4 * The contents of this file are subject to the terms of the
5 * Common Development and Distribution License (the "License").
6 * You may not use this file except in compliance with the License.
8 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
9 * or http://www.opensolaris.org/os/licensing.
10 * See the License for the specific language governing permissions
11 * and limitations under the License.
13 * When distributing Covered Code, include this CDDL HEADER in each
14 * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
15 * If applicable, add the following below this CDDL HEADER, with the
16 * fields enclosed by brackets "[]" replaced with your own identifying
17 * information: Portions Copyright [yyyy] [name of copyright owner]
23 * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
35 #include <sys/efi_partition.h>
37 #include <sys/zfs_ioctl.h>
40 #include "zfs_namecheck.h"
42 #include "libzfs_impl.h"
43 #include "zfs_comutil.h"
45 static int read_efi_label(nvlist_t *config, diskaddr_t *sb);
47 #if defined(__i386) || defined(__amd64)
48 #define BOOTCMD "installgrub(1M)"
50 #define BOOTCMD "installboot(1M)"
53 #define DISK_ROOT "/dev/dsk"
54 #define RDISK_ROOT "/dev/rdsk"
55 #define BACKUP_SLICE "s2"
58 * ====================================================================
59 * zpool property functions
60 * ====================================================================
64 zpool_get_all_props(zpool_handle_t *zhp)
67 libzfs_handle_t *hdl = zhp->zpool_hdl;
69 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
71 if (zcmd_alloc_dst_nvlist(hdl, &zc, 0) != 0)
74 while (ioctl(hdl->libzfs_fd, ZFS_IOC_POOL_GET_PROPS, &zc) != 0) {
75 if (errno == ENOMEM) {
76 if (zcmd_expand_dst_nvlist(hdl, &zc) != 0) {
77 zcmd_free_nvlists(&zc);
81 zcmd_free_nvlists(&zc);
86 if (zcmd_read_dst_nvlist(hdl, &zc, &zhp->zpool_props) != 0) {
87 zcmd_free_nvlists(&zc);
91 zcmd_free_nvlists(&zc);
97 zpool_props_refresh(zpool_handle_t *zhp)
101 old_props = zhp->zpool_props;
103 if (zpool_get_all_props(zhp) != 0)
106 nvlist_free(old_props);
111 zpool_get_prop_string(zpool_handle_t *zhp, zpool_prop_t prop,
117 zprop_source_t source;
119 nvl = zhp->zpool_props;
120 if (nvlist_lookup_nvlist(nvl, zpool_prop_to_name(prop), &nv) == 0) {
121 verify(nvlist_lookup_uint64(nv, ZPROP_SOURCE, &ival) == 0);
123 verify(nvlist_lookup_string(nv, ZPROP_VALUE, &value) == 0);
125 source = ZPROP_SRC_DEFAULT;
126 if ((value = (char *)zpool_prop_default_string(prop)) == NULL)
137 zpool_get_prop_int(zpool_handle_t *zhp, zpool_prop_t prop, zprop_source_t *src)
141 zprop_source_t source;
143 if (zhp->zpool_props == NULL && zpool_get_all_props(zhp)) {
145 * zpool_get_all_props() has most likely failed because
146 * the pool is faulted, but if all we need is the top level
147 * vdev's guid then get it from the zhp config nvlist.
149 if ((prop == ZPOOL_PROP_GUID) &&
150 (nvlist_lookup_nvlist(zhp->zpool_config,
151 ZPOOL_CONFIG_VDEV_TREE, &nv) == 0) &&
152 (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID, &value)
156 return (zpool_prop_default_numeric(prop));
159 nvl = zhp->zpool_props;
160 if (nvlist_lookup_nvlist(nvl, zpool_prop_to_name(prop), &nv) == 0) {
161 verify(nvlist_lookup_uint64(nv, ZPROP_SOURCE, &value) == 0);
163 verify(nvlist_lookup_uint64(nv, ZPROP_VALUE, &value) == 0);
165 source = ZPROP_SRC_DEFAULT;
166 value = zpool_prop_default_numeric(prop);
176 * Map VDEV STATE to printed strings.
179 zpool_state_to_name(vdev_state_t state, vdev_aux_t aux)
182 case VDEV_STATE_CLOSED:
183 case VDEV_STATE_OFFLINE:
184 return (gettext("OFFLINE"));
185 case VDEV_STATE_REMOVED:
186 return (gettext("REMOVED"));
187 case VDEV_STATE_CANT_OPEN:
188 if (aux == VDEV_AUX_CORRUPT_DATA || aux == VDEV_AUX_BAD_LOG)
189 return (gettext("FAULTED"));
190 else if (aux == VDEV_AUX_SPLIT_POOL)
191 return (gettext("SPLIT"));
193 return (gettext("UNAVAIL"));
194 case VDEV_STATE_FAULTED:
195 return (gettext("FAULTED"));
196 case VDEV_STATE_DEGRADED:
197 return (gettext("DEGRADED"));
198 case VDEV_STATE_HEALTHY:
199 return (gettext("ONLINE"));
202 return (gettext("UNKNOWN"));
206 * Get a zpool property value for 'prop' and return the value in
207 * a pre-allocated buffer.
210 zpool_get_prop(zpool_handle_t *zhp, zpool_prop_t prop, char *buf, size_t len,
211 zprop_source_t *srctype)
215 zprop_source_t src = ZPROP_SRC_NONE;
220 if (zpool_get_state(zhp) == POOL_STATE_UNAVAIL) {
222 case ZPOOL_PROP_NAME:
223 (void) strlcpy(buf, zpool_get_name(zhp), len);
226 case ZPOOL_PROP_HEALTH:
227 (void) strlcpy(buf, "FAULTED", len);
230 case ZPOOL_PROP_GUID:
231 intval = zpool_get_prop_int(zhp, prop, &src);
232 (void) snprintf(buf, len, "%llu", intval);
235 case ZPOOL_PROP_ALTROOT:
236 case ZPOOL_PROP_CACHEFILE:
237 if (zhp->zpool_props != NULL ||
238 zpool_get_all_props(zhp) == 0) {
240 zpool_get_prop_string(zhp, prop, &src),
248 (void) strlcpy(buf, "-", len);
257 if (zhp->zpool_props == NULL && zpool_get_all_props(zhp) &&
258 prop != ZPOOL_PROP_NAME)
261 switch (zpool_prop_get_type(prop)) {
262 case PROP_TYPE_STRING:
263 (void) strlcpy(buf, zpool_get_prop_string(zhp, prop, &src),
267 case PROP_TYPE_NUMBER:
268 intval = zpool_get_prop_int(zhp, prop, &src);
271 case ZPOOL_PROP_SIZE:
272 case ZPOOL_PROP_ALLOCATED:
273 case ZPOOL_PROP_FREE:
274 (void) zfs_nicenum(intval, buf, len);
277 case ZPOOL_PROP_CAPACITY:
278 (void) snprintf(buf, len, "%llu%%",
279 (u_longlong_t)intval);
282 case ZPOOL_PROP_DEDUPRATIO:
283 (void) snprintf(buf, len, "%llu.%02llux",
284 (u_longlong_t)(intval / 100),
285 (u_longlong_t)(intval % 100));
288 case ZPOOL_PROP_HEALTH:
289 verify(nvlist_lookup_nvlist(zpool_get_config(zhp, NULL),
290 ZPOOL_CONFIG_VDEV_TREE, &nvroot) == 0);
291 verify(nvlist_lookup_uint64_array(nvroot,
292 ZPOOL_CONFIG_VDEV_STATS, (uint64_t **)&vs, &vsc)
295 (void) strlcpy(buf, zpool_state_to_name(intval,
299 (void) snprintf(buf, len, "%llu", intval);
303 case PROP_TYPE_INDEX:
304 intval = zpool_get_prop_int(zhp, prop, &src);
305 if (zpool_prop_index_to_string(prop, intval, &strval)
308 (void) strlcpy(buf, strval, len);
322 * Check if the bootfs name has the same pool name as it is set to.
323 * Assuming bootfs is a valid dataset name.
326 bootfs_name_valid(const char *pool, char *bootfs)
328 int len = strlen(pool);
330 if (!zfs_name_valid(bootfs, ZFS_TYPE_FILESYSTEM|ZFS_TYPE_SNAPSHOT))
333 if (strncmp(pool, bootfs, len) == 0 &&
334 (bootfs[len] == '/' || bootfs[len] == '\0'))
341 * Inspect the configuration to determine if any of the devices contain
345 pool_uses_efi(nvlist_t *config)
350 if (nvlist_lookup_nvlist_array(config, ZPOOL_CONFIG_CHILDREN,
351 &child, &children) != 0)
352 return (read_efi_label(config, NULL) >= 0);
354 for (c = 0; c < children; c++) {
355 if (pool_uses_efi(child[c]))
362 pool_is_bootable(zpool_handle_t *zhp)
364 char bootfs[ZPOOL_MAXNAMELEN];
366 return (zpool_get_prop(zhp, ZPOOL_PROP_BOOTFS, bootfs,
367 sizeof (bootfs), NULL) == 0 && strncmp(bootfs, "-",
368 sizeof (bootfs)) != 0);
373 * Given an nvlist of zpool properties to be set, validate that they are
374 * correct, and parse any numeric properties (index, boolean, etc) if they are
375 * specified as strings.
378 zpool_valid_proplist(libzfs_handle_t *hdl, const char *poolname,
379 nvlist_t *props, uint64_t version, boolean_t create_or_import, char *errbuf)
387 struct stat64 statbuf;
391 if (nvlist_alloc(&retprops, NV_UNIQUE_NAME, 0) != 0) {
392 (void) no_memory(hdl);
397 while ((elem = nvlist_next_nvpair(props, elem)) != NULL) {
398 const char *propname = nvpair_name(elem);
401 * Make sure this property is valid and applies to this type.
403 if ((prop = zpool_name_to_prop(propname)) == ZPROP_INVAL) {
404 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
405 "invalid property '%s'"), propname);
406 (void) zfs_error(hdl, EZFS_BADPROP, errbuf);
410 if (zpool_prop_readonly(prop)) {
411 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "'%s' "
412 "is readonly"), propname);
413 (void) zfs_error(hdl, EZFS_PROPREADONLY, errbuf);
417 if (zprop_parse_value(hdl, elem, prop, ZFS_TYPE_POOL, retprops,
418 &strval, &intval, errbuf) != 0)
422 * Perform additional checking for specific properties.
425 case ZPOOL_PROP_VERSION:
426 if (intval < version || intval > SPA_VERSION) {
427 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
428 "property '%s' number %d is invalid."),
430 (void) zfs_error(hdl, EZFS_BADVERSION, errbuf);
435 case ZPOOL_PROP_BOOTFS:
436 if (create_or_import) {
437 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
438 "property '%s' cannot be set at creation "
439 "or import time"), propname);
440 (void) zfs_error(hdl, EZFS_BADPROP, errbuf);
444 if (version < SPA_VERSION_BOOTFS) {
445 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
446 "pool must be upgraded to support "
447 "'%s' property"), propname);
448 (void) zfs_error(hdl, EZFS_BADVERSION, errbuf);
453 * bootfs property value has to be a dataset name and
454 * the dataset has to be in the same pool as it sets to.
456 if (strval[0] != '\0' && !bootfs_name_valid(poolname,
458 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "'%s' "
459 "is an invalid name"), strval);
460 (void) zfs_error(hdl, EZFS_INVALIDNAME, errbuf);
464 if ((zhp = zpool_open_canfail(hdl, poolname)) == NULL) {
465 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
466 "could not open pool '%s'"), poolname);
467 (void) zfs_error(hdl, EZFS_OPENFAILED, errbuf);
470 verify(nvlist_lookup_nvlist(zpool_get_config(zhp, NULL),
471 ZPOOL_CONFIG_VDEV_TREE, &nvroot) == 0);
474 * bootfs property cannot be set on a disk which has
477 if (pool_uses_efi(nvroot)) {
478 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
479 "property '%s' not supported on "
480 "EFI labeled devices"), propname);
481 (void) zfs_error(hdl, EZFS_POOL_NOTSUP, errbuf);
488 case ZPOOL_PROP_ALTROOT:
489 if (!create_or_import) {
490 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
491 "property '%s' can only be set during pool "
492 "creation or import"), propname);
493 (void) zfs_error(hdl, EZFS_BADPROP, errbuf);
497 if (strval[0] != '/') {
498 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
499 "bad alternate root '%s'"), strval);
500 (void) zfs_error(hdl, EZFS_BADPATH, errbuf);
505 case ZPOOL_PROP_CACHEFILE:
506 if (strval[0] == '\0')
509 if (strcmp(strval, "none") == 0)
512 if (strval[0] != '/') {
513 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
514 "property '%s' must be empty, an "
515 "absolute path, or 'none'"), propname);
516 (void) zfs_error(hdl, EZFS_BADPATH, errbuf);
520 slash = strrchr(strval, '/');
522 if (slash[1] == '\0' || strcmp(slash, "/.") == 0 ||
523 strcmp(slash, "/..") == 0) {
524 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
525 "'%s' is not a valid file"), strval);
526 (void) zfs_error(hdl, EZFS_BADPATH, errbuf);
532 if (strval[0] != '\0' &&
533 (stat64(strval, &statbuf) != 0 ||
534 !S_ISDIR(statbuf.st_mode))) {
535 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
536 "'%s' is not a valid directory"),
538 (void) zfs_error(hdl, EZFS_BADPATH, errbuf);
549 nvlist_free(retprops);
554 * Set zpool property : propname=propval.
