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);
340 } else if (literal) {
341 (void) snprintf(buf, len, "%llu",
342 (u_longlong_t)intval);
344 (void) snprintf(buf, len, "%llu%%",
345 (u_longlong_t)intval);
349 case ZPOOL_PROP_DEDUPRATIO:
351 (void) snprintf(buf, len, "%llu.%02llu",
352 (u_longlong_t)(intval / 100),
353 (u_longlong_t)(intval % 100));
355 (void) snprintf(buf, len, "%llu.%02llux",
356 (u_longlong_t)(intval / 100),
357 (u_longlong_t)(intval % 100));
360 case ZPOOL_PROP_HEALTH:
361 verify(nvlist_lookup_nvlist(zpool_get_config(zhp, NULL),
362 ZPOOL_CONFIG_VDEV_TREE, &nvroot) == 0);
363 verify(nvlist_lookup_uint64_array(nvroot,
364 ZPOOL_CONFIG_VDEV_STATS, (uint64_t **)&vs, &vsc)
367 (void) strlcpy(buf, zpool_state_to_name(intval,
370 case ZPOOL_PROP_VERSION:
371 if (intval >= SPA_VERSION_FEATURES) {
372 (void) snprintf(buf, len, "-");
377 (void) snprintf(buf, len, "%llu", (u_longlong_t)intval);
381 case PROP_TYPE_INDEX:
382 intval = zpool_get_prop_int(zhp, prop, &src);
383 if (zpool_prop_index_to_string(prop, intval, &strval)
386 (void) strlcpy(buf, strval, len);
400 * Check if the bootfs name has the same pool name as it is set to.
401 * Assuming bootfs is a valid dataset name.
404 bootfs_name_valid(const char *pool, char *bootfs)
406 int len = strlen(pool);
408 if (!zfs_name_valid(bootfs, ZFS_TYPE_FILESYSTEM|ZFS_TYPE_SNAPSHOT))
411 if (strncmp(pool, bootfs, len) == 0 &&
412 (bootfs[len] == '/' || bootfs[len] == '\0'))
418 #if defined(__sun__) || defined(__sun)
420 * Inspect the configuration to determine if any of the devices contain
424 pool_uses_efi(nvlist_t *config)
429 if (nvlist_lookup_nvlist_array(config, ZPOOL_CONFIG_CHILDREN,
430 &child, &children) != 0)
431 return (read_efi_label(config, NULL) >= 0);
433 for (c = 0; c < children; c++) {
434 if (pool_uses_efi(child[c]))
442 zpool_is_bootable(zpool_handle_t *zhp)
444 char bootfs[ZFS_MAX_DATASET_NAME_LEN];
446 return (zpool_get_prop(zhp, ZPOOL_PROP_BOOTFS, bootfs,
447 sizeof (bootfs), NULL, B_FALSE) == 0 && strncmp(bootfs, "-",
448 sizeof (bootfs)) != 0);
453 * Given an nvlist of zpool properties to be set, validate that they are
454 * correct, and parse any numeric properties (index, boolean, etc) if they are
455 * specified as strings.
458 zpool_valid_proplist(libzfs_handle_t *hdl, const char *poolname,
459 nvlist_t *props, uint64_t version, prop_flags_t flags, char *errbuf)
467 struct stat64 statbuf;
471 if (nvlist_alloc(&retprops, NV_UNIQUE_NAME, 0) != 0) {
472 (void) no_memory(hdl);
477 while ((elem = nvlist_next_nvpair(props, elem)) != NULL) {
478 const char *propname = nvpair_name(elem);
480 prop = zpool_name_to_prop(propname);
481 if (prop == ZPROP_INVAL && zpool_prop_feature(propname)) {
483 char *fname = strchr(propname, '@') + 1;
485 err = zfeature_lookup_name(fname, NULL);
487 ASSERT3U(err, ==, ENOENT);
488 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
489 "invalid feature '%s'"), fname);
490 (void) zfs_error(hdl, EZFS_BADPROP, errbuf);
494 if (nvpair_type(elem) != DATA_TYPE_STRING) {
495 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
496 "'%s' must be a string"), propname);
497 (void) zfs_error(hdl, EZFS_BADPROP, errbuf);
501 (void) nvpair_value_string(elem, &strval);
502 if (strcmp(strval, ZFS_FEATURE_ENABLED) != 0) {
503 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
504 "property '%s' can only be set to "
505 "'enabled'"), propname);
506 (void) zfs_error(hdl, EZFS_BADPROP, errbuf);
510 if (nvlist_add_uint64(retprops, propname, 0) != 0) {
511 (void) no_memory(hdl);
518 * Make sure this property is valid and applies to this type.
520 if (prop == ZPROP_INVAL) {
521 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
522 "invalid property '%s'"), propname);
523 (void) zfs_error(hdl, EZFS_BADPROP, errbuf);
527 if (zpool_prop_readonly(prop)) {
528 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "'%s' "
529 "is readonly"), propname);
530 (void) zfs_error(hdl, EZFS_PROPREADONLY, errbuf);
534 if (zprop_parse_value(hdl, elem, prop, ZFS_TYPE_POOL, retprops,
535 &strval, &intval, errbuf) != 0)
539 * Perform additional checking for specific properties.
544 case ZPOOL_PROP_VERSION:
545 if (intval < version ||
546 !SPA_VERSION_IS_SUPPORTED(intval)) {
547 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
548 "property '%s' number %d is invalid."),
550 (void) zfs_error(hdl, EZFS_BADVERSION, errbuf);
555 case ZPOOL_PROP_ASHIFT:
557 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
558 "property '%s' can only be set at "
559 "creation time"), propname);
560 (void) zfs_error(hdl, EZFS_BADPROP, errbuf);
564 if (intval != 0 && (intval < 9 || intval > 13)) {
565 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
566 "property '%s' number %d is invalid."),
568 (void) zfs_error(hdl, EZFS_BADPROP, errbuf);
573 case ZPOOL_PROP_BOOTFS:
574 if (flags.create || flags.import) {
575 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
576 "property '%s' cannot be set at creation "
577 "or import time"), propname);
578 (void) zfs_error(hdl, EZFS_BADPROP, errbuf);
582 if (version < SPA_VERSION_BOOTFS) {
583 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
584 "pool must be upgraded to support "
585 "'%s' property"), propname);
586 (void) zfs_error(hdl, EZFS_BADVERSION, errbuf);
591 * bootfs property value has to be a dataset name and
592 * the dataset has to be in the same pool as it sets to.
594 if (strval[0] != '\0' && !bootfs_name_valid(poolname,
596 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "'%s' "
597 "is an invalid name"), strval);
598 (void) zfs_error(hdl, EZFS_INVALIDNAME, errbuf);
602 if ((zhp = zpool_open_canfail(hdl, poolname)) == NULL) {
603 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
604 "could not open pool '%s'"), poolname);
605 (void) zfs_error(hdl, EZFS_OPENFAILED, errbuf);
608 verify(nvlist_lookup_nvlist(zpool_get_config(zhp, NULL),
609 ZPOOL_CONFIG_VDEV_TREE, &nvroot) == 0);
611 #if defined(__sun__) || defined(__sun)
613 * bootfs property cannot be set on a disk which has
616 if (pool_uses_efi(nvroot)) {
617 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
618 "property '%s' not supported on "
619 "EFI labeled devices"), propname);
620 (void) zfs_error(hdl, EZFS_POOL_NOTSUP, errbuf);
628 case ZPOOL_PROP_ALTROOT:
629 if (!flags.create && !flags.import) {
630 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
631 "property '%s' can only be set during pool "
632 "creation or import"), propname);
633 (void) zfs_error(hdl, EZFS_BADPROP, errbuf);
637 if (strval[0] != '/') {
638 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
639 "bad alternate root '%s'"), strval);
640 (void) zfs_error(hdl, EZFS_BADPATH, errbuf);
645 case ZPOOL_PROP_CACHEFILE:
646 if (strval[0] == '\0')
649 if (strcmp(strval, "none") == 0)
652 if (strval[0] != '/') {
653 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
654 "property '%s' must be empty, an "
655 "absolute path, or 'none'"), propname);
656 (void) zfs_error(hdl, EZFS_BADPATH, errbuf);
660 slash = strrchr(strval, '/');
662 if (slash[1] == '\0' || strcmp(slash, "/.") == 0 ||
663 strcmp(slash, "/..") == 0) {
664 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
665 "'%s' is not a valid file"), strval);
666 (void) zfs_error(hdl, EZFS_BADPATH, errbuf);
672 if (strval[0] != '\0' &&
673 (stat64(strval, &statbuf) != 0 ||
674 !S_ISDIR(statbuf.st_mode))) {
675 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
676 "'%s' is not a valid directory"),
678 (void) zfs_error(hdl, EZFS_BADPATH, errbuf);
685 case ZPOOL_PROP_COMMENT:
686 for (check = strval; *check != '\0'; check++) {
687 if (!isprint(*check)) {
689 dgettext(TEXT_DOMAIN,
690 "comment may only have printable "
692 (void) zfs_error(hdl, EZFS_BADPROP,
697 if (strlen(strval) > ZPROP_MAX_COMMENT) {
698 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
699 "comment must not exceed %d characters"),
701 (void) zfs_error(hdl, EZFS_BADPROP, errbuf);
705 case ZPOOL_PROP_READONLY:
707 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
708 "property '%s' can only be set at "
709 "import time"), propname);
710 (void) zfs_error(hdl, EZFS_BADPROP, errbuf);
714 case ZPOOL_PROP_TNAME:
716 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
717 "property '%s' can only be set at "
718 "creation time"), propname);
719 (void) zfs_error(hdl, EZFS_BADPROP, errbuf);
728 nvlist_free(retprops);
733 * Set zpool property : propname=propval.
736 zpool_set_prop(zpool_handle_t *zhp, const char *propname, const char *propval)
738 zfs_cmd_t zc = {"\0"};
741 nvlist_t *nvl = NULL;
744 prop_flags_t flags = { 0 };
746 (void) snprintf(errbuf, sizeof (errbuf),
747 dgettext(TEXT_DOMAIN, "cannot set property for '%s'"),
750 if (nvlist_alloc(&nvl, NV_UNIQUE_NAME, 0) != 0)
751 return (no_memory(zhp->zpool_hdl));
753 if (nvlist_add_string(nvl, propname, propval) != 0) {
755 return (no_memory(zhp->zpool_hdl));
758 version = zpool_get_prop_int(zhp, ZPOOL_PROP_VERSION, NULL);
759 if ((realprops = zpool_valid_proplist(zhp->zpool_hdl,
760 zhp->zpool_name, nvl, version, flags, errbuf)) == NULL) {
769 * Execute the corresponding ioctl() to set this property.
771 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
773 if (zcmd_write_src_nvlist(zhp->zpool_hdl, &zc, nvl) != 0) {
778 ret = zfs_ioctl(zhp->zpool_hdl, ZFS_IOC_POOL_SET_PROPS, &zc);
780 zcmd_free_nvlists(&zc);
784 (void) zpool_standard_error(zhp->zpool_hdl, errno, errbuf);
786 (void) zpool_props_refresh(zhp);
792 zpool_expand_proplist(zpool_handle_t *zhp, zprop_list_t **plp)
794 libzfs_handle_t *hdl = zhp->zpool_hdl;
796 char buf[ZFS_MAXPROPLEN];
797 nvlist_t *features = NULL;
800 boolean_t firstexpand = (NULL == *plp);
803 if (zprop_expand_list(hdl, plp, ZFS_TYPE_POOL) != 0)
807 while (*last != NULL)
808 last = &(*last)->pl_next;
811 features = zpool_get_features(zhp);
813 if ((*plp)->pl_all && firstexpand) {
814 for (i = 0; i < SPA_FEATURES; i++) {
815 zprop_list_t *entry = zfs_alloc(hdl,
816 sizeof (zprop_list_t));
817 entry->pl_prop = ZPROP_INVAL;
818 entry->pl_user_prop = zfs_asprintf(hdl, "feature@%s",
819 spa_feature_table[i].fi_uname);
820 entry->pl_width = strlen(entry->pl_user_prop);
821 entry->pl_all = B_TRUE;
824 last = &entry->pl_next;
828 /* add any unsupported features */
829 for (nvp = nvlist_next_nvpair(features, NULL);
830 nvp != NULL; nvp = nvlist_next_nvpair(features, nvp)) {
835 if (zfeature_is_supported(nvpair_name(nvp)))
838 propname = zfs_asprintf(hdl, "unsupported@%s",
842 * Before adding the property to the list make sure that no
843 * other pool already added the same property.
847 while (entry != NULL) {
848 if (entry->pl_user_prop != NULL &&
849 strcmp(propname, entry->pl_user_prop) == 0) {
853 entry = entry->pl_next;
860 entry = zfs_alloc(hdl, sizeof (zprop_list_t));
861 entry->pl_prop = ZPROP_INVAL;
862 entry->pl_user_prop = propname;
863 entry->pl_width = strlen(entry->pl_user_prop);
864 entry->pl_all = B_TRUE;
867 last = &entry->pl_next;
870 for (entry = *plp; entry != NULL; entry = entry->pl_next) {
875 if (entry->pl_prop != ZPROP_INVAL &&
876 zpool_get_prop(zhp, entry->pl_prop, buf, sizeof (buf),
877 NULL, B_FALSE) == 0) {
878 if (strlen(buf) > entry->pl_width)
879 entry->pl_width = strlen(buf);
887 * Get the state for the given feature on the given ZFS pool.
