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 https://opensource.org/licenses/CDDL-1.0.
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) 2011, 2019 by Delphix. All rights reserved.
25 * Copyright (c) 2014 Integros [integros.com]
26 * Copyright 2016 Nexenta Systems, Inc.
27 * Copyright (c) 2017, 2018 Lawrence Livermore National Security, LLC.
28 * Copyright (c) 2015, 2017, Intel Corporation.
29 * Copyright (c) 2020 Datto Inc.
30 * Copyright (c) 2020, The FreeBSD Foundation [1]
32 * [1] Portions of this software were developed by Allan Jude
33 * under sponsorship from the FreeBSD Foundation.
34 * Copyright (c) 2021 Allan Jude
35 * Copyright (c) 2021 Toomas Soome <tsoome@me.com>
43 #include <openssl/evp.h>
44 #include <sys/zfs_context.h>
46 #include <sys/spa_impl.h>
49 #include <sys/fs/zfs.h>
50 #include <sys/zfs_znode.h>
51 #include <sys/zfs_sa.h>
53 #include <sys/sa_impl.h>
55 #include <sys/vdev_impl.h>
56 #include <sys/metaslab_impl.h>
57 #include <sys/dmu_objset.h>
58 #include <sys/dsl_dir.h>
59 #include <sys/dsl_dataset.h>
60 #include <sys/dsl_pool.h>
61 #include <sys/dsl_bookmark.h>
64 #include <sys/zil_impl.h>
66 #include <sys/resource.h>
67 #include <sys/dmu_send.h>
68 #include <sys/dmu_traverse.h>
69 #include <sys/zio_checksum.h>
70 #include <sys/zio_compress.h>
71 #include <sys/zfs_fuid.h>
73 #include <sys/arc_impl.h>
75 #include <sys/zfeature.h>
77 #include <sys/blkptr.h>
78 #include <sys/dsl_crypt.h>
79 #include <sys/dsl_scan.h>
80 #include <sys/btree.h>
81 #include <zfs_comutil.h>
82 #include <sys/zstd/zstd.h>
84 #include <libnvpair.h>
89 #define ZDB_COMPRESS_NAME(idx) ((idx) < ZIO_COMPRESS_FUNCTIONS ? \
90 zio_compress_table[(idx)].ci_name : "UNKNOWN")
91 #define ZDB_CHECKSUM_NAME(idx) ((idx) < ZIO_CHECKSUM_FUNCTIONS ? \
92 zio_checksum_table[(idx)].ci_name : "UNKNOWN")
93 #define ZDB_OT_TYPE(idx) ((idx) < DMU_OT_NUMTYPES ? (idx) : \
94 (idx) == DMU_OTN_ZAP_DATA || (idx) == DMU_OTN_ZAP_METADATA ? \
96 (idx) == DMU_OTN_UINT64_DATA || (idx) == DMU_OTN_UINT64_METADATA ? \
97 DMU_OT_UINT64_OTHER : DMU_OT_NUMTYPES)
99 /* Some platforms require part of inode IDs to be remapped */
101 #define ZDB_MAP_OBJECT_ID(obj) INO_XNUTOZFS(obj, 2)
103 #define ZDB_MAP_OBJECT_ID(obj) (obj)
107 zdb_ot_name(dmu_object_type_t type
)
109 if (type
< DMU_OT_NUMTYPES
)
110 return (dmu_ot
[type
].ot_name
);
111 else if ((type
& DMU_OT_NEWTYPE
) &&
112 ((type
& DMU_OT_BYTESWAP_MASK
) < DMU_BSWAP_NUMFUNCS
))
113 return (dmu_ot_byteswap
[type
& DMU_OT_BYTESWAP_MASK
].ob_name
);
118 extern int reference_tracking_enable
;
119 extern int zfs_recover
;
120 extern uint_t zfs_vdev_async_read_max_active
;
121 extern boolean_t spa_load_verify_dryrun
;
122 extern boolean_t spa_mode_readable_spacemaps
;
123 extern uint_t zfs_reconstruct_indirect_combinations_max
;
124 extern uint_t zfs_btree_verify_intensity
;
126 static const char cmdname
[] = "zdb";
127 uint8_t dump_opt
[256];
129 typedef void object_viewer_t(objset_t
*, uint64_t, void *data
, size_t size
);
131 static uint64_t *zopt_metaslab
= NULL
;
132 static unsigned zopt_metaslab_args
= 0;
134 typedef struct zopt_object_range
{
135 uint64_t zor_obj_start
;
136 uint64_t zor_obj_end
;
138 } zopt_object_range_t
;
140 static zopt_object_range_t
*zopt_object_ranges
= NULL
;
141 static unsigned zopt_object_args
= 0;
143 static int flagbits
[256];
145 #define ZOR_FLAG_PLAIN_FILE 0x0001
146 #define ZOR_FLAG_DIRECTORY 0x0002
147 #define ZOR_FLAG_SPACE_MAP 0x0004
148 #define ZOR_FLAG_ZAP 0x0008
149 #define ZOR_FLAG_ALL_TYPES -1
150 #define ZOR_SUPPORTED_FLAGS (ZOR_FLAG_PLAIN_FILE | \
151 ZOR_FLAG_DIRECTORY | \
152 ZOR_FLAG_SPACE_MAP | \
155 #define ZDB_FLAG_CHECKSUM 0x0001
156 #define ZDB_FLAG_DECOMPRESS 0x0002
157 #define ZDB_FLAG_BSWAP 0x0004
158 #define ZDB_FLAG_GBH 0x0008
159 #define ZDB_FLAG_INDIRECT 0x0010
160 #define ZDB_FLAG_RAW 0x0020
161 #define ZDB_FLAG_PRINT_BLKPTR 0x0040
162 #define ZDB_FLAG_VERBOSE 0x0080
164 static uint64_t max_inflight_bytes
= 256 * 1024 * 1024; /* 256MB */
165 static int leaked_objects
= 0;
166 static range_tree_t
*mos_refd_objs
;
168 static void snprintf_blkptr_compact(char *, size_t, const blkptr_t
*,
170 static void mos_obj_refd(uint64_t);
171 static void mos_obj_refd_multiple(uint64_t);
172 static int dump_bpobj_cb(void *arg
, const blkptr_t
*bp
, boolean_t free
,
175 typedef struct sublivelist_verify
{
176 /* FREE's that haven't yet matched to an ALLOC, in one sub-livelist */
179 /* ALLOC's without a matching FREE, accumulates across sub-livelists */
180 zfs_btree_t sv_leftover
;
181 } sublivelist_verify_t
;
184 livelist_compare(const void *larg
, const void *rarg
)
186 const blkptr_t
*l
= larg
;
187 const blkptr_t
*r
= rarg
;
189 /* Sort them according to dva[0] */
190 uint64_t l_dva0_vdev
, r_dva0_vdev
;
191 l_dva0_vdev
= DVA_GET_VDEV(&l
->blk_dva
[0]);
192 r_dva0_vdev
= DVA_GET_VDEV(&r
->blk_dva
[0]);
193 if (l_dva0_vdev
< r_dva0_vdev
)
195 else if (l_dva0_vdev
> r_dva0_vdev
)
198 /* if vdevs are equal, sort by offsets. */
199 uint64_t l_dva0_offset
;
200 uint64_t r_dva0_offset
;
201 l_dva0_offset
= DVA_GET_OFFSET(&l
->blk_dva
[0]);
202 r_dva0_offset
= DVA_GET_OFFSET(&r
->blk_dva
[0]);
203 if (l_dva0_offset
< r_dva0_offset
) {
205 } else if (l_dva0_offset
> r_dva0_offset
) {
210 * Since we're storing blkptrs without cancelling FREE/ALLOC pairs,
211 * it's possible the offsets are equal. In that case, sort by txg
213 if (l
->blk_birth
< r
->blk_birth
) {
215 } else if (l
->blk_birth
> r
->blk_birth
) {
221 typedef struct sublivelist_verify_block
{
225 * We need this to check if the block marked as allocated
226 * in the livelist was freed (and potentially reallocated)
227 * in the metaslab spacemaps at a later TXG.
229 uint64_t svb_allocated_txg
;
230 } sublivelist_verify_block_t
;
232 static void zdb_print_blkptr(const blkptr_t
*bp
, int flags
);
234 typedef struct sublivelist_verify_block_refcnt
{
235 /* block pointer entry in livelist being verified */
239 * Refcount gets incremented to 1 when we encounter the first
240 * FREE entry for the svfbr block pointer and a node for it
241 * is created in our ZDB verification/tracking metadata.
243 * As we encounter more FREE entries we increment this counter
244 * and similarly decrement it whenever we find the respective
245 * ALLOC entries for this block.
247 * When the refcount gets to 0 it means that all the FREE and
248 * ALLOC entries of this block have paired up and we no longer
249 * need to track it in our verification logic (e.g. the node
250 * containing this struct in our verification data structure
253 * [refer to sublivelist_verify_blkptr() for the actual code]
255 uint32_t svbr_refcnt
;
256 } sublivelist_verify_block_refcnt_t
;
259 sublivelist_block_refcnt_compare(const void *larg
, const void *rarg
)
261 const sublivelist_verify_block_refcnt_t
*l
= larg
;
262 const sublivelist_verify_block_refcnt_t
*r
= rarg
;
263 return (livelist_compare(&l
->svbr_blk
, &r
->svbr_blk
));
267 sublivelist_verify_blkptr(void *arg
, const blkptr_t
*bp
, boolean_t free
,
270 ASSERT3P(tx
, ==, NULL
);
271 struct sublivelist_verify
*sv
= arg
;
272 sublivelist_verify_block_refcnt_t current
= {
276 * Start with 1 in case this is the first free entry.
277 * This field is not used for our B-Tree comparisons
283 zfs_btree_index_t where
;
284 sublivelist_verify_block_refcnt_t
*pair
=
285 zfs_btree_find(&sv
->sv_pair
, ¤t
, &where
);
288 /* first free entry for this block pointer */
289 zfs_btree_add(&sv
->sv_pair
, ¤t
);
295 /* block that is currently marked as allocated */
296 for (int i
= 0; i
< SPA_DVAS_PER_BP
; i
++) {
297 if (DVA_IS_EMPTY(&bp
->blk_dva
[i
]))
299 sublivelist_verify_block_t svb
= {
300 .svb_dva
= bp
->blk_dva
[i
],
301 .svb_allocated_txg
= bp
->blk_birth
304 if (zfs_btree_find(&sv
->sv_leftover
, &svb
,
306 zfs_btree_add_idx(&sv
->sv_leftover
,
311 /* alloc matches a free entry */
313 if (pair
->svbr_refcnt
== 0) {
314 /* all allocs and frees have been matched */
315 zfs_btree_remove_idx(&sv
->sv_pair
, &where
);
324 sublivelist_verify_func(void *args
, dsl_deadlist_entry_t
*dle
)
327 struct sublivelist_verify
*sv
= args
;
329 zfs_btree_create(&sv
->sv_pair
, sublivelist_block_refcnt_compare
,
330 sizeof (sublivelist_verify_block_refcnt_t
));
332 err
= bpobj_iterate_nofree(&dle
->dle_bpobj
, sublivelist_verify_blkptr
,
335 sublivelist_verify_block_refcnt_t
*e
;
336 zfs_btree_index_t
*cookie
= NULL
;
337 while ((e
= zfs_btree_destroy_nodes(&sv
->sv_pair
, &cookie
)) != NULL
) {
338 char blkbuf
[BP_SPRINTF_LEN
];
339 snprintf_blkptr_compact(blkbuf
, sizeof (blkbuf
),
340 &e
->svbr_blk
, B_TRUE
);
341 (void) printf("\tERROR: %d unmatched FREE(s): %s\n",
342 e
->svbr_refcnt
, blkbuf
);
344 zfs_btree_destroy(&sv
->sv_pair
);
350 livelist_block_compare(const void *larg
, const void *rarg
)
352 const sublivelist_verify_block_t
*l
= larg
;
353 const sublivelist_verify_block_t
*r
= rarg
;
355 if (DVA_GET_VDEV(&l
->svb_dva
) < DVA_GET_VDEV(&r
->svb_dva
))
357 else if (DVA_GET_VDEV(&l
->svb_dva
) > DVA_GET_VDEV(&r
->svb_dva
))
360 if (DVA_GET_OFFSET(&l
->svb_dva
) < DVA_GET_OFFSET(&r
->svb_dva
))
362 else if (DVA_GET_OFFSET(&l
->svb_dva
) > DVA_GET_OFFSET(&r
->svb_dva
))
365 if (DVA_GET_ASIZE(&l
->svb_dva
) < DVA_GET_ASIZE(&r
->svb_dva
))
367 else if (DVA_GET_ASIZE(&l
->svb_dva
) > DVA_GET_ASIZE(&r
->svb_dva
))
374 * Check for errors in a livelist while tracking all unfreed ALLOCs in the
375 * sublivelist_verify_t: sv->sv_leftover
378 livelist_verify(dsl_deadlist_t
*dl
, void *arg
)
380 sublivelist_verify_t
*sv
= arg
;
381 dsl_deadlist_iterate(dl
, sublivelist_verify_func
, sv
);
385 * Check for errors in the livelist entry and discard the intermediary
389 sublivelist_verify_lightweight(void *args
, dsl_deadlist_entry_t
*dle
)
392 sublivelist_verify_t sv
;
393 zfs_btree_create(&sv
.sv_leftover
, livelist_block_compare
,
394 sizeof (sublivelist_verify_block_t
));
395 int err
= sublivelist_verify_func(&sv
, dle
);
396 zfs_btree_clear(&sv
.sv_leftover
);
397 zfs_btree_destroy(&sv
.sv_leftover
);
401 typedef struct metaslab_verify
{
403 * Tree containing all the leftover ALLOCs from the livelists
404 * that are part of this metaslab.
406 zfs_btree_t mv_livelist_allocs
;
409 * Metaslab information.
417 * What's currently allocated for this metaslab.
419 range_tree_t
*mv_allocated
;
422 typedef void ll_iter_t(dsl_deadlist_t
*ll
, void *arg
);
424 typedef int (*zdb_log_sm_cb_t
)(spa_t
*spa
, space_map_entry_t
*sme
, uint64_t txg
,
427 typedef struct unflushed_iter_cb_arg
{
431 zdb_log_sm_cb_t uic_cb
;
432 } unflushed_iter_cb_arg_t
;
435 iterate_through_spacemap_logs_cb(space_map_entry_t
*sme
, void *arg
)
437 unflushed_iter_cb_arg_t
*uic
= arg
;
438 return (uic
->uic_cb(uic
->uic_spa
, sme
, uic
->uic_txg
, uic
->uic_arg
));
442 iterate_through_spacemap_logs(spa_t
*spa
, zdb_log_sm_cb_t cb
, void *arg
)
444 if (!spa_feature_is_active(spa
, SPA_FEATURE_LOG_SPACEMAP
))
447 spa_config_enter(spa
, SCL_CONFIG
, FTAG
, RW_READER
);
448 for (spa_log_sm_t
*sls
= avl_first(&spa
->spa_sm_logs_by_txg
);
449 sls
; sls
= AVL_NEXT(&spa
->spa_sm_logs_by_txg
, sls
)) {
450 space_map_t
*sm
= NULL
;
451 VERIFY0(space_map_open(&sm
, spa_meta_objset(spa
),
452 sls
->sls_sm_obj
, 0, UINT64_MAX
, SPA_MINBLOCKSHIFT
));
454 unflushed_iter_cb_arg_t uic
= {
456 .uic_txg
= sls
->sls_txg
,
460 VERIFY0(space_map_iterate(sm
, space_map_length(sm
),
461 iterate_through_spacemap_logs_cb
, &uic
));
464 spa_config_exit(spa
, SCL_CONFIG
, FTAG
);
468 verify_livelist_allocs(metaslab_verify_t
*mv
, uint64_t txg
,
469 uint64_t offset
, uint64_t size
)
471 sublivelist_verify_block_t svb
= {{{0}}};
472 DVA_SET_VDEV(&svb
.svb_dva
, mv
->mv_vdid
);
473 DVA_SET_OFFSET(&svb
.svb_dva
, offset
);
474 DVA_SET_ASIZE(&svb
.svb_dva
, size
);
475 zfs_btree_index_t where
;
476 uint64_t end_offset
= offset
+ size
;
479 * Look for an exact match for spacemap entry in the livelist entries.
480 * Then, look for other livelist entries that fall within the range
481 * of the spacemap entry as it may have been condensed
483 sublivelist_verify_block_t
*found
=
484 zfs_btree_find(&mv
->mv_livelist_allocs
, &svb
, &where
);
486 found
= zfs_btree_next(&mv
->mv_livelist_allocs
, &where
, &where
);
488 for (; found
!= NULL
&& DVA_GET_VDEV(&found
->svb_dva
) == mv
->mv_vdid
&&
489 DVA_GET_OFFSET(&found
->svb_dva
) < end_offset
;
490 found
= zfs_btree_next(&mv
->mv_livelist_allocs
, &where
, &where
)) {
491 if (found
->svb_allocated_txg
<= txg
) {
492 (void) printf("ERROR: Livelist ALLOC [%llx:%llx] "
493 "from TXG %llx FREED at TXG %llx\n",
494 (u_longlong_t
)DVA_GET_OFFSET(&found
->svb_dva
),
495 (u_longlong_t
)DVA_GET_ASIZE(&found
->svb_dva
),
496 (u_longlong_t
)found
->svb_allocated_txg
,
503 metaslab_spacemap_validation_cb(space_map_entry_t
*sme
, void *arg
)
505 metaslab_verify_t
*mv
= arg
;
506 uint64_t offset
= sme
->sme_offset
;
507 uint64_t size
= sme
->sme_run
;
508 uint64_t txg
= sme
->sme_txg
;
510 if (sme
->sme_type
== SM_ALLOC
) {
511 if (range_tree_contains(mv
->mv_allocated
,
513 (void) printf("ERROR: DOUBLE ALLOC: "
515 "%llu:%llu LOG_SM\n",
516 (u_longlong_t
)txg
, (u_longlong_t
)offset
,
517 (u_longlong_t
)size
, (u_longlong_t
)mv
->mv_vdid
,
518 (u_longlong_t
)mv
->mv_msid
);
520 range_tree_add(mv
->mv_allocated
,
524 if (!range_tree_contains(mv
->mv_allocated
,
526 (void) printf("ERROR: DOUBLE FREE: "
528 "%llu:%llu LOG_SM\n",
529 (u_longlong_t
)txg
, (u_longlong_t
)offset
,
530 (u_longlong_t
)size
, (u_longlong_t
)mv
->mv_vdid
,
531 (u_longlong_t
)mv
->mv_msid
);
533 range_tree_remove(mv
->mv_allocated
,
538 if (sme
->sme_type
!= SM_ALLOC
) {
540 * If something is freed in the spacemap, verify that
541 * it is not listed as allocated in the livelist.
543 verify_livelist_allocs(mv
, txg
, offset
, size
);
549 spacemap_check_sm_log_cb(spa_t
*spa
, space_map_entry_t
*sme
,
550 uint64_t txg
, void *arg
)
552 metaslab_verify_t
*mv
= arg
;
553 uint64_t offset
= sme
->sme_offset
;
554 uint64_t vdev_id
= sme
->sme_vdev
;
556 vdev_t
*vd
= vdev_lookup_top(spa
, vdev_id
);
558 /* skip indirect vdevs */
559 if (!vdev_is_concrete(vd
))
562 if (vdev_id
!= mv
->mv_vdid
)
565 metaslab_t
*ms
= vd
->vdev_ms
[offset
>> vd
->vdev_ms_shift
];
566 if (ms
->ms_id
!= mv
->mv_msid
)
569 if (txg
< metaslab_unflushed_txg(ms
))
573 ASSERT3U(txg
, ==, sme
->sme_txg
);
574 return (metaslab_spacemap_validation_cb(sme
, mv
));
578 spacemap_check_sm_log(spa_t
*spa
, metaslab_verify_t
*mv
)
580 iterate_through_spacemap_logs(spa
, spacemap_check_sm_log_cb
, mv
);
584 spacemap_check_ms_sm(space_map_t
*sm
, metaslab_verify_t
*mv
)
589 VERIFY0(space_map_iterate(sm
, space_map_length(sm
),
590 metaslab_spacemap_validation_cb
, mv
));
593 static void iterate_deleted_livelists(spa_t
*spa
, ll_iter_t func
, void *arg
);
596 * Transfer blocks from sv_leftover tree to the mv_livelist_allocs if
597 * they are part of that metaslab (mv_msid).
600 mv_populate_livelist_allocs(metaslab_verify_t
*mv
, sublivelist_verify_t
*sv
)
602 zfs_btree_index_t where
;
603 sublivelist_verify_block_t
*svb
;
604 ASSERT3U(zfs_btree_numnodes(&mv
->mv_livelist_allocs
), ==, 0);
605 for (svb
= zfs_btree_first(&sv
->sv_leftover
, &where
);
607 svb
= zfs_btree_next(&sv
->sv_leftover
, &where
, &where
)) {
608 if (DVA_GET_VDEV(&svb
->svb_dva
) != mv
->mv_vdid
)
611 if (DVA_GET_OFFSET(&svb
->svb_dva
) < mv
->mv_start
&&
612 (DVA_GET_OFFSET(&svb
->svb_dva
) +
613 DVA_GET_ASIZE(&svb
->svb_dva
)) > mv
->mv_start
) {
614 (void) printf("ERROR: Found block that crosses "
615 "metaslab boundary: <%llu:%llx:%llx>\n",
616 (u_longlong_t
)DVA_GET_VDEV(&svb
->svb_dva
),
617 (u_longlong_t
)DVA_GET_OFFSET(&svb
->svb_dva
),
618 (u_longlong_t
)DVA_GET_ASIZE(&svb
->svb_dva
));
622 if (DVA_GET_OFFSET(&svb
->svb_dva
) < mv
->mv_start
)
625 if (DVA_GET_OFFSET(&svb
->svb_dva
) >= mv
->mv_end
)
628 if ((DVA_GET_OFFSET(&svb
->svb_dva
) +
629 DVA_GET_ASIZE(&svb
->svb_dva
)) > mv
->mv_end
) {
630 (void) printf("ERROR: Found block that crosses "
631 "metaslab boundary: <%llu:%llx:%llx>\n",
632 (u_longlong_t
)DVA_GET_VDEV(&svb
->svb_dva
),
633 (u_longlong_t
)DVA_GET_OFFSET(&svb
->svb_dva
),
634 (u_longlong_t
)DVA_GET_ASIZE(&svb
->svb_dva
));
638 zfs_btree_add(&mv
->mv_livelist_allocs
, svb
);
641 for (svb
= zfs_btree_first(&mv
->mv_livelist_allocs
, &where
);
643 svb
= zfs_btree_next(&mv
->mv_livelist_allocs
, &where
, &where
)) {
644 zfs_btree_remove(&sv
->sv_leftover
, svb
);
650 * Iterate through all the sublivelists and:
651 * - report leftover frees (**)
652 * - record leftover ALLOCs together with their TXG [see Cross Check]
654 * (**) Note: Double ALLOCs are valid in datasets that have dedup
655 * enabled. Similarly double FREEs are allowed as well but
656 * only if they pair up with a corresponding ALLOC entry once
657 * we our done with our sublivelist iteration.
661 * - iterate over spacemap and then the metaslab's entries in the
662 * spacemap log, then report any double FREEs and ALLOCs (do not
666 * After finishing the Livelist Check phase and while being in the
667 * Spacemap Check phase, we find all the recorded leftover ALLOCs
668 * of the livelist check that are part of the metaslab that we are
669 * currently looking at in the Spacemap Check. We report any entries
670 * that are marked as ALLOCs in the livelists but have been actually
671 * freed (and potentially allocated again) after their TXG stamp in
672 * the spacemaps. Also report any ALLOCs from the livelists that
673 * belong to indirect vdevs (e.g. their vdev completed removal).
675 * Note that this will miss Log Spacemap entries that cancelled each other
676 * out before being flushed to the metaslab, so we are not guaranteed
677 * to match all erroneous ALLOCs.
680 livelist_metaslab_validate(spa_t
*spa
)
682 (void) printf("Verifying deleted livelist entries\n");
684 sublivelist_verify_t sv
;
685 zfs_btree_create(&sv
.sv_leftover
, livelist_block_compare
,
686 sizeof (sublivelist_verify_block_t
));
687 iterate_deleted_livelists(spa
, livelist_verify
, &sv
);
689 (void) printf("Verifying metaslab entries\n");
690 vdev_t
*rvd
= spa
->spa_root_vdev
;
691 for (uint64_t c
= 0; c
< rvd
->vdev_children
; c
++) {
692 vdev_t
*vd
= rvd
->vdev_child
[c
];
694 if (!vdev_is_concrete(vd
))
697 for (uint64_t mid
= 0; mid
< vd
->vdev_ms_count
; mid
++) {
698 metaslab_t
*m
= vd
->vdev_ms
[mid
];
700 (void) fprintf(stderr
,
701 "\rverifying concrete vdev %llu, "
702 "metaslab %llu of %llu ...",
703 (longlong_t
)vd
->vdev_id
,
705 (longlong_t
)vd
->vdev_ms_count
);
707 uint64_t shift
, start
;
708 range_seg_type_t type
=
709 metaslab_calculate_range_tree_type(vd
, m
,
711 metaslab_verify_t mv
;
712 mv
.mv_allocated
= range_tree_create(NULL
,
713 type
, NULL
, start
, shift
);
714 mv
.mv_vdid
= vd
->vdev_id
;
715 mv
.mv_msid
= m
->ms_id
;
716 mv
.mv_start
= m
->ms_start
;
717 mv
.mv_end
= m
->ms_start
+ m
->ms_size
;
718 zfs_btree_create(&mv
.mv_livelist_allocs
,
719 livelist_block_compare
,
720 sizeof (sublivelist_verify_block_t
));
722 mv_populate_livelist_allocs(&mv
, &sv
);
724 spacemap_check_ms_sm(m
->ms_sm
, &mv
);
725 spacemap_check_sm_log(spa
, &mv
);
727 range_tree_vacate(mv
.mv_allocated
, NULL
, NULL
);
728 range_tree_destroy(mv
.mv_allocated
);
729 zfs_btree_clear(&mv
.mv_livelist_allocs
);
730 zfs_btree_destroy(&mv
.mv_livelist_allocs
);
733 (void) fprintf(stderr
, "\n");
736 * If there are any segments in the leftover tree after we walked
737 * through all the metaslabs in the concrete vdevs then this means
738 * that we have segments in the livelists that belong to indirect
739 * vdevs and are marked as allocated.
741 if (zfs_btree_numnodes(&sv
.sv_leftover
) == 0) {
742 zfs_btree_destroy(&sv
.sv_leftover
);
745 (void) printf("ERROR: Found livelist blocks marked as allocated "
746 "for indirect vdevs:\n");
748 zfs_btree_index_t
*where
= NULL
;
749 sublivelist_verify_block_t
*svb
;
750 while ((svb
= zfs_btree_destroy_nodes(&sv
.sv_leftover
, &where
)) !=
752 int vdev_id
= DVA_GET_VDEV(&svb
->svb_dva
);
753 ASSERT3U(vdev_id
, <, rvd
->vdev_children
);
754 vdev_t
*vd
= rvd
->vdev_child
[vdev_id
];
755 ASSERT(!vdev_is_concrete(vd
));
756 (void) printf("<%d:%llx:%llx> TXG %llx\n",
757 vdev_id
, (u_longlong_t
)DVA_GET_OFFSET(&svb
->svb_dva
),
758 (u_longlong_t
)DVA_GET_ASIZE(&svb
->svb_dva
),
759 (u_longlong_t
)svb
->svb_allocated_txg
);
762 zfs_btree_destroy(&sv
.sv_leftover
);
766 * These libumem hooks provide a reasonable set of defaults for the allocator's
767 * debugging facilities.
770 _umem_debug_init(void)
772 return ("default,verbose"); /* $UMEM_DEBUG setting */
776 _umem_logging_init(void)
778 return ("fail,contents"); /* $UMEM_LOGGING setting */
784 (void) fprintf(stderr
,
785 "Usage:\t%s [-AbcdDFGhikLMPsvXy] [-e [-V] [-p <path> ...]] "
786 "[-I <inflight I/Os>]\n"
787 "\t\t[-o <var>=<value>]... [-t <txg>] [-U <cache>] [-x <dumpdir>]\n"
789 "\t\t[<poolname>[/<dataset | objset id>] [<object | range> ...]]\n"
790 "\t%s [-AdiPv] [-e [-V] [-p <path> ...]] [-U <cache>] [-K <key>]\n"
791 "\t\t[<poolname>[/<dataset | objset id>] [<object | range> ...]\n"
792 "\t%s [-v] <bookmark>\n"
793 "\t%s -C [-A] [-U <cache>]\n"
794 "\t%s -l [-Aqu] <device>\n"
795 "\t%s -m [-AFLPX] [-e [-V] [-p <path> ...]] [-t <txg>] "
796 "[-U <cache>]\n\t\t<poolname> [<vdev> [<metaslab> ...]]\n"
797 "\t%s -O [-K <key>] <dataset> <path>\n"
798 "\t%s -r [-K <key>] <dataset> <path> <destination>\n"
799 "\t%s -R [-A] [-e [-V] [-p <path> ...]] [-U <cache>]\n"
800 "\t\t<poolname> <vdev>:<offset>:<size>[:<flags>]\n"
801 "\t%s -E [-A] word0:word1:...:word15\n"
802 "\t%s -S [-AP] [-e [-V] [-p <path> ...]] [-U <cache>] "
804 cmdname
, cmdname
, cmdname
, cmdname
, cmdname
, cmdname
, cmdname
,
805 cmdname
, cmdname
, cmdname
, cmdname
);
807 (void) fprintf(stderr
, " Dataset name must include at least one "
808 "separator character '/' or '@'\n");
809 (void) fprintf(stderr
, " If dataset name is specified, only that "
810 "dataset is dumped\n");
811 (void) fprintf(stderr
, " If object numbers or object number "
812 "ranges are specified, only those\n"
813 " objects or ranges are dumped.\n\n");
814 (void) fprintf(stderr
,
815 " Object ranges take the form <start>:<end>[:<flags>]\n"
816 " start Starting object number\n"
817 " end Ending object number, or -1 for no upper bound\n"
818 " flags Optional flags to select object types:\n"
819 " A All objects (this is the default)\n"
820 " d ZFS directories\n"
822 " m SPA space maps\n"
824 " - Negate effect of next flag\n\n");
825 (void) fprintf(stderr
, " Options to control amount of output:\n");
826 (void) fprintf(stderr
, " -b --block-stats "
827 "block statistics\n");
828 (void) fprintf(stderr
, " -c --checksum "
829 "checksum all metadata (twice for all data) blocks\n");
830 (void) fprintf(stderr
, " -C --config "
831 "config (or cachefile if alone)\n");
832 (void) fprintf(stderr
, " -d --datasets "
834 (void) fprintf(stderr
, " -D --dedup-stats "
835 "dedup statistics\n");
836 (void) fprintf(stderr
, " -E --embedded-block-pointer=INTEGER\n"
837 " decode and display block "
838 "from an embedded block pointer\n");
839 (void) fprintf(stderr
, " -h --history "
841 (void) fprintf(stderr
, " -i --intent-logs "
843 (void) fprintf(stderr
, " -l --label "
844 "read label contents\n");
845 (void) fprintf(stderr
, " -k --checkpointed-state "
846 "examine the checkpointed state of the pool\n");
847 (void) fprintf(stderr
, " -L --disable-leak-tracking "
848 "disable leak tracking (do not load spacemaps)\n");
849 (void) fprintf(stderr
, " -m --metaslabs "
851 (void) fprintf(stderr
, " -M --metaslab-groups "
852 "metaslab groups\n");
853 (void) fprintf(stderr
, " -O --object-lookups "
854 "perform object lookups by path\n");
855 (void) fprintf(stderr
, " -r --copy-object "
856 "copy an object by path to file\n");
857 (void) fprintf(stderr
, " -R --read-block "
858 "read and display block from a device\n");
859 (void) fprintf(stderr
, " -s --io-stats "
860 "report stats on zdb's I/O\n");
861 (void) fprintf(stderr
, " -S --simulate-dedup "
862 "simulate dedup to measure effect\n");
863 (void) fprintf(stderr
, " -v --verbose "
864 "verbose (applies to all others)\n");
865 (void) fprintf(stderr
, " -y --livelist "
866 "perform livelist and metaslab validation on any livelists being "
868 (void) fprintf(stderr
, " Below options are intended for use "
869 "with other options:\n");
870 (void) fprintf(stderr
, " -A --ignore-assertions "
871 "ignore assertions (-A), enable panic recovery (-AA) or both "
873 (void) fprintf(stderr
, " -e --exported "
874 "pool is exported/destroyed/has altroot/not in a cachefile\n");
875 (void) fprintf(stderr
, " -F --automatic-rewind "
876 "attempt automatic rewind within safe range of transaction "
878 (void) fprintf(stderr
, " -G --dump-debug-msg "
879 "dump zfs_dbgmsg buffer before exiting\n");
880 (void) fprintf(stderr
, " -I --inflight=INTEGER "
881 "specify the maximum number of checksumming I/Os "
882 "[default is 200]\n");
883 (void) fprintf(stderr
, " -K --key=KEY "
884 "decryption key for encrypted dataset\n");
885 (void) fprintf(stderr
, " -o --option=\"OPTION=INTEGER\" "
886 "set global variable to an unsigned 32-bit integer\n");
887 (void) fprintf(stderr
, " -p --path==PATH "
888 "use one or more with -e to specify path to vdev dir\n");
889 (void) fprintf(stderr
, " -P --parseable "
890 "print numbers in parseable form\n");
891 (void) fprintf(stderr
, " -q --skip-label "
892 "don't print label contents\n");
893 (void) fprintf(stderr
, " -t --txg=INTEGER "
894 "highest txg to use when searching for uberblocks\n");
895 (void) fprintf(stderr
, " -u --uberblock "
897 (void) fprintf(stderr
, " -U --cachefile=PATH "
898 "use alternate cachefile\n");
899 (void) fprintf(stderr
, " -V --verbatim "
900 "do verbatim import\n");
901 (void) fprintf(stderr
, " -x --dump-blocks=PATH "
902 "dump all read blocks into specified directory\n");
903 (void) fprintf(stderr
, " -X --extreme-rewind "
904 "attempt extreme rewind (does not work with dataset)\n");
905 (void) fprintf(stderr
, " -Y --all-reconstruction "
906 "attempt all reconstruction combinations for split blocks\n");
907 (void) fprintf(stderr
, " -Z --zstd-headers "
908 "show ZSTD headers \n");
909 (void) fprintf(stderr
, "Specify an option more than once (e.g. -bb) "
910 "to make only that option verbose\n");
911 (void) fprintf(stderr
, "Default is to dump everything non-verbosely\n");
916 dump_debug_buffer(void)
920 (void) fflush(stdout
);
921 zfs_dbgmsg_print("zdb");
926 * Called for usage errors that are discovered after a call to spa_open(),
927 * dmu_bonus_hold(), or pool_match(). abort() is called for other errors.
931 fatal(const char *fmt
, ...)
936 (void) fprintf(stderr
, "%s: ", cmdname
);
937 (void) vfprintf(stderr
, fmt
, ap
);
939 (void) fprintf(stderr
, "\n");
947 dump_packed_nvlist(objset_t
*os
, uint64_t object
, void *data
, size_t size
)
951 size_t nvsize
= *(uint64_t *)data
;
952 char *packed
= umem_alloc(nvsize
, UMEM_NOFAIL
);
954 VERIFY(0 == dmu_read(os
, object
, 0, nvsize
, packed
, DMU_READ_PREFETCH
));
956 VERIFY(nvlist_unpack(packed
, nvsize
, &nv
, 0) == 0);
958 umem_free(packed
, nvsize
);
966 dump_history_offsets(objset_t
*os
, uint64_t object
, void *data
, size_t size
)
968 (void) os
, (void) object
, (void) size
;
969 spa_history_phys_t
*shp
= data
;
974 (void) printf("\t\tpool_create_len = %llu\n",
975 (u_longlong_t
)shp
->sh_pool_create_len
);
976 (void) printf("\t\tphys_max_off = %llu\n",
977 (u_longlong_t
)shp
->sh_phys_max_off
);
978 (void) printf("\t\tbof = %llu\n",
979 (u_longlong_t
)shp
->sh_bof
);
980 (void) printf("\t\teof = %llu\n",
981 (u_longlong_t
)shp
->sh_eof
);
982 (void) printf("\t\trecords_lost = %llu\n",
983 (u_longlong_t
)shp
->sh_records_lost
);
987 zdb_nicenum(uint64_t num
, char *buf
, size_t buflen
)
990 (void) snprintf(buf
, buflen
, "%llu", (longlong_t
)num
);
992 nicenum(num
, buf
, buflen
);
995 static const char histo_stars
[] = "****************************************";
996 static const uint64_t histo_width
= sizeof (histo_stars
) - 1;
999 dump_histogram(const uint64_t *histo
, int size
, int offset
)
1002 int minidx
= size
- 1;
1006 for (i
= 0; i
< size
; i
++) {
1017 if (max
< histo_width
)
1020 for (i
= minidx
; i
<= maxidx
; i
++) {
1021 (void) printf("\t\t\t%3u: %6llu %s\n",
1022 i
+ offset
, (u_longlong_t
)histo
[i
],
1023 &histo_stars
[(max
- histo
[i
]) * histo_width
/ max
]);
1028 dump_zap_stats(objset_t
*os
, uint64_t object
)
1033 error
= zap_get_stats(os
, object
, &zs
);
1037 if (zs
.zs_ptrtbl_len
== 0) {
1038 ASSERT(zs
.zs_num_blocks
== 1);
1039 (void) printf("\tmicrozap: %llu bytes, %llu entries\n",
1040 (u_longlong_t
)zs
.zs_blocksize
,
1041 (u_longlong_t
)zs
.zs_num_entries
);
1045 (void) printf("\tFat ZAP stats:\n");
1047 (void) printf("\t\tPointer table:\n");
1048 (void) printf("\t\t\t%llu elements\n",
1049 (u_longlong_t
)zs
.zs_ptrtbl_len
);
1050 (void) printf("\t\t\tzt_blk: %llu\n",
1051 (u_longlong_t
)zs
.zs_ptrtbl_zt_blk
);
1052 (void) printf("\t\t\tzt_numblks: %llu\n",
1053 (u_longlong_t
)zs
.zs_ptrtbl_zt_numblks
);
1054 (void) printf("\t\t\tzt_shift: %llu\n",
1055 (u_longlong_t
)zs
.zs_ptrtbl_zt_shift
);
1056 (void) printf("\t\t\tzt_blks_copied: %llu\n",
1057 (u_longlong_t
)zs
.zs_ptrtbl_blks_copied
);
1058 (void) printf("\t\t\tzt_nextblk: %llu\n",
1059 (u_longlong_t
)zs
.zs_ptrtbl_nextblk
);
1061 (void) printf("\t\tZAP entries: %llu\n",
1062 (u_longlong_t
)zs
.zs_num_entries
);
1063 (void) printf("\t\tLeaf blocks: %llu\n",
1064 (u_longlong_t
)zs
.zs_num_leafs
);
1065 (void) printf("\t\tTotal blocks: %llu\n",
1066 (u_longlong_t
)zs
.zs_num_blocks
);
1067 (void) printf("\t\tzap_block_type: 0x%llx\n",
1068 (u_longlong_t
)zs
.zs_block_type
);
1069 (void) printf("\t\tzap_magic: 0x%llx\n",
1070 (u_longlong_t
)zs
.zs_magic
);
1071 (void) printf("\t\tzap_salt: 0x%llx\n",
1072 (u_longlong_t
)zs
.zs_salt
);
1074 (void) printf("\t\tLeafs with 2^n pointers:\n");
1075 dump_histogram(zs
.zs_leafs_with_2n_pointers
, ZAP_HISTOGRAM_SIZE
, 0);
1077 (void) printf("\t\tBlocks with n*5 entries:\n");
1078 dump_histogram(zs
.zs_blocks_with_n5_entries
, ZAP_HISTOGRAM_SIZE
, 0);
1080 (void) printf("\t\tBlocks n/10 full:\n");
1081 dump_histogram(zs
.zs_blocks_n_tenths_full
, ZAP_HISTOGRAM_SIZE
, 0);
1083 (void) printf("\t\tEntries with n chunks:\n");
1084 dump_histogram(zs
.zs_entries_using_n_chunks
, ZAP_HISTOGRAM_SIZE
, 0);
1086 (void) printf("\t\tBuckets with n entries:\n");
1087 dump_histogram(zs
.zs_buckets_with_n_entries
, ZAP_HISTOGRAM_SIZE
, 0);
1091 dump_none(objset_t
*os
, uint64_t object
, void *data
, size_t size
)
1093 (void) os
, (void) object
, (void) data
, (void) size
;
1097 dump_unknown(objset_t
*os
, uint64_t object
, void *data
, size_t size
)
1099 (void) os
, (void) object
, (void) data
, (void) size
;
1100 (void) printf("\tUNKNOWN OBJECT TYPE\n");
1104 dump_uint8(objset_t
*os
, uint64_t object
, void *data
, size_t size
)
1106 (void) os
, (void) object
, (void) data
, (void) size
;
1110 dump_uint64(objset_t
*os
, uint64_t object
, void *data
, size_t size
)
1114 if (dump_opt
['d'] < 6)
1118 dmu_object_info_t doi
;
1120 VERIFY0(dmu_object_info(os
, object
, &doi
));
1121 size
= doi
.doi_max_offset
;
1123 * We cap the size at 1 mebibyte here to prevent
1124 * allocation failures and nigh-infinite printing if the
1125 * object is extremely large.
1127 oursize
= MIN(size
, 1 << 20);
1128 arr
= kmem_alloc(oursize
, KM_SLEEP
);
1130 int err
= dmu_read(os
, object
, 0, oursize
, arr
, 0);
1132 (void) printf("got error %u from dmu_read\n", err
);
1133 kmem_free(arr
, oursize
);
1138 * Even though the allocation is already done in this code path,
1139 * we still cap the size to prevent excessive printing.
