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]
22 * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
23 * Copyright 2011 Nexenta Systems, Inc. All rights reserved.
24 * Copyright (c) 2012 by Delphix. All rights reserved.
28 * Pool import support functions.
30 * To import a pool, we rely on reading the configuration information from the
31 * ZFS label of each device. If we successfully read the label, then we
32 * organize the configuration information in the following hierarchy:
34 * pool guid -> toplevel vdev guid -> label txg
36 * Duplicate entries matching this same tuple will be discarded. Once we have
37 * examined every device, we pick the best label txg config for each toplevel
38 * vdev. We then arrange these toplevel vdevs into a complete pool config, and
39 * update any paths that have changed. Finally, we attempt to import the pool
40 * using our derived config, and record the results.
55 #include <sys/dktp/fdisk.h>
56 #include <sys/efi_partition.h>
58 #include <sys/vdev_impl.h>
60 #include <blkid/blkid.h>
64 #include "libzfs_impl.h"
67 * Intermediate structures used to gather configuration information.
69 typedef struct config_entry
{
72 struct config_entry
*ce_next
;
75 typedef struct vdev_entry
{
77 config_entry_t
*ve_configs
;
78 struct vdev_entry
*ve_next
;
81 typedef struct pool_entry
{
83 vdev_entry_t
*pe_vdevs
;
84 struct pool_entry
*pe_next
;
87 typedef struct name_entry
{
91 struct name_entry
*ne_next
;
94 typedef struct pool_list
{
100 get_devid(const char *path
)
106 if ((fd
= open(path
, O_RDONLY
)) < 0)
111 if (devid_get(fd
, &devid
) == 0) {
112 if (devid_get_minor_name(fd
, &minor
) == 0)
113 ret
= devid_str_encode(devid
, minor
);
115 devid_str_free(minor
);
125 * Go through and fix up any path and/or devid information for the given vdev
129 fix_paths(nvlist_t
*nv
, name_entry_t
*names
)
134 name_entry_t
*ne
, *best
;
137 if (nvlist_lookup_nvlist_array(nv
, ZPOOL_CONFIG_CHILDREN
,
138 &child
, &children
) == 0) {
139 for (c
= 0; c
< children
; c
++)
140 if (fix_paths(child
[c
], names
) != 0)
146 * This is a leaf (file or disk) vdev. In either case, go through
147 * the name list and see if we find a matching guid. If so, replace
148 * the path and see if we can calculate a new devid.
150 * There may be multiple names associated with a particular guid, in
151 * which case we have overlapping partitions or multiple paths to the
152 * same disk. In this case we prefer to use the path name which
153 * matches the ZPOOL_CONFIG_PATH. If no matching entry is found we
154 * use the lowest order device which corresponds to the first match
155 * while traversing the ZPOOL_IMPORT_PATH search path.
157 verify(nvlist_lookup_uint64(nv
, ZPOOL_CONFIG_GUID
, &guid
) == 0);
158 if (nvlist_lookup_string(nv
, ZPOOL_CONFIG_PATH
, &path
) != 0)
162 for (ne
= names
; ne
!= NULL
; ne
= ne
->ne_next
) {
163 if (ne
->ne_guid
== guid
) {
170 if ((strlen(path
) == strlen(ne
->ne_name
)) &&
171 !strncmp(path
, ne
->ne_name
, strlen(path
))) {
176 if (best
== NULL
|| ne
->ne_order
< best
->ne_order
)
184 if (nvlist_add_string(nv
, ZPOOL_CONFIG_PATH
, best
->ne_name
) != 0)
187 if ((devid
= get_devid(best
->ne_name
)) == NULL
) {
188 (void) nvlist_remove_all(nv
, ZPOOL_CONFIG_DEVID
);
190 if (nvlist_add_string(nv
, ZPOOL_CONFIG_DEVID
, devid
) != 0)
192 devid_str_free(devid
);
199 * Add the given configuration to the list of known devices.
202 add_config(libzfs_handle_t
*hdl
, pool_list_t
*pl
, const char *path
,
203 int order
, nvlist_t
*config
)
205 uint64_t pool_guid
, vdev_guid
, top_guid
, txg
, state
;
212 * If this is a hot spare not currently in use or level 2 cache
213 * device, add it to the list of names to translate, but don't do
216 if (nvlist_lookup_uint64(config
, ZPOOL_CONFIG_POOL_STATE
,
218 (state
== POOL_STATE_SPARE
|| state
== POOL_STATE_L2CACHE
) &&
219 nvlist_lookup_uint64(config
, ZPOOL_CONFIG_GUID
, &vdev_guid
) == 0) {
220 if ((ne
= zfs_alloc(hdl
, sizeof (name_entry_t
))) == NULL
)
223 if ((ne
->ne_name
= zfs_strdup(hdl
, path
)) == NULL
) {
227 ne
->ne_guid
= vdev_guid
;
228 ne
->ne_order
= order
;
229 ne
->ne_next
= pl
->names
;
235 * If we have a valid config but cannot read any of these fields, then
236 * it means we have a half-initialized label. In vdev_label_init()
237 * we write a label with txg == 0 so that we can identify the device
238 * in case the user refers to the same disk later on. If we fail to
239 * create the pool, we'll be left with a label in this state
240 * which should not be considered part of a valid pool.
