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 2009 Sun Microsystems, Inc. All rights reserved.
24 * Use is subject to license terms.
40 #include <sys/efi_partition.h>
42 #include <sys/zfs_ioctl.h>
46 #include "zfs_namecheck.h"
48 #include "libzfs_impl.h"
50 static int read_efi_label(nvlist_t
*config
, diskaddr_t
*sb
);
52 #if defined(__i386) || defined(__amd64)
53 #define BOOTCMD "installgrub(1M)"
55 #define BOOTCMD "installboot(1M)"
59 * ====================================================================
60 * zpool property functions
61 * ====================================================================
65 zpool_get_all_props(zpool_handle_t
*zhp
)
68 libzfs_handle_t
*hdl
= zhp
->zpool_hdl
;
70 (void) strlcpy(zc
.zc_name
, zhp
->zpool_name
, sizeof (zc
.zc_name
));
72 if (zcmd_alloc_dst_nvlist(hdl
, &zc
, 0) != 0)
75 while (ioctl(hdl
->libzfs_fd
, ZFS_IOC_POOL_GET_PROPS
, &zc
) != 0) {
76 if (errno
== ENOMEM
) {
77 if (zcmd_expand_dst_nvlist(hdl
, &zc
) != 0) {
78 zcmd_free_nvlists(&zc
);
82 zcmd_free_nvlists(&zc
);
87 if (zcmd_read_dst_nvlist(hdl
, &zc
, &zhp
->zpool_props
) != 0) {
88 zcmd_free_nvlists(&zc
);
92 zcmd_free_nvlists(&zc
);
98 zpool_props_refresh(zpool_handle_t
*zhp
)
102 old_props
= zhp
->zpool_props
;
104 if (zpool_get_all_props(zhp
) != 0)
107 nvlist_free(old_props
);
112 zpool_get_prop_string(zpool_handle_t
*zhp
, zpool_prop_t prop
,
118 zprop_source_t source
;
120 nvl
= zhp
->zpool_props
;
121 if (nvlist_lookup_nvlist(nvl
, zpool_prop_to_name(prop
), &nv
) == 0) {
122 verify(nvlist_lookup_uint64(nv
, ZPROP_SOURCE
, &ival
) == 0);
124 verify(nvlist_lookup_string(nv
, ZPROP_VALUE
, &value
) == 0);
126 source
= ZPROP_SRC_DEFAULT
;
127 if ((value
= (char *)zpool_prop_default_string(prop
)) == NULL
)
138 zpool_get_prop_int(zpool_handle_t
*zhp
, zpool_prop_t prop
, zprop_source_t
*src
)
142 zprop_source_t source
;
144 if (zhp
->zpool_props
== NULL
&& zpool_get_all_props(zhp
)) {
146 * zpool_get_all_props() has most likely failed because
147 * the pool is faulted, but if all we need is the top level
148 * vdev's guid then get it from the zhp config nvlist.
150 if ((prop
== ZPOOL_PROP_GUID
) &&
151 (nvlist_lookup_nvlist(zhp
->zpool_config
,
152 ZPOOL_CONFIG_VDEV_TREE
, &nv
) == 0) &&
153 (nvlist_lookup_uint64(nv
, ZPOOL_CONFIG_GUID
, &value
)
157 return (zpool_prop_default_numeric(prop
));
160 nvl
= zhp
->zpool_props
;
161 if (nvlist_lookup_nvlist(nvl
, zpool_prop_to_name(prop
), &nv
) == 0) {
162 verify(nvlist_lookup_uint64(nv
, ZPROP_SOURCE
, &value
) == 0);
164 verify(nvlist_lookup_uint64(nv
, ZPROP_VALUE
, &value
) == 0);
166 source
= ZPROP_SRC_DEFAULT
;
167 value
= zpool_prop_default_numeric(prop
);
177 * Map VDEV STATE to printed strings.
180 zpool_state_to_name(vdev_state_t state
, vdev_aux_t aux
)
183 case VDEV_STATE_CLOSED
:
184 case VDEV_STATE_OFFLINE
:
185 return (gettext("OFFLINE"));
186 case VDEV_STATE_REMOVED
:
187 return (gettext("REMOVED"));
188 case VDEV_STATE_CANT_OPEN
:
189 if (aux
== VDEV_AUX_CORRUPT_DATA
|| aux
== VDEV_AUX_BAD_LOG
)
190 return (gettext("FAULTED"));
192 return (gettext("UNAVAIL"));
193 case VDEV_STATE_FAULTED
:
194 return (gettext("FAULTED"));
195 case VDEV_STATE_DEGRADED
:
196 return (gettext("DEGRADED"));
197 case VDEV_STATE_HEALTHY
:
198 return (gettext("ONLINE"));
201 return (gettext("UNKNOWN"));
205 * Get a zpool property value for 'prop' and return the value in
206 * a pre-allocated buffer.
209 zpool_get_prop(zpool_handle_t
*zhp
, zpool_prop_t prop
, char *buf
, size_t len
,
210 zprop_source_t
*srctype
)
214 zprop_source_t src
= ZPROP_SRC_NONE
;
219 if (zpool_get_state(zhp
) == POOL_STATE_UNAVAIL
) {
221 case ZPOOL_PROP_NAME
:
222 (void) strlcpy(buf
, zpool_get_name(zhp
), len
);
225 case ZPOOL_PROP_HEALTH
:
226 (void) strlcpy(buf
, "FAULTED", len
);
229 case ZPOOL_PROP_GUID
:
230 intval
= zpool_get_prop_int(zhp
, prop
, &src
);
231 (void) snprintf(buf
, len
, "%llu", intval
);
234 case ZPOOL_PROP_ALTROOT
:
235 case ZPOOL_PROP_CACHEFILE
:
236 if (zhp
->zpool_props
!= NULL
||
237 zpool_get_all_props(zhp
) == 0) {
239 zpool_get_prop_string(zhp
, prop
, &src
),
247 (void) strlcpy(buf
, "-", len
);
256 if (zhp
->zpool_props
== NULL
&& zpool_get_all_props(zhp
) &&
257 prop
!= ZPOOL_PROP_NAME
)
260 switch (zpool_prop_get_type(prop
)) {
261 case PROP_TYPE_STRING
:
262 (void) strlcpy(buf
, zpool_get_prop_string(zhp
, prop
, &src
),
266 case PROP_TYPE_NUMBER
:
267 intval
= zpool_get_prop_int(zhp
, prop
, &src
);
270 case ZPOOL_PROP_SIZE
:
271 case ZPOOL_PROP_USED
:
272 case ZPOOL_PROP_AVAILABLE
:
273 (void) zfs_nicenum(intval
, buf
, len
);
276 case ZPOOL_PROP_CAPACITY
:
277 (void) snprintf(buf
, len
, "%llu%%",
278 (u_longlong_t
)intval
);
281 case ZPOOL_PROP_HEALTH
:
282 verify(nvlist_lookup_nvlist(zpool_get_config(zhp
, NULL
),
283 ZPOOL_CONFIG_VDEV_TREE
, &nvroot
) == 0);
284 verify(nvlist_lookup_uint64_array(nvroot
,
285 ZPOOL_CONFIG_STATS
, (uint64_t **)&vs
, &vsc
) == 0);
287 (void) strlcpy(buf
, zpool_state_to_name(intval
,
291 (void) snprintf(buf
, len
, "%llu", intval
);
295 case PROP_TYPE_INDEX
:
296 intval
= zpool_get_prop_int(zhp
, prop
, &src
);
297 if (zpool_prop_index_to_string(prop
, intval
, &strval
)
300 (void) strlcpy(buf
, strval
, len
);
314 * Check if the bootfs name has the same pool name as it is set to.
315 * Assuming bootfs is a valid dataset name.
318 bootfs_name_valid(const char *pool
, char *bootfs
)
320 int len
= strlen(pool
);
322 if (!zfs_name_valid(bootfs
, ZFS_TYPE_FILESYSTEM
|ZFS_TYPE_SNAPSHOT
))
325 if (strncmp(pool
, bootfs
, len
) == 0 &&
326 (bootfs
[len
] == '/' || bootfs
[len
] == '\0'))
333 * Inspect the configuration to determine if any of the devices contain
337 pool_uses_efi(nvlist_t
*config
)
342 if (nvlist_lookup_nvlist_array(config
, ZPOOL_CONFIG_CHILDREN
,
343 &child
, &children
) != 0)
344 return (read_efi_label(config
, NULL
) >= 0);
346 for (c
= 0; c
< children
; c
++) {
347 if (pool_uses_efi(child
[c
]))
354 pool_is_bootable(zpool_handle_t
*zhp
)
356 char bootfs
[ZPOOL_MAXNAMELEN
];
358 return (zpool_get_prop(zhp
, ZPOOL_PROP_BOOTFS
, bootfs
,
359 sizeof (bootfs
), NULL
) == 0 && strncmp(bootfs
, "-",
360 sizeof (bootfs
)) != 0);
365 * Given an nvlist of zpool properties to be set, validate that they are
366 * correct, and parse any numeric properties (index, boolean, etc) if they are
367 * specified as strings.
370 zpool_valid_proplist(libzfs_handle_t
*hdl
, const char *poolname
,
371 nvlist_t
*props
, uint64_t version
, boolean_t create_or_import
, char *errbuf
)
379 struct stat64 statbuf
;
383 if (nvlist_alloc(&retprops
, NV_UNIQUE_NAME
, 0) != 0) {
384 (void) no_memory(hdl
);
389 while ((elem
= nvlist_next_nvpair(props
, elem
)) != NULL
) {
390 const char *propname
= nvpair_name(elem
);
393 * Make sure this property is valid and applies to this type.
395 if ((prop
= zpool_name_to_prop(propname
)) == ZPROP_INVAL
) {
396 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
397 "invalid property '%s'"), propname
);
398 (void) zfs_error(hdl
, EZFS_BADPROP
, errbuf
);
402 if (zpool_prop_readonly(prop
)) {
403 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
, "'%s' "
404 "is readonly"), propname
);
405 (void) zfs_error(hdl
, EZFS_PROPREADONLY
, errbuf
);
409 if (zprop_parse_value(hdl
, elem
, prop
, ZFS_TYPE_POOL
, retprops
,
410 &strval
, &intval
, errbuf
) != 0)
414 * Perform additional checking for specific properties.
417 case ZPOOL_PROP_VERSION
:
418 if (intval
< version
|| intval
> SPA_VERSION
) {
419 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
420 "property '%s' number %d is invalid."),
422 (void) zfs_error(hdl
, EZFS_BADVERSION
, errbuf
);
427 case ZPOOL_PROP_BOOTFS
:
428 if (create_or_import
) {
429 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
430 "property '%s' cannot be set at creation "
431 "or import time"), propname
);
432 (void) zfs_error(hdl
, EZFS_BADPROP
, errbuf
);
436 if (version
< SPA_VERSION_BOOTFS
) {
437 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
438 "pool must be upgraded to support "
439 "'%s' property"), propname
);
440 (void) zfs_error(hdl
, EZFS_BADVERSION
, errbuf
);
445 * bootfs property value has to be a dataset name and
446 * the dataset has to be in the same pool as it sets to.
