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 by Delphix. All rights reserved.
39 #include <sys/efi_partition.h>
41 #include <sys/zfs_ioctl.h>
44 #include "zfs_namecheck.h"
46 #include "libzfs_impl.h"
47 #include "zfs_comutil.h"
49 static int read_efi_label(nvlist_t
*config
, diskaddr_t
*sb
);
51 typedef struct prop_flags
{
52 int create
:1; /* Validate property on creation */
53 int import
:1; /* Validate property on import */
57 * ====================================================================
58 * zpool property functions
59 * ====================================================================
63 zpool_get_all_props(zpool_handle_t
*zhp
)
65 zfs_cmd_t zc
= { "\0", "\0", "\0", "\0", 0 };
66 libzfs_handle_t
*hdl
= zhp
->zpool_hdl
;
68 (void) strlcpy(zc
.zc_name
, zhp
->zpool_name
, sizeof (zc
.zc_name
));
70 if (zcmd_alloc_dst_nvlist(hdl
, &zc
, 0) != 0)
73 while (ioctl(hdl
->libzfs_fd
, ZFS_IOC_POOL_GET_PROPS
, &zc
) != 0) {
74 if (errno
== ENOMEM
) {
75 if (zcmd_expand_dst_nvlist(hdl
, &zc
) != 0) {
76 zcmd_free_nvlists(&zc
);
80 zcmd_free_nvlists(&zc
);
85 if (zcmd_read_dst_nvlist(hdl
, &zc
, &zhp
->zpool_props
) != 0) {
86 zcmd_free_nvlists(&zc
);
90 zcmd_free_nvlists(&zc
);
96 zpool_props_refresh(zpool_handle_t
*zhp
)
100 old_props
= zhp
->zpool_props
;
102 if (zpool_get_all_props(zhp
) != 0)
105 nvlist_free(old_props
);
110 zpool_get_prop_string(zpool_handle_t
*zhp
, zpool_prop_t prop
,
116 zprop_source_t source
;
118 nvl
= zhp
->zpool_props
;
119 if (nvlist_lookup_nvlist(nvl
, zpool_prop_to_name(prop
), &nv
) == 0) {
120 verify(nvlist_lookup_uint64(nv
, ZPROP_SOURCE
, &ival
) == 0);
122 verify(nvlist_lookup_string(nv
, ZPROP_VALUE
, &value
) == 0);
124 source
= ZPROP_SRC_DEFAULT
;
125 if ((value
= (char *)zpool_prop_default_string(prop
)) == NULL
)
136 zpool_get_prop_int(zpool_handle_t
*zhp
, zpool_prop_t prop
, zprop_source_t
*src
)
140 zprop_source_t source
;
142 if (zhp
->zpool_props
== NULL
&& zpool_get_all_props(zhp
)) {
144 * zpool_get_all_props() has most likely failed because
145 * the pool is faulted, but if all we need is the top level
146 * vdev's guid then get it from the zhp config nvlist.
148 if ((prop
== ZPOOL_PROP_GUID
) &&
149 (nvlist_lookup_nvlist(zhp
->zpool_config
,
150 ZPOOL_CONFIG_VDEV_TREE
, &nv
) == 0) &&
151 (nvlist_lookup_uint64(nv
, ZPOOL_CONFIG_GUID
, &value
)
155 return (zpool_prop_default_numeric(prop
));
158 nvl
= zhp
->zpool_props
;
159 if (nvlist_lookup_nvlist(nvl
, zpool_prop_to_name(prop
), &nv
) == 0) {
160 verify(nvlist_lookup_uint64(nv
, ZPROP_SOURCE
, &value
) == 0);
162 verify(nvlist_lookup_uint64(nv
, ZPROP_VALUE
, &value
) == 0);
164 source
= ZPROP_SRC_DEFAULT
;
165 value
= zpool_prop_default_numeric(prop
);
175 * Map VDEV STATE to printed strings.
178 zpool_state_to_name(vdev_state_t state
, vdev_aux_t aux
)
183 case VDEV_STATE_CLOSED
:
184 case VDEV_STATE_OFFLINE
:
185 return (gettext("OFFLINE"));
186 case VDEV_STATE_REMOVED
:
187 return (gettext("REMOVED"));
188 case VDEV_STATE_CANT_OPEN
:
189 if (aux
== VDEV_AUX_CORRUPT_DATA
|| aux
== VDEV_AUX_BAD_LOG
)
190 return (gettext("FAULTED"));
191 else if (aux
== VDEV_AUX_SPLIT_POOL
)
192 return (gettext("SPLIT"));
194 return (gettext("UNAVAIL"));
195 case VDEV_STATE_FAULTED
:
196 return (gettext("FAULTED"));
197 case VDEV_STATE_DEGRADED
:
198 return (gettext("DEGRADED"));
199 case VDEV_STATE_HEALTHY
:
200 return (gettext("ONLINE"));
203 return (gettext("UNKNOWN"));
207 * Get a zpool property value for 'prop' and return the value in
208 * a pre-allocated buffer.
211 zpool_get_prop(zpool_handle_t
*zhp
, zpool_prop_t prop
, char *buf
, size_t len
,
212 zprop_source_t
*srctype
)
216 zprop_source_t src
= ZPROP_SRC_NONE
;
221 if (zpool_get_state(zhp
) == POOL_STATE_UNAVAIL
) {
223 case ZPOOL_PROP_NAME
:
224 (void) strlcpy(buf
, zpool_get_name(zhp
), len
);
227 case ZPOOL_PROP_HEALTH
:
228 (void) strlcpy(buf
, "FAULTED", len
);
231 case ZPOOL_PROP_GUID
:
232 intval
= zpool_get_prop_int(zhp
, prop
, &src
);
233 (void) snprintf(buf
, len
, "%llu", (u_longlong_t
)intval
);
236 case ZPOOL_PROP_ALTROOT
:
237 case ZPOOL_PROP_CACHEFILE
:
238 if (zhp
->zpool_props
!= NULL
||
239 zpool_get_all_props(zhp
) == 0) {
241 zpool_get_prop_string(zhp
, prop
, &src
),
249 (void) strlcpy(buf
, "-", len
);
258 if (zhp
->zpool_props
== NULL
&& zpool_get_all_props(zhp
) &&
259 prop
!= ZPOOL_PROP_NAME
)
262 switch (zpool_prop_get_type(prop
)) {
263 case PROP_TYPE_STRING
:
264 (void) strlcpy(buf
, zpool_get_prop_string(zhp
, prop
, &src
),
268 case PROP_TYPE_NUMBER
:
269 intval
= zpool_get_prop_int(zhp
, prop
, &src
);
272 case ZPOOL_PROP_SIZE
:
273 case ZPOOL_PROP_ALLOCATED
:
274 case ZPOOL_PROP_FREE
:
275 case ZPOOL_PROP_ASHIFT
:
276 (void) zfs_nicenum(intval
, buf
, len
);
279 case ZPOOL_PROP_CAPACITY
:
280 (void) snprintf(buf
, len
, "%llu%%",
281 (u_longlong_t
)intval
);
284 case ZPOOL_PROP_DEDUPRATIO
:
285 (void) snprintf(buf
, len
, "%llu.%02llux",
286 (u_longlong_t
)(intval
/ 100),
287 (u_longlong_t
)(intval
% 100));
290 case ZPOOL_PROP_HEALTH
:
291 verify(nvlist_lookup_nvlist(zpool_get_config(zhp
, NULL
),
292 ZPOOL_CONFIG_VDEV_TREE
, &nvroot
) == 0);
293 verify(nvlist_lookup_uint64_array(nvroot
,
294 ZPOOL_CONFIG_VDEV_STATS
, (uint64_t **)&vs
, &vsc
)
297 (void) strlcpy(buf
, zpool_state_to_name(intval
,
301 (void) snprintf(buf
, len
, "%llu", (u_longlong_t
)intval
);
305 case PROP_TYPE_INDEX
:
306 intval
= zpool_get_prop_int(zhp
, prop
, &src
);
307 if (zpool_prop_index_to_string(prop
, intval
, &strval
)
310 (void) strlcpy(buf
, strval
, len
);
324 * Check if the bootfs name has the same pool name as it is set to.
325 * Assuming bootfs is a valid dataset name.
328 bootfs_name_valid(const char *pool
, char *bootfs
)
330 int len
= strlen(pool
);
332 if (!zfs_name_valid(bootfs
, ZFS_TYPE_FILESYSTEM
|ZFS_TYPE_SNAPSHOT
))
335 if (strncmp(pool
, bootfs
, len
) == 0 &&
336 (bootfs
[len
] == '/' || bootfs
[len
] == '\0'))
343 * Inspect the configuration to determine if any of the devices contain
347 pool_uses_efi(nvlist_t
*config
)
352 if (nvlist_lookup_nvlist_array(config
, ZPOOL_CONFIG_CHILDREN
,
353 &child
, &children
) != 0)
354 return (read_efi_label(config
, NULL
) >= 0);
356 for (c
= 0; c
< children
; c
++) {
357 if (pool_uses_efi(child
[c
]))
364 pool_is_bootable(zpool_handle_t
*zhp
)
366 char bootfs
[ZPOOL_MAXNAMELEN
];
368 return (zpool_get_prop(zhp
, ZPOOL_PROP_BOOTFS
, bootfs
,
369 sizeof (bootfs
), NULL
) == 0 && strncmp(bootfs
, "-",
370 sizeof (bootfs
)) != 0);
375 * Given an nvlist of zpool properties to be set, validate that they are
376 * correct, and parse any numeric properties (index, boolean, etc) if they are
377 * specified as strings.
380 zpool_valid_proplist(libzfs_handle_t
*hdl
, const char *poolname
,
381 nvlist_t
*props
, uint64_t version
, prop_flags_t flags
, char *errbuf
)
389 struct stat64 statbuf
;
393 if (nvlist_alloc(&retprops
, NV_UNIQUE_NAME
, 0) != 0) {
394 (void) no_memory(hdl
);
399 while ((elem
= nvlist_next_nvpair(props
, elem
)) != NULL
) {
400 const char *propname
= nvpair_name(elem
);
403 * Make sure this property is valid and applies to this type.
405 if ((prop
= zpool_name_to_prop(propname
)) == ZPROP_INVAL
) {
406 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
407 "invalid property '%s'"), propname
);
408 (void) zfs_error(hdl
, EZFS_BADPROP
, errbuf
);
412 if (zpool_prop_readonly(prop
)) {
413 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
, "'%s' "
414 "is readonly"), propname
);
415 (void) zfs_error(hdl
, EZFS_PROPREADONLY
, errbuf
);
419 if (zprop_parse_value(hdl
, elem
, prop
, ZFS_TYPE_POOL
, retprops
,
420 &strval
, &intval
, errbuf
) != 0)
424 * Perform additional checking for specific properties.
429 case ZPOOL_PROP_VERSION
:
430 if (intval
< version
|| intval
> SPA_VERSION
) {
431 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
432 "property '%s' number %d is invalid."),
434 (void) zfs_error(hdl
, EZFS_BADVERSION
, errbuf
);
439 case ZPOOL_PROP_ASHIFT
:
441 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
442 "property '%s' can only be set at "
443 "creation time"), propname
);
444 (void) zfs_error(hdl
, EZFS_BADPROP
, errbuf
);
448 if (intval
!= 0 && (intval
< 9 || intval
> 13)) {
449 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
450 "property '%s' number %d is invalid."),
452 (void) zfs_error(hdl
, EZFS_BADPROP
, errbuf
);
457 case ZPOOL_PROP_BOOTFS
:
458 if (flags
.create
|| flags
.import
) {
459 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
460 "property '%s' cannot be set at creation "
461 "or import time"), propname
);
462 (void) zfs_error(hdl
, EZFS_BADPROP
, errbuf
);
466 if (version
< SPA_VERSION_BOOTFS
) {
467 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
468 "pool must be upgraded to support "
469 "'%s' property"), propname
);
470 (void) zfs_error(hdl
, EZFS_BADVERSION
, errbuf
);
475 * bootfs property value has to be a dataset name and
476 * the dataset has to be in the same pool as it sets to.
478 if (strval
[0] != '\0' && !bootfs_name_valid(poolname
,
480 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
, "'%s' "
481 "is an invalid name"), strval
);
482 (void) zfs_error(hdl
, EZFS_INVALIDNAME
, errbuf
);
486 if ((zhp
= zpool_open_canfail(hdl
, poolname
)) == NULL
) {
487 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
488 "could not open pool '%s'"), poolname
);
489 (void) zfs_error(hdl
, EZFS_OPENFAILED
, errbuf
);
492 verify(nvlist_lookup_nvlist(zpool_get_config(zhp
, NULL
),
493 ZPOOL_CONFIG_VDEV_TREE
, &nvroot
) == 0);
495 #if defined(__sun__) || defined(__sun)
497 * bootfs property cannot be set on a disk which has
500 if (pool_uses_efi(nvroot
)) {
501 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
502 "property '%s' not supported on "
503 "EFI labeled devices"), propname
);
504 (void) zfs_error(hdl
, EZFS_POOL_NOTSUP
, errbuf
);
512 case ZPOOL_PROP_ALTROOT
:
513 if (!flags
.create
&& !flags
.import
) {
514 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
515 "property '%s' can only be set during pool "
516 "creation or import"), propname
);
517 (void) zfs_error(hdl
, EZFS_BADPROP
, errbuf
);
521 if (strval
[0] != '/') {
522 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
523 "bad alternate root '%s'"), strval
);
524 (void) zfs_error(hdl
, EZFS_BADPATH
, errbuf
);
529 case ZPOOL_PROP_CACHEFILE
:
530 if (strval
[0] == '\0')
533 if (strcmp(strval
, "none") == 0)
536 if (strval
[0] != '/') {
537 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
538 "property '%s' must be empty, an "
539 "absolute path, or 'none'"), propname
);
540 (void) zfs_error(hdl
, EZFS_BADPATH
, errbuf
);
544 slash
= strrchr(strval
, '/');
546 if (slash
[1] == '\0' || strcmp(slash
, "/.") == 0 ||
547 strcmp(slash
, "/..") == 0) {
548 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
549 "'%s' is not a valid file"), strval
);
550 (void) zfs_error(hdl
, EZFS_BADPATH
, errbuf
);
556 if (strval
[0] != '\0' &&
557 (stat64(strval
, &statbuf
) != 0 ||
558 !S_ISDIR(statbuf
.st_mode
))) {
559 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
560 "'%s' is not a valid directory"),
562 (void) zfs_error(hdl
, EZFS_BADPATH
, errbuf
);
569 case ZPOOL_PROP_READONLY
:
571 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
572 "property '%s' can only be set at "
573 "import time"), propname
);
574 (void) zfs_error(hdl
, EZFS_BADPROP
, errbuf
);
583 nvlist_free(retprops
);
588 * Set zpool property : propname=propval.
