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 2015 Nexenta Systems, Inc. All rights reserved.
24 * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
25 * Copyright (c) 2013 by Delphix. All rights reserved.
29 * Pool import support functions.
31 * To import a pool, we rely on reading the configuration information from the
32 * ZFS label of each device. If we successfully read the label, then we
33 * organize the configuration information in the following hierarchy:
35 * pool guid -> toplevel vdev guid -> label txg
37 * Duplicate entries matching this same tuple will be discarded. Once we have
38 * examined every device, we pick the best label txg config for each toplevel
39 * vdev. We then arrange these toplevel vdevs into a complete pool config, and
40 * update any paths that have changed. Finally, we attempt to import the pool
41 * using our derived config, and record the results.
56 #include <sys/dktp/fdisk.h>
57 #include <sys/efi_partition.h>
58 #include <thread_pool.h>
60 #include <sys/vdev_impl.h>
63 #include "libzfs_impl.h"
66 * Intermediate structures used to gather configuration information.
68 typedef struct config_entry
{
71 struct config_entry
*ce_next
;
74 typedef struct vdev_entry
{
76 config_entry_t
*ve_configs
;
77 struct vdev_entry
*ve_next
;
80 typedef struct pool_entry
{
82 vdev_entry_t
*pe_vdevs
;
83 struct pool_entry
*pe_next
;
86 typedef struct name_entry
{
89 struct name_entry
*ne_next
;
92 typedef struct pool_list
{
98 get_devid(const char *path
)
104 if ((fd
= open(path
, O_RDONLY
)) < 0)
109 if (devid_get(fd
, &devid
) == 0) {
110 if (devid_get_minor_name(fd
, &minor
) == 0)
111 ret
= devid_str_encode(devid
, minor
);
113 devid_str_free(minor
);
123 * Go through and fix up any path and/or devid information for the given vdev
127 fix_paths(nvlist_t
*nv
, name_entry_t
*names
)
132 name_entry_t
*ne
, *best
;
136 if (nvlist_lookup_nvlist_array(nv
, ZPOOL_CONFIG_CHILDREN
,
137 &child
, &children
) == 0) {
138 for (c
= 0; c
< children
; c
++)
139 if (fix_paths(child
[c
], names
) != 0)
145 * This is a leaf (file or disk) vdev. In either case, go through
146 * the name list and see if we find a matching guid. If so, replace
147 * the path and see if we can calculate a new devid.
149 * There may be multiple names associated with a particular guid, in
150 * which case we have overlapping slices or multiple paths to the same
151 * disk. If this is the case, then we want to pick the path that is
152 * the most similar to the original, where "most similar" is the number
153 * of matching characters starting from the end of the path. This will
154 * preserve slice numbers even if the disks have been reorganized, and
155 * will also catch preferred disk names if multiple paths exist.
157 verify(nvlist_lookup_uint64(nv
, ZPOOL_CONFIG_GUID
, &guid
) == 0);
158 if (nvlist_lookup_string(nv
, ZPOOL_CONFIG_PATH
, &path
) != 0)
163 for (ne
= names
; ne
!= NULL
; ne
= ne
->ne_next
) {
164 if (ne
->ne_guid
== guid
) {
165 const char *src
, *dst
;
173 src
= ne
->ne_name
+ strlen(ne
->ne_name
) - 1;
174 dst
= path
+ strlen(path
) - 1;
175 for (count
= 0; src
>= ne
->ne_name
&& dst
>= path
;
176 src
--, dst
--, count
++)
181 * At this point, 'count' is the number of characters
182 * matched from the end.
184 if (count
> matched
|| best
== NULL
) {
194 if (nvlist_add_string(nv
, ZPOOL_CONFIG_PATH
, best
->ne_name
) != 0)
197 if ((devid
= get_devid(best
->ne_name
)) == NULL
) {
198 (void) nvlist_remove_all(nv
, ZPOOL_CONFIG_DEVID
);
200 if (nvlist_add_string(nv
, ZPOOL_CONFIG_DEVID
, devid
) != 0) {
201 devid_str_free(devid
);
204 devid_str_free(devid
);
211 * Add the given configuration to the list of known devices.
214 add_config(libzfs_handle_t
*hdl
, pool_list_t
*pl
, const char *path
,
217 uint64_t pool_guid
, vdev_guid
, top_guid
, txg
, state
;
224 * If this is a hot spare not currently in use or level 2 cache
225 * device, add it to the list of names to translate, but don't do
228 if (nvlist_lookup_uint64(config
, ZPOOL_CONFIG_POOL_STATE
,
230 (state
== POOL_STATE_SPARE
|| state
== POOL_STATE_L2CACHE
) &&
231 nvlist_lookup_uint64(config
, ZPOOL_CONFIG_GUID
, &vdev_guid
) == 0) {
232 if ((ne
= zfs_alloc(hdl
, sizeof (name_entry_t
))) == NULL
)
235 if ((ne
->ne_name
= zfs_strdup(hdl
, path
)) == NULL
) {
239 ne
->ne_guid
= vdev_guid
;
240 ne
->ne_next
= pl
->names
;
246 * If we have a valid config but cannot read any of these fields, then
247 * it means we have a half-initialized label. In vdev_label_init()
248 * we write a label with txg == 0 so that we can identify the device
249 * in case the user refers to the same disk later on. If we fail to
250 * create the pool, we'll be left with a label in this state
251 * which should not be considered part of a valid pool.
