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 (c) 2011, 2017 by Delphix. All rights reserved.
25 * Copyright (c) 2013, Joyent, Inc. All rights reserved.
26 * Copyright 2016 Nexenta Systems, Inc.
27 * Copyright 2016 Igor Kozhukhov <ikozhukhov@gmail.com>
28 * Copyright (c) 2017 Datto Inc.
41 #include <sys/efi_partition.h>
43 #include <sys/zfs_ioctl.h>
46 #include "zfs_namecheck.h"
48 #include "libzfs_impl.h"
49 #include "zfs_comutil.h"
50 #include "zfeature_common.h"
52 static int read_efi_label(nvlist_t
*, diskaddr_t
*, boolean_t
*);
53 static boolean_t
zpool_vdev_is_interior(const char *name
);
55 #define BACKUP_SLICE "s2"
57 typedef struct prop_flags
{
58 int create
:1; /* Validate property on creation */
59 int import
:1; /* Validate property on import */
63 * ====================================================================
64 * zpool property functions
65 * ====================================================================
69 zpool_get_all_props(zpool_handle_t
*zhp
)
72 libzfs_handle_t
*hdl
= zhp
->zpool_hdl
;
74 (void) strlcpy(zc
.zc_name
, zhp
->zpool_name
, sizeof (zc
.zc_name
));
76 if (zcmd_alloc_dst_nvlist(hdl
, &zc
, 0) != 0)
79 while (ioctl(hdl
->libzfs_fd
, ZFS_IOC_POOL_GET_PROPS
, &zc
) != 0) {
80 if (errno
== ENOMEM
) {
81 if (zcmd_expand_dst_nvlist(hdl
, &zc
) != 0) {
82 zcmd_free_nvlists(&zc
);
86 zcmd_free_nvlists(&zc
);
91 if (zcmd_read_dst_nvlist(hdl
, &zc
, &zhp
->zpool_props
) != 0) {
92 zcmd_free_nvlists(&zc
);
96 zcmd_free_nvlists(&zc
);
102 zpool_props_refresh(zpool_handle_t
*zhp
)
106 old_props
= zhp
->zpool_props
;
108 if (zpool_get_all_props(zhp
) != 0)
111 nvlist_free(old_props
);
116 zpool_get_prop_string(zpool_handle_t
*zhp
, zpool_prop_t prop
,
122 zprop_source_t source
;
124 nvl
= zhp
->zpool_props
;
125 if (nvlist_lookup_nvlist(nvl
, zpool_prop_to_name(prop
), &nv
) == 0) {
126 verify(nvlist_lookup_uint64(nv
, ZPROP_SOURCE
, &ival
) == 0);
128 verify(nvlist_lookup_string(nv
, ZPROP_VALUE
, &value
) == 0);
130 source
= ZPROP_SRC_DEFAULT
;
131 if ((value
= (char *)zpool_prop_default_string(prop
)) == NULL
)
142 zpool_get_prop_int(zpool_handle_t
*zhp
, zpool_prop_t prop
, zprop_source_t
*src
)
146 zprop_source_t source
;
148 if (zhp
->zpool_props
== NULL
&& zpool_get_all_props(zhp
)) {
150 * zpool_get_all_props() has most likely failed because
151 * the pool is faulted, but if all we need is the top level
152 * vdev's guid then get it from the zhp config nvlist.
154 if ((prop
== ZPOOL_PROP_GUID
) &&
155 (nvlist_lookup_nvlist(zhp
->zpool_config
,
156 ZPOOL_CONFIG_VDEV_TREE
, &nv
) == 0) &&
157 (nvlist_lookup_uint64(nv
, ZPOOL_CONFIG_GUID
, &value
)
161 return (zpool_prop_default_numeric(prop
));
164 nvl
= zhp
->zpool_props
;
165 if (nvlist_lookup_nvlist(nvl
, zpool_prop_to_name(prop
), &nv
) == 0) {
166 verify(nvlist_lookup_uint64(nv
, ZPROP_SOURCE
, &value
) == 0);
168 verify(nvlist_lookup_uint64(nv
, ZPROP_VALUE
, &value
) == 0);
170 source
= ZPROP_SRC_DEFAULT
;
171 value
= zpool_prop_default_numeric(prop
);
181 * Map VDEV STATE to printed strings.
184 zpool_state_to_name(vdev_state_t state
, vdev_aux_t aux
)
187 case VDEV_STATE_CLOSED
:
188 case VDEV_STATE_OFFLINE
:
189 return (gettext("OFFLINE"));
190 case VDEV_STATE_REMOVED
:
191 return (gettext("REMOVED"));
192 case VDEV_STATE_CANT_OPEN
:
193 if (aux
== VDEV_AUX_CORRUPT_DATA
|| aux
== VDEV_AUX_BAD_LOG
)
194 return (gettext("FAULTED"));
195 else if (aux
== VDEV_AUX_SPLIT_POOL
)
196 return (gettext("SPLIT"));
198 return (gettext("UNAVAIL"));
199 case VDEV_STATE_FAULTED
:
200 return (gettext("FAULTED"));
201 case VDEV_STATE_DEGRADED
:
202 return (gettext("DEGRADED"));
203 case VDEV_STATE_HEALTHY
:
204 return (gettext("ONLINE"));
210 return (gettext("UNKNOWN"));
214 * Map POOL STATE to printed strings.
217 zpool_pool_state_to_name(pool_state_t state
)
220 case POOL_STATE_ACTIVE
:
221 return (gettext("ACTIVE"));
222 case POOL_STATE_EXPORTED
:
223 return (gettext("EXPORTED"));
224 case POOL_STATE_DESTROYED
:
225 return (gettext("DESTROYED"));
226 case POOL_STATE_SPARE
:
227 return (gettext("SPARE"));
228 case POOL_STATE_L2CACHE
:
229 return (gettext("L2CACHE"));
230 case POOL_STATE_UNINITIALIZED
:
231 return (gettext("UNINITIALIZED"));
232 case POOL_STATE_UNAVAIL
:
233 return (gettext("UNAVAIL"));
234 case POOL_STATE_POTENTIALLY_ACTIVE
:
235 return (gettext("POTENTIALLY_ACTIVE"));
238 return (gettext("UNKNOWN"));
242 * Get a zpool property value for 'prop' and return the value in
243 * a pre-allocated buffer.
246 zpool_get_prop(zpool_handle_t
*zhp
, zpool_prop_t prop
, char *buf
, size_t len
,
247 zprop_source_t
*srctype
, boolean_t literal
)
251 zprop_source_t src
= ZPROP_SRC_NONE
;
256 if (zpool_get_state(zhp
) == POOL_STATE_UNAVAIL
) {
258 case ZPOOL_PROP_NAME
:
259 (void) strlcpy(buf
, zpool_get_name(zhp
), len
);
262 case ZPOOL_PROP_HEALTH
:
263 (void) strlcpy(buf
, "FAULTED", len
);
266 case ZPOOL_PROP_GUID
:
267 intval
= zpool_get_prop_int(zhp
, prop
, &src
);
268 (void) snprintf(buf
, len
, "%llu", intval
);
271 case ZPOOL_PROP_ALTROOT
:
272 case ZPOOL_PROP_CACHEFILE
:
273 case ZPOOL_PROP_COMMENT
:
274 if (zhp
->zpool_props
!= NULL
||
275 zpool_get_all_props(zhp
) == 0) {
277 zpool_get_prop_string(zhp
, prop
, &src
),
283 (void) strlcpy(buf
, "-", len
);
292 if (zhp
->zpool_props
== NULL
&& zpool_get_all_props(zhp
) &&
293 prop
!= ZPOOL_PROP_NAME
)
296 switch (zpool_prop_get_type(prop
)) {
297 case PROP_TYPE_STRING
:
298 (void) strlcpy(buf
, zpool_get_prop_string(zhp
, prop
, &src
),
302 case PROP_TYPE_NUMBER
:
303 intval
= zpool_get_prop_int(zhp
, prop
, &src
);
306 case ZPOOL_PROP_SIZE
:
307 case ZPOOL_PROP_ALLOCATED
:
308 case ZPOOL_PROP_FREE
:
309 case ZPOOL_PROP_FREEING
:
310 case ZPOOL_PROP_LEAKED
:
312 (void) snprintf(buf
, len
, "%llu",
313 (u_longlong_t
)intval
);
315 (void) zfs_nicenum(intval
, buf
, len
);
318 case ZPOOL_PROP_BOOTSIZE
:
319 case ZPOOL_PROP_EXPANDSZ
:
321 (void) strlcpy(buf
, "-", len
);
322 } else if (literal
) {
323 (void) snprintf(buf
, len
, "%llu",
324 (u_longlong_t
)intval
);
326 (void) zfs_nicenum(intval
, buf
, len
);
329 case ZPOOL_PROP_CAPACITY
:
331 (void) snprintf(buf
, len
, "%llu",
332 (u_longlong_t
)intval
);
334 (void) snprintf(buf
, len
, "%llu%%",
335 (u_longlong_t
)intval
);
338 case ZPOOL_PROP_FRAGMENTATION
:
339 if (intval
== UINT64_MAX
) {
340 (void) strlcpy(buf
, "-", len
);
342 (void) snprintf(buf
, len
, "%llu%%",
343 (u_longlong_t
)intval
);
346 case ZPOOL_PROP_DEDUPRATIO
:
347 (void) snprintf(buf
, len
, "%llu.%02llux",
348 (u_longlong_t
)(intval
/ 100),
349 (u_longlong_t
)(intval
% 100));
351 case ZPOOL_PROP_HEALTH
:
352 verify(nvlist_lookup_nvlist(zpool_get_config(zhp
, NULL
),
353 ZPOOL_CONFIG_VDEV_TREE
, &nvroot
) == 0);
354 verify(nvlist_lookup_uint64_array(nvroot
,
355 ZPOOL_CONFIG_VDEV_STATS
, (uint64_t **)&vs
, &vsc
)
358 (void) strlcpy(buf
, zpool_state_to_name(intval
,
361 case ZPOOL_PROP_VERSION
:
362 if (intval
>= SPA_VERSION_FEATURES
) {
363 (void) snprintf(buf
, len
, "-");
368 (void) snprintf(buf
, len
, "%llu", intval
);
372 case PROP_TYPE_INDEX
:
373 intval
= zpool_get_prop_int(zhp
, prop
, &src
);
374 if (zpool_prop_index_to_string(prop
, intval
, &strval
)
377 (void) strlcpy(buf
, strval
, len
);
391 * Check if the bootfs name has the same pool name as it is set to.
392 * Assuming bootfs is a valid dataset name.
395 bootfs_name_valid(const char *pool
, char *bootfs
)
397 int len
= strlen(pool
);
399 if (!zfs_name_valid(bootfs
, ZFS_TYPE_FILESYSTEM
|ZFS_TYPE_SNAPSHOT
))
402 if (strncmp(pool
, bootfs
, len
) == 0 &&
403 (bootfs
[len
] == '/' || bootfs
[len
] == '\0'))
410 zpool_is_bootable(zpool_handle_t
*zhp
)
412 char bootfs
[ZFS_MAX_DATASET_NAME_LEN
];
414 return (zpool_get_prop(zhp
, ZPOOL_PROP_BOOTFS
, bootfs
,
415 sizeof (bootfs
), NULL
, B_FALSE
) == 0 && strncmp(bootfs
, "-",
416 sizeof (bootfs
)) != 0);
421 * Given an nvlist of zpool properties to be set, validate that they are
422 * correct, and parse any numeric properties (index, boolean, etc) if they are
423 * specified as strings.
426 zpool_valid_proplist(libzfs_handle_t
*hdl
, const char *poolname
,
427 nvlist_t
*props
, uint64_t version
, prop_flags_t flags
, char *errbuf
)
435 struct stat64 statbuf
;
438 if (nvlist_alloc(&retprops
, NV_UNIQUE_NAME
, 0) != 0) {
439 (void) no_memory(hdl
);
444 while ((elem
= nvlist_next_nvpair(props
, elem
)) != NULL
) {
445 const char *propname
= nvpair_name(elem
);
447 prop
= zpool_name_to_prop(propname
);
448 if (prop
== ZPOOL_PROP_INVAL
&& zpool_prop_feature(propname
)) {
450 char *fname
= strchr(propname
, '@') + 1;
452 err
= zfeature_lookup_name(fname
, NULL
);
454 ASSERT3U(err
, ==, ENOENT
);
455 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
456 "invalid feature '%s'"), fname
);
457 (void) zfs_error(hdl
, EZFS_BADPROP
, errbuf
);
461 if (nvpair_type(elem
) != DATA_TYPE_STRING
) {
462 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
463 "'%s' must be a string"), propname
);
464 (void) zfs_error(hdl
, EZFS_BADPROP
, errbuf
);
468 (void) nvpair_value_string(elem
, &strval
);
469 if (strcmp(strval
, ZFS_FEATURE_ENABLED
) != 0) {
470 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
471 "property '%s' can only be set to "
472 "'enabled'"), propname
);
473 (void) zfs_error(hdl
, EZFS_BADPROP
, errbuf
);
477 if (nvlist_add_uint64(retprops
, propname
, 0) != 0) {
478 (void) no_memory(hdl
);
485 * Make sure this property is valid and applies to this type.
487 if (prop
== ZPOOL_PROP_INVAL
) {
488 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
489 "invalid property '%s'"), propname
);
490 (void) zfs_error(hdl
, EZFS_BADPROP
, errbuf
);
494 if (zpool_prop_readonly(prop
)) {
495 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
, "'%s' "
496 "is readonly"), propname
);
497 (void) zfs_error(hdl
, EZFS_PROPREADONLY
, errbuf
);
501 if (zprop_parse_value(hdl
, elem
, prop
, ZFS_TYPE_POOL
, retprops
,
502 &strval
, &intval
, errbuf
) != 0)
506 * Perform additional checking for specific properties.
509 case ZPOOL_PROP_VERSION
:
510 if (intval
< version
||
511 !SPA_VERSION_IS_SUPPORTED(intval
)) {
512 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
513 "property '%s' number %d is invalid."),
515 (void) zfs_error(hdl
, EZFS_BADVERSION
, errbuf
);
520 case ZPOOL_PROP_BOOTSIZE
:
522 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
523 "property '%s' can only be set during pool "
524 "creation"), propname
);
525 (void) zfs_error(hdl
, EZFS_BADPROP
, errbuf
);
530 case ZPOOL_PROP_BOOTFS
:
531 if (flags
.create
|| flags
.import
) {
532 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
533 "property '%s' cannot be set at creation "
534 "or import time"), propname
);
535 (void) zfs_error(hdl
, EZFS_BADPROP
, errbuf
);
539 if (version
< SPA_VERSION_BOOTFS
) {
540 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
541 "pool must be upgraded to support "
542 "'%s' property"), propname
);
543 (void) zfs_error(hdl
, EZFS_BADVERSION
, errbuf
);
548 * bootfs property value has to be a dataset name and
549 * the dataset has to be in the same pool as it sets to.
551 if (strval
[0] != '\0' && !bootfs_name_valid(poolname
,
553 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
, "'%s' "
554 "is an invalid name"), strval
);
555 (void) zfs_error(hdl
, EZFS_INVALIDNAME
, errbuf
);
559 if ((zhp
= zpool_open_canfail(hdl
, poolname
)) == NULL
) {
560 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
561 "could not open pool '%s'"), poolname
);
562 (void) zfs_error(hdl
, EZFS_OPENFAILED
, errbuf
);
568 case ZPOOL_PROP_ALTROOT
:
569 if (!flags
.create
&& !flags
.import
) {
570 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
571 "property '%s' can only be set during pool "
572 "creation or import"), propname
);
573 (void) zfs_error(hdl
, EZFS_BADPROP
, errbuf
);
577 if (strval
[0] != '/') {
578 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
579 "bad alternate root '%s'"), strval
);
580 (void) zfs_error(hdl
, EZFS_BADPATH
, errbuf
);
585 case ZPOOL_PROP_CACHEFILE
:
586 if (strval
[0] == '\0')
589 if (strcmp(strval
, "none") == 0)
592 if (strval
[0] != '/') {
593 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
594 "property '%s' must be empty, an "
595 "absolute path, or 'none'"), propname
);
596 (void) zfs_error(hdl
, EZFS_BADPATH
, errbuf
);
600 slash
= strrchr(strval
, '/');
602 if (slash
[1] == '\0' || strcmp(slash
, "/.") == 0 ||
603 strcmp(slash
, "/..") == 0) {
604 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
605 "'%s' is not a valid file"), strval
);
606 (void) zfs_error(hdl
, EZFS_BADPATH
, errbuf
);
612 if (strval
[0] != '\0' &&
613 (stat64(strval
, &statbuf
) != 0 ||
614 !S_ISDIR(statbuf
.st_mode
))) {
615 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
616 "'%s' is not a valid directory"),
618 (void) zfs_error(hdl
, EZFS_BADPATH
, errbuf
);
625 case ZPOOL_PROP_COMMENT
:
626 for (check
= strval
; *check
!= '\0'; check
++) {
627 if (!isprint(*check
)) {
629 dgettext(TEXT_DOMAIN
,
630 "comment may only have printable "
632 (void) zfs_error(hdl
, EZFS_BADPROP
,
637 if (strlen(strval
) > ZPROP_MAX_COMMENT
) {
638 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
639 "comment must not exceed %d characters"),
641 (void) zfs_error(hdl
, EZFS_BADPROP
, errbuf
);
645 case ZPOOL_PROP_READONLY
:
647 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
648 "property '%s' can only be set at "
649 "import time"), propname
);
650 (void) zfs_error(hdl
, EZFS_BADPROP
, errbuf
);
656 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
657 "property '%s'(%d) not defined"), propname
, prop
);
664 nvlist_free(retprops
);
669 * Set zpool property : propname=propval.