591 zpool_set_prop(zpool_handle_t
*zhp
, const char *propname
, const char *propval
)
593 zfs_cmd_t zc
= { "\0", "\0", "\0", "\0", 0 };
596 nvlist_t
*nvl
= NULL
;
599 prop_flags_t flags
= { 0 };
601 (void) snprintf(errbuf
, sizeof (errbuf
),
602 dgettext(TEXT_DOMAIN
, "cannot set property for '%s'"),
605 if (nvlist_alloc(&nvl
, NV_UNIQUE_NAME
, 0) != 0)
606 return (no_memory(zhp
->zpool_hdl
));
608 if (nvlist_add_string(nvl
, propname
, propval
) != 0) {
610 return (no_memory(zhp
->zpool_hdl
));
613 version
= zpool_get_prop_int(zhp
, ZPOOL_PROP_VERSION
, NULL
);
614 if ((realprops
= zpool_valid_proplist(zhp
->zpool_hdl
,
615 zhp
->zpool_name
, nvl
, version
, flags
, errbuf
)) == NULL
) {
624 * Execute the corresponding ioctl() to set this property.
626 (void) strlcpy(zc
.zc_name
, zhp
->zpool_name
, sizeof (zc
.zc_name
));
628 if (zcmd_write_src_nvlist(zhp
->zpool_hdl
, &zc
, nvl
) != 0) {
633 ret
= zfs_ioctl(zhp
->zpool_hdl
, ZFS_IOC_POOL_SET_PROPS
, &zc
);
635 zcmd_free_nvlists(&zc
);
639 (void) zpool_standard_error(zhp
->zpool_hdl
, errno
, errbuf
);
641 (void) zpool_props_refresh(zhp
);
647 zpool_expand_proplist(zpool_handle_t
*zhp
, zprop_list_t
**plp
)
649 libzfs_handle_t
*hdl
= zhp
->zpool_hdl
;
651 char buf
[ZFS_MAXPROPLEN
];
653 if (zprop_expand_list(hdl
, plp
, ZFS_TYPE_POOL
) != 0)
656 for (entry
= *plp
; entry
!= NULL
; entry
= entry
->pl_next
) {
661 if (entry
->pl_prop
!= ZPROP_INVAL
&&
662 zpool_get_prop(zhp
, entry
->pl_prop
, buf
, sizeof (buf
),
664 if (strlen(buf
) > entry
->pl_width
)
665 entry
->pl_width
= strlen(buf
);
674 * Don't start the slice at the default block of 34; many storage
675 * devices will use a stripe width of 128k, other vendors prefer a 1m
676 * alignment. It is best to play it safe and ensure a 1m alignment
677 * given 512B blocks. When the block size is larger by a power of 2
678 * we will still be 1m aligned. Some devices are sensitive to the
679 * partition ending alignment as well.
681 #define NEW_START_BLOCK 2048
682 #define PARTITION_END_ALIGNMENT 2048
685 * Validate the given pool name, optionally putting an extended error message in
689 zpool_name_valid(libzfs_handle_t
*hdl
, boolean_t isopen
, const char *pool
)
695 ret
= pool_namecheck(pool
, &why
, &what
);
698 * The rules for reserved pool names were extended at a later point.
699 * But we need to support users with existing pools that may now be
700 * invalid. So we only check for this expanded set of names during a
701 * create (or import), and only in userland.
703 if (ret
== 0 && !isopen
&&
704 (strncmp(pool
, "mirror", 6) == 0 ||
705 strncmp(pool
, "raidz", 5) == 0 ||
706 strncmp(pool
, "spare", 5) == 0 ||
707 strcmp(pool
, "log") == 0)) {
710 dgettext(TEXT_DOMAIN
, "name is reserved"));
718 case NAME_ERR_TOOLONG
:
720 dgettext(TEXT_DOMAIN
, "name is too long"));
723 case NAME_ERR_INVALCHAR
:
725 dgettext(TEXT_DOMAIN
, "invalid character "
726 "'%c' in pool name"), what
);
729 case NAME_ERR_NOLETTER
:
730 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
731 "name must begin with a letter"));
734 case NAME_ERR_RESERVED
:
735 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
736 "name is reserved"));
739 case NAME_ERR_DISKLIKE
:
740 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
741 "pool name is reserved"));
744 case NAME_ERR_LEADING_SLASH
:
745 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
746 "leading slash in name"));
749 case NAME_ERR_EMPTY_COMPONENT
:
750 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
751 "empty component in name"));
754 case NAME_ERR_TRAILING_SLASH
:
755 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
756 "trailing slash in name"));
759 case NAME_ERR_MULTIPLE_AT
:
760 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
761 "multiple '@' delimiters in name"));
764 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
765 "permission set is missing '@'"));
776 * Open a handle to the given pool, even if the pool is currently in the FAULTED
780 zpool_open_canfail(libzfs_handle_t
*hdl
, const char *pool
)
786 * Make sure the pool name is valid.
788 if (!zpool_name_valid(hdl
, B_TRUE
, pool
)) {
789 (void) zfs_error_fmt(hdl
, EZFS_INVALIDNAME
,
790 dgettext(TEXT_DOMAIN
, "cannot open '%s'"),
795 if ((zhp
= zfs_alloc(hdl
, sizeof (zpool_handle_t
))) == NULL
)
798 zhp
->zpool_hdl
= hdl
;
799 (void) strlcpy(zhp
->zpool_name
, pool
, sizeof (zhp
->zpool_name
));
801 if (zpool_refresh_stats(zhp
, &missing
) != 0) {
807 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
, "no such pool"));
808 (void) zfs_error_fmt(hdl
, EZFS_NOENT
,
809 dgettext(TEXT_DOMAIN
, "cannot open '%s'"), pool
);
818 * Like the above, but silent on error. Used when iterating over pools (because
819 * the configuration cache may be out of date).
822 zpool_open_silent(libzfs_handle_t
*hdl
, const char *pool
, zpool_handle_t
**ret
)
827 if ((zhp
= zfs_alloc(hdl
, sizeof (zpool_handle_t
))) == NULL
)
830 zhp
->zpool_hdl
= hdl
;
831 (void) strlcpy(zhp
->zpool_name
, pool
, sizeof (zhp
->zpool_name
));
833 if (zpool_refresh_stats(zhp
, &missing
) != 0) {
849 * Similar to zpool_open_canfail(), but refuses to open pools in the faulted
853 zpool_open(libzfs_handle_t
*hdl
, const char *pool
)
857 if ((zhp
= zpool_open_canfail(hdl
, pool
)) == NULL
)
860 if (zhp
->zpool_state
== POOL_STATE_UNAVAIL
) {
861 (void) zfs_error_fmt(hdl
, EZFS_POOLUNAVAIL
,
862 dgettext(TEXT_DOMAIN
, "cannot open '%s'"), zhp
->zpool_name
);
871 * Close the handle. Simply frees the memory associated with the handle.
874 zpool_close(zpool_handle_t
*zhp
)
876 if (zhp
->zpool_config
)
877 nvlist_free(zhp
->zpool_config
);
878 if (zhp
->zpool_old_config
)
879 nvlist_free(zhp
->zpool_old_config
);
880 if (zhp
->zpool_props
)
881 nvlist_free(zhp
->zpool_props
);
886 * Return the name of the pool.
889 zpool_get_name(zpool_handle_t
*zhp
)
891 return (zhp
->zpool_name
);
896 * Return the state of the pool (ACTIVE or UNAVAILABLE)
899 zpool_get_state(zpool_handle_t
*zhp
)
901 return (zhp
->zpool_state
);
905 * Create the named pool, using the provided vdev list. It is assumed
906 * that the consumer has already validated the contents of the nvlist, so we
907 * don't have to worry about error semantics.
910 zpool_create(libzfs_handle_t
*hdl
, const char *pool
, nvlist_t
*nvroot
,
911 nvlist_t
*props
, nvlist_t
*fsprops
)
913 zfs_cmd_t zc
= { "\0", "\0", "\0", "\0", 0 };
914 nvlist_t
*zc_fsprops
= NULL
;
915 nvlist_t
*zc_props
= NULL
;
920 (void) snprintf(msg
, sizeof (msg
), dgettext(TEXT_DOMAIN
,
921 "cannot create '%s'"), pool
);
923 if (!zpool_name_valid(hdl
, B_FALSE
, pool
))
924 return (zfs_error(hdl
, EZFS_INVALIDNAME
, msg
));
926 if (zcmd_write_conf_nvlist(hdl
, &zc
, nvroot
) != 0)
930 prop_flags_t flags
= { .create
= B_TRUE
, .import
= B_FALSE
};
932 if ((zc_props
= zpool_valid_proplist(hdl
, pool
, props
,
933 SPA_VERSION_1
, flags
, msg
)) == NULL
) {
942 zoned
= ((nvlist_lookup_string(fsprops
,
943 zfs_prop_to_name(ZFS_PROP_ZONED
), &zonestr
) == 0) &&
944 strcmp(zonestr
, "on") == 0);
946 if ((zc_fsprops
= zfs_valid_proplist(hdl
,
947 ZFS_TYPE_FILESYSTEM
, fsprops
, zoned
, NULL
, msg
)) == NULL
) {
951 (nvlist_alloc(&zc_props
, NV_UNIQUE_NAME
, 0) != 0)) {
954 if (nvlist_add_nvlist(zc_props
,
955 ZPOOL_ROOTFS_PROPS
, zc_fsprops
) != 0) {
960 if (zc_props
&& zcmd_write_src_nvlist(hdl
, &zc
, zc_props
) != 0)
963 (void) strlcpy(zc
.zc_name
, pool
, sizeof (zc
.zc_name
));
965 if ((ret
= zfs_ioctl(hdl
, ZFS_IOC_POOL_CREATE
, &zc
)) != 0) {
967 zcmd_free_nvlists(&zc
);
968 nvlist_free(zc_props
);
969 nvlist_free(zc_fsprops
);
974 * This can happen if the user has specified the same
975 * device multiple times. We can't reliably detect this
976 * until we try to add it and see we already have a
977 * label. This can also happen under if the device is
978 * part of an active md or lvm device.
980 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
981 "one or more vdevs refer to the same device, or one of\n"
982 "the devices is part of an active md or lvm device"));
983 return (zfs_error(hdl
, EZFS_BADDEV
, msg
));
987 * This occurs when one of the devices is below
988 * SPA_MINDEVSIZE. Unfortunately, we can't detect which
989 * device was the problem device since there's no
990 * reliable way to determine device size from userland.
995 zfs_nicenum(SPA_MINDEVSIZE
, buf
, sizeof (buf
));
997 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
998 "one or more devices is less than the "
999 "minimum size (%s)"), buf
);
1001 return (zfs_error(hdl
, EZFS_BADDEV
, msg
));
1004 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
1005 "one or more devices is out of space"));
1006 return (zfs_error(hdl
, EZFS_BADDEV
, msg
));
1009 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
1010 "cache device must be a disk or disk slice"));
1011 return (zfs_error(hdl
, EZFS_BADDEV
, msg
));
1014 return (zpool_standard_error(hdl
, errno
, msg
));
1019 * If this is an alternate root pool, then we automatically set the
1020 * mountpoint of the root dataset to be '/'.
1022 if (nvlist_lookup_string(props
, zpool_prop_to_name(ZPOOL_PROP_ALTROOT
),
1026 verify((zhp
= zfs_open(hdl
, pool
, ZFS_TYPE_DATASET
)) != NULL
);
1027 verify(zfs_prop_set(zhp
, zfs_prop_to_name(ZFS_PROP_MOUNTPOINT
),
1034 zcmd_free_nvlists(&zc
);
1035 nvlist_free(zc_props
);
1036 nvlist_free(zc_fsprops
);
1041 * Destroy the given pool. It is up to the caller to ensure that there are no
1042 * datasets left in the pool.
1045 zpool_destroy(zpool_handle_t
*zhp
)
1047 zfs_cmd_t zc
= { "\0", "\0", "\0", "\0", 0 };
1048 zfs_handle_t
*zfp
= NULL
;
1049 libzfs_handle_t
*hdl
= zhp
->zpool_hdl
;
1052 if (zhp
->zpool_state
== POOL_STATE_ACTIVE
&&
1053 (zfp
= zfs_open(hdl
, zhp
->zpool_name
, ZFS_TYPE_FILESYSTEM
)) == NULL
)
1056 (void) strlcpy(zc
.zc_name
, zhp
->zpool_name
, sizeof (zc
.zc_name
));
1058 if (zfs_ioctl(hdl
, ZFS_IOC_POOL_DESTROY
, &zc
) != 0) {
1059 (void) snprintf(msg
, sizeof (msg
), dgettext(TEXT_DOMAIN
,
1060 "cannot destroy '%s'"), zhp
->zpool_name
);
1062 if (errno
== EROFS
) {
1063 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
1064 "one or more devices is read only"));
1065 (void) zfs_error(hdl
, EZFS_BADDEV
, msg
);
1067 (void) zpool_standard_error(hdl
, errno
, msg
);
1076 remove_mountpoint(zfp
);
1084 * Add the given vdevs to the pool. The caller must have already performed the
1085 * necessary verification to ensure that the vdev specification is well-formed.
1088 zpool_add(zpool_handle_t
*zhp
, nvlist_t
*nvroot
)
1090 zfs_cmd_t zc
= { "\0", "\0", "\0", "\0", 0 };
1092 libzfs_handle_t
*hdl
= zhp
->zpool_hdl
;
1094 nvlist_t
**spares
, **l2cache
;
1095 uint_t nspares
, nl2cache
;
1097 (void) snprintf(msg
, sizeof (msg
), dgettext(TEXT_DOMAIN
,
1098 "cannot add to '%s'"), zhp
->zpool_name
);
1100 if (zpool_get_prop_int(zhp
, ZPOOL_PROP_VERSION
, NULL
) <
1101 SPA_VERSION_SPARES
&&
1102 nvlist_lookup_nvlist_array(nvroot
, ZPOOL_CONFIG_SPARES
,
1103 &spares
, &nspares
) == 0) {
1104 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
, "pool must be "
1105 "upgraded to add hot spares"));
1106 return (zfs_error(hdl
, EZFS_BADVERSION
, msg
));
1109 if (pool_is_bootable(zhp
) && nvlist_lookup_nvlist_array(nvroot
,
1110 ZPOOL_CONFIG_SPARES
, &spares
, &nspares
) == 0) {
1113 for (s
= 0; s
< nspares
; s
++) {
1116 if (nvlist_lookup_string(spares
[s
], ZPOOL_CONFIG_PATH
,
1117 &path
) == 0 && pool_uses_efi(spares
[s
])) {
1118 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
1119 "device '%s' contains an EFI label and "
1120 "cannot be used on root pools."),
1121 zpool_vdev_name(hdl
, NULL
, spares
[s
],
1123 return (zfs_error(hdl
, EZFS_POOL_NOTSUP
, msg
));
1128 if (zpool_get_prop_int(zhp
, ZPOOL_PROP_VERSION
, NULL
) <
1129 SPA_VERSION_L2CACHE
&&
1130 nvlist_lookup_nvlist_array(nvroot
, ZPOOL_CONFIG_L2CACHE
,
1131 &l2cache
, &nl2cache
) == 0) {
1132 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
, "pool must be "
1133 "upgraded to add cache devices"));
1134 return (zfs_error(hdl
, EZFS_BADVERSION
, msg
));
1137 if (zcmd_write_conf_nvlist(hdl
, &zc
, nvroot
) != 0)
1139 (void) strlcpy(zc
.zc_name
, zhp
->zpool_name
, sizeof (zc
.zc_name
));
1141 if (zfs_ioctl(hdl
, ZFS_IOC_VDEV_ADD
, &zc
) != 0) {
1145 * This can happen if the user has specified the same
1146 * device multiple times. We can't reliably detect this
1147 * until we try to add it and see we already have a
1150 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
1151 "one or more vdevs refer to the same device"));
1152 (void) zfs_error(hdl
, EZFS_BADDEV
, msg
);
1157 * This occurrs when one of the devices is below
1158 * SPA_MINDEVSIZE. Unfortunately, we can't detect which
1159 * device was the problem device since there's no
1160 * reliable way to determine device size from userland.