1141 oursize
= MIN(size
, 1 << 20);
1147 kmem_free(arr
, oursize
);
1148 (void) printf("\t\t[]\n");
1152 (void) printf("\t\t[%0llx", (u_longlong_t
)arr
[0]);
1153 for (size_t i
= 1; i
* sizeof (uint64_t) < oursize
; i
++) {
1155 (void) printf(", %0llx", (u_longlong_t
)arr
[i
]);
1157 (void) printf(",\n\t\t%0llx", (u_longlong_t
)arr
[i
]);
1159 if (oursize
!= size
)
1160 (void) printf(", ... ");
1161 (void) printf("]\n");
1164 kmem_free(arr
, oursize
);
1168 dump_zap(objset_t
*os
, uint64_t object
, void *data
, size_t size
)
1170 (void) data
, (void) size
;
1172 zap_attribute_t attr
;
1176 dump_zap_stats(os
, object
);
1177 (void) printf("\n");
1179 for (zap_cursor_init(&zc
, os
, object
);
1180 zap_cursor_retrieve(&zc
, &attr
) == 0;
1181 zap_cursor_advance(&zc
)) {
1182 (void) printf("\t\t%s = ", attr
.za_name
);
1183 if (attr
.za_num_integers
== 0) {
1184 (void) printf("\n");
1187 prop
= umem_zalloc(attr
.za_num_integers
*
1188 attr
.za_integer_length
, UMEM_NOFAIL
);
1189 (void) zap_lookup(os
, object
, attr
.za_name
,
1190 attr
.za_integer_length
, attr
.za_num_integers
, prop
);
1191 if (attr
.za_integer_length
== 1) {
1192 if (strcmp(attr
.za_name
,
1193 DSL_CRYPTO_KEY_MASTER_KEY
) == 0 ||
1194 strcmp(attr
.za_name
,
1195 DSL_CRYPTO_KEY_HMAC_KEY
) == 0 ||
1196 strcmp(attr
.za_name
, DSL_CRYPTO_KEY_IV
) == 0 ||
1197 strcmp(attr
.za_name
, DSL_CRYPTO_KEY_MAC
) == 0 ||
1198 strcmp(attr
.za_name
, DMU_POOL_CHECKSUM_SALT
) == 0) {
1201 for (i
= 0; i
< attr
.za_num_integers
; i
++) {
1202 (void) printf("%02x", u8
[i
]);
1205 (void) printf("%s", (char *)prop
);
1208 for (i
= 0; i
< attr
.za_num_integers
; i
++) {
1209 switch (attr
.za_integer_length
) {
1211 (void) printf("%u ",
1212 ((uint16_t *)prop
)[i
]);
1215 (void) printf("%u ",
1216 ((uint32_t *)prop
)[i
]);
1219 (void) printf("%lld ",
1220 (u_longlong_t
)((int64_t *)prop
)[i
]);
1225 (void) printf("\n");
1226 umem_free(prop
, attr
.za_num_integers
* attr
.za_integer_length
);
1228 zap_cursor_fini(&zc
);
1232 dump_bpobj(objset_t
*os
, uint64_t object
, void *data
, size_t size
)
1234 bpobj_phys_t
*bpop
= data
;
1236 char bytes
[32], comp
[32], uncomp
[32];
1238 /* make sure the output won't get truncated */
1239 _Static_assert(sizeof (bytes
) >= NN_NUMBUF_SZ
, "bytes truncated");
1240 _Static_assert(sizeof (comp
) >= NN_NUMBUF_SZ
, "comp truncated");
1241 _Static_assert(sizeof (uncomp
) >= NN_NUMBUF_SZ
, "uncomp truncated");
1246 zdb_nicenum(bpop
->bpo_bytes
, bytes
, sizeof (bytes
));
1247 zdb_nicenum(bpop
->bpo_comp
, comp
, sizeof (comp
));
1248 zdb_nicenum(bpop
->bpo_uncomp
, uncomp
, sizeof (uncomp
));
1250 (void) printf("\t\tnum_blkptrs = %llu\n",
1251 (u_longlong_t
)bpop
->bpo_num_blkptrs
);
1252 (void) printf("\t\tbytes = %s\n", bytes
);
1253 if (size
>= BPOBJ_SIZE_V1
) {
1254 (void) printf("\t\tcomp = %s\n", comp
);
1255 (void) printf("\t\tuncomp = %s\n", uncomp
);
1257 if (size
>= BPOBJ_SIZE_V2
) {
1258 (void) printf("\t\tsubobjs = %llu\n",
1259 (u_longlong_t
)bpop
->bpo_subobjs
);
1260 (void) printf("\t\tnum_subobjs = %llu\n",
1261 (u_longlong_t
)bpop
->bpo_num_subobjs
);
1263 if (size
>= sizeof (*bpop
)) {
1264 (void) printf("\t\tnum_freed = %llu\n",
1265 (u_longlong_t
)bpop
->bpo_num_freed
);
1268 if (dump_opt
['d'] < 5)
1271 for (i
= 0; i
< bpop
->bpo_num_blkptrs
; i
++) {
1272 char blkbuf
[BP_SPRINTF_LEN
];
1275 int err
= dmu_read(os
, object
,
1276 i
* sizeof (bp
), sizeof (bp
), &bp
, 0);
1278 (void) printf("got error %u from dmu_read\n", err
);
1281 snprintf_blkptr_compact(blkbuf
, sizeof (blkbuf
), &bp
,
1283 (void) printf("\t%s\n", blkbuf
);
1288 dump_bpobj_subobjs(objset_t
*os
, uint64_t object
, void *data
, size_t size
)
1290 (void) data
, (void) size
;
1291 dmu_object_info_t doi
;
1294 VERIFY0(dmu_object_info(os
, object
, &doi
));
1295 uint64_t *subobjs
= kmem_alloc(doi
.doi_max_offset
, KM_SLEEP
);
1297 int err
= dmu_read(os
, object
, 0, doi
.doi_max_offset
, subobjs
, 0);
1299 (void) printf("got error %u from dmu_read\n", err
);
1300 kmem_free(subobjs
, doi
.doi_max_offset
);
1304 int64_t last_nonzero
= -1;
1305 for (i
= 0; i
< doi
.doi_max_offset
/ 8; i
++) {
1306 if (subobjs
[i
] != 0)
1310 for (i
= 0; i
<= last_nonzero
; i
++) {
1311 (void) printf("\t%llu\n", (u_longlong_t
)subobjs
[i
]);
1313 kmem_free(subobjs
, doi
.doi_max_offset
);
1317 dump_ddt_zap(objset_t
*os
, uint64_t object
, void *data
, size_t size
)
1319 (void) data
, (void) size
;
1320 dump_zap_stats(os
, object
);
1321 /* contents are printed elsewhere, properly decoded */
1325 dump_sa_attrs(objset_t
*os
, uint64_t object
, void *data
, size_t size
)
1327 (void) data
, (void) size
;
1329 zap_attribute_t attr
;
1331 dump_zap_stats(os
, object
);
1332 (void) printf("\n");
1334 for (zap_cursor_init(&zc
, os
, object
);
1335 zap_cursor_retrieve(&zc
, &attr
) == 0;
1336 zap_cursor_advance(&zc
)) {
1337 (void) printf("\t\t%s = ", attr
.za_name
);
1338 if (attr
.za_num_integers
== 0) {
1339 (void) printf("\n");
1342 (void) printf(" %llx : [%d:%d:%d]\n",
1343 (u_longlong_t
)attr
.za_first_integer
,
1344 (int)ATTR_LENGTH(attr
.za_first_integer
),
1345 (int)ATTR_BSWAP(attr
.za_first_integer
),
1346 (int)ATTR_NUM(attr
.za_first_integer
));
1348 zap_cursor_fini(&zc
);
1352 dump_sa_layouts(objset_t
*os
, uint64_t object
, void *data
, size_t size
)
1354 (void) data
, (void) size
;
1356 zap_attribute_t attr
;
1357 uint16_t *layout_attrs
;
1360 dump_zap_stats(os
, object
);
1361 (void) printf("\n");
1363 for (zap_cursor_init(&zc
, os
, object
);
1364 zap_cursor_retrieve(&zc
, &attr
) == 0;
1365 zap_cursor_advance(&zc
)) {
1366 (void) printf("\t\t%s = [", attr
.za_name
);
1367 if (attr
.za_num_integers
== 0) {
1368 (void) printf("\n");
1372 VERIFY(attr
.za_integer_length
== 2);
1373 layout_attrs
= umem_zalloc(attr
.za_num_integers
*
1374 attr
.za_integer_length
, UMEM_NOFAIL
);
1376 VERIFY(zap_lookup(os
, object
, attr
.za_name
,
1377 attr
.za_integer_length
,
1378 attr
.za_num_integers
, layout_attrs
) == 0);
1380 for (i
= 0; i
!= attr
.za_num_integers
; i
++)
1381 (void) printf(" %d ", (int)layout_attrs
[i
]);
1382 (void) printf("]\n");
1383 umem_free(layout_attrs
,
1384 attr
.za_num_integers
* attr
.za_integer_length
);
1386 zap_cursor_fini(&zc
);
1390 dump_zpldir(objset_t
*os
, uint64_t object
, void *data
, size_t size
)
1392 (void) data
, (void) size
;
1394 zap_attribute_t attr
;
1395 const char *typenames
[] = {
1396 /* 0 */ "not specified",
1398 /* 2 */ "Character Device",
1399 /* 3 */ "3 (invalid)",
1400 /* 4 */ "Directory",
1401 /* 5 */ "5 (invalid)",
1402 /* 6 */ "Block Device",
1403 /* 7 */ "7 (invalid)",
1404 /* 8 */ "Regular File",
1405 /* 9 */ "9 (invalid)",
1406 /* 10 */ "Symbolic Link",
1407 /* 11 */ "11 (invalid)",
1410 /* 14 */ "Event Port",
1411 /* 15 */ "15 (invalid)",
1414 dump_zap_stats(os
, object
);
1415 (void) printf("\n");
1417 for (zap_cursor_init(&zc
, os
, object
);
1418 zap_cursor_retrieve(&zc
, &attr
) == 0;
1419 zap_cursor_advance(&zc
)) {
1420 (void) printf("\t\t%s = %lld (type: %s)\n",
1421 attr
.za_name
, ZFS_DIRENT_OBJ(attr
.za_first_integer
),
1422 typenames
[ZFS_DIRENT_TYPE(attr
.za_first_integer
)]);
1424 zap_cursor_fini(&zc
);
1428 get_dtl_refcount(vdev_t
*vd
)
1432 if (vd
->vdev_ops
->vdev_op_leaf
) {
1433 space_map_t
*sm
= vd
->vdev_dtl_sm
;
1436 sm
->sm_dbuf
->db_size
== sizeof (space_map_phys_t
))
1441 for (unsigned c
= 0; c
< vd
->vdev_children
; c
++)
1442 refcount
+= get_dtl_refcount(vd
->vdev_child
[c
]);
1447 get_metaslab_refcount(vdev_t
*vd
)
1451 if (vd
->vdev_top
== vd
) {
1452 for (uint64_t m
= 0; m
< vd
->vdev_ms_count
; m
++) {
1453 space_map_t
*sm
= vd
->vdev_ms
[m
]->ms_sm
;
1456 sm
->sm_dbuf
->db_size
== sizeof (space_map_phys_t
))
1460 for (unsigned c
= 0; c
< vd
->vdev_children
; c
++)
1461 refcount
+= get_metaslab_refcount(vd
->vdev_child
[c
]);
1467 get_obsolete_refcount(vdev_t
*vd
)
1469 uint64_t obsolete_sm_object
;
1472 VERIFY0(vdev_obsolete_sm_object(vd
, &obsolete_sm_object
));
1473 if (vd
->vdev_top
== vd
&& obsolete_sm_object
!= 0) {
1474 dmu_object_info_t doi
;
1475 VERIFY0(dmu_object_info(vd
->vdev_spa
->spa_meta_objset
,
1476 obsolete_sm_object
, &doi
));
1477 if (doi
.doi_bonus_size
== sizeof (space_map_phys_t
)) {
1481 ASSERT3P(vd
->vdev_obsolete_sm
, ==, NULL
);
1482 ASSERT3U(obsolete_sm_object
, ==, 0);
1484 for (unsigned c
= 0; c
< vd
->vdev_children
; c
++) {
1485 refcount
+= get_obsolete_refcount(vd
->vdev_child
[c
]);
1492 get_prev_obsolete_spacemap_refcount(spa_t
*spa
)
1495 spa
->spa_condensing_indirect_phys
.scip_prev_obsolete_sm_object
;
1496 if (prev_obj
!= 0) {
1497 dmu_object_info_t doi
;
1498 VERIFY0(dmu_object_info(spa
->spa_meta_objset
, prev_obj
, &doi
));
1499 if (doi
.doi_bonus_size
== sizeof (space_map_phys_t
)) {
1507 get_checkpoint_refcount(vdev_t
*vd
)
1511 if (vd
->vdev_top
== vd
&& vd
->vdev_top_zap
!= 0 &&
1512 zap_contains(spa_meta_objset(vd
->vdev_spa
),
1513 vd
->vdev_top_zap
, VDEV_TOP_ZAP_POOL_CHECKPOINT_SM
) == 0)
1516 for (uint64_t c
= 0; c
< vd
->vdev_children
; c
++)
1517 refcount
+= get_checkpoint_refcount(vd
->vdev_child
[c
]);
1523 get_log_spacemap_refcount(spa_t
*spa
)
1525 return (avl_numnodes(&spa
->spa_sm_logs_by_txg
));
1529 verify_spacemap_refcounts(spa_t
*spa
)
1531 uint64_t expected_refcount
= 0;
1532 uint64_t actual_refcount
;
1534 (void) feature_get_refcount(spa
,
1535 &spa_feature_table
[SPA_FEATURE_SPACEMAP_HISTOGRAM
],
1536 &expected_refcount
);
1537 actual_refcount
= get_dtl_refcount(spa
->spa_root_vdev
);
1538 actual_refcount
+= get_metaslab_refcount(spa
->spa_root_vdev
);
1539 actual_refcount
+= get_obsolete_refcount(spa
->spa_root_vdev
);
1540 actual_refcount
+= get_prev_obsolete_spacemap_refcount(spa
);
1541 actual_refcount
+= get_checkpoint_refcount(spa
->spa_root_vdev
);
1542 actual_refcount
+= get_log_spacemap_refcount(spa
);
1544 if (expected_refcount
!= actual_refcount
) {
1545 (void) printf("space map refcount mismatch: expected %lld != "
1547 (longlong_t
)expected_refcount
,
1548 (longlong_t
)actual_refcount
);
1555 dump_spacemap(objset_t
*os
, space_map_t
*sm
)
1557 const char *ddata
[] = { "ALLOC", "FREE", "CONDENSE", "INVALID",
1558 "INVALID", "INVALID", "INVALID", "INVALID" };
1563 (void) printf("space map object %llu:\n",
1564 (longlong_t
)sm
->sm_object
);
1565 (void) printf(" smp_length = 0x%llx\n",
1566 (longlong_t
)sm
->sm_phys
->smp_length
);
1567 (void) printf(" smp_alloc = 0x%llx\n",
1568 (longlong_t
)sm
->sm_phys
->smp_alloc
);
1570 if (dump_opt
['d'] < 6 && dump_opt
['m'] < 4)
1574 * Print out the freelist entries in both encoded and decoded form.
1576 uint8_t mapshift
= sm
->sm_shift
;
1578 uint64_t word
, entry_id
= 0;
1579 for (uint64_t offset
= 0; offset
< space_map_length(sm
);
1580 offset
+= sizeof (word
)) {
1582 VERIFY0(dmu_read(os
, space_map_object(sm
), offset
,
1583 sizeof (word
), &word
, DMU_READ_PREFETCH
));
1585 if (sm_entry_is_debug(word
)) {
1586 uint64_t de_txg
= SM_DEBUG_TXG_DECODE(word
);
1587 uint64_t de_sync_pass
= SM_DEBUG_SYNCPASS_DECODE(word
);
1590 "\t [%6llu] PADDING\n",
1591 (u_longlong_t
)entry_id
);
1594 "\t [%6llu] %s: txg %llu pass %llu\n",
1595 (u_longlong_t
)entry_id
,
1596 ddata
[SM_DEBUG_ACTION_DECODE(word
)],
1597 (u_longlong_t
)de_txg
,
1598 (u_longlong_t
)de_sync_pass
);
1606 uint64_t entry_off
, entry_run
, entry_vdev
= SM_NO_VDEVID
;
1608 if (sm_entry_is_single_word(word
)) {
1609 entry_type
= (SM_TYPE_DECODE(word
) == SM_ALLOC
) ?
1611 entry_off
= (SM_OFFSET_DECODE(word
) << mapshift
) +
1613 entry_run
= SM_RUN_DECODE(word
) << mapshift
;
1616 /* it is a two-word entry so we read another word */
1617 ASSERT(sm_entry_is_double_word(word
));
1619 uint64_t extra_word
;
1620 offset
+= sizeof (extra_word
);
1621 VERIFY0(dmu_read(os
, space_map_object(sm
), offset
,
1622 sizeof (extra_word
), &extra_word
,
1623 DMU_READ_PREFETCH
));
1625 ASSERT3U(offset
, <=, space_map_length(sm
));
1627 entry_run
= SM2_RUN_DECODE(word
) << mapshift
;
1628 entry_vdev
= SM2_VDEV_DECODE(word
);
1629 entry_type
= (SM2_TYPE_DECODE(extra_word
) == SM_ALLOC
) ?
1631 entry_off
= (SM2_OFFSET_DECODE(extra_word
) <<
1632 mapshift
) + sm
->sm_start
;
1636 (void) printf("\t [%6llu] %c range:"
1637 " %010llx-%010llx size: %06llx vdev: %06llu words: %u\n",
1638 (u_longlong_t
)entry_id
,
1639 entry_type
, (u_longlong_t
)entry_off
,
1640 (u_longlong_t
)(entry_off
+ entry_run
),
1641 (u_longlong_t
)entry_run
,
1642 (u_longlong_t
)entry_vdev
, words
);
1644 if (entry_type
== 'A')
1650 if (alloc
!= space_map_allocated(sm
)) {
1651 (void) printf("space_map_object alloc (%lld) INCONSISTENT "
1652 "with space map summary (%lld)\n",
1653 (longlong_t
)space_map_allocated(sm
), (longlong_t
)alloc
);
1658 dump_metaslab_stats(metaslab_t
*msp
)
1661 range_tree_t
*rt
= msp
->ms_allocatable
;
1662 zfs_btree_t
*t
= &msp
->ms_allocatable_by_size
;
1663 int free_pct
= range_tree_space(rt
) * 100 / msp
->ms_size
;
1665 /* max sure nicenum has enough space */
1666 _Static_assert(sizeof (maxbuf
) >= NN_NUMBUF_SZ
, "maxbuf truncated");
1668 zdb_nicenum(metaslab_largest_allocatable(msp
), maxbuf
, sizeof (maxbuf
));
1670 (void) printf("\t %25s %10lu %7s %6s %4s %4d%%\n",
1671 "segments", zfs_btree_numnodes(t
), "maxsize", maxbuf
,
1672 "freepct", free_pct
);
1673 (void) printf("\tIn-memory histogram:\n");
1674 dump_histogram(rt
->rt_histogram
, RANGE_TREE_HISTOGRAM_SIZE
, 0);
1678 dump_metaslab(metaslab_t
*msp
)
1680 vdev_t
*vd
= msp
->ms_group
->mg_vd
;
1681 spa_t
*spa
= vd
->vdev_spa
;
1682 space_map_t
*sm
= msp
->ms_sm
;
1685 zdb_nicenum(msp
->ms_size
- space_map_allocated(sm
), freebuf
,
1689 "\tmetaslab %6llu offset %12llx spacemap %6llu free %5s\n",
1690 (u_longlong_t
)msp
->ms_id
, (u_longlong_t
)msp
->ms_start
,
1691 (u_longlong_t
)space_map_object(sm
), freebuf
);
1693 if (dump_opt
['m'] > 2 && !dump_opt
['L']) {
1694 mutex_enter(&msp
->ms_lock
);
1695 VERIFY0(metaslab_load(msp
));
1696 range_tree_stat_verify(msp
->ms_allocatable
);
1697 dump_metaslab_stats(msp
);
1698 metaslab_unload(msp
);
1699 mutex_exit(&msp
->ms_lock
);
1702 if (dump_opt
['m'] > 1 && sm
!= NULL
&&
1703 spa_feature_is_active(spa
, SPA_FEATURE_SPACEMAP_HISTOGRAM
)) {
1705 * The space map histogram represents free space in chunks
1706 * of sm_shift (i.e. bucket 0 refers to 2^sm_shift).
1708 (void) printf("\tOn-disk histogram:\t\tfragmentation %llu\n",
1709 (u_longlong_t
)msp
->ms_fragmentation
);
1710 dump_histogram(sm
->sm_phys
->smp_histogram
,
1711 SPACE_MAP_HISTOGRAM_SIZE
, sm
->sm_shift
);
1714 if (vd
->vdev_ops
== &vdev_draid_ops
)
1715 ASSERT3U(msp
->ms_size
, <=, 1ULL << vd
->vdev_ms_shift
);
1717 ASSERT3U(msp
->ms_size
, ==, 1ULL << vd
->vdev_ms_shift
);
1719 dump_spacemap(spa
->spa_meta_objset
, msp
->ms_sm
);
1721 if (spa_feature_is_active(spa
, SPA_FEATURE_LOG_SPACEMAP
)) {
1722 (void) printf("\tFlush data:\n\tunflushed txg=%llu\n\n",
1723 (u_longlong_t
)metaslab_unflushed_txg(msp
));
1728 print_vdev_metaslab_header(vdev_t
*vd
)
1730 vdev_alloc_bias_t alloc_bias
= vd
->vdev_alloc_bias
;
1731 const char *bias_str
= "";
1732 if (alloc_bias
== VDEV_BIAS_LOG
|| vd
->vdev_islog
) {
1733 bias_str
= VDEV_ALLOC_BIAS_LOG
;
1734 } else if (alloc_bias
== VDEV_BIAS_SPECIAL
) {
1735 bias_str
= VDEV_ALLOC_BIAS_SPECIAL
;
1736 } else if (alloc_bias
== VDEV_BIAS_DEDUP
) {
1737 bias_str
= VDEV_ALLOC_BIAS_DEDUP
;
1740 uint64_t ms_flush_data_obj
= 0;
1741 if (vd
->vdev_top_zap
!= 0) {
1742 int error
= zap_lookup(spa_meta_objset(vd
->vdev_spa
),
1743 vd
->vdev_top_zap
, VDEV_TOP_ZAP_MS_UNFLUSHED_PHYS_TXGS
,
1744 sizeof (uint64_t), 1, &ms_flush_data_obj
);
1745 if (error
!= ENOENT
) {
1750 (void) printf("\tvdev %10llu %s",
1751 (u_longlong_t
)vd
->vdev_id
, bias_str
);
1753 if (ms_flush_data_obj
!= 0) {
1754 (void) printf(" ms_unflushed_phys object %llu",
1755 (u_longlong_t
)ms_flush_data_obj
);
1758 (void) printf("\n\t%-10s%5llu %-19s %-15s %-12s\n",
1759 "metaslabs", (u_longlong_t
)vd
->vdev_ms_count
,
1760 "offset", "spacemap", "free");
1761 (void) printf("\t%15s %19s %15s %12s\n",
1762 "---------------", "-------------------",
1763 "---------------", "------------");
1767 dump_metaslab_groups(spa_t
*spa
, boolean_t show_special
)
1769 vdev_t
*rvd
= spa
->spa_root_vdev
;
1770 metaslab_class_t
*mc
= spa_normal_class(spa
);
1771 metaslab_class_t
*smc
= spa_special_class(spa
);
1772 uint64_t fragmentation
;
1774 metaslab_class_histogram_verify(mc
);
1776 for (unsigned c
= 0; c
< rvd
->vdev_children
; c
++) {
1777 vdev_t
*tvd
= rvd
->vdev_child
[c
];
1778 metaslab_group_t
*mg
= tvd
->vdev_mg
;
1780 if (mg
== NULL
|| (mg
->mg_class
!= mc
&&
1781 (!show_special
|| mg
->mg_class
!= smc
)))
1784 metaslab_group_histogram_verify(mg
);
1785 mg
->mg_fragmentation
= metaslab_group_fragmentation(mg
);
1787 (void) printf("\tvdev %10llu\t\tmetaslabs%5llu\t\t"
1789 (u_longlong_t
)tvd
->vdev_id
,
1790 (u_longlong_t
)tvd
->vdev_ms_count
);
1791 if (mg
->mg_fragmentation
== ZFS_FRAG_INVALID
) {
1792 (void) printf("%3s\n", "-");
1794 (void) printf("%3llu%%\n",
1795 (u_longlong_t
)mg
->mg_fragmentation
);
1797 dump_histogram(mg
->mg_histogram
, RANGE_TREE_HISTOGRAM_SIZE
, 0);
1800 (void) printf("\tpool %s\tfragmentation", spa_name(spa
));
1801 fragmentation
= metaslab_class_fragmentation(mc
);
1802 if (fragmentation
== ZFS_FRAG_INVALID
)
1803 (void) printf("\t%3s\n", "-");
1805 (void) printf("\t%3llu%%\n", (u_longlong_t
)fragmentation
);
1806 dump_histogram(mc
->mc_histogram
, RANGE_TREE_HISTOGRAM_SIZE
, 0);
1810 print_vdev_indirect(vdev_t
*vd
)
1812 vdev_indirect_config_t
*vic
= &vd
->vdev_indirect_config
;
1813 vdev_indirect_mapping_t
*vim
= vd
->vdev_indirect_mapping
;
1814 vdev_indirect_births_t
*vib
= vd
->vdev_indirect_births
;
1817 ASSERT3P(vib
, ==, NULL
);
1821 ASSERT3U(vdev_indirect_mapping_object(vim
), ==,
1822 vic
->vic_mapping_object
);
1823 ASSERT3U(vdev_indirect_births_object(vib
), ==,
1824 vic
->vic_births_object
);
1826 (void) printf("indirect births obj %llu:\n",
1827 (longlong_t
)vic
->vic_births_object
);
1828 (void) printf(" vib_count = %llu\n",
1829 (longlong_t
)vdev_indirect_births_count(vib
));
1830 for (uint64_t i
= 0; i
< vdev_indirect_births_count(vib
); i
++) {
1831 vdev_indirect_birth_entry_phys_t
*cur_vibe
=
1832 &vib
->vib_entries
[i
];
1833 (void) printf("\toffset %llx -> txg %llu\n",
1834 (longlong_t
)cur_vibe
->vibe_offset
,
1835 (longlong_t
)cur_vibe
->vibe_phys_birth_txg
);
1837 (void) printf("\n");
1839 (void) printf("indirect mapping obj %llu:\n",
1840 (longlong_t
)vic
->vic_mapping_object
);
1841 (void) printf(" vim_max_offset = 0x%llx\n",
1842 (longlong_t
)vdev_indirect_mapping_max_offset(vim
));
1843 (void) printf(" vim_bytes_mapped = 0x%llx\n",
1844 (longlong_t
)vdev_indirect_mapping_bytes_mapped(vim
));
1845 (void) printf(" vim_count = %llu\n",
1846 (longlong_t
)vdev_indirect_mapping_num_entries(vim
));
1848 if (dump_opt
['d'] <= 5 && dump_opt
['m'] <= 3)
1851 uint32_t *counts
= vdev_indirect_mapping_load_obsolete_counts(vim
);
1853 for (uint64_t i
= 0; i
< vdev_indirect_mapping_num_entries(vim
); i
++) {
1854 vdev_indirect_mapping_entry_phys_t
*vimep
=
1855 &vim
->vim_entries
[i
];
1856 (void) printf("\t<%llx:%llx:%llx> -> "
1857 "<%llx:%llx:%llx> (%x obsolete)\n",
1858 (longlong_t
)vd
->vdev_id
,
1859 (longlong_t
)DVA_MAPPING_GET_SRC_OFFSET(vimep
),
1860 (longlong_t
)DVA_GET_ASIZE(&vimep
->vimep_dst
),
1861 (longlong_t
)DVA_GET_VDEV(&vimep
->vimep_dst
),
1862 (longlong_t
)DVA_GET_OFFSET(&vimep
->vimep_dst
),
1863 (longlong_t
)DVA_GET_ASIZE(&vimep
->vimep_dst
),
1866 (void) printf("\n");
1868 uint64_t obsolete_sm_object
;
1869 VERIFY0(vdev_obsolete_sm_object(vd
, &obsolete_sm_object
));
1870 if (obsolete_sm_object
!= 0) {
1871 objset_t
*mos
= vd
->vdev_spa
->spa_meta_objset
;
1872 (void) printf("obsolete space map object %llu:\n",
1873 (u_longlong_t
)obsolete_sm_object
);
1874 ASSERT(vd
->vdev_obsolete_sm
!= NULL
);
1875 ASSERT3U(space_map_object(vd
->vdev_obsolete_sm
), ==,
1876 obsolete_sm_object
);
1877 dump_spacemap(mos
, vd
->vdev_obsolete_sm
);
1878 (void) printf("\n");
1883 dump_metaslabs(spa_t
*spa
)
1885 vdev_t
*vd
, *rvd
= spa
->spa_root_vdev
;
1886 uint64_t m
, c
= 0, children
= rvd
->vdev_children
;
1888 (void) printf("\nMetaslabs:\n");
1890 if (!dump_opt
['d'] && zopt_metaslab_args
> 0) {
1891 c
= zopt_metaslab
[0];
1894 (void) fatal("bad vdev id: %llu", (u_longlong_t
)c
);
1896 if (zopt_metaslab_args
> 1) {
1897 vd
= rvd
->vdev_child
[c
];
1898 print_vdev_metaslab_header(vd
);
1900 for (m
= 1; m
< zopt_metaslab_args
; m
++) {
1901 if (zopt_metaslab
[m
] < vd
->vdev_ms_count
)
1903 vd
->vdev_ms
[zopt_metaslab
[m
]]);
1905 (void) fprintf(stderr
, "bad metaslab "
1907 (u_longlong_t
)zopt_metaslab
[m
]);
1909 (void) printf("\n");
1914 for (; c
< children
; c
++) {
1915 vd
= rvd
->vdev_child
[c
];
1916 print_vdev_metaslab_header(vd
);
1918 print_vdev_indirect(vd
);
1920 for (m
= 0; m
< vd
->vdev_ms_count
; m
++)
1921 dump_metaslab(vd
->vdev_ms
[m
]);
1922 (void) printf("\n");
1927 dump_log_spacemaps(spa_t
*spa
)
1929 if (!spa_feature_is_active(spa
, SPA_FEATURE_LOG_SPACEMAP
))
1932 (void) printf("\nLog Space Maps in Pool:\n");
1933 for (spa_log_sm_t
*sls
= avl_first(&spa
->spa_sm_logs_by_txg
);
1934 sls
; sls
= AVL_NEXT(&spa
->spa_sm_logs_by_txg
, sls
)) {
1935 space_map_t
*sm
= NULL
;
1936 VERIFY0(space_map_open(&sm
, spa_meta_objset(spa
),
1937 sls
->sls_sm_obj
, 0, UINT64_MAX
, SPA_MINBLOCKSHIFT
));
1939 (void) printf("Log Spacemap object %llu txg %llu\n",
1940 (u_longlong_t
)sls
->sls_sm_obj
, (u_longlong_t
)sls
->sls_txg
);
1941 dump_spacemap(spa
->spa_meta_objset
, sm
);
1942 space_map_close(sm
);
1944 (void) printf("\n");
1948 dump_dde(const ddt_t
*ddt
, const ddt_entry_t
*dde
, uint64_t index
)
1950 const ddt_phys_t
*ddp
= dde
->dde_phys
;
1951 const ddt_key_t
*ddk
= &dde
->dde_key
;
1952 const char *types
[4] = { "ditto", "single", "double", "triple" };
1953 char blkbuf
[BP_SPRINTF_LEN
];
1957 for (p
= 0; p
< DDT_PHYS_TYPES
; p
++, ddp
++) {
1958 if (ddp
->ddp_phys_birth
== 0)
1960 ddt_bp_create(ddt
->ddt_checksum
, ddk
, ddp
, &blk
);
1961 snprintf_blkptr(blkbuf
, sizeof (blkbuf
), &blk
);
1962 (void) printf("index %llx refcnt %llu %s %s\n",
1963 (u_longlong_t
)index
, (u_longlong_t
)ddp
->ddp_refcnt
,
1969 dump_dedup_ratio(const ddt_stat_t
*dds
)
1971 double rL
, rP
, rD
, D
, dedup
, compress
, copies
;
1973 if (dds
->dds_blocks
== 0)
1976 rL
= (double)dds
->dds_ref_lsize
;
1977 rP
= (double)dds
->dds_ref_psize
;
1978 rD
= (double)dds
->dds_ref_dsize
;
1979 D
= (double)dds
->dds_dsize
;
1985 (void) printf("dedup = %.2f, compress = %.2f, copies = %.2f, "
1986 "dedup * compress / copies = %.2f\n\n",
1987 dedup
, compress
, copies
, dedup
* compress
/ copies
);
1991 dump_ddt(ddt_t
*ddt
, enum ddt_type type
, enum ddt_class
class)
1993 char name
[DDT_NAMELEN
];
1996 dmu_object_info_t doi
;
1997 uint64_t count
, dspace
, mspace
;
2000 error
= ddt_object_info(ddt
, type
, class, &doi
);
2002 if (error
== ENOENT
)
2006 error
= ddt_object_count(ddt
, type
, class, &count
);
2011 dspace
= doi
.doi_physical_blocks_512
<< 9;
2012 mspace
= doi
.doi_fill_count
* doi
.doi_data_block_size
;
2014 ddt_object_name(ddt
, type
, class, name
);
2016 (void) printf("%s: %llu entries, size %llu on disk, %llu in core\n",
2018 (u_longlong_t
)count
,
2019 (u_longlong_t
)(dspace
/ count
),
2020 (u_longlong_t
)(mspace
/ count
));
2022 if (dump_opt
['D'] < 3)
2025 zpool_dump_ddt(NULL
, &ddt
->ddt_histogram
[type
][class]);
2027 if (dump_opt
['D'] < 4)
2030 if (dump_opt
['D'] < 5 && class == DDT_CLASS_UNIQUE
)
2033 (void) printf("%s contents:\n\n", name
);
2035 while ((error
= ddt_object_walk(ddt
, type
, class, &walk
, &dde
)) == 0)
2036 dump_dde(ddt
, &dde
, walk
);
2038 ASSERT3U(error
, ==, ENOENT
);
2040 (void) printf("\n");
2044 dump_all_ddts(spa_t
*spa
)
2046 ddt_histogram_t ddh_total
= {{{0}}};
2047 ddt_stat_t dds_total
= {0};
2049 for (enum zio_checksum c
= 0; c
< ZIO_CHECKSUM_FUNCTIONS
; c
++) {
2050 ddt_t
*ddt
= spa
->spa_ddt
[c
];
2051 for (enum ddt_type type
= 0; type
< DDT_TYPES
; type
++) {
2052 for (enum ddt_class
class = 0; class < DDT_CLASSES
;
2054 dump_ddt(ddt
, type
, class);
2059 ddt_get_dedup_stats(spa
, &dds_total
);
2061 if (dds_total
.dds_blocks
== 0) {
2062 (void) printf("All DDTs are empty\n");
2066 (void) printf("\n");
2068 if (dump_opt
['D'] > 1) {
2069 (void) printf("DDT histogram (aggregated over all DDTs):\n");
2070 ddt_get_dedup_histogram(spa
, &ddh_total
);
2071 zpool_dump_ddt(&dds_total
, &ddh_total
);
2074 dump_dedup_ratio(&dds_total
);
2078 dump_dtl_seg(void *arg
, uint64_t start
, uint64_t size
)
2082 (void) printf("%s [%llu,%llu) length %llu\n",
2084 (u_longlong_t
)start
,
2085 (u_longlong_t
)(start
+ size
),
2086 (u_longlong_t
)(size
));
2090 dump_dtl(vdev_t
*vd
, int indent
)
2092 spa_t
*spa
= vd
->vdev_spa
;
2094 const char *name
[DTL_TYPES
] = { "missing", "partial", "scrub",
2098 spa_vdev_state_enter(spa
, SCL_NONE
);
2099 required
= vdev_dtl_required(vd
);
2100 (void) spa_vdev_state_exit(spa
, NULL
, 0);
2103 (void) printf("\nDirty time logs:\n\n");
2105 (void) printf("\t%*s%s [%s]\n", indent
, "",
2106 vd
->vdev_path
? vd
->vdev_path
:
2107 vd
->vdev_parent
? vd
->vdev_ops
->vdev_op_type
: spa_name(spa
),
2108 required
? "DTL-required" : "DTL-expendable");
2110 for (int t
= 0; t
< DTL_TYPES
; t
++) {
2111 range_tree_t
*rt
= vd
->vdev_dtl
[t
];
2112 if (range_tree_space(rt
) == 0)
2114 (void) snprintf(prefix
, sizeof (prefix
), "\t%*s%s",
2115 indent
+ 2, "", name
[t
]);
2116 range_tree_walk(rt
, dump_dtl_seg
, prefix
);
2117 if (dump_opt
['d'] > 5 && vd
->vdev_children
== 0)
2118 dump_spacemap(spa
->spa_meta_objset
,
2122 for (unsigned c
= 0; c
< vd
->vdev_children
; c
++)
2123 dump_dtl(vd
->vdev_child
[c
], indent
+ 4);
2127 dump_history(spa_t
*spa
)
2129 nvlist_t
**events
= NULL
;
2131 uint64_t resid
, len
, off
= 0;
2136 if ((buf
= malloc(SPA_OLD_MAXBLOCKSIZE
)) == NULL
) {
2137 (void) fprintf(stderr
, "%s: unable to allocate I/O buffer\n",
2143 len
= SPA_OLD_MAXBLOCKSIZE
;
2145 if ((error
= spa_history_get(spa
, &off
, &len
, buf
)) != 0) {
2146 (void) fprintf(stderr
, "Unable to read history: "
2147 "error %d\n", error
);
2152 if (zpool_history_unpack(buf
, len
, &resid
, &events
, &num
) != 0)
2158 (void) printf("\nHistory:\n");
2159 for (unsigned i
= 0; i
< num
; i
++) {
2160 boolean_t printed
= B_FALSE
;
2162 if (nvlist_exists(events
[i
], ZPOOL_HIST_TIME
)) {
2166 tsec
= fnvlist_lookup_uint64(events
[i
],
2168 (void) localtime_r(&tsec
, &t
);
2169 (void) strftime(tbuf
, sizeof (tbuf
), "%F.%T", &t
);
2174 if (nvlist_exists(events
[i
], ZPOOL_HIST_CMD
)) {
2175 (void) printf("%s %s\n", tbuf
,
2176 fnvlist_lookup_string(events
[i
], ZPOOL_HIST_CMD
));
2177 } else if (nvlist_exists(events
[i
], ZPOOL_HIST_INT_EVENT
)) {
2180 ievent
= fnvlist_lookup_uint64(events
[i
],
2181 ZPOOL_HIST_INT_EVENT
);
2182 if (ievent
>= ZFS_NUM_LEGACY_HISTORY_EVENTS
)
2185 (void) printf(" %s [internal %s txg:%ju] %s\n",
2187 zfs_history_event_names
[ievent
],
2188 fnvlist_lookup_uint64(events
[i
],
2190 fnvlist_lookup_string(events
[i
],
2191 ZPOOL_HIST_INT_STR
));
2192 } else if (nvlist_exists(events
[i
], ZPOOL_HIST_INT_NAME
)) {
2193 (void) printf("%s [txg:%ju] %s", tbuf
,
2194 fnvlist_lookup_uint64(events
[i
],
2196 fnvlist_lookup_string(events
[i
],
2197 ZPOOL_HIST_INT_NAME
));
2199 if (nvlist_exists(events
[i
], ZPOOL_HIST_DSNAME
)) {
2200 (void) printf(" %s (%llu)",
2201 fnvlist_lookup_string(events
[i
],
2203 (u_longlong_t
)fnvlist_lookup_uint64(
2208 (void) printf(" %s\n", fnvlist_lookup_string(events
[i
],
2209 ZPOOL_HIST_INT_STR
));
2210 } else if (nvlist_exists(events
[i
], ZPOOL_HIST_IOCTL
)) {
2211 (void) printf("%s ioctl %s\n", tbuf
,
2212 fnvlist_lookup_string(events
[i
],
2215 if (nvlist_exists(events
[i
], ZPOOL_HIST_INPUT_NVL
)) {
2216 (void) printf(" input:\n");
2217 dump_nvlist(fnvlist_lookup_nvlist(events
[i
],
2218 ZPOOL_HIST_INPUT_NVL
), 8);
2220 if (nvlist_exists(events
[i
], ZPOOL_HIST_OUTPUT_NVL
)) {
2221 (void) printf(" output:\n");