890 zpool_prop_get_feature(zpool_handle_t *zhp, const char *propname, char *buf,
894 boolean_t found = B_FALSE;
895 nvlist_t *features = zpool_get_features(zhp);
897 const char *feature = strchr(propname, '@') + 1;
899 supported = zpool_prop_feature(propname);
900 ASSERT(supported || zpool_prop_unsupported(propname));
903 * Convert from feature name to feature guid. This conversion is
904 * unecessary for unsupported@... properties because they already
911 ret = zfeature_lookup_name(feature, &fid);
913 (void) strlcpy(buf, "-", len);
916 feature = spa_feature_table[fid].fi_guid;
919 if (nvlist_lookup_uint64(features, feature, &refcount) == 0)
924 (void) strlcpy(buf, ZFS_FEATURE_DISABLED, len);
927 (void) strlcpy(buf, ZFS_FEATURE_ENABLED, len);
929 (void) strlcpy(buf, ZFS_FEATURE_ACTIVE, len);
934 (void) strcpy(buf, ZFS_UNSUPPORTED_INACTIVE);
936 (void) strcpy(buf, ZFS_UNSUPPORTED_READONLY);
939 (void) strlcpy(buf, "-", len);
948 * Don't start the slice at the default block of 34; many storage
949 * devices will use a stripe width of 128k, other vendors prefer a 1m
950 * alignment. It is best to play it safe and ensure a 1m alignment
951 * given 512B blocks. When the block size is larger by a power of 2
952 * we will still be 1m aligned. Some devices are sensitive to the
953 * partition ending alignment as well.
955 #define NEW_START_BLOCK 2048
956 #define PARTITION_END_ALIGNMENT 2048
959 * Validate the given pool name, optionally putting an extended error message in
963 zpool_name_valid(libzfs_handle_t *hdl, boolean_t isopen, const char *pool)
969 ret = pool_namecheck(pool, &why, &what);
972 * The rules for reserved pool names were extended at a later point.
973 * But we need to support users with existing pools that may now be
974 * invalid. So we only check for this expanded set of names during a
975 * create (or import), and only in userland.
977 if (ret == 0 && !isopen &&
978 (strncmp(pool, "mirror", 6) == 0 ||
979 strncmp(pool, "raidz", 5) == 0 ||
980 strncmp(pool, "spare", 5) == 0 ||
981 strcmp(pool, "log") == 0)) {
984 dgettext(TEXT_DOMAIN, "name is reserved"));
992 case NAME_ERR_TOOLONG:
994 dgettext(TEXT_DOMAIN, "name is too long"));
997 case NAME_ERR_INVALCHAR:
999 dgettext(TEXT_DOMAIN, "invalid character "
1000 "'%c' in pool name"), what);
1003 case NAME_ERR_NOLETTER:
1004 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1005 "name must begin with a letter"));
1008 case NAME_ERR_RESERVED:
1009 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1010 "name is reserved"));
1013 case NAME_ERR_DISKLIKE:
1014 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1015 "pool name is reserved"));
1018 case NAME_ERR_LEADING_SLASH:
1019 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1020 "leading slash in name"));
1023 case NAME_ERR_EMPTY_COMPONENT:
1024 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1025 "empty component in name"));
1028 case NAME_ERR_TRAILING_SLASH:
1029 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1030 "trailing slash in name"));
1033 case NAME_ERR_MULTIPLE_AT:
1034 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1035 "multiple '@' delimiters in name"));
1037 case NAME_ERR_NO_AT:
1038 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1039 "permission set is missing '@'"));
1050 * Open a handle to the given pool, even if the pool is currently in the FAULTED
1054 zpool_open_canfail(libzfs_handle_t *hdl, const char *pool)
1056 zpool_handle_t *zhp;
1060 * Make sure the pool name is valid.
1062 if (!zpool_name_valid(hdl, B_TRUE, pool)) {
1063 (void) zfs_error_fmt(hdl, EZFS_INVALIDNAME,
1064 dgettext(TEXT_DOMAIN, "cannot open '%s'"),
1069 if ((zhp = zfs_alloc(hdl, sizeof (zpool_handle_t))) == NULL)
1072 zhp->zpool_hdl = hdl;
1073 (void) strlcpy(zhp->zpool_name, pool, sizeof (zhp->zpool_name));
1075 if (zpool_refresh_stats(zhp, &missing) != 0) {
1081 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "no such pool"));
1082 (void) zfs_error_fmt(hdl, EZFS_NOENT,
1083 dgettext(TEXT_DOMAIN, "cannot open '%s'"), pool);
1092 * Like the above, but silent on error. Used when iterating over pools (because
1093 * the configuration cache may be out of date).
1096 zpool_open_silent(libzfs_handle_t *hdl, const char *pool, zpool_handle_t **ret)
1098 zpool_handle_t *zhp;
1101 if ((zhp = zfs_alloc(hdl, sizeof (zpool_handle_t))) == NULL)
1104 zhp->zpool_hdl = hdl;
1105 (void) strlcpy(zhp->zpool_name, pool, sizeof (zhp->zpool_name));
1107 if (zpool_refresh_stats(zhp, &missing) != 0) {
1123 * Similar to zpool_open_canfail(), but refuses to open pools in the faulted
1127 zpool_open(libzfs_handle_t *hdl, const char *pool)
1129 zpool_handle_t *zhp;
1131 if ((zhp = zpool_open_canfail(hdl, pool)) == NULL)
1134 if (zhp->zpool_state == POOL_STATE_UNAVAIL) {
1135 (void) zfs_error_fmt(hdl, EZFS_POOLUNAVAIL,
1136 dgettext(TEXT_DOMAIN, "cannot open '%s'"), zhp->zpool_name);
1145 * Close the handle. Simply frees the memory associated with the handle.
1148 zpool_close(zpool_handle_t *zhp)
1150 nvlist_free(zhp->zpool_config);
1151 nvlist_free(zhp->zpool_old_config);
1152 nvlist_free(zhp->zpool_props);
1157 * Return the name of the pool.
1160 zpool_get_name(zpool_handle_t *zhp)
1162 return (zhp->zpool_name);
1167 * Return the state of the pool (ACTIVE or UNAVAILABLE)
1170 zpool_get_state(zpool_handle_t *zhp)
1172 return (zhp->zpool_state);
1176 * Create the named pool, using the provided vdev list. It is assumed
1177 * that the consumer has already validated the contents of the nvlist, so we
1178 * don't have to worry about error semantics.
1181 zpool_create(libzfs_handle_t *hdl, const char *pool, nvlist_t *nvroot,
1182 nvlist_t *props, nvlist_t *fsprops)
1184 zfs_cmd_t zc = {"\0"};
1185 nvlist_t *zc_fsprops = NULL;
1186 nvlist_t *zc_props = NULL;
1190 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN,
1191 "cannot create '%s'"), pool);
1193 if (!zpool_name_valid(hdl, B_FALSE, pool))
1194 return (zfs_error(hdl, EZFS_INVALIDNAME, msg));
1196 if (zcmd_write_conf_nvlist(hdl, &zc, nvroot) != 0)
1200 prop_flags_t flags = { .create = B_TRUE, .import = B_FALSE };
1202 if ((zc_props = zpool_valid_proplist(hdl, pool, props,
1203 SPA_VERSION_1, flags, msg)) == NULL) {
1212 zoned = ((nvlist_lookup_string(fsprops,
1213 zfs_prop_to_name(ZFS_PROP_ZONED), &zonestr) == 0) &&
1214 strcmp(zonestr, "on") == 0);
1216 if ((zc_fsprops = zfs_valid_proplist(hdl, ZFS_TYPE_FILESYSTEM,
1217 fsprops, zoned, NULL, NULL, msg)) == NULL) {
1221 (nvlist_alloc(&zc_props, NV_UNIQUE_NAME, 0) != 0)) {
1224 if (nvlist_add_nvlist(zc_props,
1225 ZPOOL_ROOTFS_PROPS, zc_fsprops) != 0) {
1230 if (zc_props && zcmd_write_src_nvlist(hdl, &zc, zc_props) != 0)
1233 (void) strlcpy(zc.zc_name, pool, sizeof (zc.zc_name));
1235 if ((ret = zfs_ioctl(hdl, ZFS_IOC_POOL_CREATE, &zc)) != 0) {
1237 zcmd_free_nvlists(&zc);
1238 nvlist_free(zc_props);
1239 nvlist_free(zc_fsprops);
1244 * This can happen if the user has specified the same
1245 * device multiple times. We can't reliably detect this
1246 * until we try to add it and see we already have a
1247 * label. This can also happen under if the device is
1248 * part of an active md or lvm device.
1250 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1251 "one or more vdevs refer to the same device, or "
1252 "one of\nthe devices is part of an active md or "
1254 return (zfs_error(hdl, EZFS_BADDEV, msg));
1258 * This happens if the record size is smaller or larger
1259 * than the allowed size range, or not a power of 2.
1261 * NOTE: although zfs_valid_proplist is called earlier,
1262 * this case may have slipped through since the
1263 * pool does not exist yet and it is therefore
1264 * impossible to read properties e.g. max blocksize
1267 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1268 "record size invalid"));
1269 return (zfs_error(hdl, EZFS_BADPROP, msg));
1273 * This occurs when one of the devices is below
1274 * SPA_MINDEVSIZE. Unfortunately, we can't detect which
1275 * device was the problem device since there's no
1276 * reliable way to determine device size from userland.
1281 zfs_nicenum(SPA_MINDEVSIZE, buf, sizeof (buf));
1283 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1284 "one or more devices is less than the "
1285 "minimum size (%s)"), buf);
1287 return (zfs_error(hdl, EZFS_BADDEV, msg));
1290 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1291 "one or more devices is out of space"));
1292 return (zfs_error(hdl, EZFS_BADDEV, msg));
1295 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1296 "cache device must be a disk or disk slice"));
1297 return (zfs_error(hdl, EZFS_BADDEV, msg));
1300 return (zpool_standard_error(hdl, errno, msg));
1305 zcmd_free_nvlists(&zc);
1306 nvlist_free(zc_props);
1307 nvlist_free(zc_fsprops);
1312 * Destroy the given pool. It is up to the caller to ensure that there are no
1313 * datasets left in the pool.
1316 zpool_destroy(zpool_handle_t *zhp, const char *log_str)
1318 zfs_cmd_t zc = {"\0"};
1319 zfs_handle_t *zfp = NULL;
1320 libzfs_handle_t *hdl = zhp->zpool_hdl;
1323 if (zhp->zpool_state == POOL_STATE_ACTIVE &&
1324 (zfp = zfs_open(hdl, zhp->zpool_name, ZFS_TYPE_FILESYSTEM)) == NULL)
1327 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
1328 zc.zc_history = (uint64_t)(uintptr_t)log_str;
1330 if (zfs_ioctl(hdl, ZFS_IOC_POOL_DESTROY, &zc) != 0) {
1331 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN,
1332 "cannot destroy '%s'"), zhp->zpool_name);
1334 if (errno == EROFS) {
1335 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1336 "one or more devices is read only"));
1337 (void) zfs_error(hdl, EZFS_BADDEV, msg);
1339 (void) zpool_standard_error(hdl, errno, msg);
1348 remove_mountpoint(zfp);
1356 * Add the given vdevs to the pool. The caller must have already performed the
1357 * necessary verification to ensure that the vdev specification is well-formed.
1360 zpool_add(zpool_handle_t *zhp, nvlist_t *nvroot)
1362 zfs_cmd_t zc = {"\0"};
1364 libzfs_handle_t *hdl = zhp->zpool_hdl;
1366 nvlist_t **spares, **l2cache;
1367 uint_t nspares, nl2cache;
1369 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN,
1370 "cannot add to '%s'"), zhp->zpool_name);
1372 if (zpool_get_prop_int(zhp, ZPOOL_PROP_VERSION, NULL) <
1373 SPA_VERSION_SPARES &&
1374 nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_SPARES,
1375 &spares, &nspares) == 0) {
1376 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "pool must be "
1377 "upgraded to add hot spares"));
1378 return (zfs_error(hdl, EZFS_BADVERSION, msg));
1381 #if defined(__sun__) || defined(__sun)
1382 if (zpool_is_bootable(zhp) && nvlist_lookup_nvlist_array(nvroot,
1383 ZPOOL_CONFIG_SPARES, &spares, &nspares) == 0) {
1386 for (s = 0; s < nspares; s++) {
1389 if (nvlist_lookup_string(spares[s], ZPOOL_CONFIG_PATH,
1390 &path) == 0 && pool_uses_efi(spares[s])) {
1391 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1392 "device '%s' contains an EFI label and "
1393 "cannot be used on root pools."),
1394 zpool_vdev_name(hdl, NULL, spares[s], 0));
1395 return (zfs_error(hdl, EZFS_POOL_NOTSUP, msg));
1401 if (zpool_get_prop_int(zhp, ZPOOL_PROP_VERSION, NULL) <
1402 SPA_VERSION_L2CACHE &&
1403 nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_L2CACHE,
1404 &l2cache, &nl2cache) == 0) {
1405 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "pool must be "
1406 "upgraded to add cache devices"));
1407 return (zfs_error(hdl, EZFS_BADVERSION, msg));
1410 if (zcmd_write_conf_nvlist(hdl, &zc, nvroot) != 0)
1412 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
1414 if (zfs_ioctl(hdl, ZFS_IOC_VDEV_ADD, &zc) != 0) {
1418 * This can happen if the user has specified the same
1419 * device multiple times. We can't reliably detect this
1420 * until we try to add it and see we already have a
1423 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1424 "one or more vdevs refer to the same device"));
1425 (void) zfs_error(hdl, EZFS_BADDEV, msg);
1430 * This occurrs when one of the devices is below
1431 * SPA_MINDEVSIZE. Unfortunately, we can't detect which
1432 * device was the problem device since there's no
1433 * reliable way to determine device size from userland.
