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.
37 #include <sys/efi_partition.h>
39 #include <sys/zfs_ioctl.h>
42 #include "zfs_namecheck.h"
44 #include "libzfs_impl.h"
45 #include "zfs_comutil.h"
47 static int read_efi_label(nvlist_t *config, diskaddr_t *sb);
49 typedef struct prop_flags {
50 int create:1; /* Validate property on creation */
51 int import:1; /* Validate property on import */
55 * ====================================================================
56 * zpool property functions
57 * ====================================================================
61 zpool_get_all_props(zpool_handle_t *zhp)
63 zfs_cmd_t zc = { "\0", "\0", "\0", "\0", 0 };
64 libzfs_handle_t *hdl = zhp->zpool_hdl;
66 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
68 if (zcmd_alloc_dst_nvlist(hdl, &zc, 0) != 0)
71 while (ioctl(hdl->libzfs_fd, ZFS_IOC_POOL_GET_PROPS, &zc) != 0) {
72 if (errno == ENOMEM) {
73 if (zcmd_expand_dst_nvlist(hdl, &zc) != 0) {
74 zcmd_free_nvlists(&zc);
78 zcmd_free_nvlists(&zc);
83 if (zcmd_read_dst_nvlist(hdl, &zc, &zhp->zpool_props) != 0) {
84 zcmd_free_nvlists(&zc);
88 zcmd_free_nvlists(&zc);
94 zpool_props_refresh(zpool_handle_t *zhp)
98 old_props = zhp->zpool_props;
100 if (zpool_get_all_props(zhp) != 0)
103 nvlist_free(old_props);
108 zpool_get_prop_string(zpool_handle_t *zhp, zpool_prop_t prop,
114 zprop_source_t source;
116 nvl = zhp->zpool_props;
117 if (nvlist_lookup_nvlist(nvl, zpool_prop_to_name(prop), &nv) == 0) {
118 verify(nvlist_lookup_uint64(nv, ZPROP_SOURCE, &ival) == 0);
120 verify(nvlist_lookup_string(nv, ZPROP_VALUE, &value) == 0);
122 source = ZPROP_SRC_DEFAULT;
123 if ((value = (char *)zpool_prop_default_string(prop)) == NULL)
134 zpool_get_prop_int(zpool_handle_t *zhp, zpool_prop_t prop, zprop_source_t *src)
138 zprop_source_t source;
140 if (zhp->zpool_props == NULL && zpool_get_all_props(zhp)) {
142 * zpool_get_all_props() has most likely failed because
143 * the pool is faulted, but if all we need is the top level
144 * vdev's guid then get it from the zhp config nvlist.
146 if ((prop == ZPOOL_PROP_GUID) &&
147 (nvlist_lookup_nvlist(zhp->zpool_config,
148 ZPOOL_CONFIG_VDEV_TREE, &nv) == 0) &&
149 (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID, &value)
153 return (zpool_prop_default_numeric(prop));
156 nvl = zhp->zpool_props;
157 if (nvlist_lookup_nvlist(nvl, zpool_prop_to_name(prop), &nv) == 0) {
158 verify(nvlist_lookup_uint64(nv, ZPROP_SOURCE, &value) == 0);
160 verify(nvlist_lookup_uint64(nv, ZPROP_VALUE, &value) == 0);
162 source = ZPROP_SRC_DEFAULT;
163 value = zpool_prop_default_numeric(prop);
173 * Map VDEV STATE to printed strings.
176 zpool_state_to_name(vdev_state_t state, vdev_aux_t aux)
181 case VDEV_STATE_CLOSED:
182 case VDEV_STATE_OFFLINE:
183 return (gettext("OFFLINE"));
184 case VDEV_STATE_REMOVED:
185 return (gettext("REMOVED"));
186 case VDEV_STATE_CANT_OPEN:
187 if (aux == VDEV_AUX_CORRUPT_DATA || aux == VDEV_AUX_BAD_LOG)
188 return (gettext("FAULTED"));
189 else if (aux == VDEV_AUX_SPLIT_POOL)
190 return (gettext("SPLIT"));
192 return (gettext("UNAVAIL"));
193 case VDEV_STATE_FAULTED:
194 return (gettext("FAULTED"));
195 case VDEV_STATE_DEGRADED:
196 return (gettext("DEGRADED"));
197 case VDEV_STATE_HEALTHY:
198 return (gettext("ONLINE"));
201 return (gettext("UNKNOWN"));
205 * Get a zpool property value for 'prop' and return the value in
206 * a pre-allocated buffer.
209 zpool_get_prop(zpool_handle_t *zhp, zpool_prop_t prop, char *buf, size_t len,
210 zprop_source_t *srctype)
214 zprop_source_t src = ZPROP_SRC_NONE;
219 if (zpool_get_state(zhp) == POOL_STATE_UNAVAIL) {
221 case ZPOOL_PROP_NAME:
222 (void) strlcpy(buf, zpool_get_name(zhp), len);
225 case ZPOOL_PROP_HEALTH:
226 (void) strlcpy(buf, "FAULTED", len);
229 case ZPOOL_PROP_GUID:
230 intval = zpool_get_prop_int(zhp, prop, &src);
231 (void) snprintf(buf, len, "%llu", (u_longlong_t)intval);
234 case ZPOOL_PROP_ALTROOT:
235 case ZPOOL_PROP_CACHEFILE:
236 if (zhp->zpool_props != NULL ||
237 zpool_get_all_props(zhp) == 0) {
239 zpool_get_prop_string(zhp, prop, &src),
247 (void) strlcpy(buf, "-", len);
256 if (zhp->zpool_props == NULL && zpool_get_all_props(zhp) &&
257 prop != ZPOOL_PROP_NAME)
260 switch (zpool_prop_get_type(prop)) {
261 case PROP_TYPE_STRING:
262 (void) strlcpy(buf, zpool_get_prop_string(zhp, prop, &src),
266 case PROP_TYPE_NUMBER:
267 intval = zpool_get_prop_int(zhp, prop, &src);
270 case ZPOOL_PROP_SIZE:
271 case ZPOOL_PROP_ALLOCATED:
272 case ZPOOL_PROP_FREE:
273 case ZPOOL_PROP_ASHIFT:
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", (u_longlong_t)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, prop_flags_t flags, 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.
427 case ZPOOL_PROP_VERSION:
428 if (intval < version || intval > SPA_VERSION) {
429 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
430 "property '%s' number %d is invalid."),
432 (void) zfs_error(hdl, EZFS_BADVERSION, errbuf);
437 case ZPOOL_PROP_ASHIFT:
439 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
440 "property '%s' can only be set at "
441 "creation time"), propname);
442 (void) zfs_error(hdl, EZFS_BADPROP, errbuf);
446 if (intval != 0 && (intval < 9 || intval > 17)) {
447 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
448 "property '%s' number %d is invalid."),
450 (void) zfs_error(hdl, EZFS_BADPROP, errbuf);
455 case ZPOOL_PROP_BOOTFS:
456 if (flags.create || flags.import) {
457 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
458 "property '%s' cannot be set at creation "
459 "or import time"), propname);
460 (void) zfs_error(hdl, EZFS_BADPROP, errbuf);
464 if (version < SPA_VERSION_BOOTFS) {
465 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
466 "pool must be upgraded to support "
467 "'%s' property"), propname);
468 (void) zfs_error(hdl, EZFS_BADVERSION, errbuf);
473 * bootfs property value has to be a dataset name and
474 * the dataset has to be in the same pool as it sets to.
476 if (strval[0] != '\0' && !bootfs_name_valid(poolname,
478 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "'%s' "
479 "is an invalid name"), strval);
480 (void) zfs_error(hdl, EZFS_INVALIDNAME, errbuf);
484 if ((zhp = zpool_open_canfail(hdl, poolname)) == NULL) {
485 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
486 "could not open pool '%s'"), poolname);
487 (void) zfs_error(hdl, EZFS_OPENFAILED, errbuf);
490 verify(nvlist_lookup_nvlist(zpool_get_config(zhp, NULL),
491 ZPOOL_CONFIG_VDEV_TREE, &nvroot) == 0);
494 * bootfs property cannot be set on a disk which has
497 if (pool_uses_efi(nvroot)) {
498 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
499 "property '%s' not supported on "
500 "EFI labeled devices"), propname);
501 (void) zfs_error(hdl, EZFS_POOL_NOTSUP, errbuf);
508 case ZPOOL_PROP_ALTROOT:
509 if (!flags.create && !flags.import) {
510 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
511 "property '%s' can only be set during pool "
512 "creation or import"), propname);
513 (void) zfs_error(hdl, EZFS_BADPROP, errbuf);
517 if (strval[0] != '/') {
518 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
519 "bad alternate root '%s'"), strval);
520 (void) zfs_error(hdl, EZFS_BADPATH, errbuf);
525 case ZPOOL_PROP_CACHEFILE:
526 if (strval[0] == '\0')
529 if (strcmp(strval, "none") == 0)
532 if (strval[0] != '/') {
533 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
534 "property '%s' must be empty, an "
535 "absolute path, or 'none'"), propname);
536 (void) zfs_error(hdl, EZFS_BADPATH, errbuf);
540 slash = strrchr(strval, '/');
542 if (slash[1] == '\0' || strcmp(slash, "/.") == 0 ||
543 strcmp(slash, "/..") == 0) {
544 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
545 "'%s' is not a valid file"), strval);
546 (void) zfs_error(hdl, EZFS_BADPATH, errbuf);
552 if (strval[0] != '\0' &&
553 (stat64(strval, &statbuf) != 0 ||
554 !S_ISDIR(statbuf.st_mode))) {
555 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
556 "'%s' is not a valid directory"),
558 (void) zfs_error(hdl, EZFS_BADPATH, errbuf);
565 case ZPOOL_PROP_READONLY:
567 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
568 "property '%s' can only be set at "
569 "import time"), propname);
570 (void) zfs_error(hdl, EZFS_BADPROP, errbuf);
579 nvlist_free(retprops);
584 * Set zpool property : propname=propval.
587 zpool_set_prop(zpool_handle_t *zhp, const char *propname, const char *propval)
589 zfs_cmd_t zc = { "\0", "\0", "\0", "\0", 0 };
592 nvlist_t *nvl = NULL;
595 prop_flags_t flags = { 0 };
597 (void) snprintf(errbuf, sizeof (errbuf),
598 dgettext(TEXT_DOMAIN, "cannot set property for '%s'"),
601 if (nvlist_alloc(&nvl, NV_UNIQUE_NAME, 0) != 0)
602 return (no_memory(zhp->zpool_hdl));
604 if (nvlist_add_string(nvl, propname, propval) != 0) {
606 return (no_memory(zhp->zpool_hdl));
609 version = zpool_get_prop_int(zhp, ZPOOL_PROP_VERSION, NULL);
610 if ((realprops = zpool_valid_proplist(zhp->zpool_hdl,
611 zhp->zpool_name, nvl, version, flags, errbuf)) == NULL) {
620 * Execute the corresponding ioctl() to set this property.
622 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
624 if (zcmd_write_src_nvlist(zhp->zpool_hdl, &zc, nvl) != 0) {
629 ret = zfs_ioctl(zhp->zpool_hdl, ZFS_IOC_POOL_SET_PROPS, &zc);
631 zcmd_free_nvlists(&zc);
635 (void) zpool_standard_error(zhp->zpool_hdl, errno, errbuf);
637 (void) zpool_props_refresh(zhp);
643 zpool_expand_proplist(zpool_handle_t *zhp, zprop_list_t **plp)
645 libzfs_handle_t *hdl = zhp->zpool_hdl;
647 char buf[ZFS_MAXPROPLEN];
649 if (zprop_expand_list(hdl, plp, ZFS_TYPE_POOL) != 0)
652 for (entry = *plp; entry != NULL; entry = entry->pl_next) {
657 if (entry->pl_prop != ZPROP_INVAL &&
658 zpool_get_prop(zhp, entry->pl_prop, buf, sizeof (buf),
660 if (strlen(buf) > entry->pl_width)
661 entry->pl_width = strlen(buf);
670 * Don't start the slice at the default block of 34; many storage
671 * devices will use a stripe width of 128k, other vendors prefer a 1m
672 * alignment. It is best to play it safe and ensure a 1m alignment
673 * give 512b blocks. When the block size is larger by a power of 2
674 * we will still be 1m aligned.
676 #define NEW_START_BLOCK 2048
679 * Validate the given pool name, optionally putting an extended error message in
683 zpool_name_valid(libzfs_handle_t *hdl, boolean_t isopen, const char *pool)
689 ret = pool_namecheck(pool, &why, &what);
692 * The rules for reserved pool names were extended at a later point.
