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, 2015 by Delphix. All rights reserved.
25 * Copyright 2015 RackTop Systems.
26 * Copyright 2016 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 get_devid(const char *path
)
105 if ((fd
= open(path
, O_RDONLY
)) < 0)
110 if (devid_get(fd
, &devid
) == 0) {
111 if (devid_get_minor_name(fd
, &minor
) == 0)
112 ret
= devid_str_encode(devid
, minor
);
114 devid_str_free(minor
);
124 * Go through and fix up any path and/or devid information for the given vdev
128 fix_paths(nvlist_t
*nv
, name_entry_t
*names
)
133 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 slices or multiple paths to the same
152 * disk. If this is the case, then we want to pick the path that is
153 * the most similar to the original, where "most similar" is the number
154 * of matching characters starting from the end of the path. This will
155 * preserve slice numbers even if the disks have been reorganized, and
156 * will also catch preferred disk names if multiple paths exist.
158 verify(nvlist_lookup_uint64(nv
, ZPOOL_CONFIG_GUID
, &guid
) == 0);
159 if (nvlist_lookup_string(nv
, ZPOOL_CONFIG_PATH
, &path
) != 0)
164 for (ne
= names
; ne
!= NULL
; ne
= ne
->ne_next
) {
165 if (ne
->ne_guid
== guid
) {
166 const char *src
, *dst
;
174 src
= ne
->ne_name
+ strlen(ne
->ne_name
) - 1;
175 dst
= path
+ strlen(path
) - 1;
176 for (count
= 0; src
>= ne
->ne_name
&& dst
>= path
;
177 src
--, dst
--, count
++)
182 * At this point, 'count' is the number of characters
183 * matched from the end.
185 if (count
> matched
|| best
== NULL
) {
195 if (nvlist_add_string(nv
, ZPOOL_CONFIG_PATH
, best
->ne_name
) != 0)
198 if ((devid
= get_devid(best
->ne_name
)) == NULL
) {
199 (void) nvlist_remove_all(nv
, ZPOOL_CONFIG_DEVID
);
201 if (nvlist_add_string(nv
, ZPOOL_CONFIG_DEVID
, devid
) != 0) {
202 devid_str_free(devid
);
205 devid_str_free(devid
);
212 * Add the given configuration to the list of known devices.
215 add_config(libzfs_handle_t
*hdl
, pool_list_t
*pl
, const char *path
,
218 uint64_t pool_guid
, vdev_guid
, top_guid
, txg
, state
;
225 * If this is a hot spare not currently in use or level 2 cache
226 * device, add it to the list of names to translate, but don't do
229 if (nvlist_lookup_uint64(config
, ZPOOL_CONFIG_POOL_STATE
,
231 (state
== POOL_STATE_SPARE
|| state
== POOL_STATE_L2CACHE
) &&
232 nvlist_lookup_uint64(config
, ZPOOL_CONFIG_GUID
, &vdev_guid
) == 0) {
233 if ((ne
= zfs_alloc(hdl
, sizeof (name_entry_t
))) == NULL
)
236 if ((ne
->ne_name
= zfs_strdup(hdl
, path
)) == NULL
) {
240 ne
->ne_guid
= vdev_guid
;
241 ne
->ne_next
= pl
->names
;
247 * If we have a valid config but cannot read any of these fields, then
248 * it means we have a half-initialized label. In vdev_label_init()
249 * we write a label with txg == 0 so that we can identify the device
250 * in case the user refers to the same disk later on. If we fail to
251 * create the pool, we'll be left with a label in this state
252 * which should not be considered part of a valid pool.
254 if (nvlist_lookup_uint64(config
, ZPOOL_CONFIG_POOL_GUID
,
256 nvlist_lookup_uint64(config
, ZPOOL_CONFIG_GUID
,
258 nvlist_lookup_uint64(config
, ZPOOL_CONFIG_TOP_GUID
,
260 nvlist_lookup_uint64(config
, ZPOOL_CONFIG_POOL_TXG
,
261 &txg
) != 0 || txg
== 0) {
267 * First, see if we know about this pool. If not, then add it to the
268 * list of known pools.
270 for (pe
= pl
->pools
; pe
!= NULL
; pe
= pe
->pe_next
) {
271 if (pe
->pe_guid
== pool_guid
)
276 if ((pe
= zfs_alloc(hdl
, sizeof (pool_entry_t
))) == NULL
) {
280 pe
->pe_guid
= pool_guid
;
281 pe
->pe_next
= pl
->pools
;
286 * Second, see if we know about this toplevel vdev. Add it if its
289 for (ve
= pe
->pe_vdevs
; ve
!= NULL
; ve
= ve
->ve_next
) {
290 if (ve
->ve_guid
== top_guid
)
295 if ((ve
= zfs_alloc(hdl
, sizeof (vdev_entry_t
))) == NULL
) {
299 ve
->ve_guid
= top_guid
;
300 ve
->ve_next
= pe
->pe_vdevs
;
305 * Third, see if we have a config with a matching transaction group. If
306 * so, then we do nothing. Otherwise, add it to the list of known
309 for (ce
= ve
->ve_configs
; ce
!= NULL
; ce
= ce
->ce_next
) {
310 if (ce
->ce_txg
== txg
)
315 if ((ce
= zfs_alloc(hdl
, sizeof (config_entry_t
))) == NULL
) {
320 ce
->ce_config
= config
;
321 ce
->ce_next
= ve
->ve_configs
;
328 * At this point we've successfully added our config to the list of
329 * known configs. The last thing to do is add the vdev guid -> path
330 * mappings so that we can fix up the configuration as necessary before
333 if ((ne
= zfs_alloc(hdl
, sizeof (name_entry_t
))) == NULL
)
336 if ((ne
->ne_name
= zfs_strdup(hdl
, path
)) == NULL
) {
341 ne
->ne_guid
= vdev_guid
;
342 ne
->ne_next
= pl
->names
;
349 * Returns true if the named pool matches the given GUID.
