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
) {
3325 boolean_t can_print
= !dump_opt
['P'];
3329 (void) printf("\t\t%s = ", nvpair_name(elem
));
3330 nvpair_value_byte_array(elem
, &value
, &cnt
);
3332 for (idx
= 0; idx
< cnt
; ++idx
) {
3333 if (!isprint(value
[idx
])) {
3334 can_print
= B_FALSE
;
3339 for (idx
= 0; idx
< cnt
; ++idx
) {
3341 (void) putchar(value
[idx
]);
3343 (void) printf("\\%3.3o", value
[idx
]);
3345 (void) putchar('\n');
3348 nvlist_free(sa_xattr
);
3349 free(sa_xattr_packed
);
3353 dump_znode_symlink(sa_handle_t
*hdl
)
3355 int sa_symlink_size
= 0;
3356 char linktarget
[MAXPATHLEN
];
3359 error
= sa_size(hdl
, sa_attr_table
[ZPL_SYMLINK
], &sa_symlink_size
);
3360 if (error
|| sa_symlink_size
== 0) {
3363 if (sa_symlink_size
>= sizeof (linktarget
)) {
3364 (void) printf("symlink size %d is too large\n",
3368 linktarget
[sa_symlink_size
] = '\0';
3369 if (sa_lookup(hdl
, sa_attr_table
[ZPL_SYMLINK
],
3370 &linktarget
, sa_symlink_size
) == 0)
3371 (void) printf("\ttarget %s\n", linktarget
);
3375 dump_znode(objset_t
*os
, uint64_t object
, void *data
, size_t size
)
3377 (void) data
, (void) size
;
3378 char path
[MAXPATHLEN
* 2]; /* allow for xattr and failure prefix */
3380 uint64_t xattr
, rdev
, gen
;
3381 uint64_t uid
, gid
, mode
, fsize
, parent
, links
;
3383 uint64_t acctm
[2], modtm
[2], chgtm
[2], crtm
[2];
3384 time_t z_crtime
, z_atime
, z_mtime
, z_ctime
;
3385 sa_bulk_attr_t bulk
[12];
3389 VERIFY3P(os
, ==, sa_os
);
3390 if (sa_handle_get(os
, object
, NULL
, SA_HDL_PRIVATE
, &hdl
)) {
3391 (void) printf("Failed to get handle for SA znode\n");
3395 SA_ADD_BULK_ATTR(bulk
, idx
, sa_attr_table
[ZPL_UID
], NULL
, &uid
, 8);
3396 SA_ADD_BULK_ATTR(bulk
, idx
, sa_attr_table
[ZPL_GID
], NULL
, &gid
, 8);
3397 SA_ADD_BULK_ATTR(bulk
, idx
, sa_attr_table
[ZPL_LINKS
], NULL
,
3399 SA_ADD_BULK_ATTR(bulk
, idx
, sa_attr_table
[ZPL_GEN
], NULL
, &gen
, 8);
3400 SA_ADD_BULK_ATTR(bulk
, idx
, sa_attr_table
[ZPL_MODE
], NULL
,
3402 SA_ADD_BULK_ATTR(bulk
, idx
, sa_attr_table
[ZPL_PARENT
],
3404 SA_ADD_BULK_ATTR(bulk
, idx
, sa_attr_table
[ZPL_SIZE
], NULL
,
3406 SA_ADD_BULK_ATTR(bulk
, idx
, sa_attr_table
[ZPL_ATIME
], NULL
,
3408 SA_ADD_BULK_ATTR(bulk
, idx
, sa_attr_table
[ZPL_MTIME
], NULL
,
3410 SA_ADD_BULK_ATTR(bulk
, idx
, sa_attr_table
[ZPL_CRTIME
], NULL
,
3412 SA_ADD_BULK_ATTR(bulk
, idx
, sa_attr_table
[ZPL_CTIME
], NULL
,
3414 SA_ADD_BULK_ATTR(bulk
, idx
, sa_attr_table
[ZPL_FLAGS
], NULL
,
3417 if (sa_bulk_lookup(hdl
, bulk
, idx
)) {
3418 (void) sa_handle_destroy(hdl
);
3422 z_crtime
= (time_t)crtm
[0];
3423 z_atime
= (time_t)acctm
[0];
3424 z_mtime
= (time_t)modtm
[0];
3425 z_ctime
= (time_t)chgtm
[0];
3427 if (dump_opt
['d'] > 4) {
3428 error
= zfs_obj_to_path(os
, object
, path
, sizeof (path
));
3429 if (error
== ESTALE
) {
3430 (void) snprintf(path
, sizeof (path
), "on delete queue");
3431 } else if (error
!= 0) {
3433 (void) snprintf(path
, sizeof (path
),
3434 "path not found, possibly leaked");
3436 (void) printf("\tpath %s\n", path
);
3440 dump_znode_symlink(hdl
);
3441 dump_uidgid(os
, uid
, gid
);
3442 (void) printf("\tatime %s", ctime(&z_atime
));
3443 (void) printf("\tmtime %s", ctime(&z_mtime
));
3444 (void) printf("\tctime %s", ctime(&z_ctime
));
3445 (void) printf("\tcrtime %s", ctime(&z_crtime
));
3446 (void) printf("\tgen %llu\n", (u_longlong_t
)gen
);
3447 (void) printf("\tmode %llo\n", (u_longlong_t
)mode
);
3448 (void) printf("\tsize %llu\n", (u_longlong_t
)fsize
);
3449 (void) printf("\tparent %llu\n", (u_longlong_t
)parent
);
3450 (void) printf("\tlinks %llu\n", (u_longlong_t
)links
);
3451 (void) printf("\tpflags %llx\n", (u_longlong_t
)pflags
);
3452 if (dmu_objset_projectquota_enabled(os
) && (pflags
& ZFS_PROJID
)) {
3455 if (sa_lookup(hdl
, sa_attr_table
[ZPL_PROJID
], &projid
,
3456 sizeof (uint64_t)) == 0)
3457 (void) printf("\tprojid %llu\n", (u_longlong_t
)projid
);
3459 if (sa_lookup(hdl
, sa_attr_table
[ZPL_XATTR
], &xattr
,
3460 sizeof (uint64_t)) == 0)
3461 (void) printf("\txattr %llu\n", (u_longlong_t
)xattr
);
3462 if (sa_lookup(hdl
, sa_attr_table
[ZPL_RDEV
], &rdev
,
3463 sizeof (uint64_t)) == 0)
3464 (void) printf("\trdev 0x%016llx\n", (u_longlong_t
)rdev
);
3465 dump_znode_sa_xattr(hdl
);
3466 sa_handle_destroy(hdl
);
3470 dump_acl(objset_t
*os
, uint64_t object
, void *data
, size_t size
)
3472 (void) os
, (void) object
, (void) data
, (void) size
;
3476 dump_dmu_objset(objset_t
*os
, uint64_t object
, void *data
, size_t size
)
3478 (void) os
, (void) object
, (void) data
, (void) size
;
3481 static object_viewer_t
*object_viewer
[DMU_OT_NUMTYPES
+ 1] = {
3482 dump_none
, /* unallocated */
3483 dump_zap
, /* object directory */
3484 dump_uint64
, /* object array */
3485 dump_none
, /* packed nvlist */
3486 dump_packed_nvlist
, /* packed nvlist size */
3487 dump_none
, /* bpobj */
3488 dump_bpobj
, /* bpobj header */
3489 dump_none
, /* SPA space map header */
3490 dump_none
, /* SPA space map */
3491 dump_none
, /* ZIL intent log */
3492 dump_dnode
, /* DMU dnode */
3493 dump_dmu_objset
, /* DMU objset */
3494 dump_dsl_dir
, /* DSL directory */
3495 dump_zap
, /* DSL directory child map */
3496 dump_zap
, /* DSL dataset snap map */
3497 dump_zap
, /* DSL props */
3498 dump_dsl_dataset
, /* DSL dataset */
3499 dump_znode
, /* ZFS znode */
3500 dump_acl
, /* ZFS V0 ACL */
3501 dump_uint8
, /* ZFS plain file */
3502 dump_zpldir
, /* ZFS directory */
3503 dump_zap
, /* ZFS master node */
3504 dump_zap
, /* ZFS delete queue */
3505 dump_uint8
, /* zvol object */
3506 dump_zap
, /* zvol prop */
3507 dump_uint8
, /* other uint8[] */
3508 dump_uint64
, /* other uint64[] */
3509 dump_zap
, /* other ZAP */
3510 dump_zap
, /* persistent error log */
3511 dump_uint8
, /* SPA history */
3512 dump_history_offsets
, /* SPA history offsets */
3513 dump_zap
, /* Pool properties */
3514 dump_zap
, /* DSL permissions */
3515 dump_acl
, /* ZFS ACL */
3516 dump_uint8
, /* ZFS SYSACL */
3517 dump_none
, /* FUID nvlist */
3518 dump_packed_nvlist
, /* FUID nvlist size */
3519 dump_zap
, /* DSL dataset next clones */
3520 dump_zap
, /* DSL scrub queue */
3521 dump_zap
, /* ZFS user/group/project used */
3522 dump_zap
, /* ZFS user/group/project quota */
3523 dump_zap
, /* snapshot refcount tags */
3524 dump_ddt_zap
, /* DDT ZAP object */
3525 dump_zap
, /* DDT statistics */
3526 dump_znode
, /* SA object */
3527 dump_zap
, /* SA Master Node */
3528 dump_sa_attrs
, /* SA attribute registration */
3529 dump_sa_layouts
, /* SA attribute layouts */
3530 dump_zap
, /* DSL scrub translations */
3531 dump_none
, /* fake dedup BP */
3532 dump_zap
, /* deadlist */
3533 dump_none
, /* deadlist hdr */
3534 dump_zap
, /* dsl clones */
3535 dump_bpobj_subobjs
, /* bpobj subobjs */
3536 dump_unknown
, /* Unknown type, must be last */
3540 match_object_type(dmu_object_type_t obj_type
, uint64_t flags
)
3542 boolean_t match
= B_TRUE
;
3545 case DMU_OT_DIRECTORY_CONTENTS
:
3546 if (!(flags
& ZOR_FLAG_DIRECTORY
))
3549 case DMU_OT_PLAIN_FILE_CONTENTS
:
3550 if (!(flags
& ZOR_FLAG_PLAIN_FILE
))
3553 case DMU_OT_SPACE_MAP
:
3554 if (!(flags
& ZOR_FLAG_SPACE_MAP
))
3558 if (strcmp(zdb_ot_name(obj_type
), "zap") == 0) {
3559 if (!(flags
& ZOR_FLAG_ZAP
))
3565 * If all bits except some of the supported flags are
3566 * set, the user combined the all-types flag (A) with
3567 * a negated flag to exclude some types (e.g. A-f to
3568 * show all object types except plain files).
3570 if ((flags
| ZOR_SUPPORTED_FLAGS
) != ZOR_FLAG_ALL_TYPES
)
3580 dump_object(objset_t
*os
, uint64_t object
, int verbosity
,
3581 boolean_t
*print_header
, uint64_t *dnode_slots_used
, uint64_t flags
)
3583 dmu_buf_t
*db
= NULL
;
3584 dmu_object_info_t doi
;
3586 boolean_t dnode_held
= B_FALSE
;
3589 char iblk
[32], dblk
[32], lsize
[32], asize
[32], fill
[32], dnsize
[32];
3590 char bonus_size
[32];
3594 /* make sure nicenum has enough space */
3595 _Static_assert(sizeof (iblk
) >= NN_NUMBUF_SZ
, "iblk truncated");
3596 _Static_assert(sizeof (dblk
) >= NN_NUMBUF_SZ
, "dblk truncated");
3597 _Static_assert(sizeof (lsize
) >= NN_NUMBUF_SZ
, "lsize truncated");
3598 _Static_assert(sizeof (asize
) >= NN_NUMBUF_SZ
, "asize truncated");
3599 _Static_assert(sizeof (bonus_size
) >= NN_NUMBUF_SZ
,
3600 "bonus_size truncated");
3602 if (*print_header
) {
3603 (void) printf("\n%10s %3s %5s %5s %5s %6s %5s %6s %s\n",
3604 "Object", "lvl", "iblk", "dblk", "dsize", "dnsize",
3605 "lsize", "%full", "type");
3610 dn
= DMU_META_DNODE(os
);
3611 dmu_object_info_from_dnode(dn
, &doi
);
3614 * Encrypted datasets will have sensitive bonus buffers
3615 * encrypted. Therefore we cannot hold the bonus buffer and
3616 * must hold the dnode itself instead.
3618 error
= dmu_object_info(os
, object
, &doi
);
3620 fatal("dmu_object_info() failed, errno %u", error
);
3622 if (!key_loaded
&& os
->os_encrypted
&&
3623 DMU_OT_IS_ENCRYPTED(doi
.doi_bonus_type
)) {
3624 error
= dnode_hold(os
, object
, FTAG
, &dn
);
3626 fatal("dnode_hold() failed, errno %u", error
);
3627 dnode_held
= B_TRUE
;
3629 error
= dmu_bonus_hold(os
, object
, FTAG
, &db
);
3631 fatal("dmu_bonus_hold(%llu) failed, errno %u",
3633 bonus
= db
->db_data
;
3634 bsize
= db
->db_size
;
3635 dn
= DB_DNODE((dmu_buf_impl_t
*)db
);
3640 * Default to showing all object types if no flags were specified.
3642 if (flags
!= 0 && flags
!= ZOR_FLAG_ALL_TYPES
&&
3643 !match_object_type(doi
.doi_type
, flags
))
3646 if (dnode_slots_used
)
3647 *dnode_slots_used
= doi
.doi_dnodesize
/ DNODE_MIN_SIZE
;
3649 zdb_nicenum(doi
.doi_metadata_block_size
, iblk
, sizeof (iblk
));
3650 zdb_nicenum(doi
.doi_data_block_size
, dblk
, sizeof (dblk
));
3651 zdb_nicenum(doi
.doi_max_offset
, lsize
, sizeof (lsize
));
3652 zdb_nicenum(doi
.doi_physical_blocks_512
<< 9, asize
, sizeof (asize
));
3653 zdb_nicenum(doi
.doi_bonus_size
, bonus_size
, sizeof (bonus_size
));
3654 zdb_nicenum(doi
.doi_dnodesize
, dnsize
, sizeof (dnsize
));
3655 (void) snprintf(fill
, sizeof (fill
), "%6.2f", 100.0 *
3656 doi
.doi_fill_count
* doi
.doi_data_block_size
/ (object
== 0 ?
3657 DNODES_PER_BLOCK
: 1) / doi
.doi_max_offset
);
3661 if (doi
.doi_checksum
!= ZIO_CHECKSUM_INHERIT
|| verbosity
>= 6) {
3662 (void) snprintf(aux
+ strlen(aux
), sizeof (aux
) - strlen(aux
),
3663 " (K=%s)", ZDB_CHECKSUM_NAME(doi
.doi_checksum
));
3666 if (doi
.doi_compress
== ZIO_COMPRESS_INHERIT
&&
3667 ZIO_COMPRESS_HASLEVEL(os
->os_compress
) && verbosity
>= 6) {
3668 const char *compname
= NULL
;
3669 if (zfs_prop_index_to_string(ZFS_PROP_COMPRESSION
,
3670 ZIO_COMPRESS_RAW(os
->os_compress
, os
->os_complevel
),
3672 (void) snprintf(aux
+ strlen(aux
),
3673 sizeof (aux
) - strlen(aux
), " (Z=inherit=%s)",
3676 (void) snprintf(aux
+ strlen(aux
),
3677 sizeof (aux
) - strlen(aux
),
3678 " (Z=inherit=%s-unknown)",
3679 ZDB_COMPRESS_NAME(os
->os_compress
));
3681 } else if (doi
.doi_compress
== ZIO_COMPRESS_INHERIT
&& verbosity
>= 6) {
3682 (void) snprintf(aux
+ strlen(aux
), sizeof (aux
) - strlen(aux
),
3683 " (Z=inherit=%s)", ZDB_COMPRESS_NAME(os
->os_compress
));
3684 } else if (doi
.doi_compress
!= ZIO_COMPRESS_INHERIT
|| verbosity
>= 6) {
3685 (void) snprintf(aux
+ strlen(aux
), sizeof (aux
) - strlen(aux
),
3686 " (Z=%s)", ZDB_COMPRESS_NAME(doi
.doi_compress
));
3689 (void) printf("%10lld %3u %5s %5s %5s %6s %5s %6s %s%s\n",
3690 (u_longlong_t
)object
, doi
.doi_indirection
, iblk
, dblk
,
3691 asize
, dnsize
, lsize
, fill
, zdb_ot_name(doi
.doi_type
), aux
);
3693 if (doi
.doi_bonus_type
!= DMU_OT_NONE
&& verbosity
> 3) {
3694 (void) printf("%10s %3s %5s %5s %5s %5s %5s %6s %s\n",
3695 "", "", "", "", "", "", bonus_size
, "bonus",
3696 zdb_ot_name(doi
.doi_bonus_type
));
3699 if (verbosity
>= 4) {
3700 (void) printf("\tdnode flags: %s%s%s%s\n",
3701 (dn
->dn_phys
->dn_flags
& DNODE_FLAG_USED_BYTES
) ?
3703 (dn
->dn_phys
->dn_flags
& DNODE_FLAG_USERUSED_ACCOUNTED
) ?
3704 "USERUSED_ACCOUNTED " : "",
3705 (dn
->dn_phys
->dn_flags
& DNODE_FLAG_USEROBJUSED_ACCOUNTED
) ?
3706 "USEROBJUSED_ACCOUNTED " : "",
3707 (dn
->dn_phys
->dn_flags
& DNODE_FLAG_SPILL_BLKPTR
) ?
3708 "SPILL_BLKPTR" : "");
3709 (void) printf("\tdnode maxblkid: %llu\n",
3710 (longlong_t
)dn
->dn_phys
->dn_maxblkid
);
3713 object_viewer
[ZDB_OT_TYPE(doi
.doi_bonus_type
)](os
,
3714 object
, bonus
, bsize
);
3716 (void) printf("\t\t(bonus encrypted)\n");
3720 (!os
->os_encrypted
|| !DMU_OT_IS_ENCRYPTED(doi
.doi_type
))) {
3721 object_viewer
[ZDB_OT_TYPE(doi
.doi_type
)](os
, object
,
3724 (void) printf("\t\t(object encrypted)\n");
3727 *print_header
= B_TRUE
;
3730 if (verbosity
>= 5) {
3731 if (dn
->dn_phys
->dn_flags
& DNODE_FLAG_SPILL_BLKPTR
) {
3732 char blkbuf
[BP_SPRINTF_LEN
];
3733 snprintf_blkptr_compact(blkbuf
, sizeof (blkbuf
),
3734 DN_SPILL_BLKPTR(dn
->dn_phys
), B_FALSE
);
3735 (void) printf("\nSpill block: %s\n", blkbuf
);
3740 if (verbosity
>= 5) {
3742 * Report the list of segments that comprise the object.
3746 uint64_t blkfill
= 1;
3749 if (dn
->dn_type
== DMU_OT_DNODE
) {
3751 blkfill
= DNODES_PER_BLOCK
;
3756 /* make sure nicenum has enough space */
3757 _Static_assert(sizeof (segsize
) >= NN_NUMBUF_SZ
,
3758 "segsize truncated");
3759 error
= dnode_next_offset(dn
,
3760 0, &start
, minlvl
, blkfill
, 0);
3764 error
= dnode_next_offset(dn
,
3765 DNODE_FIND_HOLE
, &end
, minlvl
, blkfill
, 0);
3766 zdb_nicenum(end
- start
, segsize
, sizeof (segsize
));
3767 (void) printf("\t\tsegment [%016llx, %016llx)"
3768 " size %5s\n", (u_longlong_t
)start
,
3769 (u_longlong_t
)end
, segsize
);
3778 dmu_buf_rele(db
, FTAG
);
3780 dnode_rele(dn
, FTAG
);
3784 count_dir_mos_objects(dsl_dir_t
*dd
)
3786 mos_obj_refd(dd
->dd_object
);
3787 mos_obj_refd(dsl_dir_phys(dd
)->dd_child_dir_zapobj
);
3788 mos_obj_refd(dsl_dir_phys(dd
)->dd_deleg_zapobj
);
3789 mos_obj_refd(dsl_dir_phys(dd
)->dd_props_zapobj
);
3790 mos_obj_refd(dsl_dir_phys(dd
)->dd_clones
);
3793 * The dd_crypto_obj can be referenced by multiple dsl_dir's.
3794 * Ignore the references after the first one.
3796 mos_obj_refd_multiple(dd
->dd_crypto_obj
);
3800 count_ds_mos_objects(dsl_dataset_t
*ds
)
3802 mos_obj_refd(ds
->ds_object
);
3803 mos_obj_refd(dsl_dataset_phys(ds
)->ds_next_clones_obj
);
3804 mos_obj_refd(dsl_dataset_phys(ds
)->ds_props_obj
);
3805 mos_obj_refd(dsl_dataset_phys(ds
)->ds_userrefs_obj
);
3806 mos_obj_refd(dsl_dataset_phys(ds
)->ds_snapnames_zapobj
);
3807 mos_obj_refd(ds
->ds_bookmarks_obj
);
3809 if (!dsl_dataset_is_snapshot(ds
)) {
3810 count_dir_mos_objects(ds
->ds_dir
);
3814 static const char *const objset_types
[DMU_OST_NUMTYPES
] = {
3815 "NONE", "META", "ZPL", "ZVOL", "OTHER", "ANY" };
3818 * Parse a string denoting a range of object IDs of the form
3819 * <start>[:<end>[:flags]], and store the results in zor.
3820 * Return 0 on success. On error, return 1 and update the msg
3821 * pointer to point to a descriptive error message.