693 * But we need to support users with existing pools that may now be
694 * invalid. So we only check for this expanded set of names during a
695 * create (or import), and only in userland.
697 if (ret == 0 && !isopen &&
698 (strncmp(pool, "mirror", 6) == 0 ||
699 strncmp(pool, "raidz", 5) == 0 ||
700 strncmp(pool, "spare", 5) == 0 ||
701 strcmp(pool, "log") == 0)) {
704 dgettext(TEXT_DOMAIN, "name is reserved"));
712 case NAME_ERR_TOOLONG:
714 dgettext(TEXT_DOMAIN, "name is too long"));
717 case NAME_ERR_INVALCHAR:
719 dgettext(TEXT_DOMAIN, "invalid character "
720 "'%c' in pool name"), what);
723 case NAME_ERR_NOLETTER:
724 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
725 "name must begin with a letter"));
728 case NAME_ERR_RESERVED:
729 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
730 "name is reserved"));
733 case NAME_ERR_DISKLIKE:
734 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
735 "pool name is reserved"));
738 case NAME_ERR_LEADING_SLASH:
739 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
740 "leading slash in name"));
743 case NAME_ERR_EMPTY_COMPONENT:
744 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
745 "empty component in name"));
748 case NAME_ERR_TRAILING_SLASH:
749 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
750 "trailing slash in name"));
753 case NAME_ERR_MULTIPLE_AT:
754 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
755 "multiple '@' delimiters in name"));
758 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
759 "permission set is missing '@'"));
770 * Open a handle to the given pool, even if the pool is currently in the FAULTED
774 zpool_open_canfail(libzfs_handle_t *hdl, const char *pool)
780 * Make sure the pool name is valid.
782 if (!zpool_name_valid(hdl, B_TRUE, pool)) {
783 (void) zfs_error_fmt(hdl, EZFS_INVALIDNAME,
784 dgettext(TEXT_DOMAIN, "cannot open '%s'"),
789 if ((zhp = zfs_alloc(hdl, sizeof (zpool_handle_t))) == NULL)
792 zhp->zpool_hdl = hdl;
793 (void) strlcpy(zhp->zpool_name, pool, sizeof (zhp->zpool_name));
795 if (zpool_refresh_stats(zhp, &missing) != 0) {
801 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "no such pool"));
802 (void) zfs_error_fmt(hdl, EZFS_NOENT,
803 dgettext(TEXT_DOMAIN, "cannot open '%s'"), pool);
812 * Like the above, but silent on error. Used when iterating over pools (because
813 * the configuration cache may be out of date).
816 zpool_open_silent(libzfs_handle_t *hdl, const char *pool, zpool_handle_t **ret)
821 if ((zhp = zfs_alloc(hdl, sizeof (zpool_handle_t))) == NULL)
824 zhp->zpool_hdl = hdl;
825 (void) strlcpy(zhp->zpool_name, pool, sizeof (zhp->zpool_name));
827 if (zpool_refresh_stats(zhp, &missing) != 0) {
843 * Similar to zpool_open_canfail(), but refuses to open pools in the faulted
847 zpool_open(libzfs_handle_t *hdl, const char *pool)
851 if ((zhp = zpool_open_canfail(hdl, pool)) == NULL)
854 if (zhp->zpool_state == POOL_STATE_UNAVAIL) {
855 (void) zfs_error_fmt(hdl, EZFS_POOLUNAVAIL,
856 dgettext(TEXT_DOMAIN, "cannot open '%s'"), zhp->zpool_name);
865 * Close the handle. Simply frees the memory associated with the handle.
868 zpool_close(zpool_handle_t *zhp)
870 if (zhp->zpool_config)
871 nvlist_free(zhp->zpool_config);
872 if (zhp->zpool_old_config)
873 nvlist_free(zhp->zpool_old_config);
874 if (zhp->zpool_props)
875 nvlist_free(zhp->zpool_props);
880 * Return the name of the pool.
883 zpool_get_name(zpool_handle_t *zhp)
885 return (zhp->zpool_name);
890 * Return the state of the pool (ACTIVE or UNAVAILABLE)
893 zpool_get_state(zpool_handle_t *zhp)
895 return (zhp->zpool_state);
899 * Create the named pool, using the provided vdev list. It is assumed
900 * that the consumer has already validated the contents of the nvlist, so we
901 * don't have to worry about error semantics.
904 zpool_create(libzfs_handle_t *hdl, const char *pool, nvlist_t *nvroot,
905 nvlist_t *props, nvlist_t *fsprops)
907 zfs_cmd_t zc = { "\0", "\0", "\0", "\0", 0 };
908 nvlist_t *zc_fsprops = NULL;
909 nvlist_t *zc_props = NULL;
914 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN,
915 "cannot create '%s'"), pool);
917 if (!zpool_name_valid(hdl, B_FALSE, pool))
918 return (zfs_error(hdl, EZFS_INVALIDNAME, msg));
920 if (zcmd_write_conf_nvlist(hdl, &zc, nvroot) != 0)
924 prop_flags_t flags = { .create = B_TRUE, .import = B_FALSE };
926 if ((zc_props = zpool_valid_proplist(hdl, pool, props,
927 SPA_VERSION_1, flags, msg)) == NULL) {
936 zoned = ((nvlist_lookup_string(fsprops,
937 zfs_prop_to_name(ZFS_PROP_ZONED), &zonestr) == 0) &&
938 strcmp(zonestr, "on") == 0);
940 if ((zc_fsprops = zfs_valid_proplist(hdl,
941 ZFS_TYPE_FILESYSTEM, fsprops, zoned, NULL, msg)) == NULL) {
945 (nvlist_alloc(&zc_props, NV_UNIQUE_NAME, 0) != 0)) {
948 if (nvlist_add_nvlist(zc_props,
949 ZPOOL_ROOTFS_PROPS, zc_fsprops) != 0) {
954 if (zc_props && zcmd_write_src_nvlist(hdl, &zc, zc_props) != 0)
957 (void) strlcpy(zc.zc_name, pool, sizeof (zc.zc_name));
959 if ((ret = zfs_ioctl(hdl, ZFS_IOC_POOL_CREATE, &zc)) != 0) {
961 zcmd_free_nvlists(&zc);
962 nvlist_free(zc_props);
963 nvlist_free(zc_fsprops);
968 * This can happen if the user has specified the same
969 * device multiple times. We can't reliably detect this
970 * until we try to add it and see we already have a
971 * label. This can also happen under if the device is
972 * part of an active md or lvm device.
974 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
975 "one or more vdevs refer to the same device, or one of\n"
976 "the devices is part of an active md or lvm device"));
977 return (zfs_error(hdl, EZFS_BADDEV, msg));
981 * This occurs when one of the devices is below
982 * SPA_MINDEVSIZE. Unfortunately, we can't detect which
983 * device was the problem device since there's no
984 * reliable way to determine device size from userland.
989 zfs_nicenum(SPA_MINDEVSIZE, buf, sizeof (buf));
991 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
992 "one or more devices is less than the "
993 "minimum size (%s)"), buf);
995 return (zfs_error(hdl, EZFS_BADDEV, msg));
998 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
999 "one or more devices is out of space"));
1000 return (zfs_error(hdl, EZFS_BADDEV, msg));
1003 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1004 "cache device must be a disk or disk slice"));
1005 return (zfs_error(hdl, EZFS_BADDEV, msg));
1008 return (zpool_standard_error(hdl, errno, msg));
1013 * If this is an alternate root pool, then we automatically set the
1014 * mountpoint of the root dataset to be '/'.
1016 if (nvlist_lookup_string(props, zpool_prop_to_name(ZPOOL_PROP_ALTROOT),
1020 verify((zhp = zfs_open(hdl, pool, ZFS_TYPE_DATASET)) != NULL);
1021 verify(zfs_prop_set(zhp, zfs_prop_to_name(ZFS_PROP_MOUNTPOINT),
1028 zcmd_free_nvlists(&zc);
1029 nvlist_free(zc_props);
1030 nvlist_free(zc_fsprops);
1035 * Destroy the given pool. It is up to the caller to ensure that there are no
1036 * datasets left in the pool.
1039 zpool_destroy(zpool_handle_t *zhp)
1041 zfs_cmd_t zc = { "\0", "\0", "\0", "\0", 0 };
1042 zfs_handle_t *zfp = NULL;
1043 libzfs_handle_t *hdl = zhp->zpool_hdl;
1046 if (zhp->zpool_state == POOL_STATE_ACTIVE &&
1047 (zfp = zfs_open(hdl, zhp->zpool_name, ZFS_TYPE_FILESYSTEM)) == NULL)
1050 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
1052 if (zfs_ioctl(hdl, ZFS_IOC_POOL_DESTROY, &zc) != 0) {
1053 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN,
1054 "cannot destroy '%s'"), zhp->zpool_name);
1056 if (errno == EROFS) {
1057 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1058 "one or more devices is read only"));
1059 (void) zfs_error(hdl, EZFS_BADDEV, msg);
1061 (void) zpool_standard_error(hdl, errno, msg);
1070 remove_mountpoint(zfp);
1078 * Add the given vdevs to the pool. The caller must have already performed the
1079 * necessary verification to ensure that the vdev specification is well-formed.
1082 zpool_add(zpool_handle_t *zhp, nvlist_t *nvroot)
1084 zfs_cmd_t zc = { "\0", "\0", "\0", "\0", 0 };
1086 libzfs_handle_t *hdl = zhp->zpool_hdl;
1088 nvlist_t **spares, **l2cache;
1089 uint_t nspares, nl2cache;
1091 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN,
1092 "cannot add to '%s'"), zhp->zpool_name);
1094 if (zpool_get_prop_int(zhp, ZPOOL_PROP_VERSION, NULL) <
1095 SPA_VERSION_SPARES &&
1096 nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_SPARES,
1097 &spares, &nspares) == 0) {
1098 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "pool must be "
1099 "upgraded to add hot spares"));
1100 return (zfs_error(hdl, EZFS_BADVERSION, msg));
1103 if (pool_is_bootable(zhp) && nvlist_lookup_nvlist_array(nvroot,
1104 ZPOOL_CONFIG_SPARES, &spares, &nspares) == 0) {
1107 for (s = 0; s < nspares; s++) {
1110 if (nvlist_lookup_string(spares[s], ZPOOL_CONFIG_PATH,
1111 &path) == 0 && pool_uses_efi(spares[s])) {
1112 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1113 "device '%s' contains an EFI label and "
1114 "cannot be used on root pools."),
1115 zpool_vdev_name(hdl, NULL, spares[s],
1117 return (zfs_error(hdl, EZFS_POOL_NOTSUP, msg));
1122 if (zpool_get_prop_int(zhp, ZPOOL_PROP_VERSION, NULL) <
1123 SPA_VERSION_L2CACHE &&
1124 nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_L2CACHE,
1125 &l2cache, &nl2cache) == 0) {
1126 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "pool must be "
1127 "upgraded to add cache devices"));
1128 return (zfs_error(hdl, EZFS_BADVERSION, msg));
1131 if (zcmd_write_conf_nvlist(hdl, &zc, nvroot) != 0)
1133 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
1135 if (zfs_ioctl(hdl, ZFS_IOC_VDEV_ADD, &zc) != 0) {
1139 * This can happen if the user has specified the same
1140 * device multiple times. We can't reliably detect this
1141 * until we try to add it and see we already have a
1144 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1145 "one or more vdevs refer to the same device"));
1146 (void) zfs_error(hdl, EZFS_BADDEV, msg);
1151 * This occurrs when one of the devices is below
1152 * SPA_MINDEVSIZE. Unfortunately, we can't detect which
1153 * device was the problem device since there's no
1154 * reliable way to determine device size from userland.
1159 zfs_nicenum(SPA_MINDEVSIZE, buf, sizeof (buf));
1161 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1162 "device is less than the minimum "
1165 (void) zfs_error(hdl, EZFS_BADDEV, msg);
1169 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1170 "pool must be upgraded to add these vdevs"));
1171 (void) zfs_error(hdl, EZFS_BADVERSION, msg);
1175 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1176 "root pool can not have multiple vdevs"
1177 " or separate logs"));
1178 (void) zfs_error(hdl, EZFS_POOL_NOTSUP, msg);
1182 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1183 "cache device must be a disk or disk slice"));
1184 (void) zfs_error(hdl, EZFS_BADDEV, msg);
1188 (void) zpool_standard_error(hdl, errno, msg);
1196 zcmd_free_nvlists(&zc);
1202 * Exports the pool from the system. The caller must ensure that there are no
1203 * mounted datasets in the pool.