672 zpool_set_prop(zpool_handle_t
*zhp
, const char *propname
, const char *propval
)
674 zfs_cmd_t zc
= { 0 };
677 nvlist_t
*nvl
= NULL
;
680 prop_flags_t flags
= { 0 };
682 (void) snprintf(errbuf
, sizeof (errbuf
),
683 dgettext(TEXT_DOMAIN
, "cannot set property for '%s'"),
686 if (nvlist_alloc(&nvl
, NV_UNIQUE_NAME
, 0) != 0)
687 return (no_memory(zhp
->zpool_hdl
));
689 if (nvlist_add_string(nvl
, propname
, propval
) != 0) {
691 return (no_memory(zhp
->zpool_hdl
));
694 version
= zpool_get_prop_int(zhp
, ZPOOL_PROP_VERSION
, NULL
);
695 if ((realprops
= zpool_valid_proplist(zhp
->zpool_hdl
,
696 zhp
->zpool_name
, nvl
, version
, flags
, errbuf
)) == NULL
) {
705 * Execute the corresponding ioctl() to set this property.
707 (void) strlcpy(zc
.zc_name
, zhp
->zpool_name
, sizeof (zc
.zc_name
));
709 if (zcmd_write_src_nvlist(zhp
->zpool_hdl
, &zc
, nvl
) != 0) {
714 ret
= zfs_ioctl(zhp
->zpool_hdl
, ZFS_IOC_POOL_SET_PROPS
, &zc
);
716 zcmd_free_nvlists(&zc
);
720 (void) zpool_standard_error(zhp
->zpool_hdl
, errno
, errbuf
);
722 (void) zpool_props_refresh(zhp
);
728 zpool_expand_proplist(zpool_handle_t
*zhp
, zprop_list_t
**plp
)
730 libzfs_handle_t
*hdl
= zhp
->zpool_hdl
;
732 char buf
[ZFS_MAXPROPLEN
];
733 nvlist_t
*features
= NULL
;
735 boolean_t firstexpand
= (NULL
== *plp
);
737 if (zprop_expand_list(hdl
, plp
, ZFS_TYPE_POOL
) != 0)
741 while (*last
!= NULL
)
742 last
= &(*last
)->pl_next
;
745 features
= zpool_get_features(zhp
);
747 if ((*plp
)->pl_all
&& firstexpand
) {
748 for (int i
= 0; i
< SPA_FEATURES
; i
++) {
749 zprop_list_t
*entry
= zfs_alloc(hdl
,
750 sizeof (zprop_list_t
));
751 entry
->pl_prop
= ZPROP_INVAL
;
752 entry
->pl_user_prop
= zfs_asprintf(hdl
, "feature@%s",
753 spa_feature_table
[i
].fi_uname
);
754 entry
->pl_width
= strlen(entry
->pl_user_prop
);
755 entry
->pl_all
= B_TRUE
;
758 last
= &entry
->pl_next
;
762 /* add any unsupported features */
763 for (nvpair_t
*nvp
= nvlist_next_nvpair(features
, NULL
);
764 nvp
!= NULL
; nvp
= nvlist_next_nvpair(features
, nvp
)) {
769 if (zfeature_is_supported(nvpair_name(nvp
)))
772 propname
= zfs_asprintf(hdl
, "unsupported@%s",
776 * Before adding the property to the list make sure that no
777 * other pool already added the same property.
781 while (entry
!= NULL
) {
782 if (entry
->pl_user_prop
!= NULL
&&
783 strcmp(propname
, entry
->pl_user_prop
) == 0) {
787 entry
= entry
->pl_next
;
794 entry
= zfs_alloc(hdl
, sizeof (zprop_list_t
));
795 entry
->pl_prop
= ZPROP_INVAL
;
796 entry
->pl_user_prop
= propname
;
797 entry
->pl_width
= strlen(entry
->pl_user_prop
);
798 entry
->pl_all
= B_TRUE
;
801 last
= &entry
->pl_next
;
804 for (entry
= *plp
; entry
!= NULL
; entry
= entry
->pl_next
) {
809 if (entry
->pl_prop
!= ZPROP_INVAL
&&
810 zpool_get_prop(zhp
, entry
->pl_prop
, buf
, sizeof (buf
),
811 NULL
, B_FALSE
) == 0) {
812 if (strlen(buf
) > entry
->pl_width
)
813 entry
->pl_width
= strlen(buf
);
821 * Get the state for the given feature on the given ZFS pool.
824 zpool_prop_get_feature(zpool_handle_t
*zhp
, const char *propname
, char *buf
,
828 boolean_t found
= B_FALSE
;
829 nvlist_t
*features
= zpool_get_features(zhp
);
831 const char *feature
= strchr(propname
, '@') + 1;
833 supported
= zpool_prop_feature(propname
);
834 ASSERT(supported
|| zpool_prop_unsupported(propname
));
837 * Convert from feature name to feature guid. This conversion is
838 * unecessary for unsupported@... properties because they already
845 ret
= zfeature_lookup_name(feature
, &fid
);
847 (void) strlcpy(buf
, "-", len
);
850 feature
= spa_feature_table
[fid
].fi_guid
;
853 if (nvlist_lookup_uint64(features
, feature
, &refcount
) == 0)
858 (void) strlcpy(buf
, ZFS_FEATURE_DISABLED
, len
);
861 (void) strlcpy(buf
, ZFS_FEATURE_ENABLED
, len
);
863 (void) strlcpy(buf
, ZFS_FEATURE_ACTIVE
, len
);
868 (void) strcpy(buf
, ZFS_UNSUPPORTED_INACTIVE
);
870 (void) strcpy(buf
, ZFS_UNSUPPORTED_READONLY
);
873 (void) strlcpy(buf
, "-", len
);
882 * Don't start the slice at the default block of 34; many storage
883 * devices will use a stripe width of 128k, so start there instead.
885 #define NEW_START_BLOCK 256
888 * Validate the given pool name, optionally putting an extended error message in
892 zpool_name_valid(libzfs_handle_t
*hdl
, boolean_t isopen
, const char *pool
)
898 ret
= pool_namecheck(pool
, &why
, &what
);
901 * The rules for reserved pool names were extended at a later point.
902 * But we need to support users with existing pools that may now be
903 * invalid. So we only check for this expanded set of names during a
904 * create (or import), and only in userland.
906 if (ret
== 0 && !isopen
&&
907 (strncmp(pool
, "mirror", 6) == 0 ||
908 strncmp(pool
, "raidz", 5) == 0 ||
909 strncmp(pool
, "spare", 5) == 0 ||
910 strcmp(pool
, "log") == 0)) {
913 dgettext(TEXT_DOMAIN
, "name is reserved"));
921 case NAME_ERR_TOOLONG
:
923 dgettext(TEXT_DOMAIN
, "name is too long"));
926 case NAME_ERR_INVALCHAR
:
928 dgettext(TEXT_DOMAIN
, "invalid character "
929 "'%c' in pool name"), what
);
932 case NAME_ERR_NOLETTER
:
933 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
934 "name must begin with a letter"));
937 case NAME_ERR_RESERVED
:
938 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
939 "name is reserved"));
942 case NAME_ERR_DISKLIKE
:
943 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
944 "pool name is reserved"));
947 case NAME_ERR_LEADING_SLASH
:
948 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
949 "leading slash in name"));
952 case NAME_ERR_EMPTY_COMPONENT
:
953 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
954 "empty component in name"));
957 case NAME_ERR_TRAILING_SLASH
:
958 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
959 "trailing slash in name"));
962 case NAME_ERR_MULTIPLE_DELIMITERS
:
963 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
964 "multiple '@' and/or '#' delimiters in "
969 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
970 "(%d) not defined"), why
);
981 * Open a handle to the given pool, even if the pool is currently in the FAULTED
985 zpool_open_canfail(libzfs_handle_t
*hdl
, const char *pool
)
991 * Make sure the pool name is valid.
993 if (!zpool_name_valid(hdl
, B_TRUE
, pool
)) {
994 (void) zfs_error_fmt(hdl
, EZFS_INVALIDNAME
,
995 dgettext(TEXT_DOMAIN
, "cannot open '%s'"),
1000 if ((zhp
= zfs_alloc(hdl
, sizeof (zpool_handle_t
))) == NULL
)
1003 zhp
->zpool_hdl
= hdl
;
1004 (void) strlcpy(zhp
->zpool_name
, pool
, sizeof (zhp
->zpool_name
));
1006 if (zpool_refresh_stats(zhp
, &missing
) != 0) {
1012 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
, "no such pool"));
1013 (void) zfs_error_fmt(hdl
, EZFS_NOENT
,
1014 dgettext(TEXT_DOMAIN
, "cannot open '%s'"), pool
);
1023 * Like the above, but silent on error. Used when iterating over pools (because
1024 * the configuration cache may be out of date).
1027 zpool_open_silent(libzfs_handle_t
*hdl
, const char *pool
, zpool_handle_t
**ret
)
1029 zpool_handle_t
*zhp
;
1032 if ((zhp
= zfs_alloc(hdl
, sizeof (zpool_handle_t
))) == NULL
)
1035 zhp
->zpool_hdl
= hdl
;
1036 (void) strlcpy(zhp
->zpool_name
, pool
, sizeof (zhp
->zpool_name
));
1038 if (zpool_refresh_stats(zhp
, &missing
) != 0) {
1054 * Similar to zpool_open_canfail(), but refuses to open pools in the faulted
1058 zpool_open(libzfs_handle_t
*hdl
, const char *pool
)
1060 zpool_handle_t
*zhp
;
1062 if ((zhp
= zpool_open_canfail(hdl
, pool
)) == NULL
)
1065 if (zhp
->zpool_state
== POOL_STATE_UNAVAIL
) {
1066 (void) zfs_error_fmt(hdl
, EZFS_POOLUNAVAIL
,
1067 dgettext(TEXT_DOMAIN
, "cannot open '%s'"), zhp
->zpool_name
);
1076 * Close the handle. Simply frees the memory associated with the handle.
1079 zpool_close(zpool_handle_t
*zhp
)
1081 nvlist_free(zhp
->zpool_config
);
1082 nvlist_free(zhp
->zpool_old_config
);
1083 nvlist_free(zhp
->zpool_props
);
1088 * Return the name of the pool.
1091 zpool_get_name(zpool_handle_t
*zhp
)
1093 return (zhp
->zpool_name
);
1098 * Return the state of the pool (ACTIVE or UNAVAILABLE)
1101 zpool_get_state(zpool_handle_t
*zhp
)
1103 return (zhp
->zpool_state
);
1107 * Create the named pool, using the provided vdev list. It is assumed
1108 * that the consumer has already validated the contents of the nvlist, so we
1109 * don't have to worry about error semantics.
1112 zpool_create(libzfs_handle_t
*hdl
, const char *pool
, nvlist_t
*nvroot
,
1113 nvlist_t
*props
, nvlist_t
*fsprops
)
1115 zfs_cmd_t zc
= { 0 };
1116 nvlist_t
*zc_fsprops
= NULL
;
1117 nvlist_t
*zc_props
= NULL
;
1121 (void) snprintf(msg
, sizeof (msg
), dgettext(TEXT_DOMAIN
,
1122 "cannot create '%s'"), pool
);
1124 if (!zpool_name_valid(hdl
, B_FALSE
, pool
))
1125 return (zfs_error(hdl
, EZFS_INVALIDNAME
, msg
));
1127 if (zcmd_write_conf_nvlist(hdl
, &zc
, nvroot
) != 0)
1131 prop_flags_t flags
= { .create
= B_TRUE
, .import
= B_FALSE
};
1133 if ((zc_props
= zpool_valid_proplist(hdl
, pool
, props
,
1134 SPA_VERSION_1
, flags
, msg
)) == NULL
) {
1143 zoned
= ((nvlist_lookup_string(fsprops
,
1144 zfs_prop_to_name(ZFS_PROP_ZONED
), &zonestr
) == 0) &&
1145 strcmp(zonestr
, "on") == 0);
1147 if ((zc_fsprops
= zfs_valid_proplist(hdl
, ZFS_TYPE_FILESYSTEM
,
1148 fsprops
, zoned
, NULL
, NULL
, msg
)) == NULL
) {
1152 (nvlist_alloc(&zc_props
, NV_UNIQUE_NAME
, 0) != 0)) {
1155 if (nvlist_add_nvlist(zc_props
,
1156 ZPOOL_ROOTFS_PROPS
, zc_fsprops
) != 0) {
1161 if (zc_props
&& zcmd_write_src_nvlist(hdl
, &zc
, zc_props
) != 0)
1164 (void) strlcpy(zc
.zc_name
, pool
, sizeof (zc
.zc_name
));
1166 if ((ret
= zfs_ioctl(hdl
, ZFS_IOC_POOL_CREATE
, &zc
)) != 0) {
1168 zcmd_free_nvlists(&zc
);
1169 nvlist_free(zc_props
);
1170 nvlist_free(zc_fsprops
);
1175 * This can happen if the user has specified the same
1176 * device multiple times. We can't reliably detect this
1177 * until we try to add it and see we already have a
1180 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
1181 "one or more vdevs refer to the same device"));
1182 return (zfs_error(hdl
, EZFS_BADDEV
, msg
));
1186 * This happens if the record size is smaller or larger
1187 * than the allowed size range, or not a power of 2.
1189 * NOTE: although zfs_valid_proplist is called earlier,
1190 * this case may have slipped through since the
1191 * pool does not exist yet and it is therefore
1192 * impossible to read properties e.g. max blocksize
1195 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
1196 "record size invalid"));
1197 return (zfs_error(hdl
, EZFS_BADPROP
, msg
));
1201 * This occurs when one of the devices is below
1202 * SPA_MINDEVSIZE. Unfortunately, we can't detect which
1203 * device was the problem device since there's no
1204 * reliable way to determine device size from userland.
1209 zfs_nicenum(SPA_MINDEVSIZE
, buf
, sizeof (buf
));
1211 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
1212 "one or more devices is less than the "
1213 "minimum size (%s)"), buf
);
1215 return (zfs_error(hdl
, EZFS_BADDEV
, msg
));
1218 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
1219 "one or more devices is out of space"));
1220 return (zfs_error(hdl
, EZFS_BADDEV
, msg
));
1223 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
1224 "cache device must be a disk or disk slice"));
1225 return (zfs_error(hdl
, EZFS_BADDEV
, msg
));
1228 return (zpool_standard_error(hdl
, errno
, msg
));
1233 zcmd_free_nvlists(&zc
);
1234 nvlist_free(zc_props
);
1235 nvlist_free(zc_fsprops
);
1240 * Destroy the given pool. It is up to the caller to ensure that there are no
1241 * datasets left in the pool.
