4 * The contents of this file are subject to the terms of the
5 * Common Development and Distribution License (the "License").
6 * You may not use this file except in compliance with the License.
8 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
9 * or http://www.opensolaris.org/os/licensing.
10 * See the License for the specific language governing permissions
11 * and limitations under the License.
13 * When distributing Covered Code, include this CDDL HEADER in each
14 * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
15 * If applicable, add the following below this CDDL HEADER, with the
16 * fields enclosed by brackets "[]" replaced with your own identifying
17 * information: Portions Copyright [yyyy] [name of copyright owner]
23 * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
37 #include <sys/efi_partition.h>
39 #include <sys/zfs_ioctl.h>
42 #include "zfs_namecheck.h"
44 #include "libzfs_impl.h"
45 #include "zfs_comutil.h"
47 static int read_efi_label(nvlist_t
*config
, diskaddr_t
*sb
);
49 typedef struct prop_flags
{
50 int create
:1; /* Validate property on creation */
51 int import
:1; /* Validate property on import */
55 * ====================================================================
56 * zpool property functions
57 * ====================================================================
61 zpool_get_all_props(zpool_handle_t
*zhp
)
63 zfs_cmd_t zc
= { "\0", "\0", "\0", "\0", 0 };
64 libzfs_handle_t
*hdl
= zhp
->zpool_hdl
;
66 (void) strlcpy(zc
.zc_name
, zhp
->zpool_name
, sizeof (zc
.zc_name
));
68 if (zcmd_alloc_dst_nvlist(hdl
, &zc
, 0) != 0)
71 while (ioctl(hdl
->libzfs_fd
, ZFS_IOC_POOL_GET_PROPS
, &zc
) != 0) {
72 if (errno
== ENOMEM
) {
73 if (zcmd_expand_dst_nvlist(hdl
, &zc
) != 0) {
74 zcmd_free_nvlists(&zc
);
78 zcmd_free_nvlists(&zc
);
83 if (zcmd_read_dst_nvlist(hdl
, &zc
, &zhp
->zpool_props
) != 0) {
84 zcmd_free_nvlists(&zc
);
88 zcmd_free_nvlists(&zc
);
94 zpool_props_refresh(zpool_handle_t
*zhp
)
98 old_props
= zhp
->zpool_props
;
100 if (zpool_get_all_props(zhp
) != 0)
103 nvlist_free(old_props
);
108 zpool_get_prop_string(zpool_handle_t
*zhp
, zpool_prop_t prop
,
114 zprop_source_t source
;
116 nvl
= zhp
->zpool_props
;
117 if (nvlist_lookup_nvlist(nvl
, zpool_prop_to_name(prop
), &nv
) == 0) {
118 verify(nvlist_lookup_uint64(nv
, ZPROP_SOURCE
, &ival
) == 0);
120 verify(nvlist_lookup_string(nv
, ZPROP_VALUE
, &value
) == 0);
122 source
= ZPROP_SRC_DEFAULT
;
123 if ((value
= (char *)zpool_prop_default_string(prop
)) == NULL
)
134 zpool_get_prop_int(zpool_handle_t
*zhp
, zpool_prop_t prop
, zprop_source_t
*src
)
138 zprop_source_t source
;
140 if (zhp
->zpool_props
== NULL
&& zpool_get_all_props(zhp
)) {
142 * zpool_get_all_props() has most likely failed because
143 * the pool is faulted, but if all we need is the top level
144 * vdev's guid then get it from the zhp config nvlist.
146 if ((prop
== ZPOOL_PROP_GUID
) &&
147 (nvlist_lookup_nvlist(zhp
->zpool_config
,
148 ZPOOL_CONFIG_VDEV_TREE
, &nv
) == 0) &&
149 (nvlist_lookup_uint64(nv
, ZPOOL_CONFIG_GUID
, &value
)
153 return (zpool_prop_default_numeric(prop
));
156 nvl
= zhp
->zpool_props
;
157 if (nvlist_lookup_nvlist(nvl
, zpool_prop_to_name(prop
), &nv
) == 0) {
158 verify(nvlist_lookup_uint64(nv
, ZPROP_SOURCE
, &value
) == 0);
160 verify(nvlist_lookup_uint64(nv
, ZPROP_VALUE
, &value
) == 0);
162 source
= ZPROP_SRC_DEFAULT
;
163 value
= zpool_prop_default_numeric(prop
);
173 * Map VDEV STATE to printed strings.
176 zpool_state_to_name(vdev_state_t state
, vdev_aux_t aux
)
181 case VDEV_STATE_CLOSED
:
182 case VDEV_STATE_OFFLINE
:
183 return (gettext("OFFLINE"));
184 case VDEV_STATE_REMOVED
:
185 return (gettext("REMOVED"));
186 case VDEV_STATE_CANT_OPEN
:
187 if (aux
== VDEV_AUX_CORRUPT_DATA
|| aux
== VDEV_AUX_BAD_LOG
)
188 return (gettext("FAULTED"));
189 else if (aux
== VDEV_AUX_SPLIT_POOL
)
190 return (gettext("SPLIT"));
192 return (gettext("UNAVAIL"));
193 case VDEV_STATE_FAULTED
:
194 return (gettext("FAULTED"));
195 case VDEV_STATE_DEGRADED
:
196 return (gettext("DEGRADED"));
197 case VDEV_STATE_HEALTHY
:
198 return (gettext("ONLINE"));
201 return (gettext("UNKNOWN"));
205 * Get a zpool property value for 'prop' and return the value in
206 * a pre-allocated buffer.
209 zpool_get_prop(zpool_handle_t
*zhp
, zpool_prop_t prop
, char *buf
, size_t len
,
210 zprop_source_t
*srctype
)
214 zprop_source_t src
= ZPROP_SRC_NONE
;
219 if (zpool_get_state(zhp
) == POOL_STATE_UNAVAIL
) {
221 case ZPOOL_PROP_NAME
:
222 (void) strlcpy(buf
, zpool_get_name(zhp
), len
);
225 case ZPOOL_PROP_HEALTH
:
226 (void) strlcpy(buf
, "FAULTED", len
);
229 case ZPOOL_PROP_GUID
:
230 intval
= zpool_get_prop_int(zhp
, prop
, &src
);
231 (void) snprintf(buf
, len
, "%llu", (u_longlong_t
)intval
);
234 case ZPOOL_PROP_ALTROOT
:
235 case ZPOOL_PROP_CACHEFILE
:
236 if (zhp
->zpool_props
!= NULL
||
237 zpool_get_all_props(zhp
) == 0) {
239 zpool_get_prop_string(zhp
, prop
, &src
),
247 (void) strlcpy(buf
, "-", len
);
256 if (zhp
->zpool_props
== NULL
&& zpool_get_all_props(zhp
) &&
257 prop
!= ZPOOL_PROP_NAME
)
260 switch (zpool_prop_get_type(prop
)) {
261 case PROP_TYPE_STRING
:
262 (void) strlcpy(buf
, zpool_get_prop_string(zhp
, prop
, &src
),
266 case PROP_TYPE_NUMBER
:
267 intval
= zpool_get_prop_int(zhp
, prop
, &src
);
270 case ZPOOL_PROP_SIZE
:
271 case ZPOOL_PROP_ALLOCATED
:
272 case ZPOOL_PROP_FREE
:
273 case ZPOOL_PROP_ASHIFT
:
274 (void) zfs_nicenum(intval
, buf
, len
);
277 case ZPOOL_PROP_CAPACITY
:
278 (void) snprintf(buf
, len
, "%llu%%",
279 (u_longlong_t
)intval
);
282 case ZPOOL_PROP_DEDUPRATIO
:
283 (void) snprintf(buf
, len
, "%llu.%02llux",
284 (u_longlong_t
)(intval
/ 100),
285 (u_longlong_t
)(intval
% 100));
288 case ZPOOL_PROP_HEALTH
:
289 verify(nvlist_lookup_nvlist(zpool_get_config(zhp
, NULL
),
290 ZPOOL_CONFIG_VDEV_TREE
, &nvroot
) == 0);
291 verify(nvlist_lookup_uint64_array(nvroot
,
292 ZPOOL_CONFIG_VDEV_STATS
, (uint64_t **)&vs
, &vsc
)
295 (void) strlcpy(buf
, zpool_state_to_name(intval
,
299 (void) snprintf(buf
, len
, "%llu", (u_longlong_t
)intval
);
303 case PROP_TYPE_INDEX
:
304 intval
= zpool_get_prop_int(zhp
, prop
, &src
);
305 if (zpool_prop_index_to_string(prop
, intval
, &strval
)
308 (void) strlcpy(buf
, strval
, len
);
322 * Check if the bootfs name has the same pool name as it is set to.
323 * Assuming bootfs is a valid dataset name.
326 bootfs_name_valid(const char *pool
, char *bootfs
)
328 int len
= strlen(pool
);
330 if (!zfs_name_valid(bootfs
, ZFS_TYPE_FILESYSTEM
|ZFS_TYPE_SNAPSHOT
))
333 if (strncmp(pool
, bootfs
, len
) == 0 &&
334 (bootfs
[len
] == '/' || bootfs
[len
] == '\0'))
341 * Inspect the configuration to determine if any of the devices contain
345 pool_uses_efi(nvlist_t
*config
)
350 if (nvlist_lookup_nvlist_array(config
, ZPOOL_CONFIG_CHILDREN
,
351 &child
, &children
) != 0)
352 return (read_efi_label(config
, NULL
) >= 0);
354 for (c
= 0; c
< children
; c
++) {
355 if (pool_uses_efi(child
[c
]))
362 pool_is_bootable(zpool_handle_t
*zhp
)
364 char bootfs
[ZPOOL_MAXNAMELEN
];
366 return (zpool_get_prop(zhp
, ZPOOL_PROP_BOOTFS
, bootfs
,
367 sizeof (bootfs
), NULL
) == 0 && strncmp(bootfs
, "-",
368 sizeof (bootfs
)) != 0);
373 * Given an nvlist of zpool properties to be set, validate that they are
374 * correct, and parse any numeric properties (index, boolean, etc) if they are
375 * specified as strings.
378 zpool_valid_proplist(libzfs_handle_t
*hdl
, const char *poolname
,
379 nvlist_t
*props
, uint64_t version
, prop_flags_t flags
, char *errbuf
)
387 struct stat64 statbuf
;
391 if (nvlist_alloc(&retprops
, NV_UNIQUE_NAME
, 0) != 0) {
392 (void) no_memory(hdl
);
397 while ((elem
= nvlist_next_nvpair(props
, elem
)) != NULL
) {
398 const char *propname
= nvpair_name(elem
);
401 * Make sure this property is valid and applies to this type.
403 if ((prop
= zpool_name_to_prop(propname
)) == ZPROP_INVAL
) {
404 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
405 "invalid property '%s'"), propname
);
406 (void) zfs_error(hdl
, EZFS_BADPROP
, errbuf
);
410 if (zpool_prop_readonly(prop
)) {
411 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
, "'%s' "
412 "is readonly"), propname
);
413 (void) zfs_error(hdl
, EZFS_PROPREADONLY
, errbuf
);
417 if (zprop_parse_value(hdl
, elem
, prop
, ZFS_TYPE_POOL
, retprops
,
418 &strval
, &intval
, errbuf
) != 0)
422 * Perform additional checking for specific properties.
427 case ZPOOL_PROP_VERSION
:
428 if (intval
< version
|| intval
> SPA_VERSION
) {
429 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
430 "property '%s' number %d is invalid."),
432 (void) zfs_error(hdl
, EZFS_BADVERSION
, errbuf
);
437 case ZPOOL_PROP_ASHIFT
:
439 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
440 "property '%s' can only be set at "
441 "creation time"), propname
);
442 (void) zfs_error(hdl
, EZFS_BADPROP
, errbuf
);
446 if (intval
!= 0 && (intval
< 9 || intval
> 12)) {
447 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
448 "property '%s' number %d is invalid."),
450 (void) zfs_error(hdl
, EZFS_BADPROP
, errbuf
);
455 case ZPOOL_PROP_BOOTFS
:
456 if (flags
.create
|| flags
.import
) {
457 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
458 "property '%s' cannot be set at creation "
459 "or import time"), propname
);
460 (void) zfs_error(hdl
, EZFS_BADPROP
, errbuf
);
464 if (version
< SPA_VERSION_BOOTFS
) {
465 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
466 "pool must be upgraded to support "
467 "'%s' property"), propname
);
468 (void) zfs_error(hdl
, EZFS_BADVERSION
, errbuf
);
473 * bootfs property value has to be a dataset name and
474 * the dataset has to be in the same pool as it sets to.
476 if (strval
[0] != '\0' && !bootfs_name_valid(poolname
,
478 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
, "'%s' "
479 "is an invalid name"), strval
);
480 (void) zfs_error(hdl
, EZFS_INVALIDNAME
, errbuf
);
484 if ((zhp
= zpool_open_canfail(hdl
, poolname
)) == NULL
) {
485 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
486 "could not open pool '%s'"), poolname
);
487 (void) zfs_error(hdl
, EZFS_OPENFAILED
, errbuf
);
490 verify(nvlist_lookup_nvlist(zpool_get_config(zhp
, NULL
),
491 ZPOOL_CONFIG_VDEV_TREE
, &nvroot
) == 0);
493 #if defined(__sun__) || defined(__sun)
495 * bootfs property cannot be set on a disk which has
498 if (pool_uses_efi(nvroot
)) {
499 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
500 "property '%s' not supported on "
501 "EFI labeled devices"), propname
);
502 (void) zfs_error(hdl
, EZFS_POOL_NOTSUP
, errbuf
);
510 case ZPOOL_PROP_ALTROOT
:
511 if (!flags
.create
&& !flags
.import
) {
512 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
513 "property '%s' can only be set during pool "
514 "creation or import"), propname
);
515 (void) zfs_error(hdl
, EZFS_BADPROP
, errbuf
);
519 if (strval
[0] != '/') {
520 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
521 "bad alternate root '%s'"), strval
);
522 (void) zfs_error(hdl
, EZFS_BADPATH
, errbuf
);
527 case ZPOOL_PROP_CACHEFILE
:
528 if (strval
[0] == '\0')
531 if (strcmp(strval
, "none") == 0)
534 if (strval
[0] != '/') {
535 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
536 "property '%s' must be empty, an "
537 "absolute path, or 'none'"), propname
);
538 (void) zfs_error(hdl
, EZFS_BADPATH
, errbuf
);
542 slash
= strrchr(strval
, '/');
544 if (slash
[1] == '\0' || strcmp(slash
, "/.") == 0 ||
545 strcmp(slash
, "/..") == 0) {
546 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
547 "'%s' is not a valid file"), strval
);
548 (void) zfs_error(hdl
, EZFS_BADPATH
, errbuf
);
554 if (strval
[0] != '\0' &&
555 (stat64(strval
, &statbuf
) != 0 ||
556 !S_ISDIR(statbuf
.st_mode
))) {
557 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
558 "'%s' is not a valid directory"),
560 (void) zfs_error(hdl
, EZFS_BADPATH
, errbuf
);
567 case ZPOOL_PROP_READONLY
:
569 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
570 "property '%s' can only be set at "
571 "import time"), propname
);
572 (void) zfs_error(hdl
, EZFS_BADPROP
, errbuf
);
581 nvlist_free(retprops
);
586 * Set zpool property : propname=propval.