448 if (strval
[0] != '\0' && !bootfs_name_valid(poolname
,
450 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
, "'%s' "
451 "is an invalid name"), strval
);
452 (void) zfs_error(hdl
, EZFS_INVALIDNAME
, errbuf
);
456 if ((zhp
= zpool_open_canfail(hdl
, poolname
)) == NULL
) {
457 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
458 "could not open pool '%s'"), poolname
);
459 (void) zfs_error(hdl
, EZFS_OPENFAILED
, errbuf
);
462 verify(nvlist_lookup_nvlist(zpool_get_config(zhp
, NULL
),
463 ZPOOL_CONFIG_VDEV_TREE
, &nvroot
) == 0);
466 * bootfs property cannot be set on a disk which has
469 if (pool_uses_efi(nvroot
)) {
470 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
471 "property '%s' not supported on "
472 "EFI labeled devices"), propname
);
473 (void) zfs_error(hdl
, EZFS_POOL_NOTSUP
, errbuf
);
480 case ZPOOL_PROP_ALTROOT
:
481 if (!create_or_import
) {
482 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
483 "property '%s' can only be set during pool "
484 "creation or import"), propname
);
485 (void) zfs_error(hdl
, EZFS_BADPROP
, errbuf
);
489 if (strval
[0] != '/') {
490 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
491 "bad alternate root '%s'"), strval
);
492 (void) zfs_error(hdl
, EZFS_BADPATH
, errbuf
);
497 case ZPOOL_PROP_CACHEFILE
:
498 if (strval
[0] == '\0')
501 if (strcmp(strval
, "none") == 0)
504 if (strval
[0] != '/') {
505 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
506 "property '%s' must be empty, an "
507 "absolute path, or 'none'"), propname
);
508 (void) zfs_error(hdl
, EZFS_BADPATH
, errbuf
);
512 slash
= strrchr(strval
, '/');
514 if (slash
[1] == '\0' || strcmp(slash
, "/.") == 0 ||
515 strcmp(slash
, "/..") == 0) {
516 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
517 "'%s' is not a valid file"), strval
);
518 (void) zfs_error(hdl
, EZFS_BADPATH
, errbuf
);
524 if (strval
[0] != '\0' &&
525 (stat64(strval
, &statbuf
) != 0 ||
526 !S_ISDIR(statbuf
.st_mode
))) {
527 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
528 "'%s' is not a valid directory"),
530 (void) zfs_error(hdl
, EZFS_BADPATH
, errbuf
);
541 nvlist_free(retprops
);
546 * Set zpool property : propname=propval.
549 zpool_set_prop(zpool_handle_t
*zhp
, const char *propname
, const char *propval
)
551 zfs_cmd_t zc
= { 0 };
554 nvlist_t
*nvl
= NULL
;
558 (void) snprintf(errbuf
, sizeof (errbuf
),
559 dgettext(TEXT_DOMAIN
, "cannot set property for '%s'"),
562 if (nvlist_alloc(&nvl
, NV_UNIQUE_NAME
, 0) != 0)
563 return (no_memory(zhp
->zpool_hdl
));
565 if (nvlist_add_string(nvl
, propname
, propval
) != 0) {
567 return (no_memory(zhp
->zpool_hdl
));
570 version
= zpool_get_prop_int(zhp
, ZPOOL_PROP_VERSION
, NULL
);
571 if ((realprops
= zpool_valid_proplist(zhp
->zpool_hdl
,
572 zhp
->zpool_name
, nvl
, version
, B_FALSE
, errbuf
)) == NULL
) {
581 * Execute the corresponding ioctl() to set this property.
583 (void) strlcpy(zc
.zc_name
, zhp
->zpool_name
, sizeof (zc
.zc_name
));
585 if (zcmd_write_src_nvlist(zhp
->zpool_hdl
, &zc
, nvl
) != 0) {
590 ret
= zfs_ioctl(zhp
->zpool_hdl
, ZFS_IOC_POOL_SET_PROPS
, &zc
);
592 zcmd_free_nvlists(&zc
);
596 (void) zpool_standard_error(zhp
->zpool_hdl
, errno
, errbuf
);
598 (void) zpool_props_refresh(zhp
);
604 zpool_expand_proplist(zpool_handle_t
*zhp
, zprop_list_t
**plp
)
606 libzfs_handle_t
*hdl
= zhp
->zpool_hdl
;
608 char buf
[ZFS_MAXPROPLEN
];
610 if (zprop_expand_list(hdl
, plp
, ZFS_TYPE_POOL
) != 0)
613 for (entry
= *plp
; entry
!= NULL
; entry
= entry
->pl_next
) {
618 if (entry
->pl_prop
!= ZPROP_INVAL
&&
619 zpool_get_prop(zhp
, entry
->pl_prop
, buf
, sizeof (buf
),
621 if (strlen(buf
) > entry
->pl_width
)
622 entry
->pl_width
= strlen(buf
);
631 * Validate the given pool name, optionally putting an extended error message in
635 zpool_name_valid(libzfs_handle_t
*hdl
, boolean_t isopen
, const char *pool
)
641 ret
= pool_namecheck(pool
, &why
, &what
);
644 * The rules for reserved pool names were extended at a later point.
645 * But we need to support users with existing pools that may now be
646 * invalid. So we only check for this expanded set of names during a
647 * create (or import), and only in userland.
649 if (ret
== 0 && !isopen
&&
650 (strncmp(pool
, "mirror", 6) == 0 ||
651 strncmp(pool
, "raidz", 5) == 0 ||
652 strncmp(pool
, "spare", 5) == 0 ||
653 strcmp(pool
, "log") == 0)) {
656 dgettext(TEXT_DOMAIN
, "name is reserved"));
664 case NAME_ERR_TOOLONG
:
666 dgettext(TEXT_DOMAIN
, "name is too long"));
669 case NAME_ERR_INVALCHAR
:
671 dgettext(TEXT_DOMAIN
, "invalid character "
672 "'%c' in pool name"), what
);
675 case NAME_ERR_NOLETTER
:
676 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
677 "name must begin with a letter"));
680 case NAME_ERR_RESERVED
:
681 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
682 "name is reserved"));
685 case NAME_ERR_DISKLIKE
:
686 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
687 "pool name is reserved"));
690 case NAME_ERR_LEADING_SLASH
:
691 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
692 "leading slash in name"));
695 case NAME_ERR_EMPTY_COMPONENT
:
696 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
697 "empty component in name"));
700 case NAME_ERR_TRAILING_SLASH
:
701 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
702 "trailing slash in name"));
705 case NAME_ERR_MULTIPLE_AT
:
706 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
707 "multiple '@' delimiters in name"));
719 * Open a handle to the given pool, even if the pool is currently in the FAULTED
723 zpool_open_canfail(libzfs_handle_t
*hdl
, const char *pool
)
729 * Make sure the pool name is valid.
731 if (!zpool_name_valid(hdl
, B_TRUE
, pool
)) {
732 (void) zfs_error_fmt(hdl
, EZFS_INVALIDNAME
,
733 dgettext(TEXT_DOMAIN
, "cannot open '%s'"),
738 if ((zhp
= zfs_alloc(hdl
, sizeof (zpool_handle_t
))) == NULL
)
741 zhp
->zpool_hdl
= hdl
;
742 (void) strlcpy(zhp
->zpool_name
, pool
, sizeof (zhp
->zpool_name
));
744 if (zpool_refresh_stats(zhp
, &missing
) != 0) {
750 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
, "no such pool"));
751 (void) zfs_error_fmt(hdl
, EZFS_NOENT
,
752 dgettext(TEXT_DOMAIN
, "cannot open '%s'"), pool
);
761 * Like the above, but silent on error. Used when iterating over pools (because
762 * the configuration cache may be out of date).
765 zpool_open_silent(libzfs_handle_t
*hdl
, const char *pool
, zpool_handle_t
**ret
)
770 if ((zhp
= zfs_alloc(hdl
, sizeof (zpool_handle_t
))) == NULL
)
773 zhp
->zpool_hdl
= hdl
;
774 (void) strlcpy(zhp
->zpool_name
, pool
, sizeof (zhp
->zpool_name
));
776 if (zpool_refresh_stats(zhp
, &missing
) != 0) {
792 * Similar to zpool_open_canfail(), but refuses to open pools in the faulted
796 zpool_open(libzfs_handle_t
*hdl
, const char *pool
)
800 if ((zhp
= zpool_open_canfail(hdl
, pool
)) == NULL
)
803 if (zhp
->zpool_state
== POOL_STATE_UNAVAIL
) {
804 (void) zfs_error_fmt(hdl
, EZFS_POOLUNAVAIL
,
805 dgettext(TEXT_DOMAIN
, "cannot open '%s'"), zhp
->zpool_name
);
814 * Close the handle. Simply frees the memory associated with the handle.
817 zpool_close(zpool_handle_t
*zhp
)
819 if (zhp
->zpool_config
)
820 nvlist_free(zhp
->zpool_config
);
821 if (zhp
->zpool_old_config
)
822 nvlist_free(zhp
->zpool_old_config
);
823 if (zhp
->zpool_props
)
824 nvlist_free(zhp
->zpool_props
);
829 * Return the name of the pool.
832 zpool_get_name(zpool_handle_t
*zhp
)
834 return (zhp
->zpool_name
);
839 * Return the state of the pool (ACTIVE or UNAVAILABLE)
842 zpool_get_state(zpool_handle_t
*zhp
)
844 return (zhp
->zpool_state
);
848 * Create the named pool, using the provided vdev list. It is assumed
849 * that the consumer has already validated the contents of the nvlist, so we
850 * don't have to worry about error semantics.
853 zpool_create(libzfs_handle_t
*hdl
, const char *pool
, nvlist_t
*nvroot
,
854 nvlist_t
*props
, nvlist_t
*fsprops
)
856 zfs_cmd_t zc
= { 0 };
857 nvlist_t
*zc_fsprops
= NULL
;
858 nvlist_t
*zc_props
= NULL
;
863 (void) snprintf(msg
, sizeof (msg
), dgettext(TEXT_DOMAIN
,
864 "cannot create '%s'"), pool
);
866 if (!zpool_name_valid(hdl
, B_FALSE
, pool
))
867 return (zfs_error(hdl
, EZFS_INVALIDNAME
, msg
));
869 if (zcmd_write_conf_nvlist(hdl
, &zc
, nvroot
) != 0)
873 if ((zc_props
= zpool_valid_proplist(hdl
, pool
, props
,
874 SPA_VERSION_1
, B_TRUE
, msg
)) == NULL
) {
883 zoned
= ((nvlist_lookup_string(fsprops
,
884 zfs_prop_to_name(ZFS_PROP_ZONED
), &zonestr
) == 0) &&
885 strcmp(zonestr
, "on") == 0);
887 if ((zc_fsprops
= zfs_valid_proplist(hdl
,
888 ZFS_TYPE_FILESYSTEM
, fsprops
, zoned
, NULL
, msg
)) == NULL
) {
892 (nvlist_alloc(&zc_props
, NV_UNIQUE_NAME
, 0) != 0)) {
895 if (nvlist_add_nvlist(zc_props
,
896 ZPOOL_ROOTFS_PROPS
, zc_fsprops
) != 0) {
901 if (zc_props
&& zcmd_write_src_nvlist(hdl
, &zc
, zc_props
) != 0)
904 (void) strlcpy(zc
.zc_name
, pool
, sizeof (zc
.zc_name
));
906 if ((ret
= zfs_ioctl(hdl
, ZFS_IOC_POOL_CREATE
, &zc
)) != 0) {
908 zcmd_free_nvlists(&zc
);
909 nvlist_free(zc_props
);
910 nvlist_free(zc_fsprops
);
915 * This can happen if the user has specified the same
916 * device multiple times. We can't reliably detect this
917 * until we try to add it and see we already have a
920 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
921 "one or more vdevs refer to the same device"));
922 return (zfs_error(hdl
, EZFS_BADDEV
, msg
));
926 * This occurs when one of the devices is below
927 * SPA_MINDEVSIZE. Unfortunately, we can't detect which
928 * device was the problem device since there's no
929 * reliable way to determine device size from userland.