1244 zpool_destroy(zpool_handle_t
*zhp
, const char *log_str
)
1246 zfs_cmd_t zc
= { 0 };
1247 zfs_handle_t
*zfp
= NULL
;
1248 libzfs_handle_t
*hdl
= zhp
->zpool_hdl
;
1251 if (zhp
->zpool_state
== POOL_STATE_ACTIVE
&&
1252 (zfp
= zfs_open(hdl
, zhp
->zpool_name
, ZFS_TYPE_FILESYSTEM
)) == NULL
)
1255 (void) strlcpy(zc
.zc_name
, zhp
->zpool_name
, sizeof (zc
.zc_name
));
1256 zc
.zc_history
= (uint64_t)(uintptr_t)log_str
;
1258 if (zfs_ioctl(hdl
, ZFS_IOC_POOL_DESTROY
, &zc
) != 0) {
1259 (void) snprintf(msg
, sizeof (msg
), dgettext(TEXT_DOMAIN
,
1260 "cannot destroy '%s'"), zhp
->zpool_name
);
1262 if (errno
== EROFS
) {
1263 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
1264 "one or more devices is read only"));
1265 (void) zfs_error(hdl
, EZFS_BADDEV
, msg
);
1267 (void) zpool_standard_error(hdl
, errno
, msg
);
1276 remove_mountpoint(zfp
);
1284 * Add the given vdevs to the pool. The caller must have already performed the
1285 * necessary verification to ensure that the vdev specification is well-formed.
1288 zpool_add(zpool_handle_t
*zhp
, nvlist_t
*nvroot
)
1290 zfs_cmd_t zc
= { 0 };
1292 libzfs_handle_t
*hdl
= zhp
->zpool_hdl
;
1294 nvlist_t
**spares
, **l2cache
;
1295 uint_t nspares
, nl2cache
;
1297 (void) snprintf(msg
, sizeof (msg
), dgettext(TEXT_DOMAIN
,
1298 "cannot add to '%s'"), zhp
->zpool_name
);
1300 if (zpool_get_prop_int(zhp
, ZPOOL_PROP_VERSION
, NULL
) <
1301 SPA_VERSION_SPARES
&&
1302 nvlist_lookup_nvlist_array(nvroot
, ZPOOL_CONFIG_SPARES
,
1303 &spares
, &nspares
) == 0) {
1304 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
, "pool must be "
1305 "upgraded to add hot spares"));
1306 return (zfs_error(hdl
, EZFS_BADVERSION
, msg
));
1309 if (zpool_get_prop_int(zhp
, ZPOOL_PROP_VERSION
, NULL
) <
1310 SPA_VERSION_L2CACHE
&&
1311 nvlist_lookup_nvlist_array(nvroot
, ZPOOL_CONFIG_L2CACHE
,
1312 &l2cache
, &nl2cache
) == 0) {
1313 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
, "pool must be "
1314 "upgraded to add cache devices"));
1315 return (zfs_error(hdl
, EZFS_BADVERSION
, msg
));
1318 if (zcmd_write_conf_nvlist(hdl
, &zc
, nvroot
) != 0)
1320 (void) strlcpy(zc
.zc_name
, zhp
->zpool_name
, sizeof (zc
.zc_name
));
1322 if (zfs_ioctl(hdl
, ZFS_IOC_VDEV_ADD
, &zc
) != 0) {
1326 * This can happen if the user has specified the same
1327 * device multiple times. We can't reliably detect this
1328 * until we try to add it and see we already have a
1331 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
1332 "one or more vdevs refer to the same device"));
1333 (void) zfs_error(hdl
, EZFS_BADDEV
, msg
);
1337 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
1338 "invalid config; a pool with removing/removed "
1339 "vdevs does not support adding raidz vdevs"));
1340 (void) zfs_error(hdl
, EZFS_BADDEV
, msg
);
1345 * This occurrs when one of the devices is below
1346 * SPA_MINDEVSIZE. Unfortunately, we can't detect which
1347 * device was the problem device since there's no
1348 * reliable way to determine device size from userland.
1353 zfs_nicenum(SPA_MINDEVSIZE
, buf
, sizeof (buf
));
1355 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
1356 "device is less than the minimum "
1359 (void) zfs_error(hdl
, EZFS_BADDEV
, msg
);
1363 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
1364 "pool must be upgraded to add these vdevs"));
1365 (void) zfs_error(hdl
, EZFS_BADVERSION
, msg
);
1369 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
1370 "root pool can not have multiple vdevs"
1371 " or separate logs"));
1372 (void) zfs_error(hdl
, EZFS_POOL_NOTSUP
, msg
);
1376 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
1377 "cache device must be a disk or disk slice"));
1378 (void) zfs_error(hdl
, EZFS_BADDEV
, msg
);
1382 (void) zpool_standard_error(hdl
, errno
, msg
);
1390 zcmd_free_nvlists(&zc
);
1396 * Exports the pool from the system. The caller must ensure that there are no
1397 * mounted datasets in the pool.
1400 zpool_export_common(zpool_handle_t
*zhp
, boolean_t force
, boolean_t hardforce
,
1401 const char *log_str
)
1403 zfs_cmd_t zc
= { 0 };
1406 (void) snprintf(msg
, sizeof (msg
), dgettext(TEXT_DOMAIN
,
1407 "cannot export '%s'"), zhp
->zpool_name
);
1409 (void) strlcpy(zc
.zc_name
, zhp
->zpool_name
, sizeof (zc
.zc_name
));
1410 zc
.zc_cookie
= force
;
1411 zc
.zc_guid
= hardforce
;
1412 zc
.zc_history
= (uint64_t)(uintptr_t)log_str
;
1414 if (zfs_ioctl(zhp
->zpool_hdl
, ZFS_IOC_POOL_EXPORT
, &zc
) != 0) {
1417 zfs_error_aux(zhp
->zpool_hdl
, dgettext(TEXT_DOMAIN
,
1418 "use '-f' to override the following errors:\n"
1419 "'%s' has an active shared spare which could be"
1420 " used by other pools once '%s' is exported."),
1421 zhp
->zpool_name
, zhp
->zpool_name
);
1422 return (zfs_error(zhp
->zpool_hdl
, EZFS_ACTIVE_SPARE
,
1425 return (zpool_standard_error_fmt(zhp
->zpool_hdl
, errno
,
1434 zpool_export(zpool_handle_t
*zhp
, boolean_t force
, const char *log_str
)
1436 return (zpool_export_common(zhp
, force
, B_FALSE
, log_str
));
1440 zpool_export_force(zpool_handle_t
*zhp
, const char *log_str
)
1442 return (zpool_export_common(zhp
, B_TRUE
, B_TRUE
, log_str
));
1446 zpool_rewind_exclaim(libzfs_handle_t
*hdl
, const char *name
, boolean_t dryrun
,
1449 nvlist_t
*nv
= NULL
;
1455 if (!hdl
->libzfs_printerr
|| config
== NULL
)
1458 if (nvlist_lookup_nvlist(config
, ZPOOL_CONFIG_LOAD_INFO
, &nv
) != 0 ||
1459 nvlist_lookup_nvlist(nv
, ZPOOL_CONFIG_REWIND_INFO
, &nv
) != 0) {
1463 if (nvlist_lookup_uint64(nv
, ZPOOL_CONFIG_LOAD_TIME
, &rewindto
) != 0)
1465 (void) nvlist_lookup_int64(nv
, ZPOOL_CONFIG_REWIND_TIME
, &loss
);
1467 if (localtime_r((time_t *)&rewindto
, &t
) != NULL
&&
1468 strftime(timestr
, 128, 0, &t
) != 0) {
1470 (void) printf(dgettext(TEXT_DOMAIN
,
1471 "Would be able to return %s "
1472 "to its state as of %s.\n"),
1475 (void) printf(dgettext(TEXT_DOMAIN
,
1476 "Pool %s returned to its state as of %s.\n"),
1480 (void) printf(dgettext(TEXT_DOMAIN
,
1481 "%s approximately %lld "),
1482 dryrun
? "Would discard" : "Discarded",
1484 (void) printf(dgettext(TEXT_DOMAIN
,
1485 "minutes of transactions.\n"));
1486 } else if (loss
> 0) {
1487 (void) printf(dgettext(TEXT_DOMAIN
,
1488 "%s approximately %lld "),
1489 dryrun
? "Would discard" : "Discarded", loss
);
1490 (void) printf(dgettext(TEXT_DOMAIN
,
1491 "seconds of transactions.\n"));
1497 zpool_explain_recover(libzfs_handle_t
*hdl
, const char *name
, int reason
,
1500 nvlist_t
*nv
= NULL
;
1502 uint64_t edata
= UINT64_MAX
;
1507 if (!hdl
->libzfs_printerr
)
1511 (void) printf(dgettext(TEXT_DOMAIN
, "action: "));
1513 (void) printf(dgettext(TEXT_DOMAIN
, "\t"));
1515 /* All attempted rewinds failed if ZPOOL_CONFIG_LOAD_TIME missing */
1516 if (nvlist_lookup_nvlist(config
, ZPOOL_CONFIG_LOAD_INFO
, &nv
) != 0 ||
1517 nvlist_lookup_nvlist(nv
, ZPOOL_CONFIG_REWIND_INFO
, &nv
) != 0 ||
1518 nvlist_lookup_uint64(nv
, ZPOOL_CONFIG_LOAD_TIME
, &rewindto
) != 0)
1521 (void) nvlist_lookup_int64(nv
, ZPOOL_CONFIG_REWIND_TIME
, &loss
);
1522 (void) nvlist_lookup_uint64(nv
, ZPOOL_CONFIG_LOAD_DATA_ERRORS
,
1525 (void) printf(dgettext(TEXT_DOMAIN
,
1526 "Recovery is possible, but will result in some data loss.\n"));
1528 if (localtime_r((time_t *)&rewindto
, &t
) != NULL
&&
1529 strftime(timestr
, 128, 0, &t
) != 0) {
1530 (void) printf(dgettext(TEXT_DOMAIN
,
1531 "\tReturning the pool to its state as of %s\n"
1532 "\tshould correct the problem. "),
1535 (void) printf(dgettext(TEXT_DOMAIN
,
1536 "\tReverting the pool to an earlier state "
1537 "should correct the problem.\n\t"));
1541 (void) printf(dgettext(TEXT_DOMAIN
,
1542 "Approximately %lld minutes of data\n"
1543 "\tmust be discarded, irreversibly. "), (loss
+ 30) / 60);
1544 } else if (loss
> 0) {
1545 (void) printf(dgettext(TEXT_DOMAIN
,
1546 "Approximately %lld seconds of data\n"
1547 "\tmust be discarded, irreversibly. "), loss
);
1549 if (edata
!= 0 && edata
!= UINT64_MAX
) {
1551 (void) printf(dgettext(TEXT_DOMAIN
,
1552 "After rewind, at least\n"
1553 "\tone persistent user-data error will remain. "));
1555 (void) printf(dgettext(TEXT_DOMAIN
,
1556 "After rewind, several\n"
1557 "\tpersistent user-data errors will remain. "));
1560 (void) printf(dgettext(TEXT_DOMAIN
,
1561 "Recovery can be attempted\n\tby executing 'zpool %s -F %s'. "),
1562 reason
>= 0 ? "clear" : "import", name
);
1564 (void) printf(dgettext(TEXT_DOMAIN
,
1565 "A scrub of the pool\n"
1566 "\tis strongly recommended after recovery.\n"));
1570 (void) printf(dgettext(TEXT_DOMAIN
,
1571 "Destroy and re-create the pool from\n\ta backup source.\n"));
1575 * zpool_import() is a contracted interface. Should be kept the same
1578 * Applications should use zpool_import_props() to import a pool with
1579 * new properties value to be set.
1582 zpool_import(libzfs_handle_t
*hdl
, nvlist_t
*config
, const char *newname
,
1585 nvlist_t
*props
= NULL
;
1588 if (altroot
!= NULL
) {
1589 if (nvlist_alloc(&props
, NV_UNIQUE_NAME
, 0) != 0) {
1590 return (zfs_error_fmt(hdl
, EZFS_NOMEM
,
1591 dgettext(TEXT_DOMAIN
, "cannot import '%s'"),
1595 if (nvlist_add_string(props
,
1596 zpool_prop_to_name(ZPOOL_PROP_ALTROOT
), altroot
) != 0 ||
1597 nvlist_add_string(props
,
1598 zpool_prop_to_name(ZPOOL_PROP_CACHEFILE
), "none") != 0) {
1600 return (zfs_error_fmt(hdl
, EZFS_NOMEM
,
1601 dgettext(TEXT_DOMAIN
, "cannot import '%s'"),
1606 ret
= zpool_import_props(hdl
, config
, newname
, props
,
1613 print_vdev_tree(libzfs_handle_t
*hdl
, const char *name
, nvlist_t
*nv
,
1619 uint64_t is_log
= 0;
1621 (void) nvlist_lookup_uint64(nv
, ZPOOL_CONFIG_IS_LOG
,
1625 (void) printf("\t%*s%s%s\n", indent
, "", name
,
1626 is_log
? " [log]" : "");
1628 if (nvlist_lookup_nvlist_array(nv
, ZPOOL_CONFIG_CHILDREN
,
1629 &child
, &children
) != 0)
1632 for (c
= 0; c
< children
; c
++) {
1633 vname
= zpool_vdev_name(hdl
, NULL
, child
[c
], B_TRUE
);
1634 print_vdev_tree(hdl
, vname
, child
[c
], indent
+ 2);
1640 zpool_print_unsup_feat(nvlist_t
*config
)
1642 nvlist_t
*nvinfo
, *unsup_feat
;
1644 verify(nvlist_lookup_nvlist(config
, ZPOOL_CONFIG_LOAD_INFO
, &nvinfo
) ==
1646 verify(nvlist_lookup_nvlist(nvinfo
, ZPOOL_CONFIG_UNSUP_FEAT
,
1649 for (nvpair_t
*nvp
= nvlist_next_nvpair(unsup_feat
, NULL
); nvp
!= NULL
;
1650 nvp
= nvlist_next_nvpair(unsup_feat
, nvp
)) {
1653 verify(nvpair_type(nvp
) == DATA_TYPE_STRING
);
1654 verify(nvpair_value_string(nvp
, &desc
) == 0);
1656 if (strlen(desc
) > 0)
1657 (void) printf("\t%s (%s)\n", nvpair_name(nvp
), desc
);
1659 (void) printf("\t%s\n", nvpair_name(nvp
));
1664 * Import the given pool using the known configuration and a list of
1665 * properties to be set. The configuration should have come from
1666 * zpool_find_import(). The 'newname' parameters control whether the pool
1667 * is imported with a different name.
1670 zpool_import_props(libzfs_handle_t
*hdl
, nvlist_t
*config
, const char *newname
,
1671 nvlist_t
*props
, int flags
)
1673 zfs_cmd_t zc
= { 0 };
1674 zpool_rewind_policy_t policy
;
1675 nvlist_t
*nv
= NULL
;
1676 nvlist_t
*nvinfo
= NULL
;
1677 nvlist_t
*missing
= NULL
;
1684 verify(nvlist_lookup_string(config
, ZPOOL_CONFIG_POOL_NAME
,
1687 (void) snprintf(errbuf
, sizeof (errbuf
), dgettext(TEXT_DOMAIN
,
1688 "cannot import pool '%s'"), origname
);
1690 if (newname
!= NULL
) {
1691 if (!zpool_name_valid(hdl
, B_FALSE
, newname
))
1692 return (zfs_error_fmt(hdl
, EZFS_INVALIDNAME
,
1693 dgettext(TEXT_DOMAIN
, "cannot import '%s'"),
1695 thename
= (char *)newname
;
1700 if (props
!= NULL
) {
1702 prop_flags_t flags
= { .create
= B_FALSE
, .import
= B_TRUE
};
1704 verify(nvlist_lookup_uint64(config
, ZPOOL_CONFIG_VERSION
,
1707 if ((props
= zpool_valid_proplist(hdl
, origname
,
1708 props
, version
, flags
, errbuf
)) == NULL
)
1710 if (zcmd_write_src_nvlist(hdl
, &zc
, props
) != 0) {
1717 (void) strlcpy(zc
.zc_name
, thename
, sizeof (zc
.zc_name
));
1719 verify(nvlist_lookup_uint64(config
, ZPOOL_CONFIG_POOL_GUID
,
1722 if (zcmd_write_conf_nvlist(hdl
, &zc
, config
) != 0) {
1723 zcmd_free_nvlists(&zc
);
1726 if (zcmd_alloc_dst_nvlist(hdl
, &zc
, zc
.zc_nvlist_conf_size
* 2) != 0) {
1727 zcmd_free_nvlists(&zc
);
1731 zc
.zc_cookie
= flags
;
1732 while ((ret
= zfs_ioctl(hdl
, ZFS_IOC_POOL_IMPORT
, &zc
)) != 0 &&
1734 if (zcmd_expand_dst_nvlist(hdl
, &zc
) != 0) {
1735 zcmd_free_nvlists(&zc
);
1742 (void) zcmd_read_dst_nvlist(hdl
, &zc
, &nv
);
1744 zcmd_free_nvlists(&zc
);
1746 zpool_get_rewind_policy(config
, &policy
);
1752 * Dry-run failed, but we print out what success
1753 * looks like if we found a best txg
1755 if (policy
.zrp_request
& ZPOOL_TRY_REWIND
) {
1756 zpool_rewind_exclaim(hdl
, newname
? origname
: thename
,
1762 if (newname
== NULL
)
1763 (void) snprintf(desc
, sizeof (desc
),
1764 dgettext(TEXT_DOMAIN
, "cannot import '%s'"),
1767 (void) snprintf(desc
, sizeof (desc
),
1768 dgettext(TEXT_DOMAIN
, "cannot import '%s' as '%s'"),
1773 if (nv
!= NULL
&& nvlist_lookup_nvlist(nv
,
1774 ZPOOL_CONFIG_LOAD_INFO
, &nvinfo
) == 0 &&
1775 nvlist_exists(nvinfo
, ZPOOL_CONFIG_UNSUP_FEAT
)) {
1776 (void) printf(dgettext(TEXT_DOMAIN
, "This "
1777 "pool uses the following feature(s) not "
1778 "supported by this system:\n"));
1779 zpool_print_unsup_feat(nv
);
1780 if (nvlist_exists(nvinfo
,
1781 ZPOOL_CONFIG_CAN_RDONLY
)) {
1782 (void) printf(dgettext(TEXT_DOMAIN
,
1783 "All unsupported features are only "
1784 "required for writing to the pool."
