgt = zpool_find_vdev(zhp, path, &avail_spare,
3188 &l2cache, NULL)) == 0)
3189 return (zfs_error(hdl, EZFS_NODEVICE, msg));
3192 * Don't allow error clearing for hot spares. Do allow
3193 * error clearing for l2cache devices.
3196 return (zfs_error(hdl, EZFS_ISSPARE, msg));
3198 verify(nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID,
3202 zpool_get_rewind_policy(rewindnvl, &policy);
3203 zc.zc_cookie = policy.zrp_request;
3205 if (zcmd_alloc_dst_nvlist(hdl, &zc, zhp->zpool_config_size * 2) != 0)
3208 if (zcmd_write_src_nvlist(hdl, &zc, rewindnvl) != 0)
3211 while ((error = zfs_ioctl(hdl, ZFS_IOC_CLEAR, &zc)) != 0 &&
3213 if (zcmd_expand_dst_nvlist(hdl, &zc) != 0) {
3214 zcmd_free_nvlists(&zc);
3219 if (!error || ((policy.zrp_request & ZPOOL_TRY_REWIND) &&
3220 errno != EPERM && errno != EACCES)) {
3221 if (policy.zrp_request &
3222 (ZPOOL_DO_REWIND | ZPOOL_TRY_REWIND)) {
3223 (void) zcmd_read_dst_nvlist(hdl, &zc, &nvi);
3224 zpool_rewind_exclaim(hdl, zc.zc_name,
3225 ((policy.zrp_request & ZPOOL_TRY_REWIND) != 0),
3229 zcmd_free_nvlists(&zc);
3233 zcmd_free_nvlists(&zc);
3234 return (zpool_standard_error(hdl, errno, msg));
3238 * Similar to zpool_clear(), but takes a GUID (used by fmd).
3241 zpool_vdev_clear(zpool_handle_t *zhp, uint64_t guid)
3243 zfs_cmd_t zc = {"\0"};
3245 libzfs_handle_t *hdl = zhp->zpool_hdl;
3247 (void) snprintf(msg, sizeof (msg),
3248 dgettext(TEXT_DOMAIN, "cannot clear errors for %llx"),
3249 (u_longlong_t)guid);
3251 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
3253 zc.zc_cookie = ZPOOL_NO_REWIND;
3255 if (ioctl(hdl->libzfs_fd, ZFS_IOC_CLEAR, &zc) == 0)
3258 return (zpool_standard_error(hdl, errno, msg));
3262 * Change the GUID for a pool.
3265 zpool_reguid(zpool_handle_t *zhp)
3268 libzfs_handle_t *hdl = zhp->zpool_hdl;
3269 zfs_cmd_t zc = {"\0"};
3271 (void) snprintf(msg, sizeof (msg),
3272 dgettext(TEXT_DOMAIN, "cannot reguid '%s'"), zhp->zpool_name);
3274 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
3275 if (zfs_ioctl(hdl, ZFS_IOC_POOL_REGUID, &zc) == 0)
3278 return (zpool_standard_error(hdl, errno, msg));
3285 zpool_reopen(zpool_handle_t *zhp)
3287 zfs_cmd_t zc = {"\0"};
3289 libzfs_handle_t *hdl = zhp->zpool_hdl;
3291 (void) snprintf(msg, sizeof (msg),
3292 dgettext(TEXT_DOMAIN, "cannot reopen '%s'"),
3295 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
3296 if (zfs_ioctl(hdl, ZFS_IOC_POOL_REOPEN, &zc) == 0)
3298 return (zpool_standard_error(hdl, errno, msg));
3301 #if defined(__sun__) || defined(__sun)
3303 * Convert from a devid string to a path.
3306 devid_to_path(char *devid_str)
3311 devid_nmlist_t *list = NULL;
3314 if (devid_str_decode(devid_str, &devid, &minor) != 0)
3317 ret = devid_deviceid_to_nmlist("/dev", devid, minor, &list);
3319 devid_str_free(minor);
3326 * In a case the strdup() fails, we will just return NULL below.
3328 path = strdup(list[0].devname);
3330 devid_free_nmlist(list);
3336 * Convert from a path to a devid string.