934 zfs_nicenum(SPA_MINDEVSIZE
, buf
, sizeof (buf
));
936 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
937 "one or more devices is less than the "
938 "minimum size (%s)"), buf
);
940 return (zfs_error(hdl
, EZFS_BADDEV
, msg
));
943 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
944 "one or more devices is out of space"));
945 return (zfs_error(hdl
, EZFS_BADDEV
, msg
));
948 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
949 "cache device must be a disk or disk slice"));
950 return (zfs_error(hdl
, EZFS_BADDEV
, msg
));
953 return (zpool_standard_error(hdl
, errno
, msg
));
958 * If this is an alternate root pool, then we automatically set the
959 * mountpoint of the root dataset to be '/'.
961 if (nvlist_lookup_string(props
, zpool_prop_to_name(ZPOOL_PROP_ALTROOT
),
965 verify((zhp
= zfs_open(hdl
, pool
, ZFS_TYPE_DATASET
)) != NULL
);
966 verify(zfs_prop_set(zhp
, zfs_prop_to_name(ZFS_PROP_MOUNTPOINT
),
973 zcmd_free_nvlists(&zc
);
974 nvlist_free(zc_props
);
975 nvlist_free(zc_fsprops
);
980 * Destroy the given pool. It is up to the caller to ensure that there are no
981 * datasets left in the pool.
984 zpool_destroy(zpool_handle_t
*zhp
)
986 zfs_cmd_t zc
= { 0 };
987 zfs_handle_t
*zfp
= NULL
;
988 libzfs_handle_t
*hdl
= zhp
->zpool_hdl
;
991 if (zhp
->zpool_state
== POOL_STATE_ACTIVE
&&
992 (zfp
= zfs_open(zhp
->zpool_hdl
, zhp
->zpool_name
,
993 ZFS_TYPE_FILESYSTEM
)) == NULL
)
996 if (zpool_remove_zvol_links(zhp
) != 0)
999 (void) strlcpy(zc
.zc_name
, zhp
->zpool_name
, sizeof (zc
.zc_name
));
1001 if (zfs_ioctl(zhp
->zpool_hdl
, ZFS_IOC_POOL_DESTROY
, &zc
) != 0) {
1002 (void) snprintf(msg
, sizeof (msg
), dgettext(TEXT_DOMAIN
,
1003 "cannot destroy '%s'"), zhp
->zpool_name
);
1005 if (errno
== EROFS
) {
1006 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
1007 "one or more devices is read only"));
1008 (void) zfs_error(hdl
, EZFS_BADDEV
, msg
);
1010 (void) zpool_standard_error(hdl
, errno
, msg
);
1019 remove_mountpoint(zfp
);
1027 * Add the given vdevs to the pool. The caller must have already performed the
1028 * necessary verification to ensure that the vdev specification is well-formed.
1031 zpool_add(zpool_handle_t
*zhp
, nvlist_t
*nvroot
)
1033 zfs_cmd_t zc
= { 0 };
1035 libzfs_handle_t
*hdl
= zhp
->zpool_hdl
;
1037 nvlist_t
**spares
, **l2cache
;
1038 uint_t nspares
, nl2cache
;
1040 (void) snprintf(msg
, sizeof (msg
), dgettext(TEXT_DOMAIN
,
1041 "cannot add to '%s'"), zhp
->zpool_name
);
1043 if (zpool_get_prop_int(zhp
, ZPOOL_PROP_VERSION
, NULL
) <
1044 SPA_VERSION_SPARES
&&
1045 nvlist_lookup_nvlist_array(nvroot
, ZPOOL_CONFIG_SPARES
,
1046 &spares
, &nspares
) == 0) {
1047 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
, "pool must be "
1048 "upgraded to add hot spares"));
1049 return (zfs_error(hdl
, EZFS_BADVERSION
, msg
));
1052 if (pool_is_bootable(zhp
) && nvlist_lookup_nvlist_array(nvroot
,
1053 ZPOOL_CONFIG_SPARES
, &spares
, &nspares
) == 0) {
1056 for (s
= 0; s
< nspares
; s
++) {
1059 if (nvlist_lookup_string(spares
[s
], ZPOOL_CONFIG_PATH
,
1060 &path
) == 0 && pool_uses_efi(spares
[s
])) {
1061 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
1062 "device '%s' contains an EFI label and "
1063 "cannot be used on root pools."),
1064 zpool_vdev_name(hdl
, NULL
, spares
[s
]));
1065 return (zfs_error(hdl
, EZFS_POOL_NOTSUP
, msg
));
1070 if (zpool_get_prop_int(zhp
, ZPOOL_PROP_VERSION
, NULL
) <
1071 SPA_VERSION_L2CACHE
&&
1072 nvlist_lookup_nvlist_array(nvroot
, ZPOOL_CONFIG_L2CACHE
,
1073 &l2cache
, &nl2cache
) == 0) {
1074 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
, "pool must be "
1075 "upgraded to add cache devices"));
1076 return (zfs_error(hdl
, EZFS_BADVERSION
, msg
));
1079 if (zcmd_write_conf_nvlist(hdl
, &zc
, nvroot
) != 0)
1081 (void) strlcpy(zc
.zc_name
, zhp
->zpool_name
, sizeof (zc
.zc_name
));
1083 if (zfs_ioctl(zhp
->zpool_hdl
, ZFS_IOC_VDEV_ADD
, &zc
) != 0) {
1087 * This can happen if the user has specified the same
1088 * device multiple times. We can't reliably detect this
1089 * until we try to add it and see we already have a
1092 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
1093 "one or more vdevs refer to the same device"));
1094 (void) zfs_error(hdl
, EZFS_BADDEV
, msg
);
1099 * This occurrs when one of the devices is below
1100 * SPA_MINDEVSIZE. Unfortunately, we can't detect which
1101 * device was the problem device since there's no
1102 * reliable way to determine device size from userland.
1107 zfs_nicenum(SPA_MINDEVSIZE
, buf
, sizeof (buf
));
1109 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
1110 "device is less than the minimum "
1113 (void) zfs_error(hdl
, EZFS_BADDEV
, msg
);
1117 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
1118 "pool must be upgraded to add these vdevs"));
1119 (void) zfs_error(hdl
, EZFS_BADVERSION
, msg
);
1123 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
1124 "root pool can not have multiple vdevs"
1125 " or separate logs"));
1126 (void) zfs_error(hdl
, EZFS_POOL_NOTSUP
, msg
);
1130 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
1131 "cache device must be a disk or disk slice"));
1132 (void) zfs_error(hdl
, EZFS_BADDEV
, msg
);
1136 (void) zpool_standard_error(hdl
, errno
, msg
);
1144 zcmd_free_nvlists(&zc
);
1150 * Exports the pool from the system. The caller must ensure that there are no
1151 * mounted datasets in the pool.
1154 zpool_export_common(zpool_handle_t
*zhp
, boolean_t force
, boolean_t hardforce
)
1156 zfs_cmd_t zc
= { 0 };
1159 if (zpool_remove_zvol_links(zhp
) != 0)
1162 (void) snprintf(msg
, sizeof (msg
), dgettext(TEXT_DOMAIN
,
1163 "cannot export '%s'"), zhp
->zpool_name
);
1165 (void) strlcpy(zc
.zc_name
, zhp
->zpool_name
, sizeof (zc
.zc_name
));
1166 zc
.zc_cookie
= force
;
1167 zc
.zc_guid
= hardforce
;
1169 if (zfs_ioctl(zhp
->zpool_hdl
, ZFS_IOC_POOL_EXPORT
, &zc
) != 0) {
1172 zfs_error_aux(zhp
->zpool_hdl
, dgettext(TEXT_DOMAIN
,
1173 "use '-f' to override the following errors:\n"
1174 "'%s' has an active shared spare which could be"
1175 " used by other pools once '%s' is exported."),
1176 zhp
->zpool_name
, zhp
->zpool_name
);
1177 return (zfs_error(zhp
->zpool_hdl
, EZFS_ACTIVE_SPARE
,
1180 return (zpool_standard_error_fmt(zhp
->zpool_hdl
, errno
,
1189 zpool_export(zpool_handle_t
*zhp
, boolean_t force
)
1191 return (zpool_export_common(zhp
, force
, B_FALSE
));
1195 zpool_export_force(zpool_handle_t
*zhp
)
1197 return (zpool_export_common(zhp
, B_TRUE
, B_TRUE
));
1201 * zpool_import() is a contracted interface. Should be kept the same
1204 * Applications should use zpool_import_props() to import a pool with
1205 * new properties value to be set.
1208 zpool_import(libzfs_handle_t
*hdl
, nvlist_t
*config
, const char *newname
,
1211 nvlist_t
*props
= NULL
;
1214 if (altroot
!= NULL
) {
1215 if (nvlist_alloc(&props
, NV_UNIQUE_NAME
, 0) != 0) {
1216 return (zfs_error_fmt(hdl
, EZFS_NOMEM
,
1217 dgettext(TEXT_DOMAIN
, "cannot import '%s'"),
1221 if (nvlist_add_string(props
,
1222 zpool_prop_to_name(ZPOOL_PROP_ALTROOT
), altroot
) != 0 ||
1223 nvlist_add_string(props
,
1224 zpool_prop_to_name(ZPOOL_PROP_CACHEFILE
), "none") != 0) {
1226 return (zfs_error_fmt(hdl
, EZFS_NOMEM
,
1227 dgettext(TEXT_DOMAIN
, "cannot import '%s'"),
1232 ret
= zpool_import_props(hdl
, config
, newname
, props
, B_FALSE
);
1239 * Import the given pool using the known configuration and a list of
1240 * properties to be set. The configuration should have come from
1241 * zpool_find_import(). The 'newname' parameters control whether the pool
1242 * is imported with a different name.
