4 * The contents of this file are subject to the terms of the
5 * Common Development and Distribution License (the "License").
6 * You may not use this file except in compliance with the License.
8 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
9 * or http://www.opensolaris.org/os/licensing.
10 * See the License for the specific language governing permissions
11 * and limitations under the License.
13 * When distributing Covered Code, include this CDDL HEADER in each
14 * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
15 * If applicable, add the following below this CDDL HEADER, with the
16 * fields enclosed by brackets "[]" replaced with your own identifying
17 * information: Portions Copyright [yyyy] [name of copyright owner]
23 * Copyright 2015 Nexenta Systems, Inc. All rights reserved.
24 * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
25 * Copyright (c) 2011, 2014 by Delphix. All rights reserved.
40 #include <sys/efi_partition.h>
42 #include <sys/zfs_ioctl.h>
45 #include "zfs_namecheck.h"
47 #include "libzfs_impl.h"
48 #include "zfs_comutil.h"
49 #include "zfeature_common.h"
51 static int read_efi_label(nvlist_t
*config
, diskaddr_t
*sb
);
53 typedef struct prop_flags
{
54 int create
:1; /* Validate property on creation */
55 int import
:1; /* Validate property on import */
59 * ====================================================================
60 * zpool property functions
61 * ====================================================================
65 zpool_get_all_props(zpool_handle_t
*zhp
)
67 zfs_cmd_t zc
= {"\0"};
68 libzfs_handle_t
*hdl
= zhp
->zpool_hdl
;
70 (void) strlcpy(zc
.zc_name
, zhp
->zpool_name
, sizeof (zc
.zc_name
));
72 if (zcmd_alloc_dst_nvlist(hdl
, &zc
, 0) != 0)
75 while (ioctl(hdl
->libzfs_fd
, ZFS_IOC_POOL_GET_PROPS
, &zc
) != 0) {
76 if (errno
== ENOMEM
) {
77 if (zcmd_expand_dst_nvlist(hdl
, &zc
) != 0) {
78 zcmd_free_nvlists(&zc
);
82 zcmd_free_nvlists(&zc
);
87 if (zcmd_read_dst_nvlist(hdl
, &zc
, &zhp
->zpool_props
) != 0) {
88 zcmd_free_nvlists(&zc
);
92 zcmd_free_nvlists(&zc
);
98 zpool_props_refresh(zpool_handle_t
*zhp
)
102 old_props
= zhp
->zpool_props
;
104 if (zpool_get_all_props(zhp
) != 0)
107 nvlist_free(old_props
);
112 zpool_get_prop_string(zpool_handle_t
*zhp
, zpool_prop_t prop
,
118 zprop_source_t source
;
120 nvl
= zhp
->zpool_props
;
121 if (nvlist_lookup_nvlist(nvl
, zpool_prop_to_name(prop
), &nv
) == 0) {
122 verify(nvlist_lookup_uint64(nv
, ZPROP_SOURCE
, &ival
) == 0);
124 verify(nvlist_lookup_string(nv
, ZPROP_VALUE
, &value
) == 0);
126 source
= ZPROP_SRC_DEFAULT
;
127 if ((value
= (char *)zpool_prop_default_string(prop
)) == NULL
)
138 zpool_get_prop_int(zpool_handle_t
*zhp
, zpool_prop_t prop
, zprop_source_t
*src
)
142 zprop_source_t source
;
144 if (zhp
->zpool_props
== NULL
&& zpool_get_all_props(zhp
)) {
146 * zpool_get_all_props() has most likely failed because
147 * the pool is faulted, but if all we need is the top level
148 * vdev's guid then get it from the zhp config nvlist.
150 if ((prop
== ZPOOL_PROP_GUID
) &&
151 (nvlist_lookup_nvlist(zhp
->zpool_config
,
152 ZPOOL_CONFIG_VDEV_TREE
, &nv
) == 0) &&
153 (nvlist_lookup_uint64(nv
, ZPOOL_CONFIG_GUID
, &value
)
157 return (zpool_prop_default_numeric(prop
));
160 nvl
= zhp
->zpool_props
;
161 if (nvlist_lookup_nvlist(nvl
, zpool_prop_to_name(prop
), &nv
) == 0) {
162 verify(nvlist_lookup_uint64(nv
, ZPROP_SOURCE
, &value
) == 0);
164 verify(nvlist_lookup_uint64(nv
, ZPROP_VALUE
, &value
) == 0);
166 source
= ZPROP_SRC_DEFAULT
;
167 value
= zpool_prop_default_numeric(prop
);
177 * Map VDEV STATE to printed strings.
180 zpool_state_to_name(vdev_state_t state
, vdev_aux_t aux
)
185 case VDEV_STATE_CLOSED
:
186 case VDEV_STATE_OFFLINE
:
187 return (gettext("OFFLINE"));
188 case VDEV_STATE_REMOVED
:
189 return (gettext("REMOVED"));
190 case VDEV_STATE_CANT_OPEN
:
191 if (aux
== VDEV_AUX_CORRUPT_DATA
|| aux
== VDEV_AUX_BAD_LOG
)
192 return (gettext("FAULTED"));
193 else if (aux
== VDEV_AUX_SPLIT_POOL
)
194 return (gettext("SPLIT"));
196 return (gettext("UNAVAIL"));
197 case VDEV_STATE_FAULTED
:
198 return (gettext("FAULTED"));
199 case VDEV_STATE_DEGRADED
:
200 return (gettext("DEGRADED"));
201 case VDEV_STATE_HEALTHY
:
202 return (gettext("ONLINE"));
205 return (gettext("UNKNOWN"));
209 * Map POOL STATE to printed strings.
212 zpool_pool_state_to_name(pool_state_t state
)
217 case POOL_STATE_ACTIVE
:
218 return (gettext("ACTIVE"));
219 case POOL_STATE_EXPORTED
:
220 return (gettext("EXPORTED"));
221 case POOL_STATE_DESTROYED
:
222 return (gettext("DESTROYED"));
223 case POOL_STATE_SPARE
:
224 return (gettext("SPARE"));
225 case POOL_STATE_L2CACHE
:
226 return (gettext("L2CACHE"));
227 case POOL_STATE_UNINITIALIZED
:
228 return (gettext("UNINITIALIZED"));
229 case POOL_STATE_UNAVAIL
:
230 return (gettext("UNAVAIL"));
231 case POOL_STATE_POTENTIALLY_ACTIVE
:
232 return (gettext("POTENTIALLY_ACTIVE"));
235 return (gettext("UNKNOWN"));
239 * API compatibility wrapper around zpool_get_prop_literal
242 zpool_get_prop(zpool_handle_t
*zhp
, zpool_prop_t prop
, char *buf
, size_t len
,
243 zprop_source_t
*srctype
)
245 return (zpool_get_prop_literal(zhp
, prop
, buf
, len
, srctype
, B_FALSE
));
249 * Get a zpool property value for 'prop' and return the value in
250 * a pre-allocated buffer.
253 zpool_get_prop_literal(zpool_handle_t
*zhp
, zpool_prop_t prop
, char *buf
,
254 size_t len
, zprop_source_t
*srctype
, boolean_t literal
)
258 zprop_source_t src
= ZPROP_SRC_NONE
;
263 if (zpool_get_state(zhp
) == POOL_STATE_UNAVAIL
) {
265 case ZPOOL_PROP_NAME
:
266 (void) strlcpy(buf
, zpool_get_name(zhp
), len
);
269 case ZPOOL_PROP_HEALTH
:
270 (void) strlcpy(buf
, "FAULTED", len
);
273 case ZPOOL_PROP_GUID
:
274 intval
= zpool_get_prop_int(zhp
, prop
, &src
);
275 (void) snprintf(buf
, len
, "%llu", (u_longlong_t
)intval
);
278 case ZPOOL_PROP_ALTROOT
:
279 case ZPOOL_PROP_CACHEFILE
:
280 case ZPOOL_PROP_COMMENT
:
281 if (zhp
->zpool_props
!= NULL
||
282 zpool_get_all_props(zhp
) == 0) {
284 zpool_get_prop_string(zhp
, prop
, &src
),
292 (void) strlcpy(buf
, "-", len
);
301 if (zhp
->zpool_props
== NULL
&& zpool_get_all_props(zhp
) &&
302 prop
!= ZPOOL_PROP_NAME
)
305 switch (zpool_prop_get_type(prop
)) {
306 case PROP_TYPE_STRING
:
307 (void) strlcpy(buf
, zpool_get_prop_string(zhp
, prop
, &src
),
311 case PROP_TYPE_NUMBER
:
312 intval
= zpool_get_prop_int(zhp
, prop
, &src
);
315 case ZPOOL_PROP_SIZE
:
316 case ZPOOL_PROP_ALLOCATED
:
317 case ZPOOL_PROP_FREE
:
318 case ZPOOL_PROP_FREEING
:
319 case ZPOOL_PROP_LEAKED
:
320 case ZPOOL_PROP_ASHIFT
:
322 (void) snprintf(buf
, len
, "%llu",
323 (u_longlong_t
)intval
);
325 (void) zfs_nicenum(intval
, buf
, len
);
328 case ZPOOL_PROP_EXPANDSZ
:
330 (void) strlcpy(buf
, "-", len
);
331 } else if (literal
) {
332 (void) snprintf(buf
, len
, "%llu",
333 (u_longlong_t
)intval
);
335 (void) zfs_nicenum(intval
, buf
, len
);
339 case ZPOOL_PROP_CAPACITY
:
340 (void) snprintf(buf
, len
, "%llu%%",
341 (u_longlong_t
)intval
);
344 case ZPOOL_PROP_FRAGMENTATION
:
345 if (intval
== UINT64_MAX
) {
346 (void) strlcpy(buf
, "-", len
);
348 (void) snprintf(buf
, len
, "%llu%%",
349 (u_longlong_t
)intval
);
353 case ZPOOL_PROP_DEDUPRATIO
:
354 (void) snprintf(buf
, len
, "%llu.%02llux",
355 (u_longlong_t
)(intval
/ 100),
356 (u_longlong_t
)(intval
% 100));
359 case ZPOOL_PROP_HEALTH
:
360 verify(nvlist_lookup_nvlist(zpool_get_config(zhp
, NULL
),
361 ZPOOL_CONFIG_VDEV_TREE
, &nvroot
) == 0);
362 verify(nvlist_lookup_uint64_array(nvroot
,
363 ZPOOL_CONFIG_VDEV_STATS
, (uint64_t **)&vs
, &vsc
)
366 (void) strlcpy(buf
, zpool_state_to_name(intval
,
369 case ZPOOL_PROP_VERSION
:
370 if (intval
>= SPA_VERSION_FEATURES
) {
371 (void) snprintf(buf
, len
, "-");
376 (void) snprintf(buf
, len
, "%llu", (u_longlong_t
)intval
);
380 case PROP_TYPE_INDEX
:
381 intval
= zpool_get_prop_int(zhp
, prop
, &src
);
382 if (zpool_prop_index_to_string(prop
, intval
, &strval
)
385 (void) strlcpy(buf
, strval
, len
);
399 * Check if the bootfs name has the same pool name as it is set to.
400 * Assuming bootfs is a valid dataset name.
403 bootfs_name_valid(const char *pool
, char *bootfs
)
405 int len
= strlen(pool
);
407 if (!zfs_name_valid(bootfs
, ZFS_TYPE_FILESYSTEM
|ZFS_TYPE_SNAPSHOT
))
410 if (strncmp(pool
, bootfs
, len
) == 0 &&
411 (bootfs
[len
] == '/' || bootfs
[len
] == '\0'))
417 #if defined(__sun__) || defined(__sun)
419 * Inspect the configuration to determine if any of the devices contain
423 pool_uses_efi(nvlist_t
*config
)
428 if (nvlist_lookup_nvlist_array(config
, ZPOOL_CONFIG_CHILDREN
,
429 &child
, &children
) != 0)
430 return (read_efi_label(config
, NULL
) >= 0);
432 for (c
= 0; c
< children
; c
++) {
433 if (pool_uses_efi(child
[c
]))
441 zpool_is_bootable(zpool_handle_t
*zhp
)
443 char bootfs
[ZPOOL_MAXNAMELEN
];
445 return (zpool_get_prop(zhp
, ZPOOL_PROP_BOOTFS
, bootfs
,
446 sizeof (bootfs
), NULL
) == 0 && strncmp(bootfs
, "-",
447 sizeof (bootfs
)) != 0);
452 * Given an nvlist of zpool properties to be set, validate that they are
453 * correct, and parse any numeric properties (index, boolean, etc) if they are
454 * specified as strings.
457 zpool_valid_proplist(libzfs_handle_t
*hdl
, const char *poolname
,
458 nvlist_t
*props
, uint64_t version
, prop_flags_t flags
, char *errbuf
)
466 struct stat64 statbuf
;
470 if (nvlist_alloc(&retprops
, NV_UNIQUE_NAME
, 0) != 0) {
471 (void) no_memory(hdl
);
476 while ((elem
= nvlist_next_nvpair(props
, elem
)) != NULL
) {
477 const char *propname
= nvpair_name(elem
);
479 prop
= zpool_name_to_prop(propname
);
480 if (prop
== ZPROP_INVAL
&& zpool_prop_feature(propname
)) {
482 char *fname
= strchr(propname
, '@') + 1;
484 err
= zfeature_lookup_name(fname
, NULL
);
486 ASSERT3U(err
, ==, ENOENT
);
487 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
488 "invalid feature '%s'"), fname
);
489 (void) zfs_error(hdl
, EZFS_BADPROP
, errbuf
);
493 if (nvpair_type(elem
) != DATA_TYPE_STRING
) {
494 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
495 "'%s' must be a string"), propname
);
496 (void) zfs_error(hdl
, EZFS_BADPROP
, errbuf
);
500 (void) nvpair_value_string(elem
, &strval
);
501 if (strcmp(strval
, ZFS_FEATURE_ENABLED
) != 0) {
502 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
503 "property '%s' can only be set to "
504 "'enabled'"), propname
);
505 (void) zfs_error(hdl
, EZFS_BADPROP
, errbuf
);
509 if (nvlist_add_uint64(retprops
, propname
, 0) != 0) {
510 (void) no_memory(hdl
);
517 * Make sure this property is valid and applies to this type.
519 if (prop
== ZPROP_INVAL
) {
520 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
521 "invalid property '%s'"), propname
);
522 (void) zfs_error(hdl
, EZFS_BADPROP
, errbuf
);
526 if (zpool_prop_readonly(prop
)) {
527 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
, "'%s' "
528 "is readonly"), propname
);
529 (void) zfs_error(hdl
, EZFS_PROPREADONLY
, errbuf
);
533 if (zprop_parse_value(hdl
, elem
, prop
, ZFS_TYPE_POOL
, retprops
,
534 &strval
, &intval
, errbuf
) != 0)
538 * Perform additional checking for specific properties.
543 case ZPOOL_PROP_VERSION
:
544 if (intval
< version
||
545 !SPA_VERSION_IS_SUPPORTED(intval
)) {
546 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
547 "property '%s' number %d is invalid."),
549 (void) zfs_error(hdl
, EZFS_BADVERSION
, errbuf
);
554 case ZPOOL_PROP_ASHIFT
:
556 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
557 "property '%s' can only be set at "
558 "creation time"), propname
);
559 (void) zfs_error(hdl
, EZFS_BADPROP
, errbuf
);
563 if (intval
!= 0 && (intval
< 9 || intval
> 13)) {
564 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
565 "property '%s' number %d is invalid."),
567 (void) zfs_error(hdl
, EZFS_BADPROP
, errbuf
);
572 case ZPOOL_PROP_BOOTFS
:
573 if (flags
.create
|| flags
.import
) {
574 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
575 "property '%s' cannot be set at creation "
576 "or import time"), propname
);
577 (void) zfs_error(hdl
, EZFS_BADPROP
, errbuf
);
581 if (version
< SPA_VERSION_BOOTFS
) {
582 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
583 "pool must be upgraded to support "
584 "'%s' property"), propname
);
585 (void) zfs_error(hdl
, EZFS_BADVERSION
, errbuf
);
590 * bootfs property value has to be a dataset name and
591 * the dataset has to be in the same pool as it sets to.
