4 * The contents of this file are subject to the terms of the
5 * Common Development and Distribution License (the "License").
6 * You may not use this file except in compliance with the License.
8 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
9 * or http://www.opensolaris.org/os/licensing.
10 * See the License for the specific language governing permissions
11 * and limitations under the License.
13 * When distributing Covered Code, include this CDDL HEADER in each
14 * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
15 * If applicable, add the following below this CDDL HEADER, with the
16 * fields enclosed by brackets "[]" replaced with your own identifying
17 * information: Portions Copyright [yyyy] [name of copyright owner]
23 * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
24 * Copyright (c) 2012, 2016 by Delphix. All rights reserved.
25 * Copyright 2015 RackTop Systems.
26 * Copyright 2017 Nexenta Systems, Inc.
30 * Pool import support functions.
32 * To import a pool, we rely on reading the configuration information from the
33 * ZFS label of each device. If we successfully read the label, then we
34 * organize the configuration information in the following hierarchy:
36 * pool guid -> toplevel vdev guid -> label txg
38 * Duplicate entries matching this same tuple will be discarded. Once we have
39 * examined every device, we pick the best label txg config for each toplevel
40 * vdev. We then arrange these toplevel vdevs into a complete pool config, and
41 * update any paths that have changed. Finally, we attempt to import the pool
42 * using our derived config, and record the results.
57 #include <sys/dktp/fdisk.h>
58 #include <sys/efi_partition.h>
59 #include <thread_pool.h>
61 #include <sys/vdev_impl.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
{
90 struct name_entry
*ne_next
;
93 typedef struct pool_list
{
99 * Go through and fix up any path and/or devid information for the given vdev
103 fix_paths(nvlist_t
*nv
, name_entry_t
*names
)
108 name_entry_t
*ne
, *best
;
112 if (nvlist_lookup_nvlist_array(nv
, ZPOOL_CONFIG_CHILDREN
,
113 &child
, &children
) == 0) {
114 for (c
= 0; c
< children
; c
++)
115 if (fix_paths(child
[c
], names
) != 0)
121 * This is a leaf (file or disk) vdev. In either case, go through
122 * the name list and see if we find a matching guid. If so, replace
123 * the path and see if we can calculate a new devid.
125 * There may be multiple names associated with a particular guid, in
126 * which case we have overlapping slices or multiple paths to the same
127 * disk. If this is the case, then we want to pick the path that is
128 * the most similar to the original, where "most similar" is the number
129 * of matching characters starting from the end of the path. This will
130 * preserve slice numbers even if the disks have been reorganized, and
131 * will also catch preferred disk names if multiple paths exist.
133 verify(nvlist_lookup_uint64(nv
, ZPOOL_CONFIG_GUID
, &guid
) == 0);
134 if (nvlist_lookup_string(nv
, ZPOOL_CONFIG_PATH
, &path
) != 0)
139 for (ne
= names
; ne
!= NULL
; ne
= ne
->ne_next
) {
140 if (ne
->ne_guid
== guid
) {
141 const char *src
, *dst
;
149 src
= ne
->ne_name
+ strlen(ne
->ne_name
) - 1;
150 dst
= path
+ strlen(path
) - 1;
151 for (count
= 0; src
>= ne
->ne_name
&& dst
>= path
;
152 src
--, dst
--, count
++)
157 * At this point, 'count' is the number of characters
158 * matched from the end.
160 if (count
> matched
|| best
== NULL
) {
170 if (nvlist_add_string(nv
, ZPOOL_CONFIG_PATH
, best
->ne_name
) != 0)
173 if ((devid
= devid_str_from_path(best
->ne_name
)) == NULL
) {
174 (void) nvlist_remove_all(nv
, ZPOOL_CONFIG_DEVID
);
176 if (nvlist_add_string(nv
, ZPOOL_CONFIG_DEVID
, devid
) != 0) {
177 devid_str_free(devid
);
180 devid_str_free(devid
);
187 * Add the given configuration to the list of known devices.
190 add_config(libzfs_handle_t
*hdl
, pool_list_t
*pl
, const char *path
,
193 uint64_t pool_guid
, vdev_guid
, top_guid
, txg
, state
;
200 * If this is a hot spare not currently in use or level 2 cache
201 * device, add it to the list of names to translate, but don't do
204 if (nvlist_lookup_uint64(config
, ZPOOL_CONFIG_POOL_STATE
,
206 (state
== POOL_STATE_SPARE
|| state
== POOL_STATE_L2CACHE
) &&
207 nvlist_lookup_uint64(config
, ZPOOL_CONFIG_GUID
, &vdev_guid
) == 0) {
208 if ((ne
= zfs_alloc(hdl
, sizeof (name_entry_t
))) == NULL
)
211 if ((ne
->ne_name
= zfs_strdup(hdl
, path
)) == NULL
) {
216 ne
->ne_guid
= vdev_guid
;
217 ne
->ne_next
= pl
->names
;
225 * If we have a valid config but cannot read any of these fields, then
226 * it means we have a half-initialized label. In vdev_label_init()
227 * we write a label with txg == 0 so that we can identify the device
228 * in case the user refers to the same disk later on. If we fail to
229 * create the pool, we'll be left with a label in this state
230 * which should not be considered part of a valid pool.
232 if (nvlist_lookup_uint64(config
, ZPOOL_CONFIG_POOL_GUID
,
234 nvlist_lookup_uint64(config
, ZPOOL_CONFIG_GUID
,
236 nvlist_lookup_uint64(config
, ZPOOL_CONFIG_TOP_GUID
,
238 nvlist_lookup_uint64(config
, ZPOOL_CONFIG_POOL_TXG
,
239 &txg
) != 0 || txg
== 0) {
245 * First, see if we know about this pool. If not, then add it to the
246 * list of known pools.