3339 path_to_devid(const char *path)
3345 if ((fd = open(path, O_RDONLY)) < 0)
3350 if (devid_get(fd, &devid) == 0) {
3351 if (devid_get_minor_name(fd, &minor) == 0)
3352 ret = devid_str_encode(devid, minor);
3354 devid_str_free(minor);
3363 * Issue the necessary ioctl() to update the stored path value for the vdev. We
3364 * ignore any failure here, since a common case is for an unprivileged user to
3365 * type 'zpool status', and we'll display the correct information anyway.
3368 set_path(zpool_handle_t *zhp, nvlist_t *nv, const char *path)
3370 zfs_cmd_t zc = {"\0"};
3372 (void) strncpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
3373 (void) strncpy(zc.zc_value, path, sizeof (zc.zc_value));
3374 verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID,
3377 (void) ioctl(zhp->zpool_hdl->libzfs_fd, ZFS_IOC_VDEV_SETPATH, &zc);
3382 * Remove partition suffix from a vdev path. Partition suffixes may take three
3383 * forms: "-partX", "pX", or "X", where X is a string of digits. The second
3384 * case only occurs when the suffix is preceded by a digit, i.e. "md0p0" The
3385 * third case only occurs when preceded by a string matching the regular
3386 * expression "^([hsv]|xv)d[a-z]+", i.e. a scsi, ide, virtio or xen disk.
3389 strip_partition(libzfs_handle_t *hdl, char *path)
3391 char *tmp = zfs_strdup(hdl, path);
3392 char *part = NULL, *d = NULL;
3394 if ((part = strstr(tmp, "-part")) && part != tmp) {
3396 } else if ((part = strrchr(tmp, 'p')) &&
3397 part > tmp + 1 && isdigit(*(part-1))) {
3399 } else if ((tmp[0] == 'h' || tmp[0] == 's' || tmp[0] == 'v') &&
3401 for (d = &tmp[2]; isalpha(*d); part = ++d);
3402 } else if (strncmp("xvd", tmp, 3) == 0) {
3403 for (d = &tmp[3]; isalpha(*d); part = ++d);
3405 if (part && d && *d != '\0') {
3406 for (; isdigit(*d); d++);
3413 #define PATH_BUF_LEN 64
3416 * Given a vdev, return the name to display in iostat. If the vdev has a path,
3417 * we use that, stripping off any leading "/dev/dsk/"; if not, we use the type.
3418 * We also check if this is a whole disk, in which case we strip off the
3419 * trailing 's0' slice name.
3421 * This routine is also responsible for identifying when disks have been
3422 * reconfigured in a new location. The kernel will have opened the device by
3423 * devid, but the path will still refer to the old location. To catch this, we
3424 * first do a path -> devid translation (which is fast for the common case). If
3425 * the devid matches, we're done. If not, we do a reverse devid -> path
3426 * translation and issue the appropriate ioctl() to update the path of the vdev.
3427 * If 'zhp' is NULL, then this is an exported pool, and we don't need to do any
3431 zpool_vdev_name(libzfs_handle_t *hdl, zpool_handle_t *zhp, nvlist_t *nv,
3434 char *path, *type, *env;
3436 char buf[PATH_BUF_LEN];
3437 char tmpbuf[PATH_BUF_LEN];
3439 env = getenv("ZPOOL_VDEV_NAME_PATH");
3440 if (env && (strtoul(env, NULL, 0) > 0 ||
3441 !strncasecmp(env, "YES", 3) || !strncasecmp(env, "ON", 2)))
3442 name_flags |= VDEV_NAME_PATH;
3444 env = getenv("ZPOOL_VDEV_NAME_GUID");
3445 if (env && (strtoul(env, NULL, 0) > 0 ||
3446 !strncasecmp(env, "YES", 3) || !strncasecmp(env, "ON", 2)))
3447 name_flags |= VDEV_NAME_GUID;
3449 env = getenv("ZPOOL_VDEV_NAME_FOLLOW_LINKS");
3450 if (env && (strtoul(env, NULL, 0) > 0 ||
3451 !strncasecmp(env, "YES", 3) || !strncasecmp(env, "ON", 2)))
3452 name_flags |= VDEV_NAME_FOLLOW_LINKS;
3454 if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_NOT_PRESENT, &value) == 0 ||
3455 name_flags & VDEV_NAME_GUID) {
3456 nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID, &value);
3457 (void) snprintf(buf, sizeof (buf), "%llu", (u_longlong_t)value);
3459 } else if (nvlist_lookup_string(nv, ZPOOL_CONFIG_PATH, &path) == 0) {
3460 #if defined(__sun__) || defined(__sun)
3462 * Live VDEV path updates to a kernel VDEV during a
3463 * zpool_vdev_name lookup are not supported on Linux.