593 if (strval
[0] != '\0' && !bootfs_name_valid(poolname
,
595 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
, "'%s' "
596 "is an invalid name"), strval
);
597 (void) zfs_error(hdl
, EZFS_INVALIDNAME
, errbuf
);
601 if ((zhp
= zpool_open_canfail(hdl
, poolname
)) == NULL
) {
602 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
603 "could not open pool '%s'"), poolname
);
604 (void) zfs_error(hdl
, EZFS_OPENFAILED
, errbuf
);
607 verify(nvlist_lookup_nvlist(zpool_get_config(zhp
, NULL
),
608 ZPOOL_CONFIG_VDEV_TREE
, &nvroot
) == 0);
610 #if defined(__sun__) || defined(__sun)
612 * bootfs property cannot be set on a disk which has
615 if (pool_uses_efi(nvroot
)) {
616 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
617 "property '%s' not supported on "
618 "EFI labeled devices"), propname
);
619 (void) zfs_error(hdl
, EZFS_POOL_NOTSUP
, errbuf
);
627 case ZPOOL_PROP_ALTROOT
:
628 if (!flags
.create
&& !flags
.import
) {
629 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
630 "property '%s' can only be set during pool "
631 "creation or import"), propname
);
632 (void) zfs_error(hdl
, EZFS_BADPROP
, errbuf
);
636 if (strval
[0] != '/') {
637 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
638 "bad alternate root '%s'"), strval
);
639 (void) zfs_error(hdl
, EZFS_BADPATH
, errbuf
);
644 case ZPOOL_PROP_CACHEFILE
:
645 if (strval
[0] == '\0')
648 if (strcmp(strval
, "none") == 0)
651 if (strval
[0] != '/') {
652 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
653 "property '%s' must be empty, an "
654 "absolute path, or 'none'"), propname
);
655 (void) zfs_error(hdl
, EZFS_BADPATH
, errbuf
);
659 slash
= strrchr(strval
, '/');
661 if (slash
[1] == '\0' || strcmp(slash
, "/.") == 0 ||
662 strcmp(slash
, "/..") == 0) {
663 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
664 "'%s' is not a valid file"), strval
);
665 (void) zfs_error(hdl
, EZFS_BADPATH
, errbuf
);
671 if (strval
[0] != '\0' &&
672 (stat64(strval
, &statbuf
) != 0 ||
673 !S_ISDIR(statbuf
.st_mode
))) {
674 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
675 "'%s' is not a valid directory"),
677 (void) zfs_error(hdl
, EZFS_BADPATH
, errbuf
);
684 case ZPOOL_PROP_COMMENT
:
685 for (check
= strval
; *check
!= '\0'; check
++) {
686 if (!isprint(*check
)) {
688 dgettext(TEXT_DOMAIN
,
689 "comment may only have printable "
691 (void) zfs_error(hdl
, EZFS_BADPROP
,
696 if (strlen(strval
) > ZPROP_MAX_COMMENT
) {
697 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
698 "comment must not exceed %d characters"),
700 (void) zfs_error(hdl
, EZFS_BADPROP
, errbuf
);
704 case ZPOOL_PROP_READONLY
:
706 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
707 "property '%s' can only be set at "
708 "import time"), propname
);
709 (void) zfs_error(hdl
, EZFS_BADPROP
, errbuf
);
713 case ZPOOL_PROP_TNAME
:
715 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
716 "property '%s' can only be set at "
717 "creation time"), propname
);
718 (void) zfs_error(hdl
, EZFS_BADPROP
, errbuf
);
727 nvlist_free(retprops
);
732 * Set zpool property : propname=propval.
735 zpool_set_prop(zpool_handle_t
*zhp
, const char *propname
, const char *propval
)
737 zfs_cmd_t zc
= {"\0"};
740 nvlist_t
*nvl
= NULL
;
743 prop_flags_t flags
= { 0 };
745 (void) snprintf(errbuf
, sizeof (errbuf
),
746 dgettext(TEXT_DOMAIN
, "cannot set property for '%s'"),
749 if (nvlist_alloc(&nvl
, NV_UNIQUE_NAME
, 0) != 0)
750 return (no_memory(zhp
->zpool_hdl
));
752 if (nvlist_add_string(nvl
, propname
, propval
) != 0) {
754 return (no_memory(zhp
->zpool_hdl
));
757 version
= zpool_get_prop_int(zhp
, ZPOOL_PROP_VERSION
, NULL
);
758 if ((realprops
= zpool_valid_proplist(zhp
->zpool_hdl
,
759 zhp
->zpool_name
, nvl
, version
, flags
, errbuf
)) == NULL
) {
768 * Execute the corresponding ioctl() to set this property.
770 (void) strlcpy(zc
.zc_name
, zhp
->zpool_name
, sizeof (zc
.zc_name
));
772 if (zcmd_write_src_nvlist(zhp
->zpool_hdl
, &zc
, nvl
) != 0) {
777 ret
= zfs_ioctl(zhp
->zpool_hdl
, ZFS_IOC_POOL_SET_PROPS
, &zc
);
779 zcmd_free_nvlists(&zc
);
783 (void) zpool_standard_error(zhp
->zpool_hdl
, errno
, errbuf
);
785 (void) zpool_props_refresh(zhp
);
791 zpool_expand_proplist(zpool_handle_t
*zhp
, zprop_list_t
**plp
)
793 libzfs_handle_t
*hdl
= zhp
->zpool_hdl
;
795 char buf
[ZFS_MAXPROPLEN
];
796 nvlist_t
*features
= NULL
;
799 boolean_t firstexpand
= (NULL
== *plp
);
802 if (zprop_expand_list(hdl
, plp
, ZFS_TYPE_POOL
) != 0)
806 while (*last
!= NULL
)
807 last
= &(*last
)->pl_next
;
810 features
= zpool_get_features(zhp
);
812 if ((*plp
)->pl_all
&& firstexpand
) {
813 for (i
= 0; i
< SPA_FEATURES
; i
++) {
814 zprop_list_t
*entry
= zfs_alloc(hdl
,
815 sizeof (zprop_list_t
));
816 entry
->pl_prop
= ZPROP_INVAL
;
817 entry
->pl_user_prop
= zfs_asprintf(hdl
, "feature@%s",
818 spa_feature_table
[i
].fi_uname
);
819 entry
->pl_width
= strlen(entry
->pl_user_prop
);
820 entry
->pl_all
= B_TRUE
;
823 last
= &entry
->pl_next
;
827 /* add any unsupported features */
828 for (nvp
= nvlist_next_nvpair(features
, NULL
);
829 nvp
!= NULL
; nvp
= nvlist_next_nvpair(features
, nvp
)) {
834 if (zfeature_is_supported(nvpair_name(nvp
)))
837 propname
= zfs_asprintf(hdl
, "unsupported@%s",
841 * Before adding the property to the list make sure that no
842 * other pool already added the same property.
846 while (entry
!= NULL
) {
847 if (entry
->pl_user_prop
!= NULL
&&
848 strcmp(propname
, entry
->pl_user_prop
) == 0) {
852 entry
= entry
->pl_next
;
859 entry
= zfs_alloc(hdl
, sizeof (zprop_list_t
));
860 entry
->pl_prop
= ZPROP_INVAL
;
861 entry
->pl_user_prop
= propname
;
862 entry
->pl_width
= strlen(entry
->pl_user_prop
);
863 entry
->pl_all
= B_TRUE
;
866 last
= &entry
->pl_next
;
869 for (entry
= *plp
; entry
!= NULL
; entry
= entry
->pl_next
) {
874 if (entry
->pl_prop
!= ZPROP_INVAL
&&
875 zpool_get_prop(zhp
, entry
->pl_prop
, buf
, sizeof (buf
),
877 if (strlen(buf
) > entry
->pl_width
)
878 entry
->pl_width
= strlen(buf
);
886 * Get the state for the given feature on the given ZFS pool.
889 zpool_prop_get_feature(zpool_handle_t
*zhp
, const char *propname
, char *buf
,
893 boolean_t found
= B_FALSE
;
894 nvlist_t
*features
= zpool_get_features(zhp
);
896 const char *feature
= strchr(propname
, '@') + 1;
898 supported
= zpool_prop_feature(propname
);
899 ASSERT(supported
|| zpool_prop_unsupported(propname
));
902 * Convert from feature name to feature guid. This conversion is
903 * unecessary for unsupported@... properties because they already
910 ret
= zfeature_lookup_name(feature
, &fid
);
912 (void) strlcpy(buf
, "-", len
);
915 feature
= spa_feature_table
[fid
].fi_guid
;
918 if (nvlist_lookup_uint64(features
, feature
, &refcount
) == 0)
923 (void) strlcpy(buf
, ZFS_FEATURE_DISABLED
, len
);
926 (void) strlcpy(buf
, ZFS_FEATURE_ENABLED
, len
);
928 (void) strlcpy(buf
, ZFS_FEATURE_ACTIVE
, len
);
933 (void) strcpy(buf
, ZFS_UNSUPPORTED_INACTIVE
);
935 (void) strcpy(buf
, ZFS_UNSUPPORTED_READONLY
);
938 (void) strlcpy(buf
, "-", len
);
947 * Don't start the slice at the default block of 34; many storage
948 * devices will use a stripe width of 128k, other vendors prefer a 1m
949 * alignment. It is best to play it safe and ensure a 1m alignment
950 * given 512B blocks. When the block size is larger by a power of 2
951 * we will still be 1m aligned. Some devices are sensitive to the
952 * partition ending alignment as well.
954 #define NEW_START_BLOCK 2048
955 #define PARTITION_END_ALIGNMENT 2048
958 * Validate the given pool name, optionally putting an extended error message in
962 zpool_name_valid(libzfs_handle_t
*hdl
, boolean_t isopen
, const char *pool
)
968 ret
= pool_namecheck(pool
, &why
, &what
);
971 * The rules for reserved pool names were extended at a later point.
972 * But we need to support users with existing pools that may now be
973 * invalid. So we only check for this expanded set of names during a
974 * create (or import), and only in userland.
976 if (ret
== 0 && !isopen
&&
977 (strncmp(pool
, "mirror", 6) == 0 ||
978 strncmp(pool
, "raidz", 5) == 0 ||
979 strncmp(pool
, "spare", 5) == 0 ||
980 strcmp(pool
, "log") == 0)) {
983 dgettext(TEXT_DOMAIN
, "name is reserved"));
991 case NAME_ERR_TOOLONG
:
993 dgettext(TEXT_DOMAIN
, "name is too long"));
996 case NAME_ERR_INVALCHAR
:
998 dgettext(TEXT_DOMAIN
, "invalid character "
999 "'%c' in pool name"), what
);
1002 case NAME_ERR_NOLETTER
:
1003 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
1004 "name must begin with a letter"));
1007 case NAME_ERR_RESERVED
:
1008 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
1009 "name is reserved"));
1012 case NAME_ERR_DISKLIKE
:
1013 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
1014 "pool name is reserved"));
1017 case NAME_ERR_LEADING_SLASH
:
1018 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
1019 "leading slash in name"));
1022 case NAME_ERR_EMPTY_COMPONENT
:
1023 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
1024 "empty component in name"));
1027 case NAME_ERR_TRAILING_SLASH
:
1028 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
1029 "trailing slash in name"));
1032 case NAME_ERR_MULTIPLE_AT
:
1033 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
1034 "multiple '@' delimiters in name"));
1036 case NAME_ERR_NO_AT
:
1037 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
1038 "permission set is missing '@'"));
1049 * Open a handle to the given pool, even if the pool is currently in the FAULTED
1053 zpool_open_canfail(libzfs_handle_t
*hdl
, const char *pool
)
1055 zpool_handle_t
*zhp
;
1059 * Make sure the pool name is valid.
1061 if (!zpool_name_valid(hdl
, B_TRUE
, pool
)) {
1062 (void) zfs_error_fmt(hdl
, EZFS_INVALIDNAME
,
1063 dgettext(TEXT_DOMAIN
, "cannot open '%s'"),
1068 if ((zhp
= zfs_alloc(hdl
, sizeof (zpool_handle_t
))) == NULL
)
1071 zhp
->zpool_hdl
= hdl
;
1072 (void) strlcpy(zhp
->zpool_name
, pool
, sizeof (zhp
->zpool_name
));
1074 if (zpool_refresh_stats(zhp
, &missing
) != 0) {
1080 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
, "no such pool"));
1081 (void) zfs_error_fmt(hdl
, EZFS_NOENT
,
1082 dgettext(TEXT_DOMAIN
, "cannot open '%s'"), pool
);
1091 * Like the above, but silent on error. Used when iterating over pools (because
1092 * the configuration cache may be out of date).
1095 zpool_open_silent(libzfs_handle_t
*hdl
, const char *pool
, zpool_handle_t
**ret
)
1097 zpool_handle_t
*zhp
;
1100 if ((zhp
= zfs_alloc(hdl
, sizeof (zpool_handle_t
))) == NULL
)
1103 zhp
->zpool_hdl
= hdl
;
1104 (void) strlcpy(zhp
->zpool_name
, pool
, sizeof (zhp
->zpool_name
));
1106 if (zpool_refresh_stats(zhp
, &missing
) != 0) {
1122 * Similar to zpool_open_canfail(), but refuses to open pools in the faulted
1126 zpool_open(libzfs_handle_t
*hdl
, const char *pool
)
1128 zpool_handle_t
*zhp
;
1130 if ((zhp
= zpool_open_canfail(hdl
, pool
)) == NULL
)
1133 if (zhp
->zpool_state
== POOL_STATE_UNAVAIL
) {
1134 (void) zfs_error_fmt(hdl
, EZFS_POOLUNAVAIL
,
1135 dgettext(TEXT_DOMAIN
, "cannot open '%s'"), zhp
->zpool_name
);
1144 * Close the handle. Simply frees the memory associated with the handle.
1147 zpool_close(zpool_handle_t
*zhp
)
1149 nvlist_free(zhp
->zpool_config
);
1150 nvlist_free(zhp
->zpool_old_config
);
1151 nvlist_free(zhp
->zpool_props
);
1156 * Return the name of the pool.
1159 zpool_get_name(zpool_handle_t
*zhp
)
1161 return (zhp
->zpool_name
);
1166 * Return the state of the pool (ACTIVE or UNAVAILABLE)
1169 zpool_get_state(zpool_handle_t
*zhp
)
1171 return (zhp
->zpool_state
);
1175 * Create the named pool, using the provided vdev list. It is assumed
1176 * that the consumer has already validated the contents of the nvlist, so we
1177 * don't have to worry about error semantics.
1180 zpool_create(libzfs_handle_t
*hdl
, const char *pool
, nvlist_t
*nvroot
,
1181 nvlist_t
*props
, nvlist_t
*fsprops
)
1183 zfs_cmd_t zc
= {"\0"};
1184 nvlist_t
*zc_fsprops
= NULL
;
1185 nvlist_t
*zc_props
= NULL
;
1189 (void) snprintf(msg
, sizeof (msg
), dgettext(TEXT_DOMAIN
,
1190 "cannot create '%s'"), pool
);
1192 if (!zpool_name_valid(hdl
, B_FALSE
, pool
))
1193 return (zfs_error(hdl
, EZFS_INVALIDNAME
, msg
));
1195 if (zcmd_write_conf_nvlist(hdl
, &zc
, nvroot
) != 0)
1199 prop_flags_t flags
= { .create
= B_TRUE
, .import
= B_FALSE
};
1201 if ((zc_props
= zpool_valid_proplist(hdl
, pool
, props
,
1202 SPA_VERSION_1
, flags
, msg
)) == NULL
) {
1211 zoned
= ((nvlist_lookup_string(fsprops
,
1212 zfs_prop_to_name(ZFS_PROP_ZONED
), &zonestr
) == 0) &&
1213 strcmp(zonestr
, "on") == 0);
1215 if ((zc_fsprops
= zfs_valid_proplist(hdl
, ZFS_TYPE_FILESYSTEM
,
1216 fsprops
, zoned
, NULL
, NULL
, msg
)) == NULL
) {
1220 (nvlist_alloc(&zc_props
, NV_UNIQUE_NAME
, 0) != 0)) {
1223 if (nvlist_add_nvlist(zc_props
,
1224 ZPOOL_ROOTFS_PROPS
, zc_fsprops
) != 0) {
1229 if (zc_props
&& zcmd_write_src_nvlist(hdl
, &zc
, zc_props
) != 0)
1232 (void) strlcpy(zc
.zc_name
, pool
, sizeof (zc
.zc_name
));
1234 if ((ret
= zfs_ioctl(hdl
, ZFS_IOC_POOL_CREATE
, &zc
)) != 0) {
1236 zcmd_free_nvlists(&zc
);
1237 nvlist_free(zc_props
);
1238 nvlist_free(zc_fsprops
);
1243 * This can happen if the user has specified the same
1244 * device multiple times. We can't reliably detect this
1245 * until we try to add it and see we already have a
1246 * label. This can also happen under if the device is
1247 * part of an active md or lvm device.
