4 * The contents of this file are subject to the terms of the
5 * Common Development and Distribution License (the "License").
6 * You may not use this file except in compliance with the License.
8 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
9 * or http://www.opensolaris.org/os/licensing.
10 * See the License for the specific language governing permissions
11 * and limitations under the License.
13 * When distributing Covered Code, include this CDDL HEADER in each
14 * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
15 * If applicable, add the following below this CDDL HEADER, with the
16 * fields enclosed by brackets "[]" replaced with your own identifying
17 * information: Portions Copyright [yyyy] [name of copyright owner]
23 * Copyright 2015 Nexenta Systems, Inc. All rights reserved.
24 * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
25 * Copyright (c) 2011, 2018 by Delphix. All rights reserved.
26 * Copyright 2016 Igor Kozhukhov <ikozhukhov@gmail.com>
27 * Copyright (c) 2018 Datto Inc.
28 * Copyright (c) 2017 Open-E, Inc. All Rights Reserved.
29 * Copyright (c) 2017, Intel Corporation.
30 * Copyright (c) 2018, loli10K <ezomori.nozomu@gmail.com>
43 #include <sys/efi_partition.h>
44 #include <sys/systeminfo.h>
46 #include <sys/zfs_ioctl.h>
47 #include <sys/vdev_disk.h>
51 #include "zfs_namecheck.h"
53 #include "libzfs_impl.h"
54 #include "zfs_comutil.h"
55 #include "zfeature_common.h"
57 static int read_efi_label(nvlist_t *config, diskaddr_t *sb);
58 static boolean_t zpool_vdev_is_interior(const char *name);
60 typedef struct prop_flags {
61 int create:1; /* Validate property on creation */
62 int import:1; /* Validate property on import */
66 * ====================================================================
67 * zpool property functions
68 * ====================================================================
72 zpool_get_all_props(zpool_handle_t *zhp)
74 zfs_cmd_t zc = {"\0"};
75 libzfs_handle_t *hdl = zhp->zpool_hdl;
77 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
79 if (zcmd_alloc_dst_nvlist(hdl, &zc, 0) != 0)
82 while (ioctl(hdl->libzfs_fd, ZFS_IOC_POOL_GET_PROPS, &zc) != 0) {
83 if (errno == ENOMEM) {
84 if (zcmd_expand_dst_nvlist(hdl, &zc) != 0) {
85 zcmd_free_nvlists(&zc);
89 zcmd_free_nvlists(&zc);
94 if (zcmd_read_dst_nvlist(hdl, &zc, &zhp->zpool_props) != 0) {
95 zcmd_free_nvlists(&zc);
99 zcmd_free_nvlists(&zc);
105 zpool_props_refresh(zpool_handle_t *zhp)
109 old_props = zhp->zpool_props;
111 if (zpool_get_all_props(zhp) != 0)
114 nvlist_free(old_props);
119 zpool_get_prop_string(zpool_handle_t *zhp, zpool_prop_t prop,
125 zprop_source_t source;
127 nvl = zhp->zpool_props;
128 if (nvlist_lookup_nvlist(nvl, zpool_prop_to_name(prop), &nv) == 0) {
129 verify(nvlist_lookup_uint64(nv, ZPROP_SOURCE, &ival) == 0);
131 verify(nvlist_lookup_string(nv, ZPROP_VALUE, &value) == 0);
133 source = ZPROP_SRC_DEFAULT;
134 if ((value = (char *)zpool_prop_default_string(prop)) == NULL)
145 zpool_get_prop_int(zpool_handle_t *zhp, zpool_prop_t prop, zprop_source_t *src)
149 zprop_source_t source;
151 if (zhp->zpool_props == NULL && zpool_get_all_props(zhp)) {
153 * zpool_get_all_props() has most likely failed because
154 * the pool is faulted, but if all we need is the top level
155 * vdev's guid then get it from the zhp config nvlist.
157 if ((prop == ZPOOL_PROP_GUID) &&
158 (nvlist_lookup_nvlist(zhp->zpool_config,
159 ZPOOL_CONFIG_VDEV_TREE, &nv) == 0) &&
160 (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID, &value)
164 return (zpool_prop_default_numeric(prop));
167 nvl = zhp->zpool_props;
168 if (nvlist_lookup_nvlist(nvl, zpool_prop_to_name(prop), &nv) == 0) {
169 verify(nvlist_lookup_uint64(nv, ZPROP_SOURCE, &value) == 0);
171 verify(nvlist_lookup_uint64(nv, ZPROP_VALUE, &value) == 0);
173 source = ZPROP_SRC_DEFAULT;
174 value = zpool_prop_default_numeric(prop);
184 * Map VDEV STATE to printed strings.
187 zpool_state_to_name(vdev_state_t state, vdev_aux_t aux)
190 case VDEV_STATE_CLOSED:
191 case VDEV_STATE_OFFLINE:
192 return (gettext("OFFLINE"));
193 case VDEV_STATE_REMOVED:
194 return (gettext("REMOVED"));
195 case VDEV_STATE_CANT_OPEN:
196 if (aux == VDEV_AUX_CORRUPT_DATA || aux == VDEV_AUX_BAD_LOG)
197 return (gettext("FAULTED"));
198 else if (aux == VDEV_AUX_SPLIT_POOL)
199 return (gettext("SPLIT"));
201 return (gettext("UNAVAIL"));
202 case VDEV_STATE_FAULTED:
203 return (gettext("FAULTED"));
204 case VDEV_STATE_DEGRADED:
205 return (gettext("DEGRADED"));
206 case VDEV_STATE_HEALTHY:
207 return (gettext("ONLINE"));
213 return (gettext("UNKNOWN"));
217 * Map POOL STATE to printed strings.
220 zpool_pool_state_to_name(pool_state_t state)
225 case POOL_STATE_ACTIVE:
226 return (gettext("ACTIVE"));
227 case POOL_STATE_EXPORTED:
228 return (gettext("EXPORTED"));
229 case POOL_STATE_DESTROYED:
230 return (gettext("DESTROYED"));
231 case POOL_STATE_SPARE:
232 return (gettext("SPARE"));
233 case POOL_STATE_L2CACHE:
234 return (gettext("L2CACHE"));
235 case POOL_STATE_UNINITIALIZED:
236 return (gettext("UNINITIALIZED"));
237 case POOL_STATE_UNAVAIL:
238 return (gettext("UNAVAIL"));
239 case POOL_STATE_POTENTIALLY_ACTIVE:
240 return (gettext("POTENTIALLY_ACTIVE"));
243 return (gettext("UNKNOWN"));
247 * Given a pool handle, return the pool health string ("ONLINE", "DEGRADED",
251 zpool_get_state_str(zpool_handle_t *zhp)
253 zpool_errata_t errata;
254 zpool_status_t status;
260 status = zpool_get_status(zhp, NULL, &errata);
262 if (zpool_get_state(zhp) == POOL_STATE_UNAVAIL) {
263 str = gettext("FAULTED");
264 } else if (status == ZPOOL_STATUS_IO_FAILURE_WAIT ||
265 status == ZPOOL_STATUS_IO_FAILURE_MMP) {
266 str = gettext("SUSPENDED");
268 verify(nvlist_lookup_nvlist(zpool_get_config(zhp, NULL),
269 ZPOOL_CONFIG_VDEV_TREE, &nvroot) == 0);
270 verify(nvlist_lookup_uint64_array(nvroot,
271 ZPOOL_CONFIG_VDEV_STATS, (uint64_t **)&vs, &vsc)
273 str = zpool_state_to_name(vs->vs_state, vs->vs_aux);
279 * Get a zpool property value for 'prop' and return the value in
280 * a pre-allocated buffer.
283 zpool_get_prop(zpool_handle_t *zhp, zpool_prop_t prop, char *buf,
284 size_t len, zprop_source_t *srctype, boolean_t literal)
288 zprop_source_t src = ZPROP_SRC_NONE;
290 if (zpool_get_state(zhp) == POOL_STATE_UNAVAIL) {
292 case ZPOOL_PROP_NAME:
293 (void) strlcpy(buf, zpool_get_name(zhp), len);
296 case ZPOOL_PROP_HEALTH:
297 (void) strlcpy(buf, zpool_get_state_str(zhp), len);
300 case ZPOOL_PROP_GUID:
301 intval = zpool_get_prop_int(zhp, prop, &src);
302 (void) snprintf(buf, len, "%llu", (u_longlong_t)intval);
305 case ZPOOL_PROP_ALTROOT:
306 case ZPOOL_PROP_CACHEFILE:
307 case ZPOOL_PROP_COMMENT:
308 if (zhp->zpool_props != NULL ||
309 zpool_get_all_props(zhp) == 0) {
311 zpool_get_prop_string(zhp, prop, &src),
317 (void) strlcpy(buf, "-", len);
326 if (zhp->zpool_props == NULL && zpool_get_all_props(zhp) &&
327 prop != ZPOOL_PROP_NAME)
330 switch (zpool_prop_get_type(prop)) {
331 case PROP_TYPE_STRING:
332 (void) strlcpy(buf, zpool_get_prop_string(zhp, prop, &src),
336 case PROP_TYPE_NUMBER:
337 intval = zpool_get_prop_int(zhp, prop, &src);
340 case ZPOOL_PROP_SIZE:
341 case ZPOOL_PROP_ALLOCATED:
342 case ZPOOL_PROP_FREE:
343 case ZPOOL_PROP_FREEING:
344 case ZPOOL_PROP_LEAKED:
345 case ZPOOL_PROP_ASHIFT:
347 (void) snprintf(buf, len, "%llu",
348 (u_longlong_t)intval);
350 (void) zfs_nicenum(intval, buf, len);
353 case ZPOOL_PROP_EXPANDSZ:
354 case ZPOOL_PROP_CHECKPOINT:
356 (void) strlcpy(buf, "-", len);
357 } else if (literal) {
358 (void) snprintf(buf, len, "%llu",
359 (u_longlong_t)intval);
361 (void) zfs_nicebytes(intval, buf, len);
365 case ZPOOL_PROP_CAPACITY:
367 (void) snprintf(buf, len, "%llu",
368 (u_longlong_t)intval);
370 (void) snprintf(buf, len, "%llu%%",
371 (u_longlong_t)intval);
375 case ZPOOL_PROP_FRAGMENTATION:
376 if (intval == UINT64_MAX) {
377 (void) strlcpy(buf, "-", len);
378 } else if (literal) {
379 (void) snprintf(buf, len, "%llu",
380 (u_longlong_t)intval);
382 (void) snprintf(buf, len, "%llu%%",
383 (u_longlong_t)intval);
387 case ZPOOL_PROP_DEDUPRATIO:
389 (void) snprintf(buf, len, "%llu.%02llu",
390 (u_longlong_t)(intval / 100),
391 (u_longlong_t)(intval % 100));
393 (void) snprintf(buf, len, "%llu.%02llux",
394 (u_longlong_t)(intval / 100),
395 (u_longlong_t)(intval % 100));
398 case ZPOOL_PROP_HEALTH:
399 (void) strlcpy(buf, zpool_get_state_str(zhp), len);
401 case ZPOOL_PROP_VERSION:
402 if (intval >= SPA_VERSION_FEATURES) {
403 (void) snprintf(buf, len, "-");
408 (void) snprintf(buf, len, "%llu", (u_longlong_t)intval);
412 case PROP_TYPE_INDEX:
413 intval = zpool_get_prop_int(zhp, prop, &src);
414 if (zpool_prop_index_to_string(prop, intval, &strval)
417 (void) strlcpy(buf, strval, len);
431 * Check if the bootfs name has the same pool name as it is set to.
432 * Assuming bootfs is a valid dataset name.
435 bootfs_name_valid(const char *pool, char *bootfs)
437 int len = strlen(pool);
438 if (bootfs[0] == '\0')
441 if (!zfs_name_valid(bootfs, ZFS_TYPE_FILESYSTEM|ZFS_TYPE_SNAPSHOT))
444 if (strncmp(pool, bootfs, len) == 0 &&
445 (bootfs[len] == '/' || bootfs[len] == '\0'))
452 zpool_is_bootable(zpool_handle_t *zhp)
454 char bootfs[ZFS_MAX_DATASET_NAME_LEN];
456 return (zpool_get_prop(zhp, ZPOOL_PROP_BOOTFS, bootfs,
457 sizeof (bootfs), NULL, B_FALSE) == 0 && strncmp(bootfs, "-",
458 sizeof (bootfs)) != 0);
463 * Given an nvlist of zpool properties to be set, validate that they are
464 * correct, and parse any numeric properties (index, boolean, etc) if they are
465 * specified as strings.
468 zpool_valid_proplist(libzfs_handle_t *hdl, const char *poolname,
469 nvlist_t *props, uint64_t version, prop_flags_t flags, char *errbuf)
477 struct stat64 statbuf;
480 if (nvlist_alloc(&retprops, NV_UNIQUE_NAME, 0) != 0) {
481 (void) no_memory(hdl);
486 while ((elem = nvlist_next_nvpair(props, elem)) != NULL) {
487 const char *propname = nvpair_name(elem);
489 prop = zpool_name_to_prop(propname);
490 if (prop == ZPOOL_PROP_INVAL && zpool_prop_feature(propname)) {
492 char *fname = strchr(propname, '@') + 1;
494 err = zfeature_lookup_name(fname, NULL);
496 ASSERT3U(err, ==, ENOENT);
497 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
498 "invalid feature '%s'"), fname);
499 (void) zfs_error(hdl, EZFS_BADPROP, errbuf);
503 if (nvpair_type(elem) != DATA_TYPE_STRING) {
504 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
505 "'%s' must be a string"), propname);
506 (void) zfs_error(hdl, EZFS_BADPROP, errbuf);
510 (void) nvpair_value_string(elem, &strval);
511 if (strcmp(strval, ZFS_FEATURE_ENABLED) != 0 &&
512 strcmp(strval, ZFS_FEATURE_DISABLED) != 0) {
513 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
514 "property '%s' can only be set to "
515 "'enabled' or 'disabled'"), propname);
516 (void) zfs_error(hdl, EZFS_BADPROP, errbuf);
521 strcmp(strval, ZFS_FEATURE_DISABLED) == 0) {
522 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
523 "property '%s' can only be set to "
524 "'disabled' at creation time"), propname);
525 (void) zfs_error(hdl, EZFS_BADPROP, errbuf);
529 if (nvlist_add_uint64(retprops, propname, 0) != 0) {
530 (void) no_memory(hdl);
537 * Make sure this property is valid and applies to this type.
539 if (prop == ZPOOL_PROP_INVAL) {
540 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
541 "invalid property '%s'"), propname);
542 (void) zfs_error(hdl, EZFS_BADPROP, errbuf);
546 if (zpool_prop_readonly(prop)) {
547 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "'%s' "
548 "is readonly"), propname);
549 (void) zfs_error(hdl, EZFS_PROPREADONLY, errbuf);
553 if (!flags.create && zpool_prop_setonce(prop)) {
554 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
555 "property '%s' can only be set at "
556 "creation time"), propname);
557 (void) zfs_error(hdl, EZFS_BADPROP, errbuf);
561 if (zprop_parse_value(hdl, elem, prop, ZFS_TYPE_POOL, retprops,
562 &strval, &intval, errbuf) != 0)
566 * Perform additional checking for specific properties.