1206 zpool_export_common(zpool_handle_t *zhp, boolean_t force, boolean_t hardforce)
1208 zfs_cmd_t zc = { "\0", "\0", "\0", "\0", 0 };
1211 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN,
1212 "cannot export '%s'"), zhp->zpool_name);
1214 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
1215 zc.zc_cookie = force;
1216 zc.zc_guid = hardforce;
1218 if (zfs_ioctl(zhp->zpool_hdl, ZFS_IOC_POOL_EXPORT, &zc) != 0) {
1221 zfs_error_aux(zhp->zpool_hdl, dgettext(TEXT_DOMAIN,
1222 "use '-f' to override the following errors:\n"
1223 "'%s' has an active shared spare which could be"
1224 " used by other pools once '%s' is exported."),
1225 zhp->zpool_name, zhp->zpool_name);
1226 return (zfs_error(zhp->zpool_hdl, EZFS_ACTIVE_SPARE,
1229 return (zpool_standard_error_fmt(zhp->zpool_hdl, errno,
1238 zpool_export(zpool_handle_t *zhp, boolean_t force)
1240 return (zpool_export_common(zhp, force, B_FALSE));
1244 zpool_export_force(zpool_handle_t *zhp)
1246 return (zpool_export_common(zhp, B_TRUE, B_TRUE));
1250 zpool_rewind_exclaim(libzfs_handle_t *hdl, const char *name, boolean_t dryrun,
1253 nvlist_t *nv = NULL;
1259 if (!hdl->libzfs_printerr || config == NULL)
1262 if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_LOAD_INFO, &nv) != 0)
1265 if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_LOAD_TIME, &rewindto) != 0)
1267 (void) nvlist_lookup_int64(nv, ZPOOL_CONFIG_REWIND_TIME, &loss);
1269 if (localtime_r((time_t *)&rewindto, &t) != NULL &&
1270 strftime(timestr, 128, "%c", &t) != 0) {
1272 (void) printf(dgettext(TEXT_DOMAIN,
1273 "Would be able to return %s "
1274 "to its state as of %s.\n"),
1277 (void) printf(dgettext(TEXT_DOMAIN,
1278 "Pool %s returned to its state as of %s.\n"),
1282 (void) printf(dgettext(TEXT_DOMAIN,
1283 "%s approximately %lld "),
1284 dryrun ? "Would discard" : "Discarded",
1285 ((longlong_t)loss + 30) / 60);
1286 (void) printf(dgettext(TEXT_DOMAIN,
1287 "minutes of transactions.\n"));
1288 } else if (loss > 0) {
1289 (void) printf(dgettext(TEXT_DOMAIN,
1290 "%s approximately %lld "),
1291 dryrun ? "Would discard" : "Discarded",
1293 (void) printf(dgettext(TEXT_DOMAIN,
1294 "seconds of transactions.\n"));
1300 zpool_explain_recover(libzfs_handle_t *hdl, const char *name, int reason,
1303 nvlist_t *nv = NULL;
1305 uint64_t edata = UINT64_MAX;
1310 if (!hdl->libzfs_printerr)
1314 (void) printf(dgettext(TEXT_DOMAIN, "action: "));
1316 (void) printf(dgettext(TEXT_DOMAIN, "\t"));
1318 /* All attempted rewinds failed if ZPOOL_CONFIG_LOAD_TIME missing */
1319 if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_LOAD_INFO, &nv) != 0 ||
1320 nvlist_lookup_uint64(nv, ZPOOL_CONFIG_LOAD_TIME, &rewindto) != 0)
1323 (void) nvlist_lookup_int64(nv, ZPOOL_CONFIG_REWIND_TIME, &loss);
1324 (void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_LOAD_DATA_ERRORS,
1327 (void) printf(dgettext(TEXT_DOMAIN,
1328 "Recovery is possible, but will result in some data loss.\n"));
1330 if (localtime_r((time_t *)&rewindto, &t) != NULL &&
1331 strftime(timestr, 128, "%c", &t) != 0) {
1332 (void) printf(dgettext(TEXT_DOMAIN,
1333 "\tReturning the pool to its state as of %s\n"
1334 "\tshould correct the problem. "),
1337 (void) printf(dgettext(TEXT_DOMAIN,
1338 "\tReverting the pool to an earlier state "
1339 "should correct the problem.\n\t"));
1343 (void) printf(dgettext(TEXT_DOMAIN,
1344 "Approximately %lld minutes of data\n"
1345 "\tmust be discarded, irreversibly. "),
1346 ((longlong_t)loss + 30) / 60);
1347 } else if (loss > 0) {
1348 (void) printf(dgettext(TEXT_DOMAIN,
1349 "Approximately %lld seconds of data\n"
1350 "\tmust be discarded, irreversibly. "),
1353 if (edata != 0 && edata != UINT64_MAX) {
1355 (void) printf(dgettext(TEXT_DOMAIN,
1356 "After rewind, at least\n"
1357 "\tone persistent user-data error will remain. "));
1359 (void) printf(dgettext(TEXT_DOMAIN,
1360 "After rewind, several\n"
1361 "\tpersistent user-data errors will remain. "));
1364 (void) printf(dgettext(TEXT_DOMAIN,
1365 "Recovery can be attempted\n\tby executing 'zpool %s -F %s'. "),
1366 reason >= 0 ? "clear" : "import", name);
1368 (void) printf(dgettext(TEXT_DOMAIN,
1369 "A scrub of the pool\n"
1370 "\tis strongly recommended after recovery.\n"));
1374 (void) printf(dgettext(TEXT_DOMAIN,
1375 "Destroy and re-create the pool from\n\ta backup source.\n"));
1379 * zpool_import() is a contracted interface. Should be kept the same
1382 * Applications should use zpool_import_props() to import a pool with
1383 * new properties value to be set.
1386 zpool_import(libzfs_handle_t *hdl, nvlist_t *config, const char *newname,
1389 nvlist_t *props = NULL;
1392 if (altroot != NULL) {
1393 if (nvlist_alloc(&props, NV_UNIQUE_NAME, 0) != 0) {
1394 return (zfs_error_fmt(hdl, EZFS_NOMEM,
1395 dgettext(TEXT_DOMAIN, "cannot import '%s'"),
1399 if (nvlist_add_string(props,
1400 zpool_prop_to_name(ZPOOL_PROP_ALTROOT), altroot) != 0 ||
1401 nvlist_add_string(props,
1402 zpool_prop_to_name(ZPOOL_PROP_CACHEFILE), "none") != 0) {
1404 return (zfs_error_fmt(hdl, EZFS_NOMEM,
1405 dgettext(TEXT_DOMAIN, "cannot import '%s'"),
1410 ret = zpool_import_props(hdl, config, newname, props,
1418 print_vdev_tree(libzfs_handle_t *hdl, const char *name, nvlist_t *nv,
1424 uint64_t is_log = 0;
1426 (void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_IS_LOG,
1430 (void) printf("\t%*s%s%s\n", indent, "", name,
1431 is_log ? " [log]" : "");
1433 if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN,
1434 &child, &children) != 0)
1437 for (c = 0; c < children; c++) {
1438 vname = zpool_vdev_name(hdl, NULL, child[c], B_TRUE);
1439 print_vdev_tree(hdl, vname, child[c], indent + 2);
1445 * Import the given pool using the known configuration and a list of
1446 * properties to be set. The configuration should have come from
1447 * zpool_find_import(). The 'newname' parameters control whether the pool
1448 * is imported with a different name.
1451 zpool_import_props(libzfs_handle_t *hdl, nvlist_t *config, const char *newname,
1452 nvlist_t *props, int flags)
1454 zfs_cmd_t zc = { "\0", "\0", "\0", "\0", 0 };
1455 zpool_rewind_policy_t policy;
1456 nvlist_t *nv = NULL;
1457 nvlist_t *nvinfo = NULL;
1458 nvlist_t *missing = NULL;
1465 verify(nvlist_lookup_string(config, ZPOOL_CONFIG_POOL_NAME,
1468 (void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN,
1469 "cannot import pool '%s'"), origname);
1471 if (newname != NULL) {
1472 if (!zpool_name_valid(hdl, B_FALSE, newname))
1473 return (zfs_error_fmt(hdl, EZFS_INVALIDNAME,
1474 dgettext(TEXT_DOMAIN, "cannot import '%s'"),
1476 thename = (char *)newname;
1483 prop_flags_t flags = { .create = B_FALSE, .import = B_TRUE };
1485 verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_VERSION,
1488 if ((props = zpool_valid_proplist(hdl, origname,
1489 props, version, flags, errbuf)) == NULL) {
1491 } else if (zcmd_write_src_nvlist(hdl, &zc, props) != 0) {
1497 (void) strlcpy(zc.zc_name, thename, sizeof (zc.zc_name));
1499 verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID,
1502 if (zcmd_write_conf_nvlist(hdl, &zc, config) != 0) {
1506 if (zcmd_alloc_dst_nvlist(hdl, &zc, zc.zc_nvlist_conf_size * 2) != 0) {
1511 zc.zc_cookie = flags;
1512 while ((ret = zfs_ioctl(hdl, ZFS_IOC_POOL_IMPORT, &zc)) != 0 &&
1514 if (zcmd_expand_dst_nvlist(hdl, &zc) != 0) {
1515 zcmd_free_nvlists(&zc);
1522 (void) zcmd_read_dst_nvlist(hdl, &zc, &nv);
1523 zpool_get_rewind_policy(config, &policy);
1529 * Dry-run failed, but we print out what success
1530 * looks like if we found a best txg
1532 if (policy.zrp_request & ZPOOL_TRY_REWIND) {
1533 zpool_rewind_exclaim(hdl, newname ? origname : thename,
1539 if (newname == NULL)
1540 (void) snprintf(desc, sizeof (desc),
1541 dgettext(TEXT_DOMAIN, "cannot import '%s'"),
1544 (void) snprintf(desc, sizeof (desc),
1545 dgettext(TEXT_DOMAIN, "cannot import '%s' as '%s'"),
1551 * Unsupported version.
1553 (void) zfs_error(hdl, EZFS_BADVERSION, desc);
1557 (void) zfs_error(hdl, EZFS_INVALCONFIG, desc);
1561 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1562 "one or more devices is read only"));
1563 (void) zfs_error(hdl, EZFS_BADDEV, desc);
1567 if (nv && nvlist_lookup_nvlist(nv,
1568 ZPOOL_CONFIG_LOAD_INFO, &nvinfo) == 0 &&
1569 nvlist_lookup_nvlist(nvinfo,
1570 ZPOOL_CONFIG_MISSING_DEVICES, &missing) == 0) {
1571 (void) printf(dgettext(TEXT_DOMAIN,
1572 "The devices below are missing, use "
1573 "'-m' to import the pool anyway:\n"));
1574 print_vdev_tree(hdl, NULL, missing, 2);
1575 (void) printf("\n");
1577 (void) zpool_standard_error(hdl, error, desc);
1581 (void) zpool_standard_error(hdl, error, desc);
1585 (void) zpool_standard_error(hdl, error, desc);
1586 zpool_explain_recover(hdl,
1587 newname ? origname : thename, -error, nv);
1594 zpool_handle_t *zhp;
1597 * This should never fail, but play it safe anyway.
1599 if (zpool_open_silent(hdl, thename, &zhp) != 0)
1601 else if (zhp != NULL)
1603 if (policy.zrp_request &
1604 (ZPOOL_DO_REWIND | ZPOOL_TRY_REWIND)) {
1605 zpool_rewind_exclaim(hdl, newname ? origname : thename,
1606 ((policy.zrp_request & ZPOOL_TRY_REWIND) != 0), nv);
1612 zcmd_free_nvlists(&zc);
1622 zpool_scan(zpool_handle_t *zhp, pool_scan_func_t func)
1624 zfs_cmd_t zc = { "\0", "\0", "\0", "\0", 0 };
1626 libzfs_handle_t *hdl = zhp->zpool_hdl;
1628 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
1629 zc.zc_cookie = func;
1631 if (zfs_ioctl(hdl, ZFS_IOC_POOL_SCAN, &zc) == 0 ||
1632 (errno == ENOENT && func != POOL_SCAN_NONE))
1635 if (func == POOL_SCAN_SCRUB) {
1636 (void) snprintf(msg, sizeof (msg),
1637 dgettext(TEXT_DOMAIN, "cannot scrub %s"), zc.zc_name);
1638 } else if (func == POOL_SCAN_NONE) {
1639 (void) snprintf(msg, sizeof (msg),
1640 dgettext(TEXT_DOMAIN, "cannot cancel scrubbing %s"),
1643 assert(!"unexpected result");
1646 if (errno == EBUSY) {
1648 pool_scan_stat_t *ps = NULL;
1651 verify(nvlist_lookup_nvlist(zhp->zpool_config,
1652 ZPOOL_CONFIG_VDEV_TREE, &nvroot) == 0);
1653 (void) nvlist_lookup_uint64_array(nvroot,
1654 ZPOOL_CONFIG_SCAN_STATS, (uint64_t **)&ps, &psc);
1655 if (ps && ps->pss_func == POOL_SCAN_SCRUB)
1656 return (zfs_error(hdl, EZFS_SCRUBBING, msg));
1658 return (zfs_error(hdl, EZFS_RESILVERING, msg));
1659 } else if (errno == ENOENT) {
1660 return (zfs_error(hdl, EZFS_NO_SCRUB, msg));
1662 return (zpool_standard_error(hdl, errno, msg));
1667 * Find a vdev that matches the search criteria specified. We use the
1668 * the nvpair name to determine how we should look for the device.