1249 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
1250 "one or more vdevs refer to the same device, or "
1251 "one of\nthe devices is part of an active md or "
1253 return (zfs_error(hdl
, EZFS_BADDEV
, msg
));
1257 * This happens if the record size is smaller or larger
1258 * than the allowed size range, or not a power of 2.
1260 * NOTE: although zfs_valid_proplist is called earlier,
1261 * this case may have slipped through since the
1262 * pool does not exist yet and it is therefore
1263 * impossible to read properties e.g. max blocksize
1266 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
1267 "record size invalid"));
1268 return (zfs_error(hdl
, EZFS_BADPROP
, msg
));
1272 * This occurs when one of the devices is below
1273 * SPA_MINDEVSIZE. Unfortunately, we can't detect which
1274 * device was the problem device since there's no
1275 * reliable way to determine device size from userland.
1280 zfs_nicenum(SPA_MINDEVSIZE
, buf
, sizeof (buf
));
1282 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
1283 "one or more devices is less than the "
1284 "minimum size (%s)"), buf
);
1286 return (zfs_error(hdl
, EZFS_BADDEV
, msg
));
1289 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
1290 "one or more devices is out of space"));
1291 return (zfs_error(hdl
, EZFS_BADDEV
, msg
));
1294 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
1295 "cache device must be a disk or disk slice"));
1296 return (zfs_error(hdl
, EZFS_BADDEV
, msg
));
1299 return (zpool_standard_error(hdl
, errno
, msg
));
1304 zcmd_free_nvlists(&zc
);
1305 nvlist_free(zc_props
);
1306 nvlist_free(zc_fsprops
);
1311 * Destroy the given pool. It is up to the caller to ensure that there are no
1312 * datasets left in the pool.
1315 zpool_destroy(zpool_handle_t
*zhp
, const char *log_str
)
1317 zfs_cmd_t zc
= {"\0"};
1318 zfs_handle_t
*zfp
= NULL
;
1319 libzfs_handle_t
*hdl
= zhp
->zpool_hdl
;
1322 if (zhp
->zpool_state
== POOL_STATE_ACTIVE
&&
1323 (zfp
= zfs_open(hdl
, zhp
->zpool_name
, ZFS_TYPE_FILESYSTEM
)) == NULL
)
1326 (void) strlcpy(zc
.zc_name
, zhp
->zpool_name
, sizeof (zc
.zc_name
));
1327 zc
.zc_history
= (uint64_t)(uintptr_t)log_str
;
1329 if (zfs_ioctl(hdl
, ZFS_IOC_POOL_DESTROY
, &zc
) != 0) {
1330 (void) snprintf(msg
, sizeof (msg
), dgettext(TEXT_DOMAIN
,
1331 "cannot destroy '%s'"), zhp
->zpool_name
);
1333 if (errno
== EROFS
) {
1334 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
1335 "one or more devices is read only"));
1336 (void) zfs_error(hdl
, EZFS_BADDEV
, msg
);
1338 (void) zpool_standard_error(hdl
, errno
, msg
);
1347 remove_mountpoint(zfp
);
1355 * Add the given vdevs to the pool. The caller must have already performed the
1356 * necessary verification to ensure that the vdev specification is well-formed.
1359 zpool_add(zpool_handle_t
*zhp
, nvlist_t
*nvroot
)
1361 zfs_cmd_t zc
= {"\0"};
1363 libzfs_handle_t
*hdl
= zhp
->zpool_hdl
;
1365 nvlist_t
**spares
, **l2cache
;
1366 uint_t nspares
, nl2cache
;
1368 (void) snprintf(msg
, sizeof (msg
), dgettext(TEXT_DOMAIN
,
1369 "cannot add to '%s'"), zhp
->zpool_name
);
1371 if (zpool_get_prop_int(zhp
, ZPOOL_PROP_VERSION
, NULL
) <
1372 SPA_VERSION_SPARES
&&
1373 nvlist_lookup_nvlist_array(nvroot
, ZPOOL_CONFIG_SPARES
,
1374 &spares
, &nspares
) == 0) {
1375 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
, "pool must be "
1376 "upgraded to add hot spares"));
1377 return (zfs_error(hdl
, EZFS_BADVERSION
, msg
));
1380 #if defined(__sun__) || defined(__sun)
1381 if (zpool_is_bootable(zhp
) && nvlist_lookup_nvlist_array(nvroot
,
1382 ZPOOL_CONFIG_SPARES
, &spares
, &nspares
) == 0) {
1385 for (s
= 0; s
< nspares
; s
++) {
1388 if (nvlist_lookup_string(spares
[s
], ZPOOL_CONFIG_PATH
,
1389 &path
) == 0 && pool_uses_efi(spares
[s
])) {
1390 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
1391 "device '%s' contains an EFI label and "
1392 "cannot be used on root pools."),
1393 zpool_vdev_name(hdl
, NULL
, spares
[s
], 0));
1394 return (zfs_error(hdl
, EZFS_POOL_NOTSUP
, msg
));
1400 if (zpool_get_prop_int(zhp
, ZPOOL_PROP_VERSION
, NULL
) <
1401 SPA_VERSION_L2CACHE
&&
1402 nvlist_lookup_nvlist_array(nvroot
, ZPOOL_CONFIG_L2CACHE
,
1403 &l2cache
, &nl2cache
) == 0) {
1404 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
, "pool must be "
1405 "upgraded to add cache devices"));
1406 return (zfs_error(hdl
, EZFS_BADVERSION
, msg
));
1409 if (zcmd_write_conf_nvlist(hdl
, &zc
, nvroot
) != 0)
1411 (void) strlcpy(zc
.zc_name
, zhp
->zpool_name
, sizeof (zc
.zc_name
));
1413 if (zfs_ioctl(hdl
, ZFS_IOC_VDEV_ADD
, &zc
) != 0) {
1417 * This can happen if the user has specified the same
1418 * device multiple times. We can't reliably detect this
1419 * until we try to add it and see we already have a
1422 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
1423 "one or more vdevs refer to the same device"));
1424 (void) zfs_error(hdl
, EZFS_BADDEV
, msg
);
1429 * This occurrs when one of the devices is below
1430 * SPA_MINDEVSIZE. Unfortunately, we can't detect which
1431 * device was the problem device since there's no
1432 * reliable way to determine device size from userland.
1437 zfs_nicenum(SPA_MINDEVSIZE
, buf
, sizeof (buf
));
1439 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
1440 "device is less than the minimum "
1443 (void) zfs_error(hdl
, EZFS_BADDEV
, msg
);
1447 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
1448 "pool must be upgraded to add these vdevs"));
1449 (void) zfs_error(hdl
, EZFS_BADVERSION
, msg
);
1453 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
1454 "cache device must be a disk or disk slice"));
1455 (void) zfs_error(hdl
, EZFS_BADDEV
, msg
);
1459 (void) zpool_standard_error(hdl
, errno
, msg
);
1467 zcmd_free_nvlists(&zc
);
1473 * Exports the pool from the system. The caller must ensure that there are no
1474 * mounted datasets in the pool.
1477 zpool_export_common(zpool_handle_t
*zhp
, boolean_t force
, boolean_t hardforce
,
1478 const char *log_str
)
1480 zfs_cmd_t zc
= {"\0"};
1483 (void) snprintf(msg
, sizeof (msg
), dgettext(TEXT_DOMAIN
,
1484 "cannot export '%s'"), zhp
->zpool_name
);
1486 (void) strlcpy(zc
.zc_name
, zhp
->zpool_name
, sizeof (zc
.zc_name
));
1487 zc
.zc_cookie
= force
;
1488 zc
.zc_guid
= hardforce
;
1489 zc
.zc_history
= (uint64_t)(uintptr_t)log_str
;
1491 if (zfs_ioctl(zhp
->zpool_hdl
, ZFS_IOC_POOL_EXPORT
, &zc
) != 0) {
1494 zfs_error_aux(zhp
->zpool_hdl
, dgettext(TEXT_DOMAIN
,
1495 "use '-f' to override the following errors:\n"
1496 "'%s' has an active shared spare which could be"
1497 " used by other pools once '%s' is exported."),
1498 zhp
->zpool_name
, zhp
->zpool_name
);
1499 return (zfs_error(zhp
->zpool_hdl
, EZFS_ACTIVE_SPARE
,
1502 return (zpool_standard_error_fmt(zhp
->zpool_hdl
, errno
,
1511 zpool_export(zpool_handle_t
*zhp
, boolean_t force
, const char *log_str
)
1513 return (zpool_export_common(zhp
, force
, B_FALSE
, log_str
));
1517 zpool_export_force(zpool_handle_t
*zhp
, const char *log_str
)
1519 return (zpool_export_common(zhp
, B_TRUE
, B_TRUE
, log_str
));
1523 zpool_rewind_exclaim(libzfs_handle_t
*hdl
, const char *name
, boolean_t dryrun
,
1526 nvlist_t
*nv
= NULL
;
1532 if (!hdl
->libzfs_printerr
|| config
== NULL
)
1535 if (nvlist_lookup_nvlist(config
, ZPOOL_CONFIG_LOAD_INFO
, &nv
) != 0 ||
1536 nvlist_lookup_nvlist(nv
, ZPOOL_CONFIG_REWIND_INFO
, &nv
) != 0) {
1540 if (nvlist_lookup_uint64(nv
, ZPOOL_CONFIG_LOAD_TIME
, &rewindto
) != 0)
1542 (void) nvlist_lookup_int64(nv
, ZPOOL_CONFIG_REWIND_TIME
, &loss
);
1544 if (localtime_r((time_t *)&rewindto
, &t
) != NULL
&&
1545 strftime(timestr
, 128, "%c", &t
) != 0) {
1547 (void) printf(dgettext(TEXT_DOMAIN
,
1548 "Would be able to return %s "
1549 "to its state as of %s.\n"),
1552 (void) printf(dgettext(TEXT_DOMAIN
,
1553 "Pool %s returned to its state as of %s.\n"),
1557 (void) printf(dgettext(TEXT_DOMAIN
,
1558 "%s approximately %lld "),
1559 dryrun
? "Would discard" : "Discarded",
1560 ((longlong_t
)loss
+ 30) / 60);
1561 (void) printf(dgettext(TEXT_DOMAIN
,
1562 "minutes of transactions.\n"));
1563 } else if (loss
> 0) {
1564 (void) printf(dgettext(TEXT_DOMAIN
,
1565 "%s approximately %lld "),
1566 dryrun
? "Would discard" : "Discarded",
1568 (void) printf(dgettext(TEXT_DOMAIN
,
1569 "seconds of transactions.\n"));
1575 zpool_explain_recover(libzfs_handle_t
*hdl
, const char *name
, int reason
,
1578 nvlist_t
*nv
= NULL
;
1580 uint64_t edata
= UINT64_MAX
;
1585 if (!hdl
->libzfs_printerr
)
1589 (void) printf(dgettext(TEXT_DOMAIN
, "action: "));
1591 (void) printf(dgettext(TEXT_DOMAIN
, "\t"));
1593 /* All attempted rewinds failed if ZPOOL_CONFIG_LOAD_TIME missing */
1594 if (nvlist_lookup_nvlist(config
, ZPOOL_CONFIG_LOAD_INFO
, &nv
) != 0 ||
1595 nvlist_lookup_nvlist(nv
, ZPOOL_CONFIG_REWIND_INFO
, &nv
) != 0 ||
1596 nvlist_lookup_uint64(nv
, ZPOOL_CONFIG_LOAD_TIME
, &rewindto
) != 0)
1599 (void) nvlist_lookup_int64(nv
, ZPOOL_CONFIG_REWIND_TIME
, &loss
);
1600 (void) nvlist_lookup_uint64(nv
, ZPOOL_CONFIG_LOAD_DATA_ERRORS
,
1603 (void) printf(dgettext(TEXT_DOMAIN
,
1604 "Recovery is possible, but will result in some data loss.\n"));
1606 if (localtime_r((time_t *)&rewindto
, &t
) != NULL
&&
1607 strftime(timestr
, 128, "%c", &t
) != 0) {
1608 (void) printf(dgettext(TEXT_DOMAIN
,
1609 "\tReturning the pool to its state as of %s\n"
1610 "\tshould correct the problem. "),
1613 (void) printf(dgettext(TEXT_DOMAIN
,
1614 "\tReverting the pool to an earlier state "
1615 "should correct the problem.\n\t"));
1619 (void) printf(dgettext(TEXT_DOMAIN
,
1620 "Approximately %lld minutes of data\n"
1621 "\tmust be discarded, irreversibly. "),
1622 ((longlong_t
)loss
+ 30) / 60);
1623 } else if (loss
> 0) {
1624 (void) printf(dgettext(TEXT_DOMAIN
,
1625 "Approximately %lld seconds of data\n"
1626 "\tmust be discarded, irreversibly. "),
1629 if (edata
!= 0 && edata
!= UINT64_MAX
) {
1631 (void) printf(dgettext(TEXT_DOMAIN
,
1632 "After rewind, at least\n"
1633 "\tone persistent user-data error will remain. "));
1635 (void) printf(dgettext(TEXT_DOMAIN
,
1636 "After rewind, several\n"
1637 "\tpersistent user-data errors will remain. "));
1640 (void) printf(dgettext(TEXT_DOMAIN
,
1641 "Recovery can be attempted\n\tby executing 'zpool %s -F %s'. "),
1642 reason
>= 0 ? "clear" : "import", name
);
1644 (void) printf(dgettext(TEXT_DOMAIN
,
1645 "A scrub of the pool\n"
1646 "\tis strongly recommended after recovery.\n"));
1650 (void) printf(dgettext(TEXT_DOMAIN
,
1651 "Destroy and re-create the pool from\n\ta backup source.\n"));
1655 * zpool_import() is a contracted interface. Should be kept the same
1658 * Applications should use zpool_import_props() to import a pool with
1659 * new properties value to be set.
1662 zpool_import(libzfs_handle_t
*hdl
, nvlist_t
*config
, const char *newname
,
1665 nvlist_t
*props
= NULL
;
1668 if (altroot
!= NULL
) {
1669 if (nvlist_alloc(&props
, NV_UNIQUE_NAME
, 0) != 0) {
1670 return (zfs_error_fmt(hdl
, EZFS_NOMEM
,
1671 dgettext(TEXT_DOMAIN
, "cannot import '%s'"),
1675 if (nvlist_add_string(props
,
1676 zpool_prop_to_name(ZPOOL_PROP_ALTROOT
), altroot
) != 0 ||
1677 nvlist_add_string(props
,
1678 zpool_prop_to_name(ZPOOL_PROP_CACHEFILE
), "none") != 0) {
1680 return (zfs_error_fmt(hdl
, EZFS_NOMEM
,
1681 dgettext(TEXT_DOMAIN
, "cannot import '%s'"),
1686 ret
= zpool_import_props(hdl
, config
, newname
, props
,
1693 print_vdev_tree(libzfs_handle_t
*hdl
, const char *name
, nvlist_t
*nv
,
1699 uint64_t is_log
= 0;
1701 (void) nvlist_lookup_uint64(nv
, ZPOOL_CONFIG_IS_LOG
,
1705 (void) printf("\t%*s%s%s\n", indent
, "", name
,
1706 is_log
? " [log]" : "");
1708 if (nvlist_lookup_nvlist_array(nv
, ZPOOL_CONFIG_CHILDREN
,
1709 &child
, &children
) != 0)
1712 for (c
= 0; c
< children
; c
++) {
1713 vname
= zpool_vdev_name(hdl
, NULL
, child
[c
], VDEV_NAME_TYPE_ID
);
1714 print_vdev_tree(hdl
, vname
, child
[c
], indent
+ 2);
1720 zpool_print_unsup_feat(nvlist_t
*config
)
1722 nvlist_t
*nvinfo
, *unsup_feat
;
1725 verify(nvlist_lookup_nvlist(config
, ZPOOL_CONFIG_LOAD_INFO
, &nvinfo
) ==
1727 verify(nvlist_lookup_nvlist(nvinfo
, ZPOOL_CONFIG_UNSUP_FEAT
,
1730 for (nvp
= nvlist_next_nvpair(unsup_feat
, NULL
); nvp
!= NULL
;
1731 nvp
= nvlist_next_nvpair(unsup_feat
, nvp
)) {
1734 verify(nvpair_type(nvp
) == DATA_TYPE_STRING
);
1735 verify(nvpair_value_string(nvp
, &desc
) == 0);
1737 if (strlen(desc
) > 0)
1738 (void) printf("\t%s (%s)\n", nvpair_name(nvp
), desc
);
1740 (void) printf("\t%s\n", nvpair_name(nvp
));
1745 * Import the given pool using the known configuration and a list of
1746 * properties to be set. The configuration should have come from
1747 * zpool_find_import(). The 'newname' parameters control whether the pool
1748 * is imported with a different name.