242 if (nvlist_lookup_uint64(config
, ZPOOL_CONFIG_POOL_GUID
,
244 nvlist_lookup_uint64(config
, ZPOOL_CONFIG_GUID
,
246 nvlist_lookup_uint64(config
, ZPOOL_CONFIG_TOP_GUID
,
248 nvlist_lookup_uint64(config
, ZPOOL_CONFIG_POOL_TXG
,
249 &txg
) != 0 || txg
== 0) {
255 * First, see if we know about this pool. If not, then add it to the
256 * list of known pools.
258 for (pe
= pl
->pools
; pe
!= NULL
; pe
= pe
->pe_next
) {
259 if (pe
->pe_guid
== pool_guid
)
264 if ((pe
= zfs_alloc(hdl
, sizeof (pool_entry_t
))) == NULL
) {
268 pe
->pe_guid
= pool_guid
;
269 pe
->pe_next
= pl
->pools
;
274 * Second, see if we know about this toplevel vdev. Add it if its
277 for (ve
= pe
->pe_vdevs
; ve
!= NULL
; ve
= ve
->ve_next
) {
278 if (ve
->ve_guid
== top_guid
)
283 if ((ve
= zfs_alloc(hdl
, sizeof (vdev_entry_t
))) == NULL
) {
287 ve
->ve_guid
= top_guid
;
288 ve
->ve_next
= pe
->pe_vdevs
;
293 * Third, see if we have a config with a matching transaction group. If
294 * so, then we do nothing. Otherwise, add it to the list of known
297 for (ce
= ve
->ve_configs
; ce
!= NULL
; ce
= ce
->ce_next
) {
298 if (ce
->ce_txg
== txg
)
303 if ((ce
= zfs_alloc(hdl
, sizeof (config_entry_t
))) == NULL
) {
308 ce
->ce_config
= config
;
309 ce
->ce_next
= ve
->ve_configs
;
316 * At this point we've successfully added our config to the list of
317 * known configs. The last thing to do is add the vdev guid -> path
318 * mappings so that we can fix up the configuration as necessary before
321 if ((ne
= zfs_alloc(hdl
, sizeof (name_entry_t
))) == NULL
)
324 if ((ne
->ne_name
= zfs_strdup(hdl
, path
)) == NULL
) {
329 ne
->ne_guid
= vdev_guid
;
330 ne
->ne_order
= order
;
331 ne
->ne_next
= pl
->names
;
338 * Returns true if the named pool matches the given GUID.
341 pool_active(libzfs_handle_t
*hdl
, const char *name
, uint64_t guid
,
347 if (zpool_open_silent(hdl
, name
, &zhp
) != 0)
355 verify(nvlist_lookup_uint64(zhp
->zpool_config
, ZPOOL_CONFIG_POOL_GUID
,
360 *isactive
= (theguid
== guid
);
365 refresh_config(libzfs_handle_t
*hdl
, nvlist_t
*config
)
368 zfs_cmd_t zc
= { "\0", "\0", "\0", "\0", 0 };
371 if (zcmd_write_conf_nvlist(hdl
, &zc
, config
) != 0)
374 if (zcmd_alloc_dst_nvlist(hdl
, &zc
,
375 zc
.zc_nvlist_conf_size
* 2) != 0) {
376 zcmd_free_nvlists(&zc
);
380 while ((err
= ioctl(hdl
->libzfs_fd
, ZFS_IOC_POOL_TRYIMPORT
,
381 &zc
)) != 0 && errno
== ENOMEM
) {
382 if (zcmd_expand_dst_nvlist(hdl
, &zc
) != 0) {
383 zcmd_free_nvlists(&zc
);
389 zcmd_free_nvlists(&zc
);
393 if (zcmd_read_dst_nvlist(hdl
, &zc
, &nvl
) != 0) {
394 zcmd_free_nvlists(&zc
);
398 zcmd_free_nvlists(&zc
);
403 * Determine if the vdev id is a hole in the namespace.
406 vdev_is_hole(uint64_t *hole_array
, uint_t holes
, uint_t id
)
410 for (c
= 0; c
< holes
; c
++) {
412 /* Top-level is a hole */
413 if (hole_array
[c
] == id
)
420 * Convert our list of pools into the definitive set of configurations. We
421 * start by picking the best config for each toplevel vdev. Once that's done,
422 * we assemble the toplevel vdevs into a full config for the pool. We make a
423 * pass to fix up any incorrect paths, and then add it to the main list to
424 * return to the user.
427 get_configs(libzfs_handle_t
*hdl
, pool_list_t
*pl
, boolean_t active_ok
)
432 nvlist_t
*ret
= NULL
, *config
= NULL
, *tmp
= NULL
, *nvtop
, *nvroot
;
433 nvlist_t
**spares
, **l2cache
;
434 uint_t i
, nspares
, nl2cache
;
435 boolean_t config_seen
;
437 char *name
, *hostname
= NULL
;
440 nvlist_t
**child
= NULL
;
442 uint64_t *hole_array
, max_id
;
447 boolean_t found_one
= B_FALSE
;
448 boolean_t valid_top_config
= B_FALSE
;
450 if (nvlist_alloc(&ret
, 0, 0) != 0)
453 for (pe
= pl
->pools
; pe
!= NULL
; pe
= pe
->pe_next
) {
454 uint64_t id
, max_txg
= 0;
456 if (nvlist_alloc(&config
, NV_UNIQUE_NAME
, 0) != 0)
458 config_seen
= B_FALSE
;
461 * Iterate over all toplevel vdevs. Grab the pool configuration
462 * from the first one we find, and then go through the rest and
463 * add them as necessary to the 'vdevs' member of the config.