557 zpool_set_prop(zpool_handle_t *zhp, const char *propname, const char *propval)
559 zfs_cmd_t zc = { 0 };
562 nvlist_t *nvl = NULL;
566 (void) snprintf(errbuf, sizeof (errbuf),
567 dgettext(TEXT_DOMAIN, "cannot set property for '%s'"),
570 if (nvlist_alloc(&nvl, NV_UNIQUE_NAME, 0) != 0)
571 return (no_memory(zhp->zpool_hdl));
573 if (nvlist_add_string(nvl, propname, propval) != 0) {
575 return (no_memory(zhp->zpool_hdl));
578 version = zpool_get_prop_int(zhp, ZPOOL_PROP_VERSION, NULL);
579 if ((realprops = zpool_valid_proplist(zhp->zpool_hdl,
580 zhp->zpool_name, nvl, version, B_FALSE, errbuf)) == NULL) {
589 * Execute the corresponding ioctl() to set this property.
591 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
593 if (zcmd_write_src_nvlist(zhp->zpool_hdl, &zc, nvl) != 0) {
598 ret = zfs_ioctl(zhp->zpool_hdl, ZFS_IOC_POOL_SET_PROPS, &zc);
600 zcmd_free_nvlists(&zc);
604 (void) zpool_standard_error(zhp->zpool_hdl, errno, errbuf);
606 (void) zpool_props_refresh(zhp);
612 zpool_expand_proplist(zpool_handle_t *zhp, zprop_list_t **plp)
614 libzfs_handle_t *hdl = zhp->zpool_hdl;
616 char buf[ZFS_MAXPROPLEN];
618 if (zprop_expand_list(hdl, plp, ZFS_TYPE_POOL) != 0)
621 for (entry = *plp; entry != NULL; entry = entry->pl_next) {
626 if (entry->pl_prop != ZPROP_INVAL &&
627 zpool_get_prop(zhp, entry->pl_prop, buf, sizeof (buf),
629 if (strlen(buf) > entry->pl_width)
630 entry->pl_width = strlen(buf);
639 * Don't start the slice at the default block of 34; many storage
640 * devices will use a stripe width of 128k, so start there instead.
642 #define NEW_START_BLOCK 256
645 * Validate the given pool name, optionally putting an extended error message in
649 zpool_name_valid(libzfs_handle_t *hdl, boolean_t isopen, const char *pool)
655 ret = pool_namecheck(pool, &why, &what);
658 * The rules for reserved pool names were extended at a later point.
659 * But we need to support users with existing pools that may now be
660 * invalid. So we only check for this expanded set of names during a
661 * create (or import), and only in userland.
663 if (ret == 0 && !isopen &&
664 (strncmp(pool, "mirror", 6) == 0 ||
665 strncmp(pool, "raidz", 5) == 0 ||
666 strncmp(pool, "spare", 5) == 0 ||
667 strcmp(pool, "log") == 0)) {
670 dgettext(TEXT_DOMAIN, "name is reserved"));
678 case NAME_ERR_TOOLONG:
680 dgettext(TEXT_DOMAIN, "name is too long"));
683 case NAME_ERR_INVALCHAR:
685 dgettext(TEXT_DOMAIN, "invalid character "
686 "'%c' in pool name"), what);
689 case NAME_ERR_NOLETTER:
690 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
691 "name must begin with a letter"));
694 case NAME_ERR_RESERVED:
695 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
696 "name is reserved"));
699 case NAME_ERR_DISKLIKE:
700 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
701 "pool name is reserved"));
704 case NAME_ERR_LEADING_SLASH:
705 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
706 "leading slash in name"));
709 case NAME_ERR_EMPTY_COMPONENT:
710 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
711 "empty component in name"));
714 case NAME_ERR_TRAILING_SLASH:
715 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
716 "trailing slash in name"));
719 case NAME_ERR_MULTIPLE_AT:
720 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
721 "multiple '@' delimiters in name"));
733 * Open a handle to the given pool, even if the pool is currently in the FAULTED
737 zpool_open_canfail(libzfs_handle_t *hdl, const char *pool)
743 * Make sure the pool name is valid.
745 if (!zpool_name_valid(hdl, B_TRUE, pool)) {
746 (void) zfs_error_fmt(hdl, EZFS_INVALIDNAME,
747 dgettext(TEXT_DOMAIN, "cannot open '%s'"),
752 if ((zhp = zfs_alloc(hdl, sizeof (zpool_handle_t))) == NULL)
755 zhp->zpool_hdl = hdl;
756 (void) strlcpy(zhp->zpool_name, pool, sizeof (zhp->zpool_name));
758 if (zpool_refresh_stats(zhp, &missing) != 0) {
764 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "no such pool"));
765 (void) zfs_error_fmt(hdl, EZFS_NOENT,
766 dgettext(TEXT_DOMAIN, "cannot open '%s'"), pool);
775 * Like the above, but silent on error. Used when iterating over pools (because
776 * the configuration cache may be out of date).
779 zpool_open_silent(libzfs_handle_t *hdl, const char *pool, zpool_handle_t **ret)
784 if ((zhp = zfs_alloc(hdl, sizeof (zpool_handle_t))) == NULL)
787 zhp->zpool_hdl = hdl;
788 (void) strlcpy(zhp->zpool_name, pool, sizeof (zhp->zpool_name));
790 if (zpool_refresh_stats(zhp, &missing) != 0) {
806 * Similar to zpool_open_canfail(), but refuses to open pools in the faulted
810 zpool_open(libzfs_handle_t *hdl, const char *pool)
814 if ((zhp = zpool_open_canfail(hdl, pool)) == NULL)
817 if (zhp->zpool_state == POOL_STATE_UNAVAIL) {
818 (void) zfs_error_fmt(hdl, EZFS_POOLUNAVAIL,
819 dgettext(TEXT_DOMAIN, "cannot open '%s'"), zhp->zpool_name);
828 * Close the handle. Simply frees the memory associated with the handle.
831 zpool_close(zpool_handle_t *zhp)
833 if (zhp->zpool_config)
834 nvlist_free(zhp->zpool_config);
835 if (zhp->zpool_old_config)
836 nvlist_free(zhp->zpool_old_config);
837 if (zhp->zpool_props)
838 nvlist_free(zhp->zpool_props);
843 * Return the name of the pool.
846 zpool_get_name(zpool_handle_t *zhp)
848 return (zhp->zpool_name);
853 * Return the state of the pool (ACTIVE or UNAVAILABLE)
856 zpool_get_state(zpool_handle_t *zhp)
858 return (zhp->zpool_state);
862 * Create the named pool, using the provided vdev list. It is assumed
863 * that the consumer has already validated the contents of the nvlist, so we
864 * don't have to worry about error semantics.
867 zpool_create(libzfs_handle_t *hdl, const char *pool, nvlist_t *nvroot,
868 nvlist_t *props, nvlist_t *fsprops)
870 zfs_cmd_t zc = { 0 };
871 nvlist_t *zc_fsprops = NULL;
872 nvlist_t *zc_props = NULL;
877 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN,
878 "cannot create '%s'"), pool);
880 if (!zpool_name_valid(hdl, B_FALSE, pool))
881 return (zfs_error(hdl, EZFS_INVALIDNAME, msg));
883 if (zcmd_write_conf_nvlist(hdl, &zc, nvroot) != 0)
887 if ((zc_props = zpool_valid_proplist(hdl, pool, props,
888 SPA_VERSION_1, B_TRUE, msg)) == NULL) {
897 zoned = ((nvlist_lookup_string(fsprops,
898 zfs_prop_to_name(ZFS_PROP_ZONED), &zonestr) == 0) &&
899 strcmp(zonestr, "on") == 0);
901 if ((zc_fsprops = zfs_valid_proplist(hdl,
902 ZFS_TYPE_FILESYSTEM, fsprops, zoned, NULL, msg)) == NULL) {
906 (nvlist_alloc(&zc_props, NV_UNIQUE_NAME, 0) != 0)) {
909 if (nvlist_add_nvlist(zc_props,
910 ZPOOL_ROOTFS_PROPS, zc_fsprops) != 0) {
915 if (zc_props && zcmd_write_src_nvlist(hdl, &zc, zc_props) != 0)
918 (void) strlcpy(zc.zc_name, pool, sizeof (zc.zc_name));
920 if ((ret = zfs_ioctl(hdl, ZFS_IOC_POOL_CREATE, &zc)) != 0) {
922 zcmd_free_nvlists(&zc);
923 nvlist_free(zc_props);
924 nvlist_free(zc_fsprops);
929 * This can happen if the user has specified the same
930 * device multiple times. We can't reliably detect this
931 * until we try to add it and see we already have a
934 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
935 "one or more vdevs refer to the same device"));
936 return (zfs_error(hdl, EZFS_BADDEV, msg));
940 * This occurs when one of the devices is below
941 * SPA_MINDEVSIZE. Unfortunately, we can't detect which
942 * device was the problem device since there's no
943 * reliable way to determine device size from userland.
948 zfs_nicenum(SPA_MINDEVSIZE, buf, sizeof (buf));
950 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
951 "one or more devices is less than the "
952 "minimum size (%s)"), buf);
954 return (zfs_error(hdl, EZFS_BADDEV, msg));
957 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
958 "one or more devices is out of space"));
959 return (zfs_error(hdl, EZFS_BADDEV, msg));
962 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
963 "cache device must be a disk or disk slice"));
964 return (zfs_error(hdl, EZFS_BADDEV, msg));
967 return (zpool_standard_error(hdl, errno, msg));
972 * If this is an alternate root pool, then we automatically set the
973 * mountpoint of the root dataset to be '/'.
975 if (nvlist_lookup_string(props, zpool_prop_to_name(ZPOOL_PROP_ALTROOT),
979 verify((zhp = zfs_open(hdl, pool, ZFS_TYPE_DATASET)) != NULL);
980 verify(zfs_prop_set(zhp, zfs_prop_to_name(ZFS_PROP_MOUNTPOINT),
987 zcmd_free_nvlists(&zc);
988 nvlist_free(zc_props);
989 nvlist_free(zc_fsprops);
994 * Destroy the given pool. It is up to the caller to ensure that there are no
995 * datasets left in the pool.
998 zpool_destroy(zpool_handle_t *zhp)
1000 zfs_cmd_t zc = { 0 };
1001 zfs_handle_t *zfp = NULL;
1002 libzfs_handle_t *hdl = zhp->zpool_hdl;
1005 if (zhp->zpool_state == POOL_STATE_ACTIVE &&
1006 (zfp = zfs_open(zhp->zpool_hdl, zhp->zpool_name,
1007 ZFS_TYPE_FILESYSTEM)) == NULL)
1010 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
1012 if (zfs_ioctl(zhp->zpool_hdl, ZFS_IOC_POOL_DESTROY, &zc) != 0) {
1013 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN,
1014 "cannot destroy '%s'"), zhp->zpool_name);
1016 if (errno == EROFS) {
1017 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1018 "one or more devices is read only"));
1019 (void) zfs_error(hdl, EZFS_BADDEV, msg);
1021 (void) zpool_standard_error(hdl, errno, msg);
1030 remove_mountpoint(zfp);
1038 * Add the given vdevs to the pool. The caller must have already performed the
1039 * necessary verification to ensure that the vdev specification is well-formed.