1245 zpool_import_props(libzfs_handle_t
*hdl
, nvlist_t
*config
, const char *newname
,
1246 nvlist_t
*props
, boolean_t importfaulted
)
1248 zfs_cmd_t zc
= { 0 };
1254 verify(nvlist_lookup_string(config
, ZPOOL_CONFIG_POOL_NAME
,
1257 (void) snprintf(errbuf
, sizeof (errbuf
), dgettext(TEXT_DOMAIN
,
1258 "cannot import pool '%s'"), origname
);
1260 if (newname
!= NULL
) {
1261 if (!zpool_name_valid(hdl
, B_FALSE
, newname
))
1262 return (zfs_error_fmt(hdl
, EZFS_INVALIDNAME
,
1263 dgettext(TEXT_DOMAIN
, "cannot import '%s'"),
1265 thename
= (char *)newname
;
1273 verify(nvlist_lookup_uint64(config
, ZPOOL_CONFIG_VERSION
,
1276 if ((props
= zpool_valid_proplist(hdl
, origname
,
1277 props
, version
, B_TRUE
, errbuf
)) == NULL
) {
1279 } else if (zcmd_write_src_nvlist(hdl
, &zc
, props
) != 0) {
1285 (void) strlcpy(zc
.zc_name
, thename
, sizeof (zc
.zc_name
));
1287 verify(nvlist_lookup_uint64(config
, ZPOOL_CONFIG_POOL_GUID
,
1290 if (zcmd_write_conf_nvlist(hdl
, &zc
, config
) != 0) {
1295 zc
.zc_cookie
= (uint64_t)importfaulted
;
1297 if (zfs_ioctl(hdl
, ZFS_IOC_POOL_IMPORT
, &zc
) != 0) {
1299 if (newname
== NULL
)
1300 (void) snprintf(desc
, sizeof (desc
),
1301 dgettext(TEXT_DOMAIN
, "cannot import '%s'"),
1304 (void) snprintf(desc
, sizeof (desc
),
1305 dgettext(TEXT_DOMAIN
, "cannot import '%s' as '%s'"),
1311 * Unsupported version.
1313 (void) zfs_error(hdl
, EZFS_BADVERSION
, desc
);
1317 (void) zfs_error(hdl
, EZFS_INVALCONFIG
, desc
);
1321 (void) zpool_standard_error(hdl
, errno
, desc
);
1326 zpool_handle_t
*zhp
;
1329 * This should never fail, but play it safe anyway.
1331 if (zpool_open_silent(hdl
, thename
, &zhp
) != 0) {
1333 } else if (zhp
!= NULL
) {
1334 ret
= zpool_create_zvol_links(zhp
);
1340 zcmd_free_nvlists(&zc
);
1350 zpool_scrub(zpool_handle_t
*zhp
, pool_scrub_type_t type
)
1352 zfs_cmd_t zc
= { 0 };
1354 libzfs_handle_t
*hdl
= zhp
->zpool_hdl
;
1356 (void) strlcpy(zc
.zc_name
, zhp
->zpool_name
, sizeof (zc
.zc_name
));
1357 zc
.zc_cookie
= type
;
1359 if (zfs_ioctl(zhp
->zpool_hdl
, ZFS_IOC_POOL_SCRUB
, &zc
) == 0)
1362 (void) snprintf(msg
, sizeof (msg
),
1363 dgettext(TEXT_DOMAIN
, "cannot scrub %s"), zc
.zc_name
);
1366 return (zfs_error(hdl
, EZFS_RESILVERING
, msg
));
1368 return (zpool_standard_error(hdl
, errno
, msg
));
1372 * 'avail_spare' is set to TRUE if the provided guid refers to an AVAIL
1373 * spare; but FALSE if its an INUSE spare.
1376 vdev_to_nvlist_iter(nvlist_t
*nv
, const char *search
, uint64_t guid
,
1377 boolean_t
*avail_spare
, boolean_t
*l2cache
, boolean_t
*log
)
1381 uint64_t theguid
, present
;
1383 uint64_t wholedisk
= 0;
1387 verify(nvlist_lookup_uint64(nv
, ZPOOL_CONFIG_GUID
, &theguid
) == 0);
1389 if (search
== NULL
&&
1390 nvlist_lookup_uint64(nv
, ZPOOL_CONFIG_NOT_PRESENT
, &present
) == 0) {
1392 * If the device has never been present since import, the only
1393 * reliable way to match the vdev is by GUID.
1395 if (theguid
== guid
)
1397 } else if (search
!= NULL
&&
1398 nvlist_lookup_string(nv
, ZPOOL_CONFIG_PATH
, &path
) == 0) {
1399 (void) nvlist_lookup_uint64(nv
, ZPOOL_CONFIG_WHOLE_DISK
,
1403 * For whole disks, the internal path has 's0', but the
1404 * path passed in by the user doesn't.
1406 if (strlen(search
) == strlen(path
) - 2 &&
1407 strncmp(search
, path
, strlen(search
)) == 0)
1409 } else if (strcmp(search
, path
) == 0) {
1414 if (nvlist_lookup_nvlist_array(nv
, ZPOOL_CONFIG_CHILDREN
,
1415 &child
, &children
) != 0)
1418 for (c
= 0; c
< children
; c
++) {
1419 if ((ret
= vdev_to_nvlist_iter(child
[c
], search
, guid
,
1420 avail_spare
, l2cache
, NULL
)) != NULL
) {
1422 * The 'is_log' value is only set for the toplevel
1423 * vdev, not the leaf vdevs. So we always lookup the
1424 * log device from the root of the vdev tree (where
1425 * 'log' is non-NULL).
1428 nvlist_lookup_uint64(child
[c
],
1429 ZPOOL_CONFIG_IS_LOG
, &is_log
) == 0 &&
1437 if (nvlist_lookup_nvlist_array(nv
, ZPOOL_CONFIG_SPARES
,
1438 &child
, &children
) == 0) {
1439 for (c
= 0; c
< children
; c
++) {
1440 if ((ret
= vdev_to_nvlist_iter(child
[c
], search
, guid
,
1441 avail_spare
, l2cache
, NULL
)) != NULL
) {
1442 *avail_spare
= B_TRUE
;
1448 if (nvlist_lookup_nvlist_array(nv
, ZPOOL_CONFIG_L2CACHE
,
1449 &child
, &children
) == 0) {
1450 for (c
= 0; c
< children
; c
++) {
1451 if ((ret
= vdev_to_nvlist_iter(child
[c
], search
, guid
,
1452 avail_spare
, l2cache
, NULL
)) != NULL
) {
1463 zpool_find_vdev(zpool_handle_t
*zhp
, const char *path
, boolean_t
*avail_spare
,
1464 boolean_t
*l2cache
, boolean_t
*log
)
1466 char buf
[MAXPATHLEN
];
1472 guid
= strtoull(path
, &end
, 10);
1473 if (guid
!= 0 && *end
== '\0') {
1475 } else if (path
[0] != '/') {
1476 (void) snprintf(buf
, sizeof (buf
), "%s%s", "/dev/dsk/", path
);
1482 verify(nvlist_lookup_nvlist(zhp
->zpool_config
, ZPOOL_CONFIG_VDEV_TREE
,
1485 *avail_spare
= B_FALSE
;
1489 return (vdev_to_nvlist_iter(nvroot
, search
, guid
, avail_spare
,
1494 vdev_online(nvlist_t
*nv
)
1498 if (nvlist_lookup_uint64(nv
, ZPOOL_CONFIG_OFFLINE
, &ival
) == 0 ||
1499 nvlist_lookup_uint64(nv
, ZPOOL_CONFIG_FAULTED
, &ival
) == 0 ||
1500 nvlist_lookup_uint64(nv
, ZPOOL_CONFIG_REMOVED
, &ival
) == 0)
1507 * Get phys_path for a root pool
1508 * Return 0 on success; non-zeron on failure.
1511 zpool_get_physpath(zpool_handle_t
*zhp
, char *physpath
)
1513 nvlist_t
*vdev_root
;
1519 * Make sure this is a root pool, as phys_path doesn't mean
1520 * anything to a non-root pool.
1522 if (!pool_is_bootable(zhp
))
1525 verify(nvlist_lookup_nvlist(zhp
->zpool_config
,
1526 ZPOOL_CONFIG_VDEV_TREE
, &vdev_root
) == 0);
1528 if (nvlist_lookup_nvlist_array(vdev_root
, ZPOOL_CONFIG_CHILDREN
,
1529 &child
, &count
) != 0)
1532 for (i
= 0; i
< count
; i
++) {
1539 if (nvlist_lookup_string(child
[i
], ZPOOL_CONFIG_TYPE
, &type
)
1543 if (strcmp(type
, VDEV_TYPE_DISK
) == 0) {
1544 if (!vdev_online(child
[i
]))
1546 verify(nvlist_lookup_string(child
[i
],
1547 ZPOOL_CONFIG_PHYS_PATH
, &tmppath
) == 0);
1548 (void) strncpy(physpath
, tmppath
, strlen(tmppath
));
1549 } else if (strcmp(type
, VDEV_TYPE_MIRROR
) == 0) {
1550 if (nvlist_lookup_nvlist_array(child
[i
],
1551 ZPOOL_CONFIG_CHILDREN
, &child2
, &count2
) != 0)
1554 for (j
= 0; j
< count2
; j
++) {
1555 if (!vdev_online(child2
[j
]))
1557 if (nvlist_lookup_string(child2
[j
],
1558 ZPOOL_CONFIG_PHYS_PATH
, &tmppath
) != 0)
1561 if ((strlen(physpath
) + strlen(tmppath
)) >
1565 if (strlen(physpath
) == 0) {
1566 (void) strncpy(physpath
, tmppath
,
1569 (void) strcat(physpath
, " ");
1570 (void) strcat(physpath
, tmppath
);
1582 * Returns TRUE if the given guid corresponds to the given type.
1583 * This is used to check for hot spares (INUSE or not), and level 2 cache
1587 is_guid_type(zpool_handle_t
*zhp
, uint64_t guid
, const char *type
)
1589 uint64_t target_guid
;
1595 verify(nvlist_lookup_nvlist(zhp
->zpool_config
, ZPOOL_CONFIG_VDEV_TREE
,
1597 if (nvlist_lookup_nvlist_array(nvroot
, type
, &list
, &count
) == 0) {
1598 for (i
= 0; i
< count
; i
++) {
1599 verify(nvlist_lookup_uint64(list
[i
], ZPOOL_CONFIG_GUID
,
1600 &target_guid
) == 0);
1601 if (guid
== target_guid
)
1610 * Bring the specified vdev online. The 'flags' parameter is a set of the
1611 * ZFS_ONLINE_* flags.