569 case ZPOOL_PROP_VERSION:
570 if (intval < version ||
571 !SPA_VERSION_IS_SUPPORTED(intval)) {
572 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
573 "property '%s' number %d is invalid."),
575 (void) zfs_error(hdl, EZFS_BADVERSION, errbuf);
580 case ZPOOL_PROP_ASHIFT:
582 (intval < ASHIFT_MIN || intval > ASHIFT_MAX)) {
583 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
584 "property '%s' number %d is invalid, only "
585 "values between %" PRId32 " and "
586 "%" PRId32 " are allowed."),
587 propname, intval, ASHIFT_MIN, ASHIFT_MAX);
588 (void) zfs_error(hdl, EZFS_BADPROP, errbuf);
593 case ZPOOL_PROP_BOOTFS:
594 if (flags.create || flags.import) {
595 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
596 "property '%s' cannot be set at creation "
597 "or import time"), propname);
598 (void) zfs_error(hdl, EZFS_BADPROP, errbuf);
602 if (version < SPA_VERSION_BOOTFS) {
603 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
604 "pool must be upgraded to support "
605 "'%s' property"), propname);
606 (void) zfs_error(hdl, EZFS_BADVERSION, errbuf);
611 * bootfs property value has to be a dataset name and
612 * the dataset has to be in the same pool as it sets to.
614 if (!bootfs_name_valid(poolname, strval)) {
615 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "'%s' "
616 "is an invalid name"), strval);
617 (void) zfs_error(hdl, EZFS_INVALIDNAME, errbuf);
621 if ((zhp = zpool_open_canfail(hdl, poolname)) == NULL) {
622 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
623 "could not open pool '%s'"), poolname);
624 (void) zfs_error(hdl, EZFS_OPENFAILED, errbuf);
630 case ZPOOL_PROP_ALTROOT:
631 if (!flags.create && !flags.import) {
632 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
633 "property '%s' can only be set during pool "
634 "creation or import"), propname);
635 (void) zfs_error(hdl, EZFS_BADPROP, errbuf);
639 if (strval[0] != '/') {
640 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
641 "bad alternate root '%s'"), strval);
642 (void) zfs_error(hdl, EZFS_BADPATH, errbuf);
647 case ZPOOL_PROP_CACHEFILE:
648 if (strval[0] == '\0')
651 if (strcmp(strval, "none") == 0)
654 if (strval[0] != '/') {
655 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
656 "property '%s' must be empty, an "
657 "absolute path, or 'none'"), propname);
658 (void) zfs_error(hdl, EZFS_BADPATH, errbuf);
662 slash = strrchr(strval, '/');
664 if (slash[1] == '\0' || strcmp(slash, "/.") == 0 ||
665 strcmp(slash, "/..") == 0) {
666 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
667 "'%s' is not a valid file"), strval);
668 (void) zfs_error(hdl, EZFS_BADPATH, errbuf);
674 if (strval[0] != '\0' &&
675 (stat64(strval, &statbuf) != 0 ||
676 !S_ISDIR(statbuf.st_mode))) {
677 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
678 "'%s' is not a valid directory"),
680 (void) zfs_error(hdl, EZFS_BADPATH, errbuf);
687 case ZPOOL_PROP_COMMENT:
688 for (check = strval; *check != '\0'; check++) {
689 if (!isprint(*check)) {
691 dgettext(TEXT_DOMAIN,
692 "comment may only have printable "
694 (void) zfs_error(hdl, EZFS_BADPROP,
699 if (strlen(strval) > ZPROP_MAX_COMMENT) {
700 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
701 "comment must not exceed %d characters"),
703 (void) zfs_error(hdl, EZFS_BADPROP, errbuf);
707 case ZPOOL_PROP_READONLY:
709 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
710 "property '%s' can only be set at "
711 "import time"), propname);
712 (void) zfs_error(hdl, EZFS_BADPROP, errbuf);
716 case ZPOOL_PROP_MULTIHOST:
717 if (get_system_hostid() == 0) {
718 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
719 "requires a non-zero system hostid"));
720 (void) zfs_error(hdl, EZFS_BADPROP, errbuf);
724 case ZPOOL_PROP_DEDUPDITTO:
725 if (intval < ZIO_DEDUPDITTO_MIN && intval != 0) {
726 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
727 "property '%s' value %d is invalid; only "
728 "values of 0 or >= %" PRId32 " are allowed "
729 "for this property."),
730 propname, intval, ZIO_DEDUPDITTO_MIN);
731 (void) zfs_error(hdl, EZFS_BADPROP, errbuf);
737 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
738 "property '%s'(%d) not defined"), propname, prop);
745 nvlist_free(retprops);
750 * Set zpool property : propname=propval.
753 zpool_set_prop(zpool_handle_t *zhp, const char *propname, const char *propval)
755 zfs_cmd_t zc = {"\0"};
758 nvlist_t *nvl = NULL;
761 prop_flags_t flags = { 0 };
763 (void) snprintf(errbuf, sizeof (errbuf),
764 dgettext(TEXT_DOMAIN, "cannot set property for '%s'"),
767 if (nvlist_alloc(&nvl, NV_UNIQUE_NAME, 0) != 0)
768 return (no_memory(zhp->zpool_hdl));
770 if (nvlist_add_string(nvl, propname, propval) != 0) {
772 return (no_memory(zhp->zpool_hdl));
775 version = zpool_get_prop_int(zhp, ZPOOL_PROP_VERSION, NULL);
776 if ((realprops = zpool_valid_proplist(zhp->zpool_hdl,
777 zhp->zpool_name, nvl, version, flags, errbuf)) == NULL) {
786 * Execute the corresponding ioctl() to set this property.
788 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
790 if (zcmd_write_src_nvlist(zhp->zpool_hdl, &zc, nvl) != 0) {
795 ret = zfs_ioctl(zhp->zpool_hdl, ZFS_IOC_POOL_SET_PROPS, &zc);
797 zcmd_free_nvlists(&zc);
801 (void) zpool_standard_error(zhp->zpool_hdl, errno, errbuf);
803 (void) zpool_props_refresh(zhp);
809 zpool_expand_proplist(zpool_handle_t *zhp, zprop_list_t **plp)
811 libzfs_handle_t *hdl = zhp->zpool_hdl;
813 char buf[ZFS_MAXPROPLEN];
814 nvlist_t *features = NULL;
817 boolean_t firstexpand = (NULL == *plp);
820 if (zprop_expand_list(hdl, plp, ZFS_TYPE_POOL) != 0)
824 while (*last != NULL)
825 last = &(*last)->pl_next;
828 features = zpool_get_features(zhp);
830 if ((*plp)->pl_all && firstexpand) {
831 for (i = 0; i < SPA_FEATURES; i++) {
832 zprop_list_t *entry = zfs_alloc(hdl,
833 sizeof (zprop_list_t));
834 entry->pl_prop = ZPROP_INVAL;
835 entry->pl_user_prop = zfs_asprintf(hdl, "feature@%s",
836 spa_feature_table[i].fi_uname);
837 entry->pl_width = strlen(entry->pl_user_prop);
838 entry->pl_all = B_TRUE;
841 last = &entry->pl_next;
845 /* add any unsupported features */
846 for (nvp = nvlist_next_nvpair(features, NULL);
847 nvp != NULL; nvp = nvlist_next_nvpair(features, nvp)) {
852 if (zfeature_is_supported(nvpair_name(nvp)))
855 propname = zfs_asprintf(hdl, "unsupported@%s",
859 * Before adding the property to the list make sure that no
860 * other pool already added the same property.
864 while (entry != NULL) {
865 if (entry->pl_user_prop != NULL &&
866 strcmp(propname, entry->pl_user_prop) == 0) {
870 entry = entry->pl_next;
877 entry = zfs_alloc(hdl, sizeof (zprop_list_t));
878 entry->pl_prop = ZPROP_INVAL;
879 entry->pl_user_prop = propname;
880 entry->pl_width = strlen(entry->pl_user_prop);
881 entry->pl_all = B_TRUE;
884 last = &entry->pl_next;
887 for (entry = *plp; entry != NULL; entry = entry->pl_next) {
892 if (entry->pl_prop != ZPROP_INVAL &&
893 zpool_get_prop(zhp, entry->pl_prop, buf, sizeof (buf),
894 NULL, B_FALSE) == 0) {
895 if (strlen(buf) > entry->pl_width)
896 entry->pl_width = strlen(buf);
904 * Get the state for the given feature on the given ZFS pool.
907 zpool_prop_get_feature(zpool_handle_t *zhp, const char *propname, char *buf,
911 boolean_t found = B_FALSE;
912 nvlist_t *features = zpool_get_features(zhp);
914 const char *feature = strchr(propname, '@') + 1;
916 supported = zpool_prop_feature(propname);
917 ASSERT(supported || zpool_prop_unsupported(propname));
920 * Convert from feature name to feature guid. This conversion is
921 * unnecessary for unsupported@... properties because they already
928 ret = zfeature_lookup_name(feature, &fid);
930 (void) strlcpy(buf, "-", len);
933 feature = spa_feature_table[fid].fi_guid;
936 if (nvlist_lookup_uint64(features, feature, &refcount) == 0)
941 (void) strlcpy(buf, ZFS_FEATURE_DISABLED, len);
944 (void) strlcpy(buf, ZFS_FEATURE_ENABLED, len);
946 (void) strlcpy(buf, ZFS_FEATURE_ACTIVE, len);
951 (void) strcpy(buf, ZFS_UNSUPPORTED_INACTIVE);
953 (void) strcpy(buf, ZFS_UNSUPPORTED_READONLY);
956 (void) strlcpy(buf, "-", len);
965 * Validate the given pool name, optionally putting an extended error message in
969 zpool_name_valid(libzfs_handle_t *hdl, boolean_t isopen, const char *pool)
975 ret = pool_namecheck(pool, &why, &what);
978 * The rules for reserved pool names were extended at a later point.
979 * But we need to support users with existing pools that may now be
980 * invalid. So we only check for this expanded set of names during a
981 * create (or import), and only in userland.
983 if (ret == 0 && !isopen &&
984 (strncmp(pool, "mirror", 6) == 0 ||
985 strncmp(pool, "raidz", 5) == 0 ||
986 strncmp(pool, "spare", 5) == 0 ||
987 strcmp(pool, "log") == 0)) {
990 dgettext(TEXT_DOMAIN, "name is reserved"));
998 case NAME_ERR_TOOLONG:
1000 dgettext(TEXT_DOMAIN, "name is too long"));
1003 case NAME_ERR_INVALCHAR:
1005 dgettext(TEXT_DOMAIN, "invalid character "
1006 "'%c' in pool name"), what);
1009 case NAME_ERR_NOLETTER:
1010 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1011 "name must begin with a letter"));
1014 case NAME_ERR_RESERVED:
1015 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1016 "name is reserved"));
1019 case NAME_ERR_DISKLIKE:
1020 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1021 "pool name is reserved"));
1024 case NAME_ERR_LEADING_SLASH:
1025 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1026 "leading slash in name"));
1029 case NAME_ERR_EMPTY_COMPONENT:
1030 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1031 "empty component in name"));
1034 case NAME_ERR_TRAILING_SLASH:
1035 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1036 "trailing slash in name"));
1039 case NAME_ERR_MULTIPLE_DELIMITERS:
1040 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1041 "multiple '@' and/or '#' delimiters in "
1045 case NAME_ERR_NO_AT:
1046 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1047 "permission set is missing '@'"));
1051 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1052 "(%d) not defined"), why);
1063 * Open a handle to the given pool, even if the pool is currently in the FAULTED
1067 zpool_open_canfail(libzfs_handle_t *hdl, const char *pool)
1069 zpool_handle_t *zhp;
1073 * Make sure the pool name is valid.
1075 if (!zpool_name_valid(hdl, B_TRUE, pool)) {
1076 (void) zfs_error_fmt(hdl, EZFS_INVALIDNAME,
1077 dgettext(TEXT_DOMAIN, "cannot open '%s'"),
1082 if ((zhp = zfs_alloc(hdl, sizeof (zpool_handle_t))) == NULL)
1085 zhp->zpool_hdl = hdl;
1086 (void) strlcpy(zhp->zpool_name, pool, sizeof (zhp->zpool_name));
1088 if (zpool_refresh_stats(zhp, &missing) != 0) {
1094 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "no such pool"));
1095 (void) zfs_error_fmt(hdl, EZFS_NOENT,
1096 dgettext(TEXT_DOMAIN, "cannot open '%s'"), pool);
1105 * Like the above, but silent on error. Used when iterating over pools (because
1106 * the configuration cache may be out of date).
1109 zpool_open_silent(libzfs_handle_t *hdl, const char *pool, zpool_handle_t **ret)
1111 zpool_handle_t *zhp;
1114 if ((zhp = zfs_alloc(hdl, sizeof (zpool_handle_t))) == NULL)
1117 zhp->zpool_hdl = hdl;
1118 (void) strlcpy(zhp->zpool_name, pool, sizeof (zhp->zpool_name));
1120 if (zpool_refresh_stats(zhp, &missing) != 0) {
1136 * Similar to zpool_open_canfail(), but refuses to open pools in the faulted
1140 zpool_open(libzfs_handle_t *hdl, const char *pool)
1142 zpool_handle_t *zhp;
1144 if ((zhp = zpool_open_canfail(hdl, pool)) == NULL)
1147 if (zhp->zpool_state == POOL_STATE_UNAVAIL) {
1148 (void) zfs_error_fmt(hdl, EZFS_POOLUNAVAIL,
1149 dgettext(TEXT_DOMAIN, "cannot open '%s'"), zhp->zpool_name);
1158 * Close the handle. Simply frees the memory associated with the handle.
1161 zpool_close(zpool_handle_t *zhp)
1163 nvlist_free(zhp->zpool_config);
1164 nvlist_free(zhp->zpool_old_config);
1165 nvlist_free(zhp->zpool_props);
1170 * Return the name of the pool.
1173 zpool_get_name(zpool_handle_t *zhp)
1175 return (zhp->zpool_name);
1180 * Return the state of the pool (ACTIVE or UNAVAILABLE)
1183 zpool_get_state(zpool_handle_t *zhp)
1185 return (zhp->zpool_state);
1189 * Check if vdev list contains a special vdev
1192 zpool_has_special_vdev(nvlist_t *nvroot)
1197 if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_CHILDREN, &child,
1199 for (uint_t c = 0; c < children; c++) {
1202 if (nvlist_lookup_string(child[c],
1203 ZPOOL_CONFIG_ALLOCATION_BIAS, &bias) == 0 &&
1204 strcmp(bias, VDEV_ALLOC_BIAS_SPECIAL) == 0) {
1213 * Create the named pool, using the provided vdev list. It is assumed
1214 * that the consumer has already validated the contents of the nvlist, so we
1215 * don't have to worry about error semantics.