1042 zpool_add(zpool_handle_t *zhp, nvlist_t *nvroot)
1044 zfs_cmd_t zc = { 0 };
1046 libzfs_handle_t *hdl = zhp->zpool_hdl;
1048 nvlist_t **spares, **l2cache;
1049 uint_t nspares, nl2cache;
1051 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN,
1052 "cannot add to '%s'"), zhp->zpool_name);
1054 if (zpool_get_prop_int(zhp, ZPOOL_PROP_VERSION, NULL) <
1055 SPA_VERSION_SPARES &&
1056 nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_SPARES,
1057 &spares, &nspares) == 0) {
1058 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "pool must be "
1059 "upgraded to add hot spares"));
1060 return (zfs_error(hdl, EZFS_BADVERSION, msg));
1063 if (pool_is_bootable(zhp) && nvlist_lookup_nvlist_array(nvroot,
1064 ZPOOL_CONFIG_SPARES, &spares, &nspares) == 0) {
1067 for (s = 0; s < nspares; s++) {
1070 if (nvlist_lookup_string(spares[s], ZPOOL_CONFIG_PATH,
1071 &path) == 0 && pool_uses_efi(spares[s])) {
1072 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1073 "device '%s' contains an EFI label and "
1074 "cannot be used on root pools."),
1075 zpool_vdev_name(hdl, NULL, spares[s],
1077 return (zfs_error(hdl, EZFS_POOL_NOTSUP, msg));
1082 if (zpool_get_prop_int(zhp, ZPOOL_PROP_VERSION, NULL) <
1083 SPA_VERSION_L2CACHE &&
1084 nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_L2CACHE,
1085 &l2cache, &nl2cache) == 0) {
1086 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "pool must be "
1087 "upgraded to add cache devices"));
1088 return (zfs_error(hdl, EZFS_BADVERSION, msg));
1091 if (zcmd_write_conf_nvlist(hdl, &zc, nvroot) != 0)
1093 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
1095 if (zfs_ioctl(zhp->zpool_hdl, ZFS_IOC_VDEV_ADD, &zc) != 0) {
1099 * This can happen if the user has specified the same
1100 * device multiple times. We can't reliably detect this
1101 * until we try to add it and see we already have a
1104 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1105 "one or more vdevs refer to the same device"));
1106 (void) zfs_error(hdl, EZFS_BADDEV, msg);
1111 * This occurrs when one of the devices is below
1112 * SPA_MINDEVSIZE. Unfortunately, we can't detect which
1113 * device was the problem device since there's no
1114 * reliable way to determine device size from userland.
1119 zfs_nicenum(SPA_MINDEVSIZE, buf, sizeof (buf));
1121 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1122 "device is less than the minimum "
1125 (void) zfs_error(hdl, EZFS_BADDEV, msg);
1129 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1130 "pool must be upgraded to add these vdevs"));
1131 (void) zfs_error(hdl, EZFS_BADVERSION, msg);
1135 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1136 "root pool can not have multiple vdevs"
1137 " or separate logs"));
1138 (void) zfs_error(hdl, EZFS_POOL_NOTSUP, msg);
1142 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1143 "cache device must be a disk or disk slice"));
1144 (void) zfs_error(hdl, EZFS_BADDEV, msg);
1148 (void) zpool_standard_error(hdl, errno, msg);
1156 zcmd_free_nvlists(&zc);
1162 * Exports the pool from the system. The caller must ensure that there are no
1163 * mounted datasets in the pool.
1166 zpool_export_common(zpool_handle_t *zhp, boolean_t force, boolean_t hardforce)
1168 zfs_cmd_t zc = { 0 };
1171 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN,
1172 "cannot export '%s'"), zhp->zpool_name);
1174 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
1175 zc.zc_cookie = force;
1176 zc.zc_guid = hardforce;
1178 if (zfs_ioctl(zhp->zpool_hdl, ZFS_IOC_POOL_EXPORT, &zc) != 0) {
1181 zfs_error_aux(zhp->zpool_hdl, dgettext(TEXT_DOMAIN,
1182 "use '-f' to override the following errors:\n"
1183 "'%s' has an active shared spare which could be"
1184 " used by other pools once '%s' is exported."),
1185 zhp->zpool_name, zhp->zpool_name);
1186 return (zfs_error(zhp->zpool_hdl, EZFS_ACTIVE_SPARE,
1189 return (zpool_standard_error_fmt(zhp->zpool_hdl, errno,
1198 zpool_export(zpool_handle_t *zhp, boolean_t force)
1200 return (zpool_export_common(zhp, force, B_FALSE));
1204 zpool_export_force(zpool_handle_t *zhp)
1206 return (zpool_export_common(zhp, B_TRUE, B_TRUE));
1210 zpool_rewind_exclaim(libzfs_handle_t *hdl, const char *name, boolean_t dryrun,
1218 if (!hdl->libzfs_printerr || rbi == NULL)
1221 if (nvlist_lookup_uint64(rbi, ZPOOL_CONFIG_LOAD_TIME, &rewindto) != 0)
1223 (void) nvlist_lookup_int64(rbi, ZPOOL_CONFIG_REWIND_TIME, &loss);
1225 if (localtime_r((time_t *)&rewindto, &t) != NULL &&
1226 strftime(timestr, 128, 0, &t) != 0) {
1228 (void) printf(dgettext(TEXT_DOMAIN,
1229 "Would be able to return %s "
1230 "to its state as of %s.\n"),
1233 (void) printf(dgettext(TEXT_DOMAIN,
1234 "Pool %s returned to its state as of %s.\n"),
1238 (void) printf(dgettext(TEXT_DOMAIN,
1239 "%s approximately %lld "),
1240 dryrun ? "Would discard" : "Discarded",
1242 (void) printf(dgettext(TEXT_DOMAIN,
1243 "minutes of transactions.\n"));
1244 } else if (loss > 0) {
1245 (void) printf(dgettext(TEXT_DOMAIN,
1246 "%s approximately %lld "),
1247 dryrun ? "Would discard" : "Discarded", loss);
1248 (void) printf(dgettext(TEXT_DOMAIN,
1249 "seconds of transactions.\n"));
1255 zpool_explain_recover(libzfs_handle_t *hdl, const char *name, int reason,
1259 uint64_t edata = UINT64_MAX;
1264 if (!hdl->libzfs_printerr)
1268 (void) printf(dgettext(TEXT_DOMAIN, "action: "));
1270 (void) printf(dgettext(TEXT_DOMAIN, "\t"));
1272 /* All attempted rewinds failed if ZPOOL_CONFIG_LOAD_TIME missing */
1273 if (nvlist_lookup_uint64(config,
1274 ZPOOL_CONFIG_LOAD_TIME, &rewindto) != 0)
1277 (void) nvlist_lookup_int64(config, ZPOOL_CONFIG_REWIND_TIME, &loss);
1278 (void) nvlist_lookup_uint64(config, ZPOOL_CONFIG_LOAD_DATA_ERRORS,
1281 (void) printf(dgettext(TEXT_DOMAIN,
1282 "Recovery is possible, but will result in some data loss.\n"));
1284 if (localtime_r((time_t *)&rewindto, &t) != NULL &&
1285 strftime(timestr, 128, 0, &t) != 0) {
1286 (void) printf(dgettext(TEXT_DOMAIN,
1287 "\tReturning the pool to its state as of %s\n"
1288 "\tshould correct the problem. "),
1291 (void) printf(dgettext(TEXT_DOMAIN,
1292 "\tReverting the pool to an earlier state "
1293 "should correct the problem.\n\t"));
1297 (void) printf(dgettext(TEXT_DOMAIN,
1298 "Approximately %lld minutes of data\n"
1299 "\tmust be discarded, irreversibly. "), (loss + 30) / 60);
1300 } else if (loss > 0) {
1301 (void) printf(dgettext(TEXT_DOMAIN,
1302 "Approximately %lld seconds of data\n"
1303 "\tmust be discarded, irreversibly. "), loss);
1305 if (edata != 0 && edata != UINT64_MAX) {
1307 (void) printf(dgettext(TEXT_DOMAIN,
1308 "After rewind, at least\n"
1309 "\tone persistent user-data error will remain. "));
1311 (void) printf(dgettext(TEXT_DOMAIN,
1312 "After rewind, several\n"
1313 "\tpersistent user-data errors will remain. "));
1316 (void) printf(dgettext(TEXT_DOMAIN,
1317 "Recovery can be attempted\n\tby executing 'zpool %s -F %s'. "),
1318 reason >= 0 ? "clear" : "import", name);
1320 (void) printf(dgettext(TEXT_DOMAIN,
1321 "A scrub of the pool\n"
1322 "\tis strongly recommended after recovery.\n"));
1326 (void) printf(dgettext(TEXT_DOMAIN,
1327 "Destroy and re-create the pool from\n\ta backup source.\n"));
1331 * zpool_import() is a contracted interface. Should be kept the same
1334 * Applications should use zpool_import_props() to import a pool with
1335 * new properties value to be set.
1338 zpool_import(libzfs_handle_t *hdl, nvlist_t *config, const char *newname,
1341 nvlist_t *props = NULL;
1344 if (altroot != NULL) {
1345 if (nvlist_alloc(&props, NV_UNIQUE_NAME, 0) != 0) {
1346 return (zfs_error_fmt(hdl, EZFS_NOMEM,
1347 dgettext(TEXT_DOMAIN, "cannot import '%s'"),
1351 if (nvlist_add_string(props,
1352 zpool_prop_to_name(ZPOOL_PROP_ALTROOT), altroot) != 0 ||
1353 nvlist_add_string(props,
1354 zpool_prop_to_name(ZPOOL_PROP_CACHEFILE), "none") != 0) {
1356 return (zfs_error_fmt(hdl, EZFS_NOMEM,
1357 dgettext(TEXT_DOMAIN, "cannot import '%s'"),
1362 ret = zpool_import_props(hdl, config, newname, props, B_FALSE);
1369 * Import the given pool using the known configuration and a list of
1370 * properties to be set. The configuration should have come from
1371 * zpool_find_import(). The 'newname' parameters control whether the pool
1372 * is imported with a different name.
1375 zpool_import_props(libzfs_handle_t *hdl, nvlist_t *config, const char *newname,
1376 nvlist_t *props, boolean_t importfaulted)
1378 zfs_cmd_t zc = { 0 };
1379 zpool_rewind_policy_t policy;
1380 nvlist_t *nvi = NULL;
1383 uint64_t returned_size;
1387 verify(nvlist_lookup_string(config, ZPOOL_CONFIG_POOL_NAME,
1390 (void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN,
1391 "cannot import pool '%s'"), origname);
1393 if (newname != NULL) {
1394 if (!zpool_name_valid(hdl, B_FALSE, newname))
1395 return (zfs_error_fmt(hdl, EZFS_INVALIDNAME,
1396 dgettext(TEXT_DOMAIN, "cannot import '%s'"),
1398 thename = (char *)newname;
1406 verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_VERSION,
1409 if ((props = zpool_valid_proplist(hdl, origname,
1410 props, version, B_TRUE, errbuf)) == NULL) {
1412 } else if (zcmd_write_src_nvlist(hdl, &zc, props) != 0) {
1418 (void) strlcpy(zc.zc_name, thename, sizeof (zc.zc_name));
1420 verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID,
1423 if (zcmd_write_conf_nvlist(hdl, &zc, config) != 0) {
1427 returned_size = zc.zc_nvlist_conf_size + 512;
1428 if (zcmd_alloc_dst_nvlist(hdl, &zc, returned_size) != 0) {
1433 zc.zc_cookie = (uint64_t)importfaulted;
1435 if (zfs_ioctl(hdl, ZFS_IOC_POOL_IMPORT, &zc) != 0) {
1438 (void) zcmd_read_dst_nvlist(hdl, &zc, &nvi);
1439 zpool_get_rewind_policy(config, &policy);
1441 * Dry-run failed, but we print out what success
1442 * looks like if we found a best txg
1444 if ((policy.zrp_request & ZPOOL_TRY_REWIND) && nvi) {
1445 zpool_rewind_exclaim(hdl, newname ? origname : thename,
1451 if (newname == NULL)
1452 (void) snprintf(desc, sizeof (desc),
1453 dgettext(TEXT_DOMAIN, "cannot import '%s'"),
1456 (void) snprintf(desc, sizeof (desc),
1457 dgettext(TEXT_DOMAIN, "cannot import '%s' as '%s'"),
1463 * Unsupported version.
1465 (void) zfs_error(hdl, EZFS_BADVERSION, desc);
1469 (void) zfs_error(hdl, EZFS_INVALCONFIG, desc);
1473 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1474 "one or more devices is read only"));
1475 (void) zfs_error(hdl, EZFS_BADDEV, desc);
1479 (void) zcmd_read_dst_nvlist(hdl, &zc, &nvi);
1480 (void) zpool_standard_error(hdl, errno, desc);
1481 zpool_explain_recover(hdl,
1482 newname ? origname : thename, -errno, nvi);
1489 zpool_handle_t *zhp;
1492 * This should never fail, but play it safe anyway.