248 for (pe
= pl
->pools
; pe
!= NULL
; pe
= pe
->pe_next
) {
249 if (pe
->pe_guid
== pool_guid
)
254 if ((pe
= zfs_alloc(hdl
, sizeof (pool_entry_t
))) == NULL
) {
258 pe
->pe_guid
= pool_guid
;
259 pe
->pe_next
= pl
->pools
;
264 * Second, see if we know about this toplevel vdev. Add it if its
267 for (ve
= pe
->pe_vdevs
; ve
!= NULL
; ve
= ve
->ve_next
) {
268 if (ve
->ve_guid
== top_guid
)
273 if ((ve
= zfs_alloc(hdl
, sizeof (vdev_entry_t
))) == NULL
) {
277 ve
->ve_guid
= top_guid
;
278 ve
->ve_next
= pe
->pe_vdevs
;
283 * Third, see if we have a config with a matching transaction group. If
284 * so, then we do nothing. Otherwise, add it to the list of known
287 for (ce
= ve
->ve_configs
; ce
!= NULL
; ce
= ce
->ce_next
) {
288 if (ce
->ce_txg
== txg
)
293 if ((ce
= zfs_alloc(hdl
, sizeof (config_entry_t
))) == NULL
) {
298 ce
->ce_config
= config
;
299 ce
->ce_next
= ve
->ve_configs
;
306 * At this point we've successfully added our config to the list of
307 * known configs. The last thing to do is add the vdev guid -> path
308 * mappings so that we can fix up the configuration as necessary before
311 if ((ne
= zfs_alloc(hdl
, sizeof (name_entry_t
))) == NULL
)
314 if ((ne
->ne_name
= zfs_strdup(hdl
, path
)) == NULL
) {
319 ne
->ne_guid
= vdev_guid
;
320 ne
->ne_next
= pl
->names
;
327 * Returns true if the named pool matches the given GUID.
330 pool_active(libzfs_handle_t
*hdl
, const char *name
, uint64_t guid
,
336 if (zpool_open_silent(hdl
, name
, &zhp
) != 0)
344 verify(nvlist_lookup_uint64(zhp
->zpool_config
, ZPOOL_CONFIG_POOL_GUID
,
349 *isactive
= (theguid
== guid
);
354 refresh_config(libzfs_handle_t
*hdl
, nvlist_t
*config
)
357 zfs_cmd_t zc
= { 0 };
358 int err
, dstbuf_size
;
360 if (zcmd_write_conf_nvlist(hdl
, &zc
, config
) != 0)
363 dstbuf_size
= MAX(CONFIG_BUF_MINSIZE
, zc
.zc_nvlist_conf_size
* 4);
365 if (zcmd_alloc_dst_nvlist(hdl
, &zc
, dstbuf_size
) != 0) {
366 zcmd_free_nvlists(&zc
);
370 while ((err
= ioctl(hdl
->libzfs_fd
, ZFS_IOC_POOL_TRYIMPORT
,
371 &zc
)) != 0 && errno
== ENOMEM
) {
372 if (zcmd_expand_dst_nvlist(hdl
, &zc
) != 0) {
373 zcmd_free_nvlists(&zc
);
379 zcmd_free_nvlists(&zc
);
383 if (zcmd_read_dst_nvlist(hdl
, &zc
, &nvl
) != 0) {
384 zcmd_free_nvlists(&zc
);
388 zcmd_free_nvlists(&zc
);
393 * Determine if the vdev id is a hole in the namespace.
396 vdev_is_hole(uint64_t *hole_array
, uint_t holes
, uint_t id
)
398 for (int c
= 0; c
< holes
; c
++) {
400 /* Top-level is a hole */
401 if (hole_array
[c
] == id
)
408 * Convert our list of pools into the definitive set of configurations. We
409 * start by picking the best config for each toplevel vdev. Once that's done,
410 * we assemble the toplevel vdevs into a full config for the pool. We make a
411 * pass to fix up any incorrect paths, and then add it to the main list to
412 * return to the user.
415 get_configs(libzfs_handle_t
*hdl
, pool_list_t
*pl
, boolean_t active_ok
,
421 nvlist_t
*ret
= NULL
, *config
= NULL
, *tmp
= NULL
, *nvtop
, *nvroot
;
422 nvlist_t
**spares
, **l2cache
;
423 uint_t i
, nspares
, nl2cache
;
424 boolean_t config_seen
;
426 char *name
, *hostname
= NULL
;
429 nvlist_t
**child
= NULL
;
431 uint64_t *hole_array
, max_id
;
436 boolean_t found_one
= B_FALSE
;
437 boolean_t valid_top_config
= B_FALSE
;
439 if (nvlist_alloc(&ret
, 0, 0) != 0)
442 for (pe
= pl
->pools
; pe
!= NULL
; pe
= pe
->pe_next
) {
443 uint64_t id
, max_txg
= 0;
445 if (nvlist_alloc(&config
, NV_UNIQUE_NAME
, 0) != 0)
447 config_seen
= B_FALSE
;
450 * Iterate over all toplevel vdevs. Grab the pool configuration
451 * from the first one we find, and then go through the rest and
452 * add them as necessary to the 'vdevs' member of the config.
454 for (ve
= pe
->pe_vdevs
; ve
!= NULL
; ve
= ve
->ve_next
) {
457 * Determine the best configuration for this vdev by
458 * selecting the config with the latest transaction
462 for (ce
= ve
->ve_configs
; ce
!= NULL
;
465 if (ce
->ce_txg
> best_txg
) {
467 best_txg
= ce
->ce_txg
;
472 * We rely on the fact that the max txg for the
473 * pool will contain the most up-to-date information
474 * about the valid top-levels in the vdev namespace.