1438 zfs_nicenum(SPA_MINDEVSIZE, buf, sizeof (buf));
1440 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1441 "device is less than the minimum "
1444 (void) zfs_error(hdl, EZFS_BADDEV, msg);
1448 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1449 "pool must be upgraded to add these vdevs"));
1450 (void) zfs_error(hdl, EZFS_BADVERSION, msg);
1454 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1455 "cache device must be a disk or disk slice"));
1456 (void) zfs_error(hdl, EZFS_BADDEV, msg);
1460 (void) zpool_standard_error(hdl, errno, msg);
1468 zcmd_free_nvlists(&zc);
1474 * Exports the pool from the system. The caller must ensure that there are no
1475 * mounted datasets in the pool.
1478 zpool_export_common(zpool_handle_t *zhp, boolean_t force, boolean_t hardforce,
1479 const char *log_str)
1481 zfs_cmd_t zc = {"\0"};
1484 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN,
1485 "cannot export '%s'"), zhp->zpool_name);
1487 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
1488 zc.zc_cookie = force;
1489 zc.zc_guid = hardforce;
1490 zc.zc_history = (uint64_t)(uintptr_t)log_str;
1492 if (zfs_ioctl(zhp->zpool_hdl, ZFS_IOC_POOL_EXPORT, &zc) != 0) {
1495 zfs_error_aux(zhp->zpool_hdl, dgettext(TEXT_DOMAIN,
1496 "use '-f' to override the following errors:\n"
1497 "'%s' has an active shared spare which could be"
1498 " used by other pools once '%s' is exported."),
1499 zhp->zpool_name, zhp->zpool_name);
1500 return (zfs_error(zhp->zpool_hdl, EZFS_ACTIVE_SPARE,
1503 return (zpool_standard_error_fmt(zhp->zpool_hdl, errno,
1512 zpool_export(zpool_handle_t *zhp, boolean_t force, const char *log_str)
1514 return (zpool_export_common(zhp, force, B_FALSE, log_str));
1518 zpool_export_force(zpool_handle_t *zhp, const char *log_str)
1520 return (zpool_export_common(zhp, B_TRUE, B_TRUE, log_str));
1524 zpool_rewind_exclaim(libzfs_handle_t *hdl, const char *name, boolean_t dryrun,
1527 nvlist_t *nv = NULL;
1533 if (!hdl->libzfs_printerr || config == NULL)
1536 if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_LOAD_INFO, &nv) != 0 ||
1537 nvlist_lookup_nvlist(nv, ZPOOL_CONFIG_REWIND_INFO, &nv) != 0) {
1541 if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_LOAD_TIME, &rewindto) != 0)
1543 (void) nvlist_lookup_int64(nv, ZPOOL_CONFIG_REWIND_TIME, &loss);
1545 if (localtime_r((time_t *)&rewindto, &t) != NULL &&
1546 strftime(timestr, 128, "%c", &t) != 0) {
1548 (void) printf(dgettext(TEXT_DOMAIN,
1549 "Would be able to return %s "
1550 "to its state as of %s.\n"),
1553 (void) printf(dgettext(TEXT_DOMAIN,
1554 "Pool %s returned to its state as of %s.\n"),
1558 (void) printf(dgettext(TEXT_DOMAIN,
1559 "%s approximately %lld "),
1560 dryrun ? "Would discard" : "Discarded",
1561 ((longlong_t)loss + 30) / 60);
1562 (void) printf(dgettext(TEXT_DOMAIN,
1563 "minutes of transactions.\n"));
1564 } else if (loss > 0) {
1565 (void) printf(dgettext(TEXT_DOMAIN,
1566 "%s approximately %lld "),
1567 dryrun ? "Would discard" : "Discarded",
1569 (void) printf(dgettext(TEXT_DOMAIN,
1570 "seconds of transactions.\n"));
1576 zpool_explain_recover(libzfs_handle_t *hdl, const char *name, int reason,
1579 nvlist_t *nv = NULL;
1581 uint64_t edata = UINT64_MAX;
1586 if (!hdl->libzfs_printerr)
1590 (void) printf(dgettext(TEXT_DOMAIN, "action: "));
1592 (void) printf(dgettext(TEXT_DOMAIN, "\t"));
1594 /* All attempted rewinds failed if ZPOOL_CONFIG_LOAD_TIME missing */
1595 if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_LOAD_INFO, &nv) != 0 ||
1596 nvlist_lookup_nvlist(nv, ZPOOL_CONFIG_REWIND_INFO, &nv) != 0 ||
1597 nvlist_lookup_uint64(nv, ZPOOL_CONFIG_LOAD_TIME, &rewindto) != 0)
1600 (void) nvlist_lookup_int64(nv, ZPOOL_CONFIG_REWIND_TIME, &loss);
1601 (void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_LOAD_DATA_ERRORS,
1604 (void) printf(dgettext(TEXT_DOMAIN,
1605 "Recovery is possible, but will result in some data loss.\n"));
1607 if (localtime_r((time_t *)&rewindto, &t) != NULL &&
1608 strftime(timestr, 128, "%c", &t) != 0) {
1609 (void) printf(dgettext(TEXT_DOMAIN,
1610 "\tReturning the pool to its state as of %s\n"
1611 "\tshould correct the problem. "),
1614 (void) printf(dgettext(TEXT_DOMAIN,
1615 "\tReverting the pool to an earlier state "
1616 "should correct the problem.\n\t"));
1620 (void) printf(dgettext(TEXT_DOMAIN,
1621 "Approximately %lld minutes of data\n"
1622 "\tmust be discarded, irreversibly. "),
1623 ((longlong_t)loss + 30) / 60);
1624 } else if (loss > 0) {
1625 (void) printf(dgettext(TEXT_DOMAIN,
1626 "Approximately %lld seconds of data\n"
1627 "\tmust be discarded, irreversibly. "),
1630 if (edata != 0 && edata != UINT64_MAX) {
1632 (void) printf(dgettext(TEXT_DOMAIN,
1633 "After rewind, at least\n"
1634 "\tone persistent user-data error will remain. "));
1636 (void) printf(dgettext(TEXT_DOMAIN,
1637 "After rewind, several\n"
1638 "\tpersistent user-data errors will remain. "));
1641 (void) printf(dgettext(TEXT_DOMAIN,
1642 "Recovery can be attempted\n\tby executing 'zpool %s -F %s'. "),
1643 reason >= 0 ? "clear" : "import", name);
1645 (void) printf(dgettext(TEXT_DOMAIN,
1646 "A scrub of the pool\n"
1647 "\tis strongly recommended after recovery.\n"));
1651 (void) printf(dgettext(TEXT_DOMAIN,
1652 "Destroy and re-create the pool from\n\ta backup source.\n"));
1656 * zpool_import() is a contracted interface. Should be kept the same
1659 * Applications should use zpool_import_props() to import a pool with
1660 * new properties value to be set.
1663 zpool_import(libzfs_handle_t *hdl, nvlist_t *config, const char *newname,
1666 nvlist_t *props = NULL;
1669 if (altroot != NULL) {
1670 if (nvlist_alloc(&props, NV_UNIQUE_NAME, 0) != 0) {
1671 return (zfs_error_fmt(hdl, EZFS_NOMEM,
1672 dgettext(TEXT_DOMAIN, "cannot import '%s'"),
1676 if (nvlist_add_string(props,
1677 zpool_prop_to_name(ZPOOL_PROP_ALTROOT), altroot) != 0 ||
1678 nvlist_add_string(props,
1679 zpool_prop_to_name(ZPOOL_PROP_CACHEFILE), "none") != 0) {
1681 return (zfs_error_fmt(hdl, EZFS_NOMEM,
1682 dgettext(TEXT_DOMAIN, "cannot import '%s'"),
1687 ret = zpool_import_props(hdl, config, newname, props,
1694 print_vdev_tree(libzfs_handle_t *hdl, const char *name, nvlist_t *nv,
1700 uint64_t is_log = 0;
1702 (void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_IS_LOG,
1706 (void) printf("\t%*s%s%s\n", indent, "", name,
1707 is_log ? " [log]" : "");
1709 if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN,
1710 &child, &children) != 0)
1713 for (c = 0; c < children; c++) {
1714 vname = zpool_vdev_name(hdl, NULL, child[c], VDEV_NAME_TYPE_ID);
1715 print_vdev_tree(hdl, vname, child[c], indent + 2);
1721 zpool_print_unsup_feat(nvlist_t *config)
1723 nvlist_t *nvinfo, *unsup_feat;
1726 verify(nvlist_lookup_nvlist(config, ZPOOL_CONFIG_LOAD_INFO, &nvinfo) ==
1728 verify(nvlist_lookup_nvlist(nvinfo, ZPOOL_CONFIG_UNSUP_FEAT,
1731 for (nvp = nvlist_next_nvpair(unsup_feat, NULL); nvp != NULL;
1732 nvp = nvlist_next_nvpair(unsup_feat, nvp)) {
1735 verify(nvpair_type(nvp) == DATA_TYPE_STRING);
1736 verify(nvpair_value_string(nvp, &desc) == 0);
1738 if (strlen(desc) > 0)
1739 (void) printf("\t%s (%s)\n", nvpair_name(nvp), desc);
1741 (void) printf("\t%s\n", nvpair_name(nvp));
1746 * Import the given pool using the known configuration and a list of
1747 * properties to be set. The configuration should have come from
1748 * zpool_find_import(). The 'newname' parameters control whether the pool
1749 * is imported with a different name.
1752 zpool_import_props(libzfs_handle_t *hdl, nvlist_t *config, const char *newname,
1753 nvlist_t *props, int flags)
1755 zfs_cmd_t zc = {"\0"};
1756 zpool_rewind_policy_t policy;
1757 nvlist_t *nv = NULL;
1758 nvlist_t *nvinfo = NULL;
1759 nvlist_t *missing = NULL;
1766 verify(nvlist_lookup_string(config, ZPOOL_CONFIG_POOL_NAME,
1769 (void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN,
1770 "cannot import pool '%s'"), origname);
1772 if (newname != NULL) {
1773 if (!zpool_name_valid(hdl, B_FALSE, newname))
1774 return (zfs_error_fmt(hdl, EZFS_INVALIDNAME,
1775 dgettext(TEXT_DOMAIN, "cannot import '%s'"),
1777 thename = (char *)newname;
1782 if (props != NULL) {
1784 prop_flags_t flags = { .create = B_FALSE, .import = B_TRUE };
1786 verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_VERSION,
1789 if ((props = zpool_valid_proplist(hdl, origname,
1790 props, version, flags, errbuf)) == NULL)
1792 if (zcmd_write_src_nvlist(hdl, &zc, props) != 0) {
1799 (void) strlcpy(zc.zc_name, thename, sizeof (zc.zc_name));
1801 verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID,
1804 if (zcmd_write_conf_nvlist(hdl, &zc, config) != 0) {
1805 zcmd_free_nvlists(&zc);
1808 if (zcmd_alloc_dst_nvlist(hdl, &zc, zc.zc_nvlist_conf_size * 2) != 0) {
1809 zcmd_free_nvlists(&zc);
1813 zc.zc_cookie = flags;
1814 while ((ret = zfs_ioctl(hdl, ZFS_IOC_POOL_IMPORT, &zc)) != 0 &&
1816 if (zcmd_expand_dst_nvlist(hdl, &zc) != 0) {
1817 zcmd_free_nvlists(&zc);
1824 (void) zcmd_read_dst_nvlist(hdl, &zc, &nv);
1826 zcmd_free_nvlists(&zc);
1828 zpool_get_rewind_policy(config, &policy);
1834 * Dry-run failed, but we print out what success
1835 * looks like if we found a best txg
1837 if (policy.zrp_request & ZPOOL_TRY_REWIND) {
1838 zpool_rewind_exclaim(hdl, newname ? origname : thename,
1844 if (newname == NULL)
1845 (void) snprintf(desc, sizeof (desc),
1846 dgettext(TEXT_DOMAIN, "cannot import '%s'"),
1849 (void) snprintf(desc, sizeof (desc),
1850 dgettext(TEXT_DOMAIN, "cannot import '%s' as '%s'"),
1855 if (nv != NULL && nvlist_lookup_nvlist(nv,
1856 ZPOOL_CONFIG_LOAD_INFO, &nvinfo) == 0 &&
1857 nvlist_exists(nvinfo, ZPOOL_CONFIG_UNSUP_FEAT)) {
1858 (void) printf(dgettext(TEXT_DOMAIN, "This "
1859 "pool uses the following feature(s) not "
1860 "supported by this system:\n"));
1861 zpool_print_unsup_feat(nv);
1862 if (nvlist_exists(nvinfo,
1863 ZPOOL_CONFIG_CAN_RDONLY)) {
1864 (void) printf(dgettext(TEXT_DOMAIN,
1865 "All unsupported features are only "
1866 "required for writing to the pool."
1867 "\nThe pool can be imported using "
1868 "'-o readonly=on'.\n"));
1872 * Unsupported version.
1874 (void) zfs_error(hdl, EZFS_BADVERSION, desc);
1878 (void) zfs_error(hdl, EZFS_INVALCONFIG, desc);
1882 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1883 "one or more devices is read only"));
1884 (void) zfs_error(hdl, EZFS_BADDEV, desc);
1888 if (nv && nvlist_lookup_nvlist(nv,
1889 ZPOOL_CONFIG_LOAD_INFO, &nvinfo) == 0 &&
1890 nvlist_lookup_nvlist(nvinfo,
1891 ZPOOL_CONFIG_MISSING_DEVICES, &missing) == 0) {
1892 (void) printf(dgettext(TEXT_DOMAIN,
1893 "The devices below are missing, use "
1894 "'-m' to import the pool anyway:\n"));
1895 print_vdev_tree(hdl, NULL, missing, 2);
1896 (void) printf("\n");
1898 (void) zpool_standard_error(hdl, error, desc);
1902 (void) zpool_standard_error(hdl, error, desc);
1906 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1907 "one or more devices are already in use\n"));
1908 (void) zfs_error(hdl, EZFS_BADDEV, desc);
1911 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1912 "new name of at least one dataset is longer than "
1913 "the maximum allowable length"));
1914 (void) zfs_error(hdl, EZFS_NAMETOOLONG, desc);
1917 (void) zpool_standard_error(hdl, error, desc);
1918 zpool_explain_recover(hdl,
1919 newname ? origname : thename, -error, nv);
1926 zpool_handle_t *zhp;
1929 * This should never fail, but play it safe anyway.