2222 dump_nvlist(fnvlist_lookup_nvlist(events
[i
],
2223 ZPOOL_HIST_OUTPUT_NVL
), 8);
2225 if (nvlist_exists(events
[i
], ZPOOL_HIST_ERRNO
)) {
2226 (void) printf(" errno: %lld\n",
2227 (longlong_t
)fnvlist_lookup_int64(events
[i
],
2236 if (dump_opt
['h'] > 1) {
2238 (void) printf("unrecognized record:\n");
2239 dump_nvlist(events
[i
], 2);
2246 dump_dnode(objset_t
*os
, uint64_t object
, void *data
, size_t size
)
2248 (void) os
, (void) object
, (void) data
, (void) size
;
2252 blkid2offset(const dnode_phys_t
*dnp
, const blkptr_t
*bp
,
2253 const zbookmark_phys_t
*zb
)
2256 ASSERT(zb
->zb_level
< 0);
2257 if (zb
->zb_object
== 0)
2258 return (zb
->zb_blkid
);
2259 return (zb
->zb_blkid
* BP_GET_LSIZE(bp
));
2262 ASSERT(zb
->zb_level
>= 0);
2264 return ((zb
->zb_blkid
<<
2265 (zb
->zb_level
* (dnp
->dn_indblkshift
- SPA_BLKPTRSHIFT
))) *
2266 dnp
->dn_datablkszsec
<< SPA_MINBLOCKSHIFT
);
2270 snprintf_zstd_header(spa_t
*spa
, char *blkbuf
, size_t buflen
,
2276 zfs_zstdhdr_t zstd_hdr
;
2279 if (BP_GET_COMPRESS(bp
) != ZIO_COMPRESS_ZSTD
)
2285 if (BP_IS_EMBEDDED(bp
)) {
2286 buf
= malloc(SPA_MAXBLOCKSIZE
);
2288 (void) fprintf(stderr
, "out of memory\n");
2291 decode_embedded_bp_compressed(bp
, buf
);
2292 memcpy(&zstd_hdr
, buf
, sizeof (zstd_hdr
));
2294 zstd_hdr
.c_len
= BE_32(zstd_hdr
.c_len
);
2295 zstd_hdr
.raw_version_level
= BE_32(zstd_hdr
.raw_version_level
);
2296 (void) snprintf(blkbuf
+ strlen(blkbuf
),
2297 buflen
- strlen(blkbuf
),
2298 " ZSTD:size=%u:version=%u:level=%u:EMBEDDED",
2299 zstd_hdr
.c_len
, zfs_get_hdrversion(&zstd_hdr
),
2300 zfs_get_hdrlevel(&zstd_hdr
));
2304 pabd
= abd_alloc_for_io(SPA_MAXBLOCKSIZE
, B_FALSE
);
2305 zio
= zio_root(spa
, NULL
, NULL
, 0);
2307 /* Decrypt but don't decompress so we can read the compression header */
2308 zio_nowait(zio_read(zio
, spa
, bp
, pabd
, BP_GET_PSIZE(bp
), NULL
, NULL
,
2309 ZIO_PRIORITY_SYNC_READ
, ZIO_FLAG_CANFAIL
| ZIO_FLAG_RAW_COMPRESS
,
2311 error
= zio_wait(zio
);
2313 (void) fprintf(stderr
, "read failed: %d\n", error
);
2316 buf
= abd_borrow_buf_copy(pabd
, BP_GET_LSIZE(bp
));
2317 memcpy(&zstd_hdr
, buf
, sizeof (zstd_hdr
));
2318 zstd_hdr
.c_len
= BE_32(zstd_hdr
.c_len
);
2319 zstd_hdr
.raw_version_level
= BE_32(zstd_hdr
.raw_version_level
);
2321 (void) snprintf(blkbuf
+ strlen(blkbuf
),
2322 buflen
- strlen(blkbuf
),
2323 " ZSTD:size=%u:version=%u:level=%u:NORMAL",
2324 zstd_hdr
.c_len
, zfs_get_hdrversion(&zstd_hdr
),
2325 zfs_get_hdrlevel(&zstd_hdr
));
2327 abd_return_buf_copy(pabd
, buf
, BP_GET_LSIZE(bp
));
2331 snprintf_blkptr_compact(char *blkbuf
, size_t buflen
, const blkptr_t
*bp
,
2334 const dva_t
*dva
= bp
->blk_dva
;
2335 int ndvas
= dump_opt
['d'] > 5 ? BP_GET_NDVAS(bp
) : 1;
2338 if (dump_opt
['b'] >= 6) {
2339 snprintf_blkptr(blkbuf
, buflen
, bp
);
2341 (void) snprintf(blkbuf
+ strlen(blkbuf
),
2342 buflen
- strlen(blkbuf
), " %s", "FREE");
2347 if (BP_IS_EMBEDDED(bp
)) {
2348 (void) sprintf(blkbuf
,
2349 "EMBEDDED et=%u %llxL/%llxP B=%llu",
2350 (int)BPE_GET_ETYPE(bp
),
2351 (u_longlong_t
)BPE_GET_LSIZE(bp
),
2352 (u_longlong_t
)BPE_GET_PSIZE(bp
),
2353 (u_longlong_t
)bp
->blk_birth
);
2359 for (i
= 0; i
< ndvas
; i
++)
2360 (void) snprintf(blkbuf
+ strlen(blkbuf
),
2361 buflen
- strlen(blkbuf
), "%llu:%llx:%llx ",
2362 (u_longlong_t
)DVA_GET_VDEV(&dva
[i
]),
2363 (u_longlong_t
)DVA_GET_OFFSET(&dva
[i
]),
2364 (u_longlong_t
)DVA_GET_ASIZE(&dva
[i
]));
2366 if (BP_IS_HOLE(bp
)) {
2367 (void) snprintf(blkbuf
+ strlen(blkbuf
),
2368 buflen
- strlen(blkbuf
),
2370 (u_longlong_t
)BP_GET_LSIZE(bp
),
2371 (u_longlong_t
)bp
->blk_birth
);
2373 (void) snprintf(blkbuf
+ strlen(blkbuf
),
2374 buflen
- strlen(blkbuf
),
2375 "%llxL/%llxP F=%llu B=%llu/%llu",
2376 (u_longlong_t
)BP_GET_LSIZE(bp
),
2377 (u_longlong_t
)BP_GET_PSIZE(bp
),
2378 (u_longlong_t
)BP_GET_FILL(bp
),
2379 (u_longlong_t
)bp
->blk_birth
,
2380 (u_longlong_t
)BP_PHYSICAL_BIRTH(bp
));
2382 (void) snprintf(blkbuf
+ strlen(blkbuf
),
2383 buflen
- strlen(blkbuf
), " %s", "FREE");
2384 (void) snprintf(blkbuf
+ strlen(blkbuf
),
2385 buflen
- strlen(blkbuf
),
2386 " cksum=%016llx:%016llx:%016llx:%016llx",
2387 (u_longlong_t
)bp
->blk_cksum
.zc_word
[0],
2388 (u_longlong_t
)bp
->blk_cksum
.zc_word
[1],
2389 (u_longlong_t
)bp
->blk_cksum
.zc_word
[2],
2390 (u_longlong_t
)bp
->blk_cksum
.zc_word
[3]);
2395 print_indirect(spa_t
*spa
, blkptr_t
*bp
, const zbookmark_phys_t
*zb
,
2396 const dnode_phys_t
*dnp
)
2398 char blkbuf
[BP_SPRINTF_LEN
];
2401 if (!BP_IS_EMBEDDED(bp
)) {
2402 ASSERT3U(BP_GET_TYPE(bp
), ==, dnp
->dn_type
);
2403 ASSERT3U(BP_GET_LEVEL(bp
), ==, zb
->zb_level
);
2406 (void) printf("%16llx ", (u_longlong_t
)blkid2offset(dnp
, bp
, zb
));
2408 ASSERT(zb
->zb_level
>= 0);
2410 for (l
= dnp
->dn_nlevels
- 1; l
>= -1; l
--) {
2411 if (l
== zb
->zb_level
) {
2412 (void) printf("L%llx", (u_longlong_t
)zb
->zb_level
);
2418 snprintf_blkptr_compact(blkbuf
, sizeof (blkbuf
), bp
, B_FALSE
);
2419 if (dump_opt
['Z'] && BP_GET_COMPRESS(bp
) == ZIO_COMPRESS_ZSTD
)
2420 snprintf_zstd_header(spa
, blkbuf
, sizeof (blkbuf
), bp
);
2421 (void) printf("%s\n", blkbuf
);
2425 visit_indirect(spa_t
*spa
, const dnode_phys_t
*dnp
,
2426 blkptr_t
*bp
, const zbookmark_phys_t
*zb
)
2430 if (bp
->blk_birth
== 0)
2433 print_indirect(spa
, bp
, zb
, dnp
);
2435 if (BP_GET_LEVEL(bp
) > 0 && !BP_IS_HOLE(bp
)) {
2436 arc_flags_t flags
= ARC_FLAG_WAIT
;
2439 int epb
= BP_GET_LSIZE(bp
) >> SPA_BLKPTRSHIFT
;
2442 ASSERT(!BP_IS_REDACTED(bp
));
2444 err
= arc_read(NULL
, spa
, bp
, arc_getbuf_func
, &buf
,
2445 ZIO_PRIORITY_ASYNC_READ
, ZIO_FLAG_CANFAIL
, &flags
, zb
);
2448 ASSERT(buf
->b_data
);
2450 /* recursively visit blocks below this */
2452 for (i
= 0; i
< epb
; i
++, cbp
++) {
2453 zbookmark_phys_t czb
;
2455 SET_BOOKMARK(&czb
, zb
->zb_objset
, zb
->zb_object
,
2457 zb
->zb_blkid
* epb
+ i
);
2458 err
= visit_indirect(spa
, dnp
, cbp
, &czb
);
2461 fill
+= BP_GET_FILL(cbp
);
2464 ASSERT3U(fill
, ==, BP_GET_FILL(bp
));
2465 arc_buf_destroy(buf
, &buf
);
2472 dump_indirect(dnode_t
*dn
)
2474 dnode_phys_t
*dnp
= dn
->dn_phys
;
2475 zbookmark_phys_t czb
;
2477 (void) printf("Indirect blocks:\n");
2479 SET_BOOKMARK(&czb
, dmu_objset_id(dn
->dn_objset
),
2480 dn
->dn_object
, dnp
->dn_nlevels
- 1, 0);
2481 for (int j
= 0; j
< dnp
->dn_nblkptr
; j
++) {
2483 (void) visit_indirect(dmu_objset_spa(dn
->dn_objset
), dnp
,
2484 &dnp
->dn_blkptr
[j
], &czb
);
2487 (void) printf("\n");
2491 dump_dsl_dir(objset_t
*os
, uint64_t object
, void *data
, size_t size
)
2493 (void) os
, (void) object
;
2494 dsl_dir_phys_t
*dd
= data
;
2498 /* make sure nicenum has enough space */
2499 _Static_assert(sizeof (nice
) >= NN_NUMBUF_SZ
, "nice truncated");
2504 ASSERT3U(size
, >=, sizeof (dsl_dir_phys_t
));
2506 crtime
= dd
->dd_creation_time
;
2507 (void) printf("\t\tcreation_time = %s", ctime(&crtime
));
2508 (void) printf("\t\thead_dataset_obj = %llu\n",
2509 (u_longlong_t
)dd
->dd_head_dataset_obj
);
2510 (void) printf("\t\tparent_dir_obj = %llu\n",
2511 (u_longlong_t
)dd
->dd_parent_obj
);
2512 (void) printf("\t\torigin_obj = %llu\n",
2513 (u_longlong_t
)dd
->dd_origin_obj
);
2514 (void) printf("\t\tchild_dir_zapobj = %llu\n",
2515 (u_longlong_t
)dd
->dd_child_dir_zapobj
);
2516 zdb_nicenum(dd
->dd_used_bytes
, nice
, sizeof (nice
));
2517 (void) printf("\t\tused_bytes = %s\n", nice
);
2518 zdb_nicenum(dd
->dd_compressed_bytes
, nice
, sizeof (nice
));
2519 (void) printf("\t\tcompressed_bytes = %s\n", nice
);
2520 zdb_nicenum(dd
->dd_uncompressed_bytes
, nice
, sizeof (nice
));
2521 (void) printf("\t\tuncompressed_bytes = %s\n", nice
);
2522 zdb_nicenum(dd
->dd_quota
, nice
, sizeof (nice
));
2523 (void) printf("\t\tquota = %s\n", nice
);
2524 zdb_nicenum(dd
->dd_reserved
, nice
, sizeof (nice
));
2525 (void) printf("\t\treserved = %s\n", nice
);
2526 (void) printf("\t\tprops_zapobj = %llu\n",
2527 (u_longlong_t
)dd
->dd_props_zapobj
);
2528 (void) printf("\t\tdeleg_zapobj = %llu\n",
2529 (u_longlong_t
)dd
->dd_deleg_zapobj
);
2530 (void) printf("\t\tflags = %llx\n",
2531 (u_longlong_t
)dd
->dd_flags
);
2534 zdb_nicenum(dd->dd_used_breakdown[DD_USED_ ## which], nice, \
2536 (void) printf("\t\tused_breakdown[" #which "] = %s\n", nice)
2543 (void) printf("\t\tclones = %llu\n",
2544 (u_longlong_t
)dd
->dd_clones
);
2548 dump_dsl_dataset(objset_t
*os
, uint64_t object
, void *data
, size_t size
)
2550 (void) os
, (void) object
;
2551 dsl_dataset_phys_t
*ds
= data
;
2553 char used
[32], compressed
[32], uncompressed
[32], unique
[32];
2554 char blkbuf
[BP_SPRINTF_LEN
];
2556 /* make sure nicenum has enough space */
2557 _Static_assert(sizeof (used
) >= NN_NUMBUF_SZ
, "used truncated");
2558 _Static_assert(sizeof (compressed
) >= NN_NUMBUF_SZ
,
2559 "compressed truncated");
2560 _Static_assert(sizeof (uncompressed
) >= NN_NUMBUF_SZ
,
2561 "uncompressed truncated");
2562 _Static_assert(sizeof (unique
) >= NN_NUMBUF_SZ
, "unique truncated");
2567 ASSERT(size
== sizeof (*ds
));
2568 crtime
= ds
->ds_creation_time
;
2569 zdb_nicenum(ds
->ds_referenced_bytes
, used
, sizeof (used
));
2570 zdb_nicenum(ds
->ds_compressed_bytes
, compressed
, sizeof (compressed
));
2571 zdb_nicenum(ds
->ds_uncompressed_bytes
, uncompressed
,
2572 sizeof (uncompressed
));
2573 zdb_nicenum(ds
->ds_unique_bytes
, unique
, sizeof (unique
));
2574 snprintf_blkptr(blkbuf
, sizeof (blkbuf
), &ds
->ds_bp
);
2576 (void) printf("\t\tdir_obj = %llu\n",
2577 (u_longlong_t
)ds
->ds_dir_obj
);
2578 (void) printf("\t\tprev_snap_obj = %llu\n",
2579 (u_longlong_t
)ds
->ds_prev_snap_obj
);
2580 (void) printf("\t\tprev_snap_txg = %llu\n",
2581 (u_longlong_t
)ds
->ds_prev_snap_txg
);
2582 (void) printf("\t\tnext_snap_obj = %llu\n",
2583 (u_longlong_t
)ds
->ds_next_snap_obj
);
2584 (void) printf("\t\tsnapnames_zapobj = %llu\n",
2585 (u_longlong_t
)ds
->ds_snapnames_zapobj
);
2586 (void) printf("\t\tnum_children = %llu\n",
2587 (u_longlong_t
)ds
->ds_num_children
);
2588 (void) printf("\t\tuserrefs_obj = %llu\n",
2589 (u_longlong_t
)ds
->ds_userrefs_obj
);
2590 (void) printf("\t\tcreation_time = %s", ctime(&crtime
));
2591 (void) printf("\t\tcreation_txg = %llu\n",
2592 (u_longlong_t
)ds
->ds_creation_txg
);
2593 (void) printf("\t\tdeadlist_obj = %llu\n",
2594 (u_longlong_t
)ds
->ds_deadlist_obj
);
2595 (void) printf("\t\tused_bytes = %s\n", used
);
2596 (void) printf("\t\tcompressed_bytes = %s\n", compressed
);
2597 (void) printf("\t\tuncompressed_bytes = %s\n", uncompressed
);
2598 (void) printf("\t\tunique = %s\n", unique
);
2599 (void) printf("\t\tfsid_guid = %llu\n",
2600 (u_longlong_t
)ds
->ds_fsid_guid
);
2601 (void) printf("\t\tguid = %llu\n",
2602 (u_longlong_t
)ds
->ds_guid
);
2603 (void) printf("\t\tflags = %llx\n",
2604 (u_longlong_t
)ds
->ds_flags
);
2605 (void) printf("\t\tnext_clones_obj = %llu\n",
2606 (u_longlong_t
)ds
->ds_next_clones_obj
);
2607 (void) printf("\t\tprops_obj = %llu\n",
2608 (u_longlong_t
)ds
->ds_props_obj
);
2609 (void) printf("\t\tbp = %s\n", blkbuf
);
2613 dump_bptree_cb(void *arg
, const blkptr_t
*bp
, dmu_tx_t
*tx
)
2615 (void) arg
, (void) tx
;
2616 char blkbuf
[BP_SPRINTF_LEN
];
2618 if (bp
->blk_birth
!= 0) {
2619 snprintf_blkptr(blkbuf
, sizeof (blkbuf
), bp
);
2620 (void) printf("\t%s\n", blkbuf
);
2626 dump_bptree(objset_t
*os
, uint64_t obj
, const char *name
)
2632 /* make sure nicenum has enough space */
2633 _Static_assert(sizeof (bytes
) >= NN_NUMBUF_SZ
, "bytes truncated");
2635 if (dump_opt
['d'] < 3)
2638 VERIFY3U(0, ==, dmu_bonus_hold(os
, obj
, FTAG
, &db
));
2640 zdb_nicenum(bt
->bt_bytes
, bytes
, sizeof (bytes
));
2641 (void) printf("\n %s: %llu datasets, %s\n",
2642 name
, (unsigned long long)(bt
->bt_end
- bt
->bt_begin
), bytes
);
2643 dmu_buf_rele(db
, FTAG
);
2645 if (dump_opt
['d'] < 5)
2648 (void) printf("\n");
2650 (void) bptree_iterate(os
, obj
, B_FALSE
, dump_bptree_cb
, NULL
, NULL
);
2654 dump_bpobj_cb(void *arg
, const blkptr_t
*bp
, boolean_t bp_freed
, dmu_tx_t
*tx
)
2656 (void) arg
, (void) tx
;
2657 char blkbuf
[BP_SPRINTF_LEN
];
2659 ASSERT(bp
->blk_birth
!= 0);
2660 snprintf_blkptr_compact(blkbuf
, sizeof (blkbuf
), bp
, bp_freed
);
2661 (void) printf("\t%s\n", blkbuf
);
2666 dump_full_bpobj(bpobj_t
*bpo
, const char *name
, int indent
)
2673 /* make sure nicenum has enough space */
2674 _Static_assert(sizeof (bytes
) >= NN_NUMBUF_SZ
, "bytes truncated");
2675 _Static_assert(sizeof (comp
) >= NN_NUMBUF_SZ
, "comp truncated");
2676 _Static_assert(sizeof (uncomp
) >= NN_NUMBUF_SZ
, "uncomp truncated");
2678 if (dump_opt
['d'] < 3)
2681 zdb_nicenum(bpo
->bpo_phys
->bpo_bytes
, bytes
, sizeof (bytes
));
2682 if (bpo
->bpo_havesubobj
&& bpo
->bpo_phys
->bpo_subobjs
!= 0) {
2683 zdb_nicenum(bpo
->bpo_phys
->bpo_comp
, comp
, sizeof (comp
));
2684 zdb_nicenum(bpo
->bpo_phys
->bpo_uncomp
, uncomp
, sizeof (uncomp
));
2685 if (bpo
->bpo_havefreed
) {
2686 (void) printf(" %*s: object %llu, %llu local "
2687 "blkptrs, %llu freed, %llu subobjs in object %llu, "
2688 "%s (%s/%s comp)\n",
2690 (u_longlong_t
)bpo
->bpo_object
,
2691 (u_longlong_t
)bpo
->bpo_phys
->bpo_num_blkptrs
,
2692 (u_longlong_t
)bpo
->bpo_phys
->bpo_num_freed
,
2693 (u_longlong_t
)bpo
->bpo_phys
->bpo_num_subobjs
,
2694 (u_longlong_t
)bpo
->bpo_phys
->bpo_subobjs
,
2695 bytes
, comp
, uncomp
);
2697 (void) printf(" %*s: object %llu, %llu local "
2698 "blkptrs, %llu subobjs in object %llu, "
2699 "%s (%s/%s comp)\n",
2701 (u_longlong_t
)bpo
->bpo_object
,
2702 (u_longlong_t
)bpo
->bpo_phys
->bpo_num_blkptrs
,
2703 (u_longlong_t
)bpo
->bpo_phys
->bpo_num_subobjs
,
2704 (u_longlong_t
)bpo
->bpo_phys
->bpo_subobjs
,
2705 bytes
, comp
, uncomp
);
2708 for (i
= 0; i
< bpo
->bpo_phys
->bpo_num_subobjs
; i
++) {
2712 VERIFY0(dmu_read(bpo
->bpo_os
,
2713 bpo
->bpo_phys
->bpo_subobjs
,
2714 i
* sizeof (subobj
), sizeof (subobj
), &subobj
, 0));
2715 error
= bpobj_open(&subbpo
, bpo
->bpo_os
, subobj
);
2717 (void) printf("ERROR %u while trying to open "
2719 error
, (u_longlong_t
)subobj
);
2722 dump_full_bpobj(&subbpo
, "subobj", indent
+ 1);
2723 bpobj_close(&subbpo
);
2726 if (bpo
->bpo_havefreed
) {
2727 (void) printf(" %*s: object %llu, %llu blkptrs, "
2730 (u_longlong_t
)bpo
->bpo_object
,
2731 (u_longlong_t
)bpo
->bpo_phys
->bpo_num_blkptrs
,
2732 (u_longlong_t
)bpo
->bpo_phys
->bpo_num_freed
,
2735 (void) printf(" %*s: object %llu, %llu blkptrs, "
2738 (u_longlong_t
)bpo
->bpo_object
,
2739 (u_longlong_t
)bpo
->bpo_phys
->bpo_num_blkptrs
,
2744 if (dump_opt
['d'] < 5)
2749 (void) bpobj_iterate_nofree(bpo
, dump_bpobj_cb
, NULL
, NULL
);
2750 (void) printf("\n");
2755 dump_bookmark(dsl_pool_t
*dp
, char *name
, boolean_t print_redact
,
2756 boolean_t print_list
)
2759 zfs_bookmark_phys_t prop
;
2760 objset_t
*mos
= dp
->dp_spa
->spa_meta_objset
;
2761 err
= dsl_bookmark_lookup(dp
, name
, NULL
, &prop
);
2767 (void) printf("\t#%s: ", strchr(name
, '#') + 1);
2768 (void) printf("{guid: %llx creation_txg: %llu creation_time: "
2769 "%llu redaction_obj: %llu}\n", (u_longlong_t
)prop
.zbm_guid
,
2770 (u_longlong_t
)prop
.zbm_creation_txg
,
2771 (u_longlong_t
)prop
.zbm_creation_time
,
2772 (u_longlong_t
)prop
.zbm_redaction_obj
);
2774 IMPLY(print_list
, print_redact
);
2775 if (!print_redact
|| prop
.zbm_redaction_obj
== 0)
2778 redaction_list_t
*rl
;
2779 VERIFY0(dsl_redaction_list_hold_obj(dp
,
2780 prop
.zbm_redaction_obj
, FTAG
, &rl
));
2782 redaction_list_phys_t
*rlp
= rl
->rl_phys
;
2783 (void) printf("\tRedacted:\n\t\tProgress: ");
2784 if (rlp
->rlp_last_object
!= UINT64_MAX
||
2785 rlp
->rlp_last_blkid
!= UINT64_MAX
) {
2786 (void) printf("%llu %llu (incomplete)\n",
2787 (u_longlong_t
)rlp
->rlp_last_object
,
2788 (u_longlong_t
)rlp
->rlp_last_blkid
);
2790 (void) printf("complete\n");
2792 (void) printf("\t\tSnapshots: [");
2793 for (unsigned int i
= 0; i
< rlp
->rlp_num_snaps
; i
++) {
2795 (void) printf(", ");
2796 (void) printf("%0llu",
2797 (u_longlong_t
)rlp
->rlp_snaps
[i
]);
2799 (void) printf("]\n\t\tLength: %llu\n",
2800 (u_longlong_t
)rlp
->rlp_num_entries
);
2803 dsl_redaction_list_rele(rl
, FTAG
);
2807 if (rlp
->rlp_num_entries
== 0) {
2808 dsl_redaction_list_rele(rl
, FTAG
);
2809 (void) printf("\t\tRedaction List: []\n\n");
2813 redact_block_phys_t
*rbp_buf
;
2815 dmu_object_info_t doi
;
2817 VERIFY0(dmu_object_info(mos
, prop
.zbm_redaction_obj
, &doi
));
2818 size
= doi
.doi_max_offset
;
2819 rbp_buf
= kmem_alloc(size
, KM_SLEEP
);
2821 err
= dmu_read(mos
, prop
.zbm_redaction_obj
, 0, size
,
2824 dsl_redaction_list_rele(rl
, FTAG
);
2825 kmem_free(rbp_buf
, size
);
2829 (void) printf("\t\tRedaction List: [{object: %llx, offset: "
2830 "%llx, blksz: %x, count: %llx}",
2831 (u_longlong_t
)rbp_buf
[0].rbp_object
,
2832 (u_longlong_t
)rbp_buf
[0].rbp_blkid
,
2833 (uint_t
)(redact_block_get_size(&rbp_buf
[0])),
2834 (u_longlong_t
)redact_block_get_count(&rbp_buf
[0]));
2836 for (size_t i
= 1; i
< rlp
->rlp_num_entries
; i
++) {
2837 (void) printf(",\n\t\t{object: %llx, offset: %llx, "
2838 "blksz: %x, count: %llx}",
2839 (u_longlong_t
)rbp_buf
[i
].rbp_object
,
2840 (u_longlong_t
)rbp_buf
[i
].rbp_blkid
,
2841 (uint_t
)(redact_block_get_size(&rbp_buf
[i
])),
2842 (u_longlong_t
)redact_block_get_count(&rbp_buf
[i
]));
2844 dsl_redaction_list_rele(rl
, FTAG
);
2845 kmem_free(rbp_buf
, size
);
2846 (void) printf("]\n\n");
2851 dump_bookmarks(objset_t
*os
, int verbosity
)
2854 zap_attribute_t attr
;
2855 dsl_dataset_t
*ds
= dmu_objset_ds(os
);
2856 dsl_pool_t
*dp
= spa_get_dsl(os
->os_spa
);
2857 objset_t
*mos
= os
->os_spa
->spa_meta_objset
;
2860 dsl_pool_config_enter(dp
, FTAG
);
2862 for (zap_cursor_init(&zc
, mos
, ds
->ds_bookmarks_obj
);
2863 zap_cursor_retrieve(&zc
, &attr
) == 0;
2864 zap_cursor_advance(&zc
)) {
2865 char osname
[ZFS_MAX_DATASET_NAME_LEN
];
2866 char buf
[ZFS_MAX_DATASET_NAME_LEN
];
2868 dmu_objset_name(os
, osname
);
2869 len
= snprintf(buf
, sizeof (buf
), "%s#%s", osname
,
2871 VERIFY3S(len
, <, ZFS_MAX_DATASET_NAME_LEN
);
2872 (void) dump_bookmark(dp
, buf
, verbosity
>= 5, verbosity
>= 6);
2874 zap_cursor_fini(&zc
);
2875 dsl_pool_config_exit(dp
, FTAG
);
2879 bpobj_count_refd(bpobj_t
*bpo
)
2881 mos_obj_refd(bpo
->bpo_object
);
2883 if (bpo
->bpo_havesubobj
&& bpo
->bpo_phys
->bpo_subobjs
!= 0) {
2884 mos_obj_refd(bpo
->bpo_phys
->bpo_subobjs
);
2885 for (uint64_t i
= 0; i
< bpo
->bpo_phys
->bpo_num_subobjs
; i
++) {
2889 VERIFY0(dmu_read(bpo
->bpo_os
,
2890 bpo
->bpo_phys
->bpo_subobjs
,
2891 i
* sizeof (subobj
), sizeof (subobj
), &subobj
, 0));
2892 error
= bpobj_open(&subbpo
, bpo
->bpo_os
, subobj
);
2894 (void) printf("ERROR %u while trying to open "
2896 error
, (u_longlong_t
)subobj
);
2899 bpobj_count_refd(&subbpo
);
2900 bpobj_close(&subbpo
);
2906 dsl_deadlist_entry_count_refd(void *arg
, dsl_deadlist_entry_t
*dle
)
2909 uint64_t empty_bpobj
= spa
->spa_dsl_pool
->dp_empty_bpobj
;
2910 if (dle
->dle_bpobj
.bpo_object
!= empty_bpobj
)
2911 bpobj_count_refd(&dle
->dle_bpobj
);
2916 dsl_deadlist_entry_dump(void *arg
, dsl_deadlist_entry_t
*dle
)
2918 ASSERT(arg
== NULL
);
2919 if (dump_opt
['d'] >= 5) {
2921 (void) snprintf(buf
, sizeof (buf
),
2922 "mintxg %llu -> obj %llu",
2923 (longlong_t
)dle
->dle_mintxg
,
2924 (longlong_t
)dle
->dle_bpobj
.bpo_object
);
2926 dump_full_bpobj(&dle
->dle_bpobj
, buf
, 0);
2928 (void) printf("mintxg %llu -> obj %llu\n",
2929 (longlong_t
)dle
->dle_mintxg
,
2930 (longlong_t
)dle
->dle_bpobj
.bpo_object
);
2936 dump_blkptr_list(dsl_deadlist_t
*dl
, const char *name
)
2942 spa_t
*spa
= dmu_objset_spa(dl
->dl_os
);
2943 uint64_t empty_bpobj
= spa
->spa_dsl_pool
->dp_empty_bpobj
;
2945 if (dl
->dl_oldfmt
) {
2946 if (dl
->dl_bpobj
.bpo_object
!= empty_bpobj
)
2947 bpobj_count_refd(&dl
->dl_bpobj
);
2949 mos_obj_refd(dl
->dl_object
);
2950 dsl_deadlist_iterate(dl
, dsl_deadlist_entry_count_refd
, spa
);
2953 /* make sure nicenum has enough space */
2954 _Static_assert(sizeof (bytes
) >= NN_NUMBUF_SZ
, "bytes truncated");
2955 _Static_assert(sizeof (comp
) >= NN_NUMBUF_SZ
, "comp truncated");
2956 _Static_assert(sizeof (uncomp
) >= NN_NUMBUF_SZ
, "uncomp truncated");
2957 _Static_assert(sizeof (entries
) >= NN_NUMBUF_SZ
, "entries truncated");
2959 if (dump_opt
['d'] < 3)
2962 if (dl
->dl_oldfmt
) {
2963 dump_full_bpobj(&dl
->dl_bpobj
, "old-format deadlist", 0);
2967 zdb_nicenum(dl
->dl_phys
->dl_used
, bytes
, sizeof (bytes
));
2968 zdb_nicenum(dl
->dl_phys
->dl_comp
, comp
, sizeof (comp
));
2969 zdb_nicenum(dl
->dl_phys
->dl_uncomp
, uncomp
, sizeof (uncomp
));
2970 zdb_nicenum(avl_numnodes(&dl
->dl_tree
), entries
, sizeof (entries
));
2971 (void) printf("\n %s: %s (%s/%s comp), %s entries\n",
2972 name
, bytes
, comp
, uncomp
, entries
);
2974 if (dump_opt
['d'] < 4)
2977 (void) putchar('\n');
2979 dsl_deadlist_iterate(dl
, dsl_deadlist_entry_dump
, NULL
);
2983 verify_dd_livelist(objset_t
*os
)
2985 uint64_t ll_used
, used
, ll_comp
, comp
, ll_uncomp
, uncomp
;
2986 dsl_pool_t
*dp
= spa_get_dsl(os
->os_spa
);
2987 dsl_dir_t
*dd
= os
->os_dsl_dataset
->ds_dir
;
2989 ASSERT(!dmu_objset_is_snapshot(os
));
2990 if (!dsl_deadlist_is_open(&dd
->dd_livelist
))
2993 /* Iterate through the livelist to check for duplicates */
2994 dsl_deadlist_iterate(&dd
->dd_livelist
, sublivelist_verify_lightweight
,
2997 dsl_pool_config_enter(dp
, FTAG
);
2998 dsl_deadlist_space(&dd
->dd_livelist
, &ll_used
,
2999 &ll_comp
, &ll_uncomp
);
3001 dsl_dataset_t
*origin_ds
;
3002 ASSERT(dsl_pool_config_held(dp
));
3003 VERIFY0(dsl_dataset_hold_obj(dp
,
3004 dsl_dir_phys(dd
)->dd_origin_obj
, FTAG
, &origin_ds
));
3005 VERIFY0(dsl_dataset_space_written(origin_ds
, os
->os_dsl_dataset
,
3006 &used
, &comp
, &uncomp
));
3007 dsl_dataset_rele(origin_ds
, FTAG
);
3008 dsl_pool_config_exit(dp
, FTAG
);
3010 * It's possible that the dataset's uncomp space is larger than the
3011 * livelist's because livelists do not track embedded block pointers
3013 if (used
!= ll_used
|| comp
!= ll_comp
|| uncomp
< ll_uncomp
) {
3014 char nice_used
[32], nice_comp
[32], nice_uncomp
[32];
3015 (void) printf("Discrepancy in space accounting:\n");
3016 zdb_nicenum(used
, nice_used
, sizeof (nice_used
));
3017 zdb_nicenum(comp
, nice_comp
, sizeof (nice_comp
));
3018 zdb_nicenum(uncomp
, nice_uncomp
, sizeof (nice_uncomp
));
3019 (void) printf("dir: used %s, comp %s, uncomp %s\n",
3020 nice_used
, nice_comp
, nice_uncomp
);
3021 zdb_nicenum(ll_used
, nice_used
, sizeof (nice_used
));
3022 zdb_nicenum(ll_comp
, nice_comp
, sizeof (nice_comp
));
3023 zdb_nicenum(ll_uncomp
, nice_uncomp
, sizeof (nice_uncomp
));
3024 (void) printf("livelist: used %s, comp %s, uncomp %s\n",
3025 nice_used
, nice_comp
, nice_uncomp
);
3031 static char *key_material
= NULL
;
3034 zdb_derive_key(dsl_dir_t
*dd
, uint8_t *key_out
)
3036 uint64_t keyformat
, salt
, iters
;
3040 VERIFY0(zap_lookup(dd
->dd_pool
->dp_meta_objset
, dd
->dd_crypto_obj
,
3041 zfs_prop_to_name(ZFS_PROP_KEYFORMAT
), sizeof (uint64_t),
3044 switch (keyformat
) {
3045 case ZFS_KEYFORMAT_HEX
:
3046 for (i
= 0; i
< WRAPPING_KEY_LEN
* 2; i
+= 2) {
3047 if (!isxdigit(key_material
[i
]) ||
3048 !isxdigit(key_material
[i
+1]))
3050 if (sscanf(&key_material
[i
], "%02hhx", &c
) != 1)
3056 case ZFS_KEYFORMAT_PASSPHRASE
:
3057 VERIFY0(zap_lookup(dd
->dd_pool
->dp_meta_objset
,
3058 dd
->dd_crypto_obj
, zfs_prop_to_name(ZFS_PROP_PBKDF2_SALT
),
3059 sizeof (uint64_t), 1, &salt
));
3060 VERIFY0(zap_lookup(dd
->dd_pool
->dp_meta_objset
,
3061 dd
->dd_crypto_obj
, zfs_prop_to_name(ZFS_PROP_PBKDF2_ITERS
),
3062 sizeof (uint64_t), 1, &iters
));
3064 if (PKCS5_PBKDF2_HMAC_SHA1(key_material
, strlen(key_material
),
3065 ((uint8_t *)&salt
), sizeof (uint64_t), iters
,
3066 WRAPPING_KEY_LEN
, key_out
) != 1)
3072 fatal("no support for key format %u\n",
3073 (unsigned int) keyformat
);
3079 static char encroot
[ZFS_MAX_DATASET_NAME_LEN
];
3080 static boolean_t key_loaded
= B_FALSE
;
3083 zdb_load_key(objset_t
*os
)
3086 dsl_dir_t
*dd
, *rdd
;
3087 uint8_t key
[WRAPPING_KEY_LEN
];
3091 dp
= spa_get_dsl(os
->os_spa
);
3092 dd
= os
->os_dsl_dataset
->ds_dir
;
3094 dsl_pool_config_enter(dp
, FTAG
);
3095 VERIFY0(zap_lookup(dd
->dd_pool
->dp_meta_objset
, dd
->dd_crypto_obj
,
3096 DSL_CRYPTO_KEY_ROOT_DDOBJ
, sizeof (uint64_t), 1, &rddobj
));
3097 VERIFY0(dsl_dir_hold_obj(dd
->dd_pool
, rddobj
, NULL
, FTAG
, &rdd
));
3098 dsl_dir_name(rdd
, encroot
);
3099 dsl_dir_rele(rdd
, FTAG
);
3101 if (!zdb_derive_key(dd
, key
))
3102 fatal("couldn't derive encryption key");
3104 dsl_pool_config_exit(dp
, FTAG
);
3106 ASSERT3U(dsl_dataset_get_keystatus(dd
), ==, ZFS_KEYSTATUS_UNAVAILABLE
);
3108 dsl_crypto_params_t
*dcp
;
3109 nvlist_t
*crypto_args
;
3111 crypto_args
= fnvlist_alloc();
3112 fnvlist_add_uint8_array(crypto_args
, "wkeydata",
3113 (uint8_t *)key
, WRAPPING_KEY_LEN
);
3114 VERIFY0(dsl_crypto_params_create_nvlist(DCP_CMD_NONE
,
3115 NULL
, crypto_args
, &dcp
));
3116 err
= spa_keystore_load_wkey(encroot
, dcp
, B_FALSE
);
3118 dsl_crypto_params_free(dcp
, (err
!= 0));
3119 fnvlist_free(crypto_args
);
3123 "couldn't load encryption key for %s: %s",
3124 encroot
, err
== ZFS_ERR_CRYPTO_NOTSUP
?
3125 "crypto params not supported" : strerror(err
));
3127 ASSERT3U(dsl_dataset_get_keystatus(dd
), ==, ZFS_KEYSTATUS_AVAILABLE
);
3129 printf("Unlocked encryption root: %s\n", encroot
);
3130 key_loaded
= B_TRUE
;
3134 zdb_unload_key(void)
3139 VERIFY0(spa_keystore_unload_wkey(encroot
));
3140 key_loaded
= B_FALSE
;
3143 static avl_tree_t idx_tree
;
3144 static avl_tree_t domain_tree
;
3145 static boolean_t fuid_table_loaded
;
3146 static objset_t
*sa_os
= NULL
;
3147 static sa_attr_type_t
*sa_attr_table
= NULL
;
3150 open_objset(const char *path
, const void *tag
, objset_t
**osp
)
3153 uint64_t sa_attrs
= 0;
3154 uint64_t version
= 0;
3156 VERIFY3P(sa_os
, ==, NULL
);
3159 * We can't own an objset if it's redacted. Therefore, we do this
3160 * dance: hold the objset, then acquire a long hold on its dataset, then
3161 * release the pool (which is held as part of holding the objset).
3164 if (dump_opt
['K']) {
3165 /* decryption requested, try to load keys */
3166 err
= dmu_objset_hold(path
, tag
, osp
);
3168 (void) fprintf(stderr
, "failed to hold dataset "
3170 path
, strerror(err
));
3173 dsl_dataset_long_hold(dmu_objset_ds(*osp
), tag
);
3174 dsl_pool_rele(dmu_objset_pool(*osp
), tag
);
3176 /* succeeds or dies */
3179 /* release it all */
3180 dsl_dataset_long_rele(dmu_objset_ds(*osp
), tag
);
3181 dsl_dataset_rele(dmu_objset_ds(*osp
), tag
);
3184 int ds_hold_flags
= key_loaded
? DS_HOLD_FLAG_DECRYPT
: 0;
3186 err
= dmu_objset_hold_flags(path
, ds_hold_flags
, tag
, osp
);
3188 (void) fprintf(stderr
, "failed to hold dataset '%s': %s\n",
3189 path
, strerror(err
));
3192 dsl_dataset_long_hold(dmu_objset_ds(*osp
), tag
);
3193 dsl_pool_rele(dmu_objset_pool(*osp
), tag
);
3195 if (dmu_objset_type(*osp
) == DMU_OST_ZFS
&&
3196 (key_loaded
|| !(*osp
)->os_encrypted
)) {
3197 (void) zap_lookup(*osp
, MASTER_NODE_OBJ
, ZPL_VERSION_STR
,
3199 if (version
>= ZPL_VERSION_SA
) {
3200 (void) zap_lookup(*osp
, MASTER_NODE_OBJ
, ZFS_SA_ATTRS
,
3203 err
= sa_setup(*osp
, sa_attrs
, zfs_attr_table
, ZPL_END
,
3206 (void) fprintf(stderr
, "sa_setup failed: %s\n",
3208 dsl_dataset_long_rele(dmu_objset_ds(*osp
), tag
);
3209 dsl_dataset_rele_flags(dmu_objset_ds(*osp
),
3210 ds_hold_flags
, tag
);
3220 close_objset(objset_t
*os
, const void *tag
)
3222 VERIFY3P(os
, ==, sa_os
);
3223 if (os
->os_sa
!= NULL
)
3225 dsl_dataset_long_rele(dmu_objset_ds(os
), tag
);
3226 dsl_dataset_rele_flags(dmu_objset_ds(os
),
3227 key_loaded
? DS_HOLD_FLAG_DECRYPT
: 0, tag
);
3228 sa_attr_table
= NULL
;
3235 fuid_table_destroy(void)
3237 if (fuid_table_loaded
) {
3238 zfs_fuid_table_destroy(&idx_tree
, &domain_tree
);
3239 fuid_table_loaded
= B_FALSE
;
3244 * print uid or gid information.
3245 * For normal POSIX id just the id is printed in decimal format.
3246 * For CIFS files with FUID the fuid is printed in hex followed by
3247 * the domain-rid string.