1218 zpool_create(libzfs_handle_t *hdl, const char *pool, nvlist_t *nvroot,
1219 nvlist_t *props, nvlist_t *fsprops)
1221 zfs_cmd_t zc = {"\0"};
1222 nvlist_t *zc_fsprops = NULL;
1223 nvlist_t *zc_props = NULL;
1224 nvlist_t *hidden_args = NULL;
1225 uint8_t *wkeydata = NULL;
1230 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN,
1231 "cannot create '%s'"), pool);
1233 if (!zpool_name_valid(hdl, B_FALSE, pool))
1234 return (zfs_error(hdl, EZFS_INVALIDNAME, msg));
1236 if (zcmd_write_conf_nvlist(hdl, &zc, nvroot) != 0)
1240 prop_flags_t flags = { .create = B_TRUE, .import = B_FALSE };
1242 if ((zc_props = zpool_valid_proplist(hdl, pool, props,
1243 SPA_VERSION_1, flags, msg)) == NULL) {
1252 zoned = ((nvlist_lookup_string(fsprops,
1253 zfs_prop_to_name(ZFS_PROP_ZONED), &zonestr) == 0) &&
1254 strcmp(zonestr, "on") == 0);
1256 if ((zc_fsprops = zfs_valid_proplist(hdl, ZFS_TYPE_FILESYSTEM,
1257 fsprops, zoned, NULL, NULL, B_TRUE, msg)) == NULL) {
1261 if (nvlist_exists(zc_fsprops,
1262 zfs_prop_to_name(ZFS_PROP_SPECIAL_SMALL_BLOCKS)) &&
1263 !zpool_has_special_vdev(nvroot)) {
1264 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1265 "%s property requires a special vdev"),
1266 zfs_prop_to_name(ZFS_PROP_SPECIAL_SMALL_BLOCKS));
1267 (void) zfs_error(hdl, EZFS_BADPROP, msg);
1272 (nvlist_alloc(&zc_props, NV_UNIQUE_NAME, 0) != 0)) {
1275 if (zfs_crypto_create(hdl, NULL, zc_fsprops, props, B_TRUE,
1276 &wkeydata, &wkeylen) != 0) {
1277 zfs_error(hdl, EZFS_CRYPTOFAILED, msg);
1280 if (nvlist_add_nvlist(zc_props,
1281 ZPOOL_ROOTFS_PROPS, zc_fsprops) != 0) {
1284 if (wkeydata != NULL) {
1285 if (nvlist_alloc(&hidden_args, NV_UNIQUE_NAME, 0) != 0)
1288 if (nvlist_add_uint8_array(hidden_args, "wkeydata",
1289 wkeydata, wkeylen) != 0)
1292 if (nvlist_add_nvlist(zc_props, ZPOOL_HIDDEN_ARGS,
1298 if (zc_props && zcmd_write_src_nvlist(hdl, &zc, zc_props) != 0)
1301 (void) strlcpy(zc.zc_name, pool, sizeof (zc.zc_name));
1303 if ((ret = zfs_ioctl(hdl, ZFS_IOC_POOL_CREATE, &zc)) != 0) {
1305 zcmd_free_nvlists(&zc);
1306 nvlist_free(zc_props);
1307 nvlist_free(zc_fsprops);
1308 nvlist_free(hidden_args);
1309 if (wkeydata != NULL)
1315 * This can happen if the user has specified the same
1316 * device multiple times. We can't reliably detect this
1317 * until we try to add it and see we already have a
1318 * label. This can also happen under if the device is
1319 * part of an active md or lvm device.
1321 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1322 "one or more vdevs refer to the same device, or "
1323 "one of\nthe devices is part of an active md or "
1325 return (zfs_error(hdl, EZFS_BADDEV, msg));
1329 * This happens if the record size is smaller or larger
1330 * than the allowed size range, or not a power of 2.
1332 * NOTE: although zfs_valid_proplist is called earlier,
1333 * this case may have slipped through since the
1334 * pool does not exist yet and it is therefore
1335 * impossible to read properties e.g. max blocksize
1338 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1339 "record size invalid"));
1340 return (zfs_error(hdl, EZFS_BADPROP, msg));
1344 * This occurs when one of the devices is below
1345 * SPA_MINDEVSIZE. Unfortunately, we can't detect which
1346 * device was the problem device since there's no
1347 * reliable way to determine device size from userland.
1352 zfs_nicebytes(SPA_MINDEVSIZE, buf,
1355 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1356 "one or more devices is less than the "
1357 "minimum size (%s)"), buf);
1359 return (zfs_error(hdl, EZFS_BADDEV, msg));
1362 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1363 "one or more devices is out of space"));
1364 return (zfs_error(hdl, EZFS_BADDEV, msg));
1367 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1368 "cache device must be a disk or disk slice"));
1369 return (zfs_error(hdl, EZFS_BADDEV, msg));
1372 return (zpool_standard_error(hdl, errno, msg));
1377 zcmd_free_nvlists(&zc);
1378 nvlist_free(zc_props);
1379 nvlist_free(zc_fsprops);
1380 nvlist_free(hidden_args);
1381 if (wkeydata != NULL)
1387 * Destroy the given pool. It is up to the caller to ensure that there are no
1388 * datasets left in the pool.
1391 zpool_destroy(zpool_handle_t *zhp, const char *log_str)
1393 zfs_cmd_t zc = {"\0"};
1394 zfs_handle_t *zfp = NULL;
1395 libzfs_handle_t *hdl = zhp->zpool_hdl;
1398 if (zhp->zpool_state == POOL_STATE_ACTIVE &&
1399 (zfp = zfs_open(hdl, zhp->zpool_name, ZFS_TYPE_FILESYSTEM)) == NULL)
1402 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
1403 zc.zc_history = (uint64_t)(uintptr_t)log_str;
1405 if (zfs_ioctl(hdl, ZFS_IOC_POOL_DESTROY, &zc) != 0) {
1406 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN,
1407 "cannot destroy '%s'"), zhp->zpool_name);
1409 if (errno == EROFS) {
1410 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1411 "one or more devices is read only"));
1412 (void) zfs_error(hdl, EZFS_BADDEV, msg);
1414 (void) zpool_standard_error(hdl, errno, msg);
1423 remove_mountpoint(zfp);
1431 * Create a checkpoint in the given pool.
1434 zpool_checkpoint(zpool_handle_t *zhp)
1436 libzfs_handle_t *hdl = zhp->zpool_hdl;
1440 error = lzc_pool_checkpoint(zhp->zpool_name);
1442 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN,
1443 "cannot checkpoint '%s'"), zhp->zpool_name);
1444 (void) zpool_standard_error(hdl, error, msg);
1452 * Discard the checkpoint from the given pool.
1455 zpool_discard_checkpoint(zpool_handle_t *zhp)
1457 libzfs_handle_t *hdl = zhp->zpool_hdl;
1461 error = lzc_pool_checkpoint_discard(zhp->zpool_name);
1463 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN,
1464 "cannot discard checkpoint in '%s'"), zhp->zpool_name);
1465 (void) zpool_standard_error(hdl, error, msg);
1473 * Add the given vdevs to the pool. The caller must have already performed the
1474 * necessary verification to ensure that the vdev specification is well-formed.
1477 zpool_add(zpool_handle_t *zhp, nvlist_t *nvroot)
1479 zfs_cmd_t zc = {"\0"};
1481 libzfs_handle_t *hdl = zhp->zpool_hdl;
1483 nvlist_t **spares, **l2cache;
1484 uint_t nspares, nl2cache;
1486 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN,
1487 "cannot add to '%s'"), zhp->zpool_name);
1489 if (zpool_get_prop_int(zhp, ZPOOL_PROP_VERSION, NULL) <
1490 SPA_VERSION_SPARES &&
1491 nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_SPARES,
1492 &spares, &nspares) == 0) {
1493 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "pool must be "
1494 "upgraded to add hot spares"));
1495 return (zfs_error(hdl, EZFS_BADVERSION, msg));
1498 if (zpool_get_prop_int(zhp, ZPOOL_PROP_VERSION, NULL) <
1499 SPA_VERSION_L2CACHE &&
1500 nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_L2CACHE,
1501 &l2cache, &nl2cache) == 0) {
1502 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "pool must be "
1503 "upgraded to add cache devices"));
1504 return (zfs_error(hdl, EZFS_BADVERSION, msg));
1507 if (zcmd_write_conf_nvlist(hdl, &zc, nvroot) != 0)
1509 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
1511 if (zfs_ioctl(hdl, ZFS_IOC_VDEV_ADD, &zc) != 0) {
1515 * This can happen if the user has specified the same
1516 * device multiple times. We can't reliably detect this
1517 * until we try to add it and see we already have a
1520 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1521 "one or more vdevs refer to the same device"));
1522 (void) zfs_error(hdl, EZFS_BADDEV, msg);
1526 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1527 "invalid config; a pool with removing/removed "
1528 "vdevs does not support adding raidz vdevs"));
1529 (void) zfs_error(hdl, EZFS_BADDEV, msg);
1534 * This occurrs when one of the devices is below
1535 * SPA_MINDEVSIZE. Unfortunately, we can't detect which
1536 * device was the problem device since there's no
1537 * reliable way to determine device size from userland.
1542 zfs_nicebytes(SPA_MINDEVSIZE, buf,
1545 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1546 "device is less than the minimum "
1549 (void) zfs_error(hdl, EZFS_BADDEV, msg);
1553 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1554 "pool must be upgraded to add these vdevs"));
1555 (void) zfs_error(hdl, EZFS_BADVERSION, msg);
1559 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1560 "cache device must be a disk or disk slice"));
1561 (void) zfs_error(hdl, EZFS_BADDEV, msg);
1565 (void) zpool_standard_error(hdl, errno, msg);
1573 zcmd_free_nvlists(&zc);
1579 * Exports the pool from the system. The caller must ensure that there are no
1580 * mounted datasets in the pool.
1583 zpool_export_common(zpool_handle_t *zhp, boolean_t force, boolean_t hardforce,
1584 const char *log_str)
1586 zfs_cmd_t zc = {"\0"};
1589 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN,
1590 "cannot export '%s'"), zhp->zpool_name);
1592 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
1593 zc.zc_cookie = force;
1594 zc.zc_guid = hardforce;
1595 zc.zc_history = (uint64_t)(uintptr_t)log_str;
1597 if (zfs_ioctl(zhp->zpool_hdl, ZFS_IOC_POOL_EXPORT, &zc) != 0) {
1600 zfs_error_aux(zhp->zpool_hdl, dgettext(TEXT_DOMAIN,
1601 "use '-f' to override the following errors:\n"
1602 "'%s' has an active shared spare which could be"
1603 " used by other pools once '%s' is exported."),
1604 zhp->zpool_name, zhp->zpool_name);
1605 return (zfs_error(zhp->zpool_hdl, EZFS_ACTIVE_SPARE,
1608 return (zpool_standard_error_fmt(zhp->zpool_hdl, errno,
1617 zpool_export(zpool_handle_t *zhp, boolean_t force, const char *log_str)
1619 return (zpool_export_common(zhp, force, B_FALSE, log_str));
1623 zpool_export_force(zpool_handle_t *zhp, const char *log_str)
1625 return (zpool_export_common(zhp, B_TRUE, B_TRUE, log_str));
1629 zpool_rewind_exclaim(libzfs_handle_t *hdl, const char *name, boolean_t dryrun,
1632 nvlist_t *nv = NULL;
1638 if (!hdl->libzfs_printerr || config == NULL)
1641 if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_LOAD_INFO, &nv) != 0 ||
1642 nvlist_lookup_nvlist(nv, ZPOOL_CONFIG_REWIND_INFO, &nv) != 0) {
1646 if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_LOAD_TIME, &rewindto) != 0)
1648 (void) nvlist_lookup_int64(nv, ZPOOL_CONFIG_REWIND_TIME, &loss);
1650 if (localtime_r((time_t *)&rewindto, &t) != NULL &&
1651 strftime(timestr, 128, "%c", &t) != 0) {
1653 (void) printf(dgettext(TEXT_DOMAIN,
1654 "Would be able to return %s "
1655 "to its state as of %s.\n"),
1658 (void) printf(dgettext(TEXT_DOMAIN,
1659 "Pool %s returned to its state as of %s.\n"),
1663 (void) printf(dgettext(TEXT_DOMAIN,
1664 "%s approximately %lld "),
1665 dryrun ? "Would discard" : "Discarded",
1666 ((longlong_t)loss + 30) / 60);
1667 (void) printf(dgettext(TEXT_DOMAIN,
1668 "minutes of transactions.\n"));
1669 } else if (loss > 0) {
1670 (void) printf(dgettext(TEXT_DOMAIN,
1671 "%s approximately %lld "),
1672 dryrun ? "Would discard" : "Discarded",
1674 (void) printf(dgettext(TEXT_DOMAIN,
1675 "seconds of transactions.\n"));
1681 zpool_explain_recover(libzfs_handle_t *hdl, const char *name, int reason,
1684 nvlist_t *nv = NULL;
1686 uint64_t edata = UINT64_MAX;
1691 if (!hdl->libzfs_printerr)
1695 (void) printf(dgettext(TEXT_DOMAIN, "action: "));
1697 (void) printf(dgettext(TEXT_DOMAIN, "\t"));
1699 /* All attempted rewinds failed if ZPOOL_CONFIG_LOAD_TIME missing */
1700 if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_LOAD_INFO, &nv) != 0 ||
1701 nvlist_lookup_nvlist(nv, ZPOOL_CONFIG_REWIND_INFO, &nv) != 0 ||
1702 nvlist_lookup_uint64(nv, ZPOOL_CONFIG_LOAD_TIME, &rewindto) != 0)
1705 (void) nvlist_lookup_int64(nv, ZPOOL_CONFIG_REWIND_TIME, &loss);
1706 (void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_LOAD_DATA_ERRORS,
1709 (void) printf(dgettext(TEXT_DOMAIN,
1710 "Recovery is possible, but will result in some data loss.\n"));
1712 if (localtime_r((time_t *)&rewindto, &t) != NULL &&
1713 strftime(timestr, 128, "%c", &t) != 0) {
1714 (void) printf(dgettext(TEXT_DOMAIN,
1715 "\tReturning the pool to its state as of %s\n"
1716 "\tshould correct the problem. "),
1719 (void) printf(dgettext(TEXT_DOMAIN,
1720 "\tReverting the pool to an earlier state "
1721 "should correct the problem.\n\t"));
1725 (void) printf(dgettext(TEXT_DOMAIN,
1726 "Approximately %lld minutes of data\n"
1727 "\tmust be discarded, irreversibly. "),
1728 ((longlong_t)loss + 30) / 60);
1729 } else if (loss > 0) {
1730 (void) printf(dgettext(TEXT_DOMAIN,
1731 "Approximately %lld seconds of data\n"
1732 "\tmust be discarded, irreversibly. "),
1735 if (edata != 0 && edata != UINT64_MAX) {
1737 (void) printf(dgettext(TEXT_DOMAIN,
1738 "After rewind, at least\n"
1739 "\tone persistent user-data error will remain. "));
1741 (void) printf(dgettext(TEXT_DOMAIN,
1742 "After rewind, several\n"
1743 "\tpersistent user-data errors will remain. "));
1746 (void) printf(dgettext(TEXT_DOMAIN,
1747 "Recovery can be attempted\n\tby executing 'zpool %s -F %s'. "),
1748 reason >= 0 ? "clear" : "import", name);
1750 (void) printf(dgettext(TEXT_DOMAIN,
1751 "A scrub of the pool\n"
1752 "\tis strongly recommended after recovery.\n"));
1756 (void) printf(dgettext(TEXT_DOMAIN,
1757 "Destroy and re-create the pool from\n\ta backup source.\n"));
1761 * zpool_import() is a contracted interface. Should be kept the same
1764 * Applications should use zpool_import_props() to import a pool with
1765 * new properties value to be set.