1751 zpool_import_props(libzfs_handle_t
*hdl
, nvlist_t
*config
, const char *newname
,
1752 nvlist_t
*props
, int flags
)
1754 zfs_cmd_t zc
= {"\0"};
1755 zpool_rewind_policy_t policy
;
1756 nvlist_t
*nv
= NULL
;
1757 nvlist_t
*nvinfo
= NULL
;
1758 nvlist_t
*missing
= NULL
;
1765 verify(nvlist_lookup_string(config
, ZPOOL_CONFIG_POOL_NAME
,
1768 (void) snprintf(errbuf
, sizeof (errbuf
), dgettext(TEXT_DOMAIN
,
1769 "cannot import pool '%s'"), origname
);
1771 if (newname
!= NULL
) {
1772 if (!zpool_name_valid(hdl
, B_FALSE
, newname
))
1773 return (zfs_error_fmt(hdl
, EZFS_INVALIDNAME
,
1774 dgettext(TEXT_DOMAIN
, "cannot import '%s'"),
1776 thename
= (char *)newname
;
1781 if (props
!= NULL
) {
1783 prop_flags_t flags
= { .create
= B_FALSE
, .import
= B_TRUE
};
1785 verify(nvlist_lookup_uint64(config
, ZPOOL_CONFIG_VERSION
,
1788 if ((props
= zpool_valid_proplist(hdl
, origname
,
1789 props
, version
, flags
, errbuf
)) == NULL
)
1791 if (zcmd_write_src_nvlist(hdl
, &zc
, props
) != 0) {
1798 (void) strlcpy(zc
.zc_name
, thename
, sizeof (zc
.zc_name
));
1800 verify(nvlist_lookup_uint64(config
, ZPOOL_CONFIG_POOL_GUID
,
1803 if (zcmd_write_conf_nvlist(hdl
, &zc
, config
) != 0) {
1804 zcmd_free_nvlists(&zc
);
1807 if (zcmd_alloc_dst_nvlist(hdl
, &zc
, zc
.zc_nvlist_conf_size
* 2) != 0) {
1808 zcmd_free_nvlists(&zc
);
1812 zc
.zc_cookie
= flags
;
1813 while ((ret
= zfs_ioctl(hdl
, ZFS_IOC_POOL_IMPORT
, &zc
)) != 0 &&
1815 if (zcmd_expand_dst_nvlist(hdl
, &zc
) != 0) {
1816 zcmd_free_nvlists(&zc
);
1823 (void) zcmd_read_dst_nvlist(hdl
, &zc
, &nv
);
1825 zcmd_free_nvlists(&zc
);
1827 zpool_get_rewind_policy(config
, &policy
);
1833 * Dry-run failed, but we print out what success
1834 * looks like if we found a best txg
1836 if (policy
.zrp_request
& ZPOOL_TRY_REWIND
) {
1837 zpool_rewind_exclaim(hdl
, newname
? origname
: thename
,
1843 if (newname
== NULL
)
1844 (void) snprintf(desc
, sizeof (desc
),
1845 dgettext(TEXT_DOMAIN
, "cannot import '%s'"),
1848 (void) snprintf(desc
, sizeof (desc
),
1849 dgettext(TEXT_DOMAIN
, "cannot import '%s' as '%s'"),
1854 if (nv
!= NULL
&& nvlist_lookup_nvlist(nv
,
1855 ZPOOL_CONFIG_LOAD_INFO
, &nvinfo
) == 0 &&
1856 nvlist_exists(nvinfo
, ZPOOL_CONFIG_UNSUP_FEAT
)) {
1857 (void) printf(dgettext(TEXT_DOMAIN
, "This "
1858 "pool uses the following feature(s) not "
1859 "supported by this system:\n"));
1860 zpool_print_unsup_feat(nv
);
1861 if (nvlist_exists(nvinfo
,
1862 ZPOOL_CONFIG_CAN_RDONLY
)) {
1863 (void) printf(dgettext(TEXT_DOMAIN
,
1864 "All unsupported features are only "
1865 "required for writing to the pool."
1866 "\nThe pool can be imported using "
1867 "'-o readonly=on'.\n"));
1871 * Unsupported version.
1873 (void) zfs_error(hdl
, EZFS_BADVERSION
, desc
);
1877 (void) zfs_error(hdl
, EZFS_INVALCONFIG
, desc
);
1881 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
1882 "one or more devices is read only"));
1883 (void) zfs_error(hdl
, EZFS_BADDEV
, desc
);
1887 if (nv
&& nvlist_lookup_nvlist(nv
,
1888 ZPOOL_CONFIG_LOAD_INFO
, &nvinfo
) == 0 &&
1889 nvlist_lookup_nvlist(nvinfo
,
1890 ZPOOL_CONFIG_MISSING_DEVICES
, &missing
) == 0) {
1891 (void) printf(dgettext(TEXT_DOMAIN
,
1892 "The devices below are missing, use "
1893 "'-m' to import the pool anyway:\n"));
1894 print_vdev_tree(hdl
, NULL
, missing
, 2);
1895 (void) printf("\n");
1897 (void) zpool_standard_error(hdl
, error
, desc
);
1901 (void) zpool_standard_error(hdl
, error
, desc
);
1905 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
1906 "one or more devices are already in use\n"));
1907 (void) zfs_error(hdl
, EZFS_BADDEV
, desc
);
1911 (void) zpool_standard_error(hdl
, error
, desc
);
1912 zpool_explain_recover(hdl
,
1913 newname
? origname
: thename
, -error
, nv
);
1920 zpool_handle_t
*zhp
;
1923 * This should never fail, but play it safe anyway.
1925 if (zpool_open_silent(hdl
, thename
, &zhp
) != 0)
1927 else if (zhp
!= NULL
)
1929 if (policy
.zrp_request
&
1930 (ZPOOL_DO_REWIND
| ZPOOL_TRY_REWIND
)) {
1931 zpool_rewind_exclaim(hdl
, newname
? origname
: thename
,
1932 ((policy
.zrp_request
& ZPOOL_TRY_REWIND
) != 0), nv
);
1945 zpool_scan(zpool_handle_t
*zhp
, pool_scan_func_t func
)
1947 zfs_cmd_t zc
= {"\0"};
1949 libzfs_handle_t
*hdl
= zhp
->zpool_hdl
;
1951 (void) strlcpy(zc
.zc_name
, zhp
->zpool_name
, sizeof (zc
.zc_name
));
1952 zc
.zc_cookie
= func
;
1954 if (zfs_ioctl(hdl
, ZFS_IOC_POOL_SCAN
, &zc
) == 0 ||
1955 (errno
== ENOENT
&& func
!= POOL_SCAN_NONE
))
1958 if (func
== POOL_SCAN_SCRUB
) {
1959 (void) snprintf(msg
, sizeof (msg
),
1960 dgettext(TEXT_DOMAIN
, "cannot scrub %s"), zc
.zc_name
);
1961 } else if (func
== POOL_SCAN_NONE
) {
1962 (void) snprintf(msg
, sizeof (msg
),
1963 dgettext(TEXT_DOMAIN
, "cannot cancel scrubbing %s"),
1966 assert(!"unexpected result");
1969 if (errno
== EBUSY
) {
1971 pool_scan_stat_t
*ps
= NULL
;
1974 verify(nvlist_lookup_nvlist(zhp
->zpool_config
,
1975 ZPOOL_CONFIG_VDEV_TREE
, &nvroot
) == 0);
1976 (void) nvlist_lookup_uint64_array(nvroot
,
1977 ZPOOL_CONFIG_SCAN_STATS
, (uint64_t **)&ps
, &psc
);
1978 if (ps
&& ps
->pss_func
== POOL_SCAN_SCRUB
)
1979 return (zfs_error(hdl
, EZFS_SCRUBBING
, msg
));
1981 return (zfs_error(hdl
, EZFS_RESILVERING
, msg
));
1982 } else if (errno
== ENOENT
) {
1983 return (zfs_error(hdl
, EZFS_NO_SCRUB
, msg
));
1985 return (zpool_standard_error(hdl
, errno
, msg
));
1990 * Find a vdev that matches the search criteria specified. We use the
1991 * the nvpair name to determine how we should look for the device.
1992 * 'avail_spare' is set to TRUE if the provided guid refers to an AVAIL
1993 * spare; but FALSE if its an INUSE spare.
1996 vdev_to_nvlist_iter(nvlist_t
*nv
, nvlist_t
*search
, boolean_t
*avail_spare
,
1997 boolean_t
*l2cache
, boolean_t
*log
)
2004 nvpair_t
*pair
= nvlist_next_nvpair(search
, NULL
);
2006 /* Nothing to look for */
2007 if (search
== NULL
|| pair
== NULL
)
2010 /* Obtain the key we will use to search */
2011 srchkey
= nvpair_name(pair
);
2013 switch (nvpair_type(pair
)) {
2014 case DATA_TYPE_UINT64
:
2015 if (strcmp(srchkey
, ZPOOL_CONFIG_GUID
) == 0) {
2016 uint64_t srchval
, theguid
;
2018 verify(nvpair_value_uint64(pair
, &srchval
) == 0);
2019 verify(nvlist_lookup_uint64(nv
, ZPOOL_CONFIG_GUID
,
2021 if (theguid
== srchval
)
2026 case DATA_TYPE_STRING
: {
2027 char *srchval
, *val
;
2029 verify(nvpair_value_string(pair
, &srchval
) == 0);
2030 if (nvlist_lookup_string(nv
, srchkey
, &val
) != 0)
2034 * Search for the requested value. Special cases:
2036 * - ZPOOL_CONFIG_PATH for whole disk entries. These end in
2037 * "-part1", or "p1". The suffix is hidden from the user,
2038 * but included in the string, so this matches around it.
2039 * - ZPOOL_CONFIG_PATH for short names zfs_strcmp_shortname()
2040 * is used to check all possible expanded paths.
2041 * - looking for a top-level vdev name (i.e. ZPOOL_CONFIG_TYPE).
2043 * Otherwise, all other searches are simple string compares.
2045 if (strcmp(srchkey
, ZPOOL_CONFIG_PATH
) == 0) {
2046 uint64_t wholedisk
= 0;
2048 (void) nvlist_lookup_uint64(nv
, ZPOOL_CONFIG_WHOLE_DISK
,
2050 if (zfs_strcmp_pathname(srchval
, val
, wholedisk
) == 0)
2053 } else if (strcmp(srchkey
, ZPOOL_CONFIG_TYPE
) == 0 && val
) {
2054 char *type
, *idx
, *end
, *p
;
2055 uint64_t id
, vdev_id
;
2058 * Determine our vdev type, keeping in mind
2059 * that the srchval is composed of a type and
2060 * vdev id pair (i.e. mirror-4).
2062 if ((type
= strdup(srchval
)) == NULL
)
2065 if ((p
= strrchr(type
, '-')) == NULL
) {
2073 * If the types don't match then keep looking.
2075 if (strncmp(val
, type
, strlen(val
)) != 0) {
2080 verify(strncmp(type
, VDEV_TYPE_RAIDZ
,
2081 strlen(VDEV_TYPE_RAIDZ
)) == 0 ||
2082 strncmp(type
, VDEV_TYPE_MIRROR
,
2083 strlen(VDEV_TYPE_MIRROR
)) == 0);
2084 verify(nvlist_lookup_uint64(nv
, ZPOOL_CONFIG_ID
,
2088 vdev_id
= strtoull(idx
, &end
, 10);
2095 * Now verify that we have the correct vdev id.
2104 if (strcmp(srchval
, val
) == 0)
2113 if (nvlist_lookup_nvlist_array(nv
, ZPOOL_CONFIG_CHILDREN
,
2114 &child
, &children
) != 0)
2117 for (c
= 0; c
< children
; c
++) {
2118 if ((ret
= vdev_to_nvlist_iter(child
[c
], search
,
2119 avail_spare
, l2cache
, NULL
)) != NULL
) {
2121 * The 'is_log' value is only set for the toplevel
2122 * vdev, not the leaf vdevs. So we always lookup the
2123 * log device from the root of the vdev tree (where
2124 * 'log' is non-NULL).
2127 nvlist_lookup_uint64(child
[c
],
2128 ZPOOL_CONFIG_IS_LOG
, &is_log
) == 0 &&
2136 if (nvlist_lookup_nvlist_array(nv
, ZPOOL_CONFIG_SPARES
,
2137 &child
, &children
) == 0) {
2138 for (c
= 0; c
< children
; c
++) {
2139 if ((ret
= vdev_to_nvlist_iter(child
[c
], search
,
2140 avail_spare
, l2cache
, NULL
)) != NULL
) {
2141 *avail_spare
= B_TRUE
;
2147 if (nvlist_lookup_nvlist_array(nv
, ZPOOL_CONFIG_L2CACHE
,
2148 &child
, &children
) == 0) {
2149 for (c
= 0; c
< children
; c
++) {
2150 if ((ret
= vdev_to_nvlist_iter(child
[c
], search
,
2151 avail_spare
, l2cache
, NULL
)) != NULL
) {
2162 * Given a physical path (minus the "/devices" prefix), find the
2166 zpool_find_vdev_by_physpath(zpool_handle_t
*zhp
, const char *ppath
,
2167 boolean_t
*avail_spare
, boolean_t
*l2cache
, boolean_t
*log
)
2169 nvlist_t
*search
, *nvroot
, *ret
;
2171 verify(nvlist_alloc(&search
, NV_UNIQUE_NAME
, KM_SLEEP
) == 0);
2172 verify(nvlist_add_string(search
, ZPOOL_CONFIG_PHYS_PATH
, ppath
) == 0);
2174 verify(nvlist_lookup_nvlist(zhp
->zpool_config
, ZPOOL_CONFIG_VDEV_TREE
,
2177 *avail_spare
= B_FALSE
;
2181 ret
= vdev_to_nvlist_iter(nvroot
, search
, avail_spare
, l2cache
, log
);
2182 nvlist_free(search
);
2188 * Determine if we have an "interior" top-level vdev (i.e mirror/raidz).
2191 zpool_vdev_is_interior(const char *name
)
2193 if (strncmp(name
, VDEV_TYPE_RAIDZ
, strlen(VDEV_TYPE_RAIDZ
)) == 0 ||
2194 strncmp(name
, VDEV_TYPE_MIRROR
, strlen(VDEV_TYPE_MIRROR
)) == 0)
2200 zpool_find_vdev(zpool_handle_t
*zhp
, const char *path
, boolean_t
*avail_spare
,
2201 boolean_t
*l2cache
, boolean_t
*log
)
2204 nvlist_t
*nvroot
, *search
, *ret
;
2207 verify(nvlist_alloc(&search
, NV_UNIQUE_NAME
, KM_SLEEP
) == 0);
2209 guid
= strtoull(path
, &end
, 0);
2210 if (guid
!= 0 && *end
== '\0') {
2211 verify(nvlist_add_uint64(search
, ZPOOL_CONFIG_GUID
, guid
) == 0);
2212 } else if (zpool_vdev_is_interior(path
)) {
2213 verify(nvlist_add_string(search
, ZPOOL_CONFIG_TYPE
, path
) == 0);
2215 verify(nvlist_add_string(search
, ZPOOL_CONFIG_PATH
, path
) == 0);
2218 verify(nvlist_lookup_nvlist(zhp
->zpool_config
, ZPOOL_CONFIG_VDEV_TREE
,
2221 *avail_spare
= B_FALSE
;
2225 ret
= vdev_to_nvlist_iter(nvroot
, search
, avail_spare
, l2cache
, log
);
2226 nvlist_free(search
);
2232 vdev_online(nvlist_t
*nv
)
2236 if (nvlist_lookup_uint64(nv
, ZPOOL_CONFIG_OFFLINE
, &ival
) == 0 ||
2237 nvlist_lookup_uint64(nv
, ZPOOL_CONFIG_FAULTED
, &ival
) == 0 ||
2238 nvlist_lookup_uint64(nv
, ZPOOL_CONFIG_REMOVED
, &ival
) == 0)
2245 * Helper function for zpool_get_physpaths().