1785 "\nThe pool can be imported using "
1786 "'-o readonly=on'.\n"));
1790 * Unsupported version.
1792 (void) zfs_error(hdl
, EZFS_BADVERSION
, desc
);
1796 (void) zfs_error(hdl
, EZFS_INVALCONFIG
, desc
);
1800 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
1801 "one or more devices is read only"));
1802 (void) zfs_error(hdl
, EZFS_BADDEV
, desc
);
1806 if (nv
&& nvlist_lookup_nvlist(nv
,
1807 ZPOOL_CONFIG_LOAD_INFO
, &nvinfo
) == 0 &&
1808 nvlist_lookup_nvlist(nvinfo
,
1809 ZPOOL_CONFIG_MISSING_DEVICES
, &missing
) == 0) {
1810 (void) printf(dgettext(TEXT_DOMAIN
,
1811 "The devices below are missing or "
1812 "corrupted, use '-m' to import the pool "
1814 print_vdev_tree(hdl
, NULL
, missing
, 2);
1815 (void) printf("\n");
1817 (void) zpool_standard_error(hdl
, error
, desc
);
1821 (void) zpool_standard_error(hdl
, error
, desc
);
1824 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
1825 "new name of at least one dataset is longer than "
1826 "the maximum allowable length"));
1827 (void) zfs_error(hdl
, EZFS_NAMETOOLONG
, desc
);
1830 (void) zpool_standard_error(hdl
, error
, desc
);
1831 zpool_explain_recover(hdl
,
1832 newname
? origname
: thename
, -error
, nv
);
1839 zpool_handle_t
*zhp
;
1842 * This should never fail, but play it safe anyway.
1844 if (zpool_open_silent(hdl
, thename
, &zhp
) != 0)
1846 else if (zhp
!= NULL
)
1848 if (policy
.zrp_request
&
1849 (ZPOOL_DO_REWIND
| ZPOOL_TRY_REWIND
)) {
1850 zpool_rewind_exclaim(hdl
, newname
? origname
: thename
,
1851 ((policy
.zrp_request
& ZPOOL_TRY_REWIND
) != 0), nv
);
1864 zpool_scan(zpool_handle_t
*zhp
, pool_scan_func_t func
, pool_scrub_cmd_t cmd
)
1866 zfs_cmd_t zc
= { 0 };
1869 libzfs_handle_t
*hdl
= zhp
->zpool_hdl
;
1871 (void) strlcpy(zc
.zc_name
, zhp
->zpool_name
, sizeof (zc
.zc_name
));
1872 zc
.zc_cookie
= func
;
1875 if (zfs_ioctl(hdl
, ZFS_IOC_POOL_SCAN
, &zc
) == 0)
1880 /* ECANCELED on a scrub means we resumed a paused scrub */
1881 if (err
== ECANCELED
&& func
== POOL_SCAN_SCRUB
&&
1882 cmd
== POOL_SCRUB_NORMAL
)
1885 if (err
== ENOENT
&& func
!= POOL_SCAN_NONE
&& cmd
== POOL_SCRUB_NORMAL
)
1888 if (func
== POOL_SCAN_SCRUB
) {
1889 if (cmd
== POOL_SCRUB_PAUSE
) {
1890 (void) snprintf(msg
, sizeof (msg
), dgettext(TEXT_DOMAIN
,
1891 "cannot pause scrubbing %s"), zc
.zc_name
);
1893 assert(cmd
== POOL_SCRUB_NORMAL
);
1894 (void) snprintf(msg
, sizeof (msg
), dgettext(TEXT_DOMAIN
,
1895 "cannot scrub %s"), zc
.zc_name
);
1897 } else if (func
== POOL_SCAN_NONE
) {
1898 (void) snprintf(msg
, sizeof (msg
),
1899 dgettext(TEXT_DOMAIN
, "cannot cancel scrubbing %s"),
1902 assert(!"unexpected result");
1907 pool_scan_stat_t
*ps
= NULL
;
1910 verify(nvlist_lookup_nvlist(zhp
->zpool_config
,
1911 ZPOOL_CONFIG_VDEV_TREE
, &nvroot
) == 0);
1912 (void) nvlist_lookup_uint64_array(nvroot
,
1913 ZPOOL_CONFIG_SCAN_STATS
, (uint64_t **)&ps
, &psc
);
1914 if (ps
&& ps
->pss_func
== POOL_SCAN_SCRUB
) {
1915 if (cmd
== POOL_SCRUB_PAUSE
)
1916 return (zfs_error(hdl
, EZFS_SCRUB_PAUSED
, msg
));
1918 return (zfs_error(hdl
, EZFS_SCRUBBING
, msg
));
1920 return (zfs_error(hdl
, EZFS_RESILVERING
, msg
));
1922 } else if (err
== ENOENT
) {
1923 return (zfs_error(hdl
, EZFS_NO_SCRUB
, msg
));
1925 return (zpool_standard_error(hdl
, err
, msg
));
1930 * This provides a very minimal check whether a given string is likely a
1931 * c#t#d# style string. Users of this are expected to do their own
1932 * verification of the s# part.
1934 #define CTD_CHECK(str) (str && str[0] == 'c' && isdigit(str[1]))
1937 * More elaborate version for ones which may start with "/dev/dsk/"
1941 ctd_check_path(char *str
)
1944 * If it starts with a slash, check the last component.
1946 if (str
&& str
[0] == '/') {
1947 char *tmp
= strrchr(str
, '/');
1950 * If it ends in "/old", check the second-to-last
1951 * component of the string instead.
1953 if (tmp
!= str
&& strcmp(tmp
, "/old") == 0) {
1954 for (tmp
--; *tmp
!= '/'; tmp
--)
1959 return (CTD_CHECK(str
));
1963 * Find a vdev that matches the search criteria specified. We use the
1964 * the nvpair name to determine how we should look for the device.
1965 * 'avail_spare' is set to TRUE if the provided guid refers to an AVAIL
1966 * spare; but FALSE if its an INUSE spare.
1969 vdev_to_nvlist_iter(nvlist_t
*nv
, nvlist_t
*search
, boolean_t
*avail_spare
,
1970 boolean_t
*l2cache
, boolean_t
*log
)
1977 nvpair_t
*pair
= nvlist_next_nvpair(search
, NULL
);
1979 /* Nothing to look for */
1980 if (search
== NULL
|| pair
== NULL
)
1983 /* Obtain the key we will use to search */
1984 srchkey
= nvpair_name(pair
);
1986 switch (nvpair_type(pair
)) {
1987 case DATA_TYPE_UINT64
:
1988 if (strcmp(srchkey
, ZPOOL_CONFIG_GUID
) == 0) {
1989 uint64_t srchval
, theguid
;
1991 verify(nvpair_value_uint64(pair
, &srchval
) == 0);
1992 verify(nvlist_lookup_uint64(nv
, ZPOOL_CONFIG_GUID
,
1994 if (theguid
== srchval
)
1999 case DATA_TYPE_STRING
: {
2000 char *srchval
, *val
;
2002 verify(nvpair_value_string(pair
, &srchval
) == 0);
2003 if (nvlist_lookup_string(nv
, srchkey
, &val
) != 0)
2007 * Search for the requested value. Special cases:
2009 * - ZPOOL_CONFIG_PATH for whole disk entries. To support
2010 * UEFI boot, these end in "s0" or "s0/old" or "s1" or
2011 * "s1/old". The "s0" or "s1" part is hidden from the user,
2012 * but included in the string, so this matches around it.
2013 * - looking for a top-level vdev name (i.e. ZPOOL_CONFIG_TYPE).
2015 * Otherwise, all other searches are simple string compares.
2017 if (strcmp(srchkey
, ZPOOL_CONFIG_PATH
) == 0 &&
2018 ctd_check_path(val
)) {
2019 uint64_t wholedisk
= 0;
2021 (void) nvlist_lookup_uint64(nv
, ZPOOL_CONFIG_WHOLE_DISK
,
2024 int slen
= strlen(srchval
);
2025 int vlen
= strlen(val
);
2027 if (slen
!= vlen
- 2)
2031 * make_leaf_vdev() should only set
2032 * wholedisk for ZPOOL_CONFIG_PATHs which
2033 * will include "/dev/dsk/", giving plenty of
2034 * room for the indices used next.
2039 * strings identical except trailing "s0"
2041 if ((strcmp(&val
[vlen
- 2], "s0") == 0 ||
2042 strcmp(&val
[vlen
- 2], "s1") == 0) &&
2043 strncmp(srchval
, val
, slen
) == 0)
2047 * strings identical except trailing "s0/old"
2049 if ((strcmp(&val
[vlen
- 6], "s0/old") == 0 ||
2050 strcmp(&val
[vlen
- 6], "s1/old") == 0) &&
2051 strcmp(&srchval
[slen
- 4], "/old") == 0 &&
2052 strncmp(srchval
, val
, slen
- 4) == 0)
2057 } else if (strcmp(srchkey
, ZPOOL_CONFIG_TYPE
) == 0 && val
) {
2058 char *type
, *idx
, *end
, *p
;
2059 uint64_t id
, vdev_id
;
2062 * Determine our vdev type, keeping in mind
2063 * that the srchval is composed of a type and
2064 * vdev id pair (i.e. mirror-4).
2066 if ((type
= strdup(srchval
)) == NULL
)
2069 if ((p
= strrchr(type
, '-')) == NULL
) {
2077 * If the types don't match then keep looking.
2079 if (strncmp(val
, type
, strlen(val
)) != 0) {
2084 verify(zpool_vdev_is_interior(type
));
2085 verify(nvlist_lookup_uint64(nv
, ZPOOL_CONFIG_ID
,
2089 vdev_id
= strtoull(idx
, &end
, 10);
2096 * Now verify that we have the correct vdev id.
2105 if (strcmp(srchval
, val
) == 0)
2114 if (nvlist_lookup_nvlist_array(nv
, ZPOOL_CONFIG_CHILDREN
,
2115 &child
, &children
) != 0)
2118 for (c
= 0; c
< children
; c
++) {
2119 if ((ret
= vdev_to_nvlist_iter(child
[c
], search
,
2120 avail_spare
, l2cache
, NULL
)) != NULL
) {
2122 * The 'is_log' value is only set for the toplevel
2123 * vdev, not the leaf vdevs. So we always lookup the
2124 * log device from the root of the vdev tree (where
2125 * 'log' is non-NULL).
2128 nvlist_lookup_uint64(child
[c
],
2129 ZPOOL_CONFIG_IS_LOG
, &is_log
) == 0 &&
2137 if (nvlist_lookup_nvlist_array(nv
, ZPOOL_CONFIG_SPARES
,
2138 &child
, &children
) == 0) {
2139 for (c
= 0; c
< children
; c
++) {
2140 if ((ret
= vdev_to_nvlist_iter(child
[c
], search
,
2141 avail_spare
, l2cache
, NULL
)) != NULL
) {
2142 *avail_spare
= B_TRUE
;
2148 if (nvlist_lookup_nvlist_array(nv
, ZPOOL_CONFIG_L2CACHE
,
2149 &child
, &children
) == 0) {
2150 for (c
= 0; c
< children
; c
++) {
2151 if ((ret
= vdev_to_nvlist_iter(child
[c
], search
,
2152 avail_spare
, l2cache
, NULL
)) != NULL
) {
2163 * Given a physical path (minus the "/devices" prefix), find the
2167 zpool_find_vdev_by_physpath(zpool_handle_t
*zhp
, const char *ppath
,
2168 boolean_t
*avail_spare
, boolean_t
*l2cache
, boolean_t
*log
)
2170 nvlist_t
*search
, *nvroot
, *ret
;
2172 verify(nvlist_alloc(&search
, NV_UNIQUE_NAME
, KM_SLEEP
) == 0);
2173 verify(nvlist_add_string(search
, ZPOOL_CONFIG_PHYS_PATH
, ppath
) == 0);
2175 verify(nvlist_lookup_nvlist(zhp
->zpool_config
, ZPOOL_CONFIG_VDEV_TREE
,
2178 *avail_spare
= B_FALSE
;
2182 ret
= vdev_to_nvlist_iter(nvroot
, search
, avail_spare
, l2cache
, log
);
2183 nvlist_free(search
);
2189 * Determine if we have an "interior" top-level vdev (i.e mirror/raidz).
2192 zpool_vdev_is_interior(const char *name
)
2194 if (strncmp(name
, VDEV_TYPE_RAIDZ
, strlen(VDEV_TYPE_RAIDZ
)) == 0 ||
2195 strncmp(name
, VDEV_TYPE_SPARE
, strlen(VDEV_TYPE_SPARE
)) == 0 ||
2197 VDEV_TYPE_REPLACING
, strlen(VDEV_TYPE_REPLACING
)) == 0 ||
2198 strncmp(name
, VDEV_TYPE_MIRROR
, strlen(VDEV_TYPE_MIRROR
)) == 0)
2204 zpool_find_vdev(zpool_handle_t
*zhp
, const char *path
, boolean_t
*avail_spare
,
2205 boolean_t
*l2cache
, boolean_t
*log
)
2207 char buf
[MAXPATHLEN
];
2209 nvlist_t
*nvroot
, *search
, *ret
;
2212 verify(nvlist_alloc(&search
, NV_UNIQUE_NAME
, KM_SLEEP
) == 0);
2214 guid
= strtoull(path
, &end
, 10);
2215 if (guid
!= 0 && *end
== '\0') {
2216 verify(nvlist_add_uint64(search
, ZPOOL_CONFIG_GUID
, guid
) == 0);
2217 } else if (zpool_vdev_is_interior(path
)) {
2218 verify(nvlist_add_string(search
, ZPOOL_CONFIG_TYPE
, path
) == 0);
2219 } else if (path
[0] != '/') {
2220 (void) snprintf(buf
, sizeof (buf
), "%s/%s", ZFS_DISK_ROOT
,
2222 verify(nvlist_add_string(search
, ZPOOL_CONFIG_PATH
, buf
) == 0);
2224 verify(nvlist_add_string(search
, ZPOOL_CONFIG_PATH
, path
) == 0);
2227 verify(nvlist_lookup_nvlist(zhp
->zpool_config
, ZPOOL_CONFIG_VDEV_TREE
,
2230 *avail_spare
= B_FALSE
;
2234 ret
= vdev_to_nvlist_iter(nvroot
, search
, avail_spare
, l2cache
, log
);
2235 nvlist_free(search
);
2241 vdev_online(nvlist_t
*nv
)
2245 if (nvlist_lookup_uint64(nv
, ZPOOL_CONFIG_OFFLINE
, &ival
) == 0 ||
2246 nvlist_lookup_uint64(nv
, ZPOOL_CONFIG_FAULTED
, &ival
) == 0 ||
2247 nvlist_lookup_uint64(nv
, ZPOOL_CONFIG_REMOVED
, &ival
) == 0)
2254 * Helper function for zpool_get_physpaths().