476 if (best_txg
> max_txg
) {
477 (void) nvlist_remove(config
,
478 ZPOOL_CONFIG_VDEV_CHILDREN
,
480 (void) nvlist_remove(config
,
481 ZPOOL_CONFIG_HOLE_ARRAY
,
482 DATA_TYPE_UINT64_ARRAY
);
488 valid_top_config
= B_FALSE
;
490 if (nvlist_lookup_uint64(tmp
,
491 ZPOOL_CONFIG_VDEV_CHILDREN
, &max_id
) == 0) {
492 verify(nvlist_add_uint64(config
,
493 ZPOOL_CONFIG_VDEV_CHILDREN
,
495 valid_top_config
= B_TRUE
;
498 if (nvlist_lookup_uint64_array(tmp
,
499 ZPOOL_CONFIG_HOLE_ARRAY
, &hole_array
,
501 verify(nvlist_add_uint64_array(config
,
502 ZPOOL_CONFIG_HOLE_ARRAY
,
503 hole_array
, holes
) == 0);
509 * Copy the relevant pieces of data to the pool
515 * comment (if available)
517 * hostid (if available)
518 * hostname (if available)
520 uint64_t state
, version
;
521 char *comment
= NULL
;
523 version
= fnvlist_lookup_uint64(tmp
,
524 ZPOOL_CONFIG_VERSION
);
525 fnvlist_add_uint64(config
,
526 ZPOOL_CONFIG_VERSION
, version
);
527 guid
= fnvlist_lookup_uint64(tmp
,
528 ZPOOL_CONFIG_POOL_GUID
);
529 fnvlist_add_uint64(config
,
530 ZPOOL_CONFIG_POOL_GUID
, guid
);
531 name
= fnvlist_lookup_string(tmp
,
532 ZPOOL_CONFIG_POOL_NAME
);
533 fnvlist_add_string(config
,
534 ZPOOL_CONFIG_POOL_NAME
, name
);
536 if (nvlist_lookup_string(tmp
,
537 ZPOOL_CONFIG_COMMENT
, &comment
) == 0)
538 fnvlist_add_string(config
,
539 ZPOOL_CONFIG_COMMENT
, comment
);
541 state
= fnvlist_lookup_uint64(tmp
,
542 ZPOOL_CONFIG_POOL_STATE
);
543 fnvlist_add_uint64(config
,
544 ZPOOL_CONFIG_POOL_STATE
, state
);
547 if (nvlist_lookup_uint64(tmp
,
548 ZPOOL_CONFIG_HOSTID
, &hostid
) == 0) {
549 fnvlist_add_uint64(config
,
550 ZPOOL_CONFIG_HOSTID
, hostid
);
551 hostname
= fnvlist_lookup_string(tmp
,
552 ZPOOL_CONFIG_HOSTNAME
);
553 fnvlist_add_string(config
,
554 ZPOOL_CONFIG_HOSTNAME
, hostname
);
557 config_seen
= B_TRUE
;
561 * Add this top-level vdev to the child array.
563 verify(nvlist_lookup_nvlist(tmp
,
564 ZPOOL_CONFIG_VDEV_TREE
, &nvtop
) == 0);
565 verify(nvlist_lookup_uint64(nvtop
, ZPOOL_CONFIG_ID
,
568 if (id
>= children
) {
571 newchild
= zfs_alloc(hdl
, (id
+ 1) *
572 sizeof (nvlist_t
*));
573 if (newchild
== NULL
)
576 for (c
= 0; c
< children
; c
++)
577 newchild
[c
] = child
[c
];
583 if (nvlist_dup(nvtop
, &child
[id
], 0) != 0)
589 * If we have information about all the top-levels then
590 * clean up the nvlist which we've constructed. This
591 * means removing any extraneous devices that are
592 * beyond the valid range or adding devices to the end
593 * of our array which appear to be missing.
595 if (valid_top_config
) {
596 if (max_id
< children
) {
597 for (c
= max_id
; c
< children
; c
++)
598 nvlist_free(child
[c
]);
600 } else if (max_id
> children
) {
603 newchild
= zfs_alloc(hdl
, (max_id
) *
604 sizeof (nvlist_t
*));
605 if (newchild
== NULL
)
608 for (c
= 0; c
< children
; c
++)
609 newchild
[c
] = child
[c
];
617 verify(nvlist_lookup_uint64(config
, ZPOOL_CONFIG_POOL_GUID
,
621 * The vdev namespace may contain holes as a result of
622 * device removal. We must add them back into the vdev
623 * tree before we process any missing devices.
626 ASSERT(valid_top_config
);
628 for (c
= 0; c
< children
; c
++) {
631 if (child
[c
] != NULL
||
632 !vdev_is_hole(hole_array
, holes
, c
))
635 if (nvlist_alloc(&holey
, NV_UNIQUE_NAME
,
640 * Holes in the namespace are treated as
641 * "hole" top-level vdevs and have a
642 * special flag set on them.
644 if (nvlist_add_string(holey
,
646 VDEV_TYPE_HOLE
) != 0 ||
647 nvlist_add_uint64(holey
,
648 ZPOOL_CONFIG_ID
, c
) != 0 ||
649 nvlist_add_uint64(holey
,
650 ZPOOL_CONFIG_GUID
, 0ULL) != 0) {
659 * Look for any missing top-level vdevs. If this is the case,
660 * create a faked up 'missing' vdev as a placeholder. We cannot
661 * simply compress the child array, because the kernel performs
662 * certain checks to make sure the vdev IDs match their location
663 * in the configuration.