1669 * 'avail_spare' is set to TRUE if the provided guid refers to an AVAIL
1670 * spare; but FALSE if its an INUSE spare.
1673 vdev_to_nvlist_iter(nvlist_t *nv, nvlist_t *search, boolean_t *avail_spare,
1674 boolean_t *l2cache, boolean_t *log)
1681 nvpair_t *pair = nvlist_next_nvpair(search, NULL);
1683 /* Nothing to look for */
1684 if (search == NULL || pair == NULL)
1687 /* Obtain the key we will use to search */
1688 srchkey = nvpair_name(pair);
1690 switch (nvpair_type(pair)) {
1691 case DATA_TYPE_UINT64:
1692 if (strcmp(srchkey, ZPOOL_CONFIG_GUID) == 0) {
1693 uint64_t srchval, theguid;
1695 verify(nvpair_value_uint64(pair, &srchval) == 0);
1696 verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID,
1698 if (theguid == srchval)
1703 case DATA_TYPE_STRING: {
1704 char *srchval, *val;
1706 verify(nvpair_value_string(pair, &srchval) == 0);
1707 if (nvlist_lookup_string(nv, srchkey, &val) != 0)
1711 * Search for the requested value. Special cases:
1713 * - ZPOOL_CONFIG_PATH for whole disk entries. These end in with a
1714 * partition suffix "1", "-part1", or "p1". The suffix is hidden
1715 * from the user, but included in the string, so this matches around
1717 * - looking for a top-level vdev name (i.e. ZPOOL_CONFIG_TYPE).
1719 * Otherwise, all other searches are simple string compares.
1721 if (strcmp(srchkey, ZPOOL_CONFIG_PATH) == 0) {
1722 uint64_t wholedisk = 0;
1724 (void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_WHOLE_DISK,
1727 char buf[MAXPATHLEN];
1729 zfs_append_partition(srchval, buf, sizeof (buf));
1730 if (strcmp(val, buf) == 0)
1735 } else if (strcmp(srchkey, ZPOOL_CONFIG_TYPE) == 0 && val) {
1736 char *type, *idx, *end, *p;
1737 uint64_t id, vdev_id;
1740 * Determine our vdev type, keeping in mind
1741 * that the srchval is composed of a type and
1742 * vdev id pair (i.e. mirror-4).
1744 if ((type = strdup(srchval)) == NULL)
1747 if ((p = strrchr(type, '-')) == NULL) {
1755 * If the types don't match then keep looking.
1757 if (strncmp(val, type, strlen(val)) != 0) {
1762 verify(strncmp(type, VDEV_TYPE_RAIDZ,
1763 strlen(VDEV_TYPE_RAIDZ)) == 0 ||
1764 strncmp(type, VDEV_TYPE_MIRROR,
1765 strlen(VDEV_TYPE_MIRROR)) == 0);
1766 verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_ID,
1770 vdev_id = strtoull(idx, &end, 10);
1777 * Now verify that we have the correct vdev id.
1786 if (strcmp(srchval, val) == 0)
1795 if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN,
1796 &child, &children) != 0)
1799 for (c = 0; c < children; c++) {
1800 if ((ret = vdev_to_nvlist_iter(child[c], search,
1801 avail_spare, l2cache, NULL)) != NULL) {
1803 * The 'is_log' value is only set for the toplevel
1804 * vdev, not the leaf vdevs. So we always lookup the
1805 * log device from the root of the vdev tree (where
1806 * 'log' is non-NULL).
1809 nvlist_lookup_uint64(child[c],
1810 ZPOOL_CONFIG_IS_LOG, &is_log) == 0 &&
1818 if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_SPARES,
1819 &child, &children) == 0) {
1820 for (c = 0; c < children; c++) {
1821 if ((ret = vdev_to_nvlist_iter(child[c], search,
1822 avail_spare, l2cache, NULL)) != NULL) {
1823 *avail_spare = B_TRUE;
1829 if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_L2CACHE,
1830 &child, &children) == 0) {
1831 for (c = 0; c < children; c++) {
1832 if ((ret = vdev_to_nvlist_iter(child[c], search,
1833 avail_spare, l2cache, NULL)) != NULL) {
1844 * Given a physical path (minus the "/devices" prefix), find the
1848 zpool_find_vdev_by_physpath(zpool_handle_t *zhp, const char *ppath,
1849 boolean_t *avail_spare, boolean_t *l2cache, boolean_t *log)
1851 nvlist_t *search, *nvroot, *ret;
1853 verify(nvlist_alloc(&search, NV_UNIQUE_NAME, KM_SLEEP) == 0);
1854 verify(nvlist_add_string(search, ZPOOL_CONFIG_PHYS_PATH, ppath) == 0);
1856 verify(nvlist_lookup_nvlist(zhp->zpool_config, ZPOOL_CONFIG_VDEV_TREE,
1859 *avail_spare = B_FALSE;
1863 ret = vdev_to_nvlist_iter(nvroot, search, avail_spare, l2cache, log);
1864 nvlist_free(search);
1870 * Determine if we have an "interior" top-level vdev (i.e mirror/raidz).
1873 zpool_vdev_is_interior(const char *name)
1875 if (strncmp(name, VDEV_TYPE_RAIDZ, strlen(VDEV_TYPE_RAIDZ)) == 0 ||
1876 strncmp(name, VDEV_TYPE_MIRROR, strlen(VDEV_TYPE_MIRROR)) == 0)
1882 zpool_find_vdev(zpool_handle_t *zhp, const char *path, boolean_t *avail_spare,
1883 boolean_t *l2cache, boolean_t *log)
1885 char buf[MAXPATHLEN];
1887 nvlist_t *nvroot, *search, *ret;
1890 verify(nvlist_alloc(&search, NV_UNIQUE_NAME, KM_SLEEP) == 0);
1892 guid = strtoull(path, &end, 10);
1893 if (guid != 0 && *end == '\0') {
1894 verify(nvlist_add_uint64(search, ZPOOL_CONFIG_GUID, guid) == 0);
1895 } else if (zpool_vdev_is_interior(path)) {
1896 verify(nvlist_add_string(search, ZPOOL_CONFIG_TYPE, path) == 0);
1897 } else if (path[0] != '/') {
1898 if (zfs_resolve_shortname(path, buf, sizeof (buf)) < 0) {
1899 nvlist_free(search);
1902 verify(nvlist_add_string(search, ZPOOL_CONFIG_PATH, buf) == 0);
1904 verify(nvlist_add_string(search, ZPOOL_CONFIG_PATH, path) == 0);
1907 verify(nvlist_lookup_nvlist(zhp->zpool_config, ZPOOL_CONFIG_VDEV_TREE,
1910 *avail_spare = B_FALSE;
1914 ret = vdev_to_nvlist_iter(nvroot, search, avail_spare, l2cache, log);
1915 nvlist_free(search);
1921 vdev_online(nvlist_t *nv)
1925 if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_OFFLINE, &ival) == 0 ||
1926 nvlist_lookup_uint64(nv, ZPOOL_CONFIG_FAULTED, &ival) == 0 ||
1927 nvlist_lookup_uint64(nv, ZPOOL_CONFIG_REMOVED, &ival) == 0)
1934 * Helper function for zpool_get_physpaths().
1937 vdev_get_one_physpath(nvlist_t *config, char *physpath, size_t physpath_size,
1938 size_t *bytes_written)
1940 size_t bytes_left, pos, rsz;
1944 if (nvlist_lookup_string(config, ZPOOL_CONFIG_PHYS_PATH,
1946 return (EZFS_NODEVICE);
1948 pos = *bytes_written;
1949 bytes_left = physpath_size - pos;
1950 format = (pos == 0) ? "%s" : " %s";
1952 rsz = snprintf(physpath + pos, bytes_left, format, tmppath);
1953 *bytes_written += rsz;
1955 if (rsz >= bytes_left) {
1956 /* if physpath was not copied properly, clear it */
1957 if (bytes_left != 0) {
1960 return (EZFS_NOSPC);
1966 vdev_get_physpaths(nvlist_t *nv, char *physpath, size_t phypath_size,
1967 size_t *rsz, boolean_t is_spare)
1972 if (nvlist_lookup_string(nv, ZPOOL_CONFIG_TYPE, &type) != 0)
1973 return (EZFS_INVALCONFIG);
1975 if (strcmp(type, VDEV_TYPE_DISK) == 0) {
1977 * An active spare device has ZPOOL_CONFIG_IS_SPARE set.
1978 * For a spare vdev, we only want to boot from the active
1983 (void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_IS_SPARE,
1986 return (EZFS_INVALCONFIG);
1989 if (vdev_online(nv)) {
1990 if ((ret = vdev_get_one_physpath(nv, physpath,
1991 phypath_size, rsz)) != 0)
1994 } else if (strcmp(type, VDEV_TYPE_MIRROR) == 0 ||
1995 strcmp(type, VDEV_TYPE_REPLACING) == 0 ||
1996 (is_spare = (strcmp(type, VDEV_TYPE_SPARE) == 0))) {
2001 if (nvlist_lookup_nvlist_array(nv,
2002 ZPOOL_CONFIG_CHILDREN, &child, &count) != 0)
2003 return (EZFS_INVALCONFIG);
2005 for (i = 0; i < count; i++) {
2006 ret = vdev_get_physpaths(child[i], physpath,
2007 phypath_size, rsz, is_spare);
2008 if (ret == EZFS_NOSPC)
2013 return (EZFS_POOL_INVALARG);
2017 * Get phys_path for a root pool config.
2018 * Return 0 on success; non-zero on failure.
2021 zpool_get_config_physpath(nvlist_t *config, char *physpath, size_t phypath_size)
2024 nvlist_t *vdev_root;
2031 if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE,
2033 return (EZFS_INVALCONFIG);
2035 if (nvlist_lookup_string(vdev_root, ZPOOL_CONFIG_TYPE, &type) != 0 ||
2036 nvlist_lookup_nvlist_array(vdev_root, ZPOOL_CONFIG_CHILDREN,
2037 &child, &count) != 0)
2038 return (EZFS_INVALCONFIG);
2041 * root pool can not have EFI labeled disks and can only have
2042 * a single top-level vdev.
2044 if (strcmp(type, VDEV_TYPE_ROOT) != 0 || count != 1 ||
2045 pool_uses_efi(vdev_root))
2046 return (EZFS_POOL_INVALARG);
2048 (void) vdev_get_physpaths(child[0], physpath, phypath_size, &rsz,
2051 /* No online devices */
2053 return (EZFS_NODEVICE);
2059 * Get phys_path for a root pool
2060 * Return 0 on success; non-zero on failure.
2063 zpool_get_physpath(zpool_handle_t *zhp, char *physpath, size_t phypath_size)
2065 return (zpool_get_config_physpath(zhp->zpool_config, physpath,
2070 * If the device has being dynamically expanded then we need to relabel
2071 * the disk to use the new unallocated space.
2074 zpool_relabel_disk(libzfs_handle_t *hdl, const char *path)
2079 if ((fd = open(path, O_RDWR|O_DIRECT)) < 0) {
2080 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "cannot "
2081 "relabel '%s': unable to open device"), path);
2082 return (zfs_error(hdl, EZFS_OPENFAILED, errbuf));
2086 * It's possible that we might encounter an error if the device
2087 * does not have any unallocated space left. If so, we simply
2088 * ignore that error and continue on.
2090 error = efi_use_whole_disk(fd);
2092 if (error && error != VT_ENOSPC) {
2093 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "cannot "
2094 "relabel '%s': unable to read disk capacity"), path);
2095 return (zfs_error(hdl, EZFS_NOCAP, errbuf));
2101 * Bring the specified vdev online. The 'flags' parameter is a set of the
2102 * ZFS_ONLINE_* flags.