352 pool_active(libzfs_handle_t
*hdl
, const char *name
, uint64_t guid
,
358 if (zpool_open_silent(hdl
, name
, &zhp
) != 0)
366 verify(nvlist_lookup_uint64(zhp
->zpool_config
, ZPOOL_CONFIG_POOL_GUID
,
371 *isactive
= (theguid
== guid
);
376 refresh_config(libzfs_handle_t
*hdl
, nvlist_t
*config
)
379 zfs_cmd_t zc
= { 0 };
382 if (zcmd_write_conf_nvlist(hdl
, &zc
, config
) != 0)
385 if (zcmd_alloc_dst_nvlist(hdl
, &zc
,
386 zc
.zc_nvlist_conf_size
* 2) != 0) {
387 zcmd_free_nvlists(&zc
);
391 while ((err
= ioctl(hdl
->libzfs_fd
, ZFS_IOC_POOL_TRYIMPORT
,
392 &zc
)) != 0 && errno
== ENOMEM
) {
393 if (zcmd_expand_dst_nvlist(hdl
, &zc
) != 0) {
394 zcmd_free_nvlists(&zc
);
400 zcmd_free_nvlists(&zc
);
404 if (zcmd_read_dst_nvlist(hdl
, &zc
, &nvl
) != 0) {
405 zcmd_free_nvlists(&zc
);
409 zcmd_free_nvlists(&zc
);
414 * Determine if the vdev id is a hole in the namespace.
417 vdev_is_hole(uint64_t *hole_array
, uint_t holes
, uint_t id
)
419 for (int c
= 0; c
< holes
; c
++) {
421 /* Top-level is a hole */
422 if (hole_array
[c
] == id
)
429 * Convert our list of pools into the definitive set of configurations. We
430 * start by picking the best config for each toplevel vdev. Once that's done,
431 * we assemble the toplevel vdevs into a full config for the pool. We make a
432 * pass to fix up any incorrect paths, and then add it to the main list to
433 * return to the user.
436 get_configs(libzfs_handle_t
*hdl
, pool_list_t
*pl
, boolean_t active_ok
)
441 nvlist_t
*ret
= NULL
, *config
= NULL
, *tmp
= NULL
, *nvtop
, *nvroot
;
442 nvlist_t
**spares
, **l2cache
;
443 uint_t i
, nspares
, nl2cache
;
444 boolean_t config_seen
;
446 char *name
, *hostname
= NULL
;
449 nvlist_t
**child
= NULL
;
451 uint64_t *hole_array
, max_id
;
456 boolean_t found_one
= B_FALSE
;
457 boolean_t valid_top_config
= B_FALSE
;
459 if (nvlist_alloc(&ret
, 0, 0) != 0)
462 for (pe
= pl
->pools
; pe
!= NULL
; pe
= pe
->pe_next
) {
463 uint64_t id
, max_txg
= 0;
465 if (nvlist_alloc(&config
, NV_UNIQUE_NAME
, 0) != 0)
467 config_seen
= B_FALSE
;
470 * Iterate over all toplevel vdevs. Grab the pool configuration
471 * from the first one we find, and then go through the rest and
472 * add them as necessary to the 'vdevs' member of the config.
474 for (ve
= pe
->pe_vdevs
; ve
!= NULL
; ve
= ve
->ve_next
) {
477 * Determine the best configuration for this vdev by
478 * selecting the config with the latest transaction
482 for (ce
= ve
->ve_configs
; ce
!= NULL
;
485 if (ce
->ce_txg
> best_txg
) {
487 best_txg
= ce
->ce_txg
;
492 * We rely on the fact that the max txg for the
493 * pool will contain the most up-to-date information
494 * about the valid top-levels in the vdev namespace.