253 if (nvlist_lookup_uint64(config
, ZPOOL_CONFIG_POOL_GUID
,
255 nvlist_lookup_uint64(config
, ZPOOL_CONFIG_GUID
,
257 nvlist_lookup_uint64(config
, ZPOOL_CONFIG_TOP_GUID
,
259 nvlist_lookup_uint64(config
, ZPOOL_CONFIG_POOL_TXG
,
260 &txg
) != 0 || txg
== 0) {
266 * First, see if we know about this pool. If not, then add it to the
267 * list of known pools.
269 for (pe
= pl
->pools
; pe
!= NULL
; pe
= pe
->pe_next
) {
270 if (pe
->pe_guid
== pool_guid
)
275 if ((pe
= zfs_alloc(hdl
, sizeof (pool_entry_t
))) == NULL
) {
279 pe
->pe_guid
= pool_guid
;
280 pe
->pe_next
= pl
->pools
;
285 * Second, see if we know about this toplevel vdev. Add it if its
288 for (ve
= pe
->pe_vdevs
; ve
!= NULL
; ve
= ve
->ve_next
) {
289 if (ve
->ve_guid
== top_guid
)
294 if ((ve
= zfs_alloc(hdl
, sizeof (vdev_entry_t
))) == NULL
) {
298 ve
->ve_guid
= top_guid
;
299 ve
->ve_next
= pe
->pe_vdevs
;
304 * Third, see if we have a config with a matching transaction group. If
305 * so, then we do nothing. Otherwise, add it to the list of known
308 for (ce
= ve
->ve_configs
; ce
!= NULL
; ce
= ce
->ce_next
) {
309 if (ce
->ce_txg
== txg
)
314 if ((ce
= zfs_alloc(hdl
, sizeof (config_entry_t
))) == NULL
) {
319 ce
->ce_config
= config
;
320 ce
->ce_next
= ve
->ve_configs
;
327 * At this point we've successfully added our config to the list of
328 * known configs. The last thing to do is add the vdev guid -> path
329 * mappings so that we can fix up the configuration as necessary before
332 if ((ne
= zfs_alloc(hdl
, sizeof (name_entry_t
))) == NULL
)
335 if ((ne
->ne_name
= zfs_strdup(hdl
, path
)) == NULL
) {
340 ne
->ne_guid
= vdev_guid
;
341 ne
->ne_next
= pl
->names
;
348 * Returns true if the named pool matches the given GUID.
351 pool_active(libzfs_handle_t
*hdl
, const char *name
, uint64_t guid
,
357 if (zpool_open_silent(hdl
, name
, &zhp
) != 0)
365 verify(nvlist_lookup_uint64(zhp
->zpool_config
, ZPOOL_CONFIG_POOL_GUID
,
370 *isactive
= (theguid
== guid
);
375 refresh_config(libzfs_handle_t
*hdl
, nvlist_t
*config
)
378 zfs_cmd_t zc
= { 0 };
381 if (zcmd_write_conf_nvlist(hdl
, &zc
, config
) != 0)
384 if (zcmd_alloc_dst_nvlist(hdl
, &zc
,
385 zc
.zc_nvlist_conf_size
* 2) != 0) {
386 zcmd_free_nvlists(&zc
);
390 while ((err
= ioctl(hdl
->libzfs_fd
, ZFS_IOC_POOL_TRYIMPORT
,
391 &zc
)) != 0 && errno
== ENOMEM
) {
392 if (zcmd_expand_dst_nvlist(hdl
, &zc
) != 0) {
393 zcmd_free_nvlists(&zc
);
399 zcmd_free_nvlists(&zc
);
403 if (zcmd_read_dst_nvlist(hdl
, &zc
, &nvl
) != 0) {
404 zcmd_free_nvlists(&zc
);
408 zcmd_free_nvlists(&zc
);
413 * Determine if the vdev id is a hole in the namespace.
416 vdev_is_hole(uint64_t *hole_array
, uint_t holes
, uint_t id
)
418 for (int c
= 0; c
< holes
; c
++) {
420 /* Top-level is a hole */
421 if (hole_array
[c
] == id
)
428 * Convert our list of pools into the definitive set of configurations. We
429 * start by picking the best config for each toplevel vdev. Once that's done,
430 * we assemble the toplevel vdevs into a full config for the pool. We make a
431 * pass to fix up any incorrect paths, and then add it to the main list to
432 * return to the user.
435 get_configs(libzfs_handle_t
*hdl
, pool_list_t
*pl
, boolean_t active_ok
)
440 nvlist_t
*ret
= NULL
, *config
= NULL
, *tmp
, *nvtop
, *nvroot
;
441 nvlist_t
**spares
, **l2cache
;
442 uint_t i
, nspares
, nl2cache
;
443 boolean_t config_seen
;
445 char *name
, *hostname
;
448 nvlist_t
**child
= NULL
;
450 uint64_t *hole_array
, max_id
;
455 boolean_t found_one
= B_FALSE
;
456 boolean_t valid_top_config
= B_FALSE
;
458 if (nvlist_alloc(&ret
, 0, 0) != 0)
461 for (pe
= pl
->pools
; pe
!= NULL
; pe
= pe
->pe_next
) {
462 uint64_t id
, max_txg
= 0;
464 if (nvlist_alloc(&config
, NV_UNIQUE_NAME
, 0) != 0)
466 config_seen
= B_FALSE
;
469 * Iterate over all toplevel vdevs. Grab the pool configuration
470 * from the first one we find, and then go through the rest and
471 * add them as necessary to the 'vdevs' member of the config.
473 for (ve
= pe
->pe_vdevs
; ve
!= NULL
; ve
= ve
->ve_next
) {
476 * Determine the best configuration for this vdev by
477 * selecting the config with the latest transaction
481 for (ce
= ve
->ve_configs
; ce
!= NULL
;
484 if (ce
->ce_txg
> best_txg
) {
486 best_txg
= ce
->ce_txg
;
491 * We rely on the fact that the max txg for the
492 * pool will contain the most up-to-date information
493 * about the valid top-levels in the vdev namespace.
