4 * The contents of this file are subject to the terms of the
5 * Common Development and Distribution License (the "License").
6 * You may not use this file except in compliance with the License.
8 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
9 * or http://www.opensolaris.org/os/licensing.
10 * See the License for the specific language governing permissions
11 * and limitations under the License.
13 * When distributing Covered Code, include this CDDL HEADER in each
14 * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
15 * If applicable, add the following below this CDDL HEADER, with the
16 * fields enclosed by brackets "[]" replaced with your own identifying
17 * information: Portions Copyright [yyyy] [name of copyright owner]
23 * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
24 * Copyright 2011 Nexenta Systems, Inc. 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
);
718 nvlist_free(retprops
);
723 * Set zpool property : propname=propval.
726 zpool_set_prop(zpool_handle_t
*zhp
, const char *propname
, const char *propval
)
728 zfs_cmd_t zc
= {"\0"};
731 nvlist_t
*nvl
= NULL
;
734 prop_flags_t flags
= { 0 };
736 (void) snprintf(errbuf
, sizeof (errbuf
),
737 dgettext(TEXT_DOMAIN
, "cannot set property for '%s'"),
740 if (nvlist_alloc(&nvl
, NV_UNIQUE_NAME
, 0) != 0)
741 return (no_memory(zhp
->zpool_hdl
));
743 if (nvlist_add_string(nvl
, propname
, propval
) != 0) {
745 return (no_memory(zhp
->zpool_hdl
));
748 version
= zpool_get_prop_int(zhp
, ZPOOL_PROP_VERSION
, NULL
);
749 if ((realprops
= zpool_valid_proplist(zhp
->zpool_hdl
,
750 zhp
->zpool_name
, nvl
, version
, flags
, errbuf
)) == NULL
) {
759 * Execute the corresponding ioctl() to set this property.
761 (void) strlcpy(zc
.zc_name
, zhp
->zpool_name
, sizeof (zc
.zc_name
));
763 if (zcmd_write_src_nvlist(zhp
->zpool_hdl
, &zc
, nvl
) != 0) {
768 ret
= zfs_ioctl(zhp
->zpool_hdl
, ZFS_IOC_POOL_SET_PROPS
, &zc
);
770 zcmd_free_nvlists(&zc
);
774 (void) zpool_standard_error(zhp
->zpool_hdl
, errno
, errbuf
);
776 (void) zpool_props_refresh(zhp
);
782 zpool_expand_proplist(zpool_handle_t
*zhp
, zprop_list_t
**plp
)
784 libzfs_handle_t
*hdl
= zhp
->zpool_hdl
;
786 char buf
[ZFS_MAXPROPLEN
];
787 nvlist_t
*features
= NULL
;
790 boolean_t firstexpand
= (NULL
== *plp
);
793 if (zprop_expand_list(hdl
, plp
, ZFS_TYPE_POOL
) != 0)
797 while (*last
!= NULL
)
798 last
= &(*last
)->pl_next
;
801 features
= zpool_get_features(zhp
);
803 if ((*plp
)->pl_all
&& firstexpand
) {
804 for (i
= 0; i
< SPA_FEATURES
; i
++) {
805 zprop_list_t
*entry
= zfs_alloc(hdl
,
806 sizeof (zprop_list_t
));
807 entry
->pl_prop
= ZPROP_INVAL
;
808 entry
->pl_user_prop
= zfs_asprintf(hdl
, "feature@%s",
809 spa_feature_table
[i
].fi_uname
);
810 entry
->pl_width
= strlen(entry
->pl_user_prop
);
811 entry
->pl_all
= B_TRUE
;
814 last
= &entry
->pl_next
;
818 /* add any unsupported features */
819 for (nvp
= nvlist_next_nvpair(features
, NULL
);
820 nvp
!= NULL
; nvp
= nvlist_next_nvpair(features
, nvp
)) {
825 if (zfeature_is_supported(nvpair_name(nvp
)))
828 propname
= zfs_asprintf(hdl
, "unsupported@%s",
832 * Before adding the property to the list make sure that no
833 * other pool already added the same property.
837 while (entry
!= NULL
) {
838 if (entry
->pl_user_prop
!= NULL
&&
839 strcmp(propname
, entry
->pl_user_prop
) == 0) {
843 entry
= entry
->pl_next
;
850 entry
= zfs_alloc(hdl
, sizeof (zprop_list_t
));
851 entry
->pl_prop
= ZPROP_INVAL
;
852 entry
->pl_user_prop
= propname
;
853 entry
->pl_width
= strlen(entry
->pl_user_prop
);
854 entry
->pl_all
= B_TRUE
;
857 last
= &entry
->pl_next
;
860 for (entry
= *plp
; entry
!= NULL
; entry
= entry
->pl_next
) {
865 if (entry
->pl_prop
!= ZPROP_INVAL
&&
866 zpool_get_prop(zhp
, entry
->pl_prop
, buf
, sizeof (buf
),
868 if (strlen(buf
) > entry
->pl_width
)
869 entry
->pl_width
= strlen(buf
);
877 * Get the state for the given feature on the given ZFS pool.
880 zpool_prop_get_feature(zpool_handle_t
*zhp
, const char *propname
, char *buf
,
884 boolean_t found
= B_FALSE
;
885 nvlist_t
*features
= zpool_get_features(zhp
);
887 const char *feature
= strchr(propname
, '@') + 1;
889 supported
= zpool_prop_feature(propname
);
890 ASSERT(supported
|| zpool_prop_unsupported(propname
));
893 * Convert from feature name to feature guid. This conversion is
894 * unecessary for unsupported@... properties because they already
901 ret
= zfeature_lookup_name(feature
, &fid
);
903 (void) strlcpy(buf
, "-", len
);
906 feature
= spa_feature_table
[fid
].fi_guid
;
909 if (nvlist_lookup_uint64(features
, feature
, &refcount
) == 0)
914 (void) strlcpy(buf
, ZFS_FEATURE_DISABLED
, len
);
917 (void) strlcpy(buf
, ZFS_FEATURE_ENABLED
, len
);
919 (void) strlcpy(buf
, ZFS_FEATURE_ACTIVE
, len
);
924 (void) strcpy(buf
, ZFS_UNSUPPORTED_INACTIVE
);
926 (void) strcpy(buf
, ZFS_UNSUPPORTED_READONLY
);
929 (void) strlcpy(buf
, "-", len
);
938 * Don't start the slice at the default block of 34; many storage
939 * devices will use a stripe width of 128k, other vendors prefer a 1m
940 * alignment. It is best to play it safe and ensure a 1m alignment
941 * given 512B blocks. When the block size is larger by a power of 2
942 * we will still be 1m aligned. Some devices are sensitive to the
943 * partition ending alignment as well.
945 #define NEW_START_BLOCK 2048
946 #define PARTITION_END_ALIGNMENT 2048
949 * Validate the given pool name, optionally putting an extended error message in
953 zpool_name_valid(libzfs_handle_t
*hdl
, boolean_t isopen
, const char *pool
)
959 ret
= pool_namecheck(pool
, &why
, &what
);
962 * The rules for reserved pool names were extended at a later point.
963 * But we need to support users with existing pools that may now be
964 * invalid. So we only check for this expanded set of names during a
965 * create (or import), and only in userland.
967 if (ret
== 0 && !isopen
&&
968 (strncmp(pool
, "mirror", 6) == 0 ||
969 strncmp(pool
, "raidz", 5) == 0 ||
970 strncmp(pool
, "spare", 5) == 0 ||
971 strcmp(pool
, "log") == 0)) {
974 dgettext(TEXT_DOMAIN
, "name is reserved"));
982 case NAME_ERR_TOOLONG
:
984 dgettext(TEXT_DOMAIN
, "name is too long"));
987 case NAME_ERR_INVALCHAR
:
989 dgettext(TEXT_DOMAIN
, "invalid character "
990 "'%c' in pool name"), what
);
993 case NAME_ERR_NOLETTER
:
994 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
995 "name must begin with a letter"));
998 case NAME_ERR_RESERVED
:
999 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
1000 "name is reserved"));
1003 case NAME_ERR_DISKLIKE
:
1004 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
1005 "pool name is reserved"));
1008 case NAME_ERR_LEADING_SLASH
:
1009 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
1010 "leading slash in name"));
1013 case NAME_ERR_EMPTY_COMPONENT
:
1014 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
1015 "empty component in name"));
1018 case NAME_ERR_TRAILING_SLASH
:
1019 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
1020 "trailing slash in name"));
1023 case NAME_ERR_MULTIPLE_AT
:
1024 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
1025 "multiple '@' delimiters in name"));
1027 case NAME_ERR_NO_AT
:
1028 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
1029 "permission set is missing '@'"));
1040 * Open a handle to the given pool, even if the pool is currently in the FAULTED
1044 zpool_open_canfail(libzfs_handle_t
*hdl
, const char *pool
)
1046 zpool_handle_t
*zhp
;
1050 * Make sure the pool name is valid.
1052 if (!zpool_name_valid(hdl
, B_TRUE
, pool
)) {
1053 (void) zfs_error_fmt(hdl
, EZFS_INVALIDNAME
,
1054 dgettext(TEXT_DOMAIN
, "cannot open '%s'"),
1059 if ((zhp
= zfs_alloc(hdl
, sizeof (zpool_handle_t
))) == NULL
)
1062 zhp
->zpool_hdl
= hdl
;
1063 (void) strlcpy(zhp
->zpool_name
, pool
, sizeof (zhp
->zpool_name
));
1065 if (zpool_refresh_stats(zhp
, &missing
) != 0) {
1071 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
, "no such pool"));
1072 (void) zfs_error_fmt(hdl
, EZFS_NOENT
,
1073 dgettext(TEXT_DOMAIN
, "cannot open '%s'"), pool
);
1082 * Like the above, but silent on error. Used when iterating over pools (because
1083 * the configuration cache may be out of date).
1086 zpool_open_silent(libzfs_handle_t
*hdl
, const char *pool
, zpool_handle_t
**ret
)
1088 zpool_handle_t
*zhp
;
1091 if ((zhp
= zfs_alloc(hdl
, sizeof (zpool_handle_t
))) == NULL
)
1094 zhp
->zpool_hdl
= hdl
;
1095 (void) strlcpy(zhp
->zpool_name
, pool
, sizeof (zhp
->zpool_name
));
1097 if (zpool_refresh_stats(zhp
, &missing
) != 0) {
1113 * Similar to zpool_open_canfail(), but refuses to open pools in the faulted
1117 zpool_open(libzfs_handle_t
*hdl
, const char *pool
)
1119 zpool_handle_t
*zhp
;
1121 if ((zhp
= zpool_open_canfail(hdl
, pool
)) == NULL
)
1124 if (zhp
->zpool_state
== POOL_STATE_UNAVAIL
) {
1125 (void) zfs_error_fmt(hdl
, EZFS_POOLUNAVAIL
,
1126 dgettext(TEXT_DOMAIN
, "cannot open '%s'"), zhp
->zpool_name
);
1135 * Close the handle. Simply frees the memory associated with the handle.
1138 zpool_close(zpool_handle_t
*zhp
)
1140 if (zhp
->zpool_config
)
1141 nvlist_free(zhp
->zpool_config
);
1142 if (zhp
->zpool_old_config
)
1143 nvlist_free(zhp
->zpool_old_config
);
1144 if (zhp
->zpool_props
)
1145 nvlist_free(zhp
->zpool_props
);
1150 * Return the name of the pool.
1153 zpool_get_name(zpool_handle_t
*zhp
)
1155 return (zhp
->zpool_name
);
1160 * Return the state of the pool (ACTIVE or UNAVAILABLE)
1163 zpool_get_state(zpool_handle_t
*zhp
)
1165 return (zhp
->zpool_state
);
1169 * Create the named pool, using the provided vdev list. It is assumed
1170 * that the consumer has already validated the contents of the nvlist, so we
1171 * don't have to worry about error semantics.
1174 zpool_create(libzfs_handle_t
*hdl
, const char *pool
, nvlist_t
*nvroot
,
1175 nvlist_t
*props
, nvlist_t
*fsprops
)
1177 zfs_cmd_t zc
= {"\0"};
1178 nvlist_t
*zc_fsprops
= NULL
;
1179 nvlist_t
*zc_props
= NULL
;
1183 (void) snprintf(msg
, sizeof (msg
), dgettext(TEXT_DOMAIN
,
1184 "cannot create '%s'"), pool
);
1186 if (!zpool_name_valid(hdl
, B_FALSE
, pool
))
1187 return (zfs_error(hdl
, EZFS_INVALIDNAME
, msg
));
1189 if (zcmd_write_conf_nvlist(hdl
, &zc
, nvroot
) != 0)
1193 prop_flags_t flags
= { .create
= B_TRUE
, .import
= B_FALSE
};
1195 if ((zc_props
= zpool_valid_proplist(hdl
, pool
, props
,
1196 SPA_VERSION_1
, flags
, msg
)) == NULL
) {
1205 zoned
= ((nvlist_lookup_string(fsprops
,
1206 zfs_prop_to_name(ZFS_PROP_ZONED
), &zonestr
) == 0) &&
1207 strcmp(zonestr
, "on") == 0);
1209 if ((zc_fsprops
= zfs_valid_proplist(hdl
,
1210 ZFS_TYPE_FILESYSTEM
, fsprops
, zoned
, NULL
, msg
)) == NULL
) {
1214 (nvlist_alloc(&zc_props
, NV_UNIQUE_NAME
, 0) != 0)) {
1217 if (nvlist_add_nvlist(zc_props
,
1218 ZPOOL_ROOTFS_PROPS
, zc_fsprops
) != 0) {
1223 if (zc_props
&& zcmd_write_src_nvlist(hdl
, &zc
, zc_props
) != 0)
1226 (void) strlcpy(zc
.zc_name
, pool
, sizeof (zc
.zc_name
));
1228 if ((ret
= zfs_ioctl(hdl
, ZFS_IOC_POOL_CREATE
, &zc
)) != 0) {
1230 zcmd_free_nvlists(&zc
);
1231 nvlist_free(zc_props
);
1232 nvlist_free(zc_fsprops
);
1237 * This can happen if the user has specified the same
1238 * device multiple times. We can't reliably detect this
1239 * until we try to add it and see we already have a
1240 * label. This can also happen under if the device is
1241 * part of an active md or lvm device.