1768 zpool_import(libzfs_handle_t *hdl, nvlist_t *config, const char *newname,
1771 nvlist_t *props = NULL;
1774 if (altroot != NULL) {
1775 if (nvlist_alloc(&props, NV_UNIQUE_NAME, 0) != 0) {
1776 return (zfs_error_fmt(hdl, EZFS_NOMEM,
1777 dgettext(TEXT_DOMAIN, "cannot import '%s'"),
1781 if (nvlist_add_string(props,
1782 zpool_prop_to_name(ZPOOL_PROP_ALTROOT), altroot) != 0 ||
1783 nvlist_add_string(props,
1784 zpool_prop_to_name(ZPOOL_PROP_CACHEFILE), "none") != 0) {
1786 return (zfs_error_fmt(hdl, EZFS_NOMEM,
1787 dgettext(TEXT_DOMAIN, "cannot import '%s'"),
1792 ret = zpool_import_props(hdl, config, newname, props,
1799 print_vdev_tree(libzfs_handle_t *hdl, const char *name, nvlist_t *nv,
1805 uint64_t is_log = 0;
1807 (void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_IS_LOG,
1811 (void) printf("\t%*s%s%s\n", indent, "", name,
1812 is_log ? " [log]" : "");
1814 if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN,
1815 &child, &children) != 0)
1818 for (c = 0; c < children; c++) {
1819 vname = zpool_vdev_name(hdl, NULL, child[c], VDEV_NAME_TYPE_ID);
1820 print_vdev_tree(hdl, vname, child[c], indent + 2);
1826 zpool_print_unsup_feat(nvlist_t *config)
1828 nvlist_t *nvinfo, *unsup_feat;
1831 verify(nvlist_lookup_nvlist(config, ZPOOL_CONFIG_LOAD_INFO, &nvinfo) ==
1833 verify(nvlist_lookup_nvlist(nvinfo, ZPOOL_CONFIG_UNSUP_FEAT,
1836 for (nvp = nvlist_next_nvpair(unsup_feat, NULL); nvp != NULL;
1837 nvp = nvlist_next_nvpair(unsup_feat, nvp)) {
1840 verify(nvpair_type(nvp) == DATA_TYPE_STRING);
1841 verify(nvpair_value_string(nvp, &desc) == 0);
1843 if (strlen(desc) > 0)
1844 (void) printf("\t%s (%s)\n", nvpair_name(nvp), desc);
1846 (void) printf("\t%s\n", nvpair_name(nvp));
1851 * Import the given pool using the known configuration and a list of
1852 * properties to be set. The configuration should have come from
1853 * zpool_find_import(). The 'newname' parameters control whether the pool
1854 * is imported with a different name.
1857 zpool_import_props(libzfs_handle_t *hdl, nvlist_t *config, const char *newname,
1858 nvlist_t *props, int flags)
1860 zfs_cmd_t zc = {"\0"};
1861 zpool_load_policy_t policy;
1862 nvlist_t *nv = NULL;
1863 nvlist_t *nvinfo = NULL;
1864 nvlist_t *missing = NULL;
1871 verify(nvlist_lookup_string(config, ZPOOL_CONFIG_POOL_NAME,
1874 (void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN,
1875 "cannot import pool '%s'"), origname);
1877 if (newname != NULL) {
1878 if (!zpool_name_valid(hdl, B_FALSE, newname))
1879 return (zfs_error_fmt(hdl, EZFS_INVALIDNAME,
1880 dgettext(TEXT_DOMAIN, "cannot import '%s'"),
1882 thename = (char *)newname;
1887 if (props != NULL) {
1889 prop_flags_t flags = { .create = B_FALSE, .import = B_TRUE };
1891 verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_VERSION,
1894 if ((props = zpool_valid_proplist(hdl, origname,
1895 props, version, flags, errbuf)) == NULL)
1897 if (zcmd_write_src_nvlist(hdl, &zc, props) != 0) {
1904 (void) strlcpy(zc.zc_name, thename, sizeof (zc.zc_name));
1906 verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID,
1909 if (zcmd_write_conf_nvlist(hdl, &zc, config) != 0) {
1910 zcmd_free_nvlists(&zc);
1913 if (zcmd_alloc_dst_nvlist(hdl, &zc, zc.zc_nvlist_conf_size * 2) != 0) {
1914 zcmd_free_nvlists(&zc);
1918 zc.zc_cookie = flags;
1919 while ((ret = zfs_ioctl(hdl, ZFS_IOC_POOL_IMPORT, &zc)) != 0 &&
1921 if (zcmd_expand_dst_nvlist(hdl, &zc) != 0) {
1922 zcmd_free_nvlists(&zc);
1929 (void) zcmd_read_dst_nvlist(hdl, &zc, &nv);
1931 zcmd_free_nvlists(&zc);
1933 zpool_get_load_policy(config, &policy);
1940 * Dry-run failed, but we print out what success
1941 * looks like if we found a best txg
1943 if (policy.zlp_rewind & ZPOOL_TRY_REWIND) {
1944 zpool_rewind_exclaim(hdl, newname ? origname : thename,
1950 if (newname == NULL)
1951 (void) snprintf(desc, sizeof (desc),
1952 dgettext(TEXT_DOMAIN, "cannot import '%s'"),
1955 (void) snprintf(desc, sizeof (desc),
1956 dgettext(TEXT_DOMAIN, "cannot import '%s' as '%s'"),
1961 if (nv != NULL && nvlist_lookup_nvlist(nv,
1962 ZPOOL_CONFIG_LOAD_INFO, &nvinfo) == 0 &&
1963 nvlist_exists(nvinfo, ZPOOL_CONFIG_UNSUP_FEAT)) {
1964 (void) printf(dgettext(TEXT_DOMAIN, "This "
1965 "pool uses the following feature(s) not "
1966 "supported by this system:\n"));
1967 zpool_print_unsup_feat(nv);
1968 if (nvlist_exists(nvinfo,
1969 ZPOOL_CONFIG_CAN_RDONLY)) {
1970 (void) printf(dgettext(TEXT_DOMAIN,
1971 "All unsupported features are only "
1972 "required for writing to the pool."
1973 "\nThe pool can be imported using "
1974 "'-o readonly=on'.\n"));
1978 * Unsupported version.
1980 (void) zfs_error(hdl, EZFS_BADVERSION, desc);
1984 if (nv != NULL && nvlist_lookup_nvlist(nv,
1985 ZPOOL_CONFIG_LOAD_INFO, &nvinfo) == 0) {
1986 char *hostname = "<unknown>";
1987 uint64_t hostid = 0;
1988 mmp_state_t mmp_state;
1990 mmp_state = fnvlist_lookup_uint64(nvinfo,
1991 ZPOOL_CONFIG_MMP_STATE);
1993 if (nvlist_exists(nvinfo,
1994 ZPOOL_CONFIG_MMP_HOSTNAME))
1995 hostname = fnvlist_lookup_string(nvinfo,
1996 ZPOOL_CONFIG_MMP_HOSTNAME);
1998 if (nvlist_exists(nvinfo,
1999 ZPOOL_CONFIG_MMP_HOSTID))
2000 hostid = fnvlist_lookup_uint64(nvinfo,
2001 ZPOOL_CONFIG_MMP_HOSTID);
2003 if (mmp_state == MMP_STATE_ACTIVE) {
2004 (void) snprintf(aux, sizeof (aux),
2005 dgettext(TEXT_DOMAIN, "pool is imp"
2006 "orted on host '%s' (hostid=%lx).\n"
2007 "Export the pool on the other "
2008 "system, then run 'zpool import'."),
2009 hostname, (unsigned long) hostid);
2010 } else if (mmp_state == MMP_STATE_NO_HOSTID) {
2011 (void) snprintf(aux, sizeof (aux),
2012 dgettext(TEXT_DOMAIN, "pool has "
2013 "the multihost property on and "
2014 "the\nsystem's hostid is not set. "
2015 "Set a unique system hostid with "
2016 "the zgenhostid(8) command.\n"));
2019 (void) zfs_error_aux(hdl, aux);
2021 (void) zfs_error(hdl, EZFS_ACTIVE_POOL, desc);
2025 (void) zfs_error(hdl, EZFS_INVALCONFIG, desc);
2029 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2030 "one or more devices is read only"));
2031 (void) zfs_error(hdl, EZFS_BADDEV, desc);
2035 if (nv && nvlist_lookup_nvlist(nv,
2036 ZPOOL_CONFIG_LOAD_INFO, &nvinfo) == 0 &&
2037 nvlist_lookup_nvlist(nvinfo,
2038 ZPOOL_CONFIG_MISSING_DEVICES, &missing) == 0) {
2039 (void) printf(dgettext(TEXT_DOMAIN,
2040 "The devices below are missing or "
2041 "corrupted, use '-m' to import the pool "
2043 print_vdev_tree(hdl, NULL, missing, 2);
2044 (void) printf("\n");
2046 (void) zpool_standard_error(hdl, error, desc);
2050 (void) zpool_standard_error(hdl, error, desc);
2054 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2055 "one or more devices are already in use\n"));
2056 (void) zfs_error(hdl, EZFS_BADDEV, desc);
2059 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2060 "new name of at least one dataset is longer than "
2061 "the maximum allowable length"));
2062 (void) zfs_error(hdl, EZFS_NAMETOOLONG, desc);
2065 (void) zpool_standard_error(hdl, error, desc);
2066 zpool_explain_recover(hdl,
2067 newname ? origname : thename, -error, nv);
2074 zpool_handle_t *zhp;
2077 * This should never fail, but play it safe anyway.
2079 if (zpool_open_silent(hdl, thename, &zhp) != 0)
2081 else if (zhp != NULL)
2083 if (policy.zlp_rewind &
2084 (ZPOOL_DO_REWIND | ZPOOL_TRY_REWIND)) {
2085 zpool_rewind_exclaim(hdl, newname ? origname : thename,
2086 ((policy.zlp_rewind & ZPOOL_TRY_REWIND) != 0), nv);
2099 zpool_scan(zpool_handle_t *zhp, pool_scan_func_t func, pool_scrub_cmd_t cmd)
2101 zfs_cmd_t zc = {"\0"};
2104 libzfs_handle_t *hdl = zhp->zpool_hdl;
2106 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
2107 zc.zc_cookie = func;
2110 if (zfs_ioctl(hdl, ZFS_IOC_POOL_SCAN, &zc) == 0)
2115 /* ECANCELED on a scrub means we resumed a paused scrub */
2116 if (err == ECANCELED && func == POOL_SCAN_SCRUB &&
2117 cmd == POOL_SCRUB_NORMAL)
2120 if (err == ENOENT && func != POOL_SCAN_NONE && cmd == POOL_SCRUB_NORMAL)
2123 if (func == POOL_SCAN_SCRUB) {
2124 if (cmd == POOL_SCRUB_PAUSE) {
2125 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN,
2126 "cannot pause scrubbing %s"), zc.zc_name);
2128 assert(cmd == POOL_SCRUB_NORMAL);
2129 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN,
2130 "cannot scrub %s"), zc.zc_name);
2132 } else if (func == POOL_SCAN_NONE) {
2133 (void) snprintf(msg, sizeof (msg),
2134 dgettext(TEXT_DOMAIN, "cannot cancel scrubbing %s"),
2137 assert(!"unexpected result");
2142 pool_scan_stat_t *ps = NULL;
2145 verify(nvlist_lookup_nvlist(zhp->zpool_config,
2146 ZPOOL_CONFIG_VDEV_TREE, &nvroot) == 0);
2147 (void) nvlist_lookup_uint64_array(nvroot,
2148 ZPOOL_CONFIG_SCAN_STATS, (uint64_t **)&ps, &psc);
2149 if (ps && ps->pss_func == POOL_SCAN_SCRUB) {
2150 if (cmd == POOL_SCRUB_PAUSE)
2151 return (zfs_error(hdl, EZFS_SCRUB_PAUSED, msg));
2153 return (zfs_error(hdl, EZFS_SCRUBBING, msg));
2155 return (zfs_error(hdl, EZFS_RESILVERING, msg));
2157 } else if (err == ENOENT) {
2158 return (zfs_error(hdl, EZFS_NO_SCRUB, msg));
2160 return (zpool_standard_error(hdl, err, msg));
2165 * Find a vdev that matches the search criteria specified. We use the
2166 * the nvpair name to determine how we should look for the device.
2167 * 'avail_spare' is set to TRUE if the provided guid refers to an AVAIL
2168 * spare; but FALSE if its an INUSE spare.
2171 vdev_to_nvlist_iter(nvlist_t *nv, nvlist_t *search, boolean_t *avail_spare,
2172 boolean_t *l2cache, boolean_t *log)
2179 nvpair_t *pair = nvlist_next_nvpair(search, NULL);
2181 /* Nothing to look for */
2182 if (search == NULL || pair == NULL)
2185 /* Obtain the key we will use to search */
2186 srchkey = nvpair_name(pair);
2188 switch (nvpair_type(pair)) {
2189 case DATA_TYPE_UINT64:
2190 if (strcmp(srchkey, ZPOOL_CONFIG_GUID) == 0) {
2191 uint64_t srchval, theguid;
2193 verify(nvpair_value_uint64(pair, &srchval) == 0);
2194 verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID,
2196 if (theguid == srchval)
2201 case DATA_TYPE_STRING: {
2202 char *srchval, *val;
2204 verify(nvpair_value_string(pair, &srchval) == 0);
2205 if (nvlist_lookup_string(nv, srchkey, &val) != 0)
2209 * Search for the requested value. Special cases:
2211 * - ZPOOL_CONFIG_PATH for whole disk entries. These end in
2212 * "-part1", or "p1". The suffix is hidden from the user,
2213 * but included in the string, so this matches around it.
2214 * - ZPOOL_CONFIG_PATH for short names zfs_strcmp_shortname()
2215 * is used to check all possible expanded paths.
2216 * - looking for a top-level vdev name (i.e. ZPOOL_CONFIG_TYPE).
2218 * Otherwise, all other searches are simple string compares.
2220 if (strcmp(srchkey, ZPOOL_CONFIG_PATH) == 0) {
2221 uint64_t wholedisk = 0;
2223 (void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_WHOLE_DISK,
2225 if (zfs_strcmp_pathname(srchval, val, wholedisk) == 0)
2228 } else if (strcmp(srchkey, ZPOOL_CONFIG_TYPE) == 0 && val) {
2229 char *type, *idx, *end, *p;
2230 uint64_t id, vdev_id;
2233 * Determine our vdev type, keeping in mind
2234 * that the srchval is composed of a type and
2235 * vdev id pair (i.e. mirror-4).
2237 if ((type = strdup(srchval)) == NULL)
2240 if ((p = strrchr(type, '-')) == NULL) {
2248 * If the types don't match then keep looking.
2250 if (strncmp(val, type, strlen(val)) != 0) {
2255 verify(zpool_vdev_is_interior(type));
2256 verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_ID,
2260 vdev_id = strtoull(idx, &end, 10);
2267 * Now verify that we have the correct vdev id.
2276 if (strcmp(srchval, val) == 0)
2285 if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN,
2286 &child, &children) != 0)
2289 for (c = 0; c < children; c++) {
2290 if ((ret = vdev_to_nvlist_iter(child[c], search,
2291 avail_spare, l2cache, NULL)) != NULL) {
2293 * The 'is_log' value is only set for the toplevel
2294 * vdev, not the leaf vdevs. So we always lookup the
2295 * log device from the root of the vdev tree (where
2296 * 'log' is non-NULL).
2299 nvlist_lookup_uint64(child[c],
2300 ZPOOL_CONFIG_IS_LOG, &is_log) == 0 &&
2308 if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_SPARES,
2309 &child, &children) == 0) {
2310 for (c = 0; c < children; c++) {
2311 if ((ret = vdev_to_nvlist_iter(child[c], search,
2312 avail_spare, l2cache, NULL)) != NULL) {
2313 *avail_spare = B_TRUE;
2319 if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_L2CACHE,
2320 &child, &children) == 0) {
2321 for (c = 0; c < children; c++) {
2322 if ((ret = vdev_to_nvlist_iter(child[c], search,
2323 avail_spare, l2cache, NULL)) != NULL) {
2334 * Given a physical path or guid, find the associated vdev.
2337 zpool_find_vdev_by_physpath(zpool_handle_t *zhp, const char *ppath,
2338 boolean_t *avail_spare, boolean_t *l2cache, boolean_t *log)
2340 nvlist_t *search, *nvroot, *ret;
2344 verify(nvlist_alloc(&search, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2346 guid = strtoull(ppath, &end, 0);
2347 if (guid != 0 && *end == '\0') {
2348 verify(nvlist_add_uint64(search, ZPOOL_CONFIG_GUID, guid) == 0);
2350 verify(nvlist_add_string(search, ZPOOL_CONFIG_PHYS_PATH,
2354 verify(nvlist_lookup_nvlist(zhp->zpool_config, ZPOOL_CONFIG_VDEV_TREE,
2357 *avail_spare = B_FALSE;
2361 ret = vdev_to_nvlist_iter(nvroot, search, avail_spare, l2cache, log);
2362 nvlist_free(search);
2368 * Determine if we have an "interior" top-level vdev (i.e mirror/raidz).