1494 if (zpool_open_silent(hdl, thename, &zhp) != 0)
1496 else if (zhp != NULL)
1498 (void) zcmd_read_dst_nvlist(hdl, &zc, &nvi);
1499 zpool_get_rewind_policy(config, &policy);
1500 if (policy.zrp_request &
1501 (ZPOOL_DO_REWIND | ZPOOL_TRY_REWIND)) {
1502 zpool_rewind_exclaim(hdl, newname ? origname : thename,
1503 ((policy.zrp_request & ZPOOL_TRY_REWIND) != 0),
1510 zcmd_free_nvlists(&zc);
1520 zpool_scan(zpool_handle_t *zhp, pool_scan_func_t func)
1522 zfs_cmd_t zc = { 0 };
1524 libzfs_handle_t *hdl = zhp->zpool_hdl;
1526 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
1527 zc.zc_cookie = func;
1529 if (zfs_ioctl(zhp->zpool_hdl, ZFS_IOC_POOL_SCAN, &zc) == 0 ||
1530 (errno == ENOENT && func != POOL_SCAN_NONE))
1533 if (func == POOL_SCAN_SCRUB) {
1534 (void) snprintf(msg, sizeof (msg),
1535 dgettext(TEXT_DOMAIN, "cannot scrub %s"), zc.zc_name);
1536 } else if (func == POOL_SCAN_NONE) {
1537 (void) snprintf(msg, sizeof (msg),
1538 dgettext(TEXT_DOMAIN, "cannot cancel scrubbing %s"),
1541 assert(!"unexpected result");
1544 if (errno == EBUSY) {
1546 pool_scan_stat_t *ps = NULL;
1549 verify(nvlist_lookup_nvlist(zhp->zpool_config,
1550 ZPOOL_CONFIG_VDEV_TREE, &nvroot) == 0);
1551 (void) nvlist_lookup_uint64_array(nvroot,
1552 ZPOOL_CONFIG_SCAN_STATS, (uint64_t **)&ps, &psc);
1553 if (ps && ps->pss_func == POOL_SCAN_SCRUB)
1554 return (zfs_error(hdl, EZFS_SCRUBBING, msg));
1556 return (zfs_error(hdl, EZFS_RESILVERING, msg));
1557 } else if (errno == ENOENT) {
1558 return (zfs_error(hdl, EZFS_NO_SCRUB, msg));
1560 return (zpool_standard_error(hdl, errno, msg));
1565 * This provides a very minimal check whether a given string is likely a
1566 * c#t#d# style string. Users of this are expected to do their own
1567 * verification of the s# part.
1569 #define CTD_CHECK(str) (str && str[0] == 'c' && isdigit(str[1]))
1572 * More elaborate version for ones which may start with "/dev/dsk/"
1576 ctd_check_path(char *str) {
1578 * If it starts with a slash, check the last component.
1580 if (str && str[0] == '/') {
1581 char *tmp = strrchr(str, '/');
1584 * If it ends in "/old", check the second-to-last
1585 * component of the string instead.
1587 if (tmp != str && strcmp(tmp, "/old") == 0) {
1588 for (tmp--; *tmp != '/'; tmp--)
1593 return (CTD_CHECK(str));
1597 * Find a vdev that matches the search criteria specified. We use the
1598 * the nvpair name to determine how we should look for the device.
1599 * 'avail_spare' is set to TRUE if the provided guid refers to an AVAIL
1600 * spare; but FALSE if its an INUSE spare.
1603 vdev_to_nvlist_iter(nvlist_t *nv, nvlist_t *search, boolean_t *avail_spare,
1604 boolean_t *l2cache, boolean_t *log)
1611 nvpair_t *pair = nvlist_next_nvpair(search, NULL);
1613 /* Nothing to look for */
1614 if (search == NULL || pair == NULL)
1617 /* Obtain the key we will use to search */
1618 srchkey = nvpair_name(pair);
1620 switch (nvpair_type(pair)) {
1621 case DATA_TYPE_UINT64: {
1622 uint64_t srchval, theguid, present;
1624 verify(nvpair_value_uint64(pair, &srchval) == 0);
1625 if (strcmp(srchkey, ZPOOL_CONFIG_GUID) == 0) {
1626 if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_NOT_PRESENT,
1629 * If the device has never been present since
1630 * import, the only reliable way to match the
1633 verify(nvlist_lookup_uint64(nv,
1634 ZPOOL_CONFIG_GUID, &theguid) == 0);
1635 if (theguid == srchval)
1642 case DATA_TYPE_STRING: {
1643 char *srchval, *val;
1645 verify(nvpair_value_string(pair, &srchval) == 0);
1646 if (nvlist_lookup_string(nv, srchkey, &val) != 0)
1650 * Search for the requested value. Special cases:
1652 * - ZPOOL_CONFIG_PATH for whole disk entries. These end in
1653 * "s0" or "s0/old". The "s0" part is hidden from the user,
1654 * but included in the string, so this matches around it.
1655 * - looking for a top-level vdev name (i.e. ZPOOL_CONFIG_TYPE).
1657 * Otherwise, all other searches are simple string compares.
1659 if (strcmp(srchkey, ZPOOL_CONFIG_PATH) == 0 &&
1660 ctd_check_path(val)) {
1661 uint64_t wholedisk = 0;
1663 (void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_WHOLE_DISK,
1666 int slen = strlen(srchval);
1667 int vlen = strlen(val);
1669 if (slen != vlen - 2)
1673 * make_leaf_vdev() should only set
1674 * wholedisk for ZPOOL_CONFIG_PATHs which
1675 * will include "/dev/dsk/", giving plenty of
1676 * room for the indices used next.
1681 * strings identical except trailing "s0"
1683 if (strcmp(&val[vlen - 2], "s0") == 0 &&
1684 strncmp(srchval, val, slen) == 0)
1688 * strings identical except trailing "s0/old"
1690 if (strcmp(&val[vlen - 6], "s0/old") == 0 &&
1691 strcmp(&srchval[slen - 4], "/old") == 0 &&
1692 strncmp(srchval, val, slen - 4) == 0)
1697 } else if (strcmp(srchkey, ZPOOL_CONFIG_TYPE) == 0 && val) {
1698 char *type, *idx, *end, *p;
1699 uint64_t id, vdev_id;
1702 * Determine our vdev type, keeping in mind
1703 * that the srchval is composed of a type and
1704 * vdev id pair (i.e. mirror-4).
1706 if ((type = strdup(srchval)) == NULL)
1709 if ((p = strrchr(type, '-')) == NULL) {
1717 * If the types don't match then keep looking.
1719 if (strncmp(val, type, strlen(val)) != 0) {
1724 verify(strncmp(type, VDEV_TYPE_RAIDZ,
1725 strlen(VDEV_TYPE_RAIDZ)) == 0 ||
1726 strncmp(type, VDEV_TYPE_MIRROR,
1727 strlen(VDEV_TYPE_MIRROR)) == 0);
1728 verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_ID,
1732 vdev_id = strtoull(idx, &end, 10);
1739 * Now verify that we have the correct vdev id.
1748 if (strcmp(srchval, val) == 0)
1757 if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN,
1758 &child, &children) != 0)
1761 for (c = 0; c < children; c++) {
1762 if ((ret = vdev_to_nvlist_iter(child[c], search,
1763 avail_spare, l2cache, NULL)) != NULL) {
1765 * The 'is_log' value is only set for the toplevel
1766 * vdev, not the leaf vdevs. So we always lookup the
1767 * log device from the root of the vdev tree (where
1768 * 'log' is non-NULL).
1771 nvlist_lookup_uint64(child[c],
1772 ZPOOL_CONFIG_IS_LOG, &is_log) == 0 &&
1780 if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_SPARES,
1781 &child, &children) == 0) {
1782 for (c = 0; c < children; c++) {
1783 if ((ret = vdev_to_nvlist_iter(child[c], search,
1784 avail_spare, l2cache, NULL)) != NULL) {
1785 *avail_spare = B_TRUE;
1791 if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_L2CACHE,
1792 &child, &children) == 0) {
1793 for (c = 0; c < children; c++) {
1794 if ((ret = vdev_to_nvlist_iter(child[c], search,
1795 avail_spare, l2cache, NULL)) != NULL) {
1806 * Given a physical path (minus the "/devices" prefix), find the
1810 zpool_find_vdev_by_physpath(zpool_handle_t *zhp, const char *ppath,
1811 boolean_t *avail_spare, boolean_t *l2cache, boolean_t *log)
1813 nvlist_t *search, *nvroot, *ret;
1815 verify(nvlist_alloc(&search, NV_UNIQUE_NAME, KM_SLEEP) == 0);
1816 verify(nvlist_add_string(search, ZPOOL_CONFIG_PHYS_PATH, ppath) == 0);
1818 verify(nvlist_lookup_nvlist(zhp->zpool_config, ZPOOL_CONFIG_VDEV_TREE,
1821 *avail_spare = B_FALSE;
1822 ret = vdev_to_nvlist_iter(nvroot, search, avail_spare, l2cache, log);
1823 nvlist_free(search);
1829 * Determine if we have an "interior" top-level vdev (i.e mirror/raidz).
1832 zpool_vdev_is_interior(const char *name)
1834 if (strncmp(name, VDEV_TYPE_RAIDZ, strlen(VDEV_TYPE_RAIDZ)) == 0 ||
1835 strncmp(name, VDEV_TYPE_MIRROR, strlen(VDEV_TYPE_MIRROR)) == 0)
1841 zpool_find_vdev(zpool_handle_t *zhp, const char *path, boolean_t *avail_spare,
1842 boolean_t *l2cache, boolean_t *log)
1844 char buf[MAXPATHLEN];
1846 nvlist_t *nvroot, *search, *ret;
1849 verify(nvlist_alloc(&search, NV_UNIQUE_NAME, KM_SLEEP) == 0);
1851 guid = strtoull(path, &end, 10);
1852 if (guid != 0 && *end == '\0') {
1853 verify(nvlist_add_uint64(search, ZPOOL_CONFIG_GUID, guid) == 0);
1854 } else if (zpool_vdev_is_interior(path)) {
1855 verify(nvlist_add_string(search, ZPOOL_CONFIG_TYPE, path) == 0);
1856 } else if (path[0] != '/') {
1857 (void) snprintf(buf, sizeof (buf), "%s%s", "/dev/dsk/", path);
1858 verify(nvlist_add_string(search, ZPOOL_CONFIG_PATH, buf) == 0);
1860 verify(nvlist_add_string(search, ZPOOL_CONFIG_PATH, path) == 0);
1863 verify(nvlist_lookup_nvlist(zhp->zpool_config, ZPOOL_CONFIG_VDEV_TREE,
1866 *avail_spare = B_FALSE;
1870 ret = vdev_to_nvlist_iter(nvroot, search, avail_spare, l2cache, log);
1871 nvlist_free(search);
1877 vdev_online(nvlist_t *nv)
1881 if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_OFFLINE, &ival) == 0 ||
1882 nvlist_lookup_uint64(nv, ZPOOL_CONFIG_FAULTED, &ival) == 0 ||
1883 nvlist_lookup_uint64(nv, ZPOOL_CONFIG_REMOVED, &ival) == 0)
1890 * Helper function for zpool_get_physpaths().
1893 vdev_get_one_physpath(nvlist_t *config, char *physpath, size_t physpath_size,
1894 size_t *bytes_written)
1896 size_t bytes_left, pos, rsz;
1900 if (nvlist_lookup_string(config, ZPOOL_CONFIG_PHYS_PATH,
1902 return (EZFS_NODEVICE);
1904 pos = *bytes_written;
1905 bytes_left = physpath_size - pos;
1906 format = (pos == 0) ? "%s" : " %s";
1908 rsz = snprintf(physpath + pos, bytes_left, format, tmppath);
1909 *bytes_written += rsz;
1911 if (rsz >= bytes_left) {
1912 /* if physpath was not copied properly, clear it */
1913 if (bytes_left != 0) {
1916 return (EZFS_NOSPC);
1922 vdev_get_physpaths(nvlist_t *nv, char *physpath, size_t phypath_size,
1923 size_t *rsz, boolean_t is_spare)
1928 if (nvlist_lookup_string(nv, ZPOOL_CONFIG_TYPE, &type) != 0)
1929 return (EZFS_INVALCONFIG);
1931 if (strcmp(type, VDEV_TYPE_DISK) == 0) {
1933 * An active spare device has ZPOOL_CONFIG_IS_SPARE set.
1934 * For a spare vdev, we only want to boot from the active
1939 (void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_IS_SPARE,
1942 return (EZFS_INVALCONFIG);
1945 if (vdev_online(nv)) {
1946 if ((ret = vdev_get_one_physpath(nv, physpath,
1947 phypath_size, rsz)) != 0)
1950 } else if (strcmp(type, VDEV_TYPE_MIRROR) == 0 ||
1951 strcmp(type, VDEV_TYPE_REPLACING) == 0 ||
1952 (is_spare = (strcmp(type, VDEV_TYPE_SPARE) == 0))) {
1957 if (nvlist_lookup_nvlist_array(nv,
1958 ZPOOL_CONFIG_CHILDREN, &child, &count) != 0)
1959 return (EZFS_INVALCONFIG);
1961 for (i = 0; i < count; i++) {
1962 ret = vdev_get_physpaths(child[i], physpath,
1963 phypath_size, rsz, is_spare);
1964 if (ret == EZFS_NOSPC)
1969 return (EZFS_POOL_INVALARG);
1973 * Get phys_path for a root pool config.