465 for (ve
= pe
->pe_vdevs
; ve
!= NULL
; ve
= ve
->ve_next
) {
468 * Determine the best configuration for this vdev by
469 * selecting the config with the latest transaction
473 for (ce
= ve
->ve_configs
; ce
!= NULL
;
476 if (ce
->ce_txg
> best_txg
) {
478 best_txg
= ce
->ce_txg
;
483 * We rely on the fact that the max txg for the
484 * pool will contain the most up-to-date information
485 * about the valid top-levels in the vdev namespace.
487 if (best_txg
> max_txg
) {
488 (void) nvlist_remove(config
,
489 ZPOOL_CONFIG_VDEV_CHILDREN
,
491 (void) nvlist_remove(config
,
492 ZPOOL_CONFIG_HOLE_ARRAY
,
493 DATA_TYPE_UINT64_ARRAY
);
499 valid_top_config
= B_FALSE
;
501 if (nvlist_lookup_uint64(tmp
,
502 ZPOOL_CONFIG_VDEV_CHILDREN
, &max_id
) == 0) {
503 verify(nvlist_add_uint64(config
,
504 ZPOOL_CONFIG_VDEV_CHILDREN
,
506 valid_top_config
= B_TRUE
;
509 if (nvlist_lookup_uint64_array(tmp
,
510 ZPOOL_CONFIG_HOLE_ARRAY
, &hole_array
,
512 verify(nvlist_add_uint64_array(config
,
513 ZPOOL_CONFIG_HOLE_ARRAY
,
514 hole_array
, holes
) == 0);
520 * Copy the relevant pieces of data to the pool
526 * pool txg (if available)
527 * comment (if available)
529 * hostid (if available)
530 * hostname (if available)
532 uint64_t state
, version
, pool_txg
;
533 char *comment
= NULL
;
535 version
= fnvlist_lookup_uint64(tmp
,
536 ZPOOL_CONFIG_VERSION
);
537 fnvlist_add_uint64(config
,
538 ZPOOL_CONFIG_VERSION
, version
);
539 guid
= fnvlist_lookup_uint64(tmp
,
540 ZPOOL_CONFIG_POOL_GUID
);
541 fnvlist_add_uint64(config
,
542 ZPOOL_CONFIG_POOL_GUID
, guid
);
543 name
= fnvlist_lookup_string(tmp
,
544 ZPOOL_CONFIG_POOL_NAME
);
545 fnvlist_add_string(config
,
546 ZPOOL_CONFIG_POOL_NAME
, name
);
548 if (nvlist_lookup_uint64(tmp
,
549 ZPOOL_CONFIG_POOL_TXG
, &pool_txg
) == 0)
550 fnvlist_add_uint64(config
,
551 ZPOOL_CONFIG_POOL_TXG
, pool_txg
);
553 if (nvlist_lookup_string(tmp
,
554 ZPOOL_CONFIG_COMMENT
, &comment
) == 0)
555 fnvlist_add_string(config
,
556 ZPOOL_CONFIG_COMMENT
, comment
);
558 state
= fnvlist_lookup_uint64(tmp
,
559 ZPOOL_CONFIG_POOL_STATE
);
560 fnvlist_add_uint64(config
,
561 ZPOOL_CONFIG_POOL_STATE
, state
);
564 if (nvlist_lookup_uint64(tmp
,
565 ZPOOL_CONFIG_HOSTID
, &hostid
) == 0) {
566 fnvlist_add_uint64(config
,
567 ZPOOL_CONFIG_HOSTID
, hostid
);
568 hostname
= fnvlist_lookup_string(tmp
,
569 ZPOOL_CONFIG_HOSTNAME
);
570 fnvlist_add_string(config
,
571 ZPOOL_CONFIG_HOSTNAME
, hostname
);
574 config_seen
= B_TRUE
;
578 * Add this top-level vdev to the child array.
580 verify(nvlist_lookup_nvlist(tmp
,
581 ZPOOL_CONFIG_VDEV_TREE
, &nvtop
) == 0);
582 verify(nvlist_lookup_uint64(nvtop
, ZPOOL_CONFIG_ID
,
585 if (id
>= children
) {
588 newchild
= zfs_alloc(hdl
, (id
+ 1) *
589 sizeof (nvlist_t
*));
590 if (newchild
== NULL
)
593 for (c
= 0; c
< children
; c
++)
594 newchild
[c
] = child
[c
];
600 if (nvlist_dup(nvtop
, &child
[id
], 0) != 0)
606 * If we have information about all the top-levels then
607 * clean up the nvlist which we've constructed. This
608 * means removing any extraneous devices that are
609 * beyond the valid range or adding devices to the end
610 * of our array which appear to be missing.
612 if (valid_top_config
) {
613 if (max_id
< children
) {
614 for (c
= max_id
; c
< children
; c
++)
615 nvlist_free(child
[c
]);
617 } else if (max_id
> children
) {
620 newchild
= zfs_alloc(hdl
, (max_id
) *
621 sizeof (nvlist_t
*));
622 if (newchild
== NULL
)
625 for (c
= 0; c
< children
; c
++)
626 newchild
[c
] = child
[c
];
634 verify(nvlist_lookup_uint64(config
, ZPOOL_CONFIG_POOL_GUID
,
638 * The vdev namespace may contain holes as a result of
639 * device removal. We must add them back into the vdev
640 * tree before we process any missing devices.