3250 print_idstr(uint64_t id
, const char *id_type
)
3252 if (FUID_INDEX(id
)) {
3253 const char *domain
=
3254 zfs_fuid_idx_domain(&idx_tree
, FUID_INDEX(id
));
3255 (void) printf("\t%s %llx [%s-%d]\n", id_type
,
3256 (u_longlong_t
)id
, domain
, (int)FUID_RID(id
));
3258 (void) printf("\t%s %llu\n", id_type
, (u_longlong_t
)id
);
3264 dump_uidgid(objset_t
*os
, uint64_t uid
, uint64_t gid
)
3266 uint32_t uid_idx
, gid_idx
;
3268 uid_idx
= FUID_INDEX(uid
);
3269 gid_idx
= FUID_INDEX(gid
);
3271 /* Load domain table, if not already loaded */
3272 if (!fuid_table_loaded
&& (uid_idx
|| gid_idx
)) {
3275 /* first find the fuid object. It lives in the master node */
3276 VERIFY(zap_lookup(os
, MASTER_NODE_OBJ
, ZFS_FUID_TABLES
,
3277 8, 1, &fuid_obj
) == 0);
3278 zfs_fuid_avl_tree_create(&idx_tree
, &domain_tree
);
3279 (void) zfs_fuid_table_load(os
, fuid_obj
,
3280 &idx_tree
, &domain_tree
);
3281 fuid_table_loaded
= B_TRUE
;
3284 print_idstr(uid
, "uid");
3285 print_idstr(gid
, "gid");
3289 dump_znode_sa_xattr(sa_handle_t
*hdl
)
3292 nvpair_t
*elem
= NULL
;
3293 int sa_xattr_size
= 0;
3294 int sa_xattr_entries
= 0;
3296 char *sa_xattr_packed
;
3298 error
= sa_size(hdl
, sa_attr_table
[ZPL_DXATTR
], &sa_xattr_size
);
3299 if (error
|| sa_xattr_size
== 0)
3302 sa_xattr_packed
= malloc(sa_xattr_size
);
3303 if (sa_xattr_packed
== NULL
)
3306 error
= sa_lookup(hdl
, sa_attr_table
[ZPL_DXATTR
],
3307 sa_xattr_packed
, sa_xattr_size
);
3309 free(sa_xattr_packed
);
3313 error
= nvlist_unpack(sa_xattr_packed
, sa_xattr_size
, &sa_xattr
, 0);
3315 free(sa_xattr_packed
);
3319 while ((elem
= nvlist_next_nvpair(sa_xattr
, elem
)) != NULL
)
3322 (void) printf("\tSA xattrs: %d bytes, %d entries\n\n",
3323 sa_xattr_size
, sa_xattr_entries
);
3324 while ((elem
= nvlist_next_nvpair(sa_xattr
, elem
)) != NULL
) {
3328 (void) printf("\t\t%s = ", nvpair_name(elem
));
3329 nvpair_value_byte_array(elem
, &value
, &cnt
);
3330 for (idx
= 0; idx
< cnt
; ++idx
) {
3331 if (isprint(value
[idx
]))
3332 (void) putchar(value
[idx
]);
3334 (void) printf("\\%3.3o", value
[idx
]);
3336 (void) putchar('\n');
3339 nvlist_free(sa_xattr
);
3340 free(sa_xattr_packed
);
3344 dump_znode_symlink(sa_handle_t
*hdl
)
3346 int sa_symlink_size
= 0;
3347 char linktarget
[MAXPATHLEN
];
3350 error
= sa_size(hdl
, sa_attr_table
[ZPL_SYMLINK
], &sa_symlink_size
);
3351 if (error
|| sa_symlink_size
== 0) {
3354 if (sa_symlink_size
>= sizeof (linktarget
)) {
3355 (void) printf("symlink size %d is too large\n",
3359 linktarget
[sa_symlink_size
] = '\0';
3360 if (sa_lookup(hdl
, sa_attr_table
[ZPL_SYMLINK
],
3361 &linktarget
, sa_symlink_size
) == 0)
3362 (void) printf("\ttarget %s\n", linktarget
);
3366 dump_znode(objset_t
*os
, uint64_t object
, void *data
, size_t size
)
3368 (void) data
, (void) size
;
3369 char path
[MAXPATHLEN
* 2]; /* allow for xattr and failure prefix */
3371 uint64_t xattr
, rdev
, gen
;
3372 uint64_t uid
, gid
, mode
, fsize
, parent
, links
;
3374 uint64_t acctm
[2], modtm
[2], chgtm
[2], crtm
[2];
3375 time_t z_crtime
, z_atime
, z_mtime
, z_ctime
;
3376 sa_bulk_attr_t bulk
[12];
3380 VERIFY3P(os
, ==, sa_os
);
3381 if (sa_handle_get(os
, object
, NULL
, SA_HDL_PRIVATE
, &hdl
)) {
3382 (void) printf("Failed to get handle for SA znode\n");
3386 SA_ADD_BULK_ATTR(bulk
, idx
, sa_attr_table
[ZPL_UID
], NULL
, &uid
, 8);
3387 SA_ADD_BULK_ATTR(bulk
, idx
, sa_attr_table
[ZPL_GID
], NULL
, &gid
, 8);
3388 SA_ADD_BULK_ATTR(bulk
, idx
, sa_attr_table
[ZPL_LINKS
], NULL
,
3390 SA_ADD_BULK_ATTR(bulk
, idx
, sa_attr_table
[ZPL_GEN
], NULL
, &gen
, 8);
3391 SA_ADD_BULK_ATTR(bulk
, idx
, sa_attr_table
[ZPL_MODE
], NULL
,
3393 SA_ADD_BULK_ATTR(bulk
, idx
, sa_attr_table
[ZPL_PARENT
],
3395 SA_ADD_BULK_ATTR(bulk
, idx
, sa_attr_table
[ZPL_SIZE
], NULL
,
3397 SA_ADD_BULK_ATTR(bulk
, idx
, sa_attr_table
[ZPL_ATIME
], NULL
,
3399 SA_ADD_BULK_ATTR(bulk
, idx
, sa_attr_table
[ZPL_MTIME
], NULL
,
3401 SA_ADD_BULK_ATTR(bulk
, idx
, sa_attr_table
[ZPL_CRTIME
], NULL
,
3403 SA_ADD_BULK_ATTR(bulk
, idx
, sa_attr_table
[ZPL_CTIME
], NULL
,
3405 SA_ADD_BULK_ATTR(bulk
, idx
, sa_attr_table
[ZPL_FLAGS
], NULL
,
3408 if (sa_bulk_lookup(hdl
, bulk
, idx
)) {
3409 (void) sa_handle_destroy(hdl
);
3413 z_crtime
= (time_t)crtm
[0];
3414 z_atime
= (time_t)acctm
[0];
3415 z_mtime
= (time_t)modtm
[0];
3416 z_ctime
= (time_t)chgtm
[0];
3418 if (dump_opt
['d'] > 4) {
3419 error
= zfs_obj_to_path(os
, object
, path
, sizeof (path
));
3420 if (error
== ESTALE
) {
3421 (void) snprintf(path
, sizeof (path
), "on delete queue");
3422 } else if (error
!= 0) {
3424 (void) snprintf(path
, sizeof (path
),
3425 "path not found, possibly leaked");
3427 (void) printf("\tpath %s\n", path
);
3431 dump_znode_symlink(hdl
);
3432 dump_uidgid(os
, uid
, gid
);
3433 (void) printf("\tatime %s", ctime(&z_atime
));
3434 (void) printf("\tmtime %s", ctime(&z_mtime
));
3435 (void) printf("\tctime %s", ctime(&z_ctime
));
3436 (void) printf("\tcrtime %s", ctime(&z_crtime
));
3437 (void) printf("\tgen %llu\n", (u_longlong_t
)gen
);
3438 (void) printf("\tmode %llo\n", (u_longlong_t
)mode
);
3439 (void) printf("\tsize %llu\n", (u_longlong_t
)fsize
);
3440 (void) printf("\tparent %llu\n", (u_longlong_t
)parent
);
3441 (void) printf("\tlinks %llu\n", (u_longlong_t
)links
);
3442 (void) printf("\tpflags %llx\n", (u_longlong_t
)pflags
);
3443 if (dmu_objset_projectquota_enabled(os
) && (pflags
& ZFS_PROJID
)) {
3446 if (sa_lookup(hdl
, sa_attr_table
[ZPL_PROJID
], &projid
,
3447 sizeof (uint64_t)) == 0)
3448 (void) printf("\tprojid %llu\n", (u_longlong_t
)projid
);
3450 if (sa_lookup(hdl
, sa_attr_table
[ZPL_XATTR
], &xattr
,
3451 sizeof (uint64_t)) == 0)
3452 (void) printf("\txattr %llu\n", (u_longlong_t
)xattr
);
3453 if (sa_lookup(hdl
, sa_attr_table
[ZPL_RDEV
], &rdev
,
3454 sizeof (uint64_t)) == 0)
3455 (void) printf("\trdev 0x%016llx\n", (u_longlong_t
)rdev
);
3456 dump_znode_sa_xattr(hdl
);
3457 sa_handle_destroy(hdl
);
3461 dump_acl(objset_t
*os
, uint64_t object
, void *data
, size_t size
)
3463 (void) os
, (void) object
, (void) data
, (void) size
;
3467 dump_dmu_objset(objset_t
*os
, uint64_t object
, void *data
, size_t size
)
3469 (void) os
, (void) object
, (void) data
, (void) size
;
3472 static object_viewer_t
*object_viewer
[DMU_OT_NUMTYPES
+ 1] = {
3473 dump_none
, /* unallocated */
3474 dump_zap
, /* object directory */
3475 dump_uint64
, /* object array */
3476 dump_none
, /* packed nvlist */
3477 dump_packed_nvlist
, /* packed nvlist size */
3478 dump_none
, /* bpobj */
3479 dump_bpobj
, /* bpobj header */
3480 dump_none
, /* SPA space map header */
3481 dump_none
, /* SPA space map */
3482 dump_none
, /* ZIL intent log */
3483 dump_dnode
, /* DMU dnode */
3484 dump_dmu_objset
, /* DMU objset */
3485 dump_dsl_dir
, /* DSL directory */
3486 dump_zap
, /* DSL directory child map */
3487 dump_zap
, /* DSL dataset snap map */
3488 dump_zap
, /* DSL props */
3489 dump_dsl_dataset
, /* DSL dataset */
3490 dump_znode
, /* ZFS znode */
3491 dump_acl
, /* ZFS V0 ACL */
3492 dump_uint8
, /* ZFS plain file */
3493 dump_zpldir
, /* ZFS directory */
3494 dump_zap
, /* ZFS master node */
3495 dump_zap
, /* ZFS delete queue */
3496 dump_uint8
, /* zvol object */
3497 dump_zap
, /* zvol prop */
3498 dump_uint8
, /* other uint8[] */
3499 dump_uint64
, /* other uint64[] */
3500 dump_zap
, /* other ZAP */
3501 dump_zap
, /* persistent error log */
3502 dump_uint8
, /* SPA history */
3503 dump_history_offsets
, /* SPA history offsets */
3504 dump_zap
, /* Pool properties */
3505 dump_zap
, /* DSL permissions */
3506 dump_acl
, /* ZFS ACL */
3507 dump_uint8
, /* ZFS SYSACL */
3508 dump_none
, /* FUID nvlist */
3509 dump_packed_nvlist
, /* FUID nvlist size */
3510 dump_zap
, /* DSL dataset next clones */
3511 dump_zap
, /* DSL scrub queue */
3512 dump_zap
, /* ZFS user/group/project used */
3513 dump_zap
, /* ZFS user/group/project quota */
3514 dump_zap
, /* snapshot refcount tags */
3515 dump_ddt_zap
, /* DDT ZAP object */
3516 dump_zap
, /* DDT statistics */
3517 dump_znode
, /* SA object */
3518 dump_zap
, /* SA Master Node */
3519 dump_sa_attrs
, /* SA attribute registration */
3520 dump_sa_layouts
, /* SA attribute layouts */
3521 dump_zap
, /* DSL scrub translations */
3522 dump_none
, /* fake dedup BP */
3523 dump_zap
, /* deadlist */
3524 dump_none
, /* deadlist hdr */
3525 dump_zap
, /* dsl clones */
3526 dump_bpobj_subobjs
, /* bpobj subobjs */
3527 dump_unknown
, /* Unknown type, must be last */
3531 match_object_type(dmu_object_type_t obj_type
, uint64_t flags
)
3533 boolean_t match
= B_TRUE
;
3536 case DMU_OT_DIRECTORY_CONTENTS
:
3537 if (!(flags
& ZOR_FLAG_DIRECTORY
))
3540 case DMU_OT_PLAIN_FILE_CONTENTS
:
3541 if (!(flags
& ZOR_FLAG_PLAIN_FILE
))
3544 case DMU_OT_SPACE_MAP
:
3545 if (!(flags
& ZOR_FLAG_SPACE_MAP
))
3549 if (strcmp(zdb_ot_name(obj_type
), "zap") == 0) {
3550 if (!(flags
& ZOR_FLAG_ZAP
))
3556 * If all bits except some of the supported flags are
3557 * set, the user combined the all-types flag (A) with
3558 * a negated flag to exclude some types (e.g. A-f to
3559 * show all object types except plain files).
3561 if ((flags
| ZOR_SUPPORTED_FLAGS
) != ZOR_FLAG_ALL_TYPES
)
3571 dump_object(objset_t
*os
, uint64_t object
, int verbosity
,
3572 boolean_t
*print_header
, uint64_t *dnode_slots_used
, uint64_t flags
)
3574 dmu_buf_t
*db
= NULL
;
3575 dmu_object_info_t doi
;
3577 boolean_t dnode_held
= B_FALSE
;
3580 char iblk
[32], dblk
[32], lsize
[32], asize
[32], fill
[32], dnsize
[32];
3581 char bonus_size
[32];
3585 /* make sure nicenum has enough space */
3586 _Static_assert(sizeof (iblk
) >= NN_NUMBUF_SZ
, "iblk truncated");
3587 _Static_assert(sizeof (dblk
) >= NN_NUMBUF_SZ
, "dblk truncated");
3588 _Static_assert(sizeof (lsize
) >= NN_NUMBUF_SZ
, "lsize truncated");
3589 _Static_assert(sizeof (asize
) >= NN_NUMBUF_SZ
, "asize truncated");
3590 _Static_assert(sizeof (bonus_size
) >= NN_NUMBUF_SZ
,
3591 "bonus_size truncated");
3593 if (*print_header
) {
3594 (void) printf("\n%10s %3s %5s %5s %5s %6s %5s %6s %s\n",
3595 "Object", "lvl", "iblk", "dblk", "dsize", "dnsize",
3596 "lsize", "%full", "type");
3601 dn
= DMU_META_DNODE(os
);
3602 dmu_object_info_from_dnode(dn
, &doi
);
3605 * Encrypted datasets will have sensitive bonus buffers
3606 * encrypted. Therefore we cannot hold the bonus buffer and
3607 * must hold the dnode itself instead.
3609 error
= dmu_object_info(os
, object
, &doi
);
3611 fatal("dmu_object_info() failed, errno %u", error
);
3613 if (!key_loaded
&& os
->os_encrypted
&&
3614 DMU_OT_IS_ENCRYPTED(doi
.doi_bonus_type
)) {
3615 error
= dnode_hold(os
, object
, FTAG
, &dn
);
3617 fatal("dnode_hold() failed, errno %u", error
);
3618 dnode_held
= B_TRUE
;
3620 error
= dmu_bonus_hold(os
, object
, FTAG
, &db
);
3622 fatal("dmu_bonus_hold(%llu) failed, errno %u",
3624 bonus
= db
->db_data
;
3625 bsize
= db
->db_size
;
3626 dn
= DB_DNODE((dmu_buf_impl_t
*)db
);
3631 * Default to showing all object types if no flags were specified.
3633 if (flags
!= 0 && flags
!= ZOR_FLAG_ALL_TYPES
&&
3634 !match_object_type(doi
.doi_type
, flags
))
3637 if (dnode_slots_used
)
3638 *dnode_slots_used
= doi
.doi_dnodesize
/ DNODE_MIN_SIZE
;
3640 zdb_nicenum(doi
.doi_metadata_block_size
, iblk
, sizeof (iblk
));
3641 zdb_nicenum(doi
.doi_data_block_size
, dblk
, sizeof (dblk
));
3642 zdb_nicenum(doi
.doi_max_offset
, lsize
, sizeof (lsize
));
3643 zdb_nicenum(doi
.doi_physical_blocks_512
<< 9, asize
, sizeof (asize
));
3644 zdb_nicenum(doi
.doi_bonus_size
, bonus_size
, sizeof (bonus_size
));
3645 zdb_nicenum(doi
.doi_dnodesize
, dnsize
, sizeof (dnsize
));
3646 (void) snprintf(fill
, sizeof (fill
), "%6.2f", 100.0 *
3647 doi
.doi_fill_count
* doi
.doi_data_block_size
/ (object
== 0 ?
3648 DNODES_PER_BLOCK
: 1) / doi
.doi_max_offset
);
3652 if (doi
.doi_checksum
!= ZIO_CHECKSUM_INHERIT
|| verbosity
>= 6) {
3653 (void) snprintf(aux
+ strlen(aux
), sizeof (aux
) - strlen(aux
),
3654 " (K=%s)", ZDB_CHECKSUM_NAME(doi
.doi_checksum
));
3657 if (doi
.doi_compress
== ZIO_COMPRESS_INHERIT
&&
3658 ZIO_COMPRESS_HASLEVEL(os
->os_compress
) && verbosity
>= 6) {
3659 const char *compname
= NULL
;
3660 if (zfs_prop_index_to_string(ZFS_PROP_COMPRESSION
,
3661 ZIO_COMPRESS_RAW(os
->os_compress
, os
->os_complevel
),
3663 (void) snprintf(aux
+ strlen(aux
),
3664 sizeof (aux
) - strlen(aux
), " (Z=inherit=%s)",
3667 (void) snprintf(aux
+ strlen(aux
),
3668 sizeof (aux
) - strlen(aux
),
3669 " (Z=inherit=%s-unknown)",
3670 ZDB_COMPRESS_NAME(os
->os_compress
));
3672 } else if (doi
.doi_compress
== ZIO_COMPRESS_INHERIT
&& verbosity
>= 6) {
3673 (void) snprintf(aux
+ strlen(aux
), sizeof (aux
) - strlen(aux
),
3674 " (Z=inherit=%s)", ZDB_COMPRESS_NAME(os
->os_compress
));
3675 } else if (doi
.doi_compress
!= ZIO_COMPRESS_INHERIT
|| verbosity
>= 6) {
3676 (void) snprintf(aux
+ strlen(aux
), sizeof (aux
) - strlen(aux
),
3677 " (Z=%s)", ZDB_COMPRESS_NAME(doi
.doi_compress
));
3680 (void) printf("%10lld %3u %5s %5s %5s %6s %5s %6s %s%s\n",
3681 (u_longlong_t
)object
, doi
.doi_indirection
, iblk
, dblk
,
3682 asize
, dnsize
, lsize
, fill
, zdb_ot_name(doi
.doi_type
), aux
);
3684 if (doi
.doi_bonus_type
!= DMU_OT_NONE
&& verbosity
> 3) {
3685 (void) printf("%10s %3s %5s %5s %5s %5s %5s %6s %s\n",
3686 "", "", "", "", "", "", bonus_size
, "bonus",
3687 zdb_ot_name(doi
.doi_bonus_type
));
3690 if (verbosity
>= 4) {
3691 (void) printf("\tdnode flags: %s%s%s%s\n",
3692 (dn
->dn_phys
->dn_flags
& DNODE_FLAG_USED_BYTES
) ?
3694 (dn
->dn_phys
->dn_flags
& DNODE_FLAG_USERUSED_ACCOUNTED
) ?
3695 "USERUSED_ACCOUNTED " : "",
3696 (dn
->dn_phys
->dn_flags
& DNODE_FLAG_USEROBJUSED_ACCOUNTED
) ?
3697 "USEROBJUSED_ACCOUNTED " : "",
3698 (dn
->dn_phys
->dn_flags
& DNODE_FLAG_SPILL_BLKPTR
) ?
3699 "SPILL_BLKPTR" : "");
3700 (void) printf("\tdnode maxblkid: %llu\n",
3701 (longlong_t
)dn
->dn_phys
->dn_maxblkid
);
3704 object_viewer
[ZDB_OT_TYPE(doi
.doi_bonus_type
)](os
,
3705 object
, bonus
, bsize
);
3707 (void) printf("\t\t(bonus encrypted)\n");
3711 (!os
->os_encrypted
|| !DMU_OT_IS_ENCRYPTED(doi
.doi_type
))) {
3712 object_viewer
[ZDB_OT_TYPE(doi
.doi_type
)](os
, object
,
3715 (void) printf("\t\t(object encrypted)\n");
3718 *print_header
= B_TRUE
;
3721 if (verbosity
>= 5) {
3722 if (dn
->dn_phys
->dn_flags
& DNODE_FLAG_SPILL_BLKPTR
) {
3723 char blkbuf
[BP_SPRINTF_LEN
];
3724 snprintf_blkptr_compact(blkbuf
, sizeof (blkbuf
),
3725 DN_SPILL_BLKPTR(dn
->dn_phys
), B_FALSE
);
3726 (void) printf("\nSpill block: %s\n", blkbuf
);
3731 if (verbosity
>= 5) {
3733 * Report the list of segments that comprise the object.
3737 uint64_t blkfill
= 1;
3740 if (dn
->dn_type
== DMU_OT_DNODE
) {
3742 blkfill
= DNODES_PER_BLOCK
;
3747 /* make sure nicenum has enough space */
3748 _Static_assert(sizeof (segsize
) >= NN_NUMBUF_SZ
,
3749 "segsize truncated");
3750 error
= dnode_next_offset(dn
,
3751 0, &start
, minlvl
, blkfill
, 0);
3755 error
= dnode_next_offset(dn
,
3756 DNODE_FIND_HOLE
, &end
, minlvl
, blkfill
, 0);
3757 zdb_nicenum(end
- start
, segsize
, sizeof (segsize
));
3758 (void) printf("\t\tsegment [%016llx, %016llx)"
3759 " size %5s\n", (u_longlong_t
)start
,
3760 (u_longlong_t
)end
, segsize
);
3769 dmu_buf_rele(db
, FTAG
);
3771 dnode_rele(dn
, FTAG
);
3775 count_dir_mos_objects(dsl_dir_t
*dd
)
3777 mos_obj_refd(dd
->dd_object
);
3778 mos_obj_refd(dsl_dir_phys(dd
)->dd_child_dir_zapobj
);
3779 mos_obj_refd(dsl_dir_phys(dd
)->dd_deleg_zapobj
);
3780 mos_obj_refd(dsl_dir_phys(dd
)->dd_props_zapobj
);
3781 mos_obj_refd(dsl_dir_phys(dd
)->dd_clones
);
3784 * The dd_crypto_obj can be referenced by multiple dsl_dir's.
3785 * Ignore the references after the first one.
3787 mos_obj_refd_multiple(dd
->dd_crypto_obj
);
3791 count_ds_mos_objects(dsl_dataset_t
*ds
)
3793 mos_obj_refd(ds
->ds_object
);
3794 mos_obj_refd(dsl_dataset_phys(ds
)->ds_next_clones_obj
);
3795 mos_obj_refd(dsl_dataset_phys(ds
)->ds_props_obj
);
3796 mos_obj_refd(dsl_dataset_phys(ds
)->ds_userrefs_obj
);
3797 mos_obj_refd(dsl_dataset_phys(ds
)->ds_snapnames_zapobj
);
3798 mos_obj_refd(ds
->ds_bookmarks_obj
);
3800 if (!dsl_dataset_is_snapshot(ds
)) {
3801 count_dir_mos_objects(ds
->ds_dir
);
3805 static const char *const objset_types
[DMU_OST_NUMTYPES
] = {
3806 "NONE", "META", "ZPL", "ZVOL", "OTHER", "ANY" };
3809 * Parse a string denoting a range of object IDs of the form
3810 * <start>[:<end>[:flags]], and store the results in zor.
3811 * Return 0 on success. On error, return 1 and update the msg
3812 * pointer to point to a descriptive error message.
3815 parse_object_range(char *range
, zopt_object_range_t
*zor
, const char **msg
)
3818 char *p
, *s
, *dup
, *flagstr
, *tmp
= NULL
;
3823 if (strchr(range
, ':') == NULL
) {
3824 zor
->zor_obj_start
= strtoull(range
, &p
, 0);
3826 *msg
= "Invalid characters in object ID";
3829 zor
->zor_obj_start
= ZDB_MAP_OBJECT_ID(zor
->zor_obj_start
);
3830 zor
->zor_obj_end
= zor
->zor_obj_start
;
3834 if (strchr(range
, ':') == range
) {
3835 *msg
= "Invalid leading colon";
3840 len
= strlen(range
);
3841 if (range
[len
- 1] == ':') {
3842 *msg
= "Invalid trailing colon";
3847 dup
= strdup(range
);
3848 s
= strtok_r(dup
, ":", &tmp
);
3849 zor
->zor_obj_start
= strtoull(s
, &p
, 0);
3852 *msg
= "Invalid characters in start object ID";
3857 s
= strtok_r(NULL
, ":", &tmp
);
3858 zor
->zor_obj_end
= strtoull(s
, &p
, 0);
3861 *msg
= "Invalid characters in end object ID";
3866 if (zor
->zor_obj_start
> zor
->zor_obj_end
) {
3867 *msg
= "Start object ID may not exceed end object ID";
3872 s
= strtok_r(NULL
, ":", &tmp
);
3874 zor
->zor_flags
= ZOR_FLAG_ALL_TYPES
;
3876 } else if (strtok_r(NULL
, ":", &tmp
) != NULL
) {
3877 *msg
= "Invalid colon-delimited field after flags";
3883 for (i
= 0; flagstr
[i
]; i
++) {
3885 boolean_t negation
= (flagstr
[i
] == '-');
3889 if (flagstr
[i
] == '\0') {
3890 *msg
= "Invalid trailing negation operator";
3895 bit
= flagbits
[(uchar_t
)flagstr
[i
]];
3897 *msg
= "Invalid flag";
3906 zor
->zor_flags
= flags
;
3908 zor
->zor_obj_start
= ZDB_MAP_OBJECT_ID(zor
->zor_obj_start
);
3909 zor
->zor_obj_end
= ZDB_MAP_OBJECT_ID(zor
->zor_obj_end
);
3917 dump_objset(objset_t
*os
)
3919 dmu_objset_stats_t dds
= { 0 };
3920 uint64_t object
, object_count
;
3921 uint64_t refdbytes
, usedobjs
, scratch
;
3923 char blkbuf
[BP_SPRINTF_LEN
+ 20];
3924 char osname
[ZFS_MAX_DATASET_NAME_LEN
];
3925 const char *type
= "UNKNOWN";
3926 int verbosity
= dump_opt
['d'];
3927 boolean_t print_header
;
3930 uint64_t total_slots_used
= 0;
3931 uint64_t max_slot_used
= 0;
3932 uint64_t dnode_slots
;
3937 /* make sure nicenum has enough space */
3938 _Static_assert(sizeof (numbuf
) >= NN_NUMBUF_SZ
, "numbuf truncated");
3940 dsl_pool_config_enter(dmu_objset_pool(os
), FTAG
);
3941 dmu_objset_fast_stat(os
, &dds
);
3942 dsl_pool_config_exit(dmu_objset_pool(os
), FTAG
);
3944 print_header
= B_TRUE
;
3946 if (dds
.dds_type
< DMU_OST_NUMTYPES
)
3947 type
= objset_types
[dds
.dds_type
];
3949 if (dds
.dds_type
== DMU_OST_META
) {
3950 dds
.dds_creation_txg
= TXG_INITIAL
;
3951 usedobjs
= BP_GET_FILL(os
->os_rootbp
);
3952 refdbytes
= dsl_dir_phys(os
->os_spa
->spa_dsl_pool
->dp_mos_dir
)->
3955 dmu_objset_space(os
, &refdbytes
, &scratch
, &usedobjs
, &scratch
);
3958 ASSERT3U(usedobjs
, ==, BP_GET_FILL(os
->os_rootbp
));
3960 zdb_nicenum(refdbytes
, numbuf
, sizeof (numbuf
));
3962 if (verbosity
>= 4) {
3963 (void) snprintf(blkbuf
, sizeof (blkbuf
), ", rootbp ");
3964 (void) snprintf_blkptr(blkbuf
+ strlen(blkbuf
),
3965 sizeof (blkbuf
) - strlen(blkbuf
), os
->os_rootbp
);
3970 dmu_objset_name(os
, osname
);
3972 (void) printf("Dataset %s [%s], ID %llu, cr_txg %llu, "
3973 "%s, %llu objects%s%s\n",
3974 osname
, type
, (u_longlong_t
)dmu_objset_id(os
),
3975 (u_longlong_t
)dds
.dds_creation_txg
,
3976 numbuf
, (u_longlong_t
)usedobjs
, blkbuf
,
3977 (dds
.dds_inconsistent
) ? " (inconsistent)" : "");
3979 for (i
= 0; i
< zopt_object_args
; i
++) {
3980 obj_start
= zopt_object_ranges
[i
].zor_obj_start
;
3981 obj_end
= zopt_object_ranges
[i
].zor_obj_end
;
3982 flags
= zopt_object_ranges
[i
].zor_flags
;
3985 if (object
== 0 || obj_start
== obj_end
)
3986 dump_object(os
, object
, verbosity
, &print_header
, NULL
,
3991 while ((dmu_object_next(os
, &object
, B_FALSE
, 0) == 0) &&
3992 object
<= obj_end
) {
3993 dump_object(os
, object
, verbosity
, &print_header
, NULL
,
3998 if (zopt_object_args
> 0) {
3999 (void) printf("\n");
4003 if (dump_opt
['i'] != 0 || verbosity
>= 2)
4004 dump_intent_log(dmu_objset_zil(os
));
4006 if (dmu_objset_ds(os
) != NULL
) {
4007 dsl_dataset_t
*ds
= dmu_objset_ds(os
);
4008 dump_blkptr_list(&ds
->ds_deadlist
, "Deadlist");
4009 if (dsl_deadlist_is_open(&ds
->ds_dir
->dd_livelist
) &&
4010 !dmu_objset_is_snapshot(os
)) {
4011 dump_blkptr_list(&ds
->ds_dir
->dd_livelist
, "Livelist");
4012 if (verify_dd_livelist(os
) != 0)
4013 fatal("livelist is incorrect");
4016 if (dsl_dataset_remap_deadlist_exists(ds
)) {
4017 (void) printf("ds_remap_deadlist:\n");
4018 dump_blkptr_list(&ds
->ds_remap_deadlist
, "Deadlist");
4020 count_ds_mos_objects(ds
);
4023 if (dmu_objset_ds(os
) != NULL
)
4024 dump_bookmarks(os
, verbosity
);
4029 if (BP_IS_HOLE(os
->os_rootbp
))
4032 dump_object(os
, 0, verbosity
, &print_header
, NULL
, 0);
4034 if (DMU_USERUSED_DNODE(os
) != NULL
&&
4035 DMU_USERUSED_DNODE(os
)->dn_type
!= 0) {
4036 dump_object(os
, DMU_USERUSED_OBJECT
, verbosity
, &print_header
,
4038 dump_object(os
, DMU_GROUPUSED_OBJECT
, verbosity
, &print_header
,
4042 if (DMU_PROJECTUSED_DNODE(os
) != NULL
&&
4043 DMU_PROJECTUSED_DNODE(os
)->dn_type
!= 0)
4044 dump_object(os
, DMU_PROJECTUSED_OBJECT
, verbosity
,
4045 &print_header
, NULL
, 0);
4048 while ((error
= dmu_object_next(os
, &object
, B_FALSE
, 0)) == 0) {
4049 dump_object(os
, object
, verbosity
, &print_header
, &dnode_slots
,
4052 total_slots_used
+= dnode_slots
;
4053 max_slot_used
= object
+ dnode_slots
- 1;
4056 (void) printf("\n");
4058 (void) printf(" Dnode slots:\n");
4059 (void) printf("\tTotal used: %10llu\n",
4060 (u_longlong_t
)total_slots_used
);
4061 (void) printf("\tMax used: %10llu\n",
4062 (u_longlong_t
)max_slot_used
);
4063 (void) printf("\tPercent empty: %10lf\n",
4064 (double)(max_slot_used
- total_slots_used
)*100 /
4065 (double)max_slot_used
);
4066 (void) printf("\n");
4068 if (error
!= ESRCH
) {
4069 (void) fprintf(stderr
, "dmu_object_next() = %d\n", error
);
4073 ASSERT3U(object_count
, ==, usedobjs
);
4075 if (leaked_objects
!= 0) {
4076 (void) printf("%d potentially leaked objects detected\n",
4083 dump_uberblock(uberblock_t
*ub
, const char *header
, const char *footer
)
4085 time_t timestamp
= ub
->ub_timestamp
;
4087 (void) printf("%s", header
? header
: "");
4088 (void) printf("\tmagic = %016llx\n", (u_longlong_t
)ub
->ub_magic
);
4089 (void) printf("\tversion = %llu\n", (u_longlong_t
)ub
->ub_version
);
4090 (void) printf("\ttxg = %llu\n", (u_longlong_t
)ub
->ub_txg
);
4091 (void) printf("\tguid_sum = %llu\n", (u_longlong_t
)ub
->ub_guid_sum
);
4092 (void) printf("\ttimestamp = %llu UTC = %s",
4093 (u_longlong_t
)ub
->ub_timestamp
, ctime(×tamp
));
4095 (void) printf("\tmmp_magic = %016llx\n",
4096 (u_longlong_t
)ub
->ub_mmp_magic
);
4097 if (MMP_VALID(ub
)) {
4098 (void) printf("\tmmp_delay = %0llu\n",
4099 (u_longlong_t
)ub
->ub_mmp_delay
);
4100 if (MMP_SEQ_VALID(ub
))
4101 (void) printf("\tmmp_seq = %u\n",
4102 (unsigned int) MMP_SEQ(ub
));
4103 if (MMP_FAIL_INT_VALID(ub
))
4104 (void) printf("\tmmp_fail = %u\n",
4105 (unsigned int) MMP_FAIL_INT(ub
));
4106 if (MMP_INTERVAL_VALID(ub
))
4107 (void) printf("\tmmp_write = %u\n",
4108 (unsigned int) MMP_INTERVAL(ub
));
4109 /* After MMP_* to make summarize_uberblock_mmp cleaner */
4110 (void) printf("\tmmp_valid = %x\n",
4111 (unsigned int) ub
->ub_mmp_config
& 0xFF);
4114 if (dump_opt
['u'] >= 4) {
4115 char blkbuf
[BP_SPRINTF_LEN
];
4116 snprintf_blkptr(blkbuf
, sizeof (blkbuf
), &ub
->ub_rootbp
);
4117 (void) printf("\trootbp = %s\n", blkbuf
);
4119 (void) printf("\tcheckpoint_txg = %llu\n",
4120 (u_longlong_t
)ub
->ub_checkpoint_txg
);
4121 (void) printf("%s", footer
? footer
: "");
4125 dump_config(spa_t
*spa
)
4132 error
= dmu_bonus_hold(spa
->spa_meta_objset
,
4133 spa
->spa_config_object
, FTAG
, &db
);
4136 nvsize
= *(uint64_t *)db
->db_data
;
4137 dmu_buf_rele(db
, FTAG
);
4139 (void) printf("\nMOS Configuration:\n");
4140 dump_packed_nvlist(spa
->spa_meta_objset
,
4141 spa
->spa_config_object
, (void *)&nvsize
, 1);
4143 (void) fprintf(stderr
, "dmu_bonus_hold(%llu) failed, errno %d",
4144 (u_longlong_t
)spa
->spa_config_object
, error
);
4149 dump_cachefile(const char *cachefile
)
4152 struct stat64 statbuf
;
4156 if ((fd
= open64(cachefile
, O_RDONLY
)) < 0) {
4157 (void) printf("cannot open '%s': %s\n", cachefile
,
4162 if (fstat64(fd
, &statbuf
) != 0) {
4163 (void) printf("failed to stat '%s': %s\n", cachefile
,
4168 if ((buf
= malloc(statbuf
.st_size
)) == NULL
) {
4169 (void) fprintf(stderr
, "failed to allocate %llu bytes\n",
4170 (u_longlong_t
)statbuf
.st_size
);
4174 if (read(fd
, buf
, statbuf
.st_size
) != statbuf
.st_size
) {
4175 (void) fprintf(stderr
, "failed to read %llu bytes\n",
4176 (u_longlong_t
)statbuf
.st_size
);
4182 if (nvlist_unpack(buf
, statbuf
.st_size
, &config
, 0) != 0) {
4183 (void) fprintf(stderr
, "failed to unpack nvlist\n");
4189 dump_nvlist(config
, 0);
4191 nvlist_free(config
);
4195 * ZFS label nvlist stats
4197 typedef struct zdb_nvl_stats
{
4200 size_t zns_leaf_largest
;
4201 size_t zns_leaf_total
;
4202 nvlist_t
*zns_string
;
4203 nvlist_t
*zns_uint64
;
4204 nvlist_t
*zns_boolean
;
4208 collect_nvlist_stats(nvlist_t
*nvl
, zdb_nvl_stats_t
*stats
)
4210 nvlist_t
*list
, **array
;
4211 nvpair_t
*nvp
= NULL
;
4215 stats
->zns_list_count
++;
4217 while ((nvp
= nvlist_next_nvpair(nvl
, nvp
)) != NULL
) {
4218 name
= nvpair_name(nvp
);
4220 switch (nvpair_type(nvp
)) {
4221 case DATA_TYPE_STRING
:
4222 fnvlist_add_string(stats
->zns_string
, name
,
4223 fnvpair_value_string(nvp
));
4225 case DATA_TYPE_UINT64
:
4226 fnvlist_add_uint64(stats
->zns_uint64
, name
,
4227 fnvpair_value_uint64(nvp
));
4229 case DATA_TYPE_BOOLEAN
:
4230 fnvlist_add_boolean(stats
->zns_boolean
, name
);
4232 case DATA_TYPE_NVLIST
:
4233 if (nvpair_value_nvlist(nvp
, &list
) == 0)
4234 collect_nvlist_stats(list
, stats
);
4236 case DATA_TYPE_NVLIST_ARRAY
:
4237 if (nvpair_value_nvlist_array(nvp
, &array
, &items
) != 0)
4240 for (i
= 0; i
< items
; i
++) {
4241 collect_nvlist_stats(array
[i
], stats
);
4243 /* collect stats on leaf vdev */
4244 if (strcmp(name
, "children") == 0) {
4247 (void) nvlist_size(array
[i
], &size
,
4249 stats
->zns_leaf_total
+= size
;
4250 if (size
> stats
->zns_leaf_largest
)
4251 stats
->zns_leaf_largest
= size
;
4252 stats
->zns_leaf_count
++;
4257 (void) printf("skip type %d!\n", (int)nvpair_type(nvp
));
4263 dump_nvlist_stats(nvlist_t
*nvl
, size_t cap
)
4265 zdb_nvl_stats_t stats
= { 0 };
4266 size_t size
, sum
= 0, total
;
4269 /* requires nvlist with non-unique names for stat collection */
4270 VERIFY0(nvlist_alloc(&stats
.zns_string
, 0, 0));
4271 VERIFY0(nvlist_alloc(&stats
.zns_uint64
, 0, 0));
4272 VERIFY0(nvlist_alloc(&stats
.zns_boolean
, 0, 0));
4273 VERIFY0(nvlist_size(stats
.zns_boolean
, &noise
, NV_ENCODE_XDR
));
4275 (void) printf("\n\nZFS Label NVList Config Stats:\n");
4277 VERIFY0(nvlist_size(nvl
, &total
, NV_ENCODE_XDR
));
4278 (void) printf(" %d bytes used, %d bytes free (using %4.1f%%)\n\n",
4279 (int)total
, (int)(cap
- total
), 100.0 * total
/ cap
);
4281 collect_nvlist_stats(nvl
, &stats
);
4283 VERIFY0(nvlist_size(stats
.zns_uint64
, &size
, NV_ENCODE_XDR
));
4286 (void) printf("%12s %4d %6d bytes (%5.2f%%)\n", "integers:",
4287 (int)fnvlist_num_pairs(stats
.zns_uint64
),
4288 (int)size
, 100.0 * size
/ total
);
4290 VERIFY0(nvlist_size(stats
.zns_string
, &size
, NV_ENCODE_XDR
));
4293 (void) printf("%12s %4d %6d bytes (%5.2f%%)\n", "strings:",
4294 (int)fnvlist_num_pairs(stats
.zns_string
),
4295 (int)size
, 100.0 * size
/ total
);
4297 VERIFY0(nvlist_size(stats
.zns_boolean
, &size
, NV_ENCODE_XDR
));
4300 (void) printf("%12s %4d %6d bytes (%5.2f%%)\n", "booleans:",
4301 (int)fnvlist_num_pairs(stats
.zns_boolean
),
4302 (int)size
, 100.0 * size
/ total
);
4304 size
= total
- sum
; /* treat remainder as nvlist overhead */
4305 (void) printf("%12s %4d %6d bytes (%5.2f%%)\n\n", "nvlists:",
4306 stats
.zns_list_count
, (int)size
, 100.0 * size
/ total
);
4308 if (stats
.zns_leaf_count
> 0) {
4309 size_t average
= stats
.zns_leaf_total
/ stats
.zns_leaf_count
;
4311 (void) printf("%12s %4d %6d bytes average\n", "leaf vdevs:",
4312 stats
.zns_leaf_count
, (int)average
);
4313 (void) printf("%24d bytes largest\n",
4314 (int)stats
.zns_leaf_largest
);
4316 if (dump_opt
['l'] >= 3 && average
> 0)
4317 (void) printf(" space for %d additional leaf vdevs\n",
4318 (int)((cap
- total
) / average
));
4320 (void) printf("\n");
4322 nvlist_free(stats
.zns_string
);
4323 nvlist_free(stats
.zns_uint64
);
4324 nvlist_free(stats
.zns_boolean
);
4327 typedef struct cksum_record
{
4329 boolean_t labels
[VDEV_LABELS
];
4334 cksum_record_compare(const void *x1
, const void *x2
)
4336 const cksum_record_t
*l
= (cksum_record_t
*)x1
;
4337 const cksum_record_t
*r
= (cksum_record_t
*)x2
;
4338 int arraysize
= ARRAY_SIZE(l
->cksum
.zc_word
);
4341 for (int i
= 0; i
< arraysize
; i
++) {
4342 difference
= TREE_CMP(l
->cksum
.zc_word
[i
], r
->cksum
.zc_word
[i
]);
4347 return (difference
);
4350 static cksum_record_t
*
4351 cksum_record_alloc(zio_cksum_t
*cksum
, int l
)
4353 cksum_record_t
*rec
;
4355 rec
= umem_zalloc(sizeof (*rec
), UMEM_NOFAIL
);
4356 rec
->cksum
= *cksum
;
4357 rec
->labels
[l
] = B_TRUE
;
4362 static cksum_record_t
*
4363 cksum_record_lookup(avl_tree_t
*tree
, zio_cksum_t
*cksum
)
4365 cksum_record_t lookup
= { .cksum
= *cksum
};
4368 return (avl_find(tree
, &lookup
, &where
));
4371 static cksum_record_t
*
4372 cksum_record_insert(avl_tree_t
*tree
, zio_cksum_t
*cksum
, int l
)
4374 cksum_record_t
*rec
;
4376 rec
= cksum_record_lookup(tree
, cksum
);
4378 rec
->labels
[l
] = B_TRUE
;
4380 rec
= cksum_record_alloc(cksum
, l
);
4388 first_label(cksum_record_t
*rec
)
4390 for (int i
= 0; i
< VDEV_LABELS
; i
++)
4398 print_label_numbers(const char *prefix
, const cksum_record_t
*rec
)
4400 fputs(prefix
, stdout
);
4401 for (int i
= 0; i
< VDEV_LABELS
; i
++)
4402 if (rec
->labels
[i
] == B_TRUE
)
4407 #define MAX_UBERBLOCK_COUNT (VDEV_UBERBLOCK_RING >> UBERBLOCK_SHIFT)
4409 typedef struct zdb_label
{
4411 uint64_t label_offset
;
4412 nvlist_t
*config_nv
;
4413 cksum_record_t
*config
;
4414 cksum_record_t
*uberblocks
[MAX_UBERBLOCK_COUNT
];
4415 boolean_t header_printed
;
4416 boolean_t read_failed
;
4417 boolean_t cksum_valid
;
4421 print_label_header(zdb_label_t
*label
, int l
)
4427 if (label
->header_printed
== B_TRUE
)
4430 (void) printf("------------------------------------\n");
4431 (void) printf("LABEL %d %s\n", l
,
4432 label
->cksum_valid
? "" : "(Bad label cksum)");
4433 (void) printf("------------------------------------\n");
4435 label
->header_printed
= B_TRUE
;
4439 print_l2arc_header(void)
4441 (void) printf("------------------------------------\n");
4442 (void) printf("L2ARC device header\n");
4443 (void) printf("------------------------------------\n");
4447 print_l2arc_log_blocks(void)
4449 (void) printf("------------------------------------\n");
4450 (void) printf("L2ARC device log blocks\n");
4451 (void) printf("------------------------------------\n");
4455 dump_l2arc_log_entries(uint64_t log_entries
,
4456 l2arc_log_ent_phys_t
*le
, uint64_t i
)
4458 for (int j
= 0; j
< log_entries
; j
++) {
4459 dva_t dva
= le
[j
].le_dva
;
4460 (void) printf("lb[%4llu]\tle[%4d]\tDVA asize: %llu, "
4461 "vdev: %llu, offset: %llu\n",
4462 (u_longlong_t
)i
, j
+ 1,
4463 (u_longlong_t
)DVA_GET_ASIZE(&dva
),
4464 (u_longlong_t
)DVA_GET_VDEV(&dva
),
4465 (u_longlong_t
)DVA_GET_OFFSET(&dva
));
4466 (void) printf("|\t\t\t\tbirth: %llu\n",
4467 (u_longlong_t
)le
[j
].le_birth
);
4468 (void) printf("|\t\t\t\tlsize: %llu\n",
4469 (u_longlong_t
)L2BLK_GET_LSIZE((&le
[j
])->le_prop
));
4470 (void) printf("|\t\t\t\tpsize: %llu\n",
4471 (u_longlong_t
)L2BLK_GET_PSIZE((&le
[j
])->le_prop
));
4472 (void) printf("|\t\t\t\tcompr: %llu\n",
4473 (u_longlong_t
)L2BLK_GET_COMPRESS((&le
[j
])->le_prop
));
4474 (void) printf("|\t\t\t\tcomplevel: %llu\n",
4475 (u_longlong_t
)(&le
[j
])->le_complevel
);
4476 (void) printf("|\t\t\t\ttype: %llu\n",
4477 (u_longlong_t
)L2BLK_GET_TYPE((&le
[j
])->le_prop
));
4478 (void) printf("|\t\t\t\tprotected: %llu\n",
4479 (u_longlong_t
)L2BLK_GET_PROTECTED((&le
[j
])->le_prop
));
4480 (void) printf("|\t\t\t\tprefetch: %llu\n",
4481 (u_longlong_t
)L2BLK_GET_PREFETCH((&le
[j
])->le_prop
));
4482 (void) printf("|\t\t\t\taddress: %llu\n",
4483 (u_longlong_t
)le
[j
].le_daddr
);
4484 (void) printf("|\t\t\t\tARC state: %llu\n",
4485 (u_longlong_t
)L2BLK_GET_STATE((&le
[j
])->le_prop
));
4486 (void) printf("|\n");
4488 (void) printf("\n");
4492 dump_l2arc_log_blkptr(const l2arc_log_blkptr_t
*lbps
)
4494 (void) printf("|\t\tdaddr: %llu\n", (u_longlong_t
)lbps
->lbp_daddr
);
4495 (void) printf("|\t\tpayload_asize: %llu\n",
4496 (u_longlong_t
)lbps
->lbp_payload_asize
);
4497 (void) printf("|\t\tpayload_start: %llu\n",
4498 (u_longlong_t
)lbps
->lbp_payload_start
);
4499 (void) printf("|\t\tlsize: %llu\n",
4500 (u_longlong_t
)L2BLK_GET_LSIZE(lbps
->lbp_prop
));
4501 (void) printf("|\t\tasize: %llu\n",
4502 (u_longlong_t
)L2BLK_GET_PSIZE(lbps
->lbp_prop
));
4503 (void) printf("|\t\tcompralgo: %llu\n",
4504 (u_longlong_t
)L2BLK_GET_COMPRESS(lbps
->lbp_prop
));
4505 (void) printf("|\t\tcksumalgo: %llu\n",
4506 (u_longlong_t
)L2BLK_GET_CHECKSUM(lbps
->lbp_prop
));
4507 (void) printf("|\n\n");
4511 dump_l2arc_log_blocks(int fd
, const l2arc_dev_hdr_phys_t
*l2dhdr
,
4512 l2arc_dev_hdr_phys_t
*rebuild
)
4514 l2arc_log_blk_phys_t this_lb
;
4516 l2arc_log_blkptr_t lbps
[2];
4523 print_l2arc_log_blocks();
4524 memcpy(lbps
, l2dhdr
->dh_start_lbps
, sizeof (lbps
));
4526 dev
.l2ad_evict
= l2dhdr
->dh_evict
;
4527 dev
.l2ad_start
= l2dhdr
->dh_start
;
4528 dev
.l2ad_end
= l2dhdr
->dh_end
;
4530 if (l2dhdr
->dh_start_lbps
[0].lbp_daddr
== 0) {
4531 /* no log blocks to read */
4532 if (!dump_opt
['q']) {
4533 (void) printf("No log blocks to read\n");
4534 (void) printf("\n");
4538 dev
.l2ad_hand
= lbps
[0].lbp_daddr
+
4539 L2BLK_GET_PSIZE((&lbps
[0])->lbp_prop
);
4542 dev
.l2ad_first
= !!(l2dhdr
->dh_flags
& L2ARC_DEV_HDR_EVICT_FIRST
);
4545 if (!l2arc_log_blkptr_valid(&dev
, &lbps
[0]))
4548 /* L2BLK_GET_PSIZE returns aligned size for log blocks */
4549 asize
= L2BLK_GET_PSIZE((&lbps
[0])->lbp_prop
);
4550 if (pread64(fd
, &this_lb
, asize
, lbps
[0].lbp_daddr
) != asize
) {
4551 if (!dump_opt
['q']) {
4552 (void) printf("Error while reading next log "
4558 fletcher_4_native_varsize(&this_lb
, asize
, &cksum
);
4559 if (!ZIO_CHECKSUM_EQUAL(cksum
, lbps
[0].lbp_cksum
)) {
4561 if (!dump_opt
['q']) {
4562 (void) printf("Invalid cksum\n");
4563 dump_l2arc_log_blkptr(&lbps
[0]);
4568 switch (L2BLK_GET_COMPRESS((&lbps
[0])->lbp_prop
)) {
4569 case ZIO_COMPRESS_OFF
:
4572 abd
= abd_alloc_for_io(asize
, B_TRUE
);
4573 abd_copy_from_buf_off(abd
, &this_lb
, 0, asize
);
4574 if (zio_decompress_data(L2BLK_GET_COMPRESS(
4575 (&lbps
[0])->lbp_prop
), abd
, &this_lb
,
4576 asize
, sizeof (this_lb
), NULL
) != 0) {
4577 (void) printf("L2ARC block decompression "
4586 if (this_lb
.lb_magic
== BSWAP_64(L2ARC_LOG_BLK_MAGIC
))
4587 byteswap_uint64_array(&this_lb
, sizeof (this_lb
));
4588 if (this_lb
.lb_magic
!= L2ARC_LOG_BLK_MAGIC
) {
4590 (void) printf("Invalid log block magic\n\n");
4594 rebuild
->dh_lb_count
++;
4595 rebuild
->dh_lb_asize
+= asize
;
4596 if (dump_opt
['l'] > 1 && !dump_opt
['q']) {
4597 (void) printf("lb[%4llu]\tmagic: %llu\n",
4598 (u_longlong_t
)rebuild
->dh_lb_count
,
4599 (u_longlong_t
)this_lb
.lb_magic
);
4600 dump_l2arc_log_blkptr(&lbps
[0]);
4603 if (dump_opt
['l'] > 2 && !dump_opt
['q'])
4604 dump_l2arc_log_entries(l2dhdr
->dh_log_entries
,
4606 rebuild
->dh_lb_count
);
4608 if (l2arc_range_check_overlap(lbps
[1].lbp_payload_start
,
4609 lbps
[0].lbp_payload_start
, dev
.l2ad_evict
) &&
4614 lbps
[1] = this_lb
.lb_prev_lbp
;
4617 if (!dump_opt
['q']) {
4618 (void) printf("log_blk_count:\t %llu with valid cksum\n",
4619 (u_longlong_t
)rebuild
->dh_lb_count
);
4620 (void) printf("\t\t %d with invalid cksum\n", failed
);
4621 (void) printf("log_blk_asize:\t %llu\n\n",
4622 (u_longlong_t
)rebuild
->dh_lb_asize
);
4627 dump_l2arc_header(int fd
)
4629 l2arc_dev_hdr_phys_t l2dhdr
= {0}, rebuild
= {0};
4630 int error
= B_FALSE
;
4632 if (pread64(fd
, &l2dhdr
, sizeof (l2dhdr
),
4633 VDEV_LABEL_START_SIZE
) != sizeof (l2dhdr
)) {
4636 if (l2dhdr
.dh_magic
== BSWAP_64(L2ARC_DEV_HDR_MAGIC
))
4637 byteswap_uint64_array(&l2dhdr
, sizeof (l2dhdr
));
4639 if (l2dhdr
.dh_magic
!= L2ARC_DEV_HDR_MAGIC
)
4644 (void) printf("L2ARC device header not found\n\n");
4645 /* Do not return an error here for backward compatibility */
4647 } else if (!dump_opt
['q']) {
4648 print_l2arc_header();
4650 (void) printf(" magic: %llu\n",
4651 (u_longlong_t
)l2dhdr
.dh_magic
);
4652 (void) printf(" version: %llu\n",
4653 (u_longlong_t
)l2dhdr
.dh_version
);
4654 (void) printf(" pool_guid: %llu\n",
4655 (u_longlong_t
)l2dhdr
.dh_spa_guid
);
4656 (void) printf(" flags: %llu\n",
4657 (u_longlong_t
)l2dhdr
.dh_flags
);
4658 (void) printf(" start_lbps[0]: %llu\n",
4660 l2dhdr
.dh_start_lbps
[0].lbp_daddr
);
4661 (void) printf(" start_lbps[1]: %llu\n",
4663 l2dhdr
.dh_start_lbps
[1].lbp_daddr
);
4664 (void) printf(" log_blk_ent: %llu\n",
4665 (u_longlong_t
)l2dhdr
.dh_log_entries
);
4666 (void) printf(" start: %llu\n",
4667 (u_longlong_t
)l2dhdr
.dh_start
);
4668 (void) printf(" end: %llu\n",
4669 (u_longlong_t
)l2dhdr
.dh_end
);
4670 (void) printf(" evict: %llu\n",
4671 (u_longlong_t
)l2dhdr
.dh_evict
);
4672 (void) printf(" lb_asize_refcount: %llu\n",
4673 (u_longlong_t
)l2dhdr
.dh_lb_asize
);
4674 (void) printf(" lb_count_refcount: %llu\n",
4675 (u_longlong_t
)l2dhdr
.dh_lb_count
);
4676 (void) printf(" trim_action_time: %llu\n",
4677 (u_longlong_t
)l2dhdr
.dh_trim_action_time
);
4678 (void) printf(" trim_state: %llu\n\n",
4679 (u_longlong_t
)l2dhdr
.dh_trim_state
);
4682 dump_l2arc_log_blocks(fd
, &l2dhdr
, &rebuild
);
4684 * The total aligned size of log blocks and the number of log blocks
4685 * reported in the header of the device may be less than what zdb
4686 * reports by dump_l2arc_log_blocks() which emulates l2arc_rebuild().