589 zpool_set_prop(zpool_handle_t
*zhp
, const char *propname
, const char *propval
)
591 zfs_cmd_t zc
= { "\0", "\0", "\0", "\0", 0 };
594 nvlist_t
*nvl
= NULL
;
597 prop_flags_t flags
= { 0 };
599 (void) snprintf(errbuf
, sizeof (errbuf
),
600 dgettext(TEXT_DOMAIN
, "cannot set property for '%s'"),
603 if (nvlist_alloc(&nvl
, NV_UNIQUE_NAME
, 0) != 0)
604 return (no_memory(zhp
->zpool_hdl
));
606 if (nvlist_add_string(nvl
, propname
, propval
) != 0) {
608 return (no_memory(zhp
->zpool_hdl
));
611 version
= zpool_get_prop_int(zhp
, ZPOOL_PROP_VERSION
, NULL
);
612 if ((realprops
= zpool_valid_proplist(zhp
->zpool_hdl
,
613 zhp
->zpool_name
, nvl
, version
, flags
, errbuf
)) == NULL
) {
622 * Execute the corresponding ioctl() to set this property.
624 (void) strlcpy(zc
.zc_name
, zhp
->zpool_name
, sizeof (zc
.zc_name
));
626 if (zcmd_write_src_nvlist(zhp
->zpool_hdl
, &zc
, nvl
) != 0) {
631 ret
= zfs_ioctl(zhp
->zpool_hdl
, ZFS_IOC_POOL_SET_PROPS
, &zc
);
633 zcmd_free_nvlists(&zc
);
637 (void) zpool_standard_error(zhp
->zpool_hdl
, errno
, errbuf
);
639 (void) zpool_props_refresh(zhp
);
645 zpool_expand_proplist(zpool_handle_t
*zhp
, zprop_list_t
**plp
)
647 libzfs_handle_t
*hdl
= zhp
->zpool_hdl
;
649 char buf
[ZFS_MAXPROPLEN
];
651 if (zprop_expand_list(hdl
, plp
, ZFS_TYPE_POOL
) != 0)
654 for (entry
= *plp
; entry
!= NULL
; entry
= entry
->pl_next
) {
659 if (entry
->pl_prop
!= ZPROP_INVAL
&&
660 zpool_get_prop(zhp
, entry
->pl_prop
, buf
, sizeof (buf
),
662 if (strlen(buf
) > entry
->pl_width
)
663 entry
->pl_width
= strlen(buf
);
672 * Don't start the slice at the default block of 34; many storage
673 * devices will use a stripe width of 128k, other vendors prefer a 1m
674 * alignment. It is best to play it safe and ensure a 1m alignment
675 * give 512b blocks. When the block size is larger by a power of 2
676 * we will still be 1m aligned.
678 #define NEW_START_BLOCK 2048
681 * Validate the given pool name, optionally putting an extended error message in
685 zpool_name_valid(libzfs_handle_t
*hdl
, boolean_t isopen
, const char *pool
)
691 ret
= pool_namecheck(pool
, &why
, &what
);
694 * The rules for reserved pool names were extended at a later point.
695 * But we need to support users with existing pools that may now be
696 * invalid. So we only check for this expanded set of names during a
697 * create (or import), and only in userland.
699 if (ret
== 0 && !isopen
&&
700 (strncmp(pool
, "mirror", 6) == 0 ||
701 strncmp(pool
, "raidz", 5) == 0 ||
702 strncmp(pool
, "spare", 5) == 0 ||
703 strcmp(pool
, "log") == 0)) {
706 dgettext(TEXT_DOMAIN
, "name is reserved"));
714 case NAME_ERR_TOOLONG
:
716 dgettext(TEXT_DOMAIN
, "name is too long"));
719 case NAME_ERR_INVALCHAR
:
721 dgettext(TEXT_DOMAIN
, "invalid character "
722 "'%c' in pool name"), what
);
725 case NAME_ERR_NOLETTER
:
726 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
727 "name must begin with a letter"));
730 case NAME_ERR_RESERVED
:
731 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
732 "name is reserved"));
735 case NAME_ERR_DISKLIKE
:
736 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
737 "pool name is reserved"));
740 case NAME_ERR_LEADING_SLASH
:
741 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
742 "leading slash in name"));
745 case NAME_ERR_EMPTY_COMPONENT
:
746 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
747 "empty component in name"));
750 case NAME_ERR_TRAILING_SLASH
:
751 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
752 "trailing slash in name"));
755 case NAME_ERR_MULTIPLE_AT
:
756 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
757 "multiple '@' delimiters in name"));
760 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
761 "permission set is missing '@'"));
772 * Open a handle to the given pool, even if the pool is currently in the FAULTED
776 zpool_open_canfail(libzfs_handle_t
*hdl
, const char *pool
)
782 * Make sure the pool name is valid.
784 if (!zpool_name_valid(hdl
, B_TRUE
, pool
)) {
785 (void) zfs_error_fmt(hdl
, EZFS_INVALIDNAME
,
786 dgettext(TEXT_DOMAIN
, "cannot open '%s'"),
791 if ((zhp
= zfs_alloc(hdl
, sizeof (zpool_handle_t
))) == NULL
)
794 zhp
->zpool_hdl
= hdl
;
795 (void) strlcpy(zhp
->zpool_name
, pool
, sizeof (zhp
->zpool_name
));
797 if (zpool_refresh_stats(zhp
, &missing
) != 0) {
803 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
, "no such pool"));
804 (void) zfs_error_fmt(hdl
, EZFS_NOENT
,
805 dgettext(TEXT_DOMAIN
, "cannot open '%s'"), pool
);
814 * Like the above, but silent on error. Used when iterating over pools (because
815 * the configuration cache may be out of date).
818 zpool_open_silent(libzfs_handle_t
*hdl
, const char *pool
, zpool_handle_t
**ret
)
823 if ((zhp
= zfs_alloc(hdl
, sizeof (zpool_handle_t
))) == NULL
)
826 zhp
->zpool_hdl
= hdl
;
827 (void) strlcpy(zhp
->zpool_name
, pool
, sizeof (zhp
->zpool_name
));
829 if (zpool_refresh_stats(zhp
, &missing
) != 0) {
845 * Similar to zpool_open_canfail(), but refuses to open pools in the faulted
849 zpool_open(libzfs_handle_t
*hdl
, const char *pool
)
853 if ((zhp
= zpool_open_canfail(hdl
, pool
)) == NULL
)
856 if (zhp
->zpool_state
== POOL_STATE_UNAVAIL
) {
857 (void) zfs_error_fmt(hdl
, EZFS_POOLUNAVAIL
,
858 dgettext(TEXT_DOMAIN
, "cannot open '%s'"), zhp
->zpool_name
);
867 * Close the handle. Simply frees the memory associated with the handle.
870 zpool_close(zpool_handle_t
*zhp
)
872 if (zhp
->zpool_config
)
873 nvlist_free(zhp
->zpool_config
);
874 if (zhp
->zpool_old_config
)
875 nvlist_free(zhp
->zpool_old_config
);
876 if (zhp
->zpool_props
)
877 nvlist_free(zhp
->zpool_props
);
882 * Return the name of the pool.
885 zpool_get_name(zpool_handle_t
*zhp
)
887 return (zhp
->zpool_name
);
892 * Return the state of the pool (ACTIVE or UNAVAILABLE)
895 zpool_get_state(zpool_handle_t
*zhp
)
897 return (zhp
->zpool_state
);
901 * Create the named pool, using the provided vdev list. It is assumed
902 * that the consumer has already validated the contents of the nvlist, so we
903 * don't have to worry about error semantics.
906 zpool_create(libzfs_handle_t
*hdl
, const char *pool
, nvlist_t
*nvroot
,
907 nvlist_t
*props
, nvlist_t
*fsprops
)
909 zfs_cmd_t zc
= { "\0", "\0", "\0", "\0", 0 };
910 nvlist_t
*zc_fsprops
= NULL
;
911 nvlist_t
*zc_props
= NULL
;
916 (void) snprintf(msg
, sizeof (msg
), dgettext(TEXT_DOMAIN
,
917 "cannot create '%s'"), pool
);
919 if (!zpool_name_valid(hdl
, B_FALSE
, pool
))
920 return (zfs_error(hdl
, EZFS_INVALIDNAME
, msg
));
922 if (zcmd_write_conf_nvlist(hdl
, &zc
, nvroot
) != 0)
926 prop_flags_t flags
= { .create
= B_TRUE
, .import
= B_FALSE
};
928 if ((zc_props
= zpool_valid_proplist(hdl
, pool
, props
,
929 SPA_VERSION_1
, flags
, msg
)) == NULL
) {
938 zoned
= ((nvlist_lookup_string(fsprops
,
939 zfs_prop_to_name(ZFS_PROP_ZONED
), &zonestr
) == 0) &&
940 strcmp(zonestr
, "on") == 0);
942 if ((zc_fsprops
= zfs_valid_proplist(hdl
,
943 ZFS_TYPE_FILESYSTEM
, fsprops
, zoned
, NULL
, msg
)) == NULL
) {
947 (nvlist_alloc(&zc_props
, NV_UNIQUE_NAME
, 0) != 0)) {
950 if (nvlist_add_nvlist(zc_props
,
951 ZPOOL_ROOTFS_PROPS
, zc_fsprops
) != 0) {
956 if (zc_props
&& zcmd_write_src_nvlist(hdl
, &zc
, zc_props
) != 0)
959 (void) strlcpy(zc
.zc_name
, pool
, sizeof (zc
.zc_name
));
961 if ((ret
= zfs_ioctl(hdl
, ZFS_IOC_POOL_CREATE
, &zc
)) != 0) {
963 zcmd_free_nvlists(&zc
);
964 nvlist_free(zc_props
);
965 nvlist_free(zc_fsprops
);
970 * This can happen if the user has specified the same
971 * device multiple times. We can't reliably detect this
972 * until we try to add it and see we already have a
973 * label. This can also happen under if the device is
974 * part of an active md or lvm device.
976 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
977 "one or more vdevs refer to the same device, or one of\n"
978 "the devices is part of an active md or lvm device"));
979 return (zfs_error(hdl
, EZFS_BADDEV
, msg
));
983 * This occurs when one of the devices is below
984 * SPA_MINDEVSIZE. Unfortunately, we can't detect which
985 * device was the problem device since there's no
986 * reliable way to determine device size from userland.
991 zfs_nicenum(SPA_MINDEVSIZE
, buf
, sizeof (buf
));
993 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
994 "one or more devices is less than the "
995 "minimum size (%s)"), buf
);
997 return (zfs_error(hdl
, EZFS_BADDEV
, msg
));
1000 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
1001 "one or more devices is out of space"));
1002 return (zfs_error(hdl
, EZFS_BADDEV
, msg
));
1005 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
1006 "cache device must be a disk or disk slice"));
1007 return (zfs_error(hdl
, EZFS_BADDEV
, msg
));
1010 return (zpool_standard_error(hdl
, errno
, msg
));
1015 * If this is an alternate root pool, then we automatically set the
1016 * mountpoint of the root dataset to be '/'.
1018 if (nvlist_lookup_string(props
, zpool_prop_to_name(ZPOOL_PROP_ALTROOT
),
1022 verify((zhp
= zfs_open(hdl
, pool
, ZFS_TYPE_DATASET
)) != NULL
);
1023 verify(zfs_prop_set(zhp
, zfs_prop_to_name(ZFS_PROP_MOUNTPOINT
),
1030 zcmd_free_nvlists(&zc
);
1031 nvlist_free(zc_props
);
1032 nvlist_free(zc_fsprops
);
1037 * Destroy the given pool. It is up to the caller to ensure that there are no
1038 * datasets left in the pool.
1041 zpool_destroy(zpool_handle_t
*zhp
)
1043 zfs_cmd_t zc
= { "\0", "\0", "\0", "\0", 0 };
1044 zfs_handle_t
*zfp
= NULL
;
1045 libzfs_handle_t
*hdl
= zhp
->zpool_hdl
;
1048 if (zhp
->zpool_state
== POOL_STATE_ACTIVE
&&
1049 (zfp
= zfs_open(hdl
, zhp
->zpool_name
, ZFS_TYPE_FILESYSTEM
)) == NULL
)
1052 (void) strlcpy(zc
.zc_name
, zhp
->zpool_name
, sizeof (zc
.zc_name
));
1054 if (zfs_ioctl(hdl
, ZFS_IOC_POOL_DESTROY
, &zc
) != 0) {
1055 (void) snprintf(msg
, sizeof (msg
), dgettext(TEXT_DOMAIN
,
1056 "cannot destroy '%s'"), zhp
->zpool_name
);
1058 if (errno
== EROFS
) {
1059 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
1060 "one or more devices is read only"));
1061 (void) zfs_error(hdl
, EZFS_BADDEV
, msg
);
1063 (void) zpool_standard_error(hdl
, errno
, msg
);
1072 remove_mountpoint(zfp
);
1080 * Add the given vdevs to the pool. The caller must have already performed the
1081 * necessary verification to ensure that the vdev specification is well-formed.