495 if (best_txg
> max_txg
) {
496 (void) nvlist_remove(config
,
497 ZPOOL_CONFIG_VDEV_CHILDREN
,
499 (void) nvlist_remove(config
,
500 ZPOOL_CONFIG_HOLE_ARRAY
,
501 DATA_TYPE_UINT64_ARRAY
);
507 valid_top_config
= B_FALSE
;
509 if (nvlist_lookup_uint64(tmp
,
510 ZPOOL_CONFIG_VDEV_CHILDREN
, &max_id
) == 0) {
511 verify(nvlist_add_uint64(config
,
512 ZPOOL_CONFIG_VDEV_CHILDREN
,
514 valid_top_config
= B_TRUE
;
517 if (nvlist_lookup_uint64_array(tmp
,
518 ZPOOL_CONFIG_HOLE_ARRAY
, &hole_array
,
520 verify(nvlist_add_uint64_array(config
,
521 ZPOOL_CONFIG_HOLE_ARRAY
,
522 hole_array
, holes
) == 0);
528 * Copy the relevant pieces of data to the pool
534 * comment (if available)
536 * hostid (if available)
537 * hostname (if available)
539 uint64_t state
, version
;
540 char *comment
= NULL
;
542 version
= fnvlist_lookup_uint64(tmp
,
543 ZPOOL_CONFIG_VERSION
);
544 fnvlist_add_uint64(config
,
545 ZPOOL_CONFIG_VERSION
, version
);
546 guid
= fnvlist_lookup_uint64(tmp
,
547 ZPOOL_CONFIG_POOL_GUID
);
548 fnvlist_add_uint64(config
,
549 ZPOOL_CONFIG_POOL_GUID
, guid
);
550 name
= fnvlist_lookup_string(tmp
,
551 ZPOOL_CONFIG_POOL_NAME
);
552 fnvlist_add_string(config
,
553 ZPOOL_CONFIG_POOL_NAME
, name
);
555 if (nvlist_lookup_string(tmp
,
556 ZPOOL_CONFIG_COMMENT
, &comment
) == 0)
557 fnvlist_add_string(config
,
558 ZPOOL_CONFIG_COMMENT
, comment
);
560 state
= fnvlist_lookup_uint64(tmp
,
561 ZPOOL_CONFIG_POOL_STATE
);
562 fnvlist_add_uint64(config
,
563 ZPOOL_CONFIG_POOL_STATE
, state
);
566 if (nvlist_lookup_uint64(tmp
,
567 ZPOOL_CONFIG_HOSTID
, &hostid
) == 0) {
568 fnvlist_add_uint64(config
,
569 ZPOOL_CONFIG_HOSTID
, hostid
);
570 hostname
= fnvlist_lookup_string(tmp
,
571 ZPOOL_CONFIG_HOSTNAME
);
572 fnvlist_add_string(config
,
573 ZPOOL_CONFIG_HOSTNAME
, hostname
);
576 config_seen
= B_TRUE
;
580 * Add this top-level vdev to the child array.
582 verify(nvlist_lookup_nvlist(tmp
,
583 ZPOOL_CONFIG_VDEV_TREE
, &nvtop
) == 0);
584 verify(nvlist_lookup_uint64(nvtop
, ZPOOL_CONFIG_ID
,
587 if (id
>= children
) {
590 newchild
= zfs_alloc(hdl
, (id
+ 1) *
591 sizeof (nvlist_t
*));
592 if (newchild
== NULL
)
595 for (c
= 0; c
< children
; c
++)
596 newchild
[c
] = child
[c
];
602 if (nvlist_dup(nvtop
, &child
[id
], 0) != 0)
608 * If we have information about all the top-levels then
609 * clean up the nvlist which we've constructed. This
610 * means removing any extraneous devices that are
611 * beyond the valid range or adding devices to the end
612 * of our array which appear to be missing.
614 if (valid_top_config
) {
615 if (max_id
< children
) {
616 for (c
= max_id
; c
< children
; c
++)
617 nvlist_free(child
[c
]);
619 } else if (max_id
> children
) {
622 newchild
= zfs_alloc(hdl
, (max_id
) *
623 sizeof (nvlist_t
*));
624 if (newchild
== NULL
)
627 for (c
= 0; c
< children
; c
++)
628 newchild
[c
] = child
[c
];
636 verify(nvlist_lookup_uint64(config
, ZPOOL_CONFIG_POOL_GUID
,
640 * The vdev namespace may contain holes as a result of
641 * device removal. We must add them back into the vdev
642 * tree before we process any missing devices.
645 ASSERT(valid_top_config
);
647 for (c
= 0; c
< children
; c
++) {
650 if (child
[c
] != NULL
||
651 !vdev_is_hole(hole_array
, holes
, c
))
654 if (nvlist_alloc(&holey
, NV_UNIQUE_NAME
,
659 * Holes in the namespace are treated as
660 * "hole" top-level vdevs and have a
661 * special flag set on them.
663 if (nvlist_add_string(holey
,
665 VDEV_TYPE_HOLE
) != 0 ||
666 nvlist_add_uint64(holey
,
667 ZPOOL_CONFIG_ID
, c
) != 0 ||
668 nvlist_add_uint64(holey
,
669 ZPOOL_CONFIG_GUID
, 0ULL) != 0) {
678 * Look for any missing top-level vdevs. If this is the case,
679 * create a faked up 'missing' vdev as a placeholder. We cannot
680 * simply compress the child array, because the kernel performs
681 * certain checks to make sure the vdev IDs match their location
682 * in the configuration.