4687 * This happens because dump_l2arc_log_blocks() lacks the memory
4688 * pressure valve that l2arc_rebuild() has. Thus, if we are on a system
4689 * with low memory, l2arc_rebuild will exit prematurely and dh_lb_asize
4690 * and dh_lb_count will be lower to begin with than what exists on the
4691 * device. This is normal and zdb should not exit with an error. The
4692 * opposite case should never happen though, the values reported in the
4693 * header should never be higher than what dump_l2arc_log_blocks() and
4694 * l2arc_rebuild() report. If this happens there is a leak in the
4695 * accounting of log blocks.
4697 if (l2dhdr
.dh_lb_asize
> rebuild
.dh_lb_asize
||
4698 l2dhdr
.dh_lb_count
> rebuild
.dh_lb_count
)
4705 dump_config_from_label(zdb_label_t
*label
, size_t buflen
, int l
)
4710 if ((dump_opt
['l'] < 3) && (first_label(label
->config
) != l
))
4713 print_label_header(label
, l
);
4714 dump_nvlist(label
->config_nv
, 4);
4715 print_label_numbers(" labels = ", label
->config
);
4717 if (dump_opt
['l'] >= 2)
4718 dump_nvlist_stats(label
->config_nv
, buflen
);
4721 #define ZDB_MAX_UB_HEADER_SIZE 32
4724 dump_label_uberblocks(zdb_label_t
*label
, uint64_t ashift
, int label_num
)
4728 char header
[ZDB_MAX_UB_HEADER_SIZE
];
4730 vd
.vdev_ashift
= ashift
;
4733 for (int i
= 0; i
< VDEV_UBERBLOCK_COUNT(&vd
); i
++) {
4734 uint64_t uoff
= VDEV_UBERBLOCK_OFFSET(&vd
, i
);
4735 uberblock_t
*ub
= (void *)((char *)&label
->label
+ uoff
);
4736 cksum_record_t
*rec
= label
->uberblocks
[i
];
4739 if (dump_opt
['u'] >= 2) {
4740 print_label_header(label
, label_num
);
4741 (void) printf(" Uberblock[%d] invalid\n", i
);
4746 if ((dump_opt
['u'] < 3) && (first_label(rec
) != label_num
))
4749 if ((dump_opt
['u'] < 4) &&
4750 (ub
->ub_mmp_magic
== MMP_MAGIC
) && ub
->ub_mmp_delay
&&
4751 (i
>= VDEV_UBERBLOCK_COUNT(&vd
) - MMP_BLOCKS_PER_LABEL
))
4754 print_label_header(label
, label_num
);
4755 (void) snprintf(header
, ZDB_MAX_UB_HEADER_SIZE
,
4756 " Uberblock[%d]\n", i
);
4757 dump_uberblock(ub
, header
, "");
4758 print_label_numbers(" labels = ", rec
);
4762 static char curpath
[PATH_MAX
];
4765 * Iterate through the path components, recursively passing
4766 * current one's obj and remaining path until we find the obj
4770 dump_path_impl(objset_t
*os
, uint64_t obj
, char *name
, uint64_t *retobj
)
4773 boolean_t header
= B_TRUE
;
4777 dmu_object_info_t doi
;
4779 if ((s
= strchr(name
, '/')) != NULL
)
4781 err
= zap_lookup(os
, obj
, name
, 8, 1, &child_obj
);
4783 (void) strlcat(curpath
, name
, sizeof (curpath
));
4786 (void) fprintf(stderr
, "failed to lookup %s: %s\n",
4787 curpath
, strerror(err
));
4791 child_obj
= ZFS_DIRENT_OBJ(child_obj
);
4792 err
= sa_buf_hold(os
, child_obj
, FTAG
, &db
);
4794 (void) fprintf(stderr
,
4795 "failed to get SA dbuf for obj %llu: %s\n",
4796 (u_longlong_t
)child_obj
, strerror(err
));
4799 dmu_object_info_from_db(db
, &doi
);
4800 sa_buf_rele(db
, FTAG
);
4802 if (doi
.doi_bonus_type
!= DMU_OT_SA
&&
4803 doi
.doi_bonus_type
!= DMU_OT_ZNODE
) {
4804 (void) fprintf(stderr
, "invalid bonus type %d for obj %llu\n",
4805 doi
.doi_bonus_type
, (u_longlong_t
)child_obj
);
4809 if (dump_opt
['v'] > 6) {
4810 (void) printf("obj=%llu %s type=%d bonustype=%d\n",
4811 (u_longlong_t
)child_obj
, curpath
, doi
.doi_type
,
4812 doi
.doi_bonus_type
);
4815 (void) strlcat(curpath
, "/", sizeof (curpath
));
4817 switch (doi
.doi_type
) {
4818 case DMU_OT_DIRECTORY_CONTENTS
:
4819 if (s
!= NULL
&& *(s
+ 1) != '\0')
4820 return (dump_path_impl(os
, child_obj
, s
+ 1, retobj
));
4822 case DMU_OT_PLAIN_FILE_CONTENTS
:
4823 if (retobj
!= NULL
) {
4824 *retobj
= child_obj
;
4826 dump_object(os
, child_obj
, dump_opt
['v'], &header
,
4831 (void) fprintf(stderr
, "object %llu has non-file/directory "
4832 "type %d\n", (u_longlong_t
)obj
, doi
.doi_type
);
4840 * Dump the blocks for the object specified by path inside the dataset.
4843 dump_path(char *ds
, char *path
, uint64_t *retobj
)
4849 err
= open_objset(ds
, FTAG
, &os
);
4853 err
= zap_lookup(os
, MASTER_NODE_OBJ
, ZFS_ROOT_OBJ
, 8, 1, &root_obj
);
4855 (void) fprintf(stderr
, "can't lookup root znode: %s\n",
4857 close_objset(os
, FTAG
);
4861 (void) snprintf(curpath
, sizeof (curpath
), "dataset=%s path=/", ds
);
4863 err
= dump_path_impl(os
, root_obj
, path
, retobj
);
4865 close_objset(os
, FTAG
);
4870 zdb_copy_object(objset_t
*os
, uint64_t srcobj
, char *destfile
)
4873 uint64_t size
, readsize
, oursize
, offset
;
4877 (void) printf("Copying object %" PRIu64
" to file %s\n", srcobj
,
4880 VERIFY3P(os
, ==, sa_os
);
4881 if ((err
= sa_handle_get(os
, srcobj
, NULL
, SA_HDL_PRIVATE
, &hdl
))) {
4882 (void) printf("Failed to get handle for SA znode\n");
4885 if ((err
= sa_lookup(hdl
, sa_attr_table
[ZPL_SIZE
], &size
, 8))) {
4886 (void) sa_handle_destroy(hdl
);
4889 (void) sa_handle_destroy(hdl
);
4891 (void) printf("Object %" PRIu64
" is %" PRIu64
" bytes\n", srcobj
,
4897 int fd
= open(destfile
, O_WRONLY
| O_CREAT
| O_TRUNC
, 0644);
4901 * We cap the size at 1 mebibyte here to prevent
4902 * allocation failures and nigh-infinite printing if the
4903 * object is extremely large.
4905 oursize
= MIN(size
, 1 << 20);
4907 char *buf
= kmem_alloc(oursize
, KM_NOSLEEP
);
4913 while (offset
< size
) {
4914 readsize
= MIN(size
- offset
, 1 << 20);
4915 err
= dmu_read(os
, srcobj
, offset
, readsize
, buf
, 0);
4917 (void) printf("got error %u from dmu_read\n", err
);
4918 kmem_free(buf
, oursize
);
4922 if (dump_opt
['v'] > 3) {
4923 (void) printf("Read offset=%" PRIu64
" size=%" PRIu64
4924 " error=%d\n", offset
, readsize
, err
);
4927 writesize
= write(fd
, buf
, readsize
);
4928 if (writesize
< 0) {
4931 } else if (writesize
!= readsize
) {
4932 /* Incomplete write */
4933 (void) fprintf(stderr
, "Short write, only wrote %llu of"
4934 " %" PRIu64
" bytes, exiting...\n",
4935 (u_longlong_t
)writesize
, readsize
);
4945 kmem_free(buf
, oursize
);
4951 label_cksum_valid(vdev_label_t
*label
, uint64_t offset
)
4953 zio_checksum_info_t
*ci
= &zio_checksum_table
[ZIO_CHECKSUM_LABEL
];
4954 zio_cksum_t expected_cksum
;
4955 zio_cksum_t actual_cksum
;
4956 zio_cksum_t verifier
;
4960 void *data
= (char *)label
+ offsetof(vdev_label_t
, vl_vdev_phys
);
4961 eck
= (zio_eck_t
*)((char *)(data
) + VDEV_PHYS_SIZE
) - 1;
4963 offset
+= offsetof(vdev_label_t
, vl_vdev_phys
);
4964 ZIO_SET_CHECKSUM(&verifier
, offset
, 0, 0, 0);
4966 byteswap
= (eck
->zec_magic
== BSWAP_64(ZEC_MAGIC
));
4968 byteswap_uint64_array(&verifier
, sizeof (zio_cksum_t
));
4970 expected_cksum
= eck
->zec_cksum
;
4971 eck
->zec_cksum
= verifier
;
4973 abd_t
*abd
= abd_get_from_buf(data
, VDEV_PHYS_SIZE
);
4974 ci
->ci_func
[byteswap
](abd
, VDEV_PHYS_SIZE
, NULL
, &actual_cksum
);
4978 byteswap_uint64_array(&expected_cksum
, sizeof (zio_cksum_t
));
4980 if (ZIO_CHECKSUM_EQUAL(actual_cksum
, expected_cksum
))
4987 dump_label(const char *dev
)
4989 char path
[MAXPATHLEN
];
4990 zdb_label_t labels
[VDEV_LABELS
] = {{{{0}}}};
4991 uint64_t psize
, ashift
, l2cache
;
4992 struct stat64 statbuf
;
4993 boolean_t config_found
= B_FALSE
;
4994 boolean_t error
= B_FALSE
;
4995 boolean_t read_l2arc_header
= B_FALSE
;
4996 avl_tree_t config_tree
;
4997 avl_tree_t uberblock_tree
;
4998 void *node
, *cookie
;
5002 * Check if we were given absolute path and use it as is.
5003 * Otherwise if the provided vdev name doesn't point to a file,
5004 * try prepending expected disk paths and partition numbers.
5006 (void) strlcpy(path
, dev
, sizeof (path
));
5007 if (dev
[0] != '/' && stat64(path
, &statbuf
) != 0) {
5010 error
= zfs_resolve_shortname(dev
, path
, MAXPATHLEN
);
5011 if (error
== 0 && zfs_dev_is_whole_disk(path
)) {
5012 if (zfs_append_partition(path
, MAXPATHLEN
) == -1)
5016 if (error
|| (stat64(path
, &statbuf
) != 0)) {
5017 (void) printf("failed to find device %s, try "
5018 "specifying absolute path instead\n", dev
);
5023 if ((fd
= open64(path
, O_RDONLY
)) < 0) {
5024 (void) printf("cannot open '%s': %s\n", path
, strerror(errno
));
5028 if (fstat64_blk(fd
, &statbuf
) != 0) {
5029 (void) printf("failed to stat '%s': %s\n", path
,
5035 if (S_ISBLK(statbuf
.st_mode
) && zfs_dev_flush(fd
) != 0)
5036 (void) printf("failed to invalidate cache '%s' : %s\n", path
,
5039 avl_create(&config_tree
, cksum_record_compare
,
5040 sizeof (cksum_record_t
), offsetof(cksum_record_t
, link
));
5041 avl_create(&uberblock_tree
, cksum_record_compare
,
5042 sizeof (cksum_record_t
), offsetof(cksum_record_t
, link
));
5044 psize
= statbuf
.st_size
;
5045 psize
= P2ALIGN(psize
, (uint64_t)sizeof (vdev_label_t
));
5046 ashift
= SPA_MINBLOCKSHIFT
;
5049 * 1. Read the label from disk
5050 * 2. Verify label cksum
5051 * 3. Unpack the configuration and insert in config tree.
5052 * 4. Traverse all uberblocks and insert in uberblock tree.
5054 for (int l
= 0; l
< VDEV_LABELS
; l
++) {
5055 zdb_label_t
*label
= &labels
[l
];
5056 char *buf
= label
->label
.vl_vdev_phys
.vp_nvlist
;
5057 size_t buflen
= sizeof (label
->label
.vl_vdev_phys
.vp_nvlist
);
5059 cksum_record_t
*rec
;
5063 label
->label_offset
= vdev_label_offset(psize
, l
, 0);
5065 if (pread64(fd
, &label
->label
, sizeof (label
->label
),
5066 label
->label_offset
) != sizeof (label
->label
)) {
5068 (void) printf("failed to read label %d\n", l
);
5069 label
->read_failed
= B_TRUE
;
5074 label
->read_failed
= B_FALSE
;
5075 label
->cksum_valid
= label_cksum_valid(&label
->label
,
5076 label
->label_offset
);
5078 if (nvlist_unpack(buf
, buflen
, &config
, 0) == 0) {
5079 nvlist_t
*vdev_tree
= NULL
;
5082 if ((nvlist_lookup_nvlist(config
,
5083 ZPOOL_CONFIG_VDEV_TREE
, &vdev_tree
) != 0) ||
5084 (nvlist_lookup_uint64(vdev_tree
,
5085 ZPOOL_CONFIG_ASHIFT
, &ashift
) != 0))
5086 ashift
= SPA_MINBLOCKSHIFT
;
5088 if (nvlist_size(config
, &size
, NV_ENCODE_XDR
) != 0)
5091 /* If the device is a cache device clear the header. */
5092 if (!read_l2arc_header
) {
5093 if (nvlist_lookup_uint64(config
,
5094 ZPOOL_CONFIG_POOL_STATE
, &l2cache
) == 0 &&
5095 l2cache
== POOL_STATE_L2CACHE
) {
5096 read_l2arc_header
= B_TRUE
;
5100 fletcher_4_native_varsize(buf
, size
, &cksum
);
5101 rec
= cksum_record_insert(&config_tree
, &cksum
, l
);
5103 label
->config
= rec
;
5104 label
->config_nv
= config
;
5105 config_found
= B_TRUE
;
5110 vd
.vdev_ashift
= ashift
;
5113 for (int i
= 0; i
< VDEV_UBERBLOCK_COUNT(&vd
); i
++) {
5114 uint64_t uoff
= VDEV_UBERBLOCK_OFFSET(&vd
, i
);
5115 uberblock_t
*ub
= (void *)((char *)label
+ uoff
);
5117 if (uberblock_verify(ub
))
5120 fletcher_4_native_varsize(ub
, sizeof (*ub
), &cksum
);
5121 rec
= cksum_record_insert(&uberblock_tree
, &cksum
, l
);
5123 label
->uberblocks
[i
] = rec
;
5128 * Dump the label and uberblocks.
5130 for (int l
= 0; l
< VDEV_LABELS
; l
++) {
5131 zdb_label_t
*label
= &labels
[l
];
5132 size_t buflen
= sizeof (label
->label
.vl_vdev_phys
.vp_nvlist
);
5134 if (label
->read_failed
== B_TRUE
)
5137 if (label
->config_nv
) {
5138 dump_config_from_label(label
, buflen
, l
);
5141 (void) printf("failed to unpack label %d\n", l
);
5145 dump_label_uberblocks(label
, ashift
, l
);
5147 nvlist_free(label
->config_nv
);
5151 * Dump the L2ARC header, if existent.
5153 if (read_l2arc_header
)
5154 error
|= dump_l2arc_header(fd
);
5157 while ((node
= avl_destroy_nodes(&config_tree
, &cookie
)) != NULL
)
5158 umem_free(node
, sizeof (cksum_record_t
));
5161 while ((node
= avl_destroy_nodes(&uberblock_tree
, &cookie
)) != NULL
)
5162 umem_free(node
, sizeof (cksum_record_t
));
5164 avl_destroy(&config_tree
);
5165 avl_destroy(&uberblock_tree
);
5169 return (config_found
== B_FALSE
? 2 :
5170 (error
== B_TRUE
? 1 : 0));
5173 static uint64_t dataset_feature_count
[SPA_FEATURES
];
5174 static uint64_t global_feature_count
[SPA_FEATURES
];
5175 static uint64_t remap_deadlist_count
= 0;
5178 dump_one_objset(const char *dsname
, void *arg
)
5185 error
= open_objset(dsname
, FTAG
, &os
);
5189 for (f
= 0; f
< SPA_FEATURES
; f
++) {
5190 if (!dsl_dataset_feature_is_active(dmu_objset_ds(os
), f
))
5192 ASSERT(spa_feature_table
[f
].fi_flags
&
5193 ZFEATURE_FLAG_PER_DATASET
);
5194 dataset_feature_count
[f
]++;
5197 if (dsl_dataset_remap_deadlist_exists(dmu_objset_ds(os
))) {
5198 remap_deadlist_count
++;
5201 for (dsl_bookmark_node_t
*dbn
=
5202 avl_first(&dmu_objset_ds(os
)->ds_bookmarks
); dbn
!= NULL
;
5203 dbn
= AVL_NEXT(&dmu_objset_ds(os
)->ds_bookmarks
, dbn
)) {
5204 mos_obj_refd(dbn
->dbn_phys
.zbm_redaction_obj
);
5205 if (dbn
->dbn_phys
.zbm_redaction_obj
!= 0)
5206 global_feature_count
[SPA_FEATURE_REDACTION_BOOKMARKS
]++;
5207 if (dbn
->dbn_phys
.zbm_flags
& ZBM_FLAG_HAS_FBN
)
5208 global_feature_count
[SPA_FEATURE_BOOKMARK_WRITTEN
]++;
5211 if (dsl_deadlist_is_open(&dmu_objset_ds(os
)->ds_dir
->dd_livelist
) &&
5212 !dmu_objset_is_snapshot(os
)) {
5213 global_feature_count
[SPA_FEATURE_LIVELIST
]++;
5217 close_objset(os
, FTAG
);
5218 fuid_table_destroy();
5225 #define PSIZE_HISTO_SIZE (SPA_OLD_MAXBLOCKSIZE / SPA_MINBLOCKSIZE + 2)
5226 typedef struct zdb_blkstats
{
5232 uint64_t zb_ditto_samevdev
;
5233 uint64_t zb_ditto_same_ms
;
5234 uint64_t zb_psize_histogram
[PSIZE_HISTO_SIZE
];
5238 * Extended object types to report deferred frees and dedup auto-ditto blocks.
5240 #define ZDB_OT_DEFERRED (DMU_OT_NUMTYPES + 0)
5241 #define ZDB_OT_DITTO (DMU_OT_NUMTYPES + 1)
5242 #define ZDB_OT_OTHER (DMU_OT_NUMTYPES + 2)
5243 #define ZDB_OT_TOTAL (DMU_OT_NUMTYPES + 3)
5245 static const char *zdb_ot_extname
[] = {
5252 #define ZB_TOTAL DN_MAX_LEVELS
5253 #define SPA_MAX_FOR_16M (SPA_MAXBLOCKSHIFT+1)
5255 typedef struct zdb_cb
{
5256 zdb_blkstats_t zcb_type
[ZB_TOTAL
+ 1][ZDB_OT_TOTAL
+ 1];
5257 uint64_t zcb_removing_size
;
5258 uint64_t zcb_checkpoint_size
;
5259 uint64_t zcb_dedup_asize
;
5260 uint64_t zcb_dedup_blocks
;
5261 uint64_t zcb_psize_count
[SPA_MAX_FOR_16M
];
5262 uint64_t zcb_lsize_count
[SPA_MAX_FOR_16M
];
5263 uint64_t zcb_asize_count
[SPA_MAX_FOR_16M
];
5264 uint64_t zcb_psize_len
[SPA_MAX_FOR_16M
];
5265 uint64_t zcb_lsize_len
[SPA_MAX_FOR_16M
];
5266 uint64_t zcb_asize_len
[SPA_MAX_FOR_16M
];
5267 uint64_t zcb_psize_total
;
5268 uint64_t zcb_lsize_total
;
5269 uint64_t zcb_asize_total
;
5270 uint64_t zcb_embedded_blocks
[NUM_BP_EMBEDDED_TYPES
];
5271 uint64_t zcb_embedded_histogram
[NUM_BP_EMBEDDED_TYPES
]
5272 [BPE_PAYLOAD_SIZE
+ 1];
5274 hrtime_t zcb_lastprint
;
5275 uint64_t zcb_totalasize
;
5276 uint64_t zcb_errors
[256];
5280 uint32_t **zcb_vd_obsolete_counts
;
5283 /* test if two DVA offsets from same vdev are within the same metaslab */
5285 same_metaslab(spa_t
*spa
, uint64_t vdev
, uint64_t off1
, uint64_t off2
)
5287 vdev_t
*vd
= vdev_lookup_top(spa
, vdev
);
5288 uint64_t ms_shift
= vd
->vdev_ms_shift
;
5290 return ((off1
>> ms_shift
) == (off2
>> ms_shift
));
5294 * Used to simplify reporting of the histogram data.
5296 typedef struct one_histo
{
5300 uint64_t cumulative
;
5304 * The number of separate histograms processed for psize, lsize and asize.
5309 * This routine will create a fixed column size output of three different
5310 * histograms showing by blocksize of 512 - 2^ SPA_MAX_FOR_16M
5311 * the count, length and cumulative length of the psize, lsize and
5314 * All three types of blocks are listed on a single line
5316 * By default the table is printed in nicenumber format (e.g. 123K) but
5317 * if the '-P' parameter is specified then the full raw number (parseable)
5321 dump_size_histograms(zdb_cb_t
*zcb
)
5324 * A temporary buffer that allows us to convert a number into
5325 * a string using zdb_nicenumber to allow either raw or human
5326 * readable numbers to be output.
5331 * Define titles which are used in the headers of the tables
5332 * printed by this routine.
5334 const char blocksize_title1
[] = "block";
5335 const char blocksize_title2
[] = "size";
5336 const char count_title
[] = "Count";
5337 const char length_title
[] = "Size";
5338 const char cumulative_title
[] = "Cum.";
5341 * Setup the histogram arrays (psize, lsize, and asize).
5343 one_histo_t parm_histo
[NUM_HISTO
];
5345 parm_histo
[0].name
= "psize";
5346 parm_histo
[0].count
= zcb
->zcb_psize_count
;
5347 parm_histo
[0].len
= zcb
->zcb_psize_len
;
5348 parm_histo
[0].cumulative
= 0;
5350 parm_histo
[1].name
= "lsize";
5351 parm_histo
[1].count
= zcb
->zcb_lsize_count
;
5352 parm_histo
[1].len
= zcb
->zcb_lsize_len
;
5353 parm_histo
[1].cumulative
= 0;
5355 parm_histo
[2].name
= "asize";
5356 parm_histo
[2].count
= zcb
->zcb_asize_count
;
5357 parm_histo
[2].len
= zcb
->zcb_asize_len
;
5358 parm_histo
[2].cumulative
= 0;
5361 (void) printf("\nBlock Size Histogram\n");
5363 * Print the first line titles
5366 (void) printf("\n%s\t", blocksize_title1
);
5368 (void) printf("\n%7s ", blocksize_title1
);
5370 for (int j
= 0; j
< NUM_HISTO
; j
++) {
5371 if (dump_opt
['P']) {
5372 if (j
< NUM_HISTO
- 1) {
5373 (void) printf("%s\t\t\t", parm_histo
[j
].name
);
5375 /* Don't print trailing spaces */
5376 (void) printf(" %s", parm_histo
[j
].name
);
5379 if (j
< NUM_HISTO
- 1) {
5380 /* Left aligned strings in the output */
5381 (void) printf("%-7s ",
5382 parm_histo
[j
].name
);
5384 /* Don't print trailing spaces */
5385 (void) printf("%s", parm_histo
[j
].name
);
5389 (void) printf("\n");
5392 * Print the second line titles
5394 if (dump_opt
['P']) {
5395 (void) printf("%s\t", blocksize_title2
);
5397 (void) printf("%7s ", blocksize_title2
);
5400 for (int i
= 0; i
< NUM_HISTO
; i
++) {
5401 if (dump_opt
['P']) {
5402 (void) printf("%s\t%s\t%s\t",
5403 count_title
, length_title
, cumulative_title
);
5405 (void) printf("%7s%7s%7s",
5406 count_title
, length_title
, cumulative_title
);
5409 (void) printf("\n");
5414 for (int i
= SPA_MINBLOCKSHIFT
; i
< SPA_MAX_FOR_16M
; i
++) {
5417 * Print the first column showing the blocksize
5419 zdb_nicenum((1ULL << i
), numbuf
, sizeof (numbuf
));
5421 if (dump_opt
['P']) {
5422 printf("%s", numbuf
);
5424 printf("%7s:", numbuf
);
5428 * Print the remaining set of 3 columns per size:
5429 * for psize, lsize and asize
5431 for (int j
= 0; j
< NUM_HISTO
; j
++) {
5432 parm_histo
[j
].cumulative
+= parm_histo
[j
].len
[i
];
5434 zdb_nicenum(parm_histo
[j
].count
[i
],
5435 numbuf
, sizeof (numbuf
));
5437 (void) printf("\t%s", numbuf
);
5439 (void) printf("%7s", numbuf
);
5441 zdb_nicenum(parm_histo
[j
].len
[i
],
5442 numbuf
, sizeof (numbuf
));
5444 (void) printf("\t%s", numbuf
);
5446 (void) printf("%7s", numbuf
);
5448 zdb_nicenum(parm_histo
[j
].cumulative
,
5449 numbuf
, sizeof (numbuf
));
5451 (void) printf("\t%s", numbuf
);
5453 (void) printf("%7s", numbuf
);
5455 (void) printf("\n");
5460 zdb_count_block(zdb_cb_t
*zcb
, zilog_t
*zilog
, const blkptr_t
*bp
,
5461 dmu_object_type_t type
)
5463 uint64_t refcnt
= 0;
5466 ASSERT(type
< ZDB_OT_TOTAL
);
5468 if (zilog
&& zil_bp_tree_add(zilog
, bp
) != 0)
5471 spa_config_enter(zcb
->zcb_spa
, SCL_CONFIG
, FTAG
, RW_READER
);
5473 for (i
= 0; i
< 4; i
++) {
5474 int l
= (i
< 2) ? BP_GET_LEVEL(bp
) : ZB_TOTAL
;
5475 int t
= (i
& 1) ? type
: ZDB_OT_TOTAL
;
5477 zdb_blkstats_t
*zb
= &zcb
->zcb_type
[l
][t
];
5479 zb
->zb_asize
+= BP_GET_ASIZE(bp
);
5480 zb
->zb_lsize
+= BP_GET_LSIZE(bp
);
5481 zb
->zb_psize
+= BP_GET_PSIZE(bp
);
5485 * The histogram is only big enough to record blocks up to
5486 * SPA_OLD_MAXBLOCKSIZE; larger blocks go into the last,
5489 unsigned idx
= BP_GET_PSIZE(bp
) >> SPA_MINBLOCKSHIFT
;
5490 idx
= MIN(idx
, SPA_OLD_MAXBLOCKSIZE
/ SPA_MINBLOCKSIZE
+ 1);
5491 zb
->zb_psize_histogram
[idx
]++;
5493 zb
->zb_gangs
+= BP_COUNT_GANG(bp
);
5495 switch (BP_GET_NDVAS(bp
)) {
5497 if (DVA_GET_VDEV(&bp
->blk_dva
[0]) ==
5498 DVA_GET_VDEV(&bp
->blk_dva
[1])) {
5499 zb
->zb_ditto_samevdev
++;
5501 if (same_metaslab(zcb
->zcb_spa
,
5502 DVA_GET_VDEV(&bp
->blk_dva
[0]),
5503 DVA_GET_OFFSET(&bp
->blk_dva
[0]),
5504 DVA_GET_OFFSET(&bp
->blk_dva
[1])))
5505 zb
->zb_ditto_same_ms
++;
5509 equal
= (DVA_GET_VDEV(&bp
->blk_dva
[0]) ==
5510 DVA_GET_VDEV(&bp
->blk_dva
[1])) +
5511 (DVA_GET_VDEV(&bp
->blk_dva
[0]) ==
5512 DVA_GET_VDEV(&bp
->blk_dva
[2])) +
5513 (DVA_GET_VDEV(&bp
->blk_dva
[1]) ==
5514 DVA_GET_VDEV(&bp
->blk_dva
[2]));
5516 zb
->zb_ditto_samevdev
++;
5518 if (DVA_GET_VDEV(&bp
->blk_dva
[0]) ==
5519 DVA_GET_VDEV(&bp
->blk_dva
[1]) &&
5520 same_metaslab(zcb
->zcb_spa
,
5521 DVA_GET_VDEV(&bp
->blk_dva
[0]),
5522 DVA_GET_OFFSET(&bp
->blk_dva
[0]),
5523 DVA_GET_OFFSET(&bp
->blk_dva
[1])))
5524 zb
->zb_ditto_same_ms
++;
5525 else if (DVA_GET_VDEV(&bp
->blk_dva
[0]) ==
5526 DVA_GET_VDEV(&bp
->blk_dva
[2]) &&
5527 same_metaslab(zcb
->zcb_spa
,
5528 DVA_GET_VDEV(&bp
->blk_dva
[0]),
5529 DVA_GET_OFFSET(&bp
->blk_dva
[0]),
5530 DVA_GET_OFFSET(&bp
->blk_dva
[2])))
5531 zb
->zb_ditto_same_ms
++;
5532 else if (DVA_GET_VDEV(&bp
->blk_dva
[1]) ==
5533 DVA_GET_VDEV(&bp
->blk_dva
[2]) &&
5534 same_metaslab(zcb
->zcb_spa
,
5535 DVA_GET_VDEV(&bp
->blk_dva
[1]),
5536 DVA_GET_OFFSET(&bp
->blk_dva
[1]),
5537 DVA_GET_OFFSET(&bp
->blk_dva
[2])))
5538 zb
->zb_ditto_same_ms
++;
5544 spa_config_exit(zcb
->zcb_spa
, SCL_CONFIG
, FTAG
);
5546 if (BP_IS_EMBEDDED(bp
)) {
5547 zcb
->zcb_embedded_blocks
[BPE_GET_ETYPE(bp
)]++;
5548 zcb
->zcb_embedded_histogram
[BPE_GET_ETYPE(bp
)]
5549 [BPE_GET_PSIZE(bp
)]++;
5553 * The binning histogram bins by powers of two up to
5554 * SPA_MAXBLOCKSIZE rather than creating bins for
5555 * every possible blocksize found in the pool.