665 for (c
= 0; c
< children
; c
++) {
666 if (child
[c
] == NULL
) {
668 if (nvlist_alloc(&missing
, NV_UNIQUE_NAME
,
671 if (nvlist_add_string(missing
,
673 VDEV_TYPE_MISSING
) != 0 ||
674 nvlist_add_uint64(missing
,
675 ZPOOL_CONFIG_ID
, c
) != 0 ||
676 nvlist_add_uint64(missing
,
677 ZPOOL_CONFIG_GUID
, 0ULL) != 0) {
678 nvlist_free(missing
);
686 * Put all of this pool's top-level vdevs into a root vdev.
688 if (nvlist_alloc(&nvroot
, NV_UNIQUE_NAME
, 0) != 0)
690 if (nvlist_add_string(nvroot
, ZPOOL_CONFIG_TYPE
,
691 VDEV_TYPE_ROOT
) != 0 ||
692 nvlist_add_uint64(nvroot
, ZPOOL_CONFIG_ID
, 0ULL) != 0 ||
693 nvlist_add_uint64(nvroot
, ZPOOL_CONFIG_GUID
, guid
) != 0 ||
694 nvlist_add_nvlist_array(nvroot
, ZPOOL_CONFIG_CHILDREN
,
695 child
, children
) != 0) {
700 for (c
= 0; c
< children
; c
++)
701 nvlist_free(child
[c
]);
707 * Go through and fix up any paths and/or devids based on our
708 * known list of vdev GUID -> path mappings.
710 if (fix_paths(nvroot
, pl
->names
) != 0) {
716 * Add the root vdev to this pool's configuration.
718 if (nvlist_add_nvlist(config
, ZPOOL_CONFIG_VDEV_TREE
,
726 * zdb uses this path to report on active pools that were
727 * imported or created using -R.
733 * Determine if this pool is currently active, in which case we
734 * can't actually import it.
736 verify(nvlist_lookup_string(config
, ZPOOL_CONFIG_POOL_NAME
,
738 verify(nvlist_lookup_uint64(config
, ZPOOL_CONFIG_POOL_GUID
,
741 if (pool_active(hdl
, name
, guid
, &isactive
) != 0)
750 if (policy
!= NULL
) {
751 if (nvlist_add_nvlist(config
, ZPOOL_REWIND_POLICY
,
756 if ((nvl
= refresh_config(hdl
, config
)) == NULL
) {
766 * Go through and update the paths for spares, now that we have
769 verify(nvlist_lookup_nvlist(config
, ZPOOL_CONFIG_VDEV_TREE
,
771 if (nvlist_lookup_nvlist_array(nvroot
, ZPOOL_CONFIG_SPARES
,
772 &spares
, &nspares
) == 0) {
773 for (i
= 0; i
< nspares
; i
++) {
774 if (fix_paths(spares
[i
], pl
->names
) != 0)
780 * Update the paths for l2cache devices.
782 if (nvlist_lookup_nvlist_array(nvroot
, ZPOOL_CONFIG_L2CACHE
,
783 &l2cache
, &nl2cache
) == 0) {
784 for (i
= 0; i
< nl2cache
; i
++) {
785 if (fix_paths(l2cache
[i
], pl
->names
) != 0)
791 * Restore the original information read from the actual label.
793 (void) nvlist_remove(config
, ZPOOL_CONFIG_HOSTID
,
795 (void) nvlist_remove(config
, ZPOOL_CONFIG_HOSTNAME
,
798 verify(nvlist_add_uint64(config
, ZPOOL_CONFIG_HOSTID
,
800 verify(nvlist_add_string(config
, ZPOOL_CONFIG_HOSTNAME
,
806 * Add this pool to the list of configs.
808 verify(nvlist_lookup_string(config
, ZPOOL_CONFIG_POOL_NAME
,
810 if (nvlist_add_nvlist(ret
, name
, config
) != 0)
826 (void) no_memory(hdl
);
830 for (c
= 0; c
< children
; c
++)
831 nvlist_free(child
[c
]);
838 * Return the offset of the given label.
841 label_offset(uint64_t size
, int l
)
843 ASSERT(P2PHASE_TYPED(size
, sizeof (vdev_label_t
), uint64_t) == 0);
844 return (l
* sizeof (vdev_label_t
) + (l
< VDEV_LABELS
/ 2 ?
845 0 : size
- VDEV_LABELS
* sizeof (vdev_label_t
)));
849 * Given a file descriptor, read the label information and return an nvlist
850 * describing the configuration, if there is one.