3470 * If the device is dead (faulted, offline, etc) then don't
3471 * bother opening it. Otherwise we may be forcing the user to
3472 * open a misbehaving device, which can have undesirable
3475 if ((nvlist_lookup_uint64_array(nv, ZPOOL_CONFIG_VDEV_STATS,
3476 (uint64_t **)&vs, &vsc) != 0 ||
3477 vs->vs_state >= VDEV_STATE_DEGRADED) &&
3479 nvlist_lookup_string(nv, ZPOOL_CONFIG_DEVID, &devid) == 0) {
3481 * Determine if the current path is correct.
3483 char *newdevid = path_to_devid(path);
3485 if (newdevid == NULL ||
3486 strcmp(devid, newdevid) != 0) {
3489 if ((newpath = devid_to_path(devid)) != NULL) {
3491 * Update the path appropriately.
3493 set_path(zhp, nv, newpath);
3494 if (nvlist_add_string(nv,
3495 ZPOOL_CONFIG_PATH, newpath) == 0)
3496 verify(nvlist_lookup_string(nv,
3504 devid_str_free(newdevid);
3508 if (name_flags & VDEV_NAME_FOLLOW_LINKS) {
3509 char *rp = realpath(path, NULL);
3511 strlcpy(buf, rp, sizeof (buf));
3518 * For a block device only use the name.
3520 verify(nvlist_lookup_string(nv, ZPOOL_CONFIG_TYPE, &type) == 0);
3521 if ((strcmp(type, VDEV_TYPE_DISK) == 0) &&
3522 !(name_flags & VDEV_NAME_PATH)) {
3523 path = strrchr(path, '/');
3528 * Remove the partition from the path it this is a whole disk.
3530 if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_WHOLE_DISK, &value)
3531 == 0 && value && !(name_flags & VDEV_NAME_PATH)) {
3532 return (strip_partition(hdl, path));
3535 verify(nvlist_lookup_string(nv, ZPOOL_CONFIG_TYPE, &path) == 0);
3538 * If it's a raidz device, we need to stick in the parity level.
3540 if (strcmp(path, VDEV_TYPE_RAIDZ) == 0) {
3541 verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_NPARITY,
3543 (void) snprintf(buf, sizeof (buf), "%s%llu", path,
3544 (u_longlong_t)value);
3549 * We identify each top-level vdev by using a <type-id>
3550 * naming convention.
3552 if (name_flags & VDEV_NAME_TYPE_ID) {
3554 verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_ID,
3556 (void) snprintf(tmpbuf, sizeof (tmpbuf), "%s-%llu",
3557 path, (u_longlong_t)id);
3562 return (zfs_strdup(hdl, path));
3566 zbookmark_mem_compare(const void *a, const void *b)
3568 return (memcmp(a, b, sizeof (zbookmark_phys_t)));
3572 * Retrieve the persistent error log, uniquify the members, and return to the
3576 zpool_get_errlog(zpool_handle_t *zhp, nvlist_t **nverrlistp)
3578 zfs_cmd_t zc = {"\0"};
3580 zbookmark_phys_t *zb = NULL;
3584 * Retrieve the raw error list from the kernel. If the number of errors
3585 * has increased, allocate more space and continue until we get the
3588 verify(nvlist_lookup_uint64(zhp->zpool_config, ZPOOL_CONFIG_ERRCOUNT,
3592 if ((zc.zc_nvlist_dst = (uintptr_t)zfs_alloc(zhp->zpool_hdl,
3593 count * sizeof (zbookmark_phys_t))) == (uintptr_t)NULL)
3595 zc.zc_nvlist_dst_size = count;
3596 (void) strcpy(zc.zc_name, zhp->zpool_name);
3598 if (ioctl(zhp->zpool_hdl->libzfs_fd, ZFS_IOC_ERROR_LOG,
3600 free((void *)(uintptr_t)zc.zc_nvlist_dst);
3601 if (errno == ENOMEM) {
3604 count = zc.zc_nvlist_dst_size;
3605 dst = zfs_alloc(zhp->zpool_hdl, count *
3606 sizeof (zbookmark_phys_t));
3609 zc.zc_nvlist_dst = (uintptr_t)dst;
3619 * Sort the resulting bookmarks. This is a little confusing due to the
3620 * implementation of ZFS_IOC_ERROR_LOG. The bookmarks are copied last
3621 * to first, and 'zc_nvlist_dst_size' indicates the number of boomarks
3622 * _not_ copied as part of the process. So we point the start of our
3623 * array appropriate and decrement the total number of elements.