5557 int bin
= highbit64(BP_GET_PSIZE(bp
)) - 1;
5559 zcb
->zcb_psize_count
[bin
]++;
5560 zcb
->zcb_psize_len
[bin
] += BP_GET_PSIZE(bp
);
5561 zcb
->zcb_psize_total
+= BP_GET_PSIZE(bp
);
5563 bin
= highbit64(BP_GET_LSIZE(bp
)) - 1;
5565 zcb
->zcb_lsize_count
[bin
]++;
5566 zcb
->zcb_lsize_len
[bin
] += BP_GET_LSIZE(bp
);
5567 zcb
->zcb_lsize_total
+= BP_GET_LSIZE(bp
);
5569 bin
= highbit64(BP_GET_ASIZE(bp
)) - 1;
5571 zcb
->zcb_asize_count
[bin
]++;
5572 zcb
->zcb_asize_len
[bin
] += BP_GET_ASIZE(bp
);
5573 zcb
->zcb_asize_total
+= BP_GET_ASIZE(bp
);
5578 if (BP_GET_DEDUP(bp
)) {
5582 ddt
= ddt_select(zcb
->zcb_spa
, bp
);
5584 dde
= ddt_lookup(ddt
, bp
, B_FALSE
);
5589 ddt_phys_t
*ddp
= ddt_phys_select(dde
, bp
);
5590 ddt_phys_decref(ddp
);
5591 refcnt
= ddp
->ddp_refcnt
;
5592 if (ddt_phys_total_refcnt(dde
) == 0)
5593 ddt_remove(ddt
, dde
);
5598 VERIFY3U(zio_wait(zio_claim(NULL
, zcb
->zcb_spa
,
5599 refcnt
? 0 : spa_min_claim_txg(zcb
->zcb_spa
),
5600 bp
, NULL
, NULL
, ZIO_FLAG_CANFAIL
)), ==, 0);
5604 zdb_blkptr_done(zio_t
*zio
)
5606 spa_t
*spa
= zio
->io_spa
;
5607 blkptr_t
*bp
= zio
->io_bp
;
5608 int ioerr
= zio
->io_error
;
5609 zdb_cb_t
*zcb
= zio
->io_private
;
5610 zbookmark_phys_t
*zb
= &zio
->io_bookmark
;
5612 mutex_enter(&spa
->spa_scrub_lock
);
5613 spa
->spa_load_verify_bytes
-= BP_GET_PSIZE(bp
);
5614 cv_broadcast(&spa
->spa_scrub_io_cv
);
5616 if (ioerr
&& !(zio
->io_flags
& ZIO_FLAG_SPECULATIVE
)) {
5617 char blkbuf
[BP_SPRINTF_LEN
];
5619 zcb
->zcb_haderrors
= 1;
5620 zcb
->zcb_errors
[ioerr
]++;
5622 if (dump_opt
['b'] >= 2)
5623 snprintf_blkptr(blkbuf
, sizeof (blkbuf
), bp
);
5627 (void) printf("zdb_blkptr_cb: "
5628 "Got error %d reading "
5629 "<%llu, %llu, %lld, %llx> %s -- skipping\n",
5631 (u_longlong_t
)zb
->zb_objset
,
5632 (u_longlong_t
)zb
->zb_object
,
5633 (u_longlong_t
)zb
->zb_level
,
5634 (u_longlong_t
)zb
->zb_blkid
,
5637 mutex_exit(&spa
->spa_scrub_lock
);
5639 abd_free(zio
->io_abd
);
5643 zdb_blkptr_cb(spa_t
*spa
, zilog_t
*zilog
, const blkptr_t
*bp
,
5644 const zbookmark_phys_t
*zb
, const dnode_phys_t
*dnp
, void *arg
)
5646 zdb_cb_t
*zcb
= arg
;
5647 dmu_object_type_t type
;
5648 boolean_t is_metadata
;
5650 if (zb
->zb_level
== ZB_DNODE_LEVEL
)
5653 if (dump_opt
['b'] >= 5 && bp
->blk_birth
> 0) {
5654 char blkbuf
[BP_SPRINTF_LEN
];
5655 snprintf_blkptr(blkbuf
, sizeof (blkbuf
), bp
);
5656 (void) printf("objset %llu object %llu "
5657 "level %lld offset 0x%llx %s\n",
5658 (u_longlong_t
)zb
->zb_objset
,
5659 (u_longlong_t
)zb
->zb_object
,
5660 (longlong_t
)zb
->zb_level
,
5661 (u_longlong_t
)blkid2offset(dnp
, bp
, zb
),
5665 if (BP_IS_HOLE(bp
) || BP_IS_REDACTED(bp
))
5668 type
= BP_GET_TYPE(bp
);
5670 zdb_count_block(zcb
, zilog
, bp
,
5671 (type
& DMU_OT_NEWTYPE
) ? ZDB_OT_OTHER
: type
);
5673 is_metadata
= (BP_GET_LEVEL(bp
) != 0 || DMU_OT_IS_METADATA(type
));
5675 if (!BP_IS_EMBEDDED(bp
) &&
5676 (dump_opt
['c'] > 1 || (dump_opt
['c'] && is_metadata
))) {
5677 size_t size
= BP_GET_PSIZE(bp
);
5678 abd_t
*abd
= abd_alloc(size
, B_FALSE
);
5679 int flags
= ZIO_FLAG_CANFAIL
| ZIO_FLAG_SCRUB
| ZIO_FLAG_RAW
;
5681 /* If it's an intent log block, failure is expected. */
5682 if (zb
->zb_level
== ZB_ZIL_LEVEL
)
5683 flags
|= ZIO_FLAG_SPECULATIVE
;
5685 mutex_enter(&spa
->spa_scrub_lock
);
5686 while (spa
->spa_load_verify_bytes
> max_inflight_bytes
)
5687 cv_wait(&spa
->spa_scrub_io_cv
, &spa
->spa_scrub_lock
);
5688 spa
->spa_load_verify_bytes
+= size
;
5689 mutex_exit(&spa
->spa_scrub_lock
);
5691 zio_nowait(zio_read(NULL
, spa
, bp
, abd
, size
,
5692 zdb_blkptr_done
, zcb
, ZIO_PRIORITY_ASYNC_READ
, flags
, zb
));
5695 zcb
->zcb_readfails
= 0;
5697 /* only call gethrtime() every 100 blocks */
5704 if (dump_opt
['b'] < 5 && gethrtime() > zcb
->zcb_lastprint
+ NANOSEC
) {
5705 uint64_t now
= gethrtime();
5707 uint64_t bytes
= zcb
->zcb_type
[ZB_TOTAL
][ZDB_OT_TOTAL
].zb_asize
;
5708 uint64_t kb_per_sec
=
5709 1 + bytes
/ (1 + ((now
- zcb
->zcb_start
) / 1000 / 1000));
5710 uint64_t sec_remaining
=
5711 (zcb
->zcb_totalasize
- bytes
) / 1024 / kb_per_sec
;
5713 /* make sure nicenum has enough space */
5714 _Static_assert(sizeof (buf
) >= NN_NUMBUF_SZ
, "buf truncated");
5716 zfs_nicebytes(bytes
, buf
, sizeof (buf
));
5717 (void) fprintf(stderr
,
5718 "\r%5s completed (%4"PRIu64
"MB/s) "
5719 "estimated time remaining: "
5720 "%"PRIu64
"hr %02"PRIu64
"min %02"PRIu64
"sec ",
5721 buf
, kb_per_sec
/ 1024,
5722 sec_remaining
/ 60 / 60,
5723 sec_remaining
/ 60 % 60,
5724 sec_remaining
% 60);
5726 zcb
->zcb_lastprint
= now
;
5733 zdb_leak(void *arg
, uint64_t start
, uint64_t size
)
5737 (void) printf("leaked space: vdev %llu, offset 0x%llx, size %llu\n",
5738 (u_longlong_t
)vd
->vdev_id
, (u_longlong_t
)start
, (u_longlong_t
)size
);
5741 static metaslab_ops_t zdb_metaslab_ops
= {
5746 load_unflushed_svr_segs_cb(spa_t
*spa
, space_map_entry_t
*sme
,
5747 uint64_t txg
, void *arg
)
5749 spa_vdev_removal_t
*svr
= arg
;
5751 uint64_t offset
= sme
->sme_offset
;
5752 uint64_t size
= sme
->sme_run
;
5754 /* skip vdevs we don't care about */
5755 if (sme
->sme_vdev
!= svr
->svr_vdev_id
)
5758 vdev_t
*vd
= vdev_lookup_top(spa
, sme
->sme_vdev
);
5759 metaslab_t
*ms
= vd
->vdev_ms
[offset
>> vd
->vdev_ms_shift
];
5760 ASSERT(sme
->sme_type
== SM_ALLOC
|| sme
->sme_type
== SM_FREE
);
5762 if (txg
< metaslab_unflushed_txg(ms
))
5765 if (sme
->sme_type
== SM_ALLOC
)
5766 range_tree_add(svr
->svr_allocd_segs
, offset
, size
);
5768 range_tree_remove(svr
->svr_allocd_segs
, offset
, size
);
5774 claim_segment_impl_cb(uint64_t inner_offset
, vdev_t
*vd
, uint64_t offset
,
5775 uint64_t size
, void *arg
)
5777 (void) inner_offset
, (void) arg
;
5780 * This callback was called through a remap from
5781 * a device being removed. Therefore, the vdev that
5782 * this callback is applied to is a concrete
5785 ASSERT(vdev_is_concrete(vd
));
5787 VERIFY0(metaslab_claim_impl(vd
, offset
, size
,
5788 spa_min_claim_txg(vd
->vdev_spa
)));
5792 claim_segment_cb(void *arg
, uint64_t offset
, uint64_t size
)
5796 vdev_indirect_ops
.vdev_op_remap(vd
, offset
, size
,
5797 claim_segment_impl_cb
, NULL
);
5801 * After accounting for all allocated blocks that are directly referenced,
5802 * we might have missed a reference to a block from a partially complete
5803 * (and thus unused) indirect mapping object. We perform a secondary pass
5804 * through the metaslabs we have already mapped and claim the destination
5808 zdb_claim_removing(spa_t
*spa
, zdb_cb_t
*zcb
)
5813 if (spa
->spa_vdev_removal
== NULL
)
5816 spa_config_enter(spa
, SCL_CONFIG
, FTAG
, RW_READER
);
5818 spa_vdev_removal_t
*svr
= spa
->spa_vdev_removal
;
5819 vdev_t
*vd
= vdev_lookup_top(spa
, svr
->svr_vdev_id
);
5820 vdev_indirect_mapping_t
*vim
= vd
->vdev_indirect_mapping
;
5822 ASSERT0(range_tree_space(svr
->svr_allocd_segs
));
5824 range_tree_t
*allocs
= range_tree_create(NULL
, RANGE_SEG64
, NULL
, 0, 0);
5825 for (uint64_t msi
= 0; msi
< vd
->vdev_ms_count
; msi
++) {
5826 metaslab_t
*msp
= vd
->vdev_ms
[msi
];
5828 ASSERT0(range_tree_space(allocs
));
5829 if (msp
->ms_sm
!= NULL
)
5830 VERIFY0(space_map_load(msp
->ms_sm
, allocs
, SM_ALLOC
));
5831 range_tree_vacate(allocs
, range_tree_add
, svr
->svr_allocd_segs
);
5833 range_tree_destroy(allocs
);
5835 iterate_through_spacemap_logs(spa
, load_unflushed_svr_segs_cb
, svr
);
5838 * Clear everything past what has been synced,
5839 * because we have not allocated mappings for
5842 range_tree_clear(svr
->svr_allocd_segs
,
5843 vdev_indirect_mapping_max_offset(vim
),
5844 vd
->vdev_asize
- vdev_indirect_mapping_max_offset(vim
));
5846 zcb
->zcb_removing_size
+= range_tree_space(svr
->svr_allocd_segs
);
5847 range_tree_vacate(svr
->svr_allocd_segs
, claim_segment_cb
, vd
);
5849 spa_config_exit(spa
, SCL_CONFIG
, FTAG
);
5853 increment_indirect_mapping_cb(void *arg
, const blkptr_t
*bp
, boolean_t bp_freed
,
5857 zdb_cb_t
*zcb
= arg
;
5858 spa_t
*spa
= zcb
->zcb_spa
;
5860 const dva_t
*dva
= &bp
->blk_dva
[0];
5863 ASSERT(!dump_opt
['L']);
5864 ASSERT3U(BP_GET_NDVAS(bp
), ==, 1);
5866 spa_config_enter(spa
, SCL_VDEV
, FTAG
, RW_READER
);
5867 vd
= vdev_lookup_top(zcb
->zcb_spa
, DVA_GET_VDEV(dva
));
5868 ASSERT3P(vd
, !=, NULL
);
5869 spa_config_exit(spa
, SCL_VDEV
, FTAG
);
5871 ASSERT(vd
->vdev_indirect_config
.vic_mapping_object
!= 0);
5872 ASSERT3P(zcb
->zcb_vd_obsolete_counts
[vd
->vdev_id
], !=, NULL
);
5874 vdev_indirect_mapping_increment_obsolete_count(
5875 vd
->vdev_indirect_mapping
,
5876 DVA_GET_OFFSET(dva
), DVA_GET_ASIZE(dva
),
5877 zcb
->zcb_vd_obsolete_counts
[vd
->vdev_id
]);
5883 zdb_load_obsolete_counts(vdev_t
*vd
)
5885 vdev_indirect_mapping_t
*vim
= vd
->vdev_indirect_mapping
;
5886 spa_t
*spa
= vd
->vdev_spa
;
5887 spa_condensing_indirect_phys_t
*scip
=
5888 &spa
->spa_condensing_indirect_phys
;
5889 uint64_t obsolete_sm_object
;
5892 VERIFY0(vdev_obsolete_sm_object(vd
, &obsolete_sm_object
));
5893 EQUIV(obsolete_sm_object
!= 0, vd
->vdev_obsolete_sm
!= NULL
);
5894 counts
= vdev_indirect_mapping_load_obsolete_counts(vim
);
5895 if (vd
->vdev_obsolete_sm
!= NULL
) {
5896 vdev_indirect_mapping_load_obsolete_spacemap(vim
, counts
,
5897 vd
->vdev_obsolete_sm
);
5899 if (scip
->scip_vdev
== vd
->vdev_id
&&
5900 scip
->scip_prev_obsolete_sm_object
!= 0) {
5901 space_map_t
*prev_obsolete_sm
= NULL
;
5902 VERIFY0(space_map_open(&prev_obsolete_sm
, spa
->spa_meta_objset
,
5903 scip
->scip_prev_obsolete_sm_object
, 0, vd
->vdev_asize
, 0));
5904 vdev_indirect_mapping_load_obsolete_spacemap(vim
, counts
,
5906 space_map_close(prev_obsolete_sm
);
5912 zdb_ddt_leak_init(spa_t
*spa
, zdb_cb_t
*zcb
)
5914 ddt_bookmark_t ddb
= {0};
5919 ASSERT(!dump_opt
['L']);
5921 while ((error
= ddt_walk(spa
, &ddb
, &dde
)) == 0) {
5923 ddt_phys_t
*ddp
= dde
.dde_phys
;
5925 if (ddb
.ddb_class
== DDT_CLASS_UNIQUE
)
5928 ASSERT(ddt_phys_total_refcnt(&dde
) > 1);
5930 for (p
= 0; p
< DDT_PHYS_TYPES
; p
++, ddp
++) {
5931 if (ddp
->ddp_phys_birth
== 0)
5933 ddt_bp_create(ddb
.ddb_checksum
,
5934 &dde
.dde_key
, ddp
, &blk
);
5935 if (p
== DDT_PHYS_DITTO
) {
5936 zdb_count_block(zcb
, NULL
, &blk
, ZDB_OT_DITTO
);
5938 zcb
->zcb_dedup_asize
+=
5939 BP_GET_ASIZE(&blk
) * (ddp
->ddp_refcnt
- 1);
5940 zcb
->zcb_dedup_blocks
++;
5943 ddt_t
*ddt
= spa
->spa_ddt
[ddb
.ddb_checksum
];
5945 VERIFY(ddt_lookup(ddt
, &blk
, B_TRUE
) != NULL
);
5949 ASSERT(error
== ENOENT
);
5952 typedef struct checkpoint_sm_exclude_entry_arg
{
5954 uint64_t cseea_checkpoint_size
;
5955 } checkpoint_sm_exclude_entry_arg_t
;
5958 checkpoint_sm_exclude_entry_cb(space_map_entry_t
*sme
, void *arg
)
5960 checkpoint_sm_exclude_entry_arg_t
*cseea
= arg
;
5961 vdev_t
*vd
= cseea
->cseea_vd
;
5962 metaslab_t
*ms
= vd
->vdev_ms
[sme
->sme_offset
>> vd
->vdev_ms_shift
];
5963 uint64_t end
= sme
->sme_offset
+ sme
->sme_run
;
5965 ASSERT(sme
->sme_type
== SM_FREE
);
5968 * Since the vdev_checkpoint_sm exists in the vdev level
5969 * and the ms_sm space maps exist in the metaslab level,
5970 * an entry in the checkpoint space map could theoretically
5971 * cross the boundaries of the metaslab that it belongs.
5973 * In reality, because of the way that we populate and
5974 * manipulate the checkpoint's space maps currently,
5975 * there shouldn't be any entries that cross metaslabs.
5976 * Hence the assertion below.
5978 * That said, there is no fundamental requirement that
5979 * the checkpoint's space map entries should not cross
5980 * metaslab boundaries. So if needed we could add code
5981 * that handles metaslab-crossing segments in the future.
5983 VERIFY3U(sme
->sme_offset
, >=, ms
->ms_start
);
5984 VERIFY3U(end
, <=, ms
->ms_start
+ ms
->ms_size
);
5987 * By removing the entry from the allocated segments we
5988 * also verify that the entry is there to begin with.
5990 mutex_enter(&ms
->ms_lock
);
5991 range_tree_remove(ms
->ms_allocatable
, sme
->sme_offset
, sme
->sme_run
);
5992 mutex_exit(&ms
->ms_lock
);
5994 cseea
->cseea_checkpoint_size
+= sme
->sme_run
;
5999 zdb_leak_init_vdev_exclude_checkpoint(vdev_t
*vd
, zdb_cb_t
*zcb
)
6001 spa_t
*spa
= vd
->vdev_spa
;
6002 space_map_t
*checkpoint_sm
= NULL
;
6003 uint64_t checkpoint_sm_obj
;
6006 * If there is no vdev_top_zap, we are in a pool whose
6007 * version predates the pool checkpoint feature.
6009 if (vd
->vdev_top_zap
== 0)
6013 * If there is no reference of the vdev_checkpoint_sm in
6014 * the vdev_top_zap, then one of the following scenarios
6017 * 1] There is no checkpoint
6018 * 2] There is a checkpoint, but no checkpointed blocks
6019 * have been freed yet
6020 * 3] The current vdev is indirect
6022 * In these cases we return immediately.
6024 if (zap_contains(spa_meta_objset(spa
), vd
->vdev_top_zap
,
6025 VDEV_TOP_ZAP_POOL_CHECKPOINT_SM
) != 0)
6028 VERIFY0(zap_lookup(spa_meta_objset(spa
), vd
->vdev_top_zap
,
6029 VDEV_TOP_ZAP_POOL_CHECKPOINT_SM
, sizeof (uint64_t), 1,
6030 &checkpoint_sm_obj
));
6032 checkpoint_sm_exclude_entry_arg_t cseea
;
6033 cseea
.cseea_vd
= vd
;
6034 cseea
.cseea_checkpoint_size
= 0;
6036 VERIFY0(space_map_open(&checkpoint_sm
, spa_meta_objset(spa
),
6037 checkpoint_sm_obj
, 0, vd
->vdev_asize
, vd
->vdev_ashift
));
6039 VERIFY0(space_map_iterate(checkpoint_sm
,
6040 space_map_length(checkpoint_sm
),
6041 checkpoint_sm_exclude_entry_cb
, &cseea
));
6042 space_map_close(checkpoint_sm
);
6044 zcb
->zcb_checkpoint_size
+= cseea
.cseea_checkpoint_size
;
6048 zdb_leak_init_exclude_checkpoint(spa_t
*spa
, zdb_cb_t
*zcb
)
6050 ASSERT(!dump_opt
['L']);
6052 vdev_t
*rvd
= spa
->spa_root_vdev
;
6053 for (uint64_t c
= 0; c
< rvd
->vdev_children
; c
++) {
6054 ASSERT3U(c
, ==, rvd
->vdev_child
[c
]->vdev_id
);
6055 zdb_leak_init_vdev_exclude_checkpoint(rvd
->vdev_child
[c
], zcb
);
6060 count_unflushed_space_cb(spa_t
*spa
, space_map_entry_t
*sme
,
6061 uint64_t txg
, void *arg
)
6063 int64_t *ualloc_space
= arg
;
6065 uint64_t offset
= sme
->sme_offset
;
6066 uint64_t vdev_id
= sme
->sme_vdev
;
6068 vdev_t
*vd
= vdev_lookup_top(spa
, vdev_id
);
6069 if (!vdev_is_concrete(vd
))
6072 metaslab_t
*ms
= vd
->vdev_ms
[offset
>> vd
->vdev_ms_shift
];
6073 ASSERT(sme
->sme_type
== SM_ALLOC
|| sme
->sme_type
== SM_FREE
);
6075 if (txg
< metaslab_unflushed_txg(ms
))
6078 if (sme
->sme_type
== SM_ALLOC
)
6079 *ualloc_space
+= sme
->sme_run
;
6081 *ualloc_space
-= sme
->sme_run
;
6087 get_unflushed_alloc_space(spa_t
*spa
)
6092 int64_t ualloc_space
= 0;
6093 iterate_through_spacemap_logs(spa
, count_unflushed_space_cb
,
6095 return (ualloc_space
);
6099 load_unflushed_cb(spa_t
*spa
, space_map_entry_t
*sme
, uint64_t txg
, void *arg
)
6101 maptype_t
*uic_maptype
= arg
;
6103 uint64_t offset
= sme
->sme_offset
;
6104 uint64_t size
= sme
->sme_run
;
6105 uint64_t vdev_id
= sme
->sme_vdev
;
6107 vdev_t
*vd
= vdev_lookup_top(spa
, vdev_id
);
6109 /* skip indirect vdevs */
6110 if (!vdev_is_concrete(vd
))
6113 metaslab_t
*ms
= vd
->vdev_ms
[offset
>> vd
->vdev_ms_shift
];
6115 ASSERT(sme
->sme_type
== SM_ALLOC
|| sme
->sme_type
== SM_FREE
);
6116 ASSERT(*uic_maptype
== SM_ALLOC
|| *uic_maptype
== SM_FREE
);
6118 if (txg
< metaslab_unflushed_txg(ms
))
6121 if (*uic_maptype
== sme
->sme_type
)
6122 range_tree_add(ms
->ms_allocatable
, offset
, size
);
6124 range_tree_remove(ms
->ms_allocatable
, offset
, size
);
6130 load_unflushed_to_ms_allocatables(spa_t
*spa
, maptype_t maptype
)
6132 iterate_through_spacemap_logs(spa
, load_unflushed_cb
, &maptype
);
6136 load_concrete_ms_allocatable_trees(spa_t
*spa
, maptype_t maptype
)
6138 vdev_t
*rvd
= spa
->spa_root_vdev
;
6139 for (uint64_t i
= 0; i
< rvd
->vdev_children
; i
++) {
6140 vdev_t
*vd
= rvd
->vdev_child
[i
];
6142 ASSERT3U(i
, ==, vd
->vdev_id
);
6144 if (vd
->vdev_ops
== &vdev_indirect_ops
)
6147 for (uint64_t m
= 0; m
< vd
->vdev_ms_count
; m
++) {
6148 metaslab_t
*msp
= vd
->vdev_ms
[m
];
6150 (void) fprintf(stderr
,
6151 "\rloading concrete vdev %llu, "
6152 "metaslab %llu of %llu ...",
6153 (longlong_t
)vd
->vdev_id
,
6154 (longlong_t
)msp
->ms_id
,
6155 (longlong_t
)vd
->vdev_ms_count
);
6157 mutex_enter(&msp
->ms_lock
);
6158 range_tree_vacate(msp
->ms_allocatable
, NULL
, NULL
);
6161 * We don't want to spend the CPU manipulating the
6162 * size-ordered tree, so clear the range_tree ops.
6164 msp
->ms_allocatable
->rt_ops
= NULL
;
6166 if (msp
->ms_sm
!= NULL
) {
6167 VERIFY0(space_map_load(msp
->ms_sm
,
6168 msp
->ms_allocatable
, maptype
));
6170 if (!msp
->ms_loaded
)
6171 msp
->ms_loaded
= B_TRUE
;
6172 mutex_exit(&msp
->ms_lock
);
6176 load_unflushed_to_ms_allocatables(spa
, maptype
);
6180 * vm_idxp is an in-out parameter which (for indirect vdevs) is the
6181 * index in vim_entries that has the first entry in this metaslab.
6182 * On return, it will be set to the first entry after this metaslab.
6185 load_indirect_ms_allocatable_tree(vdev_t
*vd
, metaslab_t
*msp
,
6188 vdev_indirect_mapping_t
*vim
= vd
->vdev_indirect_mapping
;
6190 mutex_enter(&msp
->ms_lock
);
6191 range_tree_vacate(msp
->ms_allocatable
, NULL
, NULL
);
6194 * We don't want to spend the CPU manipulating the
6195 * size-ordered tree, so clear the range_tree ops.
6197 msp
->ms_allocatable
->rt_ops
= NULL
;
6199 for (; *vim_idxp
< vdev_indirect_mapping_num_entries(vim
);
6201 vdev_indirect_mapping_entry_phys_t
*vimep
=
6202 &vim
->vim_entries
[*vim_idxp
];
6203 uint64_t ent_offset
= DVA_MAPPING_GET_SRC_OFFSET(vimep
);
6204 uint64_t ent_len
= DVA_GET_ASIZE(&vimep
->vimep_dst
);
6205 ASSERT3U(ent_offset
, >=, msp
->ms_start
);
6206 if (ent_offset
>= msp
->ms_start
+ msp
->ms_size
)
6210 * Mappings do not cross metaslab boundaries,
6211 * because we create them by walking the metaslabs.
6213 ASSERT3U(ent_offset
+ ent_len
, <=,
6214 msp
->ms_start
+ msp
->ms_size
);
6215 range_tree_add(msp
->ms_allocatable
, ent_offset
, ent_len
);
6218 if (!msp
->ms_loaded
)
6219 msp
->ms_loaded
= B_TRUE
;
6220 mutex_exit(&msp
->ms_lock
);
6224 zdb_leak_init_prepare_indirect_vdevs(spa_t
*spa
, zdb_cb_t
*zcb
)
6226 ASSERT(!dump_opt
['L']);
6228 vdev_t
*rvd
= spa
->spa_root_vdev
;
6229 for (uint64_t c
= 0; c
< rvd
->vdev_children
; c
++) {
6230 vdev_t
*vd
= rvd
->vdev_child
[c
];
6232 ASSERT3U(c
, ==, vd
->vdev_id
);
6234 if (vd
->vdev_ops
!= &vdev_indirect_ops
)
6238 * Note: we don't check for mapping leaks on
6239 * removing vdevs because their ms_allocatable's
6240 * are used to look for leaks in allocated space.
6242 zcb
->zcb_vd_obsolete_counts
[c
] = zdb_load_obsolete_counts(vd
);
6245 * Normally, indirect vdevs don't have any
6246 * metaslabs. We want to set them up for
6249 vdev_metaslab_group_create(vd
);
6250 VERIFY0(vdev_metaslab_init(vd
, 0));
6252 vdev_indirect_mapping_t
*vim __maybe_unused
=
6253 vd
->vdev_indirect_mapping
;
6254 uint64_t vim_idx
= 0;
6255 for (uint64_t m
= 0; m
< vd
->vdev_ms_count
; m
++) {
6257 (void) fprintf(stderr
,
6258 "\rloading indirect vdev %llu, "
6259 "metaslab %llu of %llu ...",
6260 (longlong_t
)vd
->vdev_id
,
6261 (longlong_t
)vd
->vdev_ms
[m
]->ms_id
,
6262 (longlong_t
)vd
->vdev_ms_count
);
6264 load_indirect_ms_allocatable_tree(vd
, vd
->vdev_ms
[m
],
6267 ASSERT3U(vim_idx
, ==, vdev_indirect_mapping_num_entries(vim
));
6272 zdb_leak_init(spa_t
*spa
, zdb_cb_t
*zcb
)
6279 dsl_pool_t
*dp
= spa
->spa_dsl_pool
;
6280 vdev_t
*rvd
= spa
->spa_root_vdev
;
6283 * We are going to be changing the meaning of the metaslab's
6284 * ms_allocatable. Ensure that the allocator doesn't try to
6287 spa
->spa_normal_class
->mc_ops
= &zdb_metaslab_ops
;
6288 spa
->spa_log_class
->mc_ops
= &zdb_metaslab_ops
;
6289 spa
->spa_embedded_log_class
->mc_ops
= &zdb_metaslab_ops
;
6291 zcb
->zcb_vd_obsolete_counts
=
6292 umem_zalloc(rvd
->vdev_children
* sizeof (uint32_t *),
6296 * For leak detection, we overload the ms_allocatable trees
6297 * to contain allocated segments instead of free segments.
6298 * As a result, we can't use the normal metaslab_load/unload
6301 zdb_leak_init_prepare_indirect_vdevs(spa
, zcb
);
6302 load_concrete_ms_allocatable_trees(spa
, SM_ALLOC
);
6305 * On load_concrete_ms_allocatable_trees() we loaded all the
6306 * allocated entries from the ms_sm to the ms_allocatable for
6307 * each metaslab. If the pool has a checkpoint or is in the
6308 * middle of discarding a checkpoint, some of these blocks
6309 * may have been freed but their ms_sm may not have been
6310 * updated because they are referenced by the checkpoint. In
6311 * order to avoid false-positives during leak-detection, we
6312 * go through the vdev's checkpoint space map and exclude all
6313 * its entries from their relevant ms_allocatable.
6315 * We also aggregate the space held by the checkpoint and add
6316 * it to zcb_checkpoint_size.
6318 * Note that at this point we are also verifying that all the
6319 * entries on the checkpoint_sm are marked as allocated in
6320 * the ms_sm of their relevant metaslab.
6321 * [see comment in checkpoint_sm_exclude_entry_cb()]
6323 zdb_leak_init_exclude_checkpoint(spa
, zcb
);
6324 ASSERT3U(zcb
->zcb_checkpoint_size
, ==, spa_get_checkpoint_space(spa
));
6326 /* for cleaner progress output */
6327 (void) fprintf(stderr
, "\n");
6329 if (bpobj_is_open(&dp
->dp_obsolete_bpobj
)) {
6330 ASSERT(spa_feature_is_enabled(spa
,
6331 SPA_FEATURE_DEVICE_REMOVAL
));
6332 (void) bpobj_iterate_nofree(&dp
->dp_obsolete_bpobj
,
6333 increment_indirect_mapping_cb
, zcb
, NULL
);
6336 spa_config_enter(spa
, SCL_CONFIG
, FTAG
, RW_READER
);
6337 zdb_ddt_leak_init(spa
, zcb
);
6338 spa_config_exit(spa
, SCL_CONFIG
, FTAG
);
6342 zdb_check_for_obsolete_leaks(vdev_t
*vd
, zdb_cb_t
*zcb
)
6344 boolean_t leaks
= B_FALSE
;
6345 vdev_indirect_mapping_t
*vim
= vd
->vdev_indirect_mapping
;
6346 uint64_t total_leaked
= 0;
6347 boolean_t are_precise
= B_FALSE
;
6349 ASSERT(vim
!= NULL
);
6351 for (uint64_t i
= 0; i
< vdev_indirect_mapping_num_entries(vim
); i
++) {
6352 vdev_indirect_mapping_entry_phys_t
*vimep
=
6353 &vim
->vim_entries
[i
];
6354 uint64_t obsolete_bytes
= 0;
6355 uint64_t offset
= DVA_MAPPING_GET_SRC_OFFSET(vimep
);
6356 metaslab_t
*msp
= vd
->vdev_ms
[offset
>> vd
->vdev_ms_shift
];
6359 * This is not very efficient but it's easy to
6360 * verify correctness.
6362 for (uint64_t inner_offset
= 0;
6363 inner_offset
< DVA_GET_ASIZE(&vimep
->vimep_dst
);
6364 inner_offset
+= 1ULL << vd
->vdev_ashift
) {
6365 if (range_tree_contains(msp
->ms_allocatable
,
6366 offset
+ inner_offset
, 1ULL << vd
->vdev_ashift
)) {
6367 obsolete_bytes
+= 1ULL << vd
->vdev_ashift
;
6371 int64_t bytes_leaked
= obsolete_bytes
-
6372 zcb
->zcb_vd_obsolete_counts
[vd
->vdev_id
][i
];
6373 ASSERT3U(DVA_GET_ASIZE(&vimep
->vimep_dst
), >=,
6374 zcb
->zcb_vd_obsolete_counts
[vd
->vdev_id
][i
]);
6376 VERIFY0(vdev_obsolete_counts_are_precise(vd
, &are_precise
));
6377 if (bytes_leaked
!= 0 && (are_precise
|| dump_opt
['d'] >= 5)) {
6378 (void) printf("obsolete indirect mapping count "
6379 "mismatch on %llu:%llx:%llx : %llx bytes leaked\n",
6380 (u_longlong_t
)vd
->vdev_id
,
6381 (u_longlong_t
)DVA_MAPPING_GET_SRC_OFFSET(vimep
),
6382 (u_longlong_t
)DVA_GET_ASIZE(&vimep
->vimep_dst
),
6383 (u_longlong_t
)bytes_leaked
);
6385 total_leaked
+= ABS(bytes_leaked
);
6388 VERIFY0(vdev_obsolete_counts_are_precise(vd
, &are_precise
));
6389 if (!are_precise
&& total_leaked
> 0) {
6390 int pct_leaked
= total_leaked
* 100 /
6391 vdev_indirect_mapping_bytes_mapped(vim
);
6392 (void) printf("cannot verify obsolete indirect mapping "
6393 "counts of vdev %llu because precise feature was not "
6394 "enabled when it was removed: %d%% (%llx bytes) of mapping"
6396 (u_longlong_t
)vd
->vdev_id
, pct_leaked
,
6397 (u_longlong_t
)total_leaked
);
6398 } else if (total_leaked
> 0) {
6399 (void) printf("obsolete indirect mapping count mismatch "
6400 "for vdev %llu -- %llx total bytes mismatched\n",
6401 (u_longlong_t
)vd
->vdev_id
,
6402 (u_longlong_t
)total_leaked
);
6406 vdev_indirect_mapping_free_obsolete_counts(vim
,
6407 zcb
->zcb_vd_obsolete_counts
[vd
->vdev_id
]);
6408 zcb
->zcb_vd_obsolete_counts
[vd
->vdev_id
] = NULL
;
6414 zdb_leak_fini(spa_t
*spa
, zdb_cb_t
*zcb
)
6419 boolean_t leaks
= B_FALSE
;
6420 vdev_t
*rvd
= spa
->spa_root_vdev
;
6421 for (unsigned c
= 0; c
< rvd
->vdev_children
; c
++) {
6422 vdev_t
*vd
= rvd
->vdev_child
[c
];
6424 if (zcb
->zcb_vd_obsolete_counts
[c
] != NULL
) {
6425 leaks
|= zdb_check_for_obsolete_leaks(vd
, zcb
);
6428 for (uint64_t m
= 0; m
< vd
->vdev_ms_count
; m
++) {
6429 metaslab_t
*msp
= vd
->vdev_ms
[m
];
6430 ASSERT3P(msp
->ms_group
, ==, (msp
->ms_group
->mg_class
==
6431 spa_embedded_log_class(spa
)) ?
6432 vd
->vdev_log_mg
: vd
->vdev_mg
);
6435 * ms_allocatable has been overloaded
6436 * to contain allocated segments. Now that
6437 * we finished traversing all blocks, any
6438 * block that remains in the ms_allocatable
6439 * represents an allocated block that we
6440 * did not claim during the traversal.
6441 * Claimed blocks would have been removed
6442 * from the ms_allocatable. For indirect
6443 * vdevs, space remaining in the tree
6444 * represents parts of the mapping that are
6445 * not referenced, which is not a bug.
6447 if (vd
->vdev_ops
== &vdev_indirect_ops
) {
6448 range_tree_vacate(msp
->ms_allocatable
,
6451 range_tree_vacate(msp
->ms_allocatable
,
6454 if (msp
->ms_loaded
) {
6455 msp
->ms_loaded
= B_FALSE
;
6460 umem_free(zcb
->zcb_vd_obsolete_counts
,
6461 rvd
->vdev_children
* sizeof (uint32_t *));
6462 zcb
->zcb_vd_obsolete_counts
= NULL
;
6468 count_block_cb(void *arg
, const blkptr_t
*bp
, dmu_tx_t
*tx
)
6471 zdb_cb_t
*zcb
= arg
;
6473 if (dump_opt
['b'] >= 5) {
6474 char blkbuf
[BP_SPRINTF_LEN
];
6475 snprintf_blkptr(blkbuf
, sizeof (blkbuf
), bp
);
6476 (void) printf("[%s] %s\n",
6477 "deferred free", blkbuf
);
6479 zdb_count_block(zcb
, NULL
, bp
, ZDB_OT_DEFERRED
);
6484 * Iterate over livelists which have been destroyed by the user but
6485 * are still present in the MOS, waiting to be freed
6488 iterate_deleted_livelists(spa_t
*spa
, ll_iter_t func
, void *arg
)
6490 objset_t
*mos
= spa
->spa_meta_objset
;
6492 int err
= zap_lookup(mos
, DMU_POOL_DIRECTORY_OBJECT
,
6493 DMU_POOL_DELETED_CLONES
, sizeof (uint64_t), 1, &zap_obj
);
6499 zap_attribute_t attr
;
6501 /* NULL out os prior to dsl_deadlist_open in case it's garbage */
6503 for (zap_cursor_init(&zc
, mos
, zap_obj
);
6504 zap_cursor_retrieve(&zc
, &attr
) == 0;
6505 (void) zap_cursor_advance(&zc
)) {
6506 dsl_deadlist_open(&ll
, mos
, attr
.za_first_integer
);
6508 dsl_deadlist_close(&ll
);
6510 zap_cursor_fini(&zc
);
6514 bpobj_count_block_cb(void *arg
, const blkptr_t
*bp
, boolean_t bp_freed
,
6518 return (count_block_cb(arg
, bp
, tx
));
6522 livelist_entry_count_blocks_cb(void *args
, dsl_deadlist_entry_t
*dle
)
6524 zdb_cb_t
*zbc
= args
;
6526 bplist_create(&blks
);
6527 /* determine which blocks have been alloc'd but not freed */
6528 VERIFY0(dsl_process_sub_livelist(&dle
->dle_bpobj
, &blks
, NULL
, NULL
));
6529 /* count those blocks */
6530 (void) bplist_iterate(&blks
, count_block_cb
, zbc
, NULL
);
6531 bplist_destroy(&blks
);
6536 livelist_count_blocks(dsl_deadlist_t
*ll
, void *arg
)
6538 dsl_deadlist_iterate(ll
, livelist_entry_count_blocks_cb
, arg
);
6542 * Count the blocks in the livelists that have been destroyed by the user
6543 * but haven't yet been freed.
6546 deleted_livelists_count_blocks(spa_t
*spa
, zdb_cb_t
*zbc
)
6548 iterate_deleted_livelists(spa
, livelist_count_blocks
, zbc
);
6552 dump_livelist_cb(dsl_deadlist_t
*ll
, void *arg
)
6554 ASSERT3P(arg
, ==, NULL
);
6555 global_feature_count
[SPA_FEATURE_LIVELIST
]++;
6556 dump_blkptr_list(ll
, "Deleted Livelist");
6557 dsl_deadlist_iterate(ll
, sublivelist_verify_lightweight
, NULL
);
6561 * Print out, register object references to, and increment feature counts for
6562 * livelists that have been destroyed by the user but haven't yet been freed.
6565 deleted_livelists_dump_mos(spa_t
*spa
)
6568 objset_t
*mos
= spa
->spa_meta_objset
;
6569 int err
= zap_lookup(mos
, DMU_POOL_DIRECTORY_OBJECT
,
6570 DMU_POOL_DELETED_CLONES
, sizeof (uint64_t), 1, &zap_obj
);
6573 mos_obj_refd(zap_obj
);
6574 iterate_deleted_livelists(spa
, dump_livelist_cb
, NULL
);
6578 dump_block_stats(spa_t
*spa
)
6581 zdb_blkstats_t
*zb
, *tzb
;
6582 uint64_t norm_alloc
, norm_space
, total_alloc
, total_found
;
6583 int flags
= TRAVERSE_PRE
| TRAVERSE_PREFETCH_METADATA
|
6584 TRAVERSE_NO_DECRYPT
| TRAVERSE_HARD
;
6585 boolean_t leaks
= B_FALSE
;
6587 bp_embedded_type_t i
;
6589 zcb
= umem_zalloc(sizeof (zdb_cb_t
), UMEM_NOFAIL
);
6591 (void) printf("\nTraversing all blocks %s%s%s%s%s...\n\n",
6592 (dump_opt
['c'] || !dump_opt
['L']) ? "to verify " : "",
6593 (dump_opt
['c'] == 1) ? "metadata " : "",
6594 dump_opt
['c'] ? "checksums " : "",
6595 (dump_opt
['c'] && !dump_opt
['L']) ? "and verify " : "",
6596 !dump_opt
['L'] ? "nothing leaked " : "");
6599 * When leak detection is enabled we load all space maps as SM_ALLOC
6600 * maps, then traverse the pool claiming each block we discover. If
6601 * the pool is perfectly consistent, the segment trees will be empty
6602 * when we're done. Anything left over is a leak; any block we can't
6603 * claim (because it's not part of any space map) is a double
6604 * allocation, reference to a freed block, or an unclaimed log block.
6606 * When leak detection is disabled (-L option) we still traverse the
6607 * pool claiming each block we discover, but we skip opening any space
6610 zdb_leak_init(spa
, zcb
);
6613 * If there's a deferred-free bplist, process that first.
6615 (void) bpobj_iterate_nofree(&spa
->spa_deferred_bpobj
,
6616 bpobj_count_block_cb
, zcb
, NULL
);
6618 if (spa_version(spa
) >= SPA_VERSION_DEADLISTS
) {
6619 (void) bpobj_iterate_nofree(&spa
->spa_dsl_pool
->dp_free_bpobj
,
6620 bpobj_count_block_cb
, zcb
, NULL
);
6623 zdb_claim_removing(spa
, zcb
);
6625 if (spa_feature_is_active(spa
, SPA_FEATURE_ASYNC_DESTROY
)) {
6626 VERIFY3U(0, ==, bptree_iterate(spa
->spa_meta_objset
,
6627 spa
->spa_dsl_pool
->dp_bptree_obj
, B_FALSE
, count_block_cb
,
6631 deleted_livelists_count_blocks(spa
, zcb
);
6633 if (dump_opt
['c'] > 1)
6634 flags
|= TRAVERSE_PREFETCH_DATA
;
6636 zcb
->zcb_totalasize
= metaslab_class_get_alloc(spa_normal_class(spa
));
6637 zcb
->zcb_totalasize
+= metaslab_class_get_alloc(spa_special_class(spa
));
6638 zcb
->zcb_totalasize
+= metaslab_class_get_alloc(spa_dedup_class(spa
));
6639 zcb
->zcb_totalasize
+=
6640 metaslab_class_get_alloc(spa_embedded_log_class(spa
));
6641 zcb
->zcb_start
= zcb
->zcb_lastprint
= gethrtime();
6642 err
= traverse_pool(spa
, 0, flags
, zdb_blkptr_cb
, zcb
);
6645 * If we've traversed the data blocks then we need to wait for those
6646 * I/Os to complete. We leverage "The Godfather" zio to wait on
6647 * all async I/Os to complete.
6649 if (dump_opt
['c']) {
6650 for (c
= 0; c
< max_ncpus
; c
++) {
6651 (void) zio_wait(spa
->spa_async_zio_root
[c
]);
6652 spa
->spa_async_zio_root
[c
] = zio_root(spa
, NULL
, NULL
,
6653 ZIO_FLAG_CANFAIL
| ZIO_FLAG_SPECULATIVE
|
6654 ZIO_FLAG_GODFATHER
);
6657 ASSERT0(spa
->spa_load_verify_bytes
);
6660 * Done after zio_wait() since zcb_haderrors is modified in
6663 zcb
->zcb_haderrors
|= err
;
6665 if (zcb
->zcb_haderrors
) {
6666 (void) printf("\nError counts:\n\n");
6667 (void) printf("\t%5s %s\n", "errno", "count");
6668 for (e
= 0; e
< 256; e
++) {
6669 if (zcb
->zcb_errors
[e
] != 0) {
6670 (void) printf("\t%5d %llu\n",
6671 e
, (u_longlong_t
)zcb
->zcb_errors
[e
]);
6677 * Report any leaked segments.