2371 zpool_vdev_is_interior(const char *name)
2373 if (strncmp(name, VDEV_TYPE_RAIDZ, strlen(VDEV_TYPE_RAIDZ)) == 0 ||
2374 strncmp(name, VDEV_TYPE_SPARE, strlen(VDEV_TYPE_SPARE)) == 0 ||
2376 VDEV_TYPE_REPLACING, strlen(VDEV_TYPE_REPLACING)) == 0 ||
2377 strncmp(name, VDEV_TYPE_MIRROR, strlen(VDEV_TYPE_MIRROR)) == 0)
2383 zpool_find_vdev(zpool_handle_t *zhp, const char *path, boolean_t *avail_spare,
2384 boolean_t *l2cache, boolean_t *log)
2387 nvlist_t *nvroot, *search, *ret;
2390 verify(nvlist_alloc(&search, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2392 guid = strtoull(path, &end, 0);
2393 if (guid != 0 && *end == '\0') {
2394 verify(nvlist_add_uint64(search, ZPOOL_CONFIG_GUID, guid) == 0);
2395 } else if (zpool_vdev_is_interior(path)) {
2396 verify(nvlist_add_string(search, ZPOOL_CONFIG_TYPE, path) == 0);
2398 verify(nvlist_add_string(search, ZPOOL_CONFIG_PATH, path) == 0);
2401 verify(nvlist_lookup_nvlist(zhp->zpool_config, ZPOOL_CONFIG_VDEV_TREE,
2404 *avail_spare = B_FALSE;
2408 ret = vdev_to_nvlist_iter(nvroot, search, avail_spare, l2cache, log);
2409 nvlist_free(search);
2415 vdev_is_online(nvlist_t *nv)
2419 if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_OFFLINE, &ival) == 0 ||
2420 nvlist_lookup_uint64(nv, ZPOOL_CONFIG_FAULTED, &ival) == 0 ||
2421 nvlist_lookup_uint64(nv, ZPOOL_CONFIG_REMOVED, &ival) == 0)
2428 * Helper function for zpool_get_physpaths().
2431 vdev_get_one_physpath(nvlist_t *config, char *physpath, size_t physpath_size,
2432 size_t *bytes_written)
2434 size_t bytes_left, pos, rsz;
2438 if (nvlist_lookup_string(config, ZPOOL_CONFIG_PHYS_PATH,
2440 return (EZFS_NODEVICE);
2442 pos = *bytes_written;
2443 bytes_left = physpath_size - pos;
2444 format = (pos == 0) ? "%s" : " %s";
2446 rsz = snprintf(physpath + pos, bytes_left, format, tmppath);
2447 *bytes_written += rsz;
2449 if (rsz >= bytes_left) {
2450 /* if physpath was not copied properly, clear it */
2451 if (bytes_left != 0) {
2454 return (EZFS_NOSPC);
2460 vdev_get_physpaths(nvlist_t *nv, char *physpath, size_t phypath_size,
2461 size_t *rsz, boolean_t is_spare)
2466 if (nvlist_lookup_string(nv, ZPOOL_CONFIG_TYPE, &type) != 0)
2467 return (EZFS_INVALCONFIG);
2469 if (strcmp(type, VDEV_TYPE_DISK) == 0) {
2471 * An active spare device has ZPOOL_CONFIG_IS_SPARE set.
2472 * For a spare vdev, we only want to boot from the active
2477 (void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_IS_SPARE,
2480 return (EZFS_INVALCONFIG);
2483 if (vdev_is_online(nv)) {
2484 if ((ret = vdev_get_one_physpath(nv, physpath,
2485 phypath_size, rsz)) != 0)
2488 } else if (strcmp(type, VDEV_TYPE_MIRROR) == 0 ||
2489 strcmp(type, VDEV_TYPE_RAIDZ) == 0 ||
2490 strcmp(type, VDEV_TYPE_REPLACING) == 0 ||
2491 (is_spare = (strcmp(type, VDEV_TYPE_SPARE) == 0))) {
2496 if (nvlist_lookup_nvlist_array(nv,
2497 ZPOOL_CONFIG_CHILDREN, &child, &count) != 0)
2498 return (EZFS_INVALCONFIG);
2500 for (i = 0; i < count; i++) {
2501 ret = vdev_get_physpaths(child[i], physpath,
2502 phypath_size, rsz, is_spare);
2503 if (ret == EZFS_NOSPC)
2508 return (EZFS_POOL_INVALARG);
2512 * Get phys_path for a root pool config.
2513 * Return 0 on success; non-zero on failure.
2516 zpool_get_config_physpath(nvlist_t *config, char *physpath, size_t phypath_size)
2519 nvlist_t *vdev_root;
2526 if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE,
2528 return (EZFS_INVALCONFIG);
2530 if (nvlist_lookup_string(vdev_root, ZPOOL_CONFIG_TYPE, &type) != 0 ||
2531 nvlist_lookup_nvlist_array(vdev_root, ZPOOL_CONFIG_CHILDREN,
2532 &child, &count) != 0)
2533 return (EZFS_INVALCONFIG);
2536 * root pool can only have a single top-level vdev.
2538 if (strcmp(type, VDEV_TYPE_ROOT) != 0 || count != 1)
2539 return (EZFS_POOL_INVALARG);
2541 (void) vdev_get_physpaths(child[0], physpath, phypath_size, &rsz,
2544 /* No online devices */
2546 return (EZFS_NODEVICE);
2552 * Get phys_path for a root pool
2553 * Return 0 on success; non-zero on failure.
2556 zpool_get_physpath(zpool_handle_t *zhp, char *physpath, size_t phypath_size)
2558 return (zpool_get_config_physpath(zhp->zpool_config, physpath,
2563 * If the device has being dynamically expanded then we need to relabel
2564 * the disk to use the new unallocated space.
2567 zpool_relabel_disk(libzfs_handle_t *hdl, const char *path, const char *msg)
2571 if ((fd = open(path, O_RDWR|O_DIRECT)) < 0) {
2572 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "cannot "
2573 "relabel '%s': unable to open device: %d"), path, errno);
2574 return (zfs_error(hdl, EZFS_OPENFAILED, msg));
2578 * It's possible that we might encounter an error if the device
2579 * does not have any unallocated space left. If so, we simply
2580 * ignore that error and continue on.
2582 * Also, we don't call efi_rescan() - that would just return EBUSY.
2583 * The module will do it for us in vdev_disk_open().
2585 error = efi_use_whole_disk(fd);
2587 /* Flush the buffers to disk and invalidate the page cache. */
2589 (void) ioctl(fd, BLKFLSBUF);
2592 if (error && error != VT_ENOSPC) {
2593 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "cannot "
2594 "relabel '%s': unable to read disk capacity"), path);
2595 return (zfs_error(hdl, EZFS_NOCAP, msg));
2602 * Convert a vdev path to a GUID. Returns GUID or 0 on error.
2604 * If is_spare, is_l2cache, or is_log is non-NULL, then store within it
2605 * if the VDEV is a spare, l2cache, or log device. If they're NULL then
2609 zpool_vdev_path_to_guid_impl(zpool_handle_t *zhp, const char *path,
2610 boolean_t *is_spare, boolean_t *is_l2cache, boolean_t *is_log)
2613 boolean_t spare = B_FALSE, l2cache = B_FALSE, log = B_FALSE;
2616 if ((tgt = zpool_find_vdev(zhp, path, &spare, &l2cache,
2620 verify(nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID, &guid) == 0);
2621 if (is_spare != NULL)
2623 if (is_l2cache != NULL)
2624 *is_l2cache = l2cache;
2631 /* Convert a vdev path to a GUID. Returns GUID or 0 on error. */
2633 zpool_vdev_path_to_guid(zpool_handle_t *zhp, const char *path)
2635 return (zpool_vdev_path_to_guid_impl(zhp, path, NULL, NULL, NULL));
2639 * Bring the specified vdev online. The 'flags' parameter is a set of the
2640 * ZFS_ONLINE_* flags.
2643 zpool_vdev_online(zpool_handle_t *zhp, const char *path, int flags,
2644 vdev_state_t *newstate)
2646 zfs_cmd_t zc = {"\0"};
2650 boolean_t avail_spare, l2cache, islog;
2651 libzfs_handle_t *hdl = zhp->zpool_hdl;
2654 if (flags & ZFS_ONLINE_EXPAND) {
2655 (void) snprintf(msg, sizeof (msg),
2656 dgettext(TEXT_DOMAIN, "cannot expand %s"), path);
2658 (void) snprintf(msg, sizeof (msg),
2659 dgettext(TEXT_DOMAIN, "cannot online %s"), path);
2662 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
2663 if ((tgt = zpool_find_vdev(zhp, path, &avail_spare, &l2cache,
2665 return (zfs_error(hdl, EZFS_NODEVICE, msg));
2667 verify(nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID, &zc.zc_guid) == 0);
2670 return (zfs_error(hdl, EZFS_ISSPARE, msg));
2672 if ((flags & ZFS_ONLINE_EXPAND ||
2673 zpool_get_prop_int(zhp, ZPOOL_PROP_AUTOEXPAND, NULL)) &&
2674 nvlist_lookup_string(tgt, ZPOOL_CONFIG_PATH, &pathname) == 0) {
2675 uint64_t wholedisk = 0;
2677 (void) nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_WHOLE_DISK,
2681 * XXX - L2ARC 1.0 devices can't support expansion.
2684 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2685 "cannot expand cache devices"));
2686 return (zfs_error(hdl, EZFS_VDEVNOTSUP, msg));
2690 const char *fullpath = path;
2691 char buf[MAXPATHLEN];
2693 if (path[0] != '/') {
2694 error = zfs_resolve_shortname(path, buf,
2697 return (zfs_error(hdl, EZFS_NODEVICE,
2703 error = zpool_relabel_disk(hdl, fullpath, msg);
2709 zc.zc_cookie = VDEV_STATE_ONLINE;
2712 if (zfs_ioctl(hdl, ZFS_IOC_VDEV_SET_STATE, &zc) != 0) {
2713 if (errno == EINVAL) {
2714 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "was split "
2715 "from this pool into a new one. Use '%s' "
2716 "instead"), "zpool detach");
2717 return (zfs_error(hdl, EZFS_POSTSPLIT_ONLINE, msg));
2719 return (zpool_standard_error(hdl, errno, msg));
2722 *newstate = zc.zc_cookie;
2727 * Take the specified vdev offline
2730 zpool_vdev_offline(zpool_handle_t *zhp, const char *path, boolean_t istmp)
2732 zfs_cmd_t zc = {"\0"};
2735 boolean_t avail_spare, l2cache;
2736 libzfs_handle_t *hdl = zhp->zpool_hdl;
2738 (void) snprintf(msg, sizeof (msg),
2739 dgettext(TEXT_DOMAIN, "cannot offline %s"), path);
2741 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
2742 if ((tgt = zpool_find_vdev(zhp, path, &avail_spare, &l2cache,
2744 return (zfs_error(hdl, EZFS_NODEVICE, msg));
2746 verify(nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID, &zc.zc_guid) == 0);
2749 return (zfs_error(hdl, EZFS_ISSPARE, msg));
2751 zc.zc_cookie = VDEV_STATE_OFFLINE;
2752 zc.zc_obj = istmp ? ZFS_OFFLINE_TEMPORARY : 0;
2754 if (zfs_ioctl(hdl, ZFS_IOC_VDEV_SET_STATE, &zc) == 0)
2761 * There are no other replicas of this device.
2763 return (zfs_error(hdl, EZFS_NOREPLICAS, msg));
2767 * The log device has unplayed logs
2769 return (zfs_error(hdl, EZFS_UNPLAYED_LOGS, msg));
2772 return (zpool_standard_error(hdl, errno, msg));
2777 * Mark the given vdev faulted.
2780 zpool_vdev_fault(zpool_handle_t *zhp, uint64_t guid, vdev_aux_t aux)
2782 zfs_cmd_t zc = {"\0"};
2784 libzfs_handle_t *hdl = zhp->zpool_hdl;
2786 (void) snprintf(msg, sizeof (msg),
2787 dgettext(TEXT_DOMAIN, "cannot fault %llu"), (u_longlong_t)guid);
2789 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
2791 zc.zc_cookie = VDEV_STATE_FAULTED;
2794 if (ioctl(hdl->libzfs_fd, ZFS_IOC_VDEV_SET_STATE, &zc) == 0)
2801 * There are no other replicas of this device.
2803 return (zfs_error(hdl, EZFS_NOREPLICAS, msg));
2806 return (zpool_standard_error(hdl, errno, msg));
2812 * Mark the given vdev degraded.
2815 zpool_vdev_degrade(zpool_handle_t *zhp, uint64_t guid, vdev_aux_t aux)
2817 zfs_cmd_t zc = {"\0"};
2819 libzfs_handle_t *hdl = zhp->zpool_hdl;
2821 (void) snprintf(msg, sizeof (msg),
2822 dgettext(TEXT_DOMAIN, "cannot degrade %llu"), (u_longlong_t)guid);
2824 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
2826 zc.zc_cookie = VDEV_STATE_DEGRADED;
2829 if (ioctl(hdl->libzfs_fd, ZFS_IOC_VDEV_SET_STATE, &zc) == 0)
2832 return (zpool_standard_error(hdl, errno, msg));
2836 * Returns TRUE if the given nvlist is a vdev that was originally swapped in as
2840 is_replacing_spare(nvlist_t *search, nvlist_t *tgt, int which)
2846 if (nvlist_lookup_nvlist_array(search, ZPOOL_CONFIG_CHILDREN, &child,
2848 verify(nvlist_lookup_string(search, ZPOOL_CONFIG_TYPE,
2851 if (strcmp(type, VDEV_TYPE_SPARE) == 0 &&
2852 children == 2 && child[which] == tgt)
2855 for (c = 0; c < children; c++)
2856 if (is_replacing_spare(child[c], tgt, which))
2864 * Attach new_disk (fully described by nvroot) to old_disk.
2865 * If 'replacing' is specified, the new disk will replace the old one.
2868 zpool_vdev_attach(zpool_handle_t *zhp,
2869 const char *old_disk, const char *new_disk, nvlist_t *nvroot, int replacing)
2871 zfs_cmd_t zc = {"\0"};
2875 boolean_t avail_spare, l2cache, islog;
2880 nvlist_t *config_root;
2881 libzfs_handle_t *hdl = zhp->zpool_hdl;
2882 boolean_t rootpool = zpool_is_bootable(zhp);
2885 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN,
2886 "cannot replace %s with %s"), old_disk, new_disk);
2888 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN,
2889 "cannot attach %s to %s"), new_disk, old_disk);
2891 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
2892 if ((tgt = zpool_find_vdev(zhp, old_disk, &avail_spare, &l2cache,
2894 return (zfs_error(hdl, EZFS_NODEVICE, msg));
2897 return (zfs_error(hdl, EZFS_ISSPARE, msg));
2900 return (zfs_error(hdl, EZFS_ISL2CACHE, msg));
2902 verify(nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID, &zc.zc_guid) == 0);
2903 zc.zc_cookie = replacing;
2905 if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_CHILDREN,
2906 &child, &children) != 0 || children != 1) {
2907 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2908 "new device must be a single disk"));
2909 return (zfs_error(hdl, EZFS_INVALCONFIG, msg));
2912 verify(nvlist_lookup_nvlist(zpool_get_config(zhp, NULL),
2913 ZPOOL_CONFIG_VDEV_TREE, &config_root) == 0);
2915 if ((newname = zpool_vdev_name(NULL, NULL, child[0], 0)) == NULL)
2919 * If the target is a hot spare that has been swapped in, we can only
2920 * replace it with another hot spare.