1165 zfs_nicenum(SPA_MINDEVSIZE
, buf
, sizeof (buf
));
1167 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
1168 "device is less than the minimum "
1171 (void) zfs_error(hdl
, EZFS_BADDEV
, msg
);
1175 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
1176 "pool must be upgraded to add these vdevs"));
1177 (void) zfs_error(hdl
, EZFS_BADVERSION
, msg
);
1181 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
1182 "root pool can not have multiple vdevs"
1183 " or separate logs"));
1184 (void) zfs_error(hdl
, EZFS_POOL_NOTSUP
, msg
);
1188 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
1189 "cache device must be a disk or disk slice"));
1190 (void) zfs_error(hdl
, EZFS_BADDEV
, msg
);
1194 (void) zpool_standard_error(hdl
, errno
, msg
);
1202 zcmd_free_nvlists(&zc
);
1208 * Exports the pool from the system. The caller must ensure that there are no
1209 * mounted datasets in the pool.
1212 zpool_export_common(zpool_handle_t
*zhp
, boolean_t force
, boolean_t hardforce
)
1214 zfs_cmd_t zc
= { "\0", "\0", "\0", "\0", 0 };
1217 (void) snprintf(msg
, sizeof (msg
), dgettext(TEXT_DOMAIN
,
1218 "cannot export '%s'"), zhp
->zpool_name
);
1220 (void) strlcpy(zc
.zc_name
, zhp
->zpool_name
, sizeof (zc
.zc_name
));
1221 zc
.zc_cookie
= force
;
1222 zc
.zc_guid
= hardforce
;
1224 if (zfs_ioctl(zhp
->zpool_hdl
, ZFS_IOC_POOL_EXPORT
, &zc
) != 0) {
1227 zfs_error_aux(zhp
->zpool_hdl
, dgettext(TEXT_DOMAIN
,
1228 "use '-f' to override the following errors:\n"
1229 "'%s' has an active shared spare which could be"
1230 " used by other pools once '%s' is exported."),
1231 zhp
->zpool_name
, zhp
->zpool_name
);
1232 return (zfs_error(zhp
->zpool_hdl
, EZFS_ACTIVE_SPARE
,
1235 return (zpool_standard_error_fmt(zhp
->zpool_hdl
, errno
,
1244 zpool_export(zpool_handle_t
*zhp
, boolean_t force
)
1246 return (zpool_export_common(zhp
, force
, B_FALSE
));
1250 zpool_export_force(zpool_handle_t
*zhp
)
1252 return (zpool_export_common(zhp
, B_TRUE
, B_TRUE
));
1256 zpool_rewind_exclaim(libzfs_handle_t
*hdl
, const char *name
, boolean_t dryrun
,
1259 nvlist_t
*nv
= NULL
;
1265 if (!hdl
->libzfs_printerr
|| config
== NULL
)
1268 if (nvlist_lookup_nvlist(config
, ZPOOL_CONFIG_LOAD_INFO
, &nv
) != 0)
1271 if (nvlist_lookup_uint64(nv
, ZPOOL_CONFIG_LOAD_TIME
, &rewindto
) != 0)
1273 (void) nvlist_lookup_int64(nv
, ZPOOL_CONFIG_REWIND_TIME
, &loss
);
1275 if (localtime_r((time_t *)&rewindto
, &t
) != NULL
&&
1276 strftime(timestr
, 128, "%c", &t
) != 0) {
1278 (void) printf(dgettext(TEXT_DOMAIN
,
1279 "Would be able to return %s "
1280 "to its state as of %s.\n"),
1283 (void) printf(dgettext(TEXT_DOMAIN
,
1284 "Pool %s returned to its state as of %s.\n"),
1288 (void) printf(dgettext(TEXT_DOMAIN
,
1289 "%s approximately %lld "),
1290 dryrun
? "Would discard" : "Discarded",
1291 ((longlong_t
)loss
+ 30) / 60);
1292 (void) printf(dgettext(TEXT_DOMAIN
,
1293 "minutes of transactions.\n"));
1294 } else if (loss
> 0) {
1295 (void) printf(dgettext(TEXT_DOMAIN
,
1296 "%s approximately %lld "),
1297 dryrun
? "Would discard" : "Discarded",
1299 (void) printf(dgettext(TEXT_DOMAIN
,
1300 "seconds of transactions.\n"));
1306 zpool_explain_recover(libzfs_handle_t
*hdl
, const char *name
, int reason
,
1309 nvlist_t
*nv
= NULL
;
1311 uint64_t edata
= UINT64_MAX
;
1316 if (!hdl
->libzfs_printerr
)
1320 (void) printf(dgettext(TEXT_DOMAIN
, "action: "));
1322 (void) printf(dgettext(TEXT_DOMAIN
, "\t"));
1324 /* All attempted rewinds failed if ZPOOL_CONFIG_LOAD_TIME missing */
1325 if (nvlist_lookup_nvlist(config
, ZPOOL_CONFIG_LOAD_INFO
, &nv
) != 0 ||
1326 nvlist_lookup_uint64(nv
, ZPOOL_CONFIG_LOAD_TIME
, &rewindto
) != 0)
1329 (void) nvlist_lookup_int64(nv
, ZPOOL_CONFIG_REWIND_TIME
, &loss
);
1330 (void) nvlist_lookup_uint64(nv
, ZPOOL_CONFIG_LOAD_DATA_ERRORS
,
1333 (void) printf(dgettext(TEXT_DOMAIN
,
1334 "Recovery is possible, but will result in some data loss.\n"));
1336 if (localtime_r((time_t *)&rewindto
, &t
) != NULL
&&
1337 strftime(timestr
, 128, "%c", &t
) != 0) {
1338 (void) printf(dgettext(TEXT_DOMAIN
,
1339 "\tReturning the pool to its state as of %s\n"
1340 "\tshould correct the problem. "),
1343 (void) printf(dgettext(TEXT_DOMAIN
,
1344 "\tReverting the pool to an earlier state "
1345 "should correct the problem.\n\t"));
1349 (void) printf(dgettext(TEXT_DOMAIN
,
1350 "Approximately %lld minutes of data\n"
1351 "\tmust be discarded, irreversibly. "),
1352 ((longlong_t
)loss
+ 30) / 60);
1353 } else if (loss
> 0) {
1354 (void) printf(dgettext(TEXT_DOMAIN
,
1355 "Approximately %lld seconds of data\n"
1356 "\tmust be discarded, irreversibly. "),
1359 if (edata
!= 0 && edata
!= UINT64_MAX
) {
1361 (void) printf(dgettext(TEXT_DOMAIN
,
1362 "After rewind, at least\n"
1363 "\tone persistent user-data error will remain. "));
1365 (void) printf(dgettext(TEXT_DOMAIN
,
1366 "After rewind, several\n"
1367 "\tpersistent user-data errors will remain. "));
1370 (void) printf(dgettext(TEXT_DOMAIN
,
1371 "Recovery can be attempted\n\tby executing 'zpool %s -F %s'. "),
1372 reason
>= 0 ? "clear" : "import", name
);
1374 (void) printf(dgettext(TEXT_DOMAIN
,
1375 "A scrub of the pool\n"
1376 "\tis strongly recommended after recovery.\n"));
1380 (void) printf(dgettext(TEXT_DOMAIN
,
1381 "Destroy and re-create the pool from\n\ta backup source.\n"));
1385 * zpool_import() is a contracted interface. Should be kept the same
1388 * Applications should use zpool_import_props() to import a pool with
1389 * new properties value to be set.
1392 zpool_import(libzfs_handle_t
*hdl
, nvlist_t
*config
, const char *newname
,
1395 nvlist_t
*props
= NULL
;
1398 if (altroot
!= NULL
) {
1399 if (nvlist_alloc(&props
, NV_UNIQUE_NAME
, 0) != 0) {
1400 return (zfs_error_fmt(hdl
, EZFS_NOMEM
,
1401 dgettext(TEXT_DOMAIN
, "cannot import '%s'"),
1405 if (nvlist_add_string(props
,
1406 zpool_prop_to_name(ZPOOL_PROP_ALTROOT
), altroot
) != 0 ||
1407 nvlist_add_string(props
,
1408 zpool_prop_to_name(ZPOOL_PROP_CACHEFILE
), "none") != 0) {
1410 return (zfs_error_fmt(hdl
, EZFS_NOMEM
,
1411 dgettext(TEXT_DOMAIN
, "cannot import '%s'"),
1416 ret
= zpool_import_props(hdl
, config
, newname
, props
,
1424 print_vdev_tree(libzfs_handle_t
*hdl
, const char *name
, nvlist_t
*nv
,
1430 uint64_t is_log
= 0;
1432 (void) nvlist_lookup_uint64(nv
, ZPOOL_CONFIG_IS_LOG
,
1436 (void) printf("\t%*s%s%s\n", indent
, "", name
,
1437 is_log
? " [log]" : "");
1439 if (nvlist_lookup_nvlist_array(nv
, ZPOOL_CONFIG_CHILDREN
,
1440 &child
, &children
) != 0)
1443 for (c
= 0; c
< children
; c
++) {
1444 vname
= zpool_vdev_name(hdl
, NULL
, child
[c
], B_TRUE
);
1445 print_vdev_tree(hdl
, vname
, child
[c
], indent
+ 2);
1451 * Import the given pool using the known configuration and a list of
1452 * properties to be set. The configuration should have come from
1453 * zpool_find_import(). The 'newname' parameters control whether the pool
1454 * is imported with a different name.
1457 zpool_import_props(libzfs_handle_t
*hdl
, nvlist_t
*config
, const char *newname
,
1458 nvlist_t
*props
, int flags
)
1460 zfs_cmd_t zc
= { "\0", "\0", "\0", "\0", 0 };
1461 zpool_rewind_policy_t policy
;
1462 nvlist_t
*nv
= NULL
;
1463 nvlist_t
*nvinfo
= NULL
;
1464 nvlist_t
*missing
= NULL
;
1471 verify(nvlist_lookup_string(config
, ZPOOL_CONFIG_POOL_NAME
,
1474 (void) snprintf(errbuf
, sizeof (errbuf
), dgettext(TEXT_DOMAIN
,
1475 "cannot import pool '%s'"), origname
);
1477 if (newname
!= NULL
) {
1478 if (!zpool_name_valid(hdl
, B_FALSE
, newname
))
1479 return (zfs_error_fmt(hdl
, EZFS_INVALIDNAME
,
1480 dgettext(TEXT_DOMAIN
, "cannot import '%s'"),
1482 thename
= (char *)newname
;
1489 prop_flags_t flags
= { .create
= B_FALSE
, .import
= B_TRUE
};
1491 verify(nvlist_lookup_uint64(config
, ZPOOL_CONFIG_VERSION
,
1494 if ((props
= zpool_valid_proplist(hdl
, origname
,
1495 props
, version
, flags
, errbuf
)) == NULL
) {
1497 } else if (zcmd_write_src_nvlist(hdl
, &zc
, props
) != 0) {
1503 (void) strlcpy(zc
.zc_name
, thename
, sizeof (zc
.zc_name
));
1505 verify(nvlist_lookup_uint64(config
, ZPOOL_CONFIG_POOL_GUID
,
1508 if (zcmd_write_conf_nvlist(hdl
, &zc
, config
) != 0) {
1512 if (zcmd_alloc_dst_nvlist(hdl
, &zc
, zc
.zc_nvlist_conf_size
* 2) != 0) {
1517 zc
.zc_cookie
= flags
;
1518 while ((ret
= zfs_ioctl(hdl
, ZFS_IOC_POOL_IMPORT
, &zc
)) != 0 &&
1520 if (zcmd_expand_dst_nvlist(hdl
, &zc
) != 0) {
1521 zcmd_free_nvlists(&zc
);
1528 (void) zcmd_read_dst_nvlist(hdl
, &zc
, &nv
);
1529 zpool_get_rewind_policy(config
, &policy
);
1535 * Dry-run failed, but we print out what success
1536 * looks like if we found a best txg
1538 if (policy
.zrp_request
& ZPOOL_TRY_REWIND
) {
1539 zpool_rewind_exclaim(hdl
, newname
? origname
: thename
,
1545 if (newname
== NULL
)
1546 (void) snprintf(desc
, sizeof (desc
),
1547 dgettext(TEXT_DOMAIN
, "cannot import '%s'"),
1550 (void) snprintf(desc
, sizeof (desc
),
1551 dgettext(TEXT_DOMAIN
, "cannot import '%s' as '%s'"),
1557 * Unsupported version.
1559 (void) zfs_error(hdl
, EZFS_BADVERSION
, desc
);
1563 (void) zfs_error(hdl
, EZFS_INVALCONFIG
, desc
);
1567 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
1568 "one or more devices is read only"));
1569 (void) zfs_error(hdl
, EZFS_BADDEV
, desc
);
1573 if (nv
&& nvlist_lookup_nvlist(nv
,
1574 ZPOOL_CONFIG_LOAD_INFO
, &nvinfo
) == 0 &&
1575 nvlist_lookup_nvlist(nvinfo
,
1576 ZPOOL_CONFIG_MISSING_DEVICES
, &missing
) == 0) {
1577 (void) printf(dgettext(TEXT_DOMAIN
,
1578 "The devices below are missing, use "
1579 "'-m' to import the pool anyway:\n"));
1580 print_vdev_tree(hdl
, NULL
, missing
, 2);
1581 (void) printf("\n");
1583 (void) zpool_standard_error(hdl
, error
, desc
);
1587 (void) zpool_standard_error(hdl
, error
, desc
);
1591 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
1592 "one or more devices are already in use\n"));
1593 (void) zfs_error(hdl
, EZFS_BADDEV
, desc
);
1597 (void) zpool_standard_error(hdl
, error
, desc
);
1598 zpool_explain_recover(hdl
,
1599 newname
? origname
: thename
, -error
, nv
);
1606 zpool_handle_t
*zhp
;
1609 * This should never fail, but play it safe anyway.