6679 leaks
|= zdb_leak_fini(spa
, zcb
);
6681 tzb
= &zcb
->zcb_type
[ZB_TOTAL
][ZDB_OT_TOTAL
];
6683 norm_alloc
= metaslab_class_get_alloc(spa_normal_class(spa
));
6684 norm_space
= metaslab_class_get_space(spa_normal_class(spa
));
6686 total_alloc
= norm_alloc
+
6687 metaslab_class_get_alloc(spa_log_class(spa
)) +
6688 metaslab_class_get_alloc(spa_embedded_log_class(spa
)) +
6689 metaslab_class_get_alloc(spa_special_class(spa
)) +
6690 metaslab_class_get_alloc(spa_dedup_class(spa
)) +
6691 get_unflushed_alloc_space(spa
);
6692 total_found
= tzb
->zb_asize
- zcb
->zcb_dedup_asize
+
6693 zcb
->zcb_removing_size
+ zcb
->zcb_checkpoint_size
;
6695 if (total_found
== total_alloc
&& !dump_opt
['L']) {
6696 (void) printf("\n\tNo leaks (block sum matches space"
6697 " maps exactly)\n");
6698 } else if (!dump_opt
['L']) {
6699 (void) printf("block traversal size %llu != alloc %llu "
6701 (u_longlong_t
)total_found
,
6702 (u_longlong_t
)total_alloc
,
6703 (dump_opt
['L']) ? "unreachable" : "leaked",
6704 (longlong_t
)(total_alloc
- total_found
));
6708 if (tzb
->zb_count
== 0) {
6709 umem_free(zcb
, sizeof (zdb_cb_t
));
6713 (void) printf("\n");
6714 (void) printf("\t%-16s %14llu\n", "bp count:",
6715 (u_longlong_t
)tzb
->zb_count
);
6716 (void) printf("\t%-16s %14llu\n", "ganged count:",
6717 (longlong_t
)tzb
->zb_gangs
);
6718 (void) printf("\t%-16s %14llu avg: %6llu\n", "bp logical:",
6719 (u_longlong_t
)tzb
->zb_lsize
,
6720 (u_longlong_t
)(tzb
->zb_lsize
/ tzb
->zb_count
));
6721 (void) printf("\t%-16s %14llu avg: %6llu compression: %6.2f\n",
6722 "bp physical:", (u_longlong_t
)tzb
->zb_psize
,
6723 (u_longlong_t
)(tzb
->zb_psize
/ tzb
->zb_count
),
6724 (double)tzb
->zb_lsize
/ tzb
->zb_psize
);
6725 (void) printf("\t%-16s %14llu avg: %6llu compression: %6.2f\n",
6726 "bp allocated:", (u_longlong_t
)tzb
->zb_asize
,
6727 (u_longlong_t
)(tzb
->zb_asize
/ tzb
->zb_count
),
6728 (double)tzb
->zb_lsize
/ tzb
->zb_asize
);
6729 (void) printf("\t%-16s %14llu ref>1: %6llu deduplication: %6.2f\n",
6730 "bp deduped:", (u_longlong_t
)zcb
->zcb_dedup_asize
,
6731 (u_longlong_t
)zcb
->zcb_dedup_blocks
,
6732 (double)zcb
->zcb_dedup_asize
/ tzb
->zb_asize
+ 1.0);
6733 (void) printf("\t%-16s %14llu used: %5.2f%%\n", "Normal class:",
6734 (u_longlong_t
)norm_alloc
, 100.0 * norm_alloc
/ norm_space
);
6736 if (spa_special_class(spa
)->mc_allocator
[0].mca_rotor
!= NULL
) {
6737 uint64_t alloc
= metaslab_class_get_alloc(
6738 spa_special_class(spa
));
6739 uint64_t space
= metaslab_class_get_space(
6740 spa_special_class(spa
));
6742 (void) printf("\t%-16s %14llu used: %5.2f%%\n",
6743 "Special class", (u_longlong_t
)alloc
,
6744 100.0 * alloc
/ space
);
6747 if (spa_dedup_class(spa
)->mc_allocator
[0].mca_rotor
!= NULL
) {
6748 uint64_t alloc
= metaslab_class_get_alloc(
6749 spa_dedup_class(spa
));
6750 uint64_t space
= metaslab_class_get_space(
6751 spa_dedup_class(spa
));
6753 (void) printf("\t%-16s %14llu used: %5.2f%%\n",
6754 "Dedup class", (u_longlong_t
)alloc
,
6755 100.0 * alloc
/ space
);
6758 if (spa_embedded_log_class(spa
)->mc_allocator
[0].mca_rotor
!= NULL
) {
6759 uint64_t alloc
= metaslab_class_get_alloc(
6760 spa_embedded_log_class(spa
));
6761 uint64_t space
= metaslab_class_get_space(
6762 spa_embedded_log_class(spa
));
6764 (void) printf("\t%-16s %14llu used: %5.2f%%\n",
6765 "Embedded log class", (u_longlong_t
)alloc
,
6766 100.0 * alloc
/ space
);
6769 for (i
= 0; i
< NUM_BP_EMBEDDED_TYPES
; i
++) {
6770 if (zcb
->zcb_embedded_blocks
[i
] == 0)
6772 (void) printf("\n");
6773 (void) printf("\tadditional, non-pointer bps of type %u: "
6775 i
, (u_longlong_t
)zcb
->zcb_embedded_blocks
[i
]);
6777 if (dump_opt
['b'] >= 3) {
6778 (void) printf("\t number of (compressed) bytes: "
6780 dump_histogram(zcb
->zcb_embedded_histogram
[i
],
6781 sizeof (zcb
->zcb_embedded_histogram
[i
]) /
6782 sizeof (zcb
->zcb_embedded_histogram
[i
][0]), 0);
6786 if (tzb
->zb_ditto_samevdev
!= 0) {
6787 (void) printf("\tDittoed blocks on same vdev: %llu\n",
6788 (longlong_t
)tzb
->zb_ditto_samevdev
);
6790 if (tzb
->zb_ditto_same_ms
!= 0) {
6791 (void) printf("\tDittoed blocks in same metaslab: %llu\n",
6792 (longlong_t
)tzb
->zb_ditto_same_ms
);
6795 for (uint64_t v
= 0; v
< spa
->spa_root_vdev
->vdev_children
; v
++) {
6796 vdev_t
*vd
= spa
->spa_root_vdev
->vdev_child
[v
];
6797 vdev_indirect_mapping_t
*vim
= vd
->vdev_indirect_mapping
;
6804 zdb_nicenum(vdev_indirect_mapping_num_entries(vim
),
6805 mem
, vdev_indirect_mapping_size(vim
));
6807 (void) printf("\tindirect vdev id %llu has %llu segments "
6809 (longlong_t
)vd
->vdev_id
,
6810 (longlong_t
)vdev_indirect_mapping_num_entries(vim
), mem
);
6813 if (dump_opt
['b'] >= 2) {
6815 (void) printf("\nBlocks\tLSIZE\tPSIZE\tASIZE"
6816 "\t avg\t comp\t%%Total\tType\n");
6818 for (t
= 0; t
<= ZDB_OT_TOTAL
; t
++) {
6819 char csize
[32], lsize
[32], psize
[32], asize
[32];
6820 char avg
[32], gang
[32];
6821 const char *typename
;
6823 /* make sure nicenum has enough space */
6824 _Static_assert(sizeof (csize
) >= NN_NUMBUF_SZ
,
6826 _Static_assert(sizeof (lsize
) >= NN_NUMBUF_SZ
,
6828 _Static_assert(sizeof (psize
) >= NN_NUMBUF_SZ
,
6830 _Static_assert(sizeof (asize
) >= NN_NUMBUF_SZ
,
6832 _Static_assert(sizeof (avg
) >= NN_NUMBUF_SZ
,
6834 _Static_assert(sizeof (gang
) >= NN_NUMBUF_SZ
,
6837 if (t
< DMU_OT_NUMTYPES
)
6838 typename
= dmu_ot
[t
].ot_name
;
6840 typename
= zdb_ot_extname
[t
- DMU_OT_NUMTYPES
];
6842 if (zcb
->zcb_type
[ZB_TOTAL
][t
].zb_asize
== 0) {
6843 (void) printf("%6s\t%5s\t%5s\t%5s"
6844 "\t%5s\t%5s\t%6s\t%s\n",
6856 for (l
= ZB_TOTAL
- 1; l
>= -1; l
--) {
6857 level
= (l
== -1 ? ZB_TOTAL
: l
);
6858 zb
= &zcb
->zcb_type
[level
][t
];
6860 if (zb
->zb_asize
== 0)
6863 if (dump_opt
['b'] < 3 && level
!= ZB_TOTAL
)
6866 if (level
== 0 && zb
->zb_asize
==
6867 zcb
->zcb_type
[ZB_TOTAL
][t
].zb_asize
)
6870 zdb_nicenum(zb
->zb_count
, csize
,
6872 zdb_nicenum(zb
->zb_lsize
, lsize
,
6874 zdb_nicenum(zb
->zb_psize
, psize
,
6876 zdb_nicenum(zb
->zb_asize
, asize
,
6878 zdb_nicenum(zb
->zb_asize
/ zb
->zb_count
, avg
,
6880 zdb_nicenum(zb
->zb_gangs
, gang
, sizeof (gang
));
6882 (void) printf("%6s\t%5s\t%5s\t%5s\t%5s"
6884 csize
, lsize
, psize
, asize
, avg
,
6885 (double)zb
->zb_lsize
/ zb
->zb_psize
,
6886 100.0 * zb
->zb_asize
/ tzb
->zb_asize
);
6888 if (level
== ZB_TOTAL
)
6889 (void) printf("%s\n", typename
);
6891 (void) printf(" L%d %s\n",
6894 if (dump_opt
['b'] >= 3 && zb
->zb_gangs
> 0) {
6895 (void) printf("\t number of ganged "
6896 "blocks: %s\n", gang
);
6899 if (dump_opt
['b'] >= 4) {
6900 (void) printf("psize "
6901 "(in 512-byte sectors): "
6902 "number of blocks\n");
6903 dump_histogram(zb
->zb_psize_histogram
,
6904 PSIZE_HISTO_SIZE
, 0);
6909 /* Output a table summarizing block sizes in the pool */
6910 if (dump_opt
['b'] >= 2) {
6911 dump_size_histograms(zcb
);
6915 (void) printf("\n");
6918 umem_free(zcb
, sizeof (zdb_cb_t
));
6922 if (zcb
->zcb_haderrors
) {
6923 umem_free(zcb
, sizeof (zdb_cb_t
));
6927 umem_free(zcb
, sizeof (zdb_cb_t
));
6931 typedef struct zdb_ddt_entry
{
6933 uint64_t zdde_ref_blocks
;
6934 uint64_t zdde_ref_lsize
;
6935 uint64_t zdde_ref_psize
;
6936 uint64_t zdde_ref_dsize
;
6937 avl_node_t zdde_node
;
6941 zdb_ddt_add_cb(spa_t
*spa
, zilog_t
*zilog
, const blkptr_t
*bp
,
6942 const zbookmark_phys_t
*zb
, const dnode_phys_t
*dnp
, void *arg
)
6944 (void) zilog
, (void) dnp
;
6945 avl_tree_t
*t
= arg
;
6947 zdb_ddt_entry_t
*zdde
, zdde_search
;
6949 if (zb
->zb_level
== ZB_DNODE_LEVEL
|| BP_IS_HOLE(bp
) ||
6953 if (dump_opt
['S'] > 1 && zb
->zb_level
== ZB_ROOT_LEVEL
) {
6954 (void) printf("traversing objset %llu, %llu objects, "
6955 "%lu blocks so far\n",
6956 (u_longlong_t
)zb
->zb_objset
,
6957 (u_longlong_t
)BP_GET_FILL(bp
),
6961 if (BP_IS_HOLE(bp
) || BP_GET_CHECKSUM(bp
) == ZIO_CHECKSUM_OFF
||
6962 BP_GET_LEVEL(bp
) > 0 || DMU_OT_IS_METADATA(BP_GET_TYPE(bp
)))
6965 ddt_key_fill(&zdde_search
.zdde_key
, bp
);
6967 zdde
= avl_find(t
, &zdde_search
, &where
);
6970 zdde
= umem_zalloc(sizeof (*zdde
), UMEM_NOFAIL
);
6971 zdde
->zdde_key
= zdde_search
.zdde_key
;
6972 avl_insert(t
, zdde
, where
);
6975 zdde
->zdde_ref_blocks
+= 1;
6976 zdde
->zdde_ref_lsize
+= BP_GET_LSIZE(bp
);
6977 zdde
->zdde_ref_psize
+= BP_GET_PSIZE(bp
);
6978 zdde
->zdde_ref_dsize
+= bp_get_dsize_sync(spa
, bp
);
6984 dump_simulated_ddt(spa_t
*spa
)
6987 void *cookie
= NULL
;
6988 zdb_ddt_entry_t
*zdde
;
6989 ddt_histogram_t ddh_total
= {{{0}}};
6990 ddt_stat_t dds_total
= {0};
6992 avl_create(&t
, ddt_entry_compare
,
6993 sizeof (zdb_ddt_entry_t
), offsetof(zdb_ddt_entry_t
, zdde_node
));
6995 spa_config_enter(spa
, SCL_CONFIG
, FTAG
, RW_READER
);
6997 (void) traverse_pool(spa
, 0, TRAVERSE_PRE
| TRAVERSE_PREFETCH_METADATA
|
6998 TRAVERSE_NO_DECRYPT
, zdb_ddt_add_cb
, &t
);
7000 spa_config_exit(spa
, SCL_CONFIG
, FTAG
);
7002 while ((zdde
= avl_destroy_nodes(&t
, &cookie
)) != NULL
) {
7004 uint64_t refcnt
= zdde
->zdde_ref_blocks
;
7005 ASSERT(refcnt
!= 0);
7007 dds
.dds_blocks
= zdde
->zdde_ref_blocks
/ refcnt
;
7008 dds
.dds_lsize
= zdde
->zdde_ref_lsize
/ refcnt
;
7009 dds
.dds_psize
= zdde
->zdde_ref_psize
/ refcnt
;
7010 dds
.dds_dsize
= zdde
->zdde_ref_dsize
/ refcnt
;
7012 dds
.dds_ref_blocks
= zdde
->zdde_ref_blocks
;
7013 dds
.dds_ref_lsize
= zdde
->zdde_ref_lsize
;
7014 dds
.dds_ref_psize
= zdde
->zdde_ref_psize
;
7015 dds
.dds_ref_dsize
= zdde
->zdde_ref_dsize
;
7017 ddt_stat_add(&ddh_total
.ddh_stat
[highbit64(refcnt
) - 1],
7020 umem_free(zdde
, sizeof (*zdde
));
7025 ddt_histogram_stat(&dds_total
, &ddh_total
);
7027 (void) printf("Simulated DDT histogram:\n");
7029 zpool_dump_ddt(&dds_total
, &ddh_total
);
7031 dump_dedup_ratio(&dds_total
);
7035 verify_device_removal_feature_counts(spa_t
*spa
)
7037 uint64_t dr_feature_refcount
= 0;
7038 uint64_t oc_feature_refcount
= 0;
7039 uint64_t indirect_vdev_count
= 0;
7040 uint64_t precise_vdev_count
= 0;
7041 uint64_t obsolete_counts_object_count
= 0;
7042 uint64_t obsolete_sm_count
= 0;
7043 uint64_t obsolete_counts_count
= 0;
7044 uint64_t scip_count
= 0;
7045 uint64_t obsolete_bpobj_count
= 0;
7048 spa_condensing_indirect_phys_t
*scip
=
7049 &spa
->spa_condensing_indirect_phys
;
7050 if (scip
->scip_next_mapping_object
!= 0) {
7051 vdev_t
*vd
= spa
->spa_root_vdev
->vdev_child
[scip
->scip_vdev
];
7052 ASSERT(scip
->scip_prev_obsolete_sm_object
!= 0);
7053 ASSERT3P(vd
->vdev_ops
, ==, &vdev_indirect_ops
);
7055 (void) printf("Condensing indirect vdev %llu: new mapping "
7056 "object %llu, prev obsolete sm %llu\n",
7057 (u_longlong_t
)scip
->scip_vdev
,
7058 (u_longlong_t
)scip
->scip_next_mapping_object
,
7059 (u_longlong_t
)scip
->scip_prev_obsolete_sm_object
);
7060 if (scip
->scip_prev_obsolete_sm_object
!= 0) {
7061 space_map_t
*prev_obsolete_sm
= NULL
;
7062 VERIFY0(space_map_open(&prev_obsolete_sm
,
7063 spa
->spa_meta_objset
,
7064 scip
->scip_prev_obsolete_sm_object
,
7065 0, vd
->vdev_asize
, 0));
7066 dump_spacemap(spa
->spa_meta_objset
, prev_obsolete_sm
);
7067 (void) printf("\n");
7068 space_map_close(prev_obsolete_sm
);
7074 for (uint64_t i
= 0; i
< spa
->spa_root_vdev
->vdev_children
; i
++) {
7075 vdev_t
*vd
= spa
->spa_root_vdev
->vdev_child
[i
];
7076 vdev_indirect_config_t
*vic
= &vd
->vdev_indirect_config
;
7078 if (vic
->vic_mapping_object
!= 0) {
7079 ASSERT(vd
->vdev_ops
== &vdev_indirect_ops
||
7081 indirect_vdev_count
++;
7083 if (vd
->vdev_indirect_mapping
->vim_havecounts
) {
7084 obsolete_counts_count
++;
7088 boolean_t are_precise
;
7089 VERIFY0(vdev_obsolete_counts_are_precise(vd
, &are_precise
));
7091 ASSERT(vic
->vic_mapping_object
!= 0);
7092 precise_vdev_count
++;
7095 uint64_t obsolete_sm_object
;
7096 VERIFY0(vdev_obsolete_sm_object(vd
, &obsolete_sm_object
));
7097 if (obsolete_sm_object
!= 0) {
7098 ASSERT(vic
->vic_mapping_object
!= 0);
7099 obsolete_sm_count
++;
7103 (void) feature_get_refcount(spa
,
7104 &spa_feature_table
[SPA_FEATURE_DEVICE_REMOVAL
],
7105 &dr_feature_refcount
);
7106 (void) feature_get_refcount(spa
,
7107 &spa_feature_table
[SPA_FEATURE_OBSOLETE_COUNTS
],
7108 &oc_feature_refcount
);
7110 if (dr_feature_refcount
!= indirect_vdev_count
) {
7112 (void) printf("Number of indirect vdevs (%llu) " \
7113 "does not match feature count (%llu)\n",
7114 (u_longlong_t
)indirect_vdev_count
,
7115 (u_longlong_t
)dr_feature_refcount
);
7117 (void) printf("Verified device_removal feature refcount " \
7118 "of %llu is correct\n",
7119 (u_longlong_t
)dr_feature_refcount
);
7122 if (zap_contains(spa_meta_objset(spa
), DMU_POOL_DIRECTORY_OBJECT
,
7123 DMU_POOL_OBSOLETE_BPOBJ
) == 0) {
7124 obsolete_bpobj_count
++;
7128 obsolete_counts_object_count
= precise_vdev_count
;
7129 obsolete_counts_object_count
+= obsolete_sm_count
;
7130 obsolete_counts_object_count
+= obsolete_counts_count
;
7131 obsolete_counts_object_count
+= scip_count
;
7132 obsolete_counts_object_count
+= obsolete_bpobj_count
;
7133 obsolete_counts_object_count
+= remap_deadlist_count
;
7135 if (oc_feature_refcount
!= obsolete_counts_object_count
) {
7137 (void) printf("Number of obsolete counts objects (%llu) " \
7138 "does not match feature count (%llu)\n",
7139 (u_longlong_t
)obsolete_counts_object_count
,
7140 (u_longlong_t
)oc_feature_refcount
);
7141 (void) printf("pv:%llu os:%llu oc:%llu sc:%llu "
7142 "ob:%llu rd:%llu\n",
7143 (u_longlong_t
)precise_vdev_count
,
7144 (u_longlong_t
)obsolete_sm_count
,
7145 (u_longlong_t
)obsolete_counts_count
,
7146 (u_longlong_t
)scip_count
,
7147 (u_longlong_t
)obsolete_bpobj_count
,
7148 (u_longlong_t
)remap_deadlist_count
);
7150 (void) printf("Verified indirect_refcount feature refcount " \
7151 "of %llu is correct\n",
7152 (u_longlong_t
)oc_feature_refcount
);
7158 zdb_set_skip_mmp(char *target
)
7163 * Disable the activity check to allow examination of
7166 mutex_enter(&spa_namespace_lock
);
7167 if ((spa
= spa_lookup(target
)) != NULL
) {
7168 spa
->spa_import_flags
|= ZFS_IMPORT_SKIP_MMP
;
7170 mutex_exit(&spa_namespace_lock
);
7173 #define BOGUS_SUFFIX "_CHECKPOINTED_UNIVERSE"
7175 * Import the checkpointed state of the pool specified by the target
7176 * parameter as readonly. The function also accepts a pool config
7177 * as an optional parameter, else it attempts to infer the config by
7178 * the name of the target pool.
7180 * Note that the checkpointed state's pool name will be the name of
7181 * the original pool with the above suffix appended to it. In addition,
7182 * if the target is not a pool name (e.g. a path to a dataset) then
7183 * the new_path parameter is populated with the updated path to
7184 * reflect the fact that we are looking into the checkpointed state.
7186 * The function returns a newly-allocated copy of the name of the
7187 * pool containing the checkpointed state. When this copy is no
7188 * longer needed it should be freed with free(3C). Same thing
7189 * applies to the new_path parameter if allocated.
7192 import_checkpointed_state(char *target
, nvlist_t
*cfg
, char **new_path
)
7195 char *poolname
, *bogus_name
= NULL
;
7196 boolean_t freecfg
= B_FALSE
;
7198 /* If the target is not a pool, the extract the pool name */
7199 char *path_start
= strchr(target
, '/');
7200 if (path_start
!= NULL
) {
7201 size_t poolname_len
= path_start
- target
;
7202 poolname
= strndup(target
, poolname_len
);
7208 zdb_set_skip_mmp(poolname
);
7209 error
= spa_get_stats(poolname
, &cfg
, NULL
, 0);
7211 fatal("Tried to read config of pool \"%s\" but "
7212 "spa_get_stats() failed with error %d\n",
7218 if (asprintf(&bogus_name
, "%s%s", poolname
, BOGUS_SUFFIX
) == -1) {
7219 if (target
!= poolname
)
7223 fnvlist_add_string(cfg
, ZPOOL_CONFIG_POOL_NAME
, bogus_name
);
7225 error
= spa_import(bogus_name
, cfg
, NULL
,
7226 ZFS_IMPORT_MISSING_LOG
| ZFS_IMPORT_CHECKPOINT
|
7227 ZFS_IMPORT_SKIP_MMP
);
7231 fatal("Tried to import pool \"%s\" but spa_import() failed "
7232 "with error %d\n", bogus_name
, error
);
7235 if (new_path
!= NULL
&& path_start
!= NULL
) {
7236 if (asprintf(new_path
, "%s%s", bogus_name
, path_start
) == -1) {
7238 if (path_start
!= NULL
)
7244 if (target
!= poolname
)
7247 return (bogus_name
);
7250 typedef struct verify_checkpoint_sm_entry_cb_arg
{
7253 /* the following fields are only used for printing progress */
7254 uint64_t vcsec_entryid
;
7255 uint64_t vcsec_num_entries
;
7256 } verify_checkpoint_sm_entry_cb_arg_t
;
7258 #define ENTRIES_PER_PROGRESS_UPDATE 10000
7261 verify_checkpoint_sm_entry_cb(space_map_entry_t
*sme
, void *arg
)
7263 verify_checkpoint_sm_entry_cb_arg_t
*vcsec
= arg
;
7264 vdev_t
*vd
= vcsec
->vcsec_vd
;
7265 metaslab_t
*ms
= vd
->vdev_ms
[sme
->sme_offset
>> vd
->vdev_ms_shift
];
7266 uint64_t end
= sme
->sme_offset
+ sme
->sme_run
;
7268 ASSERT(sme
->sme_type
== SM_FREE
);
7270 if ((vcsec
->vcsec_entryid
% ENTRIES_PER_PROGRESS_UPDATE
) == 0) {
7271 (void) fprintf(stderr
,
7272 "\rverifying vdev %llu, space map entry %llu of %llu ...",
7273 (longlong_t
)vd
->vdev_id
,
7274 (longlong_t
)vcsec
->vcsec_entryid
,
7275 (longlong_t
)vcsec
->vcsec_num_entries
);
7277 vcsec
->vcsec_entryid
++;
7280 * See comment in checkpoint_sm_exclude_entry_cb()
7282 VERIFY3U(sme
->sme_offset
, >=, ms
->ms_start
);
7283 VERIFY3U(end
, <=, ms
->ms_start
+ ms
->ms_size
);
7286 * The entries in the vdev_checkpoint_sm should be marked as
7287 * allocated in the checkpointed state of the pool, therefore
7288 * their respective ms_allocateable trees should not contain them.
7290 mutex_enter(&ms
->ms_lock
);
7291 range_tree_verify_not_present(ms
->ms_allocatable
,
7292 sme
->sme_offset
, sme
->sme_run
);
7293 mutex_exit(&ms
->ms_lock
);
7299 * Verify that all segments in the vdev_checkpoint_sm are allocated
7300 * according to the checkpoint's ms_sm (i.e. are not in the checkpoint's
7303 * Do so by comparing the checkpoint space maps (vdev_checkpoint_sm) of
7304 * each vdev in the current state of the pool to the metaslab space maps
7305 * (ms_sm) of the checkpointed state of the pool.
7307 * Note that the function changes the state of the ms_allocatable
7308 * trees of the current spa_t. The entries of these ms_allocatable
7309 * trees are cleared out and then repopulated from with the free
7310 * entries of their respective ms_sm space maps.
7313 verify_checkpoint_vdev_spacemaps(spa_t
*checkpoint
, spa_t
*current
)
7315 vdev_t
*ckpoint_rvd
= checkpoint
->spa_root_vdev
;
7316 vdev_t
*current_rvd
= current
->spa_root_vdev
;
7318 load_concrete_ms_allocatable_trees(checkpoint
, SM_FREE
);
7320 for (uint64_t c
= 0; c
< ckpoint_rvd
->vdev_children
; c
++) {
7321 vdev_t
*ckpoint_vd
= ckpoint_rvd
->vdev_child
[c
];
7322 vdev_t
*current_vd
= current_rvd
->vdev_child
[c
];
7324 space_map_t
*checkpoint_sm
= NULL
;
7325 uint64_t checkpoint_sm_obj
;
7327 if (ckpoint_vd
->vdev_ops
== &vdev_indirect_ops
) {
7329 * Since we don't allow device removal in a pool
7330 * that has a checkpoint, we expect that all removed
7331 * vdevs were removed from the pool before the
7334 ASSERT3P(current_vd
->vdev_ops
, ==, &vdev_indirect_ops
);
7339 * If the checkpoint space map doesn't exist, then nothing
7340 * here is checkpointed so there's nothing to verify.
7342 if (current_vd
->vdev_top_zap
== 0 ||
7343 zap_contains(spa_meta_objset(current
),
7344 current_vd
->vdev_top_zap
,
7345 VDEV_TOP_ZAP_POOL_CHECKPOINT_SM
) != 0)
7348 VERIFY0(zap_lookup(spa_meta_objset(current
),
7349 current_vd
->vdev_top_zap
, VDEV_TOP_ZAP_POOL_CHECKPOINT_SM
,
7350 sizeof (uint64_t), 1, &checkpoint_sm_obj
));
7352 VERIFY0(space_map_open(&checkpoint_sm
, spa_meta_objset(current
),
7353 checkpoint_sm_obj
, 0, current_vd
->vdev_asize
,
7354 current_vd
->vdev_ashift
));
7356 verify_checkpoint_sm_entry_cb_arg_t vcsec
;
7357 vcsec
.vcsec_vd
= ckpoint_vd
;
7358 vcsec
.vcsec_entryid
= 0;
7359 vcsec
.vcsec_num_entries
=
7360 space_map_length(checkpoint_sm
) / sizeof (uint64_t);
7361 VERIFY0(space_map_iterate(checkpoint_sm
,
7362 space_map_length(checkpoint_sm
),
7363 verify_checkpoint_sm_entry_cb
, &vcsec
));
7364 if (dump_opt
['m'] > 3)
7365 dump_spacemap(current
->spa_meta_objset
, checkpoint_sm
);
7366 space_map_close(checkpoint_sm
);
7370 * If we've added vdevs since we took the checkpoint, ensure
7371 * that their checkpoint space maps are empty.
7373 if (ckpoint_rvd
->vdev_children
< current_rvd
->vdev_children
) {
7374 for (uint64_t c
= ckpoint_rvd
->vdev_children
;
7375 c
< current_rvd
->vdev_children
; c
++) {
7376 vdev_t
*current_vd
= current_rvd
->vdev_child
[c
];
7377 VERIFY3P(current_vd
->vdev_checkpoint_sm
, ==, NULL
);
7381 /* for cleaner progress output */
7382 (void) fprintf(stderr
, "\n");
7386 * Verifies that all space that's allocated in the checkpoint is
7387 * still allocated in the current version, by checking that everything
7388 * in checkpoint's ms_allocatable (which is actually allocated, not
7389 * allocatable/free) is not present in current's ms_allocatable.
7391 * Note that the function changes the state of the ms_allocatable
7392 * trees of both spas when called. The entries of all ms_allocatable
7393 * trees are cleared out and then repopulated from their respective
7394 * ms_sm space maps. In the checkpointed state we load the allocated
7395 * entries, and in the current state we load the free entries.
7398 verify_checkpoint_ms_spacemaps(spa_t
*checkpoint
, spa_t
*current
)
7400 vdev_t
*ckpoint_rvd
= checkpoint
->spa_root_vdev
;
7401 vdev_t
*current_rvd
= current
->spa_root_vdev
;
7403 load_concrete_ms_allocatable_trees(checkpoint
, SM_ALLOC
);
7404 load_concrete_ms_allocatable_trees(current
, SM_FREE
);
7406 for (uint64_t i
= 0; i
< ckpoint_rvd
->vdev_children
; i
++) {
7407 vdev_t
*ckpoint_vd
= ckpoint_rvd
->vdev_child
[i
];
7408 vdev_t
*current_vd
= current_rvd
->vdev_child
[i
];
7410 if (ckpoint_vd
->vdev_ops
== &vdev_indirect_ops
) {
7412 * See comment in verify_checkpoint_vdev_spacemaps()
7414 ASSERT3P(current_vd
->vdev_ops
, ==, &vdev_indirect_ops
);
7418 for (uint64_t m
= 0; m
< ckpoint_vd
->vdev_ms_count
; m
++) {
7419 metaslab_t
*ckpoint_msp
= ckpoint_vd
->vdev_ms
[m
];
7420 metaslab_t
*current_msp
= current_vd
->vdev_ms
[m
];
7422 (void) fprintf(stderr
,
7423 "\rverifying vdev %llu of %llu, "
7424 "metaslab %llu of %llu ...",
7425 (longlong_t
)current_vd
->vdev_id
,
7426 (longlong_t
)current_rvd
->vdev_children
,
7427 (longlong_t
)current_vd
->vdev_ms
[m
]->ms_id
,
7428 (longlong_t
)current_vd
->vdev_ms_count
);
7431 * We walk through the ms_allocatable trees that
7432 * are loaded with the allocated blocks from the
7433 * ms_sm spacemaps of the checkpoint. For each
7434 * one of these ranges we ensure that none of them
7435 * exists in the ms_allocatable trees of the
7436 * current state which are loaded with the ranges
7437 * that are currently free.
7439 * This way we ensure that none of the blocks that
7440 * are part of the checkpoint were freed by mistake.
7442 range_tree_walk(ckpoint_msp
->ms_allocatable
,
7443 (range_tree_func_t
*)range_tree_verify_not_present
,
7444 current_msp
->ms_allocatable
);
7448 /* for cleaner progress output */
7449 (void) fprintf(stderr
, "\n");
7453 verify_checkpoint_blocks(spa_t
*spa
)
7455 ASSERT(!dump_opt
['L']);
7457 spa_t
*checkpoint_spa
;
7458 char *checkpoint_pool
;
7462 * We import the checkpointed state of the pool (under a different
7463 * name) so we can do verification on it against the current state
7466 checkpoint_pool
= import_checkpointed_state(spa
->spa_name
, NULL
,
7468 ASSERT(strcmp(spa
->spa_name
, checkpoint_pool
) != 0);
7470 error
= spa_open(checkpoint_pool
, &checkpoint_spa
, FTAG
);
7472 fatal("Tried to open pool \"%s\" but spa_open() failed with "
7473 "error %d\n", checkpoint_pool
, error
);
7477 * Ensure that ranges in the checkpoint space maps of each vdev
7478 * are allocated according to the checkpointed state's metaslab
7481 verify_checkpoint_vdev_spacemaps(checkpoint_spa
, spa
);
7484 * Ensure that allocated ranges in the checkpoint's metaslab
7485 * space maps remain allocated in the metaslab space maps of
7486 * the current state.
7488 verify_checkpoint_ms_spacemaps(checkpoint_spa
, spa
);
7491 * Once we are done, we get rid of the checkpointed state.
7493 spa_close(checkpoint_spa
, FTAG
);
7494 free(checkpoint_pool
);
7498 dump_leftover_checkpoint_blocks(spa_t
*spa
)
7500 vdev_t
*rvd
= spa
->spa_root_vdev
;
7502 for (uint64_t i
= 0; i
< rvd
->vdev_children
; i
++) {
7503 vdev_t
*vd
= rvd
->vdev_child
[i
];
7505 space_map_t
*checkpoint_sm
= NULL
;
7506 uint64_t checkpoint_sm_obj
;
7508 if (vd
->vdev_top_zap
== 0)
7511 if (zap_contains(spa_meta_objset(spa
), vd
->vdev_top_zap
,
7512 VDEV_TOP_ZAP_POOL_CHECKPOINT_SM
) != 0)
7515 VERIFY0(zap_lookup(spa_meta_objset(spa
), vd
->vdev_top_zap
,
7516 VDEV_TOP_ZAP_POOL_CHECKPOINT_SM
,
7517 sizeof (uint64_t), 1, &checkpoint_sm_obj
));
7519 VERIFY0(space_map_open(&checkpoint_sm
, spa_meta_objset(spa
),
7520 checkpoint_sm_obj
, 0, vd
->vdev_asize
, vd
->vdev_ashift
));
7521 dump_spacemap(spa
->spa_meta_objset
, checkpoint_sm
);
7522 space_map_close(checkpoint_sm
);
7527 verify_checkpoint(spa_t
*spa
)
7529 uberblock_t checkpoint
;
7532 if (!spa_feature_is_active(spa
, SPA_FEATURE_POOL_CHECKPOINT
))
7535 error
= zap_lookup(spa
->spa_meta_objset
, DMU_POOL_DIRECTORY_OBJECT
,
7536 DMU_POOL_ZPOOL_CHECKPOINT
, sizeof (uint64_t),
7537 sizeof (uberblock_t
) / sizeof (uint64_t), &checkpoint
);
7539 if (error
== ENOENT
&& !dump_opt
['L']) {
7541 * If the feature is active but the uberblock is missing
7542 * then we must be in the middle of discarding the
7545 (void) printf("\nPartially discarded checkpoint "
7547 if (dump_opt
['m'] > 3)
7548 dump_leftover_checkpoint_blocks(spa
);
7550 } else if (error
!= 0) {
7551 (void) printf("lookup error %d when looking for "
7552 "checkpointed uberblock in MOS\n", error
);
7555 dump_uberblock(&checkpoint
, "\nCheckpointed uberblock found:\n", "\n");
7557 if (checkpoint
.ub_checkpoint_txg
== 0) {
7558 (void) printf("\nub_checkpoint_txg not set in checkpointed "
7563 if (error
== 0 && !dump_opt
['L'])
7564 verify_checkpoint_blocks(spa
);
7570 mos_leaks_cb(void *arg
, uint64_t start
, uint64_t size
)
7573 for (uint64_t i
= start
; i
< size
; i
++) {
7574 (void) printf("MOS object %llu referenced but not allocated\n",
7580 mos_obj_refd(uint64_t obj
)
7582 if (obj
!= 0 && mos_refd_objs
!= NULL
)
7583 range_tree_add(mos_refd_objs
, obj
, 1);
7587 * Call on a MOS object that may already have been referenced.
7590 mos_obj_refd_multiple(uint64_t obj
)
7592 if (obj
!= 0 && mos_refd_objs
!= NULL
&&
7593 !range_tree_contains(mos_refd_objs
, obj
, 1))
7594 range_tree_add(mos_refd_objs
, obj
, 1);
7598 mos_leak_vdev_top_zap(vdev_t
*vd
)
7600 uint64_t ms_flush_data_obj
;
7601 int error
= zap_lookup(spa_meta_objset(vd
->vdev_spa
),
7602 vd
->vdev_top_zap
, VDEV_TOP_ZAP_MS_UNFLUSHED_PHYS_TXGS
,
7603 sizeof (ms_flush_data_obj
), 1, &ms_flush_data_obj
);
7604 if (error
== ENOENT
)
7608 mos_obj_refd(ms_flush_data_obj
);
7612 mos_leak_vdev(vdev_t
*vd
)
7614 mos_obj_refd(vd
->vdev_dtl_object
);
7615 mos_obj_refd(vd
->vdev_ms_array
);
7616 mos_obj_refd(vd
->vdev_indirect_config
.vic_births_object
);
7617 mos_obj_refd(vd
->vdev_indirect_config
.vic_mapping_object
);
7618 mos_obj_refd(vd
->vdev_leaf_zap
);
7619 if (vd
->vdev_checkpoint_sm
!= NULL
)
7620 mos_obj_refd(vd
->vdev_checkpoint_sm
->sm_object
);
7621 if (vd
->vdev_indirect_mapping
!= NULL
) {
7622 mos_obj_refd(vd
->vdev_indirect_mapping
->
7623 vim_phys
->vimp_counts_object
);
7625 if (vd
->vdev_obsolete_sm
!= NULL
)
7626 mos_obj_refd(vd
->vdev_obsolete_sm
->sm_object
);
7628 for (uint64_t m
= 0; m
< vd
->vdev_ms_count
; m
++) {
7629 metaslab_t
*ms
= vd
->vdev_ms
[m
];
7630 mos_obj_refd(space_map_object(ms
->ms_sm
));
7633 if (vd
->vdev_root_zap
!= 0)
7634 mos_obj_refd(vd
->vdev_root_zap
);
7636 if (vd
->vdev_top_zap
!= 0) {
7637 mos_obj_refd(vd
->vdev_top_zap
);
7638 mos_leak_vdev_top_zap(vd
);
7641 for (uint64_t c
= 0; c
< vd
->vdev_children
; c
++) {
7642 mos_leak_vdev(vd
->vdev_child
[c
]);
7647 mos_leak_log_spacemaps(spa_t
*spa
)
7649 uint64_t spacemap_zap
;
7650 int error
= zap_lookup(spa_meta_objset(spa
),
7651 DMU_POOL_DIRECTORY_OBJECT
, DMU_POOL_LOG_SPACEMAP_ZAP
,
7652 sizeof (spacemap_zap
), 1, &spacemap_zap
);
7653 if (error
== ENOENT
)
7657 mos_obj_refd(spacemap_zap
);
7658 for (spa_log_sm_t
*sls
= avl_first(&spa
->spa_sm_logs_by_txg
);
7659 sls
; sls
= AVL_NEXT(&spa
->spa_sm_logs_by_txg
, sls
))
7660 mos_obj_refd(sls
->sls_sm_obj
);
7664 errorlog_count_refd(objset_t
*mos
, uint64_t errlog
)
7668 for (zap_cursor_init(&zc
, mos
, errlog
);
7669 zap_cursor_retrieve(&zc
, &za
) == 0;
7670 zap_cursor_advance(&zc
)) {
7671 mos_obj_refd(za
.za_first_integer
);
7673 zap_cursor_fini(&zc
);
7677 dump_mos_leaks(spa_t
*spa
)
7680 objset_t
*mos
= spa
->spa_meta_objset
;
7681 dsl_pool_t
*dp
= spa
->spa_dsl_pool
;
7683 /* Visit and mark all referenced objects in the MOS */
7685 mos_obj_refd(DMU_POOL_DIRECTORY_OBJECT
);
7686 mos_obj_refd(spa
->spa_pool_props_object
);
7687 mos_obj_refd(spa
->spa_config_object
);
7688 mos_obj_refd(spa
->spa_ddt_stat_object
);
7689 mos_obj_refd(spa
->spa_feat_desc_obj
);
7690 mos_obj_refd(spa
->spa_feat_enabled_txg_obj
);
7691 mos_obj_refd(spa
->spa_feat_for_read_obj
);
7692 mos_obj_refd(spa
->spa_feat_for_write_obj
);
7693 mos_obj_refd(spa
->spa_history
);
7694 mos_obj_refd(spa
->spa_errlog_last
);
7695 mos_obj_refd(spa
->spa_errlog_scrub
);
7697 if (spa_feature_is_enabled(spa
, SPA_FEATURE_HEAD_ERRLOG
)) {
7698 errorlog_count_refd(mos
, spa
->spa_errlog_last
);
7699 errorlog_count_refd(mos
, spa
->spa_errlog_scrub
);
7702 mos_obj_refd(spa
->spa_all_vdev_zaps
);
7703 mos_obj_refd(spa
->spa_dsl_pool
->dp_bptree_obj
);
7704 mos_obj_refd(spa
->spa_dsl_pool
->dp_tmp_userrefs_obj
);
7705 mos_obj_refd(spa
->spa_dsl_pool
->dp_scan
->scn_phys
.scn_queue_obj
);
7706 bpobj_count_refd(&spa
->spa_deferred_bpobj
);
7707 mos_obj_refd(dp
->dp_empty_bpobj
);
7708 bpobj_count_refd(&dp
->dp_obsolete_bpobj
);
7709 bpobj_count_refd(&dp
->dp_free_bpobj
);
7710 mos_obj_refd(spa
->spa_l2cache
.sav_object
);
7711 mos_obj_refd(spa
->spa_spares
.sav_object
);
7713 if (spa
->spa_syncing_log_sm
!= NULL
)
7714 mos_obj_refd(spa
->spa_syncing_log_sm
->sm_object
);
7715 mos_leak_log_spacemaps(spa
);
7717 mos_obj_refd(spa
->spa_condensing_indirect_phys
.
7718 scip_next_mapping_object
);
7719 mos_obj_refd(spa
->spa_condensing_indirect_phys
.
7720 scip_prev_obsolete_sm_object
);
7721 if (spa
->spa_condensing_indirect_phys
.scip_next_mapping_object
!= 0) {
7722 vdev_indirect_mapping_t
*vim
=
7723 vdev_indirect_mapping_open(mos
,
7724 spa
->spa_condensing_indirect_phys
.scip_next_mapping_object
);
7725 mos_obj_refd(vim
->vim_phys
->vimp_counts_object
);
7726 vdev_indirect_mapping_close(vim
);
7728 deleted_livelists_dump_mos(spa
);
7730 if (dp
->dp_origin_snap
!= NULL
) {
7733 dsl_pool_config_enter(dp
, FTAG
);
7734 VERIFY0(dsl_dataset_hold_obj(dp
,
7735 dsl_dataset_phys(dp
->dp_origin_snap
)->ds_next_snap_obj
,
7737 count_ds_mos_objects(ds
);
7738 dump_blkptr_list(&ds
->ds_deadlist
, "Deadlist");
7739 dsl_dataset_rele(ds
, FTAG
);
7740 dsl_pool_config_exit(dp
, FTAG
);
7742 count_ds_mos_objects(dp
->dp_origin_snap
);
7743 dump_blkptr_list(&dp
->dp_origin_snap
->ds_deadlist
, "Deadlist");
7745 count_dir_mos_objects(dp
->dp_mos_dir
);
7746 if (dp
->dp_free_dir
!= NULL
)
7747 count_dir_mos_objects(dp
->dp_free_dir
);
7748 if (dp
->dp_leak_dir
!= NULL
)
7749 count_dir_mos_objects(dp
->dp_leak_dir
);
7751 mos_leak_vdev(spa
->spa_root_vdev
);
7753 for (uint64_t class = 0; class < DDT_CLASSES
; class++) {
7754 for (uint64_t type
= 0; type
< DDT_TYPES
; type
++) {
7755 for (uint64_t cksum
= 0;
7756 cksum
< ZIO_CHECKSUM_FUNCTIONS
; cksum
++) {
7757 ddt_t
*ddt
= spa
->spa_ddt
[cksum
];
7758 mos_obj_refd(ddt
->ddt_object
[type
][class]);
7764 * Visit all allocated objects and make sure they are referenced.