851 * Return 0 on success, or -1 on failure
854 zpool_read_label(int fd
, nvlist_t
**config
)
856 struct stat64 statbuf
;
859 uint64_t state
, txg
, size
;
863 if (fstat64(fd
, &statbuf
) == -1)
865 size
= P2ALIGN_TYPED(statbuf
.st_size
, sizeof (vdev_label_t
), uint64_t);
867 if ((label
= malloc(sizeof (vdev_label_t
))) == NULL
)
870 for (l
= 0; l
< VDEV_LABELS
; l
++) {
871 if (pread64(fd
, label
, sizeof (vdev_label_t
),
872 label_offset(size
, l
)) != sizeof (vdev_label_t
))
875 if (nvlist_unpack(label
->vl_vdev_phys
.vp_nvlist
,
876 sizeof (label
->vl_vdev_phys
.vp_nvlist
), config
, 0) != 0)
879 if (nvlist_lookup_uint64(*config
, ZPOOL_CONFIG_POOL_STATE
,
880 &state
) != 0 || state
> POOL_STATE_L2CACHE
) {
881 nvlist_free(*config
);
885 if (state
!= POOL_STATE_SPARE
&& state
!= POOL_STATE_L2CACHE
&&
886 (nvlist_lookup_uint64(*config
, ZPOOL_CONFIG_POOL_TXG
,
887 &txg
) != 0 || txg
== 0)) {
888 nvlist_free(*config
);
901 typedef struct rdsk_node
{
904 libzfs_handle_t
*rn_hdl
;
908 boolean_t rn_nozpool
;
912 slice_cache_compare(const void *arg1
, const void *arg2
)
914 const char *nm1
= ((rdsk_node_t
*)arg1
)->rn_name
;
915 const char *nm2
= ((rdsk_node_t
*)arg2
)->rn_name
;
916 char *nm1slice
, *nm2slice
;
920 * slices zero and two are the most likely to provide results,
923 nm1slice
= strstr(nm1
, "s0");
924 nm2slice
= strstr(nm2
, "s0");
925 if (nm1slice
&& !nm2slice
) {
928 if (!nm1slice
&& nm2slice
) {
931 nm1slice
= strstr(nm1
, "s2");
932 nm2slice
= strstr(nm2
, "s2");
933 if (nm1slice
&& !nm2slice
) {
936 if (!nm1slice
&& nm2slice
) {
940 rv
= strcmp(nm1
, nm2
);
943 return (rv
> 0 ? 1 : -1);
947 check_one_slice(avl_tree_t
*r
, char *diskname
, uint_t partno
,
948 diskaddr_t size
, uint_t blksz
)
952 char sname
[MAXNAMELEN
];
954 tmpnode
.rn_name
= &sname
[0];
955 (void) snprintf(tmpnode
.rn_name
, MAXNAMELEN
, "%s%u",
958 * protect against division by zero for disk labels that
959 * contain a bogus sector size
963 /* too small to contain a zpool? */
964 if ((size
< (SPA_MINDEVSIZE
/ blksz
)) &&
965 (node
= avl_find(r
, &tmpnode
, NULL
)))
966 node
->rn_nozpool
= B_TRUE
;
970 nozpool_all_slices(avl_tree_t
*r
, const char *sname
)
972 char diskname
[MAXNAMELEN
];
976 (void) strncpy(diskname
, sname
, MAXNAMELEN
);
977 if (((ptr
= strrchr(diskname
, 's')) == NULL
) &&
978 ((ptr
= strrchr(diskname
, 'p')) == NULL
))
982 for (i
= 0; i
< NDKMAP
; i
++)
983 check_one_slice(r
, diskname
, i
, 0, 1);
985 for (i
= 0; i
<= FD_NUMPART
; i
++)
986 check_one_slice(r
, diskname
, i
, 0, 1);
990 check_slices(avl_tree_t
*r
, int fd
, const char *sname
)
994 char diskname
[MAXNAMELEN
];
998 (void) strncpy(diskname
, sname
, MAXNAMELEN
);
999 if ((ptr
= strrchr(diskname
, 's')) == NULL
|| !isdigit(ptr
[1]))
1003 if (read_extvtoc(fd
, &vtoc
) >= 0) {
1004 for (i
= 0; i
< NDKMAP
; i
++)
1005 check_one_slice(r
, diskname
, i
,
1006 vtoc
.v_part
[i
].p_size
, vtoc
.v_sectorsz
);
1007 } else if (efi_alloc_and_read(fd
, &gpt
) >= 0) {
1009 * on x86 we'll still have leftover links that point
1010 * to slices s[9-15], so use NDKMAP instead
1012 for (i
= 0; i
< NDKMAP
; i
++)
1013 check_one_slice(r
, diskname
, i
,
1014 gpt
->efi_parts
[i
].p_size
, gpt
->efi_lbasize
);
1015 /* nodes p[1-4] are never used with EFI labels */
1017 for (i
= 1; i
<= FD_NUMPART
; i
++)
1018 check_one_slice(r
, diskname
, i
, 0, 1);
1024 zpool_open_func(void *arg
)
1026 rdsk_node_t
*rn
= arg
;
1027 struct stat64 statbuf
;
1033 if ((fd
= openat64(rn
->rn_dfd
, rn
->rn_name
, O_RDONLY
)) < 0) {
1034 /* symlink to a device that's no longer there */
1035 if (errno
== ENOENT
)
1036 nozpool_all_slices(rn
->rn_avl
, rn
->rn_name
);
1040 * Ignore failed stats. We only want regular
1041 * files, character devs and block devs.
1043 if (fstat64(fd
, &statbuf
) != 0 ||
1044 (!S_ISREG(statbuf
.st_mode
) &&
1045 !S_ISCHR(statbuf
.st_mode
) &&
1046 !S_ISBLK(statbuf
.st_mode
))) {
1050 /* this file is too small to hold a zpool */
1051 if (S_ISREG(statbuf
.st_mode
) &&
1052 statbuf
.st_size
< SPA_MINDEVSIZE
) {
1055 } else if (!S_ISREG(statbuf
.st_mode
)) {
1057 * Try to read the disk label first so we don't have to
1058 * open a bunch of minor nodes that can't have a zpool.
1060 check_slices(rn
->rn_avl
, fd
, rn
->rn_name
);
1063 if ((zpool_read_label(fd
, &config
)) != 0 && errno
== ENOMEM
) {
1065 (void) no_memory(rn
->rn_hdl
);
1070 rn
->rn_config
= config
;
1074 * Given a file descriptor, clear (zero) the label information.