2105 zpool_vdev_online(zpool_handle_t *zhp, const char *path, int flags,
2106 vdev_state_t *newstate)
2108 zfs_cmd_t zc = { "\0", "\0", "\0", "\0", 0 };
2111 boolean_t avail_spare, l2cache, islog;
2112 libzfs_handle_t *hdl = zhp->zpool_hdl;
2114 if (flags & ZFS_ONLINE_EXPAND) {
2115 (void) snprintf(msg, sizeof (msg),
2116 dgettext(TEXT_DOMAIN, "cannot expand %s"), path);
2118 (void) snprintf(msg, sizeof (msg),
2119 dgettext(TEXT_DOMAIN, "cannot online %s"), path);
2122 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
2123 if ((tgt = zpool_find_vdev(zhp, path, &avail_spare, &l2cache,
2125 return (zfs_error(hdl, EZFS_NODEVICE, msg));
2127 verify(nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID, &zc.zc_guid) == 0);
2130 return (zfs_error(hdl, EZFS_ISSPARE, msg));
2132 if (flags & ZFS_ONLINE_EXPAND ||
2133 zpool_get_prop_int(zhp, ZPOOL_PROP_AUTOEXPAND, NULL)) {
2134 char *pathname = NULL;
2135 uint64_t wholedisk = 0;
2137 (void) nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_WHOLE_DISK,
2139 verify(nvlist_lookup_string(tgt, ZPOOL_CONFIG_PATH,
2143 * XXX - L2ARC 1.0 devices can't support expansion.
2146 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2147 "cannot expand cache devices"));
2148 return (zfs_error(hdl, EZFS_VDEVNOTSUP, msg));
2152 pathname += strlen(DISK_ROOT) + 1;
2153 (void) zpool_relabel_disk(hdl, pathname);
2157 zc.zc_cookie = VDEV_STATE_ONLINE;
2160 if (zfs_ioctl(hdl, ZFS_IOC_VDEV_SET_STATE, &zc) != 0) {
2161 if (errno == EINVAL) {
2162 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "was split "
2163 "from this pool into a new one. Use '%s' "
2164 "instead"), "zpool detach");
2165 return (zfs_error(hdl, EZFS_POSTSPLIT_ONLINE, msg));
2167 return (zpool_standard_error(hdl, errno, msg));
2170 *newstate = zc.zc_cookie;
2175 * Take the specified vdev offline
2178 zpool_vdev_offline(zpool_handle_t *zhp, const char *path, boolean_t istmp)
2180 zfs_cmd_t zc = { "\0", "\0", "\0", "\0", 0 };
2183 boolean_t avail_spare, l2cache;
2184 libzfs_handle_t *hdl = zhp->zpool_hdl;
2186 (void) snprintf(msg, sizeof (msg),
2187 dgettext(TEXT_DOMAIN, "cannot offline %s"), path);
2189 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
2190 if ((tgt = zpool_find_vdev(zhp, path, &avail_spare, &l2cache,
2192 return (zfs_error(hdl, EZFS_NODEVICE, msg));
2194 verify(nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID, &zc.zc_guid) == 0);
2197 return (zfs_error(hdl, EZFS_ISSPARE, msg));
2199 zc.zc_cookie = VDEV_STATE_OFFLINE;
2200 zc.zc_obj = istmp ? ZFS_OFFLINE_TEMPORARY : 0;
2202 if (zfs_ioctl(hdl, ZFS_IOC_VDEV_SET_STATE, &zc) == 0)
2209 * There are no other replicas of this device.
2211 return (zfs_error(hdl, EZFS_NOREPLICAS, msg));
2215 * The log device has unplayed logs
2217 return (zfs_error(hdl, EZFS_UNPLAYED_LOGS, msg));
2220 return (zpool_standard_error(hdl, errno, msg));
2225 * Mark the given vdev faulted.
2228 zpool_vdev_fault(zpool_handle_t *zhp, uint64_t guid, vdev_aux_t aux)
2230 zfs_cmd_t zc = { "\0", "\0", "\0", "\0", 0 };
2232 libzfs_handle_t *hdl = zhp->zpool_hdl;
2234 (void) snprintf(msg, sizeof (msg),
2235 dgettext(TEXT_DOMAIN, "cannot fault %llu"), (u_longlong_t)guid);
2237 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
2239 zc.zc_cookie = VDEV_STATE_FAULTED;
2242 if (ioctl(hdl->libzfs_fd, ZFS_IOC_VDEV_SET_STATE, &zc) == 0)
2249 * There are no other replicas of this device.
2251 return (zfs_error(hdl, EZFS_NOREPLICAS, msg));
2254 return (zpool_standard_error(hdl, errno, msg));
2260 * Mark the given vdev degraded.
2263 zpool_vdev_degrade(zpool_handle_t *zhp, uint64_t guid, vdev_aux_t aux)
2265 zfs_cmd_t zc = { "\0", "\0", "\0", "\0", 0 };
2267 libzfs_handle_t *hdl = zhp->zpool_hdl;
2269 (void) snprintf(msg, sizeof (msg),
2270 dgettext(TEXT_DOMAIN, "cannot degrade %llu"), (u_longlong_t)guid);
2272 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
2274 zc.zc_cookie = VDEV_STATE_DEGRADED;
2277 if (ioctl(hdl->libzfs_fd, ZFS_IOC_VDEV_SET_STATE, &zc) == 0)
2280 return (zpool_standard_error(hdl, errno, msg));
2284 * Returns TRUE if the given nvlist is a vdev that was originally swapped in as
2288 is_replacing_spare(nvlist_t *search, nvlist_t *tgt, int which)
2294 if (nvlist_lookup_nvlist_array(search, ZPOOL_CONFIG_CHILDREN, &child,
2296 verify(nvlist_lookup_string(search, ZPOOL_CONFIG_TYPE,
2299 if (strcmp(type, VDEV_TYPE_SPARE) == 0 &&
2300 children == 2 && child[which] == tgt)
2303 for (c = 0; c < children; c++)
2304 if (is_replacing_spare(child[c], tgt, which))
2312 * Attach new_disk (fully described by nvroot) to old_disk.
2313 * If 'replacing' is specified, the new disk will replace the old one.
2316 zpool_vdev_attach(zpool_handle_t *zhp,
2317 const char *old_disk, const char *new_disk, nvlist_t *nvroot, int replacing)
2319 zfs_cmd_t zc = { "\0", "\0", "\0", "\0", 0 };
2323 boolean_t avail_spare, l2cache, islog;
2328 nvlist_t *config_root;
2329 libzfs_handle_t *hdl = zhp->zpool_hdl;
2330 boolean_t rootpool = pool_is_bootable(zhp);
2333 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN,
2334 "cannot replace %s with %s"), old_disk, new_disk);
2336 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN,
2337 "cannot attach %s to %s"), new_disk, old_disk);
2340 * If this is a root pool, make sure that we're not attaching an
2341 * EFI labeled device.
2343 if (rootpool && pool_uses_efi(nvroot)) {
2344 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2345 "EFI labeled devices are not supported on root pools."));
2346 return (zfs_error(hdl, EZFS_POOL_NOTSUP, msg));
2349 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
2350 if ((tgt = zpool_find_vdev(zhp, old_disk, &avail_spare, &l2cache,
2352 return (zfs_error(hdl, EZFS_NODEVICE, msg));
2355 return (zfs_error(hdl, EZFS_ISSPARE, msg));
2358 return (zfs_error(hdl, EZFS_ISL2CACHE, msg));
2360 verify(nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID, &zc.zc_guid) == 0);
2361 zc.zc_cookie = replacing;
2363 if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_CHILDREN,
2364 &child, &children) != 0 || children != 1) {
2365 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2366 "new device must be a single disk"));
2367 return (zfs_error(hdl, EZFS_INVALCONFIG, msg));
2370 verify(nvlist_lookup_nvlist(zpool_get_config(zhp, NULL),
2371 ZPOOL_CONFIG_VDEV_TREE, &config_root) == 0);
2373 if ((newname = zpool_vdev_name(NULL, NULL, child[0], B_FALSE)) == NULL)
2377 * If the target is a hot spare that has been swapped in, we can only
2378 * replace it with another hot spare.
2381 nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_IS_SPARE, &val) == 0 &&
2382 (zpool_find_vdev(zhp, newname, &avail_spare, &l2cache,
2383 NULL) == NULL || !avail_spare) &&
2384 is_replacing_spare(config_root, tgt, 1)) {
2385 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2386 "can only be replaced by another hot spare"));
2388 return (zfs_error(hdl, EZFS_BADTARGET, msg));
2393 if (zcmd_write_conf_nvlist(hdl, &zc, nvroot) != 0)
2396 ret = zfs_ioctl(hdl, ZFS_IOC_VDEV_ATTACH, &zc);
2398 zcmd_free_nvlists(&zc);
2403 * XXX need a better way to prevent user from
2404 * booting up a half-baked vdev.
2406 (void) fprintf(stderr, dgettext(TEXT_DOMAIN, "Make "
2407 "sure to wait until resilver is done "
2408 "before rebooting.\n"));
2416 * Can't attach to or replace this type of vdev.
2419 uint64_t version = zpool_get_prop_int(zhp,
2420 ZPOOL_PROP_VERSION, NULL);
2423 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2424 "cannot replace a log with a spare"));
2425 else if (version >= SPA_VERSION_MULTI_REPLACE)
2426 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2427 "already in replacing/spare config; wait "
2428 "for completion or use 'zpool detach'"));
2430 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2431 "cannot replace a replacing device"));
2433 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2434 "can only attach to mirrors and top-level "
2437 (void) zfs_error(hdl, EZFS_BADTARGET, msg);
2442 * The new device must be a single disk.
2444 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2445 "new device must be a single disk"));
2446 (void) zfs_error(hdl, EZFS_INVALCONFIG, msg);
2450 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "%s is busy"),
2452 (void) zfs_error(hdl, EZFS_BADDEV, msg);
2457 * The new device is too small.
2459 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2460 "device is too small"));
2461 (void) zfs_error(hdl, EZFS_BADDEV, msg);
2466 * The new device has a different alignment requirement.
2468 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2469 "devices have different sector alignment"));
2470 (void) zfs_error(hdl, EZFS_BADDEV, msg);
2475 * The resulting top-level vdev spec won't fit in the label.
2477 (void) zfs_error(hdl, EZFS_DEVOVERFLOW, msg);
2481 (void) zpool_standard_error(hdl, errno, msg);
2488 * Detach the specified device.
2491 zpool_vdev_detach(zpool_handle_t *zhp, const char *path)
2493 zfs_cmd_t zc = { "\0", "\0", "\0", "\0", 0 };
2496 boolean_t avail_spare, l2cache;
2497 libzfs_handle_t *hdl = zhp->zpool_hdl;
2499 (void) snprintf(msg, sizeof (msg),
2500 dgettext(TEXT_DOMAIN, "cannot detach %s"), path);
2502 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
2503 if ((tgt = zpool_find_vdev(zhp, path, &avail_spare, &l2cache,
2505 return (zfs_error(hdl, EZFS_NODEVICE, msg));
2508 return (zfs_error(hdl, EZFS_ISSPARE, msg));
2511 return (zfs_error(hdl, EZFS_ISL2CACHE, msg));
2513 verify(nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID, &zc.zc_guid) == 0);
2515 if (zfs_ioctl(hdl, ZFS_IOC_VDEV_DETACH, &zc) == 0)
2522 * Can't detach from this type of vdev.
2524 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "only "
2525 "applicable to mirror and replacing vdevs"));
2526 (void) zfs_error(hdl, EZFS_BADTARGET, msg);
2531 * There are no other replicas of this device.
2533 (void) zfs_error(hdl, EZFS_NOREPLICAS, msg);
2537 (void) zpool_standard_error(hdl, errno, msg);
2544 * Find a mirror vdev in the source nvlist.
2546 * The mchild array contains a list of disks in one of the top-level mirrors
2547 * of the source pool. The schild array contains a list of disks that the
2548 * user specified on the command line. We loop over the mchild array to
2549 * see if any entry in the schild array matches.
2551 * If a disk in the mchild array is found in the schild array, we return
2552 * the index of that entry. Otherwise we return -1.