1614 zpool_vdev_online(zpool_handle_t
*zhp
, const char *path
, int flags
,
1615 vdev_state_t
*newstate
)
1617 zfs_cmd_t zc
= { 0 };
1620 boolean_t avail_spare
, l2cache
;
1621 libzfs_handle_t
*hdl
= zhp
->zpool_hdl
;
1623 (void) snprintf(msg
, sizeof (msg
),
1624 dgettext(TEXT_DOMAIN
, "cannot online %s"), path
);
1626 (void) strlcpy(zc
.zc_name
, zhp
->zpool_name
, sizeof (zc
.zc_name
));
1627 if ((tgt
= zpool_find_vdev(zhp
, path
, &avail_spare
, &l2cache
,
1629 return (zfs_error(hdl
, EZFS_NODEVICE
, msg
));
1631 verify(nvlist_lookup_uint64(tgt
, ZPOOL_CONFIG_GUID
, &zc
.zc_guid
) == 0);
1634 is_guid_type(zhp
, zc
.zc_guid
, ZPOOL_CONFIG_SPARES
) == B_TRUE
)
1635 return (zfs_error(hdl
, EZFS_ISSPARE
, msg
));
1637 zc
.zc_cookie
= VDEV_STATE_ONLINE
;
1640 if (zfs_ioctl(zhp
->zpool_hdl
, ZFS_IOC_VDEV_SET_STATE
, &zc
) != 0)
1641 return (zpool_standard_error(hdl
, errno
, msg
));
1643 *newstate
= zc
.zc_cookie
;
1648 * Take the specified vdev offline
1651 zpool_vdev_offline(zpool_handle_t
*zhp
, const char *path
, boolean_t istmp
)
1653 zfs_cmd_t zc
= { 0 };
1656 boolean_t avail_spare
, l2cache
;
1657 libzfs_handle_t
*hdl
= zhp
->zpool_hdl
;
1659 (void) snprintf(msg
, sizeof (msg
),
1660 dgettext(TEXT_DOMAIN
, "cannot offline %s"), path
);
1662 (void) strlcpy(zc
.zc_name
, zhp
->zpool_name
, sizeof (zc
.zc_name
));
1663 if ((tgt
= zpool_find_vdev(zhp
, path
, &avail_spare
, &l2cache
,
1665 return (zfs_error(hdl
, EZFS_NODEVICE
, msg
));
1667 verify(nvlist_lookup_uint64(tgt
, ZPOOL_CONFIG_GUID
, &zc
.zc_guid
) == 0);
1670 is_guid_type(zhp
, zc
.zc_guid
, ZPOOL_CONFIG_SPARES
) == B_TRUE
)
1671 return (zfs_error(hdl
, EZFS_ISSPARE
, msg
));
1673 zc
.zc_cookie
= VDEV_STATE_OFFLINE
;
1674 zc
.zc_obj
= istmp
? ZFS_OFFLINE_TEMPORARY
: 0;
1676 if (zfs_ioctl(zhp
->zpool_hdl
, ZFS_IOC_VDEV_SET_STATE
, &zc
) == 0)
1683 * There are no other replicas of this device.
1685 return (zfs_error(hdl
, EZFS_NOREPLICAS
, msg
));
1688 return (zpool_standard_error(hdl
, errno
, msg
));
1693 * Mark the given vdev faulted.
1696 zpool_vdev_fault(zpool_handle_t
*zhp
, uint64_t guid
)
1698 zfs_cmd_t zc
= { 0 };
1700 libzfs_handle_t
*hdl
= zhp
->zpool_hdl
;
1702 (void) snprintf(msg
, sizeof (msg
),
1703 dgettext(TEXT_DOMAIN
, "cannot fault %llu"), guid
);
1705 (void) strlcpy(zc
.zc_name
, zhp
->zpool_name
, sizeof (zc
.zc_name
));
1707 zc
.zc_cookie
= VDEV_STATE_FAULTED
;
1709 if (ioctl(zhp
->zpool_hdl
->libzfs_fd
, ZFS_IOC_VDEV_SET_STATE
, &zc
) == 0)
1716 * There are no other replicas of this device.
1718 return (zfs_error(hdl
, EZFS_NOREPLICAS
, msg
));
1721 return (zpool_standard_error(hdl
, errno
, msg
));
1727 * Mark the given vdev degraded.
1730 zpool_vdev_degrade(zpool_handle_t
*zhp
, uint64_t guid
)
1732 zfs_cmd_t zc
= { 0 };
1734 libzfs_handle_t
*hdl
= zhp
->zpool_hdl
;
1736 (void) snprintf(msg
, sizeof (msg
),
1737 dgettext(TEXT_DOMAIN
, "cannot degrade %llu"), guid
);
1739 (void) strlcpy(zc
.zc_name
, zhp
->zpool_name
, sizeof (zc
.zc_name
));
1741 zc
.zc_cookie
= VDEV_STATE_DEGRADED
;
1743 if (ioctl(zhp
->zpool_hdl
->libzfs_fd
, ZFS_IOC_VDEV_SET_STATE
, &zc
) == 0)
1746 return (zpool_standard_error(hdl
, errno
, msg
));
1750 * Returns TRUE if the given nvlist is a vdev that was originally swapped in as
1754 is_replacing_spare(nvlist_t
*search
, nvlist_t
*tgt
, int which
)
1760 if (nvlist_lookup_nvlist_array(search
, ZPOOL_CONFIG_CHILDREN
, &child
,
1762 verify(nvlist_lookup_string(search
, ZPOOL_CONFIG_TYPE
,
1765 if (strcmp(type
, VDEV_TYPE_SPARE
) == 0 &&
1766 children
== 2 && child
[which
] == tgt
)
1769 for (c
= 0; c
< children
; c
++)
1770 if (is_replacing_spare(child
[c
], tgt
, which
))
1778 * Attach new_disk (fully described by nvroot) to old_disk.
1779 * If 'replacing' is specified, the new disk will replace the old one.
1782 zpool_vdev_attach(zpool_handle_t
*zhp
,
1783 const char *old_disk
, const char *new_disk
, nvlist_t
*nvroot
, int replacing
)
1785 zfs_cmd_t zc
= { 0 };
1789 boolean_t avail_spare
, l2cache
, islog
;
1791 char *path
, *newname
;
1794 nvlist_t
*config_root
;
1795 libzfs_handle_t
*hdl
= zhp
->zpool_hdl
;
1796 boolean_t rootpool
= pool_is_bootable(zhp
);
1799 (void) snprintf(msg
, sizeof (msg
), dgettext(TEXT_DOMAIN
,
1800 "cannot replace %s with %s"), old_disk
, new_disk
);
1802 (void) snprintf(msg
, sizeof (msg
), dgettext(TEXT_DOMAIN
,
1803 "cannot attach %s to %s"), new_disk
, old_disk
);
1806 * If this is a root pool, make sure that we're not attaching an
1807 * EFI labeled device.
1809 if (rootpool
&& pool_uses_efi(nvroot
)) {
1810 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
1811 "EFI labeled devices are not supported on root pools."));
1812 return (zfs_error(hdl
, EZFS_POOL_NOTSUP
, msg
));
1815 (void) strlcpy(zc
.zc_name
, zhp
->zpool_name
, sizeof (zc
.zc_name
));
1816 if ((tgt
= zpool_find_vdev(zhp
, old_disk
, &avail_spare
, &l2cache
,
1818 return (zfs_error(hdl
, EZFS_NODEVICE
, msg
));
1821 return (zfs_error(hdl
, EZFS_ISSPARE
, msg
));
1824 return (zfs_error(hdl
, EZFS_ISL2CACHE
, msg
));
1826 verify(nvlist_lookup_uint64(tgt
, ZPOOL_CONFIG_GUID
, &zc
.zc_guid
) == 0);
1827 zc
.zc_cookie
= replacing
;
1829 if (nvlist_lookup_nvlist_array(nvroot
, ZPOOL_CONFIG_CHILDREN
,
1830 &child
, &children
) != 0 || children
!= 1) {
1831 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
1832 "new device must be a single disk"));
1833 return (zfs_error(hdl
, EZFS_INVALCONFIG
, msg
));
1836 verify(nvlist_lookup_nvlist(zpool_get_config(zhp
, NULL
),
1837 ZPOOL_CONFIG_VDEV_TREE
, &config_root
) == 0);
1839 if ((newname
= zpool_vdev_name(NULL
, NULL
, child
[0])) == NULL
)
1843 * If the target is a hot spare that has been swapped in, we can only
1844 * replace it with another hot spare.
1847 nvlist_lookup_uint64(tgt
, ZPOOL_CONFIG_IS_SPARE
, &val
) == 0 &&
1848 (zpool_find_vdev(zhp
, newname
, &avail_spare
, &l2cache
,
1849 NULL
) == NULL
|| !avail_spare
) &&
1850 is_replacing_spare(config_root
, tgt
, 1)) {
1851 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
1852 "can only be replaced by another hot spare"));
1854 return (zfs_error(hdl
, EZFS_BADTARGET
, msg
));
1858 * If we are attempting to replace a spare, it canot be applied to an
1859 * already spared device.
1862 nvlist_lookup_string(child
[0], ZPOOL_CONFIG_PATH
, &path
) == 0 &&
1863 zpool_find_vdev(zhp
, newname
, &avail_spare
,
1864 &l2cache
, NULL
) != NULL
&& avail_spare
&&
1865 is_replacing_spare(config_root
, tgt
, 0)) {
1866 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
1867 "device has already been replaced with a spare"));
1869 return (zfs_error(hdl
, EZFS_BADTARGET
, msg
));
1874 if (zcmd_write_conf_nvlist(hdl
, &zc
, nvroot
) != 0)
1877 ret
= zfs_ioctl(zhp
->zpool_hdl
, ZFS_IOC_VDEV_ATTACH
, &zc
);
1879 zcmd_free_nvlists(&zc
);
1884 * XXX - This should be removed once we can
1885 * automatically install the bootblocks on the
1886 * newly attached disk.
1888 (void) fprintf(stderr
, dgettext(TEXT_DOMAIN
, "Please "
1889 "be sure to invoke %s to make '%s' bootable.\n"),
1898 * Can't attach to or replace this type of vdev.
1902 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
1903 "cannot replace a log with a spare"));
1905 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
1906 "cannot replace a replacing device"));
1908 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
1909 "can only attach to mirrors and top-level "
1912 (void) zfs_error(hdl
, EZFS_BADTARGET
, msg
);
1917 * The new device must be a single disk.
1919 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
1920 "new device must be a single disk"));
1921 (void) zfs_error(hdl
, EZFS_INVALCONFIG
, msg
);
1925 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
, "%s is busy"),
1927 (void) zfs_error(hdl
, EZFS_BADDEV
, msg
);
1932 * The new device is too small.
1934 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
1935 "device is too small"));
1936 (void) zfs_error(hdl
, EZFS_BADDEV
, msg
);
1941 * The new device has a different alignment requirement.
1943 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
1944 "devices have different sector alignment"));
1945 (void) zfs_error(hdl
, EZFS_BADDEV
, msg
);
1950 * The resulting top-level vdev spec won't fit in the label.
1952 (void) zfs_error(hdl
, EZFS_DEVOVERFLOW
, msg
);
1956 (void) zpool_standard_error(hdl
, errno
, msg
);
1963 * Detach the specified device.
1966 zpool_vdev_detach(zpool_handle_t
*zhp
, const char *path
)
1968 zfs_cmd_t zc
= { 0 };
1971 boolean_t avail_spare
, l2cache
;
1972 libzfs_handle_t
*hdl
= zhp
->zpool_hdl
;
1974 (void) snprintf(msg
, sizeof (msg
),
1975 dgettext(TEXT_DOMAIN
, "cannot detach %s"), path
);
1977 (void) strlcpy(zc
.zc_name
, zhp
->zpool_name
, sizeof (zc
.zc_name
));
1978 if ((tgt
= zpool_find_vdev(zhp
, path
, &avail_spare
, &l2cache
,
1980 return (zfs_error(hdl
, EZFS_NODEVICE
, msg
));
1983 return (zfs_error(hdl
, EZFS_ISSPARE
, msg
));
1986 return (zfs_error(hdl
, EZFS_ISL2CACHE
, msg
));
1988 verify(nvlist_lookup_uint64(tgt
, ZPOOL_CONFIG_GUID
, &zc
.zc_guid
) == 0);
1990 if (zfs_ioctl(hdl
, ZFS_IOC_VDEV_DETACH
, &zc
) == 0)
1997 * Can't detach from this type of vdev.
1999 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
, "only "
2000 "applicable to mirror and replacing vdevs"));
2001 (void) zfs_error(zhp
->zpool_hdl
, EZFS_BADTARGET
, msg
);
2006 * There are no other replicas of this device.