1084 zpool_add(zpool_handle_t
*zhp
, nvlist_t
*nvroot
)
1086 zfs_cmd_t zc
= { "\0", "\0", "\0", "\0", 0 };
1088 libzfs_handle_t
*hdl
= zhp
->zpool_hdl
;
1090 nvlist_t
**spares
, **l2cache
;
1091 uint_t nspares
, nl2cache
;
1093 (void) snprintf(msg
, sizeof (msg
), dgettext(TEXT_DOMAIN
,
1094 "cannot add to '%s'"), zhp
->zpool_name
);
1096 if (zpool_get_prop_int(zhp
, ZPOOL_PROP_VERSION
, NULL
) <
1097 SPA_VERSION_SPARES
&&
1098 nvlist_lookup_nvlist_array(nvroot
, ZPOOL_CONFIG_SPARES
,
1099 &spares
, &nspares
) == 0) {
1100 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
, "pool must be "
1101 "upgraded to add hot spares"));
1102 return (zfs_error(hdl
, EZFS_BADVERSION
, msg
));
1105 if (pool_is_bootable(zhp
) && nvlist_lookup_nvlist_array(nvroot
,
1106 ZPOOL_CONFIG_SPARES
, &spares
, &nspares
) == 0) {
1109 for (s
= 0; s
< nspares
; s
++) {
1112 if (nvlist_lookup_string(spares
[s
], ZPOOL_CONFIG_PATH
,
1113 &path
) == 0 && pool_uses_efi(spares
[s
])) {
1114 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
1115 "device '%s' contains an EFI label and "
1116 "cannot be used on root pools."),
1117 zpool_vdev_name(hdl
, NULL
, spares
[s
],
1119 return (zfs_error(hdl
, EZFS_POOL_NOTSUP
, msg
));
1124 if (zpool_get_prop_int(zhp
, ZPOOL_PROP_VERSION
, NULL
) <
1125 SPA_VERSION_L2CACHE
&&
1126 nvlist_lookup_nvlist_array(nvroot
, ZPOOL_CONFIG_L2CACHE
,
1127 &l2cache
, &nl2cache
) == 0) {
1128 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
, "pool must be "
1129 "upgraded to add cache devices"));
1130 return (zfs_error(hdl
, EZFS_BADVERSION
, msg
));
1133 if (zcmd_write_conf_nvlist(hdl
, &zc
, nvroot
) != 0)
1135 (void) strlcpy(zc
.zc_name
, zhp
->zpool_name
, sizeof (zc
.zc_name
));
1137 if (zfs_ioctl(hdl
, ZFS_IOC_VDEV_ADD
, &zc
) != 0) {
1141 * This can happen if the user has specified the same
1142 * device multiple times. We can't reliably detect this
1143 * until we try to add it and see we already have a
1146 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
1147 "one or more vdevs refer to the same device"));
1148 (void) zfs_error(hdl
, EZFS_BADDEV
, msg
);
1153 * This occurrs when one of the devices is below
1154 * SPA_MINDEVSIZE. Unfortunately, we can't detect which
1155 * device was the problem device since there's no
1156 * reliable way to determine device size from userland.
1161 zfs_nicenum(SPA_MINDEVSIZE
, buf
, sizeof (buf
));
1163 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
1164 "device is less than the minimum "
1167 (void) zfs_error(hdl
, EZFS_BADDEV
, msg
);
1171 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
1172 "pool must be upgraded to add these vdevs"));
1173 (void) zfs_error(hdl
, EZFS_BADVERSION
, msg
);
1177 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
1178 "root pool can not have multiple vdevs"
1179 " or separate logs"));
1180 (void) zfs_error(hdl
, EZFS_POOL_NOTSUP
, msg
);
1184 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
1185 "cache device must be a disk or disk slice"));
1186 (void) zfs_error(hdl
, EZFS_BADDEV
, msg
);
1190 (void) zpool_standard_error(hdl
, errno
, msg
);
1198 zcmd_free_nvlists(&zc
);
1204 * Exports the pool from the system. The caller must ensure that there are no
1205 * mounted datasets in the pool.
1208 zpool_export_common(zpool_handle_t
*zhp
, boolean_t force
, boolean_t hardforce
)
1210 zfs_cmd_t zc
= { "\0", "\0", "\0", "\0", 0 };
1213 (void) snprintf(msg
, sizeof (msg
), dgettext(TEXT_DOMAIN
,
1214 "cannot export '%s'"), zhp
->zpool_name
);
1216 (void) strlcpy(zc
.zc_name
, zhp
->zpool_name
, sizeof (zc
.zc_name
));
1217 zc
.zc_cookie
= force
;
1218 zc
.zc_guid
= hardforce
;
1220 if (zfs_ioctl(zhp
->zpool_hdl
, ZFS_IOC_POOL_EXPORT
, &zc
) != 0) {
1223 zfs_error_aux(zhp
->zpool_hdl
, dgettext(TEXT_DOMAIN
,
1224 "use '-f' to override the following errors:\n"
1225 "'%s' has an active shared spare which could be"
1226 " used by other pools once '%s' is exported."),
1227 zhp
->zpool_name
, zhp
->zpool_name
);
1228 return (zfs_error(zhp
->zpool_hdl
, EZFS_ACTIVE_SPARE
,
1231 return (zpool_standard_error_fmt(zhp
->zpool_hdl
, errno
,
1240 zpool_export(zpool_handle_t
*zhp
, boolean_t force
)
1242 return (zpool_export_common(zhp
, force
, B_FALSE
));
1246 zpool_export_force(zpool_handle_t
*zhp
)
1248 return (zpool_export_common(zhp
, B_TRUE
, B_TRUE
));
1252 zpool_rewind_exclaim(libzfs_handle_t
*hdl
, const char *name
, boolean_t dryrun
,
1255 nvlist_t
*nv
= NULL
;
1261 if (!hdl
->libzfs_printerr
|| config
== NULL
)
1264 if (nvlist_lookup_nvlist(config
, ZPOOL_CONFIG_LOAD_INFO
, &nv
) != 0)
1267 if (nvlist_lookup_uint64(nv
, ZPOOL_CONFIG_LOAD_TIME
, &rewindto
) != 0)
1269 (void) nvlist_lookup_int64(nv
, ZPOOL_CONFIG_REWIND_TIME
, &loss
);
1271 if (localtime_r((time_t *)&rewindto
, &t
) != NULL
&&
1272 strftime(timestr
, 128, "%c", &t
) != 0) {
1274 (void) printf(dgettext(TEXT_DOMAIN
,
1275 "Would be able to return %s "
1276 "to its state as of %s.\n"),
1279 (void) printf(dgettext(TEXT_DOMAIN
,
1280 "Pool %s returned to its state as of %s.\n"),
1284 (void) printf(dgettext(TEXT_DOMAIN
,
1285 "%s approximately %lld "),
1286 dryrun
? "Would discard" : "Discarded",
1287 ((longlong_t
)loss
+ 30) / 60);
1288 (void) printf(dgettext(TEXT_DOMAIN
,
1289 "minutes of transactions.\n"));
1290 } else if (loss
> 0) {
1291 (void) printf(dgettext(TEXT_DOMAIN
,
1292 "%s approximately %lld "),
1293 dryrun
? "Would discard" : "Discarded",
1295 (void) printf(dgettext(TEXT_DOMAIN
,
1296 "seconds of transactions.\n"));
1302 zpool_explain_recover(libzfs_handle_t
*hdl
, const char *name
, int reason
,
1305 nvlist_t
*nv
= NULL
;
1307 uint64_t edata
= UINT64_MAX
;
1312 if (!hdl
->libzfs_printerr
)
1316 (void) printf(dgettext(TEXT_DOMAIN
, "action: "));
1318 (void) printf(dgettext(TEXT_DOMAIN
, "\t"));
1320 /* All attempted rewinds failed if ZPOOL_CONFIG_LOAD_TIME missing */
1321 if (nvlist_lookup_nvlist(config
, ZPOOL_CONFIG_LOAD_INFO
, &nv
) != 0 ||
1322 nvlist_lookup_uint64(nv
, ZPOOL_CONFIG_LOAD_TIME
, &rewindto
) != 0)
1325 (void) nvlist_lookup_int64(nv
, ZPOOL_CONFIG_REWIND_TIME
, &loss
);
1326 (void) nvlist_lookup_uint64(nv
, ZPOOL_CONFIG_LOAD_DATA_ERRORS
,
1329 (void) printf(dgettext(TEXT_DOMAIN
,
1330 "Recovery is possible, but will result in some data loss.\n"));
1332 if (localtime_r((time_t *)&rewindto
, &t
) != NULL
&&
1333 strftime(timestr
, 128, "%c", &t
) != 0) {
1334 (void) printf(dgettext(TEXT_DOMAIN
,
1335 "\tReturning the pool to its state as of %s\n"
1336 "\tshould correct the problem. "),
1339 (void) printf(dgettext(TEXT_DOMAIN
,
1340 "\tReverting the pool to an earlier state "
1341 "should correct the problem.\n\t"));
1345 (void) printf(dgettext(TEXT_DOMAIN
,
1346 "Approximately %lld minutes of data\n"
1347 "\tmust be discarded, irreversibly. "),
1348 ((longlong_t
)loss
+ 30) / 60);
1349 } else if (loss
> 0) {
1350 (void) printf(dgettext(TEXT_DOMAIN
,
1351 "Approximately %lld seconds of data\n"
1352 "\tmust be discarded, irreversibly. "),
1355 if (edata
!= 0 && edata
!= UINT64_MAX
) {
1357 (void) printf(dgettext(TEXT_DOMAIN
,
1358 "After rewind, at least\n"
1359 "\tone persistent user-data error will remain. "));
1361 (void) printf(dgettext(TEXT_DOMAIN
,
1362 "After rewind, several\n"
1363 "\tpersistent user-data errors will remain. "));
1366 (void) printf(dgettext(TEXT_DOMAIN
,
1367 "Recovery can be attempted\n\tby executing 'zpool %s -F %s'. "),
1368 reason
>= 0 ? "clear" : "import", name
);
1370 (void) printf(dgettext(TEXT_DOMAIN
,
1371 "A scrub of the pool\n"
1372 "\tis strongly recommended after recovery.\n"));
1376 (void) printf(dgettext(TEXT_DOMAIN
,
1377 "Destroy and re-create the pool from\n\ta backup source.\n"));
1381 * zpool_import() is a contracted interface. Should be kept the same
1384 * Applications should use zpool_import_props() to import a pool with
1385 * new properties value to be set.
1388 zpool_import(libzfs_handle_t
*hdl
, nvlist_t
*config
, const char *newname
,
1391 nvlist_t
*props
= NULL
;
1394 if (altroot
!= NULL
) {
1395 if (nvlist_alloc(&props
, NV_UNIQUE_NAME
, 0) != 0) {
1396 return (zfs_error_fmt(hdl
, EZFS_NOMEM
,
1397 dgettext(TEXT_DOMAIN
, "cannot import '%s'"),
1401 if (nvlist_add_string(props
,
1402 zpool_prop_to_name(ZPOOL_PROP_ALTROOT
), altroot
) != 0 ||
1403 nvlist_add_string(props
,
1404 zpool_prop_to_name(ZPOOL_PROP_CACHEFILE
), "none") != 0) {
1406 return (zfs_error_fmt(hdl
, EZFS_NOMEM
,
1407 dgettext(TEXT_DOMAIN
, "cannot import '%s'"),
1412 ret
= zpool_import_props(hdl
, config
, newname
, props
,
1420 print_vdev_tree(libzfs_handle_t
*hdl
, const char *name
, nvlist_t
*nv
,
1426 uint64_t is_log
= 0;
1428 (void) nvlist_lookup_uint64(nv
, ZPOOL_CONFIG_IS_LOG
,
1432 (void) printf("\t%*s%s%s\n", indent
, "", name
,
1433 is_log
? " [log]" : "");
1435 if (nvlist_lookup_nvlist_array(nv
, ZPOOL_CONFIG_CHILDREN
,
1436 &child
, &children
) != 0)
1439 for (c
= 0; c
< children
; c
++) {
1440 vname
= zpool_vdev_name(hdl
, NULL
, child
[c
], B_TRUE
);
1441 print_vdev_tree(hdl
, vname
, child
[c
], indent
+ 2);
1447 * Import the given pool using the known configuration and a list of
1448 * properties to be set. The configuration should have come from
1449 * zpool_find_import(). The 'newname' parameters control whether the pool
1450 * is imported with a different name.
1453 zpool_import_props(libzfs_handle_t
*hdl
, nvlist_t
*config
, const char *newname
,
1454 nvlist_t
*props
, int flags
)
1456 zfs_cmd_t zc
= { "\0", "\0", "\0", "\0", 0 };
1457 zpool_rewind_policy_t policy
;
1458 nvlist_t
*nv
= NULL
;
1459 nvlist_t
*nvinfo
= NULL
;
1460 nvlist_t
*missing
= NULL
;
1467 verify(nvlist_lookup_string(config
, ZPOOL_CONFIG_POOL_NAME
,
1470 (void) snprintf(errbuf
, sizeof (errbuf
), dgettext(TEXT_DOMAIN
,
1471 "cannot import pool '%s'"), origname
);
1473 if (newname
!= NULL
) {
1474 if (!zpool_name_valid(hdl
, B_FALSE
, newname
))
1475 return (zfs_error_fmt(hdl
, EZFS_INVALIDNAME
,
1476 dgettext(TEXT_DOMAIN
, "cannot import '%s'"),
1478 thename
= (char *)newname
;
1485 prop_flags_t flags
= { .create
= B_FALSE
, .import
= B_TRUE
};
1487 verify(nvlist_lookup_uint64(config
, ZPOOL_CONFIG_VERSION
,
1490 if ((props
= zpool_valid_proplist(hdl
, origname
,
1491 props
, version
, flags
, errbuf
)) == NULL
) {
1493 } else if (zcmd_write_src_nvlist(hdl
, &zc
, props
) != 0) {
1499 (void) strlcpy(zc
.zc_name
, thename
, sizeof (zc
.zc_name
));
1501 verify(nvlist_lookup_uint64(config
, ZPOOL_CONFIG_POOL_GUID
,
1504 if (zcmd_write_conf_nvlist(hdl
, &zc
, config
) != 0) {
1508 if (zcmd_alloc_dst_nvlist(hdl
, &zc
, zc
.zc_nvlist_conf_size
* 2) != 0) {
1513 zc
.zc_cookie
= flags
;
1514 while ((ret
= zfs_ioctl(hdl
, ZFS_IOC_POOL_IMPORT
, &zc
)) != 0 &&
1516 if (zcmd_expand_dst_nvlist(hdl
, &zc
) != 0) {
1517 zcmd_free_nvlists(&zc
);
1524 (void) zcmd_read_dst_nvlist(hdl
, &zc
, &nv
);
1525 zpool_get_rewind_policy(config
, &policy
);
1531 * Dry-run failed, but we print out what success
1532 * looks like if we found a best txg
1534 if (policy
.zrp_request
& ZPOOL_TRY_REWIND
) {
1535 zpool_rewind_exclaim(hdl
, newname
? origname
: thename
,
1541 if (newname
== NULL
)
1542 (void) snprintf(desc
, sizeof (desc
),
1543 dgettext(TEXT_DOMAIN
, "cannot import '%s'"),
1546 (void) snprintf(desc
, sizeof (desc
),
1547 dgettext(TEXT_DOMAIN
, "cannot import '%s' as '%s'"),
1553 * Unsupported version.