2248 vdev_get_one_physpath(nvlist_t
*config
, char *physpath
, size_t physpath_size
,
2249 size_t *bytes_written
)
2251 size_t bytes_left
, pos
, rsz
;
2255 if (nvlist_lookup_string(config
, ZPOOL_CONFIG_PHYS_PATH
,
2257 return (EZFS_NODEVICE
);
2259 pos
= *bytes_written
;
2260 bytes_left
= physpath_size
- pos
;
2261 format
= (pos
== 0) ? "%s" : " %s";
2263 rsz
= snprintf(physpath
+ pos
, bytes_left
, format
, tmppath
);
2264 *bytes_written
+= rsz
;
2266 if (rsz
>= bytes_left
) {
2267 /* if physpath was not copied properly, clear it */
2268 if (bytes_left
!= 0) {
2271 return (EZFS_NOSPC
);
2277 vdev_get_physpaths(nvlist_t
*nv
, char *physpath
, size_t phypath_size
,
2278 size_t *rsz
, boolean_t is_spare
)
2283 if (nvlist_lookup_string(nv
, ZPOOL_CONFIG_TYPE
, &type
) != 0)
2284 return (EZFS_INVALCONFIG
);
2286 if (strcmp(type
, VDEV_TYPE_DISK
) == 0) {
2288 * An active spare device has ZPOOL_CONFIG_IS_SPARE set.
2289 * For a spare vdev, we only want to boot from the active
2294 (void) nvlist_lookup_uint64(nv
, ZPOOL_CONFIG_IS_SPARE
,
2297 return (EZFS_INVALCONFIG
);
2300 if (vdev_online(nv
)) {
2301 if ((ret
= vdev_get_one_physpath(nv
, physpath
,
2302 phypath_size
, rsz
)) != 0)
2305 } else if (strcmp(type
, VDEV_TYPE_MIRROR
) == 0 ||
2306 strcmp(type
, VDEV_TYPE_REPLACING
) == 0 ||
2307 (is_spare
= (strcmp(type
, VDEV_TYPE_SPARE
) == 0))) {
2312 if (nvlist_lookup_nvlist_array(nv
,
2313 ZPOOL_CONFIG_CHILDREN
, &child
, &count
) != 0)
2314 return (EZFS_INVALCONFIG
);
2316 for (i
= 0; i
< count
; i
++) {
2317 ret
= vdev_get_physpaths(child
[i
], physpath
,
2318 phypath_size
, rsz
, is_spare
);
2319 if (ret
== EZFS_NOSPC
)
2324 return (EZFS_POOL_INVALARG
);
2328 * Get phys_path for a root pool config.
2329 * Return 0 on success; non-zero on failure.
2332 zpool_get_config_physpath(nvlist_t
*config
, char *physpath
, size_t phypath_size
)
2335 nvlist_t
*vdev_root
;
2342 if (nvlist_lookup_nvlist(config
, ZPOOL_CONFIG_VDEV_TREE
,
2344 return (EZFS_INVALCONFIG
);
2346 if (nvlist_lookup_string(vdev_root
, ZPOOL_CONFIG_TYPE
, &type
) != 0 ||
2347 nvlist_lookup_nvlist_array(vdev_root
, ZPOOL_CONFIG_CHILDREN
,
2348 &child
, &count
) != 0)
2349 return (EZFS_INVALCONFIG
);
2351 #if defined(__sun__) || defined(__sun)
2353 * root pool can not have EFI labeled disks and can only have
2354 * a single top-level vdev.
2356 if (strcmp(type
, VDEV_TYPE_ROOT
) != 0 || count
!= 1 ||
2357 pool_uses_efi(vdev_root
))
2358 return (EZFS_POOL_INVALARG
);
2361 (void) vdev_get_physpaths(child
[0], physpath
, phypath_size
, &rsz
,
2364 /* No online devices */
2366 return (EZFS_NODEVICE
);
2372 * Get phys_path for a root pool
2373 * Return 0 on success; non-zero on failure.
2376 zpool_get_physpath(zpool_handle_t
*zhp
, char *physpath
, size_t phypath_size
)
2378 return (zpool_get_config_physpath(zhp
->zpool_config
, physpath
,
2383 * If the device has being dynamically expanded then we need to relabel
2384 * the disk to use the new unallocated space.
2387 zpool_relabel_disk(libzfs_handle_t
*hdl
, const char *path
, const char *msg
)
2391 if ((fd
= open(path
, O_RDWR
|O_DIRECT
)) < 0) {
2392 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
, "cannot "
2393 "relabel '%s': unable to open device: %d"), path
, errno
);
2394 return (zfs_error(hdl
, EZFS_OPENFAILED
, msg
));
2398 * It's possible that we might encounter an error if the device
2399 * does not have any unallocated space left. If so, we simply
2400 * ignore that error and continue on.
2402 * Also, we don't call efi_rescan() - that would just return EBUSY.
2403 * The module will do it for us in vdev_disk_open().
2405 error
= efi_use_whole_disk(fd
);
2407 if (error
&& error
!= VT_ENOSPC
) {
2408 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
, "cannot "
2409 "relabel '%s': unable to read disk capacity"), path
);
2410 return (zfs_error(hdl
, EZFS_NOCAP
, msg
));
2416 * Bring the specified vdev online. The 'flags' parameter is a set of the
2417 * ZFS_ONLINE_* flags.
2420 zpool_vdev_online(zpool_handle_t
*zhp
, const char *path
, int flags
,
2421 vdev_state_t
*newstate
)
2423 zfs_cmd_t zc
= {"\0"};
2426 boolean_t avail_spare
, l2cache
, islog
;
2427 libzfs_handle_t
*hdl
= zhp
->zpool_hdl
;
2430 if (flags
& ZFS_ONLINE_EXPAND
) {
2431 (void) snprintf(msg
, sizeof (msg
),
2432 dgettext(TEXT_DOMAIN
, "cannot expand %s"), path
);
2434 (void) snprintf(msg
, sizeof (msg
),
2435 dgettext(TEXT_DOMAIN
, "cannot online %s"), path
);
2438 (void) strlcpy(zc
.zc_name
, zhp
->zpool_name
, sizeof (zc
.zc_name
));
2439 if ((tgt
= zpool_find_vdev(zhp
, path
, &avail_spare
, &l2cache
,
2441 return (zfs_error(hdl
, EZFS_NODEVICE
, msg
));
2443 verify(nvlist_lookup_uint64(tgt
, ZPOOL_CONFIG_GUID
, &zc
.zc_guid
) == 0);
2446 return (zfs_error(hdl
, EZFS_ISSPARE
, msg
));
2448 if (flags
& ZFS_ONLINE_EXPAND
||
2449 zpool_get_prop_int(zhp
, ZPOOL_PROP_AUTOEXPAND
, NULL
)) {
2450 uint64_t wholedisk
= 0;
2452 (void) nvlist_lookup_uint64(tgt
, ZPOOL_CONFIG_WHOLE_DISK
,
2456 * XXX - L2ARC 1.0 devices can't support expansion.
2459 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
2460 "cannot expand cache devices"));
2461 return (zfs_error(hdl
, EZFS_VDEVNOTSUP
, msg
));
2465 const char *fullpath
= path
;
2466 char buf
[MAXPATHLEN
];
2468 if (path
[0] != '/') {
2469 error
= zfs_resolve_shortname(path
, buf
,
2472 return (zfs_error(hdl
, EZFS_NODEVICE
,
2478 error
= zpool_relabel_disk(hdl
, fullpath
, msg
);
2484 zc
.zc_cookie
= VDEV_STATE_ONLINE
;
2487 if (zfs_ioctl(hdl
, ZFS_IOC_VDEV_SET_STATE
, &zc
) != 0) {
2488 if (errno
== EINVAL
) {
2489 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
, "was split "
2490 "from this pool into a new one. Use '%s' "
2491 "instead"), "zpool detach");
2492 return (zfs_error(hdl
, EZFS_POSTSPLIT_ONLINE
, msg
));
2494 return (zpool_standard_error(hdl
, errno
, msg
));
2497 *newstate
= zc
.zc_cookie
;
2502 * Take the specified vdev offline
2505 zpool_vdev_offline(zpool_handle_t
*zhp
, const char *path
, boolean_t istmp
)
2507 zfs_cmd_t zc
= {"\0"};
2510 boolean_t avail_spare
, l2cache
;
2511 libzfs_handle_t
*hdl
= zhp
->zpool_hdl
;
2513 (void) snprintf(msg
, sizeof (msg
),
2514 dgettext(TEXT_DOMAIN
, "cannot offline %s"), path
);
2516 (void) strlcpy(zc
.zc_name
, zhp
->zpool_name
, sizeof (zc
.zc_name
));
2517 if ((tgt
= zpool_find_vdev(zhp
, path
, &avail_spare
, &l2cache
,
2519 return (zfs_error(hdl
, EZFS_NODEVICE
, msg
));
2521 verify(nvlist_lookup_uint64(tgt
, ZPOOL_CONFIG_GUID
, &zc
.zc_guid
) == 0);
2524 return (zfs_error(hdl
, EZFS_ISSPARE
, msg
));
2526 zc
.zc_cookie
= VDEV_STATE_OFFLINE
;
2527 zc
.zc_obj
= istmp
? ZFS_OFFLINE_TEMPORARY
: 0;
2529 if (zfs_ioctl(hdl
, ZFS_IOC_VDEV_SET_STATE
, &zc
) == 0)
2536 * There are no other replicas of this device.
2538 return (zfs_error(hdl
, EZFS_NOREPLICAS
, msg
));
2542 * The log device has unplayed logs
2544 return (zfs_error(hdl
, EZFS_UNPLAYED_LOGS
, msg
));
2547 return (zpool_standard_error(hdl
, errno
, msg
));
2552 * Mark the given vdev faulted.
2555 zpool_vdev_fault(zpool_handle_t
*zhp
, uint64_t guid
, vdev_aux_t aux
)
2557 zfs_cmd_t zc
= {"\0"};
2559 libzfs_handle_t
*hdl
= zhp
->zpool_hdl
;
2561 (void) snprintf(msg
, sizeof (msg
),
2562 dgettext(TEXT_DOMAIN
, "cannot fault %llu"), (u_longlong_t
)guid
);
2564 (void) strlcpy(zc
.zc_name
, zhp
->zpool_name
, sizeof (zc
.zc_name
));
2566 zc
.zc_cookie
= VDEV_STATE_FAULTED
;
2569 if (ioctl(hdl
->libzfs_fd
, ZFS_IOC_VDEV_SET_STATE
, &zc
) == 0)
2576 * There are no other replicas of this device.
2578 return (zfs_error(hdl
, EZFS_NOREPLICAS
, msg
));
2581 return (zpool_standard_error(hdl
, errno
, msg
));
2587 * Mark the given vdev degraded.
2590 zpool_vdev_degrade(zpool_handle_t
*zhp
, uint64_t guid
, vdev_aux_t aux
)
2592 zfs_cmd_t zc
= {"\0"};
2594 libzfs_handle_t
*hdl
= zhp
->zpool_hdl
;
2596 (void) snprintf(msg
, sizeof (msg
),
2597 dgettext(TEXT_DOMAIN
, "cannot degrade %llu"), (u_longlong_t
)guid
);
2599 (void) strlcpy(zc
.zc_name
, zhp
->zpool_name
, sizeof (zc
.zc_name
));
2601 zc
.zc_cookie
= VDEV_STATE_DEGRADED
;
2604 if (ioctl(hdl
->libzfs_fd
, ZFS_IOC_VDEV_SET_STATE
, &zc
) == 0)
2607 return (zpool_standard_error(hdl
, errno
, msg
));
2611 * Returns TRUE if the given nvlist is a vdev that was originally swapped in as
2615 is_replacing_spare(nvlist_t
*search
, nvlist_t
*tgt
, int which
)
2621 if (nvlist_lookup_nvlist_array(search
, ZPOOL_CONFIG_CHILDREN
, &child
,
2623 verify(nvlist_lookup_string(search
, ZPOOL_CONFIG_TYPE
,
2626 if (strcmp(type
, VDEV_TYPE_SPARE
) == 0 &&
2627 children
== 2 && child
[which
] == tgt
)
2630 for (c
= 0; c
< children
; c
++)
2631 if (is_replacing_spare(child
[c
], tgt
, which
))
2639 * Attach new_disk (fully described by nvroot) to old_disk.
2640 * If 'replacing' is specified, the new disk will replace the old one.
2643 zpool_vdev_attach(zpool_handle_t
*zhp
,
2644 const char *old_disk
, const char *new_disk
, nvlist_t
*nvroot
, int replacing
)
2646 zfs_cmd_t zc
= {"\0"};
2650 boolean_t avail_spare
, l2cache
, islog
;
2655 nvlist_t
*config_root
;
2656 libzfs_handle_t
*hdl
= zhp
->zpool_hdl
;
2657 boolean_t rootpool
= zpool_is_bootable(zhp
);
2660 (void) snprintf(msg
, sizeof (msg
), dgettext(TEXT_DOMAIN
,
2661 "cannot replace %s with %s"), old_disk
, new_disk
);
2663 (void) snprintf(msg
, sizeof (msg
), dgettext(TEXT_DOMAIN
,
2664 "cannot attach %s to %s"), new_disk
, old_disk
);
2666 #if defined(__sun__) || defined(__sun)
2668 * If this is a root pool, make sure that we're not attaching an
2669 * EFI labeled device.
2671 if (rootpool
&& pool_uses_efi(nvroot
)) {
2672 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
2673 "EFI labeled devices are not supported on root pools."));
2674 return (zfs_error(hdl
, EZFS_POOL_NOTSUP
, msg
));
2678 (void) strlcpy(zc
.zc_name
, zhp
->zpool_name
, sizeof (zc
.zc_name
));
2679 if ((tgt
= zpool_find_vdev(zhp
, old_disk
, &avail_spare
, &l2cache
,
2681 return (zfs_error(hdl
, EZFS_NODEVICE
, msg
));
2684 return (zfs_error(hdl
, EZFS_ISSPARE
, msg
));
2687 return (zfs_error(hdl
, EZFS_ISL2CACHE
, msg
));
2689 verify(nvlist_lookup_uint64(tgt
, ZPOOL_CONFIG_GUID
, &zc
.zc_guid
) == 0);
2690 zc
.zc_cookie
= replacing
;
2692 if (nvlist_lookup_nvlist_array(nvroot
, ZPOOL_CONFIG_CHILDREN
,
2693 &child
, &children
) != 0 || children
!= 1) {
2694 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
2695 "new device must be a single disk"));
2696 return (zfs_error(hdl
, EZFS_INVALCONFIG
, msg
));
2699 verify(nvlist_lookup_nvlist(zpool_get_config(zhp
, NULL
),
2700 ZPOOL_CONFIG_VDEV_TREE
, &config_root
) == 0);
2702 if ((newname
= zpool_vdev_name(NULL
, NULL
, child
[0], 0)) == NULL
)
2706 * If the target is a hot spare that has been swapped in, we can only
2707 * replace it with another hot spare.
2710 nvlist_lookup_uint64(tgt
, ZPOOL_CONFIG_IS_SPARE
, &val
) == 0 &&
2711 (zpool_find_vdev(zhp
, newname
, &avail_spare
, &l2cache
,
2712 NULL
) == NULL
|| !avail_spare
) &&
2713 is_replacing_spare(config_root
, tgt
, 1)) {
2714 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
2715 "can only be replaced by another hot spare"));
2717 return (zfs_error(hdl
, EZFS_BADTARGET
, msg
));
2722 if (zcmd_write_conf_nvlist(hdl
, &zc
, nvroot
) != 0)
2725 ret
= zfs_ioctl(hdl
, ZFS_IOC_VDEV_ATTACH
, &zc
);
2727 zcmd_free_nvlists(&zc
);
2732 * XXX need a better way to prevent user from
2733 * booting up a half-baked vdev.
2735 (void) fprintf(stderr
, dgettext(TEXT_DOMAIN
, "Make "
2736 "sure to wait until resilver is done "
2737 "before rebooting.\n"));
2745 * Can't attach to or replace this type of vdev.
2748 uint64_t version
= zpool_get_prop_int(zhp
,
2749 ZPOOL_PROP_VERSION
, NULL
);
2752 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
2753 "cannot replace a log with a spare"));
2754 else if (version
>= SPA_VERSION_MULTI_REPLACE
)
2755 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
2756 "already in replacing/spare config; wait "
2757 "for completion or use 'zpool detach'"));
2759 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
2760 "cannot replace a replacing device"));
2762 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
2763 "can only attach to mirrors and top-level "
2766 (void) zfs_error(hdl
, EZFS_BADTARGET
, msg
);
2771 * The new device must be a single disk.