1931 if (zpool_open_silent(hdl, thename, &zhp) != 0)
1933 else if (zhp != NULL)
1935 if (policy.zrp_request &
1936 (ZPOOL_DO_REWIND | ZPOOL_TRY_REWIND)) {
1937 zpool_rewind_exclaim(hdl, newname ? origname : thename,
1938 ((policy.zrp_request & ZPOOL_TRY_REWIND) != 0), nv);
1951 zpool_scan(zpool_handle_t *zhp, pool_scan_func_t func)
1953 zfs_cmd_t zc = {"\0"};
1955 libzfs_handle_t *hdl = zhp->zpool_hdl;
1957 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
1958 zc.zc_cookie = func;
1960 if (zfs_ioctl(hdl, ZFS_IOC_POOL_SCAN, &zc) == 0 ||
1961 (errno == ENOENT && func != POOL_SCAN_NONE))
1964 if (func == POOL_SCAN_SCRUB) {
1965 (void) snprintf(msg, sizeof (msg),
1966 dgettext(TEXT_DOMAIN, "cannot scrub %s"), zc.zc_name);
1967 } else if (func == POOL_SCAN_NONE) {
1968 (void) snprintf(msg, sizeof (msg),
1969 dgettext(TEXT_DOMAIN, "cannot cancel scrubbing %s"),
1972 assert(!"unexpected result");
1975 if (errno == EBUSY) {
1977 pool_scan_stat_t *ps = NULL;
1980 verify(nvlist_lookup_nvlist(zhp->zpool_config,
1981 ZPOOL_CONFIG_VDEV_TREE, &nvroot) == 0);
1982 (void) nvlist_lookup_uint64_array(nvroot,
1983 ZPOOL_CONFIG_SCAN_STATS, (uint64_t **)&ps, &psc);
1984 if (ps && ps->pss_func == POOL_SCAN_SCRUB)
1985 return (zfs_error(hdl, EZFS_SCRUBBING, msg));
1987 return (zfs_error(hdl, EZFS_RESILVERING, msg));
1988 } else if (errno == ENOENT) {
1989 return (zfs_error(hdl, EZFS_NO_SCRUB, msg));
1991 return (zpool_standard_error(hdl, errno, msg));
1996 * Find a vdev that matches the search criteria specified. We use the
1997 * the nvpair name to determine how we should look for the device.
1998 * 'avail_spare' is set to TRUE if the provided guid refers to an AVAIL
1999 * spare; but FALSE if its an INUSE spare.
2002 vdev_to_nvlist_iter(nvlist_t *nv, nvlist_t *search, boolean_t *avail_spare,
2003 boolean_t *l2cache, boolean_t *log)
2010 nvpair_t *pair = nvlist_next_nvpair(search, NULL);
2012 /* Nothing to look for */
2013 if (search == NULL || pair == NULL)
2016 /* Obtain the key we will use to search */
2017 srchkey = nvpair_name(pair);
2019 switch (nvpair_type(pair)) {
2020 case DATA_TYPE_UINT64:
2021 if (strcmp(srchkey, ZPOOL_CONFIG_GUID) == 0) {
2022 uint64_t srchval, theguid;
2024 verify(nvpair_value_uint64(pair, &srchval) == 0);
2025 verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID,
2027 if (theguid == srchval)
2032 case DATA_TYPE_STRING: {
2033 char *srchval, *val;
2035 verify(nvpair_value_string(pair, &srchval) == 0);
2036 if (nvlist_lookup_string(nv, srchkey, &val) != 0)
2040 * Search for the requested value. Special cases:
2042 * - ZPOOL_CONFIG_PATH for whole disk entries. These end in
2043 * "-part1", or "p1". The suffix is hidden from the user,
2044 * but included in the string, so this matches around it.
2045 * - ZPOOL_CONFIG_PATH for short names zfs_strcmp_shortname()
2046 * is used to check all possible expanded paths.
2047 * - looking for a top-level vdev name (i.e. ZPOOL_CONFIG_TYPE).
2049 * Otherwise, all other searches are simple string compares.
2051 if (strcmp(srchkey, ZPOOL_CONFIG_PATH) == 0) {
2052 uint64_t wholedisk = 0;
2054 (void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_WHOLE_DISK,
2056 if (zfs_strcmp_pathname(srchval, val, wholedisk) == 0)
2059 } else if (strcmp(srchkey, ZPOOL_CONFIG_TYPE) == 0 && val) {
2060 char *type, *idx, *end, *p;
2061 uint64_t id, vdev_id;
2064 * Determine our vdev type, keeping in mind
2065 * that the srchval is composed of a type and
2066 * vdev id pair (i.e. mirror-4).
2068 if ((type = strdup(srchval)) == NULL)
2071 if ((p = strrchr(type, '-')) == NULL) {
2079 * If the types don't match then keep looking.
2081 if (strncmp(val, type, strlen(val)) != 0) {
2086 verify(strncmp(type, VDEV_TYPE_RAIDZ,
2087 strlen(VDEV_TYPE_RAIDZ)) == 0 ||
2088 strncmp(type, VDEV_TYPE_MIRROR,
2089 strlen(VDEV_TYPE_MIRROR)) == 0);
2090 verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_ID,
2094 vdev_id = strtoull(idx, &end, 10);
2101 * Now verify that we have the correct vdev id.
2110 if (strcmp(srchval, val) == 0)
2119 if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN,
2120 &child, &children) != 0)
2123 for (c = 0; c < children; c++) {
2124 if ((ret = vdev_to_nvlist_iter(child[c], search,
2125 avail_spare, l2cache, NULL)) != NULL) {
2127 * The 'is_log' value is only set for the toplevel
2128 * vdev, not the leaf vdevs. So we always lookup the
2129 * log device from the root of the vdev tree (where
2130 * 'log' is non-NULL).
2133 nvlist_lookup_uint64(child[c],
2134 ZPOOL_CONFIG_IS_LOG, &is_log) == 0 &&
2142 if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_SPARES,
2143 &child, &children) == 0) {
2144 for (c = 0; c < children; c++) {
2145 if ((ret = vdev_to_nvlist_iter(child[c], search,
2146 avail_spare, l2cache, NULL)) != NULL) {
2147 *avail_spare = B_TRUE;
2153 if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_L2CACHE,
2154 &child, &children) == 0) {
2155 for (c = 0; c < children; c++) {
2156 if ((ret = vdev_to_nvlist_iter(child[c], search,
2157 avail_spare, l2cache, NULL)) != NULL) {
2168 * Given a physical path (minus the "/devices" prefix), find the
2172 zpool_find_vdev_by_physpath(zpool_handle_t *zhp, const char *ppath,
2173 boolean_t *avail_spare, boolean_t *l2cache, boolean_t *log)
2175 nvlist_t *search, *nvroot, *ret;
2177 verify(nvlist_alloc(&search, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2178 verify(nvlist_add_string(search, ZPOOL_CONFIG_PHYS_PATH, ppath) == 0);
2180 verify(nvlist_lookup_nvlist(zhp->zpool_config, ZPOOL_CONFIG_VDEV_TREE,
2183 *avail_spare = B_FALSE;
2187 ret = vdev_to_nvlist_iter(nvroot, search, avail_spare, l2cache, log);
2188 nvlist_free(search);
2194 * Determine if we have an "interior" top-level vdev (i.e mirror/raidz).
2197 zpool_vdev_is_interior(const char *name)
2199 if (strncmp(name, VDEV_TYPE_RAIDZ, strlen(VDEV_TYPE_RAIDZ)) == 0 ||
2200 strncmp(name, VDEV_TYPE_MIRROR, strlen(VDEV_TYPE_MIRROR)) == 0)
2206 zpool_find_vdev(zpool_handle_t *zhp, const char *path, boolean_t *avail_spare,
2207 boolean_t *l2cache, boolean_t *log)
2210 nvlist_t *nvroot, *search, *ret;
2213 verify(nvlist_alloc(&search, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2215 guid = strtoull(path, &end, 0);
2216 if (guid != 0 && *end == '\0') {
2217 verify(nvlist_add_uint64(search, ZPOOL_CONFIG_GUID, guid) == 0);
2218 } else if (zpool_vdev_is_interior(path)) {
2219 verify(nvlist_add_string(search, ZPOOL_CONFIG_TYPE, path) == 0);
2221 verify(nvlist_add_string(search, ZPOOL_CONFIG_PATH, path) == 0);
2224 verify(nvlist_lookup_nvlist(zhp->zpool_config, ZPOOL_CONFIG_VDEV_TREE,
2227 *avail_spare = B_FALSE;
2231 ret = vdev_to_nvlist_iter(nvroot, search, avail_spare, l2cache, log);
2232 nvlist_free(search);
2238 vdev_online(nvlist_t *nv)
2242 if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_OFFLINE, &ival) == 0 ||
2243 nvlist_lookup_uint64(nv, ZPOOL_CONFIG_FAULTED, &ival) == 0 ||
2244 nvlist_lookup_uint64(nv, ZPOOL_CONFIG_REMOVED, &ival) == 0)
2251 * Helper function for zpool_get_physpaths().
2254 vdev_get_one_physpath(nvlist_t *config, char *physpath, size_t physpath_size,
2255 size_t *bytes_written)
2257 size_t bytes_left, pos, rsz;
2261 if (nvlist_lookup_string(config, ZPOOL_CONFIG_PHYS_PATH,
2263 return (EZFS_NODEVICE);
2265 pos = *bytes_written;
2266 bytes_left = physpath_size - pos;
2267 format = (pos == 0) ? "%s" : " %s";
2269 rsz = snprintf(physpath + pos, bytes_left, format, tmppath);
2270 *bytes_written += rsz;
2272 if (rsz >= bytes_left) {
2273 /* if physpath was not copied properly, clear it */
2274 if (bytes_left != 0) {
2277 return (EZFS_NOSPC);
2283 vdev_get_physpaths(nvlist_t *nv, char *physpath, size_t phypath_size,
2284 size_t *rsz, boolean_t is_spare)
2289 if (nvlist_lookup_string(nv, ZPOOL_CONFIG_TYPE, &type) != 0)
2290 return (EZFS_INVALCONFIG);
2292 if (strcmp(type, VDEV_TYPE_DISK) == 0) {
2294 * An active spare device has ZPOOL_CONFIG_IS_SPARE set.
2295 * For a spare vdev, we only want to boot from the active
2300 (void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_IS_SPARE,
2303 return (EZFS_INVALCONFIG);
2306 if (vdev_online(nv)) {
2307 if ((ret = vdev_get_one_physpath(nv, physpath,
2308 phypath_size, rsz)) != 0)
2311 } else if (strcmp(type, VDEV_TYPE_MIRROR) == 0 ||
2312 strcmp(type, VDEV_TYPE_REPLACING) == 0 ||
2313 (is_spare = (strcmp(type, VDEV_TYPE_SPARE) == 0))) {
2318 if (nvlist_lookup_nvlist_array(nv,
2319 ZPOOL_CONFIG_CHILDREN, &child, &count) != 0)
2320 return (EZFS_INVALCONFIG);
2322 for (i = 0; i < count; i++) {
2323 ret = vdev_get_physpaths(child[i], physpath,
2324 phypath_size, rsz, is_spare);
2325 if (ret == EZFS_NOSPC)
2330 return (EZFS_POOL_INVALARG);
2334 * Get phys_path for a root pool config.
2335 * Return 0 on success; non-zero on failure.
2338 zpool_get_config_physpath(nvlist_t *config, char *physpath, size_t phypath_size)
2341 nvlist_t *vdev_root;
2348 if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE,
2350 return (EZFS_INVALCONFIG);
2352 if (nvlist_lookup_string(vdev_root, ZPOOL_CONFIG_TYPE, &type) != 0 ||
2353 nvlist_lookup_nvlist_array(vdev_root, ZPOOL_CONFIG_CHILDREN,
2354 &child, &count) != 0)
2355 return (EZFS_INVALCONFIG);
2357 #if defined(__sun__) || defined(__sun)
2359 * root pool can not have EFI labeled disks and can only have
2360 * a single top-level vdev.
2362 if (strcmp(type, VDEV_TYPE_ROOT) != 0 || count != 1 ||
2363 pool_uses_efi(vdev_root))
2364 return (EZFS_POOL_INVALARG);
2367 (void) vdev_get_physpaths(child[0], physpath, phypath_size, &rsz,
2370 /* No online devices */
2372 return (EZFS_NODEVICE);
2378 * Get phys_path for a root pool
2379 * Return 0 on success; non-zero on failure.
2382 zpool_get_physpath(zpool_handle_t *zhp, char *physpath, size_t phypath_size)
2384 return (zpool_get_config_physpath(zhp->zpool_config, physpath,
2389 * If the device has being dynamically expanded then we need to relabel
2390 * the disk to use the new unallocated space.
2393 zpool_relabel_disk(libzfs_handle_t *hdl, const char *path, const char *msg)
2397 if ((fd = open(path, O_RDWR|O_DIRECT)) < 0) {
2398 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "cannot "
2399 "relabel '%s': unable to open device: %d"), path, errno);
2400 return (zfs_error(hdl, EZFS_OPENFAILED, msg));
2404 * It's possible that we might encounter an error if the device
2405 * does not have any unallocated space left. If so, we simply
2406 * ignore that error and continue on.