1555 (void) zfs_error(hdl
, EZFS_BADVERSION
, desc
);
1559 (void) zfs_error(hdl
, EZFS_INVALCONFIG
, desc
);
1563 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
1564 "one or more devices is read only"));
1565 (void) zfs_error(hdl
, EZFS_BADDEV
, desc
);
1569 if (nv
&& nvlist_lookup_nvlist(nv
,
1570 ZPOOL_CONFIG_LOAD_INFO
, &nvinfo
) == 0 &&
1571 nvlist_lookup_nvlist(nvinfo
,
1572 ZPOOL_CONFIG_MISSING_DEVICES
, &missing
) == 0) {
1573 (void) printf(dgettext(TEXT_DOMAIN
,
1574 "The devices below are missing, use "
1575 "'-m' to import the pool anyway:\n"));
1576 print_vdev_tree(hdl
, NULL
, missing
, 2);
1577 (void) printf("\n");
1579 (void) zpool_standard_error(hdl
, error
, desc
);
1583 (void) zpool_standard_error(hdl
, error
, desc
);
1587 (void) zpool_standard_error(hdl
, error
, desc
);
1588 zpool_explain_recover(hdl
,
1589 newname
? origname
: thename
, -error
, nv
);
1596 zpool_handle_t
*zhp
;
1599 * This should never fail, but play it safe anyway.
1601 if (zpool_open_silent(hdl
, thename
, &zhp
) != 0)
1603 else if (zhp
!= NULL
)
1605 if (policy
.zrp_request
&
1606 (ZPOOL_DO_REWIND
| ZPOOL_TRY_REWIND
)) {
1607 zpool_rewind_exclaim(hdl
, newname
? origname
: thename
,
1608 ((policy
.zrp_request
& ZPOOL_TRY_REWIND
) != 0), nv
);
1614 zcmd_free_nvlists(&zc
);
1624 zpool_scan(zpool_handle_t
*zhp
, pool_scan_func_t func
)
1626 zfs_cmd_t zc
= { "\0", "\0", "\0", "\0", 0 };
1628 libzfs_handle_t
*hdl
= zhp
->zpool_hdl
;
1630 (void) strlcpy(zc
.zc_name
, zhp
->zpool_name
, sizeof (zc
.zc_name
));
1631 zc
.zc_cookie
= func
;
1633 if (zfs_ioctl(hdl
, ZFS_IOC_POOL_SCAN
, &zc
) == 0 ||
1634 (errno
== ENOENT
&& func
!= POOL_SCAN_NONE
))
1637 if (func
== POOL_SCAN_SCRUB
) {
1638 (void) snprintf(msg
, sizeof (msg
),
1639 dgettext(TEXT_DOMAIN
, "cannot scrub %s"), zc
.zc_name
);
1640 } else if (func
== POOL_SCAN_NONE
) {
1641 (void) snprintf(msg
, sizeof (msg
),
1642 dgettext(TEXT_DOMAIN
, "cannot cancel scrubbing %s"),
1645 assert(!"unexpected result");
1648 if (errno
== EBUSY
) {
1650 pool_scan_stat_t
*ps
= NULL
;
1653 verify(nvlist_lookup_nvlist(zhp
->zpool_config
,
1654 ZPOOL_CONFIG_VDEV_TREE
, &nvroot
) == 0);
1655 (void) nvlist_lookup_uint64_array(nvroot
,
1656 ZPOOL_CONFIG_SCAN_STATS
, (uint64_t **)&ps
, &psc
);
1657 if (ps
&& ps
->pss_func
== POOL_SCAN_SCRUB
)
1658 return (zfs_error(hdl
, EZFS_SCRUBBING
, msg
));
1660 return (zfs_error(hdl
, EZFS_RESILVERING
, msg
));
1661 } else if (errno
== ENOENT
) {
1662 return (zfs_error(hdl
, EZFS_NO_SCRUB
, msg
));
1664 return (zpool_standard_error(hdl
, errno
, msg
));
1669 * Find a vdev that matches the search criteria specified. We use the
1670 * the nvpair name to determine how we should look for the device.
1671 * 'avail_spare' is set to TRUE if the provided guid refers to an AVAIL
1672 * spare; but FALSE if its an INUSE spare.
1675 vdev_to_nvlist_iter(nvlist_t
*nv
, nvlist_t
*search
, boolean_t
*avail_spare
,
1676 boolean_t
*l2cache
, boolean_t
*log
)
1683 nvpair_t
*pair
= nvlist_next_nvpair(search
, NULL
);
1685 /* Nothing to look for */
1686 if (search
== NULL
|| pair
== NULL
)
1689 /* Obtain the key we will use to search */
1690 srchkey
= nvpair_name(pair
);
1692 switch (nvpair_type(pair
)) {
1693 case DATA_TYPE_UINT64
:
1694 if (strcmp(srchkey
, ZPOOL_CONFIG_GUID
) == 0) {
1695 uint64_t srchval
, theguid
;
1697 verify(nvpair_value_uint64(pair
, &srchval
) == 0);
1698 verify(nvlist_lookup_uint64(nv
, ZPOOL_CONFIG_GUID
,
1700 if (theguid
== srchval
)
1705 case DATA_TYPE_STRING
: {
1706 char *srchval
, *val
;
1708 verify(nvpair_value_string(pair
, &srchval
) == 0);
1709 if (nvlist_lookup_string(nv
, srchkey
, &val
) != 0)
1713 * Search for the requested value. Special cases:
1715 * - ZPOOL_CONFIG_PATH for whole disk entries. These end in with a
1716 * partition suffix "1", "-part1", or "p1". The suffix is hidden
1717 * from the user, but included in the string, so this matches around
1719 * - looking for a top-level vdev name (i.e. ZPOOL_CONFIG_TYPE).
1721 * Otherwise, all other searches are simple string compares.
1723 if (strcmp(srchkey
, ZPOOL_CONFIG_PATH
) == 0) {
1724 uint64_t wholedisk
= 0;
1726 (void) nvlist_lookup_uint64(nv
, ZPOOL_CONFIG_WHOLE_DISK
,
1729 char buf
[MAXPATHLEN
];
1731 zfs_append_partition(srchval
, buf
, sizeof (buf
));
1732 if (strcmp(val
, buf
) == 0)
1737 } else if (strcmp(srchkey
, ZPOOL_CONFIG_TYPE
) == 0 && val
) {
1738 char *type
, *idx
, *end
, *p
;
1739 uint64_t id
, vdev_id
;
1742 * Determine our vdev type, keeping in mind
1743 * that the srchval is composed of a type and
1744 * vdev id pair (i.e. mirror-4).
1746 if ((type
= strdup(srchval
)) == NULL
)
1749 if ((p
= strrchr(type
, '-')) == NULL
) {
1757 * If the types don't match then keep looking.
1759 if (strncmp(val
, type
, strlen(val
)) != 0) {
1764 verify(strncmp(type
, VDEV_TYPE_RAIDZ
,
1765 strlen(VDEV_TYPE_RAIDZ
)) == 0 ||
1766 strncmp(type
, VDEV_TYPE_MIRROR
,
1767 strlen(VDEV_TYPE_MIRROR
)) == 0);
1768 verify(nvlist_lookup_uint64(nv
, ZPOOL_CONFIG_ID
,
1772 vdev_id
= strtoull(idx
, &end
, 10);
1779 * Now verify that we have the correct vdev id.
1788 if (strcmp(srchval
, val
) == 0)
1797 if (nvlist_lookup_nvlist_array(nv
, ZPOOL_CONFIG_CHILDREN
,
1798 &child
, &children
) != 0)
1801 for (c
= 0; c
< children
; c
++) {
1802 if ((ret
= vdev_to_nvlist_iter(child
[c
], search
,
1803 avail_spare
, l2cache
, NULL
)) != NULL
) {
1805 * The 'is_log' value is only set for the toplevel
1806 * vdev, not the leaf vdevs. So we always lookup the
1807 * log device from the root of the vdev tree (where
1808 * 'log' is non-NULL).
1811 nvlist_lookup_uint64(child
[c
],
1812 ZPOOL_CONFIG_IS_LOG
, &is_log
) == 0 &&
1820 if (nvlist_lookup_nvlist_array(nv
, ZPOOL_CONFIG_SPARES
,
1821 &child
, &children
) == 0) {
1822 for (c
= 0; c
< children
; c
++) {
1823 if ((ret
= vdev_to_nvlist_iter(child
[c
], search
,
1824 avail_spare
, l2cache
, NULL
)) != NULL
) {
1825 *avail_spare
= B_TRUE
;
1831 if (nvlist_lookup_nvlist_array(nv
, ZPOOL_CONFIG_L2CACHE
,
1832 &child
, &children
) == 0) {
1833 for (c
= 0; c
< children
; c
++) {
1834 if ((ret
= vdev_to_nvlist_iter(child
[c
], search
,
1835 avail_spare
, l2cache
, NULL
)) != NULL
) {
1846 * Given a physical path (minus the "/devices" prefix), find the
1850 zpool_find_vdev_by_physpath(zpool_handle_t
*zhp
, const char *ppath
,
1851 boolean_t
*avail_spare
, boolean_t
*l2cache
, boolean_t
*log
)
1853 nvlist_t
*search
, *nvroot
, *ret
;
1855 verify(nvlist_alloc(&search
, NV_UNIQUE_NAME
, KM_SLEEP
) == 0);
1856 verify(nvlist_add_string(search
, ZPOOL_CONFIG_PHYS_PATH
, ppath
) == 0);
1858 verify(nvlist_lookup_nvlist(zhp
->zpool_config
, ZPOOL_CONFIG_VDEV_TREE
,
1861 *avail_spare
= B_FALSE
;
1865 ret
= vdev_to_nvlist_iter(nvroot
, search
, avail_spare
, l2cache
, log
);
1866 nvlist_free(search
);
1872 * Determine if we have an "interior" top-level vdev (i.e mirror/raidz).
1875 zpool_vdev_is_interior(const char *name
)
1877 if (strncmp(name
, VDEV_TYPE_RAIDZ
, strlen(VDEV_TYPE_RAIDZ
)) == 0 ||
1878 strncmp(name
, VDEV_TYPE_MIRROR
, strlen(VDEV_TYPE_MIRROR
)) == 0)
1884 zpool_find_vdev(zpool_handle_t
*zhp
, const char *path
, boolean_t
*avail_spare
,
1885 boolean_t
*l2cache
, boolean_t
*log
)
1887 char buf
[MAXPATHLEN
];
1889 nvlist_t
*nvroot
, *search
, *ret
;
1892 verify(nvlist_alloc(&search
, NV_UNIQUE_NAME
, KM_SLEEP
) == 0);
1894 guid
= strtoull(path
, &end
, 10);
1895 if (guid
!= 0 && *end
== '\0') {
1896 verify(nvlist_add_uint64(search
, ZPOOL_CONFIG_GUID
, guid
) == 0);
1897 } else if (zpool_vdev_is_interior(path
)) {
1898 verify(nvlist_add_string(search
, ZPOOL_CONFIG_TYPE
, path
) == 0);
1899 } else if (path
[0] != '/') {
1900 if (zfs_resolve_shortname(path
, buf
, sizeof (buf
)) < 0) {
1901 nvlist_free(search
);
1904 verify(nvlist_add_string(search
, ZPOOL_CONFIG_PATH
, buf
) == 0);
1906 verify(nvlist_add_string(search
, ZPOOL_CONFIG_PATH
, path
) == 0);
1909 verify(nvlist_lookup_nvlist(zhp
->zpool_config
, ZPOOL_CONFIG_VDEV_TREE
,
1912 *avail_spare
= B_FALSE
;
1916 ret
= vdev_to_nvlist_iter(nvroot
, search
, avail_spare
, l2cache
, log
);
1917 nvlist_free(search
);
1923 vdev_online(nvlist_t
*nv
)
1927 if (nvlist_lookup_uint64(nv
, ZPOOL_CONFIG_OFFLINE
, &ival
) == 0 ||
1928 nvlist_lookup_uint64(nv
, ZPOOL_CONFIG_FAULTED
, &ival
) == 0 ||
1929 nvlist_lookup_uint64(nv
, ZPOOL_CONFIG_REMOVED
, &ival
) == 0)
1936 * Helper function for zpool_get_physpaths().
1939 vdev_get_one_physpath(nvlist_t
*config
, char *physpath
, size_t physpath_size
,
1940 size_t *bytes_written
)
1942 size_t bytes_left
, pos
, rsz
;
1946 if (nvlist_lookup_string(config
, ZPOOL_CONFIG_PHYS_PATH
,
1948 return (EZFS_NODEVICE
);
1950 pos
= *bytes_written
;
1951 bytes_left
= physpath_size
- pos
;
1952 format
= (pos
== 0) ? "%s" : " %s";
1954 rsz
= snprintf(physpath
+ pos
, bytes_left
, format
, tmppath
);
1955 *bytes_written
+= rsz
;
1957 if (rsz
>= bytes_left
) {
1958 /* if physpath was not copied properly, clear it */
1959 if (bytes_left
!= 0) {
1962 return (EZFS_NOSPC
);
1968 vdev_get_physpaths(nvlist_t
*nv
, char *physpath
, size_t phypath_size
,
1969 size_t *rsz
, boolean_t is_spare
)
1974 if (nvlist_lookup_string(nv
, ZPOOL_CONFIG_TYPE
, &type
) != 0)
1975 return (EZFS_INVALCONFIG
);
1977 if (strcmp(type
, VDEV_TYPE_DISK
) == 0) {
1979 * An active spare device has ZPOOL_CONFIG_IS_SPARE set.
1980 * For a spare vdev, we only want to boot from the active
1985 (void) nvlist_lookup_uint64(nv
, ZPOOL_CONFIG_IS_SPARE
,
1988 return (EZFS_INVALCONFIG
);
1991 if (vdev_online(nv
)) {
1992 if ((ret
= vdev_get_one_physpath(nv
, physpath
,
1993 phypath_size
, rsz
)) != 0)
1996 } else if (strcmp(type
, VDEV_TYPE_MIRROR
) == 0 ||
1997 strcmp(type
, VDEV_TYPE_REPLACING
) == 0 ||
1998 (is_spare
= (strcmp(type
, VDEV_TYPE_SPARE
) == 0))) {
2003 if (nvlist_lookup_nvlist_array(nv
,
2004 ZPOOL_CONFIG_CHILDREN
, &child
, &count
) != 0)
2005 return (EZFS_INVALCONFIG
);
2007 for (i
= 0; i
< count
; i
++) {
2008 ret
= vdev_get_physpaths(child
[i
], physpath
,
2009 phypath_size
, rsz
, is_spare
);
2010 if (ret
== EZFS_NOSPC
)
2015 return (EZFS_POOL_INVALARG
);
2019 * Get phys_path for a root pool config.
2020 * Return 0 on success; non-zero on failure.