3625 zb = ((zbookmark_phys_t *)(uintptr_t)zc.zc_nvlist_dst) +
3626 zc.zc_nvlist_dst_size;
3627 count -= zc.zc_nvlist_dst_size;
3629 qsort(zb, count, sizeof (zbookmark_phys_t), zbookmark_mem_compare);
3631 verify(nvlist_alloc(nverrlistp, 0, KM_SLEEP) == 0);
3634 * Fill in the nverrlistp with nvlist's of dataset and object numbers.
3636 for (i = 0; i < count; i++) {
3639 /* ignoring zb_blkid and zb_level for now */
3640 if (i > 0 && zb[i-1].zb_objset == zb[i].zb_objset &&
3641 zb[i-1].zb_object == zb[i].zb_object)
3644 if (nvlist_alloc(&nv, NV_UNIQUE_NAME, KM_SLEEP) != 0)
3646 if (nvlist_add_uint64(nv, ZPOOL_ERR_DATASET,
3647 zb[i].zb_objset) != 0) {
3651 if (nvlist_add_uint64(nv, ZPOOL_ERR_OBJECT,
3652 zb[i].zb_object) != 0) {
3656 if (nvlist_add_nvlist(*nverrlistp, "ejk", nv) != 0) {
3663 free((void *)(uintptr_t)zc.zc_nvlist_dst);
3667 free((void *)(uintptr_t)zc.zc_nvlist_dst);
3668 return (no_memory(zhp->zpool_hdl));
3672 * Upgrade a ZFS pool to the latest on-disk version.
3675 zpool_upgrade(zpool_handle_t *zhp, uint64_t new_version)
3677 zfs_cmd_t zc = {"\0"};
3678 libzfs_handle_t *hdl = zhp->zpool_hdl;
3680 (void) strcpy(zc.zc_name, zhp->zpool_name);
3681 zc.zc_cookie = new_version;
3683 if (zfs_ioctl(hdl, ZFS_IOC_POOL_UPGRADE, &zc) != 0)
3684 return (zpool_standard_error_fmt(hdl, errno,
3685 dgettext(TEXT_DOMAIN, "cannot upgrade '%s'"),
3691 zfs_save_arguments(int argc, char **argv, char *string, int len)
3695 (void) strlcpy(string, basename(argv[0]), len);
3696 for (i = 1; i < argc; i++) {
3697 (void) strlcat(string, " ", len);
3698 (void) strlcat(string, argv[i], len);
3703 zpool_log_history(libzfs_handle_t *hdl, const char *message)
3705 zfs_cmd_t zc = {"\0"};
3709 args = fnvlist_alloc();
3710 fnvlist_add_string(args, "message", message);
3711 err = zcmd_write_src_nvlist(hdl, &zc, args);
3713 err = ioctl(hdl->libzfs_fd, ZFS_IOC_LOG_HISTORY, &zc);
3715 zcmd_free_nvlists(&zc);
3720 * Perform ioctl to get some command history of a pool.
3722 * 'buf' is the buffer to fill up to 'len' bytes. 'off' is the
3723 * logical offset of the history buffer to start reading from.
3725 * Upon return, 'off' is the next logical offset to read from and
3726 * 'len' is the actual amount of bytes read into 'buf'.