684 for (c
= 0; c
< children
; c
++) {
685 if (child
[c
] == NULL
) {
687 if (nvlist_alloc(&missing
, NV_UNIQUE_NAME
,
690 if (nvlist_add_string(missing
,
692 VDEV_TYPE_MISSING
) != 0 ||
693 nvlist_add_uint64(missing
,
694 ZPOOL_CONFIG_ID
, c
) != 0 ||
695 nvlist_add_uint64(missing
,
696 ZPOOL_CONFIG_GUID
, 0ULL) != 0) {
697 nvlist_free(missing
);
705 * Put all of this pool's top-level vdevs into a root vdev.
707 if (nvlist_alloc(&nvroot
, NV_UNIQUE_NAME
, 0) != 0)
709 if (nvlist_add_string(nvroot
, ZPOOL_CONFIG_TYPE
,
710 VDEV_TYPE_ROOT
) != 0 ||
711 nvlist_add_uint64(nvroot
, ZPOOL_CONFIG_ID
, 0ULL) != 0 ||
712 nvlist_add_uint64(nvroot
, ZPOOL_CONFIG_GUID
, guid
) != 0 ||
713 nvlist_add_nvlist_array(nvroot
, ZPOOL_CONFIG_CHILDREN
,
714 child
, children
) != 0) {
719 for (c
= 0; c
< children
; c
++)
720 nvlist_free(child
[c
]);
726 * Go through and fix up any paths and/or devids based on our
727 * known list of vdev GUID -> path mappings.
729 if (fix_paths(nvroot
, pl
->names
) != 0) {
735 * Add the root vdev to this pool's configuration.
737 if (nvlist_add_nvlist(config
, ZPOOL_CONFIG_VDEV_TREE
,
745 * zdb uses this path to report on active pools that were
746 * imported or created using -R.
752 * Determine if this pool is currently active, in which case we
753 * can't actually import it.
755 verify(nvlist_lookup_string(config
, ZPOOL_CONFIG_POOL_NAME
,
757 verify(nvlist_lookup_uint64(config
, ZPOOL_CONFIG_POOL_GUID
,
760 if (pool_active(hdl
, name
, guid
, &isactive
) != 0)
769 if ((nvl
= refresh_config(hdl
, config
)) == NULL
) {
779 * Go through and update the paths for spares, now that we have
782 verify(nvlist_lookup_nvlist(config
, ZPOOL_CONFIG_VDEV_TREE
,
784 if (nvlist_lookup_nvlist_array(nvroot
, ZPOOL_CONFIG_SPARES
,
785 &spares
, &nspares
) == 0) {
786 for (i
= 0; i
< nspares
; i
++) {
787 if (fix_paths(spares
[i
], pl
->names
) != 0)
793 * Update the paths for l2cache devices.
795 if (nvlist_lookup_nvlist_array(nvroot
, ZPOOL_CONFIG_L2CACHE
,
796 &l2cache
, &nl2cache
) == 0) {
797 for (i
= 0; i
< nl2cache
; i
++) {
798 if (fix_paths(l2cache
[i
], pl
->names
) != 0)
804 * Restore the original information read from the actual label.
806 (void) nvlist_remove(config
, ZPOOL_CONFIG_HOSTID
,
808 (void) nvlist_remove(config
, ZPOOL_CONFIG_HOSTNAME
,
811 verify(nvlist_add_uint64(config
, ZPOOL_CONFIG_HOSTID
,
813 verify(nvlist_add_string(config
, ZPOOL_CONFIG_HOSTNAME
,
819 * Add this pool to the list of configs.
821 verify(nvlist_lookup_string(config
, ZPOOL_CONFIG_POOL_NAME
,
823 if (nvlist_add_nvlist(ret
, name
, config
) != 0)
839 (void) no_memory(hdl
);
843 for (c
= 0; c
< children
; c
++)
844 nvlist_free(child
[c
]);
851 * Return the offset of the given label.
854 label_offset(uint64_t size
, int l
)
856 ASSERT(P2PHASE_TYPED(size
, sizeof (vdev_label_t
), uint64_t) == 0);
857 return (l
* sizeof (vdev_label_t
) + (l
< VDEV_LABELS
/ 2 ?
858 0 : size
- VDEV_LABELS
* sizeof (vdev_label_t
)));
862 * Given a file descriptor, read the label information and return an nvlist
863 * describing the configuration, if there is one.