3824 parse_object_range(char *range
, zopt_object_range_t
*zor
, const char **msg
)
3827 char *p
, *s
, *dup
, *flagstr
, *tmp
= NULL
;
3832 if (strchr(range
, ':') == NULL
) {
3833 zor
->zor_obj_start
= strtoull(range
, &p
, 0);
3835 *msg
= "Invalid characters in object ID";
3838 zor
->zor_obj_start
= ZDB_MAP_OBJECT_ID(zor
->zor_obj_start
);
3839 zor
->zor_obj_end
= zor
->zor_obj_start
;
3843 if (strchr(range
, ':') == range
) {
3844 *msg
= "Invalid leading colon";
3849 len
= strlen(range
);
3850 if (range
[len
- 1] == ':') {
3851 *msg
= "Invalid trailing colon";
3856 dup
= strdup(range
);
3857 s
= strtok_r(dup
, ":", &tmp
);
3858 zor
->zor_obj_start
= strtoull(s
, &p
, 0);
3861 *msg
= "Invalid characters in start object ID";
3866 s
= strtok_r(NULL
, ":", &tmp
);
3867 zor
->zor_obj_end
= strtoull(s
, &p
, 0);
3870 *msg
= "Invalid characters in end object ID";
3875 if (zor
->zor_obj_start
> zor
->zor_obj_end
) {
3876 *msg
= "Start object ID may not exceed end object ID";
3881 s
= strtok_r(NULL
, ":", &tmp
);
3883 zor
->zor_flags
= ZOR_FLAG_ALL_TYPES
;
3885 } else if (strtok_r(NULL
, ":", &tmp
) != NULL
) {
3886 *msg
= "Invalid colon-delimited field after flags";
3892 for (i
= 0; flagstr
[i
]; i
++) {
3894 boolean_t negation
= (flagstr
[i
] == '-');
3898 if (flagstr
[i
] == '\0') {
3899 *msg
= "Invalid trailing negation operator";
3904 bit
= flagbits
[(uchar_t
)flagstr
[i
]];
3906 *msg
= "Invalid flag";
3915 zor
->zor_flags
= flags
;
3917 zor
->zor_obj_start
= ZDB_MAP_OBJECT_ID(zor
->zor_obj_start
);
3918 zor
->zor_obj_end
= ZDB_MAP_OBJECT_ID(zor
->zor_obj_end
);
3926 dump_objset(objset_t
*os
)
3928 dmu_objset_stats_t dds
= { 0 };
3929 uint64_t object
, object_count
;
3930 uint64_t refdbytes
, usedobjs
, scratch
;
3932 char blkbuf
[BP_SPRINTF_LEN
+ 20];
3933 char osname
[ZFS_MAX_DATASET_NAME_LEN
];
3934 const char *type
= "UNKNOWN";
3935 int verbosity
= dump_opt
['d'];
3936 boolean_t print_header
;
3939 uint64_t total_slots_used
= 0;
3940 uint64_t max_slot_used
= 0;
3941 uint64_t dnode_slots
;
3946 /* make sure nicenum has enough space */
3947 _Static_assert(sizeof (numbuf
) >= NN_NUMBUF_SZ
, "numbuf truncated");
3949 dsl_pool_config_enter(dmu_objset_pool(os
), FTAG
);
3950 dmu_objset_fast_stat(os
, &dds
);
3951 dsl_pool_config_exit(dmu_objset_pool(os
), FTAG
);
3953 print_header
= B_TRUE
;
3955 if (dds
.dds_type
< DMU_OST_NUMTYPES
)
3956 type
= objset_types
[dds
.dds_type
];
3958 if (dds
.dds_type
== DMU_OST_META
) {
3959 dds
.dds_creation_txg
= TXG_INITIAL
;
3960 usedobjs
= BP_GET_FILL(os
->os_rootbp
);
3961 refdbytes
= dsl_dir_phys(os
->os_spa
->spa_dsl_pool
->dp_mos_dir
)->
3964 dmu_objset_space(os
, &refdbytes
, &scratch
, &usedobjs
, &scratch
);
3967 ASSERT3U(usedobjs
, ==, BP_GET_FILL(os
->os_rootbp
));
3969 zdb_nicenum(refdbytes
, numbuf
, sizeof (numbuf
));
3971 if (verbosity
>= 4) {
3972 (void) snprintf(blkbuf
, sizeof (blkbuf
), ", rootbp ");
3973 (void) snprintf_blkptr(blkbuf
+ strlen(blkbuf
),
3974 sizeof (blkbuf
) - strlen(blkbuf
), os
->os_rootbp
);
3979 dmu_objset_name(os
, osname
);
3981 (void) printf("Dataset %s [%s], ID %llu, cr_txg %llu, "
3982 "%s, %llu objects%s%s\n",
3983 osname
, type
, (u_longlong_t
)dmu_objset_id(os
),
3984 (u_longlong_t
)dds
.dds_creation_txg
,
3985 numbuf
, (u_longlong_t
)usedobjs
, blkbuf
,
3986 (dds
.dds_inconsistent
) ? " (inconsistent)" : "");
3988 for (i
= 0; i
< zopt_object_args
; i
++) {
3989 obj_start
= zopt_object_ranges
[i
].zor_obj_start
;
3990 obj_end
= zopt_object_ranges
[i
].zor_obj_end
;
3991 flags
= zopt_object_ranges
[i
].zor_flags
;
3994 if (object
== 0 || obj_start
== obj_end
)
3995 dump_object(os
, object
, verbosity
, &print_header
, NULL
,
4000 while ((dmu_object_next(os
, &object
, B_FALSE
, 0) == 0) &&
4001 object
<= obj_end
) {
4002 dump_object(os
, object
, verbosity
, &print_header
, NULL
,
4007 if (zopt_object_args
> 0) {
4008 (void) printf("\n");
4012 if (dump_opt
['i'] != 0 || verbosity
>= 2)
4013 dump_intent_log(dmu_objset_zil(os
));
4015 if (dmu_objset_ds(os
) != NULL
) {
4016 dsl_dataset_t
*ds
= dmu_objset_ds(os
);
4017 dump_blkptr_list(&ds
->ds_deadlist
, "Deadlist");
4018 if (dsl_deadlist_is_open(&ds
->ds_dir
->dd_livelist
) &&
4019 !dmu_objset_is_snapshot(os
)) {
4020 dump_blkptr_list(&ds
->ds_dir
->dd_livelist
, "Livelist");
4021 if (verify_dd_livelist(os
) != 0)
4022 fatal("livelist is incorrect");
4025 if (dsl_dataset_remap_deadlist_exists(ds
)) {
4026 (void) printf("ds_remap_deadlist:\n");
4027 dump_blkptr_list(&ds
->ds_remap_deadlist
, "Deadlist");
4029 count_ds_mos_objects(ds
);
4032 if (dmu_objset_ds(os
) != NULL
)
4033 dump_bookmarks(os
, verbosity
);
4038 if (BP_IS_HOLE(os
->os_rootbp
))
4041 dump_object(os
, 0, verbosity
, &print_header
, NULL
, 0);
4043 if (DMU_USERUSED_DNODE(os
) != NULL
&&
4044 DMU_USERUSED_DNODE(os
)->dn_type
!= 0) {
4045 dump_object(os
, DMU_USERUSED_OBJECT
, verbosity
, &print_header
,
4047 dump_object(os
, DMU_GROUPUSED_OBJECT
, verbosity
, &print_header
,
4051 if (DMU_PROJECTUSED_DNODE(os
) != NULL
&&
4052 DMU_PROJECTUSED_DNODE(os
)->dn_type
!= 0)
4053 dump_object(os
, DMU_PROJECTUSED_OBJECT
, verbosity
,
4054 &print_header
, NULL
, 0);
4057 while ((error
= dmu_object_next(os
, &object
, B_FALSE
, 0)) == 0) {
4058 dump_object(os
, object
, verbosity
, &print_header
, &dnode_slots
,
4061 total_slots_used
+= dnode_slots
;
4062 max_slot_used
= object
+ dnode_slots
- 1;
4065 (void) printf("\n");
4067 (void) printf(" Dnode slots:\n");
4068 (void) printf("\tTotal used: %10llu\n",
4069 (u_longlong_t
)total_slots_used
);
4070 (void) printf("\tMax used: %10llu\n",
4071 (u_longlong_t
)max_slot_used
);
4072 (void) printf("\tPercent empty: %10lf\n",
4073 (double)(max_slot_used
- total_slots_used
)*100 /
4074 (double)max_slot_used
);
4075 (void) printf("\n");
4077 if (error
!= ESRCH
) {
4078 (void) fprintf(stderr
, "dmu_object_next() = %d\n", error
);
4082 ASSERT3U(object_count
, ==, usedobjs
);
4084 if (leaked_objects
!= 0) {
4085 (void) printf("%d potentially leaked objects detected\n",
4092 dump_uberblock(uberblock_t
*ub
, const char *header
, const char *footer
)
4094 time_t timestamp
= ub
->ub_timestamp
;
4096 (void) printf("%s", header
? header
: "");
4097 (void) printf("\tmagic = %016llx\n", (u_longlong_t
)ub
->ub_magic
);
4098 (void) printf("\tversion = %llu\n", (u_longlong_t
)ub
->ub_version
);
4099 (void) printf("\ttxg = %llu\n", (u_longlong_t
)ub
->ub_txg
);
4100 (void) printf("\tguid_sum = %llu\n", (u_longlong_t
)ub
->ub_guid_sum
);
4101 (void) printf("\ttimestamp = %llu UTC = %s",
4102 (u_longlong_t
)ub
->ub_timestamp
, ctime(×tamp
));
4104 (void) printf("\tmmp_magic = %016llx\n",
4105 (u_longlong_t
)ub
->ub_mmp_magic
);
4106 if (MMP_VALID(ub
)) {
4107 (void) printf("\tmmp_delay = %0llu\n",
4108 (u_longlong_t
)ub
->ub_mmp_delay
);
4109 if (MMP_SEQ_VALID(ub
))
4110 (void) printf("\tmmp_seq = %u\n",
4111 (unsigned int) MMP_SEQ(ub
));
4112 if (MMP_FAIL_INT_VALID(ub
))
4113 (void) printf("\tmmp_fail = %u\n",
4114 (unsigned int) MMP_FAIL_INT(ub
));
4115 if (MMP_INTERVAL_VALID(ub
))
4116 (void) printf("\tmmp_write = %u\n",
4117 (unsigned int) MMP_INTERVAL(ub
));
4118 /* After MMP_* to make summarize_uberblock_mmp cleaner */
4119 (void) printf("\tmmp_valid = %x\n",
4120 (unsigned int) ub
->ub_mmp_config
& 0xFF);
4123 if (dump_opt
['u'] >= 4) {
4124 char blkbuf
[BP_SPRINTF_LEN
];
4125 snprintf_blkptr(blkbuf
, sizeof (blkbuf
), &ub
->ub_rootbp
);
4126 (void) printf("\trootbp = %s\n", blkbuf
);
4128 (void) printf("\tcheckpoint_txg = %llu\n",
4129 (u_longlong_t
)ub
->ub_checkpoint_txg
);
4130 (void) printf("%s", footer
? footer
: "");
4134 dump_config(spa_t
*spa
)
4141 error
= dmu_bonus_hold(spa
->spa_meta_objset
,
4142 spa
->spa_config_object
, FTAG
, &db
);
4145 nvsize
= *(uint64_t *)db
->db_data
;
4146 dmu_buf_rele(db
, FTAG
);
4148 (void) printf("\nMOS Configuration:\n");
4149 dump_packed_nvlist(spa
->spa_meta_objset
,
4150 spa
->spa_config_object
, (void *)&nvsize
, 1);
4152 (void) fprintf(stderr
, "dmu_bonus_hold(%llu) failed, errno %d",
4153 (u_longlong_t
)spa
->spa_config_object
, error
);
4158 dump_cachefile(const char *cachefile
)
4161 struct stat64 statbuf
;
4165 if ((fd
= open64(cachefile
, O_RDONLY
)) < 0) {
4166 (void) printf("cannot open '%s': %s\n", cachefile
,
4171 if (fstat64(fd
, &statbuf
) != 0) {
4172 (void) printf("failed to stat '%s': %s\n", cachefile
,
4177 if ((buf
= malloc(statbuf
.st_size
)) == NULL
) {
4178 (void) fprintf(stderr
, "failed to allocate %llu bytes\n",
4179 (u_longlong_t
)statbuf
.st_size
);
4183 if (read(fd
, buf
, statbuf
.st_size
) != statbuf
.st_size
) {
4184 (void) fprintf(stderr
, "failed to read %llu bytes\n",
4185 (u_longlong_t
)statbuf
.st_size
);
4191 if (nvlist_unpack(buf
, statbuf
.st_size
, &config
, 0) != 0) {
4192 (void) fprintf(stderr
, "failed to unpack nvlist\n");
4198 dump_nvlist(config
, 0);
4200 nvlist_free(config
);
4204 * ZFS label nvlist stats
4206 typedef struct zdb_nvl_stats
{
4209 size_t zns_leaf_largest
;
4210 size_t zns_leaf_total
;
4211 nvlist_t
*zns_string
;
4212 nvlist_t
*zns_uint64
;
4213 nvlist_t
*zns_boolean
;
4217 collect_nvlist_stats(nvlist_t
*nvl
, zdb_nvl_stats_t
*stats
)
4219 nvlist_t
*list
, **array
;
4220 nvpair_t
*nvp
= NULL
;
4224 stats
->zns_list_count
++;
4226 while ((nvp
= nvlist_next_nvpair(nvl
, nvp
)) != NULL
) {
4227 name
= nvpair_name(nvp
);
4229 switch (nvpair_type(nvp
)) {
4230 case DATA_TYPE_STRING
:
4231 fnvlist_add_string(stats
->zns_string
, name
,
4232 fnvpair_value_string(nvp
));
4234 case DATA_TYPE_UINT64
:
4235 fnvlist_add_uint64(stats
->zns_uint64
, name
,
4236 fnvpair_value_uint64(nvp
));
4238 case DATA_TYPE_BOOLEAN
:
4239 fnvlist_add_boolean(stats
->zns_boolean
, name
);
4241 case DATA_TYPE_NVLIST
:
4242 if (nvpair_value_nvlist(nvp
, &list
) == 0)
4243 collect_nvlist_stats(list
, stats
);
4245 case DATA_TYPE_NVLIST_ARRAY
:
4246 if (nvpair_value_nvlist_array(nvp
, &array
, &items
) != 0)
4249 for (i
= 0; i
< items
; i
++) {
4250 collect_nvlist_stats(array
[i
], stats
);
4252 /* collect stats on leaf vdev */
4253 if (strcmp(name
, "children") == 0) {
4256 (void) nvlist_size(array
[i
], &size
,
4258 stats
->zns_leaf_total
+= size
;
4259 if (size
> stats
->zns_leaf_largest
)
4260 stats
->zns_leaf_largest
= size
;
4261 stats
->zns_leaf_count
++;
4266 (void) printf("skip type %d!\n", (int)nvpair_type(nvp
));
4272 dump_nvlist_stats(nvlist_t
*nvl
, size_t cap
)
4274 zdb_nvl_stats_t stats
= { 0 };
4275 size_t size
, sum
= 0, total
;
4278 /* requires nvlist with non-unique names for stat collection */
4279 VERIFY0(nvlist_alloc(&stats
.zns_string
, 0, 0));
4280 VERIFY0(nvlist_alloc(&stats
.zns_uint64
, 0, 0));
4281 VERIFY0(nvlist_alloc(&stats
.zns_boolean
, 0, 0));
4282 VERIFY0(nvlist_size(stats
.zns_boolean
, &noise
, NV_ENCODE_XDR
));
4284 (void) printf("\n\nZFS Label NVList Config Stats:\n");
4286 VERIFY0(nvlist_size(nvl
, &total
, NV_ENCODE_XDR
));
4287 (void) printf(" %d bytes used, %d bytes free (using %4.1f%%)\n\n",
4288 (int)total
, (int)(cap
- total
), 100.0 * total
/ cap
);
4290 collect_nvlist_stats(nvl
, &stats
);
4292 VERIFY0(nvlist_size(stats
.zns_uint64
, &size
, NV_ENCODE_XDR
));
4295 (void) printf("%12s %4d %6d bytes (%5.2f%%)\n", "integers:",
4296 (int)fnvlist_num_pairs(stats
.zns_uint64
),
4297 (int)size
, 100.0 * size
/ total
);
4299 VERIFY0(nvlist_size(stats
.zns_string
, &size
, NV_ENCODE_XDR
));
4302 (void) printf("%12s %4d %6d bytes (%5.2f%%)\n", "strings:",
4303 (int)fnvlist_num_pairs(stats
.zns_string
),
4304 (int)size
, 100.0 * size
/ total
);
4306 VERIFY0(nvlist_size(stats
.zns_boolean
, &size
, NV_ENCODE_XDR
));
4309 (void) printf("%12s %4d %6d bytes (%5.2f%%)\n", "booleans:",
4310 (int)fnvlist_num_pairs(stats
.zns_boolean
),
4311 (int)size
, 100.0 * size
/ total
);
4313 size
= total
- sum
; /* treat remainder as nvlist overhead */
4314 (void) printf("%12s %4d %6d bytes (%5.2f%%)\n\n", "nvlists:",
4315 stats
.zns_list_count
, (int)size
, 100.0 * size
/ total
);
4317 if (stats
.zns_leaf_count
> 0) {
4318 size_t average
= stats
.zns_leaf_total
/ stats
.zns_leaf_count
;
4320 (void) printf("%12s %4d %6d bytes average\n", "leaf vdevs:",
4321 stats
.zns_leaf_count
, (int)average
);
4322 (void) printf("%24d bytes largest\n",
4323 (int)stats
.zns_leaf_largest
);
4325 if (dump_opt
['l'] >= 3 && average
> 0)
4326 (void) printf(" space for %d additional leaf vdevs\n",
4327 (int)((cap
- total
) / average
));
4329 (void) printf("\n");
4331 nvlist_free(stats
.zns_string
);
4332 nvlist_free(stats
.zns_uint64
);
4333 nvlist_free(stats
.zns_boolean
);
4336 typedef struct cksum_record
{
4338 boolean_t labels
[VDEV_LABELS
];
4343 cksum_record_compare(const void *x1
, const void *x2
)
4345 const cksum_record_t
*l
= (cksum_record_t
*)x1
;
4346 const cksum_record_t
*r
= (cksum_record_t
*)x2
;
4347 int arraysize
= ARRAY_SIZE(l
->cksum
.zc_word
);
4350 for (int i
= 0; i
< arraysize
; i
++) {
4351 difference
= TREE_CMP(l
->cksum
.zc_word
[i
], r
->cksum
.zc_word
[i
]);
4356 return (difference
);
4359 static cksum_record_t
*
4360 cksum_record_alloc(zio_cksum_t
*cksum
, int l
)
4362 cksum_record_t
*rec
;
4364 rec
= umem_zalloc(sizeof (*rec
), UMEM_NOFAIL
);
4365 rec
->cksum
= *cksum
;
4366 rec
->labels
[l
] = B_TRUE
;
4371 static cksum_record_t
*
4372 cksum_record_lookup(avl_tree_t
*tree
, zio_cksum_t
*cksum
)
4374 cksum_record_t lookup
= { .cksum
= *cksum
};
4377 return (avl_find(tree
, &lookup
, &where
));
4380 static cksum_record_t
*
4381 cksum_record_insert(avl_tree_t
*tree
, zio_cksum_t
*cksum
, int l
)
4383 cksum_record_t
*rec
;
4385 rec
= cksum_record_lookup(tree
, cksum
);
4387 rec
->labels
[l
] = B_TRUE
;
4389 rec
= cksum_record_alloc(cksum
, l
);
4397 first_label(cksum_record_t
*rec
)
4399 for (int i
= 0; i
< VDEV_LABELS
; i
++)
4407 print_label_numbers(const char *prefix
, const cksum_record_t
*rec
)
4409 fputs(prefix
, stdout
);
4410 for (int i
= 0; i
< VDEV_LABELS
; i
++)
4411 if (rec
->labels
[i
] == B_TRUE
)
4416 #define MAX_UBERBLOCK_COUNT (VDEV_UBERBLOCK_RING >> UBERBLOCK_SHIFT)
4418 typedef struct zdb_label
{
4420 uint64_t label_offset
;
4421 nvlist_t
*config_nv
;
4422 cksum_record_t
*config
;
4423 cksum_record_t
*uberblocks
[MAX_UBERBLOCK_COUNT
];
4424 boolean_t header_printed
;
4425 boolean_t read_failed
;
4426 boolean_t cksum_valid
;
4430 print_label_header(zdb_label_t
*label
, int l
)
4436 if (label
->header_printed
== B_TRUE
)
4439 (void) printf("------------------------------------\n");
4440 (void) printf("LABEL %d %s\n", l
,
4441 label
->cksum_valid
? "" : "(Bad label cksum)");
4442 (void) printf("------------------------------------\n");
4444 label
->header_printed
= B_TRUE
;
4448 print_l2arc_header(void)
4450 (void) printf("------------------------------------\n");
4451 (void) printf("L2ARC device header\n");
4452 (void) printf("------------------------------------\n");
4456 print_l2arc_log_blocks(void)
4458 (void) printf("------------------------------------\n");
4459 (void) printf("L2ARC device log blocks\n");
4460 (void) printf("------------------------------------\n");
4464 dump_l2arc_log_entries(uint64_t log_entries
,
4465 l2arc_log_ent_phys_t
*le
, uint64_t i
)
4467 for (int j
= 0; j
< log_entries
; j
++) {
4468 dva_t dva
= le
[j
].le_dva
;
4469 (void) printf("lb[%4llu]\tle[%4d]\tDVA asize: %llu, "
4470 "vdev: %llu, offset: %llu\n",
4471 (u_longlong_t
)i
, j
+ 1,
4472 (u_longlong_t
)DVA_GET_ASIZE(&dva
),
4473 (u_longlong_t
)DVA_GET_VDEV(&dva
),
4474 (u_longlong_t
)DVA_GET_OFFSET(&dva
));
4475 (void) printf("|\t\t\t\tbirth: %llu\n",
4476 (u_longlong_t
)le
[j
].le_birth
);
4477 (void) printf("|\t\t\t\tlsize: %llu\n",
4478 (u_longlong_t
)L2BLK_GET_LSIZE((&le
[j
])->le_prop
));
4479 (void) printf("|\t\t\t\tpsize: %llu\n",
4480 (u_longlong_t
)L2BLK_GET_PSIZE((&le
[j
])->le_prop
));
4481 (void) printf("|\t\t\t\tcompr: %llu\n",
4482 (u_longlong_t
)L2BLK_GET_COMPRESS((&le
[j
])->le_prop
));
4483 (void) printf("|\t\t\t\tcomplevel: %llu\n",
4484 (u_longlong_t
)(&le
[j
])->le_complevel
);
4485 (void) printf("|\t\t\t\ttype: %llu\n",
4486 (u_longlong_t
)L2BLK_GET_TYPE((&le
[j
])->le_prop
));
4487 (void) printf("|\t\t\t\tprotected: %llu\n",
4488 (u_longlong_t
)L2BLK_GET_PROTECTED((&le
[j
])->le_prop
));
4489 (void) printf("|\t\t\t\tprefetch: %llu\n",
4490 (u_longlong_t
)L2BLK_GET_PREFETCH((&le
[j
])->le_prop
));
4491 (void) printf("|\t\t\t\taddress: %llu\n",
4492 (u_longlong_t
)le
[j
].le_daddr
);
4493 (void) printf("|\t\t\t\tARC state: %llu\n",
4494 (u_longlong_t
)L2BLK_GET_STATE((&le
[j
])->le_prop
));
4495 (void) printf("|\n");
4497 (void) printf("\n");
4501 dump_l2arc_log_blkptr(const l2arc_log_blkptr_t
*lbps
)
4503 (void) printf("|\t\tdaddr: %llu\n", (u_longlong_t
)lbps
->lbp_daddr
);
4504 (void) printf("|\t\tpayload_asize: %llu\n",
4505 (u_longlong_t
)lbps
->lbp_payload_asize
);
4506 (void) printf("|\t\tpayload_start: %llu\n",
4507 (u_longlong_t
)lbps
->lbp_payload_start
);
4508 (void) printf("|\t\tlsize: %llu\n",
4509 (u_longlong_t
)L2BLK_GET_LSIZE(lbps
->lbp_prop
));
4510 (void) printf("|\t\tasize: %llu\n",
4511 (u_longlong_t
)L2BLK_GET_PSIZE(lbps
->lbp_prop
));
4512 (void) printf("|\t\tcompralgo: %llu\n",
4513 (u_longlong_t
)L2BLK_GET_COMPRESS(lbps
->lbp_prop
));
4514 (void) printf("|\t\tcksumalgo: %llu\n",
4515 (u_longlong_t
)L2BLK_GET_CHECKSUM(lbps
->lbp_prop
));
4516 (void) printf("|\n\n");
4520 dump_l2arc_log_blocks(int fd
, const l2arc_dev_hdr_phys_t
*l2dhdr
,
4521 l2arc_dev_hdr_phys_t
*rebuild
)
4523 l2arc_log_blk_phys_t this_lb
;
4525 l2arc_log_blkptr_t lbps
[2];
4532 print_l2arc_log_blocks();
4533 memcpy(lbps
, l2dhdr
->dh_start_lbps
, sizeof (lbps
));
4535 dev
.l2ad_evict
= l2dhdr
->dh_evict
;
4536 dev
.l2ad_start
= l2dhdr
->dh_start
;
4537 dev
.l2ad_end
= l2dhdr
->dh_end
;
4539 if (l2dhdr
->dh_start_lbps
[0].lbp_daddr
== 0) {
4540 /* no log blocks to read */
4541 if (!dump_opt
['q']) {
4542 (void) printf("No log blocks to read\n");
4543 (void) printf("\n");
4547 dev
.l2ad_hand
= lbps
[0].lbp_daddr
+
4548 L2BLK_GET_PSIZE((&lbps
[0])->lbp_prop
);
4551 dev
.l2ad_first
= !!(l2dhdr
->dh_flags
& L2ARC_DEV_HDR_EVICT_FIRST
);
4554 if (!l2arc_log_blkptr_valid(&dev
, &lbps
[0]))
4557 /* L2BLK_GET_PSIZE returns aligned size for log blocks */
4558 asize
= L2BLK_GET_PSIZE((&lbps
[0])->lbp_prop
);
4559 if (pread64(fd
, &this_lb
, asize
, lbps
[0].lbp_daddr
) != asize
) {
4560 if (!dump_opt
['q']) {
4561 (void) printf("Error while reading next log "
4567 fletcher_4_native_varsize(&this_lb
, asize
, &cksum
);
4568 if (!ZIO_CHECKSUM_EQUAL(cksum
, lbps
[0].lbp_cksum
)) {
4570 if (!dump_opt
['q']) {
4571 (void) printf("Invalid cksum\n");
4572 dump_l2arc_log_blkptr(&lbps
[0]);
4577 switch (L2BLK_GET_COMPRESS((&lbps
[0])->lbp_prop
)) {
4578 case ZIO_COMPRESS_OFF
:
4581 abd
= abd_alloc_for_io(asize
, B_TRUE
);
4582 abd_copy_from_buf_off(abd
, &this_lb
, 0, asize
);
4583 if (zio_decompress_data(L2BLK_GET_COMPRESS(
4584 (&lbps
[0])->lbp_prop
), abd
, &this_lb
,
4585 asize
, sizeof (this_lb
), NULL
) != 0) {
4586 (void) printf("L2ARC block decompression "
4595 if (this_lb
.lb_magic
== BSWAP_64(L2ARC_LOG_BLK_MAGIC
))
4596 byteswap_uint64_array(&this_lb
, sizeof (this_lb
));
4597 if (this_lb
.lb_magic
!= L2ARC_LOG_BLK_MAGIC
) {
4599 (void) printf("Invalid log block magic\n\n");
4603 rebuild
->dh_lb_count
++;
4604 rebuild
->dh_lb_asize
+= asize
;
4605 if (dump_opt
['l'] > 1 && !dump_opt
['q']) {
4606 (void) printf("lb[%4llu]\tmagic: %llu\n",
4607 (u_longlong_t
)rebuild
->dh_lb_count
,
4608 (u_longlong_t
)this_lb
.lb_magic
);
4609 dump_l2arc_log_blkptr(&lbps
[0]);
4612 if (dump_opt
['l'] > 2 && !dump_opt
['q'])
4613 dump_l2arc_log_entries(l2dhdr
->dh_log_entries
,
4615 rebuild
->dh_lb_count
);
4617 if (l2arc_range_check_overlap(lbps
[1].lbp_payload_start
,
4618 lbps
[0].lbp_payload_start
, dev
.l2ad_evict
) &&
4623 lbps
[1] = this_lb
.lb_prev_lbp
;
4626 if (!dump_opt
['q']) {
4627 (void) printf("log_blk_count:\t %llu with valid cksum\n",
4628 (u_longlong_t
)rebuild
->dh_lb_count
);
4629 (void) printf("\t\t %d with invalid cksum\n", failed
);
4630 (void) printf("log_blk_asize:\t %llu\n\n",
4631 (u_longlong_t
)rebuild
->dh_lb_asize
);
4636 dump_l2arc_header(int fd
)
4638 l2arc_dev_hdr_phys_t l2dhdr
= {0}, rebuild
= {0};
4639 int error
= B_FALSE
;
4641 if (pread64(fd
, &l2dhdr
, sizeof (l2dhdr
),
4642 VDEV_LABEL_START_SIZE
) != sizeof (l2dhdr
)) {
4645 if (l2dhdr
.dh_magic
== BSWAP_64(L2ARC_DEV_HDR_MAGIC
))
4646 byteswap_uint64_array(&l2dhdr
, sizeof (l2dhdr
));
4648 if (l2dhdr
.dh_magic
!= L2ARC_DEV_HDR_MAGIC
)
4653 (void) printf("L2ARC device header not found\n\n");
4654 /* Do not return an error here for backward compatibility */
4656 } else if (!dump_opt
['q']) {
4657 print_l2arc_header();
4659 (void) printf(" magic: %llu\n",
4660 (u_longlong_t
)l2dhdr
.dh_magic
);
4661 (void) printf(" version: %llu\n",
4662 (u_longlong_t
)l2dhdr
.dh_version
);
4663 (void) printf(" pool_guid: %llu\n",
4664 (u_longlong_t
)l2dhdr
.dh_spa_guid
);
4665 (void) printf(" flags: %llu\n",
4666 (u_longlong_t
)l2dhdr
.dh_flags
);
4667 (void) printf(" start_lbps[0]: %llu\n",
4669 l2dhdr
.dh_start_lbps
[0].lbp_daddr
);
4670 (void) printf(" start_lbps[1]: %llu\n",
4672 l2dhdr
.dh_start_lbps
[1].lbp_daddr
);
4673 (void) printf(" log_blk_ent: %llu\n",
4674 (u_longlong_t
)l2dhdr
.dh_log_entries
);
4675 (void) printf(" start: %llu\n",
4676 (u_longlong_t
)l2dhdr
.dh_start
);
4677 (void) printf(" end: %llu\n",
4678 (u_longlong_t
)l2dhdr
.dh_end
);
4679 (void) printf(" evict: %llu\n",
4680 (u_longlong_t
)l2dhdr
.dh_evict
);
4681 (void) printf(" lb_asize_refcount: %llu\n",
4682 (u_longlong_t
)l2dhdr
.dh_lb_asize
);
4683 (void) printf(" lb_count_refcount: %llu\n",
4684 (u_longlong_t
)l2dhdr
.dh_lb_count
);
4685 (void) printf(" trim_action_time: %llu\n",
4686 (u_longlong_t
)l2dhdr
.dh_trim_action_time
);
4687 (void) printf(" trim_state: %llu\n\n",
4688 (u_longlong_t
)l2dhdr
.dh_trim_state
);
4691 dump_l2arc_log_blocks(fd
, &l2dhdr
, &rebuild
);
4693 * The total aligned size of log blocks and the number of log blocks
4694 * reported in the header of the device may be less than what zdb
4695 * reports by dump_l2arc_log_blocks() which emulates l2arc_rebuild().