2773 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
2774 "new device must be a single disk"));
2775 (void) zfs_error(hdl
, EZFS_INVALCONFIG
, msg
);
2779 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
, "%s is busy"),
2781 (void) zfs_error(hdl
, EZFS_BADDEV
, msg
);
2786 * The new device is too small.
2788 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
2789 "device is too small"));
2790 (void) zfs_error(hdl
, EZFS_BADDEV
, msg
);
2795 * The new device has a different optimal sector size.
2797 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
2798 "new device has a different optimal sector size; use the "
2799 "option '-o ashift=N' to override the optimal size"));
2800 (void) zfs_error(hdl
, EZFS_BADDEV
, msg
);
2805 * The resulting top-level vdev spec won't fit in the label.
2807 (void) zfs_error(hdl
, EZFS_DEVOVERFLOW
, msg
);
2811 (void) zpool_standard_error(hdl
, errno
, msg
);
2818 * Detach the specified device.
2821 zpool_vdev_detach(zpool_handle_t
*zhp
, const char *path
)
2823 zfs_cmd_t zc
= {"\0"};
2826 boolean_t avail_spare
, l2cache
;
2827 libzfs_handle_t
*hdl
= zhp
->zpool_hdl
;
2829 (void) snprintf(msg
, sizeof (msg
),
2830 dgettext(TEXT_DOMAIN
, "cannot detach %s"), path
);
2832 (void) strlcpy(zc
.zc_name
, zhp
->zpool_name
, sizeof (zc
.zc_name
));
2833 if ((tgt
= zpool_find_vdev(zhp
, path
, &avail_spare
, &l2cache
,
2835 return (zfs_error(hdl
, EZFS_NODEVICE
, msg
));
2838 return (zfs_error(hdl
, EZFS_ISSPARE
, msg
));
2841 return (zfs_error(hdl
, EZFS_ISL2CACHE
, msg
));
2843 verify(nvlist_lookup_uint64(tgt
, ZPOOL_CONFIG_GUID
, &zc
.zc_guid
) == 0);
2845 if (zfs_ioctl(hdl
, ZFS_IOC_VDEV_DETACH
, &zc
) == 0)
2852 * Can't detach from this type of vdev.
2854 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
, "only "
2855 "applicable to mirror and replacing vdevs"));
2856 (void) zfs_error(hdl
, EZFS_BADTARGET
, msg
);
2861 * There are no other replicas of this device.
2863 (void) zfs_error(hdl
, EZFS_NOREPLICAS
, msg
);
2867 (void) zpool_standard_error(hdl
, errno
, msg
);
2874 * Find a mirror vdev in the source nvlist.
2876 * The mchild array contains a list of disks in one of the top-level mirrors
2877 * of the source pool. The schild array contains a list of disks that the
2878 * user specified on the command line. We loop over the mchild array to
2879 * see if any entry in the schild array matches.
2881 * If a disk in the mchild array is found in the schild array, we return
2882 * the index of that entry. Otherwise we return -1.
2885 find_vdev_entry(zpool_handle_t
*zhp
, nvlist_t
**mchild
, uint_t mchildren
,
2886 nvlist_t
**schild
, uint_t schildren
)
2890 for (mc
= 0; mc
< mchildren
; mc
++) {
2892 char *mpath
= zpool_vdev_name(zhp
->zpool_hdl
, zhp
,
2895 for (sc
= 0; sc
< schildren
; sc
++) {
2896 char *spath
= zpool_vdev_name(zhp
->zpool_hdl
, zhp
,
2898 boolean_t result
= (strcmp(mpath
, spath
) == 0);
2914 * Split a mirror pool. If newroot points to null, then a new nvlist
2915 * is generated and it is the responsibility of the caller to free it.
2918 zpool_vdev_split(zpool_handle_t
*zhp
, char *newname
, nvlist_t
**newroot
,
2919 nvlist_t
*props
, splitflags_t flags
)
2921 zfs_cmd_t zc
= {"\0"};
2923 nvlist_t
*tree
, *config
, **child
, **newchild
, *newconfig
= NULL
;
2924 nvlist_t
**varray
= NULL
, *zc_props
= NULL
;
2925 uint_t c
, children
, newchildren
, lastlog
= 0, vcount
, found
= 0;
2926 libzfs_handle_t
*hdl
= zhp
->zpool_hdl
;
2928 boolean_t freelist
= B_FALSE
, memory_err
= B_TRUE
;
2931 (void) snprintf(msg
, sizeof (msg
),
2932 dgettext(TEXT_DOMAIN
, "Unable to split %s"), zhp
->zpool_name
);
2934 if (!zpool_name_valid(hdl
, B_FALSE
, newname
))
2935 return (zfs_error(hdl
, EZFS_INVALIDNAME
, msg
));
2937 if ((config
= zpool_get_config(zhp
, NULL
)) == NULL
) {
2938 (void) fprintf(stderr
, gettext("Internal error: unable to "
2939 "retrieve pool configuration\n"));
2943 verify(nvlist_lookup_nvlist(config
, ZPOOL_CONFIG_VDEV_TREE
, &tree
)
2945 verify(nvlist_lookup_uint64(config
, ZPOOL_CONFIG_VERSION
, &vers
) == 0);
2948 prop_flags_t flags
= { .create
= B_FALSE
, .import
= B_TRUE
};
2949 if ((zc_props
= zpool_valid_proplist(hdl
, zhp
->zpool_name
,
2950 props
, vers
, flags
, msg
)) == NULL
)
2954 if (nvlist_lookup_nvlist_array(tree
, ZPOOL_CONFIG_CHILDREN
, &child
,
2956 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
2957 "Source pool is missing vdev tree"));
2958 nvlist_free(zc_props
);
2962 varray
= zfs_alloc(hdl
, children
* sizeof (nvlist_t
*));
2965 if (*newroot
== NULL
||
2966 nvlist_lookup_nvlist_array(*newroot
, ZPOOL_CONFIG_CHILDREN
,
2967 &newchild
, &newchildren
) != 0)
2970 for (c
= 0; c
< children
; c
++) {
2971 uint64_t is_log
= B_FALSE
, is_hole
= B_FALSE
;
2973 nvlist_t
**mchild
, *vdev
;
2978 * Unlike cache & spares, slogs are stored in the
2979 * ZPOOL_CONFIG_CHILDREN array. We filter them out here.
2981 (void) nvlist_lookup_uint64(child
[c
], ZPOOL_CONFIG_IS_LOG
,
2983 (void) nvlist_lookup_uint64(child
[c
], ZPOOL_CONFIG_IS_HOLE
,
2985 if (is_log
|| is_hole
) {
2987 * Create a hole vdev and put it in the config.
2989 if (nvlist_alloc(&vdev
, NV_UNIQUE_NAME
, 0) != 0)
2991 if (nvlist_add_string(vdev
, ZPOOL_CONFIG_TYPE
,
2992 VDEV_TYPE_HOLE
) != 0)
2994 if (nvlist_add_uint64(vdev
, ZPOOL_CONFIG_IS_HOLE
,
2999 varray
[vcount
++] = vdev
;
3003 verify(nvlist_lookup_string(child
[c
], ZPOOL_CONFIG_TYPE
, &type
)
3005 if (strcmp(type
, VDEV_TYPE_MIRROR
) != 0) {
3006 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
3007 "Source pool must be composed only of mirrors\n"));
3008 retval
= zfs_error(hdl
, EZFS_INVALCONFIG
, msg
);
3012 verify(nvlist_lookup_nvlist_array(child
[c
],
3013 ZPOOL_CONFIG_CHILDREN
, &mchild
, &mchildren
) == 0);
3015 /* find or add an entry for this top-level vdev */
3016 if (newchildren
> 0 &&
3017 (entry
= find_vdev_entry(zhp
, mchild
, mchildren
,
3018 newchild
, newchildren
)) >= 0) {
3019 /* We found a disk that the user specified. */
3020 vdev
= mchild
[entry
];
3023 /* User didn't specify a disk for this vdev. */
3024 vdev
= mchild
[mchildren
- 1];
3027 if (nvlist_dup(vdev
, &varray
[vcount
++], 0) != 0)
3031 /* did we find every disk the user specified? */
3032 if (found
!= newchildren
) {
3033 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
, "Device list must "
3034 "include at most one disk from each mirror"));
3035 retval
= zfs_error(hdl
, EZFS_INVALCONFIG
, msg
);
3039 /* Prepare the nvlist for populating. */
3040 if (*newroot
== NULL
) {
3041 if (nvlist_alloc(newroot
, NV_UNIQUE_NAME
, 0) != 0)
3044 if (nvlist_add_string(*newroot
, ZPOOL_CONFIG_TYPE
,
3045 VDEV_TYPE_ROOT
) != 0)
3048 verify(nvlist_remove_all(*newroot
, ZPOOL_CONFIG_CHILDREN
) == 0);
3051 /* Add all the children we found */
3052 if (nvlist_add_nvlist_array(*newroot
, ZPOOL_CONFIG_CHILDREN
, varray
,
3053 lastlog
== 0 ? vcount
: lastlog
) != 0)
3057 * If we're just doing a dry run, exit now with success.
3060 memory_err
= B_FALSE
;
3065 /* now build up the config list & call the ioctl */
3066 if (nvlist_alloc(&newconfig
, NV_UNIQUE_NAME
, 0) != 0)
3069 if (nvlist_add_nvlist(newconfig
,
3070 ZPOOL_CONFIG_VDEV_TREE
, *newroot
) != 0 ||
3071 nvlist_add_string(newconfig
,
3072 ZPOOL_CONFIG_POOL_NAME
, newname
) != 0 ||
3073 nvlist_add_uint64(newconfig
, ZPOOL_CONFIG_VERSION
, vers
) != 0)
3077 * The new pool is automatically part of the namespace unless we
3078 * explicitly export it.
3081 zc
.zc_cookie
= ZPOOL_EXPORT_AFTER_SPLIT
;
3082 (void) strlcpy(zc
.zc_name
, zhp
->zpool_name
, sizeof (zc
.zc_name
));
3083 (void) strlcpy(zc
.zc_string
, newname
, sizeof (zc
.zc_string
));
3084 if (zcmd_write_conf_nvlist(hdl
, &zc
, newconfig
) != 0)
3086 if (zc_props
!= NULL
&& zcmd_write_src_nvlist(hdl
, &zc
, zc_props
) != 0)
3089 if (zfs_ioctl(hdl
, ZFS_IOC_VDEV_SPLIT
, &zc
) != 0) {
3090 retval
= zpool_standard_error(hdl
, errno
, msg
);
3095 memory_err
= B_FALSE
;
3098 if (varray
!= NULL
) {
3101 for (v
= 0; v
< vcount
; v
++)
3102 nvlist_free(varray
[v
]);
3105 zcmd_free_nvlists(&zc
);
3106 nvlist_free(zc_props
);
3107 nvlist_free(newconfig
);
3109 nvlist_free(*newroot
);
3117 return (no_memory(hdl
));
3123 * Remove the given device. Currently, this is supported only for hot spares
3124 * and level 2 cache devices.
3127 zpool_vdev_remove(zpool_handle_t
*zhp
, const char *path
)
3129 zfs_cmd_t zc
= {"\0"};
3132 boolean_t avail_spare
, l2cache
, islog
;
3133 libzfs_handle_t
*hdl
= zhp
->zpool_hdl
;
3136 (void) snprintf(msg
, sizeof (msg
),
3137 dgettext(TEXT_DOMAIN
, "cannot remove %s"), path
);
3139 (void) strlcpy(zc
.zc_name
, zhp
->zpool_name
, sizeof (zc
.zc_name
));
3140 if ((tgt
= zpool_find_vdev(zhp
, path
, &avail_spare
, &l2cache
,
3142 return (zfs_error(hdl
, EZFS_NODEVICE
, msg
));
3144 * XXX - this should just go away.
3146 if (!avail_spare
&& !l2cache
&& !islog
) {
3147 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
3148 "only inactive hot spares, cache, top-level, "
3149 "or log devices can be removed"));
3150 return (zfs_error(hdl
, EZFS_NODEVICE
, msg
));
3153 version
= zpool_get_prop_int(zhp
, ZPOOL_PROP_VERSION
, NULL
);
3154 if (islog
&& version
< SPA_VERSION_HOLES
) {
3155 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
3156 "pool must be upgrade to support log removal"));
3157 return (zfs_error(hdl
, EZFS_BADVERSION
, msg
));
3160 verify(nvlist_lookup_uint64(tgt
, ZPOOL_CONFIG_GUID
, &zc
.zc_guid
) == 0);
3162 if (zfs_ioctl(hdl
, ZFS_IOC_VDEV_REMOVE
, &zc
) == 0)
3165 return (zpool_standard_error(hdl
, errno
, msg
));
3169 * Clear the errors for the pool, or the particular device if specified.
3172 zpool_clear(zpool_handle_t
*zhp
, const char *path
, nvlist_t
*rewindnvl
)
3174 zfs_cmd_t zc
= {"\0"};
3177 zpool_rewind_policy_t policy
;
3178 boolean_t avail_spare
, l2cache
;
3179 libzfs_handle_t
*hdl
= zhp
->zpool_hdl
;
3180 nvlist_t
*nvi
= NULL
;
3184 (void) snprintf(msg
, sizeof (msg
),
3185 dgettext(TEXT_DOMAIN
, "cannot clear errors for %s"),
3188 (void) snprintf(msg
, sizeof (msg
),
3189 dgettext(TEXT_DOMAIN
, "cannot clear errors for %s"),
3192 (void) strlcpy(zc
.zc_name
, zhp
->zpool_name
, sizeof (zc
.zc_name
));
3194 if ((tgt
= zpool_find_vdev(zhp
, path
, &avail_spare
,
3195 &l2cache
, NULL
)) == 0)
3196 return (zfs_error(hdl
, EZFS_NODEVICE
, msg
));
3199 * Don't allow error clearing for hot spares. Do allow
3200 * error clearing for l2cache devices.
3203 return (zfs_error(hdl
, EZFS_ISSPARE
, msg
));
3205 verify(nvlist_lookup_uint64(tgt
, ZPOOL_CONFIG_GUID
,
3209 zpool_get_rewind_policy(rewindnvl
, &policy
);
3210 zc
.zc_cookie
= policy
.zrp_request
;
3212 if (zcmd_alloc_dst_nvlist(hdl
, &zc
, zhp
->zpool_config_size
* 2) != 0)
3215 if (zcmd_write_src_nvlist(hdl
, &zc
, rewindnvl
) != 0)
3218 while ((error
= zfs_ioctl(hdl
, ZFS_IOC_CLEAR
, &zc
)) != 0 &&
3220 if (zcmd_expand_dst_nvlist(hdl
, &zc
) != 0) {
3221 zcmd_free_nvlists(&zc
);
3226 if (!error
|| ((policy
.zrp_request
& ZPOOL_TRY_REWIND
) &&
3227 errno
!= EPERM
&& errno
!= EACCES
)) {
3228 if (policy
.zrp_request
&
3229 (ZPOOL_DO_REWIND
| ZPOOL_TRY_REWIND
)) {
3230 (void) zcmd_read_dst_nvlist(hdl
, &zc
, &nvi
);
3231 zpool_rewind_exclaim(hdl
, zc
.zc_name
,
3232 ((policy
.zrp_request
& ZPOOL_TRY_REWIND
) != 0),
3236 zcmd_free_nvlists(&zc
);
3240 zcmd_free_nvlists(&zc
);
3241 return (zpool_standard_error(hdl
, errno
, msg
));
3245 * Similar to zpool_clear(), but takes a GUID (used by fmd).
3248 zpool_vdev_clear(zpool_handle_t
*zhp
, uint64_t guid
)
3250 zfs_cmd_t zc
= {"\0"};
3252 libzfs_handle_t
*hdl
= zhp
->zpool_hdl
;
3254 (void) snprintf(msg
, sizeof (msg
),
3255 dgettext(TEXT_DOMAIN
, "cannot clear errors for %llx"),
3256 (u_longlong_t
)guid
);
3258 (void) strlcpy(zc
.zc_name
, zhp
->zpool_name
, sizeof (zc
.zc_name
));
3260 zc
.zc_cookie
= ZPOOL_NO_REWIND
;
3262 if (ioctl(hdl
->libzfs_fd
, ZFS_IOC_CLEAR
, &zc
) == 0)
3265 return (zpool_standard_error(hdl
, errno
, msg
));
3269 * Change the GUID for a pool.