3729 get_history(zpool_handle_t *zhp, char *buf, uint64_t *off, uint64_t *len)
3731 zfs_cmd_t zc = {"\0"};
3732 libzfs_handle_t *hdl = zhp->zpool_hdl;
3734 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
3736 zc.zc_history = (uint64_t)(uintptr_t)buf;
3737 zc.zc_history_len = *len;
3738 zc.zc_history_offset = *off;
3740 if (ioctl(hdl->libzfs_fd, ZFS_IOC_POOL_GET_HISTORY, &zc) != 0) {
3743 return (zfs_error_fmt(hdl, EZFS_PERM,
3744 dgettext(TEXT_DOMAIN,
3745 "cannot show history for pool '%s'"),
3748 return (zfs_error_fmt(hdl, EZFS_NOHISTORY,
3749 dgettext(TEXT_DOMAIN, "cannot get history for pool "
3750 "'%s'"), zhp->zpool_name));
3752 return (zfs_error_fmt(hdl, EZFS_BADVERSION,
3753 dgettext(TEXT_DOMAIN, "cannot get history for pool "
3754 "'%s', pool must be upgraded"), zhp->zpool_name));
3756 return (zpool_standard_error_fmt(hdl, errno,
3757 dgettext(TEXT_DOMAIN,
3758 "cannot get history for '%s'"), zhp->zpool_name));
3762 *len = zc.zc_history_len;
3763 *off = zc.zc_history_offset;
3769 * Process the buffer of nvlists, unpacking and storing each nvlist record
3770 * into 'records'. 'leftover' is set to the number of bytes that weren't
3771 * processed as there wasn't a complete record.
3774 zpool_history_unpack(char *buf, uint64_t bytes_read, uint64_t *leftover,
3775 nvlist_t ***records, uint_t *numrecords)
3781 while (bytes_read > sizeof (reclen)) {
3783 /* get length of packed record (stored as little endian) */
3784 for (i = 0, reclen = 0; i < sizeof (reclen); i++)
3785 reclen += (uint64_t)(((uchar_t *)buf)[i]) << (8*i);
3787 if (bytes_read < sizeof (reclen) + reclen)
3791 if (nvlist_unpack(buf + sizeof (reclen), reclen, &nv, 0) != 0)
3793 bytes_read -= sizeof (reclen) + reclen;
3794 buf += sizeof (reclen) + reclen;
3796 /* add record to nvlist array */
3798 if (ISP2(*numrecords + 1)) {
3799 *records = realloc(*records,
3800 *numrecords * 2 * sizeof (nvlist_t *));
3802 (*records)[*numrecords - 1] = nv;
3805 *leftover = bytes_read;
3810 * Retrieve the command history of a pool.
3813 zpool_get_history(zpool_handle_t *zhp, nvlist_t **nvhisp)
3816 int buflen = 128 * 1024;
3818 nvlist_t **records = NULL;
3819 uint_t numrecords = 0;
3822 buf = malloc(buflen);
3826 uint64_t bytes_read = buflen;
3829 if ((err = get_history(zhp, buf, &off, &bytes_read)) != 0)
3832 /* if nothing else was read in, we're at EOF, just return */
3836 if ((err = zpool_history_unpack(buf, bytes_read,
3837 &leftover, &records, &numrecords)) != 0)
3840 if (leftover == bytes_read) {
3842 * no progress made, because buffer is not big enough
3843 * to hold this record; resize and retry.
3847 buf = malloc(buflen);
3858 verify(nvlist_alloc(nvhisp, NV_UNIQUE_NAME, 0) == 0);
3859 verify(nvlist_add_nvlist_array(*nvhisp, ZPOOL_HIST_RECORD,
3860 records, numrecords) == 0);
3862 for (i = 0; i < numrecords; i++)
3863 nvlist_free(records[i]);
3870 * Retrieve the next event given the passed 'zevent_fd' file descriptor.
3871 * If there is a new event available 'nvp' will contain a newly allocated
3872 * nvlist and 'dropped' will be set to the number of missed events since
3873 * the last call to this function. When 'nvp' is set to NULL it indicates
3874 * no new events are available. In either case the function returns 0 and
3875 * it is up to the caller to free 'nvp'. In the case of a fatal error the
3876 * function will return a non-zero value. When the function is called in
3877 * blocking mode (the default, unless the ZEVENT_NONBLOCK flag is passed),
3878 * it will not return until a new event is available.