643 ASSERT(valid_top_config
);
645 for (c
= 0; c
< children
; c
++) {
648 if (child
[c
] != NULL
||
649 !vdev_is_hole(hole_array
, holes
, c
))
652 if (nvlist_alloc(&holey
, NV_UNIQUE_NAME
,
657 * Holes in the namespace are treated as
658 * "hole" top-level vdevs and have a
659 * special flag set on them.
661 if (nvlist_add_string(holey
,
663 VDEV_TYPE_HOLE
) != 0 ||
664 nvlist_add_uint64(holey
,
665 ZPOOL_CONFIG_ID
, c
) != 0 ||
666 nvlist_add_uint64(holey
,
667 ZPOOL_CONFIG_GUID
, 0ULL) != 0)
674 * Look for any missing top-level vdevs. If this is the case,
675 * create a faked up 'missing' vdev as a placeholder. We cannot
676 * simply compress the child array, because the kernel performs
677 * certain checks to make sure the vdev IDs match their location
678 * in the configuration.
680 for (c
= 0; c
< children
; c
++) {
681 if (child
[c
] == NULL
) {
683 if (nvlist_alloc(&missing
, NV_UNIQUE_NAME
,
686 if (nvlist_add_string(missing
,
688 VDEV_TYPE_MISSING
) != 0 ||
689 nvlist_add_uint64(missing
,
690 ZPOOL_CONFIG_ID
, c
) != 0 ||
691 nvlist_add_uint64(missing
,
692 ZPOOL_CONFIG_GUID
, 0ULL) != 0) {
693 nvlist_free(missing
);
701 * Put all of this pool's top-level vdevs into a root vdev.
703 if (nvlist_alloc(&nvroot
, NV_UNIQUE_NAME
, 0) != 0)
705 if (nvlist_add_string(nvroot
, ZPOOL_CONFIG_TYPE
,
706 VDEV_TYPE_ROOT
) != 0 ||
707 nvlist_add_uint64(nvroot
, ZPOOL_CONFIG_ID
, 0ULL) != 0 ||
708 nvlist_add_uint64(nvroot
, ZPOOL_CONFIG_GUID
, guid
) != 0 ||
709 nvlist_add_nvlist_array(nvroot
, ZPOOL_CONFIG_CHILDREN
,
710 child
, children
) != 0) {
715 for (c
= 0; c
< children
; c
++)
716 nvlist_free(child
[c
]);
722 * Go through and fix up any paths and/or devids based on our
723 * known list of vdev GUID -> path mappings.
725 if (fix_paths(nvroot
, pl
->names
) != 0) {
731 * Add the root vdev to this pool's configuration.
733 if (nvlist_add_nvlist(config
, ZPOOL_CONFIG_VDEV_TREE
,
741 * zdb uses this path to report on active pools that were
742 * imported or created using -R.
748 * Determine if this pool is currently active, in which case we
749 * can't actually import it.
751 verify(nvlist_lookup_string(config
, ZPOOL_CONFIG_POOL_NAME
,
753 verify(nvlist_lookup_uint64(config
, ZPOOL_CONFIG_POOL_GUID
,
756 if (pool_active(hdl
, name
, guid
, &isactive
) != 0)
765 if ((nvl
= refresh_config(hdl
, config
)) == NULL
) {
775 * Go through and update the paths for spares, now that we have
778 verify(nvlist_lookup_nvlist(config
, ZPOOL_CONFIG_VDEV_TREE
,
780 if (nvlist_lookup_nvlist_array(nvroot
, ZPOOL_CONFIG_SPARES
,
781 &spares
, &nspares
) == 0) {
782 for (i
= 0; i
< nspares
; i
++) {
783 if (fix_paths(spares
[i
], pl
->names
) != 0)
789 * Update the paths for l2cache devices.
791 if (nvlist_lookup_nvlist_array(nvroot
, ZPOOL_CONFIG_L2CACHE
,
792 &l2cache
, &nl2cache
) == 0) {
793 for (i
= 0; i
< nl2cache
; i
++) {
794 if (fix_paths(l2cache
[i
], pl
->names
) != 0)
800 * Restore the original information read from the actual label.
802 (void) nvlist_remove(config
, ZPOOL_CONFIG_HOSTID
,
804 (void) nvlist_remove(config
, ZPOOL_CONFIG_HOSTNAME
,
807 verify(nvlist_add_uint64(config
, ZPOOL_CONFIG_HOSTID
,
809 verify(nvlist_add_string(config
, ZPOOL_CONFIG_HOSTNAME
,
815 * Add this pool to the list of configs.
817 verify(nvlist_lookup_string(config
, ZPOOL_CONFIG_POOL_NAME
,
819 if (nvlist_add_nvlist(ret
, name
, config
) != 0)
835 (void) no_memory(hdl
);
839 for (c
= 0; c
< children
; c
++)
840 nvlist_free(child
[c
]);
847 * Return the offset of the given label.