3272 zpool_reguid(zpool_handle_t
*zhp
)
3275 libzfs_handle_t
*hdl
= zhp
->zpool_hdl
;
3276 zfs_cmd_t zc
= {"\0"};
3278 (void) snprintf(msg
, sizeof (msg
),
3279 dgettext(TEXT_DOMAIN
, "cannot reguid '%s'"), zhp
->zpool_name
);
3281 (void) strlcpy(zc
.zc_name
, zhp
->zpool_name
, sizeof (zc
.zc_name
));
3282 if (zfs_ioctl(hdl
, ZFS_IOC_POOL_REGUID
, &zc
) == 0)
3285 return (zpool_standard_error(hdl
, errno
, msg
));
3292 zpool_reopen(zpool_handle_t
*zhp
)
3294 zfs_cmd_t zc
= {"\0"};
3296 libzfs_handle_t
*hdl
= zhp
->zpool_hdl
;
3298 (void) snprintf(msg
, sizeof (msg
),
3299 dgettext(TEXT_DOMAIN
, "cannot reopen '%s'"),
3302 (void) strlcpy(zc
.zc_name
, zhp
->zpool_name
, sizeof (zc
.zc_name
));
3303 if (zfs_ioctl(hdl
, ZFS_IOC_POOL_REOPEN
, &zc
) == 0)
3305 return (zpool_standard_error(hdl
, errno
, msg
));
3308 #if defined(__sun__) || defined(__sun)
3310 * Convert from a devid string to a path.
3313 devid_to_path(char *devid_str
)
3318 devid_nmlist_t
*list
= NULL
;
3321 if (devid_str_decode(devid_str
, &devid
, &minor
) != 0)
3324 ret
= devid_deviceid_to_nmlist("/dev", devid
, minor
, &list
);
3326 devid_str_free(minor
);
3333 * In a case the strdup() fails, we will just return NULL below.
3335 path
= strdup(list
[0].devname
);
3337 devid_free_nmlist(list
);
3343 * Convert from a path to a devid string.
3346 path_to_devid(const char *path
)
3352 if ((fd
= open(path
, O_RDONLY
)) < 0)
3357 if (devid_get(fd
, &devid
) == 0) {
3358 if (devid_get_minor_name(fd
, &minor
) == 0)
3359 ret
= devid_str_encode(devid
, minor
);
3361 devid_str_free(minor
);
3370 * Issue the necessary ioctl() to update the stored path value for the vdev. We
3371 * ignore any failure here, since a common case is for an unprivileged user to
3372 * type 'zpool status', and we'll display the correct information anyway.
3375 set_path(zpool_handle_t
*zhp
, nvlist_t
*nv
, const char *path
)
3377 zfs_cmd_t zc
= {"\0"};
3379 (void) strncpy(zc
.zc_name
, zhp
->zpool_name
, sizeof (zc
.zc_name
));
3380 (void) strncpy(zc
.zc_value
, path
, sizeof (zc
.zc_value
));
3381 verify(nvlist_lookup_uint64(nv
, ZPOOL_CONFIG_GUID
,
3384 (void) ioctl(zhp
->zpool_hdl
->libzfs_fd
, ZFS_IOC_VDEV_SETPATH
, &zc
);
3389 * Remove partition suffix from a vdev path. Partition suffixes may take three
3390 * forms: "-partX", "pX", or "X", where X is a string of digits. The second
3391 * case only occurs when the suffix is preceded by a digit, i.e. "md0p0" The
3392 * third case only occurs when preceded by a string matching the regular
3393 * expression "^([hsv]|xv)d[a-z]+", i.e. a scsi, ide, virtio or xen disk.
3396 strip_partition(libzfs_handle_t
*hdl
, char *path
)
3398 char *tmp
= zfs_strdup(hdl
, path
);
3399 char *part
= NULL
, *d
= NULL
;
3401 if ((part
= strstr(tmp
, "-part")) && part
!= tmp
) {
3403 } else if ((part
= strrchr(tmp
, 'p')) &&
3404 part
> tmp
+ 1 && isdigit(*(part
-1))) {
3406 } else if ((tmp
[0] == 'h' || tmp
[0] == 's' || tmp
[0] == 'v') &&
3408 for (d
= &tmp
[2]; isalpha(*d
); part
= ++d
);
3409 } else if (strncmp("xvd", tmp
, 3) == 0) {
3410 for (d
= &tmp
[3]; isalpha(*d
); part
= ++d
);
3412 if (part
&& d
&& *d
!= '\0') {
3413 for (; isdigit(*d
); d
++);
3420 #define PATH_BUF_LEN 64
3423 * Given a vdev, return the name to display in iostat. If the vdev has a path,
3424 * we use that, stripping off any leading "/dev/dsk/"; if not, we use the type.
3425 * We also check if this is a whole disk, in which case we strip off the
3426 * trailing 's0' slice name.
3428 * This routine is also responsible for identifying when disks have been
3429 * reconfigured in a new location. The kernel will have opened the device by
3430 * devid, but the path will still refer to the old location. To catch this, we
3431 * first do a path -> devid translation (which is fast for the common case). If
3432 * the devid matches, we're done. If not, we do a reverse devid -> path
3433 * translation and issue the appropriate ioctl() to update the path of the vdev.
3434 * If 'zhp' is NULL, then this is an exported pool, and we don't need to do any
3438 zpool_vdev_name(libzfs_handle_t
*hdl
, zpool_handle_t
*zhp
, nvlist_t
*nv
,
3441 char *path
, *type
, *env
;
3443 char buf
[PATH_BUF_LEN
];
3444 char tmpbuf
[PATH_BUF_LEN
];
3446 env
= getenv("ZPOOL_VDEV_NAME_PATH");
3447 if (env
&& (strtoul(env
, NULL
, 0) > 0 ||
3448 !strncasecmp(env
, "YES", 3) || !strncasecmp(env
, "ON", 2)))
3449 name_flags
|= VDEV_NAME_PATH
;
3451 env
= getenv("ZPOOL_VDEV_NAME_GUID");
3452 if (env
&& (strtoul(env
, NULL
, 0) > 0 ||
3453 !strncasecmp(env
, "YES", 3) || !strncasecmp(env
, "ON", 2)))
3454 name_flags
|= VDEV_NAME_GUID
;
3456 env
= getenv("ZPOOL_VDEV_NAME_FOLLOW_LINKS");
3457 if (env
&& (strtoul(env
, NULL
, 0) > 0 ||
3458 !strncasecmp(env
, "YES", 3) || !strncasecmp(env
, "ON", 2)))
3459 name_flags
|= VDEV_NAME_FOLLOW_LINKS
;
3461 if (nvlist_lookup_uint64(nv
, ZPOOL_CONFIG_NOT_PRESENT
, &value
) == 0 ||
3462 name_flags
& VDEV_NAME_GUID
) {
3463 nvlist_lookup_uint64(nv
, ZPOOL_CONFIG_GUID
, &value
);
3464 (void) snprintf(buf
, sizeof (buf
), "%llu", (u_longlong_t
)value
);
3466 } else if (nvlist_lookup_string(nv
, ZPOOL_CONFIG_PATH
, &path
) == 0) {
3467 #if defined(__sun__) || defined(__sun)
3469 * Live VDEV path updates to a kernel VDEV during a
3470 * zpool_vdev_name lookup are not supported on Linux.
3477 * If the device is dead (faulted, offline, etc) then don't
3478 * bother opening it. Otherwise we may be forcing the user to
3479 * open a misbehaving device, which can have undesirable
3482 if ((nvlist_lookup_uint64_array(nv
, ZPOOL_CONFIG_VDEV_STATS
,
3483 (uint64_t **)&vs
, &vsc
) != 0 ||
3484 vs
->vs_state
>= VDEV_STATE_DEGRADED
) &&
3486 nvlist_lookup_string(nv
, ZPOOL_CONFIG_DEVID
, &devid
) == 0) {
3488 * Determine if the current path is correct.
3490 char *newdevid
= path_to_devid(path
);
3492 if (newdevid
== NULL
||
3493 strcmp(devid
, newdevid
) != 0) {
3496 if ((newpath
= devid_to_path(devid
)) != NULL
) {
3498 * Update the path appropriately.
3500 set_path(zhp
, nv
, newpath
);
3501 if (nvlist_add_string(nv
,
3502 ZPOOL_CONFIG_PATH
, newpath
) == 0)
3503 verify(nvlist_lookup_string(nv
,
3511 devid_str_free(newdevid
);
3515 if (name_flags
& VDEV_NAME_FOLLOW_LINKS
) {
3516 char *rp
= realpath(path
, NULL
);
3518 strlcpy(buf
, rp
, sizeof (buf
));
3525 * For a block device only use the name.
3527 verify(nvlist_lookup_string(nv
, ZPOOL_CONFIG_TYPE
, &type
) == 0);
3528 if ((strcmp(type
, VDEV_TYPE_DISK
) == 0) &&
3529 !(name_flags
& VDEV_NAME_PATH
)) {
3530 path
= strrchr(path
, '/');
3535 * Remove the partition from the path it this is a whole disk.
3537 if (nvlist_lookup_uint64(nv
, ZPOOL_CONFIG_WHOLE_DISK
, &value
)
3538 == 0 && value
&& !(name_flags
& VDEV_NAME_PATH
)) {
3539 return (strip_partition(hdl
, path
));
3542 verify(nvlist_lookup_string(nv
, ZPOOL_CONFIG_TYPE
, &path
) == 0);
3545 * If it's a raidz device, we need to stick in the parity level.
3547 if (strcmp(path
, VDEV_TYPE_RAIDZ
) == 0) {
3549 verify(nvlist_lookup_uint64(nv
, ZPOOL_CONFIG_NPARITY
,
3551 (void) snprintf(buf
, sizeof (buf
), "%s%llu", path
,
3552 (u_longlong_t
)value
);
3557 * We identify each top-level vdev by using a <type-id>
3558 * naming convention.
3560 if (name_flags
& VDEV_NAME_TYPE_ID
) {
3563 verify(nvlist_lookup_uint64(nv
, ZPOOL_CONFIG_ID
,
3565 (void) snprintf(tmpbuf
, sizeof (tmpbuf
), "%s-%llu",
3566 path
, (u_longlong_t
)id
);
3571 return (zfs_strdup(hdl
, path
));
3575 zbookmark_mem_compare(const void *a
, const void *b
)
3577 return (memcmp(a
, b
, sizeof (zbookmark_phys_t
)));
3581 * Retrieve the persistent error log, uniquify the members, and return to the
3585 zpool_get_errlog(zpool_handle_t
*zhp
, nvlist_t
**nverrlistp
)
3587 zfs_cmd_t zc
= {"\0"};
3589 zbookmark_phys_t
*zb
= NULL
;
3593 * Retrieve the raw error list from the kernel. If the number of errors
3594 * has increased, allocate more space and continue until we get the
3597 verify(nvlist_lookup_uint64(zhp
->zpool_config
, ZPOOL_CONFIG_ERRCOUNT
,
3601 if ((zc
.zc_nvlist_dst
= (uintptr_t)zfs_alloc(zhp
->zpool_hdl
,
3602 count
* sizeof (zbookmark_phys_t
))) == (uintptr_t)NULL
)
3604 zc
.zc_nvlist_dst_size
= count
;
3605 (void) strcpy(zc
.zc_name
, zhp
->zpool_name
);
3607 if (ioctl(zhp
->zpool_hdl
->libzfs_fd
, ZFS_IOC_ERROR_LOG
,
3609 free((void *)(uintptr_t)zc
.zc_nvlist_dst
);
3610 if (errno
== ENOMEM
) {
3613 count
= zc
.zc_nvlist_dst_size
;
3614 dst
= zfs_alloc(zhp
->zpool_hdl
, count
*
3615 sizeof (zbookmark_phys_t
));
3618 zc
.zc_nvlist_dst
= (uintptr_t)dst
;
3628 * Sort the resulting bookmarks. This is a little confusing due to the
3629 * implementation of ZFS_IOC_ERROR_LOG. The bookmarks are copied last
3630 * to first, and 'zc_nvlist_dst_size' indicates the number of boomarks
3631 * _not_ copied as part of the process. So we point the start of our
3632 * array appropriate and decrement the total number of elements.
3634 zb
= ((zbookmark_phys_t
*)(uintptr_t)zc
.zc_nvlist_dst
) +
3635 zc
.zc_nvlist_dst_size
;
3636 count
-= zc
.zc_nvlist_dst_size
;
3638 qsort(zb
, count
, sizeof (zbookmark_phys_t
), zbookmark_mem_compare
);
3640 verify(nvlist_alloc(nverrlistp
, 0, KM_SLEEP
) == 0);
3643 * Fill in the nverrlistp with nvlist's of dataset and object numbers.
3645 for (i
= 0; i
< count
; i
++) {
3648 /* ignoring zb_blkid and zb_level for now */
3649 if (i
> 0 && zb
[i
-1].zb_objset
== zb
[i
].zb_objset
&&
3650 zb
[i
-1].zb_object
== zb
[i
].zb_object
)
3653 if (nvlist_alloc(&nv
, NV_UNIQUE_NAME
, KM_SLEEP
) != 0)
3655 if (nvlist_add_uint64(nv
, ZPOOL_ERR_DATASET
,
3656 zb
[i
].zb_objset
) != 0) {
3660 if (nvlist_add_uint64(nv
, ZPOOL_ERR_OBJECT
,
3661 zb
[i
].zb_object
) != 0) {
3665 if (nvlist_add_nvlist(*nverrlistp
, "ejk", nv
) != 0) {
3672 free((void *)(uintptr_t)zc
.zc_nvlist_dst
);
3676 free((void *)(uintptr_t)zc
.zc_nvlist_dst
);
3677 return (no_memory(zhp
->zpool_hdl
));
3681 * Upgrade a ZFS pool to the latest on-disk version.
3684 zpool_upgrade(zpool_handle_t
*zhp
, uint64_t new_version
)
3686 zfs_cmd_t zc
= {"\0"};
3687 libzfs_handle_t
*hdl
= zhp
->zpool_hdl
;
3689 (void) strcpy(zc
.zc_name
, zhp
->zpool_name
);
3690 zc
.zc_cookie
= new_version
;
3692 if (zfs_ioctl(hdl
, ZFS_IOC_POOL_UPGRADE
, &zc
) != 0)
3693 return (zpool_standard_error_fmt(hdl
, errno
,
3694 dgettext(TEXT_DOMAIN
, "cannot upgrade '%s'"),
3700 zfs_save_arguments(int argc
, char **argv
, char *string
, int len
)
3704 (void) strlcpy(string
, basename(argv
[0]), len
);
3705 for (i
= 1; i
< argc
; i
++) {
3706 (void) strlcat(string
, " ", len
);
3707 (void) strlcat(string
, argv
[i
], len
);
3712 zpool_log_history(libzfs_handle_t
*hdl
, const char *message
)
3714 zfs_cmd_t zc
= {"\0"};
3718 args
= fnvlist_alloc();
3719 fnvlist_add_string(args
, "message", message
);
3720 err
= zcmd_write_src_nvlist(hdl
, &zc
, args
);
3722 err
= ioctl(hdl
->libzfs_fd
, ZFS_IOC_LOG_HISTORY
, &zc
);
3724 zcmd_free_nvlists(&zc
);
3729 * Perform ioctl to get some command history of a pool.
3731 * 'buf' is the buffer to fill up to 'len' bytes. 'off' is the
3732 * logical offset of the history buffer to start reading from.
3734 * Upon return, 'off' is the next logical offset to read from and
3735 * 'len' is the actual amount of bytes read into 'buf'.
3738 get_history(zpool_handle_t
*zhp
, char *buf
, uint64_t *off
, uint64_t *len
)
3740 zfs_cmd_t zc
= {"\0"};
3741 libzfs_handle_t
*hdl
= zhp
->zpool_hdl
;
3743 (void) strlcpy(zc
.zc_name
, zhp
->zpool_name
, sizeof (zc
.zc_name
));
3745 zc
.zc_history
= (uint64_t)(uintptr_t)buf
;
3746 zc
.zc_history_len
= *len
;
3747 zc
.zc_history_offset
= *off
;
3749 if (ioctl(hdl
->libzfs_fd
, ZFS_IOC_POOL_GET_HISTORY
, &zc
) != 0) {
3752 return (zfs_error_fmt(hdl
, EZFS_PERM
,
3753 dgettext(TEXT_DOMAIN
,
3754 "cannot show history for pool '%s'"),
3757 return (zfs_error_fmt(hdl
, EZFS_NOHISTORY
,
3758 dgettext(TEXT_DOMAIN
, "cannot get history for pool "
3759 "'%s'"), zhp
->zpool_name
));
3761 return (zfs_error_fmt(hdl
, EZFS_BADVERSION
,
3762 dgettext(TEXT_DOMAIN
, "cannot get history for pool "
3763 "'%s', pool must be upgraded"), zhp
->zpool_name
));
3765 return (zpool_standard_error_fmt(hdl
, errno
,
3766 dgettext(TEXT_DOMAIN
,
3767 "cannot get history for '%s'"), zhp
->zpool_name
));
3771 *len
= zc
.zc_history_len
;
3772 *off
= zc
.zc_history_offset
;
3778 * Process the buffer of nvlists, unpacking and storing each nvlist record
3779 * into 'records'. 'leftover' is set to the number of bytes that weren't
3780 * processed as there wasn't a complete record.