1611 if (zpool_open_silent(hdl
, thename
, &zhp
) != 0)
1613 else if (zhp
!= NULL
)
1615 if (policy
.zrp_request
&
1616 (ZPOOL_DO_REWIND
| ZPOOL_TRY_REWIND
)) {
1617 zpool_rewind_exclaim(hdl
, newname
? origname
: thename
,
1618 ((policy
.zrp_request
& ZPOOL_TRY_REWIND
) != 0), nv
);
1624 zcmd_free_nvlists(&zc
);
1634 zpool_scan(zpool_handle_t
*zhp
, pool_scan_func_t func
)
1636 zfs_cmd_t zc
= { "\0", "\0", "\0", "\0", 0 };
1638 libzfs_handle_t
*hdl
= zhp
->zpool_hdl
;
1640 (void) strlcpy(zc
.zc_name
, zhp
->zpool_name
, sizeof (zc
.zc_name
));
1641 zc
.zc_cookie
= func
;
1643 if (zfs_ioctl(hdl
, ZFS_IOC_POOL_SCAN
, &zc
) == 0 ||
1644 (errno
== ENOENT
&& func
!= POOL_SCAN_NONE
))
1647 if (func
== POOL_SCAN_SCRUB
) {
1648 (void) snprintf(msg
, sizeof (msg
),
1649 dgettext(TEXT_DOMAIN
, "cannot scrub %s"), zc
.zc_name
);
1650 } else if (func
== POOL_SCAN_NONE
) {
1651 (void) snprintf(msg
, sizeof (msg
),
1652 dgettext(TEXT_DOMAIN
, "cannot cancel scrubbing %s"),
1655 assert(!"unexpected result");
1658 if (errno
== EBUSY
) {
1660 pool_scan_stat_t
*ps
= NULL
;
1663 verify(nvlist_lookup_nvlist(zhp
->zpool_config
,
1664 ZPOOL_CONFIG_VDEV_TREE
, &nvroot
) == 0);
1665 (void) nvlist_lookup_uint64_array(nvroot
,
1666 ZPOOL_CONFIG_SCAN_STATS
, (uint64_t **)&ps
, &psc
);
1667 if (ps
&& ps
->pss_func
== POOL_SCAN_SCRUB
)
1668 return (zfs_error(hdl
, EZFS_SCRUBBING
, msg
));
1670 return (zfs_error(hdl
, EZFS_RESILVERING
, msg
));
1671 } else if (errno
== ENOENT
) {
1672 return (zfs_error(hdl
, EZFS_NO_SCRUB
, msg
));
1674 return (zpool_standard_error(hdl
, errno
, msg
));
1679 * Find a vdev that matches the search criteria specified. We use the
1680 * the nvpair name to determine how we should look for the device.
1681 * 'avail_spare' is set to TRUE if the provided guid refers to an AVAIL
1682 * spare; but FALSE if its an INUSE spare.
1685 vdev_to_nvlist_iter(nvlist_t
*nv
, nvlist_t
*search
, boolean_t
*avail_spare
,
1686 boolean_t
*l2cache
, boolean_t
*log
)
1693 nvpair_t
*pair
= nvlist_next_nvpair(search
, NULL
);
1695 /* Nothing to look for */
1696 if (search
== NULL
|| pair
== NULL
)
1699 /* Obtain the key we will use to search */
1700 srchkey
= nvpair_name(pair
);
1702 switch (nvpair_type(pair
)) {
1703 case DATA_TYPE_UINT64
:
1704 if (strcmp(srchkey
, ZPOOL_CONFIG_GUID
) == 0) {
1705 uint64_t srchval
, theguid
;
1707 verify(nvpair_value_uint64(pair
, &srchval
) == 0);
1708 verify(nvlist_lookup_uint64(nv
, ZPOOL_CONFIG_GUID
,
1710 if (theguid
== srchval
)
1715 case DATA_TYPE_STRING
: {
1716 char *srchval
, *val
;
1718 verify(nvpair_value_string(pair
, &srchval
) == 0);
1719 if (nvlist_lookup_string(nv
, srchkey
, &val
) != 0)
1723 * Search for the requested value. Special cases:
1725 * - ZPOOL_CONFIG_PATH for whole disk entries. These end in with a
1726 * partition suffix "1", "-part1", or "p1". The suffix is hidden
1727 * from the user, but included in the string, so this matches around
1729 * - looking for a top-level vdev name (i.e. ZPOOL_CONFIG_TYPE).
1731 * Otherwise, all other searches are simple string compares.
1733 if (strcmp(srchkey
, ZPOOL_CONFIG_PATH
) == 0) {
1734 uint64_t wholedisk
= 0;
1736 (void) nvlist_lookup_uint64(nv
, ZPOOL_CONFIG_WHOLE_DISK
,
1739 char buf
[MAXPATHLEN
];
1741 zfs_append_partition(srchval
, buf
, sizeof (buf
));
1742 if (strcmp(val
, buf
) == 0)
1747 } else if (strcmp(srchkey
, ZPOOL_CONFIG_TYPE
) == 0 && val
) {
1748 char *type
, *idx
, *end
, *p
;
1749 uint64_t id
, vdev_id
;
1752 * Determine our vdev type, keeping in mind
1753 * that the srchval is composed of a type and
1754 * vdev id pair (i.e. mirror-4).
1756 if ((type
= strdup(srchval
)) == NULL
)
1759 if ((p
= strrchr(type
, '-')) == NULL
) {
1767 * If the types don't match then keep looking.
1769 if (strncmp(val
, type
, strlen(val
)) != 0) {
1774 verify(strncmp(type
, VDEV_TYPE_RAIDZ
,
1775 strlen(VDEV_TYPE_RAIDZ
)) == 0 ||
1776 strncmp(type
, VDEV_TYPE_MIRROR
,
1777 strlen(VDEV_TYPE_MIRROR
)) == 0);
1778 verify(nvlist_lookup_uint64(nv
, ZPOOL_CONFIG_ID
,
1782 vdev_id
= strtoull(idx
, &end
, 10);
1789 * Now verify that we have the correct vdev id.
1798 if (strcmp(srchval
, val
) == 0)
1807 if (nvlist_lookup_nvlist_array(nv
, ZPOOL_CONFIG_CHILDREN
,
1808 &child
, &children
) != 0)
1811 for (c
= 0; c
< children
; c
++) {
1812 if ((ret
= vdev_to_nvlist_iter(child
[c
], search
,
1813 avail_spare
, l2cache
, NULL
)) != NULL
) {
1815 * The 'is_log' value is only set for the toplevel
1816 * vdev, not the leaf vdevs. So we always lookup the
1817 * log device from the root of the vdev tree (where
1818 * 'log' is non-NULL).
1821 nvlist_lookup_uint64(child
[c
],
1822 ZPOOL_CONFIG_IS_LOG
, &is_log
) == 0 &&
1830 if (nvlist_lookup_nvlist_array(nv
, ZPOOL_CONFIG_SPARES
,
1831 &child
, &children
) == 0) {
1832 for (c
= 0; c
< children
; c
++) {
1833 if ((ret
= vdev_to_nvlist_iter(child
[c
], search
,
1834 avail_spare
, l2cache
, NULL
)) != NULL
) {
1835 *avail_spare
= B_TRUE
;
1841 if (nvlist_lookup_nvlist_array(nv
, ZPOOL_CONFIG_L2CACHE
,
1842 &child
, &children
) == 0) {
1843 for (c
= 0; c
< children
; c
++) {
1844 if ((ret
= vdev_to_nvlist_iter(child
[c
], search
,
1845 avail_spare
, l2cache
, NULL
)) != NULL
) {
1856 * Given a physical path (minus the "/devices" prefix), find the
1860 zpool_find_vdev_by_physpath(zpool_handle_t
*zhp
, const char *ppath
,
1861 boolean_t
*avail_spare
, boolean_t
*l2cache
, boolean_t
*log
)
1863 nvlist_t
*search
, *nvroot
, *ret
;
1865 verify(nvlist_alloc(&search
, NV_UNIQUE_NAME
, KM_SLEEP
) == 0);
1866 verify(nvlist_add_string(search
, ZPOOL_CONFIG_PHYS_PATH
, ppath
) == 0);
1868 verify(nvlist_lookup_nvlist(zhp
->zpool_config
, ZPOOL_CONFIG_VDEV_TREE
,
1871 *avail_spare
= B_FALSE
;
1875 ret
= vdev_to_nvlist_iter(nvroot
, search
, avail_spare
, l2cache
, log
);
1876 nvlist_free(search
);
1882 * Determine if we have an "interior" top-level vdev (i.e mirror/raidz).
1885 zpool_vdev_is_interior(const char *name
)
1887 if (strncmp(name
, VDEV_TYPE_RAIDZ
, strlen(VDEV_TYPE_RAIDZ
)) == 0 ||
1888 strncmp(name
, VDEV_TYPE_MIRROR
, strlen(VDEV_TYPE_MIRROR
)) == 0)
1894 zpool_find_vdev(zpool_handle_t
*zhp
, const char *path
, boolean_t
*avail_spare
,
1895 boolean_t
*l2cache
, boolean_t
*log
)
1897 char buf
[MAXPATHLEN
];
1899 nvlist_t
*nvroot
, *search
, *ret
;
1902 verify(nvlist_alloc(&search
, NV_UNIQUE_NAME
, KM_SLEEP
) == 0);
1904 guid
= strtoull(path
, &end
, 10);
1905 if (guid
!= 0 && *end
== '\0') {
1906 verify(nvlist_add_uint64(search
, ZPOOL_CONFIG_GUID
, guid
) == 0);
1907 } else if (zpool_vdev_is_interior(path
)) {
1908 verify(nvlist_add_string(search
, ZPOOL_CONFIG_TYPE
, path
) == 0);
1909 } else if (path
[0] != '/') {
1910 if (zfs_resolve_shortname(path
, buf
, sizeof (buf
)) < 0) {
1911 nvlist_free(search
);
1914 verify(nvlist_add_string(search
, ZPOOL_CONFIG_PATH
, buf
) == 0);
1916 verify(nvlist_add_string(search
, ZPOOL_CONFIG_PATH
, path
) == 0);
1919 verify(nvlist_lookup_nvlist(zhp
->zpool_config
, ZPOOL_CONFIG_VDEV_TREE
,
1922 *avail_spare
= B_FALSE
;
1926 ret
= vdev_to_nvlist_iter(nvroot
, search
, avail_spare
, l2cache
, log
);
1927 nvlist_free(search
);
1933 vdev_online(nvlist_t
*nv
)
1937 if (nvlist_lookup_uint64(nv
, ZPOOL_CONFIG_OFFLINE
, &ival
) == 0 ||
1938 nvlist_lookup_uint64(nv
, ZPOOL_CONFIG_FAULTED
, &ival
) == 0 ||
1939 nvlist_lookup_uint64(nv
, ZPOOL_CONFIG_REMOVED
, &ival
) == 0)
1946 * Helper function for zpool_get_physpaths().
1949 vdev_get_one_physpath(nvlist_t
*config
, char *physpath
, size_t physpath_size
,
1950 size_t *bytes_written
)
1952 size_t bytes_left
, pos
, rsz
;
1956 if (nvlist_lookup_string(config
, ZPOOL_CONFIG_PHYS_PATH
,
1958 return (EZFS_NODEVICE
);
1960 pos
= *bytes_written
;
1961 bytes_left
= physpath_size
- pos
;
1962 format
= (pos
== 0) ? "%s" : " %s";
1964 rsz
= snprintf(physpath
+ pos
, bytes_left
, format
, tmppath
);
1965 *bytes_written
+= rsz
;
1967 if (rsz
>= bytes_left
) {
1968 /* if physpath was not copied properly, clear it */
1969 if (bytes_left
!= 0) {
1972 return (EZFS_NOSPC
);
1978 vdev_get_physpaths(nvlist_t
*nv
, char *physpath
, size_t phypath_size
,
1979 size_t *rsz
, boolean_t is_spare
)
1984 if (nvlist_lookup_string(nv
, ZPOOL_CONFIG_TYPE
, &type
) != 0)
1985 return (EZFS_INVALCONFIG
);
1987 if (strcmp(type
, VDEV_TYPE_DISK
) == 0) {
1989 * An active spare device has ZPOOL_CONFIG_IS_SPARE set.
1990 * For a spare vdev, we only want to boot from the active
1995 (void) nvlist_lookup_uint64(nv
, ZPOOL_CONFIG_IS_SPARE
,
1998 return (EZFS_INVALCONFIG
);
2001 if (vdev_online(nv
)) {
2002 if ((ret
= vdev_get_one_physpath(nv
, physpath
,
2003 phypath_size
, rsz
)) != 0)
2006 } else if (strcmp(type
, VDEV_TYPE_MIRROR
) == 0 ||
2007 strcmp(type
, VDEV_TYPE_REPLACING
) == 0 ||
2008 (is_spare
= (strcmp(type
, VDEV_TYPE_SPARE
) == 0))) {
2013 if (nvlist_lookup_nvlist_array(nv
,
2014 ZPOOL_CONFIG_CHILDREN
, &child
, &count
) != 0)
2015 return (EZFS_INVALCONFIG
);
2017 for (i
= 0; i
< count
; i
++) {
2018 ret
= vdev_get_physpaths(child
[i
], physpath
,
2019 phypath_size
, rsz
, is_spare
);
2020 if (ret
== EZFS_NOSPC
)
2025 return (EZFS_POOL_INVALARG
);
2029 * Get phys_path for a root pool config.
2030 * Return 0 on success; non-zero on failure.
2033 zpool_get_config_physpath(nvlist_t
*config
, char *physpath
, size_t phypath_size
)
2036 nvlist_t
*vdev_root
;
2043 if (nvlist_lookup_nvlist(config
, ZPOOL_CONFIG_VDEV_TREE
,
2045 return (EZFS_INVALCONFIG
);
2047 if (nvlist_lookup_string(vdev_root
, ZPOOL_CONFIG_TYPE
, &type
) != 0 ||
2048 nvlist_lookup_nvlist_array(vdev_root
, ZPOOL_CONFIG_CHILDREN
,
2049 &child
, &count
) != 0)
2050 return (EZFS_INVALCONFIG
);
2053 * root pool can not have EFI labeled disks and can only have
2054 * a single top-level vdev.
2056 if (strcmp(type
, VDEV_TYPE_ROOT
) != 0 || count
!= 1 ||
2057 pool_uses_efi(vdev_root
))
2058 return (EZFS_POOL_INVALARG
);
2060 (void) vdev_get_physpaths(child
[0], physpath
, phypath_size
, &rsz
,
2063 /* No online devices */
2065 return (EZFS_NODEVICE
);
2071 * Get phys_path for a root pool
2072 * Return 0 on success; non-zero on failure.