1243 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
1244 "one or more vdevs refer to the same device, or "
1245 "one of\nthe devices is part of an active md or "
1247 return (zfs_error(hdl
, EZFS_BADDEV
, msg
));
1251 * This occurs when one of the devices is below
1252 * SPA_MINDEVSIZE. Unfortunately, we can't detect which
1253 * device was the problem device since there's no
1254 * reliable way to determine device size from userland.
1259 zfs_nicenum(SPA_MINDEVSIZE
, buf
, sizeof (buf
));
1261 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
1262 "one or more devices is less than the "
1263 "minimum size (%s)"), buf
);
1265 return (zfs_error(hdl
, EZFS_BADDEV
, msg
));
1268 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
1269 "one or more devices is out of space"));
1270 return (zfs_error(hdl
, EZFS_BADDEV
, msg
));
1273 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
1274 "cache device must be a disk or disk slice"));
1275 return (zfs_error(hdl
, EZFS_BADDEV
, msg
));
1278 return (zpool_standard_error(hdl
, errno
, msg
));
1283 zcmd_free_nvlists(&zc
);
1284 nvlist_free(zc_props
);
1285 nvlist_free(zc_fsprops
);
1290 * Destroy the given pool. It is up to the caller to ensure that there are no
1291 * datasets left in the pool.
1294 zpool_destroy(zpool_handle_t
*zhp
, const char *log_str
)
1296 zfs_cmd_t zc
= {"\0"};
1297 zfs_handle_t
*zfp
= NULL
;
1298 libzfs_handle_t
*hdl
= zhp
->zpool_hdl
;
1301 if (zhp
->zpool_state
== POOL_STATE_ACTIVE
&&
1302 (zfp
= zfs_open(hdl
, zhp
->zpool_name
, ZFS_TYPE_FILESYSTEM
)) == NULL
)
1305 (void) strlcpy(zc
.zc_name
, zhp
->zpool_name
, sizeof (zc
.zc_name
));
1306 zc
.zc_history
= (uint64_t)(uintptr_t)log_str
;
1308 if (zfs_ioctl(hdl
, ZFS_IOC_POOL_DESTROY
, &zc
) != 0) {
1309 (void) snprintf(msg
, sizeof (msg
), dgettext(TEXT_DOMAIN
,
1310 "cannot destroy '%s'"), zhp
->zpool_name
);
1312 if (errno
== EROFS
) {
1313 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
1314 "one or more devices is read only"));
1315 (void) zfs_error(hdl
, EZFS_BADDEV
, msg
);
1317 (void) zpool_standard_error(hdl
, errno
, msg
);
1326 remove_mountpoint(zfp
);
1334 * Add the given vdevs to the pool. The caller must have already performed the
1335 * necessary verification to ensure that the vdev specification is well-formed.
1338 zpool_add(zpool_handle_t
*zhp
, nvlist_t
*nvroot
)
1340 zfs_cmd_t zc
= {"\0"};
1342 libzfs_handle_t
*hdl
= zhp
->zpool_hdl
;
1344 nvlist_t
**spares
, **l2cache
;
1345 uint_t nspares
, nl2cache
;
1347 (void) snprintf(msg
, sizeof (msg
), dgettext(TEXT_DOMAIN
,
1348 "cannot add to '%s'"), zhp
->zpool_name
);
1350 if (zpool_get_prop_int(zhp
, ZPOOL_PROP_VERSION
, NULL
) <
1351 SPA_VERSION_SPARES
&&
1352 nvlist_lookup_nvlist_array(nvroot
, ZPOOL_CONFIG_SPARES
,
1353 &spares
, &nspares
) == 0) {
1354 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
, "pool must be "
1355 "upgraded to add hot spares"));
1356 return (zfs_error(hdl
, EZFS_BADVERSION
, msg
));
1359 #if defined(__sun__) || defined(__sun)
1360 if (zpool_is_bootable(zhp
) && nvlist_lookup_nvlist_array(nvroot
,
1361 ZPOOL_CONFIG_SPARES
, &spares
, &nspares
) == 0) {
1364 for (s
= 0; s
< nspares
; s
++) {
1367 if (nvlist_lookup_string(spares
[s
], ZPOOL_CONFIG_PATH
,
1368 &path
) == 0 && pool_uses_efi(spares
[s
])) {
1369 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
1370 "device '%s' contains an EFI label and "
1371 "cannot be used on root pools."),
1372 zpool_vdev_name(hdl
, NULL
, spares
[s
],
1374 return (zfs_error(hdl
, EZFS_POOL_NOTSUP
, msg
));
1380 if (zpool_get_prop_int(zhp
, ZPOOL_PROP_VERSION
, NULL
) <
1381 SPA_VERSION_L2CACHE
&&
1382 nvlist_lookup_nvlist_array(nvroot
, ZPOOL_CONFIG_L2CACHE
,
1383 &l2cache
, &nl2cache
) == 0) {
1384 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
, "pool must be "
1385 "upgraded to add cache devices"));
1386 return (zfs_error(hdl
, EZFS_BADVERSION
, msg
));
1389 if (zcmd_write_conf_nvlist(hdl
, &zc
, nvroot
) != 0)
1391 (void) strlcpy(zc
.zc_name
, zhp
->zpool_name
, sizeof (zc
.zc_name
));
1393 if (zfs_ioctl(hdl
, ZFS_IOC_VDEV_ADD
, &zc
) != 0) {
1397 * This can happen if the user has specified the same
1398 * device multiple times. We can't reliably detect this
1399 * until we try to add it and see we already have a
1402 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
1403 "one or more vdevs refer to the same device"));
1404 (void) zfs_error(hdl
, EZFS_BADDEV
, msg
);
1409 * This occurrs when one of the devices is below
1410 * SPA_MINDEVSIZE. Unfortunately, we can't detect which
1411 * device was the problem device since there's no
1412 * reliable way to determine device size from userland.
1417 zfs_nicenum(SPA_MINDEVSIZE
, buf
, sizeof (buf
));
1419 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
1420 "device is less than the minimum "
1423 (void) zfs_error(hdl
, EZFS_BADDEV
, msg
);
1427 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
1428 "pool must be upgraded to add these vdevs"));
1429 (void) zfs_error(hdl
, EZFS_BADVERSION
, msg
);
1433 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
1434 "cache device must be a disk or disk slice"));
1435 (void) zfs_error(hdl
, EZFS_BADDEV
, msg
);
1439 (void) zpool_standard_error(hdl
, errno
, msg
);
1447 zcmd_free_nvlists(&zc
);
1453 * Exports the pool from the system. The caller must ensure that there are no
1454 * mounted datasets in the pool.
1457 zpool_export_common(zpool_handle_t
*zhp
, boolean_t force
, boolean_t hardforce
,
1458 const char *log_str
)
1460 zfs_cmd_t zc
= {"\0"};
1463 (void) snprintf(msg
, sizeof (msg
), dgettext(TEXT_DOMAIN
,
1464 "cannot export '%s'"), zhp
->zpool_name
);
1466 (void) strlcpy(zc
.zc_name
, zhp
->zpool_name
, sizeof (zc
.zc_name
));
1467 zc
.zc_cookie
= force
;
1468 zc
.zc_guid
= hardforce
;
1469 zc
.zc_history
= (uint64_t)(uintptr_t)log_str
;
1471 if (zfs_ioctl(zhp
->zpool_hdl
, ZFS_IOC_POOL_EXPORT
, &zc
) != 0) {
1474 zfs_error_aux(zhp
->zpool_hdl
, dgettext(TEXT_DOMAIN
,
1475 "use '-f' to override the following errors:\n"
1476 "'%s' has an active shared spare which could be"
1477 " used by other pools once '%s' is exported."),
1478 zhp
->zpool_name
, zhp
->zpool_name
);
1479 return (zfs_error(zhp
->zpool_hdl
, EZFS_ACTIVE_SPARE
,
1482 return (zpool_standard_error_fmt(zhp
->zpool_hdl
, errno
,
1491 zpool_export(zpool_handle_t
*zhp
, boolean_t force
, const char *log_str
)
1493 return (zpool_export_common(zhp
, force
, B_FALSE
, log_str
));
1497 zpool_export_force(zpool_handle_t
*zhp
, const char *log_str
)
1499 return (zpool_export_common(zhp
, B_TRUE
, B_TRUE
, log_str
));
1503 zpool_rewind_exclaim(libzfs_handle_t
*hdl
, const char *name
, boolean_t dryrun
,
1506 nvlist_t
*nv
= NULL
;
1512 if (!hdl
->libzfs_printerr
|| config
== NULL
)
1515 if (nvlist_lookup_nvlist(config
, ZPOOL_CONFIG_LOAD_INFO
, &nv
) != 0 ||
1516 nvlist_lookup_nvlist(nv
, ZPOOL_CONFIG_REWIND_INFO
, &nv
) != 0) {
1520 if (nvlist_lookup_uint64(nv
, ZPOOL_CONFIG_LOAD_TIME
, &rewindto
) != 0)
1522 (void) nvlist_lookup_int64(nv
, ZPOOL_CONFIG_REWIND_TIME
, &loss
);
1524 if (localtime_r((time_t *)&rewindto
, &t
) != NULL
&&
1525 strftime(timestr
, 128, "%c", &t
) != 0) {
1527 (void) printf(dgettext(TEXT_DOMAIN
,
1528 "Would be able to return %s "
1529 "to its state as of %s.\n"),
1532 (void) printf(dgettext(TEXT_DOMAIN
,
1533 "Pool %s returned to its state as of %s.\n"),
1537 (void) printf(dgettext(TEXT_DOMAIN
,
1538 "%s approximately %lld "),
1539 dryrun
? "Would discard" : "Discarded",
1540 ((longlong_t
)loss
+ 30) / 60);
1541 (void) printf(dgettext(TEXT_DOMAIN
,
1542 "minutes of transactions.\n"));
1543 } else if (loss
> 0) {
1544 (void) printf(dgettext(TEXT_DOMAIN
,
1545 "%s approximately %lld "),
1546 dryrun
? "Would discard" : "Discarded",
1548 (void) printf(dgettext(TEXT_DOMAIN
,
1549 "seconds of transactions.\n"));
1555 zpool_explain_recover(libzfs_handle_t
*hdl
, const char *name
, int reason
,
1558 nvlist_t
*nv
= NULL
;
1560 uint64_t edata
= UINT64_MAX
;
1565 if (!hdl
->libzfs_printerr
)
1569 (void) printf(dgettext(TEXT_DOMAIN
, "action: "));
1571 (void) printf(dgettext(TEXT_DOMAIN
, "\t"));
1573 /* All attempted rewinds failed if ZPOOL_CONFIG_LOAD_TIME missing */
1574 if (nvlist_lookup_nvlist(config
, ZPOOL_CONFIG_LOAD_INFO
, &nv
) != 0 ||
1575 nvlist_lookup_nvlist(nv
, ZPOOL_CONFIG_REWIND_INFO
, &nv
) != 0 ||
1576 nvlist_lookup_uint64(nv
, ZPOOL_CONFIG_LOAD_TIME
, &rewindto
) != 0)
1579 (void) nvlist_lookup_int64(nv
, ZPOOL_CONFIG_REWIND_TIME
, &loss
);
1580 (void) nvlist_lookup_uint64(nv
, ZPOOL_CONFIG_LOAD_DATA_ERRORS
,
1583 (void) printf(dgettext(TEXT_DOMAIN
,
1584 "Recovery is possible, but will result in some data loss.\n"));
1586 if (localtime_r((time_t *)&rewindto
, &t
) != NULL
&&
1587 strftime(timestr
, 128, "%c", &t
) != 0) {
1588 (void) printf(dgettext(TEXT_DOMAIN
,
1589 "\tReturning the pool to its state as of %s\n"
1590 "\tshould correct the problem. "),
1593 (void) printf(dgettext(TEXT_DOMAIN
,
1594 "\tReverting the pool to an earlier state "
1595 "should correct the problem.\n\t"));
1599 (void) printf(dgettext(TEXT_DOMAIN
,
1600 "Approximately %lld minutes of data\n"
1601 "\tmust be discarded, irreversibly. "),
1602 ((longlong_t
)loss
+ 30) / 60);
1603 } else if (loss
> 0) {
1604 (void) printf(dgettext(TEXT_DOMAIN
,
1605 "Approximately %lld seconds of data\n"
1606 "\tmust be discarded, irreversibly. "),
1609 if (edata
!= 0 && edata
!= UINT64_MAX
) {
1611 (void) printf(dgettext(TEXT_DOMAIN
,
1612 "After rewind, at least\n"
1613 "\tone persistent user-data error will remain. "));
1615 (void) printf(dgettext(TEXT_DOMAIN
,
1616 "After rewind, several\n"
1617 "\tpersistent user-data errors will remain. "));
1620 (void) printf(dgettext(TEXT_DOMAIN
,
1621 "Recovery can be attempted\n\tby executing 'zpool %s -F %s'. "),
1622 reason
>= 0 ? "clear" : "import", name
);
1624 (void) printf(dgettext(TEXT_DOMAIN
,
1625 "A scrub of the pool\n"
1626 "\tis strongly recommended after recovery.\n"));
1630 (void) printf(dgettext(TEXT_DOMAIN
,
1631 "Destroy and re-create the pool from\n\ta backup source.\n"));
1635 * zpool_import() is a contracted interface. Should be kept the same
1638 * Applications should use zpool_import_props() to import a pool with
1639 * new properties value to be set.