850 label_offset(uint64_t size
, int l
)
852 ASSERT(P2PHASE_TYPED(size
, sizeof (vdev_label_t
), uint64_t) == 0);
853 return (l
* sizeof (vdev_label_t
) + (l
< VDEV_LABELS
/ 2 ?
854 0 : size
- VDEV_LABELS
* sizeof (vdev_label_t
)));
858 * Given a file descriptor, read the label information and return an nvlist
859 * describing the configuration, if there is one.
862 zpool_read_label(int fd
, nvlist_t
**config
)
864 struct stat64 statbuf
;
867 uint64_t state
, txg
, size
;
871 if (fstat64(fd
, &statbuf
) == -1)
873 size
= P2ALIGN_TYPED(statbuf
.st_size
, sizeof (vdev_label_t
), uint64_t);
875 if ((label
= malloc(sizeof (vdev_label_t
))) == NULL
)
878 for (l
= 0; l
< VDEV_LABELS
; l
++) {
879 if (pread64(fd
, label
, sizeof (vdev_label_t
),
880 label_offset(size
, l
)) != sizeof (vdev_label_t
))
883 if (nvlist_unpack(label
->vl_vdev_phys
.vp_nvlist
,
884 sizeof (label
->vl_vdev_phys
.vp_nvlist
), config
, 0) != 0)
887 if (nvlist_lookup_uint64(*config
, ZPOOL_CONFIG_POOL_STATE
,
888 &state
) != 0 || state
> POOL_STATE_L2CACHE
) {
889 nvlist_free(*config
);
893 if (state
!= POOL_STATE_SPARE
&& state
!= POOL_STATE_L2CACHE
&&
894 (nvlist_lookup_uint64(*config
, ZPOOL_CONFIG_POOL_TXG
,
895 &txg
) != 0 || txg
== 0)) {
896 nvlist_free(*config
);
911 * Use libblkid to quickly search for zfs devices
914 zpool_find_import_blkid(libzfs_handle_t
*hdl
, pool_list_t
*pools
)
917 blkid_dev_iterate iter
;
923 err
= blkid_get_cache(&cache
, NULL
);
925 (void) zfs_error_fmt(hdl
, EZFS_BADCACHE
,
926 dgettext(TEXT_DOMAIN
, "blkid_get_cache() %d"), err
);
930 err
= blkid_probe_all(cache
);
932 (void) zfs_error_fmt(hdl
, EZFS_BADCACHE
,
933 dgettext(TEXT_DOMAIN
, "blkid_probe_all() %d"), err
);
937 iter
= blkid_dev_iterate_begin(cache
);
939 (void) zfs_error_fmt(hdl
, EZFS_BADCACHE
,
940 dgettext(TEXT_DOMAIN
, "blkid_dev_iterate_begin()"));
944 err
= blkid_dev_set_search(iter
, "TYPE", "zfs");
946 (void) zfs_error_fmt(hdl
, EZFS_BADCACHE
,
947 dgettext(TEXT_DOMAIN
, "blkid_dev_set_search() %d"), err
);
951 while (blkid_dev_next(iter
, &dev
) == 0) {
952 devname
= blkid_dev_devname(dev
);
953 if ((fd
= open64(devname
, O_RDONLY
)) < 0)
956 err
= zpool_read_label(fd
, &config
);
960 (void) no_memory(hdl
);
964 if (config
!= NULL
) {
965 err
= add_config(hdl
, pools
, devname
, 0, config
);
972 blkid_dev_iterate_end(iter
);
974 blkid_put_cache(cache
);
978 #endif /* HAVE_LIBBLKID */
981 zpool_default_import_path
[DEFAULT_IMPORT_PATH_SIZE
] = {
982 "/dev/disk/by-vdev", /* Custom rules, use first if they exist */
983 "/dev/mapper", /* Use multipath devices before components */
984 "/dev/disk/by-uuid", /* Single unique entry and persistent */
985 "/dev/disk/by-id", /* May be multiple entries and persistent */
986 "/dev/disk/by-path", /* Encodes physical location and persistent */
987 "/dev/disk/by-label", /* Custom persistent labels */
988 "/dev" /* UNSAFE device names will change */
992 * Given a list of directories to search, find all pools stored on disk. This
993 * includes partial pools which are not available to import. If no args are
994 * given (argc is 0), then the default directory (/dev/dsk) is searched.
995 * poolname or guid (but not both) are provided by the caller when trying
996 * to import a specific pool.
999 zpool_find_import_impl(libzfs_handle_t
*hdl
, importargs_t
*iarg
)
1001 int i
, dirs
= iarg
->paths
;
1003 struct dirent64
*dp
;
1004 char path
[MAXPATHLEN
];
1005 char *end
, **dir
= iarg
->path
;
1007 struct stat64 statbuf
;
1008 nvlist_t
*ret
= NULL
, *config
;
1010 pool_list_t pools
= { 0 };
1011 pool_entry_t
*pe
, *penext
;
1012 vdev_entry_t
*ve
, *venext
;
1013 config_entry_t
*ce
, *cenext
;
1014 name_entry_t
*ne
, *nenext
;
1016 verify(iarg
->poolname
== NULL
|| iarg
->guid
== 0);
1019 #ifdef HAVE_LIBBLKID
1020 /* Use libblkid to scan all device for their type */
1021 if (zpool_find_import_blkid(hdl
, &pools
) == 0)
1024 (void) zfs_error_fmt(hdl
, EZFS_BADCACHE
,
1025 dgettext(TEXT_DOMAIN
, "blkid failure falling back "
1026 "to manual probing"));
1027 #endif /* HAVE_LIBBLKID */
1029 dir
= zpool_default_import_path
;
1030 dirs
= DEFAULT_IMPORT_PATH_SIZE
;
1034 * Go through and read the label configuration information from every
1035 * possible device, organizing the information according to pool GUID
1036 * and toplevel GUID.