2075 zpool_get_physpath(zpool_handle_t
*zhp
, char *physpath
, size_t phypath_size
)
2077 return (zpool_get_config_physpath(zhp
->zpool_config
, physpath
,
2082 * If the device has being dynamically expanded then we need to relabel
2083 * the disk to use the new unallocated space.
2086 zpool_relabel_disk(libzfs_handle_t
*hdl
, const char *path
, const char *msg
)
2090 if ((fd
= open(path
, O_RDWR
|O_DIRECT
)) < 0) {
2091 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
, "cannot "
2092 "relabel '%s': unable to open device: %d"), path
, errno
);
2093 return (zfs_error(hdl
, EZFS_OPENFAILED
, msg
));
2097 * It's possible that we might encounter an error if the device
2098 * does not have any unallocated space left. If so, we simply
2099 * ignore that error and continue on.
2101 * Also, we don't call efi_rescan() - that would just return EBUSY.
2102 * The module will do it for us in vdev_disk_open().
2104 error
= efi_use_whole_disk(fd
);
2106 if (error
&& error
!= VT_ENOSPC
) {
2107 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
, "cannot "
2108 "relabel '%s': unable to read disk capacity"), path
);
2109 return (zfs_error(hdl
, EZFS_NOCAP
, msg
));
2115 * Bring the specified vdev online. The 'flags' parameter is a set of the
2116 * ZFS_ONLINE_* flags.
2119 zpool_vdev_online(zpool_handle_t
*zhp
, const char *path
, int flags
,
2120 vdev_state_t
*newstate
)
2122 zfs_cmd_t zc
= { "\0", "\0", "\0", "\0", 0 };
2125 boolean_t avail_spare
, l2cache
, islog
;
2126 libzfs_handle_t
*hdl
= zhp
->zpool_hdl
;
2129 if (flags
& ZFS_ONLINE_EXPAND
) {
2130 (void) snprintf(msg
, sizeof (msg
),
2131 dgettext(TEXT_DOMAIN
, "cannot expand %s"), path
);
2133 (void) snprintf(msg
, sizeof (msg
),
2134 dgettext(TEXT_DOMAIN
, "cannot online %s"), path
);
2137 (void) strlcpy(zc
.zc_name
, zhp
->zpool_name
, sizeof (zc
.zc_name
));
2138 if ((tgt
= zpool_find_vdev(zhp
, path
, &avail_spare
, &l2cache
,
2140 return (zfs_error(hdl
, EZFS_NODEVICE
, msg
));
2142 verify(nvlist_lookup_uint64(tgt
, ZPOOL_CONFIG_GUID
, &zc
.zc_guid
) == 0);
2145 return (zfs_error(hdl
, EZFS_ISSPARE
, msg
));
2147 if (flags
& ZFS_ONLINE_EXPAND
||
2148 zpool_get_prop_int(zhp
, ZPOOL_PROP_AUTOEXPAND
, NULL
)) {
2149 uint64_t wholedisk
= 0;
2151 (void) nvlist_lookup_uint64(tgt
, ZPOOL_CONFIG_WHOLE_DISK
,
2155 * XXX - L2ARC 1.0 devices can't support expansion.
2158 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
2159 "cannot expand cache devices"));
2160 return (zfs_error(hdl
, EZFS_VDEVNOTSUP
, msg
));
2164 const char *fullpath
= path
;
2165 char buf
[MAXPATHLEN
];
2167 if (path
[0] != '/') {
2168 error
= zfs_resolve_shortname(path
, buf
,
2171 return (zfs_error(hdl
, EZFS_NODEVICE
,
2177 error
= zpool_relabel_disk(hdl
, fullpath
, msg
);
2183 zc
.zc_cookie
= VDEV_STATE_ONLINE
;
2186 if (zfs_ioctl(hdl
, ZFS_IOC_VDEV_SET_STATE
, &zc
) != 0) {
2187 if (errno
== EINVAL
) {
2188 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
, "was split "
2189 "from this pool into a new one. Use '%s' "
2190 "instead"), "zpool detach");
2191 return (zfs_error(hdl
, EZFS_POSTSPLIT_ONLINE
, msg
));
2193 return (zpool_standard_error(hdl
, errno
, msg
));
2196 *newstate
= zc
.zc_cookie
;
2201 * Take the specified vdev offline
2204 zpool_vdev_offline(zpool_handle_t
*zhp
, const char *path
, boolean_t istmp
)
2206 zfs_cmd_t zc
= { "\0", "\0", "\0", "\0", 0 };
2209 boolean_t avail_spare
, l2cache
;
2210 libzfs_handle_t
*hdl
= zhp
->zpool_hdl
;
2212 (void) snprintf(msg
, sizeof (msg
),
2213 dgettext(TEXT_DOMAIN
, "cannot offline %s"), path
);
2215 (void) strlcpy(zc
.zc_name
, zhp
->zpool_name
, sizeof (zc
.zc_name
));
2216 if ((tgt
= zpool_find_vdev(zhp
, path
, &avail_spare
, &l2cache
,
2218 return (zfs_error(hdl
, EZFS_NODEVICE
, msg
));
2220 verify(nvlist_lookup_uint64(tgt
, ZPOOL_CONFIG_GUID
, &zc
.zc_guid
) == 0);
2223 return (zfs_error(hdl
, EZFS_ISSPARE
, msg
));
2225 zc
.zc_cookie
= VDEV_STATE_OFFLINE
;
2226 zc
.zc_obj
= istmp
? ZFS_OFFLINE_TEMPORARY
: 0;
2228 if (zfs_ioctl(hdl
, ZFS_IOC_VDEV_SET_STATE
, &zc
) == 0)
2235 * There are no other replicas of this device.
2237 return (zfs_error(hdl
, EZFS_NOREPLICAS
, msg
));
2241 * The log device has unplayed logs
2243 return (zfs_error(hdl
, EZFS_UNPLAYED_LOGS
, msg
));
2246 return (zpool_standard_error(hdl
, errno
, msg
));
2251 * Mark the given vdev faulted.
2254 zpool_vdev_fault(zpool_handle_t
*zhp
, uint64_t guid
, vdev_aux_t aux
)
2256 zfs_cmd_t zc
= { "\0", "\0", "\0", "\0", 0 };
2258 libzfs_handle_t
*hdl
= zhp
->zpool_hdl
;
2260 (void) snprintf(msg
, sizeof (msg
),
2261 dgettext(TEXT_DOMAIN
, "cannot fault %llu"), (u_longlong_t
)guid
);
2263 (void) strlcpy(zc
.zc_name
, zhp
->zpool_name
, sizeof (zc
.zc_name
));
2265 zc
.zc_cookie
= VDEV_STATE_FAULTED
;
2268 if (ioctl(hdl
->libzfs_fd
, ZFS_IOC_VDEV_SET_STATE
, &zc
) == 0)
2275 * There are no other replicas of this device.
2277 return (zfs_error(hdl
, EZFS_NOREPLICAS
, msg
));
2280 return (zpool_standard_error(hdl
, errno
, msg
));
2286 * Mark the given vdev degraded.
2289 zpool_vdev_degrade(zpool_handle_t
*zhp
, uint64_t guid
, vdev_aux_t aux
)
2291 zfs_cmd_t zc
= { "\0", "\0", "\0", "\0", 0 };
2293 libzfs_handle_t
*hdl
= zhp
->zpool_hdl
;
2295 (void) snprintf(msg
, sizeof (msg
),
2296 dgettext(TEXT_DOMAIN
, "cannot degrade %llu"), (u_longlong_t
)guid
);
2298 (void) strlcpy(zc
.zc_name
, zhp
->zpool_name
, sizeof (zc
.zc_name
));
2300 zc
.zc_cookie
= VDEV_STATE_DEGRADED
;
2303 if (ioctl(hdl
->libzfs_fd
, ZFS_IOC_VDEV_SET_STATE
, &zc
) == 0)
2306 return (zpool_standard_error(hdl
, errno
, msg
));
2310 * Returns TRUE if the given nvlist is a vdev that was originally swapped in as
2314 is_replacing_spare(nvlist_t
*search
, nvlist_t
*tgt
, int which
)
2320 if (nvlist_lookup_nvlist_array(search
, ZPOOL_CONFIG_CHILDREN
, &child
,
2322 verify(nvlist_lookup_string(search
, ZPOOL_CONFIG_TYPE
,
2325 if (strcmp(type
, VDEV_TYPE_SPARE
) == 0 &&
2326 children
== 2 && child
[which
] == tgt
)
2329 for (c
= 0; c
< children
; c
++)
2330 if (is_replacing_spare(child
[c
], tgt
, which
))
2338 * Attach new_disk (fully described by nvroot) to old_disk.
2339 * If 'replacing' is specified, the new disk will replace the old one.
2342 zpool_vdev_attach(zpool_handle_t
*zhp
,
2343 const char *old_disk
, const char *new_disk
, nvlist_t
*nvroot
, int replacing
)
2345 zfs_cmd_t zc
= { "\0", "\0", "\0", "\0", 0 };
2349 boolean_t avail_spare
, l2cache
, islog
;
2354 nvlist_t
*config_root
;
2355 libzfs_handle_t
*hdl
= zhp
->zpool_hdl
;
2356 boolean_t rootpool
= pool_is_bootable(zhp
);
2359 (void) snprintf(msg
, sizeof (msg
), dgettext(TEXT_DOMAIN
,
2360 "cannot replace %s with %s"), old_disk
, new_disk
);
2362 (void) snprintf(msg
, sizeof (msg
), dgettext(TEXT_DOMAIN
,
2363 "cannot attach %s to %s"), new_disk
, old_disk
);
2366 * If this is a root pool, make sure that we're not attaching an
2367 * EFI labeled device.
2369 if (rootpool
&& pool_uses_efi(nvroot
)) {
2370 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
2371 "EFI labeled devices are not supported on root pools."));
2372 return (zfs_error(hdl
, EZFS_POOL_NOTSUP
, msg
));
2375 (void) strlcpy(zc
.zc_name
, zhp
->zpool_name
, sizeof (zc
.zc_name
));
2376 if ((tgt
= zpool_find_vdev(zhp
, old_disk
, &avail_spare
, &l2cache
,
2378 return (zfs_error(hdl
, EZFS_NODEVICE
, msg
));
2381 return (zfs_error(hdl
, EZFS_ISSPARE
, msg
));
2384 return (zfs_error(hdl
, EZFS_ISL2CACHE
, msg
));
2386 verify(nvlist_lookup_uint64(tgt
, ZPOOL_CONFIG_GUID
, &zc
.zc_guid
) == 0);
2387 zc
.zc_cookie
= replacing
;
2389 if (nvlist_lookup_nvlist_array(nvroot
, ZPOOL_CONFIG_CHILDREN
,
2390 &child
, &children
) != 0 || children
!= 1) {
2391 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
2392 "new device must be a single disk"));
2393 return (zfs_error(hdl
, EZFS_INVALCONFIG
, msg
));
2396 verify(nvlist_lookup_nvlist(zpool_get_config(zhp
, NULL
),
2397 ZPOOL_CONFIG_VDEV_TREE
, &config_root
) == 0);
2399 if ((newname
= zpool_vdev_name(NULL
, NULL
, child
[0], B_FALSE
)) == NULL
)
2403 * If the target is a hot spare that has been swapped in, we can only
2404 * replace it with another hot spare.
2407 nvlist_lookup_uint64(tgt
, ZPOOL_CONFIG_IS_SPARE
, &val
) == 0 &&
2408 (zpool_find_vdev(zhp
, newname
, &avail_spare
, &l2cache
,
2409 NULL
) == NULL
|| !avail_spare
) &&
2410 is_replacing_spare(config_root
, tgt
, 1)) {
2411 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
2412 "can only be replaced by another hot spare"));
2414 return (zfs_error(hdl
, EZFS_BADTARGET
, msg
));
2419 if (zcmd_write_conf_nvlist(hdl
, &zc
, nvroot
) != 0)
2422 ret
= zfs_ioctl(hdl
, ZFS_IOC_VDEV_ATTACH
, &zc
);
2424 zcmd_free_nvlists(&zc
);
2429 * XXX need a better way to prevent user from
2430 * booting up a half-baked vdev.
2432 (void) fprintf(stderr
, dgettext(TEXT_DOMAIN
, "Make "
2433 "sure to wait until resilver is done "
2434 "before rebooting.\n"));
2442 * Can't attach to or replace this type of vdev.
2445 uint64_t version
= zpool_get_prop_int(zhp
,
2446 ZPOOL_PROP_VERSION
, NULL
);
2449 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
2450 "cannot replace a log with a spare"));
2451 else if (version
>= SPA_VERSION_MULTI_REPLACE
)
2452 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
2453 "already in replacing/spare config; wait "
2454 "for completion or use 'zpool detach'"));
2456 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
2457 "cannot replace a replacing device"));
2459 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
2460 "can only attach to mirrors and top-level "
2463 (void) zfs_error(hdl
, EZFS_BADTARGET
, msg
);
2468 * The new device must be a single disk.
2470 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
2471 "new device must be a single disk"));
2472 (void) zfs_error(hdl
, EZFS_INVALCONFIG
, msg
);
2476 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
, "%s is busy"),
2478 (void) zfs_error(hdl
, EZFS_BADDEV
, msg
);
2483 * The new device is too small.
2485 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
2486 "device is too small"));
2487 (void) zfs_error(hdl
, EZFS_BADDEV
, msg
);
2492 * The new device has a different alignment requirement.
2494 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
2495 "devices have different sector alignment"));
2496 (void) zfs_error(hdl
, EZFS_BADDEV
, msg
);
2501 * The resulting top-level vdev spec won't fit in the label.
2503 (void) zfs_error(hdl
, EZFS_DEVOVERFLOW
, msg
);
2507 (void) zpool_standard_error(hdl
, errno
, msg
);
2514 * Detach the specified device.
2517 zpool_vdev_detach(zpool_handle_t
*zhp
, const char *path
)
2519 zfs_cmd_t zc
= { "\0", "\0", "\0", "\0", 0 };
2522 boolean_t avail_spare
, l2cache
;
2523 libzfs_handle_t
*hdl
= zhp
->zpool_hdl
;
2525 (void) snprintf(msg
, sizeof (msg
),
2526 dgettext(TEXT_DOMAIN
, "cannot detach %s"), path
);
2528 (void) strlcpy(zc
.zc_name
, zhp
->zpool_name
, sizeof (zc
.zc_name
));
2529 if ((tgt
= zpool_find_vdev(zhp
, path
, &avail_spare
, &l2cache
,
2531 return (zfs_error(hdl
, EZFS_NODEVICE
, msg
));
2534 return (zfs_error(hdl
, EZFS_ISSPARE
, msg
));
2537 return (zfs_error(hdl
, EZFS_ISL2CACHE
, msg
));
2539 verify(nvlist_lookup_uint64(tgt
, ZPOOL_CONFIG_GUID
, &zc
.zc_guid
) == 0);
2541 if (zfs_ioctl(hdl
, ZFS_IOC_VDEV_DETACH
, &zc
) == 0)
2548 * Can't detach from this type of vdev.