2257 vdev_get_one_physpath(nvlist_t
*config
, char *physpath
, size_t physpath_size
,
2258 size_t *bytes_written
)
2260 size_t bytes_left
, pos
, rsz
;
2264 if (nvlist_lookup_string(config
, ZPOOL_CONFIG_PHYS_PATH
,
2266 return (EZFS_NODEVICE
);
2268 pos
= *bytes_written
;
2269 bytes_left
= physpath_size
- pos
;
2270 format
= (pos
== 0) ? "%s" : " %s";
2272 rsz
= snprintf(physpath
+ pos
, bytes_left
, format
, tmppath
);
2273 *bytes_written
+= rsz
;
2275 if (rsz
>= bytes_left
) {
2276 /* if physpath was not copied properly, clear it */
2277 if (bytes_left
!= 0) {
2280 return (EZFS_NOSPC
);
2286 vdev_get_physpaths(nvlist_t
*nv
, char *physpath
, size_t phypath_size
,
2287 size_t *rsz
, boolean_t is_spare
)
2292 if (nvlist_lookup_string(nv
, ZPOOL_CONFIG_TYPE
, &type
) != 0)
2293 return (EZFS_INVALCONFIG
);
2295 if (strcmp(type
, VDEV_TYPE_DISK
) == 0) {
2297 * An active spare device has ZPOOL_CONFIG_IS_SPARE set.
2298 * For a spare vdev, we only want to boot from the active
2303 (void) nvlist_lookup_uint64(nv
, ZPOOL_CONFIG_IS_SPARE
,
2306 return (EZFS_INVALCONFIG
);
2309 if (vdev_online(nv
)) {
2310 if ((ret
= vdev_get_one_physpath(nv
, physpath
,
2311 phypath_size
, rsz
)) != 0)
2314 } else if (strcmp(type
, VDEV_TYPE_MIRROR
) == 0 ||
2315 strcmp(type
, VDEV_TYPE_RAIDZ
) == 0 ||
2316 strcmp(type
, VDEV_TYPE_REPLACING
) == 0 ||
2317 (is_spare
= (strcmp(type
, VDEV_TYPE_SPARE
) == 0))) {
2322 if (nvlist_lookup_nvlist_array(nv
,
2323 ZPOOL_CONFIG_CHILDREN
, &child
, &count
) != 0)
2324 return (EZFS_INVALCONFIG
);
2326 for (i
= 0; i
< count
; i
++) {
2327 ret
= vdev_get_physpaths(child
[i
], physpath
,
2328 phypath_size
, rsz
, is_spare
);
2329 if (ret
== EZFS_NOSPC
)
2334 return (EZFS_POOL_INVALARG
);
2338 * Get phys_path for a root pool config.
2339 * Return 0 on success; non-zero on failure.
2342 zpool_get_config_physpath(nvlist_t
*config
, char *physpath
, size_t phypath_size
)
2345 nvlist_t
*vdev_root
;
2352 if (nvlist_lookup_nvlist(config
, ZPOOL_CONFIG_VDEV_TREE
,
2354 return (EZFS_INVALCONFIG
);
2356 if (nvlist_lookup_string(vdev_root
, ZPOOL_CONFIG_TYPE
, &type
) != 0 ||
2357 nvlist_lookup_nvlist_array(vdev_root
, ZPOOL_CONFIG_CHILDREN
,
2358 &child
, &count
) != 0)
2359 return (EZFS_INVALCONFIG
);
2362 * root pool can only have a single top-level vdev.
2364 if (strcmp(type
, VDEV_TYPE_ROOT
) != 0 || count
!= 1)
2365 return (EZFS_POOL_INVALARG
);
2367 (void) vdev_get_physpaths(child
[0], physpath
, phypath_size
, &rsz
,
2370 /* No online devices */
2372 return (EZFS_NODEVICE
);
2378 * Get phys_path for a root pool
2379 * Return 0 on success; non-zero on failure.
2382 zpool_get_physpath(zpool_handle_t
*zhp
, char *physpath
, size_t phypath_size
)
2384 return (zpool_get_config_physpath(zhp
->zpool_config
, physpath
,
2389 * If the device has being dynamically expanded then we need to relabel
2390 * the disk to use the new unallocated space.
2393 zpool_relabel_disk(libzfs_handle_t
*hdl
, const char *name
)
2395 char path
[MAXPATHLEN
];
2398 int (*_efi_use_whole_disk
)(int);
2400 if ((_efi_use_whole_disk
= (int (*)(int))dlsym(RTLD_DEFAULT
,
2401 "efi_use_whole_disk")) == NULL
)
2404 (void) snprintf(path
, sizeof (path
), "%s/%s", ZFS_RDISK_ROOT
, name
);
2406 if ((fd
= open(path
, O_RDWR
| O_NDELAY
)) < 0) {
2407 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
, "cannot "
2408 "relabel '%s': unable to open device"), name
);
2409 return (zfs_error(hdl
, EZFS_OPENFAILED
, errbuf
));
2413 * It's possible that we might encounter an error if the device
2414 * does not have any unallocated space left. If so, we simply
2415 * ignore that error and continue on.
2417 error
= _efi_use_whole_disk(fd
);
2419 if (error
&& error
!= VT_ENOSPC
) {
2420 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
, "cannot "
2421 "relabel '%s': unable to read disk capacity"), name
);
2422 return (zfs_error(hdl
, EZFS_NOCAP
, errbuf
));
2428 * Bring the specified vdev online. The 'flags' parameter is a set of the
2429 * ZFS_ONLINE_* flags.
2432 zpool_vdev_online(zpool_handle_t
*zhp
, const char *path
, int flags
,
2433 vdev_state_t
*newstate
)
2435 zfs_cmd_t zc
= { 0 };
2439 boolean_t avail_spare
, l2cache
, islog
;
2440 libzfs_handle_t
*hdl
= zhp
->zpool_hdl
;
2442 if (flags
& ZFS_ONLINE_EXPAND
) {
2443 (void) snprintf(msg
, sizeof (msg
),
2444 dgettext(TEXT_DOMAIN
, "cannot expand %s"), path
);
2446 (void) snprintf(msg
, sizeof (msg
),
2447 dgettext(TEXT_DOMAIN
, "cannot online %s"), path
);
2450 (void) strlcpy(zc
.zc_name
, zhp
->zpool_name
, sizeof (zc
.zc_name
));
2451 if ((tgt
= zpool_find_vdev(zhp
, path
, &avail_spare
, &l2cache
,
2453 return (zfs_error(hdl
, EZFS_NODEVICE
, msg
));
2455 verify(nvlist_lookup_uint64(tgt
, ZPOOL_CONFIG_GUID
, &zc
.zc_guid
) == 0);
2458 return (zfs_error(hdl
, EZFS_ISSPARE
, msg
));
2460 if ((flags
& ZFS_ONLINE_EXPAND
||
2461 zpool_get_prop_int(zhp
, ZPOOL_PROP_AUTOEXPAND
, NULL
)) &&
2462 nvlist_lookup_string(tgt
, ZPOOL_CONFIG_PATH
, &pathname
) == 0) {
2463 uint64_t wholedisk
= 0;
2465 (void) nvlist_lookup_uint64(tgt
, ZPOOL_CONFIG_WHOLE_DISK
,
2469 * XXX - L2ARC 1.0 devices can't support expansion.
2472 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
2473 "cannot expand cache devices"));
2474 return (zfs_error(hdl
, EZFS_VDEVNOTSUP
, msg
));
2478 pathname
+= strlen(ZFS_DISK_ROOT
) + 1;
2479 (void) zpool_relabel_disk(hdl
, pathname
);
2483 zc
.zc_cookie
= VDEV_STATE_ONLINE
;
2486 if (zfs_ioctl(hdl
, ZFS_IOC_VDEV_SET_STATE
, &zc
) != 0) {
2487 if (errno
== EINVAL
) {
2488 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
, "was split "
2489 "from this pool into a new one. Use '%s' "
2490 "instead"), "zpool detach");
2491 return (zfs_error(hdl
, EZFS_POSTSPLIT_ONLINE
, msg
));
2493 return (zpool_standard_error(hdl
, errno
, msg
));
2496 *newstate
= zc
.zc_cookie
;
2501 * Take the specified vdev offline
2504 zpool_vdev_offline(zpool_handle_t
*zhp
, const char *path
, boolean_t istmp
)
2506 zfs_cmd_t zc
= { 0 };
2509 boolean_t avail_spare
, l2cache
;
2510 libzfs_handle_t
*hdl
= zhp
->zpool_hdl
;
2512 (void) snprintf(msg
, sizeof (msg
),
2513 dgettext(TEXT_DOMAIN
, "cannot offline %s"), path
);
2515 (void) strlcpy(zc
.zc_name
, zhp
->zpool_name
, sizeof (zc
.zc_name
));
2516 if ((tgt
= zpool_find_vdev(zhp
, path
, &avail_spare
, &l2cache
,
2518 return (zfs_error(hdl
, EZFS_NODEVICE
, msg
));
2520 verify(nvlist_lookup_uint64(tgt
, ZPOOL_CONFIG_GUID
, &zc
.zc_guid
) == 0);
2523 return (zfs_error(hdl
, EZFS_ISSPARE
, msg
));
2525 zc
.zc_cookie
= VDEV_STATE_OFFLINE
;
2526 zc
.zc_obj
= istmp
? ZFS_OFFLINE_TEMPORARY
: 0;
2528 if (zfs_ioctl(hdl
, ZFS_IOC_VDEV_SET_STATE
, &zc
) == 0)
2535 * There are no other replicas of this device.
2537 return (zfs_error(hdl
, EZFS_NOREPLICAS
, msg
));
2541 * The log device has unplayed logs
2543 return (zfs_error(hdl
, EZFS_UNPLAYED_LOGS
, msg
));
2546 return (zpool_standard_error(hdl
, errno
, msg
));
2551 * Mark the given vdev faulted.
2554 zpool_vdev_fault(zpool_handle_t
*zhp
, uint64_t guid
, vdev_aux_t aux
)
2556 zfs_cmd_t zc
= { 0 };
2558 libzfs_handle_t
*hdl
= zhp
->zpool_hdl
;
2560 (void) snprintf(msg
, sizeof (msg
),
2561 dgettext(TEXT_DOMAIN
, "cannot fault %llu"), guid
);
2563 (void) strlcpy(zc
.zc_name
, zhp
->zpool_name
, sizeof (zc
.zc_name
));
2565 zc
.zc_cookie
= VDEV_STATE_FAULTED
;
2568 if (ioctl(hdl
->libzfs_fd
, ZFS_IOC_VDEV_SET_STATE
, &zc
) == 0)
2575 * There are no other replicas of this device.
2577 return (zfs_error(hdl
, EZFS_NOREPLICAS
, msg
));
2580 return (zpool_standard_error(hdl
, errno
, msg
));
2586 * Mark the given vdev degraded.
2589 zpool_vdev_degrade(zpool_handle_t
*zhp
, uint64_t guid
, vdev_aux_t aux
)
2591 zfs_cmd_t zc
= { 0 };
2593 libzfs_handle_t
*hdl
= zhp
->zpool_hdl
;
2595 (void) snprintf(msg
, sizeof (msg
),
2596 dgettext(TEXT_DOMAIN
, "cannot degrade %llu"), guid
);
2598 (void) strlcpy(zc
.zc_name
, zhp
->zpool_name
, sizeof (zc
.zc_name
));
2600 zc
.zc_cookie
= VDEV_STATE_DEGRADED
;
2603 if (ioctl(hdl
->libzfs_fd
, ZFS_IOC_VDEV_SET_STATE
, &zc
) == 0)
2606 return (zpool_standard_error(hdl
, errno
, msg
));
2610 * Returns TRUE if the given nvlist is a vdev that was originally swapped in as
2614 is_replacing_spare(nvlist_t
*search
, nvlist_t
*tgt
, int which
)
2620 if (nvlist_lookup_nvlist_array(search
, ZPOOL_CONFIG_CHILDREN
, &child
,
2622 verify(nvlist_lookup_string(search
, ZPOOL_CONFIG_TYPE
,
2625 if (strcmp(type
, VDEV_TYPE_SPARE
) == 0 &&
2626 children
== 2 && child
[which
] == tgt
)
2629 for (c
= 0; c
< children
; c
++)
2630 if (is_replacing_spare(child
[c
], tgt
, which
))
2638 * Attach new_disk (fully described by nvroot) to old_disk.
2639 * If 'replacing' is specified, the new disk will replace the old one.
2642 zpool_vdev_attach(zpool_handle_t
*zhp
,
2643 const char *old_disk
, const char *new_disk
, nvlist_t
*nvroot
, int replacing
)
2645 zfs_cmd_t zc
= { 0 };
2649 boolean_t avail_spare
, l2cache
, islog
;
2654 nvlist_t
*config_root
;
2655 libzfs_handle_t
*hdl
= zhp
->zpool_hdl
;
2656 boolean_t rootpool
= zpool_is_bootable(zhp
);
2659 (void) snprintf(msg
, sizeof (msg
), dgettext(TEXT_DOMAIN
,
2660 "cannot replace %s with %s"), old_disk
, new_disk
);
2662 (void) snprintf(msg
, sizeof (msg
), dgettext(TEXT_DOMAIN
,
2663 "cannot attach %s to %s"), new_disk
, old_disk
);
2665 (void) strlcpy(zc
.zc_name
, zhp
->zpool_name
, sizeof (zc
.zc_name
));
2666 if ((tgt
= zpool_find_vdev(zhp
, old_disk
, &avail_spare
, &l2cache
,
2668 return (zfs_error(hdl
, EZFS_NODEVICE
, msg
));
2671 return (zfs_error(hdl
, EZFS_ISSPARE
, msg
));
2674 return (zfs_error(hdl
, EZFS_ISL2CACHE
, msg
));
2676 verify(nvlist_lookup_uint64(tgt
, ZPOOL_CONFIG_GUID
, &zc
.zc_guid
) == 0);
2677 zc
.zc_cookie
= replacing
;
2679 if (nvlist_lookup_nvlist_array(nvroot
, ZPOOL_CONFIG_CHILDREN
,
2680 &child
, &children
) != 0 || children
!= 1) {
2681 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
2682 "new device must be a single disk"));
2683 return (zfs_error(hdl
, EZFS_INVALCONFIG
, msg
));
2686 verify(nvlist_lookup_nvlist(zpool_get_config(zhp
, NULL
),
2687 ZPOOL_CONFIG_VDEV_TREE
, &config_root
) == 0);
2689 if ((newname
= zpool_vdev_name(NULL
, NULL
, child
[0], B_FALSE
)) == NULL
)
2693 * If the target is a hot spare that has been swapped in, we can only
2694 * replace it with another hot spare.
2697 nvlist_lookup_uint64(tgt
, ZPOOL_CONFIG_IS_SPARE
, &val
) == 0 &&
2698 (zpool_find_vdev(zhp
, newname
, &avail_spare
, &l2cache
,
2699 NULL
) == NULL
|| !avail_spare
) &&
2700 is_replacing_spare(config_root
, tgt
, 1)) {
2701 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
2702 "can only be replaced by another hot spare"));
2704 return (zfs_error(hdl
, EZFS_BADTARGET
, msg
));
2709 if (zcmd_write_conf_nvlist(hdl
, &zc
, nvroot
) != 0)
2712 ret
= zfs_ioctl(hdl
, ZFS_IOC_VDEV_ATTACH
, &zc
);
2714 zcmd_free_nvlists(&zc
);
2719 * XXX need a better way to prevent user from
2720 * booting up a half-baked vdev.
2722 (void) fprintf(stderr
, dgettext(TEXT_DOMAIN
, "Make "
2723 "sure to wait until resilver is done "
2724 "before rebooting.\n"));
2732 * Can't attach to or replace this type of vdev.
2735 uint64_t version
= zpool_get_prop_int(zhp
,
2736 ZPOOL_PROP_VERSION
, NULL
);
2739 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
2740 "cannot replace a log with a spare"));
2741 else if (version
>= SPA_VERSION_MULTI_REPLACE
)
2742 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
2743 "already in replacing/spare config; wait "
2744 "for completion or use 'zpool detach'"));
2746 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
2747 "cannot replace a replacing device"));
2749 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
2750 "can only attach to mirrors and top-level "
2753 (void) zfs_error(hdl
, EZFS_BADTARGET
, msg
);
2758 * The new device must be a single disk.
2760 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
2761 "new device must be a single disk"));
2762 (void) zfs_error(hdl
, EZFS_INVALCONFIG
, msg
);
2766 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
, "%s is busy, "
2767 "or pool has removing/removed vdevs"),
2769 (void) zfs_error(hdl
, EZFS_BADDEV
, msg
);
2774 * The new device is too small.
2776 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
2777 "device is too small"));
2778 (void) zfs_error(hdl
, EZFS_BADDEV
, msg
);
2783 * The new device has a different alignment requirement.