4696 * This happens because dump_l2arc_log_blocks() lacks the memory
4697 * pressure valve that l2arc_rebuild() has. Thus, if we are on a system
4698 * with low memory, l2arc_rebuild will exit prematurely and dh_lb_asize
4699 * and dh_lb_count will be lower to begin with than what exists on the
4700 * device. This is normal and zdb should not exit with an error. The
4701 * opposite case should never happen though, the values reported in the
4702 * header should never be higher than what dump_l2arc_log_blocks() and
4703 * l2arc_rebuild() report. If this happens there is a leak in the
4704 * accounting of log blocks.
4706 if (l2dhdr
.dh_lb_asize
> rebuild
.dh_lb_asize
||
4707 l2dhdr
.dh_lb_count
> rebuild
.dh_lb_count
)
4714 dump_config_from_label(zdb_label_t
*label
, size_t buflen
, int l
)
4719 if ((dump_opt
['l'] < 3) && (first_label(label
->config
) != l
))
4722 print_label_header(label
, l
);
4723 dump_nvlist(label
->config_nv
, 4);
4724 print_label_numbers(" labels = ", label
->config
);
4726 if (dump_opt
['l'] >= 2)
4727 dump_nvlist_stats(label
->config_nv
, buflen
);
4730 #define ZDB_MAX_UB_HEADER_SIZE 32
4733 dump_label_uberblocks(zdb_label_t
*label
, uint64_t ashift
, int label_num
)
4737 char header
[ZDB_MAX_UB_HEADER_SIZE
];
4739 vd
.vdev_ashift
= ashift
;
4742 for (int i
= 0; i
< VDEV_UBERBLOCK_COUNT(&vd
); i
++) {
4743 uint64_t uoff
= VDEV_UBERBLOCK_OFFSET(&vd
, i
);
4744 uberblock_t
*ub
= (void *)((char *)&label
->label
+ uoff
);
4745 cksum_record_t
*rec
= label
->uberblocks
[i
];
4748 if (dump_opt
['u'] >= 2) {
4749 print_label_header(label
, label_num
);
4750 (void) printf(" Uberblock[%d] invalid\n", i
);
4755 if ((dump_opt
['u'] < 3) && (first_label(rec
) != label_num
))
4758 if ((dump_opt
['u'] < 4) &&
4759 (ub
->ub_mmp_magic
== MMP_MAGIC
) && ub
->ub_mmp_delay
&&
4760 (i
>= VDEV_UBERBLOCK_COUNT(&vd
) - MMP_BLOCKS_PER_LABEL
))
4763 print_label_header(label
, label_num
);
4764 (void) snprintf(header
, ZDB_MAX_UB_HEADER_SIZE
,
4765 " Uberblock[%d]\n", i
);
4766 dump_uberblock(ub
, header
, "");
4767 print_label_numbers(" labels = ", rec
);
4771 static char curpath
[PATH_MAX
];
4774 * Iterate through the path components, recursively passing
4775 * current one's obj and remaining path until we find the obj
4779 dump_path_impl(objset_t
*os
, uint64_t obj
, char *name
, uint64_t *retobj
)
4782 boolean_t header
= B_TRUE
;
4786 dmu_object_info_t doi
;
4788 if ((s
= strchr(name
, '/')) != NULL
)
4790 err
= zap_lookup(os
, obj
, name
, 8, 1, &child_obj
);
4792 (void) strlcat(curpath
, name
, sizeof (curpath
));
4795 (void) fprintf(stderr
, "failed to lookup %s: %s\n",
4796 curpath
, strerror(err
));
4800 child_obj
= ZFS_DIRENT_OBJ(child_obj
);
4801 err
= sa_buf_hold(os
, child_obj
, FTAG
, &db
);
4803 (void) fprintf(stderr
,
4804 "failed to get SA dbuf for obj %llu: %s\n",
4805 (u_longlong_t
)child_obj
, strerror(err
));
4808 dmu_object_info_from_db(db
, &doi
);
4809 sa_buf_rele(db
, FTAG
);
4811 if (doi
.doi_bonus_type
!= DMU_OT_SA
&&
4812 doi
.doi_bonus_type
!= DMU_OT_ZNODE
) {
4813 (void) fprintf(stderr
, "invalid bonus type %d for obj %llu\n",
4814 doi
.doi_bonus_type
, (u_longlong_t
)child_obj
);
4818 if (dump_opt
['v'] > 6) {
4819 (void) printf("obj=%llu %s type=%d bonustype=%d\n",
4820 (u_longlong_t
)child_obj
, curpath
, doi
.doi_type
,
4821 doi
.doi_bonus_type
);
4824 (void) strlcat(curpath
, "/", sizeof (curpath
));
4826 switch (doi
.doi_type
) {
4827 case DMU_OT_DIRECTORY_CONTENTS
:
4828 if (s
!= NULL
&& *(s
+ 1) != '\0')
4829 return (dump_path_impl(os
, child_obj
, s
+ 1, retobj
));
4831 case DMU_OT_PLAIN_FILE_CONTENTS
:
4832 if (retobj
!= NULL
) {
4833 *retobj
= child_obj
;
4835 dump_object(os
, child_obj
, dump_opt
['v'], &header
,
4840 (void) fprintf(stderr
, "object %llu has non-file/directory "
4841 "type %d\n", (u_longlong_t
)obj
, doi
.doi_type
);
4849 * Dump the blocks for the object specified by path inside the dataset.
4852 dump_path(char *ds
, char *path
, uint64_t *retobj
)
4858 err
= open_objset(ds
, FTAG
, &os
);
4862 err
= zap_lookup(os
, MASTER_NODE_OBJ
, ZFS_ROOT_OBJ
, 8, 1, &root_obj
);
4864 (void) fprintf(stderr
, "can't lookup root znode: %s\n",
4866 close_objset(os
, FTAG
);
4870 (void) snprintf(curpath
, sizeof (curpath
), "dataset=%s path=/", ds
);
4872 err
= dump_path_impl(os
, root_obj
, path
, retobj
);
4874 close_objset(os
, FTAG
);
4879 zdb_copy_object(objset_t
*os
, uint64_t srcobj
, char *destfile
)
4882 uint64_t size
, readsize
, oursize
, offset
;
4886 (void) printf("Copying object %" PRIu64
" to file %s\n", srcobj
,
4889 VERIFY3P(os
, ==, sa_os
);
4890 if ((err
= sa_handle_get(os
, srcobj
, NULL
, SA_HDL_PRIVATE
, &hdl
))) {
4891 (void) printf("Failed to get handle for SA znode\n");
4894 if ((err
= sa_lookup(hdl
, sa_attr_table
[ZPL_SIZE
], &size
, 8))) {
4895 (void) sa_handle_destroy(hdl
);
4898 (void) sa_handle_destroy(hdl
);
4900 (void) printf("Object %" PRIu64
" is %" PRIu64
" bytes\n", srcobj
,
4906 int fd
= open(destfile
, O_WRONLY
| O_CREAT
| O_TRUNC
, 0644);
4910 * We cap the size at 1 mebibyte here to prevent
4911 * allocation failures and nigh-infinite printing if the
4912 * object is extremely large.
4914 oursize
= MIN(size
, 1 << 20);
4916 char *buf
= kmem_alloc(oursize
, KM_NOSLEEP
);
4922 while (offset
< size
) {
4923 readsize
= MIN(size
- offset
, 1 << 20);
4924 err
= dmu_read(os
, srcobj
, offset
, readsize
, buf
, 0);
4926 (void) printf("got error %u from dmu_read\n", err
);
4927 kmem_free(buf
, oursize
);
4931 if (dump_opt
['v'] > 3) {
4932 (void) printf("Read offset=%" PRIu64
" size=%" PRIu64
4933 " error=%d\n", offset
, readsize
, err
);
4936 writesize
= write(fd
, buf
, readsize
);
4937 if (writesize
< 0) {
4940 } else if (writesize
!= readsize
) {
4941 /* Incomplete write */
4942 (void) fprintf(stderr
, "Short write, only wrote %llu of"
4943 " %" PRIu64
" bytes, exiting...\n",
4944 (u_longlong_t
)writesize
, readsize
);
4954 kmem_free(buf
, oursize
);
4960 label_cksum_valid(vdev_label_t
*label
, uint64_t offset
)
4962 zio_checksum_info_t
*ci
= &zio_checksum_table
[ZIO_CHECKSUM_LABEL
];
4963 zio_cksum_t expected_cksum
;
4964 zio_cksum_t actual_cksum
;
4965 zio_cksum_t verifier
;
4969 void *data
= (char *)label
+ offsetof(vdev_label_t
, vl_vdev_phys
);
4970 eck
= (zio_eck_t
*)((char *)(data
) + VDEV_PHYS_SIZE
) - 1;
4972 offset
+= offsetof(vdev_label_t
, vl_vdev_phys
);
4973 ZIO_SET_CHECKSUM(&verifier
, offset
, 0, 0, 0);
4975 byteswap
= (eck
->zec_magic
== BSWAP_64(ZEC_MAGIC
));
4977 byteswap_uint64_array(&verifier
, sizeof (zio_cksum_t
));
4979 expected_cksum
= eck
->zec_cksum
;
4980 eck
->zec_cksum
= verifier
;
4982 abd_t
*abd
= abd_get_from_buf(data
, VDEV_PHYS_SIZE
);
4983 ci
->ci_func
[byteswap
](abd
, VDEV_PHYS_SIZE
, NULL
, &actual_cksum
);
4987 byteswap_uint64_array(&expected_cksum
, sizeof (zio_cksum_t
));
4989 if (ZIO_CHECKSUM_EQUAL(actual_cksum
, expected_cksum
))
4996 dump_label(const char *dev
)
4998 char path
[MAXPATHLEN
];
4999 zdb_label_t labels
[VDEV_LABELS
] = {{{{0}}}};
5000 uint64_t psize
, ashift
, l2cache
;
5001 struct stat64 statbuf
;
5002 boolean_t config_found
= B_FALSE
;
5003 boolean_t error
= B_FALSE
;
5004 boolean_t read_l2arc_header
= B_FALSE
;
5005 avl_tree_t config_tree
;
5006 avl_tree_t uberblock_tree
;
5007 void *node
, *cookie
;
5011 * Check if we were given absolute path and use it as is.
5012 * Otherwise if the provided vdev name doesn't point to a file,
5013 * try prepending expected disk paths and partition numbers.
5015 (void) strlcpy(path
, dev
, sizeof (path
));
5016 if (dev
[0] != '/' && stat64(path
, &statbuf
) != 0) {
5019 error
= zfs_resolve_shortname(dev
, path
, MAXPATHLEN
);
5020 if (error
== 0 && zfs_dev_is_whole_disk(path
)) {
5021 if (zfs_append_partition(path
, MAXPATHLEN
) == -1)
5025 if (error
|| (stat64(path
, &statbuf
) != 0)) {
5026 (void) printf("failed to find device %s, try "
5027 "specifying absolute path instead\n", dev
);
5032 if ((fd
= open64(path
, O_RDONLY
)) < 0) {
5033 (void) printf("cannot open '%s': %s\n", path
, strerror(errno
));
5037 if (fstat64_blk(fd
, &statbuf
) != 0) {
5038 (void) printf("failed to stat '%s': %s\n", path
,
5044 if (S_ISBLK(statbuf
.st_mode
) && zfs_dev_flush(fd
) != 0)
5045 (void) printf("failed to invalidate cache '%s' : %s\n", path
,
5048 avl_create(&config_tree
, cksum_record_compare
,
5049 sizeof (cksum_record_t
), offsetof(cksum_record_t
, link
));
5050 avl_create(&uberblock_tree
, cksum_record_compare
,
5051 sizeof (cksum_record_t
), offsetof(cksum_record_t
, link
));
5053 psize
= statbuf
.st_size
;
5054 psize
= P2ALIGN(psize
, (uint64_t)sizeof (vdev_label_t
));
5055 ashift
= SPA_MINBLOCKSHIFT
;
5058 * 1. Read the label from disk
5059 * 2. Verify label cksum
5060 * 3. Unpack the configuration and insert in config tree.
5061 * 4. Traverse all uberblocks and insert in uberblock tree.
5063 for (int l
= 0; l
< VDEV_LABELS
; l
++) {
5064 zdb_label_t
*label
= &labels
[l
];
5065 char *buf
= label
->label
.vl_vdev_phys
.vp_nvlist
;
5066 size_t buflen
= sizeof (label
->label
.vl_vdev_phys
.vp_nvlist
);
5068 cksum_record_t
*rec
;
5072 label
->label_offset
= vdev_label_offset(psize
, l
, 0);
5074 if (pread64(fd
, &label
->label
, sizeof (label
->label
),
5075 label
->label_offset
) != sizeof (label
->label
)) {
5077 (void) printf("failed to read label %d\n", l
);
5078 label
->read_failed
= B_TRUE
;
5083 label
->read_failed
= B_FALSE
;
5084 label
->cksum_valid
= label_cksum_valid(&label
->label
,
5085 label
->label_offset
);
5087 if (nvlist_unpack(buf
, buflen
, &config
, 0) == 0) {
5088 nvlist_t
*vdev_tree
= NULL
;
5091 if ((nvlist_lookup_nvlist(config
,
5092 ZPOOL_CONFIG_VDEV_TREE
, &vdev_tree
) != 0) ||
5093 (nvlist_lookup_uint64(vdev_tree
,
5094 ZPOOL_CONFIG_ASHIFT
, &ashift
) != 0))
5095 ashift
= SPA_MINBLOCKSHIFT
;
5097 if (nvlist_size(config
, &size
, NV_ENCODE_XDR
) != 0)
5100 /* If the device is a cache device clear the header. */
5101 if (!read_l2arc_header
) {
5102 if (nvlist_lookup_uint64(config
,
5103 ZPOOL_CONFIG_POOL_STATE
, &l2cache
) == 0 &&
5104 l2cache
== POOL_STATE_L2CACHE
) {
5105 read_l2arc_header
= B_TRUE
;
5109 fletcher_4_native_varsize(buf
, size
, &cksum
);
5110 rec
= cksum_record_insert(&config_tree
, &cksum
, l
);
5112 label
->config
= rec
;
5113 label
->config_nv
= config
;
5114 config_found
= B_TRUE
;
5119 vd
.vdev_ashift
= ashift
;
5122 for (int i
= 0; i
< VDEV_UBERBLOCK_COUNT(&vd
); i
++) {
5123 uint64_t uoff
= VDEV_UBERBLOCK_OFFSET(&vd
, i
);
5124 uberblock_t
*ub
= (void *)((char *)label
+ uoff
);
5126 if (uberblock_verify(ub
))
5129 fletcher_4_native_varsize(ub
, sizeof (*ub
), &cksum
);
5130 rec
= cksum_record_insert(&uberblock_tree
, &cksum
, l
);
5132 label
->uberblocks
[i
] = rec
;
5137 * Dump the label and uberblocks.
5139 for (int l
= 0; l
< VDEV_LABELS
; l
++) {
5140 zdb_label_t
*label
= &labels
[l
];
5141 size_t buflen
= sizeof (label
->label
.vl_vdev_phys
.vp_nvlist
);
5143 if (label
->read_failed
== B_TRUE
)
5146 if (label
->config_nv
) {
5147 dump_config_from_label(label
, buflen
, l
);
5150 (void) printf("failed to unpack label %d\n", l
);
5154 dump_label_uberblocks(label
, ashift
, l
);
5156 nvlist_free(label
->config_nv
);
5160 * Dump the L2ARC header, if existent.
5162 if (read_l2arc_header
)
5163 error
|= dump_l2arc_header(fd
);
5166 while ((node
= avl_destroy_nodes(&config_tree
, &cookie
)) != NULL
)
5167 umem_free(node
, sizeof (cksum_record_t
));
5170 while ((node
= avl_destroy_nodes(&uberblock_tree
, &cookie
)) != NULL
)
5171 umem_free(node
, sizeof (cksum_record_t
));
5173 avl_destroy(&config_tree
);
5174 avl_destroy(&uberblock_tree
);
5178 return (config_found
== B_FALSE
? 2 :
5179 (error
== B_TRUE
? 1 : 0));
5182 static uint64_t dataset_feature_count
[SPA_FEATURES
];
5183 static uint64_t global_feature_count
[SPA_FEATURES
];
5184 static uint64_t remap_deadlist_count
= 0;
5187 dump_one_objset(const char *dsname
, void *arg
)
5194 error
= open_objset(dsname
, FTAG
, &os
);
5198 for (f
= 0; f
< SPA_FEATURES
; f
++) {
5199 if (!dsl_dataset_feature_is_active(dmu_objset_ds(os
), f
))
5201 ASSERT(spa_feature_table
[f
].fi_flags
&
5202 ZFEATURE_FLAG_PER_DATASET
);
5203 dataset_feature_count
[f
]++;
5206 if (dsl_dataset_remap_deadlist_exists(dmu_objset_ds(os
))) {
5207 remap_deadlist_count
++;
5210 for (dsl_bookmark_node_t
*dbn
=
5211 avl_first(&dmu_objset_ds(os
)->ds_bookmarks
); dbn
!= NULL
;
5212 dbn
= AVL_NEXT(&dmu_objset_ds(os
)->ds_bookmarks
, dbn
)) {
5213 mos_obj_refd(dbn
->dbn_phys
.zbm_redaction_obj
);
5214 if (dbn
->dbn_phys
.zbm_redaction_obj
!= 0)
5215 global_feature_count
[SPA_FEATURE_REDACTION_BOOKMARKS
]++;
5216 if (dbn
->dbn_phys
.zbm_flags
& ZBM_FLAG_HAS_FBN
)
5217 global_feature_count
[SPA_FEATURE_BOOKMARK_WRITTEN
]++;
5220 if (dsl_deadlist_is_open(&dmu_objset_ds(os
)->ds_dir
->dd_livelist
) &&
5221 !dmu_objset_is_snapshot(os
)) {
5222 global_feature_count
[SPA_FEATURE_LIVELIST
]++;
5226 close_objset(os
, FTAG
);
5227 fuid_table_destroy();
5234 #define PSIZE_HISTO_SIZE (SPA_OLD_MAXBLOCKSIZE / SPA_MINBLOCKSIZE + 2)
5235 typedef struct zdb_blkstats
{
5241 uint64_t zb_ditto_samevdev
;
5242 uint64_t zb_ditto_same_ms
;
5243 uint64_t zb_psize_histogram
[PSIZE_HISTO_SIZE
];
5247 * Extended object types to report deferred frees and dedup auto-ditto blocks.
5249 #define ZDB_OT_DEFERRED (DMU_OT_NUMTYPES + 0)
5250 #define ZDB_OT_DITTO (DMU_OT_NUMTYPES + 1)
5251 #define ZDB_OT_OTHER (DMU_OT_NUMTYPES + 2)
5252 #define ZDB_OT_TOTAL (DMU_OT_NUMTYPES + 3)
5254 static const char *zdb_ot_extname
[] = {
5261 #define ZB_TOTAL DN_MAX_LEVELS
5262 #define SPA_MAX_FOR_16M (SPA_MAXBLOCKSHIFT+1)
5264 typedef struct zdb_cb
{
5265 zdb_blkstats_t zcb_type
[ZB_TOTAL
+ 1][ZDB_OT_TOTAL
+ 1];
5266 uint64_t zcb_removing_size
;
5267 uint64_t zcb_checkpoint_size
;
5268 uint64_t zcb_dedup_asize
;
5269 uint64_t zcb_dedup_blocks
;
5270 uint64_t zcb_psize_count
[SPA_MAX_FOR_16M
];
5271 uint64_t zcb_lsize_count
[SPA_MAX_FOR_16M
];
5272 uint64_t zcb_asize_count
[SPA_MAX_FOR_16M
];
5273 uint64_t zcb_psize_len
[SPA_MAX_FOR_16M
];
5274 uint64_t zcb_lsize_len
[SPA_MAX_FOR_16M
];
5275 uint64_t zcb_asize_len
[SPA_MAX_FOR_16M
];
5276 uint64_t zcb_psize_total
;
5277 uint64_t zcb_lsize_total
;
5278 uint64_t zcb_asize_total
;
5279 uint64_t zcb_embedded_blocks
[NUM_BP_EMBEDDED_TYPES
];
5280 uint64_t zcb_embedded_histogram
[NUM_BP_EMBEDDED_TYPES
]
5281 [BPE_PAYLOAD_SIZE
+ 1];
5283 hrtime_t zcb_lastprint
;
5284 uint64_t zcb_totalasize
;
5285 uint64_t zcb_errors
[256];
5289 uint32_t **zcb_vd_obsolete_counts
;
5292 /* test if two DVA offsets from same vdev are within the same metaslab */
5294 same_metaslab(spa_t
*spa
, uint64_t vdev
, uint64_t off1
, uint64_t off2
)
5296 vdev_t
*vd
= vdev_lookup_top(spa
, vdev
);
5297 uint64_t ms_shift
= vd
->vdev_ms_shift
;
5299 return ((off1
>> ms_shift
) == (off2
>> ms_shift
));
5303 * Used to simplify reporting of the histogram data.
5305 typedef struct one_histo
{
5309 uint64_t cumulative
;
5313 * The number of separate histograms processed for psize, lsize and asize.
5318 * This routine will create a fixed column size output of three different
5319 * histograms showing by blocksize of 512 - 2^ SPA_MAX_FOR_16M
5320 * the count, length and cumulative length of the psize, lsize and
5323 * All three types of blocks are listed on a single line
5325 * By default the table is printed in nicenumber format (e.g. 123K) but
5326 * if the '-P' parameter is specified then the full raw number (parseable)
5330 dump_size_histograms(zdb_cb_t
*zcb
)
5333 * A temporary buffer that allows us to convert a number into
5334 * a string using zdb_nicenumber to allow either raw or human
5335 * readable numbers to be output.
5340 * Define titles which are used in the headers of the tables
5341 * printed by this routine.
5343 const char blocksize_title1
[] = "block";
5344 const char blocksize_title2
[] = "size";
5345 const char count_title
[] = "Count";
5346 const char length_title
[] = "Size";
5347 const char cumulative_title
[] = "Cum.";
5350 * Setup the histogram arrays (psize, lsize, and asize).