2008 (void) zfs_error(hdl
, EZFS_NOREPLICAS
, msg
);
2012 (void) zpool_standard_error(hdl
, errno
, msg
);
2019 * Remove the given device. Currently, this is supported only for hot spares
2020 * and level 2 cache devices.
2023 zpool_vdev_remove(zpool_handle_t
*zhp
, const char *path
)
2025 zfs_cmd_t zc
= { 0 };
2028 boolean_t avail_spare
, l2cache
;
2029 libzfs_handle_t
*hdl
= zhp
->zpool_hdl
;
2031 (void) snprintf(msg
, sizeof (msg
),
2032 dgettext(TEXT_DOMAIN
, "cannot remove %s"), path
);
2034 (void) strlcpy(zc
.zc_name
, zhp
->zpool_name
, sizeof (zc
.zc_name
));
2035 if ((tgt
= zpool_find_vdev(zhp
, path
, &avail_spare
, &l2cache
,
2037 return (zfs_error(hdl
, EZFS_NODEVICE
, msg
));
2039 if (!avail_spare
&& !l2cache
) {
2040 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
2041 "only inactive hot spares or cache devices "
2043 return (zfs_error(hdl
, EZFS_NODEVICE
, msg
));
2046 verify(nvlist_lookup_uint64(tgt
, ZPOOL_CONFIG_GUID
, &zc
.zc_guid
) == 0);
2048 if (zfs_ioctl(hdl
, ZFS_IOC_VDEV_REMOVE
, &zc
) == 0)
2051 return (zpool_standard_error(hdl
, errno
, msg
));
2055 * Clear the errors for the pool, or the particular device if specified.
2058 zpool_clear(zpool_handle_t
*zhp
, const char *path
)
2060 zfs_cmd_t zc
= { 0 };
2063 boolean_t avail_spare
, l2cache
;
2064 libzfs_handle_t
*hdl
= zhp
->zpool_hdl
;
2067 (void) snprintf(msg
, sizeof (msg
),
2068 dgettext(TEXT_DOMAIN
, "cannot clear errors for %s"),
2071 (void) snprintf(msg
, sizeof (msg
),
2072 dgettext(TEXT_DOMAIN
, "cannot clear errors for %s"),
2075 (void) strlcpy(zc
.zc_name
, zhp
->zpool_name
, sizeof (zc
.zc_name
));
2077 if ((tgt
= zpool_find_vdev(zhp
, path
, &avail_spare
,
2078 &l2cache
, NULL
)) == 0)
2079 return (zfs_error(hdl
, EZFS_NODEVICE
, msg
));
2082 * Don't allow error clearing for hot spares. Do allow
2083 * error clearing for l2cache devices.
2086 return (zfs_error(hdl
, EZFS_ISSPARE
, msg
));
2088 verify(nvlist_lookup_uint64(tgt
, ZPOOL_CONFIG_GUID
,
2092 if (zfs_ioctl(hdl
, ZFS_IOC_CLEAR
, &zc
) == 0)
2095 return (zpool_standard_error(hdl
, errno
, msg
));
2099 * Similar to zpool_clear(), but takes a GUID (used by fmd).
2102 zpool_vdev_clear(zpool_handle_t
*zhp
, uint64_t guid
)
2104 zfs_cmd_t zc
= { 0 };
2106 libzfs_handle_t
*hdl
= zhp
->zpool_hdl
;
2108 (void) snprintf(msg
, sizeof (msg
),
2109 dgettext(TEXT_DOMAIN
, "cannot clear errors for %llx"),
2112 (void) strlcpy(zc
.zc_name
, zhp
->zpool_name
, sizeof (zc
.zc_name
));
2115 if (ioctl(hdl
->libzfs_fd
, ZFS_IOC_CLEAR
, &zc
) == 0)
2118 return (zpool_standard_error(hdl
, errno
, msg
));
2122 * Iterate over all zvols in a given pool by walking the /dev/zvol/dsk/<pool>
2126 zpool_iter_zvol(zpool_handle_t
*zhp
, int (*cb
)(const char *, void *),
2129 libzfs_handle_t
*hdl
= zhp
->zpool_hdl
;
2130 char (*paths
)[MAXPATHLEN
];
2132 int curr
, fd
, base
, ret
= 0;
2137 if ((base
= open("/dev/zvol/dsk", O_RDONLY
)) < 0)
2138 return (errno
== ENOENT
? 0 : -1);
2140 if (fstatat(base
, zhp
->zpool_name
, &st
, 0) != 0) {
2143 return (err
== ENOENT
? 0 : -1);
2147 * Oddly this wasn't a directory -- ignore that failure since we
2148 * know there are no links lower in the (non-existant) hierarchy.
2150 if (!S_ISDIR(st
.st_mode
)) {
2155 if ((paths
= zfs_alloc(hdl
, size
* sizeof (paths
[0]))) == NULL
) {
2160 (void) strlcpy(paths
[0], zhp
->zpool_name
, sizeof (paths
[0]));
2164 if (fstatat(base
, paths
[curr
], &st
, AT_SYMLINK_NOFOLLOW
) != 0)
2167 if (S_ISDIR(st
.st_mode
)) {
2168 if ((fd
= openat(base
, paths
[curr
], O_RDONLY
)) < 0)
2171 if ((dirp
= fdopendir(fd
)) == NULL
) {
2176 while ((dp
= readdir(dirp
)) != NULL
) {
2177 if (dp
->d_name
[0] == '.')
2180 if (curr
+ 1 == size
) {
2181 paths
= zfs_realloc(hdl
, paths
,
2182 size
* sizeof (paths
[0]),
2183 size
* 2 * sizeof (paths
[0]));
2184 if (paths
== NULL
) {
2185 (void) closedir(dirp
);
2193 (void) strlcpy(paths
[curr
+ 1], paths
[curr
],
2194 sizeof (paths
[curr
+ 1]));
2195 (void) strlcat(paths
[curr
], "/",
2196 sizeof (paths
[curr
]));
2197 (void) strlcat(paths
[curr
], dp
->d_name
,
2198 sizeof (paths
[curr
]));
2202 (void) closedir(dirp
);
2205 if ((ret
= cb(paths
[curr
], data
)) != 0)
2223 typedef struct zvol_cb
{
2224 zpool_handle_t
*zcb_pool
;
2225 boolean_t zcb_create
;
2230 do_zvol_create(zfs_handle_t
*zhp
, void *data
)
2234 if (ZFS_IS_VOLUME(zhp
)) {
2235 (void) zvol_create_link(zhp
->zfs_hdl
, zhp
->zfs_name
);
2236 ret
= zfs_iter_snapshots(zhp
, do_zvol_create
, NULL
);
2240 ret
= zfs_iter_filesystems(zhp
, do_zvol_create
, NULL
);
2248 * Iterate over all zvols in the pool and make any necessary minor nodes.
2251 zpool_create_zvol_links(zpool_handle_t
*zhp
)
2257 * If the pool is unavailable, just return success.
2259 if ((zfp
= make_dataset_handle(zhp
->zpool_hdl
,
2260 zhp
->zpool_name
)) == NULL
)
2263 ret
= zfs_iter_filesystems(zfp
, do_zvol_create
, NULL
);
2270 do_zvol_remove(const char *dataset
, void *data
)
2272 zpool_handle_t
*zhp
= data
;
2274 return (zvol_remove_link(zhp
->zpool_hdl
, dataset
));
2278 * Iterate over all zvols in the pool and remove any minor nodes. We iterate
2279 * by examining the /dev links so that a corrupted pool doesn't impede this
2283 zpool_remove_zvol_links(zpool_handle_t
*zhp
)
2285 return (zpool_iter_zvol(zhp
, do_zvol_remove
, zhp
));
2289 * Convert from a devid string to a path.
2292 devid_to_path(char *devid_str
)
2297 devid_nmlist_t
*list
= NULL
;
2300 if (devid_str_decode(devid_str
, &devid
, &minor
) != 0)
2303 ret
= devid_deviceid_to_nmlist("/dev", devid
, minor
, &list
);
2305 devid_str_free(minor
);
2311 if ((path
= strdup(list
[0].devname
)) == NULL
)
2314 devid_free_nmlist(list
);
2320 * Convert from a path to a devid string.
2323 path_to_devid(const char *path
)
2329 if ((fd
= open(path
, O_RDONLY
)) < 0)
2334 if (devid_get(fd
, &devid
) == 0) {
2335 if (devid_get_minor_name(fd
, &minor
) == 0)
2336 ret
= devid_str_encode(devid
, minor
);
2338 devid_str_free(minor
);
2347 * Issue the necessary ioctl() to update the stored path value for the vdev. We
2348 * ignore any failure here, since a common case is for an unprivileged user to
2349 * type 'zpool status', and we'll display the correct information anyway.
2352 set_path(zpool_handle_t
*zhp
, nvlist_t
*nv
, const char *path
)
2354 zfs_cmd_t zc
= { 0 };
2356 (void) strncpy(zc
.zc_name
, zhp
->zpool_name
, sizeof (zc
.zc_name
));
2357 (void) strncpy(zc
.zc_value
, path
, sizeof (zc
.zc_value
));
2358 verify(nvlist_lookup_uint64(nv
, ZPOOL_CONFIG_GUID
,
2361 (void) ioctl(zhp
->zpool_hdl
->libzfs_fd
, ZFS_IOC_VDEV_SETPATH
, &zc
);
2365 * Given a vdev, return the name to display in iostat. If the vdev has a path,
2366 * we use that, stripping off any leading "/dev/dsk/"; if not, we use the type.
2367 * We also check if this is a whole disk, in which case we strip off the
2368 * trailing 's0' slice name.
2370 * This routine is also responsible for identifying when disks have been
2371 * reconfigured in a new location. The kernel will have opened the device by
2372 * devid, but the path will still refer to the old location. To catch this, we
2373 * first do a path -> devid translation (which is fast for the common case). If
2374 * the devid matches, we're done. If not, we do a reverse devid -> path
2375 * translation and issue the appropriate ioctl() to update the path of the vdev.
2376 * If 'zhp' is NULL, then this is an exported pool, and we don't need to do any
2380 zpool_vdev_name(libzfs_handle_t
*hdl
, zpool_handle_t
*zhp
, nvlist_t
*nv
)
2388 if (nvlist_lookup_uint64(nv
, ZPOOL_CONFIG_NOT_PRESENT
,
2390 verify(nvlist_lookup_uint64(nv
, ZPOOL_CONFIG_GUID
,
2392 (void) snprintf(buf
, sizeof (buf
), "%llu",
2393 (u_longlong_t
)value
);
2395 } else if (nvlist_lookup_string(nv
, ZPOOL_CONFIG_PATH
, &path
) == 0) {
2398 * If the device is dead (faulted, offline, etc) then don't
2399 * bother opening it. Otherwise we may be forcing the user to
2400 * open a misbehaving device, which can have undesirable
2403 if ((nvlist_lookup_uint64_array(nv
, ZPOOL_CONFIG_STATS
,
2404 (uint64_t **)&vs
, &vsc
) != 0 ||
2405 vs
->vs_state
>= VDEV_STATE_DEGRADED
) &&
2407 nvlist_lookup_string(nv
, ZPOOL_CONFIG_DEVID
, &devid
) == 0) {
2409 * Determine if the current path is correct.