2550 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
, "only "
2551 "applicable to mirror and replacing vdevs"));
2552 (void) zfs_error(hdl
, EZFS_BADTARGET
, msg
);
2557 * There are no other replicas of this device.
2559 (void) zfs_error(hdl
, EZFS_NOREPLICAS
, msg
);
2563 (void) zpool_standard_error(hdl
, errno
, msg
);
2570 * Find a mirror vdev in the source nvlist.
2572 * The mchild array contains a list of disks in one of the top-level mirrors
2573 * of the source pool. The schild array contains a list of disks that the
2574 * user specified on the command line. We loop over the mchild array to
2575 * see if any entry in the schild array matches.
2577 * If a disk in the mchild array is found in the schild array, we return
2578 * the index of that entry. Otherwise we return -1.
2581 find_vdev_entry(zpool_handle_t
*zhp
, nvlist_t
**mchild
, uint_t mchildren
,
2582 nvlist_t
**schild
, uint_t schildren
)
2586 for (mc
= 0; mc
< mchildren
; mc
++) {
2588 char *mpath
= zpool_vdev_name(zhp
->zpool_hdl
, zhp
,
2589 mchild
[mc
], B_FALSE
);
2591 for (sc
= 0; sc
< schildren
; sc
++) {
2592 char *spath
= zpool_vdev_name(zhp
->zpool_hdl
, zhp
,
2593 schild
[sc
], B_FALSE
);
2594 boolean_t result
= (strcmp(mpath
, spath
) == 0);
2610 * Split a mirror pool. If newroot points to null, then a new nvlist
2611 * is generated and it is the responsibility of the caller to free it.
2614 zpool_vdev_split(zpool_handle_t
*zhp
, char *newname
, nvlist_t
**newroot
,
2615 nvlist_t
*props
, splitflags_t flags
)
2617 zfs_cmd_t zc
= { "\0", "\0", "\0", "\0", 0 };
2619 nvlist_t
*tree
, *config
, **child
, **newchild
, *newconfig
= NULL
;
2620 nvlist_t
**varray
= NULL
, *zc_props
= NULL
;
2621 uint_t c
, children
, newchildren
, lastlog
= 0, vcount
, found
= 0;
2622 libzfs_handle_t
*hdl
= zhp
->zpool_hdl
;
2624 boolean_t freelist
= B_FALSE
, memory_err
= B_TRUE
;
2627 (void) snprintf(msg
, sizeof (msg
),
2628 dgettext(TEXT_DOMAIN
, "Unable to split %s"), zhp
->zpool_name
);
2630 if (!zpool_name_valid(hdl
, B_FALSE
, newname
))
2631 return (zfs_error(hdl
, EZFS_INVALIDNAME
, msg
));
2633 if ((config
= zpool_get_config(zhp
, NULL
)) == NULL
) {
2634 (void) fprintf(stderr
, gettext("Internal error: unable to "
2635 "retrieve pool configuration\n"));
2639 verify(nvlist_lookup_nvlist(config
, ZPOOL_CONFIG_VDEV_TREE
, &tree
)
2641 verify(nvlist_lookup_uint64(config
, ZPOOL_CONFIG_VERSION
, &vers
) == 0);
2644 prop_flags_t flags
= { .create
= B_FALSE
, .import
= B_TRUE
};
2645 if ((zc_props
= zpool_valid_proplist(hdl
, zhp
->zpool_name
,
2646 props
, vers
, flags
, msg
)) == NULL
)
2650 if (nvlist_lookup_nvlist_array(tree
, ZPOOL_CONFIG_CHILDREN
, &child
,
2652 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
2653 "Source pool is missing vdev tree"));
2655 nvlist_free(zc_props
);
2659 varray
= zfs_alloc(hdl
, children
* sizeof (nvlist_t
*));
2662 if (*newroot
== NULL
||
2663 nvlist_lookup_nvlist_array(*newroot
, ZPOOL_CONFIG_CHILDREN
,
2664 &newchild
, &newchildren
) != 0)
2667 for (c
= 0; c
< children
; c
++) {
2668 uint64_t is_log
= B_FALSE
, is_hole
= B_FALSE
;
2670 nvlist_t
**mchild
, *vdev
;
2675 * Unlike cache & spares, slogs are stored in the
2676 * ZPOOL_CONFIG_CHILDREN array. We filter them out here.
2678 (void) nvlist_lookup_uint64(child
[c
], ZPOOL_CONFIG_IS_LOG
,
2680 (void) nvlist_lookup_uint64(child
[c
], ZPOOL_CONFIG_IS_HOLE
,
2682 if (is_log
|| is_hole
) {
2684 * Create a hole vdev and put it in the config.
2686 if (nvlist_alloc(&vdev
, NV_UNIQUE_NAME
, 0) != 0)
2688 if (nvlist_add_string(vdev
, ZPOOL_CONFIG_TYPE
,
2689 VDEV_TYPE_HOLE
) != 0)
2691 if (nvlist_add_uint64(vdev
, ZPOOL_CONFIG_IS_HOLE
,
2696 varray
[vcount
++] = vdev
;
2700 verify(nvlist_lookup_string(child
[c
], ZPOOL_CONFIG_TYPE
, &type
)
2702 if (strcmp(type
, VDEV_TYPE_MIRROR
) != 0) {
2703 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
2704 "Source pool must be composed only of mirrors\n"));
2705 retval
= zfs_error(hdl
, EZFS_INVALCONFIG
, msg
);
2709 verify(nvlist_lookup_nvlist_array(child
[c
],
2710 ZPOOL_CONFIG_CHILDREN
, &mchild
, &mchildren
) == 0);
2712 /* find or add an entry for this top-level vdev */
2713 if (newchildren
> 0 &&
2714 (entry
= find_vdev_entry(zhp
, mchild
, mchildren
,
2715 newchild
, newchildren
)) >= 0) {
2716 /* We found a disk that the user specified. */
2717 vdev
= mchild
[entry
];
2720 /* User didn't specify a disk for this vdev. */
2721 vdev
= mchild
[mchildren
- 1];
2724 if (nvlist_dup(vdev
, &varray
[vcount
++], 0) != 0)
2728 /* did we find every disk the user specified? */
2729 if (found
!= newchildren
) {
2730 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
, "Device list must "
2731 "include at most one disk from each mirror"));
2732 retval
= zfs_error(hdl
, EZFS_INVALCONFIG
, msg
);
2736 /* Prepare the nvlist for populating. */
2737 if (*newroot
== NULL
) {
2738 if (nvlist_alloc(newroot
, NV_UNIQUE_NAME
, 0) != 0)
2741 if (nvlist_add_string(*newroot
, ZPOOL_CONFIG_TYPE
,
2742 VDEV_TYPE_ROOT
) != 0)
2745 verify(nvlist_remove_all(*newroot
, ZPOOL_CONFIG_CHILDREN
) == 0);
2748 /* Add all the children we found */
2749 if (nvlist_add_nvlist_array(*newroot
, ZPOOL_CONFIG_CHILDREN
, varray
,
2750 lastlog
== 0 ? vcount
: lastlog
) != 0)
2754 * If we're just doing a dry run, exit now with success.
2757 memory_err
= B_FALSE
;
2762 /* now build up the config list & call the ioctl */
2763 if (nvlist_alloc(&newconfig
, NV_UNIQUE_NAME
, 0) != 0)
2766 if (nvlist_add_nvlist(newconfig
,
2767 ZPOOL_CONFIG_VDEV_TREE
, *newroot
) != 0 ||
2768 nvlist_add_string(newconfig
,
2769 ZPOOL_CONFIG_POOL_NAME
, newname
) != 0 ||
2770 nvlist_add_uint64(newconfig
, ZPOOL_CONFIG_VERSION
, vers
) != 0)
2774 * The new pool is automatically part of the namespace unless we
2775 * explicitly export it.
2778 zc
.zc_cookie
= ZPOOL_EXPORT_AFTER_SPLIT
;
2779 (void) strlcpy(zc
.zc_name
, zhp
->zpool_name
, sizeof (zc
.zc_name
));
2780 (void) strlcpy(zc
.zc_string
, newname
, sizeof (zc
.zc_string
));
2781 if (zcmd_write_conf_nvlist(hdl
, &zc
, newconfig
) != 0)
2783 if (zc_props
!= NULL
&& zcmd_write_src_nvlist(hdl
, &zc
, zc_props
) != 0)
2786 if (zfs_ioctl(hdl
, ZFS_IOC_VDEV_SPLIT
, &zc
) != 0) {
2787 retval
= zpool_standard_error(hdl
, errno
, msg
);
2792 memory_err
= B_FALSE
;
2795 if (varray
!= NULL
) {
2798 for (v
= 0; v
< vcount
; v
++)
2799 nvlist_free(varray
[v
]);
2802 zcmd_free_nvlists(&zc
);
2804 nvlist_free(zc_props
);
2806 nvlist_free(newconfig
);
2808 nvlist_free(*newroot
);
2816 return (no_memory(hdl
));
2822 * Remove the given device. Currently, this is supported only for hot spares
2823 * and level 2 cache devices.
2826 zpool_vdev_remove(zpool_handle_t
*zhp
, const char *path
)
2828 zfs_cmd_t zc
= { "\0", "\0", "\0", "\0", 0 };
2831 boolean_t avail_spare
, l2cache
, islog
;
2832 libzfs_handle_t
*hdl
= zhp
->zpool_hdl
;
2835 (void) snprintf(msg
, sizeof (msg
),
2836 dgettext(TEXT_DOMAIN
, "cannot remove %s"), path
);
2838 (void) strlcpy(zc
.zc_name
, zhp
->zpool_name
, sizeof (zc
.zc_name
));
2839 if ((tgt
= zpool_find_vdev(zhp
, path
, &avail_spare
, &l2cache
,
2841 return (zfs_error(hdl
, EZFS_NODEVICE
, msg
));
2843 * XXX - this should just go away.
2845 if (!avail_spare
&& !l2cache
&& !islog
) {
2846 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
2847 "only inactive hot spares, cache, top-level, "
2848 "or log devices can be removed"));
2849 return (zfs_error(hdl
, EZFS_NODEVICE
, msg
));
2852 version
= zpool_get_prop_int(zhp
, ZPOOL_PROP_VERSION
, NULL
);
2853 if (islog
&& version
< SPA_VERSION_HOLES
) {
2854 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
2855 "pool must be upgrade to support log removal"));
2856 return (zfs_error(hdl
, EZFS_BADVERSION
, msg
));
2859 verify(nvlist_lookup_uint64(tgt
, ZPOOL_CONFIG_GUID
, &zc
.zc_guid
) == 0);
2861 if (zfs_ioctl(hdl
, ZFS_IOC_VDEV_REMOVE
, &zc
) == 0)
2864 return (zpool_standard_error(hdl
, errno
, msg
));
2868 * Clear the errors for the pool, or the particular device if specified.
2871 zpool_clear(zpool_handle_t
*zhp
, const char *path
, nvlist_t
*rewindnvl
)
2873 zfs_cmd_t zc
= { "\0", "\0", "\0", "\0", 0 };
2876 zpool_rewind_policy_t policy
;
2877 boolean_t avail_spare
, l2cache
;
2878 libzfs_handle_t
*hdl
= zhp
->zpool_hdl
;
2879 nvlist_t
*nvi
= NULL
;
2883 (void) snprintf(msg
, sizeof (msg
),
2884 dgettext(TEXT_DOMAIN
, "cannot clear errors for %s"),
2887 (void) snprintf(msg
, sizeof (msg
),
2888 dgettext(TEXT_DOMAIN
, "cannot clear errors for %s"),
2891 (void) strlcpy(zc
.zc_name
, zhp
->zpool_name
, sizeof (zc
.zc_name
));
2893 if ((tgt
= zpool_find_vdev(zhp
, path
, &avail_spare
,
2894 &l2cache
, NULL
)) == 0)
2895 return (zfs_error(hdl
, EZFS_NODEVICE
, msg
));
2898 * Don't allow error clearing for hot spares. Do allow
2899 * error clearing for l2cache devices.
2902 return (zfs_error(hdl
, EZFS_ISSPARE
, msg
));
2904 verify(nvlist_lookup_uint64(tgt
, ZPOOL_CONFIG_GUID
,
2908 zpool_get_rewind_policy(rewindnvl
, &policy
);
2909 zc
.zc_cookie
= policy
.zrp_request
;
2911 if (zcmd_alloc_dst_nvlist(hdl
, &zc
, zhp
->zpool_config_size
* 2) != 0)
2914 if (zcmd_write_src_nvlist(hdl
, &zc
, rewindnvl
) != 0)
2917 while ((error
= zfs_ioctl(hdl
, ZFS_IOC_CLEAR
, &zc
)) != 0 &&
2919 if (zcmd_expand_dst_nvlist(hdl
, &zc
) != 0) {
2920 zcmd_free_nvlists(&zc
);
2925 if (!error
|| ((policy
.zrp_request
& ZPOOL_TRY_REWIND
) &&
2926 errno
!= EPERM
&& errno
!= EACCES
)) {
2927 if (policy
.zrp_request
&
2928 (ZPOOL_DO_REWIND
| ZPOOL_TRY_REWIND
)) {
2929 (void) zcmd_read_dst_nvlist(hdl
, &zc
, &nvi
);
2930 zpool_rewind_exclaim(hdl
, zc
.zc_name
,
2931 ((policy
.zrp_request
& ZPOOL_TRY_REWIND
) != 0),
2935 zcmd_free_nvlists(&zc
);
2939 zcmd_free_nvlists(&zc
);
2940 return (zpool_standard_error(hdl
, errno
, msg
));
2944 * Similar to zpool_clear(), but takes a GUID (used by fmd).
2947 zpool_vdev_clear(zpool_handle_t
*zhp
, uint64_t guid
)
2949 zfs_cmd_t zc
= { "\0", "\0", "\0", "\0", 0 };
2951 libzfs_handle_t
*hdl
= zhp
->zpool_hdl
;
2953 (void) snprintf(msg
, sizeof (msg
),
2954 dgettext(TEXT_DOMAIN
, "cannot clear errors for %llx"),
2955 (u_longlong_t
)guid
);
2957 (void) strlcpy(zc
.zc_name
, zhp
->zpool_name
, sizeof (zc
.zc_name
));
2959 zc
.zc_cookie
= ZPOOL_NO_REWIND
;
2961 if (ioctl(hdl
->libzfs_fd
, ZFS_IOC_CLEAR
, &zc
) == 0)
2964 return (zpool_standard_error(hdl
, errno
, msg
));
2968 * Change the GUID for a pool.