866 zpool_read_label(int fd
, nvlist_t
**config
)
868 struct stat64 statbuf
;
871 uint64_t state
, txg
, size
;
875 if (fstat64(fd
, &statbuf
) == -1)
877 size
= P2ALIGN_TYPED(statbuf
.st_size
, sizeof (vdev_label_t
), uint64_t);
879 if ((label
= malloc(sizeof (vdev_label_t
))) == NULL
)
882 for (l
= 0; l
< VDEV_LABELS
; l
++) {
883 if (pread64(fd
, label
, sizeof (vdev_label_t
),
884 label_offset(size
, l
)) != sizeof (vdev_label_t
))
887 if (nvlist_unpack(label
->vl_vdev_phys
.vp_nvlist
,
888 sizeof (label
->vl_vdev_phys
.vp_nvlist
), config
, 0) != 0)
891 if (nvlist_lookup_uint64(*config
, ZPOOL_CONFIG_POOL_STATE
,
892 &state
) != 0 || state
> POOL_STATE_L2CACHE
) {
893 nvlist_free(*config
);
897 if (state
!= POOL_STATE_SPARE
&& state
!= POOL_STATE_L2CACHE
&&
898 (nvlist_lookup_uint64(*config
, ZPOOL_CONFIG_POOL_TXG
,
899 &txg
) != 0 || txg
== 0)) {
900 nvlist_free(*config
);
913 typedef struct rdsk_node
{
916 libzfs_handle_t
*rn_hdl
;
920 boolean_t rn_nozpool
;
924 slice_cache_compare(const void *arg1
, const void *arg2
)
926 const char *nm1
= ((rdsk_node_t
*)arg1
)->rn_name
;
927 const char *nm2
= ((rdsk_node_t
*)arg2
)->rn_name
;
928 char *nm1slice
, *nm2slice
;
932 * slices zero and two are the most likely to provide results,
935 nm1slice
= strstr(nm1
, "s0");
936 nm2slice
= strstr(nm2
, "s0");
937 if (nm1slice
&& !nm2slice
) {
940 if (!nm1slice
&& nm2slice
) {
943 nm1slice
= strstr(nm1
, "s2");
944 nm2slice
= strstr(nm2
, "s2");
945 if (nm1slice
&& !nm2slice
) {
948 if (!nm1slice
&& nm2slice
) {
952 rv
= strcmp(nm1
, nm2
);
955 return (rv
> 0 ? 1 : -1);
959 check_one_slice(avl_tree_t
*r
, char *diskname
, uint_t partno
,
960 diskaddr_t size
, uint_t blksz
)
964 char sname
[MAXNAMELEN
];
966 tmpnode
.rn_name
= &sname
[0];
967 (void) snprintf(tmpnode
.rn_name
, MAXNAMELEN
, "%s%u",
970 * protect against division by zero for disk labels that
971 * contain a bogus sector size
975 /* too small to contain a zpool? */
976 if ((size
< (SPA_MINDEVSIZE
/ blksz
)) &&
977 (node
= avl_find(r
, &tmpnode
, NULL
)))
978 node
->rn_nozpool
= B_TRUE
;
982 nozpool_all_slices(avl_tree_t
*r
, const char *sname
)
984 char diskname
[MAXNAMELEN
];
988 (void) strncpy(diskname
, sname
, MAXNAMELEN
);
989 if (((ptr
= strrchr(diskname
, 's')) == NULL
) &&
990 ((ptr
= strrchr(diskname
, 'p')) == NULL
))
994 for (i
= 0; i
< NDKMAP
; i
++)
995 check_one_slice(r
, diskname
, i
, 0, 1);
997 for (i
= 0; i
<= FD_NUMPART
; i
++)
998 check_one_slice(r
, diskname
, i
, 0, 1);
1002 check_slices(avl_tree_t
*r
, int fd
, const char *sname
)
1004 struct extvtoc vtoc
;
1006 char diskname
[MAXNAMELEN
];
1010 (void) strncpy(diskname
, sname
, MAXNAMELEN
);
1011 if ((ptr
= strrchr(diskname
, 's')) == NULL
|| !isdigit(ptr
[1]))
1015 if (read_extvtoc(fd
, &vtoc
) >= 0) {
1016 for (i
= 0; i
< NDKMAP
; i
++)
1017 check_one_slice(r
, diskname
, i
,
1018 vtoc
.v_part
[i
].p_size
, vtoc
.v_sectorsz
);
1019 } else if (efi_alloc_and_read(fd
, &gpt
) >= 0) {
1021 * on x86 we'll still have leftover links that point
1022 * to slices s[9-15], so use NDKMAP instead
1024 for (i
= 0; i
< NDKMAP
; i
++)
1025 check_one_slice(r
, diskname
, i
,
1026 gpt
->efi_parts
[i
].p_size
, gpt
->efi_lbasize
);
1027 /* nodes p[1-4] are never used with EFI labels */
1029 for (i
= 1; i
<= FD_NUMPART
; i
++)
1030 check_one_slice(r
, diskname
, i
, 0, 1);
1036 zpool_open_func(void *arg
)
1038 rdsk_node_t
*rn
= arg
;
1039 struct stat64 statbuf
;
1045 if ((fd
= openat64(rn
->rn_dfd
, rn
->rn_name
, O_RDONLY
)) < 0) {
1046 /* symlink to a device that's no longer there */
1047 if (errno
== ENOENT
)
1048 nozpool_all_slices(rn
->rn_avl
, rn
->rn_name
);
1052 * Ignore failed stats. We only want regular
1053 * files, character devs and block devs.
1055 if (fstat64(fd
, &statbuf
) != 0 ||
1056 (!S_ISREG(statbuf
.st_mode
) &&
1057 !S_ISCHR(statbuf
.st_mode
) &&
1058 !S_ISBLK(statbuf
.st_mode
))) {
1062 /* this file is too small to hold a zpool */
1063 if (S_ISREG(statbuf
.st_mode
) &&
1064 statbuf
.st_size
< SPA_MINDEVSIZE
) {
1067 } else if (!S_ISREG(statbuf
.st_mode
)) {
1069 * Try to read the disk label first so we don't have to
1070 * open a bunch of minor nodes that can't have a zpool.