2555 find_vdev_entry(zpool_handle_t *zhp, nvlist_t **mchild, uint_t mchildren,
2556 nvlist_t **schild, uint_t schildren)
2560 for (mc = 0; mc < mchildren; mc++) {
2562 char *mpath = zpool_vdev_name(zhp->zpool_hdl, zhp,
2563 mchild[mc], B_FALSE);
2565 for (sc = 0; sc < schildren; sc++) {
2566 char *spath = zpool_vdev_name(zhp->zpool_hdl, zhp,
2567 schild[sc], B_FALSE);
2568 boolean_t result = (strcmp(mpath, spath) == 0);
2584 * Split a mirror pool. If newroot points to null, then a new nvlist
2585 * is generated and it is the responsibility of the caller to free it.
2588 zpool_vdev_split(zpool_handle_t *zhp, char *newname, nvlist_t **newroot,
2589 nvlist_t *props, splitflags_t flags)
2591 zfs_cmd_t zc = { "\0", "\0", "\0", "\0", 0 };
2593 nvlist_t *tree, *config, **child, **newchild, *newconfig = NULL;
2594 nvlist_t **varray = NULL, *zc_props = NULL;
2595 uint_t c, children, newchildren, lastlog = 0, vcount, found = 0;
2596 libzfs_handle_t *hdl = zhp->zpool_hdl;
2598 boolean_t freelist = B_FALSE, memory_err = B_TRUE;
2601 (void) snprintf(msg, sizeof (msg),
2602 dgettext(TEXT_DOMAIN, "Unable to split %s"), zhp->zpool_name);
2604 if (!zpool_name_valid(hdl, B_FALSE, newname))
2605 return (zfs_error(hdl, EZFS_INVALIDNAME, msg));
2607 if ((config = zpool_get_config(zhp, NULL)) == NULL) {
2608 (void) fprintf(stderr, gettext("Internal error: unable to "
2609 "retrieve pool configuration\n"));
2613 verify(nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE, &tree)
2615 verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_VERSION, &vers) == 0);
2618 prop_flags_t flags = { .create = B_FALSE, .import = B_TRUE };
2619 if ((zc_props = zpool_valid_proplist(hdl, zhp->zpool_name,
2620 props, vers, flags, msg)) == NULL)
2624 if (nvlist_lookup_nvlist_array(tree, ZPOOL_CONFIG_CHILDREN, &child,
2626 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2627 "Source pool is missing vdev tree"));
2629 nvlist_free(zc_props);
2633 varray = zfs_alloc(hdl, children * sizeof (nvlist_t *));
2636 if (*newroot == NULL ||
2637 nvlist_lookup_nvlist_array(*newroot, ZPOOL_CONFIG_CHILDREN,
2638 &newchild, &newchildren) != 0)
2641 for (c = 0; c < children; c++) {
2642 uint64_t is_log = B_FALSE, is_hole = B_FALSE;
2644 nvlist_t **mchild, *vdev;
2649 * Unlike cache & spares, slogs are stored in the
2650 * ZPOOL_CONFIG_CHILDREN array. We filter them out here.
2652 (void) nvlist_lookup_uint64(child[c], ZPOOL_CONFIG_IS_LOG,
2654 (void) nvlist_lookup_uint64(child[c], ZPOOL_CONFIG_IS_HOLE,
2656 if (is_log || is_hole) {
2658 * Create a hole vdev and put it in the config.
2660 if (nvlist_alloc(&vdev, NV_UNIQUE_NAME, 0) != 0)
2662 if (nvlist_add_string(vdev, ZPOOL_CONFIG_TYPE,
2663 VDEV_TYPE_HOLE) != 0)
2665 if (nvlist_add_uint64(vdev, ZPOOL_CONFIG_IS_HOLE,
2670 varray[vcount++] = vdev;
2674 verify(nvlist_lookup_string(child[c], ZPOOL_CONFIG_TYPE, &type)
2676 if (strcmp(type, VDEV_TYPE_MIRROR) != 0) {
2677 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2678 "Source pool must be composed only of mirrors\n"));
2679 retval = zfs_error(hdl, EZFS_INVALCONFIG, msg);
2683 verify(nvlist_lookup_nvlist_array(child[c],
2684 ZPOOL_CONFIG_CHILDREN, &mchild, &mchildren) == 0);
2686 /* find or add an entry for this top-level vdev */
2687 if (newchildren > 0 &&
2688 (entry = find_vdev_entry(zhp, mchild, mchildren,
2689 newchild, newchildren)) >= 0) {
2690 /* We found a disk that the user specified. */
2691 vdev = mchild[entry];
2694 /* User didn't specify a disk for this vdev. */
2695 vdev = mchild[mchildren - 1];
2698 if (nvlist_dup(vdev, &varray[vcount++], 0) != 0)
2702 /* did we find every disk the user specified? */
2703 if (found != newchildren) {
2704 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "Device list must "
2705 "include at most one disk from each mirror"));
2706 retval = zfs_error(hdl, EZFS_INVALCONFIG, msg);
2710 /* Prepare the nvlist for populating. */
2711 if (*newroot == NULL) {
2712 if (nvlist_alloc(newroot, NV_UNIQUE_NAME, 0) != 0)
2715 if (nvlist_add_string(*newroot, ZPOOL_CONFIG_TYPE,
2716 VDEV_TYPE_ROOT) != 0)
2719 verify(nvlist_remove_all(*newroot, ZPOOL_CONFIG_CHILDREN) == 0);
2722 /* Add all the children we found */
2723 if (nvlist_add_nvlist_array(*newroot, ZPOOL_CONFIG_CHILDREN, varray,
2724 lastlog == 0 ? vcount : lastlog) != 0)
2728 * If we're just doing a dry run, exit now with success.
2731 memory_err = B_FALSE;
2736 /* now build up the config list & call the ioctl */
2737 if (nvlist_alloc(&newconfig, NV_UNIQUE_NAME, 0) != 0)
2740 if (nvlist_add_nvlist(newconfig,
2741 ZPOOL_CONFIG_VDEV_TREE, *newroot) != 0 ||
2742 nvlist_add_string(newconfig,
2743 ZPOOL_CONFIG_POOL_NAME, newname) != 0 ||
2744 nvlist_add_uint64(newconfig, ZPOOL_CONFIG_VERSION, vers) != 0)
2748 * The new pool is automatically part of the namespace unless we
2749 * explicitly export it.
2752 zc.zc_cookie = ZPOOL_EXPORT_AFTER_SPLIT;
2753 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
2754 (void) strlcpy(zc.zc_string, newname, sizeof (zc.zc_string));
2755 if (zcmd_write_conf_nvlist(hdl, &zc, newconfig) != 0)
2757 if (zc_props != NULL && zcmd_write_src_nvlist(hdl, &zc, zc_props) != 0)
2760 if (zfs_ioctl(hdl, ZFS_IOC_VDEV_SPLIT, &zc) != 0) {
2761 retval = zpool_standard_error(hdl, errno, msg);
2766 memory_err = B_FALSE;
2769 if (varray != NULL) {
2772 for (v = 0; v < vcount; v++)
2773 nvlist_free(varray[v]);
2776 zcmd_free_nvlists(&zc);
2778 nvlist_free(zc_props);
2780 nvlist_free(newconfig);
2782 nvlist_free(*newroot);
2790 return (no_memory(hdl));
2796 * Remove the given device. Currently, this is supported only for hot spares
2797 * and level 2 cache devices.
2800 zpool_vdev_remove(zpool_handle_t *zhp, const char *path)
2802 zfs_cmd_t zc = { "\0", "\0", "\0", "\0", 0 };
2805 boolean_t avail_spare, l2cache, islog;
2806 libzfs_handle_t *hdl = zhp->zpool_hdl;
2809 (void) snprintf(msg, sizeof (msg),
2810 dgettext(TEXT_DOMAIN, "cannot remove %s"), path);
2812 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
2813 if ((tgt = zpool_find_vdev(zhp, path, &avail_spare, &l2cache,
2815 return (zfs_error(hdl, EZFS_NODEVICE, msg));
2817 * XXX - this should just go away.
2819 if (!avail_spare && !l2cache && !islog) {
2820 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2821 "only inactive hot spares, cache, top-level, "
2822 "or log devices can be removed"));
2823 return (zfs_error(hdl, EZFS_NODEVICE, msg));
2826 version = zpool_get_prop_int(zhp, ZPOOL_PROP_VERSION, NULL);
2827 if (islog && version < SPA_VERSION_HOLES) {
2828 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2829 "pool must be upgrade to support log removal"));
2830 return (zfs_error(hdl, EZFS_BADVERSION, msg));
2833 verify(nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID, &zc.zc_guid) == 0);
2835 if (zfs_ioctl(hdl, ZFS_IOC_VDEV_REMOVE, &zc) == 0)
2838 return (zpool_standard_error(hdl, errno, msg));
2842 * Clear the errors for the pool, or the particular device if specified.
2845 zpool_clear(zpool_handle_t *zhp, const char *path, nvlist_t *rewindnvl)
2847 zfs_cmd_t zc = { "\0", "\0", "\0", "\0", 0 };
2850 zpool_rewind_policy_t policy;
2851 boolean_t avail_spare, l2cache;
2852 libzfs_handle_t *hdl = zhp->zpool_hdl;
2853 nvlist_t *nvi = NULL;
2857 (void) snprintf(msg, sizeof (msg),
2858 dgettext(TEXT_DOMAIN, "cannot clear errors for %s"),
2861 (void) snprintf(msg, sizeof (msg),
2862 dgettext(TEXT_DOMAIN, "cannot clear errors for %s"),
2865 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
2867 if ((tgt = zpool_find_vdev(zhp, path, &avail_spare,
2868 &l2cache, NULL)) == 0)
2869 return (zfs_error(hdl, EZFS_NODEVICE, msg));
2872 * Don't allow error clearing for hot spares. Do allow
2873 * error clearing for l2cache devices.
2876 return (zfs_error(hdl, EZFS_ISSPARE, msg));
2878 verify(nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID,
2882 zpool_get_rewind_policy(rewindnvl, &policy);
2883 zc.zc_cookie = policy.zrp_request;
2885 if (zcmd_alloc_dst_nvlist(hdl, &zc, zhp->zpool_config_size * 2) != 0)
2888 if (zcmd_write_src_nvlist(hdl, &zc, rewindnvl) != 0)
2891 while ((error = zfs_ioctl(hdl, ZFS_IOC_CLEAR, &zc)) != 0 &&
2893 if (zcmd_expand_dst_nvlist(hdl, &zc) != 0) {
2894 zcmd_free_nvlists(&zc);
2899 if (!error || ((policy.zrp_request & ZPOOL_TRY_REWIND) &&
2900 errno != EPERM && errno != EACCES)) {
2901 if (policy.zrp_request &
2902 (ZPOOL_DO_REWIND | ZPOOL_TRY_REWIND)) {
2903 (void) zcmd_read_dst_nvlist(hdl, &zc, &nvi);
2904 zpool_rewind_exclaim(hdl, zc.zc_name,
2905 ((policy.zrp_request & ZPOOL_TRY_REWIND) != 0),
2909 zcmd_free_nvlists(&zc);
2913 zcmd_free_nvlists(&zc);
2914 return (zpool_standard_error(hdl, errno, msg));
2918 * Similar to zpool_clear(), but takes a GUID (used by fmd).
2921 zpool_vdev_clear(zpool_handle_t *zhp, uint64_t guid)
2923 zfs_cmd_t zc = { "\0", "\0", "\0", "\0", 0 };
2925 libzfs_handle_t *hdl = zhp->zpool_hdl;
2927 (void) snprintf(msg, sizeof (msg),
2928 dgettext(TEXT_DOMAIN, "cannot clear errors for %llx"),
2929 (u_longlong_t)guid);
2931 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
2933 zc.zc_cookie = ZPOOL_NO_REWIND;
2935 if (ioctl(hdl->libzfs_fd, ZFS_IOC_CLEAR, &zc) == 0)
2938 return (zpool_standard_error(hdl, errno, msg));
2942 * Convert from a devid string to a path.
2945 devid_to_path(char *devid_str)
2950 devid_nmlist_t *list = NULL;
2953 if (devid_str_decode(devid_str, &devid, &minor) != 0)
2956 ret = devid_deviceid_to_nmlist("/dev", devid, minor, &list);
2958 devid_str_free(minor);
2964 if ((path = strdup(list[0].devname)) == NULL)
2967 devid_free_nmlist(list);
2973 * Convert from a path to a devid string.
2976 path_to_devid(const char *path)
2982 if ((fd = open(path, O_RDONLY)) < 0)
2987 if (devid_get(fd, &devid) == 0) {
2988 if (devid_get_minor_name(fd, &minor) == 0)
2989 ret = devid_str_encode(devid, minor);
2991 devid_str_free(minor);
3000 * Issue the necessary ioctl() to update the stored path value for the vdev. We
3001 * ignore any failure here, since a common case is for an unprivileged user to
3002 * type 'zpool status', and we'll display the correct information anyway.