2023 zpool_get_config_physpath(nvlist_t
*config
, char *physpath
, size_t phypath_size
)
2026 nvlist_t
*vdev_root
;
2033 if (nvlist_lookup_nvlist(config
, ZPOOL_CONFIG_VDEV_TREE
,
2035 return (EZFS_INVALCONFIG
);
2037 if (nvlist_lookup_string(vdev_root
, ZPOOL_CONFIG_TYPE
, &type
) != 0 ||
2038 nvlist_lookup_nvlist_array(vdev_root
, ZPOOL_CONFIG_CHILDREN
,
2039 &child
, &count
) != 0)
2040 return (EZFS_INVALCONFIG
);
2043 * root pool can not have EFI labeled disks and can only have
2044 * a single top-level vdev.
2046 if (strcmp(type
, VDEV_TYPE_ROOT
) != 0 || count
!= 1 ||
2047 pool_uses_efi(vdev_root
))
2048 return (EZFS_POOL_INVALARG
);
2050 (void) vdev_get_physpaths(child
[0], physpath
, phypath_size
, &rsz
,
2053 /* No online devices */
2055 return (EZFS_NODEVICE
);
2061 * Get phys_path for a root pool
2062 * Return 0 on success; non-zero on failure.
2065 zpool_get_physpath(zpool_handle_t
*zhp
, char *physpath
, size_t phypath_size
)
2067 return (zpool_get_config_physpath(zhp
->zpool_config
, physpath
,
2072 * If the device has being dynamically expanded then we need to relabel
2073 * the disk to use the new unallocated space.
2076 zpool_relabel_disk(libzfs_handle_t
*hdl
, const char *path
)
2081 if ((fd
= open(path
, O_RDWR
|O_DIRECT
)) < 0) {
2082 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
, "cannot "
2083 "relabel '%s': unable to open device"), path
);
2084 return (zfs_error(hdl
, EZFS_OPENFAILED
, errbuf
));
2088 * It's possible that we might encounter an error if the device
2089 * does not have any unallocated space left. If so, we simply
2090 * ignore that error and continue on.
2092 error
= efi_use_whole_disk(fd
);
2094 if (error
&& error
!= VT_ENOSPC
) {
2095 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
, "cannot "
2096 "relabel '%s': unable to read disk capacity"), path
);
2097 return (zfs_error(hdl
, EZFS_NOCAP
, errbuf
));
2103 * Bring the specified vdev online. The 'flags' parameter is a set of the
2104 * ZFS_ONLINE_* flags.
2107 zpool_vdev_online(zpool_handle_t
*zhp
, const char *path
, int flags
,
2108 vdev_state_t
*newstate
)
2110 zfs_cmd_t zc
= { "\0", "\0", "\0", "\0", 0 };
2113 boolean_t avail_spare
, l2cache
, islog
;
2114 libzfs_handle_t
*hdl
= zhp
->zpool_hdl
;
2116 if (flags
& ZFS_ONLINE_EXPAND
) {
2117 (void) snprintf(msg
, sizeof (msg
),
2118 dgettext(TEXT_DOMAIN
, "cannot expand %s"), path
);
2120 (void) snprintf(msg
, sizeof (msg
),
2121 dgettext(TEXT_DOMAIN
, "cannot online %s"), path
);
2124 (void) strlcpy(zc
.zc_name
, zhp
->zpool_name
, sizeof (zc
.zc_name
));
2125 if ((tgt
= zpool_find_vdev(zhp
, path
, &avail_spare
, &l2cache
,
2127 return (zfs_error(hdl
, EZFS_NODEVICE
, msg
));
2129 verify(nvlist_lookup_uint64(tgt
, ZPOOL_CONFIG_GUID
, &zc
.zc_guid
) == 0);
2132 return (zfs_error(hdl
, EZFS_ISSPARE
, msg
));
2134 if (flags
& ZFS_ONLINE_EXPAND
||
2135 zpool_get_prop_int(zhp
, ZPOOL_PROP_AUTOEXPAND
, NULL
)) {
2136 char *pathname
= NULL
;
2137 uint64_t wholedisk
= 0;
2139 (void) nvlist_lookup_uint64(tgt
, ZPOOL_CONFIG_WHOLE_DISK
,
2141 verify(nvlist_lookup_string(tgt
, ZPOOL_CONFIG_PATH
,
2145 * XXX - L2ARC 1.0 devices can't support expansion.
2148 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
2149 "cannot expand cache devices"));
2150 return (zfs_error(hdl
, EZFS_VDEVNOTSUP
, msg
));
2154 pathname
+= strlen(DISK_ROOT
) + 1;
2155 (void) zpool_relabel_disk(hdl
, pathname
);
2159 zc
.zc_cookie
= VDEV_STATE_ONLINE
;
2162 if (zfs_ioctl(hdl
, ZFS_IOC_VDEV_SET_STATE
, &zc
) != 0) {
2163 if (errno
== EINVAL
) {
2164 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
, "was split "
2165 "from this pool into a new one. Use '%s' "
2166 "instead"), "zpool detach");
2167 return (zfs_error(hdl
, EZFS_POSTSPLIT_ONLINE
, msg
));
2169 return (zpool_standard_error(hdl
, errno
, msg
));
2172 *newstate
= zc
.zc_cookie
;
2177 * Take the specified vdev offline
2180 zpool_vdev_offline(zpool_handle_t
*zhp
, const char *path
, boolean_t istmp
)
2182 zfs_cmd_t zc
= { "\0", "\0", "\0", "\0", 0 };
2185 boolean_t avail_spare
, l2cache
;
2186 libzfs_handle_t
*hdl
= zhp
->zpool_hdl
;
2188 (void) snprintf(msg
, sizeof (msg
),
2189 dgettext(TEXT_DOMAIN
, "cannot offline %s"), path
);
2191 (void) strlcpy(zc
.zc_name
, zhp
->zpool_name
, sizeof (zc
.zc_name
));
2192 if ((tgt
= zpool_find_vdev(zhp
, path
, &avail_spare
, &l2cache
,
2194 return (zfs_error(hdl
, EZFS_NODEVICE
, msg
));
2196 verify(nvlist_lookup_uint64(tgt
, ZPOOL_CONFIG_GUID
, &zc
.zc_guid
) == 0);
2199 return (zfs_error(hdl
, EZFS_ISSPARE
, msg
));
2201 zc
.zc_cookie
= VDEV_STATE_OFFLINE
;
2202 zc
.zc_obj
= istmp
? ZFS_OFFLINE_TEMPORARY
: 0;
2204 if (zfs_ioctl(hdl
, ZFS_IOC_VDEV_SET_STATE
, &zc
) == 0)
2211 * There are no other replicas of this device.
2213 return (zfs_error(hdl
, EZFS_NOREPLICAS
, msg
));
2217 * The log device has unplayed logs
2219 return (zfs_error(hdl
, EZFS_UNPLAYED_LOGS
, msg
));
2222 return (zpool_standard_error(hdl
, errno
, msg
));
2227 * Mark the given vdev faulted.
2230 zpool_vdev_fault(zpool_handle_t
*zhp
, uint64_t guid
, vdev_aux_t aux
)
2232 zfs_cmd_t zc
= { "\0", "\0", "\0", "\0", 0 };
2234 libzfs_handle_t
*hdl
= zhp
->zpool_hdl
;
2236 (void) snprintf(msg
, sizeof (msg
),
2237 dgettext(TEXT_DOMAIN
, "cannot fault %llu"), (u_longlong_t
)guid
);
2239 (void) strlcpy(zc
.zc_name
, zhp
->zpool_name
, sizeof (zc
.zc_name
));
2241 zc
.zc_cookie
= VDEV_STATE_FAULTED
;
2244 if (ioctl(hdl
->libzfs_fd
, ZFS_IOC_VDEV_SET_STATE
, &zc
) == 0)
2251 * There are no other replicas of this device.
2253 return (zfs_error(hdl
, EZFS_NOREPLICAS
, msg
));
2256 return (zpool_standard_error(hdl
, errno
, msg
));
2262 * Mark the given vdev degraded.
2265 zpool_vdev_degrade(zpool_handle_t
*zhp
, uint64_t guid
, vdev_aux_t aux
)
2267 zfs_cmd_t zc
= { "\0", "\0", "\0", "\0", 0 };
2269 libzfs_handle_t
*hdl
= zhp
->zpool_hdl
;
2271 (void) snprintf(msg
, sizeof (msg
),
2272 dgettext(TEXT_DOMAIN
, "cannot degrade %llu"), (u_longlong_t
)guid
);
2274 (void) strlcpy(zc
.zc_name
, zhp
->zpool_name
, sizeof (zc
.zc_name
));
2276 zc
.zc_cookie
= VDEV_STATE_DEGRADED
;
2279 if (ioctl(hdl
->libzfs_fd
, ZFS_IOC_VDEV_SET_STATE
, &zc
) == 0)
2282 return (zpool_standard_error(hdl
, errno
, msg
));
2286 * Returns TRUE if the given nvlist is a vdev that was originally swapped in as
2290 is_replacing_spare(nvlist_t
*search
, nvlist_t
*tgt
, int which
)
2296 if (nvlist_lookup_nvlist_array(search
, ZPOOL_CONFIG_CHILDREN
, &child
,
2298 verify(nvlist_lookup_string(search
, ZPOOL_CONFIG_TYPE
,
2301 if (strcmp(type
, VDEV_TYPE_SPARE
) == 0 &&
2302 children
== 2 && child
[which
] == tgt
)
2305 for (c
= 0; c
< children
; c
++)
2306 if (is_replacing_spare(child
[c
], tgt
, which
))
2314 * Attach new_disk (fully described by nvroot) to old_disk.
2315 * If 'replacing' is specified, the new disk will replace the old one.
2318 zpool_vdev_attach(zpool_handle_t
*zhp
,
2319 const char *old_disk
, const char *new_disk
, nvlist_t
*nvroot
, int replacing
)
2321 zfs_cmd_t zc
= { "\0", "\0", "\0", "\0", 0 };
2325 boolean_t avail_spare
, l2cache
, islog
;
2330 nvlist_t
*config_root
;
2331 libzfs_handle_t
*hdl
= zhp
->zpool_hdl
;
2332 boolean_t rootpool
= pool_is_bootable(zhp
);
2335 (void) snprintf(msg
, sizeof (msg
), dgettext(TEXT_DOMAIN
,
2336 "cannot replace %s with %s"), old_disk
, new_disk
);
2338 (void) snprintf(msg
, sizeof (msg
), dgettext(TEXT_DOMAIN
,
2339 "cannot attach %s to %s"), new_disk
, old_disk
);
2342 * If this is a root pool, make sure that we're not attaching an
2343 * EFI labeled device.
2345 if (rootpool
&& pool_uses_efi(nvroot
)) {
2346 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
2347 "EFI labeled devices are not supported on root pools."));
2348 return (zfs_error(hdl
, EZFS_POOL_NOTSUP
, msg
));
2351 (void) strlcpy(zc
.zc_name
, zhp
->zpool_name
, sizeof (zc
.zc_name
));
2352 if ((tgt
= zpool_find_vdev(zhp
, old_disk
, &avail_spare
, &l2cache
,
2354 return (zfs_error(hdl
, EZFS_NODEVICE
, msg
));
2357 return (zfs_error(hdl
, EZFS_ISSPARE
, msg
));
2360 return (zfs_error(hdl
, EZFS_ISL2CACHE
, msg
));
2362 verify(nvlist_lookup_uint64(tgt
, ZPOOL_CONFIG_GUID
, &zc
.zc_guid
) == 0);
2363 zc
.zc_cookie
= replacing
;
2365 if (nvlist_lookup_nvlist_array(nvroot
, ZPOOL_CONFIG_CHILDREN
,
2366 &child
, &children
) != 0 || children
!= 1) {
2367 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
2368 "new device must be a single disk"));
2369 return (zfs_error(hdl
, EZFS_INVALCONFIG
, msg
));
2372 verify(nvlist_lookup_nvlist(zpool_get_config(zhp
, NULL
),
2373 ZPOOL_CONFIG_VDEV_TREE
, &config_root
) == 0);
2375 if ((newname
= zpool_vdev_name(NULL
, NULL
, child
[0], B_FALSE
)) == NULL
)
2379 * If the target is a hot spare that has been swapped in, we can only
2380 * replace it with another hot spare.
2383 nvlist_lookup_uint64(tgt
, ZPOOL_CONFIG_IS_SPARE
, &val
) == 0 &&
2384 (zpool_find_vdev(zhp
, newname
, &avail_spare
, &l2cache
,
2385 NULL
) == NULL
|| !avail_spare
) &&
2386 is_replacing_spare(config_root
, tgt
, 1)) {
2387 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
2388 "can only be replaced by another hot spare"));
2390 return (zfs_error(hdl
, EZFS_BADTARGET
, msg
));
2395 if (zcmd_write_conf_nvlist(hdl
, &zc
, nvroot
) != 0)
2398 ret
= zfs_ioctl(hdl
, ZFS_IOC_VDEV_ATTACH
, &zc
);
2400 zcmd_free_nvlists(&zc
);
2405 * XXX need a better way to prevent user from
2406 * booting up a half-baked vdev.
2408 (void) fprintf(stderr
, dgettext(TEXT_DOMAIN
, "Make "
2409 "sure to wait until resilver is done "
2410 "before rebooting.\n"));
2418 * Can't attach to or replace this type of vdev.
2421 uint64_t version
= zpool_get_prop_int(zhp
,
2422 ZPOOL_PROP_VERSION
, NULL
);
2425 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
2426 "cannot replace a log with a spare"));
2427 else if (version
>= SPA_VERSION_MULTI_REPLACE
)
2428 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
2429 "already in replacing/spare config; wait "
2430 "for completion or use 'zpool detach'"));
2432 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
2433 "cannot replace a replacing device"));
2435 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
2436 "can only attach to mirrors and top-level "
2439 (void) zfs_error(hdl
, EZFS_BADTARGET
, msg
);
2444 * The new device must be a single disk.
2446 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
2447 "new device must be a single disk"));
2448 (void) zfs_error(hdl
, EZFS_INVALCONFIG
, msg
);
2452 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
, "%s is busy"),
2454 (void) zfs_error(hdl
, EZFS_BADDEV
, msg
);
2459 * The new device is too small.
2461 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
2462 "device is too small"));
2463 (void) zfs_error(hdl
, EZFS_BADDEV
, msg
);
2468 * The new device has a different alignment requirement.
2470 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
2471 "devices have different sector alignment"));
2472 (void) zfs_error(hdl
, EZFS_BADDEV
, msg
);
2477 * The resulting top-level vdev spec won't fit in the label.
2479 (void) zfs_error(hdl
, EZFS_DEVOVERFLOW
, msg
);
2483 (void) zpool_standard_error(hdl
, errno
, msg
);
2490 * Detach the specified device.