2408 * Also, we don't call efi_rescan() - that would just return EBUSY.
2409 * The module will do it for us in vdev_disk_open().
2411 error = efi_use_whole_disk(fd);
2413 if (error && error != VT_ENOSPC) {
2414 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "cannot "
2415 "relabel '%s': unable to read disk capacity"), path);
2416 return (zfs_error(hdl, EZFS_NOCAP, msg));
2422 * Bring the specified vdev online. The 'flags' parameter is a set of the
2423 * ZFS_ONLINE_* flags.
2426 zpool_vdev_online(zpool_handle_t *zhp, const char *path, int flags,
2427 vdev_state_t *newstate)
2429 zfs_cmd_t zc = {"\0"};
2432 boolean_t avail_spare, l2cache, islog;
2433 libzfs_handle_t *hdl = zhp->zpool_hdl;
2436 if (flags & ZFS_ONLINE_EXPAND) {
2437 (void) snprintf(msg, sizeof (msg),
2438 dgettext(TEXT_DOMAIN, "cannot expand %s"), path);
2440 (void) snprintf(msg, sizeof (msg),
2441 dgettext(TEXT_DOMAIN, "cannot online %s"), path);
2444 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
2445 if ((tgt = zpool_find_vdev(zhp, path, &avail_spare, &l2cache,
2447 return (zfs_error(hdl, EZFS_NODEVICE, msg));
2449 verify(nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID, &zc.zc_guid) == 0);
2452 return (zfs_error(hdl, EZFS_ISSPARE, msg));
2454 if (flags & ZFS_ONLINE_EXPAND ||
2455 zpool_get_prop_int(zhp, ZPOOL_PROP_AUTOEXPAND, NULL)) {
2456 uint64_t wholedisk = 0;
2458 (void) nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_WHOLE_DISK,
2462 * XXX - L2ARC 1.0 devices can't support expansion.
2465 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2466 "cannot expand cache devices"));
2467 return (zfs_error(hdl, EZFS_VDEVNOTSUP, msg));
2471 const char *fullpath = path;
2472 char buf[MAXPATHLEN];
2474 if (path[0] != '/') {
2475 error = zfs_resolve_shortname(path, buf,
2478 return (zfs_error(hdl, EZFS_NODEVICE,
2484 error = zpool_relabel_disk(hdl, fullpath, msg);
2490 zc.zc_cookie = VDEV_STATE_ONLINE;
2493 if (zfs_ioctl(hdl, ZFS_IOC_VDEV_SET_STATE, &zc) != 0) {
2494 if (errno == EINVAL) {
2495 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "was split "
2496 "from this pool into a new one. Use '%s' "
2497 "instead"), "zpool detach");
2498 return (zfs_error(hdl, EZFS_POSTSPLIT_ONLINE, msg));
2500 return (zpool_standard_error(hdl, errno, msg));
2503 *newstate = zc.zc_cookie;
2508 * Take the specified vdev offline
2511 zpool_vdev_offline(zpool_handle_t *zhp, const char *path, boolean_t istmp)
2513 zfs_cmd_t zc = {"\0"};
2516 boolean_t avail_spare, l2cache;
2517 libzfs_handle_t *hdl = zhp->zpool_hdl;
2519 (void) snprintf(msg, sizeof (msg),
2520 dgettext(TEXT_DOMAIN, "cannot offline %s"), path);
2522 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
2523 if ((tgt = zpool_find_vdev(zhp, path, &avail_spare, &l2cache,
2525 return (zfs_error(hdl, EZFS_NODEVICE, msg));
2527 verify(nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID, &zc.zc_guid) == 0);
2530 return (zfs_error(hdl, EZFS_ISSPARE, msg));
2532 zc.zc_cookie = VDEV_STATE_OFFLINE;
2533 zc.zc_obj = istmp ? ZFS_OFFLINE_TEMPORARY : 0;
2535 if (zfs_ioctl(hdl, ZFS_IOC_VDEV_SET_STATE, &zc) == 0)
2542 * There are no other replicas of this device.
2544 return (zfs_error(hdl, EZFS_NOREPLICAS, msg));
2548 * The log device has unplayed logs
2550 return (zfs_error(hdl, EZFS_UNPLAYED_LOGS, msg));
2553 return (zpool_standard_error(hdl, errno, msg));
2558 * Mark the given vdev faulted.
2561 zpool_vdev_fault(zpool_handle_t *zhp, uint64_t guid, vdev_aux_t aux)
2563 zfs_cmd_t zc = {"\0"};
2565 libzfs_handle_t *hdl = zhp->zpool_hdl;
2567 (void) snprintf(msg, sizeof (msg),
2568 dgettext(TEXT_DOMAIN, "cannot fault %llu"), (u_longlong_t)guid);
2570 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
2572 zc.zc_cookie = VDEV_STATE_FAULTED;
2575 if (ioctl(hdl->libzfs_fd, ZFS_IOC_VDEV_SET_STATE, &zc) == 0)
2582 * There are no other replicas of this device.
2584 return (zfs_error(hdl, EZFS_NOREPLICAS, msg));
2587 return (zpool_standard_error(hdl, errno, msg));
2593 * Mark the given vdev degraded.
2596 zpool_vdev_degrade(zpool_handle_t *zhp, uint64_t guid, vdev_aux_t aux)
2598 zfs_cmd_t zc = {"\0"};
2600 libzfs_handle_t *hdl = zhp->zpool_hdl;
2602 (void) snprintf(msg, sizeof (msg),
2603 dgettext(TEXT_DOMAIN, "cannot degrade %llu"), (u_longlong_t)guid);
2605 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
2607 zc.zc_cookie = VDEV_STATE_DEGRADED;
2610 if (ioctl(hdl->libzfs_fd, ZFS_IOC_VDEV_SET_STATE, &zc) == 0)
2613 return (zpool_standard_error(hdl, errno, msg));
2617 * Returns TRUE if the given nvlist is a vdev that was originally swapped in as
2621 is_replacing_spare(nvlist_t *search, nvlist_t *tgt, int which)
2627 if (nvlist_lookup_nvlist_array(search, ZPOOL_CONFIG_CHILDREN, &child,
2629 verify(nvlist_lookup_string(search, ZPOOL_CONFIG_TYPE,
2632 if (strcmp(type, VDEV_TYPE_SPARE) == 0 &&
2633 children == 2 && child[which] == tgt)
2636 for (c = 0; c < children; c++)
2637 if (is_replacing_spare(child[c], tgt, which))
2645 * Attach new_disk (fully described by nvroot) to old_disk.
2646 * If 'replacing' is specified, the new disk will replace the old one.
2649 zpool_vdev_attach(zpool_handle_t *zhp,
2650 const char *old_disk, const char *new_disk, nvlist_t *nvroot, int replacing)
2652 zfs_cmd_t zc = {"\0"};
2656 boolean_t avail_spare, l2cache, islog;
2661 nvlist_t *config_root;
2662 libzfs_handle_t *hdl = zhp->zpool_hdl;
2663 boolean_t rootpool = zpool_is_bootable(zhp);
2666 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN,
2667 "cannot replace %s with %s"), old_disk, new_disk);
2669 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN,
2670 "cannot attach %s to %s"), new_disk, old_disk);
2672 #if defined(__sun__) || defined(__sun)
2674 * If this is a root pool, make sure that we're not attaching an
2675 * EFI labeled device.
2677 if (rootpool && pool_uses_efi(nvroot)) {
2678 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2679 "EFI labeled devices are not supported on root pools."));
2680 return (zfs_error(hdl, EZFS_POOL_NOTSUP, msg));
2684 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
2685 if ((tgt = zpool_find_vdev(zhp, old_disk, &avail_spare, &l2cache,
2687 return (zfs_error(hdl, EZFS_NODEVICE, msg));
2690 return (zfs_error(hdl, EZFS_ISSPARE, msg));
2693 return (zfs_error(hdl, EZFS_ISL2CACHE, msg));
2695 verify(nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID, &zc.zc_guid) == 0);
2696 zc.zc_cookie = replacing;
2698 if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_CHILDREN,
2699 &child, &children) != 0 || children != 1) {
2700 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2701 "new device must be a single disk"));
2702 return (zfs_error(hdl, EZFS_INVALCONFIG, msg));
2705 verify(nvlist_lookup_nvlist(zpool_get_config(zhp, NULL),
2706 ZPOOL_CONFIG_VDEV_TREE, &config_root) == 0);
2708 if ((newname = zpool_vdev_name(NULL, NULL, child[0], 0)) == NULL)
2712 * If the target is a hot spare that has been swapped in, we can only
2713 * replace it with another hot spare.
2716 nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_IS_SPARE, &val) == 0 &&
2717 (zpool_find_vdev(zhp, newname, &avail_spare, &l2cache,
2718 NULL) == NULL || !avail_spare) &&
2719 is_replacing_spare(config_root, tgt, 1)) {
2720 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2721 "can only be replaced by another hot spare"));
2723 return (zfs_error(hdl, EZFS_BADTARGET, msg));
2728 if (zcmd_write_conf_nvlist(hdl, &zc, nvroot) != 0)
2731 ret = zfs_ioctl(hdl, ZFS_IOC_VDEV_ATTACH, &zc);
2733 zcmd_free_nvlists(&zc);
2738 * XXX need a better way to prevent user from
2739 * booting up a half-baked vdev.
2741 (void) fprintf(stderr, dgettext(TEXT_DOMAIN, "Make "
2742 "sure to wait until resilver is done "
2743 "before rebooting.\n"));
2751 * Can't attach to or replace this type of vdev.
2754 uint64_t version = zpool_get_prop_int(zhp,
2755 ZPOOL_PROP_VERSION, NULL);
2758 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2759 "cannot replace a log with a spare"));
2760 else if (version >= SPA_VERSION_MULTI_REPLACE)
2761 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2762 "already in replacing/spare config; wait "
2763 "for completion or use 'zpool detach'"));
2765 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2766 "cannot replace a replacing device"));
2768 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2769 "can only attach to mirrors and top-level "
2772 (void) zfs_error(hdl, EZFS_BADTARGET, msg);
2777 * The new device must be a single disk.
2779 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2780 "new device must be a single disk"));
2781 (void) zfs_error(hdl, EZFS_INVALCONFIG, msg);
2785 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "%s is busy"),
2787 (void) zfs_error(hdl, EZFS_BADDEV, msg);
2792 * The new device is too small.
2794 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2795 "device is too small"));
2796 (void) zfs_error(hdl, EZFS_BADDEV, msg);
2801 * The new device has a different optimal sector size.
2803 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2804 "new device has a different optimal sector size; use the "
2805 "option '-o ashift=N' to override the optimal size"));
2806 (void) zfs_error(hdl, EZFS_BADDEV, msg);
2811 * The resulting top-level vdev spec won't fit in the label.
2813 (void) zfs_error(hdl, EZFS_DEVOVERFLOW, msg);
2817 (void) zpool_standard_error(hdl, errno, msg);
2824 * Detach the specified device.
2827 zpool_vdev_detach(zpool_handle_t *zhp, const char *path)
2829 zfs_cmd_t zc = {"\0"};
2832 boolean_t avail_spare, l2cache;
2833 libzfs_handle_t *hdl = zhp->zpool_hdl;
2835 (void) snprintf(msg, sizeof (msg),
2836 dgettext(TEXT_DOMAIN, "cannot detach %s"), path);
2838 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
2839 if ((tgt = zpool_find_vdev(zhp, path, &avail_spare, &l2cache,
2841 return (zfs_error(hdl, EZFS_NODEVICE, msg));
2844 return (zfs_error(hdl, EZFS_ISSPARE, msg));
2847 return (zfs_error(hdl, EZFS_ISL2CACHE, msg));
2849 verify(nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID, &zc.zc_guid) == 0);
2851 if (zfs_ioctl(hdl, ZFS_IOC_VDEV_DETACH, &zc) == 0)
2858 * Can't detach from this type of vdev.
2860 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "only "
2861 "applicable to mirror and replacing vdevs"));
2862 (void) zfs_error(hdl, EZFS_BADTARGET, msg);
2867 * There are no other replicas of this device.
2869 (void) zfs_error(hdl, EZFS_NOREPLICAS, msg);
2873 (void) zpool_standard_error(hdl, errno, msg);
2880 * Find a mirror vdev in the source nvlist.
2882 * The mchild array contains a list of disks in one of the top-level mirrors
2883 * of the source pool. The schild array contains a list of disks that the
2884 * user specified on the command line. We loop over the mchild array to
2885 * see if any entry in the schild array matches.
2887 * If a disk in the mchild array is found in the schild array, we return
2888 * the index of that entry. Otherwise we return -1.
2891 find_vdev_entry(zpool_handle_t *zhp, nvlist_t **mchild, uint_t mchildren,
2892 nvlist_t **schild, uint_t schildren)
2896 for (mc = 0; mc < mchildren; mc++) {
2898 char *mpath = zpool_vdev_name(zhp->zpool_hdl, zhp,
2901 for (sc = 0; sc < schildren; sc++) {
2902 char *spath = zpool_vdev_name(zhp->zpool_hdl, zhp,
2904 boolean_t result = (strcmp(mpath, spath) == 0);
2920 * Split a mirror pool. If newroot points to null, then a new nvlist
2921 * is generated and it is the responsibility of the caller to free it.