5352 one_histo_t parm_histo
[NUM_HISTO
];
5354 parm_histo
[0].name
= "psize";
5355 parm_histo
[0].count
= zcb
->zcb_psize_count
;
5356 parm_histo
[0].len
= zcb
->zcb_psize_len
;
5357 parm_histo
[0].cumulative
= 0;
5359 parm_histo
[1].name
= "lsize";
5360 parm_histo
[1].count
= zcb
->zcb_lsize_count
;
5361 parm_histo
[1].len
= zcb
->zcb_lsize_len
;
5362 parm_histo
[1].cumulative
= 0;
5364 parm_histo
[2].name
= "asize";
5365 parm_histo
[2].count
= zcb
->zcb_asize_count
;
5366 parm_histo
[2].len
= zcb
->zcb_asize_len
;
5367 parm_histo
[2].cumulative
= 0;
5370 (void) printf("\nBlock Size Histogram\n");
5372 * Print the first line titles
5375 (void) printf("\n%s\t", blocksize_title1
);
5377 (void) printf("\n%7s ", blocksize_title1
);
5379 for (int j
= 0; j
< NUM_HISTO
; j
++) {
5380 if (dump_opt
['P']) {
5381 if (j
< NUM_HISTO
- 1) {
5382 (void) printf("%s\t\t\t", parm_histo
[j
].name
);
5384 /* Don't print trailing spaces */
5385 (void) printf(" %s", parm_histo
[j
].name
);
5388 if (j
< NUM_HISTO
- 1) {
5389 /* Left aligned strings in the output */
5390 (void) printf("%-7s ",
5391 parm_histo
[j
].name
);
5393 /* Don't print trailing spaces */
5394 (void) printf("%s", parm_histo
[j
].name
);
5398 (void) printf("\n");
5401 * Print the second line titles
5403 if (dump_opt
['P']) {
5404 (void) printf("%s\t", blocksize_title2
);
5406 (void) printf("%7s ", blocksize_title2
);
5409 for (int i
= 0; i
< NUM_HISTO
; i
++) {
5410 if (dump_opt
['P']) {
5411 (void) printf("%s\t%s\t%s\t",
5412 count_title
, length_title
, cumulative_title
);
5414 (void) printf("%7s%7s%7s",
5415 count_title
, length_title
, cumulative_title
);
5418 (void) printf("\n");
5423 for (int i
= SPA_MINBLOCKSHIFT
; i
< SPA_MAX_FOR_16M
; i
++) {
5426 * Print the first column showing the blocksize
5428 zdb_nicenum((1ULL << i
), numbuf
, sizeof (numbuf
));
5430 if (dump_opt
['P']) {
5431 printf("%s", numbuf
);
5433 printf("%7s:", numbuf
);
5437 * Print the remaining set of 3 columns per size:
5438 * for psize, lsize and asize
5440 for (int j
= 0; j
< NUM_HISTO
; j
++) {
5441 parm_histo
[j
].cumulative
+= parm_histo
[j
].len
[i
];
5443 zdb_nicenum(parm_histo
[j
].count
[i
],
5444 numbuf
, sizeof (numbuf
));
5446 (void) printf("\t%s", numbuf
);
5448 (void) printf("%7s", numbuf
);
5450 zdb_nicenum(parm_histo
[j
].len
[i
],
5451 numbuf
, sizeof (numbuf
));
5453 (void) printf("\t%s", numbuf
);
5455 (void) printf("%7s", numbuf
);
5457 zdb_nicenum(parm_histo
[j
].cumulative
,
5458 numbuf
, sizeof (numbuf
));
5460 (void) printf("\t%s", numbuf
);
5462 (void) printf("%7s", numbuf
);
5464 (void) printf("\n");
5469 zdb_count_block(zdb_cb_t
*zcb
, zilog_t
*zilog
, const blkptr_t
*bp
,
5470 dmu_object_type_t type
)
5472 uint64_t refcnt
= 0;
5475 ASSERT(type
< ZDB_OT_TOTAL
);
5477 if (zilog
&& zil_bp_tree_add(zilog
, bp
) != 0)
5480 spa_config_enter(zcb
->zcb_spa
, SCL_CONFIG
, FTAG
, RW_READER
);
5482 for (i
= 0; i
< 4; i
++) {
5483 int l
= (i
< 2) ? BP_GET_LEVEL(bp
) : ZB_TOTAL
;
5484 int t
= (i
& 1) ? type
: ZDB_OT_TOTAL
;
5486 zdb_blkstats_t
*zb
= &zcb
->zcb_type
[l
][t
];
5488 zb
->zb_asize
+= BP_GET_ASIZE(bp
);
5489 zb
->zb_lsize
+= BP_GET_LSIZE(bp
);
5490 zb
->zb_psize
+= BP_GET_PSIZE(bp
);
5494 * The histogram is only big enough to record blocks up to
5495 * SPA_OLD_MAXBLOCKSIZE; larger blocks go into the last,
5498 unsigned idx
= BP_GET_PSIZE(bp
) >> SPA_MINBLOCKSHIFT
;
5499 idx
= MIN(idx
, SPA_OLD_MAXBLOCKSIZE
/ SPA_MINBLOCKSIZE
+ 1);
5500 zb
->zb_psize_histogram
[idx
]++;
5502 zb
->zb_gangs
+= BP_COUNT_GANG(bp
);
5504 switch (BP_GET_NDVAS(bp
)) {
5506 if (DVA_GET_VDEV(&bp
->blk_dva
[0]) ==
5507 DVA_GET_VDEV(&bp
->blk_dva
[1])) {
5508 zb
->zb_ditto_samevdev
++;
5510 if (same_metaslab(zcb
->zcb_spa
,
5511 DVA_GET_VDEV(&bp
->blk_dva
[0]),
5512 DVA_GET_OFFSET(&bp
->blk_dva
[0]),
5513 DVA_GET_OFFSET(&bp
->blk_dva
[1])))
5514 zb
->zb_ditto_same_ms
++;
5518 equal
= (DVA_GET_VDEV(&bp
->blk_dva
[0]) ==
5519 DVA_GET_VDEV(&bp
->blk_dva
[1])) +
5520 (DVA_GET_VDEV(&bp
->blk_dva
[0]) ==
5521 DVA_GET_VDEV(&bp
->blk_dva
[2])) +
5522 (DVA_GET_VDEV(&bp
->blk_dva
[1]) ==
5523 DVA_GET_VDEV(&bp
->blk_dva
[2]));
5525 zb
->zb_ditto_samevdev
++;
5527 if (DVA_GET_VDEV(&bp
->blk_dva
[0]) ==
5528 DVA_GET_VDEV(&bp
->blk_dva
[1]) &&
5529 same_metaslab(zcb
->zcb_spa
,
5530 DVA_GET_VDEV(&bp
->blk_dva
[0]),
5531 DVA_GET_OFFSET(&bp
->blk_dva
[0]),
5532 DVA_GET_OFFSET(&bp
->blk_dva
[1])))
5533 zb
->zb_ditto_same_ms
++;
5534 else if (DVA_GET_VDEV(&bp
->blk_dva
[0]) ==
5535 DVA_GET_VDEV(&bp
->blk_dva
[2]) &&
5536 same_metaslab(zcb
->zcb_spa
,
5537 DVA_GET_VDEV(&bp
->blk_dva
[0]),
5538 DVA_GET_OFFSET(&bp
->blk_dva
[0]),
5539 DVA_GET_OFFSET(&bp
->blk_dva
[2])))
5540 zb
->zb_ditto_same_ms
++;
5541 else if (DVA_GET_VDEV(&bp
->blk_dva
[1]) ==
5542 DVA_GET_VDEV(&bp
->blk_dva
[2]) &&
5543 same_metaslab(zcb
->zcb_spa
,
5544 DVA_GET_VDEV(&bp
->blk_dva
[1]),
5545 DVA_GET_OFFSET(&bp
->blk_dva
[1]),
5546 DVA_GET_OFFSET(&bp
->blk_dva
[2])))
5547 zb
->zb_ditto_same_ms
++;
5553 spa_config_exit(zcb
->zcb_spa
, SCL_CONFIG
, FTAG
);
5555 if (BP_IS_EMBEDDED(bp
)) {
5556 zcb
->zcb_embedded_blocks
[BPE_GET_ETYPE(bp
)]++;
5557 zcb
->zcb_embedded_histogram
[BPE_GET_ETYPE(bp
)]
5558 [BPE_GET_PSIZE(bp
)]++;
5562 * The binning histogram bins by powers of two up to
5563 * SPA_MAXBLOCKSIZE rather than creating bins for
5564 * every possible blocksize found in the pool.
5566 int bin
= highbit64(BP_GET_PSIZE(bp
)) - 1;
5568 zcb
->zcb_psize_count
[bin
]++;
5569 zcb
->zcb_psize_len
[bin
] += BP_GET_PSIZE(bp
);
5570 zcb
->zcb_psize_total
+= BP_GET_PSIZE(bp
);
5572 bin
= highbit64(BP_GET_LSIZE(bp
)) - 1;
5574 zcb
->zcb_lsize_count
[bin
]++;
5575 zcb
->zcb_lsize_len
[bin
] += BP_GET_LSIZE(bp
);
5576 zcb
->zcb_lsize_total
+= BP_GET_LSIZE(bp
);
5578 bin
= highbit64(BP_GET_ASIZE(bp
)) - 1;
5580 zcb
->zcb_asize_count
[bin
]++;
5581 zcb
->zcb_asize_len
[bin
] += BP_GET_ASIZE(bp
);
5582 zcb
->zcb_asize_total
+= BP_GET_ASIZE(bp
);
5587 if (BP_GET_DEDUP(bp
)) {
5591 ddt
= ddt_select(zcb
->zcb_spa
, bp
);
5593 dde
= ddt_lookup(ddt
, bp
, B_FALSE
);
5598 ddt_phys_t
*ddp
= ddt_phys_select(dde
, bp
);
5599 ddt_phys_decref(ddp
);
5600 refcnt
= ddp
->ddp_refcnt
;
5601 if (ddt_phys_total_refcnt(dde
) == 0)
5602 ddt_remove(ddt
, dde
);
5607 VERIFY3U(zio_wait(zio_claim(NULL
, zcb
->zcb_spa
,
5608 refcnt
? 0 : spa_min_claim_txg(zcb
->zcb_spa
),
5609 bp
, NULL
, NULL
, ZIO_FLAG_CANFAIL
)), ==, 0);
5613 zdb_blkptr_done(zio_t
*zio
)
5615 spa_t
*spa
= zio
->io_spa
;
5616 blkptr_t
*bp
= zio
->io_bp
;
5617 int ioerr
= zio
->io_error
;
5618 zdb_cb_t
*zcb
= zio
->io_private
;
5619 zbookmark_phys_t
*zb
= &zio
->io_bookmark
;
5621 mutex_enter(&spa
->spa_scrub_lock
);
5622 spa
->spa_load_verify_bytes
-= BP_GET_PSIZE(bp
);
5623 cv_broadcast(&spa
->spa_scrub_io_cv
);
5625 if (ioerr
&& !(zio
->io_flags
& ZIO_FLAG_SPECULATIVE
)) {
5626 char blkbuf
[BP_SPRINTF_LEN
];
5628 zcb
->zcb_haderrors
= 1;
5629 zcb
->zcb_errors
[ioerr
]++;
5631 if (dump_opt
['b'] >= 2)
5632 snprintf_blkptr(blkbuf
, sizeof (blkbuf
), bp
);
5636 (void) printf("zdb_blkptr_cb: "
5637 "Got error %d reading "
5638 "<%llu, %llu, %lld, %llx> %s -- skipping\n",
5640 (u_longlong_t
)zb
->zb_objset
,
5641 (u_longlong_t
)zb
->zb_object
,
5642 (u_longlong_t
)zb
->zb_level
,
5643 (u_longlong_t
)zb
->zb_blkid
,
5646 mutex_exit(&spa
->spa_scrub_lock
);
5648 abd_free(zio
->io_abd
);
5652 zdb_blkptr_cb(spa_t
*spa
, zilog_t
*zilog
, const blkptr_t
*bp
,
5653 const zbookmark_phys_t
*zb
, const dnode_phys_t
*dnp
, void *arg
)
5655 zdb_cb_t
*zcb
= arg
;
5656 dmu_object_type_t type
;
5657 boolean_t is_metadata
;
5659 if (zb
->zb_level
== ZB_DNODE_LEVEL
)
5662 if (dump_opt
['b'] >= 5 && bp
->blk_birth
> 0) {
5663 char blkbuf
[BP_SPRINTF_LEN
];
5664 snprintf_blkptr(blkbuf
, sizeof (blkbuf
), bp
);
5665 (void) printf("objset %llu object %llu "
5666 "level %lld offset 0x%llx %s\n",
5667 (u_longlong_t
)zb
->zb_objset
,
5668 (u_longlong_t
)zb
->zb_object
,
5669 (longlong_t
)zb
->zb_level
,
5670 (u_longlong_t
)blkid2offset(dnp
, bp
, zb
),
5674 if (BP_IS_HOLE(bp
) || BP_IS_REDACTED(bp
))
5677 type
= BP_GET_TYPE(bp
);
5679 zdb_count_block(zcb
, zilog
, bp
,
5680 (type
& DMU_OT_NEWTYPE
) ? ZDB_OT_OTHER
: type
);
5682 is_metadata
= (BP_GET_LEVEL(bp
) != 0 || DMU_OT_IS_METADATA(type
));
5684 if (!BP_IS_EMBEDDED(bp
) &&
5685 (dump_opt
['c'] > 1 || (dump_opt
['c'] && is_metadata
))) {
5686 size_t size
= BP_GET_PSIZE(bp
);
5687 abd_t
*abd
= abd_alloc(size
, B_FALSE
);
5688 int flags
= ZIO_FLAG_CANFAIL
| ZIO_FLAG_SCRUB
| ZIO_FLAG_RAW
;
5690 /* If it's an intent log block, failure is expected. */
5691 if (zb
->zb_level
== ZB_ZIL_LEVEL
)
5692 flags
|= ZIO_FLAG_SPECULATIVE
;
5694 mutex_enter(&spa
->spa_scrub_lock
);
5695 while (spa
->spa_load_verify_bytes
> max_inflight_bytes
)
5696 cv_wait(&spa
->spa_scrub_io_cv
, &spa
->spa_scrub_lock
);
5697 spa
->spa_load_verify_bytes
+= size
;
5698 mutex_exit(&spa
->spa_scrub_lock
);
5700 zio_nowait(zio_read(NULL
, spa
, bp
, abd
, size
,
5701 zdb_blkptr_done
, zcb
, ZIO_PRIORITY_ASYNC_READ
, flags
, zb
));
5704 zcb
->zcb_readfails
= 0;
5706 /* only call gethrtime() every 100 blocks */
5713 if (dump_opt
['b'] < 5 && gethrtime() > zcb
->zcb_lastprint
+ NANOSEC
) {
5714 uint64_t now
= gethrtime();
5716 uint64_t bytes
= zcb
->zcb_type
[ZB_TOTAL
][ZDB_OT_TOTAL
].zb_asize
;
5717 uint64_t kb_per_sec
=
5718 1 + bytes
/ (1 + ((now
- zcb
->zcb_start
) / 1000 / 1000));
5719 uint64_t sec_remaining
=
5720 (zcb
->zcb_totalasize
- bytes
) / 1024 / kb_per_sec
;
5722 /* make sure nicenum has enough space */
5723 _Static_assert(sizeof (buf
) >= NN_NUMBUF_SZ
, "buf truncated");
5725 zfs_nicebytes(bytes
, buf
, sizeof (buf
));
5726 (void) fprintf(stderr
,
5727 "\r%5s completed (%4"PRIu64
"MB/s) "
5728 "estimated time remaining: "
5729 "%"PRIu64
"hr %02"PRIu64
"min %02"PRIu64
"sec ",
5730 buf
, kb_per_sec
/ 1024,
5731 sec_remaining
/ 60 / 60,
5732 sec_remaining
/ 60 % 60,
5733 sec_remaining
% 60);
5735 zcb
->zcb_lastprint
= now
;
5742 zdb_leak(void *arg
, uint64_t start
, uint64_t size
)
5746 (void) printf("leaked space: vdev %llu, offset 0x%llx, size %llu\n",
5747 (u_longlong_t
)vd
->vdev_id
, (u_longlong_t
)start
, (u_longlong_t
)size
);
5750 static metaslab_ops_t zdb_metaslab_ops
= {
5755 load_unflushed_svr_segs_cb(spa_t
*spa
, space_map_entry_t
*sme
,
5756 uint64_t txg
, void *arg
)
5758 spa_vdev_removal_t
*svr
= arg
;
5760 uint64_t offset
= sme
->sme_offset
;
5761 uint64_t size
= sme
->sme_run
;
5763 /* skip vdevs we don't care about */
5764 if (sme
->sme_vdev
!= svr
->svr_vdev_id
)
5767 vdev_t
*vd
= vdev_lookup_top(spa
, sme
->sme_vdev
);
5768 metaslab_t
*ms
= vd
->vdev_ms
[offset
>> vd
->vdev_ms_shift
];
5769 ASSERT(sme
->sme_type
== SM_ALLOC
|| sme
->sme_type
== SM_FREE
);
5771 if (txg
< metaslab_unflushed_txg(ms
))
5774 if (sme
->sme_type
== SM_ALLOC
)
5775 range_tree_add(svr
->svr_allocd_segs
, offset
, size
);
5777 range_tree_remove(svr
->svr_allocd_segs
, offset
, size
);
5783 claim_segment_impl_cb(uint64_t inner_offset
, vdev_t
*vd
, uint64_t offset
,
5784 uint64_t size
, void *arg
)
5786 (void) inner_offset
, (void) arg
;
5789 * This callback was called through a remap from
5790 * a device being removed. Therefore, the vdev that
5791 * this callback is applied to is a concrete
5794 ASSERT(vdev_is_concrete(vd
));
5796 VERIFY0(metaslab_claim_impl(vd
, offset
, size
,
5797 spa_min_claim_txg(vd
->vdev_spa
)));
5801 claim_segment_cb(void *arg
, uint64_t offset
, uint64_t size
)
5805 vdev_indirect_ops
.vdev_op_remap(vd
, offset
, size
,
5806 claim_segment_impl_cb
, NULL
);
5810 * After accounting for all allocated blocks that are directly referenced,
5811 * we might have missed a reference to a block from a partially complete
5812 * (and thus unused) indirect mapping object. We perform a secondary pass
5813 * through the metaslabs we have already mapped and claim the destination
5817 zdb_claim_removing(spa_t
*spa
, zdb_cb_t
*zcb
)
5822 if (spa
->spa_vdev_removal
== NULL
)
5825 spa_config_enter(spa
, SCL_CONFIG
, FTAG
, RW_READER
);
5827 spa_vdev_removal_t
*svr
= spa
->spa_vdev_removal
;
5828 vdev_t
*vd
= vdev_lookup_top(spa
, svr
->svr_vdev_id
);
5829 vdev_indirect_mapping_t
*vim
= vd
->vdev_indirect_mapping
;
5831 ASSERT0(range_tree_space(svr
->svr_allocd_segs
));
5833 range_tree_t
*allocs
= range_tree_create(NULL
, RANGE_SEG64
, NULL
, 0, 0);
5834 for (uint64_t msi
= 0; msi
< vd
->vdev_ms_count
; msi
++) {
5835 metaslab_t
*msp
= vd
->vdev_ms
[msi
];
5837 ASSERT0(range_tree_space(allocs
));
5838 if (msp
->ms_sm
!= NULL
)
5839 VERIFY0(space_map_load(msp
->ms_sm
, allocs
, SM_ALLOC
));
5840 range_tree_vacate(allocs
, range_tree_add
, svr
->svr_allocd_segs
);
5842 range_tree_destroy(allocs
);
5844 iterate_through_spacemap_logs(spa
, load_unflushed_svr_segs_cb
, svr
);
5847 * Clear everything past what has been synced,
5848 * because we have not allocated mappings for
5851 range_tree_clear(svr
->svr_allocd_segs
,
5852 vdev_indirect_mapping_max_offset(vim
),
5853 vd
->vdev_asize
- vdev_indirect_mapping_max_offset(vim
));
5855 zcb
->zcb_removing_size
+= range_tree_space(svr
->svr_allocd_segs
);
5856 range_tree_vacate(svr
->svr_allocd_segs
, claim_segment_cb
, vd
);
5858 spa_config_exit(spa
, SCL_CONFIG
, FTAG
);
5862 increment_indirect_mapping_cb(void *arg
, const blkptr_t
*bp
, boolean_t bp_freed
,
5866 zdb_cb_t
*zcb
= arg
;
5867 spa_t
*spa
= zcb
->zcb_spa
;
5869 const dva_t
*dva
= &bp
->blk_dva
[0];
5872 ASSERT(!dump_opt
['L']);
5873 ASSERT3U(BP_GET_NDVAS(bp
), ==, 1);
5875 spa_config_enter(spa
, SCL_VDEV
, FTAG
, RW_READER
);
5876 vd
= vdev_lookup_top(zcb
->zcb_spa
, DVA_GET_VDEV(dva
));
5877 ASSERT3P(vd
, !=, NULL
);
5878 spa_config_exit(spa
, SCL_VDEV
, FTAG
);
5880 ASSERT(vd
->vdev_indirect_config
.vic_mapping_object
!= 0);
5881 ASSERT3P(zcb
->zcb_vd_obsolete_counts
[vd
->vdev_id
], !=, NULL
);
5883 vdev_indirect_mapping_increment_obsolete_count(
5884 vd
->vdev_indirect_mapping
,
5885 DVA_GET_OFFSET(dva
), DVA_GET_ASIZE(dva
),
5886 zcb
->zcb_vd_obsolete_counts
[vd
->vdev_id
]);
5892 zdb_load_obsolete_counts(vdev_t
*vd
)
5894 vdev_indirect_mapping_t
*vim
= vd
->vdev_indirect_mapping
;
5895 spa_t
*spa
= vd
->vdev_spa
;
5896 spa_condensing_indirect_phys_t
*scip
=
5897 &spa
->spa_condensing_indirect_phys
;
5898 uint64_t obsolete_sm_object
;
5901 VERIFY0(vdev_obsolete_sm_object(vd
, &obsolete_sm_object
));
5902 EQUIV(obsolete_sm_object
!= 0, vd
->vdev_obsolete_sm
!= NULL
);
5903 counts
= vdev_indirect_mapping_load_obsolete_counts(vim
);
5904 if (vd
->vdev_obsolete_sm
!= NULL
) {
5905 vdev_indirect_mapping_load_obsolete_spacemap(vim
, counts
,
5906 vd
->vdev_obsolete_sm
);
5908 if (scip
->scip_vdev
== vd
->vdev_id
&&
5909 scip
->scip_prev_obsolete_sm_object
!= 0) {
5910 space_map_t
*prev_obsolete_sm
= NULL
;
5911 VERIFY0(space_map_open(&prev_obsolete_sm
, spa
->spa_meta_objset
,
5912 scip
->scip_prev_obsolete_sm_object
, 0, vd
->vdev_asize
, 0));
5913 vdev_indirect_mapping_load_obsolete_spacemap(vim
, counts
,
5915 space_map_close(prev_obsolete_sm
);
5921 zdb_ddt_leak_init(spa_t
*spa
, zdb_cb_t
*zcb
)
5923 ddt_bookmark_t ddb
= {0};
5928 ASSERT(!dump_opt
['L']);
5930 while ((error
= ddt_walk(spa
, &ddb
, &dde
)) == 0) {
5932 ddt_phys_t
*ddp
= dde
.dde_phys
;
5934 if (ddb
.ddb_class
== DDT_CLASS_UNIQUE
)
5937 ASSERT(ddt_phys_total_refcnt(&dde
) > 1);
5939 for (p
= 0; p
< DDT_PHYS_TYPES
; p
++, ddp
++) {
5940 if (ddp
->ddp_phys_birth
== 0)
5942 ddt_bp_create(ddb
.ddb_checksum
,
5943 &dde
.dde_key
, ddp
, &blk
);
5944 if (p
== DDT_PHYS_DITTO
) {
5945 zdb_count_block(zcb
, NULL
, &blk
, ZDB_OT_DITTO
);
5947 zcb
->zcb_dedup_asize
+=
5948 BP_GET_ASIZE(&blk
) * (ddp
->ddp_refcnt
- 1);
5949 zcb
->zcb_dedup_blocks
++;
5952 ddt_t
*ddt
= spa
->spa_ddt
[ddb
.ddb_checksum
];
5954 VERIFY(ddt_lookup(ddt
, &blk
, B_TRUE
) != NULL
);
5958 ASSERT(error
== ENOENT
);
5961 typedef struct checkpoint_sm_exclude_entry_arg
{
5963 uint64_t cseea_checkpoint_size
;
5964 } checkpoint_sm_exclude_entry_arg_t
;
5967 checkpoint_sm_exclude_entry_cb(space_map_entry_t
*sme
, void *arg
)
5969 checkpoint_sm_exclude_entry_arg_t
*cseea
= arg
;
5970 vdev_t
*vd
= cseea
->cseea_vd
;
5971 metaslab_t
*ms
= vd
->vdev_ms
[sme
->sme_offset
>> vd
->vdev_ms_shift
];
5972 uint64_t end
= sme
->sme_offset
+ sme
->sme_run
;
5974 ASSERT(sme
->sme_type
== SM_FREE
);
5977 * Since the vdev_checkpoint_sm exists in the vdev level
5978 * and the ms_sm space maps exist in the metaslab level,
5979 * an entry in the checkpoint space map could theoretically
5980 * cross the boundaries of the metaslab that it belongs.
5982 * In reality, because of the way that we populate and
5983 * manipulate the checkpoint's space maps currently,
5984 * there shouldn't be any entries that cross metaslabs.
5985 * Hence the assertion below.
5987 * That said, there is no fundamental requirement that
5988 * the checkpoint's space map entries should not cross
5989 * metaslab boundaries. So if needed we could add code
5990 * that handles metaslab-crossing segments in the future.
5992 VERIFY3U(sme
->sme_offset
, >=, ms
->ms_start
);
5993 VERIFY3U(end
, <=, ms
->ms_start
+ ms
->ms_size
);
5996 * By removing the entry from the allocated segments we
5997 * also verify that the entry is there to begin with.
5999 mutex_enter(&ms
->ms_lock
);
6000 range_tree_remove(ms
->ms_allocatable
, sme
->sme_offset
, sme
->sme_run
);
6001 mutex_exit(&ms
->ms_lock
);
6003 cseea
->cseea_checkpoint_size
+= sme
->sme_run
;
6008 zdb_leak_init_vdev_exclude_checkpoint(vdev_t
*vd
, zdb_cb_t
*zcb
)
6010 spa_t
*spa
= vd
->vdev_spa
;
6011 space_map_t
*checkpoint_sm
= NULL
;
6012 uint64_t checkpoint_sm_obj
;
6015 * If there is no vdev_top_zap, we are in a pool whose
6016 * version predates the pool checkpoint feature.