3881 zpool_events_next(libzfs_handle_t *hdl, nvlist_t **nvp,
3882 int *dropped, unsigned flags, int zevent_fd)
3884 zfs_cmd_t zc = {"\0"};
3889 zc.zc_cleanup_fd = zevent_fd;
3891 if (flags & ZEVENT_NONBLOCK)
3892 zc.zc_guid = ZEVENT_NONBLOCK;
3894 if (zcmd_alloc_dst_nvlist(hdl, &zc, ZEVENT_SIZE) != 0)
3898 if (zfs_ioctl(hdl, ZFS_IOC_EVENTS_NEXT, &zc) != 0) {
3901 error = zfs_error_fmt(hdl, EZFS_POOLUNAVAIL,
3902 dgettext(TEXT_DOMAIN, "zfs shutdown"));
3905 /* Blocking error case should not occur */
3906 if (!(flags & ZEVENT_NONBLOCK))
3907 error = zpool_standard_error_fmt(hdl, errno,
3908 dgettext(TEXT_DOMAIN, "cannot get event"));
3912 if (zcmd_expand_dst_nvlist(hdl, &zc) != 0) {
3913 error = zfs_error_fmt(hdl, EZFS_NOMEM,
3914 dgettext(TEXT_DOMAIN, "cannot get event"));
3920 error = zpool_standard_error_fmt(hdl, errno,
3921 dgettext(TEXT_DOMAIN, "cannot get event"));
3926 error = zcmd_read_dst_nvlist(hdl, &zc, nvp);
3930 *dropped = (int)zc.zc_cookie;
3932 zcmd_free_nvlists(&zc);
3941 zpool_events_clear(libzfs_handle_t *hdl, int *count)
3943 zfs_cmd_t zc = {"\0"};
3946 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN,
3947 "cannot clear events"));
3949 if (zfs_ioctl(hdl, ZFS_IOC_EVENTS_CLEAR, &zc) != 0)
3950 return (zpool_standard_error_fmt(hdl, errno, msg));
3953 *count = (int)zc.zc_cookie; /* # of events cleared */
3959 * Seek to a specific EID, ZEVENT_SEEK_START, or ZEVENT_SEEK_END for
3960 * the passed zevent_fd file handle. On success zero is returned,
3961 * otherwise -1 is returned and hdl->libzfs_error is set to the errno.
3964 zpool_events_seek(libzfs_handle_t *hdl, uint64_t eid, int zevent_fd)
3966 zfs_cmd_t zc = {"\0"};
3970 zc.zc_cleanup_fd = zevent_fd;
3972 if (zfs_ioctl(hdl, ZFS_IOC_EVENTS_SEEK, &zc) != 0) {
3975 error = zfs_error_fmt(hdl, EZFS_NOENT,
3976 dgettext(TEXT_DOMAIN, "cannot get event"));
3980 error = zfs_error_fmt(hdl, EZFS_NOMEM,
3981 dgettext(TEXT_DOMAIN, "cannot get event"));
3985 error = zpool_standard_error_fmt(hdl, errno,
3986 dgettext(TEXT_DOMAIN, "cannot get event"));
3995 zpool_obj_to_path(zpool_handle_t *zhp, uint64_t dsobj, uint64_t obj,
3996 char *pathname, size_t len)
3998 zfs_cmd_t zc = {"\0"};
3999 boolean_t mounted = B_FALSE;
4000 char *mntpnt = NULL;
4001 char dsname[MAXNAMELEN];
4004 /* special case for the MOS */
4005 (void) snprintf(pathname, len, "<metadata>:<0x%llx>",
4010 /* get the dataset's name */
4011 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
4013 if (ioctl(zhp->zpool_hdl->libzfs_fd,
4014 ZFS_IOC_DSOBJ_TO_DSNAME, &zc) != 0) {
4015 /* just write out a path of two object numbers */
4016 (void) snprintf(pathname, len, "<0x%llx>:<0x%llx>",