1974 * Return 0 on success; non-zero on failure.
1977 zpool_get_config_physpath(nvlist_t *config, char *physpath, size_t phypath_size)
1980 nvlist_t *vdev_root;
1987 if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE,
1989 return (EZFS_INVALCONFIG);
1991 if (nvlist_lookup_string(vdev_root, ZPOOL_CONFIG_TYPE, &type) != 0 ||
1992 nvlist_lookup_nvlist_array(vdev_root, ZPOOL_CONFIG_CHILDREN,
1993 &child, &count) != 0)
1994 return (EZFS_INVALCONFIG);
1997 * root pool can not have EFI labeled disks and can only have
1998 * a single top-level vdev.
2000 if (strcmp(type, VDEV_TYPE_ROOT) != 0 || count != 1 ||
2001 pool_uses_efi(vdev_root))
2002 return (EZFS_POOL_INVALARG);
2004 (void) vdev_get_physpaths(child[0], physpath, phypath_size, &rsz,
2007 /* No online devices */
2009 return (EZFS_NODEVICE);
2015 * Get phys_path for a root pool
2016 * Return 0 on success; non-zero on failure.
2019 zpool_get_physpath(zpool_handle_t *zhp, char *physpath, size_t phypath_size)
2021 return (zpool_get_config_physpath(zhp->zpool_config, physpath,
2026 * If the device has being dynamically expanded then we need to relabel
2027 * the disk to use the new unallocated space.
2030 zpool_relabel_disk(libzfs_handle_t *hdl, const char *name)
2032 char path[MAXPATHLEN];
2035 int (*_efi_use_whole_disk)(int);
2037 if ((_efi_use_whole_disk = (int (*)(int))dlsym(RTLD_DEFAULT,
2038 "efi_use_whole_disk")) == NULL)
2041 (void) snprintf(path, sizeof (path), "%s/%s", RDISK_ROOT, name);
2043 if ((fd = open(path, O_RDWR | O_NDELAY)) < 0) {
2044 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "cannot "
2045 "relabel '%s': unable to open device"), name);
2046 return (zfs_error(hdl, EZFS_OPENFAILED, errbuf));
2050 * It's possible that we might encounter an error if the device
2051 * does not have any unallocated space left. If so, we simply
2052 * ignore that error and continue on.
2054 error = _efi_use_whole_disk(fd);
2056 if (error && error != VT_ENOSPC) {
2057 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "cannot "
2058 "relabel '%s': unable to read disk capacity"), name);
2059 return (zfs_error(hdl, EZFS_NOCAP, errbuf));
2065 * Bring the specified vdev online. The 'flags' parameter is a set of the
2066 * ZFS_ONLINE_* flags.
2069 zpool_vdev_online(zpool_handle_t *zhp, const char *path, int flags,
2070 vdev_state_t *newstate)
2072 zfs_cmd_t zc = { 0 };
2075 boolean_t avail_spare, l2cache, islog;
2076 libzfs_handle_t *hdl = zhp->zpool_hdl;
2078 if (flags & ZFS_ONLINE_EXPAND) {
2079 (void) snprintf(msg, sizeof (msg),
2080 dgettext(TEXT_DOMAIN, "cannot expand %s"), path);
2082 (void) snprintf(msg, sizeof (msg),
2083 dgettext(TEXT_DOMAIN, "cannot online %s"), path);
2086 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
2087 if ((tgt = zpool_find_vdev(zhp, path, &avail_spare, &l2cache,
2089 return (zfs_error(hdl, EZFS_NODEVICE, msg));
2091 verify(nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID, &zc.zc_guid) == 0);
2094 return (zfs_error(hdl, EZFS_ISSPARE, msg));
2096 if (flags & ZFS_ONLINE_EXPAND ||
2097 zpool_get_prop_int(zhp, ZPOOL_PROP_AUTOEXPAND, NULL)) {
2098 char *pathname = NULL;
2099 uint64_t wholedisk = 0;
2101 (void) nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_WHOLE_DISK,
2103 verify(nvlist_lookup_string(tgt, ZPOOL_CONFIG_PATH,
2107 * XXX - L2ARC 1.0 devices can't support expansion.
2110 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2111 "cannot expand cache devices"));
2112 return (zfs_error(hdl, EZFS_VDEVNOTSUP, msg));
2116 pathname += strlen(DISK_ROOT) + 1;
2117 (void) zpool_relabel_disk(zhp->zpool_hdl, pathname);
2121 zc.zc_cookie = VDEV_STATE_ONLINE;
2124 if (zfs_ioctl(zhp->zpool_hdl, ZFS_IOC_VDEV_SET_STATE, &zc) != 0) {
2125 if (errno == EINVAL) {
2126 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "was split "
2127 "from this pool into a new one. Use '%s' "
2128 "instead"), "zpool detach");
2129 return (zfs_error(hdl, EZFS_POSTSPLIT_ONLINE, msg));
2131 return (zpool_standard_error(hdl, errno, msg));
2134 *newstate = zc.zc_cookie;
2139 * Take the specified vdev offline
2142 zpool_vdev_offline(zpool_handle_t *zhp, const char *path, boolean_t istmp)
2144 zfs_cmd_t zc = { 0 };
2147 boolean_t avail_spare, l2cache;
2148 libzfs_handle_t *hdl = zhp->zpool_hdl;
2150 (void) snprintf(msg, sizeof (msg),
2151 dgettext(TEXT_DOMAIN, "cannot offline %s"), path);
2153 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
2154 if ((tgt = zpool_find_vdev(zhp, path, &avail_spare, &l2cache,
2156 return (zfs_error(hdl, EZFS_NODEVICE, msg));
2158 verify(nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID, &zc.zc_guid) == 0);
2161 return (zfs_error(hdl, EZFS_ISSPARE, msg));
2163 zc.zc_cookie = VDEV_STATE_OFFLINE;
2164 zc.zc_obj = istmp ? ZFS_OFFLINE_TEMPORARY : 0;
2166 if (zfs_ioctl(zhp->zpool_hdl, ZFS_IOC_VDEV_SET_STATE, &zc) == 0)
2173 * There are no other replicas of this device.
2175 return (zfs_error(hdl, EZFS_NOREPLICAS, msg));
2179 * The log device has unplayed logs
2181 return (zfs_error(hdl, EZFS_UNPLAYED_LOGS, msg));
2184 return (zpool_standard_error(hdl, errno, msg));
2189 * Mark the given vdev faulted.
2192 zpool_vdev_fault(zpool_handle_t *zhp, uint64_t guid, vdev_aux_t aux)
2194 zfs_cmd_t zc = { 0 };
2196 libzfs_handle_t *hdl = zhp->zpool_hdl;
2198 (void) snprintf(msg, sizeof (msg),
2199 dgettext(TEXT_DOMAIN, "cannot fault %llu"), guid);
2201 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
2203 zc.zc_cookie = VDEV_STATE_FAULTED;
2206 if (ioctl(zhp->zpool_hdl->libzfs_fd, ZFS_IOC_VDEV_SET_STATE, &zc) == 0)
2213 * There are no other replicas of this device.
2215 return (zfs_error(hdl, EZFS_NOREPLICAS, msg));
2218 return (zpool_standard_error(hdl, errno, msg));
2224 * Mark the given vdev degraded.
2227 zpool_vdev_degrade(zpool_handle_t *zhp, uint64_t guid, vdev_aux_t aux)
2229 zfs_cmd_t zc = { 0 };
2231 libzfs_handle_t *hdl = zhp->zpool_hdl;
2233 (void) snprintf(msg, sizeof (msg),
2234 dgettext(TEXT_DOMAIN, "cannot degrade %llu"), guid);
2236 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
2238 zc.zc_cookie = VDEV_STATE_DEGRADED;
2241 if (ioctl(zhp->zpool_hdl->libzfs_fd, ZFS_IOC_VDEV_SET_STATE, &zc) == 0)
2244 return (zpool_standard_error(hdl, errno, msg));
2248 * Returns TRUE if the given nvlist is a vdev that was originally swapped in as
2252 is_replacing_spare(nvlist_t *search, nvlist_t *tgt, int which)
2258 if (nvlist_lookup_nvlist_array(search, ZPOOL_CONFIG_CHILDREN, &child,
2260 verify(nvlist_lookup_string(search, ZPOOL_CONFIG_TYPE,
2263 if (strcmp(type, VDEV_TYPE_SPARE) == 0 &&
2264 children == 2 && child[which] == tgt)
2267 for (c = 0; c < children; c++)
2268 if (is_replacing_spare(child[c], tgt, which))
2276 * Attach new_disk (fully described by nvroot) to old_disk.
2277 * If 'replacing' is specified, the new disk will replace the old one.
2280 zpool_vdev_attach(zpool_handle_t *zhp,
2281 const char *old_disk, const char *new_disk, nvlist_t *nvroot, int replacing)
2283 zfs_cmd_t zc = { 0 };
2287 boolean_t avail_spare, l2cache, islog;
2289 char *path, *newname;
2292 nvlist_t *config_root;
2293 libzfs_handle_t *hdl = zhp->zpool_hdl;
2294 boolean_t rootpool = pool_is_bootable(zhp);
2297 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN,
2298 "cannot replace %s with %s"), old_disk, new_disk);
2300 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN,
2301 "cannot attach %s to %s"), new_disk, old_disk);
2304 * If this is a root pool, make sure that we're not attaching an
2305 * EFI labeled device.
2307 if (rootpool && pool_uses_efi(nvroot)) {
2308 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2309 "EFI labeled devices are not supported on root pools."));
2310 return (zfs_error(hdl, EZFS_POOL_NOTSUP, msg));
2313 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
2314 if ((tgt = zpool_find_vdev(zhp, old_disk, &avail_spare, &l2cache,
2316 return (zfs_error(hdl, EZFS_NODEVICE, msg));
2319 return (zfs_error(hdl, EZFS_ISSPARE, msg));
2322 return (zfs_error(hdl, EZFS_ISL2CACHE, msg));
2324 verify(nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID, &zc.zc_guid) == 0);
2325 zc.zc_cookie = replacing;
2327 if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_CHILDREN,
2328 &child, &children) != 0 || children != 1) {
2329 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2330 "new device must be a single disk"));
2331 return (zfs_error(hdl, EZFS_INVALCONFIG, msg));
2334 verify(nvlist_lookup_nvlist(zpool_get_config(zhp, NULL),
2335 ZPOOL_CONFIG_VDEV_TREE, &config_root) == 0);
2337 if ((newname = zpool_vdev_name(NULL, NULL, child[0], B_FALSE)) == NULL)
2341 * If the target is a hot spare that has been swapped in, we can only
2342 * replace it with another hot spare.
2345 nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_IS_SPARE, &val) == 0 &&
2346 (zpool_find_vdev(zhp, newname, &avail_spare, &l2cache,
2347 NULL) == NULL || !avail_spare) &&
2348 is_replacing_spare(config_root, tgt, 1)) {
2349 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2350 "can only be replaced by another hot spare"));
2352 return (zfs_error(hdl, EZFS_BADTARGET, msg));
2356 * If we are attempting to replace a spare, it canot be applied to an
2357 * already spared device.
2360 nvlist_lookup_string(child[0], ZPOOL_CONFIG_PATH, &path) == 0 &&
2361 zpool_find_vdev(zhp, newname, &avail_spare,
2362 &l2cache, NULL) != NULL && avail_spare &&
2363 is_replacing_spare(config_root, tgt, 0)) {
2364 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2365 "device has already been replaced with a spare"));
2367 return (zfs_error(hdl, EZFS_BADTARGET, msg));
2372 if (zcmd_write_conf_nvlist(hdl, &zc, nvroot) != 0)
2375 ret = zfs_ioctl(zhp->zpool_hdl, ZFS_IOC_VDEV_ATTACH, &zc);
2377 zcmd_free_nvlists(&zc);
2382 * XXX - This should be removed once we can
2383 * automatically install the bootblocks on the
2384 * newly attached disk.
2386 (void) fprintf(stderr, dgettext(TEXT_DOMAIN, "Please "
2387 "be sure to invoke %s to make '%s' bootable.\n"),
2391 * XXX need a better way to prevent user from
2392 * booting up a half-baked vdev.
2394 (void) fprintf(stderr, dgettext(TEXT_DOMAIN, "Make "
2395 "sure to wait until resilver is done "
2396 "before rebooting.\n"));
2404 * Can't attach to or replace this type of vdev.