2411 char *newdevid
= path_to_devid(path
);
2413 if (newdevid
== NULL
||
2414 strcmp(devid
, newdevid
) != 0) {
2417 if ((newpath
= devid_to_path(devid
)) != NULL
) {
2419 * Update the path appropriately.
2421 set_path(zhp
, nv
, newpath
);
2422 if (nvlist_add_string(nv
,
2423 ZPOOL_CONFIG_PATH
, newpath
) == 0)
2424 verify(nvlist_lookup_string(nv
,
2432 devid_str_free(newdevid
);
2435 if (strncmp(path
, "/dev/dsk/", 9) == 0)
2438 if (nvlist_lookup_uint64(nv
, ZPOOL_CONFIG_WHOLE_DISK
,
2439 &value
) == 0 && value
) {
2440 char *tmp
= zfs_strdup(hdl
, path
);
2443 tmp
[strlen(path
) - 2] = '\0';
2447 verify(nvlist_lookup_string(nv
, ZPOOL_CONFIG_TYPE
, &path
) == 0);
2450 * If it's a raidz device, we need to stick in the parity level.
2452 if (strcmp(path
, VDEV_TYPE_RAIDZ
) == 0) {
2453 verify(nvlist_lookup_uint64(nv
, ZPOOL_CONFIG_NPARITY
,
2455 (void) snprintf(buf
, sizeof (buf
), "%s%llu", path
,
2456 (u_longlong_t
)value
);
2461 return (zfs_strdup(hdl
, path
));
2465 zbookmark_compare(const void *a
, const void *b
)
2467 return (memcmp(a
, b
, sizeof (zbookmark_t
)));
2471 * Retrieve the persistent error log, uniquify the members, and return to the
2475 zpool_get_errlog(zpool_handle_t
*zhp
, nvlist_t
**nverrlistp
)
2477 zfs_cmd_t zc
= { 0 };
2479 zbookmark_t
*zb
= NULL
;
2483 * Retrieve the raw error list from the kernel. If the number of errors
2484 * has increased, allocate more space and continue until we get the
2487 verify(nvlist_lookup_uint64(zhp
->zpool_config
, ZPOOL_CONFIG_ERRCOUNT
,
2491 if ((zc
.zc_nvlist_dst
= (uintptr_t)zfs_alloc(zhp
->zpool_hdl
,
2492 count
* sizeof (zbookmark_t
))) == (uintptr_t)NULL
)
2494 zc
.zc_nvlist_dst_size
= count
;
2495 (void) strcpy(zc
.zc_name
, zhp
->zpool_name
);
2497 if (ioctl(zhp
->zpool_hdl
->libzfs_fd
, ZFS_IOC_ERROR_LOG
,
2499 free((void *)(uintptr_t)zc
.zc_nvlist_dst
);
2500 if (errno
== ENOMEM
) {
2501 count
= zc
.zc_nvlist_dst_size
;
2502 if ((zc
.zc_nvlist_dst
= (uintptr_t)
2503 zfs_alloc(zhp
->zpool_hdl
, count
*
2504 sizeof (zbookmark_t
))) == (uintptr_t)NULL
)
2515 * Sort the resulting bookmarks. This is a little confusing due to the
2516 * implementation of ZFS_IOC_ERROR_LOG. The bookmarks are copied last
2517 * to first, and 'zc_nvlist_dst_size' indicates the number of boomarks
2518 * _not_ copied as part of the process. So we point the start of our
2519 * array appropriate and decrement the total number of elements.
2521 zb
= ((zbookmark_t
*)(uintptr_t)zc
.zc_nvlist_dst
) +
2522 zc
.zc_nvlist_dst_size
;
2523 count
-= zc
.zc_nvlist_dst_size
;
2525 qsort(zb
, count
, sizeof (zbookmark_t
), zbookmark_compare
);
2527 verify(nvlist_alloc(nverrlistp
, 0, KM_SLEEP
) == 0);
2530 * Fill in the nverrlistp with nvlist's of dataset and object numbers.
2532 for (i
= 0; i
< count
; i
++) {
2535 /* ignoring zb_blkid and zb_level for now */
2536 if (i
> 0 && zb
[i
-1].zb_objset
== zb
[i
].zb_objset
&&
2537 zb
[i
-1].zb_object
== zb
[i
].zb_object
)
2540 if (nvlist_alloc(&nv
, NV_UNIQUE_NAME
, KM_SLEEP
) != 0)
2542 if (nvlist_add_uint64(nv
, ZPOOL_ERR_DATASET
,
2543 zb
[i
].zb_objset
) != 0) {
2547 if (nvlist_add_uint64(nv
, ZPOOL_ERR_OBJECT
,
2548 zb
[i
].zb_object
) != 0) {
2552 if (nvlist_add_nvlist(*nverrlistp
, "ejk", nv
) != 0) {
2559 free((void *)(uintptr_t)zc
.zc_nvlist_dst
);
2563 free((void *)(uintptr_t)zc
.zc_nvlist_dst
);
2564 return (no_memory(zhp
->zpool_hdl
));
2568 * Upgrade a ZFS pool to the latest on-disk version.
2571 zpool_upgrade(zpool_handle_t
*zhp
, uint64_t new_version
)
2573 zfs_cmd_t zc
= { 0 };
2574 libzfs_handle_t
*hdl
= zhp
->zpool_hdl
;
2576 (void) strcpy(zc
.zc_name
, zhp
->zpool_name
);
2577 zc
.zc_cookie
= new_version
;
2579 if (zfs_ioctl(hdl
, ZFS_IOC_POOL_UPGRADE
, &zc
) != 0)
2580 return (zpool_standard_error_fmt(hdl
, errno
,
2581 dgettext(TEXT_DOMAIN
, "cannot upgrade '%s'"),
2587 zpool_set_history_str(const char *subcommand
, int argc
, char **argv
,
2592 (void) strlcpy(history_str
, subcommand
, HIS_MAX_RECORD_LEN
);
2593 for (i
= 1; i
< argc
; i
++) {
2594 if (strlen(history_str
) + 1 + strlen(argv
[i
]) >
2597 (void) strlcat(history_str
, " ", HIS_MAX_RECORD_LEN
);
2598 (void) strlcat(history_str
, argv
[i
], HIS_MAX_RECORD_LEN
);
2603 * Stage command history for logging.
2606 zpool_stage_history(libzfs_handle_t
*hdl
, const char *history_str
)
2608 if (history_str
== NULL
)
2611 if (strlen(history_str
) > HIS_MAX_RECORD_LEN
)
2614 if (hdl
->libzfs_log_str
!= NULL
)
2615 free(hdl
->libzfs_log_str
);
2617 if ((hdl
->libzfs_log_str
= strdup(history_str
)) == NULL
)
2618 return (no_memory(hdl
));
2624 * Perform ioctl to get some command history of a pool.
2626 * 'buf' is the buffer to fill up to 'len' bytes. 'off' is the
2627 * logical offset of the history buffer to start reading from.
2629 * Upon return, 'off' is the next logical offset to read from and
2630 * 'len' is the actual amount of bytes read into 'buf'.
2633 get_history(zpool_handle_t
*zhp
, char *buf
, uint64_t *off
, uint64_t *len
)
2635 zfs_cmd_t zc
= { 0 };
2636 libzfs_handle_t
*hdl
= zhp
->zpool_hdl
;
2638 (void) strlcpy(zc
.zc_name
, zhp
->zpool_name
, sizeof (zc
.zc_name
));
2640 zc
.zc_history
= (uint64_t)(uintptr_t)buf
;
2641 zc
.zc_history_len
= *len
;
2642 zc
.zc_history_offset
= *off
;
2644 if (ioctl(hdl
->libzfs_fd
, ZFS_IOC_POOL_GET_HISTORY
, &zc
) != 0) {
2647 return (zfs_error_fmt(hdl
, EZFS_PERM
,
2648 dgettext(TEXT_DOMAIN
,
2649 "cannot show history for pool '%s'"),
2652 return (zfs_error_fmt(hdl
, EZFS_NOHISTORY
,
2653 dgettext(TEXT_DOMAIN
, "cannot get history for pool "
2654 "'%s'"), zhp
->zpool_name
));
2656 return (zfs_error_fmt(hdl
, EZFS_BADVERSION
,
2657 dgettext(TEXT_DOMAIN
, "cannot get history for pool "
2658 "'%s', pool must be upgraded"), zhp
->zpool_name
));
2660 return (zpool_standard_error_fmt(hdl
, errno
,
2661 dgettext(TEXT_DOMAIN
,
2662 "cannot get history for '%s'"), zhp
->zpool_name
));
2666 *len
= zc
.zc_history_len
;
2667 *off
= zc
.zc_history_offset
;
2673 * Process the buffer of nvlists, unpacking and storing each nvlist record
2674 * into 'records'. 'leftover' is set to the number of bytes that weren't
2675 * processed as there wasn't a complete record.
2678 zpool_history_unpack(char *buf
, uint64_t bytes_read
, uint64_t *leftover
,
2679 nvlist_t
***records
, uint_t
*numrecords
)
2685 while (bytes_read
> sizeof (reclen
)) {
2687 /* get length of packed record (stored as little endian) */
2688 for (i
= 0, reclen
= 0; i
< sizeof (reclen
); i
++)
2689 reclen
+= (uint64_t)(((uchar_t
*)buf
)[i
]) << (8*i
);
2691 if (bytes_read
< sizeof (reclen
) + reclen
)
2695 if (nvlist_unpack(buf
+ sizeof (reclen
), reclen
, &nv
, 0) != 0)
2697 bytes_read
-= sizeof (reclen
) + reclen
;
2698 buf
+= sizeof (reclen
) + reclen
;
2700 /* add record to nvlist array */
2702 if (ISP2(*numrecords
+ 1)) {
2703 *records
= realloc(*records
,
2704 *numrecords
* 2 * sizeof (nvlist_t
*));
2706 (*records
)[*numrecords
- 1] = nv
;
2709 *leftover
= bytes_read
;
2713 #define HIS_BUF_LEN (128*1024)
2716 * Retrieve the command history of a pool.