2493 zpool_vdev_detach(zpool_handle_t
*zhp
, const char *path
)
2495 zfs_cmd_t zc
= { "\0", "\0", "\0", "\0", 0 };
2498 boolean_t avail_spare
, l2cache
;
2499 libzfs_handle_t
*hdl
= zhp
->zpool_hdl
;
2501 (void) snprintf(msg
, sizeof (msg
),
2502 dgettext(TEXT_DOMAIN
, "cannot detach %s"), path
);
2504 (void) strlcpy(zc
.zc_name
, zhp
->zpool_name
, sizeof (zc
.zc_name
));
2505 if ((tgt
= zpool_find_vdev(zhp
, path
, &avail_spare
, &l2cache
,
2507 return (zfs_error(hdl
, EZFS_NODEVICE
, msg
));
2510 return (zfs_error(hdl
, EZFS_ISSPARE
, msg
));
2513 return (zfs_error(hdl
, EZFS_ISL2CACHE
, msg
));
2515 verify(nvlist_lookup_uint64(tgt
, ZPOOL_CONFIG_GUID
, &zc
.zc_guid
) == 0);
2517 if (zfs_ioctl(hdl
, ZFS_IOC_VDEV_DETACH
, &zc
) == 0)
2524 * Can't detach from this type of vdev.
2526 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
, "only "
2527 "applicable to mirror and replacing vdevs"));
2528 (void) zfs_error(hdl
, EZFS_BADTARGET
, msg
);
2533 * There are no other replicas of this device.
2535 (void) zfs_error(hdl
, EZFS_NOREPLICAS
, msg
);
2539 (void) zpool_standard_error(hdl
, errno
, msg
);
2546 * Find a mirror vdev in the source nvlist.
2548 * The mchild array contains a list of disks in one of the top-level mirrors
2549 * of the source pool. The schild array contains a list of disks that the
2550 * user specified on the command line. We loop over the mchild array to
2551 * see if any entry in the schild array matches.
2553 * If a disk in the mchild array is found in the schild array, we return
2554 * the index of that entry. Otherwise we return -1.
2557 find_vdev_entry(zpool_handle_t
*zhp
, nvlist_t
**mchild
, uint_t mchildren
,
2558 nvlist_t
**schild
, uint_t schildren
)
2562 for (mc
= 0; mc
< mchildren
; mc
++) {
2564 char *mpath
= zpool_vdev_name(zhp
->zpool_hdl
, zhp
,
2565 mchild
[mc
], B_FALSE
);
2567 for (sc
= 0; sc
< schildren
; sc
++) {
2568 char *spath
= zpool_vdev_name(zhp
->zpool_hdl
, zhp
,
2569 schild
[sc
], B_FALSE
);
2570 boolean_t result
= (strcmp(mpath
, spath
) == 0);
2586 * Split a mirror pool. If newroot points to null, then a new nvlist
2587 * is generated and it is the responsibility of the caller to free it.
2590 zpool_vdev_split(zpool_handle_t
*zhp
, char *newname
, nvlist_t
**newroot
,
2591 nvlist_t
*props
, splitflags_t flags
)
2593 zfs_cmd_t zc
= { "\0", "\0", "\0", "\0", 0 };
2595 nvlist_t
*tree
, *config
, **child
, **newchild
, *newconfig
= NULL
;
2596 nvlist_t
**varray
= NULL
, *zc_props
= NULL
;
2597 uint_t c
, children
, newchildren
, lastlog
= 0, vcount
, found
= 0;
2598 libzfs_handle_t
*hdl
= zhp
->zpool_hdl
;
2600 boolean_t freelist
= B_FALSE
, memory_err
= B_TRUE
;
2603 (void) snprintf(msg
, sizeof (msg
),
2604 dgettext(TEXT_DOMAIN
, "Unable to split %s"), zhp
->zpool_name
);
2606 if (!zpool_name_valid(hdl
, B_FALSE
, newname
))
2607 return (zfs_error(hdl
, EZFS_INVALIDNAME
, msg
));
2609 if ((config
= zpool_get_config(zhp
, NULL
)) == NULL
) {
2610 (void) fprintf(stderr
, gettext("Internal error: unable to "
2611 "retrieve pool configuration\n"));
2615 verify(nvlist_lookup_nvlist(config
, ZPOOL_CONFIG_VDEV_TREE
, &tree
)
2617 verify(nvlist_lookup_uint64(config
, ZPOOL_CONFIG_VERSION
, &vers
) == 0);
2620 prop_flags_t flags
= { .create
= B_FALSE
, .import
= B_TRUE
};
2621 if ((zc_props
= zpool_valid_proplist(hdl
, zhp
->zpool_name
,
2622 props
, vers
, flags
, msg
)) == NULL
)
2626 if (nvlist_lookup_nvlist_array(tree
, ZPOOL_CONFIG_CHILDREN
, &child
,
2628 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
2629 "Source pool is missing vdev tree"));
2631 nvlist_free(zc_props
);
2635 varray
= zfs_alloc(hdl
, children
* sizeof (nvlist_t
*));
2638 if (*newroot
== NULL
||
2639 nvlist_lookup_nvlist_array(*newroot
, ZPOOL_CONFIG_CHILDREN
,
2640 &newchild
, &newchildren
) != 0)
2643 for (c
= 0; c
< children
; c
++) {
2644 uint64_t is_log
= B_FALSE
, is_hole
= B_FALSE
;
2646 nvlist_t
**mchild
, *vdev
;
2651 * Unlike cache & spares, slogs are stored in the
2652 * ZPOOL_CONFIG_CHILDREN array. We filter them out here.
2654 (void) nvlist_lookup_uint64(child
[c
], ZPOOL_CONFIG_IS_LOG
,
2656 (void) nvlist_lookup_uint64(child
[c
], ZPOOL_CONFIG_IS_HOLE
,
2658 if (is_log
|| is_hole
) {
2660 * Create a hole vdev and put it in the config.
2662 if (nvlist_alloc(&vdev
, NV_UNIQUE_NAME
, 0) != 0)
2664 if (nvlist_add_string(vdev
, ZPOOL_CONFIG_TYPE
,
2665 VDEV_TYPE_HOLE
) != 0)
2667 if (nvlist_add_uint64(vdev
, ZPOOL_CONFIG_IS_HOLE
,
2672 varray
[vcount
++] = vdev
;
2676 verify(nvlist_lookup_string(child
[c
], ZPOOL_CONFIG_TYPE
, &type
)
2678 if (strcmp(type
, VDEV_TYPE_MIRROR
) != 0) {
2679 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
2680 "Source pool must be composed only of mirrors\n"));
2681 retval
= zfs_error(hdl
, EZFS_INVALCONFIG
, msg
);
2685 verify(nvlist_lookup_nvlist_array(child
[c
],
2686 ZPOOL_CONFIG_CHILDREN
, &mchild
, &mchildren
) == 0);
2688 /* find or add an entry for this top-level vdev */
2689 if (newchildren
> 0 &&
2690 (entry
= find_vdev_entry(zhp
, mchild
, mchildren
,
2691 newchild
, newchildren
)) >= 0) {
2692 /* We found a disk that the user specified. */
2693 vdev
= mchild
[entry
];
2696 /* User didn't specify a disk for this vdev. */
2697 vdev
= mchild
[mchildren
- 1];
2700 if (nvlist_dup(vdev
, &varray
[vcount
++], 0) != 0)
2704 /* did we find every disk the user specified? */
2705 if (found
!= newchildren
) {
2706 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
, "Device list must "
2707 "include at most one disk from each mirror"));
2708 retval
= zfs_error(hdl
, EZFS_INVALCONFIG
, msg
);
2712 /* Prepare the nvlist for populating. */
2713 if (*newroot
== NULL
) {
2714 if (nvlist_alloc(newroot
, NV_UNIQUE_NAME
, 0) != 0)
2717 if (nvlist_add_string(*newroot
, ZPOOL_CONFIG_TYPE
,
2718 VDEV_TYPE_ROOT
) != 0)
2721 verify(nvlist_remove_all(*newroot
, ZPOOL_CONFIG_CHILDREN
) == 0);
2724 /* Add all the children we found */
2725 if (nvlist_add_nvlist_array(*newroot
, ZPOOL_CONFIG_CHILDREN
, varray
,
2726 lastlog
== 0 ? vcount
: lastlog
) != 0)
2730 * If we're just doing a dry run, exit now with success.
2733 memory_err
= B_FALSE
;
2738 /* now build up the config list & call the ioctl */
2739 if (nvlist_alloc(&newconfig
, NV_UNIQUE_NAME
, 0) != 0)
2742 if (nvlist_add_nvlist(newconfig
,
2743 ZPOOL_CONFIG_VDEV_TREE
, *newroot
) != 0 ||
2744 nvlist_add_string(newconfig
,
2745 ZPOOL_CONFIG_POOL_NAME
, newname
) != 0 ||
2746 nvlist_add_uint64(newconfig
, ZPOOL_CONFIG_VERSION
, vers
) != 0)
2750 * The new pool is automatically part of the namespace unless we
2751 * explicitly export it.
2754 zc
.zc_cookie
= ZPOOL_EXPORT_AFTER_SPLIT
;
2755 (void) strlcpy(zc
.zc_name
, zhp
->zpool_name
, sizeof (zc
.zc_name
));
2756 (void) strlcpy(zc
.zc_string
, newname
, sizeof (zc
.zc_string
));
2757 if (zcmd_write_conf_nvlist(hdl
, &zc
, newconfig
) != 0)
2759 if (zc_props
!= NULL
&& zcmd_write_src_nvlist(hdl
, &zc
, zc_props
) != 0)
2762 if (zfs_ioctl(hdl
, ZFS_IOC_VDEV_SPLIT
, &zc
) != 0) {
2763 retval
= zpool_standard_error(hdl
, errno
, msg
);
2768 memory_err
= B_FALSE
;
2771 if (varray
!= NULL
) {
2774 for (v
= 0; v
< vcount
; v
++)
2775 nvlist_free(varray
[v
]);
2778 zcmd_free_nvlists(&zc
);
2780 nvlist_free(zc_props
);
2782 nvlist_free(newconfig
);
2784 nvlist_free(*newroot
);
2792 return (no_memory(hdl
));
2798 * Remove the given device. Currently, this is supported only for hot spares
2799 * and level 2 cache devices.
2802 zpool_vdev_remove(zpool_handle_t
*zhp
, const char *path
)
2804 zfs_cmd_t zc
= { "\0", "\0", "\0", "\0", 0 };
2807 boolean_t avail_spare
, l2cache
, islog
;
2808 libzfs_handle_t
*hdl
= zhp
->zpool_hdl
;
2811 (void) snprintf(msg
, sizeof (msg
),
2812 dgettext(TEXT_DOMAIN
, "cannot remove %s"), path
);
2814 (void) strlcpy(zc
.zc_name
, zhp
->zpool_name
, sizeof (zc
.zc_name
));
2815 if ((tgt
= zpool_find_vdev(zhp
, path
, &avail_spare
, &l2cache
,
2817 return (zfs_error(hdl
, EZFS_NODEVICE
, msg
));
2819 * XXX - this should just go away.
2821 if (!avail_spare
&& !l2cache
&& !islog
) {
2822 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
2823 "only inactive hot spares, cache, top-level, "
2824 "or log devices can be removed"));
2825 return (zfs_error(hdl
, EZFS_NODEVICE
, msg
));
2828 version
= zpool_get_prop_int(zhp
, ZPOOL_PROP_VERSION
, NULL
);
2829 if (islog
&& version
< SPA_VERSION_HOLES
) {
2830 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
2831 "pool must be upgrade to support log removal"));
2832 return (zfs_error(hdl
, EZFS_BADVERSION
, msg
));
2835 verify(nvlist_lookup_uint64(tgt
, ZPOOL_CONFIG_GUID
, &zc
.zc_guid
) == 0);
2837 if (zfs_ioctl(hdl
, ZFS_IOC_VDEV_REMOVE
, &zc
) == 0)
2840 return (zpool_standard_error(hdl
, errno
, msg
));
2844 * Clear the errors for the pool, or the particular device if specified.
2847 zpool_clear(zpool_handle_t
*zhp
, const char *path
, nvlist_t
*rewindnvl
)
2849 zfs_cmd_t zc
= { "\0", "\0", "\0", "\0", 0 };
2852 zpool_rewind_policy_t policy
;
2853 boolean_t avail_spare
, l2cache
;
2854 libzfs_handle_t
*hdl
= zhp
->zpool_hdl
;
2855 nvlist_t
*nvi
= NULL
;
2859 (void) snprintf(msg
, sizeof (msg
),
2860 dgettext(TEXT_DOMAIN
, "cannot clear errors for %s"),
2863 (void) snprintf(msg
, sizeof (msg
),
2864 dgettext(TEXT_DOMAIN
, "cannot clear errors for %s"),
2867 (void) strlcpy(zc
.zc_name
, zhp
->zpool_name
, sizeof (zc
.zc_name
));
2869 if ((tgt
= zpool_find_vdev(zhp
, path
, &avail_spare
,
2870 &l2cache
, NULL
)) == 0)
2871 return (zfs_error(hdl
, EZFS_NODEVICE
, msg
));
2874 * Don't allow error clearing for hot spares. Do allow
2875 * error clearing for l2cache devices.
2878 return (zfs_error(hdl
, EZFS_ISSPARE
, msg
));
2880 verify(nvlist_lookup_uint64(tgt
, ZPOOL_CONFIG_GUID
,
2884 zpool_get_rewind_policy(rewindnvl
, &policy
);
2885 zc
.zc_cookie
= policy
.zrp_request
;
2887 if (zcmd_alloc_dst_nvlist(hdl
, &zc
, zhp
->zpool_config_size
* 2) != 0)
2890 if (zcmd_write_src_nvlist(hdl
, &zc
, rewindnvl
) != 0)
2893 while ((error
= zfs_ioctl(hdl
, ZFS_IOC_CLEAR
, &zc
)) != 0 &&
2895 if (zcmd_expand_dst_nvlist(hdl
, &zc
) != 0) {
2896 zcmd_free_nvlists(&zc
);
2901 if (!error
|| ((policy
.zrp_request
& ZPOOL_TRY_REWIND
) &&
2902 errno
!= EPERM
&& errno
!= EACCES
)) {
2903 if (policy
.zrp_request
&
2904 (ZPOOL_DO_REWIND
| ZPOOL_TRY_REWIND
)) {
2905 (void) zcmd_read_dst_nvlist(hdl
, &zc
, &nvi
);
2906 zpool_rewind_exclaim(hdl
, zc
.zc_name
,
2907 ((policy
.zrp_request
& ZPOOL_TRY_REWIND
) != 0),
2911 zcmd_free_nvlists(&zc
);
2915 zcmd_free_nvlists(&zc
);
2916 return (zpool_standard_error(hdl
, errno
, msg
));
2920 * Similar to zpool_clear(), but takes a GUID (used by fmd).