496 if (best_txg
> max_txg
) {
497 (void) nvlist_remove(config
,
498 ZPOOL_CONFIG_VDEV_CHILDREN
,
500 (void) nvlist_remove(config
,
501 ZPOOL_CONFIG_HOLE_ARRAY
,
502 DATA_TYPE_UINT64_ARRAY
);
508 valid_top_config
= B_FALSE
;
510 if (nvlist_lookup_uint64(tmp
,
511 ZPOOL_CONFIG_VDEV_CHILDREN
, &max_id
) == 0) {
512 verify(nvlist_add_uint64(config
,
513 ZPOOL_CONFIG_VDEV_CHILDREN
,
515 valid_top_config
= B_TRUE
;
518 if (nvlist_lookup_uint64_array(tmp
,
519 ZPOOL_CONFIG_HOLE_ARRAY
, &hole_array
,
521 verify(nvlist_add_uint64_array(config
,
522 ZPOOL_CONFIG_HOLE_ARRAY
,
523 hole_array
, holes
) == 0);
529 * Copy the relevant pieces of data to the pool
535 * comment (if available)
537 * hostid (if available)
538 * hostname (if available)
540 uint64_t state
, version
;
541 char *comment
= NULL
;
543 version
= fnvlist_lookup_uint64(tmp
,
544 ZPOOL_CONFIG_VERSION
);
545 fnvlist_add_uint64(config
,
546 ZPOOL_CONFIG_VERSION
, version
);
547 guid
= fnvlist_lookup_uint64(tmp
,
548 ZPOOL_CONFIG_POOL_GUID
);
549 fnvlist_add_uint64(config
,
550 ZPOOL_CONFIG_POOL_GUID
, guid
);
551 name
= fnvlist_lookup_string(tmp
,
552 ZPOOL_CONFIG_POOL_NAME
);
553 fnvlist_add_string(config
,
554 ZPOOL_CONFIG_POOL_NAME
, name
);
556 if (nvlist_lookup_string(tmp
,
557 ZPOOL_CONFIG_COMMENT
, &comment
) == 0)
558 fnvlist_add_string(config
,
559 ZPOOL_CONFIG_COMMENT
, comment
);
561 state
= fnvlist_lookup_uint64(tmp
,
562 ZPOOL_CONFIG_POOL_STATE
);
563 fnvlist_add_uint64(config
,
564 ZPOOL_CONFIG_POOL_STATE
, state
);
567 if (nvlist_lookup_uint64(tmp
,
568 ZPOOL_CONFIG_HOSTID
, &hostid
) == 0) {
569 fnvlist_add_uint64(config
,
570 ZPOOL_CONFIG_HOSTID
, hostid
);
571 hostname
= fnvlist_lookup_string(tmp
,
572 ZPOOL_CONFIG_HOSTNAME
);
573 fnvlist_add_string(config
,
574 ZPOOL_CONFIG_HOSTNAME
, hostname
);
577 config_seen
= B_TRUE
;
581 * Add this top-level vdev to the child array.
583 verify(nvlist_lookup_nvlist(tmp
,
584 ZPOOL_CONFIG_VDEV_TREE
, &nvtop
) == 0);
585 verify(nvlist_lookup_uint64(nvtop
, ZPOOL_CONFIG_ID
,
588 if (id
>= children
) {
591 newchild
= zfs_alloc(hdl
, (id
+ 1) *
592 sizeof (nvlist_t
*));
593 if (newchild
== NULL
)
596 for (c
= 0; c
< children
; c
++)
597 newchild
[c
] = child
[c
];
603 if (nvlist_dup(nvtop
, &child
[id
], 0) != 0)
609 * If we have information about all the top-levels then
610 * clean up the nvlist which we've constructed. This
611 * means removing any extraneous devices that are
612 * beyond the valid range or adding devices to the end
613 * of our array which appear to be missing.
615 if (valid_top_config
) {
616 if (max_id
< children
) {
617 for (c
= max_id
; c
< children
; c
++)
618 nvlist_free(child
[c
]);
620 } else if (max_id
> children
) {
623 newchild
= zfs_alloc(hdl
, (max_id
) *
624 sizeof (nvlist_t
*));
625 if (newchild
== NULL
)
628 for (c
= 0; c
< children
; c
++)
629 newchild
[c
] = child
[c
];
637 verify(nvlist_lookup_uint64(config
, ZPOOL_CONFIG_POOL_GUID
,
641 * The vdev namespace may contain holes as a result of
642 * device removal. We must add them back into the vdev
643 * tree before we process any missing devices.
646 ASSERT(valid_top_config
);
648 for (c
= 0; c
< children
; c
++) {
651 if (child
[c
] != NULL
||
652 !vdev_is_hole(hole_array
, holes
, c
))
655 if (nvlist_alloc(&holey
, NV_UNIQUE_NAME
,
660 * Holes in the namespace are treated as
661 * "hole" top-level vdevs and have a
662 * special flag set on them.
664 if (nvlist_add_string(holey
,
666 VDEV_TYPE_HOLE
) != 0 ||
667 nvlist_add_uint64(holey
,
668 ZPOOL_CONFIG_ID
, c
) != 0 ||
669 nvlist_add_uint64(holey
,
670 ZPOOL_CONFIG_GUID
, 0ULL) != 0) {
679 * Look for any missing top-level vdevs. If this is the case,
680 * create a faked up 'missing' vdev as a placeholder. We cannot
681 * simply compress the child array, because the kernel performs
682 * certain checks to make sure the vdev IDs match their location
683 * in the configuration.
685 for (c
= 0; c
< children
; c
++) {
686 if (child
[c
] == NULL
) {
688 if (nvlist_alloc(&missing
, NV_UNIQUE_NAME
,
691 if (nvlist_add_string(missing
,
693 VDEV_TYPE_MISSING
) != 0 ||
694 nvlist_add_uint64(missing
,
695 ZPOOL_CONFIG_ID
, c
) != 0 ||
696 nvlist_add_uint64(missing
,
697 ZPOOL_CONFIG_GUID
, 0ULL) != 0) {
698 nvlist_free(missing
);
706 * Put all of this pool's top-level vdevs into a root vdev.