4017 (longlong_t)dsobj, (longlong_t)obj);
4020 (void) strlcpy(dsname, zc.zc_value, sizeof (dsname));
4022 /* find out if the dataset is mounted */
4023 mounted = is_mounted(zhp->zpool_hdl, dsname, &mntpnt);
4025 /* get the corrupted object's path */
4026 (void) strlcpy(zc.zc_name, dsname, sizeof (zc.zc_name));
4028 if (ioctl(zhp->zpool_hdl->libzfs_fd, ZFS_IOC_OBJ_TO_PATH,
4031 (void) snprintf(pathname, len, "%s%s", mntpnt,
4034 (void) snprintf(pathname, len, "%s:%s",
4035 dsname, zc.zc_value);
4038 (void) snprintf(pathname, len, "%s:<0x%llx>", dsname,
4045 * Read the EFI label from the config, if a label does not exist then
4046 * pass back the error to the caller. If the caller has passed a non-NULL
4047 * diskaddr argument then we set it to the starting address of the EFI
4051 read_efi_label(nvlist_t *config, diskaddr_t *sb)
4055 char diskname[MAXPATHLEN];
4058 if (nvlist_lookup_string(config, ZPOOL_CONFIG_PATH, &path) != 0)
4061 (void) snprintf(diskname, sizeof (diskname), "%s%s", DISK_ROOT,
4062 strrchr(path, '/'));
4063 if ((fd = open(diskname, O_RDWR|O_DIRECT)) >= 0) {
4064 struct dk_gpt *vtoc;
4066 if ((err = efi_alloc_and_read(fd, &vtoc)) >= 0) {
4068 *sb = vtoc->efi_parts[0].p_start;
4077 * determine where a partition starts on a disk in the current
4081 find_start_block(nvlist_t *config)
4085 diskaddr_t sb = MAXOFFSET_T;
4088 if (nvlist_lookup_nvlist_array(config,
4089 ZPOOL_CONFIG_CHILDREN, &child, &children) != 0) {
4090 if (nvlist_lookup_uint64(config,
4091 ZPOOL_CONFIG_WHOLE_DISK,
4092 &wholedisk) != 0 || !wholedisk) {
4093 return (MAXOFFSET_T);
4095 if (read_efi_label(config, &sb) < 0)
4100 for (c = 0; c < children; c++) {
4101 sb = find_start_block(child[c]);
4102 if (sb != MAXOFFSET_T) {
4106 return (MAXOFFSET_T);
4110 zpool_label_disk_check(char *path)
4112 struct dk_gpt *vtoc;
4115 if ((fd = open(path, O_RDWR|O_DIRECT)) < 0)
4118 if ((err = efi_alloc_and_read(fd, &vtoc)) != 0) {
4123 if (vtoc->efi_flags & EFI_GPT_PRIMARY_CORRUPT) {
4135 * Generate a unique partition name for the ZFS member. Partitions must
4136 * have unique names to ensure udev will be able to create symlinks under
4137 * /dev/disk/by-partlabel/ for all pool members. The partition names are
4138 * of the form <pool>-<unique-id>.
4141 zpool_label_name(char *label_name, int label_size)
4146 fd = open("/dev/urandom", O_RDONLY);
4148 if (read(fd, &id, sizeof (id)) != sizeof (id))
4155 id = (((uint64_t)rand()) << 32) | (uint64_t)rand();
4157 snprintf(label_name, label_size, "zfs-%016llx", (u_longlong_t) id);
4161 * Label an individual disk. The name provided is the short name,
4162 * stripped of any leading /dev path.