1072 check_slices(rn
->rn_avl
, fd
, rn
->rn_name
);
1075 if ((zpool_read_label(fd
, &config
)) != 0) {
1077 (void) no_memory(rn
->rn_hdl
);
1083 rn
->rn_config
= config
;
1084 if (config
!= NULL
) {
1085 assert(rn
->rn_nozpool
== B_FALSE
);
1090 * Given a file descriptor, clear (zero) the label information. This function
1091 * is currently only used in the appliance stack as part of the ZFS sysevent
1095 zpool_clear_label(int fd
)
1097 struct stat64 statbuf
;
1099 vdev_label_t
*label
;
1102 if (fstat64(fd
, &statbuf
) == -1)
1104 size
= P2ALIGN_TYPED(statbuf
.st_size
, sizeof (vdev_label_t
), uint64_t);
1106 if ((label
= calloc(sizeof (vdev_label_t
), 1)) == NULL
)
1109 for (l
= 0; l
< VDEV_LABELS
; l
++) {
1110 if (pwrite64(fd
, label
, sizeof (vdev_label_t
),
1111 label_offset(size
, l
)) != sizeof (vdev_label_t
)) {
1122 * Given a list of directories to search, find all pools stored on disk. This
1123 * includes partial pools which are not available to import. If no args are
1124 * given (argc is 0), then the default directory (/dev/dsk) is searched.
1125 * poolname or guid (but not both) are provided by the caller when trying
1126 * to import a specific pool.
1129 zpool_find_import_impl(libzfs_handle_t
*hdl
, importargs_t
*iarg
)
1131 int i
, dirs
= iarg
->paths
;
1132 struct dirent64
*dp
;
1133 char path
[MAXPATHLEN
];
1134 char *end
, **dir
= iarg
->path
;
1136 nvlist_t
*ret
= NULL
;
1137 static char *default_dir
= "/dev/dsk";
1138 pool_list_t pools
= { 0 };
1139 pool_entry_t
*pe
, *penext
;
1140 vdev_entry_t
*ve
, *venext
;
1141 config_entry_t
*ce
, *cenext
;
1142 name_entry_t
*ne
, *nenext
;
1143 avl_tree_t slice_cache
;
1153 * Go through and read the label configuration information from every
1154 * possible device, organizing the information according to pool GUID
1155 * and toplevel GUID.
1157 for (i
= 0; i
< dirs
; i
++) {
1161 boolean_t config_failed
= B_FALSE
;
1164 /* use realpath to normalize the path */
1165 if (realpath(dir
[i
], path
) == 0) {
1166 (void) zfs_error_fmt(hdl
, EZFS_BADPATH
,
1167 dgettext(TEXT_DOMAIN
, "cannot open '%s'"), dir
[i
]);
1170 end
= &path
[strlen(path
)];
1173 pathleft
= &path
[sizeof (path
)] - end
;
1176 * Using raw devices instead of block devices when we're
1177 * reading the labels skips a bunch of slow operations during
1178 * close(2) processing, so we replace /dev/dsk with /dev/rdsk.
1180 if (strcmp(path
, "/dev/dsk/") == 0)
1181 rdsk
= "/dev/rdsk/";
1185 if ((dfd
= open64(rdsk
, O_RDONLY
)) < 0 ||
1186 (dirp
= fdopendir(dfd
)) == NULL
) {
1189 zfs_error_aux(hdl
, strerror(errno
));
1190 (void) zfs_error_fmt(hdl
, EZFS_BADPATH
,
1191 dgettext(TEXT_DOMAIN
, "cannot open '%s'"),
1196 avl_create(&slice_cache
, slice_cache_compare
,
1197 sizeof (rdsk_node_t
), offsetof(rdsk_node_t
, rn_node
));
1199 * This is not MT-safe, but we have no MT consumers of libzfs
1201 while ((dp
= readdir64(dirp
)) != NULL
) {
1202 const char *name
= dp
->d_name
;
1203 if (name
[0] == '.' &&
1204 (name
[1] == 0 || (name
[1] == '.' && name
[2] == 0)))
1207 slice
= zfs_alloc(hdl
, sizeof (rdsk_node_t
));
1208 slice
->rn_name
= zfs_strdup(hdl
, name
);
1209 slice
->rn_avl
= &slice_cache
;
1210 slice
->rn_dfd
= dfd
;
1211 slice
->rn_hdl
= hdl
;
1212 slice
->rn_nozpool
= B_FALSE
;
1213 avl_add(&slice_cache
, slice
);
1216 * create a thread pool to do all of this in parallel;
1217 * rn_nozpool is not protected, so this is racy in that
1218 * multiple tasks could decide that the same slice can
1219 * not hold a zpool, which is benign. Also choose
1220 * double the number of processors; we hold a lot of
1221 * locks in the kernel, so going beyond this doesn't
1224 t
= tpool_create(1, 2 * sysconf(_SC_NPROCESSORS_ONLN
),
1226 for (slice
= avl_first(&slice_cache
); slice
;
1227 (slice
= avl_walk(&slice_cache
, slice
,
1229 (void) tpool_dispatch(t
, zpool_open_func
, slice
);
1234 while ((slice
= avl_destroy_nodes(&slice_cache
,
1235 &cookie
)) != NULL
) {
1236 if (slice
->rn_config
!= NULL
&& !config_failed
) {
1237 nvlist_t
*config
= slice
->rn_config
;
1238 boolean_t matched
= B_TRUE
;
1240 if (iarg
->poolname
!= NULL
) {
1243 matched
= nvlist_lookup_string(config
,
1244 ZPOOL_CONFIG_POOL_NAME
,
1246 strcmp(iarg
->poolname
, pname
) == 0;
1247 } else if (iarg
->guid
!= 0) {
1250 matched
= nvlist_lookup_uint64(config
,
1251 ZPOOL_CONFIG_POOL_GUID
,
1253 iarg
->guid
== this_guid
;
1256 nvlist_free(config
);
1259 * use the non-raw path for the config
1261 (void) strlcpy(end
, slice
->rn_name
,
1263 if (add_config(hdl
, &pools
, path
,
1265 config_failed
= B_TRUE
;
1268 free(slice
->rn_name
);
1271 avl_destroy(&slice_cache
);
1273 (void) closedir(dirp
);
1279 ret
= get_configs(hdl
, &pools
, iarg
->can_be_active
);
1282 for (pe
= pools
.pools
; pe
!= NULL
; pe
= penext
) {
1283 penext
= pe
->pe_next
;
1284 for (ve
= pe
->pe_vdevs
; ve
!= NULL
; ve
= venext
) {
1285 venext
= ve
->ve_next
;
1286 for (ce
= ve
->ve_configs
; ce
!= NULL
; ce
= cenext
) {
1287 cenext
= ce
->ce_next
;
1289 nvlist_free(ce
->ce_config
);
1297 for (ne
= pools
.names
; ne
!= NULL
; ne
= nenext
) {
1298 nenext
= ne
->ne_next
;
1307 zpool_find_import(libzfs_handle_t
*hdl
, int argc
, char **argv
)
1309 importargs_t iarg
= { 0 };
1314 return (zpool_find_import_impl(hdl
, &iarg
));
1318 * Given a cache file, return the contents as a list of importable pools.