708 if (nvlist_alloc(&nvroot
, NV_UNIQUE_NAME
, 0) != 0)
710 if (nvlist_add_string(nvroot
, ZPOOL_CONFIG_TYPE
,
711 VDEV_TYPE_ROOT
) != 0 ||
712 nvlist_add_uint64(nvroot
, ZPOOL_CONFIG_ID
, 0ULL) != 0 ||
713 nvlist_add_uint64(nvroot
, ZPOOL_CONFIG_GUID
, guid
) != 0 ||
714 nvlist_add_nvlist_array(nvroot
, ZPOOL_CONFIG_CHILDREN
,
715 child
, children
) != 0) {
720 for (c
= 0; c
< children
; c
++)
721 nvlist_free(child
[c
]);
727 * Go through and fix up any paths and/or devids based on our
728 * known list of vdev GUID -> path mappings.
730 if (fix_paths(nvroot
, pl
->names
) != 0) {
736 * Add the root vdev to this pool's configuration.
738 if (nvlist_add_nvlist(config
, ZPOOL_CONFIG_VDEV_TREE
,
746 * zdb uses this path to report on active pools that were
747 * imported or created using -R.
753 * Determine if this pool is currently active, in which case we
754 * can't actually import it.
756 verify(nvlist_lookup_string(config
, ZPOOL_CONFIG_POOL_NAME
,
758 verify(nvlist_lookup_uint64(config
, ZPOOL_CONFIG_POOL_GUID
,
761 if (pool_active(hdl
, name
, guid
, &isactive
) != 0)
770 if ((nvl
= refresh_config(hdl
, config
)) == NULL
) {
780 * Go through and update the paths for spares, now that we have
783 verify(nvlist_lookup_nvlist(config
, ZPOOL_CONFIG_VDEV_TREE
,
785 if (nvlist_lookup_nvlist_array(nvroot
, ZPOOL_CONFIG_SPARES
,
786 &spares
, &nspares
) == 0) {
787 for (i
= 0; i
< nspares
; i
++) {
788 if (fix_paths(spares
[i
], pl
->names
) != 0)
794 * Update the paths for l2cache devices.
796 if (nvlist_lookup_nvlist_array(nvroot
, ZPOOL_CONFIG_L2CACHE
,
797 &l2cache
, &nl2cache
) == 0) {
798 for (i
= 0; i
< nl2cache
; i
++) {
799 if (fix_paths(l2cache
[i
], pl
->names
) != 0)
805 * Restore the original information read from the actual label.
807 (void) nvlist_remove(config
, ZPOOL_CONFIG_HOSTID
,
809 (void) nvlist_remove(config
, ZPOOL_CONFIG_HOSTNAME
,
812 verify(nvlist_add_uint64(config
, ZPOOL_CONFIG_HOSTID
,
814 verify(nvlist_add_string(config
, ZPOOL_CONFIG_HOSTNAME
,
820 * Add this pool to the list of configs.
822 verify(nvlist_lookup_string(config
, ZPOOL_CONFIG_POOL_NAME
,
824 if (nvlist_add_nvlist(ret
, name
, config
) != 0)
840 (void) no_memory(hdl
);
844 for (c
= 0; c
< children
; c
++)
845 nvlist_free(child
[c
]);
852 * Return the offset of the given label.
855 label_offset(uint64_t size
, int l
)
857 ASSERT(P2PHASE_TYPED(size
, sizeof (vdev_label_t
), uint64_t) == 0);
858 return (l
* sizeof (vdev_label_t
) + (l
< VDEV_LABELS
/ 2 ?
859 0 : size
- VDEV_LABELS
* sizeof (vdev_label_t
)));
863 * Given a file descriptor, read the label information and return an nvlist
864 * describing the configuration, if there is one.
867 zpool_read_label(int fd
, nvlist_t
**config
)
869 struct stat64 statbuf
;
872 uint64_t state
, txg
, size
;
876 if (fstat64(fd
, &statbuf
) == -1)
878 size
= P2ALIGN_TYPED(statbuf
.st_size
, sizeof (vdev_label_t
), uint64_t);
880 if ((label
= malloc(sizeof (vdev_label_t
))) == NULL
)
883 for (l
= 0; l
< VDEV_LABELS
; l
++) {
884 if (pread64(fd
, label
, sizeof (vdev_label_t
),
885 label_offset(size
, l
)) != sizeof (vdev_label_t
))
888 if (nvlist_unpack(label
->vl_vdev_phys
.vp_nvlist
,
889 sizeof (label
->vl_vdev_phys
.vp_nvlist
), config
, 0) != 0)
892 if (nvlist_lookup_uint64(*config
, ZPOOL_CONFIG_POOL_STATE
,
893 &state
) != 0 || state
> POOL_STATE_L2CACHE
) {
894 nvlist_free(*config
);
898 if (state
!= POOL_STATE_SPARE
&& state
!= POOL_STATE_L2CACHE
&&
899 (nvlist_lookup_uint64(*config
, ZPOOL_CONFIG_POOL_TXG
,
900 &txg
) != 0 || txg
== 0)) {
901 nvlist_free(*config
);
914 typedef struct rdsk_node
{
917 libzfs_handle_t
*rn_hdl
;
921 boolean_t rn_nozpool
;
925 slice_cache_compare(const void *arg1
, const void *arg2
)
927 const char *nm1
= ((rdsk_node_t
*)arg1
)->rn_name
;
928 const char *nm2
= ((rdsk_node_t
*)arg2
)->rn_name
;
929 char *nm1slice
, *nm2slice
;
933 * slices zero and two are the most likely to provide results,
936 nm1slice