1642 zpool_import(libzfs_handle_t
*hdl
, nvlist_t
*config
, const char *newname
,
1645 nvlist_t
*props
= NULL
;
1648 if (altroot
!= NULL
) {
1649 if (nvlist_alloc(&props
, NV_UNIQUE_NAME
, 0) != 0) {
1650 return (zfs_error_fmt(hdl
, EZFS_NOMEM
,
1651 dgettext(TEXT_DOMAIN
, "cannot import '%s'"),
1655 if (nvlist_add_string(props
,
1656 zpool_prop_to_name(ZPOOL_PROP_ALTROOT
), altroot
) != 0 ||
1657 nvlist_add_string(props
,
1658 zpool_prop_to_name(ZPOOL_PROP_CACHEFILE
), "none") != 0) {
1660 return (zfs_error_fmt(hdl
, EZFS_NOMEM
,
1661 dgettext(TEXT_DOMAIN
, "cannot import '%s'"),
1666 ret
= zpool_import_props(hdl
, config
, newname
, props
,
1674 print_vdev_tree(libzfs_handle_t
*hdl
, const char *name
, nvlist_t
*nv
,
1680 uint64_t is_log
= 0;
1682 (void) nvlist_lookup_uint64(nv
, ZPOOL_CONFIG_IS_LOG
,
1686 (void) printf("\t%*s%s%s\n", indent
, "", name
,
1687 is_log
? " [log]" : "");
1689 if (nvlist_lookup_nvlist_array(nv
, ZPOOL_CONFIG_CHILDREN
,
1690 &child
, &children
) != 0)
1693 for (c
= 0; c
< children
; c
++) {
1694 vname
= zpool_vdev_name(hdl
, NULL
, child
[c
], B_TRUE
);
1695 print_vdev_tree(hdl
, vname
, child
[c
], indent
+ 2);
1701 zpool_print_unsup_feat(nvlist_t
*config
)
1703 nvlist_t
*nvinfo
, *unsup_feat
;
1706 verify(nvlist_lookup_nvlist(config
, ZPOOL_CONFIG_LOAD_INFO
, &nvinfo
) ==
1708 verify(nvlist_lookup_nvlist(nvinfo
, ZPOOL_CONFIG_UNSUP_FEAT
,
1711 for (nvp
= nvlist_next_nvpair(unsup_feat
, NULL
); nvp
!= NULL
;
1712 nvp
= nvlist_next_nvpair(unsup_feat
, nvp
)) {
1715 verify(nvpair_type(nvp
) == DATA_TYPE_STRING
);
1716 verify(nvpair_value_string(nvp
, &desc
) == 0);
1718 if (strlen(desc
) > 0)
1719 (void) printf("\t%s (%s)\n", nvpair_name(nvp
), desc
);
1721 (void) printf("\t%s\n", nvpair_name(nvp
));
1726 * Import the given pool using the known configuration and a list of
1727 * properties to be set. The configuration should have come from
1728 * zpool_find_import(). The 'newname' parameters control whether the pool
1729 * is imported with a different name.
1732 zpool_import_props(libzfs_handle_t
*hdl
, nvlist_t
*config
, const char *newname
,
1733 nvlist_t
*props
, int flags
)
1735 zfs_cmd_t zc
= {"\0"};
1736 zpool_rewind_policy_t policy
;
1737 nvlist_t
*nv
= NULL
;
1738 nvlist_t
*nvinfo
= NULL
;
1739 nvlist_t
*missing
= NULL
;
1746 verify(nvlist_lookup_string(config
, ZPOOL_CONFIG_POOL_NAME
,
1749 (void) snprintf(errbuf
, sizeof (errbuf
), dgettext(TEXT_DOMAIN
,
1750 "cannot import pool '%s'"), origname
);
1752 if (newname
!= NULL
) {
1753 if (!zpool_name_valid(hdl
, B_FALSE
, newname
))
1754 return (zfs_error_fmt(hdl
, EZFS_INVALIDNAME
,
1755 dgettext(TEXT_DOMAIN
, "cannot import '%s'"),
1757 thename
= (char *)newname
;
1764 prop_flags_t flags
= { .create
= B_FALSE
, .import
= B_TRUE
};
1766 verify(nvlist_lookup_uint64(config
, ZPOOL_CONFIG_VERSION
,
1769 if ((props
= zpool_valid_proplist(hdl
, origname
,
1770 props
, version
, flags
, errbuf
)) == NULL
) {
1772 } else if (zcmd_write_src_nvlist(hdl
, &zc
, props
) != 0) {
1778 (void) strlcpy(zc
.zc_name
, thename
, sizeof (zc
.zc_name
));
1780 verify(nvlist_lookup_uint64(config
, ZPOOL_CONFIG_POOL_GUID
,
1783 if (zcmd_write_conf_nvlist(hdl
, &zc
, config
) != 0) {
1787 if (zcmd_alloc_dst_nvlist(hdl
, &zc
, zc
.zc_nvlist_conf_size
* 2) != 0) {
1792 zc
.zc_cookie
= flags
;
1793 while ((ret
= zfs_ioctl(hdl
, ZFS_IOC_POOL_IMPORT
, &zc
)) != 0 &&
1795 if (zcmd_expand_dst_nvlist(hdl
, &zc
) != 0) {
1796 zcmd_free_nvlists(&zc
);
1803 (void) zcmd_read_dst_nvlist(hdl
, &zc
, &nv
);
1804 zpool_get_rewind_policy(config
, &policy
);
1810 * Dry-run failed, but we print out what success
1811 * looks like if we found a best txg
1813 if (policy
.zrp_request
& ZPOOL_TRY_REWIND
) {
1814 zpool_rewind_exclaim(hdl
, newname
? origname
: thename
,
1820 if (newname
== NULL
)
1821 (void) snprintf(desc
, sizeof (desc
),
1822 dgettext(TEXT_DOMAIN
, "cannot import '%s'"),
1825 (void) snprintf(desc
, sizeof (desc
),
1826 dgettext(TEXT_DOMAIN
, "cannot import '%s' as '%s'"),
1831 if (nv
!= NULL
&& nvlist_lookup_nvlist(nv
,
1832 ZPOOL_CONFIG_LOAD_INFO
, &nvinfo
) == 0 &&
1833 nvlist_exists(nvinfo
, ZPOOL_CONFIG_UNSUP_FEAT
)) {
1834 (void) printf(dgettext(TEXT_DOMAIN
, "This "
1835 "pool uses the following feature(s) not "
1836 "supported by this system:\n"));
1837 zpool_print_unsup_feat(nv
);
1838 if (nvlist_exists(nvinfo
,
1839 ZPOOL_CONFIG_CAN_RDONLY
)) {
1840 (void) printf(dgettext(TEXT_DOMAIN
,
1841 "All unsupported features are only "
1842 "required for writing to the pool."
1843 "\nThe pool can be imported using "
1844 "'-o readonly=on'.\n"));
1848 * Unsupported version.
1850 (void) zfs_error(hdl
, EZFS_BADVERSION
, desc
);
1854 (void) zfs_error(hdl
, EZFS_INVALCONFIG
, desc
);
1858 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
1859 "one or more devices is read only"));
1860 (void) zfs_error(hdl
, EZFS_BADDEV
, desc
);
1864 if (nv
&& nvlist_lookup_nvlist(nv
,
1865 ZPOOL_CONFIG_LOAD_INFO
, &nvinfo
) == 0 &&
1866 nvlist_lookup_nvlist(nvinfo
,
1867 ZPOOL_CONFIG_MISSING_DEVICES
, &missing
) == 0) {
1868 (void) printf(dgettext(TEXT_DOMAIN
,
1869 "The devices below are missing, use "
1870 "'-m' to import the pool anyway:\n"));
1871 print_vdev_tree(hdl
, NULL
, missing
, 2);
1872 (void) printf("\n");
1874 (void) zpool_standard_error(hdl
, error
, desc
);
1878 (void) zpool_standard_error(hdl
, error
, desc
);
1882 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
1883 "one or more devices are already in use\n"));
1884 (void) zfs_error(hdl
, EZFS_BADDEV
, desc
);
1888 (void) zpool_standard_error(hdl
, error
, desc
);
1889 zpool_explain_recover(hdl
,
1890 newname
? origname
: thename
, -error
, nv
);
1897 zpool_handle_t
*zhp
;
1900 * This should never fail, but play it safe anyway.
1902 if (zpool_open_silent(hdl
, thename
, &zhp
) != 0)
1904 else if (zhp
!= NULL
)
1906 if (policy
.zrp_request
&
1907 (ZPOOL_DO_REWIND
| ZPOOL_TRY_REWIND
)) {
1908 zpool_rewind_exclaim(hdl
, newname
? origname
: thename
,
1909 ((policy
.zrp_request
& ZPOOL_TRY_REWIND
) != 0), nv
);
1915 zcmd_free_nvlists(&zc
);
1925 zpool_scan(zpool_handle_t
*zhp
, pool_scan_func_t func
)
1927 zfs_cmd_t zc
= {"\0"};
1929 libzfs_handle_t
*hdl
= zhp
->zpool_hdl
;
1931 (void) strlcpy(zc
.zc_name
, zhp
->zpool_name
, sizeof (zc
.zc_name
));
1932 zc
.zc_cookie
= func
;
1934 if (zfs_ioctl(hdl
, ZFS_IOC_POOL_SCAN
, &zc
) == 0 ||
1935 (errno
== ENOENT
&& func
!= POOL_SCAN_NONE
))
1938 if (func
== POOL_SCAN_SCRUB
) {
1939 (void) snprintf(msg
, sizeof (msg
),
1940 dgettext(TEXT_DOMAIN
, "cannot scrub %s"), zc
.zc_name
);
1941 } else if (func
== POOL_SCAN_NONE
) {
1942 (void) snprintf(msg
, sizeof (msg
),
1943 dgettext(TEXT_DOMAIN
, "cannot cancel scrubbing %s"),
1946 assert(!"unexpected result");
1949 if (errno
== EBUSY
) {
1951 pool_scan_stat_t
*ps
= NULL
;
1954 verify(nvlist_lookup_nvlist(zhp
->zpool_config
,
1955 ZPOOL_CONFIG_VDEV_TREE
, &nvroot
) == 0);
1956 (void) nvlist_lookup_uint64_array(nvroot
,
1957 ZPOOL_CONFIG_SCAN_STATS
, (uint64_t **)&ps
, &psc
);
1958 if (ps
&& ps
->pss_func
== POOL_SCAN_SCRUB
)
1959 return (zfs_error(hdl
, EZFS_SCRUBBING
, msg
));
1961 return (zfs_error(hdl
, EZFS_RESILVERING
, msg
));
1962 } else if (errno
== ENOENT
) {
1963 return (zfs_error(hdl
, EZFS_NO_SCRUB
, msg
));
1965 return (zpool_standard_error(hdl
, errno
, msg
));
1970 * Find a vdev that matches the search criteria specified. We use the
1971 * the nvpair name to determine how we should look for the device.
1972 * 'avail_spare' is set to TRUE if the provided guid refers to an AVAIL
1973 * spare; but FALSE if its an INUSE spare.
1976 vdev_to_nvlist_iter(nvlist_t
*nv
, nvlist_t
*search
, boolean_t
*avail_spare
,
1977 boolean_t
*l2cache
, boolean_t
*log
)
1984 nvpair_t
*pair
= nvlist_next_nvpair(search
, NULL
);
1986 /* Nothing to look for */
1987 if (search
== NULL
|| pair
== NULL
)
1990 /* Obtain the key we will use to search */
1991 srchkey
= nvpair_name(pair
);
1993 switch (nvpair_type(pair
)) {
1994 case DATA_TYPE_UINT64
:
1995 if (strcmp(srchkey
, ZPOOL_CONFIG_GUID
) == 0) {
1996 uint64_t srchval
, theguid
;
1998 verify(nvpair_value_uint64(pair
, &srchval
) == 0);
1999 verify(nvlist_lookup_uint64(nv
, ZPOOL_CONFIG_GUID
,
2001 if (theguid
== srchval
)
2006 case DATA_TYPE_STRING
: {
2007 char *srchval
, *val
;
2009 verify(nvpair_value_string(pair
, &srchval
) == 0);
2010 if (nvlist_lookup_string(nv
, srchkey
, &val
) != 0)
2014 * Search for the requested value. Special cases:
2016 * - ZPOOL_CONFIG_PATH for whole disk entries. These end in
2017 * "-part1", or "p1". The suffix is hidden from the user,
2018 * but included in the string, so this matches around it.
2019 * - ZPOOL_CONFIG_PATH for short names zfs_strcmp_shortname()
2020 * is used to check all possible expanded paths.
2021 * - looking for a top-level vdev name (i.e. ZPOOL_CONFIG_TYPE).
2023 * Otherwise, all other searches are simple string compares.
2025 if (strcmp(srchkey
, ZPOOL_CONFIG_PATH
) == 0) {
2026 uint64_t wholedisk
= 0;
2028 (void) nvlist_lookup_uint64(nv
, ZPOOL_CONFIG_WHOLE_DISK
,
2030 if (zfs_strcmp_pathname(srchval
, val
, wholedisk
) == 0)
2033 } else if (strcmp(srchkey
, ZPOOL_CONFIG_TYPE
) == 0 && val
) {
2034 char *type
, *idx
, *end
, *p
;
2035 uint64_t id
, vdev_id
;
2038 * Determine our vdev type, keeping in mind
2039 * that the srchval is composed of a type and
2040 * vdev id pair (i.e. mirror-4).
2042 if ((type
= strdup(srchval
)) == NULL
)
2045 if ((p
= strrchr(type
, '-')) == NULL
) {
2053 * If the types don't match then keep looking.