2408 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2409 "cannot replace a log with a spare"));
2411 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2412 "cannot replace a replacing device"));
2414 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2415 "can only attach to mirrors and top-level "
2418 (void) zfs_error(hdl, EZFS_BADTARGET, msg);
2423 * The new device must be a single disk.
2425 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2426 "new device must be a single disk"));
2427 (void) zfs_error(hdl, EZFS_INVALCONFIG, msg);
2431 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "%s is busy"),
2433 (void) zfs_error(hdl, EZFS_BADDEV, msg);
2438 * The new device is too small.
2440 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2441 "device is too small"));
2442 (void) zfs_error(hdl, EZFS_BADDEV, msg);
2447 * The new device has a different alignment requirement.
2449 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2450 "devices have different sector alignment"));
2451 (void) zfs_error(hdl, EZFS_BADDEV, msg);
2456 * The resulting top-level vdev spec won't fit in the label.
2458 (void) zfs_error(hdl, EZFS_DEVOVERFLOW, msg);
2462 (void) zpool_standard_error(hdl, errno, msg);
2469 * Detach the specified device.
2472 zpool_vdev_detach(zpool_handle_t *zhp, const char *path)
2474 zfs_cmd_t zc = { 0 };
2477 boolean_t avail_spare, l2cache;
2478 libzfs_handle_t *hdl = zhp->zpool_hdl;
2480 (void) snprintf(msg, sizeof (msg),
2481 dgettext(TEXT_DOMAIN, "cannot detach %s"), path);
2483 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
2484 if ((tgt = zpool_find_vdev(zhp, path, &avail_spare, &l2cache,
2486 return (zfs_error(hdl, EZFS_NODEVICE, msg));
2489 return (zfs_error(hdl, EZFS_ISSPARE, msg));
2492 return (zfs_error(hdl, EZFS_ISL2CACHE, msg));
2494 verify(nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID, &zc.zc_guid) == 0);
2496 if (zfs_ioctl(hdl, ZFS_IOC_VDEV_DETACH, &zc) == 0)
2503 * Can't detach from this type of vdev.
2505 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "only "
2506 "applicable to mirror and replacing vdevs"));
2507 (void) zfs_error(zhp->zpool_hdl, EZFS_BADTARGET, msg);
2512 * There are no other replicas of this device.
2514 (void) zfs_error(hdl, EZFS_NOREPLICAS, msg);
2518 (void) zpool_standard_error(hdl, errno, msg);
2525 * Find a mirror vdev in the source nvlist.
2527 * The mchild array contains a list of disks in one of the top-level mirrors
2528 * of the source pool. The schild array contains a list of disks that the
2529 * user specified on the command line. We loop over the mchild array to
2530 * see if any entry in the schild array matches.
2532 * If a disk in the mchild array is found in the schild array, we return
2533 * the index of that entry. Otherwise we return -1.
2536 find_vdev_entry(zpool_handle_t *zhp, nvlist_t **mchild, uint_t mchildren,
2537 nvlist_t **schild, uint_t schildren)
2541 for (mc = 0; mc < mchildren; mc++) {
2543 char *mpath = zpool_vdev_name(zhp->zpool_hdl, zhp,
2544 mchild[mc], B_FALSE);
2546 for (sc = 0; sc < schildren; sc++) {
2547 char *spath = zpool_vdev_name(zhp->zpool_hdl, zhp,
2548 schild[sc], B_FALSE);
2549 boolean_t result = (strcmp(mpath, spath) == 0);
2565 * Split a mirror pool. If newroot points to null, then a new nvlist
2566 * is generated and it is the responsibility of the caller to free it.
2569 zpool_vdev_split(zpool_handle_t *zhp, char *newname, nvlist_t **newroot,
2570 nvlist_t *props, splitflags_t flags)
2572 zfs_cmd_t zc = { 0 };
2574 nvlist_t *tree, *config, **child, **newchild, *newconfig = NULL;
2575 nvlist_t **varray = NULL, *zc_props = NULL;
2576 uint_t c, children, newchildren, lastlog = 0, vcount, found = 0;
2577 libzfs_handle_t *hdl = zhp->zpool_hdl;
2579 boolean_t freelist = B_FALSE, memory_err = B_TRUE;
2582 (void) snprintf(msg, sizeof (msg),
2583 dgettext(TEXT_DOMAIN, "Unable to split %s"), zhp->zpool_name);
2585 if (!zpool_name_valid(hdl, B_FALSE, newname))
2586 return (zfs_error(hdl, EZFS_INVALIDNAME, msg));
2588 if ((config = zpool_get_config(zhp, NULL)) == NULL) {
2589 (void) fprintf(stderr, gettext("Internal error: unable to "
2590 "retrieve pool configuration\n"));
2594 verify(nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE, &tree)
2596 verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_VERSION, &vers) == 0);
2599 if ((zc_props = zpool_valid_proplist(hdl, zhp->zpool_name,
2600 props, vers, B_TRUE, msg)) == NULL)
2604 if (nvlist_lookup_nvlist_array(tree, ZPOOL_CONFIG_CHILDREN, &child,
2606 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2607 "Source pool is missing vdev tree"));
2609 nvlist_free(zc_props);
2613 varray = zfs_alloc(hdl, children * sizeof (nvlist_t *));
2616 if (*newroot == NULL ||
2617 nvlist_lookup_nvlist_array(*newroot, ZPOOL_CONFIG_CHILDREN,
2618 &newchild, &newchildren) != 0)
2621 for (c = 0; c < children; c++) {
2622 uint64_t is_log = B_FALSE, is_hole = B_FALSE;
2624 nvlist_t **mchild, *vdev;
2629 * Unlike cache & spares, slogs are stored in the
2630 * ZPOOL_CONFIG_CHILDREN array. We filter them out here.
2632 (void) nvlist_lookup_uint64(child[c], ZPOOL_CONFIG_IS_LOG,
2634 (void) nvlist_lookup_uint64(child[c], ZPOOL_CONFIG_IS_HOLE,
2636 if (is_log || is_hole) {
2638 * Create a hole vdev and put it in the config.
2640 if (nvlist_alloc(&vdev, NV_UNIQUE_NAME, 0) != 0)
2642 if (nvlist_add_string(vdev, ZPOOL_CONFIG_TYPE,
2643 VDEV_TYPE_HOLE) != 0)
2645 if (nvlist_add_uint64(vdev, ZPOOL_CONFIG_IS_HOLE,
2650 varray[vcount++] = vdev;
2654 verify(nvlist_lookup_string(child[c], ZPOOL_CONFIG_TYPE, &type)
2656 if (strcmp(type, VDEV_TYPE_MIRROR) != 0) {
2657 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2658 "Source pool must be composed only of mirrors\n"));
2659 retval = zfs_error(hdl, EZFS_INVALCONFIG, msg);
2663 verify(nvlist_lookup_nvlist_array(child[c],
2664 ZPOOL_CONFIG_CHILDREN, &mchild, &mchildren) == 0);
2666 /* find or add an entry for this top-level vdev */
2667 if (newchildren > 0 &&
2668 (entry = find_vdev_entry(zhp, mchild, mchildren,
2669 newchild, newchildren)) >= 0) {
2670 /* We found a disk that the user specified. */
2671 vdev = mchild[entry];
2674 /* User didn't specify a disk for this vdev. */
2675 vdev = mchild[mchildren - 1];
2678 if (nvlist_dup(vdev, &varray[vcount++], 0) != 0)
2682 /* did we find every disk the user specified? */
2683 if (found != newchildren) {
2684 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "Device list must "
2685 "include at most one disk from each mirror"));
2686 retval = zfs_error(hdl, EZFS_INVALCONFIG, msg);
2690 /* Prepare the nvlist for populating. */
2691 if (*newroot == NULL) {
2692 if (nvlist_alloc(newroot, NV_UNIQUE_NAME, 0) != 0)
2695 if (nvlist_add_string(*newroot, ZPOOL_CONFIG_TYPE,
2696 VDEV_TYPE_ROOT) != 0)
2699 verify(nvlist_remove_all(*newroot, ZPOOL_CONFIG_CHILDREN) == 0);
2702 /* Add all the children we found */
2703 if (nvlist_add_nvlist_array(*newroot, ZPOOL_CONFIG_CHILDREN, varray,
2704 lastlog == 0 ? vcount : lastlog) != 0)
2708 * If we're just doing a dry run, exit now with success.
2711 memory_err = B_FALSE;
2716 /* now build up the config list & call the ioctl */
2717 if (nvlist_alloc(&newconfig, NV_UNIQUE_NAME, 0) != 0)
2720 if (nvlist_add_nvlist(newconfig,
2721 ZPOOL_CONFIG_VDEV_TREE, *newroot) != 0 ||
2722 nvlist_add_string(newconfig,
2723 ZPOOL_CONFIG_POOL_NAME, newname) != 0 ||
2724 nvlist_add_uint64(newconfig, ZPOOL_CONFIG_VERSION, vers) != 0)
2728 * The new pool is automatically part of the namespace unless we
2729 * explicitly export it.
2732 zc.zc_cookie = ZPOOL_EXPORT_AFTER_SPLIT;
2733 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
2734 (void) strlcpy(zc.zc_string, newname, sizeof (zc.zc_string));
2735 if (zcmd_write_conf_nvlist(hdl, &zc, newconfig) != 0)
2737 if (zc_props != NULL && zcmd_write_src_nvlist(hdl, &zc, zc_props) != 0)
2740 if (zfs_ioctl(hdl, ZFS_IOC_VDEV_SPLIT, &zc) != 0) {
2741 retval = zpool_standard_error(hdl, errno, msg);
2746 memory_err = B_FALSE;
2749 if (varray != NULL) {
2752 for (v = 0; v < vcount; v++)
2753 nvlist_free(varray[v]);
2756 zcmd_free_nvlists(&zc);
2758 nvlist_free(zc_props);
2760 nvlist_free(newconfig);
2762 nvlist_free(*newroot);
2770 return (no_memory(hdl));
2776 * Remove the given device. Currently, this is supported only for hot spares
2777 * and level 2 cache devices.
2780 zpool_vdev_remove(zpool_handle_t *zhp, const char *path)
2782 zfs_cmd_t zc = { 0 };
2785 boolean_t avail_spare, l2cache, islog;
2786 libzfs_handle_t *hdl = zhp->zpool_hdl;
2789 (void) snprintf(msg, sizeof (msg),
2790 dgettext(TEXT_DOMAIN, "cannot remove %s"), path);
2792 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
2793 if ((tgt = zpool_find_vdev(zhp, path, &avail_spare, &l2cache,
2795 return (zfs_error(hdl, EZFS_NODEVICE, msg));
2797 * XXX - this should just go away.
2799 if (!avail_spare && !l2cache && !islog) {
2800 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2801 "only inactive hot spares, cache, top-level, "
2802 "or log devices can be removed"));
2803 return (zfs_error(hdl, EZFS_NODEVICE, msg));
2806 version = zpool_get_prop_int(zhp, ZPOOL_PROP_VERSION, NULL);
2807 if (islog && version < SPA_VERSION_HOLES) {
2808 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2809 "pool must be upgrade to support log removal"));
2810 return (zfs_error(hdl, EZFS_BADVERSION, msg));
2813 verify(nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID, &zc.zc_guid) == 0);
2815 if (zfs_ioctl(hdl, ZFS_IOC_VDEV_REMOVE, &zc) == 0)
2818 return (zpool_standard_error(hdl, errno, msg));
2822 * Clear the errors for the pool, or the particular device if specified.
2825 zpool_clear(zpool_handle_t *zhp, const char *path, nvlist_t *rewindnvl)
2827 zfs_cmd_t zc = { 0 };
2830 zpool_rewind_policy_t policy;
2831 boolean_t avail_spare, l2cache;
2832 libzfs_handle_t *hdl = zhp->zpool_hdl;
2833 nvlist_t *nvi = NULL;
2836 (void) snprintf(msg, sizeof (msg),
2837 dgettext(TEXT_DOMAIN, "cannot clear errors for %s"),
2840 (void) snprintf(msg, sizeof (msg),
2841 dgettext(TEXT_DOMAIN, "cannot clear errors for %s"),
2844 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
2846 if ((tgt = zpool_find_vdev(zhp, path, &avail_spare,
2847 &l2cache, NULL)) == 0)
2848 return (zfs_error(hdl, EZFS_NODEVICE, msg));
2851 * Don't allow error clearing for hot spares. Do allow
2852 * error clearing for l2cache devices.