2923 zpool_vdev_clear(zpool_handle_t
*zhp
, uint64_t guid
)
2925 zfs_cmd_t zc
= { "\0", "\0", "\0", "\0", 0 };
2927 libzfs_handle_t
*hdl
= zhp
->zpool_hdl
;
2929 (void) snprintf(msg
, sizeof (msg
),
2930 dgettext(TEXT_DOMAIN
, "cannot clear errors for %llx"),
2931 (u_longlong_t
)guid
);
2933 (void) strlcpy(zc
.zc_name
, zhp
->zpool_name
, sizeof (zc
.zc_name
));
2935 zc
.zc_cookie
= ZPOOL_NO_REWIND
;
2937 if (ioctl(hdl
->libzfs_fd
, ZFS_IOC_CLEAR
, &zc
) == 0)
2940 return (zpool_standard_error(hdl
, errno
, msg
));
2944 * Convert from a devid string to a path.
2947 devid_to_path(char *devid_str
)
2952 devid_nmlist_t
*list
= NULL
;
2955 if (devid_str_decode(devid_str
, &devid
, &minor
) != 0)
2958 ret
= devid_deviceid_to_nmlist("/dev", devid
, minor
, &list
);
2960 devid_str_free(minor
);
2966 if ((path
= strdup(list
[0].devname
)) == NULL
)
2969 devid_free_nmlist(list
);
2975 * Convert from a path to a devid string.
2978 path_to_devid(const char *path
)
2984 if ((fd
= open(path
, O_RDONLY
)) < 0)
2989 if (devid_get(fd
, &devid
) == 0) {
2990 if (devid_get_minor_name(fd
, &minor
) == 0)
2991 ret
= devid_str_encode(devid
, minor
);
2993 devid_str_free(minor
);
3002 * Issue the necessary ioctl() to update the stored path value for the vdev. We
3003 * ignore any failure here, since a common case is for an unprivileged user to
3004 * type 'zpool status', and we'll display the correct information anyway.
3007 set_path(zpool_handle_t
*zhp
, nvlist_t
*nv
, const char *path
)
3009 zfs_cmd_t zc
= { "\0", "\0", "\0", "\0", 0 };
3011 (void) strncpy(zc
.zc_name
, zhp
->zpool_name
, sizeof (zc
.zc_name
));
3012 (void) strncpy(zc
.zc_value
, path
, sizeof (zc
.zc_value
));
3013 verify(nvlist_lookup_uint64(nv
, ZPOOL_CONFIG_GUID
,
3016 (void) ioctl(zhp
->zpool_hdl
->libzfs_fd
, ZFS_IOC_VDEV_SETPATH
, &zc
);
3020 * Remove partition suffix from a vdev path. Partition suffixes may take three
3021 * forms: "-partX", "pX", or "X", where X is a string of digits. The second
3022 * case only occurs when the suffix is preceded by a digit, i.e. "md0p0" The
3023 * third case only occurs when preceded by a string matching the regular
3024 * expression "^[hs]d[a-z]+", i.e. a scsi or ide disk.
3027 strip_partition(libzfs_handle_t
*hdl
, char *path
)
3029 char *tmp
= zfs_strdup(hdl
, path
);
3030 char *part
= NULL
, *d
= NULL
;
3032 if ((part
= strstr(tmp
, "-part")) && part
!= tmp
) {
3034 } else if ((part
= strrchr(tmp
, 'p')) &&
3035 part
> tmp
+ 1 && isdigit(*(part
-1))) {
3037 } else if ((tmp
[0] == 'h' || tmp
[0] == 's') && tmp
[1] == 'd') {
3038 for (d
= &tmp
[2]; isalpha(*d
); part
= ++d
);
3040 if (part
&& d
&& *d
!= '\0') {
3041 for (; isdigit(*d
); d
++);
3048 #define PATH_BUF_LEN 64
3051 * Given a vdev, return the name to display in iostat. If the vdev has a path,
3052 * we use that, stripping off any leading "/dev/dsk/"; if not, we use the type.
3053 * We also check if this is a whole disk, in which case we strip off the
3054 * trailing 's0' slice name.
3056 * This routine is also responsible for identifying when disks have been
3057 * reconfigured in a new location. The kernel will have opened the device by
3058 * devid, but the path will still refer to the old location. To catch this, we
3059 * first do a path -> devid translation (which is fast for the common case). If
3060 * the devid matches, we're done. If not, we do a reverse devid -> path
3061 * translation and issue the appropriate ioctl() to update the path of the vdev.
3062 * If 'zhp' is NULL, then this is an exported pool, and we don't need to do any
3066 zpool_vdev_name(libzfs_handle_t
*hdl
, zpool_handle_t
*zhp
, nvlist_t
*nv
,
3069 char *path
, *devid
, *type
;
3071 char buf
[PATH_BUF_LEN
];
3075 if (nvlist_lookup_uint64(nv
, ZPOOL_CONFIG_NOT_PRESENT
,
3077 verify(nvlist_lookup_uint64(nv
, ZPOOL_CONFIG_GUID
,
3079 (void) snprintf(buf
, sizeof (buf
), "%llu",
3080 (u_longlong_t
)value
);
3082 } else if (nvlist_lookup_string(nv
, ZPOOL_CONFIG_PATH
, &path
) == 0) {
3084 * If the device is dead (faulted, offline, etc) then don't
3085 * bother opening it. Otherwise we may be forcing the user to
3086 * open a misbehaving device, which can have undesirable
3089 if ((nvlist_lookup_uint64_array(nv
, ZPOOL_CONFIG_VDEV_STATS
,
3090 (uint64_t **)&vs
, &vsc
) != 0 ||
3091 vs
->vs_state
>= VDEV_STATE_DEGRADED
) &&
3093 nvlist_lookup_string(nv
, ZPOOL_CONFIG_DEVID
, &devid
) == 0) {
3095 * Determine if the current path is correct.
3097 char *newdevid
= path_to_devid(path
);
3099 if (newdevid
== NULL
||
3100 strcmp(devid
, newdevid
) != 0) {
3103 if ((newpath
= devid_to_path(devid
)) != NULL
) {
3105 * Update the path appropriately.
3107 set_path(zhp
, nv
, newpath
);
3108 if (nvlist_add_string(nv
,
3109 ZPOOL_CONFIG_PATH
, newpath
) == 0)
3110 verify(nvlist_lookup_string(nv
,
3118 devid_str_free(newdevid
);
3122 * For a block device only use the name.
3124 verify(nvlist_lookup_string(nv
, ZPOOL_CONFIG_TYPE
, &type
) == 0);
3125 if (strcmp(type
, VDEV_TYPE_DISK
) == 0) {
3126 path
= strrchr(path
, '/');
3131 * Remove the partition from the path it this is a whole disk.
3133 if (nvlist_lookup_uint64(nv
, ZPOOL_CONFIG_WHOLE_DISK
,
3134 &value
) == 0 && value
) {
3135 return strip_partition(hdl
, path
);
3138 verify(nvlist_lookup_string(nv
, ZPOOL_CONFIG_TYPE
, &path
) == 0);
3141 * If it's a raidz device, we need to stick in the parity level.
3143 if (strcmp(path
, VDEV_TYPE_RAIDZ
) == 0) {
3144 char tmpbuf
[PATH_BUF_LEN
];
3146 verify(nvlist_lookup_uint64(nv
, ZPOOL_CONFIG_NPARITY
,
3148 (void) snprintf(tmpbuf
, sizeof (tmpbuf
), "%s%llu", path
,
3149 (u_longlong_t
)value
);
3154 * We identify each top-level vdev by using a <type-id>
3155 * naming convention.
3160 verify(nvlist_lookup_uint64(nv
, ZPOOL_CONFIG_ID
,
3162 (void) snprintf(buf
, sizeof (buf
), "%s-%llu", path
,
3168 return (zfs_strdup(hdl
, path
));
3172 zbookmark_compare(const void *a
, const void *b
)
3174 return (memcmp(a
, b
, sizeof (zbookmark_t
)));
3178 * Retrieve the persistent error log, uniquify the members, and return to the
3182 zpool_get_errlog(zpool_handle_t
*zhp
, nvlist_t
**nverrlistp
)
3184 zfs_cmd_t zc
= { "\0", "\0", "\0", "\0", 0 };
3186 zbookmark_t
*zb
= NULL
;
3190 * Retrieve the raw error list from the kernel. If the number of errors
3191 * has increased, allocate more space and continue until we get the
3194 verify(nvlist_lookup_uint64(zhp
->zpool_config
, ZPOOL_CONFIG_ERRCOUNT
,
3198 if ((zc
.zc_nvlist_dst
= (uintptr_t)zfs_alloc(zhp
->zpool_hdl
,
3199 count
* sizeof (zbookmark_t
))) == (uintptr_t)NULL
)
3201 zc
.zc_nvlist_dst_size
= count
;
3202 (void) strcpy(zc
.zc_name
, zhp
->zpool_name
);
3204 if (ioctl(zhp
->zpool_hdl
->libzfs_fd
, ZFS_IOC_ERROR_LOG
,
3206 free((void *)(uintptr_t)zc
.zc_nvlist_dst
);
3207 if (errno
== ENOMEM
) {
3208 count
= zc
.zc_nvlist_dst_size
;
3209 if ((zc
.zc_nvlist_dst
= (uintptr_t)
3210 zfs_alloc(zhp
->zpool_hdl
, count
*
3211 sizeof (zbookmark_t
))) == (uintptr_t)NULL
)
3222 * Sort the resulting bookmarks. This is a little confusing due to the
3223 * implementation of ZFS_IOC_ERROR_LOG. The bookmarks are copied last
3224 * to first, and 'zc_nvlist_dst_size' indicates the number of boomarks
3225 * _not_ copied as part of the process. So we point the start of our
3226 * array appropriate and decrement the total number of elements.
3228 zb
= ((zbookmark_t
*)(uintptr_t)zc
.zc_nvlist_dst
) +
3229 zc
.zc_nvlist_dst_size
;
3230 count
-= zc
.zc_nvlist_dst_size
;
3232 qsort(zb
, count
, sizeof (zbookmark_t
), zbookmark_compare
);
3234 verify(nvlist_alloc(nverrlistp
, 0, KM_SLEEP
) == 0);
3237 * Fill in the nverrlistp with nvlist's of dataset and object numbers.
3239 for (i
= 0; i
< count
; i
++) {
3242 /* ignoring zb_blkid and zb_level for now */
3243 if (i
> 0 && zb
[i
-1].zb_objset
== zb
[i
].zb_objset
&&
3244 zb
[i
-1].zb_object
== zb
[i
].zb_object
)
3247 if (nvlist_alloc(&nv
, NV_UNIQUE_NAME
, KM_SLEEP
) != 0)
3249 if (nvlist_add_uint64(nv
, ZPOOL_ERR_DATASET
,
3250 zb
[i
].zb_objset
) != 0) {
3254 if (nvlist_add_uint64(nv
, ZPOOL_ERR_OBJECT
,
3255 zb
[i
].zb_object
) != 0) {
3259 if (nvlist_add_nvlist(*nverrlistp
, "ejk", nv
) != 0) {
3266 free((void *)(uintptr_t)zc
.zc_nvlist_dst
);
3270 free((void *)(uintptr_t)zc
.zc_nvlist_dst
);
3271 return (no_memory(zhp
->zpool_hdl
));
3275 * Upgrade a ZFS pool to the latest on-disk version.
3278 zpool_upgrade(zpool_handle_t
*zhp
, uint64_t new_version
)
3280 zfs_cmd_t zc
= { "\0", "\0", "\0", "\0", 0 };
3281 libzfs_handle_t
*hdl
= zhp
->zpool_hdl
;
3283 (void) strcpy(zc
.zc_name
, zhp
->zpool_name
);
3284 zc
.zc_cookie
= new_version
;
3286 if (zfs_ioctl(hdl
, ZFS_IOC_POOL_UPGRADE
, &zc
) != 0)
3287 return (zpool_standard_error_fmt(hdl
, errno
,
3288 dgettext(TEXT_DOMAIN
, "cannot upgrade '%s'"),
3294 zpool_set_history_str(const char *subcommand
, int argc
, char **argv
,
3299 (void) strlcpy(history_str
, subcommand
, HIS_MAX_RECORD_LEN
);
3300 for (i
= 1; i
< argc
; i
++) {
3301 if (strlen(history_str
) + 1 + strlen(argv
[i
]) >
3304 (void) strlcat(history_str
, " ", HIS_MAX_RECORD_LEN
);
3305 (void) strlcat(history_str
, argv
[i
], HIS_MAX_RECORD_LEN
);
3310 * Stage command history for logging.
3313 zpool_stage_history(libzfs_handle_t
*hdl
, const char *history_str
)
3315 if (history_str
== NULL
)
3318 if (strlen(history_str
) > HIS_MAX_RECORD_LEN
)
3321 if (hdl
->libzfs_log_str
!= NULL
)
3322 free(hdl
->libzfs_log_str
);
3324 if ((hdl
->libzfs_log_str
= strdup(history_str
)) == NULL
)
3325 return (no_memory(hdl
));
3331 * Perform ioctl to get some command history of a pool.
3333 * 'buf' is the buffer to fill up to 'len' bytes. 'off' is the
3334 * logical offset of the history buffer to start reading from.
3336 * Upon return, 'off' is the next logical offset to read from and
3337 * 'len' is the actual amount of bytes read into 'buf'.
3340 get_history(zpool_handle_t
*zhp
, char *buf
, uint64_t *off
, uint64_t *len
)
3342 zfs_cmd_t zc
= { "\0", "\0", "\0", "\0", 0 };
3343 libzfs_handle_t
*hdl
= zhp
->zpool_hdl
;
3345 (void) strlcpy(zc
.zc_name
, zhp
->zpool_name
, sizeof (zc
.zc_name
));
3347 zc
.zc_history
= (uint64_t)(uintptr_t)buf
;
3348 zc
.zc_history_len
= *len
;
3349 zc
.zc_history_offset
= *off
;
3351 if (ioctl(hdl
->libzfs_fd
, ZFS_IOC_POOL_GET_HISTORY
, &zc
) != 0) {
3354 return (zfs_error_fmt(hdl
, EZFS_PERM
,
3355 dgettext(TEXT_DOMAIN
,
3356 "cannot show history for pool '%s'"),
3359 return (zfs_error_fmt(hdl
, EZFS_NOHISTORY
,
3360 dgettext(TEXT_DOMAIN
, "cannot get history for pool "
3361 "'%s'"), zhp
->zpool_name
));
3363 return (zfs_error_fmt(hdl
, EZFS_BADVERSION
,
3364 dgettext(TEXT_DOMAIN
, "cannot get history for pool "
3365 "'%s', pool must be upgraded"), zhp
->zpool_name
));
3367 return (zpool_standard_error_fmt(hdl
, errno
,
3368 dgettext(TEXT_DOMAIN
,
3369 "cannot get history for '%s'"), zhp
->zpool_name
));
3373 *len
= zc
.zc_history_len
;
3374 *off
= zc
.zc_history_offset
;
3380 * Process the buffer of nvlists, unpacking and storing each nvlist record
3381 * into 'records'. 'leftover' is set to the number of bytes that weren't
3382 * processed as there wasn't a complete record.