6018 if (vd
->vdev_top_zap
== 0)
6022 * If there is no reference of the vdev_checkpoint_sm in
6023 * the vdev_top_zap, then one of the following scenarios
6026 * 1] There is no checkpoint
6027 * 2] There is a checkpoint, but no checkpointed blocks
6028 * have been freed yet
6029 * 3] The current vdev is indirect
6031 * In these cases we return immediately.
6033 if (zap_contains(spa_meta_objset(spa
), vd
->vdev_top_zap
,
6034 VDEV_TOP_ZAP_POOL_CHECKPOINT_SM
) != 0)
6037 VERIFY0(zap_lookup(spa_meta_objset(spa
), vd
->vdev_top_zap
,
6038 VDEV_TOP_ZAP_POOL_CHECKPOINT_SM
, sizeof (uint64_t), 1,
6039 &checkpoint_sm_obj
));
6041 checkpoint_sm_exclude_entry_arg_t cseea
;
6042 cseea
.cseea_vd
= vd
;
6043 cseea
.cseea_checkpoint_size
= 0;
6045 VERIFY0(space_map_open(&checkpoint_sm
, spa_meta_objset(spa
),
6046 checkpoint_sm_obj
, 0, vd
->vdev_asize
, vd
->vdev_ashift
));
6048 VERIFY0(space_map_iterate(checkpoint_sm
,
6049 space_map_length(checkpoint_sm
),
6050 checkpoint_sm_exclude_entry_cb
, &cseea
));
6051 space_map_close(checkpoint_sm
);
6053 zcb
->zcb_checkpoint_size
+= cseea
.cseea_checkpoint_size
;
6057 zdb_leak_init_exclude_checkpoint(spa_t
*spa
, zdb_cb_t
*zcb
)
6059 ASSERT(!dump_opt
['L']);
6061 vdev_t
*rvd
= spa
->spa_root_vdev
;
6062 for (uint64_t c
= 0; c
< rvd
->vdev_children
; c
++) {
6063 ASSERT3U(c
, ==, rvd
->vdev_child
[c
]->vdev_id
);
6064 zdb_leak_init_vdev_exclude_checkpoint(rvd
->vdev_child
[c
], zcb
);
6069 count_unflushed_space_cb(spa_t
*spa
, space_map_entry_t
*sme
,
6070 uint64_t txg
, void *arg
)
6072 int64_t *ualloc_space
= arg
;
6074 uint64_t offset
= sme
->sme_offset
;
6075 uint64_t vdev_id
= sme
->sme_vdev
;
6077 vdev_t
*vd
= vdev_lookup_top(spa
, vdev_id
);
6078 if (!vdev_is_concrete(vd
))
6081 metaslab_t
*ms
= vd
->vdev_ms
[offset
>> vd
->vdev_ms_shift
];
6082 ASSERT(sme
->sme_type
== SM_ALLOC
|| sme
->sme_type
== SM_FREE
);
6084 if (txg
< metaslab_unflushed_txg(ms
))
6087 if (sme
->sme_type
== SM_ALLOC
)
6088 *ualloc_space
+= sme
->sme_run
;
6090 *ualloc_space
-= sme
->sme_run
;
6096 get_unflushed_alloc_space(spa_t
*spa
)
6101 int64_t ualloc_space
= 0;
6102 iterate_through_spacemap_logs(spa
, count_unflushed_space_cb
,
6104 return (ualloc_space
);
6108 load_unflushed_cb(spa_t
*spa
, space_map_entry_t
*sme
, uint64_t txg
, void *arg
)
6110 maptype_t
*uic_maptype
= arg
;
6112 uint64_t offset
= sme
->sme_offset
;
6113 uint64_t size
= sme
->sme_run
;
6114 uint64_t vdev_id
= sme
->sme_vdev
;
6116 vdev_t
*vd
= vdev_lookup_top(spa
, vdev_id
);
6118 /* skip indirect vdevs */
6119 if (!vdev_is_concrete(vd
))
6122 metaslab_t
*ms
= vd
->vdev_ms
[offset
>> vd
->vdev_ms_shift
];
6124 ASSERT(sme
->sme_type
== SM_ALLOC
|| sme
->sme_type
== SM_FREE
);
6125 ASSERT(*uic_maptype
== SM_ALLOC
|| *uic_maptype
== SM_FREE
);
6127 if (txg
< metaslab_unflushed_txg(ms
))
6130 if (*uic_maptype
== sme
->sme_type
)
6131 range_tree_add(ms
->ms_allocatable
, offset
, size
);
6133 range_tree_remove(ms
->ms_allocatable
, offset
, size
);
6139 load_unflushed_to_ms_allocatables(spa_t
*spa
, maptype_t maptype
)
6141 iterate_through_spacemap_logs(spa
, load_unflushed_cb
, &maptype
);
6145 load_concrete_ms_allocatable_trees(spa_t
*spa
, maptype_t maptype
)
6147 vdev_t
*rvd
= spa
->spa_root_vdev
;
6148 for (uint64_t i
= 0; i
< rvd
->vdev_children
; i
++) {
6149 vdev_t
*vd
= rvd
->vdev_child
[i
];
6151 ASSERT3U(i
, ==, vd
->vdev_id
);
6153 if (vd
->vdev_ops
== &vdev_indirect_ops
)
6156 for (uint64_t m
= 0; m
< vd
->vdev_ms_count
; m
++) {
6157 metaslab_t
*msp
= vd
->vdev_ms
[m
];
6159 (void) fprintf(stderr
,
6160 "\rloading concrete vdev %llu, "
6161 "metaslab %llu of %llu ...",
6162 (longlong_t
)vd
->vdev_id
,
6163 (longlong_t
)msp
->ms_id
,
6164 (longlong_t
)vd
->vdev_ms_count
);
6166 mutex_enter(&msp
->ms_lock
);
6167 range_tree_vacate(msp
->ms_allocatable
, NULL
, NULL
);
6170 * We don't want to spend the CPU manipulating the
6171 * size-ordered tree, so clear the range_tree ops.
6173 msp
->ms_allocatable
->rt_ops
= NULL
;
6175 if (msp
->ms_sm
!= NULL
) {
6176 VERIFY0(space_map_load(msp
->ms_sm
,
6177 msp
->ms_allocatable
, maptype
));
6179 if (!msp
->ms_loaded
)
6180 msp
->ms_loaded
= B_TRUE
;
6181 mutex_exit(&msp
->ms_lock
);
6185 load_unflushed_to_ms_allocatables(spa
, maptype
);
6189 * vm_idxp is an in-out parameter which (for indirect vdevs) is the
6190 * index in vim_entries that has the first entry in this metaslab.
6191 * On return, it will be set to the first entry after this metaslab.
6194 load_indirect_ms_allocatable_tree(vdev_t
*vd
, metaslab_t
*msp
,
6197 vdev_indirect_mapping_t
*vim
= vd
->vdev_indirect_mapping
;
6199 mutex_enter(&msp
->ms_lock
);
6200 range_tree_vacate(msp
->ms_allocatable
, NULL
, NULL
);
6203 * We don't want to spend the CPU manipulating the
6204 * size-ordered tree, so clear the range_tree ops.
6206 msp
->ms_allocatable
->rt_ops
= NULL
;
6208 for (; *vim_idxp
< vdev_indirect_mapping_num_entries(vim
);
6210 vdev_indirect_mapping_entry_phys_t
*vimep
=
6211 &vim
->vim_entries
[*vim_idxp
];
6212 uint64_t ent_offset
= DVA_MAPPING_GET_SRC_OFFSET(vimep
);
6213 uint64_t ent_len
= DVA_GET_ASIZE(&vimep
->vimep_dst
);
6214 ASSERT3U(ent_offset
, >=, msp
->ms_start
);
6215 if (ent_offset
>= msp
->ms_start
+ msp
->ms_size
)
6219 * Mappings do not cross metaslab boundaries,
6220 * because we create them by walking the metaslabs.
6222 ASSERT3U(ent_offset
+ ent_len
, <=,
6223 msp
->ms_start
+ msp
->ms_size
);
6224 range_tree_add(msp
->ms_allocatable
, ent_offset
, ent_len
);
6227 if (!msp
->ms_loaded
)
6228 msp
->ms_loaded
= B_TRUE
;
6229 mutex_exit(&msp
->ms_lock
);
6233 zdb_leak_init_prepare_indirect_vdevs(spa_t
*spa
, zdb_cb_t
*zcb
)
6235 ASSERT(!dump_opt
['L']);
6237 vdev_t
*rvd
= spa
->spa_root_vdev
;
6238 for (uint64_t c
= 0; c
< rvd
->vdev_children
; c
++) {
6239 vdev_t
*vd
= rvd
->vdev_child
[c
];
6241 ASSERT3U(c
, ==, vd
->vdev_id
);
6243 if (vd
->vdev_ops
!= &vdev_indirect_ops
)
6247 * Note: we don't check for mapping leaks on
6248 * removing vdevs because their ms_allocatable's
6249 * are used to look for leaks in allocated space.
6251 zcb
->zcb_vd_obsolete_counts
[c
] = zdb_load_obsolete_counts(vd
);
6254 * Normally, indirect vdevs don't have any
6255 * metaslabs. We want to set them up for
6258 vdev_metaslab_group_create(vd
);
6259 VERIFY0(vdev_metaslab_init(vd
, 0));
6261 vdev_indirect_mapping_t
*vim __maybe_unused
=
6262 vd
->vdev_indirect_mapping
;
6263 uint64_t vim_idx
= 0;
6264 for (uint64_t m
= 0; m
< vd
->vdev_ms_count
; m
++) {
6266 (void) fprintf(stderr
,
6267 "\rloading indirect vdev %llu, "
6268 "metaslab %llu of %llu ...",
6269 (longlong_t
)vd
->vdev_id
,
6270 (longlong_t
)vd
->vdev_ms
[m
]->ms_id
,
6271 (longlong_t
)vd
->vdev_ms_count
);
6273 load_indirect_ms_allocatable_tree(vd
, vd
->vdev_ms
[m
],
6276 ASSERT3U(vim_idx
, ==, vdev_indirect_mapping_num_entries(vim
));
6281 zdb_leak_init(spa_t
*spa
, zdb_cb_t
*zcb
)
6288 dsl_pool_t
*dp
= spa
->spa_dsl_pool
;
6289 vdev_t
*rvd
= spa
->spa_root_vdev
;
6292 * We are going to be changing the meaning of the metaslab's
6293 * ms_allocatable. Ensure that the allocator doesn't try to
6296 spa
->spa_normal_class
->mc_ops
= &zdb_metaslab_ops
;
6297 spa
->spa_log_class
->mc_ops
= &zdb_metaslab_ops
;
6298 spa
->spa_embedded_log_class
->mc_ops
= &zdb_metaslab_ops
;
6300 zcb
->zcb_vd_obsolete_counts
=
6301 umem_zalloc(rvd
->vdev_children
* sizeof (uint32_t *),
6305 * For leak detection, we overload the ms_allocatable trees
6306 * to contain allocated segments instead of free segments.
6307 * As a result, we can't use the normal metaslab_load/unload
6310 zdb_leak_init_prepare_indirect_vdevs(spa
, zcb
);
6311 load_concrete_ms_allocatable_trees(spa
, SM_ALLOC
);
6314 * On load_concrete_ms_allocatable_trees() we loaded all the
6315 * allocated entries from the ms_sm to the ms_allocatable for
6316 * each metaslab. If the pool has a checkpoint or is in the
6317 * middle of discarding a checkpoint, some of these blocks
6318 * may have been freed but their ms_sm may not have been
6319 * updated because they are referenced by the checkpoint. In
6320 * order to avoid false-positives during leak-detection, we
6321 * go through the vdev's checkpoint space map and exclude all
6322 * its entries from their relevant ms_allocatable.
6324 * We also aggregate the space held by the checkpoint and add
6325 * it to zcb_checkpoint_size.
6327 * Note that at this point we are also verifying that all the
6328 * entries on the checkpoint_sm are marked as allocated in
6329 * the ms_sm of their relevant metaslab.
6330 * [see comment in checkpoint_sm_exclude_entry_cb()]
6332 zdb_leak_init_exclude_checkpoint(spa
, zcb
);
6333 ASSERT3U(zcb
->zcb_checkpoint_size
, ==, spa_get_checkpoint_space(spa
));
6335 /* for cleaner progress output */
6336 (void) fprintf(stderr
, "\n");
6338 if (bpobj_is_open(&dp
->dp_obsolete_bpobj
)) {
6339 ASSERT(spa_feature_is_enabled(spa
,
6340 SPA_FEATURE_DEVICE_REMOVAL
));
6341 (void) bpobj_iterate_nofree(&dp
->dp_obsolete_bpobj
,
6342 increment_indirect_mapping_cb
, zcb
, NULL
);
6345 spa_config_enter(spa
, SCL_CONFIG
, FTAG
, RW_READER
);
6346 zdb_ddt_leak_init(spa
, zcb
);
6347 spa_config_exit(spa
, SCL_CONFIG
, FTAG
);
6351 zdb_check_for_obsolete_leaks(vdev_t
*vd
, zdb_cb_t
*zcb
)
6353 boolean_t leaks
= B_FALSE
;
6354 vdev_indirect_mapping_t
*vim
= vd
->vdev_indirect_mapping
;
6355 uint64_t total_leaked
= 0;
6356 boolean_t are_precise
= B_FALSE
;
6358 ASSERT(vim
!= NULL
);
6360 for (uint64_t i
= 0; i
< vdev_indirect_mapping_num_entries(vim
); i
++) {
6361 vdev_indirect_mapping_entry_phys_t
*vimep
=
6362 &vim
->vim_entries
[i
];
6363 uint64_t obsolete_bytes
= 0;
6364 uint64_t offset
= DVA_MAPPING_GET_SRC_OFFSET(vimep
);
6365 metaslab_t
*msp
= vd
->vdev_ms
[offset
>> vd
->vdev_ms_shift
];
6368 * This is not very efficient but it's easy to
6369 * verify correctness.
6371 for (uint64_t inner_offset
= 0;
6372 inner_offset
< DVA_GET_ASIZE(&vimep
->vimep_dst
);
6373 inner_offset
+= 1ULL << vd
->vdev_ashift
) {
6374 if (range_tree_contains(msp
->ms_allocatable
,
6375 offset
+ inner_offset
, 1ULL << vd
->vdev_ashift
)) {
6376 obsolete_bytes
+= 1ULL << vd
->vdev_ashift
;
6380 int64_t bytes_leaked
= obsolete_bytes
-
6381 zcb
->zcb_vd_obsolete_counts
[vd
->vdev_id
][i
];
6382 ASSERT3U(DVA_GET_ASIZE(&vimep
->vimep_dst
), >=,
6383 zcb
->zcb_vd_obsolete_counts
[vd
->vdev_id
][i
]);
6385 VERIFY0(vdev_obsolete_counts_are_precise(vd
, &are_precise
));
6386 if (bytes_leaked
!= 0 && (are_precise
|| dump_opt
['d'] >= 5)) {
6387 (void) printf("obsolete indirect mapping count "
6388 "mismatch on %llu:%llx:%llx : %llx bytes leaked\n",
6389 (u_longlong_t
)vd
->vdev_id
,
6390 (u_longlong_t
)DVA_MAPPING_GET_SRC_OFFSET(vimep
),
6391 (u_longlong_t
)DVA_GET_ASIZE(&vimep
->vimep_dst
),
6392 (u_longlong_t
)bytes_leaked
);
6394 total_leaked
+= ABS(bytes_leaked
);
6397 VERIFY0(vdev_obsolete_counts_are_precise(vd
, &are_precise
));
6398 if (!are_precise
&& total_leaked
> 0) {
6399 int pct_leaked
= total_leaked
* 100 /
6400 vdev_indirect_mapping_bytes_mapped(vim
);
6401 (void) printf("cannot verify obsolete indirect mapping "
6402 "counts of vdev %llu because precise feature was not "
6403 "enabled when it was removed: %d%% (%llx bytes) of mapping"
6405 (u_longlong_t
)vd
->vdev_id
, pct_leaked
,
6406 (u_longlong_t
)total_leaked
);
6407 } else if (total_leaked
> 0) {
6408 (void) printf("obsolete indirect mapping count mismatch "
6409 "for vdev %llu -- %llx total bytes mismatched\n",
6410 (u_longlong_t
)vd
->vdev_id
,
6411 (u_longlong_t
)total_leaked
);
6415 vdev_indirect_mapping_free_obsolete_counts(vim
,
6416 zcb
->zcb_vd_obsolete_counts
[vd
->vdev_id
]);
6417 zcb
->zcb_vd_obsolete_counts
[vd
->vdev_id
] = NULL
;
6423 zdb_leak_fini(spa_t
*spa
, zdb_cb_t
*zcb
)
6428 boolean_t leaks
= B_FALSE
;
6429 vdev_t
*rvd
= spa
->spa_root_vdev
;
6430 for (unsigned c
= 0; c
< rvd
->vdev_children
; c
++) {
6431 vdev_t
*vd
= rvd
->vdev_child
[c
];
6433 if (zcb
->zcb_vd_obsolete_counts
[c
] != NULL
) {
6434 leaks
|= zdb_check_for_obsolete_leaks(vd
, zcb
);
6437 for (uint64_t m
= 0; m
< vd
->vdev_ms_count
; m
++) {
6438 metaslab_t
*msp
= vd
->vdev_ms
[m
];
6439 ASSERT3P(msp
->ms_group
, ==, (msp
->ms_group
->mg_class
==
6440 spa_embedded_log_class(spa
)) ?
6441 vd
->vdev_log_mg
: vd
->vdev_mg
);
6444 * ms_allocatable has been overloaded
6445 * to contain allocated segments. Now that
6446 * we finished traversing all blocks, any
6447 * block that remains in the ms_allocatable
6448 * represents an allocated block that we
6449 * did not claim during the traversal.
6450 * Claimed blocks would have been removed
6451 * from the ms_allocatable. For indirect
6452 * vdevs, space remaining in the tree
6453 * represents parts of the mapping that are
6454 * not referenced, which is not a bug.
6456 if (vd
->vdev_ops
== &vdev_indirect_ops
) {
6457 range_tree_vacate(msp
->ms_allocatable
,
6460 range_tree_vacate(msp
->ms_allocatable
,
6463 if (msp
->ms_loaded
) {
6464 msp
->ms_loaded
= B_FALSE
;
6469 umem_free(zcb
->zcb_vd_obsolete_counts
,
6470 rvd
->vdev_children
* sizeof (uint32_t *));
6471 zcb
->zcb_vd_obsolete_counts
= NULL
;
6477 count_block_cb(void *arg
, const blkptr_t
*bp
, dmu_tx_t
*tx
)
6480 zdb_cb_t
*zcb
= arg
;
6482 if (dump_opt
['b'] >= 5) {
6483 char blkbuf
[BP_SPRINTF_LEN
];
6484 snprintf_blkptr(blkbuf
, sizeof (blkbuf
), bp
);
6485 (void) printf("[%s] %s\n",
6486 "deferred free", blkbuf
);
6488 zdb_count_block(zcb
, NULL
, bp
, ZDB_OT_DEFERRED
);
6493 * Iterate over livelists which have been destroyed by the user but
6494 * are still present in the MOS, waiting to be freed
6497 iterate_deleted_livelists(spa_t
*spa
, ll_iter_t func
, void *arg
)
6499 objset_t
*mos
= spa
->spa_meta_objset
;
6501 int err
= zap_lookup(mos
, DMU_POOL_DIRECTORY_OBJECT
,
6502 DMU_POOL_DELETED_CLONES
, sizeof (uint64_t), 1, &zap_obj
);
6508 zap_attribute_t attr
;
6510 /* NULL out os prior to dsl_deadlist_open in case it's garbage */
6512 for (zap_cursor_init(&zc
, mos
, zap_obj
);
6513 zap_cursor_retrieve(&zc
, &attr
) == 0;
6514 (void) zap_cursor_advance(&zc
)) {
6515 dsl_deadlist_open(&ll
, mos
, attr
.za_first_integer
);
6517 dsl_deadlist_close(&ll
);
6519 zap_cursor_fini(&zc
);
6523 bpobj_count_block_cb(void *arg
, const blkptr_t
*bp
, boolean_t bp_freed
,
6527 return (count_block_cb(arg
, bp
, tx
));
6531 livelist_entry_count_blocks_cb(void *args
, dsl_deadlist_entry_t
*dle
)
6533 zdb_cb_t
*zbc
= args
;
6535 bplist_create(&blks
);
6536 /* determine which blocks have been alloc'd but not freed */
6537 VERIFY0(dsl_process_sub_livelist(&dle
->dle_bpobj
, &blks
, NULL
, NULL
));
6538 /* count those blocks */
6539 (void) bplist_iterate(&blks
, count_block_cb
, zbc
, NULL
);
6540 bplist_destroy(&blks
);
6545 livelist_count_blocks(dsl_deadlist_t
*ll
, void *arg
)
6547 dsl_deadlist_iterate(ll
, livelist_entry_count_blocks_cb
, arg
);
6551 * Count the blocks in the livelists that have been destroyed by the user
6552 * but haven't yet been freed.
6555 deleted_livelists_count_blocks(spa_t
*spa
, zdb_cb_t
*zbc
)
6557 iterate_deleted_livelists(spa
, livelist_count_blocks
, zbc
);
6561 dump_livelist_cb(dsl_deadlist_t
*ll
, void *arg
)
6563 ASSERT3P(arg
, ==, NULL
);
6564 global_feature_count
[SPA_FEATURE_LIVELIST
]++;
6565 dump_blkptr_list(ll
, "Deleted Livelist");
6566 dsl_deadlist_iterate(ll
, sublivelist_verify_lightweight
, NULL
);
6570 * Print out, register object references to, and increment feature counts for
6571 * livelists that have been destroyed by the user but haven't yet been freed.
6574 deleted_livelists_dump_mos(spa_t
*spa
)
6577 objset_t
*mos
= spa
->spa_meta_objset
;
6578 int err
= zap_lookup(mos
, DMU_POOL_DIRECTORY_OBJECT
,
6579 DMU_POOL_DELETED_CLONES
, sizeof (uint64_t), 1, &zap_obj
);
6582 mos_obj_refd(zap_obj
);
6583 iterate_deleted_livelists(spa
, dump_livelist_cb
, NULL
);
6587 dump_block_stats(spa_t
*spa
)
6590 zdb_blkstats_t
*zb
, *tzb
;
6591 uint64_t norm_alloc
, norm_space
, total_alloc
, total_found
;
6592 int flags
= TRAVERSE_PRE
| TRAVERSE_PREFETCH_METADATA
|
6593 TRAVERSE_NO_DECRYPT
| TRAVERSE_HARD
;
6594 boolean_t leaks
= B_FALSE
;
6596 bp_embedded_type_t i
;
6598 zcb
= umem_zalloc(sizeof (zdb_cb_t
), UMEM_NOFAIL
);
6600 (void) printf("\nTraversing all blocks %s%s%s%s%s...\n\n",
6601 (dump_opt
['c'] || !dump_opt
['L']) ? "to verify " : "",
6602 (dump_opt
['c'] == 1) ? "metadata " : "",
6603 dump_opt
['c'] ? "checksums " : "",
6604 (dump_opt
['c'] && !dump_opt
['L']) ? "and verify " : "",
6605 !dump_opt
['L'] ? "nothing leaked " : "");
6608 * When leak detection is enabled we load all space maps as SM_ALLOC
6609 * maps, then traverse the pool claiming each block we discover. If
6610 * the pool is perfectly consistent, the segment trees will be empty
6611 * when we're done. Anything left over is a leak; any block we can't
6612 * claim (because it's not part of any space map) is a double
6613 * allocation, reference to a freed block, or an unclaimed log block.
6615 * When leak detection is disabled (-L option) we still traverse the
6616 * pool claiming each block we discover, but we skip opening any space
6619 zdb_leak_init(spa
, zcb
);
6622 * If there's a deferred-free bplist, process that first.
6624 (void) bpobj_iterate_nofree(&spa
->spa_deferred_bpobj
,
6625 bpobj_count_block_cb
, zcb
, NULL
);
6627 if (spa_version(spa
) >= SPA_VERSION_DEADLISTS
) {
6628 (void) bpobj_iterate_nofree(&spa
->spa_dsl_pool
->dp_free_bpobj
,
6629 bpobj_count_block_cb
, zcb
, NULL
);
6632 zdb_claim_removing(spa
, zcb
);
6634 if (spa_feature_is_active(spa
, SPA_FEATURE_ASYNC_DESTROY
)) {
6635 VERIFY3U(0, ==, bptree_iterate(spa
->spa_meta_objset
,
6636 spa
->spa_dsl_pool
->dp_bptree_obj
, B_FALSE
, count_block_cb
,
6640 deleted_livelists_count_blocks(spa
, zcb
);
6642 if (dump_opt
['c'] > 1)
6643 flags
|= TRAVERSE_PREFETCH_DATA
;
6645 zcb
->zcb_totalasize
= metaslab_class_get_alloc(spa_normal_class(spa
));
6646 zcb
->zcb_totalasize
+= metaslab_class_get_alloc(spa_special_class(spa
));
6647 zcb
->zcb_totalasize
+= metaslab_class_get_alloc(spa_dedup_class(spa
));
6648 zcb
->zcb_totalasize
+=
6649 metaslab_class_get_alloc(spa_embedded_log_class(spa
));
6650 zcb
->zcb_start
= zcb
->zcb_lastprint
= gethrtime();
6651 err
= traverse_pool(spa
, 0, flags
, zdb_blkptr_cb
, zcb
);
6654 * If we've traversed the data blocks then we need to wait for those
6655 * I/Os to complete. We leverage "The Godfather" zio to wait on
6656 * all async I/Os to complete.