1077 zpool_clear_label(int fd
)
1079 struct stat64 statbuf
;
1081 vdev_label_t
*label
;
1084 if (fstat64(fd
, &statbuf
) == -1)
1086 size
= P2ALIGN_TYPED(statbuf
.st_size
, sizeof (vdev_label_t
), uint64_t);
1088 if ((label
= calloc(sizeof (vdev_label_t
), 1)) == NULL
)
1091 for (l
= 0; l
< VDEV_LABELS
; l
++) {
1092 if (pwrite64(fd
, label
, sizeof (vdev_label_t
),
1093 label_offset(size
, l
)) != sizeof (vdev_label_t
)) {
1104 * Given a list of directories to search, find all pools stored on disk. This
1105 * includes partial pools which are not available to import. If no args are
1106 * given (argc is 0), then the default directory (/dev/dsk) is searched.
1107 * poolname or guid (but not both) are provided by the caller when trying
1108 * to import a specific pool.
1111 zpool_find_import_impl(libzfs_handle_t
*hdl
, importargs_t
*iarg
)
1113 int i
, dirs
= iarg
->paths
;
1114 struct dirent64
*dp
;
1115 char path
[MAXPATHLEN
];
1116 char *end
, **dir
= iarg
->path
;
1118 nvlist_t
*ret
= NULL
;
1119 static char *default_dir
= ZFS_DISK_ROOT
;
1120 pool_list_t pools
= { 0 };
1121 pool_entry_t
*pe
, *penext
;
1122 vdev_entry_t
*ve
, *venext
;
1123 config_entry_t
*ce
, *cenext
;
1124 name_entry_t
*ne
, *nenext
;
1125 avl_tree_t slice_cache
;
1135 * Go through and read the label configuration information from every
1136 * possible device, organizing the information according to pool GUID
1137 * and toplevel GUID.
1139 for (i
= 0; i
< dirs
; i
++) {
1141 char rdsk
[MAXPATHLEN
];
1143 boolean_t config_failed
= B_FALSE
;
1146 /* use realpath to normalize the path */
1147 if (realpath(dir
[i
], path
) == 0) {
1148 (void) zfs_error_fmt(hdl
, EZFS_BADPATH
,
1149 dgettext(TEXT_DOMAIN
, "cannot open '%s'"), dir
[i
]);
1152 end
= &path
[strlen(path
)];
1155 pathleft
= &path
[sizeof (path
)] - end
;
1158 * Using raw devices instead of block devices when we're
1159 * reading the labels skips a bunch of slow operations during
1160 * close(2) processing, so we replace /dev/dsk with /dev/rdsk.
1162 if (strcmp(path
, ZFS_DISK_ROOTD
) == 0)
1163 (void) strlcpy(rdsk
, ZFS_RDISK_ROOTD
, sizeof (rdsk
));
1165 (void) strlcpy(rdsk
, path
, sizeof (rdsk
));
1167 if ((dfd
= open64(rdsk
, O_RDONLY
)) < 0 ||
1168 (dirp
= fdopendir(dfd
)) == NULL
) {
1171 zfs_error_aux(hdl
, strerror(errno
));
1172 (void) zfs_error_fmt(hdl
, EZFS_BADPATH
,
1173 dgettext(TEXT_DOMAIN
, "cannot open '%s'"),
1178 avl_create(&slice_cache
, slice_cache_compare
,
1179 sizeof (rdsk_node_t
), offsetof(rdsk_node_t
, rn_node
));
1181 * This is not MT-safe, but we have no MT consumers of libzfs
1183 while ((dp
= readdir64(dirp
)) != NULL
) {
1184 const char *name
= dp
->d_name
;
1185 if (name
[0] == '.' &&
1186 (name
[1] == 0 || (name
[1] == '.' && name
[2] == 0)))
1189 slice
= zfs_alloc(hdl
, sizeof (rdsk_node_t
));
1190 slice
->rn_name
= zfs_strdup(hdl
, name
);
1191 slice
->rn_avl
= &slice_cache
;
1192 slice
->rn_dfd
= dfd
;
1193 slice
->rn_hdl
= hdl
;
1194 slice
->rn_nozpool
= B_FALSE
;
1195 avl_add(&slice_cache
, slice
);
1198 * create a thread pool to do all of this in parallel;
1199 * rn_nozpool is not protected, so this is racy in that
1200 * multiple tasks could decide that the same slice can
1201 * not hold a zpool, which is benign. Also choose
1202 * double the number of processors; we hold a lot of
1203 * locks in the kernel, so going beyond this doesn't
1206 t
= tpool_create(1, 2 * sysconf(_SC_NPROCESSORS_ONLN
),
1208 for (slice
= avl_first(&slice_cache
); slice
;
1209 (slice
= avl_walk(&slice_cache
, slice
,
1211 (void) tpool_dispatch(t
, zpool_open_func
, slice
);
1216 while ((slice
= avl_destroy_nodes(&slice_cache
,
1217 &cookie
)) != NULL
) {
1218 if (slice
->rn_config
!= NULL
&& !config_failed
) {
1219 nvlist_t
*config
= slice
->rn_config
;
1220 boolean_t matched
= B_TRUE
;
1222 if (iarg
->poolname
!= NULL
) {
1225 matched
= nvlist_lookup_string(config
,
1226 ZPOOL_CONFIG_POOL_NAME
,
1228 strcmp(iarg
->poolname
, pname
) == 0;
1229 } else if (iarg
->guid
!= 0) {
1232 matched
= nvlist_lookup_uint64(config
,
1233 ZPOOL_CONFIG_POOL_GUID
,
1235 iarg
->guid
== this_guid
;
1238 nvlist_free(config
);
1241 * use the non-raw path for the config
1243 (void) strlcpy(end
, slice
->rn_name
,
1245 if (add_config(hdl
, &pools
, path
,
1247 config_failed
= B_TRUE
;
1250 free(slice
->rn_name
);
1253 avl_destroy(&slice_cache
);
1255 (void) closedir(dirp
);
1261 ret
= get_configs(hdl
, &pools
, iarg
->can_be_active
, iarg
->policy
);
1264 for (pe
= pools
.pools
; pe
!= NULL
; pe
= penext
) {
1265 penext
= pe
->pe_next
;
1266 for (ve
= pe
->pe_vdevs
; ve
!= NULL
; ve
= venext
) {
1267 venext
= ve
->ve_next
;
1268 for (ce
= ve
->ve_configs
; ce
!= NULL
; ce
= cenext
) {
1269 cenext
= ce
->ce_next
;
1270 nvlist_free(ce
->ce_config
);
1278 for (ne
= pools
.names
; ne
!= NULL
; ne
= nenext
) {
1279 nenext
= ne
->ne_next
;
1288 zpool_find_import(libzfs_handle_t
*hdl
, int argc
, char **argv
)
1290 importargs_t iarg
= { 0 };
1295 return (zpool_find_import_impl(hdl
, &iarg
));
1299 * Given a cache file, return the contents as a list of importable pools.