= strstr(nm1
, "s0");
937 nm2slice
= strstr(nm2
, "s0");
938 if (nm1slice
&& !nm2slice
) {
941 if (!nm1slice
&& nm2slice
) {
944 nm1slice
= strstr(nm1
, "s2");
945 nm2slice
= strstr(nm2
, "s2");
946 if (nm1slice
&& !nm2slice
) {
949 if (!nm1slice
&& nm2slice
) {
953 rv
= strcmp(nm1
, nm2
);
956 return (rv
> 0 ? 1 : -1);
960 check_one_slice(avl_tree_t
*r
, char *diskname
, uint_t partno
,
961 diskaddr_t size
, uint_t blksz
)
965 char sname
[MAXNAMELEN
];
967 tmpnode
.rn_name
= &sname
[0];
968 (void) snprintf(tmpnode
.rn_name
, MAXNAMELEN
, "%s%u",
971 * protect against division by zero for disk labels that
972 * contain a bogus sector size
976 /* too small to contain a zpool? */
977 if ((size
< (SPA_MINDEVSIZE
/ blksz
)) &&
978 (node
= avl_find(r
, &tmpnode
, NULL
)))
979 node
->rn_nozpool
= B_TRUE
;
983 nozpool_all_slices(avl_tree_t
*r
, const char *sname
)
985 char diskname
[MAXNAMELEN
];
989 (void) strncpy(diskname
, sname
, MAXNAMELEN
);
990 if (((ptr
= strrchr(diskname
, 's')) == NULL
) &&
991 ((ptr
= strrchr(diskname
, 'p')) == NULL
))
995 for (i
= 0; i
< NDKMAP
; i
++)
996 check_one_slice(r
, diskname
, i
, 0, 1);
998 for (i
= 0; i
<= FD_NUMPART
; i
++)
999 check_one_slice(r
, diskname
, i
, 0, 1);
1003 check_slices(avl_tree_t
*r
, int fd
, const char *sname
)
1005 struct extvtoc vtoc
;
1007 char diskname
[MAXNAMELEN
];
1011 (void) strncpy(diskname
, sname
, MAXNAMELEN
);
1012 if ((ptr
= strrchr(diskname
, 's')) == NULL
|| !isdigit(ptr
[1]))
1016 if (read_extvtoc(fd
, &vtoc
) >= 0) {
1017 for (i
= 0; i
< NDKMAP
; i
++)
1018 check_one_slice(r
, diskname
, i
,
1019 vtoc
.v_part
[i
].p_size
, vtoc
.v_sectorsz
);
1020 } else if (efi_alloc_and_read(fd
, &gpt
) >= 0) {
1022 * on x86 we'll still have leftover links that point
1023 * to slices s[9-15], so use NDKMAP instead
1025 for (i
= 0; i
< NDKMAP
; i
++)
1026 check_one_slice(r
, diskname
, i
,
1027 gpt
->efi_parts
[i
].p_size
, gpt
->efi_lbasize
);
1028 /* nodes p[1-4] are never used with EFI labels */
1030 for (i
= 1; i
<= FD_NUMPART
; i
++)
1031 check_one_slice(r
, diskname
, i
, 0, 1);
1037 zpool_open_func(void *arg
)
1039 rdsk_node_t
*rn
= arg
;
1040 struct stat64 statbuf
;
1046 if ((fd
= openat64(rn
->rn_dfd
, rn
->rn_name
, O_RDONLY
)) < 0) {
1047 /* symlink to a device that's no longer there */
1048 if (errno
== ENOENT
)
1049 nozpool_all_slices(rn
->rn_avl
, rn
->rn_name
);
1053 * Ignore failed stats. We only want regular
1054 * files, character devs and block devs.
1056 if (fstat64(fd
, &statbuf
) != 0 ||
1057 (!S_ISREG(statbuf
.st_mode
) &&
1058 !S_ISCHR(statbuf
.st_mode
) &&
1059 !S_ISBLK(statbuf
.st_mode
))) {
1063 /* this file is too small to hold a zpool */
1064 if (S_ISREG(statbuf
.st_mode
) &&
1065 statbuf
.st_size
< SPA_MINDEVSIZE
) {
1068 } else if (!S_ISREG(statbuf
.st_mode
)) {
1070 * Try to read the disk label first so we don't have to
1071 * open a bunch of minor nodes that can't have a zpool.
1073 check_slices(rn
->rn_avl
, fd
, rn
->rn_name
);
1076 if ((zpool_read_label(fd
, &config
)) != 0) {
1078 (void) no_memory(rn
->rn_hdl
);
1083 rn
->rn_config
= config
;
1087 * Given a file descriptor, clear (zero) the label information.
1090 zpool_clear_label(int fd
)
1092 struct stat64 statbuf
;
1094 vdev_label_t
*label
;
1097 if (fstat64(fd
, &statbuf
) == -1)
1099 size
= P2ALIGN_TYPED(statbuf
.st_size
, sizeof (vdev_label_t
), uint64_t);
1101 if ((label
= calloc(sizeof (vdev_label_t
), 1)) == NULL
)
1104 for (l
= 0; l
< VDEV_LABELS
; l
++) {
1105 if (pwrite64(fd
, label
, sizeof (vdev_label_t
),
1106 label_offset(size
, l
)) != sizeof (vdev_label_t
)) {
1117 * Given a list of directories to search, find all pools stored on disk. This
1118 * includes partial pools which are not available to import. If no args are
1119 * given (argc is 0), then the default directory (/dev/dsk) is searched.
1120 * poolname or guid (but not both) are provided by the caller when trying
1121 * to import a specific pool.