3005 set_path(zpool_handle_t *zhp, nvlist_t *nv, const char *path)
3007 zfs_cmd_t zc = { "\0", "\0", "\0", "\0", 0 };
3009 (void) strncpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
3010 (void) strncpy(zc.zc_value, path, sizeof (zc.zc_value));
3011 verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID,
3014 (void) ioctl(zhp->zpool_hdl->libzfs_fd, ZFS_IOC_VDEV_SETPATH, &zc);
3018 * Remove partition suffix from a vdev path. Partition suffixes may take three
3019 * forms: "-partX", "pX", or "X", where X is a string of digits. The second
3020 * case only occurs when the suffix is preceded by a digit, i.e. "md0p0" The
3021 * third case only occurs when preceded by a string matching the regular
3022 * expression "^[hs]d[a-z]+", i.e. a scsi or ide disk.
3025 strip_partition(libzfs_handle_t *hdl, char *path)
3027 char *tmp = zfs_strdup(hdl, path);
3028 char *part = NULL, *d = NULL;
3030 if ((part = strstr(tmp, "-part")) && part != tmp) {
3032 } else if ((part = strrchr(tmp, 'p')) &&
3033 part > tmp + 1 && isdigit(*(part-1))) {
3035 } else if ((tmp[0] == 'h' || tmp[0] == 's') && tmp[1] == 'd') {
3036 for (d = &tmp[2]; isalpha(*d); part = ++d);
3038 if (part && d && *d != '\0') {
3039 for (; isdigit(*d); d++);
3046 #define PATH_BUF_LEN 64
3049 * Given a vdev, return the name to display in iostat. If the vdev has a path,
3050 * we use that, stripping off any leading "/dev/dsk/"; if not, we use the type.
3051 * We also check if this is a whole disk, in which case we strip off the
3052 * trailing 's0' slice name.
3054 * This routine is also responsible for identifying when disks have been
3055 * reconfigured in a new location. The kernel will have opened the device by
3056 * devid, but the path will still refer to the old location. To catch this, we
3057 * first do a path -> devid translation (which is fast for the common case). If
3058 * the devid matches, we're done. If not, we do a reverse devid -> path
3059 * translation and issue the appropriate ioctl() to update the path of the vdev.
3060 * If 'zhp' is NULL, then this is an exported pool, and we don't need to do any
3064 zpool_vdev_name(libzfs_handle_t *hdl, zpool_handle_t *zhp, nvlist_t *nv,
3067 char *path, *devid, *type;
3069 char buf[PATH_BUF_LEN];
3073 if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_NOT_PRESENT,
3075 verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID,
3077 (void) snprintf(buf, sizeof (buf), "%llu",
3078 (u_longlong_t)value);
3080 } else if (nvlist_lookup_string(nv, ZPOOL_CONFIG_PATH, &path) == 0) {
3082 * If the device is dead (faulted, offline, etc) then don't
3083 * bother opening it. Otherwise we may be forcing the user to
3084 * open a misbehaving device, which can have undesirable
3087 if ((nvlist_lookup_uint64_array(nv, ZPOOL_CONFIG_VDEV_STATS,
3088 (uint64_t **)&vs, &vsc) != 0 ||
3089 vs->vs_state >= VDEV_STATE_DEGRADED) &&
3091 nvlist_lookup_string(nv, ZPOOL_CONFIG_DEVID, &devid) == 0) {
3093 * Determine if the current path is correct.
3095 char *newdevid = path_to_devid(path);
3097 if (newdevid == NULL ||
3098 strcmp(devid, newdevid) != 0) {
3101 if ((newpath = devid_to_path(devid)) != NULL) {
3103 * Update the path appropriately.
3105 set_path(zhp, nv, newpath);
3106 if (nvlist_add_string(nv,
3107 ZPOOL_CONFIG_PATH, newpath) == 0)
3108 verify(nvlist_lookup_string(nv,
3116 devid_str_free(newdevid);
3120 * For a block device only use the name.
3122 verify(nvlist_lookup_string(nv, ZPOOL_CONFIG_TYPE, &type) == 0);
3123 if (strcmp(type, VDEV_TYPE_DISK) == 0) {
3124 path = strrchr(path, '/');
3129 * Remove the partition from the path it this is a whole disk.
3131 if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_WHOLE_DISK,
3132 &value) == 0 && value) {
3133 return strip_partition(hdl, path);
3136 verify(nvlist_lookup_string(nv, ZPOOL_CONFIG_TYPE, &path) == 0);
3139 * If it's a raidz device, we need to stick in the parity level.
3141 if (strcmp(path, VDEV_TYPE_RAIDZ) == 0) {
3142 char tmpbuf[PATH_BUF_LEN];
3144 verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_NPARITY,
3146 (void) snprintf(tmpbuf, sizeof (tmpbuf), "%s%llu", path,
3147 (u_longlong_t)value);
3152 * We identify each top-level vdev by using a <type-id>
3153 * naming convention.
3158 verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_ID,
3160 (void) snprintf(buf, sizeof (buf), "%s-%llu", path,
3166 return (zfs_strdup(hdl, path));
3170 zbookmark_compare(const void *a, const void *b)
3172 return (memcmp(a, b, sizeof (zbookmark_t)));
3176 * Retrieve the persistent error log, uniquify the members, and return to the
3180 zpool_get_errlog(zpool_handle_t *zhp, nvlist_t **nverrlistp)
3182 zfs_cmd_t zc = { "\0", "\0", "\0", "\0", 0 };
3184 zbookmark_t *zb = NULL;
3188 * Retrieve the raw error list from the kernel. If the number of errors
3189 * has increased, allocate more space and continue until we get the
3192 verify(nvlist_lookup_uint64(zhp->zpool_config, ZPOOL_CONFIG_ERRCOUNT,
3196 if ((zc.zc_nvlist_dst = (uintptr_t)zfs_alloc(zhp->zpool_hdl,
3197 count * sizeof (zbookmark_t))) == (uintptr_t)NULL)
3199 zc.zc_nvlist_dst_size = count;
3200 (void) strcpy(zc.zc_name, zhp->zpool_name);
3202 if (ioctl(zhp->zpool_hdl->libzfs_fd, ZFS_IOC_ERROR_LOG,
3204 free((void *)(uintptr_t)zc.zc_nvlist_dst);
3205 if (errno == ENOMEM) {
3206 count = zc.zc_nvlist_dst_size;
3207 if ((zc.zc_nvlist_dst = (uintptr_t)
3208 zfs_alloc(zhp->zpool_hdl, count *
3209 sizeof (zbookmark_t))) == (uintptr_t)NULL)
3220 * Sort the resulting bookmarks. This is a little confusing due to the
3221 * implementation of ZFS_IOC_ERROR_LOG. The bookmarks are copied last
3222 * to first, and 'zc_nvlist_dst_size' indicates the number of boomarks
3223 * _not_ copied as part of the process. So we point the start of our
3224 * array appropriate and decrement the total number of elements.
3226 zb = ((zbookmark_t *)(uintptr_t)zc.zc_nvlist_dst) +
3227 zc.zc_nvlist_dst_size;
3228 count -= zc.zc_nvlist_dst_size;
3230 qsort(zb, count, sizeof (zbookmark_t), zbookmark_compare);
3232 verify(nvlist_alloc(nverrlistp, 0, KM_SLEEP) == 0);
3235 * Fill in the nverrlistp with nvlist's of dataset and object numbers.
3237 for (i = 0; i < count; i++) {
3240 /* ignoring zb_blkid and zb_level for now */
3241 if (i > 0 && zb[i-1].zb_objset == zb[i].zb_objset &&
3242 zb[i-1].zb_object == zb[i].zb_object)
3245 if (nvlist_alloc(&nv, NV_UNIQUE_NAME, KM_SLEEP) != 0)
3247 if (nvlist_add_uint64(nv, ZPOOL_ERR_DATASET,
3248 zb[i].zb_objset) != 0) {
3252 if (nvlist_add_uint64(nv, ZPOOL_ERR_OBJECT,
3253 zb[i].zb_object) != 0) {
3257 if (nvlist_add_nvlist(*nverrlistp, "ejk", nv) != 0) {
3264 free((void *)(uintptr_t)zc.zc_nvlist_dst);
3268 free((void *)(uintptr_t)zc.zc_nvlist_dst);
3269 return (no_memory(zhp->zpool_hdl));
3273 * Upgrade a ZFS pool to the latest on-disk version.
3276 zpool_upgrade(zpool_handle_t *zhp, uint64_t new_version)
3278 zfs_cmd_t zc = { "\0", "\0", "\0", "\0", 0 };
3279 libzfs_handle_t *hdl = zhp->zpool_hdl;
3281 (void) strcpy(zc.zc_name, zhp->zpool_name);
3282 zc.zc_cookie = new_version;
3284 if (zfs_ioctl(hdl, ZFS_IOC_POOL_UPGRADE, &zc) != 0)
3285 return (zpool_standard_error_fmt(hdl, errno,
3286 dgettext(TEXT_DOMAIN, "cannot upgrade '%s'"),
3292 zpool_set_history_str(const char *subcommand, int argc, char **argv,
3297 (void) strlcpy(history_str, subcommand, HIS_MAX_RECORD_LEN);
3298 for (i = 1; i < argc; i++) {
3299 if (strlen(history_str) + 1 + strlen(argv[i]) >
3302 (void) strlcat(history_str, " ", HIS_MAX_RECORD_LEN);
3303 (void) strlcat(history_str, argv[i], HIS_MAX_RECORD_LEN);
3308 * Stage command history for logging.
3311 zpool_stage_history(libzfs_handle_t *hdl, const char *history_str)
3313 if (history_str == NULL)
3316 if (strlen(history_str) > HIS_MAX_RECORD_LEN)
3319 if (hdl->libzfs_log_str != NULL)
3320 free(hdl->libzfs_log_str);
3322 if ((hdl->libzfs_log_str = strdup(history_str)) == NULL)
3323 return (no_memory(hdl));
3329 * Perform ioctl to get some command history of a pool.
3331 * 'buf' is the buffer to fill up to 'len' bytes. 'off' is the
3332 * logical offset of the history buffer to start reading from.
3334 * Upon return, 'off' is the next logical offset to read from and
3335 * 'len' is the actual amount of bytes read into 'buf'.
3338 get_history(zpool_handle_t *zhp, char *buf, uint64_t *off, uint64_t *len)
3340 zfs_cmd_t zc = { "\0", "\0", "\0", "\0", 0 };
3341 libzfs_handle_t *hdl = zhp->zpool_hdl;
3343 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
3345 zc.zc_history = (uint64_t)(uintptr_t)buf;
3346 zc.zc_history_len = *len;
3347 zc.zc_history_offset = *off;
3349 if (ioctl(hdl->libzfs_fd, ZFS_IOC_POOL_GET_HISTORY, &zc) != 0) {
3352 return (zfs_error_fmt(hdl, EZFS_PERM,
3353 dgettext(TEXT_DOMAIN,
3354 "cannot show history for pool '%s'"),
3357 return (zfs_error_fmt(hdl, EZFS_NOHISTORY,
3358 dgettext(TEXT_DOMAIN, "cannot get history for pool "
3359 "'%s'"), zhp->zpool_name));
3361 return (zfs_error_fmt(hdl, EZFS_BADVERSION,
3362 dgettext(TEXT_DOMAIN, "cannot get history for pool "
3363 "'%s', pool must be upgraded"), zhp->zpool_name));
3365 return (zpool_standard_error_fmt(hdl, errno,
3366 dgettext(TEXT_DOMAIN,
3367 "cannot get history for '%s'"), zhp->zpool_name));
3371 *len = zc.zc_history_len;
3372 *off = zc.zc_history_offset;
3378 * Process the buffer of nvlists, unpacking and storing each nvlist record
3379 * into 'records'. 'leftover' is set to the number of bytes that weren't
3380 * processed as there wasn't a complete record.