1038 for (i
= 0; i
< dirs
; i
++) {
1042 /* use realpath to normalize the path */
1043 if (realpath(dir
[i
], path
) == 0) {
1045 /* it is safe to skip missing search paths */
1046 if (errno
== ENOENT
)
1049 zfs_error_aux(hdl
, strerror(errno
));
1050 (void) zfs_error_fmt(hdl
, EZFS_BADPATH
,
1051 dgettext(TEXT_DOMAIN
, "cannot open '%s'"), dir
[i
]);
1054 end
= &path
[strlen(path
)];
1057 pathleft
= &path
[sizeof (path
)] - end
;
1060 * Using raw devices instead of block devices when we're
1061 * reading the labels skips a bunch of slow operations during
1062 * close(2) processing, so we replace /dev/dsk with /dev/rdsk.
1064 if (strcmp(path
, "/dev/dsk/") == 0)
1065 rdsk
= "/dev/rdsk/";
1069 if ((dfd
= open64(rdsk
, O_RDONLY
)) < 0 ||
1070 (dirp
= fdopendir(dfd
)) == NULL
) {
1071 zfs_error_aux(hdl
, strerror(errno
));
1072 (void) zfs_error_fmt(hdl
, EZFS_BADPATH
,
1073 dgettext(TEXT_DOMAIN
, "cannot open '%s'"),
1079 * This is not MT-safe, but we have no MT consumers of libzfs
1081 while ((dp
= readdir64(dirp
)) != NULL
) {
1082 const char *name
= dp
->d_name
;
1083 if (name
[0] == '.' &&
1084 (name
[1] == 0 || (name
[1] == '.' && name
[2] == 0)))
1088 * Skip checking devices with well known prefixes:
1089 * watchdog - A special close is required to avoid
1090 * triggering it and resetting the system.
1091 * fuse - Fuse control device.
1092 * ppp - Generic PPP driver.
1093 * tty* - Generic serial interface.
1094 * vcs* - Virtual console memory.
1095 * parport* - Parallel port interface.
1096 * lp* - Printer interface.
1097 * fd* - Floppy interface.
1098 * hpet - High Precision Event Timer, crashes qemu
1099 * when accessed from a virtual machine.
1100 * core - Symlink to /proc/kcore, causes a crash
1101 * when access from Xen dom0.
1103 if ((strncmp(name
, "watchdog", 8) == 0) ||
1104 (strncmp(name
, "fuse", 4) == 0) ||
1105 (strncmp(name
, "ppp", 3) == 0) ||
1106 (strncmp(name
, "tty", 3) == 0) ||
1107 (strncmp(name
, "vcs", 3) == 0) ||
1108 (strncmp(name
, "parport", 7) == 0) ||
1109 (strncmp(name
, "lp", 2) == 0) ||
1110 (strncmp(name
, "fd", 2) == 0) ||
1111 (strncmp(name
, "hpet", 4) == 0) ||
1112 (strncmp(name
, "core", 4) == 0))
1116 * Ignore failed stats. We only want regular
1117 * files and block devices.
1119 if ((fstatat64(dfd
, name
, &statbuf
, 0) != 0) ||
1120 (!S_ISREG(statbuf
.st_mode
) &&
1121 !S_ISBLK(statbuf
.st_mode
)))
1124 if ((fd
= openat64(dfd
, name
, O_RDONLY
)) < 0)
1127 if ((zpool_read_label(fd
, &config
)) != 0) {
1129 (void) no_memory(hdl
);
1135 if (config
!= NULL
) {
1136 boolean_t matched
= B_TRUE
;
1139 if ((iarg
->poolname
!= NULL
) &&
1140 (nvlist_lookup_string(config
,
1141 ZPOOL_CONFIG_POOL_NAME
, &pname
) == 0)) {
1143 if (strcmp(iarg
->poolname
, pname
))
1146 } else if (iarg
->guid
!= 0) {
1149 matched
= nvlist_lookup_uint64(config
,
1150 ZPOOL_CONFIG_POOL_GUID
,
1152 iarg
->guid
== this_guid
;
1155 nvlist_free(config
);
1159 /* use the non-raw path for the config */
1160 (void) strlcpy(end
, name
, pathleft
);
1161 if (add_config(hdl
, &pools
, path
, i
+1, config
))
1166 (void) closedir(dirp
);
1170 #ifdef HAVE_LIBBLKID
1173 ret
= get_configs(hdl
, &pools
, iarg
->can_be_active
);
1176 for (pe
= pools
.pools
; pe
!= NULL
; pe
= penext
) {
1177 penext
= pe
->pe_next
;
1178 for (ve
= pe
->pe_vdevs
; ve
!= NULL
; ve
= venext
) {
1179 venext
= ve
->ve_next
;
1180 for (ce
= ve
->ve_configs
; ce
!= NULL
; ce
= cenext
) {
1181 cenext
= ce
->ce_next
;
1183 nvlist_free(ce
->ce_config
);
1191 for (ne
= pools
.names
; ne
!= NULL
; ne
= nenext
) {
1192 nenext
= ne
->ne_next
;
1199 (void) closedir(dirp
);
1205 zpool_find_import(libzfs_handle_t
*hdl
, int argc
, char **argv
)
1207 importargs_t iarg
= { 0 };
1212 return (zpool_find_import_impl(hdl
, &iarg
));
1216 * Given a cache file, return the contents as a list of importable pools.