2055 if (strncmp(val
, type
, strlen(val
)) != 0) {
2060 verify(strncmp(type
, VDEV_TYPE_RAIDZ
,
2061 strlen(VDEV_TYPE_RAIDZ
)) == 0 ||
2062 strncmp(type
, VDEV_TYPE_MIRROR
,
2063 strlen(VDEV_TYPE_MIRROR
)) == 0);
2064 verify(nvlist_lookup_uint64(nv
, ZPOOL_CONFIG_ID
,
2068 vdev_id
= strtoull(idx
, &end
, 10);
2075 * Now verify that we have the correct vdev id.
2084 if (strcmp(srchval
, val
) == 0)
2093 if (nvlist_lookup_nvlist_array(nv
, ZPOOL_CONFIG_CHILDREN
,
2094 &child
, &children
) != 0)
2097 for (c
= 0; c
< children
; c
++) {
2098 if ((ret
= vdev_to_nvlist_iter(child
[c
], search
,
2099 avail_spare
, l2cache
, NULL
)) != NULL
) {
2101 * The 'is_log' value is only set for the toplevel
2102 * vdev, not the leaf vdevs. So we always lookup the
2103 * log device from the root of the vdev tree (where
2104 * 'log' is non-NULL).
2107 nvlist_lookup_uint64(child
[c
],
2108 ZPOOL_CONFIG_IS_LOG
, &is_log
) == 0 &&
2116 if (nvlist_lookup_nvlist_array(nv
, ZPOOL_CONFIG_SPARES
,
2117 &child
, &children
) == 0) {
2118 for (c
= 0; c
< children
; c
++) {
2119 if ((ret
= vdev_to_nvlist_iter(child
[c
], search
,
2120 avail_spare
, l2cache
, NULL
)) != NULL
) {
2121 *avail_spare
= B_TRUE
;
2127 if (nvlist_lookup_nvlist_array(nv
, ZPOOL_CONFIG_L2CACHE
,
2128 &child
, &children
) == 0) {
2129 for (c
= 0; c
< children
; c
++) {
2130 if ((ret
= vdev_to_nvlist_iter(child
[c
], search
,
2131 avail_spare
, l2cache
, NULL
)) != NULL
) {
2142 * Given a physical path (minus the "/devices" prefix), find the
2146 zpool_find_vdev_by_physpath(zpool_handle_t
*zhp
, const char *ppath
,
2147 boolean_t
*avail_spare
, boolean_t
*l2cache
, boolean_t
*log
)
2149 nvlist_t
*search
, *nvroot
, *ret
;
2151 verify(nvlist_alloc(&search
, NV_UNIQUE_NAME
, KM_SLEEP
) == 0);
2152 verify(nvlist_add_string(search
, ZPOOL_CONFIG_PHYS_PATH
, ppath
) == 0);
2154 verify(nvlist_lookup_nvlist(zhp
->zpool_config
, ZPOOL_CONFIG_VDEV_TREE
,
2157 *avail_spare
= B_FALSE
;
2161 ret
= vdev_to_nvlist_iter(nvroot
, search
, avail_spare
, l2cache
, log
);
2162 nvlist_free(search
);
2168 * Determine if we have an "interior" top-level vdev (i.e mirror/raidz).
2171 zpool_vdev_is_interior(const char *name
)
2173 if (strncmp(name
, VDEV_TYPE_RAIDZ
, strlen(VDEV_TYPE_RAIDZ
)) == 0 ||
2174 strncmp(name
, VDEV_TYPE_MIRROR
, strlen(VDEV_TYPE_MIRROR
)) == 0)
2180 zpool_find_vdev(zpool_handle_t
*zhp
, const char *path
, boolean_t
*avail_spare
,
2181 boolean_t
*l2cache
, boolean_t
*log
)
2184 nvlist_t
*nvroot
, *search
, *ret
;
2187 verify(nvlist_alloc(&search
, NV_UNIQUE_NAME
, KM_SLEEP
) == 0);
2189 guid
= strtoull(path
, &end
, 0);
2190 if (guid
!= 0 && *end
== '\0') {
2191 verify(nvlist_add_uint64(search
, ZPOOL_CONFIG_GUID
, guid
) == 0);
2192 } else if (zpool_vdev_is_interior(path
)) {
2193 verify(nvlist_add_string(search
, ZPOOL_CONFIG_TYPE
, path
) == 0);
2195 verify(nvlist_add_string(search
, ZPOOL_CONFIG_PATH
, path
) == 0);
2198 verify(nvlist_lookup_nvlist(zhp
->zpool_config
, ZPOOL_CONFIG_VDEV_TREE
,
2201 *avail_spare
= B_FALSE
;
2205 ret
= vdev_to_nvlist_iter(nvroot
, search
, avail_spare
, l2cache
, log
);
2206 nvlist_free(search
);
2212 vdev_online(nvlist_t
*nv
)
2216 if (nvlist_lookup_uint64(nv
, ZPOOL_CONFIG_OFFLINE
, &ival
) == 0 ||
2217 nvlist_lookup_uint64(nv
, ZPOOL_CONFIG_FAULTED
, &ival
) == 0 ||
2218 nvlist_lookup_uint64(nv
, ZPOOL_CONFIG_REMOVED
, &ival
) == 0)
2225 * Helper function for zpool_get_physpaths().
2228 vdev_get_one_physpath(nvlist_t
*config
, char *physpath
, size_t physpath_size
,
2229 size_t *bytes_written
)
2231 size_t bytes_left
, pos
, rsz
;
2235 if (nvlist_lookup_string(config
, ZPOOL_CONFIG_PHYS_PATH
,
2237 return (EZFS_NODEVICE
);
2239 pos
= *bytes_written
;
2240 bytes_left
= physpath_size
- pos
;
2241 format
= (pos
== 0) ? "%s" : " %s";
2243 rsz
= snprintf(physpath
+ pos
, bytes_left
, format
, tmppath
);
2244 *bytes_written
+= rsz
;
2246 if (rsz
>= bytes_left
) {
2247 /* if physpath was not copied properly, clear it */
2248 if (bytes_left
!= 0) {
2251 return (EZFS_NOSPC
);
2257 vdev_get_physpaths(nvlist_t
*nv
, char *physpath
, size_t phypath_size
,
2258 size_t *rsz
, boolean_t is_spare
)
2263 if (nvlist_lookup_string(nv
, ZPOOL_CONFIG_TYPE
, &type
) != 0)
2264 return (EZFS_INVALCONFIG
);
2266 if (strcmp(type
, VDEV_TYPE_DISK
) == 0) {
2268 * An active spare device has ZPOOL_CONFIG_IS_SPARE set.
2269 * For a spare vdev, we only want to boot from the active
2274 (void) nvlist_lookup_uint64(nv
, ZPOOL_CONFIG_IS_SPARE
,
2277 return (EZFS_INVALCONFIG
);
2280 if (vdev_online(nv
)) {
2281 if ((ret
= vdev_get_one_physpath(nv
, physpath
,
2282 phypath_size
, rsz
)) != 0)
2285 } else if (strcmp(type
, VDEV_TYPE_MIRROR
) == 0 ||
2286 strcmp(type
, VDEV_TYPE_REPLACING
) == 0 ||
2287 (is_spare
= (strcmp(type
, VDEV_TYPE_SPARE
) == 0))) {
2292 if (nvlist_lookup_nvlist_array(nv
,
2293 ZPOOL_CONFIG_CHILDREN
, &child
, &count
) != 0)
2294 return (EZFS_INVALCONFIG
);
2296 for (i
= 0; i
< count
; i
++) {
2297 ret
= vdev_get_physpaths(child
[i
], physpath
,
2298 phypath_size
, rsz
, is_spare
);
2299 if (ret
== EZFS_NOSPC
)
2304 return (EZFS_POOL_INVALARG
);
2308 * Get phys_path for a root pool config.
2309 * Return 0 on success; non-zero on failure.
2312 zpool_get_config_physpath(nvlist_t
*config
, char *physpath
, size_t phypath_size
)
2315 nvlist_t
*vdev_root
;
2322 if (nvlist_lookup_nvlist(config
, ZPOOL_CONFIG_VDEV_TREE
,
2324 return (EZFS_INVALCONFIG
);
2326 if (nvlist_lookup_string(vdev_root
, ZPOOL_CONFIG_TYPE
, &type
) != 0 ||
2327 nvlist_lookup_nvlist_array(vdev_root
, ZPOOL_CONFIG_CHILDREN
,
2328 &child
, &count
) != 0)
2329 return (EZFS_INVALCONFIG
);
2331 #if defined(__sun__) || defined(__sun)
2333 * root pool can not have EFI labeled disks and can only have
2334 * a single top-level vdev.
2336 if (strcmp(type
, VDEV_TYPE_ROOT
) != 0 || count
!= 1 ||
2337 pool_uses_efi(vdev_root
))
2338 return (EZFS_POOL_INVALARG
);
2341 (void) vdev_get_physpaths(child
[0], physpath
, phypath_size
, &rsz
,
2344 /* No online devices */
2346 return (EZFS_NODEVICE
);
2352 * Get phys_path for a root pool
2353 * Return 0 on success; non-zero on failure.
2356 zpool_get_physpath(zpool_handle_t
*zhp
, char *physpath
, size_t phypath_size
)
2358 return (zpool_get_config_physpath(zhp
->zpool_config
, physpath
,
2363 * If the device has being dynamically expanded then we need to relabel
2364 * the disk to use the new unallocated space.
2367 zpool_relabel_disk(libzfs_handle_t
*hdl
, const char *path
, const char *msg
)
2371 if ((fd
= open(path
, O_RDWR
|O_DIRECT
)) < 0) {
2372 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
, "cannot "
2373 "relabel '%s': unable to open device: %d"), path
, errno
);
2374 return (zfs_error(hdl
, EZFS_OPENFAILED
, msg
));
2378 * It's possible that we might encounter an error if the device
2379 * does not have any unallocated space left. If so, we simply
2380 * ignore that error and continue on.
2382 * Also, we don't call efi_rescan() - that would just return EBUSY.
2383 * The module will do it for us in vdev_disk_open().
2385 error
= efi_use_whole_disk(fd
);
2387 if (error
&& error
!= VT_ENOSPC
) {
2388 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
, "cannot "
2389 "relabel '%s': unable to read disk capacity"), path
);
2390 return (zfs_error(hdl
, EZFS_NOCAP
, msg
));
2396 * Bring the specified vdev online. The 'flags' parameter is a set of the
2397 * ZFS_ONLINE_* flags.
2400 zpool_vdev_online(zpool_handle_t
*zhp
, const char *path
, int flags
,
2401 vdev_state_t
*newstate
)
2403 zfs_cmd_t zc
= {"\0"};
2406 boolean_t avail_spare
, l2cache
, islog
;
2407 libzfs_handle_t
*hdl
= zhp
->zpool_hdl
;
2410 if (flags
& ZFS_ONLINE_EXPAND
) {
2411 (void) snprintf(msg
, sizeof (msg
),
2412 dgettext(TEXT_DOMAIN
, "cannot expand %s"), path
);
2414 (void) snprintf(msg
, sizeof (msg
),
2415 dgettext(TEXT_DOMAIN
, "cannot online %s"), path
);
2418 (void) strlcpy(zc
.zc_name
, zhp
->zpool_name
, sizeof (zc
.zc_name
));
2419 if ((tgt
= zpool_find_vdev(zhp
, path
, &avail_spare
, &l2cache
,
2421 return (zfs_error(hdl
, EZFS_NODEVICE
, msg
));
2423 verify(nvlist_lookup_uint64(tgt
, ZPOOL_CONFIG_GUID
, &zc
.zc_guid
) == 0);
2426 return (zfs_error(hdl
, EZFS_ISSPARE
, msg
));
2428 if (flags
& ZFS_ONLINE_EXPAND
||
2429 zpool_get_prop_int(zhp
, ZPOOL_PROP_AUTOEXPAND
, NULL
)) {
2430 uint64_t wholedisk
= 0;
2432 (void) nvlist_lookup_uint64(tgt
, ZPOOL_CONFIG_WHOLE_DISK
,
2436 * XXX - L2ARC 1.0 devices can't support expansion.
2439 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
2440 "cannot expand cache devices"));
2441 return (zfs_error(hdl
, EZFS_VDEVNOTSUP
, msg
));
2445 const char *fullpath
= path
;
2446 char buf
[MAXPATHLEN
];
2448 if (path
[0] != '/') {
2449 error
= zfs_resolve_shortname(path
, buf
,
2452 return (zfs_error(hdl
, EZFS_NODEVICE
,
2458 error
= zpool_relabel_disk(hdl
, fullpath
, msg
);
2464 zc
.zc_cookie
= VDEV_STATE_ONLINE
;
2467 if (zfs_ioctl(hdl
, ZFS_IOC_VDEV_SET_STATE
, &zc
) != 0) {
2468 if (errno
== EINVAL
) {
2469 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
, "was split "
2470 "from this pool into a new one. Use '%s' "
2471 "instead"), "zpool detach");
2472 return (zfs_error(hdl
, EZFS_POSTSPLIT_ONLINE
, msg
));
2474 return (zpool_standard_error(hdl
, errno
, msg
));
2477 *newstate
= zc
.zc_cookie
;
2482 * Take the specified vdev offline
2485 zpool_vdev_offline(zpool_handle_t
*zhp
, const char *path
, boolean_t istmp
)
2487 zfs_cmd_t zc
= {"\0"};
2490 boolean_t avail_spare
, l2cache
;
2491 libzfs_handle_t
*hdl
= zhp
->zpool_hdl
;
2493 (void) snprintf(msg
, sizeof (msg
),
2494 dgettext(TEXT_DOMAIN
, "cannot offline %s"), path
);
2496 (void) strlcpy(zc
.zc_name
, zhp
->zpool_name
, sizeof (zc
.zc_name
));
2497 if ((tgt
= zpool_find_vdev(zhp
, path
, &avail_spare
, &l2cache
,
2499 return (zfs_error(hdl
, EZFS_NODEVICE
, msg
));
2501 verify(nvlist_lookup_uint64(tgt
, ZPOOL_CONFIG_GUID
, &zc
.zc_guid
) == 0);
2504 return (zfs_error(hdl
, EZFS_ISSPARE
, msg
));
2506 zc
.zc_cookie
= VDEV_STATE_OFFLINE
;
2507 zc
.zc_obj
= istmp
? ZFS_OFFLINE_TEMPORARY
: 0;
2509 if (zfs_ioctl(hdl
, ZFS_IOC_VDEV_SET_STATE
, &zc
) == 0)
2516 * There are no other replicas of this device.