7766 uint64_t object
= 0;
7767 while (dmu_object_next(mos
, &object
, B_FALSE
, 0) == 0) {
7768 if (range_tree_contains(mos_refd_objs
, object
, 1)) {
7769 range_tree_remove(mos_refd_objs
, object
, 1);
7771 dmu_object_info_t doi
;
7773 VERIFY0(dmu_object_info(mos
, object
, &doi
));
7774 if (doi
.doi_type
& DMU_OT_NEWTYPE
) {
7775 dmu_object_byteswap_t bswap
=
7776 DMU_OT_BYTESWAP(doi
.doi_type
);
7777 name
= dmu_ot_byteswap
[bswap
].ob_name
;
7779 name
= dmu_ot
[doi
.doi_type
].ot_name
;
7782 (void) printf("MOS object %llu (%s) leaked\n",
7783 (u_longlong_t
)object
, name
);
7787 (void) range_tree_walk(mos_refd_objs
, mos_leaks_cb
, NULL
);
7788 if (!range_tree_is_empty(mos_refd_objs
))
7790 range_tree_vacate(mos_refd_objs
, NULL
, NULL
);
7791 range_tree_destroy(mos_refd_objs
);
7795 typedef struct log_sm_obsolete_stats_arg
{
7796 uint64_t lsos_current_txg
;
7798 uint64_t lsos_total_entries
;
7799 uint64_t lsos_valid_entries
;
7801 uint64_t lsos_sm_entries
;
7802 uint64_t lsos_valid_sm_entries
;
7803 } log_sm_obsolete_stats_arg_t
;
7806 log_spacemap_obsolete_stats_cb(spa_t
*spa
, space_map_entry_t
*sme
,
7807 uint64_t txg
, void *arg
)
7809 log_sm_obsolete_stats_arg_t
*lsos
= arg
;
7811 uint64_t offset
= sme
->sme_offset
;
7812 uint64_t vdev_id
= sme
->sme_vdev
;
7814 if (lsos
->lsos_current_txg
== 0) {
7815 /* this is the first log */
7816 lsos
->lsos_current_txg
= txg
;
7817 } else if (lsos
->lsos_current_txg
< txg
) {
7818 /* we just changed log - print stats and reset */
7819 (void) printf("%-8llu valid entries out of %-8llu - txg %llu\n",
7820 (u_longlong_t
)lsos
->lsos_valid_sm_entries
,
7821 (u_longlong_t
)lsos
->lsos_sm_entries
,
7822 (u_longlong_t
)lsos
->lsos_current_txg
);
7823 lsos
->lsos_valid_sm_entries
= 0;
7824 lsos
->lsos_sm_entries
= 0;
7825 lsos
->lsos_current_txg
= txg
;
7827 ASSERT3U(lsos
->lsos_current_txg
, ==, txg
);
7829 lsos
->lsos_sm_entries
++;
7830 lsos
->lsos_total_entries
++;
7832 vdev_t
*vd
= vdev_lookup_top(spa
, vdev_id
);
7833 if (!vdev_is_concrete(vd
))
7836 metaslab_t
*ms
= vd
->vdev_ms
[offset
>> vd
->vdev_ms_shift
];
7837 ASSERT(sme
->sme_type
== SM_ALLOC
|| sme
->sme_type
== SM_FREE
);
7839 if (txg
< metaslab_unflushed_txg(ms
))
7841 lsos
->lsos_valid_sm_entries
++;
7842 lsos
->lsos_valid_entries
++;
7847 dump_log_spacemap_obsolete_stats(spa_t
*spa
)
7849 if (!spa_feature_is_active(spa
, SPA_FEATURE_LOG_SPACEMAP
))
7852 log_sm_obsolete_stats_arg_t lsos
= {0};
7854 (void) printf("Log Space Map Obsolete Entry Statistics:\n");
7856 iterate_through_spacemap_logs(spa
,
7857 log_spacemap_obsolete_stats_cb
, &lsos
);
7859 /* print stats for latest log */
7860 (void) printf("%-8llu valid entries out of %-8llu - txg %llu\n",
7861 (u_longlong_t
)lsos
.lsos_valid_sm_entries
,
7862 (u_longlong_t
)lsos
.lsos_sm_entries
,
7863 (u_longlong_t
)lsos
.lsos_current_txg
);
7865 (void) printf("%-8llu valid entries out of %-8llu - total\n\n",
7866 (u_longlong_t
)lsos
.lsos_valid_entries
,
7867 (u_longlong_t
)lsos
.lsos_total_entries
);
7871 dump_zpool(spa_t
*spa
)
7873 dsl_pool_t
*dp
= spa_get_dsl(spa
);
7876 if (dump_opt
['y']) {
7877 livelist_metaslab_validate(spa
);
7880 if (dump_opt
['S']) {
7881 dump_simulated_ddt(spa
);
7885 if (!dump_opt
['e'] && dump_opt
['C'] > 1) {
7886 (void) printf("\nCached configuration:\n");
7887 dump_nvlist(spa
->spa_config
, 8);
7894 dump_uberblock(&spa
->spa_uberblock
, "\nUberblock:\n", "\n");
7899 if (dump_opt
['d'] > 2 || dump_opt
['m'])
7900 dump_metaslabs(spa
);
7902 dump_metaslab_groups(spa
, dump_opt
['M'] > 1);
7903 if (dump_opt
['d'] > 2 || dump_opt
['m']) {
7904 dump_log_spacemaps(spa
);
7905 dump_log_spacemap_obsolete_stats(spa
);
7908 if (dump_opt
['d'] || dump_opt
['i']) {
7910 mos_refd_objs
= range_tree_create(NULL
, RANGE_SEG64
, NULL
, 0,
7912 dump_objset(dp
->dp_meta_objset
);
7914 if (dump_opt
['d'] >= 3) {
7915 dsl_pool_t
*dp
= spa
->spa_dsl_pool
;
7916 dump_full_bpobj(&spa
->spa_deferred_bpobj
,
7917 "Deferred frees", 0);
7918 if (spa_version(spa
) >= SPA_VERSION_DEADLISTS
) {
7919 dump_full_bpobj(&dp
->dp_free_bpobj
,
7920 "Pool snapshot frees", 0);
7922 if (bpobj_is_open(&dp
->dp_obsolete_bpobj
)) {
7923 ASSERT(spa_feature_is_enabled(spa
,
7924 SPA_FEATURE_DEVICE_REMOVAL
));
7925 dump_full_bpobj(&dp
->dp_obsolete_bpobj
,
7926 "Pool obsolete blocks", 0);
7929 if (spa_feature_is_active(spa
,
7930 SPA_FEATURE_ASYNC_DESTROY
)) {
7931 dump_bptree(spa
->spa_meta_objset
,
7933 "Pool dataset frees");
7935 dump_dtl(spa
->spa_root_vdev
, 0);
7938 for (spa_feature_t f
= 0; f
< SPA_FEATURES
; f
++)
7939 global_feature_count
[f
] = UINT64_MAX
;
7940 global_feature_count
[SPA_FEATURE_REDACTION_BOOKMARKS
] = 0;
7941 global_feature_count
[SPA_FEATURE_BOOKMARK_WRITTEN
] = 0;
7942 global_feature_count
[SPA_FEATURE_LIVELIST
] = 0;
7944 (void) dmu_objset_find(spa_name(spa
), dump_one_objset
,
7945 NULL
, DS_FIND_SNAPSHOTS
| DS_FIND_CHILDREN
);
7947 if (rc
== 0 && !dump_opt
['L'])
7948 rc
= dump_mos_leaks(spa
);
7950 for (f
= 0; f
< SPA_FEATURES
; f
++) {
7954 if (!(spa_feature_table
[f
].fi_flags
&
7955 ZFEATURE_FLAG_PER_DATASET
)) {
7956 if (global_feature_count
[f
] == UINT64_MAX
)
7958 if (!spa_feature_is_enabled(spa
, f
)) {
7959 ASSERT0(global_feature_count
[f
]);
7962 arr
= global_feature_count
;
7964 if (!spa_feature_is_enabled(spa
, f
)) {
7965 ASSERT0(dataset_feature_count
[f
]);
7968 arr
= dataset_feature_count
;
7970 if (feature_get_refcount(spa
, &spa_feature_table
[f
],
7971 &refcount
) == ENOTSUP
)
7973 if (arr
[f
] != refcount
) {
7974 (void) printf("%s feature refcount mismatch: "
7975 "%lld consumers != %lld refcount\n",
7976 spa_feature_table
[f
].fi_uname
,
7977 (longlong_t
)arr
[f
], (longlong_t
)refcount
);
7980 (void) printf("Verified %s feature refcount "
7981 "of %llu is correct\n",
7982 spa_feature_table
[f
].fi_uname
,
7983 (longlong_t
)refcount
);
7988 rc
= verify_device_removal_feature_counts(spa
);
7991 if (rc
== 0 && (dump_opt
['b'] || dump_opt
['c']))
7992 rc
= dump_block_stats(spa
);
7995 rc
= verify_spacemap_refcounts(spa
);
7998 show_pool_stats(spa
);
8004 rc
= verify_checkpoint(spa
);
8007 dump_debug_buffer();
8012 #define ZDB_FLAG_CHECKSUM 0x0001
8013 #define ZDB_FLAG_DECOMPRESS 0x0002
8014 #define ZDB_FLAG_BSWAP 0x0004
8015 #define ZDB_FLAG_GBH 0x0008
8016 #define ZDB_FLAG_INDIRECT 0x0010
8017 #define ZDB_FLAG_RAW 0x0020
8018 #define ZDB_FLAG_PRINT_BLKPTR 0x0040
8019 #define ZDB_FLAG_VERBOSE 0x0080
8021 static int flagbits
[256];
8022 static char flagbitstr
[16];
8025 zdb_print_blkptr(const blkptr_t
*bp
, int flags
)
8027 char blkbuf
[BP_SPRINTF_LEN
];
8029 if (flags
& ZDB_FLAG_BSWAP
)
8030 byteswap_uint64_array((void *)bp
, sizeof (blkptr_t
));
8032 snprintf_blkptr(blkbuf
, sizeof (blkbuf
), bp
);
8033 (void) printf("%s\n", blkbuf
);
8037 zdb_dump_indirect(blkptr_t
*bp
, int nbps
, int flags
)
8041 for (i
= 0; i
< nbps
; i
++)
8042 zdb_print_blkptr(&bp
[i
], flags
);
8046 zdb_dump_gbh(void *buf
, int flags
)
8048 zdb_dump_indirect((blkptr_t
*)buf
, SPA_GBH_NBLKPTRS
, flags
);
8052 zdb_dump_block_raw(void *buf
, uint64_t size
, int flags
)
8054 if (flags
& ZDB_FLAG_BSWAP
)
8055 byteswap_uint64_array(buf
, size
);
8056 VERIFY(write(fileno(stdout
), buf
, size
) == size
);
8060 zdb_dump_block(char *label
, void *buf
, uint64_t size
, int flags
)
8062 uint64_t *d
= (uint64_t *)buf
;
8063 unsigned nwords
= size
/ sizeof (uint64_t);
8064 int do_bswap
= !!(flags
& ZDB_FLAG_BSWAP
);
8071 hdr
= " 7 6 5 4 3 2 1 0 f e d c b a 9 8";
8073 hdr
= " 0 1 2 3 4 5 6 7 8 9 a b c d e f";
8075 (void) printf("\n%s\n%6s %s 0123456789abcdef\n", label
, "", hdr
);
8077 #ifdef _LITTLE_ENDIAN
8078 /* correct the endianness */
8079 do_bswap
= !do_bswap
;
8081 for (i
= 0; i
< nwords
; i
+= 2) {
8082 (void) printf("%06llx: %016llx %016llx ",
8083 (u_longlong_t
)(i
* sizeof (uint64_t)),
8084 (u_longlong_t
)(do_bswap
? BSWAP_64(d
[i
]) : d
[i
]),
8085 (u_longlong_t
)(do_bswap
? BSWAP_64(d
[i
+ 1]) : d
[i
+ 1]));
8088 for (j
= 0; j
< 2 * sizeof (uint64_t); j
++)
8089 (void) printf("%c", isprint(c
[j
]) ? c
[j
] : '.');
8090 (void) printf("\n");
8095 * There are two acceptable formats:
8096 * leaf_name - For example: c1t0d0 or /tmp/ztest.0a
8097 * child[.child]* - For example: 0.1.1
8099 * The second form can be used to specify arbitrary vdevs anywhere
8100 * in the hierarchy. For example, in a pool with a mirror of
8101 * RAID-Zs, you can specify either RAID-Z vdev with 0.0 or 0.1 .
8104 zdb_vdev_lookup(vdev_t
*vdev
, const char *path
)
8112 /* First, assume the x.x.x.x format */
8113 i
= strtoul(path
, &s
, 10);
8114 if (s
== path
|| (s
&& *s
!= '.' && *s
!= '\0'))
8116 if (i
>= vdev
->vdev_children
)
8119 vdev
= vdev
->vdev_child
[i
];
8120 if (s
&& *s
== '\0')
8122 return (zdb_vdev_lookup(vdev
, s
+1));
8125 for (i
= 0; i
< vdev
->vdev_children
; i
++) {
8126 vdev_t
*vc
= vdev
->vdev_child
[i
];
8128 if (vc
->vdev_path
== NULL
) {
8129 vc
= zdb_vdev_lookup(vc
, path
);
8136 p
= strrchr(vc
->vdev_path
, '/');
8137 p
= p
? p
+ 1 : vc
->vdev_path
;
8138 q
= &vc
->vdev_path
[strlen(vc
->vdev_path
) - 2];
8140 if (strcmp(vc
->vdev_path
, path
) == 0)
8142 if (strcmp(p
, path
) == 0)
8144 if (strcmp(q
, "s0") == 0 && strncmp(p
, path
, q
- p
) == 0)
8152 name_from_objset_id(spa_t
*spa
, uint64_t objset_id
, char *outstr
)
8156 dsl_pool_config_enter(spa
->spa_dsl_pool
, FTAG
);
8157 int error
= dsl_dataset_hold_obj(spa
->spa_dsl_pool
, objset_id
,
8160 (void) fprintf(stderr
, "failed to hold objset %llu: %s\n",
8161 (u_longlong_t
)objset_id
, strerror(error
));
8162 dsl_pool_config_exit(spa
->spa_dsl_pool
, FTAG
);
8165 dsl_dataset_name(ds
, outstr
);
8166 dsl_dataset_rele(ds
, NULL
);
8167 dsl_pool_config_exit(spa
->spa_dsl_pool
, FTAG
);
8172 zdb_parse_block_sizes(char *sizes
, uint64_t *lsize
, uint64_t *psize
)
8174 char *s0
, *s1
, *tmp
= NULL
;
8179 s0
= strtok_r(sizes
, "/", &tmp
);
8182 s1
= strtok_r(NULL
, "/", &tmp
);
8183 *lsize
= strtoull(s0
, NULL
, 16);
8184 *psize
= s1
? strtoull(s1
, NULL
, 16) : *lsize
;
8185 return (*lsize
>= *psize
&& *psize
> 0);
8188 #define ZIO_COMPRESS_MASK(alg) (1ULL << (ZIO_COMPRESS_##alg))
8191 zdb_decompress_block(abd_t
*pabd
, void *buf
, void *lbuf
, uint64_t lsize
,
8192 uint64_t psize
, int flags
)
8195 boolean_t exceeded
= B_FALSE
;
8197 * We don't know how the data was compressed, so just try
8198 * every decompress function at every inflated blocksize.
8200 void *lbuf2
= umem_alloc(SPA_MAXBLOCKSIZE
, UMEM_NOFAIL
);
8201 int cfuncs
[ZIO_COMPRESS_FUNCTIONS
] = { 0 };
8202 int *cfuncp
= cfuncs
;
8203 uint64_t maxlsize
= SPA_MAXBLOCKSIZE
;
8204 uint64_t mask
= ZIO_COMPRESS_MASK(ON
) | ZIO_COMPRESS_MASK(OFF
) |
8205 ZIO_COMPRESS_MASK(INHERIT
) | ZIO_COMPRESS_MASK(EMPTY
) |
8206 (getenv("ZDB_NO_ZLE") ? ZIO_COMPRESS_MASK(ZLE
) : 0);
8207 *cfuncp
++ = ZIO_COMPRESS_LZ4
;
8208 *cfuncp
++ = ZIO_COMPRESS_LZJB
;
8209 mask
|= ZIO_COMPRESS_MASK(LZ4
) | ZIO_COMPRESS_MASK(LZJB
);
8210 for (int c
= 0; c
< ZIO_COMPRESS_FUNCTIONS
; c
++)
8211 if (((1ULL << c
) & mask
) == 0)
8215 * On the one hand, with SPA_MAXBLOCKSIZE at 16MB, this
8216 * could take a while and we should let the user know
8217 * we are not stuck. On the other hand, printing progress
8218 * info gets old after a while. User can specify 'v' flag
8219 * to see the progression.
8222 lsize
+= SPA_MINBLOCKSIZE
;
8225 for (; lsize
<= maxlsize
; lsize
+= SPA_MINBLOCKSIZE
) {
8226 for (cfuncp
= cfuncs
; *cfuncp
; cfuncp
++) {
8227 if (flags
& ZDB_FLAG_VERBOSE
) {
8228 (void) fprintf(stderr
,
8229 "Trying %05llx -> %05llx (%s)\n",
8230 (u_longlong_t
)psize
,
8231 (u_longlong_t
)lsize
,
8232 zio_compress_table
[*cfuncp
].\
8237 * We randomize lbuf2, and decompress to both
8238 * lbuf and lbuf2. This way, we will know if
8239 * decompression fill exactly to lsize.
8241 VERIFY0(random_get_pseudo_bytes(lbuf2
, lsize
));
8243 if (zio_decompress_data(*cfuncp
, pabd
,
8244 lbuf
, psize
, lsize
, NULL
) == 0 &&
8245 zio_decompress_data(*cfuncp
, pabd
,
8246 lbuf2
, psize
, lsize
, NULL
) == 0 &&
8247 memcmp(lbuf
, lbuf2
, lsize
) == 0)
8253 umem_free(lbuf2
, SPA_MAXBLOCKSIZE
);
8255 if (lsize
> maxlsize
) {
8258 if (*cfuncp
== ZIO_COMPRESS_ZLE
) {
8259 printf("\nZLE decompression was selected. If you "
8260 "suspect the results are wrong,\ntry avoiding ZLE "
8261 "by setting and exporting ZDB_NO_ZLE=\"true\"\n");
8268 * Read a block from a pool and print it out. The syntax of the
8269 * block descriptor is:
8271 * pool:vdev_specifier:offset:[lsize/]psize[:flags]
8273 * pool - The name of the pool you wish to read from
8274 * vdev_specifier - Which vdev (see comment for zdb_vdev_lookup)
8275 * offset - offset, in hex, in bytes
8276 * size - Amount of data to read, in hex, in bytes
8277 * flags - A string of characters specifying options
8278 * b: Decode a blkptr at given offset within block
8279 * c: Calculate and display checksums
8280 * d: Decompress data before dumping
8281 * e: Byteswap data before dumping
8282 * g: Display data as a gang block header
8283 * i: Display as an indirect block
8284 * r: Dump raw data to stdout
8289 zdb_read_block(char *thing
, spa_t
*spa
)
8291 blkptr_t blk
, *bp
= &blk
;
8292 dva_t
*dva
= bp
->blk_dva
;
8294 uint64_t offset
= 0, psize
= 0, lsize
= 0, blkptr_offset
= 0;
8299 char *s
, *p
, *dup
, *flagstr
, *sizes
, *tmp
= NULL
;
8300 const char *vdev
, *errmsg
= NULL
;
8302 boolean_t borrowed
= B_FALSE
, found
= B_FALSE
;
8304 dup
= strdup(thing
);
8305 s
= strtok_r(dup
, ":", &tmp
);
8307 s
= strtok_r(NULL
, ":", &tmp
);
8308 offset
= strtoull(s
? s
: "", NULL
, 16);
8309 sizes
= strtok_r(NULL
, ":", &tmp
);
8310 s
= strtok_r(NULL
, ":", &tmp
);
8311 flagstr
= strdup(s
?: "");
8313 if (!zdb_parse_block_sizes(sizes
, &lsize
, &psize
))
8314 errmsg
= "invalid size(s)";
8315 if (!IS_P2ALIGNED(psize
, DEV_BSIZE
) || !IS_P2ALIGNED(lsize
, DEV_BSIZE
))
8316 errmsg
= "size must be a multiple of sector size";
8317 if (!IS_P2ALIGNED(offset
, DEV_BSIZE
))
8318 errmsg
= "offset must be a multiple of sector size";
8320 (void) printf("Invalid block specifier: %s - %s\n",
8326 for (s
= strtok_r(flagstr
, ":", &tmp
);
8328 s
= strtok_r(NULL
, ":", &tmp
)) {
8329 for (i
= 0; i
< strlen(flagstr
); i
++) {
8330 int bit
= flagbits
[(uchar_t
)flagstr
[i
]];
8333 (void) printf("***Ignoring flag: %c\n",
8334 (uchar_t
)flagstr
[i
]);
8340 p
= &flagstr
[i
+ 1];
8341 if (*p
!= ':' && *p
!= '\0') {
8342 int j
= 0, nextbit
= flagbits
[(uchar_t
)*p
];
8343 char *end
, offstr
[8] = { 0 };
8344 if ((bit
== ZDB_FLAG_PRINT_BLKPTR
) &&
8346 /* look ahead to isolate the offset */
8347 while (nextbit
== 0 &&
8348 strchr(flagbitstr
, *p
) == NULL
) {
8351 if (i
+ j
> strlen(flagstr
))
8354 nextbit
= flagbits
[(uchar_t
)*p
];
8356 blkptr_offset
= strtoull(offstr
, &end
,
8359 } else if (nextbit
== 0) {
8360 (void) printf("***Ignoring flag arg:"
8361 " '%c'\n", (uchar_t
)*p
);
8366 if (blkptr_offset
% sizeof (blkptr_t
)) {
8367 printf("Block pointer offset 0x%llx "
8368 "must be divisible by 0x%x\n",
8369 (longlong_t
)blkptr_offset
, (int)sizeof (blkptr_t
));
8372 if (found
== B_FALSE
&& strlen(flagstr
) > 0) {
8373 printf("Invalid flag arg: '%s'\n", flagstr
);
8377 vd
= zdb_vdev_lookup(spa
->spa_root_vdev
, vdev
);
8379 (void) printf("***Invalid vdev: %s\n", vdev
);
8383 (void) fprintf(stderr
, "Found vdev: %s\n",
8386 (void) fprintf(stderr
, "Found vdev type: %s\n",
8387 vd
->vdev_ops
->vdev_op_type
);
8390 pabd
= abd_alloc_for_io(SPA_MAXBLOCKSIZE
, B_FALSE
);
8391 lbuf
= umem_alloc(SPA_MAXBLOCKSIZE
, UMEM_NOFAIL
);
8395 DVA_SET_VDEV(&dva
[0], vd
->vdev_id
);
8396 DVA_SET_OFFSET(&dva
[0], offset
);
8397 DVA_SET_GANG(&dva
[0], !!(flags
& ZDB_FLAG_GBH
));
8398 DVA_SET_ASIZE(&dva
[0], vdev_psize_to_asize(vd
, psize
));
8400 BP_SET_BIRTH(bp
, TXG_INITIAL
, TXG_INITIAL
);
8402 BP_SET_LSIZE(bp
, lsize
);
8403 BP_SET_PSIZE(bp
, psize
);
8404 BP_SET_COMPRESS(bp
, ZIO_COMPRESS_OFF
);
8405 BP_SET_CHECKSUM(bp
, ZIO_CHECKSUM_OFF
);
8406 BP_SET_TYPE(bp
, DMU_OT_NONE
);
8407 BP_SET_LEVEL(bp
, 0);
8408 BP_SET_DEDUP(bp
, 0);
8409 BP_SET_BYTEORDER(bp
, ZFS_HOST_BYTEORDER
);
8411 spa_config_enter(spa
, SCL_STATE
, FTAG
, RW_READER
);
8412 zio
= zio_root(spa
, NULL
, NULL
, 0);
8414 if (vd
== vd
->vdev_top
) {
8416 * Treat this as a normal block read.
8418 zio_nowait(zio_read(zio
, spa
, bp
, pabd
, psize
, NULL
, NULL
,
8419 ZIO_PRIORITY_SYNC_READ
,
8420 ZIO_FLAG_CANFAIL
| ZIO_FLAG_RAW
, NULL
));
8423 * Treat this as a vdev child I/O.
8425 zio_nowait(zio_vdev_child_io(zio
, bp
, vd
, offset
, pabd
,
8426 psize
, ZIO_TYPE_READ
, ZIO_PRIORITY_SYNC_READ
,
8427 ZIO_FLAG_DONT_CACHE
| ZIO_FLAG_DONT_PROPAGATE
|
8428 ZIO_FLAG_DONT_RETRY
| ZIO_FLAG_CANFAIL
| ZIO_FLAG_RAW
|
8429 ZIO_FLAG_OPTIONAL
, NULL
, NULL
));
8432 error
= zio_wait(zio
);
8433 spa_config_exit(spa
, SCL_STATE
, FTAG
);
8436 (void) printf("Read of %s failed, error: %d\n", thing
, error
);
8440 uint64_t orig_lsize
= lsize
;
8442 if (flags
& ZDB_FLAG_DECOMPRESS
) {
8443 boolean_t failed
= zdb_decompress_block(pabd
, buf
, lbuf
,
8444 lsize
, psize
, flags
);
8446 (void) printf("Decompress of %s failed\n", thing
);
8450 buf
= abd_borrow_buf_copy(pabd
, lsize
);
8454 * Try to detect invalid block pointer. If invalid, try
8457 if ((flags
& ZDB_FLAG_PRINT_BLKPTR
|| flags
& ZDB_FLAG_INDIRECT
) &&
8458 !(flags
& ZDB_FLAG_DECOMPRESS
)) {
8459 const blkptr_t
*b
= (const blkptr_t
*)(void *)
8460 ((uintptr_t)buf
+ (uintptr_t)blkptr_offset
);
8461 if (zfs_blkptr_verify(spa
, b
, B_FALSE
, BLK_VERIFY_ONLY
) ==
8463 abd_return_buf_copy(pabd
, buf
, lsize
);
8466 boolean_t failed
= zdb_decompress_block(pabd
, buf
,
8467 lbuf
, lsize
, psize
, flags
);
8468 b
= (const blkptr_t
*)(void *)
8469 ((uintptr_t)buf
+ (uintptr_t)blkptr_offset
);
8470 if (failed
|| zfs_blkptr_verify(spa
, b
, B_FALSE
,
8471 BLK_VERIFY_LOG
) == B_FALSE
) {
8472 printf("invalid block pointer at this DVA\n");
8478 if (flags
& ZDB_FLAG_PRINT_BLKPTR
)
8479 zdb_print_blkptr((blkptr_t
*)(void *)
8480 ((uintptr_t)buf
+ (uintptr_t)blkptr_offset
), flags
);
8481 else if (flags
& ZDB_FLAG_RAW
)
8482 zdb_dump_block_raw(buf
, lsize
, flags
);
8483 else if (flags
& ZDB_FLAG_INDIRECT
)
8484 zdb_dump_indirect((blkptr_t
*)buf
,
8485 orig_lsize
/ sizeof (blkptr_t
), flags
);
8486 else if (flags
& ZDB_FLAG_GBH
)
8487 zdb_dump_gbh(buf
, flags
);
8489 zdb_dump_block(thing
, buf
, lsize
, flags
);
8492 * If :c was specified, iterate through the checksum table to
8493 * calculate and display each checksum for our specified
8496 if ((flags
& ZDB_FLAG_CHECKSUM
) && !(flags
& ZDB_FLAG_RAW
) &&
8497 !(flags
& ZDB_FLAG_GBH
)) {
8499 (void) printf("\n");
8500 for (enum zio_checksum ck
= ZIO_CHECKSUM_LABEL
;
8501 ck
< ZIO_CHECKSUM_FUNCTIONS
; ck
++) {
8503 if ((zio_checksum_table
[ck
].ci_flags
&
8504 ZCHECKSUM_FLAG_EMBEDDED
) ||
8505 ck
== ZIO_CHECKSUM_NOPARITY
) {
8508 BP_SET_CHECKSUM(bp
, ck
);
8509 spa_config_enter(spa
, SCL_STATE
, FTAG
, RW_READER
);
8510 czio
= zio_root(spa
, NULL
, NULL
, ZIO_FLAG_CANFAIL
);
8513 if (vd
== vd
->vdev_top
) {
8514 zio_nowait(zio_read(czio
, spa
, bp
, pabd
, psize
,
8516 ZIO_PRIORITY_SYNC_READ
,
8517 ZIO_FLAG_CANFAIL
| ZIO_FLAG_RAW
|
8518 ZIO_FLAG_DONT_RETRY
, NULL
));
8520 zio_nowait(zio_vdev_child_io(czio
, bp
, vd
,
8521 offset
, pabd
, psize
, ZIO_TYPE_READ
,
8522 ZIO_PRIORITY_SYNC_READ
,
8523 ZIO_FLAG_DONT_CACHE
|
8524 ZIO_FLAG_DONT_PROPAGATE
|
8525 ZIO_FLAG_DONT_RETRY
|
8526 ZIO_FLAG_CANFAIL
| ZIO_FLAG_RAW
|
8527 ZIO_FLAG_SPECULATIVE
|
8528 ZIO_FLAG_OPTIONAL
, NULL
, NULL
));
8530 error
= zio_wait(czio
);
8531 if (error
== 0 || error
== ECKSUM
) {
8532 zio_t
*ck_zio
= zio_root(spa
, NULL
, NULL
, 0);
8534 DVA_GET_OFFSET(&bp
->blk_dva
[0]);
8536 zio_checksum_compute(ck_zio
, ck
, pabd
, lsize
);
8539 "cksum=%016llx:%016llx:%016llx:%016llx\n",
8540 zio_checksum_table
[ck
].ci_name
,
8541 (u_longlong_t
)bp
->blk_cksum
.zc_word
[0],
8542 (u_longlong_t
)bp
->blk_cksum
.zc_word
[1],
8543 (u_longlong_t
)bp
->blk_cksum
.zc_word
[2],
8544 (u_longlong_t
)bp
->blk_cksum
.zc_word
[3]);
8547 printf("error %d reading block\n", error
);
8549 spa_config_exit(spa
, SCL_STATE
, FTAG
);
8554 abd_return_buf_copy(pabd
, buf
, lsize
);
8558 umem_free(lbuf
, SPA_MAXBLOCKSIZE
);
8565 zdb_embedded_block(char *thing
)
8567 blkptr_t bp
= {{{{0}}}};
8568 unsigned long long *words
= (void *)&bp
;
8572 err
= sscanf(thing
, "%llx:%llx:%llx:%llx:%llx:%llx:%llx:%llx:"
8573 "%llx:%llx:%llx:%llx:%llx:%llx:%llx:%llx",
8574 words
+ 0, words
+ 1, words
+ 2, words
+ 3,
8575 words
+ 4, words
+ 5, words
+ 6, words
+ 7,
8576 words
+ 8, words
+ 9, words
+ 10, words
+ 11,
8577 words
+ 12, words
+ 13, words
+ 14, words
+ 15);
8579 (void) fprintf(stderr
, "invalid input format\n");
8582 ASSERT3U(BPE_GET_LSIZE(&bp
), <=, SPA_MAXBLOCKSIZE
);
8583 buf
= malloc(SPA_MAXBLOCKSIZE
);
8585 (void) fprintf(stderr
, "out of memory\n");
8588 err
= decode_embedded_bp(&bp
, buf
, BPE_GET_LSIZE(&bp
));
8590 (void) fprintf(stderr
, "decode failed: %u\n", err
);
8593 zdb_dump_block_raw(buf
, BPE_GET_LSIZE(&bp
), 0);
8597 /* check for valid hex or decimal numeric string */
8599 zdb_numeric(char *str
)
8603 if (strlen(str
) == 0)
8605 if (strncmp(str
, "0x", 2) == 0 || strncmp(str
, "0X", 2) == 0)
8607 for (; i
< strlen(str
); i
++) {
8608 if (!isxdigit(str
[i
]))
8615 main(int argc
, char **argv
)
8619 objset_t
*os
= NULL
;
8623 char **searchdirs
= NULL
;
8625 char *target
, *target_pool
, dsname
[ZFS_MAX_DATASET_NAME_LEN
];
8626 nvlist_t
*policy
= NULL
;
8627 uint64_t max_txg
= UINT64_MAX
;
8628 int64_t objset_id
= -1;
8630 int flags
= ZFS_IMPORT_MISSING_LOG
;
8631 int rewind
= ZPOOL_NEVER_REWIND
;
8632 char *spa_config_path_env
, *objset_str
;
8633 boolean_t target_is_spa
= B_TRUE
, dataset_lookup
= B_FALSE
;
8634 nvlist_t
*cfg
= NULL
;
8636 dprintf_setup(&argc
, argv
);
8639 * If there is an environment variable SPA_CONFIG_PATH it overrides
8640 * default spa_config_path setting. If -U flag is specified it will
8641 * override this environment variable settings once again.
8643 spa_config_path_env
= getenv("SPA_CONFIG_PATH");
8644 if (spa_config_path_env
!= NULL
)
8645 spa_config_path
= spa_config_path_env
;
8648 * For performance reasons, we set this tunable down. We do so before
8649 * the arg parsing section so that the user can override this value if
8652 zfs_btree_verify_intensity
= 3;
8654 struct option long_options
[] = {
8655 {"ignore-assertions", no_argument
, NULL
, 'A'},
8656 {"block-stats", no_argument
, NULL
, 'b'},
8657 {"checksum", no_argument
, NULL
, 'c'},
8658 {"config", no_argument
, NULL
, 'C'},
8659 {"datasets", no_argument
, NULL
, 'd'},
8660 {"dedup-stats", no_argument
, NULL
, 'D'},
8661 {"exported", no_argument
, NULL
, 'e'},
8662 {"embedded-block-pointer", no_argument
, NULL
, 'E'},
8663 {"automatic-rewind", no_argument
, NULL
, 'F'},
8664 {"dump-debug-msg", no_argument
, NULL
, 'G'},
8665 {"history", no_argument
, NULL
, 'h'},
8666 {"intent-logs", no_argument
, NULL
, 'i'},
8667 {"inflight", required_argument
, NULL
, 'I'},
8668 {"checkpointed-state", no_argument
, NULL
, 'k'},
8669 {"key", required_argument
, NULL
, 'K'},
8670 {"label", no_argument
, NULL
, 'l'},
8671 {"disable-leak-tracking", no_argument
, NULL
, 'L'},
8672 {"metaslabs", no_argument
, NULL
, 'm'},
8673 {"metaslab-groups", no_argument
, NULL
, 'M'},
8674 {"numeric", no_argument
, NULL
, 'N'},
8675 {"option", required_argument
, NULL
, 'o'},
8676 {"object-lookups", no_argument
, NULL
, 'O'},
8677 {"path", required_argument
, NULL
, 'p'},
8678 {"parseable", no_argument
, NULL
, 'P'},
8679 {"skip-label", no_argument
, NULL
, 'q'},
8680 {"copy-object", no_argument
, NULL
, 'r'},
8681 {"read-block", no_argument
, NULL
, 'R'},
8682 {"io-stats", no_argument
, NULL
, 's'},
8683 {"simulate-dedup", no_argument
, NULL
, 'S'},
8684 {"txg", required_argument
, NULL
, 't'},
8685 {"uberblock", no_argument
, NULL
, 'u'},
8686 {"cachefile", required_argument
, NULL
, 'U'},
8687 {"verbose", no_argument
, NULL
, 'v'},
8688 {"verbatim", no_argument
, NULL
, 'V'},
8689 {"dump-blocks", required_argument
, NULL
, 'x'},
8690 {"extreme-rewind", no_argument
, NULL
, 'X'},
8691 {"all-reconstruction", no_argument
, NULL
, 'Y'},
8692 {"livelist", no_argument
, NULL
, 'y'},
8693 {"zstd-headers", no_argument
, NULL
, 'Z'},
8697 while ((c
= getopt_long(argc
, argv
,
8698 "AbcCdDeEFGhiI:kK:lLmMNo:Op:PqrRsSt:uU:vVx:XYyZ",
8699 long_options
, NULL
)) != -1) {
8736 zfs_reconstruct_indirect_combinations_max
= INT_MAX
;
8737 zfs_deadman_enabled
= 0;
8739 /* NB: Sort single match options below. */
8741 max_inflight_bytes
= strtoull(optarg
, NULL
, 0);
8742 if (max_inflight_bytes
== 0) {
8743 (void) fprintf(stderr
, "maximum number "
8744 "of inflight bytes must be greater "
8751 key_material
= strdup(optarg
);
8752 /* redact key material in process table */
8753 while (*optarg
!= '\0') { *optarg
++ = '*'; }
8756 error
= set_global_var(optarg
);
8761 if (searchdirs
== NULL
) {
8762 searchdirs
= umem_alloc(sizeof (char *),
8765 char **tmp
= umem_alloc((nsearch
+ 1) *
8766 sizeof (char *), UMEM_NOFAIL
);
8767 memcpy(tmp
, searchdirs
, nsearch
*
8769 umem_free(searchdirs
,
8770 nsearch
* sizeof (char *));
8773 searchdirs
[nsearch
++] = optarg
;
8776 max_txg
= strtoull(optarg
, NULL
, 0);
8777 if (max_txg
< TXG_INITIAL
) {
8778 (void) fprintf(stderr
, "incorrect txg "
8779 "specified: %s\n", optarg
);
8784 spa_config_path
= optarg
;
8785 if (spa_config_path
[0] != '/') {
8786 (void) fprintf(stderr
,
8787 "cachefile must be an absolute path "
8788 "(i.e. start with a slash)\n");
8796 flags
= ZFS_IMPORT_VERBATIM
;
8799 vn_dumpdir
= optarg
;
8807 if (!dump_opt
['e'] && searchdirs
!= NULL
) {
8808 (void) fprintf(stderr
, "-p option requires use of -e\n");
8813 * ZDB does not typically re-read blocks; therefore limit the ARC
8814 * to 256 MB, which can be used entirely for metadata.
8816 zfs_arc_min
= 2ULL << SPA_MAXBLOCKSHIFT
;
8817 zfs_arc_max
= 256 * 1024 * 1024;
8821 * "zdb -c" uses checksum-verifying scrub i/os which are async reads.
8822 * "zdb -b" uses traversal prefetch which uses async reads.
8823 * For good performance, let several of them be active at once.
8825 zfs_vdev_async_read_max_active
= 10;
8828 * Disable reference tracking for better performance.
8830 reference_tracking_enable
= B_FALSE
;
8833 * Do not fail spa_load when spa_load_verify fails. This is needed
8834 * to load non-idle pools.
8836 spa_load_verify_dryrun
= B_TRUE
;
8839 * ZDB should have ability to read spacemaps.
8841 spa_mode_readable_spacemaps
= B_TRUE
;
8843 kernel_init(SPA_MODE_READ
);
8846 verbose
= MAX(verbose
, 1);
8848 for (c
= 0; c
< 256; c
++) {
8849 if (dump_all
&& strchr("AeEFkKlLNOPrRSXy", c
) == NULL
)
8852 dump_opt
[c
] += verbose
;
8855 libspl_set_assert_ok((dump_opt
['A'] == 1) || (dump_opt
['A'] > 2));
8856 zfs_recover
= (dump_opt
['A'] > 1);
8860 if (argc
< 2 && dump_opt
['R'])
8863 if (dump_opt
['E']) {
8866 zdb_embedded_block(argv
[0]);
8871 if (!dump_opt
['e'] && dump_opt
['C']) {
8872 dump_cachefile(spa_config_path
);
8879 return (dump_label(argv
[0]));
8881 if (dump_opt
['O']) {
8884 dump_opt
['v'] = verbose
+ 3;
8885 return (dump_path(argv
[0], argv
[1], NULL
));
8887 if (dump_opt
['r']) {
8888 target_is_spa
= B_FALSE
;
8891 dump_opt
['v'] = verbose
;
8892 error
= dump_path(argv
[0], argv
[1], &object
);
8894 fatal("internal error: %s", strerror(error
));
8897 if (dump_opt
['X'] || dump_opt
['F'])
8898 rewind
= ZPOOL_DO_REWIND
|
8899 (dump_opt
['X'] ? ZPOOL_EXTREME_REWIND
: 0);
8902 if (dump_opt
['N'] && dump_opt
['d'] == 0)
8903 dump_opt
['d'] = dump_opt
['N'];
8905 if (nvlist_alloc(&policy
, NV_UNIQUE_NAME_TYPE
, 0) != 0 ||
8906 nvlist_add_uint64(policy
, ZPOOL_LOAD_REQUEST_TXG
, max_txg
) != 0 ||
8907 nvlist_add_uint32(policy
, ZPOOL_LOAD_REWIND_POLICY
, rewind
) != 0)
8908 fatal("internal error: %s", strerror(ENOMEM
));
8913 if (strpbrk(target
, "/@") != NULL
) {
8916 target_pool
= strdup(target
);
8917 *strpbrk(target_pool
, "/@") = '\0';
8919 target_is_spa
= B_FALSE
;
8920 targetlen
= strlen(target
);
8921 if (targetlen
&& target
[targetlen
- 1] == '/')
8922 target
[targetlen
- 1] = '\0';
8925 * See if an objset ID was supplied (-d <pool>/<objset ID>).
8926 * To disambiguate tank/100, consider the 100 as objsetID
8927 * if -N was given, otherwise 100 is an objsetID iff
8928 * tank/100 as a named dataset fails on lookup.
8930 objset_str
= strchr(target
, '/');
8931 if (objset_str
&& strlen(objset_str
) > 1 &&
8932 zdb_numeric(objset_str
+ 1)) {
8936 objset_id
= strtoull(objset_str
, &endptr
, 0);
8937 /* dataset 0 is the same as opening the pool */
8938 if (errno
== 0 && endptr
!= objset_str
&&
8941 dataset_lookup
= B_TRUE
;
8943 /* normal dataset name not an objset ID */
8944 if (endptr
== objset_str
) {
8947 } else if (objset_str
&& !zdb_numeric(objset_str
+ 1) &&
8949 printf("Supply a numeric objset ID with -N\n");
8953 target_pool
= target
;
8956 if (dump_opt
['e']) {
8957 importargs_t args
= { 0 };
8959 args
.paths
= nsearch
;
8960 args
.path
= searchdirs
;
8961 args
.can_be_active
= B_TRUE
;
8963 libpc_handle_t lpch
= {
8964 .lpc_lib_handle
= NULL
,
8965 .lpc_ops
= &libzpool_config_ops
,
8966 .lpc_printerr
= B_TRUE
8968 error
= zpool_find_config(&lpch
, target_pool
, &cfg
, &args
);
8972 if (nvlist_add_nvlist(cfg
,
8973 ZPOOL_LOAD_POLICY
, policy
) != 0) {
8974 fatal("can't open '%s': %s",
8975 target
, strerror(ENOMEM
));
8978 if (dump_opt
['C'] > 1) {
8979 (void) printf("\nConfiguration for import:\n");
8980 dump_nvlist(cfg
, 8);
8984 * Disable the activity check to allow examination of
8987 error
= spa_import(target_pool
, cfg
, NULL
,
8988 flags
| ZFS_IMPORT_SKIP_MMP
);
8992 if (searchdirs
!= NULL
) {
8993 umem_free(searchdirs
, nsearch
* sizeof (char *));
8998 * import_checkpointed_state makes the assumption that the
8999 * target pool that we pass it is already part of the spa
9000 * namespace. Because of that we need to make sure to call
9001 * it always after the -e option has been processed, which
9002 * imports the pool to the namespace if it's not in the
9005 char *checkpoint_pool
= NULL
;
9006 char *checkpoint_target
= NULL
;
9007 if (dump_opt
['k']) {
9008 checkpoint_pool
= import_checkpointed_state(target
, cfg
,
9009 &checkpoint_target
);
9011 if (checkpoint_target
!= NULL
)
9012 target
= checkpoint_target
;
9020 if (target_pool
!= target
)
9024 if (dump_opt
['k'] && (target_is_spa
|| dump_opt
['R'])) {
9025 ASSERT(checkpoint_pool
!= NULL
);
9026 ASSERT(checkpoint_target
== NULL
);
9028 error
= spa_open(checkpoint_pool
, &spa
, FTAG
);
9030 fatal("Tried to open pool \"%s\" but "
9031 "spa_open() failed with error %d\n",
9032 checkpoint_pool
, error
);
9035 } else if (target_is_spa
|| dump_opt
['R'] || objset_id
== 0) {
9036 zdb_set_skip_mmp(target
);
9037 error
= spa_open_rewind(target
, &spa
, FTAG
, policy
,
9041 * If we're missing the log device then
9042 * try opening the pool after clearing the
9045 mutex_enter(&spa_namespace_lock
);
9046 if ((spa
= spa_lookup(target
)) != NULL
&&
9047 spa
->spa_log_state
== SPA_LOG_MISSING
) {
9048 spa
->spa_log_state
= SPA_LOG_CLEAR
;
9051 mutex_exit(&spa_namespace_lock
);
9054 error
= spa_open_rewind(target
, &spa
,
9055 FTAG
, policy
, NULL
);
9058 } else if (strpbrk(target
, "#") != NULL
) {
9060 error
= dsl_pool_hold(target
, FTAG
, &dp
);
9062 fatal("can't dump '%s': %s", target
,
9065 error
= dump_bookmark(dp
, target
, B_TRUE
, verbose
> 1);
9066 dsl_pool_rele(dp
, FTAG
);
9068 fatal("can't dump '%s': %s", target
,
9073 target_pool
= strdup(target
);
9074 if (strpbrk(target
, "/@") != NULL
)
9075 *strpbrk(target_pool
, "/@") = '\0';
9077 zdb_set_skip_mmp(target
);
9079 * If -N was supplied, the user has indicated that
9080 * zdb -d <pool>/<objsetID> is in effect. Otherwise
9081 * we first assume that the dataset string is the
9082 * dataset name. If dmu_objset_hold fails with the
9083 * dataset string, and we have an objset_id, retry the
9084 * lookup with the objsetID.
9086 boolean_t retry
= B_TRUE
;
9088 if (dataset_lookup
== B_TRUE
) {
9090 * Use the supplied id to get the name
9093 error
= spa_open(target_pool
, &spa
, FTAG
);
9095 error
= name_from_objset_id(spa
,
9097 spa_close(spa
, FTAG
);
9103 if (objset_id
> 0 && retry
) {
9104 int err
= dmu_objset_hold(target
, FTAG
,
9107 dataset_lookup
= B_TRUE
;
9111 dmu_objset_rele(os
, FTAG
);
9114 error
= open_objset(target
, FTAG
, &os
);
9117 spa
= dmu_objset_spa(os
);
9121 nvlist_free(policy
);
9124 fatal("can't open '%s': %s", target
, strerror(error
));
9127 * Set the pool failure mode to panic in order to prevent the pool
9128 * from suspending. A suspended I/O will have no way to resume and
9129 * can prevent the zdb(8) command from terminating as expected.
9132 spa
->spa_failmode
= ZIO_FAILURE_MODE_PANIC
;
9136 if (dump_opt
['r']) {
9137 error
= zdb_copy_object(os
, object
, argv
[1]);
9138 } else if (!dump_opt
['R']) {
9139 flagbits
['d'] = ZOR_FLAG_DIRECTORY
;
9140 flagbits
['f'] = ZOR_FLAG_PLAIN_FILE
;
9141 flagbits
['m'] = ZOR_FLAG_SPACE_MAP
;
9142 flagbits
['z'] = ZOR_FLAG_ZAP
;
9143 flagbits
['A'] = ZOR_FLAG_ALL_TYPES
;
9145 if (argc
> 0 && dump_opt
['d']) {
9146 zopt_object_args
= argc
;
9147 zopt_object_ranges
= calloc(zopt_object_args
,
9148 sizeof (zopt_object_range_t
));
9149 for (unsigned i
= 0; i
< zopt_object_args
; i
++) {
9151 const char *msg
= NULL
;
9153 err
= parse_object_range(argv
[i
],
9154 &zopt_object_ranges
[i
], &msg
);
9156 fatal("Bad object or range: '%s': %s\n",
9157 argv
[i
], msg
?: "");
9159 } else if (argc
> 0 && dump_opt
['m']) {
9160 zopt_metaslab_args
= argc
;
9161 zopt_metaslab
= calloc(zopt_metaslab_args
,
9163 for (unsigned i
= 0; i
< zopt_metaslab_args
; i
++) {
9165 zopt_metaslab
[i
] = strtoull(argv
[i
], NULL
, 0);
9166 if (zopt_metaslab
[i
] == 0 && errno
!= 0)
9167 fatal("bad number %s: %s", argv
[i
],
9173 } else if (zopt_object_args
> 0 && !dump_opt
['m']) {
9174 dump_objset(spa
->spa_meta_objset
);
9179 flagbits
['b'] = ZDB_FLAG_PRINT_BLKPTR
;
9180 flagbits
['c'] = ZDB_FLAG_CHECKSUM
;
9181 flagbits
['d'] = ZDB_FLAG_DECOMPRESS
;
9182 flagbits
['e'] = ZDB_FLAG_BSWAP
;
9183 flagbits
['g'] = ZDB_FLAG_GBH
;
9184 flagbits
['i'] = ZDB_FLAG_INDIRECT
;
9185 flagbits
['r'] = ZDB_FLAG_RAW
;
9186 flagbits
['v'] = ZDB_FLAG_VERBOSE
;
9188 for (int i
= 0; i
< argc
; i
++)
9189 zdb_read_block(argv
[i
], spa
);
9192 if (dump_opt
['k']) {
9193 free(checkpoint_pool
);
9195 free(checkpoint_target
);
9199 close_objset(os
, FTAG
);
9201 spa_close(spa
, FTAG
);
9204 fuid_table_destroy();
9206 dump_debug_buffer();