2923 nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_IS_SPARE, &val) == 0 &&
2924 (zpool_find_vdev(zhp, newname, &avail_spare, &l2cache,
2925 NULL) == NULL || !avail_spare) &&
2926 is_replacing_spare(config_root, tgt, 1)) {
2927 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2928 "can only be replaced by another hot spare"));
2930 return (zfs_error(hdl, EZFS_BADTARGET, msg));
2935 if (zcmd_write_conf_nvlist(hdl, &zc, nvroot) != 0)
2938 ret = zfs_ioctl(hdl, ZFS_IOC_VDEV_ATTACH, &zc);
2940 zcmd_free_nvlists(&zc);
2945 * XXX need a better way to prevent user from
2946 * booting up a half-baked vdev.
2948 (void) fprintf(stderr, dgettext(TEXT_DOMAIN, "Make "
2949 "sure to wait until resilver is done "
2950 "before rebooting.\n"));
2958 * Can't attach to or replace this type of vdev.
2961 uint64_t version = zpool_get_prop_int(zhp,
2962 ZPOOL_PROP_VERSION, NULL);
2965 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2966 "cannot replace a log with a spare"));
2967 else if (version >= SPA_VERSION_MULTI_REPLACE)
2968 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2969 "already in replacing/spare config; wait "
2970 "for completion or use 'zpool detach'"));
2972 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2973 "cannot replace a replacing device"));
2975 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2976 "can only attach to mirrors and top-level "
2979 (void) zfs_error(hdl, EZFS_BADTARGET, msg);
2984 * The new device must be a single disk.
2986 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2987 "new device must be a single disk"));
2988 (void) zfs_error(hdl, EZFS_INVALCONFIG, msg);
2992 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "%s is busy, "
2993 "or device removal is in progress"),
2995 (void) zfs_error(hdl, EZFS_BADDEV, msg);
3000 * The new device is too small.
3002 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3003 "device is too small"));
3004 (void) zfs_error(hdl, EZFS_BADDEV, msg);
3009 * The new device has a different optimal sector size.
3011 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3012 "new device has a different optimal sector size; use the "
3013 "option '-o ashift=N' to override the optimal size"));
3014 (void) zfs_error(hdl, EZFS_BADDEV, msg);
3019 * The resulting top-level vdev spec won't fit in the label.
3021 (void) zfs_error(hdl, EZFS_DEVOVERFLOW, msg);
3025 (void) zpool_standard_error(hdl, errno, msg);
3032 * Detach the specified device.
3035 zpool_vdev_detach(zpool_handle_t *zhp, const char *path)
3037 zfs_cmd_t zc = {"\0"};
3040 boolean_t avail_spare, l2cache;
3041 libzfs_handle_t *hdl = zhp->zpool_hdl;
3043 (void) snprintf(msg, sizeof (msg),
3044 dgettext(TEXT_DOMAIN, "cannot detach %s"), path);
3046 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
3047 if ((tgt = zpool_find_vdev(zhp, path, &avail_spare, &l2cache,
3049 return (zfs_error(hdl, EZFS_NODEVICE, msg));
3052 return (zfs_error(hdl, EZFS_ISSPARE, msg));
3055 return (zfs_error(hdl, EZFS_ISL2CACHE, msg));
3057 verify(nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID, &zc.zc_guid) == 0);
3059 if (zfs_ioctl(hdl, ZFS_IOC_VDEV_DETACH, &zc) == 0)
3066 * Can't detach from this type of vdev.
3068 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "only "
3069 "applicable to mirror and replacing vdevs"));
3070 (void) zfs_error(hdl, EZFS_BADTARGET, msg);
3075 * There are no other replicas of this device.
3077 (void) zfs_error(hdl, EZFS_NOREPLICAS, msg);
3081 (void) zpool_standard_error(hdl, errno, msg);
3088 * Find a mirror vdev in the source nvlist.
3090 * The mchild array contains a list of disks in one of the top-level mirrors
3091 * of the source pool. The schild array contains a list of disks that the
3092 * user specified on the command line. We loop over the mchild array to
3093 * see if any entry in the schild array matches.
3095 * If a disk in the mchild array is found in the schild array, we return
3096 * the index of that entry. Otherwise we return -1.
3099 find_vdev_entry(zpool_handle_t *zhp, nvlist_t **mchild, uint_t mchildren,
3100 nvlist_t **schild, uint_t schildren)
3104 for (mc = 0; mc < mchildren; mc++) {
3106 char *mpath = zpool_vdev_name(zhp->zpool_hdl, zhp,
3109 for (sc = 0; sc < schildren; sc++) {
3110 char *spath = zpool_vdev_name(zhp->zpool_hdl, zhp,
3112 boolean_t result = (strcmp(mpath, spath) == 0);
3128 * Split a mirror pool. If newroot points to null, then a new nvlist
3129 * is generated and it is the responsibility of the caller to free it.
3132 zpool_vdev_split(zpool_handle_t *zhp, char *newname, nvlist_t **newroot,
3133 nvlist_t *props, splitflags_t flags)
3135 zfs_cmd_t zc = {"\0"};
3137 nvlist_t *tree, *config, **child, **newchild, *newconfig = NULL;
3138 nvlist_t **varray = NULL, *zc_props = NULL;
3139 uint_t c, children, newchildren, lastlog = 0, vcount, found = 0;
3140 libzfs_handle_t *hdl = zhp->zpool_hdl;
3141 uint64_t vers, readonly = B_FALSE;
3142 boolean_t freelist = B_FALSE, memory_err = B_TRUE;
3145 (void) snprintf(msg, sizeof (msg),
3146 dgettext(TEXT_DOMAIN, "Unable to split %s"), zhp->zpool_name);
3148 if (!zpool_name_valid(hdl, B_FALSE, newname))
3149 return (zfs_error(hdl, EZFS_INVALIDNAME, msg));
3151 if ((config = zpool_get_config(zhp, NULL)) == NULL) {
3152 (void) fprintf(stderr, gettext("Internal error: unable to "
3153 "retrieve pool configuration\n"));
3157 verify(nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE, &tree)
3159 verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_VERSION, &vers) == 0);
3162 prop_flags_t flags = { .create = B_FALSE, .import = B_TRUE };
3163 if ((zc_props = zpool_valid_proplist(hdl, zhp->zpool_name,
3164 props, vers, flags, msg)) == NULL)
3166 (void) nvlist_lookup_uint64(zc_props,
3167 zpool_prop_to_name(ZPOOL_PROP_READONLY), &readonly);
3169 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3170 "property %s can only be set at import time"),
3171 zpool_prop_to_name(ZPOOL_PROP_READONLY));
3176 if (nvlist_lookup_nvlist_array(tree, ZPOOL_CONFIG_CHILDREN, &child,
3178 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3179 "Source pool is missing vdev tree"));
3180 nvlist_free(zc_props);
3184 varray = zfs_alloc(hdl, children * sizeof (nvlist_t *));
3187 if (*newroot == NULL ||
3188 nvlist_lookup_nvlist_array(*newroot, ZPOOL_CONFIG_CHILDREN,
3189 &newchild, &newchildren) != 0)
3192 for (c = 0; c < children; c++) {
3193 uint64_t is_log = B_FALSE, is_hole = B_FALSE;
3195 nvlist_t **mchild, *vdev;
3200 * Unlike cache & spares, slogs are stored in the
3201 * ZPOOL_CONFIG_CHILDREN array. We filter them out here.
3203 (void) nvlist_lookup_uint64(child[c], ZPOOL_CONFIG_IS_LOG,
3205 (void) nvlist_lookup_uint64(child[c], ZPOOL_CONFIG_IS_HOLE,
3207 if (is_log || is_hole) {
3209 * Create a hole vdev and put it in the config.
3211 if (nvlist_alloc(&vdev, NV_UNIQUE_NAME, 0) != 0)
3213 if (nvlist_add_string(vdev, ZPOOL_CONFIG_TYPE,
3214 VDEV_TYPE_HOLE) != 0)
3216 if (nvlist_add_uint64(vdev, ZPOOL_CONFIG_IS_HOLE,
3221 varray[vcount++] = vdev;
3225 verify(nvlist_lookup_string(child[c], ZPOOL_CONFIG_TYPE, &type)
3227 if (strcmp(type, VDEV_TYPE_MIRROR) != 0) {
3228 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3229 "Source pool must be composed only of mirrors\n"));
3230 retval = zfs_error(hdl, EZFS_INVALCONFIG, msg);
3234 verify(nvlist_lookup_nvlist_array(child[c],
3235 ZPOOL_CONFIG_CHILDREN, &mchild, &mchildren) == 0);
3237 /* find or add an entry for this top-level vdev */
3238 if (newchildren > 0 &&
3239 (entry = find_vdev_entry(zhp, mchild, mchildren,
3240 newchild, newchildren)) >= 0) {
3241 /* We found a disk that the user specified. */
3242 vdev = mchild[entry];
3245 /* User didn't specify a disk for this vdev. */
3246 vdev = mchild[mchildren - 1];
3249 if (nvlist_dup(vdev, &varray[vcount++], 0) != 0)
3253 /* did we find every disk the user specified? */
3254 if (found != newchildren) {
3255 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "Device list must "
3256 "include at most one disk from each mirror"));
3257 retval = zfs_error(hdl, EZFS_INVALCONFIG, msg);
3261 /* Prepare the nvlist for populating. */
3262 if (*newroot == NULL) {
3263 if (nvlist_alloc(newroot, NV_UNIQUE_NAME, 0) != 0)
3266 if (nvlist_add_string(*newroot, ZPOOL_CONFIG_TYPE,
3267 VDEV_TYPE_ROOT) != 0)
3270 verify(nvlist_remove_all(*newroot, ZPOOL_CONFIG_CHILDREN) == 0);
3273 /* Add all the children we found */
3274 if (nvlist_add_nvlist_array(*newroot, ZPOOL_CONFIG_CHILDREN, varray,
3275 lastlog == 0 ? vcount : lastlog) != 0)
3279 * If we're just doing a dry run, exit now with success.
3282 memory_err = B_FALSE;
3287 /* now build up the config list & call the ioctl */
3288 if (nvlist_alloc(&newconfig, NV_UNIQUE_NAME, 0) != 0)
3291 if (nvlist_add_nvlist(newconfig,
3292 ZPOOL_CONFIG_VDEV_TREE, *newroot) != 0 ||
3293 nvlist_add_string(newconfig,
3294 ZPOOL_CONFIG_POOL_NAME, newname) != 0 ||
3295 nvlist_add_uint64(newconfig, ZPOOL_CONFIG_VERSION, vers) != 0)
3299 * The new pool is automatically part of the namespace unless we
3300 * explicitly export it.
3303 zc.zc_cookie = ZPOOL_EXPORT_AFTER_SPLIT;
3304 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
3305 (void) strlcpy(zc.zc_string, newname, sizeof (zc.zc_string));
3306 if (zcmd_write_conf_nvlist(hdl, &zc, newconfig) != 0)
3308 if (zc_props != NULL && zcmd_write_src_nvlist(hdl, &zc, zc_props) != 0)
3311 if (zfs_ioctl(hdl, ZFS_IOC_VDEV_SPLIT, &zc) != 0) {
3312 retval = zpool_standard_error(hdl, errno, msg);
3317 memory_err = B_FALSE;
3320 if (varray != NULL) {
3323 for (v = 0; v < vcount; v++)
3324 nvlist_free(varray[v]);
3327 zcmd_free_nvlists(&zc);
3328 nvlist_free(zc_props);
3329 nvlist_free(newconfig);
3331 nvlist_free(*newroot);
3339 return (no_memory(hdl));
3345 * Remove the given device.
3348 zpool_vdev_remove(zpool_handle_t *zhp, const char *path)
3350 zfs_cmd_t zc = {"\0"};
3353 boolean_t avail_spare, l2cache, islog;
3354 libzfs_handle_t *hdl = zhp->zpool_hdl;
3357 (void) snprintf(msg, sizeof (msg),
3358 dgettext(TEXT_DOMAIN, "cannot remove %s"), path);
3360 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
3361 if ((tgt = zpool_find_vdev(zhp, path, &avail_spare, &l2cache,
3363 return (zfs_error(hdl, EZFS_NODEVICE, msg));
3365 version = zpool_get_prop_int(zhp, ZPOOL_PROP_VERSION, NULL);
3366 if (islog && version < SPA_VERSION_HOLES) {
3367 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3368 "pool must be upgraded to support log removal"));
3369 return (zfs_error(hdl, EZFS_BADVERSION, msg));
3372 if (!islog && !avail_spare && !l2cache && zpool_is_bootable(zhp)) {
3373 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3374 "root pool can not have removed devices, "
3375 "because GRUB does not understand them"));
3376 return (zfs_error(hdl, EINVAL, msg));
3379 zc.zc_guid = fnvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID);
3381 if (zfs_ioctl(hdl, ZFS_IOC_VDEV_REMOVE, &zc) == 0)
3387 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3388 "invalid config; all top-level vdevs must "
3389 "have the same sector size and not be raidz."));
3390 (void) zfs_error(hdl, EZFS_INVALCONFIG, msg);
3395 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3396 "Mount encrypted datasets to replay logs."));
3398 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3399 "Pool busy; removal may already be in progress"));
3401 (void) zfs_error(hdl, EZFS_BUSY, msg);
3406 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3407 "Mount encrypted datasets to replay logs."));
3408 (void) zfs_error(hdl, EZFS_BUSY, msg);
3410 (void) zpool_standard_error(hdl, errno, msg);
3415 (void) zpool_standard_error(hdl, errno, msg);
3421 zpool_vdev_remove_cancel(zpool_handle_t *zhp)
3425 libzfs_handle_t *hdl = zhp->zpool_hdl;
3427 (void) snprintf(msg, sizeof (msg),
3428 dgettext(TEXT_DOMAIN, "cannot cancel removal"));
3430 bzero(&zc, sizeof (zc));
3431 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
3434 if (zfs_ioctl(hdl, ZFS_IOC_VDEV_REMOVE, &zc) == 0)
3437 return (zpool_standard_error(hdl, errno, msg));
3441 zpool_vdev_indirect_size(zpool_handle_t *zhp, const char *path,
3446 boolean_t avail_spare, l2cache, islog;
3447 libzfs_handle_t *hdl = zhp->zpool_hdl;
3449 (void) snprintf(msg, sizeof (msg),
3450 dgettext(TEXT_DOMAIN, "cannot determine indirect size of %s"),
3453 if ((tgt = zpool_find_vdev(zhp, path, &avail_spare, &l2cache,
3455 return (zfs_error(hdl, EZFS_NODEVICE, msg));
3457 if (avail_spare || l2cache || islog) {
3462 if (nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_INDIRECT_SIZE, sizep) != 0) {
3463 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3464 "indirect size not available"));
3465 return (zfs_error(hdl, EINVAL, msg));
3471 * Clear the errors for the pool, or the particular device if specified.