2518 return (zfs_error(hdl
, EZFS_NOREPLICAS
, msg
));
2522 * The log device has unplayed logs
2524 return (zfs_error(hdl
, EZFS_UNPLAYED_LOGS
, msg
));
2527 return (zpool_standard_error(hdl
, errno
, msg
));
2532 * Mark the given vdev faulted.
2535 zpool_vdev_fault(zpool_handle_t
*zhp
, uint64_t guid
, vdev_aux_t aux
)
2537 zfs_cmd_t zc
= {"\0"};
2539 libzfs_handle_t
*hdl
= zhp
->zpool_hdl
;
2541 (void) snprintf(msg
, sizeof (msg
),
2542 dgettext(TEXT_DOMAIN
, "cannot fault %llu"), (u_longlong_t
)guid
);
2544 (void) strlcpy(zc
.zc_name
, zhp
->zpool_name
, sizeof (zc
.zc_name
));
2546 zc
.zc_cookie
= VDEV_STATE_FAULTED
;
2549 if (ioctl(hdl
->libzfs_fd
, ZFS_IOC_VDEV_SET_STATE
, &zc
) == 0)
2556 * There are no other replicas of this device.
2558 return (zfs_error(hdl
, EZFS_NOREPLICAS
, msg
));
2561 return (zpool_standard_error(hdl
, errno
, msg
));
2567 * Mark the given vdev degraded.
2570 zpool_vdev_degrade(zpool_handle_t
*zhp
, uint64_t guid
, vdev_aux_t aux
)
2572 zfs_cmd_t zc
= {"\0"};
2574 libzfs_handle_t
*hdl
= zhp
->zpool_hdl
;
2576 (void) snprintf(msg
, sizeof (msg
),
2577 dgettext(TEXT_DOMAIN
, "cannot degrade %llu"), (u_longlong_t
)guid
);
2579 (void) strlcpy(zc
.zc_name
, zhp
->zpool_name
, sizeof (zc
.zc_name
));
2581 zc
.zc_cookie
= VDEV_STATE_DEGRADED
;
2584 if (ioctl(hdl
->libzfs_fd
, ZFS_IOC_VDEV_SET_STATE
, &zc
) == 0)
2587 return (zpool_standard_error(hdl
, errno
, msg
));
2591 * Returns TRUE if the given nvlist is a vdev that was originally swapped in as
2595 is_replacing_spare(nvlist_t
*search
, nvlist_t
*tgt
, int which
)
2601 if (nvlist_lookup_nvlist_array(search
, ZPOOL_CONFIG_CHILDREN
, &child
,
2603 verify(nvlist_lookup_string(search
, ZPOOL_CONFIG_TYPE
,
2606 if (strcmp(type
, VDEV_TYPE_SPARE
) == 0 &&
2607 children
== 2 && child
[which
] == tgt
)
2610 for (c
= 0; c
< children
; c
++)
2611 if (is_replacing_spare(child
[c
], tgt
, which
))
2619 * Attach new_disk (fully described by nvroot) to old_disk.
2620 * If 'replacing' is specified, the new disk will replace the old one.
2623 zpool_vdev_attach(zpool_handle_t
*zhp
,
2624 const char *old_disk
, const char *new_disk
, nvlist_t
*nvroot
, int replacing
)
2626 zfs_cmd_t zc
= {"\0"};
2630 boolean_t avail_spare
, l2cache
, islog
;
2635 nvlist_t
*config_root
;
2636 libzfs_handle_t
*hdl
= zhp
->zpool_hdl
;
2637 boolean_t rootpool
= zpool_is_bootable(zhp
);
2640 (void) snprintf(msg
, sizeof (msg
), dgettext(TEXT_DOMAIN
,
2641 "cannot replace %s with %s"), old_disk
, new_disk
);
2643 (void) snprintf(msg
, sizeof (msg
), dgettext(TEXT_DOMAIN
,
2644 "cannot attach %s to %s"), new_disk
, old_disk
);
2646 #if defined(__sun__) || defined(__sun)
2648 * If this is a root pool, make sure that we're not attaching an
2649 * EFI labeled device.
2651 if (rootpool
&& pool_uses_efi(nvroot
)) {
2652 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
2653 "EFI labeled devices are not supported on root pools."));
2654 return (zfs_error(hdl
, EZFS_POOL_NOTSUP
, msg
));
2658 (void) strlcpy(zc
.zc_name
, zhp
->zpool_name
, sizeof (zc
.zc_name
));
2659 if ((tgt
= zpool_find_vdev(zhp
, old_disk
, &avail_spare
, &l2cache
,
2661 return (zfs_error(hdl
, EZFS_NODEVICE
, msg
));
2664 return (zfs_error(hdl
, EZFS_ISSPARE
, msg
));
2667 return (zfs_error(hdl
, EZFS_ISL2CACHE
, msg
));
2669 verify(nvlist_lookup_uint64(tgt
, ZPOOL_CONFIG_GUID
, &zc
.zc_guid
) == 0);
2670 zc
.zc_cookie
= replacing
;
2672 if (nvlist_lookup_nvlist_array(nvroot
, ZPOOL_CONFIG_CHILDREN
,
2673 &child
, &children
) != 0 || children
!= 1) {
2674 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
2675 "new device must be a single disk"));
2676 return (zfs_error(hdl
, EZFS_INVALCONFIG
, msg
));
2679 verify(nvlist_lookup_nvlist(zpool_get_config(zhp
, NULL
),
2680 ZPOOL_CONFIG_VDEV_TREE
, &config_root
) == 0);
2682 if ((newname
= zpool_vdev_name(NULL
, NULL
, child
[0], B_FALSE
)) == NULL
)
2686 * If the target is a hot spare that has been swapped in, we can only
2687 * replace it with another hot spare.
2690 nvlist_lookup_uint64(tgt
, ZPOOL_CONFIG_IS_SPARE
, &val
) == 0 &&
2691 (zpool_find_vdev(zhp
, newname
, &avail_spare
, &l2cache
,
2692 NULL
) == NULL
|| !avail_spare
) &&
2693 is_replacing_spare(config_root
, tgt
, 1)) {
2694 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
2695 "can only be replaced by another hot spare"));
2697 return (zfs_error(hdl
, EZFS_BADTARGET
, msg
));
2702 if (zcmd_write_conf_nvlist(hdl
, &zc
, nvroot
) != 0)
2705 ret
= zfs_ioctl(hdl
, ZFS_IOC_VDEV_ATTACH
, &zc
);
2707 zcmd_free_nvlists(&zc
);
2712 * XXX need a better way to prevent user from
2713 * booting up a half-baked vdev.
2715 (void) fprintf(stderr
, dgettext(TEXT_DOMAIN
, "Make "
2716 "sure to wait until resilver is done "
2717 "before rebooting.\n"));
2725 * Can't attach to or replace this type of vdev.
2728 uint64_t version
= zpool_get_prop_int(zhp
,
2729 ZPOOL_PROP_VERSION
, NULL
);
2732 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
2733 "cannot replace a log with a spare"));
2734 else if (version
>= SPA_VERSION_MULTI_REPLACE
)
2735 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
2736 "already in replacing/spare config; wait "
2737 "for completion or use 'zpool detach'"));
2739 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
2740 "cannot replace a replacing device"));
2742 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
2743 "can only attach to mirrors and top-level "
2746 (void) zfs_error(hdl
, EZFS_BADTARGET
, msg
);
2751 * The new device must be a single disk.
2753 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
2754 "new device must be a single disk"));
2755 (void) zfs_error(hdl
, EZFS_INVALCONFIG
, msg
);
2759 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
, "%s is busy"),
2761 (void) zfs_error(hdl
, EZFS_BADDEV
, msg
);
2766 * The new device is too small.
2768 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
2769 "device is too small"));
2770 (void) zfs_error(hdl
, EZFS_BADDEV
, msg
);
2775 * The new device has a different optimal sector size.
2777 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
2778 "new device has a different optimal sector size; use the "
2779 "option '-o ashift=N' to override the optimal size"));
2780 (void) zfs_error(hdl
, EZFS_BADDEV
, msg
);
2785 * The resulting top-level vdev spec won't fit in the label.
2787 (void) zfs_error(hdl
, EZFS_DEVOVERFLOW
, msg
);
2791 (void) zpool_standard_error(hdl
, errno
, msg
);
2798 * Detach the specified device.
2801 zpool_vdev_detach(zpool_handle_t
*zhp
, const char *path
)
2803 zfs_cmd_t zc
= {"\0"};
2806 boolean_t avail_spare
, l2cache
;
2807 libzfs_handle_t
*hdl
= zhp
->zpool_hdl
;
2809 (void) snprintf(msg
, sizeof (msg
),
2810 dgettext(TEXT_DOMAIN
, "cannot detach %s"), path
);
2812 (void) strlcpy(zc
.zc_name
, zhp
->zpool_name
, sizeof (zc
.zc_name
));
2813 if ((tgt
= zpool_find_vdev(zhp
, path
, &avail_spare
, &l2cache
,
2815 return (zfs_error(hdl
, EZFS_NODEVICE
, msg
));
2818 return (zfs_error(hdl
, EZFS_ISSPARE
, msg
));
2821 return (zfs_error(hdl
, EZFS_ISL2CACHE
, msg
));
2823 verify(nvlist_lookup_uint64(tgt
, ZPOOL_CONFIG_GUID
, &zc
.zc_guid
) == 0);
2825 if (zfs_ioctl(hdl
, ZFS_IOC_VDEV_DETACH
, &zc
) == 0)
2832 * Can't detach from this type of vdev.
2834 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
, "only "
2835 "applicable to mirror and replacing vdevs"));
2836 (void) zfs_error(hdl
, EZFS_BADTARGET
, msg
);
2841 * There are no other replicas of this device.
2843 (void) zfs_error(hdl
, EZFS_NOREPLICAS
, msg
);
2847 (void) zpool_standard_error(hdl
, errno
, msg
);
2854 * Find a mirror vdev in the source nvlist.
2856 * The mchild array contains a list of disks in one of the top-level mirrors
2857 * of the source pool. The schild array contains a list of disks that the
2858 * user specified on the command line. We loop over the mchild array to
2859 * see if any entry in the schild array matches.
2861 * If a disk in the mchild array is found in the schild array, we return
2862 * the index of that entry. Otherwise we return -1.
2865 find_vdev_entry(zpool_handle_t
*zhp
, nvlist_t
**mchild
, uint_t mchildren
,
2866 nvlist_t
**schild
, uint_t schildren
)
2870 for (mc
= 0; mc
< mchildren
; mc
++) {
2872 char *mpath
= zpool_vdev_name(zhp
->zpool_hdl
, zhp
,
2873 mchild
[mc
], B_FALSE
);
2875 for (sc
= 0; sc
< schildren
; sc
++) {
2876 char *spath
= zpool_vdev_name(zhp
->zpool_hdl
, zhp
,
2877 schild
[sc
], B_FALSE
);
2878 boolean_t result
= (strcmp(mpath
, spath
) == 0);
2894 * Split a mirror pool. If newroot points to null, then a new nvlist
2895 * is generated and it is the responsibility of the caller to free it.
2898 zpool_vdev_split(zpool_handle_t
*zhp
, char *newname
, nvlist_t
**newroot
,
2899 nvlist_t
*props
, splitflags_t flags
)
2901 zfs_cmd_t zc
= {"\0"};
2903 nvlist_t
*tree
, *config
, **child
, **newchild
, *newconfig
= NULL
;
2904 nvlist_t
**varray
= NULL
, *zc_props
= NULL
;
2905 uint_t c
, children
, newchildren
, lastlog
= 0, vcount
, found
= 0;
2906 libzfs_handle_t
*hdl
= zhp
->zpool_hdl
;
2908 boolean_t freelist
= B_FALSE
, memory_err
= B_TRUE
;
2911 (void) snprintf(msg
, sizeof (msg
),
2912 dgettext(TEXT_DOMAIN
, "Unable to split %s"), zhp
->zpool_name
);
2914 if (!zpool_name_valid(hdl
, B_FALSE
, newname
))
2915 return (zfs_error(hdl
, EZFS_INVALIDNAME
, msg
));
2917 if ((config
= zpool_get_config(zhp
, NULL
)) == NULL
) {
2918 (void) fprintf(stderr
, gettext("Internal error: unable to "
2919 "retrieve pool configuration\n"));
2923 verify(nvlist_lookup_nvlist(config
, ZPOOL_CONFIG_VDEV_TREE
, &tree
)
2925 verify(nvlist_lookup_uint64(config
, ZPOOL_CONFIG_VERSION
, &vers
) == 0);
2928 prop_flags_t flags
= { .create
= B_FALSE
, .import
= B_TRUE
};
2929 if ((zc_props
= zpool_valid_proplist(hdl
, zhp
->zpool_name
,
2930 props
, vers
, flags
, msg
)) == NULL
)
2934 if (nvlist_lookup_nvlist_array(tree
, ZPOOL_CONFIG_CHILDREN
, &child
,
2936 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
2937 "Source pool is missing vdev tree"));
2939 nvlist_free(zc_props
);
2943 varray
= zfs_alloc(hdl
, children
* sizeof (nvlist_t
*));
2946 if (*newroot
== NULL
||
2947 nvlist_lookup_nvlist_array(*newroot
, ZPOOL_CONFIG_CHILDREN
,
2948 &newchild
, &newchildren
) != 0)
2951 for (c
= 0; c
< children
; c
++) {
2952 uint64_t is_log
= B_FALSE
, is_hole
= B_FALSE
;
2954 nvlist_t
**mchild
, *vdev
;
2959 * Unlike cache & spares, slogs are stored in the
2960 * ZPOOL_CONFIG_CHILDREN array. We filter them out here.