3783 zpool_history_unpack(char *buf
, uint64_t bytes_read
, uint64_t *leftover
,
3784 nvlist_t
***records
, uint_t
*numrecords
)
3790 while (bytes_read
> sizeof (reclen
)) {
3792 /* get length of packed record (stored as little endian) */
3793 for (i
= 0, reclen
= 0; i
< sizeof (reclen
); i
++)
3794 reclen
+= (uint64_t)(((uchar_t
*)buf
)[i
]) << (8*i
);
3796 if (bytes_read
< sizeof (reclen
) + reclen
)
3800 if (nvlist_unpack(buf
+ sizeof (reclen
), reclen
, &nv
, 0) != 0)
3802 bytes_read
-= sizeof (reclen
) + reclen
;
3803 buf
+= sizeof (reclen
) + reclen
;
3805 /* add record to nvlist array */
3807 if (ISP2(*numrecords
+ 1)) {
3808 *records
= realloc(*records
,
3809 *numrecords
* 2 * sizeof (nvlist_t
*));
3811 (*records
)[*numrecords
- 1] = nv
;
3814 *leftover
= bytes_read
;
3819 * Retrieve the command history of a pool.
3822 zpool_get_history(zpool_handle_t
*zhp
, nvlist_t
**nvhisp
)
3825 int buflen
= 128 * 1024;
3827 nvlist_t
**records
= NULL
;
3828 uint_t numrecords
= 0;
3831 buf
= malloc(buflen
);
3835 uint64_t bytes_read
= buflen
;
3838 if ((err
= get_history(zhp
, buf
, &off
, &bytes_read
)) != 0)
3841 /* if nothing else was read in, we're at EOF, just return */
3845 if ((err
= zpool_history_unpack(buf
, bytes_read
,
3846 &leftover
, &records
, &numrecords
)) != 0)
3849 if (leftover
== bytes_read
) {
3851 * no progress made, because buffer is not big enough
3852 * to hold this record; resize and retry.
3856 buf
= malloc(buflen
);
3867 verify(nvlist_alloc(nvhisp
, NV_UNIQUE_NAME
, 0) == 0);
3868 verify(nvlist_add_nvlist_array(*nvhisp
, ZPOOL_HIST_RECORD
,
3869 records
, numrecords
) == 0);
3871 for (i
= 0; i
< numrecords
; i
++)
3872 nvlist_free(records
[i
]);
3879 * Retrieve the next event given the passed 'zevent_fd' file descriptor.
3880 * If there is a new event available 'nvp' will contain a newly allocated
3881 * nvlist and 'dropped' will be set to the number of missed events since
3882 * the last call to this function. When 'nvp' is set to NULL it indicates
3883 * no new events are available. In either case the function returns 0 and
3884 * it is up to the caller to free 'nvp'. In the case of a fatal error the
3885 * function will return a non-zero value. When the function is called in
3886 * blocking mode (the default, unless the ZEVENT_NONBLOCK flag is passed),
3887 * it will not return until a new event is available.
3890 zpool_events_next(libzfs_handle_t
*hdl
, nvlist_t
**nvp
,
3891 int *dropped
, unsigned flags
, int zevent_fd
)
3893 zfs_cmd_t zc
= {"\0"};
3898 zc
.zc_cleanup_fd
= zevent_fd
;
3900 if (flags
& ZEVENT_NONBLOCK
)
3901 zc
.zc_guid
= ZEVENT_NONBLOCK
;
3903 if (zcmd_alloc_dst_nvlist(hdl
, &zc
, ZEVENT_SIZE
) != 0)
3907 if (zfs_ioctl(hdl
, ZFS_IOC_EVENTS_NEXT
, &zc
) != 0) {
3910 error
= zfs_error_fmt(hdl
, EZFS_POOLUNAVAIL
,
3911 dgettext(TEXT_DOMAIN
, "zfs shutdown"));
3914 /* Blocking error case should not occur */
3915 if (!(flags
& ZEVENT_NONBLOCK
))
3916 error
= zpool_standard_error_fmt(hdl
, errno
,
3917 dgettext(TEXT_DOMAIN
, "cannot get event"));
3921 if (zcmd_expand_dst_nvlist(hdl
, &zc
) != 0) {
3922 error
= zfs_error_fmt(hdl
, EZFS_NOMEM
,
3923 dgettext(TEXT_DOMAIN
, "cannot get event"));
3929 error
= zpool_standard_error_fmt(hdl
, errno
,
3930 dgettext(TEXT_DOMAIN
, "cannot get event"));
3935 error
= zcmd_read_dst_nvlist(hdl
, &zc
, nvp
);
3939 *dropped
= (int)zc
.zc_cookie
;
3941 zcmd_free_nvlists(&zc
);
3950 zpool_events_clear(libzfs_handle_t
*hdl
, int *count
)
3952 zfs_cmd_t zc
= {"\0"};
3955 (void) snprintf(msg
, sizeof (msg
), dgettext(TEXT_DOMAIN
,
3956 "cannot clear events"));
3958 if (zfs_ioctl(hdl
, ZFS_IOC_EVENTS_CLEAR
, &zc
) != 0)
3959 return (zpool_standard_error_fmt(hdl
, errno
, msg
));
3962 *count
= (int)zc
.zc_cookie
; /* # of events cleared */
3968 * Seek to a specific EID, ZEVENT_SEEK_START, or ZEVENT_SEEK_END for
3969 * the passed zevent_fd file handle. On success zero is returned,
3970 * otherwise -1 is returned and hdl->libzfs_error is set to the errno.
3973 zpool_events_seek(libzfs_handle_t
*hdl
, uint64_t eid
, int zevent_fd
)
3975 zfs_cmd_t zc
= {"\0"};
3979 zc
.zc_cleanup_fd
= zevent_fd
;
3981 if (zfs_ioctl(hdl
, ZFS_IOC_EVENTS_SEEK
, &zc
) != 0) {
3984 error
= zfs_error_fmt(hdl
, EZFS_NOENT
,
3985 dgettext(TEXT_DOMAIN
, "cannot get event"));
3989 error
= zfs_error_fmt(hdl
, EZFS_NOMEM
,
3990 dgettext(TEXT_DOMAIN
, "cannot get event"));
3994 error
= zpool_standard_error_fmt(hdl
, errno
,
3995 dgettext(TEXT_DOMAIN
, "cannot get event"));
4004 zpool_obj_to_path(zpool_handle_t
*zhp
, uint64_t dsobj
, uint64_t obj
,
4005 char *pathname
, size_t len
)
4007 zfs_cmd_t zc
= {"\0"};
4008 boolean_t mounted
= B_FALSE
;
4009 char *mntpnt
= NULL
;
4010 char dsname
[MAXNAMELEN
];
4013 /* special case for the MOS */
4014 (void) snprintf(pathname
, len
, "<metadata>:<0x%llx>",
4019 /* get the dataset's name */
4020 (void) strlcpy(zc
.zc_name
, zhp
->zpool_name
, sizeof (zc
.zc_name
));
4022 if (ioctl(zhp
->zpool_hdl
->libzfs_fd
,
4023 ZFS_IOC_DSOBJ_TO_DSNAME
, &zc
) != 0) {
4024 /* just write out a path of two object numbers */
4025 (void) snprintf(pathname
, len
, "<0x%llx>:<0x%llx>",
4026 (longlong_t
)dsobj
, (longlong_t
)obj
);
4029 (void) strlcpy(dsname
, zc
.zc_value
, sizeof (dsname
));
4031 /* find out if the dataset is mounted */
4032 mounted
= is_mounted(zhp
->zpool_hdl
, dsname
, &mntpnt
);
4034 /* get the corrupted object's path */
4035 (void) strlcpy(zc
.zc_name
, dsname
, sizeof (zc
.zc_name
));
4037 if (ioctl(zhp
->zpool_hdl
->libzfs_fd
, ZFS_IOC_OBJ_TO_PATH
,
4040 (void) snprintf(pathname
, len
, "%s%s", mntpnt
,
4043 (void) snprintf(pathname
, len
, "%s:%s",
4044 dsname
, zc
.zc_value
);
4047 (void) snprintf(pathname
, len
, "%s:<0x%llx>", dsname
,
4054 * Read the EFI label from the config, if a label does not exist then
4055 * pass back the error to the caller. If the caller has passed a non-NULL
4056 * diskaddr argument then we set it to the starting address of the EFI
4060 read_efi_label(nvlist_t
*config
, diskaddr_t
*sb
)
4064 char diskname
[MAXPATHLEN
];
4067 if (nvlist_lookup_string(config
, ZPOOL_CONFIG_PATH
, &path
) != 0)
4070 (void) snprintf(diskname
, sizeof (diskname
), "%s%s", DISK_ROOT
,
4071 strrchr(path
, '/'));
4072 if ((fd
= open(diskname
, O_RDWR
|O_DIRECT
)) >= 0) {
4073 struct dk_gpt
*vtoc
;
4075 if ((err
= efi_alloc_and_read(fd
, &vtoc
)) >= 0) {
4077 *sb
= vtoc
->efi_parts
[0].p_start
;
4086 * determine where a partition starts on a disk in the current
4090 find_start_block(nvlist_t
*config
)
4094 diskaddr_t sb
= MAXOFFSET_T
;
4097 if (nvlist_lookup_nvlist_array(config
,
4098 ZPOOL_CONFIG_CHILDREN
, &child
, &children
) != 0) {
4099 if (nvlist_lookup_uint64(config
,
4100 ZPOOL_CONFIG_WHOLE_DISK
,
4101 &wholedisk
) != 0 || !wholedisk
) {
4102 return (MAXOFFSET_T
);
4104 if (read_efi_label(config
, &sb
) < 0)
4109 for (c
= 0; c
< children
; c
++) {
4110 sb
= find_start_block(child
[c
]);
4111 if (sb
!= MAXOFFSET_T
) {
4115 return (MAXOFFSET_T
);
4119 zpool_label_disk_check(char *path
)
4121 struct dk_gpt
*vtoc
;
4124 if ((fd
= open(path
, O_RDWR
|O_DIRECT
)) < 0)
4127 if ((err
= efi_alloc_and_read(fd
, &vtoc
)) != 0) {
4132 if (vtoc
->efi_flags
& EFI_GPT_PRIMARY_CORRUPT
) {
4144 * Generate a unique partition name for the ZFS member. Partitions must
4145 * have unique names to ensure udev will be able to create symlinks under
4146 * /dev/disk/by-partlabel/ for all pool members. The partition names are
4147 * of the form <pool>-<unique-id>.
4150 zpool_label_name(char *label_name
, int label_size
)
4155 fd
= open("/dev/urandom", O_RDONLY
);
4157 if (read(fd
, &id
, sizeof (id
)) != sizeof (id
))
4164 id
= (((uint64_t)rand()) << 32) | (uint64_t)rand();
4166 snprintf(label_name
, label_size
, "zfs-%016llx", (u_longlong_t
) id
);
4170 * Label an individual disk. The name provided is the short name,
4171 * stripped of any leading /dev path.
4174 zpool_label_disk(libzfs_handle_t
*hdl
, zpool_handle_t
*zhp
, char *name
)
4176 char path
[MAXPATHLEN
];
4177 struct dk_gpt
*vtoc
;
4179 size_t resv
= EFI_MIN_RESV_SIZE
;
4180 uint64_t slice_size
;
4181 diskaddr_t start_block
;
4184 /* prepare an error message just in case */
4185 (void) snprintf(errbuf
, sizeof (errbuf
),
4186 dgettext(TEXT_DOMAIN
, "cannot label '%s'"), name
);
4191 #if defined(__sun__) || defined(__sun)
4192 if (zpool_is_bootable(zhp
)) {
4193 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
4194 "EFI labeled devices are not supported on root "
4196 return (zfs_error(hdl
, EZFS_POOL_NOTSUP
, errbuf
));
4200 verify(nvlist_lookup_nvlist(zhp
->zpool_config
,
4201 ZPOOL_CONFIG_VDEV_TREE
, &nvroot
) == 0);
4203 if (zhp
->zpool_start_block
== 0)
4204 start_block
= find_start_block(nvroot
);
4206 start_block
= zhp
->zpool_start_block
;
4207 zhp
->zpool_start_block
= start_block
;
4210 start_block
= NEW_START_BLOCK
;
4213 (void) snprintf(path
, sizeof (path
), "%s/%s", DISK_ROOT
, name
);
4215 if ((fd
= open(path
, O_RDWR
|O_DIRECT
)) < 0) {
4217 * This shouldn't happen. We've long since verified that this
4218 * is a valid device.
4220 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
, "cannot "
4221 "label '%s': unable to open device: %d"), path
, errno
);
4222 return (zfs_error(hdl
, EZFS_OPENFAILED
, errbuf
));
4225 if (efi_alloc_and_init(fd
, EFI_NUMPAR
, &vtoc
) != 0) {
4227 * The only way this can fail is if we run out of memory, or we
4228 * were unable to read the disk's capacity
4230 if (errno
== ENOMEM
)
4231 (void) no_memory(hdl
);
4234 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
, "cannot "
4235 "label '%s': unable to read disk capacity"), path
);
4237 return (zfs_error(hdl
, EZFS_NOCAP
, errbuf
));
4240 slice_size
= vtoc
->efi_last_u_lba
+ 1;
4241 slice_size
-= EFI_MIN_RESV_SIZE
;
4242 if (start_block
== MAXOFFSET_T
)
4243 start_block
= NEW_START_BLOCK
;
4244 slice_size
-= start_block
;
4245 slice_size
= P2ALIGN(slice_size
, PARTITION_END_ALIGNMENT
);
4247 vtoc
->efi_parts
[0].p_start
= start_block
;
4248 vtoc
->efi_parts
[0].p_size
= slice_size
;
4251 * Why we use V_USR: V_BACKUP confuses users, and is considered
4252 * disposable by some EFI utilities (since EFI doesn't have a backup
4253 * slice). V_UNASSIGNED is supposed to be used only for zero size
4254 * partitions, and efi_write() will fail if we use it. V_ROOT, V_BOOT,
4255 * etc. were all pretty specific. V_USR is as close to reality as we
4256 * can get, in the absence of V_OTHER.
4258 vtoc
->efi_parts
[0].p_tag
= V_USR
;
4259 zpool_label_name(vtoc
->efi_parts
[0].p_name
, EFI_PART_NAME_LEN
);
4261 vtoc
->efi_parts
[8].p_start
= slice_size
+ start_block
;
4262 vtoc
->efi_parts
[8].p_size
= resv
;
4263 vtoc
->efi_parts
[8].p_tag
= V_RESERVED
;
4265 if ((rval
= efi_write(fd
, vtoc
)) != 0 || (rval
= efi_rescan(fd
)) != 0) {
4267 * Some block drivers (like pcata) may not support EFI
4268 * GPT labels. Print out a helpful error message dir-
4269 * ecting the user to manually label the disk and give
4275 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
, "try using "
4276 "parted(8) and then provide a specific slice: %d"), rval
);
4277 return (zfs_error(hdl
, EZFS_LABELFAILED
, errbuf
));
4283 (void) snprintf(path
, sizeof (path
), "%s/%s", DISK_ROOT
, name
);
4284 (void) zfs_append_partition(path
, MAXPATHLEN
);
4286 /* Wait to udev to signal use the device has settled. */
4287 rval
= zpool_label_disk_wait(path
, DISK_LABEL_WAIT
);
4289 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
, "failed to "
4290 "detect device partitions on '%s': %d"), path
, rval
);
4291 return (zfs_error(hdl
, EZFS_LABELFAILED
, errbuf
));
4294 /* We can't be to paranoid. Read the label back and verify it. */
4295 (void) snprintf(path
, sizeof (path
), "%s/%s", DISK_ROOT
, name
);
4296 rval
= zpool_label_disk_check(path
);
4298 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
, "freshly written "
4299 "EFI label on '%s' is damaged. Ensure\nthis device "
4300 "is not in in use, and is functioning properly: %d"),
4302 return (zfs_error(hdl
, EZFS_LABELFAILED
, errbuf
));