1319 * poolname or guid (but not both) are provided by the caller when trying
1320 * to import a specific pool.
1323 zpool_find_import_cached(libzfs_handle_t
*hdl
, const char *cachefile
,
1324 char *poolname
, uint64_t guid
)
1328 struct stat64 statbuf
;
1329 nvlist_t
*raw
, *src
, *dst
;
1336 verify(poolname
== NULL
|| guid
== 0);
1338 if ((fd
= open(cachefile
, O_RDONLY
)) < 0) {
1339 zfs_error_aux(hdl
, "%s", strerror(errno
));
1340 (void) zfs_error(hdl
, EZFS_BADCACHE
,
1341 dgettext(TEXT_DOMAIN
, "failed to open cache file"));
1345 if (fstat64(fd
, &statbuf
) != 0) {
1346 zfs_error_aux(hdl
, "%s", strerror(errno
));
1348 (void) zfs_error(hdl
, EZFS_BADCACHE
,
1349 dgettext(TEXT_DOMAIN
, "failed to get size of cache file"));
1353 if ((buf
= zfs_alloc(hdl
, statbuf
.st_size
)) == NULL
) {
1358 if (read(fd
, buf
, statbuf
.st_size
) != statbuf
.st_size
) {
1361 (void) zfs_error(hdl
, EZFS_BADCACHE
,
1362 dgettext(TEXT_DOMAIN
,
1363 "failed to read cache file contents"));
1369 if (nvlist_unpack(buf
, statbuf
.st_size
, &raw
, 0) != 0) {
1371 (void) zfs_error(hdl
, EZFS_BADCACHE
,
1372 dgettext(TEXT_DOMAIN
,
1373 "invalid or corrupt cache file contents"));
1380 * Go through and get the current state of the pools and refresh their
1383 if (nvlist_alloc(&pools
, 0, 0) != 0) {
1384 (void) no_memory(hdl
);
1390 while ((elem
= nvlist_next_nvpair(raw
, elem
)) != NULL
) {
1391 src
= fnvpair_value_nvlist(elem
);
1393 name
= fnvlist_lookup_string(src
, ZPOOL_CONFIG_POOL_NAME
);
1394 if (poolname
!= NULL
&& strcmp(poolname
, name
) != 0)
1397 this_guid
= fnvlist_lookup_uint64(src
, ZPOOL_CONFIG_POOL_GUID
);
1398 if (guid
!= 0 && guid
!= this_guid
)
1401 if (pool_active(hdl
, name
, this_guid
, &active
) != 0) {
1410 if ((dst
= refresh_config(hdl
, src
)) == NULL
) {
1416 if (nvlist_add_nvlist(pools
, nvpair_name(elem
), dst
) != 0) {
1417 (void) no_memory(hdl
);
1431 name_or_guid_exists(zpool_handle_t
*zhp
, void *data
)
1433 importargs_t
*import
= data
;
1436 if (import
->poolname
!= NULL
) {
1439 verify(nvlist_lookup_string(zhp
->zpool_config
,
1440 ZPOOL_CONFIG_POOL_NAME
, &pool_name
) == 0);
1441 if (strcmp(pool_name
, import
->poolname
) == 0)
1446 verify(nvlist_lookup_uint64(zhp
->zpool_config
,
1447 ZPOOL_CONFIG_POOL_GUID
, &pool_guid
) == 0);
1448 if (pool_guid
== import
->guid
)
1457 zpool_search_import(libzfs_handle_t
*hdl
, importargs_t
*import
)
1459 verify(import
->poolname
== NULL
|| import
->guid
== 0);
1462 import
->exists
= zpool_iter(hdl
, name_or_guid_exists
, import
);
1464 if (import
->cachefile
!= NULL
)
1465 return (zpool_find_import_cached(hdl
, import
->cachefile
,
1466 import
->poolname
, import
->guid
));
1468 return (zpool_find_import_impl(hdl
, import
));
1472 find_guid(nvlist_t
*nv
, uint64_t guid
)
1478 verify(nvlist_lookup_uint64(nv
, ZPOOL_CONFIG_GUID
, &tmp
) == 0);
1482 if (nvlist_lookup_nvlist_array(nv
, ZPOOL_CONFIG_CHILDREN
,
1483 &child
, &children
) == 0) {
1484 for (c
= 0; c
< children
; c
++)
1485 if (find_guid(child
[c
], guid
))
1492 typedef struct aux_cbdata
{
1493 const char *cb_type
;
1495 zpool_handle_t
*cb_zhp
;
1499 find_aux(zpool_handle_t
*zhp
, void *data
)
1501 aux_cbdata_t
*cbp
= data
;
1507 verify(nvlist_lookup_nvlist(zhp
->zpool_config
, ZPOOL_CONFIG_VDEV_TREE
,
1510 if (nvlist_lookup_nvlist_array(nvroot
, cbp
->cb_type
,
1511 &list
, &count
) == 0) {
1512 for (i
= 0; i
< count
; i
++) {
1513 verify(nvlist_lookup_uint64(list
[i
],
1514 ZPOOL_CONFIG_GUID
, &guid
) == 0);
1515 if (guid
== cbp
->cb_guid
) {
1527 * Determines if the pool is in use. If so, it returns true and the state of
1528 * the pool as well as the name of the pool. Both strings are allocated and
1529 * must be freed by the caller.