2971 zpool_reguid(zpool_handle_t
*zhp
)
2974 libzfs_handle_t
*hdl
= zhp
->zpool_hdl
;
2975 zfs_cmd_t zc
= { "\0", "\0", "\0", "\0", 0 };
2977 (void) snprintf(msg
, sizeof (msg
),
2978 dgettext(TEXT_DOMAIN
, "cannot reguid '%s'"), zhp
->zpool_name
);
2980 (void) strlcpy(zc
.zc_name
, zhp
->zpool_name
, sizeof (zc
.zc_name
));
2981 if (zfs_ioctl(hdl
, ZFS_IOC_POOL_REGUID
, &zc
) == 0)
2984 return (zpool_standard_error(hdl
, errno
, msg
));
2988 * Convert from a devid string to a path.
2991 devid_to_path(char *devid_str
)
2996 devid_nmlist_t
*list
= NULL
;
2999 if (devid_str_decode(devid_str
, &devid
, &minor
) != 0)
3002 ret
= devid_deviceid_to_nmlist("/dev", devid
, minor
, &list
);
3004 devid_str_free(minor
);
3010 if ((path
= strdup(list
[0].devname
)) == NULL
)
3013 devid_free_nmlist(list
);
3019 * Convert from a path to a devid string.
3022 path_to_devid(const char *path
)
3028 if ((fd
= open(path
, O_RDONLY
)) < 0)
3033 if (devid_get(fd
, &devid
) == 0) {
3034 if (devid_get_minor_name(fd
, &minor
) == 0)
3035 ret
= devid_str_encode(devid
, minor
);
3037 devid_str_free(minor
);
3046 * Issue the necessary ioctl() to update the stored path value for the vdev. We
3047 * ignore any failure here, since a common case is for an unprivileged user to
3048 * type 'zpool status', and we'll display the correct information anyway.
3051 set_path(zpool_handle_t
*zhp
, nvlist_t
*nv
, const char *path
)
3053 zfs_cmd_t zc
= { "\0", "\0", "\0", "\0", 0 };
3055 (void) strncpy(zc
.zc_name
, zhp
->zpool_name
, sizeof (zc
.zc_name
));
3056 (void) strncpy(zc
.zc_value
, path
, sizeof (zc
.zc_value
));
3057 verify(nvlist_lookup_uint64(nv
, ZPOOL_CONFIG_GUID
,
3060 (void) ioctl(zhp
->zpool_hdl
->libzfs_fd
, ZFS_IOC_VDEV_SETPATH
, &zc
);
3064 * Remove partition suffix from a vdev path. Partition suffixes may take three
3065 * forms: "-partX", "pX", or "X", where X is a string of digits. The second
3066 * case only occurs when the suffix is preceded by a digit, i.e. "md0p0" The
3067 * third case only occurs when preceded by a string matching the regular
3068 * expression "^[hs]d[a-z]+", i.e. a scsi or ide disk.
3071 strip_partition(libzfs_handle_t
*hdl
, char *path
)
3073 char *tmp
= zfs_strdup(hdl
, path
);
3074 char *part
= NULL
, *d
= NULL
;
3076 if ((part
= strstr(tmp
, "-part")) && part
!= tmp
) {
3078 } else if ((part
= strrchr(tmp
, 'p')) &&
3079 part
> tmp
+ 1 && isdigit(*(part
-1))) {
3081 } else if ((tmp
[0] == 'h' || tmp
[0] == 's') && tmp
[1] == 'd') {
3082 for (d
= &tmp
[2]; isalpha(*d
); part
= ++d
);
3084 if (part
&& d
&& *d
!= '\0') {
3085 for (; isdigit(*d
); d
++);
3092 #define PATH_BUF_LEN 64
3095 * Given a vdev, return the name to display in iostat. If the vdev has a path,
3096 * we use that, stripping off any leading "/dev/dsk/"; if not, we use the type.
3097 * We also check if this is a whole disk, in which case we strip off the
3098 * trailing 's0' slice name.
3100 * This routine is also responsible for identifying when disks have been
3101 * reconfigured in a new location. The kernel will have opened the device by
3102 * devid, but the path will still refer to the old location. To catch this, we
3103 * first do a path -> devid translation (which is fast for the common case). If
3104 * the devid matches, we're done. If not, we do a reverse devid -> path
3105 * translation and issue the appropriate ioctl() to update the path of the vdev.
3106 * If 'zhp' is NULL, then this is an exported pool, and we don't need to do any
3110 zpool_vdev_name(libzfs_handle_t
*hdl
, zpool_handle_t
*zhp
, nvlist_t
*nv
,
3113 char *path
, *devid
, *type
;
3115 char buf
[PATH_BUF_LEN
];
3119 if (nvlist_lookup_uint64(nv
, ZPOOL_CONFIG_NOT_PRESENT
,
3121 verify(nvlist_lookup_uint64(nv
, ZPOOL_CONFIG_GUID
,
3123 (void) snprintf(buf
, sizeof (buf
), "%llu",
3124 (u_longlong_t
)value
);
3126 } else if (nvlist_lookup_string(nv
, ZPOOL_CONFIG_PATH
, &path
) == 0) {
3128 * If the device is dead (faulted, offline, etc) then don't
3129 * bother opening it. Otherwise we may be forcing the user to
3130 * open a misbehaving device, which can have undesirable
3133 if ((nvlist_lookup_uint64_array(nv
, ZPOOL_CONFIG_VDEV_STATS
,
3134 (uint64_t **)&vs
, &vsc
) != 0 ||
3135 vs
->vs_state
>= VDEV_STATE_DEGRADED
) &&
3137 nvlist_lookup_string(nv
, ZPOOL_CONFIG_DEVID
, &devid
) == 0) {
3139 * Determine if the current path is correct.
3141 char *newdevid
= path_to_devid(path
);
3143 if (newdevid
== NULL
||
3144 strcmp(devid
, newdevid
) != 0) {
3147 if ((newpath
= devid_to_path(devid
)) != NULL
) {
3149 * Update the path appropriately.
3151 set_path(zhp
, nv
, newpath
);
3152 if (nvlist_add_string(nv
,
3153 ZPOOL_CONFIG_PATH
, newpath
) == 0)
3154 verify(nvlist_lookup_string(nv
,
3162 devid_str_free(newdevid
);
3166 * For a block device only use the name.
3168 verify(nvlist_lookup_string(nv
, ZPOOL_CONFIG_TYPE
, &type
) == 0);
3169 if (strcmp(type
, VDEV_TYPE_DISK
) == 0) {
3170 path
= strrchr(path
, '/');
3175 * Remove the partition from the path it this is a whole disk.
3177 if (nvlist_lookup_uint64(nv
, ZPOOL_CONFIG_WHOLE_DISK
,
3178 &value
) == 0 && value
) {
3179 return strip_partition(hdl
, path
);
3182 verify(nvlist_lookup_string(nv
, ZPOOL_CONFIG_TYPE
, &path
) == 0);
3185 * If it's a raidz device, we need to stick in the parity level.
3187 if (strcmp(path
, VDEV_TYPE_RAIDZ
) == 0) {
3188 char tmpbuf
[PATH_BUF_LEN
];
3190 verify(nvlist_lookup_uint64(nv
, ZPOOL_CONFIG_NPARITY
,
3192 (void) snprintf(tmpbuf
, sizeof (tmpbuf
), "%s%llu", path
,
3193 (u_longlong_t
)value
);
3198 * We identify each top-level vdev by using a <type-id>
3199 * naming convention.
3204 verify(nvlist_lookup_uint64(nv
, ZPOOL_CONFIG_ID
,
3206 (void) snprintf(buf
, sizeof (buf
), "%s-%llu", path
,
3212 return (zfs_strdup(hdl
, path
));
3216 zbookmark_compare(const void *a
, const void *b
)
3218 return (memcmp(a
, b
, sizeof (zbookmark_t
)));
3222 * Retrieve the persistent error log, uniquify the members, and return to the
3226 zpool_get_errlog(zpool_handle_t
*zhp
, nvlist_t
**nverrlistp
)
3228 zfs_cmd_t zc
= { "\0", "\0", "\0", "\0", 0 };
3230 zbookmark_t
*zb
= NULL
;
3234 * Retrieve the raw error list from the kernel. If the number of errors
3235 * has increased, allocate more space and continue until we get the
3238 verify(nvlist_lookup_uint64(zhp
->zpool_config
, ZPOOL_CONFIG_ERRCOUNT
,
3242 if ((zc
.zc_nvlist_dst
= (uintptr_t)zfs_alloc(zhp
->zpool_hdl
,
3243 count
* sizeof (zbookmark_t
))) == (uintptr_t)NULL
)
3245 zc
.zc_nvlist_dst_size
= count
;
3246 (void) strcpy(zc
.zc_name
, zhp
->zpool_name
);
3248 if (ioctl(zhp
->zpool_hdl
->libzfs_fd
, ZFS_IOC_ERROR_LOG
,
3250 free((void *)(uintptr_t)zc
.zc_nvlist_dst
);
3251 if (errno
== ENOMEM
) {
3252 count
= zc
.zc_nvlist_dst_size
;
3253 if ((zc
.zc_nvlist_dst
= (uintptr_t)
3254 zfs_alloc(zhp
->zpool_hdl
, count
*
3255 sizeof (zbookmark_t
))) == (uintptr_t)NULL
)
3266 * Sort the resulting bookmarks. This is a little confusing due to the
3267 * implementation of ZFS_IOC_ERROR_LOG. The bookmarks are copied last
3268 * to first, and 'zc_nvlist_dst_size' indicates the number of boomarks
3269 * _not_ copied as part of the process. So we point the start of our
3270 * array appropriate and decrement the total number of elements.
3272 zb
= ((zbookmark_t
*)(uintptr_t)zc
.zc_nvlist_dst
) +
3273 zc
.zc_nvlist_dst_size
;
3274 count
-= zc
.zc_nvlist_dst_size
;
3276 qsort(zb
, count
, sizeof (zbookmark_t
), zbookmark_compare
);
3278 verify(nvlist_alloc(nverrlistp
, 0, KM_SLEEP
) == 0);
3281 * Fill in the nverrlistp with nvlist's of dataset and object numbers.
3283 for (i
= 0; i
< count
; i
++) {
3286 /* ignoring zb_blkid and zb_level for now */
3287 if (i
> 0 && zb
[i
-1].zb_objset
== zb
[i
].zb_objset
&&
3288 zb
[i
-1].zb_object
== zb
[i
].zb_object
)
3291 if (nvlist_alloc(&nv
, NV_UNIQUE_NAME
, KM_SLEEP
) != 0)
3293 if (nvlist_add_uint64(nv
, ZPOOL_ERR_DATASET
,
3294 zb
[i
].zb_objset
) != 0) {
3298 if (nvlist_add_uint64(nv
, ZPOOL_ERR_OBJECT
,
3299 zb
[i
].zb_object
) != 0) {
3303 if (nvlist_add_nvlist(*nverrlistp
, "ejk", nv
) != 0) {
3310 free((void *)(uintptr_t)zc
.zc_nvlist_dst
);
3314 free((void *)(uintptr_t)zc
.zc_nvlist_dst
);
3315 return (no_memory(zhp
->zpool_hdl
));
3319 * Upgrade a ZFS pool to the latest on-disk version.
3322 zpool_upgrade(zpool_handle_t
*zhp
, uint64_t new_version
)
3324 zfs_cmd_t zc
= { "\0", "\0", "\0", "\0", 0 };
3325 libzfs_handle_t
*hdl
= zhp
->zpool_hdl
;
3327 (void) strcpy(zc
.zc_name
, zhp
->zpool_name
);
3328 zc
.zc_cookie
= new_version
;
3330 if (zfs_ioctl(hdl
, ZFS_IOC_POOL_UPGRADE
, &zc
) != 0)
3331 return (zpool_standard_error_fmt(hdl
, errno
,
3332 dgettext(TEXT_DOMAIN
, "cannot upgrade '%s'"),
3338 zpool_set_history_str(const char *subcommand
, int argc
, char **argv
,
3343 (void) strlcpy(history_str
, subcommand
, HIS_MAX_RECORD_LEN
);
3344 for (i
= 1; i
< argc
; i
++) {
3345 if (strlen(history_str
) + 1 + strlen(argv
[i
]) >
3348 (void) strlcat(history_str
, " ", HIS_MAX_RECORD_LEN
);
3349 (void) strlcat(history_str
, argv
[i
], HIS_MAX_RECORD_LEN
);
3354 * Stage command history for logging.
3357 zpool_stage_history(libzfs_handle_t
*hdl
, const char *history_str
)
3359 if (history_str
== NULL
)
3362 if (strlen(history_str
) > HIS_MAX_RECORD_LEN
)
3365 if (hdl
->libzfs_log_str
!= NULL
)
3366 free(hdl
->libzfs_log_str
);
3368 if ((hdl
->libzfs_log_str
= strdup(history_str
)) == NULL
)
3369 return (no_memory(hdl
));
3375 * Perform ioctl to get some command history of a pool.
3377 * 'buf' is the buffer to fill up to 'len' bytes. 'off' is the
3378 * logical offset of the history buffer to start reading from.
3380 * Upon return, 'off' is the next logical offset to read from and
3381 * 'len' is the actual amount of bytes read into 'buf'.
3384 get_history(zpool_handle_t
*zhp
, char *buf
, uint64_t *off
, uint64_t *len
)
3386 zfs_cmd_t zc
= { "\0", "\0", "\0", "\0", 0 };
3387 libzfs_handle_t
*hdl
= zhp
->zpool_hdl
;
3389 (void) strlcpy(zc
.zc_name
, zhp
->zpool_name
, sizeof (zc
.zc_name
));
3391 zc
.zc_history
= (uint64_t)(uintptr_t)buf
;
3392 zc
.zc_history_len
= *len
;
3393 zc
.zc_history_offset
= *off
;
3395 if (ioctl(hdl
->libzfs_fd
, ZFS_IOC_POOL_GET_HISTORY
, &zc
) != 0) {
3398 return (zfs_error_fmt(hdl
, EZFS_PERM
,
3399 dgettext(TEXT_DOMAIN
,
3400 "cannot show history for pool '%s'"),
3403 return (zfs_error_fmt(hdl
, EZFS_NOHISTORY
,
3404 dgettext(TEXT_DOMAIN
, "cannot get history for pool "
3405 "'%s'"), zhp
->zpool_name
));
3407 return (zfs_error_fmt(hdl
, EZFS_BADVERSION
,
3408 dgettext(TEXT_DOMAIN
, "cannot get history for pool "
3409 "'%s', pool must be upgraded"), zhp
->zpool_name
));
3411 return (zpool_standard_error_fmt(hdl
, errno
,
3412 dgettext(TEXT_DOMAIN
,
3413 "cannot get history for '%s'"), zhp
->zpool_name
));
3417 *len
= zc
.zc_history_len
;
3418 *off
= zc
.zc_history_offset
;
3424 * Process the buffer of nvlists, unpacking and storing each nvlist record
3425 * into 'records'. 'leftover' is set to the number of bytes that weren't
3426 * processed as there wasn't a complete record.