4165 zpool_label_disk(libzfs_handle_t *hdl, zpool_handle_t *zhp, char *name)
4167 char path[MAXPATHLEN];
4168 struct dk_gpt *vtoc;
4170 size_t resv = EFI_MIN_RESV_SIZE;
4171 uint64_t slice_size;
4172 diskaddr_t start_block;
4175 /* prepare an error message just in case */
4176 (void) snprintf(errbuf, sizeof (errbuf),
4177 dgettext(TEXT_DOMAIN, "cannot label '%s'"), name);
4182 #if defined(__sun__) || defined(__sun)
4183 if (zpool_is_bootable(zhp)) {
4184 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
4185 "EFI labeled devices are not supported on root "
4187 return (zfs_error(hdl, EZFS_POOL_NOTSUP, errbuf));
4191 verify(nvlist_lookup_nvlist(zhp->zpool_config,
4192 ZPOOL_CONFIG_VDEV_TREE, &nvroot) == 0);
4194 if (zhp->zpool_start_block == 0)
4195 start_block = find_start_block(nvroot);
4197 start_block = zhp->zpool_start_block;
4198 zhp->zpool_start_block = start_block;
4201 start_block = NEW_START_BLOCK;
4204 (void) snprintf(path, sizeof (path), "%s/%s", DISK_ROOT, name);
4206 if ((fd = open(path, O_RDWR|O_DIRECT)) < 0) {
4208 * This shouldn't happen. We've long since verified that this
4209 * is a valid device.
4211 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "cannot "
4212 "label '%s': unable to open device: %d"), path, errno);
4213 return (zfs_error(hdl, EZFS_OPENFAILED, errbuf));
4216 if (efi_alloc_and_init(fd, EFI_NUMPAR, &vtoc) != 0) {
4218 * The only way this can fail is if we run out of memory, or we
4219 * were unable to read the disk's capacity
4221 if (errno == ENOMEM)
4222 (void) no_memory(hdl);
4225 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "cannot "
4226 "label '%s': unable to read disk capacity"), path);
4228 return (zfs_error(hdl, EZFS_NOCAP, errbuf));
4231 slice_size = vtoc->efi_last_u_lba + 1;
4232 slice_size -= EFI_MIN_RESV_SIZE;
4233 if (start_block == MAXOFFSET_T)
4234 start_block = NEW_START_BLOCK;
4235 slice_size -= start_block;
4236 slice_size = P2ALIGN(slice_size, PARTITION_END_ALIGNMENT);
4238 vtoc->efi_parts[0].p_start = start_block;
4239 vtoc->efi_parts[0].p_size = slice_size;
4242 * Why we use V_USR: V_BACKUP confuses users, and is considered
4243 * disposable by some EFI utilities (since EFI doesn't have a backup
4244 * slice). V_UNASSIGNED is supposed to be used only for zero size
4245 * partitions, and efi_write() will fail if we use it. V_ROOT, V_BOOT,
4246 * etc. were all pretty specific. V_USR is as close to reality as we
4247 * can get, in the absence of V_OTHER.
4249 vtoc->efi_parts[0].p_tag = V_USR;
4250 zpool_label_name(vtoc->efi_parts[0].p_name, EFI_PART_NAME_LEN);
4252 vtoc->efi_parts[8].p_start = slice_size + start_block;
4253 vtoc->efi_parts[8].p_size = resv;
4254 vtoc->efi_parts[8].p_tag = V_RESERVED;
4256 if ((rval = efi_write(fd, vtoc)) != 0 || (rval = efi_rescan(fd)) != 0) {
4258 * Some block drivers (like pcata) may not support EFI
4259 * GPT labels. Print out a helpful error message dir-
4260 * ecting the user to manually label the disk and give
4266 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "try using "
4267 "parted(8) and then provide a specific slice: %d"), rval);
4268 return (zfs_error(hdl, EZFS_LABELFAILED, errbuf));
4274 (void) snprintf(path, sizeof (path), "%s/%s", DISK_ROOT, name);
4275 (void) zfs_append_partition(path, MAXPATHLEN);
4277 /* Wait to udev to signal use the device has settled. */
4278 rval = zpool_label_disk_wait(path, DISK_LABEL_WAIT);
4280 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "failed to "
4281 "detect device partitions on '%s': %d"), path, rval);
4282 return (zfs_error(hdl, EZFS_LABELFAILED, errbuf));
4285 /* We can't be to paranoid. Read the label back and verify it. */
4286 (void) snprintf(path, sizeof (path), "%s/%s", DISK_ROOT, name);
4287 rval = zpool_label_disk_check(path);
4289 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "freshly written "
4290 "EFI label on '%s' is damaged. Ensure\nthis device "
4291 "is not in in use, and is functioning properly: %d"),
4293 return (zfs_error(hdl, EZFS_LABELFAILED, errbuf));