2855 return (zfs_error(hdl, EZFS_ISSPARE, msg));
2857 verify(nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID,
2861 zpool_get_rewind_policy(rewindnvl, &policy);
2862 zc.zc_cookie = policy.zrp_request;
2864 if (zcmd_alloc_dst_nvlist(hdl, &zc, 8192) != 0)
2867 if (zcmd_write_src_nvlist(zhp->zpool_hdl, &zc, rewindnvl) != 0)
2870 if (zfs_ioctl(hdl, ZFS_IOC_CLEAR, &zc) == 0 ||
2871 ((policy.zrp_request & ZPOOL_TRY_REWIND) &&
2872 errno != EPERM && errno != EACCES)) {
2873 if (policy.zrp_request &
2874 (ZPOOL_DO_REWIND | ZPOOL_TRY_REWIND)) {
2875 (void) zcmd_read_dst_nvlist(hdl, &zc, &nvi);
2876 zpool_rewind_exclaim(hdl, zc.zc_name,
2877 ((policy.zrp_request & ZPOOL_TRY_REWIND) != 0),
2881 zcmd_free_nvlists(&zc);
2885 zcmd_free_nvlists(&zc);
2886 return (zpool_standard_error(hdl, errno, msg));
2890 * Similar to zpool_clear(), but takes a GUID (used by fmd).
2893 zpool_vdev_clear(zpool_handle_t *zhp, uint64_t guid)
2895 zfs_cmd_t zc = { 0 };
2897 libzfs_handle_t *hdl = zhp->zpool_hdl;
2899 (void) snprintf(msg, sizeof (msg),
2900 dgettext(TEXT_DOMAIN, "cannot clear errors for %llx"),
2903 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
2905 zc.zc_cookie = ZPOOL_NO_REWIND;
2907 if (ioctl(hdl->libzfs_fd, ZFS_IOC_CLEAR, &zc) == 0)
2910 return (zpool_standard_error(hdl, errno, msg));
2914 * Convert from a devid string to a path.
2917 devid_to_path(char *devid_str)
2922 devid_nmlist_t *list = NULL;
2925 if (devid_str_decode(devid_str, &devid, &minor) != 0)
2928 ret = devid_deviceid_to_nmlist("/dev", devid, minor, &list);
2930 devid_str_free(minor);
2936 if ((path = strdup(list[0].devname)) == NULL)
2939 devid_free_nmlist(list);
2945 * Convert from a path to a devid string.
2948 path_to_devid(const char *path)
2954 if ((fd = open(path, O_RDONLY)) < 0)
2959 if (devid_get(fd, &devid) == 0) {
2960 if (devid_get_minor_name(fd, &minor) == 0)
2961 ret = devid_str_encode(devid, minor);
2963 devid_str_free(minor);
2972 * Issue the necessary ioctl() to update the stored path value for the vdev. We
2973 * ignore any failure here, since a common case is for an unprivileged user to
2974 * type 'zpool status', and we'll display the correct information anyway.
2977 set_path(zpool_handle_t *zhp, nvlist_t *nv, const char *path)
2979 zfs_cmd_t zc = { 0 };
2981 (void) strncpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
2982 (void) strncpy(zc.zc_value, path, sizeof (zc.zc_value));
2983 verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID,
2986 (void) ioctl(zhp->zpool_hdl->libzfs_fd, ZFS_IOC_VDEV_SETPATH, &zc);
2990 * Given a vdev, return the name to display in iostat. If the vdev has a path,
2991 * we use that, stripping off any leading "/dev/dsk/"; if not, we use the type.
2992 * We also check if this is a whole disk, in which case we strip off the
2993 * trailing 's0' slice name.
2995 * This routine is also responsible for identifying when disks have been
2996 * reconfigured in a new location. The kernel will have opened the device by
2997 * devid, but the path will still refer to the old location. To catch this, we
2998 * first do a path -> devid translation (which is fast for the common case). If
2999 * the devid matches, we're done. If not, we do a reverse devid -> path
3000 * translation and issue the appropriate ioctl() to update the path of the vdev.
3001 * If 'zhp' is NULL, then this is an exported pool, and we don't need to do any
3005 zpool_vdev_name(libzfs_handle_t *hdl, zpool_handle_t *zhp, nvlist_t *nv,
3014 if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_NOT_PRESENT,
3016 verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID,
3018 (void) snprintf(buf, sizeof (buf), "%llu",
3019 (u_longlong_t)value);
3021 } else if (nvlist_lookup_string(nv, ZPOOL_CONFIG_PATH, &path) == 0) {
3024 * If the device is dead (faulted, offline, etc) then don't
3025 * bother opening it. Otherwise we may be forcing the user to
3026 * open a misbehaving device, which can have undesirable
3029 if ((nvlist_lookup_uint64_array(nv, ZPOOL_CONFIG_VDEV_STATS,
3030 (uint64_t **)&vs, &vsc) != 0 ||
3031 vs->vs_state >= VDEV_STATE_DEGRADED) &&
3033 nvlist_lookup_string(nv, ZPOOL_CONFIG_DEVID, &devid) == 0) {
3035 * Determine if the current path is correct.
3037 char *newdevid = path_to_devid(path);
3039 if (newdevid == NULL ||
3040 strcmp(devid, newdevid) != 0) {
3043 if ((newpath = devid_to_path(devid)) != NULL) {
3045 * Update the path appropriately.
3047 set_path(zhp, nv, newpath);
3048 if (nvlist_add_string(nv,
3049 ZPOOL_CONFIG_PATH, newpath) == 0)
3050 verify(nvlist_lookup_string(nv,
3058 devid_str_free(newdevid);
3061 if (strncmp(path, "/dev/dsk/", 9) == 0)
3064 if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_WHOLE_DISK,
3065 &value) == 0 && value) {
3066 int pathlen = strlen(path);
3067 char *tmp = zfs_strdup(hdl, path);
3070 * If it starts with c#, and ends with "s0", chop
3071 * the "s0" off, or if it ends with "s0/old", remove
3072 * the "s0" from the middle.
3074 if (CTD_CHECK(tmp)) {
3075 if (strcmp(&tmp[pathlen - 2], "s0") == 0) {
3076 tmp[pathlen - 2] = '\0';
3077 } else if (pathlen > 6 &&
3078 strcmp(&tmp[pathlen - 6], "s0/old") == 0) {
3079 (void) strcpy(&tmp[pathlen - 6],
3086 verify(nvlist_lookup_string(nv, ZPOOL_CONFIG_TYPE, &path) == 0);
3089 * If it's a raidz device, we need to stick in the parity level.
3091 if (strcmp(path, VDEV_TYPE_RAIDZ) == 0) {
3092 verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_NPARITY,
3094 (void) snprintf(buf, sizeof (buf), "%s%llu", path,
3095 (u_longlong_t)value);
3100 * We identify each top-level vdev by using a <type-id>
3101 * naming convention.
3106 verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_ID,
3108 (void) snprintf(buf, sizeof (buf), "%s-%llu", path,
3114 return (zfs_strdup(hdl, path));
3118 zbookmark_compare(const void *a, const void *b)
3120 return (memcmp(a, b, sizeof (zbookmark_t)));
3124 * Retrieve the persistent error log, uniquify the members, and return to the
3128 zpool_get_errlog(zpool_handle_t *zhp, nvlist_t **nverrlistp)
3130 zfs_cmd_t zc = { 0 };
3132 zbookmark_t *zb = NULL;
3136 * Retrieve the raw error list from the kernel. If the number of errors
3137 * has increased, allocate more space and continue until we get the
3140 verify(nvlist_lookup_uint64(zhp->zpool_config, ZPOOL_CONFIG_ERRCOUNT,
3144 if ((zc.zc_nvlist_dst = (uintptr_t)zfs_alloc(zhp->zpool_hdl,
3145 count * sizeof (zbookmark_t))) == (uintptr_t)NULL)
3147 zc.zc_nvlist_dst_size = count;
3148 (void) strcpy(zc.zc_name, zhp->zpool_name);
3150 if (ioctl(zhp->zpool_hdl->libzfs_fd, ZFS_IOC_ERROR_LOG,
3152 free((void *)(uintptr_t)zc.zc_nvlist_dst);
3153 if (errno == ENOMEM) {
3154 count = zc.zc_nvlist_dst_size;
3155 if ((zc.zc_nvlist_dst = (uintptr_t)
3156 zfs_alloc(zhp->zpool_hdl, count *
3157 sizeof (zbookmark_t))) == (uintptr_t)NULL)
3168 * Sort the resulting bookmarks. This is a little confusing due to the
3169 * implementation of ZFS_IOC_ERROR_LOG. The bookmarks are copied last
3170 * to first, and 'zc_nvlist_dst_size' indicates the number of boomarks
3171 * _not_ copied as part of the process. So we point the start of our
3172 * array appropriate and decrement the total number of elements.
3174 zb = ((zbookmark_t *)(uintptr_t)zc.zc_nvlist_dst) +
3175 zc.zc_nvlist_dst_size;
3176 count -= zc.zc_nvlist_dst_size;
3178 qsort(zb, count, sizeof (zbookmark_t), zbookmark_compare);
3180 verify(nvlist_alloc(nverrlistp, 0, KM_SLEEP) == 0);
3183 * Fill in the nverrlistp with nvlist's of dataset and object numbers.
3185 for (i = 0; i < count; i++) {
3188 /* ignoring zb_blkid and zb_level for now */
3189 if (i > 0 && zb[i-1].zb_objset == zb[i].zb_objset &&
3190 zb[i-1].zb_object == zb[i].zb_object)
3193 if (nvlist_alloc(&nv, NV_UNIQUE_NAME, KM_SLEEP) != 0)
3195 if (nvlist_add_uint64(nv, ZPOOL_ERR_DATASET,
3196 zb[i].zb_objset) != 0) {
3200 if (nvlist_add_uint64(nv, ZPOOL_ERR_OBJECT,
3201 zb[i].zb_object) != 0) {
3205 if (nvlist_add_nvlist(*nverrlistp, "ejk", nv) != 0) {
3212 free((void *)(uintptr_t)zc.zc_nvlist_dst);
3216 free((void *)(uintptr_t)zc.zc_nvlist_dst);
3217 return (no_memory(zhp->zpool_hdl));
3221 * Upgrade a ZFS pool to the latest on-disk version.
3224 zpool_upgrade(zpool_handle_t *zhp, uint64_t new_version)
3226 zfs_cmd_t zc = { 0 };
3227 libzfs_handle_t *hdl = zhp->zpool_hdl;
3229 (void) strcpy(zc.zc_name, zhp->zpool_name);
3230 zc.zc_cookie = new_version;
3232 if (zfs_ioctl(hdl, ZFS_IOC_POOL_UPGRADE, &zc) != 0)
3233 return (zpool_standard_error_fmt(hdl, errno,
3234 dgettext(TEXT_DOMAIN, "cannot upgrade '%s'"),
3240 zpool_set_history_str(const char *subcommand, int argc, char **argv,
3245 (void) strlcpy(history_str, subcommand, HIS_MAX_RECORD_LEN);
3246 for (i = 1; i < argc; i++) {
3247 if (strlen(history_str) + 1 + strlen(argv[i]) >
3250 (void) strlcat(history_str, " ", HIS_MAX_RECORD_LEN);
3251 (void) strlcat(history_str, argv[i], HIS_MAX_RECORD_LEN);
3256 * Stage command history for logging.
3259 zpool_stage_history(libzfs_handle_t *hdl, const char *history_str)
3261 if (history_str == NULL)
3264 if (strlen(history_str) > HIS_MAX_RECORD_LEN)
3267 if (hdl->libzfs_log_str != NULL)
3268 free(hdl->libzfs_log_str);
3270 if ((hdl->libzfs_log_str = strdup(history_str)) == NULL)
3271 return (no_memory(hdl));
3277 * Perform ioctl to get some command history of a pool.
3279 * 'buf' is the buffer to fill up to 'len' bytes. 'off' is the
3280 * logical offset of the history buffer to start reading from.
3282 * Upon return, 'off' is the next logical offset to read from and
3283 * 'len' is the actual amount of bytes read into 'buf'.
3286 get_history(zpool_handle_t *zhp, char *buf, uint64_t *off, uint64_t *len)
3288 zfs_cmd_t zc = { 0 };
3289 libzfs_handle_t *hdl = zhp->zpool_hdl;
3291 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
3293 zc.zc_history = (uint64_t)(uintptr_t)buf;
3294 zc.zc_history_len = *len;
3295 zc.zc_history_offset = *off;
3297 if (ioctl(hdl->libzfs_fd, ZFS_IOC_POOL_GET_HISTORY, &zc) != 0) {
3300 return (zfs_error_fmt(hdl, EZFS_PERM,
3301 dgettext(TEXT_DOMAIN,
3302 "cannot show history for pool '%s'"),
3305 return (zfs_error_fmt(hdl, EZFS_NOHISTORY,
3306 dgettext(TEXT_DOMAIN, "cannot get history for pool "
3307 "'%s'"), zhp->zpool_name));
3309 return (zfs_error_fmt(hdl, EZFS_BADVERSION,
3310 dgettext(TEXT_DOMAIN, "cannot get history for pool "
3311 "'%s', pool must be upgraded"), zhp->zpool_name));
3313 return (zpool_standard_error_fmt(hdl, errno,
3314 dgettext(TEXT_DOMAIN,
3315 "cannot get history for '%s'"), zhp->zpool_name));
3319 *len = zc.zc_history_len;
3320 *off = zc.zc_history_offset;
3326 * Process the buffer of nvlists, unpacking and storing each nvlist record
3327 * into 'records'. 'leftover' is set to the number of bytes that weren't
3328 * processed as there wasn't a complete record.