6658 if (dump_opt
['c']) {
6659 for (c
= 0; c
< max_ncpus
; c
++) {
6660 (void) zio_wait(spa
->spa_async_zio_root
[c
]);
6661 spa
->spa_async_zio_root
[c
] = zio_root(spa
, NULL
, NULL
,
6662 ZIO_FLAG_CANFAIL
| ZIO_FLAG_SPECULATIVE
|
6663 ZIO_FLAG_GODFATHER
);
6666 ASSERT0(spa
->spa_load_verify_bytes
);
6669 * Done after zio_wait() since zcb_haderrors is modified in
6672 zcb
->zcb_haderrors
|= err
;
6674 if (zcb
->zcb_haderrors
) {
6675 (void) printf("\nError counts:\n\n");
6676 (void) printf("\t%5s %s\n", "errno", "count");
6677 for (e
= 0; e
< 256; e
++) {
6678 if (zcb
->zcb_errors
[e
] != 0) {
6679 (void) printf("\t%5d %llu\n",
6680 e
, (u_longlong_t
)zcb
->zcb_errors
[e
]);
6686 * Report any leaked segments.
6688 leaks
|= zdb_leak_fini(spa
, zcb
);
6690 tzb
= &zcb
->zcb_type
[ZB_TOTAL
][ZDB_OT_TOTAL
];
6692 norm_alloc
= metaslab_class_get_alloc(spa_normal_class(spa
));
6693 norm_space
= metaslab_class_get_space(spa_normal_class(spa
));
6695 total_alloc
= norm_alloc
+
6696 metaslab_class_get_alloc(spa_log_class(spa
)) +
6697 metaslab_class_get_alloc(spa_embedded_log_class(spa
)) +
6698 metaslab_class_get_alloc(spa_special_class(spa
)) +
6699 metaslab_class_get_alloc(spa_dedup_class(spa
)) +
6700 get_unflushed_alloc_space(spa
);
6701 total_found
= tzb
->zb_asize
- zcb
->zcb_dedup_asize
+
6702 zcb
->zcb_removing_size
+ zcb
->zcb_checkpoint_size
;
6704 if (total_found
== total_alloc
&& !dump_opt
['L']) {
6705 (void) printf("\n\tNo leaks (block sum matches space"
6706 " maps exactly)\n");
6707 } else if (!dump_opt
['L']) {
6708 (void) printf("block traversal size %llu != alloc %llu "
6710 (u_longlong_t
)total_found
,
6711 (u_longlong_t
)total_alloc
,
6712 (dump_opt
['L']) ? "unreachable" : "leaked",
6713 (longlong_t
)(total_alloc
- total_found
));
6717 if (tzb
->zb_count
== 0) {
6718 umem_free(zcb
, sizeof (zdb_cb_t
));
6722 (void) printf("\n");
6723 (void) printf("\t%-16s %14llu\n", "bp count:",
6724 (u_longlong_t
)tzb
->zb_count
);
6725 (void) printf("\t%-16s %14llu\n", "ganged count:",
6726 (longlong_t
)tzb
->zb_gangs
);
6727 (void) printf("\t%-16s %14llu avg: %6llu\n", "bp logical:",
6728 (u_longlong_t
)tzb
->zb_lsize
,
6729 (u_longlong_t
)(tzb
->zb_lsize
/ tzb
->zb_count
));
6730 (void) printf("\t%-16s %14llu avg: %6llu compression: %6.2f\n",
6731 "bp physical:", (u_longlong_t
)tzb
->zb_psize
,
6732 (u_longlong_t
)(tzb
->zb_psize
/ tzb
->zb_count
),
6733 (double)tzb
->zb_lsize
/ tzb
->zb_psize
);
6734 (void) printf("\t%-16s %14llu avg: %6llu compression: %6.2f\n",
6735 "bp allocated:", (u_longlong_t
)tzb
->zb_asize
,
6736 (u_longlong_t
)(tzb
->zb_asize
/ tzb
->zb_count
),
6737 (double)tzb
->zb_lsize
/ tzb
->zb_asize
);
6738 (void) printf("\t%-16s %14llu ref>1: %6llu deduplication: %6.2f\n",
6739 "bp deduped:", (u_longlong_t
)zcb
->zcb_dedup_asize
,
6740 (u_longlong_t
)zcb
->zcb_dedup_blocks
,
6741 (double)zcb
->zcb_dedup_asize
/ tzb
->zb_asize
+ 1.0);
6742 (void) printf("\t%-16s %14llu used: %5.2f%%\n", "Normal class:",
6743 (u_longlong_t
)norm_alloc
, 100.0 * norm_alloc
/ norm_space
);
6745 if (spa_special_class(spa
)->mc_allocator
[0].mca_rotor
!= NULL
) {
6746 uint64_t alloc
= metaslab_class_get_alloc(
6747 spa_special_class(spa
));
6748 uint64_t space
= metaslab_class_get_space(
6749 spa_special_class(spa
));
6751 (void) printf("\t%-16s %14llu used: %5.2f%%\n",
6752 "Special class", (u_longlong_t
)alloc
,
6753 100.0 * alloc
/ space
);
6756 if (spa_dedup_class(spa
)->mc_allocator
[0].mca_rotor
!= NULL
) {
6757 uint64_t alloc
= metaslab_class_get_alloc(
6758 spa_dedup_class(spa
));
6759 uint64_t space
= metaslab_class_get_space(
6760 spa_dedup_class(spa
));
6762 (void) printf("\t%-16s %14llu used: %5.2f%%\n",
6763 "Dedup class", (u_longlong_t
)alloc
,
6764 100.0 * alloc
/ space
);
6767 if (spa_embedded_log_class(spa
)->mc_allocator
[0].mca_rotor
!= NULL
) {
6768 uint64_t alloc
= metaslab_class_get_alloc(
6769 spa_embedded_log_class(spa
));
6770 uint64_t space
= metaslab_class_get_space(
6771 spa_embedded_log_class(spa
));
6773 (void) printf("\t%-16s %14llu used: %5.2f%%\n",
6774 "Embedded log class", (u_longlong_t
)alloc
,
6775 100.0 * alloc
/ space
);
6778 for (i
= 0; i
< NUM_BP_EMBEDDED_TYPES
; i
++) {
6779 if (zcb
->zcb_embedded_blocks
[i
] == 0)
6781 (void) printf("\n");
6782 (void) printf("\tadditional, non-pointer bps of type %u: "
6784 i
, (u_longlong_t
)zcb
->zcb_embedded_blocks
[i
]);
6786 if (dump_opt
['b'] >= 3) {
6787 (void) printf("\t number of (compressed) bytes: "
6789 dump_histogram(zcb
->zcb_embedded_histogram
[i
],
6790 sizeof (zcb
->zcb_embedded_histogram
[i
]) /
6791 sizeof (zcb
->zcb_embedded_histogram
[i
][0]), 0);
6795 if (tzb
->zb_ditto_samevdev
!= 0) {
6796 (void) printf("\tDittoed blocks on same vdev: %llu\n",
6797 (longlong_t
)tzb
->zb_ditto_samevdev
);
6799 if (tzb
->zb_ditto_same_ms
!= 0) {
6800 (void) printf("\tDittoed blocks in same metaslab: %llu\n",
6801 (longlong_t
)tzb
->zb_ditto_same_ms
);
6804 for (uint64_t v
= 0; v
< spa
->spa_root_vdev
->vdev_children
; v
++) {
6805 vdev_t
*vd
= spa
->spa_root_vdev
->vdev_child
[v
];
6806 vdev_indirect_mapping_t
*vim
= vd
->vdev_indirect_mapping
;
6813 zdb_nicenum(vdev_indirect_mapping_num_entries(vim
),
6814 mem
, vdev_indirect_mapping_size(vim
));
6816 (void) printf("\tindirect vdev id %llu has %llu segments "
6818 (longlong_t
)vd
->vdev_id
,
6819 (longlong_t
)vdev_indirect_mapping_num_entries(vim
), mem
);
6822 if (dump_opt
['b'] >= 2) {
6824 char csize
[32], lsize
[32], psize
[32], asize
[32];
6825 char avg
[32], gang
[32];
6826 (void) printf("\nBlocks\tLSIZE\tPSIZE\tASIZE"
6827 "\t avg\t comp\t%%Total\tType\n");
6829 zfs_blkstat_t
*mdstats
= umem_zalloc(sizeof (zfs_blkstat_t
),
6832 for (t
= 0; t
<= ZDB_OT_TOTAL
; t
++) {
6833 const char *typename
;
6835 /* make sure nicenum has enough space */
6836 _Static_assert(sizeof (csize
) >= NN_NUMBUF_SZ
,
6838 _Static_assert(sizeof (lsize
) >= NN_NUMBUF_SZ
,
6840 _Static_assert(sizeof (psize
) >= NN_NUMBUF_SZ
,
6842 _Static_assert(sizeof (asize
) >= NN_NUMBUF_SZ
,
6844 _Static_assert(sizeof (avg
) >= NN_NUMBUF_SZ
,
6846 _Static_assert(sizeof (gang
) >= NN_NUMBUF_SZ
,
6849 if (t
< DMU_OT_NUMTYPES
)
6850 typename
= dmu_ot
[t
].ot_name
;
6852 typename
= zdb_ot_extname
[t
- DMU_OT_NUMTYPES
];
6854 if (zcb
->zcb_type
[ZB_TOTAL
][t
].zb_asize
== 0) {
6855 (void) printf("%6s\t%5s\t%5s\t%5s"
6856 "\t%5s\t%5s\t%6s\t%s\n",
6868 for (l
= ZB_TOTAL
- 1; l
>= -1; l
--) {
6869 level
= (l
== -1 ? ZB_TOTAL
: l
);
6870 zb
= &zcb
->zcb_type
[level
][t
];
6872 if (zb
->zb_asize
== 0)
6875 if (level
!= ZB_TOTAL
&& t
< DMU_OT_NUMTYPES
&&
6876 (level
> 0 || DMU_OT_IS_METADATA(t
))) {
6877 mdstats
->zb_count
+= zb
->zb_count
;
6878 mdstats
->zb_lsize
+= zb
->zb_lsize
;
6879 mdstats
->zb_psize
+= zb
->zb_psize
;
6880 mdstats
->zb_asize
+= zb
->zb_asize
;
6881 mdstats
->zb_gangs
+= zb
->zb_gangs
;
6884 if (dump_opt
['b'] < 3 && level
!= ZB_TOTAL
)
6887 if (level
== 0 && zb
->zb_asize
==
6888 zcb
->zcb_type
[ZB_TOTAL
][t
].zb_asize
)
6891 zdb_nicenum(zb
->zb_count
, csize
,
6893 zdb_nicenum(zb
->zb_lsize
, lsize
,
6895 zdb_nicenum(zb
->zb_psize
, psize
,
6897 zdb_nicenum(zb
->zb_asize
, asize
,
6899 zdb_nicenum(zb
->zb_asize
/ zb
->zb_count
, avg
,
6901 zdb_nicenum(zb
->zb_gangs
, gang
, sizeof (gang
));
6903 (void) printf("%6s\t%5s\t%5s\t%5s\t%5s"
6905 csize
, lsize
, psize
, asize
, avg
,
6906 (double)zb
->zb_lsize
/ zb
->zb_psize
,
6907 100.0 * zb
->zb_asize
/ tzb
->zb_asize
);
6909 if (level
== ZB_TOTAL
)
6910 (void) printf("%s\n", typename
);
6912 (void) printf(" L%d %s\n",
6915 if (dump_opt
['b'] >= 3 && zb
->zb_gangs
> 0) {
6916 (void) printf("\t number of ganged "
6917 "blocks: %s\n", gang
);
6920 if (dump_opt
['b'] >= 4) {
6921 (void) printf("psize "
6922 "(in 512-byte sectors): "
6923 "number of blocks\n");
6924 dump_histogram(zb
->zb_psize_histogram
,
6925 PSIZE_HISTO_SIZE
, 0);
6929 zdb_nicenum(mdstats
->zb_count
, csize
,
6931 zdb_nicenum(mdstats
->zb_lsize
, lsize
,
6933 zdb_nicenum(mdstats
->zb_psize
, psize
,
6935 zdb_nicenum(mdstats
->zb_asize
, asize
,
6937 zdb_nicenum(mdstats
->zb_asize
/ mdstats
->zb_count
, avg
,
6939 zdb_nicenum(mdstats
->zb_gangs
, gang
, sizeof (gang
));
6941 (void) printf("%6s\t%5s\t%5s\t%5s\t%5s"
6943 csize
, lsize
, psize
, asize
, avg
,
6944 (double)mdstats
->zb_lsize
/ mdstats
->zb_psize
,
6945 100.0 * mdstats
->zb_asize
/ tzb
->zb_asize
);
6946 (void) printf("%s\n", "Metadata Total");
6948 /* Output a table summarizing block sizes in the pool */
6949 if (dump_opt
['b'] >= 2) {
6950 dump_size_histograms(zcb
);
6953 umem_free(mdstats
, sizeof (zfs_blkstat_t
));
6956 (void) printf("\n");
6959 umem_free(zcb
, sizeof (zdb_cb_t
));
6963 if (zcb
->zcb_haderrors
) {
6964 umem_free(zcb
, sizeof (zdb_cb_t
));
6968 umem_free(zcb
, sizeof (zdb_cb_t
));
6972 typedef struct zdb_ddt_entry
{
6974 uint64_t zdde_ref_blocks
;
6975 uint64_t zdde_ref_lsize
;
6976 uint64_t zdde_ref_psize
;
6977 uint64_t zdde_ref_dsize
;
6978 avl_node_t zdde_node
;
6982 zdb_ddt_add_cb(spa_t
*spa
, zilog_t
*zilog
, const blkptr_t
*bp
,
6983 const zbookmark_phys_t
*zb
, const dnode_phys_t
*dnp
, void *arg
)
6985 (void) zilog
, (void) dnp
;
6986 avl_tree_t
*t
= arg
;
6988 zdb_ddt_entry_t
*zdde
, zdde_search
;
6990 if (zb
->zb_level
== ZB_DNODE_LEVEL
|| BP_IS_HOLE(bp
) ||
6994 if (dump_opt
['S'] > 1 && zb
->zb_level
== ZB_ROOT_LEVEL
) {
6995 (void) printf("traversing objset %llu, %llu objects, "
6996 "%lu blocks so far\n",
6997 (u_longlong_t
)zb
->zb_objset
,
6998 (u_longlong_t
)BP_GET_FILL(bp
),
7002 if (BP_IS_HOLE(bp
) || BP_GET_CHECKSUM(bp
) == ZIO_CHECKSUM_OFF
||
7003 BP_GET_LEVEL(bp
) > 0 || DMU_OT_IS_METADATA(BP_GET_TYPE(bp
)))
7006 ddt_key_fill(&zdde_search
.zdde_key
, bp
);
7008 zdde
= avl_find(t
, &zdde_search
, &where
);
7011 zdde
= umem_zalloc(sizeof (*zdde
), UMEM_NOFAIL
);
7012 zdde
->zdde_key
= zdde_search
.zdde_key
;
7013 avl_insert(t
, zdde
, where
);
7016 zdde
->zdde_ref_blocks
+= 1;
7017 zdde
->zdde_ref_lsize
+= BP_GET_LSIZE(bp
);
7018 zdde
->zdde_ref_psize
+= BP_GET_PSIZE(bp
);
7019 zdde
->zdde_ref_dsize
+= bp_get_dsize_sync(spa
, bp
);
7025 dump_simulated_ddt(spa_t
*spa
)
7028 void *cookie
= NULL
;
7029 zdb_ddt_entry_t
*zdde
;
7030 ddt_histogram_t ddh_total
= {{{0}}};
7031 ddt_stat_t dds_total
= {0};
7033 avl_create(&t
, ddt_entry_compare
,
7034 sizeof (zdb_ddt_entry_t
), offsetof(zdb_ddt_entry_t
, zdde_node
));
7036 spa_config_enter(spa
, SCL_CONFIG
, FTAG
, RW_READER
);
7038 (void) traverse_pool(spa
, 0, TRAVERSE_PRE
| TRAVERSE_PREFETCH_METADATA
|
7039 TRAVERSE_NO_DECRYPT
, zdb_ddt_add_cb
, &t
);
7041 spa_config_exit(spa
, SCL_CONFIG
, FTAG
);
7043 while ((zdde
= avl_destroy_nodes(&t
, &cookie
)) != NULL
) {
7045 uint64_t refcnt
= zdde
->zdde_ref_blocks
;
7046 ASSERT(refcnt
!= 0);
7048 dds
.dds_blocks
= zdde
->zdde_ref_blocks
/ refcnt
;
7049 dds
.dds_lsize
= zdde
->zdde_ref_lsize
/ refcnt
;
7050 dds
.dds_psize
= zdde
->zdde_ref_psize
/ refcnt
;
7051 dds
.dds_dsize
= zdde
->zdde_ref_dsize
/ refcnt
;
7053 dds
.dds_ref_blocks
= zdde
->zdde_ref_blocks
;
7054 dds
.dds_ref_lsize
= zdde
->zdde_ref_lsize
;
7055 dds
.dds_ref_psize
= zdde
->zdde_ref_psize
;
7056 dds
.dds_ref_dsize
= zdde
->zdde_ref_dsize
;
7058 ddt_stat_add(&ddh_total
.ddh_stat
[highbit64(refcnt
) - 1],
7061 umem_free(zdde
, sizeof (*zdde
));
7066 ddt_histogram_stat(&dds_total
, &ddh_total
);
7068 (void) printf("Simulated DDT histogram:\n");
7070 zpool_dump_ddt(&dds_total
, &ddh_total
);
7072 dump_dedup_ratio(&dds_total
);
7076 verify_device_removal_feature_counts(spa_t
*spa
)
7078 uint64_t dr_feature_refcount
= 0;
7079 uint64_t oc_feature_refcount
= 0;
7080 uint64_t indirect_vdev_count
= 0;
7081 uint64_t precise_vdev_count
= 0;
7082 uint64_t obsolete_counts_object_count
= 0;
7083 uint64_t obsolete_sm_count
= 0;
7084 uint64_t obsolete_counts_count
= 0;
7085 uint64_t scip_count
= 0;
7086 uint64_t obsolete_bpobj_count
= 0;
7089 spa_condensing_indirect_phys_t
*scip
=
7090 &spa
->spa_condensing_indirect_phys
;
7091 if (scip
->scip_next_mapping_object
!= 0) {
7092 vdev_t
*vd
= spa
->spa_root_vdev
->vdev_child
[scip
->scip_vdev
];
7093 ASSERT(scip
->scip_prev_obsolete_sm_object
!= 0);
7094 ASSERT3P(vd
->vdev_ops
, ==, &vdev_indirect_ops
);
7096 (void) printf("Condensing indirect vdev %llu: new mapping "
7097 "object %llu, prev obsolete sm %llu\n",
7098 (u_longlong_t
)scip
->scip_vdev
,
7099 (u_longlong_t
)scip
->scip_next_mapping_object
,
7100 (u_longlong_t
)scip
->scip_prev_obsolete_sm_object
);
7101 if (scip
->scip_prev_obsolete_sm_object
!= 0) {
7102 space_map_t
*prev_obsolete_sm
= NULL
;
7103 VERIFY0(space_map_open(&prev_obsolete_sm
,
7104 spa
->spa_meta_objset
,
7105 scip
->scip_prev_obsolete_sm_object
,
7106 0, vd
->vdev_asize
, 0));
7107 dump_spacemap(spa
->spa_meta_objset
, prev_obsolete_sm
);
7108 (void) printf("\n");
7109 space_map_close(prev_obsolete_sm
);
7115 for (uint64_t i
= 0; i
< spa
->spa_root_vdev
->vdev_children
; i
++) {
7116 vdev_t
*vd
= spa
->spa_root_vdev
->vdev_child
[i
];
7117 vdev_indirect_config_t
*vic
= &vd
->vdev_indirect_config
;
7119 if (vic
->vic_mapping_object
!= 0) {
7120 ASSERT(vd
->vdev_ops
== &vdev_indirect_ops
||
7122 indirect_vdev_count
++;
7124 if (vd
->vdev_indirect_mapping
->vim_havecounts
) {
7125 obsolete_counts_count
++;
7129 boolean_t are_precise
;
7130 VERIFY0(vdev_obsolete_counts_are_precise(vd
, &are_precise
));
7132 ASSERT(vic
->vic_mapping_object
!= 0);
7133 precise_vdev_count
++;
7136 uint64_t obsolete_sm_object
;
7137 VERIFY0(vdev_obsolete_sm_object(vd
, &obsolete_sm_object
));
7138 if (obsolete_sm_object
!= 0) {
7139 ASSERT(vic
->vic_mapping_object
!= 0);
7140 obsolete_sm_count
++;
7144 (void) feature_get_refcount(spa
,
7145 &spa_feature_table
[SPA_FEATURE_DEVICE_REMOVAL
],
7146 &dr_feature_refcount
);
7147 (void) feature_get_refcount(spa
,
7148 &spa_feature_table
[SPA_FEATURE_OBSOLETE_COUNTS
],
7149 &oc_feature_refcount
);
7151 if (dr_feature_refcount
!= indirect_vdev_count
) {
7153 (void) printf("Number of indirect vdevs (%llu) " \
7154 "does not match feature count (%llu)\n",
7155 (u_longlong_t
)indirect_vdev_count
,
7156 (u_longlong_t
)dr_feature_refcount
);
7158 (void) printf("Verified device_removal feature refcount " \
7159 "of %llu is correct\n",
7160 (u_longlong_t
)dr_feature_refcount
);
7163 if (zap_contains(spa_meta_objset(spa
), DMU_POOL_DIRECTORY_OBJECT
,
7164 DMU_POOL_OBSOLETE_BPOBJ
) == 0) {
7165 obsolete_bpobj_count
++;
7169 obsolete_counts_object_count
= precise_vdev_count
;
7170 obsolete_counts_object_count
+= obsolete_sm_count
;
7171 obsolete_counts_object_count
+= obsolete_counts_count
;
7172 obsolete_counts_object_count
+= scip_count
;
7173 obsolete_counts_object_count
+= obsolete_bpobj_count
;
7174 obsolete_counts_object_count
+= remap_deadlist_count
;
7176 if (oc_feature_refcount
!= obsolete_counts_object_count
) {
7178 (void) printf("Number of obsolete counts objects (%llu) " \
7179 "does not match feature count (%llu)\n",
7180 (u_longlong_t
)obsolete_counts_object_count
,
7181 (u_longlong_t
)oc_feature_refcount
);
7182 (void) printf("pv:%llu os:%llu oc:%llu sc:%llu "
7183 "ob:%llu rd:%llu\n",
7184 (u_longlong_t
)precise_vdev_count
,
7185 (u_longlong_t
)obsolete_sm_count
,
7186 (u_longlong_t
)obsolete_counts_count
,
7187 (u_longlong_t
)scip_count
,
7188 (u_longlong_t
)obsolete_bpobj_count
,
7189 (u_longlong_t
)remap_deadlist_count
);
7191 (void) printf("Verified indirect_refcount feature refcount " \
7192 "of %llu is correct\n",
7193 (u_longlong_t
)oc_feature_refcount
);
7199 zdb_set_skip_mmp(char *target
)
7204 * Disable the activity check to allow examination of
7207 mutex_enter(&spa_namespace_lock
);
7208 if ((spa
= spa_lookup(target
)) != NULL
) {
7209 spa
->spa_import_flags
|= ZFS_IMPORT_SKIP_MMP
;
7211 mutex_exit(&spa_namespace_lock
);
7214 #define BOGUS_SUFFIX "_CHECKPOINTED_UNIVERSE"
7216 * Import the checkpointed state of the pool specified by the target
7217 * parameter as readonly. The function also accepts a pool config
7218 * as an optional parameter, else it attempts to infer the config by
7219 * the name of the target pool.
7221 * Note that the checkpointed state's pool name will be the name of
7222 * the original pool with the above suffix appended to it. In addition,
7223 * if the target is not a pool name (e.g. a path to a dataset) then
7224 * the new_path parameter is populated with the updated path to
7225 * reflect the fact that we are looking into the checkpointed state.
7227 * The function returns a newly-allocated copy of the name of the
7228 * pool containing the checkpointed state. When this copy is no
7229 * longer needed it should be freed with free(3C). Same thing
7230 * applies to the new_path parameter if allocated.