1532 zpool_in_use(libzfs_handle_t
*hdl
, int fd
, pool_state_t
*state
, char **namestr
,
1538 uint64_t guid
, vdev_guid
;
1539 zpool_handle_t
*zhp
;
1540 nvlist_t
*pool_config
;
1541 uint64_t stateval
, isspare
;
1542 aux_cbdata_t cb
= { 0 };
1547 if (zpool_read_label(fd
, &config
) != 0) {
1548 (void) no_memory(hdl
);
1555 verify(nvlist_lookup_uint64(config
, ZPOOL_CONFIG_POOL_STATE
,
1557 verify(nvlist_lookup_uint64(config
, ZPOOL_CONFIG_GUID
,
1560 if (stateval
!= POOL_STATE_SPARE
&& stateval
!= POOL_STATE_L2CACHE
) {
1561 verify(nvlist_lookup_string(config
, ZPOOL_CONFIG_POOL_NAME
,
1563 verify(nvlist_lookup_uint64(config
, ZPOOL_CONFIG_POOL_GUID
,
1568 case POOL_STATE_EXPORTED
:
1570 * A pool with an exported state may in fact be imported
1571 * read-only, so check the in-core state to see if it's
1572 * active and imported read-only. If it is, set
1573 * its state to active.
1575 if (pool_active(hdl
, name
, guid
, &isactive
) == 0 && isactive
&&
1576 (zhp
= zpool_open_canfail(hdl
, name
)) != NULL
) {
1577 if (zpool_get_prop_int(zhp
, ZPOOL_PROP_READONLY
, NULL
))
1578 stateval
= POOL_STATE_ACTIVE
;
1581 * All we needed the zpool handle for is the
1582 * readonly prop check.
1590 case POOL_STATE_ACTIVE
:
1592 * For an active pool, we have to determine if it's really part
1593 * of a currently active pool (in which case the pool will exist
1594 * and the guid will be the same), or whether it's part of an
1595 * active pool that was disconnected without being explicitly
1598 if (pool_active(hdl
, name
, guid
, &isactive
) != 0) {
1599 nvlist_free(config
);
1605 * Because the device may have been removed while
1606 * offlined, we only report it as active if the vdev is
1607 * still present in the config. Otherwise, pretend like
1610 if ((zhp
= zpool_open_canfail(hdl
, name
)) != NULL
&&
1611 (pool_config
= zpool_get_config(zhp
, NULL
))
1615 verify(nvlist_lookup_nvlist(pool_config
,
1616 ZPOOL_CONFIG_VDEV_TREE
, &nvroot
) == 0);
1617 ret
= find_guid(nvroot
, vdev_guid
);
1623 * If this is an active spare within another pool, we
1624 * treat it like an unused hot spare. This allows the
1625 * user to create a pool with a hot spare that currently
1626 * in use within another pool. Since we return B_TRUE,
1627 * libdiskmgt will continue to prevent generic consumers
1628 * from using the device.
1630 if (ret
&& nvlist_lookup_uint64(config
,
1631 ZPOOL_CONFIG_IS_SPARE
, &isspare
) == 0 && isspare
)
1632 stateval
= POOL_STATE_SPARE
;
1637 stateval
= POOL_STATE_POTENTIALLY_ACTIVE
;
1642 case POOL_STATE_SPARE
:
1644 * For a hot spare, it can be either definitively in use, or
1645 * potentially active. To determine if it's in use, we iterate
1646 * over all pools in the system and search for one with a spare
1647 * with a matching guid.
1649 * Due to the shared nature of spares, we don't actually report
1650 * the potentially active case as in use. This means the user
1651 * can freely create pools on the hot spares of exported pools,
1652 * but to do otherwise makes the resulting code complicated, and
1653 * we end up having to deal with this case anyway.
1656 cb
.cb_guid
= vdev_guid
;
1657 cb
.cb_type
= ZPOOL_CONFIG_SPARES
;
1658 if (zpool_iter(hdl
, find_aux
, &cb
) == 1) {
1659 name
= (char *)zpool_get_name(cb
.cb_zhp
);
1666 case POOL_STATE_L2CACHE
:
1669 * Check if any pool is currently using this l2cache device.
1672 cb
.cb_guid
= vdev_guid
;
1673 cb
.cb_type
= ZPOOL_CONFIG_L2CACHE
;
1674 if (zpool_iter(hdl
, find_aux
, &cb
) == 1) {
1675 name
= (char *)zpool_get_name(cb
.cb_zhp
);
1688 if ((*namestr
= zfs_strdup(hdl
, name
)) == NULL
) {
1690 zpool_close(cb
.cb_zhp
);
1691 nvlist_free(config
);
1694 *state
= (pool_state_t
)stateval
;
1698 zpool_close(cb
.cb_zhp
);
1700 nvlist_free(config
);