2962 (void) nvlist_lookup_uint64(child
[c
], ZPOOL_CONFIG_IS_LOG
,
2964 (void) nvlist_lookup_uint64(child
[c
], ZPOOL_CONFIG_IS_HOLE
,
2966 if (is_log
|| is_hole
) {
2968 * Create a hole vdev and put it in the config.
2970 if (nvlist_alloc(&vdev
, NV_UNIQUE_NAME
, 0) != 0)
2972 if (nvlist_add_string(vdev
, ZPOOL_CONFIG_TYPE
,
2973 VDEV_TYPE_HOLE
) != 0)
2975 if (nvlist_add_uint64(vdev
, ZPOOL_CONFIG_IS_HOLE
,
2980 varray
[vcount
++] = vdev
;
2984 verify(nvlist_lookup_string(child
[c
], ZPOOL_CONFIG_TYPE
, &type
)
2986 if (strcmp(type
, VDEV_TYPE_MIRROR
) != 0) {
2987 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
2988 "Source pool must be composed only of mirrors\n"));
2989 retval
= zfs_error(hdl
, EZFS_INVALCONFIG
, msg
);
2993 verify(nvlist_lookup_nvlist_array(child
[c
],
2994 ZPOOL_CONFIG_CHILDREN
, &mchild
, &mchildren
) == 0);
2996 /* find or add an entry for this top-level vdev */
2997 if (newchildren
> 0 &&
2998 (entry
= find_vdev_entry(zhp
, mchild
, mchildren
,
2999 newchild
, newchildren
)) >= 0) {
3000 /* We found a disk that the user specified. */
3001 vdev
= mchild
[entry
];
3004 /* User didn't specify a disk for this vdev. */
3005 vdev
= mchild
[mchildren
- 1];
3008 if (nvlist_dup(vdev
, &varray
[vcount
++], 0) != 0)
3012 /* did we find every disk the user specified? */
3013 if (found
!= newchildren
) {
3014 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
, "Device list must "
3015 "include at most one disk from each mirror"));
3016 retval
= zfs_error(hdl
, EZFS_INVALCONFIG
, msg
);
3020 /* Prepare the nvlist for populating. */
3021 if (*newroot
== NULL
) {
3022 if (nvlist_alloc(newroot
, NV_UNIQUE_NAME
, 0) != 0)
3025 if (nvlist_add_string(*newroot
, ZPOOL_CONFIG_TYPE
,
3026 VDEV_TYPE_ROOT
) != 0)
3029 verify(nvlist_remove_all(*newroot
, ZPOOL_CONFIG_CHILDREN
) == 0);
3032 /* Add all the children we found */
3033 if (nvlist_add_nvlist_array(*newroot
, ZPOOL_CONFIG_CHILDREN
, varray
,
3034 lastlog
== 0 ? vcount
: lastlog
) != 0)
3038 * If we're just doing a dry run, exit now with success.
3041 memory_err
= B_FALSE
;
3046 /* now build up the config list & call the ioctl */
3047 if (nvlist_alloc(&newconfig
, NV_UNIQUE_NAME
, 0) != 0)
3050 if (nvlist_add_nvlist(newconfig
,
3051 ZPOOL_CONFIG_VDEV_TREE
, *newroot
) != 0 ||
3052 nvlist_add_string(newconfig
,
3053 ZPOOL_CONFIG_POOL_NAME
, newname
) != 0 ||
3054 nvlist_add_uint64(newconfig
, ZPOOL_CONFIG_VERSION
, vers
) != 0)
3058 * The new pool is automatically part of the namespace unless we
3059 * explicitly export it.
3062 zc
.zc_cookie
= ZPOOL_EXPORT_AFTER_SPLIT
;
3063 (void) strlcpy(zc
.zc_name
, zhp
->zpool_name
, sizeof (zc
.zc_name
));
3064 (void) strlcpy(zc
.zc_string
, newname
, sizeof (zc
.zc_string
));
3065 if (zcmd_write_conf_nvlist(hdl
, &zc
, newconfig
) != 0)
3067 if (zc_props
!= NULL
&& zcmd_write_src_nvlist(hdl
, &zc
, zc_props
) != 0)
3070 if (zfs_ioctl(hdl
, ZFS_IOC_VDEV_SPLIT
, &zc
) != 0) {
3071 retval
= zpool_standard_error(hdl
, errno
, msg
);
3076 memory_err
= B_FALSE
;
3079 if (varray
!= NULL
) {
3082 for (v
= 0; v
< vcount
; v
++)
3083 nvlist_free(varray
[v
]);
3086 zcmd_free_nvlists(&zc
);
3088 nvlist_free(zc_props
);
3090 nvlist_free(newconfig
);
3092 nvlist_free(*newroot
);
3100 return (no_memory(hdl
));
3106 * Remove the given device. Currently, this is supported only for hot spares
3107 * and level 2 cache devices.
3110 zpool_vdev_remove(zpool_handle_t
*zhp
, const char *path
)
3112 zfs_cmd_t zc
= {"\0"};
3115 boolean_t avail_spare
, l2cache
, islog
;
3116 libzfs_handle_t
*hdl
= zhp
->zpool_hdl
;
3119 (void) snprintf(msg
, sizeof (msg
),
3120 dgettext(TEXT_DOMAIN
, "cannot remove %s"), path
);
3122 (void) strlcpy(zc
.zc_name
, zhp
->zpool_name
, sizeof (zc
.zc_name
));
3123 if ((tgt
= zpool_find_vdev(zhp
, path
, &avail_spare
, &l2cache
,
3125 return (zfs_error(hdl
, EZFS_NODEVICE
, msg
));
3127 * XXX - this should just go away.
3129 if (!avail_spare
&& !l2cache
&& !islog
) {
3130 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
3131 "only inactive hot spares, cache, top-level, "
3132 "or log devices can be removed"));
3133 return (zfs_error(hdl
, EZFS_NODEVICE
, msg
));
3136 version
= zpool_get_prop_int(zhp
, ZPOOL_PROP_VERSION
, NULL
);
3137 if (islog
&& version
< SPA_VERSION_HOLES
) {
3138 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
3139 "pool must be upgrade to support log removal"));
3140 return (zfs_error(hdl
, EZFS_BADVERSION
, msg
));
3143 verify(nvlist_lookup_uint64(tgt
, ZPOOL_CONFIG_GUID
, &zc
.zc_guid
) == 0);
3145 if (zfs_ioctl(hdl
, ZFS_IOC_VDEV_REMOVE
, &zc
) == 0)
3148 return (zpool_standard_error(hdl
, errno
, msg
));
3152 * Clear the errors for the pool, or the particular device if specified.
3155 zpool_clear(zpool_handle_t
*zhp
, const char *path
, nvlist_t
*rewindnvl
)
3157 zfs_cmd_t zc
= {"\0"};
3160 zpool_rewind_policy_t policy
;
3161 boolean_t avail_spare
, l2cache
;
3162 libzfs_handle_t
*hdl
= zhp
->zpool_hdl
;
3163 nvlist_t
*nvi
= NULL
;
3167 (void) snprintf(msg
, sizeof (msg
),
3168 dgettext(TEXT_DOMAIN
, "cannot clear errors for %s"),
3171 (void) snprintf(msg
, sizeof (msg
),
3172 dgettext(TEXT_DOMAIN
, "cannot clear errors for %s"),
3175 (void) strlcpy(zc
.zc_name
, zhp
->zpool_name
, sizeof (zc
.zc_name
));
3177 if ((tgt
= zpool_find_vdev(zhp
, path
, &avail_spare
,
3178 &l2cache
, NULL
)) == 0)
3179 return (zfs_error(hdl
, EZFS_NODEVICE
, msg
));
3182 * Don't allow error clearing for hot spares. Do allow
3183 * error clearing for l2cache devices.
3186 return (zfs_error(hdl
, EZFS_ISSPARE
, msg
));
3188 verify(nvlist_lookup_uint64(tgt
, ZPOOL_CONFIG_GUID
,
3192 zpool_get_rewind_policy(rewindnvl
, &policy
);
3193 zc
.zc_cookie
= policy
.zrp_request
;
3195 if (zcmd_alloc_dst_nvlist(hdl
, &zc
, zhp
->zpool_config_size
* 2) != 0)
3198 if (zcmd_write_src_nvlist(hdl
, &zc
, rewindnvl
) != 0)
3201 while ((error
= zfs_ioctl(hdl
, ZFS_IOC_CLEAR
, &zc
)) != 0 &&
3203 if (zcmd_expand_dst_nvlist(hdl
, &zc
) != 0) {
3204 zcmd_free_nvlists(&zc
);
3209 if (!error
|| ((policy
.zrp_request
& ZPOOL_TRY_REWIND
) &&
3210 errno
!= EPERM
&& errno
!= EACCES
)) {
3211 if (policy
.zrp_request
&
3212 (ZPOOL_DO_REWIND
| ZPOOL_TRY_REWIND
)) {
3213 (void) zcmd_read_dst_nvlist(hdl
, &zc
, &nvi
);
3214 zpool_rewind_exclaim(hdl
, zc
.zc_name
,
3215 ((policy
.zrp_request
& ZPOOL_TRY_REWIND
) != 0),
3219 zcmd_free_nvlists(&zc
);
3223 zcmd_free_nvlists(&zc
);
3224 return (zpool_standard_error(hdl
, errno
, msg
));
3228 * Similar to zpool_clear(), but takes a GUID (used by fmd).
3231 zpool_vdev_clear(zpool_handle_t
*zhp
, uint64_t guid
)
3233 zfs_cmd_t zc
= {"\0"};
3235 libzfs_handle_t
*hdl
= zhp
->zpool_hdl
;
3237 (void) snprintf(msg
, sizeof (msg
),
3238 dgettext(TEXT_DOMAIN
, "cannot clear errors for %llx"),
3239 (u_longlong_t
)guid
);
3241 (void) strlcpy(zc
.zc_name
, zhp
->zpool_name
, sizeof (zc
.zc_name
));
3243 zc
.zc_cookie
= ZPOOL_NO_REWIND
;
3245 if (ioctl(hdl
->libzfs_fd
, ZFS_IOC_CLEAR
, &zc
) == 0)
3248 return (zpool_standard_error(hdl
, errno
, msg
));
3252 * Change the GUID for a pool.
3255 zpool_reguid(zpool_handle_t
*zhp
)
3258 libzfs_handle_t
*hdl
= zhp
->zpool_hdl
;
3259 zfs_cmd_t zc
= {"\0"};
3261 (void) snprintf(msg
, sizeof (msg
),
3262 dgettext(TEXT_DOMAIN
, "cannot reguid '%s'"), zhp
->zpool_name
);
3264 (void) strlcpy(zc
.zc_name
, zhp
->zpool_name
, sizeof (zc
.zc_name
));
3265 if (zfs_ioctl(hdl
, ZFS_IOC_POOL_REGUID
, &zc
) == 0)
3268 return (zpool_standard_error(hdl
, errno
, msg
));
3275 zpool_reopen(zpool_handle_t
*zhp
)
3277 zfs_cmd_t zc
= {"\0"};
3279 libzfs_handle_t
*hdl
= zhp
->zpool_hdl
;
3281 (void) snprintf(msg
, sizeof (msg
),
3282 dgettext(TEXT_DOMAIN
, "cannot reopen '%s'"),
3285 (void) strlcpy(zc
.zc_name
, zhp
->zpool_name
, sizeof (zc
.zc_name
));
3286 if (zfs_ioctl(hdl
, ZFS_IOC_POOL_REOPEN
, &zc
) == 0)
3288 return (zpool_standard_error(hdl
, errno
, msg
));
3292 * Convert from a devid string to a path.
3295 devid_to_path(char *devid_str
)
3300 devid_nmlist_t
*list
= NULL
;
3303 if (devid_str_decode(devid_str
, &devid
, &minor
) != 0)
3306 ret
= devid_deviceid_to_nmlist("/dev", devid
, minor
, &list
);
3308 devid_str_free(minor
);
3314 if ((path
= strdup(list
[0].devname
)) == NULL
)
3317 devid_free_nmlist(list
);
3323 * Convert from a path to a devid string.
3326 path_to_devid(const char *path
)
3332 if ((fd
= open(path
, O_RDONLY
)) < 0)
3337 if (devid_get(fd
, &devid
) == 0) {
3338 if (devid_get_minor_name(fd
, &minor
) == 0)
3339 ret
= devid_str_encode(devid
, minor
);
3341 devid_str_free(minor
);
3350 * Issue the necessary ioctl() to update the stored path value for the vdev. We
3351 * ignore any failure here, since a common case is for an unprivileged user to
3352 * type 'zpool status', and we'll display the correct information anyway.
3355 set_path(zpool_handle_t
*zhp
, nvlist_t
*nv
, const char *path
)
3357 zfs_cmd_t zc
= {"\0"};
3359 (void) strncpy(zc
.zc_name
, zhp
->zpool_name
, sizeof (zc
.zc_name
));
3360 (void) strncpy(zc
.zc_value
, path
, sizeof (zc
.zc_value
));
3361 verify(nvlist_lookup_uint64(nv
, ZPOOL_CONFIG_GUID
,
3364 (void) ioctl(zhp
->zpool_hdl
->libzfs_fd
, ZFS_IOC_VDEV_SETPATH
, &zc
);
3368 * Remove partition suffix from a vdev path. Partition suffixes may take three
3369 * forms: "-partX", "pX", or "X", where X is a string of digits. The second
3370 * case only occurs when the suffix is preceded by a digit, i.e. "md0p0" The
3371 * third case only occurs when preceded by a string matching the regular
3372 * expression "^[hs]d[a-z]+", i.e. a scsi or ide disk.