2785 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
2786 "devices have different sector alignment"));
2787 (void) zfs_error(hdl
, EZFS_BADDEV
, msg
);
2792 * The resulting top-level vdev spec won't fit in the label.
2794 (void) zfs_error(hdl
, EZFS_DEVOVERFLOW
, msg
);
2798 (void) zpool_standard_error(hdl
, errno
, msg
);
2805 * Detach the specified device.
2808 zpool_vdev_detach(zpool_handle_t
*zhp
, const char *path
)
2810 zfs_cmd_t zc
= { 0 };
2813 boolean_t avail_spare
, l2cache
;
2814 libzfs_handle_t
*hdl
= zhp
->zpool_hdl
;
2816 (void) snprintf(msg
, sizeof (msg
),
2817 dgettext(TEXT_DOMAIN
, "cannot detach %s"), path
);
2819 (void) strlcpy(zc
.zc_name
, zhp
->zpool_name
, sizeof (zc
.zc_name
));
2820 if ((tgt
= zpool_find_vdev(zhp
, path
, &avail_spare
, &l2cache
,
2822 return (zfs_error(hdl
, EZFS_NODEVICE
, msg
));
2825 return (zfs_error(hdl
, EZFS_ISSPARE
, msg
));
2828 return (zfs_error(hdl
, EZFS_ISL2CACHE
, msg
));
2830 verify(nvlist_lookup_uint64(tgt
, ZPOOL_CONFIG_GUID
, &zc
.zc_guid
) == 0);
2832 if (zfs_ioctl(hdl
, ZFS_IOC_VDEV_DETACH
, &zc
) == 0)
2839 * Can't detach from this type of vdev.
2841 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
, "only "
2842 "applicable to mirror and replacing vdevs"));
2843 (void) zfs_error(hdl
, EZFS_BADTARGET
, msg
);
2848 * There are no other replicas of this device.
2850 (void) zfs_error(hdl
, EZFS_NOREPLICAS
, msg
);
2854 (void) zpool_standard_error(hdl
, errno
, msg
);
2861 * Find a mirror vdev in the source nvlist.
2863 * The mchild array contains a list of disks in one of the top-level mirrors
2864 * of the source pool. The schild array contains a list of disks that the
2865 * user specified on the command line. We loop over the mchild array to
2866 * see if any entry in the schild array matches.
2868 * If a disk in the mchild array is found in the schild array, we return
2869 * the index of that entry. Otherwise we return -1.
2872 find_vdev_entry(zpool_handle_t
*zhp
, nvlist_t
**mchild
, uint_t mchildren
,
2873 nvlist_t
**schild
, uint_t schildren
)
2877 for (mc
= 0; mc
< mchildren
; mc
++) {
2879 char *mpath
= zpool_vdev_name(zhp
->zpool_hdl
, zhp
,
2880 mchild
[mc
], B_FALSE
);
2882 for (sc
= 0; sc
< schildren
; sc
++) {
2883 char *spath
= zpool_vdev_name(zhp
->zpool_hdl
, zhp
,
2884 schild
[sc
], B_FALSE
);
2885 boolean_t result
= (strcmp(mpath
, spath
) == 0);
2901 * Split a mirror pool. If newroot points to null, then a new nvlist
2902 * is generated and it is the responsibility of the caller to free it.
2905 zpool_vdev_split(zpool_handle_t
*zhp
, char *newname
, nvlist_t
**newroot
,
2906 nvlist_t
*props
, splitflags_t flags
)
2908 zfs_cmd_t zc
= { 0 };
2910 nvlist_t
*tree
, *config
, **child
, **newchild
, *newconfig
= NULL
;
2911 nvlist_t
**varray
= NULL
, *zc_props
= NULL
;
2912 uint_t c
, children
, newchildren
, lastlog
= 0, vcount
, found
= 0;
2913 libzfs_handle_t
*hdl
= zhp
->zpool_hdl
;
2915 boolean_t freelist
= B_FALSE
, memory_err
= B_TRUE
;
2918 (void) snprintf(msg
, sizeof (msg
),
2919 dgettext(TEXT_DOMAIN
, "Unable to split %s"), zhp
->zpool_name
);
2921 if (!zpool_name_valid(hdl
, B_FALSE
, newname
))
2922 return (zfs_error(hdl
, EZFS_INVALIDNAME
, msg
));
2924 if ((config
= zpool_get_config(zhp
, NULL
)) == NULL
) {
2925 (void) fprintf(stderr
, gettext("Internal error: unable to "
2926 "retrieve pool configuration\n"));
2930 verify(nvlist_lookup_nvlist(config
, ZPOOL_CONFIG_VDEV_TREE
, &tree
)
2932 verify(nvlist_lookup_uint64(config
, ZPOOL_CONFIG_VERSION
, &vers
) == 0);
2935 prop_flags_t flags
= { .create
= B_FALSE
, .import
= B_TRUE
};
2936 if ((zc_props
= zpool_valid_proplist(hdl
, zhp
->zpool_name
,
2937 props
, vers
, flags
, msg
)) == NULL
)
2941 if (nvlist_lookup_nvlist_array(tree
, ZPOOL_CONFIG_CHILDREN
, &child
,
2943 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
2944 "Source pool is missing vdev tree"));
2945 nvlist_free(zc_props
);
2949 varray
= zfs_alloc(hdl
, children
* sizeof (nvlist_t
*));
2952 if (*newroot
== NULL
||
2953 nvlist_lookup_nvlist_array(*newroot
, ZPOOL_CONFIG_CHILDREN
,
2954 &newchild
, &newchildren
) != 0)
2957 for (c
= 0; c
< children
; c
++) {
2958 uint64_t is_log
= B_FALSE
, is_hole
= B_FALSE
;
2960 nvlist_t
**mchild
, *vdev
;
2965 * Unlike cache & spares, slogs are stored in the
2966 * ZPOOL_CONFIG_CHILDREN array. We filter them out here.
2968 (void) nvlist_lookup_uint64(child
[c
], ZPOOL_CONFIG_IS_LOG
,
2970 (void) nvlist_lookup_uint64(child
[c
], ZPOOL_CONFIG_IS_HOLE
,
2972 if (is_log
|| is_hole
) {
2974 * Create a hole vdev and put it in the config.
2976 if (nvlist_alloc(&vdev
, NV_UNIQUE_NAME
, 0) != 0)
2978 if (nvlist_add_string(vdev
, ZPOOL_CONFIG_TYPE
,
2979 VDEV_TYPE_HOLE
) != 0)
2981 if (nvlist_add_uint64(vdev
, ZPOOL_CONFIG_IS_HOLE
,
2986 varray
[vcount
++] = vdev
;
2990 verify(nvlist_lookup_string(child
[c
], ZPOOL_CONFIG_TYPE
, &type
)
2992 if (strcmp(type
, VDEV_TYPE_MIRROR
) != 0) {
2993 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
2994 "Source pool must be composed only of mirrors\n"));
2995 retval
= zfs_error(hdl
, EZFS_INVALCONFIG
, msg
);
2999 verify(nvlist_lookup_nvlist_array(child
[c
],
3000 ZPOOL_CONFIG_CHILDREN
, &mchild
, &mchildren
) == 0);
3002 /* find or add an entry for this top-level vdev */
3003 if (newchildren
> 0 &&
3004 (entry
= find_vdev_entry(zhp
, mchild
, mchildren
,
3005 newchild
, newchildren
)) >= 0) {
3006 /* We found a disk that the user specified. */
3007 vdev
= mchild
[entry
];
3010 /* User didn't specify a disk for this vdev. */
3011 vdev
= mchild
[mchildren
- 1];
3014 if (nvlist_dup(vdev
, &varray
[vcount
++], 0) != 0)
3018 /* did we find every disk the user specified? */
3019 if (found
!= newchildren
) {
3020 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
, "Device list must "
3021 "include at most one disk from each mirror"));
3022 retval
= zfs_error(hdl
, EZFS_INVALCONFIG
, msg
);
3026 /* Prepare the nvlist for populating. */
3027 if (*newroot
== NULL
) {
3028 if (nvlist_alloc(newroot
, NV_UNIQUE_NAME
, 0) != 0)
3031 if (nvlist_add_string(*newroot
, ZPOOL_CONFIG_TYPE
,
3032 VDEV_TYPE_ROOT
) != 0)
3035 verify(nvlist_remove_all(*newroot
, ZPOOL_CONFIG_CHILDREN
) == 0);
3038 /* Add all the children we found */
3039 if (nvlist_add_nvlist_array(*newroot
, ZPOOL_CONFIG_CHILDREN
, varray
,
3040 lastlog
== 0 ? vcount
: lastlog
) != 0)
3044 * If we're just doing a dry run, exit now with success.
3047 memory_err
= B_FALSE
;
3052 /* now build up the config list & call the ioctl */
3053 if (nvlist_alloc(&newconfig
, NV_UNIQUE_NAME
, 0) != 0)
3056 if (nvlist_add_nvlist(newconfig
,
3057 ZPOOL_CONFIG_VDEV_TREE
, *newroot
) != 0 ||
3058 nvlist_add_string(newconfig
,
3059 ZPOOL_CONFIG_POOL_NAME
, newname
) != 0 ||
3060 nvlist_add_uint64(newconfig
, ZPOOL_CONFIG_VERSION
, vers
) != 0)
3064 * The new pool is automatically part of the namespace unless we
3065 * explicitly export it.
3068 zc
.zc_cookie
= ZPOOL_EXPORT_AFTER_SPLIT
;
3069 (void) strlcpy(zc
.zc_name
, zhp
->zpool_name
, sizeof (zc
.zc_name
));
3070 (void) strlcpy(zc
.zc_string
, newname
, sizeof (zc
.zc_string
));
3071 if (zcmd_write_conf_nvlist(hdl
, &zc
, newconfig
) != 0)
3073 if (zc_props
!= NULL
&& zcmd_write_src_nvlist(hdl
, &zc
, zc_props
) != 0)
3076 if (zfs_ioctl(hdl
, ZFS_IOC_VDEV_SPLIT
, &zc
) != 0) {
3077 retval
= zpool_standard_error(hdl
, errno
, msg
);
3082 memory_err
= B_FALSE
;
3085 if (varray
!= NULL
) {
3088 for (v
= 0; v
< vcount
; v
++)
3089 nvlist_free(varray
[v
]);
3092 zcmd_free_nvlists(&zc
);
3093 nvlist_free(zc_props
);
3094 nvlist_free(newconfig
);
3096 nvlist_free(*newroot
);
3104 return (no_memory(hdl
));
3110 * Remove the given device.
3113 zpool_vdev_remove(zpool_handle_t
*zhp
, const char *path
)
3115 zfs_cmd_t zc
= { 0 };
3118 boolean_t avail_spare
, l2cache
, islog
;
3119 libzfs_handle_t
*hdl
= zhp
->zpool_hdl
;
3122 (void) snprintf(msg
, sizeof (msg
),
3123 dgettext(TEXT_DOMAIN
, "cannot remove %s"), path
);
3125 (void) strlcpy(zc
.zc_name
, zhp
->zpool_name
, sizeof (zc
.zc_name
));
3126 if ((tgt
= zpool_find_vdev(zhp
, path
, &avail_spare
, &l2cache
,
3128 return (zfs_error(hdl
, EZFS_NODEVICE
, msg
));
3130 version
= zpool_get_prop_int(zhp
, ZPOOL_PROP_VERSION
, NULL
);
3131 if (islog
&& version
< SPA_VERSION_HOLES
) {
3132 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
3133 "pool must be upgraded to support log removal"));
3134 return (zfs_error(hdl
, EZFS_BADVERSION
, msg
));
3137 if (!islog
&& !avail_spare
&& !l2cache
&& zpool_is_bootable(zhp
)) {
3138 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
3139 "root pool can not have removed devices, "
3140 "because GRUB does not understand them"));
3141 return (zfs_error(hdl
, EINVAL
, msg
));
3144 zc
.zc_guid
= fnvlist_lookup_uint64(tgt
, ZPOOL_CONFIG_GUID
);
3146 if (zfs_ioctl(hdl
, ZFS_IOC_VDEV_REMOVE
, &zc
) == 0)
3152 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
3153 "invalid config; all top-level vdevs must "
3154 "have the same sector size and not be raidz."));
3155 (void) zfs_error(hdl
, EZFS_INVALCONFIG
, msg
);
3159 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
3160 "Pool busy; removal may already be in progress"));
3161 (void) zfs_error(hdl
, EZFS_BUSY
, msg
);
3165 (void) zpool_standard_error(hdl
, errno
, msg
);
3171 zpool_vdev_remove_cancel(zpool_handle_t
*zhp
)
3173 zfs_cmd_t zc
= { 0 };
3175 libzfs_handle_t
*hdl
= zhp
->zpool_hdl
;
3177 (void) snprintf(msg
, sizeof (msg
),
3178 dgettext(TEXT_DOMAIN
, "cannot cancel removal"));
3180 (void) strlcpy(zc
.zc_name
, zhp
->zpool_name
, sizeof (zc
.zc_name
));
3183 if (zfs_ioctl(hdl
, ZFS_IOC_VDEV_REMOVE
, &zc
) == 0)
3186 return (zpool_standard_error(hdl
, errno
, msg
));
3190 zpool_vdev_indirect_size(zpool_handle_t
*zhp
, const char *path
,
3195 boolean_t avail_spare
, l2cache
, islog
;
3196 libzfs_handle_t
*hdl
= zhp
->zpool_hdl
;
3198 (void) snprintf(msg
, sizeof (msg
),
3199 dgettext(TEXT_DOMAIN
, "cannot determine indirect size of %s"),
3202 if ((tgt
= zpool_find_vdev(zhp
, path
, &avail_spare
, &l2cache
,
3204 return (zfs_error(hdl
, EZFS_NODEVICE
, msg
));
3206 if (avail_spare
|| l2cache
|| islog
) {
3211 if (nvlist_lookup_uint64(tgt
, ZPOOL_CONFIG_INDIRECT_SIZE
, sizep
) != 0) {
3212 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
3213 "indirect size not available"));
3214 return (zfs_error(hdl
, EINVAL
, msg
));
3220 * Clear the errors for the pool, or the particular device if specified.
3223 zpool_clear(zpool_handle_t
*zhp
, const char *path
, nvlist_t
*rewindnvl
)
3225 zfs_cmd_t zc
= { 0 };
3228 zpool_rewind_policy_t policy
;
3229 boolean_t avail_spare
, l2cache
;
3230 libzfs_handle_t
*hdl
= zhp
->zpool_hdl
;
3231 nvlist_t
*nvi
= NULL
;
3235 (void) snprintf(msg
, sizeof (msg
),
3236 dgettext(TEXT_DOMAIN
, "cannot clear errors for %s"),
3239 (void) snprintf(msg
, sizeof (msg
),
3240 dgettext(TEXT_DOMAIN
, "cannot clear errors for %s"),
3243 (void) strlcpy(zc
.zc_name
, zhp
->zpool_name
, sizeof (zc
.zc_name
));
3245 if ((tgt
= zpool_find_vdev(zhp
, path
, &avail_spare
,
3246 &l2cache
, NULL
)) == NULL
)
3247 return (zfs_error(hdl
, EZFS_NODEVICE
, msg
));
3250 * Don't allow error clearing for hot spares. Do allow
3251 * error clearing for l2cache devices.
3254 return (zfs_error(hdl
, EZFS_ISSPARE
, msg
));
3256 verify(nvlist_lookup_uint64(tgt
, ZPOOL_CONFIG_GUID
,
3260 zpool_get_rewind_policy(rewindnvl
, &policy
);
3261 zc
.zc_cookie
= policy
.zrp_request
;
3263 if (zcmd_alloc_dst_nvlist(hdl
, &zc
, zhp
->zpool_config_size
* 2) != 0)
3266 if (zcmd_write_src_nvlist(hdl
, &zc
, rewindnvl
) != 0)
3269 while ((error
= zfs_ioctl(hdl
, ZFS_IOC_CLEAR
, &zc
)) != 0 &&
3271 if (zcmd_expand_dst_nvlist(hdl
, &zc
) != 0) {
3272 zcmd_free_nvlists(&zc
);
3277 if (!error
|| ((policy
.zrp_request
& ZPOOL_TRY_REWIND
) &&
3278 errno
!= EPERM
&& errno
!= EACCES
)) {
3279 if (policy
.zrp_request
&
3280 (ZPOOL_DO_REWIND
| ZPOOL_TRY_REWIND
)) {
3281 (void) zcmd_read_dst_nvlist(hdl
, &zc
, &nvi
);
3282 zpool_rewind_exclaim(hdl
, zc
.zc_name
,
3283 ((policy
.zrp_request
& ZPOOL_TRY_REWIND
) != 0),
3287 zcmd_free_nvlists(&zc
);
3291 zcmd_free_nvlists(&zc
);
3292 return (zpool_standard_error(hdl
, errno
, msg
));
3296 * Similar to zpool_clear(), but takes a GUID (used by fmd).