1300 * poolname or guid (but not both) are provided by the caller when trying
1301 * to import a specific pool.
1304 zpool_find_import_cached(libzfs_handle_t
*hdl
, const char *cachefile
,
1305 char *poolname
, uint64_t guid
)
1309 struct stat64 statbuf
;
1310 nvlist_t
*raw
, *src
, *dst
;
1317 verify(poolname
== NULL
|| guid
== 0);
1319 if ((fd
= open(cachefile
, O_RDONLY
)) < 0) {
1320 zfs_error_aux(hdl
, "%s", strerror(errno
));
1321 (void) zfs_error(hdl
, EZFS_BADCACHE
,
1322 dgettext(TEXT_DOMAIN
, "failed to open cache file"));
1326 if (fstat64(fd
, &statbuf
) != 0) {
1327 zfs_error_aux(hdl
, "%s", strerror(errno
));
1329 (void) zfs_error(hdl
, EZFS_BADCACHE
,
1330 dgettext(TEXT_DOMAIN
, "failed to get size of cache file"));
1334 if ((buf
= zfs_alloc(hdl
, statbuf
.st_size
)) == NULL
) {
1339 if (read(fd
, buf
, statbuf
.st_size
) != statbuf
.st_size
) {
1342 (void) zfs_error(hdl
, EZFS_BADCACHE
,
1343 dgettext(TEXT_DOMAIN
,
1344 "failed to read cache file contents"));
1350 if (nvlist_unpack(buf
, statbuf
.st_size
, &raw
, 0) != 0) {
1352 (void) zfs_error(hdl
, EZFS_BADCACHE
,
1353 dgettext(TEXT_DOMAIN
,
1354 "invalid or corrupt cache file contents"));
1361 * Go through and get the current state of the pools and refresh their
1364 if (nvlist_alloc(&pools
, 0, 0) != 0) {
1365 (void) no_memory(hdl
);
1371 while ((elem
= nvlist_next_nvpair(raw
, elem
)) != NULL
) {
1372 src
= fnvpair_value_nvlist(elem
);
1374 name
= fnvlist_lookup_string(src
, ZPOOL_CONFIG_POOL_NAME
);
1375 if (poolname
!= NULL
&& strcmp(poolname
, name
) != 0)
1378 this_guid
= fnvlist_lookup_uint64(src
, ZPOOL_CONFIG_POOL_GUID
);
1379 if (guid
!= 0 && guid
!= this_guid
)
1382 if (pool_active(hdl
, name
, this_guid
, &active
) != 0) {
1391 if (nvlist_add_string(src
, ZPOOL_CONFIG_CACHEFILE
,
1393 (void) no_memory(hdl
);
1399 if ((dst
= refresh_config(hdl
, src
)) == NULL
) {
1405 if (nvlist_add_nvlist(pools
, nvpair_name(elem
), dst
) != 0) {
1406 (void) no_memory(hdl
);
1420 name_or_guid_exists(zpool_handle_t
*zhp
, void *data
)
1422 importargs_t
*import
= data
;
1425 if (import
->poolname
!= NULL
) {
1428 verify(nvlist_lookup_string(zhp
->zpool_config
,
1429 ZPOOL_CONFIG_POOL_NAME
, &pool_name
) == 0);
1430 if (strcmp(pool_name
, import
->poolname
) == 0)
1435 verify(nvlist_lookup_uint64(zhp
->zpool_config
,
1436 ZPOOL_CONFIG_POOL_GUID
, &pool_guid
) == 0);
1437 if (pool_guid
== import
->guid
)
1446 zpool_search_import(libzfs_handle_t
*hdl
, importargs_t
*import
)
1448 verify(import
->poolname
== NULL
|| import
->guid
== 0);
1451 import
->exists
= zpool_iter(hdl
, name_or_guid_exists
, import
);
1453 if (import
->cachefile
!= NULL
)
1454 return (zpool_find_import_cached(hdl
, import
->cachefile
,
1455 import
->poolname
, import
->guid
));
1457 return (zpool_find_import_impl(hdl
, import
));
1461 find_guid(nvlist_t
*nv
, uint64_t guid
)
1467 verify(nvlist_lookup_uint64(nv
, ZPOOL_CONFIG_GUID
, &tmp
) == 0);
1471 if (nvlist_lookup_nvlist_array(nv
, ZPOOL_CONFIG_CHILDREN
,
1472 &child
, &children
) == 0) {
1473 for (c
= 0; c
< children
; c
++)
1474 if (find_guid(child
[c
], guid
))
1481 typedef struct aux_cbdata
{
1482 const char *cb_type
;
1484 zpool_handle_t
*cb_zhp
;
1488 find_aux(zpool_handle_t
*zhp
, void *data
)
1490 aux_cbdata_t
*cbp
= data
;
1496 verify(nvlist_lookup_nvlist(zhp
->zpool_config
, ZPOOL_CONFIG_VDEV_TREE
,
1499 if (nvlist_lookup_nvlist_array(nvroot
, cbp
->cb_type
,
1500 &list
, &count
) == 0) {
1501 for (i
= 0; i
< count
; i
++) {
1502 verify(nvlist_lookup_uint64(list
[i
],
1503 ZPOOL_CONFIG_GUID
, &guid
) == 0);
1504 if (guid
== cbp
->cb_guid
) {
1516 * Determines if the pool is in use. If so, it returns true and the state of
1517 * the pool as well as the name of the pool. Both strings are allocated and
1518 * must be freed by the caller.