3385 zpool_history_unpack(char *buf
, uint64_t bytes_read
, uint64_t *leftover
,
3386 nvlist_t
***records
, uint_t
*numrecords
)
3392 while (bytes_read
> sizeof (reclen
)) {
3394 /* get length of packed record (stored as little endian) */
3395 for (i
= 0, reclen
= 0; i
< sizeof (reclen
); i
++)
3396 reclen
+= (uint64_t)(((uchar_t
*)buf
)[i
]) << (8*i
);
3398 if (bytes_read
< sizeof (reclen
) + reclen
)
3402 if (nvlist_unpack(buf
+ sizeof (reclen
), reclen
, &nv
, 0) != 0)
3404 bytes_read
-= sizeof (reclen
) + reclen
;
3405 buf
+= sizeof (reclen
) + reclen
;
3407 /* add record to nvlist array */
3409 if (ISP2(*numrecords
+ 1)) {
3410 *records
= realloc(*records
,
3411 *numrecords
* 2 * sizeof (nvlist_t
*));
3413 (*records
)[*numrecords
- 1] = nv
;
3416 *leftover
= bytes_read
;
3420 #define HIS_BUF_LEN (128*1024)
3423 * Retrieve the command history of a pool.
3426 zpool_get_history(zpool_handle_t
*zhp
, nvlist_t
**nvhisp
)
3428 char buf
[HIS_BUF_LEN
];
3430 nvlist_t
**records
= NULL
;
3431 uint_t numrecords
= 0;
3435 uint64_t bytes_read
= sizeof (buf
);
3438 if ((err
= get_history(zhp
, buf
, &off
, &bytes_read
)) != 0)
3441 /* if nothing else was read in, we're at EOF, just return */
3445 if ((err
= zpool_history_unpack(buf
, bytes_read
,
3446 &leftover
, &records
, &numrecords
)) != 0)
3454 verify(nvlist_alloc(nvhisp
, NV_UNIQUE_NAME
, 0) == 0);
3455 verify(nvlist_add_nvlist_array(*nvhisp
, ZPOOL_HIST_RECORD
,
3456 records
, numrecords
) == 0);
3458 for (i
= 0; i
< numrecords
; i
++)
3459 nvlist_free(records
[i
]);
3466 * Retrieve the next event. If there is a new event available 'nvp' will
3467 * contain a newly allocated nvlist and 'dropped' will be set to the number
3468 * of missed events since the last call to this function. When 'nvp' is
3469 * set to NULL it indicates no new events are available. In either case
3470 * the function returns 0 and it is up to the caller to free 'nvp'. In
3471 * the case of a fatal error the function will return a non-zero value.
3472 * When the function is called in blocking mode it will not return until
3473 * a new event is available.
3476 zpool_events_next(libzfs_handle_t
*hdl
, nvlist_t
**nvp
,
3477 int *dropped
, int block
, int cleanup_fd
)
3479 zfs_cmd_t zc
= { "\0", "\0", "\0", "\0", 0 };
3484 zc
.zc_cleanup_fd
= cleanup_fd
;
3487 zc
.zc_guid
= ZEVENT_NONBLOCK
;
3489 if (zcmd_alloc_dst_nvlist(hdl
, &zc
, ZEVENT_SIZE
) != 0)
3493 if (zfs_ioctl(hdl
, ZFS_IOC_EVENTS_NEXT
, &zc
) != 0) {
3496 error
= zfs_error_fmt(hdl
, EZFS_POOLUNAVAIL
,
3497 dgettext(TEXT_DOMAIN
, "zfs shutdown"));
3500 /* Blocking error case should not occur */
3502 error
= zpool_standard_error_fmt(hdl
, errno
,
3503 dgettext(TEXT_DOMAIN
, "cannot get event"));
3507 if (zcmd_expand_dst_nvlist(hdl
, &zc
) != 0) {
3508 error
= zfs_error_fmt(hdl
, EZFS_NOMEM
,
3509 dgettext(TEXT_DOMAIN
, "cannot get event"));
3515 error
= zpool_standard_error_fmt(hdl
, errno
,
3516 dgettext(TEXT_DOMAIN
, "cannot get event"));
3521 error
= zcmd_read_dst_nvlist(hdl
, &zc
, nvp
);
3525 *dropped
= (int)zc
.zc_cookie
;
3527 zcmd_free_nvlists(&zc
);
3536 zpool_events_clear(libzfs_handle_t
*hdl
, int *count
)
3538 zfs_cmd_t zc
= { "\0", "\0", "\0", "\0", 0 };
3541 (void) snprintf(msg
, sizeof (msg
), dgettext(TEXT_DOMAIN
,
3542 "cannot clear events"));
3544 if (zfs_ioctl(hdl
, ZFS_IOC_EVENTS_CLEAR
, &zc
) != 0)
3545 return (zpool_standard_error_fmt(hdl
, errno
, msg
));
3548 *count
= (int)zc
.zc_cookie
; /* # of events cleared */
3554 zpool_obj_to_path(zpool_handle_t
*zhp
, uint64_t dsobj
, uint64_t obj
,
3555 char *pathname
, size_t len
)
3557 zfs_cmd_t zc
= { "\0", "\0", "\0", "\0", 0 };
3558 boolean_t mounted
= B_FALSE
;
3559 char *mntpnt
= NULL
;
3560 char dsname
[MAXNAMELEN
];
3563 /* special case for the MOS */
3564 (void) snprintf(pathname
, len
, "<metadata>:<0x%llx>", (longlong_t
)obj
);
3568 /* get the dataset's name */
3569 (void) strlcpy(zc
.zc_name
, zhp
->zpool_name
, sizeof (zc
.zc_name
));
3571 if (ioctl(zhp
->zpool_hdl
->libzfs_fd
,
3572 ZFS_IOC_DSOBJ_TO_DSNAME
, &zc
) != 0) {
3573 /* just write out a path of two object numbers */
3574 (void) snprintf(pathname
, len
, "<0x%llx>:<0x%llx>",
3575 (longlong_t
)dsobj
, (longlong_t
)obj
);
3578 (void) strlcpy(dsname
, zc
.zc_value
, sizeof (dsname
));
3580 /* find out if the dataset is mounted */
3581 mounted
= is_mounted(zhp
->zpool_hdl
, dsname
, &mntpnt
);
3583 /* get the corrupted object's path */
3584 (void) strlcpy(zc
.zc_name
, dsname
, sizeof (zc
.zc_name
));
3586 if (ioctl(zhp
->zpool_hdl
->libzfs_fd
, ZFS_IOC_OBJ_TO_PATH
,
3589 (void) snprintf(pathname
, len
, "%s%s", mntpnt
,
3592 (void) snprintf(pathname
, len
, "%s:%s",
3593 dsname
, zc
.zc_value
);
3596 (void) snprintf(pathname
, len
, "%s:<0x%llx>", dsname
, (longlong_t
)obj
);
3602 * Read the EFI label from the config, if a label does not exist then
3603 * pass back the error to the caller. If the caller has passed a non-NULL
3604 * diskaddr argument then we set it to the starting address of the EFI
3608 read_efi_label(nvlist_t
*config
, diskaddr_t
*sb
)
3612 char diskname
[MAXPATHLEN
];
3615 if (nvlist_lookup_string(config
, ZPOOL_CONFIG_PATH
, &path
) != 0)
3618 (void) snprintf(diskname
, sizeof (diskname
), "%s%s", RDISK_ROOT
,
3619 strrchr(path
, '/'));
3620 if ((fd
= open(diskname
, O_RDWR
|O_DIRECT
)) >= 0) {
3621 struct dk_gpt
*vtoc
;
3623 if ((err
= efi_alloc_and_read(fd
, &vtoc
)) >= 0) {
3625 *sb
= vtoc
->efi_parts
[0].p_start
;
3634 * determine where a partition starts on a disk in the current
3638 find_start_block(nvlist_t
*config
)
3642 diskaddr_t sb
= MAXOFFSET_T
;
3645 if (nvlist_lookup_nvlist_array(config
,
3646 ZPOOL_CONFIG_CHILDREN
, &child
, &children
) != 0) {
3647 if (nvlist_lookup_uint64(config
,
3648 ZPOOL_CONFIG_WHOLE_DISK
,
3649 &wholedisk
) != 0 || !wholedisk
) {
3650 return (MAXOFFSET_T
);
3652 if (read_efi_label(config
, &sb
) < 0)
3657 for (c
= 0; c
< children
; c
++) {
3658 sb
= find_start_block(child
[c
]);
3659 if (sb
!= MAXOFFSET_T
) {
3663 return (MAXOFFSET_T
);
3667 zpool_label_disk_wait(char *path
, int timeout
)
3669 struct stat64 statbuf
;
3673 * Wait timeout miliseconds for a newly created device to be available
3674 * from the given path. There is a small window when a /dev/ device
3675 * will exist and the udev link will not, so we must wait for the
3676 * symlink. Depending on the udev rules this may take a few seconds.
3678 for (i
= 0; i
< timeout
; i
++) {
3682 if ((stat64(path
, &statbuf
) == 0) && (errno
== 0))
3690 zpool_label_disk_check(char *path
)
3692 struct dk_gpt
*vtoc
;
3695 if ((fd
= open(path
, O_RDWR
|O_DIRECT
)) < 0)
3698 if ((err
= efi_alloc_and_read(fd
, &vtoc
)) != 0) {
3703 if (vtoc
->efi_flags
& EFI_GPT_PRIMARY_CORRUPT
) {
3715 * Label an individual disk. The name provided is the short name,
3716 * stripped of any leading /dev path.
3719 zpool_label_disk(libzfs_handle_t
*hdl
, zpool_handle_t
*zhp
, char *name
)
3721 char path
[MAXPATHLEN
];
3722 struct dk_gpt
*vtoc
;
3724 size_t resv
= EFI_MIN_RESV_SIZE
;
3725 uint64_t slice_size
;
3726 diskaddr_t start_block
;
3729 /* prepare an error message just in case */
3730 (void) snprintf(errbuf
, sizeof (errbuf
),
3731 dgettext(TEXT_DOMAIN
, "cannot label '%s'"), name
);
3736 if (pool_is_bootable(zhp
)) {
3737 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
3738 "EFI labeled devices are not supported on root "
3740 return (zfs_error(hdl
, EZFS_POOL_NOTSUP
, errbuf
));
3743 verify(nvlist_lookup_nvlist(zhp
->zpool_config
,
3744 ZPOOL_CONFIG_VDEV_TREE
, &nvroot
) == 0);
3746 if (zhp
->zpool_start_block
== 0)
3747 start_block
= find_start_block(nvroot
);
3749 start_block
= zhp
->zpool_start_block
;
3750 zhp
->zpool_start_block
= start_block
;
3753 start_block
= NEW_START_BLOCK
;
3756 (void) snprintf(path
, sizeof (path
), "%s/%s%s", RDISK_ROOT
, name
,
3759 if ((fd
= open(path
, O_RDWR
|O_DIRECT
)) < 0) {
3761 * This shouldn't happen. We've long since verified that this
3762 * is a valid device.
3764 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
3765 "unable to open device '%s': %d"), path
, errno
);
3766 return (zfs_error(hdl
, EZFS_OPENFAILED
, errbuf
));
3769 if (efi_alloc_and_init(fd
, EFI_NUMPAR
, &vtoc
) != 0) {
3771 * The only way this can fail is if we run out of memory, or we
3772 * were unable to read the disk's capacity
3774 if (errno
== ENOMEM
)
3775 (void) no_memory(hdl
);
3778 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
3779 "unable to read disk capacity"), name
);
3781 return (zfs_error(hdl
, EZFS_NOCAP
, errbuf
));
3784 slice_size
= vtoc
->efi_last_u_lba
+ 1;
3785 slice_size
-= EFI_MIN_RESV_SIZE
;
3786 if (start_block
== MAXOFFSET_T
)
3787 start_block
= NEW_START_BLOCK
;
3788 slice_size
-= start_block
;
3790 vtoc
->efi_parts
[0].p_start
= start_block
;
3791 vtoc
->efi_parts
[0].p_size
= slice_size
;
3794 * Why we use V_USR: V_BACKUP confuses users, and is considered
3795 * disposable by some EFI utilities (since EFI doesn't have a backup
3796 * slice). V_UNASSIGNED is supposed to be used only for zero size
3797 * partitions, and efi_write() will fail if we use it. V_ROOT, V_BOOT,
3798 * etc. were all pretty specific. V_USR is as close to reality as we
3799 * can get, in the absence of V_OTHER.
3801 vtoc
->efi_parts
[0].p_tag
= V_USR
;
3802 (void) strcpy(vtoc
->efi_parts
[0].p_name
, "zfs");
3804 vtoc
->efi_parts
[8].p_start
= slice_size
+ start_block
;
3805 vtoc
->efi_parts
[8].p_size
= resv
;
3806 vtoc
->efi_parts
[8].p_tag
= V_RESERVED
;
3808 if ((rval
= efi_write(fd
, vtoc
)) != 0) {
3810 * Some block drivers (like pcata) may not support EFI
3811 * GPT labels. Print out a helpful error message dir-
3812 * ecting the user to manually label the disk and give
3818 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
, "try using "
3819 "parted(8) and then provide a specific slice: %d"), rval
);
3820 return (zfs_error(hdl
, EZFS_LABELFAILED
, errbuf
));
3826 /* Wait for the first expected slice to appear. */
3827 (void) snprintf(path
, sizeof (path
), "%s/%s%s%s", DISK_ROOT
, name
,
3828 isdigit(name
[strlen(name
)-1]) ? "p" : "", FIRST_SLICE
);
3829 rval
= zpool_label_disk_wait(path
, 3000);
3831 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
, "failed to "
3832 "detect device partitions on '%s': %d"), path
, rval
);
3833 return (zfs_error(hdl
, EZFS_LABELFAILED
, errbuf
));
3836 /* We can't be to paranoid. Read the label back and verify it. */
3837 (void) snprintf(path
, sizeof (path
), "%s/%s", DISK_ROOT
, name
);
3838 rval
= zpool_label_disk_check(path
);
3840 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
, "freshly written "
3841 "EFI label on '%s' is damaged. Ensure\nthis device "
3842 "is not in in use, and is functioning properly: %d"),
3844 return (zfs_error(hdl
, EZFS_LABELFAILED
, errbuf
));