3429 zpool_history_unpack(char *buf
, uint64_t bytes_read
, uint64_t *leftover
,
3430 nvlist_t
***records
, uint_t
*numrecords
)
3436 while (bytes_read
> sizeof (reclen
)) {
3438 /* get length of packed record (stored as little endian) */
3439 for (i
= 0, reclen
= 0; i
< sizeof (reclen
); i
++)
3440 reclen
+= (uint64_t)(((uchar_t
*)buf
)[i
]) << (8*i
);
3442 if (bytes_read
< sizeof (reclen
) + reclen
)
3446 if (nvlist_unpack(buf
+ sizeof (reclen
), reclen
, &nv
, 0) != 0)
3448 bytes_read
-= sizeof (reclen
) + reclen
;
3449 buf
+= sizeof (reclen
) + reclen
;
3451 /* add record to nvlist array */
3453 if (ISP2(*numrecords
+ 1)) {
3454 *records
= realloc(*records
,
3455 *numrecords
* 2 * sizeof (nvlist_t
*));
3457 (*records
)[*numrecords
- 1] = nv
;
3460 *leftover
= bytes_read
;
3464 #define HIS_BUF_LEN (128*1024)
3467 * Retrieve the command history of a pool.
3470 zpool_get_history(zpool_handle_t
*zhp
, nvlist_t
**nvhisp
)
3472 char buf
[HIS_BUF_LEN
];
3474 nvlist_t
**records
= NULL
;
3475 uint_t numrecords
= 0;
3479 uint64_t bytes_read
= sizeof (buf
);
3482 if ((err
= get_history(zhp
, buf
, &off
, &bytes_read
)) != 0)
3485 /* if nothing else was read in, we're at EOF, just return */
3489 if ((err
= zpool_history_unpack(buf
, bytes_read
,
3490 &leftover
, &records
, &numrecords
)) != 0)
3498 verify(nvlist_alloc(nvhisp
, NV_UNIQUE_NAME
, 0) == 0);
3499 verify(nvlist_add_nvlist_array(*nvhisp
, ZPOOL_HIST_RECORD
,
3500 records
, numrecords
) == 0);
3502 for (i
= 0; i
< numrecords
; i
++)
3503 nvlist_free(records
[i
]);
3510 * Retrieve the next event. If there is a new event available 'nvp' will
3511 * contain a newly allocated nvlist and 'dropped' will be set to the number
3512 * of missed events since the last call to this function. When 'nvp' is
3513 * set to NULL it indicates no new events are available. In either case
3514 * the function returns 0 and it is up to the caller to free 'nvp'. In
3515 * the case of a fatal error the function will return a non-zero value.
3516 * When the function is called in blocking mode it will not return until
3517 * a new event is available.
3520 zpool_events_next(libzfs_handle_t
*hdl
, nvlist_t
**nvp
,
3521 int *dropped
, int block
, int cleanup_fd
)
3523 zfs_cmd_t zc
= { "\0", "\0", "\0", "\0", 0 };
3528 zc
.zc_cleanup_fd
= cleanup_fd
;
3531 zc
.zc_guid
= ZEVENT_NONBLOCK
;
3533 if (zcmd_alloc_dst_nvlist(hdl
, &zc
, ZEVENT_SIZE
) != 0)
3537 if (zfs_ioctl(hdl
, ZFS_IOC_EVENTS_NEXT
, &zc
) != 0) {
3540 error
= zfs_error_fmt(hdl
, EZFS_POOLUNAVAIL
,
3541 dgettext(TEXT_DOMAIN
, "zfs shutdown"));
3544 /* Blocking error case should not occur */
3546 error
= zpool_standard_error_fmt(hdl
, errno
,
3547 dgettext(TEXT_DOMAIN
, "cannot get event"));
3551 if (zcmd_expand_dst_nvlist(hdl
, &zc
) != 0) {
3552 error
= zfs_error_fmt(hdl
, EZFS_NOMEM
,
3553 dgettext(TEXT_DOMAIN
, "cannot get event"));
3559 error
= zpool_standard_error_fmt(hdl
, errno
,
3560 dgettext(TEXT_DOMAIN
, "cannot get event"));
3565 error
= zcmd_read_dst_nvlist(hdl
, &zc
, nvp
);
3569 *dropped
= (int)zc
.zc_cookie
;
3571 zcmd_free_nvlists(&zc
);
3580 zpool_events_clear(libzfs_handle_t
*hdl
, int *count
)
3582 zfs_cmd_t zc
= { "\0", "\0", "\0", "\0", 0 };
3585 (void) snprintf(msg
, sizeof (msg
), dgettext(TEXT_DOMAIN
,
3586 "cannot clear events"));
3588 if (zfs_ioctl(hdl
, ZFS_IOC_EVENTS_CLEAR
, &zc
) != 0)
3589 return (zpool_standard_error_fmt(hdl
, errno
, msg
));
3592 *count
= (int)zc
.zc_cookie
; /* # of events cleared */
3598 zpool_obj_to_path(zpool_handle_t
*zhp
, uint64_t dsobj
, uint64_t obj
,
3599 char *pathname
, size_t len
)
3601 zfs_cmd_t zc
= { "\0", "\0", "\0", "\0", 0 };
3602 boolean_t mounted
= B_FALSE
;
3603 char *mntpnt
= NULL
;
3604 char dsname
[MAXNAMELEN
];
3607 /* special case for the MOS */
3608 (void) snprintf(pathname
, len
, "<metadata>:<0x%llx>", (longlong_t
)obj
);
3612 /* get the dataset's name */
3613 (void) strlcpy(zc
.zc_name
, zhp
->zpool_name
, sizeof (zc
.zc_name
));
3615 if (ioctl(zhp
->zpool_hdl
->libzfs_fd
,
3616 ZFS_IOC_DSOBJ_TO_DSNAME
, &zc
) != 0) {
3617 /* just write out a path of two object numbers */
3618 (void) snprintf(pathname
, len
, "<0x%llx>:<0x%llx>",
3619 (longlong_t
)dsobj
, (longlong_t
)obj
);
3622 (void) strlcpy(dsname
, zc
.zc_value
, sizeof (dsname
));
3624 /* find out if the dataset is mounted */
3625 mounted
= is_mounted(zhp
->zpool_hdl
, dsname
, &mntpnt
);
3627 /* get the corrupted object's path */
3628 (void) strlcpy(zc
.zc_name
, dsname
, sizeof (zc
.zc_name
));
3630 if (ioctl(zhp
->zpool_hdl
->libzfs_fd
, ZFS_IOC_OBJ_TO_PATH
,
3633 (void) snprintf(pathname
, len
, "%s%s", mntpnt
,
3636 (void) snprintf(pathname
, len
, "%s:%s",
3637 dsname
, zc
.zc_value
);
3640 (void) snprintf(pathname
, len
, "%s:<0x%llx>", dsname
, (longlong_t
)obj
);
3646 * Read the EFI label from the config, if a label does not exist then
3647 * pass back the error to the caller. If the caller has passed a non-NULL
3648 * diskaddr argument then we set it to the starting address of the EFI
3652 read_efi_label(nvlist_t
*config
, diskaddr_t
*sb
)
3656 char diskname
[MAXPATHLEN
];
3659 if (nvlist_lookup_string(config
, ZPOOL_CONFIG_PATH
, &path
) != 0)
3662 (void) snprintf(diskname
, sizeof (diskname
), "%s%s", RDISK_ROOT
,
3663 strrchr(path
, '/'));
3664 if ((fd
= open(diskname
, O_RDWR
|O_DIRECT
)) >= 0) {
3665 struct dk_gpt
*vtoc
;
3667 if ((err
= efi_alloc_and_read(fd
, &vtoc
)) >= 0) {
3669 *sb
= vtoc
->efi_parts
[0].p_start
;
3678 * determine where a partition starts on a disk in the current
3682 find_start_block(nvlist_t
*config
)
3686 diskaddr_t sb
= MAXOFFSET_T
;
3689 if (nvlist_lookup_nvlist_array(config
,
3690 ZPOOL_CONFIG_CHILDREN
, &child
, &children
) != 0) {
3691 if (nvlist_lookup_uint64(config
,
3692 ZPOOL_CONFIG_WHOLE_DISK
,
3693 &wholedisk
) != 0 || !wholedisk
) {
3694 return (MAXOFFSET_T
);
3696 if (read_efi_label(config
, &sb
) < 0)
3701 for (c
= 0; c
< children
; c
++) {
3702 sb
= find_start_block(child
[c
]);
3703 if (sb
!= MAXOFFSET_T
) {
3707 return (MAXOFFSET_T
);
3711 zpool_label_disk_wait(char *path
, int timeout
)
3713 struct stat64 statbuf
;
3717 * Wait timeout miliseconds for a newly created device to be available
3718 * from the given path. There is a small window when a /dev/ device
3719 * will exist and the udev link will not, so we must wait for the
3720 * symlink. Depending on the udev rules this may take a few seconds.
3722 for (i
= 0; i
< timeout
; i
++) {
3726 if ((stat64(path
, &statbuf
) == 0) && (errno
== 0))
3734 zpool_label_disk_check(char *path
)
3736 struct dk_gpt
*vtoc
;
3739 if ((fd
= open(path
, O_RDWR
|O_DIRECT
)) < 0)
3742 if ((err
= efi_alloc_and_read(fd
, &vtoc
)) != 0) {
3747 if (vtoc
->efi_flags
& EFI_GPT_PRIMARY_CORRUPT
) {
3759 * Label an individual disk. The name provided is the short name,
3760 * stripped of any leading /dev path.
3763 zpool_label_disk(libzfs_handle_t
*hdl
, zpool_handle_t
*zhp
, char *name
)
3765 char path
[MAXPATHLEN
];
3766 struct dk_gpt
*vtoc
;
3768 size_t resv
= EFI_MIN_RESV_SIZE
;
3769 uint64_t slice_size
;
3770 diskaddr_t start_block
;
3773 /* prepare an error message just in case */
3774 (void) snprintf(errbuf
, sizeof (errbuf
),
3775 dgettext(TEXT_DOMAIN
, "cannot label '%s'"), name
);
3780 if (pool_is_bootable(zhp
)) {
3781 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
3782 "EFI labeled devices are not supported on root "
3784 return (zfs_error(hdl
, EZFS_POOL_NOTSUP
, errbuf
));
3787 verify(nvlist_lookup_nvlist(zhp
->zpool_config
,
3788 ZPOOL_CONFIG_VDEV_TREE
, &nvroot
) == 0);
3790 if (zhp
->zpool_start_block
== 0)
3791 start_block
= find_start_block(nvroot
);
3793 start_block
= zhp
->zpool_start_block
;
3794 zhp
->zpool_start_block
= start_block
;
3797 start_block
= NEW_START_BLOCK
;
3800 (void) snprintf(path
, sizeof (path
), "%s/%s%s", RDISK_ROOT
, name
,
3803 if ((fd
= open(path
, O_RDWR
|O_DIRECT
)) < 0) {
3805 * This shouldn't happen. We've long since verified that this
3806 * is a valid device.
3808 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
, "cannot "
3809 "label '%s': unable to open device: %d"), path
, errno
);
3810 return (zfs_error(hdl
, EZFS_OPENFAILED
, errbuf
));
3813 if (efi_alloc_and_init(fd
, EFI_NUMPAR
, &vtoc
) != 0) {
3815 * The only way this can fail is if we run out of memory, or we
3816 * were unable to read the disk's capacity
3818 if (errno
== ENOMEM
)
3819 (void) no_memory(hdl
);
3822 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
, "cannot "
3823 "label '%s': unable to read disk capacity"), path
);
3825 return (zfs_error(hdl
, EZFS_NOCAP
, errbuf
));
3828 slice_size
= vtoc
->efi_last_u_lba
+ 1;
3829 slice_size
-= EFI_MIN_RESV_SIZE
;
3830 if (start_block
== MAXOFFSET_T
)
3831 start_block
= NEW_START_BLOCK
;
3832 slice_size
-= start_block
;
3833 slice_size
= P2ALIGN(slice_size
, PARTITION_END_ALIGNMENT
);
3835 vtoc
->efi_parts
[0].p_start
= start_block
;
3836 vtoc
->efi_parts
[0].p_size
= slice_size
;
3839 * Why we use V_USR: V_BACKUP confuses users, and is considered
3840 * disposable by some EFI utilities (since EFI doesn't have a backup
3841 * slice). V_UNASSIGNED is supposed to be used only for zero size
3842 * partitions, and efi_write() will fail if we use it. V_ROOT, V_BOOT,
3843 * etc. were all pretty specific. V_USR is as close to reality as we
3844 * can get, in the absence of V_OTHER.
3846 vtoc
->efi_parts
[0].p_tag
= V_USR
;
3847 (void) strcpy(vtoc
->efi_parts
[0].p_name
, "zfs");
3849 vtoc
->efi_parts
[8].p_start
= slice_size
+ start_block
;
3850 vtoc
->efi_parts
[8].p_size
= resv
;
3851 vtoc
->efi_parts
[8].p_tag
= V_RESERVED
;
3853 if ((rval
= efi_write(fd
, vtoc
)) != 0 || (rval
= efi_rescan(fd
)) != 0) {
3855 * Some block drivers (like pcata) may not support EFI
3856 * GPT labels. Print out a helpful error message dir-
3857 * ecting the user to manually label the disk and give
3863 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
, "try using "
3864 "parted(8) and then provide a specific slice: %d"), rval
);
3865 return (zfs_error(hdl
, EZFS_LABELFAILED
, errbuf
));
3871 /* Wait for the first expected slice to appear. */
3872 (void) snprintf(path
, sizeof (path
), "%s/%s%s%s", DISK_ROOT
, name
,
3873 isdigit(name
[strlen(name
)-1]) ? "p" : "", FIRST_SLICE
);
3874 rval
= zpool_label_disk_wait(path
, 3000);
3876 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
, "failed to "
3877 "detect device partitions on '%s': %d"), path
, rval
);
3878 return (zfs_error(hdl
, EZFS_LABELFAILED
, errbuf
));
3881 /* We can't be to paranoid. Read the label back and verify it. */
3882 (void) snprintf(path
, sizeof (path
), "%s/%s", DISK_ROOT
, name
);
3883 rval
= zpool_label_disk_check(path
);
3885 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
, "freshly written "
3886 "EFI label on '%s' is damaged. Ensure\nthis device "
3887 "is not in in use, and is functioning properly: %d"),
3889 return (zfs_error(hdl
, EZFS_LABELFAILED
, errbuf
));