3375 strip_partition(libzfs_handle_t
*hdl
, char *path
)
3377 char *tmp
= zfs_strdup(hdl
, path
);
3378 char *part
= NULL
, *d
= NULL
;
3380 if ((part
= strstr(tmp
, "-part")) && part
!= tmp
) {
3382 } else if ((part
= strrchr(tmp
, 'p')) &&
3383 part
> tmp
+ 1 && isdigit(*(part
-1))) {
3385 } else if ((tmp
[0] == 'h' || tmp
[0] == 's') && tmp
[1] == 'd') {
3386 for (d
= &tmp
[2]; isalpha(*d
); part
= ++d
);
3388 if (part
&& d
&& *d
!= '\0') {
3389 for (; isdigit(*d
); d
++);
3396 #define PATH_BUF_LEN 64
3399 * Given a vdev, return the name to display in iostat. If the vdev has a path,
3400 * we use that, stripping off any leading "/dev/dsk/"; if not, we use the type.
3401 * We also check if this is a whole disk, in which case we strip off the
3402 * trailing 's0' slice name.
3404 * This routine is also responsible for identifying when disks have been
3405 * reconfigured in a new location. The kernel will have opened the device by
3406 * devid, but the path will still refer to the old location. To catch this, we
3407 * first do a path -> devid translation (which is fast for the common case). If
3408 * the devid matches, we're done. If not, we do a reverse devid -> path
3409 * translation and issue the appropriate ioctl() to update the path of the vdev.
3410 * If 'zhp' is NULL, then this is an exported pool, and we don't need to do any
3414 zpool_vdev_name(libzfs_handle_t
*hdl
, zpool_handle_t
*zhp
, nvlist_t
*nv
,
3417 char *path
, *devid
, *type
;
3419 char buf
[PATH_BUF_LEN
];
3420 char tmpbuf
[PATH_BUF_LEN
];
3424 if (nvlist_lookup_uint64(nv
, ZPOOL_CONFIG_NOT_PRESENT
,
3426 verify(nvlist_lookup_uint64(nv
, ZPOOL_CONFIG_GUID
,
3428 (void) snprintf(buf
, sizeof (buf
), "%llu",
3429 (u_longlong_t
)value
);
3431 } else if (nvlist_lookup_string(nv
, ZPOOL_CONFIG_PATH
, &path
) == 0) {
3433 * If the device is dead (faulted, offline, etc) then don't
3434 * bother opening it. Otherwise we may be forcing the user to
3435 * open a misbehaving device, which can have undesirable
3438 if ((nvlist_lookup_uint64_array(nv
, ZPOOL_CONFIG_VDEV_STATS
,
3439 (uint64_t **)&vs
, &vsc
) != 0 ||
3440 vs
->vs_state
>= VDEV_STATE_DEGRADED
) &&
3442 nvlist_lookup_string(nv
, ZPOOL_CONFIG_DEVID
, &devid
) == 0) {
3444 * Determine if the current path is correct.
3446 char *newdevid
= path_to_devid(path
);
3448 if (newdevid
== NULL
||
3449 strcmp(devid
, newdevid
) != 0) {
3452 if ((newpath
= devid_to_path(devid
)) != NULL
) {
3454 * Update the path appropriately.
3456 set_path(zhp
, nv
, newpath
);
3457 if (nvlist_add_string(nv
,
3458 ZPOOL_CONFIG_PATH
, newpath
) == 0)
3459 verify(nvlist_lookup_string(nv
,
3467 devid_str_free(newdevid
);
3471 * For a block device only use the name.
3473 verify(nvlist_lookup_string(nv
, ZPOOL_CONFIG_TYPE
, &type
) == 0);
3474 if (strcmp(type
, VDEV_TYPE_DISK
) == 0) {
3475 path
= strrchr(path
, '/');
3480 * Remove the partition from the path it this is a whole disk.
3482 if (nvlist_lookup_uint64(nv
, ZPOOL_CONFIG_WHOLE_DISK
,
3483 &value
) == 0 && value
) {
3484 return (strip_partition(hdl
, path
));
3487 verify(nvlist_lookup_string(nv
, ZPOOL_CONFIG_TYPE
, &path
) == 0);
3490 * If it's a raidz device, we need to stick in the parity level.
3492 if (strcmp(path
, VDEV_TYPE_RAIDZ
) == 0) {
3494 verify(nvlist_lookup_uint64(nv
, ZPOOL_CONFIG_NPARITY
,
3496 (void) snprintf(buf
, sizeof (buf
), "%s%llu", path
,
3497 (u_longlong_t
)value
);
3502 * We identify each top-level vdev by using a <type-id>
3503 * naming convention.
3508 verify(nvlist_lookup_uint64(nv
, ZPOOL_CONFIG_ID
,
3510 (void) snprintf(tmpbuf
, sizeof (tmpbuf
), "%s-%llu",
3511 path
, (u_longlong_t
)id
);
3516 return (zfs_strdup(hdl
, path
));
3520 zbookmark_compare(const void *a
, const void *b
)
3522 return (memcmp(a
, b
, sizeof (zbookmark_phys_t
)));
3526 * Retrieve the persistent error log, uniquify the members, and return to the
3530 zpool_get_errlog(zpool_handle_t
*zhp
, nvlist_t
**nverrlistp
)
3532 zfs_cmd_t zc
= {"\0"};
3534 zbookmark_phys_t
*zb
= NULL
;
3538 * Retrieve the raw error list from the kernel. If the number of errors
3539 * has increased, allocate more space and continue until we get the
3542 verify(nvlist_lookup_uint64(zhp
->zpool_config
, ZPOOL_CONFIG_ERRCOUNT
,
3546 if ((zc
.zc_nvlist_dst
= (uintptr_t)zfs_alloc(zhp
->zpool_hdl
,
3547 count
* sizeof (zbookmark_phys_t
))) == (uintptr_t)NULL
)
3549 zc
.zc_nvlist_dst_size
= count
;
3550 (void) strcpy(zc
.zc_name
, zhp
->zpool_name
);
3552 if (ioctl(zhp
->zpool_hdl
->libzfs_fd
, ZFS_IOC_ERROR_LOG
,
3554 free((void *)(uintptr_t)zc
.zc_nvlist_dst
);
3555 if (errno
== ENOMEM
) {
3558 count
= zc
.zc_nvlist_dst_size
;
3559 dst
= zfs_alloc(zhp
->zpool_hdl
, count
*
3560 sizeof (zbookmark_phys_t
));
3563 zc
.zc_nvlist_dst
= (uintptr_t)dst
;
3573 * Sort the resulting bookmarks. This is a little confusing due to the
3574 * implementation of ZFS_IOC_ERROR_LOG. The bookmarks are copied last
3575 * to first, and 'zc_nvlist_dst_size' indicates the number of boomarks
3576 * _not_ copied as part of the process. So we point the start of our
3577 * array appropriate and decrement the total number of elements.
3579 zb
= ((zbookmark_phys_t
*)(uintptr_t)zc
.zc_nvlist_dst
) +
3580 zc
.zc_nvlist_dst_size
;
3581 count
-= zc
.zc_nvlist_dst_size
;
3583 qsort(zb
, count
, sizeof (zbookmark_phys_t
), zbookmark_compare
);
3585 verify(nvlist_alloc(nverrlistp
, 0, KM_SLEEP
) == 0);
3588 * Fill in the nverrlistp with nvlist's of dataset and object numbers.
3590 for (i
= 0; i
< count
; i
++) {
3593 /* ignoring zb_blkid and zb_level for now */
3594 if (i
> 0 && zb
[i
-1].zb_objset
== zb
[i
].zb_objset
&&
3595 zb
[i
-1].zb_object
== zb
[i
].zb_object
)
3598 if (nvlist_alloc(&nv
, NV_UNIQUE_NAME
, KM_SLEEP
) != 0)
3600 if (nvlist_add_uint64(nv
, ZPOOL_ERR_DATASET
,
3601 zb
[i
].zb_objset
) != 0) {
3605 if (nvlist_add_uint64(nv
, ZPOOL_ERR_OBJECT
,
3606 zb
[i
].zb_object
) != 0) {
3610 if (nvlist_add_nvlist(*nverrlistp
, "ejk", nv
) != 0) {
3617 free((void *)(uintptr_t)zc
.zc_nvlist_dst
);
3621 free((void *)(uintptr_t)zc
.zc_nvlist_dst
);
3622 return (no_memory(zhp
->zpool_hdl
));
3626 * Upgrade a ZFS pool to the latest on-disk version.
3629 zpool_upgrade(zpool_handle_t
*zhp
, uint64_t new_version
)
3631 zfs_cmd_t zc
= {"\0"};
3632 libzfs_handle_t
*hdl
= zhp
->zpool_hdl
;
3634 (void) strcpy(zc
.zc_name
, zhp
->zpool_name
);
3635 zc
.zc_cookie
= new_version
;
3637 if (zfs_ioctl(hdl
, ZFS_IOC_POOL_UPGRADE
, &zc
) != 0)
3638 return (zpool_standard_error_fmt(hdl
, errno
,
3639 dgettext(TEXT_DOMAIN
, "cannot upgrade '%s'"),
3645 zfs_save_arguments(int argc
, char **argv
, char *string
, int len
)
3649 (void) strlcpy(string
, basename(argv
[0]), len
);
3650 for (i
= 1; i
< argc
; i
++) {
3651 (void) strlcat(string
, " ", len
);
3652 (void) strlcat(string
, argv
[i
], len
);
3657 zpool_log_history(libzfs_handle_t
*hdl
, const char *message
)
3659 zfs_cmd_t zc
= {"\0"};
3663 args
= fnvlist_alloc();
3664 fnvlist_add_string(args
, "message", message
);
3665 err
= zcmd_write_src_nvlist(hdl
, &zc
, args
);
3667 err
= ioctl(hdl
->libzfs_fd
, ZFS_IOC_LOG_HISTORY
, &zc
);
3669 zcmd_free_nvlists(&zc
);
3674 * Perform ioctl to get some command history of a pool.
3676 * 'buf' is the buffer to fill up to 'len' bytes. 'off' is the
3677 * logical offset of the history buffer to start reading from.
3679 * Upon return, 'off' is the next logical offset to read from and
3680 * 'len' is the actual amount of bytes read into 'buf'.
3683 get_history(zpool_handle_t
*zhp
, char *buf
, uint64_t *off
, uint64_t *len
)
3685 zfs_cmd_t zc
= {"\0"};
3686 libzfs_handle_t
*hdl
= zhp
->zpool_hdl
;
3688 (void) strlcpy(zc
.zc_name
, zhp
->zpool_name
, sizeof (zc
.zc_name
));
3690 zc
.zc_history
= (uint64_t)(uintptr_t)buf
;
3691 zc
.zc_history_len
= *len
;
3692 zc
.zc_history_offset
= *off
;
3694 if (ioctl(hdl
->libzfs_fd
, ZFS_IOC_POOL_GET_HISTORY
, &zc
) != 0) {
3697 return (zfs_error_fmt(hdl
, EZFS_PERM
,
3698 dgettext(TEXT_DOMAIN
,
3699 "cannot show history for pool '%s'"),
3702 return (zfs_error_fmt(hdl
, EZFS_NOHISTORY
,
3703 dgettext(TEXT_DOMAIN
, "cannot get history for pool "
3704 "'%s'"), zhp
->zpool_name
));
3706 return (zfs_error_fmt(hdl
, EZFS_BADVERSION
,
3707 dgettext(TEXT_DOMAIN
, "cannot get history for pool "
3708 "'%s', pool must be upgraded"), zhp
->zpool_name
));
3710 return (zpool_standard_error_fmt(hdl
, errno
,
3711 dgettext(TEXT_DOMAIN
,
3712 "cannot get history for '%s'"), zhp
->zpool_name
));
3716 *len
= zc
.zc_history_len
;
3717 *off
= zc
.zc_history_offset
;
3723 * Process the buffer of nvlists, unpacking and storing each nvlist record
3724 * into 'records'. 'leftover' is set to the number of bytes that weren't
3725 * processed as there wasn't a complete record.
3728 zpool_history_unpack(char *buf
, uint64_t bytes_read
, uint64_t *leftover
,
3729 nvlist_t
***records
, uint_t
*numrecords
)
3735 while (bytes_read
> sizeof (reclen
)) {
3737 /* get length of packed record (stored as little endian) */
3738 for (i
= 0, reclen
= 0; i
< sizeof (reclen
); i
++)
3739 reclen
+= (uint64_t)(((uchar_t
*)buf
)[i
]) << (8*i
);
3741 if (bytes_read
< sizeof (reclen
) + reclen
)
3745 if (nvlist_unpack(buf
+ sizeof (reclen
), reclen
, &nv
, 0) != 0)
3747 bytes_read
-= sizeof (reclen
) + reclen
;
3748 buf
+= sizeof (reclen
) + reclen
;
3750 /* add record to nvlist array */
3752 if (ISP2(*numrecords
+ 1)) {
3753 *records
= realloc(*records
,
3754 *numrecords
* 2 * sizeof (nvlist_t
*));
3756 (*records
)[*numrecords
- 1] = nv
;
3759 *leftover
= bytes_read
;
3763 #define HIS_BUF_LEN (128*1024)
3766 * Retrieve the command history of a pool.
3769 zpool_get_history(zpool_handle_t
*zhp
, nvlist_t
**nvhisp
)
3771 char buf
[HIS_BUF_LEN
];
3773 nvlist_t
**records
= NULL
;
3774 uint_t numrecords
= 0;
3778 uint64_t bytes_read
= sizeof (buf
);
3781 if ((err
= get_history(zhp
, buf
, &off
, &bytes_read
)) != 0)
3784 /* if nothing else was read in, we're at EOF, just return */
3788 if ((err
= zpool_history_unpack(buf
, bytes_read
,
3789 &leftover
, &records
, &numrecords
)) != 0)
3797 verify(nvlist_alloc(nvhisp
, NV_UNIQUE_NAME
, 0) == 0);
3798 verify(nvlist_add_nvlist_array(*nvhisp
, ZPOOL_HIST_RECORD
,
3799 records
, numrecords
) == 0);
3801 for (i
= 0; i
< numrecords
; i
++)
3802 nvlist_free(records
[i
]);
3809 * Retrieve the next event given the passed 'zevent_fd' file descriptor.