3331 zpool_history_unpack(char *buf, uint64_t bytes_read, uint64_t *leftover,
3332 nvlist_t ***records, uint_t *numrecords)
3338 while (bytes_read > sizeof (reclen)) {
3340 /* get length of packed record (stored as little endian) */
3341 for (i = 0, reclen = 0; i < sizeof (reclen); i++)
3342 reclen += (uint64_t)(((uchar_t *)buf)[i]) << (8*i);
3344 if (bytes_read < sizeof (reclen) + reclen)
3348 if (nvlist_unpack(buf + sizeof (reclen), reclen, &nv, 0) != 0)
3350 bytes_read -= sizeof (reclen) + reclen;
3351 buf += sizeof (reclen) + reclen;
3353 /* add record to nvlist array */
3355 if (ISP2(*numrecords + 1)) {
3356 *records = realloc(*records,
3357 *numrecords * 2 * sizeof (nvlist_t *));
3359 (*records)[*numrecords - 1] = nv;
3362 *leftover = bytes_read;
3366 #define HIS_BUF_LEN (128*1024)
3369 * Retrieve the command history of a pool.
3372 zpool_get_history(zpool_handle_t *zhp, nvlist_t **nvhisp)
3374 char buf[HIS_BUF_LEN];
3376 nvlist_t **records = NULL;
3377 uint_t numrecords = 0;
3381 uint64_t bytes_read = sizeof (buf);
3384 if ((err = get_history(zhp, buf, &off, &bytes_read)) != 0)
3387 /* if nothing else was read in, we're at EOF, just return */
3391 if ((err = zpool_history_unpack(buf, bytes_read,
3392 &leftover, &records, &numrecords)) != 0)
3400 verify(nvlist_alloc(nvhisp, NV_UNIQUE_NAME, 0) == 0);
3401 verify(nvlist_add_nvlist_array(*nvhisp, ZPOOL_HIST_RECORD,
3402 records, numrecords) == 0);
3404 for (i = 0; i < numrecords; i++)
3405 nvlist_free(records[i]);
3412 zpool_obj_to_path(zpool_handle_t *zhp, uint64_t dsobj, uint64_t obj,
3413 char *pathname, size_t len)
3415 zfs_cmd_t zc = { 0 };
3416 boolean_t mounted = B_FALSE;
3417 char *mntpnt = NULL;
3418 char dsname[MAXNAMELEN];
3421 /* special case for the MOS */
3422 (void) snprintf(pathname, len, "<metadata>:<0x%llx>", obj);
3426 /* get the dataset's name */
3427 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
3429 if (ioctl(zhp->zpool_hdl->libzfs_fd,
3430 ZFS_IOC_DSOBJ_TO_DSNAME, &zc) != 0) {
3431 /* just write out a path of two object numbers */
3432 (void) snprintf(pathname, len, "<0x%llx>:<0x%llx>",
3436 (void) strlcpy(dsname, zc.zc_value, sizeof (dsname));
3438 /* find out if the dataset is mounted */
3439 mounted = is_mounted(zhp->zpool_hdl, dsname, &mntpnt);
3441 /* get the corrupted object's path */
3442 (void) strlcpy(zc.zc_name, dsname, sizeof (zc.zc_name));
3444 if (ioctl(zhp->zpool_hdl->libzfs_fd, ZFS_IOC_OBJ_TO_PATH,
3447 (void) snprintf(pathname, len, "%s%s", mntpnt,
3450 (void) snprintf(pathname, len, "%s:%s",
3451 dsname, zc.zc_value);
3454 (void) snprintf(pathname, len, "%s:<0x%llx>", dsname, obj);
3460 * Read the EFI label from the config, if a label does not exist then
3461 * pass back the error to the caller. If the caller has passed a non-NULL
3462 * diskaddr argument then we set it to the starting address of the EFI
3466 read_efi_label(nvlist_t *config, diskaddr_t *sb)
3470 char diskname[MAXPATHLEN];
3473 if (nvlist_lookup_string(config, ZPOOL_CONFIG_PATH, &path) != 0)
3476 (void) snprintf(diskname, sizeof (diskname), "%s%s", RDISK_ROOT,
3477 strrchr(path, '/'));
3478 if ((fd = open(diskname, O_RDONLY|O_NDELAY)) >= 0) {
3479 struct dk_gpt *vtoc;
3481 if ((err = efi_alloc_and_read(fd, &vtoc)) >= 0) {
3483 *sb = vtoc->efi_parts[0].p_start;
3492 * determine where a partition starts on a disk in the current
3496 find_start_block(nvlist_t *config)
3500 diskaddr_t sb = MAXOFFSET_T;
3503 if (nvlist_lookup_nvlist_array(config,
3504 ZPOOL_CONFIG_CHILDREN, &child, &children) != 0) {
3505 if (nvlist_lookup_uint64(config,
3506 ZPOOL_CONFIG_WHOLE_DISK,
3507 &wholedisk) != 0 || !wholedisk) {
3508 return (MAXOFFSET_T);
3510 if (read_efi_label(config, &sb) < 0)
3515 for (c = 0; c < children; c++) {
3516 sb = find_start_block(child[c]);
3517 if (sb != MAXOFFSET_T) {
3521 return (MAXOFFSET_T);
3525 * Label an individual disk. The name provided is the short name,
3526 * stripped of any leading /dev path.
3529 zpool_label_disk(libzfs_handle_t *hdl, zpool_handle_t *zhp, char *name)
3531 char path[MAXPATHLEN];
3532 struct dk_gpt *vtoc;
3534 size_t resv = EFI_MIN_RESV_SIZE;
3535 uint64_t slice_size;
3536 diskaddr_t start_block;
3539 /* prepare an error message just in case */
3540 (void) snprintf(errbuf, sizeof (errbuf),
3541 dgettext(TEXT_DOMAIN, "cannot label '%s'"), name);
3546 if (pool_is_bootable(zhp)) {
3547 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3548 "EFI labeled devices are not supported on root "
3550 return (zfs_error(hdl, EZFS_POOL_NOTSUP, errbuf));
3553 verify(nvlist_lookup_nvlist(zhp->zpool_config,
3554 ZPOOL_CONFIG_VDEV_TREE, &nvroot) == 0);
3556 if (zhp->zpool_start_block == 0)
3557 start_block = find_start_block(nvroot);
3559 start_block = zhp->zpool_start_block;
3560 zhp->zpool_start_block = start_block;
3563 start_block = NEW_START_BLOCK;
3566 (void) snprintf(path, sizeof (path), "%s/%s%s", RDISK_ROOT, name,
3569 if ((fd = open(path, O_RDWR | O_NDELAY)) < 0) {
3571 * This shouldn't happen. We've long since verified that this
3572 * is a valid device.
3575 dgettext(TEXT_DOMAIN, "unable to open device"));
3576 return (zfs_error(hdl, EZFS_OPENFAILED, errbuf));
3579 if (efi_alloc_and_init(fd, EFI_NUMPAR, &vtoc) != 0) {
3581 * The only way this can fail is if we run out of memory, or we
3582 * were unable to read the disk's capacity
3584 if (errno == ENOMEM)
3585 (void) no_memory(hdl);
3588 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3589 "unable to read disk capacity"), name);
3591 return (zfs_error(hdl, EZFS_NOCAP, errbuf));
3594 slice_size = vtoc->efi_last_u_lba + 1;
3595 slice_size -= EFI_MIN_RESV_SIZE;
3596 if (start_block == MAXOFFSET_T)
3597 start_block = NEW_START_BLOCK;
3598 slice_size -= start_block;
3600 vtoc->efi_parts[0].p_start = start_block;
3601 vtoc->efi_parts[0].p_size = slice_size;
3604 * Why we use V_USR: V_BACKUP confuses users, and is considered
3605 * disposable by some EFI utilities (since EFI doesn't have a backup
3606 * slice). V_UNASSIGNED is supposed to be used only for zero size
3607 * partitions, and efi_write() will fail if we use it. V_ROOT, V_BOOT,
3608 * etc. were all pretty specific. V_USR is as close to reality as we
3609 * can get, in the absence of V_OTHER.
3611 vtoc->efi_parts[0].p_tag = V_USR;
3612 (void) strcpy(vtoc->efi_parts[0].p_name, "zfs");
3614 vtoc->efi_parts[8].p_start = slice_size + start_block;
3615 vtoc->efi_parts[8].p_size = resv;
3616 vtoc->efi_parts[8].p_tag = V_RESERVED;
3618 if (efi_write(fd, vtoc) != 0) {
3620 * Some block drivers (like pcata) may not support EFI
3621 * GPT labels. Print out a helpful error message dir-
3622 * ecting the user to manually label the disk and give
3628 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3629 "try using fdisk(1M) and then provide a specific slice"));
3630 return (zfs_error(hdl, EZFS_LABELFAILED, errbuf));
3639 supported_dump_vdev_type(libzfs_handle_t *hdl, nvlist_t *config, char *errbuf)
3645 verify(nvlist_lookup_string(config, ZPOOL_CONFIG_TYPE, &type) == 0);
3646 if (strcmp(type, VDEV_TYPE_RAIDZ) == 0 ||
3647 strcmp(type, VDEV_TYPE_FILE) == 0 ||
3648 strcmp(type, VDEV_TYPE_LOG) == 0 ||
3649 strcmp(type, VDEV_TYPE_HOLE) == 0 ||
3650 strcmp(type, VDEV_TYPE_MISSING) == 0) {
3651 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3652 "vdev type '%s' is not supported"), type);
3653 (void) zfs_error(hdl, EZFS_VDEVNOTSUP, errbuf);
3656 if (nvlist_lookup_nvlist_array(config, ZPOOL_CONFIG_CHILDREN,
3657 &child, &children) == 0) {
3658 for (c = 0; c < children; c++) {
3659 if (!supported_dump_vdev_type(hdl, child[c], errbuf))
3667 * check if this zvol is allowable for use as a dump device; zero if
3668 * it is, > 0 if it isn't, < 0 if it isn't a zvol
3671 zvol_check_dump_config(char *arg)
3673 zpool_handle_t *zhp = NULL;
3674 nvlist_t *config, *nvroot;
3678 libzfs_handle_t *hdl;
3680 char poolname[ZPOOL_MAXNAMELEN];
3681 int pathlen = strlen(ZVOL_FULL_DEV_DIR);
3684 if (strncmp(arg, ZVOL_FULL_DEV_DIR, pathlen)) {
3688 (void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN,
3689 "dump is not supported on device '%s'"), arg);
3691 if ((hdl = libzfs_init()) == NULL)
3693 libzfs_print_on_error(hdl, B_TRUE);
3695 volname = arg + pathlen;
3697 /* check the configuration of the pool */
3698 if ((p = strchr(volname, '/')) == NULL) {
3699 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3700 "malformed dataset name"));
3701 (void) zfs_error(hdl, EZFS_INVALIDNAME, errbuf);
3703 } else if (p - volname >= ZFS_MAXNAMELEN) {
3704 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3705 "dataset name is too long"));
3706 (void) zfs_error(hdl, EZFS_NAMETOOLONG, errbuf);
3709 (void) strncpy(poolname, volname, p - volname);
3710 poolname[p - volname] = '\0';
3713 if ((zhp = zpool_open(hdl, poolname)) == NULL) {
3714 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3715 "could not open pool '%s'"), poolname);
3716 (void) zfs_error(hdl, EZFS_OPENFAILED, errbuf);
3719 config = zpool_get_config(zhp, NULL);
3720 if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE,
3722 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3723 "could not obtain vdev configuration for '%s'"), poolname);
3724 (void) zfs_error(hdl, EZFS_INVALCONFIG, errbuf);
3728 verify(nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_CHILDREN,
3729 &top, &toplevels) == 0);
3730 if (toplevels != 1) {
3731 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3732 "'%s' has multiple top level vdevs"), poolname);
3733 (void) zfs_error(hdl, EZFS_DEVOVERFLOW, errbuf);
3737 if (!supported_dump_vdev_type(hdl, top[0], errbuf)) {