1217 * poolname or guid (but not both) are provided by the caller when trying
1218 * to import a specific pool.
1221 zpool_find_import_cached(libzfs_handle_t
*hdl
, const char *cachefile
,
1222 char *poolname
, uint64_t guid
)
1226 struct stat64 statbuf
;
1227 nvlist_t
*raw
, *src
, *dst
;
1234 verify(poolname
== NULL
|| guid
== 0);
1236 if ((fd
= open(cachefile
, O_RDONLY
)) < 0) {
1237 zfs_error_aux(hdl
, "%s", strerror(errno
));
1238 (void) zfs_error(hdl
, EZFS_BADCACHE
,
1239 dgettext(TEXT_DOMAIN
, "failed to open cache file"));
1243 if (fstat64(fd
, &statbuf
) != 0) {
1244 zfs_error_aux(hdl
, "%s", strerror(errno
));
1246 (void) zfs_error(hdl
, EZFS_BADCACHE
,
1247 dgettext(TEXT_DOMAIN
, "failed to get size of cache file"));
1251 if ((buf
= zfs_alloc(hdl
, statbuf
.st_size
)) == NULL
) {
1256 if (read(fd
, buf
, statbuf
.st_size
) != statbuf
.st_size
) {
1259 (void) zfs_error(hdl
, EZFS_BADCACHE
,
1260 dgettext(TEXT_DOMAIN
,
1261 "failed to read cache file contents"));
1267 if (nvlist_unpack(buf
, statbuf
.st_size
, &raw
, 0) != 0) {
1269 (void) zfs_error(hdl
, EZFS_BADCACHE
,
1270 dgettext(TEXT_DOMAIN
,
1271 "invalid or corrupt cache file contents"));
1278 * Go through and get the current state of the pools and refresh their
1281 if (nvlist_alloc(&pools
, 0, 0) != 0) {
1282 (void) no_memory(hdl
);
1288 while ((elem
= nvlist_next_nvpair(raw
, elem
)) != NULL
) {
1289 verify(nvpair_value_nvlist(elem
, &src
) == 0);
1291 verify(nvlist_lookup_string(src
, ZPOOL_CONFIG_POOL_NAME
,
1293 if (poolname
!= NULL
&& strcmp(poolname
, name
) != 0)
1296 verify(nvlist_lookup_uint64(src
, ZPOOL_CONFIG_POOL_GUID
,
1299 verify(nvlist_lookup_uint64(src
, ZPOOL_CONFIG_POOL_GUID
,
1301 if (guid
!= this_guid
)
1305 if (pool_active(hdl
, name
, this_guid
, &active
) != 0) {
1314 if ((dst
= refresh_config(hdl
, src
)) == NULL
) {
1320 if (nvlist_add_nvlist(pools
, nvpair_name(elem
), dst
) != 0) {
1321 (void) no_memory(hdl
);
1335 name_or_guid_exists(zpool_handle_t
*zhp
, void *data
)
1337 importargs_t
*import
= data
;
1340 if (import
->poolname
!= NULL
) {
1343 verify(nvlist_lookup_string(zhp
->zpool_config
,
1344 ZPOOL_CONFIG_POOL_NAME
, &pool_name
) == 0);
1345 if (strcmp(pool_name
, import
->poolname
) == 0)
1350 verify(nvlist_lookup_uint64(zhp
->zpool_config
,
1351 ZPOOL_CONFIG_POOL_GUID
, &pool_guid
) == 0);
1352 if (pool_guid
== import
->guid
)
1361 zpool_search_import(libzfs_handle_t
*hdl
, importargs_t
*import
)
1363 verify(import
->poolname
== NULL
|| import
->guid
== 0);
1366 import
->exists
= zpool_iter(hdl
, name_or_guid_exists
, import
);
1368 if (import
->cachefile
!= NULL
)
1369 return (zpool_find_import_cached(hdl
, import
->cachefile
,
1370 import
->poolname
, import
->guid
));
1372 return (zpool_find_import_impl(hdl
, import
));
1376 find_guid(nvlist_t
*nv
, uint64_t guid
)
1382 verify(nvlist_lookup_uint64(nv
, ZPOOL_CONFIG_GUID
, &tmp
) == 0);
1386 if (nvlist_lookup_nvlist_array(nv
, ZPOOL_CONFIG_CHILDREN
,
1387 &child
, &children
) == 0) {
1388 for (c
= 0; c
< children
; c
++)
1389 if (find_guid(child
[c
], guid
))
1396 typedef struct aux_cbdata
{
1397 const char *cb_type
;
1399 zpool_handle_t
*cb_zhp
;
1403 find_aux(zpool_handle_t
*zhp
, void *data
)
1405 aux_cbdata_t
*cbp
= data
;
1411 verify(nvlist_lookup_nvlist(zhp
->zpool_config
, ZPOOL_CONFIG_VDEV_TREE
,
1414 if (nvlist_lookup_nvlist_array(nvroot
, cbp
->cb_type
,
1415 &list
, &count
) == 0) {
1416 for (i
= 0; i
< count
; i
++) {
1417 verify(nvlist_lookup_uint64(list
[i
],
1418 ZPOOL_CONFIG_GUID
, &guid
) == 0);
1419 if (guid
== cbp
->cb_guid
) {
1431 * Determines if the pool is in use. If so, it returns true and the state of
1432 * the pool as well as the name of the pool. Both strings are allocated and
1433 * must be freed by the caller.