7233 import_checkpointed_state(char *target
, nvlist_t
*cfg
, char **new_path
)
7236 char *poolname
, *bogus_name
= NULL
;
7237 boolean_t freecfg
= B_FALSE
;
7239 /* If the target is not a pool, the extract the pool name */
7240 char *path_start
= strchr(target
, '/');
7241 if (path_start
!= NULL
) {
7242 size_t poolname_len
= path_start
- target
;
7243 poolname
= strndup(target
, poolname_len
);
7249 zdb_set_skip_mmp(poolname
);
7250 error
= spa_get_stats(poolname
, &cfg
, NULL
, 0);
7252 fatal("Tried to read config of pool \"%s\" but "
7253 "spa_get_stats() failed with error %d\n",
7259 if (asprintf(&bogus_name
, "%s%s", poolname
, BOGUS_SUFFIX
) == -1) {
7260 if (target
!= poolname
)
7264 fnvlist_add_string(cfg
, ZPOOL_CONFIG_POOL_NAME
, bogus_name
);
7266 error
= spa_import(bogus_name
, cfg
, NULL
,
7267 ZFS_IMPORT_MISSING_LOG
| ZFS_IMPORT_CHECKPOINT
|
7268 ZFS_IMPORT_SKIP_MMP
);
7272 fatal("Tried to import pool \"%s\" but spa_import() failed "
7273 "with error %d\n", bogus_name
, error
);
7276 if (new_path
!= NULL
&& path_start
!= NULL
) {
7277 if (asprintf(new_path
, "%s%s", bogus_name
, path_start
) == -1) {
7279 if (path_start
!= NULL
)
7285 if (target
!= poolname
)
7288 return (bogus_name
);
7291 typedef struct verify_checkpoint_sm_entry_cb_arg
{
7294 /* the following fields are only used for printing progress */
7295 uint64_t vcsec_entryid
;
7296 uint64_t vcsec_num_entries
;
7297 } verify_checkpoint_sm_entry_cb_arg_t
;
7299 #define ENTRIES_PER_PROGRESS_UPDATE 10000
7302 verify_checkpoint_sm_entry_cb(space_map_entry_t
*sme
, void *arg
)
7304 verify_checkpoint_sm_entry_cb_arg_t
*vcsec
= arg
;
7305 vdev_t
*vd
= vcsec
->vcsec_vd
;
7306 metaslab_t
*ms
= vd
->vdev_ms
[sme
->sme_offset
>> vd
->vdev_ms_shift
];
7307 uint64_t end
= sme
->sme_offset
+ sme
->sme_run
;
7309 ASSERT(sme
->sme_type
== SM_FREE
);
7311 if ((vcsec
->vcsec_entryid
% ENTRIES_PER_PROGRESS_UPDATE
) == 0) {
7312 (void) fprintf(stderr
,
7313 "\rverifying vdev %llu, space map entry %llu of %llu ...",
7314 (longlong_t
)vd
->vdev_id
,
7315 (longlong_t
)vcsec
->vcsec_entryid
,
7316 (longlong_t
)vcsec
->vcsec_num_entries
);
7318 vcsec
->vcsec_entryid
++;
7321 * See comment in checkpoint_sm_exclude_entry_cb()
7323 VERIFY3U(sme
->sme_offset
, >=, ms
->ms_start
);
7324 VERIFY3U(end
, <=, ms
->ms_start
+ ms
->ms_size
);
7327 * The entries in the vdev_checkpoint_sm should be marked as
7328 * allocated in the checkpointed state of the pool, therefore
7329 * their respective ms_allocateable trees should not contain them.
7331 mutex_enter(&ms
->ms_lock
);
7332 range_tree_verify_not_present(ms
->ms_allocatable
,
7333 sme
->sme_offset
, sme
->sme_run
);
7334 mutex_exit(&ms
->ms_lock
);
7340 * Verify that all segments in the vdev_checkpoint_sm are allocated
7341 * according to the checkpoint's ms_sm (i.e. are not in the checkpoint's
7344 * Do so by comparing the checkpoint space maps (vdev_checkpoint_sm) of
7345 * each vdev in the current state of the pool to the metaslab space maps
7346 * (ms_sm) of the checkpointed state of the pool.
7348 * Note that the function changes the state of the ms_allocatable
7349 * trees of the current spa_t. The entries of these ms_allocatable
7350 * trees are cleared out and then repopulated from with the free
7351 * entries of their respective ms_sm space maps.
7354 verify_checkpoint_vdev_spacemaps(spa_t
*checkpoint
, spa_t
*current
)
7356 vdev_t
*ckpoint_rvd
= checkpoint
->spa_root_vdev
;
7357 vdev_t
*current_rvd
= current
->spa_root_vdev
;
7359 load_concrete_ms_allocatable_trees(checkpoint
, SM_FREE
);
7361 for (uint64_t c
= 0; c
< ckpoint_rvd
->vdev_children
; c
++) {
7362 vdev_t
*ckpoint_vd
= ckpoint_rvd
->vdev_child
[c
];
7363 vdev_t
*current_vd
= current_rvd
->vdev_child
[c
];
7365 space_map_t
*checkpoint_sm
= NULL
;
7366 uint64_t checkpoint_sm_obj
;
7368 if (ckpoint_vd
->vdev_ops
== &vdev_indirect_ops
) {
7370 * Since we don't allow device removal in a pool
7371 * that has a checkpoint, we expect that all removed
7372 * vdevs were removed from the pool before the
7375 ASSERT3P(current_vd
->vdev_ops
, ==, &vdev_indirect_ops
);
7380 * If the checkpoint space map doesn't exist, then nothing
7381 * here is checkpointed so there's nothing to verify.
7383 if (current_vd
->vdev_top_zap
== 0 ||
7384 zap_contains(spa_meta_objset(current
),
7385 current_vd
->vdev_top_zap
,
7386 VDEV_TOP_ZAP_POOL_CHECKPOINT_SM
) != 0)
7389 VERIFY0(zap_lookup(spa_meta_objset(current
),
7390 current_vd
->vdev_top_zap
, VDEV_TOP_ZAP_POOL_CHECKPOINT_SM
,
7391 sizeof (uint64_t), 1, &checkpoint_sm_obj
));
7393 VERIFY0(space_map_open(&checkpoint_sm
, spa_meta_objset(current
),
7394 checkpoint_sm_obj
, 0, current_vd
->vdev_asize
,
7395 current_vd
->vdev_ashift
));
7397 verify_checkpoint_sm_entry_cb_arg_t vcsec
;
7398 vcsec
.vcsec_vd
= ckpoint_vd
;
7399 vcsec
.vcsec_entryid
= 0;
7400 vcsec
.vcsec_num_entries
=
7401 space_map_length(checkpoint_sm
) / sizeof (uint64_t);
7402 VERIFY0(space_map_iterate(checkpoint_sm
,
7403 space_map_length(checkpoint_sm
),
7404 verify_checkpoint_sm_entry_cb
, &vcsec
));
7405 if (dump_opt
['m'] > 3)
7406 dump_spacemap(current
->spa_meta_objset
, checkpoint_sm
);
7407 space_map_close(checkpoint_sm
);
7411 * If we've added vdevs since we took the checkpoint, ensure
7412 * that their checkpoint space maps are empty.
7414 if (ckpoint_rvd
->vdev_children
< current_rvd
->vdev_children
) {
7415 for (uint64_t c
= ckpoint_rvd
->vdev_children
;
7416 c
< current_rvd
->vdev_children
; c
++) {
7417 vdev_t
*current_vd
= current_rvd
->vdev_child
[c
];
7418 VERIFY3P(current_vd
->vdev_checkpoint_sm
, ==, NULL
);
7422 /* for cleaner progress output */
7423 (void) fprintf(stderr
, "\n");
7427 * Verifies that all space that's allocated in the checkpoint is
7428 * still allocated in the current version, by checking that everything
7429 * in checkpoint's ms_allocatable (which is actually allocated, not
7430 * allocatable/free) is not present in current's ms_allocatable.
7432 * Note that the function changes the state of the ms_allocatable
7433 * trees of both spas when called. The entries of all ms_allocatable
7434 * trees are cleared out and then repopulated from their respective
7435 * ms_sm space maps. In the checkpointed state we load the allocated
7436 * entries, and in the current state we load the free entries.
7439 verify_checkpoint_ms_spacemaps(spa_t
*checkpoint
, spa_t
*current
)
7441 vdev_t
*ckpoint_rvd
= checkpoint
->spa_root_vdev
;
7442 vdev_t
*current_rvd
= current
->spa_root_vdev
;
7444 load_concrete_ms_allocatable_trees(checkpoint
, SM_ALLOC
);
7445 load_concrete_ms_allocatable_trees(current
, SM_FREE
);
7447 for (uint64_t i
= 0; i
< ckpoint_rvd
->vdev_children
; i
++) {
7448 vdev_t
*ckpoint_vd
= ckpoint_rvd
->vdev_child
[i
];
7449 vdev_t
*current_vd
= current_rvd
->vdev_child
[i
];
7451 if (ckpoint_vd
->vdev_ops
== &vdev_indirect_ops
) {
7453 * See comment in verify_checkpoint_vdev_spacemaps()
7455 ASSERT3P(current_vd
->vdev_ops
, ==, &vdev_indirect_ops
);
7459 for (uint64_t m
= 0; m
< ckpoint_vd
->vdev_ms_count
; m
++) {
7460 metaslab_t
*ckpoint_msp
= ckpoint_vd
->vdev_ms
[m
];
7461 metaslab_t
*current_msp
= current_vd
->vdev_ms
[m
];
7463 (void) fprintf(stderr
,
7464 "\rverifying vdev %llu of %llu, "
7465 "metaslab %llu of %llu ...",
7466 (longlong_t
)current_vd
->vdev_id
,
7467 (longlong_t
)current_rvd
->vdev_children
,
7468 (longlong_t
)current_vd
->vdev_ms
[m
]->ms_id
,
7469 (longlong_t
)current_vd
->vdev_ms_count
);
7472 * We walk through the ms_allocatable trees that
7473 * are loaded with the allocated blocks from the
7474 * ms_sm spacemaps of the checkpoint. For each
7475 * one of these ranges we ensure that none of them
7476 * exists in the ms_allocatable trees of the
7477 * current state which are loaded with the ranges
7478 * that are currently free.
7480 * This way we ensure that none of the blocks that
7481 * are part of the checkpoint were freed by mistake.
7483 range_tree_walk(ckpoint_msp
->ms_allocatable
,
7484 (range_tree_func_t
*)range_tree_verify_not_present
,
7485 current_msp
->ms_allocatable
);
7489 /* for cleaner progress output */
7490 (void) fprintf(stderr
, "\n");
7494 verify_checkpoint_blocks(spa_t
*spa
)
7496 ASSERT(!dump_opt
['L']);
7498 spa_t
*checkpoint_spa
;
7499 char *checkpoint_pool
;
7503 * We import the checkpointed state of the pool (under a different
7504 * name) so we can do verification on it against the current state
7507 checkpoint_pool
= import_checkpointed_state(spa
->spa_name
, NULL
,
7509 ASSERT(strcmp(spa
->spa_name
, checkpoint_pool
) != 0);
7511 error
= spa_open(checkpoint_pool
, &checkpoint_spa
, FTAG
);
7513 fatal("Tried to open pool \"%s\" but spa_open() failed with "
7514 "error %d\n", checkpoint_pool
, error
);
7518 * Ensure that ranges in the checkpoint space maps of each vdev
7519 * are allocated according to the checkpointed state's metaslab
7522 verify_checkpoint_vdev_spacemaps(checkpoint_spa
, spa
);
7525 * Ensure that allocated ranges in the checkpoint's metaslab
7526 * space maps remain allocated in the metaslab space maps of
7527 * the current state.
7529 verify_checkpoint_ms_spacemaps(checkpoint_spa
, spa
);
7532 * Once we are done, we get rid of the checkpointed state.
7534 spa_close(checkpoint_spa
, FTAG
);
7535 free(checkpoint_pool
);
7539 dump_leftover_checkpoint_blocks(spa_t
*spa
)
7541 vdev_t
*rvd
= spa
->spa_root_vdev
;
7543 for (uint64_t i
= 0; i
< rvd
->vdev_children
; i
++) {
7544 vdev_t
*vd
= rvd
->vdev_child
[i
];
7546 space_map_t
*checkpoint_sm
= NULL
;
7547 uint64_t checkpoint_sm_obj
;
7549 if (vd
->vdev_top_zap
== 0)
7552 if (zap_contains(spa_meta_objset(spa
), vd
->vdev_top_zap
,
7553 VDEV_TOP_ZAP_POOL_CHECKPOINT_SM
) != 0)
7556 VERIFY0(zap_lookup(spa_meta_objset(spa
), vd
->vdev_top_zap
,
7557 VDEV_TOP_ZAP_POOL_CHECKPOINT_SM
,
7558 sizeof (uint64_t), 1, &checkpoint_sm_obj
));
7560 VERIFY0(space_map_open(&checkpoint_sm
, spa_meta_objset(spa
),
7561 checkpoint_sm_obj
, 0, vd
->vdev_asize
, vd
->vdev_ashift
));
7562 dump_spacemap(spa
->spa_meta_objset
, checkpoint_sm
);
7563 space_map_close(checkpoint_sm
);
7568 verify_checkpoint(spa_t
*spa
)
7570 uberblock_t checkpoint
;
7573 if (!spa_feature_is_active(spa
, SPA_FEATURE_POOL_CHECKPOINT
))
7576 error
= zap_lookup(spa
->spa_meta_objset
, DMU_POOL_DIRECTORY_OBJECT
,
7577 DMU_POOL_ZPOOL_CHECKPOINT
, sizeof (uint64_t),
7578 sizeof (uberblock_t
) / sizeof (uint64_t), &checkpoint
);
7580 if (error
== ENOENT
&& !dump_opt
['L']) {
7582 * If the feature is active but the uberblock is missing
7583 * then we must be in the middle of discarding the
7586 (void) printf("\nPartially discarded checkpoint "
7588 if (dump_opt
['m'] > 3)
7589 dump_leftover_checkpoint_blocks(spa
);
7591 } else if (error
!= 0) {
7592 (void) printf("lookup error %d when looking for "
7593 "checkpointed uberblock in MOS\n", error
);
7596 dump_uberblock(&checkpoint
, "\nCheckpointed uberblock found:\n", "\n");
7598 if (checkpoint
.ub_checkpoint_txg
== 0) {
7599 (void) printf("\nub_checkpoint_txg not set in checkpointed "
7604 if (error
== 0 && !dump_opt
['L'])
7605 verify_checkpoint_blocks(spa
);
7611 mos_leaks_cb(void *arg
, uint64_t start
, uint64_t size
)
7614 for (uint64_t i
= start
; i
< size
; i
++) {
7615 (void) printf("MOS object %llu referenced but not allocated\n",
7621 mos_obj_refd(uint64_t obj
)
7623 if (obj
!= 0 && mos_refd_objs
!= NULL
)
7624 range_tree_add(mos_refd_objs
, obj
, 1);
7628 * Call on a MOS object that may already have been referenced.
7631 mos_obj_refd_multiple(uint64_t obj
)
7633 if (obj
!= 0 && mos_refd_objs
!= NULL
&&
7634 !range_tree_contains(mos_refd_objs
, obj
, 1))
7635 range_tree_add(mos_refd_objs
, obj
, 1);
7639 mos_leak_vdev_top_zap(vdev_t
*vd
)
7641 uint64_t ms_flush_data_obj
;
7642 int error
= zap_lookup(spa_meta_objset(vd
->vdev_spa
),
7643 vd
->vdev_top_zap
, VDEV_TOP_ZAP_MS_UNFLUSHED_PHYS_TXGS
,
7644 sizeof (ms_flush_data_obj
), 1, &ms_flush_data_obj
);
7645 if (error
== ENOENT
)
7649 mos_obj_refd(ms_flush_data_obj
);
7653 mos_leak_vdev(vdev_t
*vd
)
7655 mos_obj_refd(vd
->vdev_dtl_object
);
7656 mos_obj_refd(vd
->vdev_ms_array
);
7657 mos_obj_refd(vd
->vdev_indirect_config
.vic_births_object
);
7658 mos_obj_refd(vd
->vdev_indirect_config
.vic_mapping_object
);
7659 mos_obj_refd(vd
->vdev_leaf_zap
);
7660 if (vd
->vdev_checkpoint_sm
!= NULL
)
7661 mos_obj_refd(vd
->vdev_checkpoint_sm
->sm_object
);
7662 if (vd
->vdev_indirect_mapping
!= NULL
) {
7663 mos_obj_refd(vd
->vdev_indirect_mapping
->
7664 vim_phys
->vimp_counts_object
);
7666 if (vd
->vdev_obsolete_sm
!= NULL
)
7667 mos_obj_refd(vd
->vdev_obsolete_sm
->sm_object
);
7669 for (uint64_t m
= 0; m
< vd
->vdev_ms_count
; m
++) {
7670 metaslab_t
*ms
= vd
->vdev_ms
[m
];
7671 mos_obj_refd(space_map_object(ms
->ms_sm
));
7674 if (vd
->vdev_root_zap
!= 0)
7675 mos_obj_refd(vd
->vdev_root_zap
);
7677 if (vd
->vdev_top_zap
!= 0) {
7678 mos_obj_refd(vd
->vdev_top_zap
);
7679 mos_leak_vdev_top_zap(vd
);
7682 for (uint64_t c
= 0; c
< vd
->vdev_children
; c
++) {
7683 mos_leak_vdev(vd
->vdev_child
[c
]);
7688 mos_leak_log_spacemaps(spa_t
*spa
)
7690 uint64_t spacemap_zap
;
7691 int error
= zap_lookup(spa_meta_objset(spa
),
7692 DMU_POOL_DIRECTORY_OBJECT
, DMU_POOL_LOG_SPACEMAP_ZAP
,
7693 sizeof (spacemap_zap
), 1, &spacemap_zap
);
7694 if (error
== ENOENT
)
7698 mos_obj_refd(spacemap_zap
);
7699 for (spa_log_sm_t
*sls
= avl_first(&spa
->spa_sm_logs_by_txg
);
7700 sls
; sls
= AVL_NEXT(&spa
->spa_sm_logs_by_txg
, sls
))
7701 mos_obj_refd(sls
->sls_sm_obj
);
7705 errorlog_count_refd(objset_t
*mos
, uint64_t errlog
)
7709 for (zap_cursor_init(&zc
, mos
, errlog
);
7710 zap_cursor_retrieve(&zc
, &za
) == 0;
7711 zap_cursor_advance(&zc
)) {
7712 mos_obj_refd(za
.za_first_integer
);
7714 zap_cursor_fini(&zc
);
7718 dump_mos_leaks(spa_t
*spa
)
7721 objset_t
*mos
= spa
->spa_meta_objset
;
7722 dsl_pool_t
*dp
= spa
->spa_dsl_pool
;
7724 /* Visit and mark all referenced objects in the MOS */
7726 mos_obj_refd(DMU_POOL_DIRECTORY_OBJECT
);
7727 mos_obj_refd(spa
->spa_pool_props_object
);
7728 mos_obj_refd(spa
->spa_config_object
);
7729 mos_obj_refd(spa
->spa_ddt_stat_object
);
7730 mos_obj_refd(spa
->spa_feat_desc_obj
);
7731 mos_obj_refd(spa
->spa_feat_enabled_txg_obj
);
7732 mos_obj_refd(spa
->spa_feat_for_read_obj
);
7733 mos_obj_refd(spa
->spa_feat_for_write_obj
);
7734 mos_obj_refd(spa
->spa_history
);
7735 mos_obj_refd(spa
->spa_errlog_last
);
7736 mos_obj_refd(spa
->spa_errlog_scrub
);
7738 if (spa_feature_is_enabled(spa
, SPA_FEATURE_HEAD_ERRLOG
)) {
7739 errorlog_count_refd(mos
, spa
->spa_errlog_last
);
7740 errorlog_count_refd(mos
, spa
->spa_errlog_scrub
);
7743 mos_obj_refd(spa
->spa_all_vdev_zaps
);
7744 mos_obj_refd(spa
->spa_dsl_pool
->dp_bptree_obj
);
7745 mos_obj_refd(spa
->spa_dsl_pool
->dp_tmp_userrefs_obj
);
7746 mos_obj_refd(spa
->spa_dsl_pool
->dp_scan
->scn_phys
.scn_queue_obj
);
7747 bpobj_count_refd(&spa
->spa_deferred_bpobj
);
7748 mos_obj_refd(dp
->dp_empty_bpobj
);
7749 bpobj_count_refd(&dp
->dp_obsolete_bpobj
);
7750 bpobj_count_refd(&dp
->dp_free_bpobj
);
7751 mos_obj_refd(spa
->spa_l2cache
.sav_object
);
7752 mos_obj_refd(spa
->spa_spares
.sav_object
);
7754 if (spa
->spa_syncing_log_sm
!= NULL
)
7755 mos_obj_refd(spa
->spa_syncing_log_sm
->sm_object
);
7756 mos_leak_log_spacemaps(spa
);
7758 mos_obj_refd(spa
->spa_condensing_indirect_phys
.
7759 scip_next_mapping_object
);
7760 mos_obj_refd(spa
->spa_condensing_indirect_phys
.
7761 scip_prev_obsolete_sm_object
);
7762 if (spa
->spa_condensing_indirect_phys
.scip_next_mapping_object
!= 0) {
7763 vdev_indirect_mapping_t
*vim
=
7764 vdev_indirect_mapping_open(mos
,
7765 spa
->spa_condensing_indirect_phys
.scip_next_mapping_object
);
7766 mos_obj_refd(vim
->vim_phys
->vimp_counts_object
);
7767 vdev_indirect_mapping_close(vim
);
7769 deleted_livelists_dump_mos(spa
);
7771 if (dp
->dp_origin_snap
!= NULL
) {
7774 dsl_pool_config_enter(dp
, FTAG
);
7775 VERIFY0(dsl_dataset_hold_obj(dp
,
7776 dsl_dataset_phys(dp
->dp_origin_snap
)->ds_next_snap_obj
,
7778 count_ds_mos_objects(ds
);
7779 dump_blkptr_list(&ds
->ds_deadlist
, "Deadlist");
7780 dsl_dataset_rele(ds
, FTAG
);
7781 dsl_pool_config_exit(dp
, FTAG
);
7783 count_ds_mos_objects(dp
->dp_origin_snap
);
7784 dump_blkptr_list(&dp
->dp_origin_snap
->ds_deadlist
, "Deadlist");
7786 count_dir_mos_objects(dp
->dp_mos_dir
);
7787 if (dp
->dp_free_dir
!= NULL
)
7788 count_dir_mos_objects(dp
->dp_free_dir
);
7789 if (dp
->dp_leak_dir
!= NULL
)
7790 count_dir_mos_objects(dp
->dp_leak_dir
);
7792 mos_leak_vdev(spa
->spa_root_vdev
);
7794 for (uint64_t class = 0; class < DDT_CLASSES
; class++) {
7795 for (uint64_t type
= 0; type
< DDT_TYPES
; type
++) {
7796 for (uint64_t cksum
= 0;
7797 cksum
< ZIO_CHECKSUM_FUNCTIONS
; cksum
++) {
7798 ddt_t
*ddt
= spa
->spa_ddt
[cksum
];
7799 mos_obj_refd(ddt
->ddt_object
[type
][class]);
7805 * Visit all allocated objects and make sure they are referenced.