3383 zpool_history_unpack(char *buf, uint64_t bytes_read, uint64_t *leftover,
3384 nvlist_t ***records, uint_t *numrecords)
3390 while (bytes_read > sizeof (reclen)) {
3392 /* get length of packed record (stored as little endian) */
3393 for (i = 0, reclen = 0; i < sizeof (reclen); i++)
3394 reclen += (uint64_t)(((uchar_t *)buf)[i]) << (8*i);
3396 if (bytes_read < sizeof (reclen) + reclen)
3400 if (nvlist_unpack(buf + sizeof (reclen), reclen, &nv, 0) != 0)
3402 bytes_read -= sizeof (reclen) + reclen;
3403 buf += sizeof (reclen) + reclen;
3405 /* add record to nvlist array */
3407 if (ISP2(*numrecords + 1)) {
3408 *records = realloc(*records,
3409 *numrecords * 2 * sizeof (nvlist_t *));
3411 (*records)[*numrecords - 1] = nv;
3414 *leftover = bytes_read;
3418 #define HIS_BUF_LEN (128*1024)
3421 * Retrieve the command history of a pool.
3424 zpool_get_history(zpool_handle_t *zhp, nvlist_t **nvhisp)
3426 char buf[HIS_BUF_LEN];
3428 nvlist_t **records = NULL;
3429 uint_t numrecords = 0;
3433 uint64_t bytes_read = sizeof (buf);
3436 if ((err = get_history(zhp, buf, &off, &bytes_read)) != 0)
3439 /* if nothing else was read in, we're at EOF, just return */
3443 if ((err = zpool_history_unpack(buf, bytes_read,
3444 &leftover, &records, &numrecords)) != 0)
3452 verify(nvlist_alloc(nvhisp, NV_UNIQUE_NAME, 0) == 0);
3453 verify(nvlist_add_nvlist_array(*nvhisp, ZPOOL_HIST_RECORD,
3454 records, numrecords) == 0);
3456 for (i = 0; i < numrecords; i++)
3457 nvlist_free(records[i]);
3464 * Retrieve the next event. If there is a new event available 'nvp' will
3465 * contain a newly allocated nvlist and 'dropped' will be set to the number
3466 * of missed events since the last call to this function. When 'nvp' is
3467 * set to NULL it indicates no new events are available. In either case
3468 * the function returns 0 and it is up to the caller to free 'nvp'. In
3469 * the case of a fatal error the function will return a non-zero value.
3470 * When the function is called in blocking mode it will not return until
3471 * a new event is available.
3474 zpool_events_next(libzfs_handle_t *hdl, nvlist_t **nvp,
3475 int *dropped, int block, int cleanup_fd)
3477 zfs_cmd_t zc = { "\0", "\0", "\0", "\0", 0 };
3482 zc.zc_cleanup_fd = cleanup_fd;
3485 zc.zc_guid = ZEVENT_NONBLOCK;
3487 if (zcmd_alloc_dst_nvlist(hdl, &zc, ZEVENT_SIZE) != 0)
3491 if (zfs_ioctl(hdl, ZFS_IOC_EVENTS_NEXT, &zc) != 0) {
3494 error = zfs_error_fmt(hdl, EZFS_POOLUNAVAIL,
3495 dgettext(TEXT_DOMAIN, "zfs shutdown"));
3498 /* Blocking error case should not occur */
3500 error = zpool_standard_error_fmt(hdl, errno,
3501 dgettext(TEXT_DOMAIN, "cannot get event"));
3505 if (zcmd_expand_dst_nvlist(hdl, &zc) != 0) {
3506 error = zfs_error_fmt(hdl, EZFS_NOMEM,
3507 dgettext(TEXT_DOMAIN, "cannot get event"));
3513 error = zpool_standard_error_fmt(hdl, errno,
3514 dgettext(TEXT_DOMAIN, "cannot get event"));
3519 error = zcmd_read_dst_nvlist(hdl, &zc, nvp);
3523 *dropped = (int)zc.zc_cookie;
3525 zcmd_free_nvlists(&zc);
3534 zpool_events_clear(libzfs_handle_t *hdl, int *count)
3536 zfs_cmd_t zc = { "\0", "\0", "\0", "\0", 0 };
3539 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN,
3540 "cannot clear events"));
3542 if (zfs_ioctl(hdl, ZFS_IOC_EVENTS_CLEAR, &zc) != 0)
3543 return (zpool_standard_error_fmt(hdl, errno, msg));
3546 *count = (int)zc.zc_cookie; /* # of events cleared */
3552 zpool_obj_to_path(zpool_handle_t *zhp, uint64_t dsobj, uint64_t obj,
3553 char *pathname, size_t len)
3555 zfs_cmd_t zc = { "\0", "\0", "\0", "\0", 0 };
3556 boolean_t mounted = B_FALSE;
3557 char *mntpnt = NULL;
3558 char dsname[MAXNAMELEN];
3561 /* special case for the MOS */
3562 (void) snprintf(pathname, len, "<metadata>:<0x%llx>", (longlong_t)obj);
3566 /* get the dataset's name */
3567 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
3569 if (ioctl(zhp->zpool_hdl->libzfs_fd,
3570 ZFS_IOC_DSOBJ_TO_DSNAME, &zc) != 0) {
3571 /* just write out a path of two object numbers */
3572 (void) snprintf(pathname, len, "<0x%llx>:<0x%llx>",
3573 (longlong_t)dsobj, (longlong_t)obj);
3576 (void) strlcpy(dsname, zc.zc_value, sizeof (dsname));
3578 /* find out if the dataset is mounted */
3579 mounted = is_mounted(zhp->zpool_hdl, dsname, &mntpnt);
3581 /* get the corrupted object's path */
3582 (void) strlcpy(zc.zc_name, dsname, sizeof (zc.zc_name));
3584 if (ioctl(zhp->zpool_hdl->libzfs_fd, ZFS_IOC_OBJ_TO_PATH,
3587 (void) snprintf(pathname, len, "%s%s", mntpnt,
3590 (void) snprintf(pathname, len, "%s:%s",
3591 dsname, zc.zc_value);
3594 (void) snprintf(pathname, len, "%s:<0x%llx>", dsname, (longlong_t)obj);
3600 * Read the EFI label from the config, if a label does not exist then
3601 * pass back the error to the caller. If the caller has passed a non-NULL
3602 * diskaddr argument then we set it to the starting address of the EFI
3606 read_efi_label(nvlist_t *config, diskaddr_t *sb)
3610 char diskname[MAXPATHLEN];
3613 if (nvlist_lookup_string(config, ZPOOL_CONFIG_PATH, &path) != 0)
3616 (void) snprintf(diskname, sizeof (diskname), "%s%s", RDISK_ROOT,
3617 strrchr(path, '/'));
3618 if ((fd = open(diskname, O_RDWR|O_DIRECT)) >= 0) {
3619 struct dk_gpt *vtoc;
3621 if ((err = efi_alloc_and_read(fd, &vtoc)) >= 0) {
3623 *sb = vtoc->efi_parts[0].p_start;
3632 * determine where a partition starts on a disk in the current
3636 find_start_block(nvlist_t *config)
3640 diskaddr_t sb = MAXOFFSET_T;
3643 if (nvlist_lookup_nvlist_array(config,
3644 ZPOOL_CONFIG_CHILDREN, &child, &children) != 0) {
3645 if (nvlist_lookup_uint64(config,
3646 ZPOOL_CONFIG_WHOLE_DISK,
3647 &wholedisk) != 0 || !wholedisk) {
3648 return (MAXOFFSET_T);
3650 if (read_efi_label(config, &sb) < 0)
3655 for (c = 0; c < children; c++) {
3656 sb = find_start_block(child[c]);
3657 if (sb != MAXOFFSET_T) {
3661 return (MAXOFFSET_T);
3665 zpool_label_disk_wait(char *path, int timeout)
3667 struct stat64 statbuf;
3671 * Wait timeout miliseconds for a newly created device to be available
3672 * from the given path. There is a small window when a /dev/ device
3673 * will exist and the udev link will not, so we must wait for the
3674 * symlink. Depending on the udev rules this may take a few seconds.
3676 for (i = 0; i < timeout; i++) {
3680 if ((stat64(path, &statbuf) == 0) && (errno == 0))
3688 zpool_label_disk_check(char *path)
3690 struct dk_gpt *vtoc;
3693 if ((fd = open(path, O_RDWR|O_DIRECT)) < 0)
3696 if ((err = efi_alloc_and_read(fd, &vtoc)) != 0) {
3701 if (vtoc->efi_flags & EFI_GPT_PRIMARY_CORRUPT) {
3713 * Label an individual disk. The name provided is the short name,
3714 * stripped of any leading /dev path.
3717 zpool_label_disk(libzfs_handle_t *hdl, zpool_handle_t *zhp, char *name)
3719 char path[MAXPATHLEN];
3720 struct dk_gpt *vtoc;
3722 size_t resv = EFI_MIN_RESV_SIZE;
3723 uint64_t slice_size;
3724 diskaddr_t start_block;
3727 /* prepare an error message just in case */
3728 (void) snprintf(errbuf, sizeof (errbuf),
3729 dgettext(TEXT_DOMAIN, "cannot label '%s'"), name);
3734 if (pool_is_bootable(zhp)) {
3735 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3736 "EFI labeled devices are not supported on root "
3738 return (zfs_error(hdl, EZFS_POOL_NOTSUP, errbuf));
3741 verify(nvlist_lookup_nvlist(zhp->zpool_config,
3742 ZPOOL_CONFIG_VDEV_TREE, &nvroot) == 0);
3744 if (zhp->zpool_start_block == 0)
3745 start_block = find_start_block(nvroot);
3747 start_block = zhp->zpool_start_block;
3748 zhp->zpool_start_block = start_block;
3751 start_block = NEW_START_BLOCK;
3754 (void) snprintf(path, sizeof (path), "%s/%s%s", RDISK_ROOT, name,
3757 if ((fd = open(path, O_RDWR|O_DIRECT)) < 0) {
3759 * This shouldn't happen. We've long since verified that this
3760 * is a valid device.
3762 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3763 "unable to open device '%s': %d"), path, errno);
3764 return (zfs_error(hdl, EZFS_OPENFAILED, errbuf));
3767 if (efi_alloc_and_init(fd, EFI_NUMPAR, &vtoc) != 0) {
3769 * The only way this can fail is if we run out of memory, or we
3770 * were unable to read the disk's capacity
3772 if (errno == ENOMEM)
3773 (void) no_memory(hdl);
3776 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3777 "unable to read disk capacity"), name);
3779 return (zfs_error(hdl, EZFS_NOCAP, errbuf));
3782 slice_size = vtoc->efi_last_u_lba + 1;
3783 slice_size -= EFI_MIN_RESV_SIZE;
3784 if (start_block == MAXOFFSET_T)
3785 start_block = NEW_START_BLOCK;
3786 slice_size -= start_block;
3788 vtoc->efi_parts[0].p_start = start_block;
3789 vtoc->efi_parts[0].p_size = slice_size;
3792 * Why we use V_USR: V_BACKUP confuses users, and is considered
3793 * disposable by some EFI utilities (since EFI doesn't have a backup
3794 * slice). V_UNASSIGNED is supposed to be used only for zero size
3795 * partitions, and efi_write() will fail if we use it. V_ROOT, V_BOOT,
3796 * etc. were all pretty specific. V_USR is as close to reality as we
3797 * can get, in the absence of V_OTHER.
3799 vtoc->efi_parts[0].p_tag = V_USR;
3800 (void) strcpy(vtoc->efi_parts[0].p_name, "zfs");
3802 vtoc->efi_parts[8].p_start = slice_size + start_block;
3803 vtoc->efi_parts[8].p_size = resv;
3804 vtoc->efi_parts[8].p_tag = V_RESERVED;
3806 if ((rval = efi_write(fd, vtoc)) != 0) {
3808 * Some block drivers (like pcata) may not support EFI
3809 * GPT labels. Print out a helpful error message dir-
3810 * ecting the user to manually label the disk and give
3816 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "try using "
3817 "parted(8) and then provide a specific slice: %d"), rval);
3818 return (zfs_error(hdl, EZFS_LABELFAILED, errbuf));
3824 /* Wait for the first expected slice to appear. */
3825 (void) snprintf(path, sizeof (path), "%s/%s%s%s", DISK_ROOT, name,
3826 isdigit(name[strlen(name)-1]) ? "p" : "", FIRST_SLICE);
3827 rval = zpool_label_disk_wait(path, 3000);
3829 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "failed to "
3830 "detect device partitions on '%s': %d"), path, rval);
3831 return (zfs_error(hdl, EZFS_LABELFAILED, errbuf));
3834 /* We can't be to paranoid. Read the label back and verify it. */
3835 (void) snprintf(path, sizeof (path), "%s/%s", DISK_ROOT, name);
3836 rval = zpool_label_disk_check(path);
3838 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "freshly written "
3839 "EFI label on '%s' is damaged. Ensure\nthis device "
3840 "is not in in use, and is functioning properly: %d"),
3842 return (zfs_error(hdl, EZFS_LABELFAILED, errbuf));