2719 zpool_get_history(zpool_handle_t
*zhp
, nvlist_t
**nvhisp
)
2721 char buf
[HIS_BUF_LEN
];
2723 nvlist_t
**records
= NULL
;
2724 uint_t numrecords
= 0;
2728 uint64_t bytes_read
= sizeof (buf
);
2731 if ((err
= get_history(zhp
, buf
, &off
, &bytes_read
)) != 0)
2734 /* if nothing else was read in, we're at EOF, just return */
2738 if ((err
= zpool_history_unpack(buf
, bytes_read
,
2739 &leftover
, &records
, &numrecords
)) != 0)
2747 verify(nvlist_alloc(nvhisp
, NV_UNIQUE_NAME
, 0) == 0);
2748 verify(nvlist_add_nvlist_array(*nvhisp
, ZPOOL_HIST_RECORD
,
2749 records
, numrecords
) == 0);
2751 for (i
= 0; i
< numrecords
; i
++)
2752 nvlist_free(records
[i
]);
2759 zpool_obj_to_path(zpool_handle_t
*zhp
, uint64_t dsobj
, uint64_t obj
,
2760 char *pathname
, size_t len
)
2762 zfs_cmd_t zc
= { 0 };
2763 boolean_t mounted
= B_FALSE
;
2764 char *mntpnt
= NULL
;
2765 char dsname
[MAXNAMELEN
];
2768 /* special case for the MOS */
2769 (void) snprintf(pathname
, len
, "<metadata>:<0x%llx>", obj
);
2773 /* get the dataset's name */
2774 (void) strlcpy(zc
.zc_name
, zhp
->zpool_name
, sizeof (zc
.zc_name
));
2776 if (ioctl(zhp
->zpool_hdl
->libzfs_fd
,
2777 ZFS_IOC_DSOBJ_TO_DSNAME
, &zc
) != 0) {
2778 /* just write out a path of two object numbers */
2779 (void) snprintf(pathname
, len
, "<0x%llx>:<0x%llx>",
2783 (void) strlcpy(dsname
, zc
.zc_value
, sizeof (dsname
));
2785 /* find out if the dataset is mounted */
2786 mounted
= is_mounted(zhp
->zpool_hdl
, dsname
, &mntpnt
);
2788 /* get the corrupted object's path */
2789 (void) strlcpy(zc
.zc_name
, dsname
, sizeof (zc
.zc_name
));
2791 if (ioctl(zhp
->zpool_hdl
->libzfs_fd
, ZFS_IOC_OBJ_TO_PATH
,
2794 (void) snprintf(pathname
, len
, "%s%s", mntpnt
,
2797 (void) snprintf(pathname
, len
, "%s:%s",
2798 dsname
, zc
.zc_value
);
2801 (void) snprintf(pathname
, len
, "%s:<0x%llx>", dsname
, obj
);
2806 #define RDISK_ROOT "/dev/rdsk"
2807 #define BACKUP_SLICE "s2"
2809 * Don't start the slice at the default block of 34; many storage
2810 * devices will use a stripe width of 128k, so start there instead.
2812 #define NEW_START_BLOCK 256
2815 * Read the EFI label from the config, if a label does not exist then
2816 * pass back the error to the caller. If the caller has passed a non-NULL
2817 * diskaddr argument then we set it to the starting address of the EFI
2821 read_efi_label(nvlist_t
*config
, diskaddr_t
*sb
)
2825 char diskname
[MAXPATHLEN
];
2828 if (nvlist_lookup_string(config
, ZPOOL_CONFIG_PATH
, &path
) != 0)
2831 (void) snprintf(diskname
, sizeof (diskname
), "%s%s", RDISK_ROOT
,
2832 strrchr(path
, '/'));
2833 if ((fd
= open(diskname
, O_RDONLY
|O_NDELAY
)) >= 0) {
2834 struct dk_gpt
*vtoc
;
2836 if ((err
= efi_alloc_and_read(fd
, &vtoc
)) >= 0) {
2838 *sb
= vtoc
->efi_parts
[0].p_start
;
2847 * determine where a partition starts on a disk in the current
2851 find_start_block(nvlist_t
*config
)
2855 diskaddr_t sb
= MAXOFFSET_T
;
2858 if (nvlist_lookup_nvlist_array(config
,
2859 ZPOOL_CONFIG_CHILDREN
, &child
, &children
) != 0) {
2860 if (nvlist_lookup_uint64(config
,
2861 ZPOOL_CONFIG_WHOLE_DISK
,
2862 &wholedisk
) != 0 || !wholedisk
) {
2863 return (MAXOFFSET_T
);
2865 if (read_efi_label(config
, &sb
) < 0)
2870 for (c
= 0; c
< children
; c
++) {
2871 sb
= find_start_block(child
[c
]);
2872 if (sb
!= MAXOFFSET_T
) {
2876 return (MAXOFFSET_T
);
2880 * Label an individual disk. The name provided is the short name,
2881 * stripped of any leading /dev path.
2884 zpool_label_disk(libzfs_handle_t
*hdl
, zpool_handle_t
*zhp
, char *name
)
2886 char path
[MAXPATHLEN
];
2887 struct dk_gpt
*vtoc
;
2889 size_t resv
= EFI_MIN_RESV_SIZE
;
2890 uint64_t slice_size
;
2891 diskaddr_t start_block
;
2894 /* prepare an error message just in case */
2895 (void) snprintf(errbuf
, sizeof (errbuf
),
2896 dgettext(TEXT_DOMAIN
, "cannot label '%s'"), name
);
2901 if (pool_is_bootable(zhp
)) {
2902 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
2903 "EFI labeled devices are not supported on root "
2905 return (zfs_error(hdl
, EZFS_POOL_NOTSUP
, errbuf
));
2908 verify(nvlist_lookup_nvlist(zhp
->zpool_config
,
2909 ZPOOL_CONFIG_VDEV_TREE
, &nvroot
) == 0);
2911 if (zhp
->zpool_start_block
== 0)
2912 start_block
= find_start_block(nvroot
);
2914 start_block
= zhp
->zpool_start_block
;
2915 zhp
->zpool_start_block
= start_block
;
2918 start_block
= NEW_START_BLOCK
;
2921 (void) snprintf(path
, sizeof (path
), "%s/%s%s", RDISK_ROOT
, name
,
2924 if ((fd
= open(path
, O_RDWR
| O_NDELAY
)) < 0) {
2926 * This shouldn't happen. We've long since verified that this
2927 * is a valid device.
2930 dgettext(TEXT_DOMAIN
, "unable to open device"));
2931 return (zfs_error(hdl
, EZFS_OPENFAILED
, errbuf
));
2934 if (efi_alloc_and_init(fd
, EFI_NUMPAR
, &vtoc
) != 0) {
2936 * The only way this can fail is if we run out of memory, or we
2937 * were unable to read the disk's capacity
2939 if (errno
== ENOMEM
)
2940 (void) no_memory(hdl
);
2943 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
2944 "unable to read disk capacity"), name
);
2946 return (zfs_error(hdl
, EZFS_NOCAP
, errbuf
));
2949 slice_size
= vtoc
->efi_last_u_lba
+ 1;
2950 slice_size
-= EFI_MIN_RESV_SIZE
;
2951 if (start_block
== MAXOFFSET_T
)
2952 start_block
= NEW_START_BLOCK
;
2953 slice_size
-= start_block
;
2955 vtoc
->efi_parts
[0].p_start
= start_block
;
2956 vtoc
->efi_parts
[0].p_size
= slice_size
;
2959 * Why we use V_USR: V_BACKUP confuses users, and is considered
2960 * disposable by some EFI utilities (since EFI doesn't have a backup
2961 * slice). V_UNASSIGNED is supposed to be used only for zero size
2962 * partitions, and efi_write() will fail if we use it. V_ROOT, V_BOOT,
2963 * etc. were all pretty specific. V_USR is as close to reality as we
2964 * can get, in the absence of V_OTHER.
2966 vtoc
->efi_parts
[0].p_tag
= V_USR
;
2967 (void) strcpy(vtoc
->efi_parts
[0].p_name
, "zfs");
2969 vtoc
->efi_parts
[8].p_start
= slice_size
+ start_block
;
2970 vtoc
->efi_parts
[8].p_size
= resv
;
2971 vtoc
->efi_parts
[8].p_tag
= V_RESERVED
;
2973 if (efi_write(fd
, vtoc
) != 0) {
2975 * Some block drivers (like pcata) may not support EFI
2976 * GPT labels. Print out a helpful error message dir-
2977 * ecting the user to manually label the disk and give
2983 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
2984 "try using fdisk(1M) and then provide a specific slice"));
2985 return (zfs_error(hdl
, EZFS_LABELFAILED
, errbuf
));
2994 supported_dump_vdev_type(libzfs_handle_t
*hdl
, nvlist_t
*config
, char *errbuf
)
3000 verify(nvlist_lookup_string(config
, ZPOOL_CONFIG_TYPE
, &type
) == 0);
3001 if (strcmp(type
, VDEV_TYPE_RAIDZ
) == 0 ||
3002 strcmp(type
, VDEV_TYPE_FILE
) == 0 ||
3003 strcmp(type
, VDEV_TYPE_LOG
) == 0 ||
3004 strcmp(type
, VDEV_TYPE_MISSING
) == 0) {
3005 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
3006 "vdev type '%s' is not supported"), type
);
3007 (void) zfs_error(hdl
, EZFS_VDEVNOTSUP
, errbuf
);
3010 if (nvlist_lookup_nvlist_array(config
, ZPOOL_CONFIG_CHILDREN
,
3011 &child
, &children
) == 0) {
3012 for (c
= 0; c
< children
; c
++) {
3013 if (!supported_dump_vdev_type(hdl
, child
[c
], errbuf
))
3021 * check if this zvol is allowable for use as a dump device; zero if
3022 * it is, > 0 if it isn't, < 0 if it isn't a zvol
3025 zvol_check_dump_config(char *arg
)
3027 zpool_handle_t
*zhp
= NULL
;
3028 nvlist_t
*config
, *nvroot
;
3032 libzfs_handle_t
*hdl
;
3034 char poolname
[ZPOOL_MAXNAMELEN
];
3035 int pathlen
= strlen(ZVOL_FULL_DEV_DIR
);
3038 if (strncmp(arg
, ZVOL_FULL_DEV_DIR
, pathlen
)) {
3042 (void) snprintf(errbuf
, sizeof (errbuf
), dgettext(TEXT_DOMAIN
,
3043 "dump is not supported on device '%s'"), arg
);
3045 if ((hdl
= libzfs_init()) == NULL
)
3047 libzfs_print_on_error(hdl
, B_TRUE
);
3049 volname
= arg
+ pathlen
;
3051 /* check the configuration of the pool */
3052 if ((p
= strchr(volname
, '/')) == NULL
) {
3053 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
3054 "malformed dataset name"));
3055 (void) zfs_error(hdl
, EZFS_INVALIDNAME
, errbuf
);
3057 } else if (p
- volname
>= ZFS_MAXNAMELEN
) {
3058 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
3059 "dataset name is too long"));
3060 (void) zfs_error(hdl
, EZFS_NAMETOOLONG
, errbuf
);
3063 (void) strncpy(poolname
, volname
, p
- volname
);
3064 poolname
[p
- volname
] = '\0';
3067 if ((zhp
= zpool_open(hdl
, poolname
)) == NULL
) {
3068 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
3069 "could not open pool '%s'"), poolname
);
3070 (void) zfs_error(hdl
, EZFS_OPENFAILED
, errbuf
);
3073 config
= zpool_get_config(zhp
, NULL
);
3074 if (nvlist_lookup_nvlist(config
, ZPOOL_CONFIG_VDEV_TREE
,
3076 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
3077 "could not obtain vdev configuration for '%s'"), poolname
);
3078 (void) zfs_error(hdl
, EZFS_INVALCONFIG
, errbuf
);
3082 verify(nvlist_lookup_nvlist_array(nvroot
, ZPOOL_CONFIG_CHILDREN
,
3083 &top
, &toplevels
) == 0);
3084 if (toplevels
!= 1) {
3085 zfs_error_aux(hdl
, dgettext(TEXT_DOMAIN
,
3086 "'%s' has multiple top level vdevs"), poolname
);
3087 (void) zfs_error(hdl
, EZFS_DEVOVERFLOW
, errbuf
);
3091 if (!supported_dump_vdev_type(hdl
, top
[0], errbuf
)) {