1124 zpool_find_import_impl(libzfs_handle_t
*hdl
, importargs_t
*iarg
)
1126 int i
, dirs
= iarg
->paths
;
1127 struct dirent64
*dp
;
1128 char path
[MAXPATHLEN
];
1129 char *end
, **dir
= iarg
->path
;
1131 nvlist_t
*ret
= NULL
;
1132 static char *default_dir
= ZFS_DISK_ROOT
;
1133 pool_list_t pools
= { 0 };
1134 pool_entry_t
*pe
, *penext
;
1135 vdev_entry_t
*ve
, *venext
;
1136 config_entry_t
*ce
, *cenext
;
1137 name_entry_t
*ne
, *nenext
;
1138 avl_tree_t slice_cache
;
1148 * Go through and read the label configuration information from every
1149 * possible device, organizing the information according to pool GUID
1150 * and toplevel GUID.
1152 for (i
= 0; i
< dirs
; i
++) {
1154 char rdsk
[MAXPATHLEN
];
1156 boolean_t config_failed
= B_FALSE
;
1159 /* use realpath to normalize the path */
1160 if (realpath(dir
[i
], path
) == 0) {
1161 (void) zfs_error_fmt(hdl
, EZFS_BADPATH
,
1162 dgettext(TEXT_DOMAIN
, "cannot open '%s'"), dir
[i
]);
1165 end
= &path
[strlen(path
)];
1168 pathleft
= &path
[sizeof (path
)] - end
;
1171 * Using raw devices instead of block devices when we're
1172 * reading the labels skips a bunch of slow operations during
1173 * close(2) processing, so we replace /dev/dsk with /dev/rdsk.
1175 if (strcmp(path
, ZFS_DISK_ROOTD
) == 0)
1176 (void) strlcpy(rdsk
, ZFS_RDISK_ROOTD
, sizeof (rdsk
));
1178 (void) strlcpy(rdsk
, path
, sizeof (rdsk
));
1180 if ((dfd
= open64(rdsk
, O_RDONLY
)) < 0 ||
1181 (dirp
= fdopendir(dfd
)) == NULL
) {
1184 zfs_error_aux(hdl
, strerror(errno
));
1185 (void) zfs_error_fmt(hdl
, EZFS_BADPATH
,
1186 dgettext(TEXT_DOMAIN
, "cannot open '%s'"),
1191 avl_create(&slice_cache
, slice_cache_compare
,
1192 sizeof (rdsk_node_t
), offsetof(rdsk_node_t
, rn_node
));
1194 * This is not MT-safe, but we have no MT consumers of libzfs
1196 while ((dp
= readdir64(dirp
)) != NULL
) {
1197 const char *name
= dp
->d_name
;
1198 if (name
[0] == '.' &&
1199 (name
[1] == 0 || (name
[1] == '.' && name
[2] == 0)))
1202 slice
= zfs_alloc(hdl
, sizeof (rdsk_node_t
));
1203 slice
->rn_name
= zfs_strdup(hdl
, name
);
1204 slice
->rn_avl
= &slice_cache
;
1205 slice
->rn_dfd
= dfd
;
1206 slice
->rn_hdl
= hdl
;
1207 slice
->rn_nozpool
= B_FALSE
;
1208 avl_add(&slice_cache
, slice
);
1211 * create a thread pool to do all of this in parallel;
1212 * rn_nozpool is not protected, so this is racy in that
1213 * multiple tasks could decide that the same slice can
1214 * not hold a zpool, which is benign. Also choose
1215 * double the number of processors; we hold a lot of
1216 * locks in the kernel, so going beyond this doesn't
1219 t
= tpool_create(1, 2 * sysconf(_SC_NPROCESSORS_ONLN
),
1221 for (slice
= avl_first(&slice_cache
); slice
;
1222 (slice
= avl_walk(&slice_cache
, slice
,
1224 (void) tpool_dispatch(t
, zpool_open_func
, slice
);
1229 while ((slice
= avl_destroy_nodes(&slice_cache
,
1230 &cookie
)) != NULL
) {
1231 if (slice
->rn_config
!= NULL
&& !config_failed
) {
1232 nvlist_t
*config
= slice
->rn_config
;
1233 boolean_t matched
= B_TRUE
;
1235 if (iarg
->poolname
!= NULL
) {
1238 matched
= nvlist_lookup_string(config
,
1239 ZPOOL_CONFIG_POOL_NAME
,
1241 strcmp(iarg
->poolname
, pname
) == 0;
1242 } else if (iarg
->guid
!= 0) {
1245 matched
= nvlist_lookup_uint64(config
,
1246 ZPOOL_CONFIG_POOL_GUID
,
1248 iarg
->guid
== this_guid
;
1251 nvlist_free(config
);
1254 * use the non-raw path for the config
1256 (void) strlcpy(end
, slice
->rn_name
,
1258 if (add_config(hdl
, &pools
, path
,
1260 config_failed
= B_TRUE
;
1263 free(slice
->rn_name
);
1266 avl_destroy(&slice_cache
);
1268 (void) closedir(dirp
);
1274 ret
= get_configs(hdl
, &pools
, iarg
->can_be_active
);
1277 for (pe
= pools
.pools
; pe
!= NULL
; pe
= penext
) {
1278 penext
= pe
->pe_next
;
1279 for (ve
= pe
->pe_vdevs
; ve
!= NULL
; ve
= venext
) {
1280 venext
= ve
->ve_next
;
1281 for (ce
= ve
->ve_configs
; ce
!= NULL
; ce
= cenext
) {
1282 cenext
= ce
->ce_next
;
1283 nvlist_free(ce
->ce_config
);
1291 for (ne
= pools
.names
; ne
!= NULL
; ne
= nenext
) {
1292 nenext
= ne
->ne_next
;
1301 zpool_find_import(libzfs_handle_t
*hdl
, int argc
, char **argv
)
1303 importargs_t iarg
= { 0 };
1308 return (zpool_find_import_impl(hdl
, &iarg
));
1312 * Given a cache file, return the contents as a list of importable pools.