2924 zpool_vdev_split(zpool_handle_t *zhp, char *newname, nvlist_t **newroot,
2925 nvlist_t *props, splitflags_t flags)
2927 zfs_cmd_t zc = {"\0"};
2929 nvlist_t *tree, *config, **child, **newchild, *newconfig = NULL;
2930 nvlist_t **varray = NULL, *zc_props = NULL;
2931 uint_t c, children, newchildren, lastlog = 0, vcount, found = 0;
2932 libzfs_handle_t *hdl = zhp->zpool_hdl;
2934 boolean_t freelist = B_FALSE, memory_err = B_TRUE;
2937 (void) snprintf(msg, sizeof (msg),
2938 dgettext(TEXT_DOMAIN, "Unable to split %s"), zhp->zpool_name);
2940 if (!zpool_name_valid(hdl, B_FALSE, newname))
2941 return (zfs_error(hdl, EZFS_INVALIDNAME, msg));
2943 if ((config = zpool_get_config(zhp, NULL)) == NULL) {
2944 (void) fprintf(stderr, gettext("Internal error: unable to "
2945 "retrieve pool configuration\n"));
2949 verify(nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE, &tree)
2951 verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_VERSION, &vers) == 0);
2954 prop_flags_t flags = { .create = B_FALSE, .import = B_TRUE };
2955 if ((zc_props = zpool_valid_proplist(hdl, zhp->zpool_name,
2956 props, vers, flags, msg)) == NULL)
2960 if (nvlist_lookup_nvlist_array(tree, ZPOOL_CONFIG_CHILDREN, &child,
2962 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2963 "Source pool is missing vdev tree"));
2964 nvlist_free(zc_props);
2968 varray = zfs_alloc(hdl, children * sizeof (nvlist_t *));
2971 if (*newroot == NULL ||
2972 nvlist_lookup_nvlist_array(*newroot, ZPOOL_CONFIG_CHILDREN,
2973 &newchild, &newchildren) != 0)
2976 for (c = 0; c < children; c++) {
2977 uint64_t is_log = B_FALSE, is_hole = B_FALSE;
2979 nvlist_t **mchild, *vdev;
2984 * Unlike cache & spares, slogs are stored in the
2985 * ZPOOL_CONFIG_CHILDREN array. We filter them out here.
2987 (void) nvlist_lookup_uint64(child[c], ZPOOL_CONFIG_IS_LOG,
2989 (void) nvlist_lookup_uint64(child[c], ZPOOL_CONFIG_IS_HOLE,
2991 if (is_log || is_hole) {
2993 * Create a hole vdev and put it in the config.
2995 if (nvlist_alloc(&vdev, NV_UNIQUE_NAME, 0) != 0)
2997 if (nvlist_add_string(vdev, ZPOOL_CONFIG_TYPE,
2998 VDEV_TYPE_HOLE) != 0)
3000 if (nvlist_add_uint64(vdev, ZPOOL_CONFIG_IS_HOLE,
3005 varray[vcount++] = vdev;
3009 verify(nvlist_lookup_string(child[c], ZPOOL_CONFIG_TYPE, &type)
3011 if (strcmp(type, VDEV_TYPE_MIRROR) != 0) {
3012 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3013 "Source pool must be composed only of mirrors\n"));
3014 retval = zfs_error(hdl, EZFS_INVALCONFIG, msg);
3018 verify(nvlist_lookup_nvlist_array(child[c],
3019 ZPOOL_CONFIG_CHILDREN, &mchild, &mchildren) == 0);
3021 /* find or add an entry for this top-level vdev */
3022 if (newchildren > 0 &&
3023 (entry = find_vdev_entry(zhp, mchild, mchildren,
3024 newchild, newchildren)) >= 0) {
3025 /* We found a disk that the user specified. */
3026 vdev = mchild[entry];
3029 /* User didn't specify a disk for this vdev. */
3030 vdev = mchild[mchildren - 1];
3033 if (nvlist_dup(vdev, &varray[vcount++], 0) != 0)
3037 /* did we find every disk the user specified? */
3038 if (found != newchildren) {
3039 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "Device list must "
3040 "include at most one disk from each mirror"));
3041 retval = zfs_error(hdl, EZFS_INVALCONFIG, msg);
3045 /* Prepare the nvlist for populating. */
3046 if (*newroot == NULL) {
3047 if (nvlist_alloc(newroot, NV_UNIQUE_NAME, 0) != 0)
3050 if (nvlist_add_string(*newroot, ZPOOL_CONFIG_TYPE,
3051 VDEV_TYPE_ROOT) != 0)
3054 verify(nvlist_remove_all(*newroot, ZPOOL_CONFIG_CHILDREN) == 0);
3057 /* Add all the children we found */
3058 if (nvlist_add_nvlist_array(*newroot, ZPOOL_CONFIG_CHILDREN, varray,
3059 lastlog == 0 ? vcount : lastlog) != 0)
3063 * If we're just doing a dry run, exit now with success.
3066 memory_err = B_FALSE;
3071 /* now build up the config list & call the ioctl */
3072 if (nvlist_alloc(&newconfig, NV_UNIQUE_NAME, 0) != 0)
3075 if (nvlist_add_nvlist(newconfig,
3076 ZPOOL_CONFIG_VDEV_TREE, *newroot) != 0 ||
3077 nvlist_add_string(newconfig,
3078 ZPOOL_CONFIG_POOL_NAME, newname) != 0 ||
3079 nvlist_add_uint64(newconfig, ZPOOL_CONFIG_VERSION, vers) != 0)
3083 * The new pool is automatically part of the namespace unless we
3084 * explicitly export it.
3087 zc.zc_cookie = ZPOOL_EXPORT_AFTER_SPLIT;
3088 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
3089 (void) strlcpy(zc.zc_string, newname, sizeof (zc.zc_string));
3090 if (zcmd_write_conf_nvlist(hdl, &zc, newconfig) != 0)
3092 if (zc_props != NULL && zcmd_write_src_nvlist(hdl, &zc, zc_props) != 0)
3095 if (zfs_ioctl(hdl, ZFS_IOC_VDEV_SPLIT, &zc) != 0) {
3096 retval = zpool_standard_error(hdl, errno, msg);
3101 memory_err = B_FALSE;
3104 if (varray != NULL) {
3107 for (v = 0; v < vcount; v++)
3108 nvlist_free(varray[v]);
3111 zcmd_free_nvlists(&zc);
3112 nvlist_free(zc_props);
3113 nvlist_free(newconfig);
3115 nvlist_free(*newroot);
3123 return (no_memory(hdl));
3129 * Remove the given device. Currently, this is supported only for hot spares
3130 * and level 2 cache devices.
3133 zpool_vdev_remove(zpool_handle_t *zhp, const char *path)
3135 zfs_cmd_t zc = {"\0"};
3138 boolean_t avail_spare, l2cache, islog;
3139 libzfs_handle_t *hdl = zhp->zpool_hdl;
3142 (void) snprintf(msg, sizeof (msg),
3143 dgettext(TEXT_DOMAIN, "cannot remove %s"), path);
3145 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
3146 if ((tgt = zpool_find_vdev(zhp, path, &avail_spare, &l2cache,
3148 return (zfs_error(hdl, EZFS_NODEVICE, msg));
3150 * XXX - this should just go away.
3152 if (!avail_spare && !l2cache && !islog) {
3153 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3154 "only inactive hot spares, cache, top-level, "
3155 "or log devices can be removed"));
3156 return (zfs_error(hdl, EZFS_NODEVICE, msg));
3159 version = zpool_get_prop_int(zhp, ZPOOL_PROP_VERSION, NULL);
3160 if (islog && version < SPA_VERSION_HOLES) {
3161 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3162 "pool must be upgrade to support log removal"));
3163 return (zfs_error(hdl, EZFS_BADVERSION, msg));
3166 verify(nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID, &zc.zc_guid) == 0);
3168 if (zfs_ioctl(hdl, ZFS_IOC_VDEV_REMOVE, &zc) == 0)
3171 return (zpool_standard_error(hdl, errno, msg));
3175 * Clear the errors for the pool, or the particular device if specified.
3178 zpool_clear(zpool_handle_t *zhp, const char *path, nvlist_t *rewindnvl)
3180 zfs_cmd_t zc = {"\0"};
3183 zpool_rewind_policy_t policy;
3184 boolean_t avail_spare, l2cache;
3185 libzfs_handle_t *hdl = zhp->zpool_hdl;
3186 nvlist_t *nvi = NULL;
3190 (void) snprintf(msg, sizeof (msg),
3191 dgettext(TEXT_DOMAIN, "cannot clear errors for %s"),
3194 (void) snprintf(msg, sizeof (msg),
3195 dgettext(TEXT_DOMAIN, "cannot clear errors for %s"),
3198 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
3200 if ((tgt = zpool_find_vdev(zhp, path, &avail_spare,
3201 &l2cache, NULL)) == 0)
3202 return (zfs_error(hdl, EZFS_NODEVICE, msg));
3205 * Don't allow error clearing for hot spares. Do allow
3206 * error clearing for l2cache devices.
3209 return (zfs_error(hdl, EZFS_ISSPARE, msg));
3211 verify(nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID,
3215 zpool_get_rewind_policy(rewindnvl, &policy);
3216 zc.zc_cookie = policy.zrp_request;
3218 if (zcmd_alloc_dst_nvlist(hdl, &zc, zhp->zpool_config_size * 2) != 0)
3221 if (zcmd_write_src_nvlist(hdl, &zc, rewindnvl) != 0)
3224 while ((error = zfs_ioctl(hdl, ZFS_IOC_CLEAR, &zc)) != 0 &&
3226 if (zcmd_expand_dst_nvlist(hdl, &zc) != 0) {
3227 zcmd_free_nvlists(&zc);
3232 if (!error || ((policy.zrp_request & ZPOOL_TRY_REWIND) &&
3233 errno != EPERM && errno != EACCES)) {
3234 if (policy.zrp_request &
3235 (ZPOOL_DO_REWIND | ZPOOL_TRY_REWIND)) {
3236 (void) zcmd_read_dst_nvlist(hdl, &zc, &nvi);
3237 zpool_rewind_exclaim(hdl, zc.zc_name,
3238 ((policy.zrp_request & ZPOOL_TRY_REWIND) != 0),
3242 zcmd_free_nvlists(&zc);
3246 zcmd_free_nvlists(&zc);
3247 return (zpool_standard_error(hdl, errno, msg));
3251 * Similar to zpool_clear(), but takes a GUID (used by fmd).
3254 zpool_vdev_clear(zpool_handle_t *zhp, uint64_t guid)
3256 zfs_cmd_t zc = {"\0"};
3258 libzfs_handle_t *hdl = zhp->zpool_hdl;
3260 (void) snprintf(msg, sizeof (msg),
3261 dgettext(TEXT_DOMAIN, "cannot clear errors for %llx"),
3262 (u_longlong_t)guid);
3264 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
3266 zc.zc_cookie = ZPOOL_NO_REWIND;
3268 if (ioctl(hdl->libzfs_fd, ZFS_IOC_CLEAR, &zc) == 0)
3271 return (zpool_standard_error(hdl, errno, msg));
3275 * Change the GUID for a pool.
3278 zpool_reguid(zpool_handle_t *zhp)
3281 libzfs_handle_t *hdl = zhp->zpool_hdl;
3282 zfs_cmd_t zc = {"\0"};
3284 (void) snprintf(msg, sizeof (msg),
3285 dgettext(TEXT_DOMAIN, "cannot reguid '%s'"), zhp->zpool_name);
3287 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
3288 if (zfs_ioctl(hdl, ZFS_IOC_POOL_REGUID, &zc) == 0)
3291 return (zpool_standard_error(hdl, errno, msg));
3298 zpool_reopen(zpool_handle_t *zhp)
3300 zfs_cmd_t zc = {"\0"};
3302 libzfs_handle_t *hdl = zhp->zpool_hdl;
3304 (void) snprintf(msg, sizeof (msg),
3305 dgettext(TEXT_DOMAIN, "cannot reopen '%s'"),
3308 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
3309 if (zfs_ioctl(hdl, ZFS_IOC_POOL_REOPEN, &zc) == 0)
3311 return (zpool_standard_error(hdl, errno, msg));
3314 #if defined(__sun__) || defined(__sun)
3316 * Convert from a devid string to a path.
3319 devid_to_path(char *devid_str)
3324 devid_nmlist_t *list = NULL;
3327 if (devid_str_decode(devid_str, &devid, &minor) != 0)
3330 ret = devid_deviceid_to_nmlist("/dev", devid, minor, &list);
3332 devid_str_free(minor);
3339 * In a case the strdup() fails, we will just return NULL below.
3341 path = strdup(list[0].devname);
3343 devid_free_nmlist(list);
3349 * Convert from a path to a devid string.
3352 path_to_devid(const char *path)
3358 if ((fd = open(path, O_RDONLY)) < 0)
3363 if (devid_get(fd, &devid) == 0) {
3364 if (devid_get_minor_name(fd, &minor) == 0)
3365 ret = devid_str_encode(devid, minor);
3367 devid_str_free(minor);
3376 * Issue the necessary ioctl() to update the stored path value for the vdev. We
3377 * ignore any failure here, since a common case is for an unprivileged user to
3378 * type 'zpool status', and we'll display the correct information anyway.
3381 set_path(zpool_handle_t *zhp, nvlist_t *nv, const char *path)
3383 zfs_cmd_t zc = {"\0"};
3385 (void) strncpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
3386 (void) strncpy(zc.zc_value, path, sizeof (zc.zc_value));
3387 verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID,
3390 (void) ioctl(zhp->zpool_hdl->libzfs_fd, ZFS_IOC_VDEV_SETPATH, &zc);
3395 * Remove partition suffix from a vdev path. Partition suffixes may take three
3396 * forms: "-partX", "pX", or "X", where X is a string of digits. The second
3397 * case only occurs when the suffix is preceded by a digit, i.e. "md0p0" The
3398 * third case only occurs when preceded by a string matching the regular
3399 * expression "^([hsv]|xv)d[a-z]+", i.e. a scsi, ide, virtio or xen disk.