3299 zpool_vdev_clear(zpool_handle_t
*zhp
, uint64_t guid
)
3301 zfs_cmd_t zc
= { 0 };
3303 libzfs_handle_t
*hdl
= zhp
->zpool_hdl
;
3305 (void) snprintf(msg
, sizeof (msg
),
3306 dgettext(TEXT_DOMAIN
, "cannot clear errors for %llx"),
3309 (void) strlcpy(zc
.zc_name
, zhp
->zpool_name
, sizeof (zc
.zc_name
));
3311 zc
.zc_cookie
= ZPOOL_NO_REWIND
;
3313 if (ioctl(hdl
->libzfs_fd
, ZFS_IOC_CLEAR
, &zc
) == 0)
3316 return (zpool_standard_error(hdl
, errno
, msg
));
3320 * Change the GUID for a pool.
3323 zpool_reguid(zpool_handle_t
*zhp
)
3326 libzfs_handle_t
*hdl
= zhp
->zpool_hdl
;
3327 zfs_cmd_t zc
= { 0 };
3329 (void) snprintf(msg
, sizeof (msg
),
3330 dgettext(TEXT_DOMAIN
, "cannot reguid '%s'"), zhp
->zpool_name
);
3332 (void) strlcpy(zc
.zc_name
, zhp
->zpool_name
, sizeof (zc
.zc_name
));
3333 if (zfs_ioctl(hdl
, ZFS_IOC_POOL_REGUID
, &zc
) == 0)
3336 return (zpool_standard_error(hdl
, errno
, msg
));
3343 zpool_reopen(zpool_handle_t
*zhp
)
3345 zfs_cmd_t zc
= { 0 };
3347 libzfs_handle_t
*hdl
= zhp
->zpool_hdl
;
3349 (void) snprintf(msg
, sizeof (msg
),
3350 dgettext(TEXT_DOMAIN
, "cannot reopen '%s'"),
3353 (void) strlcpy(zc
.zc_name
, zhp
->zpool_name
, sizeof (zc
.zc_name
));
3354 if (zfs_ioctl(hdl
, ZFS_IOC_POOL_REOPEN
, &zc
) == 0)
3356 return (zpool_standard_error(hdl
, errno
, msg
));
3360 * Convert from a devid string to a path.
3363 devid_to_path(char *devid_str
)
3368 devid_nmlist_t
*list
= NULL
;
3371 if (devid_str_decode(devid_str
, &devid
, &minor
) != 0)
3374 ret
= devid_deviceid_to_nmlist("/dev", devid
, minor
, &list
);
3376 devid_str_free(minor
);
3383 * In a case the strdup() fails, we will just return NULL below.
3385 path
= strdup(list
[0].devname
);
3387 devid_free_nmlist(list
);
3393 * Convert from a path to a devid string.
3396 path_to_devid(const char *path
)
3402 if ((fd
= open(path
, O_RDONLY
)) < 0)
3407 if (devid_get(fd
, &devid
) == 0) {
3408 if (devid_get_minor_name(fd
, &minor
) == 0)
3409 ret
= devid_str_encode(devid
, minor
);
3411 devid_str_free(minor
);
3420 * Issue the necessary ioctl() to update the stored path value for the vdev. We
3421 * ignore any failure here, since a common case is for an unprivileged user to
3422 * type 'zpool status', and we'll display the correct information anyway.
3425 set_path(zpool_handle_t
*zhp
, nvlist_t
*nv
, const char *path
)
3427 zfs_cmd_t zc
= { 0 };
3429 (void) strncpy(zc
.zc_name
, zhp
->zpool_name
, sizeof (zc
.zc_name
));
3430 (void) strncpy(zc
.zc_value
, path
, sizeof (zc
.zc_value
));
3431 verify(nvlist_lookup_uint64(nv
, ZPOOL_CONFIG_GUID
,
3434 (void) ioctl(zhp
->zpool_hdl
->libzfs_fd
, ZFS_IOC_VDEV_SETPATH
, &zc
);
3438 * Given a vdev, return the name to display in iostat. If the vdev has a path,
3439 * we use that, stripping off any leading "/dev/dsk/"; if not, we use the type.
3440 * We also check if this is a whole disk, in which case we strip off the
3441 * trailing 's0' slice name.
3443 * This routine is also responsible for identifying when disks have been
3444 * reconfigured in a new location. The kernel will have opened the device by
3445 * devid, but the path will still refer to the old location. To catch this, we
3446 * first do a path -> devid translation (which is fast for the common case). If
3447 * the devid matches, we're done. If not, we do a reverse devid -> path
3448 * translation and issue the appropriate ioctl() to update the path of the vdev.
3449 * If 'zhp' is NULL, then this is an exported pool, and we don't need to do any
3453 zpool_vdev_name(libzfs_handle_t
*hdl
, zpool_handle_t
*zhp
, nvlist_t
*nv
,
3462 if (nvlist_lookup_uint64(nv
, ZPOOL_CONFIG_NOT_PRESENT
,
3464 verify(nvlist_lookup_uint64(nv
, ZPOOL_CONFIG_GUID
,
3466 (void) snprintf(buf
, sizeof (buf
), "%llu",
3467 (u_longlong_t
)value
);
3469 } else if (nvlist_lookup_string(nv
, ZPOOL_CONFIG_PATH
, &path
) == 0) {
3472 * If the device is dead (faulted, offline, etc) then don't
3473 * bother opening it. Otherwise we may be forcing the user to
3474 * open a misbehaving device, which can have undesirable
3477 if ((nvlist_lookup_uint64_array(nv
, ZPOOL_CONFIG_VDEV_STATS
,
3478 (uint64_t **)&vs
, &vsc
) != 0 ||
3479 vs
->vs_state
>= VDEV_STATE_DEGRADED
) &&
3481 nvlist_lookup_string(nv
, ZPOOL_CONFIG_DEVID
, &devid
) == 0) {
3483 * Determine if the current path is correct.
3485 char *newdevid
= path_to_devid(path
);
3487 if (newdevid
== NULL
||
3488 strcmp(devid
, newdevid
) != 0) {
3491 if ((newpath
= devid_to_path(devid
)) != NULL
) {
3493 * Update the path appropriately.
3495 set_path(zhp
, nv
, newpath
);
3496 if (nvlist_add_string(nv
,
3497 ZPOOL_CONFIG_PATH
, newpath
) == 0)
3498 verify(nvlist_lookup_string(nv
,
3506 devid_str_free(newdevid
);
3509 if (strncmp(path
, ZFS_DISK_ROOTD
, strlen(ZFS_DISK_ROOTD
)) == 0)
3510 path
+= strlen(ZFS_DISK_ROOTD
);
3512 if (nvlist_lookup_uint64(nv
, ZPOOL_CONFIG_WHOLE_DISK
,
3513 &value
) == 0 && value
) {
3514 int pathlen
= strlen(path
);
3515 char *tmp
= zfs_strdup(hdl
, path
);
3518 * If it starts with c#, and ends with "s0" or "s1",
3519 * chop the slice off, or if it ends with "s0/old" or
3520 * "s1/old", remove the slice from the middle.
3522 if (CTD_CHECK(tmp
)) {
3523 if (strcmp(&tmp
[pathlen
- 2], "s0") == 0 ||
3524 strcmp(&tmp
[pathlen
- 2], "s1") == 0) {
3525 tmp
[pathlen
- 2] = '\0';
3526 } else if (pathlen
> 6 &&
3527 (strcmp(&tmp
[pathlen
- 6], "s0/old") == 0 ||
3528 strcmp(&tmp
[pathlen
- 6], "s1/old") == 0)) {
3529 (void) strcpy(&tmp
[pathlen
- 6],
3536 verify(nvlist_lookup_string(nv
, ZPOOL_CONFIG_TYPE
, &path
) == 0);
3539 * If it's a raidz device, we need to stick in the parity level.
3541 if (strcmp(path
, VDEV_TYPE_RAIDZ
) == 0) {
3542 verify(nvlist_lookup_uint64(nv
, ZPOOL_CONFIG_NPARITY
,
3544 (void) snprintf(buf
, sizeof (buf
), "%s%llu", path
,
3545 (u_longlong_t
)value
);
3550 * We identify each top-level vdev by using a <type-id>
3551 * naming convention.
3556 verify(nvlist_lookup_uint64(nv
, ZPOOL_CONFIG_ID
,
3558 (void) snprintf(buf
, sizeof (buf
), "%s-%llu", path
,
3564 return (zfs_strdup(hdl
, path
));
3568 zbookmark_mem_compare(const void *a
, const void *b
)
3570 return (memcmp(a
, b
, sizeof (zbookmark_phys_t
)));
3574 * Retrieve the persistent error log, uniquify the members, and return to the
3578 zpool_get_errlog(zpool_handle_t
*zhp
, nvlist_t
**nverrlistp
)
3580 zfs_cmd_t zc
= { 0 };
3582 zbookmark_phys_t
*zb
= NULL
;
3586 * Retrieve the raw error list from the kernel. If the number of errors
3587 * has increased, allocate more space and continue until we get the
3590 verify(nvlist_lookup_uint64(zhp
->zpool_config
, ZPOOL_CONFIG_ERRCOUNT
,
3594 if ((zc
.zc_nvlist_dst
= (uintptr_t)zfs_alloc(zhp
->zpool_hdl
,
3595 count
* sizeof (zbookmark_phys_t
))) == (uintptr_t)NULL
)
3597 zc
.zc_nvlist_dst_size
= count
;
3598 (void) strcpy(zc
.zc_name
, zhp
->zpool_name
);
3600 if (ioctl(zhp
->zpool_hdl
->libzfs_fd
, ZFS_IOC_ERROR_LOG
,
3602 free((void *)(uintptr_t)zc
.zc_nvlist_dst
);
3603 if (errno
== ENOMEM
) {
3606 count
= zc
.zc_nvlist_dst_size
;
3607 dst
= zfs_alloc(zhp
->zpool_hdl
, count
*
3608 sizeof (zbookmark_phys_t
));
3611 zc
.zc_nvlist_dst
= (uintptr_t)dst
;
3621 * Sort the resulting bookmarks. This is a little confusing due to the
3622 * implementation of ZFS_IOC_ERROR_LOG. The bookmarks are copied last
3623 * to first, and 'zc_nvlist_dst_size' indicates the number of boomarks
3624 * _not_ copied as part of the process. So we point the start of our
3625 * array appropriate and decrement the total number of elements.
3627 zb
= ((zbookmark_phys_t
*)(uintptr_t)zc
.zc_nvlist_dst
) +
3628 zc
.zc_nvlist_dst_size
;
3629 count
-= zc
.zc_nvlist_dst_size
;
3631 qsort(zb
, count
, sizeof (zbookmark_phys_t
), zbookmark_mem_compare
);
3633 verify(nvlist_alloc(nverrlistp
, 0, KM_SLEEP
) == 0);
3636 * Fill in the nverrlistp with nvlist's of dataset and object numbers.
3638 for (i
= 0; i
< count
; i
++) {
3641 /* ignoring zb_blkid and zb_level for now */
3642 if (i
> 0 && zb
[i
-1].zb_objset
== zb
[i
].zb_objset
&&
3643 zb
[i
-1].zb_object
== zb
[i
].zb_object
)
3646 if (nvlist_alloc(&nv
, NV_UNIQUE_NAME
, KM_SLEEP
) != 0)
3648 if (nvlist_add_uint64(nv
, ZPOOL_ERR_DATASET
,
3649 zb
[i
].zb_objset
) != 0) {
3653 if (nvlist_add_uint64(nv
, ZPOOL_ERR_OBJECT
,
3654 zb
[i
].zb_object
) != 0) {
3658 if (nvlist_add_nvlist(*nverrlistp
, "ejk", nv
) != 0) {
3665 free((void *)(uintptr_t)zc
.zc_nvlist_dst
);
3669 free((void *)(uintptr_t)zc
.zc_nvlist_dst
);
3670 return (no_memory(zhp
->zpool_hdl
));
3674 * Upgrade a ZFS pool to the latest on-disk version.
3677 zpool_upgrade(zpool_handle_t
*zhp
, uint64_t new_version
)
3679 zfs_cmd_t zc
= { 0 };
3680 libzfs_handle_t
*hdl
= zhp
->zpool_hdl
;
3682 (void) strcpy(zc
.zc_name
, zhp
->zpool_name
);
3683 zc
.zc_cookie
= new_version
;
3685 if (zfs_ioctl(hdl
, ZFS_IOC_POOL_UPGRADE
, &zc
) != 0)
3686 return (zpool_standard_error_fmt(hdl
, errno
,
3687 dgettext(TEXT_DOMAIN
, "cannot upgrade '%s'"),
3693 zfs_save_arguments(int argc
, char **argv
, char *string
, int len
)
3695 (void) strlcpy(string
, basename(argv
[0]), len
);
3696 for (int i
= 1; i
< argc
; i
++) {
3697 (void) strlcat(string
, " ", len
);
3698 (void) strlcat(string
, argv
[i
], len
);
3703 zpool_log_history(libzfs_handle_t
*hdl
, const char *message
)
3705 zfs_cmd_t zc
= { 0 };
3709 args
= fnvlist_alloc();
3710 fnvlist_add_string(args
, "message", message
);
3711 err
= zcmd_write_src_nvlist(hdl
, &zc
, args
);
3713 err
= ioctl(hdl
->libzfs_fd
, ZFS_IOC_LOG_HISTORY
, &zc
);
3715 zcmd_free_nvlists(&zc
);
3720 * Perform ioctl to get some command history of a pool.
3722 * 'buf' is the buffer to fill up to 'len' bytes. 'off' is the
3723 * logical offset of the history buffer to start reading from.
3725 * Upon return, 'off' is the next logical offset to read from and
3726 * 'len' is the actual amount of bytes read into 'buf'.
3729 get_history(zpool_handle_t
*zhp
, char *buf
, uint64_t *off
, uint64_t *len
)
3731 zfs_cmd_t zc
= { 0 };
3732 libzfs_handle_t
*hdl
= zhp
->zpool_hdl
;
3734 (void) strlcpy(zc
.zc_name
, zhp
->zpool_name
, sizeof (zc
.zc_name
));
3736 zc
.zc_history
= (uint64_t)(uintptr_t)buf
;
3737 zc
.zc_history_len
= *len
;
3738 zc
.zc_history_offset
= *off
;
3740 if (ioctl(hdl
->libzfs_fd
, ZFS_IOC_POOL_GET_HISTORY
, &zc
) != 0) {
3743 return (zfs_error_fmt(hdl
, EZFS_PERM
,
3744 dgettext(TEXT_DOMAIN
,
3745 "cannot show history for pool '%s'"),
3748 return (zfs_error_fmt(hdl
, EZFS_NOHISTORY
,
3749 dgettext(TEXT_DOMAIN
, "cannot get history for pool "
3750 "'%s'"), zhp
->zpool_name
));
3752 return (zfs_error_fmt(hdl
, EZFS_BADVERSION
,
3753 dgettext(TEXT_DOMAIN
, "cannot get history for pool "
3754 "'%s', pool must be upgraded"), zhp
->zpool_name
));
3756 return (zpool_standard_error_fmt(hdl
, errno
,
3757 dgettext(TEXT_DOMAIN
,
3758 "cannot get history for '%s'"), zhp
->zpool_name
));
3762 *len
= zc
.zc_history_len
;
3763 *off
= zc
.zc_history_offset
;
3769 * Process the buffer of nvlists, unpacking and storing each nvlist record
3770 * into 'records'. 'leftover' is set to the number of bytes that weren't
3771 * processed as there wasn't a complete record.
3774 zpool_history_unpack(char *buf
, uint64_t bytes_read
, uint64_t *leftover
,
3775 nvlist_t
***records
, uint_t
*numrecords
)
3781 while (bytes_read
> sizeof (reclen
)) {
3783 /* get length of packed record (stored as little endian) */
3784 for (i
= 0, reclen
= 0; i
< sizeof (reclen
); i
++)
3785 reclen
+= (uint64_t)(((uchar_t
*)buf
)[i
]) << (8*i
);
3787 if (bytes_read
< sizeof (reclen
) + reclen
)
3791 if (nvlist_unpack(buf
+ sizeof (reclen
), reclen
, &nv
, 0) != 0)
3793 bytes_read
-= sizeof (reclen
) + reclen
;
3794 buf
+= sizeof (reclen
) + reclen
;
3796 /* add record to nvlist array */
3798 if (ISP2(*numrecords
+ 1)) {
3799 *records
= realloc(*records
,
3800 *numrecords
* 2 * sizeof (nvlist_t
*));
3802 (*records
)[*numrecords
- 1] = nv
;
3805 *leftover
= bytes_read
;
3810 * Retrieve the command history of a pool.