1313 * poolname or guid (but not both) are provided by the caller when trying
1314 * to import a specific pool.
1317 zpool_find_import_cached(libzfs_handle_t
*hdl
, const char *cachefile
,
1318 char *poolname
, uint64_t guid
)
1322 struct stat64 statbuf
;
1323 nvlist_t
*raw
, *src
, *dst
;
1330 verify(poolname
== NULL
|| guid
== 0);
1332 if ((fd
= open(cachefile
, O_RDONLY
)) < 0) {
1333 zfs_error_aux(hdl
, "%s", strerror(errno
));
1334 (void) zfs_error(hdl
, EZFS_BADCACHE
,
1335 dgettext(TEXT_DOMAIN
, "failed to open cache file"));
1339 if (fstat64(fd
, &statbuf
) != 0) {
1340 zfs_error_aux(hdl
, "%s", strerror(errno
));
1342 (void) zfs_error(hdl
, EZFS_BADCACHE
,
1343 dgettext(TEXT_DOMAIN
, "failed to get size of cache file"));
1347 if ((buf
= zfs_alloc(hdl
, statbuf
.st_size
)) == NULL
) {
1352 if (read(fd
, buf
, statbuf
.st_size
) != statbuf
.st_size
) {
1355 (void) zfs_error(hdl
, EZFS_BADCACHE
,
1356 dgettext(TEXT_DOMAIN
,
1357 "failed to read cache file contents"));
1363 if (nvlist_unpack(buf
, statbuf
.st_size
, &raw
, 0) != 0) {
1365 (void) zfs_error(hdl
, EZFS_BADCACHE
,
1366 dgettext(TEXT_DOMAIN
,
1367 "invalid or corrupt cache file contents"));
1374 * Go through and get the current state of the pools and refresh their
1377 if (nvlist_alloc(&pools
, 0, 0) != 0) {
1378 (void) no_memory(hdl
);
1384 while ((elem
= nvlist_next_nvpair(raw
, elem
)) != NULL
) {
1385 src
= fnvpair_value_nvlist(elem
);
1387 name
= fnvlist_lookup_string(src
, ZPOOL_CONFIG_POOL_NAME
);
1388 if (poolname
!= NULL
&& strcmp(poolname
, name
) != 0)
1391 this_guid
= fnvlist_lookup_uint64(src
, ZPOOL_CONFIG_POOL_GUID
);
1392 if (guid
!= 0 && guid
!= this_guid
)
1395 if (pool_active(hdl
, name
, this_guid
, &active
) != 0) {
1404 if ((dst
= refresh_config(hdl
, src
)) == NULL
) {
1410 if (nvlist_add_nvlist(pools
, nvpair_name(elem
), dst
) != 0) {
1411 (void) no_memory(hdl
);
1425 name_or_guid_exists(zpool_handle_t
*zhp
, void *data
)
1427 importargs_t
*import
= data
;
1430 if (import
->poolname
!= NULL
) {
1433 verify(nvlist_lookup_string(zhp
->zpool_config
,
1434 ZPOOL_CONFIG_POOL_NAME
, &pool_name
) == 0);
1435 if (strcmp(pool_name
, import
->poolname
) == 0)
1440 verify(nvlist_lookup_uint64(zhp
->zpool_config
,
1441 ZPOOL_CONFIG_POOL_GUID
, &pool_guid
) == 0);
1442 if (pool_guid
== import
->guid
)
1451 zpool_search_import(libzfs_handle_t
*hdl
, importargs_t
*import
)
1453 verify(import
->poolname
== NULL
|| import
->guid
== 0);
1456 import
->exists
= zpool_iter(hdl
, name_or_guid_exists
, import
);
1458 if (import
->cachefile
!= NULL
)
1459 return (zpool_find_import_cached(hdl
, import
->cachefile
,
1460 import
->poolname
, import
->guid
));
1462 return (zpool_find_import_impl(hdl
, import
));
1466 find_guid(nvlist_t
*nv
, uint64_t guid
)
1472 verify(nvlist_lookup_uint64(nv
, ZPOOL_CONFIG_GUID
, &tmp
) == 0);
1476 if (nvlist_lookup_nvlist_array(nv
, ZPOOL_CONFIG_CHILDREN
,
1477 &child
, &children
) == 0) {
1478 for (c
= 0; c
< children
; c
++)
1479 if (find_guid(child
[c
], guid
))
1486 typedef struct aux_cbdata
{
1487 const char *cb_type
;
1489 zpool_handle_t
*cb_zhp
;
1493 find_aux(zpool_handle_t
*zhp
, void *data
)
1495 aux_cbdata_t
*cbp
= data
;
1501 verify(nvlist_lookup_nvlist(zhp
->zpool_config
, ZPOOL_CONFIG_VDEV_TREE
,
1504 if (nvlist_lookup_nvlist_array(nvroot
, cbp
->cb_type
,
1505 &list
, &count
) == 0) {
1506 for (i
= 0; i
< count
; i
++) {
1507 verify(nvlist_lookup_uint64(list
[i
],
1508 ZPOOL_CONFIG_GUID
, &guid
) == 0);
1509 if (guid
== cbp
->cb_guid
) {
1521 * Determines if the pool is in use. If so, it returns true and the state of
1522 * the pool as well as the name of the pool. Both strings are allocated and
1523 * must be freed by the caller.