3810 * If there is a new event available 'nvp' will contain a newly allocated
3811 * nvlist and 'dropped' will be set to the number of missed events since
3812 * the last call to this function. When 'nvp' is set to NULL it indicates
3813 * no new events are available. In either case the function returns 0 and
3814 * it is up to the caller to free 'nvp'. In the case of a fatal error the
3815 * function will return a non-zero value. When the function is called in
3816 * blocking mode (the default, unless the ZEVENT_NONBLOCK flag is passed),
3817 * it will not return until a new event is available.
3820 zpool_events_next(libzfs_handle_t
*hdl
, nvlist_t
**nvp
,
3821 int *dropped
, unsigned flags
, int zevent_fd
)
3823 zfs_cmd_t zc
= {"\0"};
3828 zc
.zc_cleanup_fd
= zevent_fd
;
3830 if (flags
& ZEVENT_NONBLOCK
)
3831 zc
.zc_guid
= ZEVENT_NONBLOCK
;
3833 if (zcmd_alloc_dst_nvlist(hdl
, &zc
, ZEVENT_SIZE
) != 0)
3837 if (zfs_ioctl(hdl
, ZFS_IOC_EVENTS_NEXT
, &zc
) != 0) {
3840 error
= zfs_error_fmt(hdl
, EZFS_POOLUNAVAIL
,
3841 dgettext(TEXT_DOMAIN
, "zfs shutdown"));
3844 /* Blocking error case should not occur */
3845 if (!(flags
& ZEVENT_NONBLOCK
))
3846 error
= zpool_standard_error_fmt(hdl
, errno
,
3847 dgettext(TEXT_DOMAIN
, "cannot get event"));
3851 if (zcmd_expand_dst_nvlist(hdl
, &zc
) != 0) {
3852 error
= zfs_error_fmt(hdl
, EZFS_NOMEM
,
3853 dgettext(TEXT_DOMAIN
, "cannot get event"));
3859 error
= zpool_standard_error_fmt(hdl
, errno
,
3860 dgettext(TEXT_DOMAIN
, "cannot get event"));
3865 error
= zcmd_read_dst_nvlist(hdl
, &zc
, nvp
);
3869 *dropped
= (int)zc
.zc_cookie
;
3871 zcmd_free_nvlists(&zc
);
3880 zpool_events_clear(libzfs_handle_t
*hdl
, int *count
)
3882 zfs_cmd_t zc
= {"\0"};
3885 (void) snprintf(msg
, sizeof (msg
), dgettext(TEXT_DOMAIN
,
3886 "cannot clear events"));
3888 if (zfs_ioctl(hdl
, ZFS_IOC_EVENTS_CLEAR
, &zc
) != 0)
3889 return (zpool_standard_error_fmt(hdl
, errno
, msg
));
3892 *count
= (int)zc
.zc_cookie
; /* # of events cleared */
3898 * Seek to a specific EID, ZEVENT_SEEK_START, or ZEVENT_SEEK_END for
3899 * the passed zevent_fd file handle. On success zero is returned,
3900 * otherwise -1 is returned and hdl->libzfs_error is set to the errno.
3903 zpool_events_seek(libzfs_handle_t
*hdl
, uint64_t eid
, int zevent_fd
)
3905 zfs_cmd_t zc
= {"\0"};
3909 zc
.zc_cleanup_fd
= zevent_fd
;
3911 if (zfs_ioctl(hdl
, ZFS_IOC_EVENTS_SEEK
, &zc
) != 0) {
3914 error
= zfs_error_fmt(hdl
, EZFS_NOENT
,
3915 dgettext(TEXT_DOMAIN
, "cannot get event"));
3919 error
= zfs_error_fmt(hdl
, EZFS_NOMEM
,
3920 dgettext(TEXT_DOMAIN
, "cannot get event"));
3924 error
= zpool_standard_error_fmt(hdl
, errno
,
3925 dgettext(TEXT_DOMAIN
, "cannot get event"));
3934 zpool_obj_to_path(zpool_handle_t
*zhp
, uint64_t dsobj
, uint64_t obj
,
3935 char *pathname
, size_t len
)
3937 zfs_cmd_t zc
= {"\0"};
3938 boolean_t mounted
= B_FALSE
;
3939 char *mntpnt
= NULL
;
3940 char dsname
[MAXNAMELEN
];
3943 /* special case for the MOS */
3944 (void) snprintf(pathname
, len
, "<metadata>:<0x%llx>",
3949 /* get the dataset's name */
3950 (void) strlcpy(zc
.zc_name
, zhp
->zpool_name
, sizeof (zc
.zc_name
));
3952 if (ioctl(zhp
->zpool_hdl
->libzfs_fd
,
3953 ZFS_IOC_DSOBJ_TO_DSNAME
, &zc
) != 0) {
3954 /* just write out a path of two object numbers */
3955 (void) snprintf(pathname
, len
, "<0x%llx>:<0x%llx>",
3956 (longlong_t
)dsobj
, (longlong_t
)obj
);
3959 (void) strlcpy(dsname
, zc
.zc_value
, sizeof (dsname
));
3961 /* find out if the dataset is mounted */
3962 mounted
= is_mounted(zhp
->zpool_hdl
, dsname
, &mntpnt
);
3964 /* get the corrupted object's path */
3965 (void) strlcpy(zc
.zc_name
, dsname
, sizeof (zc
.zc_name
));
3967 if (ioctl(zhp
->zpool_hdl
->libzfs_fd
, ZFS_IOC_OBJ_TO_PATH
,
3970 (void) snprintf(pathname
, len
, "%s%s", mntpnt
,
3973 (void) snprintf(pathname
, len
, "%s:%s",
3974 dsname
, zc
.zc_value
);
3977 (void) snprintf(pathname
, len
, "%s:<0x%llx>", dsname
,
3984 * Read the EFI label from the config, if a label does not exist then
3985 * pass back the error to the caller. If the caller has passed a non-NULL
3986 * diskaddr argument then we set it to the starting address of the EFI
3990 read_efi_label(nvlist_t
*config
, diskaddr_t
*sb
)
3994 char diskname
[MAXPATHLEN
];
3997 if (nvlist_lookup_string(config
, ZPOOL_CONFIG_PATH
, &path
) != 0)
4000 (void) snprintf(diskname
, sizeof (diskname
), "%s%s", DISK_ROOT
,
4001 strrchr(path
, '/'));
4002 if ((fd
= open(diskname
, O_RDWR
|O_DIRECT
)) >= 0) {
4003 struct dk_gpt
*vtoc
;
4005 if ((err
= efi_alloc_and_read(fd
, &vtoc
)) >= 0) {
4007 *sb
= vtoc
->efi_parts
[0].p_start
;
4016 * determine where a partition starts on a disk in the current
4020 find_start_block(nvlist_t
*config
)
4024 diskaddr_t sb
= MAXOFFSET_T
;
4027 if (nvlist_lookup_nvlist_array(config
,
4028 ZPOOL_CONFIG_CHILDREN
, &child
, &children
) != 0) {
4029 if (nvlist_lookup_uint64(config
,
4030 ZPOOL_CONFIG_WHOLE_DISK
,
4031 &wholedisk
) != 0 || !wholedisk
) {
4032 return (MAXOFFSET_T
);
4034 if (read_efi_label(config
, &sb
) < 0)
4039 for (c
= 0; c
< children
; c
++) {
4040 sb
= find_start_block(child
[c
]);
4041 if (sb
!= MAXOFFSET_T
) {
4045 return (MAXOFFSET_T
);
4049 zpool_label_disk_wait(char *path
, int timeout
)
4051 struct stat64 statbuf
;
4055 * Wait timeout miliseconds for a newly created device to be available
4056 * from the given path. There is a small window when a /dev/ device
4057 * will exist and the udev link will not, so we must wait for the
4058 * symlink. Depending on the udev rules this may take a few seconds.
4060 for (i
= 0; i
< timeout
; i
++) {
4064 if ((stat64(path
, &statbuf
) == 0) && (errno
== 0))
4072 zpool_label_disk_check(char *path
)
4074 struct dk_gpt
*vtoc
;
4077 if ((fd
= open(path
, O_RDWR
|O_DIRECT
)) < 0)
4080 if ((err
= efi_alloc_and_read(fd
, &vtoc
)) != 0) {
4085 if (vtoc
->efi_flags
& EFI_GPT_PRIMARY_CORRUPT
) {
4097 * Label an individual disk. The name provided is the short name,
4098 * stripped of any leading /dev path.
4101 zpool_label_disk(libzfs_handle_t
*hdl
, zpool_handle_t
*zhp
, char *name
)
4103 char path
[MAXPATHLEN
];
4104 struct dk_gpt
*vtoc
;
4106 size_t resv
= EFI_MIN_RESV_SIZE
;
4107 uint64_t slice_size
;
4108 diskaddr_t start_block
;
4111 /* prepare an error message just in case */
4112 (void) snprintf(errbuf
, sizeof (errbuf
),
4113 dgettext(TEXT_DOMAIN
, "cannot label '%s'"), name
);
4118 #if defined(__sun__) || defined(__sun)
4119 if (zpool_is_bootable(zhp
)) {
4120 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
4121 "EFI labeled devices are not supported on root "
4123 return (zfs_error(hdl
, EZFS_POOL_NOTSUP
, errbuf
));
4127 verify(nvlist_lookup_nvlist(zhp
->zpool_config
,
4128 ZPOOL_CONFIG_VDEV_TREE
, &nvroot
) == 0);
4130 if (zhp
->zpool_start_block
== 0)
4131 start_block
= find_start_block(nvroot
);
4133 start_block
= zhp
->zpool_start_block
;
4134 zhp
->zpool_start_block
= start_block
;
4137 start_block
= NEW_START_BLOCK
;
4140 (void) snprintf(path
, sizeof (path
), "%s/%s", DISK_ROOT
, name
);
4142 if ((fd
= open(path
, O_RDWR
|O_DIRECT
)) < 0) {
4144 * This shouldn't happen. We've long since verified that this
4145 * is a valid device.
4147 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
, "cannot "
4148 "label '%s': unable to open device: %d"), path
, errno
);
4149 return (zfs_error(hdl
, EZFS_OPENFAILED
, errbuf
));
4152 if (efi_alloc_and_init(fd
, EFI_NUMPAR
, &vtoc
) != 0) {
4154 * The only way this can fail is if we run out of memory, or we
4155 * were unable to read the disk's capacity
4157 if (errno
== ENOMEM
)
4158 (void) no_memory(hdl
);
4161 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
, "cannot "
4162 "label '%s': unable to read disk capacity"), path
);
4164 return (zfs_error(hdl
, EZFS_NOCAP
, errbuf
));
4167 slice_size
= vtoc
->efi_last_u_lba
+ 1;
4168 slice_size
-= EFI_MIN_RESV_SIZE
;
4169 if (start_block
== MAXOFFSET_T
)
4170 start_block
= NEW_START_BLOCK
;
4171 slice_size
-= start_block
;
4172 slice_size
= P2ALIGN(slice_size
, PARTITION_END_ALIGNMENT
);
4174 vtoc
->efi_parts
[0].p_start
= start_block
;
4175 vtoc
->efi_parts
[0].p_size
= slice_size
;
4178 * Why we use V_USR: V_BACKUP confuses users, and is considered
4179 * disposable by some EFI utilities (since EFI doesn't have a backup
4180 * slice). V_UNASSIGNED is supposed to be used only for zero size
4181 * partitions, and efi_write() will fail if we use it. V_ROOT, V_BOOT,
4182 * etc. were all pretty specific. V_USR is as close to reality as we
4183 * can get, in the absence of V_OTHER.
4185 vtoc
->efi_parts
[0].p_tag
= V_USR
;
4186 (void) strcpy(vtoc
->efi_parts
[0].p_name
, "zfs");
4188 vtoc
->efi_parts
[8].p_start
= slice_size
+ start_block
;
4189 vtoc
->efi_parts
[8].p_size
= resv
;
4190 vtoc
->efi_parts
[8].p_tag
= V_RESERVED
;
4192 if ((rval
= efi_write(fd
, vtoc
)) != 0 || (rval
= efi_rescan(fd
)) != 0) {
4194 * Some block drivers (like pcata) may not support EFI
4195 * GPT labels. Print out a helpful error message dir-
4196 * ecting the user to manually label the disk and give
4202 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
, "try using "
4203 "parted(8) and then provide a specific slice: %d"), rval
);
4204 return (zfs_error(hdl
, EZFS_LABELFAILED
, errbuf
));
4210 /* Wait for the first expected partition to appear. */
4212 (void) snprintf(path
, sizeof (path
), "%s/%s", DISK_ROOT
, name
);
4213 (void) zfs_append_partition(path
, MAXPATHLEN
);
4215 rval
= zpool_label_disk_wait(path
, 3000);
4217 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
, "failed to "
4218 "detect device partitions on '%s': %d"), path
, rval
);
4219 return (zfs_error(hdl
, EZFS_LABELFAILED
, errbuf
));
4222 /* We can't be to paranoid. Read the label back and verify it. */
4223 (void) snprintf(path
, sizeof (path
), "%s/%s", DISK_ROOT
, name
);
4224 rval
= zpool_label_disk_check(path
);
4226 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
, "freshly written "
4227 "EFI label on '%s' is damaged. Ensure\nthis device "
4228 "is not in in use, and is functioning properly: %d"),
4230 return (zfs_error(hdl
, EZFS_LABELFAILED
, errbuf
));