3401 * caller must free the returned string
3404 zfs_strip_partition(libzfs_handle_t *hdl, char *path)
3406 char *tmp = zfs_strdup(hdl, path);
3407 char *part = NULL, *d = NULL;
3409 if ((part = strstr(tmp, "-part")) && part != tmp) {
3411 } else if ((part = strrchr(tmp, 'p')) &&
3412 part > tmp + 1 && isdigit(*(part-1))) {
3414 } else if ((tmp[0] == 'h' || tmp[0] == 's' || tmp[0] == 'v') &&
3416 for (d = &tmp[2]; isalpha(*d); part = ++d);
3417 } else if (strncmp("xvd", tmp, 3) == 0) {
3418 for (d = &tmp[3]; isalpha(*d); part = ++d);
3420 if (part && d && *d != '\0') {
3421 for (; isdigit(*d); d++);
3428 #define PATH_BUF_LEN 64
3431 * Given a vdev, return the name to display in iostat. If the vdev has a path,
3432 * we use that, stripping off any leading "/dev/dsk/"; if not, we use the type.
3433 * We also check if this is a whole disk, in which case we strip off the
3434 * trailing 's0' slice name.
3436 * This routine is also responsible for identifying when disks have been
3437 * reconfigured in a new location. The kernel will have opened the device by
3438 * devid, but the path will still refer to the old location. To catch this, we
3439 * first do a path -> devid translation (which is fast for the common case). If
3440 * the devid matches, we're done. If not, we do a reverse devid -> path
3441 * translation and issue the appropriate ioctl() to update the path of the vdev.
3442 * If 'zhp' is NULL, then this is an exported pool, and we don't need to do any
3446 zpool_vdev_name(libzfs_handle_t *hdl, zpool_handle_t *zhp, nvlist_t *nv,
3449 char *path, *type, *env;
3451 char buf[PATH_BUF_LEN];
3452 char tmpbuf[PATH_BUF_LEN];
3454 env = getenv("ZPOOL_VDEV_NAME_PATH");
3455 if (env && (strtoul(env, NULL, 0) > 0 ||
3456 !strncasecmp(env, "YES", 3) || !strncasecmp(env, "ON", 2)))
3457 name_flags |= VDEV_NAME_PATH;
3459 env = getenv("ZPOOL_VDEV_NAME_GUID");
3460 if (env && (strtoul(env, NULL, 0) > 0 ||
3461 !strncasecmp(env, "YES", 3) || !strncasecmp(env, "ON", 2)))
3462 name_flags |= VDEV_NAME_GUID;
3464 env = getenv("ZPOOL_VDEV_NAME_FOLLOW_LINKS");
3465 if (env && (strtoul(env, NULL, 0) > 0 ||
3466 !strncasecmp(env, "YES", 3) || !strncasecmp(env, "ON", 2)))
3467 name_flags |= VDEV_NAME_FOLLOW_LINKS;
3469 if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_NOT_PRESENT, &value) == 0 ||
3470 name_flags & VDEV_NAME_GUID) {
3471 (void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID, &value);
3472 (void) snprintf(buf, sizeof (buf), "%llu", (u_longlong_t)value);
3474 } else if (nvlist_lookup_string(nv, ZPOOL_CONFIG_PATH, &path) == 0) {
3475 #if defined(__sun__) || defined(__sun)
3477 * Live VDEV path updates to a kernel VDEV during a
3478 * zpool_vdev_name lookup are not supported on Linux.
3485 * If the device is dead (faulted, offline, etc) then don't
3486 * bother opening it. Otherwise we may be forcing the user to
3487 * open a misbehaving device, which can have undesirable
3490 if ((nvlist_lookup_uint64_array(nv, ZPOOL_CONFIG_VDEV_STATS,
3491 (uint64_t **)&vs, &vsc) != 0 ||
3492 vs->vs_state >= VDEV_STATE_DEGRADED) &&
3494 nvlist_lookup_string(nv, ZPOOL_CONFIG_DEVID, &devid) == 0) {
3496 * Determine if the current path is correct.
3498 char *newdevid = path_to_devid(path);
3500 if (newdevid == NULL ||
3501 strcmp(devid, newdevid) != 0) {
3504 if ((newpath = devid_to_path(devid)) != NULL) {
3506 * Update the path appropriately.
3508 set_path(zhp, nv, newpath);
3509 if (nvlist_add_string(nv,
3510 ZPOOL_CONFIG_PATH, newpath) == 0)
3511 verify(nvlist_lookup_string(nv,
3519 devid_str_free(newdevid);
3523 if (name_flags & VDEV_NAME_FOLLOW_LINKS) {
3524 char *rp = realpath(path, NULL);
3526 strlcpy(buf, rp, sizeof (buf));
3533 * For a block device only use the name.
3535 verify(nvlist_lookup_string(nv, ZPOOL_CONFIG_TYPE, &type) == 0);
3536 if ((strcmp(type, VDEV_TYPE_DISK) == 0) &&
3537 !(name_flags & VDEV_NAME_PATH)) {
3538 path = strrchr(path, '/');
3543 * Remove the partition from the path it this is a whole disk.
3545 if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_WHOLE_DISK, &value)
3546 == 0 && value && !(name_flags & VDEV_NAME_PATH)) {
3547 return (zfs_strip_partition(hdl, path));
3550 verify(nvlist_lookup_string(nv, ZPOOL_CONFIG_TYPE, &path) == 0);
3553 * If it's a raidz device, we need to stick in the parity level.
3555 if (strcmp(path, VDEV_TYPE_RAIDZ) == 0) {
3556 verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_NPARITY,
3558 (void) snprintf(buf, sizeof (buf), "%s%llu", path,
3559 (u_longlong_t)value);
3564 * We identify each top-level vdev by using a <type-id>
3565 * naming convention.
3567 if (name_flags & VDEV_NAME_TYPE_ID) {
3569 verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_ID,
3571 (void) snprintf(tmpbuf, sizeof (tmpbuf), "%s-%llu",
3572 path, (u_longlong_t)id);
3577 return (zfs_strdup(hdl, path));
3581 zbookmark_mem_compare(const void *a, const void *b)
3583 return (memcmp(a, b, sizeof (zbookmark_phys_t)));
3587 * Retrieve the persistent error log, uniquify the members, and return to the
3591 zpool_get_errlog(zpool_handle_t *zhp, nvlist_t **nverrlistp)
3593 zfs_cmd_t zc = {"\0"};
3595 zbookmark_phys_t *zb = NULL;
3599 * Retrieve the raw error list from the kernel. If the number of errors
3600 * has increased, allocate more space and continue until we get the
3603 verify(nvlist_lookup_uint64(zhp->zpool_config, ZPOOL_CONFIG_ERRCOUNT,
3607 if ((zc.zc_nvlist_dst = (uintptr_t)zfs_alloc(zhp->zpool_hdl,
3608 count * sizeof (zbookmark_phys_t))) == (uintptr_t)NULL)
3610 zc.zc_nvlist_dst_size = count;
3611 (void) strcpy(zc.zc_name, zhp->zpool_name);
3613 if (ioctl(zhp->zpool_hdl->libzfs_fd, ZFS_IOC_ERROR_LOG,
3615 free((void *)(uintptr_t)zc.zc_nvlist_dst);
3616 if (errno == ENOMEM) {
3619 count = zc.zc_nvlist_dst_size;
3620 dst = zfs_alloc(zhp->zpool_hdl, count *
3621 sizeof (zbookmark_phys_t));
3624 zc.zc_nvlist_dst = (uintptr_t)dst;
3634 * Sort the resulting bookmarks. This is a little confusing due to the
3635 * implementation of ZFS_IOC_ERROR_LOG. The bookmarks are copied last
3636 * to first, and 'zc_nvlist_dst_size' indicates the number of boomarks
3637 * _not_ copied as part of the process. So we point the start of our
3638 * array appropriate and decrement the total number of elements.
3640 zb = ((zbookmark_phys_t *)(uintptr_t)zc.zc_nvlist_dst) +
3641 zc.zc_nvlist_dst_size;
3642 count -= zc.zc_nvlist_dst_size;
3644 qsort(zb, count, sizeof (zbookmark_phys_t), zbookmark_mem_compare);
3646 verify(nvlist_alloc(nverrlistp, 0, KM_SLEEP) == 0);
3649 * Fill in the nverrlistp with nvlist's of dataset and object numbers.
3651 for (i = 0; i < count; i++) {
3654 /* ignoring zb_blkid and zb_level for now */
3655 if (i > 0 && zb[i-1].zb_objset == zb[i].zb_objset &&
3656 zb[i-1].zb_object == zb[i].zb_object)
3659 if (nvlist_alloc(&nv, NV_UNIQUE_NAME, KM_SLEEP) != 0)
3661 if (nvlist_add_uint64(nv, ZPOOL_ERR_DATASET,
3662 zb[i].zb_objset) != 0) {
3666 if (nvlist_add_uint64(nv, ZPOOL_ERR_OBJECT,
3667 zb[i].zb_object) != 0) {
3671 if (nvlist_add_nvlist(*nverrlistp, "ejk", nv) != 0) {
3678 free((void *)(uintptr_t)zc.zc_nvlist_dst);
3682 free((void *)(uintptr_t)zc.zc_nvlist_dst);
3683 return (no_memory(zhp->zpool_hdl));
3687 * Upgrade a ZFS pool to the latest on-disk version.
3690 zpool_upgrade(zpool_handle_t *zhp, uint64_t new_version)
3692 zfs_cmd_t zc = {"\0"};
3693 libzfs_handle_t *hdl = zhp->zpool_hdl;
3695 (void) strcpy(zc.zc_name, zhp->zpool_name);
3696 zc.zc_cookie = new_version;
3698 if (zfs_ioctl(hdl, ZFS_IOC_POOL_UPGRADE, &zc) != 0)
3699 return (zpool_standard_error_fmt(hdl, errno,
3700 dgettext(TEXT_DOMAIN, "cannot upgrade '%s'"),
3706 zfs_save_arguments(int argc, char **argv, char *string, int len)
3710 (void) strlcpy(string, basename(argv[0]), len);
3711 for (i = 1; i < argc; i++) {
3712 (void) strlcat(string, " ", len);
3713 (void) strlcat(string, argv[i], len);
3718 zpool_log_history(libzfs_handle_t *hdl, const char *message)
3720 zfs_cmd_t zc = {"\0"};
3724 args = fnvlist_alloc();
3725 fnvlist_add_string(args, "message", message);
3726 err = zcmd_write_src_nvlist(hdl, &zc, args);
3728 err = ioctl(hdl->libzfs_fd, ZFS_IOC_LOG_HISTORY, &zc);
3730 zcmd_free_nvlists(&zc);
3735 * Perform ioctl to get some command history of a pool.
3737 * 'buf' is the buffer to fill up to 'len' bytes. 'off' is the
3738 * logical offset of the history buffer to start reading from.
3740 * Upon return, 'off' is the next logical offset to read from and
3741 * 'len' is the actual amount of bytes read into 'buf'.
3744 get_history(zpool_handle_t *zhp, char *buf, uint64_t *off, uint64_t *len)
3746 zfs_cmd_t zc = {"\0"};
3747 libzfs_handle_t *hdl = zhp->zpool_hdl;
3749 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
3751 zc.zc_history = (uint64_t)(uintptr_t)buf;
3752 zc.zc_history_len = *len;
3753 zc.zc_history_offset = *off;
3755 if (ioctl(hdl->libzfs_fd, ZFS_IOC_POOL_GET_HISTORY, &zc) != 0) {
3758 return (zfs_error_fmt(hdl, EZFS_PERM,
3759 dgettext(TEXT_DOMAIN,
3760 "cannot show history for pool '%s'"),
3763 return (zfs_error_fmt(hdl, EZFS_NOHISTORY,
3764 dgettext(TEXT_DOMAIN, "cannot get history for pool "
3765 "'%s'"), zhp->zpool_name));
3767 return (zfs_error_fmt(hdl, EZFS_BADVERSION,
3768 dgettext(TEXT_DOMAIN, "cannot get history for pool "
3769 "'%s', pool must be upgraded"), zhp->zpool_name));
3771 return (zpool_standard_error_fmt(hdl, errno,
3772 dgettext(TEXT_DOMAIN,
3773 "cannot get history for '%s'"), zhp->zpool_name));
3777 *len = zc.zc_history_len;
3778 *off = zc.zc_history_offset;
3784 * Process the buffer of nvlists, unpacking and storing each nvlist record
3785 * into 'records'. 'leftover' is set to the number of bytes that weren't
3786 * processed as there wasn't a complete record.
3789 zpool_history_unpack(char *buf, uint64_t bytes_read, uint64_t *leftover,
3790 nvlist_t ***records, uint_t *numrecords)
3796 while (bytes_read > sizeof (reclen)) {
3798 /* get length of packed record (stored as little endian) */
3799 for (i = 0, reclen = 0; i < sizeof (reclen); i++)
3800 reclen += (uint64_t)(((uchar_t *)buf)[i]) << (8*i);
3802 if (bytes_read < sizeof (reclen) + reclen)
3806 if (nvlist_unpack(buf + sizeof (reclen), reclen, &nv, 0) != 0)
3808 bytes_read -= sizeof (reclen) + reclen;
3809 buf += sizeof (reclen) + reclen;
3811 /* add record to nvlist array */
3813 if (ISP2(*numrecords + 1)) {
3814 *records = realloc(*records,
3815 *numrecords * 2 * sizeof (nvlist_t *));
3817 (*records)[*numrecords - 1] = nv;
3820 *leftover = bytes_read;
3825 * Retrieve the command history of a pool.