1526 zpool_in_use(libzfs_handle_t
*hdl
, int fd
, pool_state_t
*state
, char **namestr
,
1532 uint64_t guid
, vdev_guid
;
1533 zpool_handle_t
*zhp
;
1534 nvlist_t
*pool_config
;
1535 uint64_t stateval
, isspare
;
1536 aux_cbdata_t cb
= { 0 };
1541 if (zpool_read_label(fd
, &config
) != 0) {
1542 (void) no_memory(hdl
);
1549 verify(nvlist_lookup_uint64(config
, ZPOOL_CONFIG_POOL_STATE
,
1551 verify(nvlist_lookup_uint64(config
, ZPOOL_CONFIG_GUID
,
1554 if (stateval
!= POOL_STATE_SPARE
&& stateval
!= POOL_STATE_L2CACHE
) {
1555 verify(nvlist_lookup_string(config
, ZPOOL_CONFIG_POOL_NAME
,
1557 verify(nvlist_lookup_uint64(config
, ZPOOL_CONFIG_POOL_GUID
,
1562 case POOL_STATE_EXPORTED
:
1564 * A pool with an exported state may in fact be imported
1565 * read-only, so check the in-core state to see if it's
1566 * active and imported read-only. If it is, set
1567 * its state to active.
1569 if (pool_active(hdl
, name
, guid
, &isactive
) == 0 && isactive
&&
1570 (zhp
= zpool_open_canfail(hdl
, name
)) != NULL
) {
1571 if (zpool_get_prop_int(zhp
, ZPOOL_PROP_READONLY
, NULL
))
1572 stateval
= POOL_STATE_ACTIVE
;
1575 * All we needed the zpool handle for is the
1576 * readonly prop check.
1584 case POOL_STATE_ACTIVE
:
1586 * For an active pool, we have to determine if it's really part
1587 * of a currently active pool (in which case the pool will exist
1588 * and the guid will be the same), or whether it's part of an
1589 * active pool that was disconnected without being explicitly
1592 if (pool_active(hdl
, name
, guid
, &isactive
) != 0) {
1593 nvlist_free(config
);
1599 * Because the device may have been removed while
1600 * offlined, we only report it as active if the vdev is
1601 * still present in the config. Otherwise, pretend like
1604 if ((zhp
= zpool_open_canfail(hdl
, name
)) != NULL
&&
1605 (pool_config
= zpool_get_config(zhp
, NULL
))
1609 verify(nvlist_lookup_nvlist(pool_config
,
1610 ZPOOL_CONFIG_VDEV_TREE
, &nvroot
) == 0);
1611 ret
= find_guid(nvroot
, vdev_guid
);
1617 * If this is an active spare within another pool, we
1618 * treat it like an unused hot spare. This allows the
1619 * user to create a pool with a hot spare that currently
1620 * in use within another pool. Since we return B_TRUE,
1621 * libdiskmgt will continue to prevent generic consumers
1622 * from using the device.
1624 if (ret
&& nvlist_lookup_uint64(config
,
1625 ZPOOL_CONFIG_IS_SPARE
, &isspare
) == 0 && isspare
)
1626 stateval
= POOL_STATE_SPARE
;
1631 stateval
= POOL_STATE_POTENTIALLY_ACTIVE
;
1636 case POOL_STATE_SPARE
:
1638 * For a hot spare, it can be either definitively in use, or
1639 * potentially active. To determine if it's in use, we iterate
1640 * over all pools in the system and search for one with a spare
1641 * with a matching guid.
1643 * Due to the shared nature of spares, we don't actually report
1644 * the potentially active case as in use. This means the user
1645 * can freely create pools on the hot spares of exported pools,
1646 * but to do otherwise makes the resulting code complicated, and
1647 * we end up having to deal with this case anyway.
1650 cb
.cb_guid
= vdev_guid
;
1651 cb
.cb_type
= ZPOOL_CONFIG_SPARES
;
1652 if (zpool_iter(hdl
, find_aux
, &cb
) == 1) {
1653 name
= (char *)zpool_get_name(cb
.cb_zhp
);
1660 case POOL_STATE_L2CACHE
:
1663 * Check if any pool is currently using this l2cache device.
1666 cb
.cb_guid
= vdev_guid
;
1667 cb
.cb_type
= ZPOOL_CONFIG_L2CACHE
;
1668 if (zpool_iter(hdl
, find_aux
, &cb
) == 1) {
1669 name
= (char *)zpool_get_name(cb
.cb_zhp
);
1682 if ((*namestr
= zfs_strdup(hdl
, name
)) == NULL
) {
1684 zpool_close(cb
.cb_zhp
);
1685 nvlist_free(config
);
1688 *state
= (pool_state_t
)stateval
;
1692 zpool_close(cb
.cb_zhp
);
1694 nvlist_free(config
);