3474 zpool_clear(zpool_handle_t *zhp, const char *path, nvlist_t *rewindnvl)
3476 zfs_cmd_t zc = {"\0"};
3479 zpool_load_policy_t policy;
3480 boolean_t avail_spare, l2cache;
3481 libzfs_handle_t *hdl = zhp->zpool_hdl;
3482 nvlist_t *nvi = NULL;
3486 (void) snprintf(msg, sizeof (msg),
3487 dgettext(TEXT_DOMAIN, "cannot clear errors for %s"),
3490 (void) snprintf(msg, sizeof (msg),
3491 dgettext(TEXT_DOMAIN, "cannot clear errors for %s"),
3494 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
3496 if ((tgt = zpool_find_vdev(zhp, path, &avail_spare,
3497 &l2cache, NULL)) == NULL)
3498 return (zfs_error(hdl, EZFS_NODEVICE, msg));
3501 * Don't allow error clearing for hot spares. Do allow
3502 * error clearing for l2cache devices.
3505 return (zfs_error(hdl, EZFS_ISSPARE, msg));
3507 verify(nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID,
3511 zpool_get_load_policy(rewindnvl, &policy);
3512 zc.zc_cookie = policy.zlp_rewind;
3514 if (zcmd_alloc_dst_nvlist(hdl, &zc, zhp->zpool_config_size * 2) != 0)
3517 if (zcmd_write_src_nvlist(hdl, &zc, rewindnvl) != 0)
3520 while ((error = zfs_ioctl(hdl, ZFS_IOC_CLEAR, &zc)) != 0 &&
3522 if (zcmd_expand_dst_nvlist(hdl, &zc) != 0) {
3523 zcmd_free_nvlists(&zc);
3528 if (!error || ((policy.zlp_rewind & ZPOOL_TRY_REWIND) &&
3529 errno != EPERM && errno != EACCES)) {
3530 if (policy.zlp_rewind &
3531 (ZPOOL_DO_REWIND | ZPOOL_TRY_REWIND)) {
3532 (void) zcmd_read_dst_nvlist(hdl, &zc, &nvi);
3533 zpool_rewind_exclaim(hdl, zc.zc_name,
3534 ((policy.zlp_rewind & ZPOOL_TRY_REWIND) != 0),
3538 zcmd_free_nvlists(&zc);
3542 zcmd_free_nvlists(&zc);
3543 return (zpool_standard_error(hdl, errno, msg));
3547 * Similar to zpool_clear(), but takes a GUID (used by fmd).
3550 zpool_vdev_clear(zpool_handle_t *zhp, uint64_t guid)
3552 zfs_cmd_t zc = {"\0"};
3554 libzfs_handle_t *hdl = zhp->zpool_hdl;
3556 (void) snprintf(msg, sizeof (msg),
3557 dgettext(TEXT_DOMAIN, "cannot clear errors for %llx"),
3558 (u_longlong_t)guid);
3560 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
3562 zc.zc_cookie = ZPOOL_NO_REWIND;
3564 if (ioctl(hdl->libzfs_fd, ZFS_IOC_CLEAR, &zc) == 0)
3567 return (zpool_standard_error(hdl, errno, msg));
3571 * Change the GUID for a pool.
3574 zpool_reguid(zpool_handle_t *zhp)
3577 libzfs_handle_t *hdl = zhp->zpool_hdl;
3578 zfs_cmd_t zc = {"\0"};
3580 (void) snprintf(msg, sizeof (msg),
3581 dgettext(TEXT_DOMAIN, "cannot reguid '%s'"), zhp->zpool_name);
3583 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
3584 if (zfs_ioctl(hdl, ZFS_IOC_POOL_REGUID, &zc) == 0)
3587 return (zpool_standard_error(hdl, errno, msg));
3594 zpool_reopen_one(zpool_handle_t *zhp, void *data)
3596 libzfs_handle_t *hdl = zpool_get_handle(zhp);
3597 const char *pool_name = zpool_get_name(zhp);
3598 boolean_t *scrub_restart = data;
3601 error = lzc_reopen(pool_name, *scrub_restart);
3603 return (zpool_standard_error_fmt(hdl, error,
3604 dgettext(TEXT_DOMAIN, "cannot reopen '%s'"), pool_name));
3610 /* call into libzfs_core to execute the sync IOCTL per pool */
3612 zpool_sync_one(zpool_handle_t *zhp, void *data)
3615 libzfs_handle_t *hdl = zpool_get_handle(zhp);
3616 const char *pool_name = zpool_get_name(zhp);
3617 boolean_t *force = data;
3618 nvlist_t *innvl = fnvlist_alloc();
3620 fnvlist_add_boolean_value(innvl, "force", *force);
3621 if ((ret = lzc_sync(pool_name, innvl, NULL)) != 0) {
3623 return (zpool_standard_error_fmt(hdl, ret,
3624 dgettext(TEXT_DOMAIN, "sync '%s' failed"), pool_name));
3631 #if defined(__sun__) || defined(__sun)
3633 * Convert from a devid string to a path.
3636 devid_to_path(char *devid_str)
3641 devid_nmlist_t *list = NULL;
3644 if (devid_str_decode(devid_str, &devid, &minor) != 0)
3647 ret = devid_deviceid_to_nmlist("/dev", devid, minor, &list);
3649 devid_str_free(minor);
3656 * In a case the strdup() fails, we will just return NULL below.
3658 path = strdup(list[0].devname);
3660 devid_free_nmlist(list);
3666 * Convert from a path to a devid string.
3669 path_to_devid(const char *path)
3675 if ((fd = open(path, O_RDONLY)) < 0)
3680 if (devid_get(fd, &devid) == 0) {
3681 if (devid_get_minor_name(fd, &minor) == 0)
3682 ret = devid_str_encode(devid, minor);
3684 devid_str_free(minor);
3693 * Issue the necessary ioctl() to update the stored path value for the vdev. We
3694 * ignore any failure here, since a common case is for an unprivileged user to
3695 * type 'zpool status', and we'll display the correct information anyway.
3698 set_path(zpool_handle_t *zhp, nvlist_t *nv, const char *path)
3700 zfs_cmd_t zc = {"\0"};
3702 (void) strncpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
3703 (void) strncpy(zc.zc_value, path, sizeof (zc.zc_value));
3704 verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID,
3707 (void) ioctl(zhp->zpool_hdl->libzfs_fd, ZFS_IOC_VDEV_SETPATH, &zc);
3711 #define PATH_BUF_LEN 64
3714 * Given a vdev, return the name to display in iostat. If the vdev has a path,
3715 * we use that, stripping off any leading "/dev/dsk/"; if not, we use the type.
3716 * We also check if this is a whole disk, in which case we strip off the
3717 * trailing 's0' slice name.
3719 * This routine is also responsible for identifying when disks have been
3720 * reconfigured in a new location. The kernel will have opened the device by
3721 * devid, but the path will still refer to the old location. To catch this, we
3722 * first do a path -> devid translation (which is fast for the common case). If
3723 * the devid matches, we're done. If not, we do a reverse devid -> path
3724 * translation and issue the appropriate ioctl() to update the path of the vdev.
3725 * If 'zhp' is NULL, then this is an exported pool, and we don't need to do any
3729 zpool_vdev_name(libzfs_handle_t *hdl, zpool_handle_t *zhp, nvlist_t *nv,
3732 char *path, *type, *env;
3734 char buf[PATH_BUF_LEN];
3735 char tmpbuf[PATH_BUF_LEN];
3738 * vdev_name will be "root"/"root-0" for the root vdev, but it is the
3739 * zpool name that will be displayed to the user.
3741 verify(nvlist_lookup_string(nv, ZPOOL_CONFIG_TYPE, &type) == 0);
3742 if (zhp != NULL && strcmp(type, "root") == 0)
3743 return (zfs_strdup(hdl, zpool_get_name(zhp)));
3745 env = getenv("ZPOOL_VDEV_NAME_PATH");
3746 if (env && (strtoul(env, NULL, 0) > 0 ||
3747 !strncasecmp(env, "YES", 3) || !strncasecmp(env, "ON", 2)))
3748 name_flags |= VDEV_NAME_PATH;
3750 env = getenv("ZPOOL_VDEV_NAME_GUID");
3751 if (env && (strtoul(env, NULL, 0) > 0 ||
3752 !strncasecmp(env, "YES", 3) || !strncasecmp(env, "ON", 2)))
3753 name_flags |= VDEV_NAME_GUID;
3755 env = getenv("ZPOOL_VDEV_NAME_FOLLOW_LINKS");
3756 if (env && (strtoul(env, NULL, 0) > 0 ||
3757 !strncasecmp(env, "YES", 3) || !strncasecmp(env, "ON", 2)))
3758 name_flags |= VDEV_NAME_FOLLOW_LINKS;
3760 if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_NOT_PRESENT, &value) == 0 ||
3761 name_flags & VDEV_NAME_GUID) {
3762 (void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID, &value);
3763 (void) snprintf(buf, sizeof (buf), "%llu", (u_longlong_t)value);
3765 } else if (nvlist_lookup_string(nv, ZPOOL_CONFIG_PATH, &path) == 0) {
3766 #if defined(__sun__) || defined(__sun)
3768 * Live VDEV path updates to a kernel VDEV during a
3769 * zpool_vdev_name lookup are not supported on Linux.
3776 * If the device is dead (faulted, offline, etc) then don't
3777 * bother opening it. Otherwise we may be forcing the user to
3778 * open a misbehaving device, which can have undesirable
3781 if ((nvlist_lookup_uint64_array(nv, ZPOOL_CONFIG_VDEV_STATS,
3782 (uint64_t **)&vs, &vsc) != 0 ||
3783 vs->vs_state >= VDEV_STATE_DEGRADED) &&
3785 nvlist_lookup_string(nv, ZPOOL_CONFIG_DEVID, &devid) == 0) {
3787 * Determine if the current path is correct.
3789 char *newdevid = path_to_devid(path);
3791 if (newdevid == NULL ||
3792 strcmp(devid, newdevid) != 0) {
3795 if ((newpath = devid_to_path(devid)) != NULL) {
3797 * Update the path appropriately.
3799 set_path(zhp, nv, newpath);
3800 if (nvlist_add_string(nv,
3801 ZPOOL_CONFIG_PATH, newpath) == 0)
3802 verify(nvlist_lookup_string(nv,
3810 devid_str_free(newdevid);
3814 if (name_flags & VDEV_NAME_FOLLOW_LINKS) {
3815 char *rp = realpath(path, NULL);
3817 strlcpy(buf, rp, sizeof (buf));
3824 * For a block device only use the name.
3826 if ((strcmp(type, VDEV_TYPE_DISK) == 0) &&
3827 !(name_flags & VDEV_NAME_PATH)) {
3828 path = strrchr(path, '/');
3833 * Remove the partition from the path it this is a whole disk.
3835 if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_WHOLE_DISK, &value)
3836 == 0 && value && !(name_flags & VDEV_NAME_PATH)) {
3837 return (zfs_strip_partition(path));
3843 * If it's a raidz device, we need to stick in the parity level.
3845 if (strcmp(path, VDEV_TYPE_RAIDZ) == 0) {
3846 verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_NPARITY,
3848 (void) snprintf(buf, sizeof (buf), "%s%llu", path,
3849 (u_longlong_t)value);
3854 * We identify each top-level vdev by using a <type-id>
3855 * naming convention.
3857 if (name_flags & VDEV_NAME_TYPE_ID) {
3859 verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_ID,
3861 (void) snprintf(tmpbuf, sizeof (tmpbuf), "%s-%llu",
3862 path, (u_longlong_t)id);
3867 return (zfs_strdup(hdl, path));
3871 zbookmark_mem_compare(const void *a, const void *b)
3873 return (memcmp(a, b, sizeof (zbookmark_phys_t)));
3877 * Retrieve the persistent error log, uniquify the members, and return to the
3881 zpool_get_errlog(zpool_handle_t *zhp, nvlist_t **nverrlistp)
3883 zfs_cmd_t zc = {"\0"};
3884 libzfs_handle_t *hdl = zhp->zpool_hdl;
3886 zbookmark_phys_t *zb = NULL;
3890 * Retrieve the raw error list from the kernel. If the number of errors
3891 * has increased, allocate more space and continue until we get the
3894 verify(nvlist_lookup_uint64(zhp->zpool_config, ZPOOL_CONFIG_ERRCOUNT,
3898 zc.zc_nvlist_dst = (uintptr_t)zfs_alloc(zhp->zpool_hdl,
3899 count * sizeof (zbookmark_phys_t));
3900 zc.zc_nvlist_dst_size = count;
3901 (void) strcpy(zc.zc_name, zhp->zpool_name);
3903 if (ioctl(zhp->zpool_hdl->libzfs_fd, ZFS_IOC_ERROR_LOG,
3905 free((void *)(uintptr_t)zc.zc_nvlist_dst);
3906 if (errno == ENOMEM) {
3909 count = zc.zc_nvlist_dst_size;
3910 dst = zfs_alloc(zhp->zpool_hdl, count *
3911 sizeof (zbookmark_phys_t));
3912 zc.zc_nvlist_dst = (uintptr_t)dst;
3914 return (zpool_standard_error_fmt(hdl, errno,
3915 dgettext(TEXT_DOMAIN, "errors: List of "
3916 "errors unavailable")));
3924 * Sort the resulting bookmarks. This is a little confusing due to the
3925 * implementation of ZFS_IOC_ERROR_LOG. The bookmarks are copied last
3926 * to first, and 'zc_nvlist_dst_size' indicates the number of boomarks
3927 * _not_ copied as part of the process. So we point the start of our
3928 * array appropriate and decrement the total number of elements.
3930 zb = ((zbookmark_phys_t *)(uintptr_t)zc.zc_nvlist_dst) +
3931 zc.zc_nvlist_dst_size;
3932 count -= zc.zc_nvlist_dst_size;
3934 qsort(zb, count, sizeof (zbookmark_phys_t), zbookmark_mem_compare);
3936 verify(nvlist_alloc(nverrlistp, 0, KM_SLEEP) == 0);
3939 * Fill in the nverrlistp with nvlist's of dataset and object numbers.
3941 for (i = 0; i < count; i++) {
3944 /* ignoring zb_blkid and zb_level for now */
3945 if (i > 0 && zb[i-1].zb_objset == zb[i].zb_objset &&
3946 zb[i-1].zb_object == zb[i].zb_object)
3949 if (nvlist_alloc(&nv, NV_UNIQUE_NAME, KM_SLEEP) != 0)
3951 if (nvlist_add_uint64(nv, ZPOOL_ERR_DATASET,
3952 zb[i].zb_objset) != 0) {
3956 if (nvlist_add_uint64(nv, ZPOOL_ERR_OBJECT,
3957 zb[i].zb_object) != 0) {
3961 if (nvlist_add_nvlist(*nverrlistp, "ejk", nv) != 0) {
3968 free((void *)(uintptr_t)zc.zc_nvlist_dst);
3972 free((void *)(uintptr_t)zc.zc_nvlist_dst);
3973 return (no_memory(zhp->zpool_hdl));
3977 * Upgrade a ZFS pool to the latest on-disk version.
3980 zpool_upgrade(zpool_handle_t *zhp, uint64_t new_version)
3982 zfs_cmd_t zc = {"\0"};
3983 libzfs_handle_t *hdl = zhp->zpool_hdl;
3985 (void) strcpy(zc.zc_name, zhp->zpool_name);
3986 zc.zc_cookie = new_version;
3988 if (zfs_ioctl(hdl, ZFS_IOC_POOL_UPGRADE, &zc) != 0)
3989 return (zpool_standard_error_fmt(hdl, errno,
3990 dgettext(TEXT_DOMAIN, "cannot upgrade '%s'"),
3996 zfs_save_arguments(int argc, char **argv, char *string, int len)
4000 (void) strlcpy(string, basename(argv[0]), len);
4001 for (i = 1; i < argc; i++) {
4002 (void) strlcat(string, " ", len);
4003 (void) strlcat(string, argv[i], len);
4008 zpool_log_history(libzfs_handle_t *hdl, const char *message)
4010 zfs_cmd_t zc = {"\0"};
4014 args = fnvlist_alloc();
4015 fnvlist_add_string(args, "message", message);
4016 err = zcmd_write_src_nvlist(hdl, &zc, args);
4018 err = ioctl(hdl->libzfs_fd, ZFS_IOC_LOG_HISTORY, &zc);
4020 zcmd_free_nvlists(&zc);
4025 * Perform ioctl to get some command history of a pool.