3828 zpool_get_history(zpool_handle_t *zhp, nvlist_t **nvhisp)
3831 int buflen = 128 * 1024;
3833 nvlist_t **records = NULL;
3834 uint_t numrecords = 0;
3837 buf = malloc(buflen);
3841 uint64_t bytes_read = buflen;
3844 if ((err = get_history(zhp, buf, &off, &bytes_read)) != 0)
3847 /* if nothing else was read in, we're at EOF, just return */
3851 if ((err = zpool_history_unpack(buf, bytes_read,
3852 &leftover, &records, &numrecords)) != 0)
3855 if (leftover == bytes_read) {
3857 * no progress made, because buffer is not big enough
3858 * to hold this record; resize and retry.
3862 buf = malloc(buflen);
3873 verify(nvlist_alloc(nvhisp, NV_UNIQUE_NAME, 0) == 0);
3874 verify(nvlist_add_nvlist_array(*nvhisp, ZPOOL_HIST_RECORD,
3875 records, numrecords) == 0);
3877 for (i = 0; i < numrecords; i++)
3878 nvlist_free(records[i]);
3885 * Retrieve the next event given the passed 'zevent_fd' file descriptor.
3886 * If there is a new event available 'nvp' will contain a newly allocated
3887 * nvlist and 'dropped' will be set to the number of missed events since
3888 * the last call to this function. When 'nvp' is set to NULL it indicates
3889 * no new events are available. In either case the function returns 0 and
3890 * it is up to the caller to free 'nvp'. In the case of a fatal error the
3891 * function will return a non-zero value. When the function is called in
3892 * blocking mode (the default, unless the ZEVENT_NONBLOCK flag is passed),
3893 * it will not return until a new event is available.
3896 zpool_events_next(libzfs_handle_t *hdl, nvlist_t **nvp,
3897 int *dropped, unsigned flags, int zevent_fd)
3899 zfs_cmd_t zc = {"\0"};
3904 zc.zc_cleanup_fd = zevent_fd;
3906 if (flags & ZEVENT_NONBLOCK)
3907 zc.zc_guid = ZEVENT_NONBLOCK;
3909 if (zcmd_alloc_dst_nvlist(hdl, &zc, ZEVENT_SIZE) != 0)
3913 if (zfs_ioctl(hdl, ZFS_IOC_EVENTS_NEXT, &zc) != 0) {
3916 error = zfs_error_fmt(hdl, EZFS_POOLUNAVAIL,
3917 dgettext(TEXT_DOMAIN, "zfs shutdown"));
3920 /* Blocking error case should not occur */
3921 if (!(flags & ZEVENT_NONBLOCK))
3922 error = zpool_standard_error_fmt(hdl, errno,
3923 dgettext(TEXT_DOMAIN, "cannot get event"));
3927 if (zcmd_expand_dst_nvlist(hdl, &zc) != 0) {
3928 error = zfs_error_fmt(hdl, EZFS_NOMEM,
3929 dgettext(TEXT_DOMAIN, "cannot get event"));
3935 error = zpool_standard_error_fmt(hdl, errno,
3936 dgettext(TEXT_DOMAIN, "cannot get event"));
3941 error = zcmd_read_dst_nvlist(hdl, &zc, nvp);
3945 *dropped = (int)zc.zc_cookie;
3947 zcmd_free_nvlists(&zc);
3956 zpool_events_clear(libzfs_handle_t *hdl, int *count)
3958 zfs_cmd_t zc = {"\0"};
3961 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN,
3962 "cannot clear events"));
3964 if (zfs_ioctl(hdl, ZFS_IOC_EVENTS_CLEAR, &zc) != 0)
3965 return (zpool_standard_error_fmt(hdl, errno, msg));
3968 *count = (int)zc.zc_cookie; /* # of events cleared */
3974 * Seek to a specific EID, ZEVENT_SEEK_START, or ZEVENT_SEEK_END for
3975 * the passed zevent_fd file handle. On success zero is returned,
3976 * otherwise -1 is returned and hdl->libzfs_error is set to the errno.
3979 zpool_events_seek(libzfs_handle_t *hdl, uint64_t eid, int zevent_fd)
3981 zfs_cmd_t zc = {"\0"};
3985 zc.zc_cleanup_fd = zevent_fd;
3987 if (zfs_ioctl(hdl, ZFS_IOC_EVENTS_SEEK, &zc) != 0) {
3990 error = zfs_error_fmt(hdl, EZFS_NOENT,
3991 dgettext(TEXT_DOMAIN, "cannot get event"));
3995 error = zfs_error_fmt(hdl, EZFS_NOMEM,
3996 dgettext(TEXT_DOMAIN, "cannot get event"));
4000 error = zpool_standard_error_fmt(hdl, errno,
4001 dgettext(TEXT_DOMAIN, "cannot get event"));
4010 zpool_obj_to_path(zpool_handle_t *zhp, uint64_t dsobj, uint64_t obj,
4011 char *pathname, size_t len)
4013 zfs_cmd_t zc = {"\0"};
4014 boolean_t mounted = B_FALSE;
4015 char *mntpnt = NULL;
4016 char dsname[ZFS_MAX_DATASET_NAME_LEN];
4019 /* special case for the MOS */
4020 (void) snprintf(pathname, len, "<metadata>:<0x%llx>",
4025 /* get the dataset's name */
4026 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
4028 if (ioctl(zhp->zpool_hdl->libzfs_fd,
4029 ZFS_IOC_DSOBJ_TO_DSNAME, &zc) != 0) {
4030 /* just write out a path of two object numbers */
4031 (void) snprintf(pathname, len, "<0x%llx>:<0x%llx>",
4032 (longlong_t)dsobj, (longlong_t)obj);
4035 (void) strlcpy(dsname, zc.zc_value, sizeof (dsname));
4037 /* find out if the dataset is mounted */
4038 mounted = is_mounted(zhp->zpool_hdl, dsname, &mntpnt);
4040 /* get the corrupted object's path */
4041 (void) strlcpy(zc.zc_name, dsname, sizeof (zc.zc_name));
4043 if (ioctl(zhp->zpool_hdl->libzfs_fd, ZFS_IOC_OBJ_TO_PATH,
4046 (void) snprintf(pathname, len, "%s%s", mntpnt,
4049 (void) snprintf(pathname, len, "%s:%s",
4050 dsname, zc.zc_value);
4053 (void) snprintf(pathname, len, "%s:<0x%llx>", dsname,
4060 * Read the EFI label from the config, if a label does not exist then
4061 * pass back the error to the caller. If the caller has passed a non-NULL
4062 * diskaddr argument then we set it to the starting address of the EFI
4066 read_efi_label(nvlist_t *config, diskaddr_t *sb)
4070 char diskname[MAXPATHLEN];
4073 if (nvlist_lookup_string(config, ZPOOL_CONFIG_PATH, &path) != 0)
4076 (void) snprintf(diskname, sizeof (diskname), "%s%s", DISK_ROOT,
4077 strrchr(path, '/'));
4078 if ((fd = open(diskname, O_RDWR|O_DIRECT)) >= 0) {
4079 struct dk_gpt *vtoc;
4081 if ((err = efi_alloc_and_read(fd, &vtoc)) >= 0) {
4083 *sb = vtoc->efi_parts[0].p_start;
4092 * determine where a partition starts on a disk in the current
4096 find_start_block(nvlist_t *config)
4100 diskaddr_t sb = MAXOFFSET_T;
4103 if (nvlist_lookup_nvlist_array(config,
4104 ZPOOL_CONFIG_CHILDREN, &child, &children) != 0) {
4105 if (nvlist_lookup_uint64(config,
4106 ZPOOL_CONFIG_WHOLE_DISK,
4107 &wholedisk) != 0 || !wholedisk) {
4108 return (MAXOFFSET_T);
4110 if (read_efi_label(config, &sb) < 0)
4115 for (c = 0; c < children; c++) {
4116 sb = find_start_block(child[c]);
4117 if (sb != MAXOFFSET_T) {
4121 return (MAXOFFSET_T);
4125 zpool_label_disk_check(char *path)
4127 struct dk_gpt *vtoc;
4130 if ((fd = open(path, O_RDWR|O_DIRECT)) < 0)
4133 if ((err = efi_alloc_and_read(fd, &vtoc)) != 0) {
4138 if (vtoc->efi_flags & EFI_GPT_PRIMARY_CORRUPT) {
4150 * Generate a unique partition name for the ZFS member. Partitions must
4151 * have unique names to ensure udev will be able to create symlinks under
4152 * /dev/disk/by-partlabel/ for all pool members. The partition names are
4153 * of the form <pool>-<unique-id>.
4156 zpool_label_name(char *label_name, int label_size)
4161 fd = open("/dev/urandom", O_RDONLY);
4163 if (read(fd, &id, sizeof (id)) != sizeof (id))
4170 id = (((uint64_t)rand()) << 32) | (uint64_t)rand();
4172 snprintf(label_name, label_size, "zfs-%016llx", (u_longlong_t) id);
4176 * Label an individual disk. The name provided is the short name,
4177 * stripped of any leading /dev path.
4180 zpool_label_disk(libzfs_handle_t *hdl, zpool_handle_t *zhp, char *name)
4182 char path[MAXPATHLEN];
4183 struct dk_gpt *vtoc;
4185 size_t resv = EFI_MIN_RESV_SIZE;
4186 uint64_t slice_size;
4187 diskaddr_t start_block;
4190 /* prepare an error message just in case */
4191 (void) snprintf(errbuf, sizeof (errbuf),
4192 dgettext(TEXT_DOMAIN, "cannot label '%s'"), name);
4197 #if defined(__sun__) || defined(__sun)
4198 if (zpool_is_bootable(zhp)) {
4199 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
4200 "EFI labeled devices are not supported on root "
4202 return (zfs_error(hdl, EZFS_POOL_NOTSUP, errbuf));
4206 verify(nvlist_lookup_nvlist(zhp->zpool_config,
4207 ZPOOL_CONFIG_VDEV_TREE, &nvroot) == 0);
4209 if (zhp->zpool_start_block == 0)
4210 start_block = find_start_block(nvroot);
4212 start_block = zhp->zpool_start_block;
4213 zhp->zpool_start_block = start_block;
4216 start_block = NEW_START_BLOCK;
4219 (void) snprintf(path, sizeof (path), "%s/%s", DISK_ROOT, name);
4221 if ((fd = open(path, O_RDWR|O_DIRECT|O_EXCL)) < 0) {
4223 * This shouldn't happen. We've long since verified that this
4224 * is a valid device.
4226 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "cannot "
4227 "label '%s': unable to open device: %d"), path, errno);
4228 return (zfs_error(hdl, EZFS_OPENFAILED, errbuf));
4231 if (efi_alloc_and_init(fd, EFI_NUMPAR, &vtoc) != 0) {
4233 * The only way this can fail is if we run out of memory, or we
4234 * were unable to read the disk's capacity
4236 if (errno == ENOMEM)
4237 (void) no_memory(hdl);
4240 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "cannot "
4241 "label '%s': unable to read disk capacity"), path);
4243 return (zfs_error(hdl, EZFS_NOCAP, errbuf));
4246 slice_size = vtoc->efi_last_u_lba + 1;
4247 slice_size -= EFI_MIN_RESV_SIZE;
4248 if (start_block == MAXOFFSET_T)
4249 start_block = NEW_START_BLOCK;
4250 slice_size -= start_block;
4251 slice_size = P2ALIGN(slice_size, PARTITION_END_ALIGNMENT);
4253 vtoc->efi_parts[0].p_start = start_block;
4254 vtoc->efi_parts[0].p_size = slice_size;
4257 * Why we use V_USR: V_BACKUP confuses users, and is considered
4258 * disposable by some EFI utilities (since EFI doesn't have a backup
4259 * slice). V_UNASSIGNED is supposed to be used only for zero size
4260 * partitions, and efi_write() will fail if we use it. V_ROOT, V_BOOT,
4261 * etc. were all pretty specific. V_USR is as close to reality as we
4262 * can get, in the absence of V_OTHER.
4264 vtoc->efi_parts[0].p_tag = V_USR;
4265 zpool_label_name(vtoc->efi_parts[0].p_name, EFI_PART_NAME_LEN);
4267 vtoc->efi_parts[8].p_start = slice_size + start_block;
4268 vtoc->efi_parts[8].p_size = resv;
4269 vtoc->efi_parts[8].p_tag = V_RESERVED;
4271 if ((rval = efi_write(fd, vtoc)) != 0 || (rval = efi_rescan(fd)) != 0) {
4273 * Some block drivers (like pcata) may not support EFI
4274 * GPT labels. Print out a helpful error message dir-
4275 * ecting the user to manually label the disk and give
4281 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "try using "
4282 "parted(8) and then provide a specific slice: %d"), rval);
4283 return (zfs_error(hdl, EZFS_LABELFAILED, errbuf));
4289 (void) snprintf(path, sizeof (path), "%s/%s", DISK_ROOT, name);
4290 (void) zfs_append_partition(path, MAXPATHLEN);
4292 /* Wait to udev to signal use the device has settled. */
4293 rval = zpool_label_disk_wait(path, DISK_LABEL_WAIT);
4295 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "failed to "
4296 "detect device partitions on '%s': %d"), path, rval);
4297 return (zfs_error(hdl, EZFS_LABELFAILED, errbuf));
4300 /* We can't be to paranoid. Read the label back and verify it. */
4301 (void) snprintf(path, sizeof (path), "%s/%s", DISK_ROOT, name);
4302 rval = zpool_label_disk_check(path);
4304 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "freshly written "
4305 "EFI label on '%s' is damaged. Ensure\nthis device "
4306 "is not in in use, and is functioning properly: %d"),
4308 return (zfs_error(hdl, EZFS_LABELFAILED, errbuf));