3813 zpool_get_history(zpool_handle_t
*zhp
, nvlist_t
**nvhisp
)
3816 int buflen
= 128 * 1024;
3818 nvlist_t
**records
= NULL
;
3819 uint_t numrecords
= 0;
3822 buf
= malloc(buflen
);
3826 uint64_t bytes_read
= buflen
;
3829 if ((err
= get_history(zhp
, buf
, &off
, &bytes_read
)) != 0)
3832 /* if nothing else was read in, we're at EOF, just return */
3836 if ((err
= zpool_history_unpack(buf
, bytes_read
,
3837 &leftover
, &records
, &numrecords
)) != 0)
3840 if (leftover
== bytes_read
) {
3842 * no progress made, because buffer is not big enough
3843 * to hold this record; resize and retry.
3847 buf
= malloc(buflen
);
3858 verify(nvlist_alloc(nvhisp
, NV_UNIQUE_NAME
, 0) == 0);
3859 verify(nvlist_add_nvlist_array(*nvhisp
, ZPOOL_HIST_RECORD
,
3860 records
, numrecords
) == 0);
3862 for (i
= 0; i
< numrecords
; i
++)
3863 nvlist_free(records
[i
]);
3870 zpool_obj_to_path(zpool_handle_t
*zhp
, uint64_t dsobj
, uint64_t obj
,
3871 char *pathname
, size_t len
)
3873 zfs_cmd_t zc
= { 0 };
3874 boolean_t mounted
= B_FALSE
;
3875 char *mntpnt
= NULL
;
3876 char dsname
[ZFS_MAX_DATASET_NAME_LEN
];
3879 /* special case for the MOS */
3880 (void) snprintf(pathname
, len
, "<metadata>:<0x%llx>", obj
);
3884 /* get the dataset's name */
3885 (void) strlcpy(zc
.zc_name
, zhp
->zpool_name
, sizeof (zc
.zc_name
));
3887 if (ioctl(zhp
->zpool_hdl
->libzfs_fd
,
3888 ZFS_IOC_DSOBJ_TO_DSNAME
, &zc
) != 0) {
3889 /* just write out a path of two object numbers */
3890 (void) snprintf(pathname
, len
, "<0x%llx>:<0x%llx>",
3894 (void) strlcpy(dsname
, zc
.zc_value
, sizeof (dsname
));
3896 /* find out if the dataset is mounted */
3897 mounted
= is_mounted(zhp
->zpool_hdl
, dsname
, &mntpnt
);
3899 /* get the corrupted object's path */
3900 (void) strlcpy(zc
.zc_name
, dsname
, sizeof (zc
.zc_name
));
3902 if (ioctl(zhp
->zpool_hdl
->libzfs_fd
, ZFS_IOC_OBJ_TO_PATH
,
3905 (void) snprintf(pathname
, len
, "%s%s", mntpnt
,
3908 (void) snprintf(pathname
, len
, "%s:%s",
3909 dsname
, zc
.zc_value
);
3912 (void) snprintf(pathname
, len
, "%s:<0x%llx>", dsname
, obj
);
3918 * Read the EFI label from the config, if a label does not exist then
3919 * pass back the error to the caller. If the caller has passed a non-NULL
3920 * diskaddr argument then we set it to the starting address of the EFI
3921 * partition. If the caller has passed a non-NULL boolean argument, then
3922 * we set it to indicate if the disk does have efi system partition.
3925 read_efi_label(nvlist_t
*config
, diskaddr_t
*sb
, boolean_t
*system
)
3929 char diskname
[MAXPATHLEN
];
3930 boolean_t boot
= B_FALSE
;
3934 if (nvlist_lookup_string(config
, ZPOOL_CONFIG_PATH
, &path
) != 0)
3937 (void) snprintf(diskname
, sizeof (diskname
), "%s%s", ZFS_RDISK_ROOT
,
3938 strrchr(path
, '/'));
3939 if ((fd
= open(diskname
, O_RDONLY
|O_NDELAY
)) >= 0) {
3940 struct dk_gpt
*vtoc
;
3942 if ((err
= efi_alloc_and_read(fd
, &vtoc
)) >= 0) {
3943 for (slice
= 0; slice
< vtoc
->efi_nparts
; slice
++) {
3944 if (vtoc
->efi_parts
[slice
].p_tag
== V_SYSTEM
)
3946 if (vtoc
->efi_parts
[slice
].p_tag
== V_USR
)
3949 if (sb
!= NULL
&& vtoc
->efi_parts
[slice
].p_tag
== V_USR
)
3950 *sb
= vtoc
->efi_parts
[slice
].p_start
;
3961 * determine where a partition starts on a disk in the current
3965 find_start_block(nvlist_t
*config
)
3969 diskaddr_t sb
= MAXOFFSET_T
;
3972 if (nvlist_lookup_nvlist_array(config
,
3973 ZPOOL_CONFIG_CHILDREN
, &child
, &children
) != 0) {
3974 if (nvlist_lookup_uint64(config
,
3975 ZPOOL_CONFIG_WHOLE_DISK
,
3976 &wholedisk
) != 0 || !wholedisk
) {
3977 return (MAXOFFSET_T
);
3979 if (read_efi_label(config
, &sb
, NULL
) < 0)
3984 for (c
= 0; c
< children
; c
++) {
3985 sb
= find_start_block(child
[c
]);
3986 if (sb
!= MAXOFFSET_T
) {
3990 return (MAXOFFSET_T
);
3994 * Label an individual disk. The name provided is the short name,
3995 * stripped of any leading /dev path.
3998 zpool_label_disk(libzfs_handle_t
*hdl
, zpool_handle_t
*zhp
, const char *name
,
3999 zpool_boot_label_t boot_type
, uint64_t boot_size
, int *slice
)
4001 char path
[MAXPATHLEN
];
4002 struct dk_gpt
*vtoc
;
4004 size_t resv
= EFI_MIN_RESV_SIZE
;
4005 uint64_t slice_size
;
4006 diskaddr_t start_block
;
4009 /* prepare an error message just in case */
4010 (void) snprintf(errbuf
, sizeof (errbuf
),
4011 dgettext(TEXT_DOMAIN
, "cannot label '%s'"), name
);
4016 verify(nvlist_lookup_nvlist(zhp
->zpool_config
,
4017 ZPOOL_CONFIG_VDEV_TREE
, &nvroot
) == 0);
4019 if (zhp
->zpool_start_block
== 0)
4020 start_block
= find_start_block(nvroot
);
4022 start_block
= zhp
->zpool_start_block
;
4023 zhp
->zpool_start_block
= start_block
;
4026 start_block
= NEW_START_BLOCK
;
4029 (void) snprintf(path
, sizeof (path
), "%s/%s%s", ZFS_RDISK_ROOT
, name
,
4032 if ((fd
= open(path
, O_RDWR
| O_NDELAY
)) < 0) {
4034 * This shouldn't happen. We've long since verified that this
4035 * is a valid device.
4038 dgettext(TEXT_DOMAIN
, "unable to open device"));
4039 return (zfs_error(hdl
, EZFS_OPENFAILED
, errbuf
));
4042 if (efi_alloc_and_init(fd
, EFI_NUMPAR
, &vtoc
) != 0) {
4044 * The only way this can fail is if we run out of memory, or we
4045 * were unable to read the disk's capacity
4047 if (errno
== ENOMEM
)
4048 (void) no_memory(hdl
);
4051 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
4052 "unable to read disk capacity"), name
);
4054 return (zfs_error(hdl
, EZFS_NOCAP
, errbuf
));
4058 * Why we use V_USR: V_BACKUP confuses users, and is considered
4059 * disposable by some EFI utilities (since EFI doesn't have a backup
4060 * slice). V_UNASSIGNED is supposed to be used only for zero size
4061 * partitions, and efi_write() will fail if we use it. V_ROOT, V_BOOT,
4062 * etc. were all pretty specific. V_USR is as close to reality as we
4063 * can get, in the absence of V_OTHER.
4065 /* first fix the partition start block */
4066 if (start_block
== MAXOFFSET_T
)
4067 start_block
= NEW_START_BLOCK
;
4070 * EFI System partition is using slice 0.
4071 * ZFS is on slice 1 and slice 8 is reserved.
4072 * We assume the GPT partition table without system
4073 * partition has zfs p_start == NEW_START_BLOCK.
4074 * If start_block != NEW_START_BLOCK, it means we have
4075 * system partition. Correct solution would be to query/cache vtoc
4076 * from existing vdev member.
4078 if (boot_type
== ZPOOL_CREATE_BOOT_LABEL
) {
4079 if (boot_size
% vtoc
->efi_lbasize
!= 0) {
4080 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
4081 "boot partition size must be a multiple of %d"),
4085 return (zfs_error(hdl
, EZFS_LABELFAILED
, errbuf
));
4088 * System partition size checks.
4089 * Note the 1MB is quite arbitrary value, since we
4090 * are creating dedicated pool, it should be enough
4091 * to hold fat + efi bootloader. May need to be
4092 * adjusted if the bootloader size will grow.
4094 if (boot_size
< 1024 * 1024) {
4096 zfs_nicenum(boot_size
, buf
, sizeof (buf
));
4097 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
4098 "Specified size %s for EFI System partition is too "
4099 "small, the minimum size is 1MB."), buf
);
4102 return (zfs_error(hdl
, EZFS_LABELFAILED
, errbuf
));
4104 /* 33MB is tested with mkfs -F pcfs */
4105 if (hdl
->libzfs_printerr
&&
4106 ((vtoc
->efi_lbasize
== 512 &&
4107 boot_size
< 33 * 1024 * 1024) ||
4108 (vtoc
->efi_lbasize
== 4096 &&
4109 boot_size
< 256 * 1024 * 1024))) {
4111 zfs_nicenum(boot_size
, buf
, sizeof (buf
));
4112 (void) fprintf(stderr
, dgettext(TEXT_DOMAIN
,
4113 "Warning: EFI System partition size %s is "
4114 "not allowing to create FAT32 file\nsystem, which "
4115 "may result in unbootable system.\n"), buf
);
4117 /* Adjust zfs partition start by size of system partition. */
4118 start_block
+= boot_size
/ vtoc
->efi_lbasize
;
4121 if (start_block
== NEW_START_BLOCK
) {
4123 * Use default layout.
4124 * ZFS is on slice 0 and slice 8 is reserved.
4126 slice_size
= vtoc
->efi_last_u_lba
+ 1;
4127 slice_size
-= EFI_MIN_RESV_SIZE
;
4128 slice_size
-= start_block
;
4132 vtoc
->efi_parts
[0].p_start
= start_block
;
4133 vtoc
->efi_parts
[0].p_size
= slice_size
;
4135 vtoc
->efi_parts
[0].p_tag
= V_USR
;
4136 (void) strcpy(vtoc
->efi_parts
[0].p_name
, "zfs");
4138 vtoc
->efi_parts
[8].p_start
= slice_size
+ start_block
;
4139 vtoc
->efi_parts
[8].p_size
= resv
;
4140 vtoc
->efi_parts
[8].p_tag
= V_RESERVED
;
4142 slice_size
= start_block
- NEW_START_BLOCK
;
4143 vtoc
->efi_parts
[0].p_start
= NEW_START_BLOCK
;
4144 vtoc
->efi_parts
[0].p_size
= slice_size
;
4145 vtoc
->efi_parts
[0].p_tag
= V_SYSTEM
;
4146 (void) strcpy(vtoc
->efi_parts
[0].p_name
, "loader");
4149 /* prepare slice 1 */
4150 slice_size
= vtoc
->efi_last_u_lba
+ 1 - slice_size
;
4152 slice_size
-= NEW_START_BLOCK
;
4153 vtoc
->efi_parts
[1].p_start
= start_block
;
4154 vtoc
->efi_parts
[1].p_size
= slice_size
;
4155 vtoc
->efi_parts
[1].p_tag
= V_USR
;
4156 (void) strcpy(vtoc
->efi_parts
[1].p_name
, "zfs");
4158 vtoc
->efi_parts
[8].p_start
= slice_size
+ start_block
;
4159 vtoc
->efi_parts
[8].p_size
= resv
;
4160 vtoc
->efi_parts
[8].p_tag
= V_RESERVED
;
4163 if (efi_write(fd
, vtoc
) != 0) {
4165 * Some block drivers (like pcata) may not support EFI
4166 * GPT labels. Print out a helpful error message dir-
4167 * ecting the user to manually label the disk and give
4173 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
4174 "try using fdisk(1M) and then provide a specific slice"));
4175 return (zfs_error(hdl
, EZFS_LABELFAILED
, errbuf
));
4184 supported_dump_vdev_type(libzfs_handle_t
*hdl
, nvlist_t
*config
, char *errbuf
)
4190 verify(nvlist_lookup_string(config
, ZPOOL_CONFIG_TYPE
, &type
) == 0);
4191 if (strcmp(type
, VDEV_TYPE_FILE
) == 0 ||
4192 strcmp(type
, VDEV_TYPE_HOLE
) == 0 ||
4193 strcmp(type
, VDEV_TYPE_MISSING
) == 0) {
4194 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
4195 "vdev type '%s' is not supported"), type
);
4196 (void) zfs_error(hdl
, EZFS_VDEVNOTSUP
, errbuf
);
4199 if (nvlist_lookup_nvlist_array(config
, ZPOOL_CONFIG_CHILDREN
,
4200 &child
, &children
) == 0) {
4201 for (c
= 0; c
< children
; c
++) {
4202 if (!supported_dump_vdev_type(hdl
, child
[c
], errbuf
))
4210 * Check if this zvol is allowable for use as a dump device; zero if
4211 * it is, > 0 if it isn't, < 0 if it isn't a zvol.
4213 * Allowable storage configurations include mirrors, all raidz variants, and
4214 * pools with log, cache, and spare devices. Pools which are backed by files or
4215 * have missing/hole vdevs are not suitable.
4218 zvol_check_dump_config(char *arg
)
4220 zpool_handle_t
*zhp
= NULL
;
4221 nvlist_t
*config
, *nvroot
;
4225 libzfs_handle_t
*hdl
;
4227 char poolname
[ZFS_MAX_DATASET_NAME_LEN
];
4228 int pathlen
= strlen(ZVOL_FULL_DEV_DIR
);
4231 if (strncmp(arg
, ZVOL_FULL_DEV_DIR
, pathlen
)) {
4235 (void) snprintf(errbuf
, sizeof (errbuf
), dgettext(TEXT_DOMAIN
,
4236 "dump is not supported on device '%s'"), arg
);
4238 if ((hdl
= libzfs_init()) == NULL
)
4240 libzfs_print_on_error(hdl
, B_TRUE
);
4242 volname
= arg
+ pathlen
;
4244 /* check the configuration of the pool */
4245 if ((p
= strchr(volname
, '/')) == NULL
) {
4246 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
4247 "malformed dataset name"));
4248 (void) zfs_error(hdl
, EZFS_INVALIDNAME
, errbuf
);
4250 } else if (p
- volname
>= ZFS_MAX_DATASET_NAME_LEN
) {
4251 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
4252 "dataset name is too long"));
4253 (void) zfs_error(hdl
, EZFS_NAMETOOLONG
, errbuf
);
4256 (void) strncpy(poolname
, volname
, p
- volname
);
4257 poolname
[p
- volname
] = '\0';
4260 if ((zhp
= zpool_open(hdl
, poolname
)) == NULL
) {
4261 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
4262 "could not open pool '%s'"), poolname
);
4263 (void) zfs_error(hdl
, EZFS_OPENFAILED
, errbuf
);
4266 config
= zpool_get_config(zhp
, NULL
);
4267 if (nvlist_lookup_nvlist(config
, ZPOOL_CONFIG_VDEV_TREE
,
4269 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
4270 "could not obtain vdev configuration for '%s'"), poolname
);
4271 (void) zfs_error(hdl
, EZFS_INVALCONFIG
, errbuf
);
4275 verify(nvlist_lookup_nvlist_array(nvroot
, ZPOOL_CONFIG_CHILDREN
,
4276 &top
, &toplevels
) == 0);
4278 if (!supported_dump_vdev_type(hdl
, top
[0], errbuf
)) {