4027 * 'buf' is the buffer to fill up to 'len' bytes. 'off' is the
4028 * logical offset of the history buffer to start reading from.
4030 * Upon return, 'off' is the next logical offset to read from and
4031 * 'len' is the actual amount of bytes read into 'buf'.
4034 get_history(zpool_handle_t *zhp, char *buf, uint64_t *off, uint64_t *len)
4036 zfs_cmd_t zc = {"\0"};
4037 libzfs_handle_t *hdl = zhp->zpool_hdl;
4039 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
4041 zc.zc_history = (uint64_t)(uintptr_t)buf;
4042 zc.zc_history_len = *len;
4043 zc.zc_history_offset = *off;
4045 if (ioctl(hdl->libzfs_fd, ZFS_IOC_POOL_GET_HISTORY, &zc) != 0) {
4048 return (zfs_error_fmt(hdl, EZFS_PERM,
4049 dgettext(TEXT_DOMAIN,
4050 "cannot show history for pool '%s'"),
4053 return (zfs_error_fmt(hdl, EZFS_NOHISTORY,
4054 dgettext(TEXT_DOMAIN, "cannot get history for pool "
4055 "'%s'"), zhp->zpool_name));
4057 return (zfs_error_fmt(hdl, EZFS_BADVERSION,
4058 dgettext(TEXT_DOMAIN, "cannot get history for pool "
4059 "'%s', pool must be upgraded"), zhp->zpool_name));
4061 return (zpool_standard_error_fmt(hdl, errno,
4062 dgettext(TEXT_DOMAIN,
4063 "cannot get history for '%s'"), zhp->zpool_name));
4067 *len = zc.zc_history_len;
4068 *off = zc.zc_history_offset;
4074 * Retrieve the command history of a pool.
4077 zpool_get_history(zpool_handle_t *zhp, nvlist_t **nvhisp)
4080 int buflen = 128 * 1024;
4082 nvlist_t **records = NULL;
4083 uint_t numrecords = 0;
4086 buf = malloc(buflen);
4090 uint64_t bytes_read = buflen;
4093 if ((err = get_history(zhp, buf, &off, &bytes_read)) != 0)
4096 /* if nothing else was read in, we're at EOF, just return */
4100 if ((err = zpool_history_unpack(buf, bytes_read,
4101 &leftover, &records, &numrecords)) != 0)
4104 if (leftover == bytes_read) {
4106 * no progress made, because buffer is not big enough
4107 * to hold this record; resize and retry.
4111 buf = malloc(buflen);
4122 verify(nvlist_alloc(nvhisp, NV_UNIQUE_NAME, 0) == 0);
4123 verify(nvlist_add_nvlist_array(*nvhisp, ZPOOL_HIST_RECORD,
4124 records, numrecords) == 0);
4126 for (i = 0; i < numrecords; i++)
4127 nvlist_free(records[i]);
4134 * Retrieve the next event given the passed 'zevent_fd' file descriptor.
4135 * If there is a new event available 'nvp' will contain a newly allocated
4136 * nvlist and 'dropped' will be set to the number of missed events since
4137 * the last call to this function. When 'nvp' is set to NULL it indicates
4138 * no new events are available. In either case the function returns 0 and
4139 * it is up to the caller to free 'nvp'. In the case of a fatal error the
4140 * function will return a non-zero value. When the function is called in
4141 * blocking mode (the default, unless the ZEVENT_NONBLOCK flag is passed),
4142 * it will not return until a new event is available.
4145 zpool_events_next(libzfs_handle_t *hdl, nvlist_t **nvp,
4146 int *dropped, unsigned flags, int zevent_fd)
4148 zfs_cmd_t zc = {"\0"};
4153 zc.zc_cleanup_fd = zevent_fd;
4155 if (flags & ZEVENT_NONBLOCK)
4156 zc.zc_guid = ZEVENT_NONBLOCK;
4158 if (zcmd_alloc_dst_nvlist(hdl, &zc, ZEVENT_SIZE) != 0)
4162 if (zfs_ioctl(hdl, ZFS_IOC_EVENTS_NEXT, &zc) != 0) {
4165 error = zfs_error_fmt(hdl, EZFS_POOLUNAVAIL,
4166 dgettext(TEXT_DOMAIN, "zfs shutdown"));
4169 /* Blocking error case should not occur */
4170 if (!(flags & ZEVENT_NONBLOCK))
4171 error = zpool_standard_error_fmt(hdl, errno,
4172 dgettext(TEXT_DOMAIN, "cannot get event"));
4176 if (zcmd_expand_dst_nvlist(hdl, &zc) != 0) {
4177 error = zfs_error_fmt(hdl, EZFS_NOMEM,
4178 dgettext(TEXT_DOMAIN, "cannot get event"));
4184 error = zpool_standard_error_fmt(hdl, errno,
4185 dgettext(TEXT_DOMAIN, "cannot get event"));
4190 error = zcmd_read_dst_nvlist(hdl, &zc, nvp);
4194 *dropped = (int)zc.zc_cookie;
4196 zcmd_free_nvlists(&zc);
4205 zpool_events_clear(libzfs_handle_t *hdl, int *count)
4207 zfs_cmd_t zc = {"\0"};
4210 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN,
4211 "cannot clear events"));
4213 if (zfs_ioctl(hdl, ZFS_IOC_EVENTS_CLEAR, &zc) != 0)
4214 return (zpool_standard_error_fmt(hdl, errno, msg));
4217 *count = (int)zc.zc_cookie; /* # of events cleared */
4223 * Seek to a specific EID, ZEVENT_SEEK_START, or ZEVENT_SEEK_END for
4224 * the passed zevent_fd file handle. On success zero is returned,
4225 * otherwise -1 is returned and hdl->libzfs_error is set to the errno.
4228 zpool_events_seek(libzfs_handle_t *hdl, uint64_t eid, int zevent_fd)
4230 zfs_cmd_t zc = {"\0"};
4234 zc.zc_cleanup_fd = zevent_fd;
4236 if (zfs_ioctl(hdl, ZFS_IOC_EVENTS_SEEK, &zc) != 0) {
4239 error = zfs_error_fmt(hdl, EZFS_NOENT,
4240 dgettext(TEXT_DOMAIN, "cannot get event"));
4244 error = zfs_error_fmt(hdl, EZFS_NOMEM,
4245 dgettext(TEXT_DOMAIN, "cannot get event"));
4249 error = zpool_standard_error_fmt(hdl, errno,
4250 dgettext(TEXT_DOMAIN, "cannot get event"));
4259 zpool_obj_to_path(zpool_handle_t *zhp, uint64_t dsobj, uint64_t obj,
4260 char *pathname, size_t len)
4262 zfs_cmd_t zc = {"\0"};
4263 boolean_t mounted = B_FALSE;
4264 char *mntpnt = NULL;
4265 char dsname[ZFS_MAX_DATASET_NAME_LEN];
4268 /* special case for the MOS */
4269 (void) snprintf(pathname, len, "<metadata>:<0x%llx>",
4274 /* get the dataset's name */
4275 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
4277 if (ioctl(zhp->zpool_hdl->libzfs_fd,
4278 ZFS_IOC_DSOBJ_TO_DSNAME, &zc) != 0) {
4279 /* just write out a path of two object numbers */
4280 (void) snprintf(pathname, len, "<0x%llx>:<0x%llx>",
4281 (longlong_t)dsobj, (longlong_t)obj);
4284 (void) strlcpy(dsname, zc.zc_value, sizeof (dsname));
4286 /* find out if the dataset is mounted */
4287 mounted = is_mounted(zhp->zpool_hdl, dsname, &mntpnt);
4289 /* get the corrupted object's path */
4290 (void) strlcpy(zc.zc_name, dsname, sizeof (zc.zc_name));
4292 if (ioctl(zhp->zpool_hdl->libzfs_fd, ZFS_IOC_OBJ_TO_PATH,
4295 (void) snprintf(pathname, len, "%s%s", mntpnt,
4298 (void) snprintf(pathname, len, "%s:%s",
4299 dsname, zc.zc_value);
4302 (void) snprintf(pathname, len, "%s:<0x%llx>", dsname,
4309 * Read the EFI label from the config, if a label does not exist then
4310 * pass back the error to the caller. If the caller has passed a non-NULL
4311 * diskaddr argument then we set it to the starting address of the EFI
4315 read_efi_label(nvlist_t *config, diskaddr_t *sb)
4319 char diskname[MAXPATHLEN];
4322 if (nvlist_lookup_string(config, ZPOOL_CONFIG_PATH, &path) != 0)
4325 (void) snprintf(diskname, sizeof (diskname), "%s%s", DISK_ROOT,
4326 strrchr(path, '/'));
4327 if ((fd = open(diskname, O_RDONLY|O_DIRECT)) >= 0) {
4328 struct dk_gpt *vtoc;
4330 if ((err = efi_alloc_and_read(fd, &vtoc)) >= 0) {
4332 *sb = vtoc->efi_parts[0].p_start;
4341 * determine where a partition starts on a disk in the current
4345 find_start_block(nvlist_t *config)
4349 diskaddr_t sb = MAXOFFSET_T;
4352 if (nvlist_lookup_nvlist_array(config,
4353 ZPOOL_CONFIG_CHILDREN, &child, &children) != 0) {
4354 if (nvlist_lookup_uint64(config,
4355 ZPOOL_CONFIG_WHOLE_DISK,
4356 &wholedisk) != 0 || !wholedisk) {
4357 return (MAXOFFSET_T);
4359 if (read_efi_label(config, &sb) < 0)
4364 for (c = 0; c < children; c++) {
4365 sb = find_start_block(child[c]);
4366 if (sb != MAXOFFSET_T) {
4370 return (MAXOFFSET_T);
4374 zpool_label_disk_check(char *path)
4376 struct dk_gpt *vtoc;
4379 if ((fd = open(path, O_RDONLY|O_DIRECT)) < 0)
4382 if ((err = efi_alloc_and_read(fd, &vtoc)) != 0) {
4387 if (vtoc->efi_flags & EFI_GPT_PRIMARY_CORRUPT) {
4399 * Generate a unique partition name for the ZFS member. Partitions must
4400 * have unique names to ensure udev will be able to create symlinks under
4401 * /dev/disk/by-partlabel/ for all pool members. The partition names are
4402 * of the form <pool>-<unique-id>.
4405 zpool_label_name(char *label_name, int label_size)
4410 fd = open("/dev/urandom", O_RDONLY);
4412 if (read(fd, &id, sizeof (id)) != sizeof (id))
4419 id = (((uint64_t)rand()) << 32) | (uint64_t)rand();
4421 snprintf(label_name, label_size, "zfs-%016llx", (u_longlong_t)id);
4425 * Label an individual disk. The name provided is the short name,
4426 * stripped of any leading /dev path.
4429 zpool_label_disk(libzfs_handle_t *hdl, zpool_handle_t *zhp, char *name)
4431 char path[MAXPATHLEN];
4432 struct dk_gpt *vtoc;
4434 size_t resv = EFI_MIN_RESV_SIZE;
4435 uint64_t slice_size;
4436 diskaddr_t start_block;
4439 /* prepare an error message just in case */
4440 (void) snprintf(errbuf, sizeof (errbuf),
4441 dgettext(TEXT_DOMAIN, "cannot label '%s'"), name);
4446 verify(nvlist_lookup_nvlist(zhp->zpool_config,
4447 ZPOOL_CONFIG_VDEV_TREE, &nvroot) == 0);
4449 if (zhp->zpool_start_block == 0)
4450 start_block = find_start_block(nvroot);
4452 start_block = zhp->zpool_start_block;
4453 zhp->zpool_start_block = start_block;
4456 start_block = NEW_START_BLOCK;
4459 (void) snprintf(path, sizeof (path), "%s/%s", DISK_ROOT, name);
4461 if ((fd = open(path, O_RDWR|O_DIRECT|O_EXCL)) < 0) {
4463 * This shouldn't happen. We've long since verified that this
4464 * is a valid device.
4466 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "cannot "
4467 "label '%s': unable to open device: %d"), path, errno);
4468 return (zfs_error(hdl, EZFS_OPENFAILED, errbuf));
4471 if (efi_alloc_and_init(fd, EFI_NUMPAR, &vtoc) != 0) {
4473 * The only way this can fail is if we run out of memory, or we
4474 * were unable to read the disk's capacity
4476 if (errno == ENOMEM)
4477 (void) no_memory(hdl);
4480 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "cannot "
4481 "label '%s': unable to read disk capacity"), path);
4483 return (zfs_error(hdl, EZFS_NOCAP, errbuf));
4486 slice_size = vtoc->efi_last_u_lba + 1;
4487 slice_size -= EFI_MIN_RESV_SIZE;
4488 if (start_block == MAXOFFSET_T)
4489 start_block = NEW_START_BLOCK;
4490 slice_size -= start_block;
4491 slice_size = P2ALIGN(slice_size, PARTITION_END_ALIGNMENT);
4493 vtoc->efi_parts[0].p_start = start_block;
4494 vtoc->efi_parts[0].p_size = slice_size;
4497 * Why we use V_USR: V_BACKUP confuses users, and is considered
4498 * disposable by some EFI utilities (since EFI doesn't have a backup
4499 * slice). V_UNASSIGNED is supposed to be used only for zero size
4500 * partitions, and efi_write() will fail if we use it. V_ROOT, V_BOOT,
4501 * etc. were all pretty specific. V_USR is as close to reality as we
4502 * can get, in the absence of V_OTHER.
4504 vtoc->efi_parts[0].p_tag = V_USR;
4505 zpool_label_name(vtoc->efi_parts[0].p_name, EFI_PART_NAME_LEN);
4507 vtoc->efi_parts[8].p_start = slice_size + start_block;
4508 vtoc->efi_parts[8].p_size = resv;
4509 vtoc->efi_parts[8].p_tag = V_RESERVED;
4511 rval = efi_write(fd, vtoc);
4513 /* Flush the buffers to disk and invalidate the page cache. */
4515 (void) ioctl(fd, BLKFLSBUF);
4518 rval = efi_rescan(fd);
4521 * Some block drivers (like pcata) may not support EFI GPT labels.
4522 * Print out a helpful error message directing the user to manually
4523 * label the disk and give a specific slice.
4529 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "try using "
4530 "parted(8) and then provide a specific slice: %d"), rval);
4531 return (zfs_error(hdl, EZFS_LABELFAILED, errbuf));
4537 (void) snprintf(path, sizeof (path), "%s/%s", DISK_ROOT, name);
4538 (void) zfs_append_partition(path, MAXPATHLEN);
4540 /* Wait to udev to signal use the device has settled. */
4541 rval = zpool_label_disk_wait(path, DISK_LABEL_WAIT);
4543 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "failed to "
4544 "detect device partitions on '%s': %d"), path, rval);
4545 return (zfs_error(hdl, EZFS_LABELFAILED, errbuf));
4548 /* We can't be to paranoid. Read the label back and verify it. */
4549 (void) snprintf(path, sizeof (path), "%s/%s", DISK_ROOT, name);
4550 rval = zpool_label_disk_check(path);
4552 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "freshly written "
4553 "EFI label on '%s' is damaged. Ensure\nthis device "
4554 "is not in in use, and is functioning properly: %d"),
4556 return (zfs_error(hdl, EZFS_LABELFAILED, errbuf));