1521 zpool_in_use(libzfs_handle_t
*hdl
, int fd
, pool_state_t
*state
, char **namestr
,
1527 uint64_t guid
, vdev_guid
;
1528 zpool_handle_t
*zhp
;
1529 nvlist_t
*pool_config
;
1530 uint64_t stateval
, isspare
;
1531 aux_cbdata_t cb
= { 0 };
1536 if (zpool_read_label(fd
, &config
) != 0 && errno
== ENOMEM
) {
1537 (void) no_memory(hdl
);
1544 verify(nvlist_lookup_uint64(config
, ZPOOL_CONFIG_POOL_STATE
,
1546 verify(nvlist_lookup_uint64(config
, ZPOOL_CONFIG_GUID
,
1549 if (stateval
!= POOL_STATE_SPARE
&& stateval
!= POOL_STATE_L2CACHE
) {
1550 verify(nvlist_lookup_string(config
, ZPOOL_CONFIG_POOL_NAME
,
1552 verify(nvlist_lookup_uint64(config
, ZPOOL_CONFIG_POOL_GUID
,
1557 case POOL_STATE_EXPORTED
:
1559 * A pool with an exported state may in fact be imported
1560 * read-only, so check the in-core state to see if it's
1561 * active and imported read-only. If it is, set
1562 * its state to active.
1564 if (pool_active(hdl
, name
, guid
, &isactive
) == 0 && isactive
&&
1565 (zhp
= zpool_open_canfail(hdl
, name
)) != NULL
) {
1566 if (zpool_get_prop_int(zhp
, ZPOOL_PROP_READONLY
, NULL
))
1567 stateval
= POOL_STATE_ACTIVE
;
1570 * All we needed the zpool handle for is the
1571 * readonly prop check.
1579 case POOL_STATE_ACTIVE
:
1581 * For an active pool, we have to determine if it's really part
1582 * of a currently active pool (in which case the pool will exist
1583 * and the guid will be the same), or whether it's part of an
1584 * active pool that was disconnected without being explicitly
1587 if (pool_active(hdl
, name
, guid
, &isactive
) != 0) {
1588 nvlist_free(config
);
1594 * Because the device may have been removed while
1595 * offlined, we only report it as active if the vdev is
1596 * still present in the config. Otherwise, pretend like
1599 if ((zhp
= zpool_open_canfail(hdl
, name
)) != NULL
&&
1600 (pool_config
= zpool_get_config(zhp
, NULL
))
1604 verify(nvlist_lookup_nvlist(pool_config
,
1605 ZPOOL_CONFIG_VDEV_TREE
, &nvroot
) == 0);
1606 ret
= find_guid(nvroot
, vdev_guid
);
1612 * If this is an active spare within another pool, we
1613 * treat it like an unused hot spare. This allows the
1614 * user to create a pool with a hot spare that currently
1615 * in use within another pool. Since we return B_TRUE,
1616 * libdiskmgt will continue to prevent generic consumers
1617 * from using the device.
1619 if (ret
&& nvlist_lookup_uint64(config
,
1620 ZPOOL_CONFIG_IS_SPARE
, &isspare
) == 0 && isspare
)
1621 stateval
= POOL_STATE_SPARE
;
1626 stateval
= POOL_STATE_POTENTIALLY_ACTIVE
;
1631 case POOL_STATE_SPARE
:
1633 * For a hot spare, it can be either definitively in use, or
1634 * potentially active. To determine if it's in use, we iterate
1635 * over all pools in the system and search for one with a spare
1636 * with a matching guid.
1638 * Due to the shared nature of spares, we don't actually report
1639 * the potentially active case as in use. This means the user
1640 * can freely create pools on the hot spares of exported pools,
1641 * but to do otherwise makes the resulting code complicated, and
1642 * we end up having to deal with this case anyway.
1645 cb
.cb_guid
= vdev_guid
;
1646 cb
.cb_type
= ZPOOL_CONFIG_SPARES
;
1647 if (zpool_iter(hdl
, find_aux
, &cb
) == 1) {
1648 name
= (char *)zpool_get_name(cb
.cb_zhp
);
1655 case POOL_STATE_L2CACHE
:
1658 * Check if any pool is currently using this l2cache device.
1661 cb
.cb_guid
= vdev_guid
;
1662 cb
.cb_type
= ZPOOL_CONFIG_L2CACHE
;
1663 if (zpool_iter(hdl
, find_aux
, &cb
) == 1) {
1664 name
= (char *)zpool_get_name(cb
.cb_zhp
);
1677 if ((*namestr
= zfs_strdup(hdl
, name
)) == NULL
) {
1679 zpool_close(cb
.cb_zhp
);
1680 nvlist_free(config
);
1683 *state
= (pool_state_t
)stateval
;
1687 zpool_close(cb
.cb_zhp
);
1689 nvlist_free(config
);