1436 zpool_in_use(libzfs_handle_t
*hdl
, int fd
, pool_state_t
*state
, char **namestr
,
1442 uint64_t guid
, vdev_guid
;
1443 zpool_handle_t
*zhp
;
1444 nvlist_t
*pool_config
;
1445 uint64_t stateval
, isspare
;
1446 aux_cbdata_t cb
= { 0 };
1451 if (zpool_read_label(fd
, &config
) != 0) {
1452 (void) no_memory(hdl
);
1459 verify(nvlist_lookup_uint64(config
, ZPOOL_CONFIG_POOL_STATE
,
1461 verify(nvlist_lookup_uint64(config
, ZPOOL_CONFIG_GUID
,
1464 if (stateval
!= POOL_STATE_SPARE
&& stateval
!= POOL_STATE_L2CACHE
) {
1465 verify(nvlist_lookup_string(config
, ZPOOL_CONFIG_POOL_NAME
,
1467 verify(nvlist_lookup_uint64(config
, ZPOOL_CONFIG_POOL_GUID
,
1472 case POOL_STATE_EXPORTED
:
1474 * A pool with an exported state may in fact be imported
1475 * read-only, so check the in-core state to see if it's
1476 * active and imported read-only. If it is, set
1477 * its state to active.
1479 if (pool_active(hdl
, name
, guid
, &isactive
) == 0 && isactive
&&
1480 (zhp
= zpool_open_canfail(hdl
, name
)) != NULL
&&
1481 zpool_get_prop_int(zhp
, ZPOOL_PROP_READONLY
, NULL
))
1482 stateval
= POOL_STATE_ACTIVE
;
1487 case POOL_STATE_ACTIVE
:
1489 * For an active pool, we have to determine if it's really part
1490 * of a currently active pool (in which case the pool will exist
1491 * and the guid will be the same), or whether it's part of an
1492 * active pool that was disconnected without being explicitly
1495 if (pool_active(hdl
, name
, guid
, &isactive
) != 0) {
1496 nvlist_free(config
);
1502 * Because the device may have been removed while
1503 * offlined, we only report it as active if the vdev is
1504 * still present in the config. Otherwise, pretend like
1507 if ((zhp
= zpool_open_canfail(hdl
, name
)) != NULL
&&
1508 (pool_config
= zpool_get_config(zhp
, NULL
))
1512 verify(nvlist_lookup_nvlist(pool_config
,
1513 ZPOOL_CONFIG_VDEV_TREE
, &nvroot
) == 0);
1514 ret
= find_guid(nvroot
, vdev_guid
);
1520 * If this is an active spare within another pool, we
1521 * treat it like an unused hot spare. This allows the
1522 * user to create a pool with a hot spare that currently
1523 * in use within another pool. Since we return B_TRUE,
1524 * libdiskmgt will continue to prevent generic consumers
1525 * from using the device.
1527 if (ret
&& nvlist_lookup_uint64(config
,
1528 ZPOOL_CONFIG_IS_SPARE
, &isspare
) == 0 && isspare
)
1529 stateval
= POOL_STATE_SPARE
;
1534 stateval
= POOL_STATE_POTENTIALLY_ACTIVE
;
1539 case POOL_STATE_SPARE
:
1541 * For a hot spare, it can be either definitively in use, or
1542 * potentially active. To determine if it's in use, we iterate
1543 * over all pools in the system and search for one with a spare
1544 * with a matching guid.
1546 * Due to the shared nature of spares, we don't actually report
1547 * the potentially active case as in use. This means the user
1548 * can freely create pools on the hot spares of exported pools,
1549 * but to do otherwise makes the resulting code complicated, and
1550 * we end up having to deal with this case anyway.
1553 cb
.cb_guid
= vdev_guid
;
1554 cb
.cb_type
= ZPOOL_CONFIG_SPARES
;
1555 if (zpool_iter(hdl
, find_aux
, &cb
) == 1) {
1556 name
= (char *)zpool_get_name(cb
.cb_zhp
);
1563 case POOL_STATE_L2CACHE
:
1566 * Check if any pool is currently using this l2cache device.
1569 cb
.cb_guid
= vdev_guid
;
1570 cb
.cb_type
= ZPOOL_CONFIG_L2CACHE
;
1571 if (zpool_iter(hdl
, find_aux
, &cb
) == 1) {
1572 name
= (char *)zpool_get_name(cb
.cb_zhp
);
1585 if ((*namestr
= zfs_strdup(hdl
, name
)) == NULL
) {
1587 zpool_close(cb
.cb_zhp
);
1588 nvlist_free(config
);
1591 *state
= (pool_state_t
)stateval
;
1595 zpool_close(cb
.cb_zhp
);
1597 nvlist_free(config
);