7807 uint64_t object
= 0;
7808 while (dmu_object_next(mos
, &object
, B_FALSE
, 0) == 0) {
7809 if (range_tree_contains(mos_refd_objs
, object
, 1)) {
7810 range_tree_remove(mos_refd_objs
, object
, 1);
7812 dmu_object_info_t doi
;
7814 VERIFY0(dmu_object_info(mos
, object
, &doi
));
7815 if (doi
.doi_type
& DMU_OT_NEWTYPE
) {
7816 dmu_object_byteswap_t bswap
=
7817 DMU_OT_BYTESWAP(doi
.doi_type
);
7818 name
= dmu_ot_byteswap
[bswap
].ob_name
;
7820 name
= dmu_ot
[doi
.doi_type
].ot_name
;
7823 (void) printf("MOS object %llu (%s) leaked\n",
7824 (u_longlong_t
)object
, name
);
7828 (void) range_tree_walk(mos_refd_objs
, mos_leaks_cb
, NULL
);
7829 if (!range_tree_is_empty(mos_refd_objs
))
7831 range_tree_vacate(mos_refd_objs
, NULL
, NULL
);
7832 range_tree_destroy(mos_refd_objs
);
7836 typedef struct log_sm_obsolete_stats_arg
{
7837 uint64_t lsos_current_txg
;
7839 uint64_t lsos_total_entries
;
7840 uint64_t lsos_valid_entries
;
7842 uint64_t lsos_sm_entries
;
7843 uint64_t lsos_valid_sm_entries
;
7844 } log_sm_obsolete_stats_arg_t
;
7847 log_spacemap_obsolete_stats_cb(spa_t
*spa
, space_map_entry_t
*sme
,
7848 uint64_t txg
, void *arg
)
7850 log_sm_obsolete_stats_arg_t
*lsos
= arg
;
7852 uint64_t offset
= sme
->sme_offset
;
7853 uint64_t vdev_id
= sme
->sme_vdev
;
7855 if (lsos
->lsos_current_txg
== 0) {
7856 /* this is the first log */
7857 lsos
->lsos_current_txg
= txg
;
7858 } else if (lsos
->lsos_current_txg
< txg
) {
7859 /* we just changed log - print stats and reset */
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
);
7864 lsos
->lsos_valid_sm_entries
= 0;
7865 lsos
->lsos_sm_entries
= 0;
7866 lsos
->lsos_current_txg
= txg
;
7868 ASSERT3U(lsos
->lsos_current_txg
, ==, txg
);
7870 lsos
->lsos_sm_entries
++;
7871 lsos
->lsos_total_entries
++;
7873 vdev_t
*vd
= vdev_lookup_top(spa
, vdev_id
);
7874 if (!vdev_is_concrete(vd
))
7877 metaslab_t
*ms
= vd
->vdev_ms
[offset
>> vd
->vdev_ms_shift
];
7878 ASSERT(sme
->sme_type
== SM_ALLOC
|| sme
->sme_type
== SM_FREE
);
7880 if (txg
< metaslab_unflushed_txg(ms
))
7882 lsos
->lsos_valid_sm_entries
++;
7883 lsos
->lsos_valid_entries
++;
7888 dump_log_spacemap_obsolete_stats(spa_t
*spa
)
7890 if (!spa_feature_is_active(spa
, SPA_FEATURE_LOG_SPACEMAP
))
7893 log_sm_obsolete_stats_arg_t lsos
= {0};
7895 (void) printf("Log Space Map Obsolete Entry Statistics:\n");
7897 iterate_through_spacemap_logs(spa
,
7898 log_spacemap_obsolete_stats_cb
, &lsos
);
7900 /* print stats for latest log */
7901 (void) printf("%-8llu valid entries out of %-8llu - txg %llu\n",
7902 (u_longlong_t
)lsos
.lsos_valid_sm_entries
,
7903 (u_longlong_t
)lsos
.lsos_sm_entries
,
7904 (u_longlong_t
)lsos
.lsos_current_txg
);
7906 (void) printf("%-8llu valid entries out of %-8llu - total\n\n",
7907 (u_longlong_t
)lsos
.lsos_valid_entries
,
7908 (u_longlong_t
)lsos
.lsos_total_entries
);
7912 dump_zpool(spa_t
*spa
)
7914 dsl_pool_t
*dp
= spa_get_dsl(spa
);
7917 if (dump_opt
['y']) {
7918 livelist_metaslab_validate(spa
);
7921 if (dump_opt
['S']) {
7922 dump_simulated_ddt(spa
);
7926 if (!dump_opt
['e'] && dump_opt
['C'] > 1) {
7927 (void) printf("\nCached configuration:\n");
7928 dump_nvlist(spa
->spa_config
, 8);
7935 dump_uberblock(&spa
->spa_uberblock
, "\nUberblock:\n", "\n");
7940 if (dump_opt
['d'] > 2 || dump_opt
['m'])
7941 dump_metaslabs(spa
);
7943 dump_metaslab_groups(spa
, dump_opt
['M'] > 1);
7944 if (dump_opt
['d'] > 2 || dump_opt
['m']) {
7945 dump_log_spacemaps(spa
);
7946 dump_log_spacemap_obsolete_stats(spa
);
7949 if (dump_opt
['d'] || dump_opt
['i']) {
7951 mos_refd_objs
= range_tree_create(NULL
, RANGE_SEG64
, NULL
, 0,
7953 dump_objset(dp
->dp_meta_objset
);
7955 if (dump_opt
['d'] >= 3) {
7956 dsl_pool_t
*dp
= spa
->spa_dsl_pool
;
7957 dump_full_bpobj(&spa
->spa_deferred_bpobj
,
7958 "Deferred frees", 0);
7959 if (spa_version(spa
) >= SPA_VERSION_DEADLISTS
) {
7960 dump_full_bpobj(&dp
->dp_free_bpobj
,
7961 "Pool snapshot frees", 0);
7963 if (bpobj_is_open(&dp
->dp_obsolete_bpobj
)) {
7964 ASSERT(spa_feature_is_enabled(spa
,
7965 SPA_FEATURE_DEVICE_REMOVAL
));
7966 dump_full_bpobj(&dp
->dp_obsolete_bpobj
,
7967 "Pool obsolete blocks", 0);
7970 if (spa_feature_is_active(spa
,
7971 SPA_FEATURE_ASYNC_DESTROY
)) {
7972 dump_bptree(spa
->spa_meta_objset
,
7974 "Pool dataset frees");
7976 dump_dtl(spa
->spa_root_vdev
, 0);
7979 for (spa_feature_t f
= 0; f
< SPA_FEATURES
; f
++)
7980 global_feature_count
[f
] = UINT64_MAX
;
7981 global_feature_count
[SPA_FEATURE_REDACTION_BOOKMARKS
] = 0;
7982 global_feature_count
[SPA_FEATURE_BOOKMARK_WRITTEN
] = 0;
7983 global_feature_count
[SPA_FEATURE_LIVELIST
] = 0;
7985 (void) dmu_objset_find(spa_name(spa
), dump_one_objset
,
7986 NULL
, DS_FIND_SNAPSHOTS
| DS_FIND_CHILDREN
);
7988 if (rc
== 0 && !dump_opt
['L'])
7989 rc
= dump_mos_leaks(spa
);
7991 for (f
= 0; f
< SPA_FEATURES
; f
++) {
7995 if (!(spa_feature_table
[f
].fi_flags
&
7996 ZFEATURE_FLAG_PER_DATASET
)) {
7997 if (global_feature_count
[f
] == UINT64_MAX
)
7999 if (!spa_feature_is_enabled(spa
, f
)) {
8000 ASSERT0(global_feature_count
[f
]);
8003 arr
= global_feature_count
;
8005 if (!spa_feature_is_enabled(spa
, f
)) {
8006 ASSERT0(dataset_feature_count
[f
]);
8009 arr
= dataset_feature_count
;
8011 if (feature_get_refcount(spa
, &spa_feature_table
[f
],
8012 &refcount
) == ENOTSUP
)
8014 if (arr
[f
] != refcount
) {
8015 (void) printf("%s feature refcount mismatch: "
8016 "%lld consumers != %lld refcount\n",
8017 spa_feature_table
[f
].fi_uname
,
8018 (longlong_t
)arr
[f
], (longlong_t
)refcount
);
8021 (void) printf("Verified %s feature refcount "
8022 "of %llu is correct\n",
8023 spa_feature_table
[f
].fi_uname
,
8024 (longlong_t
)refcount
);
8029 rc
= verify_device_removal_feature_counts(spa
);
8032 if (rc
== 0 && (dump_opt
['b'] || dump_opt
['c']))
8033 rc
= dump_block_stats(spa
);
8036 rc
= verify_spacemap_refcounts(spa
);
8039 show_pool_stats(spa
);
8045 rc
= verify_checkpoint(spa
);
8048 dump_debug_buffer();
8053 #define ZDB_FLAG_CHECKSUM 0x0001
8054 #define ZDB_FLAG_DECOMPRESS 0x0002
8055 #define ZDB_FLAG_BSWAP 0x0004
8056 #define ZDB_FLAG_GBH 0x0008
8057 #define ZDB_FLAG_INDIRECT 0x0010
8058 #define ZDB_FLAG_RAW 0x0020
8059 #define ZDB_FLAG_PRINT_BLKPTR 0x0040
8060 #define ZDB_FLAG_VERBOSE 0x0080
8062 static int flagbits
[256];
8063 static char flagbitstr
[16];
8066 zdb_print_blkptr(const blkptr_t
*bp
, int flags
)
8068 char blkbuf
[BP_SPRINTF_LEN
];
8070 if (flags
& ZDB_FLAG_BSWAP
)
8071 byteswap_uint64_array((void *)bp
, sizeof (blkptr_t
));
8073 snprintf_blkptr(blkbuf
, sizeof (blkbuf
), bp
);
8074 (void) printf("%s\n", blkbuf
);
8078 zdb_dump_indirect(blkptr_t
*bp
, int nbps
, int flags
)
8082 for (i
= 0; i
< nbps
; i
++)
8083 zdb_print_blkptr(&bp
[i
], flags
);
8087 zdb_dump_gbh(void *buf
, int flags
)
8089 zdb_dump_indirect((blkptr_t
*)buf
, SPA_GBH_NBLKPTRS
, flags
);
8093 zdb_dump_block_raw(void *buf
, uint64_t size
, int flags
)
8095 if (flags
& ZDB_FLAG_BSWAP
)
8096 byteswap_uint64_array(buf
, size
);
8097 VERIFY(write(fileno(stdout
), buf
, size
) == size
);
8101 zdb_dump_block(char *label
, void *buf
, uint64_t size
, int flags
)
8103 uint64_t *d
= (uint64_t *)buf
;
8104 unsigned nwords
= size
/ sizeof (uint64_t);
8105 int do_bswap
= !!(flags
& ZDB_FLAG_BSWAP
);
8112 hdr
= " 7 6 5 4 3 2 1 0 f e d c b a 9 8";
8114 hdr
= " 0 1 2 3 4 5 6 7 8 9 a b c d e f";
8116 (void) printf("\n%s\n%6s %s 0123456789abcdef\n", label
, "", hdr
);
8118 #ifdef _LITTLE_ENDIAN
8119 /* correct the endianness */
8120 do_bswap
= !do_bswap
;
8122 for (i
= 0; i
< nwords
; i
+= 2) {
8123 (void) printf("%06llx: %016llx %016llx ",
8124 (u_longlong_t
)(i
* sizeof (uint64_t)),
8125 (u_longlong_t
)(do_bswap
? BSWAP_64(d
[i
]) : d
[i
]),
8126 (u_longlong_t
)(do_bswap
? BSWAP_64(d
[i
+ 1]) : d
[i
+ 1]));
8129 for (j
= 0; j
< 2 * sizeof (uint64_t); j
++)
8130 (void) printf("%c", isprint(c
[j
]) ? c
[j
] : '.');
8131 (void) printf("\n");
8136 * There are two acceptable formats:
8137 * leaf_name - For example: c1t0d0 or /tmp/ztest.0a
8138 * child[.child]* - For example: 0.1.1
8140 * The second form can be used to specify arbitrary vdevs anywhere
8141 * in the hierarchy. For example, in a pool with a mirror of
8142 * RAID-Zs, you can specify either RAID-Z vdev with 0.0 or 0.1 .
8145 zdb_vdev_lookup(vdev_t
*vdev
, const char *path
)
8153 /* First, assume the x.x.x.x format */
8154 i
= strtoul(path
, &s
, 10);
8155 if (s
== path
|| (s
&& *s
!= '.' && *s
!= '\0'))
8157 if (i
>= vdev
->vdev_children
)
8160 vdev
= vdev
->vdev_child
[i
];
8161 if (s
&& *s
== '\0')
8163 return (zdb_vdev_lookup(vdev
, s
+1));
8166 for (i
= 0; i
< vdev
->vdev_children
; i
++) {
8167 vdev_t
*vc
= vdev
->vdev_child
[i
];
8169 if (vc
->vdev_path
== NULL
) {
8170 vc
= zdb_vdev_lookup(vc
, path
);
8177 p
= strrchr(vc
->vdev_path
, '/');
8178 p
= p
? p
+ 1 : vc
->vdev_path
;
8179 q
= &vc
->vdev_path
[strlen(vc
->vdev_path
) - 2];
8181 if (strcmp(vc
->vdev_path
, path
) == 0)
8183 if (strcmp(p
, path
) == 0)
8185 if (strcmp(q
, "s0") == 0 && strncmp(p
, path
, q
- p
) == 0)
8193 name_from_objset_id(spa_t
*spa
, uint64_t objset_id
, char *outstr
)
8197 dsl_pool_config_enter(spa
->spa_dsl_pool
, FTAG
);
8198 int error
= dsl_dataset_hold_obj(spa
->spa_dsl_pool
, objset_id
,
8201 (void) fprintf(stderr
, "failed to hold objset %llu: %s\n",
8202 (u_longlong_t
)objset_id
, strerror(error
));
8203 dsl_pool_config_exit(spa
->spa_dsl_pool
, FTAG
);
8206 dsl_dataset_name(ds
, outstr
);
8207 dsl_dataset_rele(ds
, NULL
);
8208 dsl_pool_config_exit(spa
->spa_dsl_pool
, FTAG
);
8213 zdb_parse_block_sizes(char *sizes
, uint64_t *lsize
, uint64_t *psize
)
8215 char *s0
, *s1
, *tmp
= NULL
;
8220 s0
= strtok_r(sizes
, "/", &tmp
);
8223 s1
= strtok_r(NULL
, "/", &tmp
);
8224 *lsize
= strtoull(s0
, NULL
, 16);
8225 *psize
= s1
? strtoull(s1
, NULL
, 16) : *lsize
;
8226 return (*lsize
>= *psize
&& *psize
> 0);
8229 #define ZIO_COMPRESS_MASK(alg) (1ULL << (ZIO_COMPRESS_##alg))
8232 zdb_decompress_block(abd_t
*pabd
, void *buf
, void *lbuf
, uint64_t lsize
,
8233 uint64_t psize
, int flags
)
8236 boolean_t exceeded
= B_FALSE
;
8238 * We don't know how the data was compressed, so just try
8239 * every decompress function at every inflated blocksize.
8241 void *lbuf2
= umem_alloc(SPA_MAXBLOCKSIZE
, UMEM_NOFAIL
);
8242 int cfuncs
[ZIO_COMPRESS_FUNCTIONS
] = { 0 };
8243 int *cfuncp
= cfuncs
;
8244 uint64_t maxlsize
= SPA_MAXBLOCKSIZE
;
8245 uint64_t mask
= ZIO_COMPRESS_MASK(ON
) | ZIO_COMPRESS_MASK(OFF
) |
8246 ZIO_COMPRESS_MASK(INHERIT
) | ZIO_COMPRESS_MASK(EMPTY
) |
8247 (getenv("ZDB_NO_ZLE") ? ZIO_COMPRESS_MASK(ZLE
) : 0);
8248 *cfuncp
++ = ZIO_COMPRESS_LZ4
;
8249 *cfuncp
++ = ZIO_COMPRESS_LZJB
;
8250 mask
|= ZIO_COMPRESS_MASK(LZ4
) | ZIO_COMPRESS_MASK(LZJB
);
8251 for (int c
= 0; c
< ZIO_COMPRESS_FUNCTIONS
; c
++)
8252 if (((1ULL << c
) & mask
) == 0)
8256 * On the one hand, with SPA_MAXBLOCKSIZE at 16MB, this
8257 * could take a while and we should let the user know
8258 * we are not stuck. On the other hand, printing progress
8259 * info gets old after a while. User can specify 'v' flag
8260 * to see the progression.
8263 lsize
+= SPA_MINBLOCKSIZE
;
8266 for (; lsize
<= maxlsize
; lsize
+= SPA_MINBLOCKSIZE
) {
8267 for (cfuncp
= cfuncs
; *cfuncp
; cfuncp
++) {
8268 if (flags
& ZDB_FLAG_VERBOSE
) {
8269 (void) fprintf(stderr
,
8270 "Trying %05llx -> %05llx (%s)\n",
8271 (u_longlong_t
)psize
,
8272 (u_longlong_t
)lsize
,
8273 zio_compress_table
[*cfuncp
].\
8278 * We randomize lbuf2, and decompress to both
8279 * lbuf and lbuf2. This way, we will know if
8280 * decompression fill exactly to lsize.
8282 VERIFY0(random_get_pseudo_bytes(lbuf2
, lsize
));
8284 if (zio_decompress_data(*cfuncp
, pabd
,
8285 lbuf
, psize
, lsize
, NULL
) == 0 &&
8286 zio_decompress_data(*cfuncp
, pabd
,
8287 lbuf2
, psize
, lsize
, NULL
) == 0 &&
8288 memcmp(lbuf
, lbuf2
, lsize
) == 0)
8294 umem_free(lbuf2
, SPA_MAXBLOCKSIZE
);
8296 if (lsize
> maxlsize
) {
8299 if (*cfuncp
== ZIO_COMPRESS_ZLE
) {
8300 printf("\nZLE decompression was selected. If you "
8301 "suspect the results are wrong,\ntry avoiding ZLE "
8302 "by setting and exporting ZDB_NO_ZLE=\"true\"\n");
8309 * Read a block from a pool and print it out. The syntax of the
8310 * block descriptor is:
8312 * pool:vdev_specifier:offset:[lsize/]psize[:flags]
8314 * pool - The name of the pool you wish to read from
8315 * vdev_specifier - Which vdev (see comment for zdb_vdev_lookup)
8316 * offset - offset, in hex, in bytes
8317 * size - Amount of data to read, in hex, in bytes
8318 * flags - A string of characters specifying options
8319 * b: Decode a blkptr at given offset within block
8320 * c: Calculate and display checksums
8321 * d: Decompress data before dumping
8322 * e: Byteswap data before dumping
8323 * g: Display data as a gang block header
8324 * i: Display as an indirect block
8325 * r: Dump raw data to stdout
8330 zdb_read_block(char *thing
, spa_t
*spa
)
8332 blkptr_t blk
, *bp
= &blk
;
8333 dva_t
*dva
= bp
->blk_dva
;
8335 uint64_t offset
= 0, psize
= 0, lsize
= 0, blkptr_offset
= 0;
8340 char *s
, *p
, *dup
, *flagstr
, *sizes
, *tmp
= NULL
;
8341 const char *vdev
, *errmsg
= NULL
;
8343 boolean_t borrowed
= B_FALSE
, found
= B_FALSE
;
8345 dup
= strdup(thing
);
8346 s
= strtok_r(dup
, ":", &tmp
);
8348 s
= strtok_r(NULL
, ":", &tmp
);
8349 offset
= strtoull(s
? s
: "", NULL
, 16);
8350 sizes
= strtok_r(NULL
, ":", &tmp
);
8351 s
= strtok_r(NULL
, ":", &tmp
);
8352 flagstr
= strdup(s
?: "");
8354 if (!zdb_parse_block_sizes(sizes
, &lsize
, &psize
))
8355 errmsg
= "invalid size(s)";
8356 if (!IS_P2ALIGNED(psize
, DEV_BSIZE
) || !IS_P2ALIGNED(lsize
, DEV_BSIZE
))
8357 errmsg
= "size must be a multiple of sector size";
8358 if (!IS_P2ALIGNED(offset
, DEV_BSIZE
))
8359 errmsg
= "offset must be a multiple of sector size";
8361 (void) printf("Invalid block specifier: %s - %s\n",
8367 for (s
= strtok_r(flagstr
, ":", &tmp
);
8369 s
= strtok_r(NULL
, ":", &tmp
)) {
8370 for (i
= 0; i
< strlen(flagstr
); i
++) {
8371 int bit
= flagbits
[(uchar_t
)flagstr
[i
]];
8374 (void) printf("***Ignoring flag: %c\n",
8375 (uchar_t
)flagstr
[i
]);
8381 p
= &flagstr
[i
+ 1];
8382 if (*p
!= ':' && *p
!= '\0') {
8383 int j
= 0, nextbit
= flagbits
[(uchar_t
)*p
];
8384 char *end
, offstr
[8] = { 0 };
8385 if ((bit
== ZDB_FLAG_PRINT_BLKPTR
) &&
8387 /* look ahead to isolate the offset */
8388 while (nextbit
== 0 &&
8389 strchr(flagbitstr
, *p
) == NULL
) {
8392 if (i
+ j
> strlen(flagstr
))
8395 nextbit
= flagbits
[(uchar_t
)*p
];
8397 blkptr_offset
= strtoull(offstr
, &end
,
8400 } else if (nextbit
== 0) {
8401 (void) printf("***Ignoring flag arg:"
8402 " '%c'\n", (uchar_t
)*p
);
8407 if (blkptr_offset
% sizeof (blkptr_t
)) {
8408 printf("Block pointer offset 0x%llx "
8409 "must be divisible by 0x%x\n",
8410 (longlong_t
)blkptr_offset
, (int)sizeof (blkptr_t
));
8413 if (found
== B_FALSE
&& strlen(flagstr
) > 0) {
8414 printf("Invalid flag arg: '%s'\n", flagstr
);
8418 vd
= zdb_vdev_lookup(spa
->spa_root_vdev
, vdev
);
8420 (void) printf("***Invalid vdev: %s\n", vdev
);
8424 (void) fprintf(stderr
, "Found vdev: %s\n",
8427 (void) fprintf(stderr
, "Found vdev type: %s\n",
8428 vd
->vdev_ops
->vdev_op_type
);
8431 pabd
= abd_alloc_for_io(SPA_MAXBLOCKSIZE
, B_FALSE
);
8432 lbuf
= umem_alloc(SPA_MAXBLOCKSIZE
, UMEM_NOFAIL
);
8436 DVA_SET_VDEV(&dva
[0], vd
->vdev_id
);
8437 DVA_SET_OFFSET(&dva
[0], offset
);
8438 DVA_SET_GANG(&dva
[0], !!(flags
& ZDB_FLAG_GBH
));
8439 DVA_SET_ASIZE(&dva
[0], vdev_psize_to_asize(vd
, psize
));
8441 BP_SET_BIRTH(bp
, TXG_INITIAL
, TXG_INITIAL
);
8443 BP_SET_LSIZE(bp
, lsize
);
8444 BP_SET_PSIZE(bp
, psize
);
8445 BP_SET_COMPRESS(bp
, ZIO_COMPRESS_OFF
);
8446 BP_SET_CHECKSUM(bp
, ZIO_CHECKSUM_OFF
);
8447 BP_SET_TYPE(bp
, DMU_OT_NONE
);
8448 BP_SET_LEVEL(bp
, 0);
8449 BP_SET_DEDUP(bp
, 0);
8450 BP_SET_BYTEORDER(bp
, ZFS_HOST_BYTEORDER
);
8452 spa_config_enter(spa
, SCL_STATE
, FTAG
, RW_READER
);
8453 zio
= zio_root(spa
, NULL
, NULL
, 0);
8455 if (vd
== vd
->vdev_top
) {
8457 * Treat this as a normal block read.
8459 zio_nowait(zio_read(zio
, spa
, bp
, pabd
, psize
, NULL
, NULL
,
8460 ZIO_PRIORITY_SYNC_READ
,
8461 ZIO_FLAG_CANFAIL
| ZIO_FLAG_RAW
, NULL
));
8464 * Treat this as a vdev child I/O.
8466 zio_nowait(zio_vdev_child_io(zio
, bp
, vd
, offset
, pabd
,
8467 psize
, ZIO_TYPE_READ
, ZIO_PRIORITY_SYNC_READ
,
8468 ZIO_FLAG_DONT_CACHE
| ZIO_FLAG_DONT_PROPAGATE
|
8469 ZIO_FLAG_DONT_RETRY
| ZIO_FLAG_CANFAIL
| ZIO_FLAG_RAW
|
8470 ZIO_FLAG_OPTIONAL
, NULL
, NULL
));
8473 error
= zio_wait(zio
);
8474 spa_config_exit(spa
, SCL_STATE
, FTAG
);
8477 (void) printf("Read of %s failed, error: %d\n", thing
, error
);
8481 uint64_t orig_lsize
= lsize
;
8483 if (flags
& ZDB_FLAG_DECOMPRESS
) {
8484 boolean_t failed
= zdb_decompress_block(pabd
, buf
, lbuf
,
8485 lsize
, psize
, flags
);
8487 (void) printf("Decompress of %s failed\n", thing
);
8491 buf
= abd_borrow_buf_copy(pabd
, lsize
);
8495 * Try to detect invalid block pointer. If invalid, try
8498 if ((flags
& ZDB_FLAG_PRINT_BLKPTR
|| flags
& ZDB_FLAG_INDIRECT
) &&
8499 !(flags
& ZDB_FLAG_DECOMPRESS
)) {
8500 const blkptr_t
*b
= (const blkptr_t
*)(void *)
8501 ((uintptr_t)buf
+ (uintptr_t)blkptr_offset
);
8502 if (zfs_blkptr_verify(spa
, b
, B_FALSE
, BLK_VERIFY_ONLY
) ==
8504 abd_return_buf_copy(pabd
, buf
, lsize
);
8507 boolean_t failed
= zdb_decompress_block(pabd
, buf
,
8508 lbuf
, lsize
, psize
, flags
);
8509 b
= (const blkptr_t
*)(void *)
8510 ((uintptr_t)buf
+ (uintptr_t)blkptr_offset
);
8511 if (failed
|| zfs_blkptr_verify(spa
, b
, B_FALSE
,
8512 BLK_VERIFY_LOG
) == B_FALSE
) {
8513 printf("invalid block pointer at this DVA\n");
8519 if (flags
& ZDB_FLAG_PRINT_BLKPTR
)
8520 zdb_print_blkptr((blkptr_t
*)(void *)
8521 ((uintptr_t)buf
+ (uintptr_t)blkptr_offset
), flags
);
8522 else if (flags
& ZDB_FLAG_RAW
)
8523 zdb_dump_block_raw(buf
, lsize
, flags
);
8524 else if (flags
& ZDB_FLAG_INDIRECT
)
8525 zdb_dump_indirect((blkptr_t
*)buf
,
8526 orig_lsize
/ sizeof (blkptr_t
), flags
);
8527 else if (flags
& ZDB_FLAG_GBH
)
8528 zdb_dump_gbh(buf
, flags
);
8530 zdb_dump_block(thing
, buf
, lsize
, flags
);
8533 * If :c was specified, iterate through the checksum table to
8534 * calculate and display each checksum for our specified
8537 if ((flags
& ZDB_FLAG_CHECKSUM
) && !(flags
& ZDB_FLAG_RAW
) &&
8538 !(flags
& ZDB_FLAG_GBH
)) {
8540 (void) printf("\n");
8541 for (enum zio_checksum ck
= ZIO_CHECKSUM_LABEL
;
8542 ck
< ZIO_CHECKSUM_FUNCTIONS
; ck
++) {
8544 if ((zio_checksum_table
[ck
].ci_flags
&
8545 ZCHECKSUM_FLAG_EMBEDDED
) ||
8546 ck
== ZIO_CHECKSUM_NOPARITY
) {
8549 BP_SET_CHECKSUM(bp
, ck
);
8550 spa_config_enter(spa
, SCL_STATE
, FTAG
, RW_READER
);
8551 czio
= zio_root(spa
, NULL
, NULL
, ZIO_FLAG_CANFAIL
);
8554 if (vd
== vd
->vdev_top
) {
8555 zio_nowait(zio_read(czio
, spa
, bp
, pabd
, psize
,
8557 ZIO_PRIORITY_SYNC_READ
,
8558 ZIO_FLAG_CANFAIL
| ZIO_FLAG_RAW
|
8559 ZIO_FLAG_DONT_RETRY
, NULL
));
8561 zio_nowait(zio_vdev_child_io(czio
, bp
, vd
,
8562 offset
, pabd
, psize
, ZIO_TYPE_READ
,
8563 ZIO_PRIORITY_SYNC_READ
,
8564 ZIO_FLAG_DONT_CACHE
|
8565 ZIO_FLAG_DONT_PROPAGATE
|
8566 ZIO_FLAG_DONT_RETRY
|
8567 ZIO_FLAG_CANFAIL
| ZIO_FLAG_RAW
|
8568 ZIO_FLAG_SPECULATIVE
|
8569 ZIO_FLAG_OPTIONAL
, NULL
, NULL
));
8571 error
= zio_wait(czio
);
8572 if (error
== 0 || error
== ECKSUM
) {
8573 zio_t
*ck_zio
= zio_root(spa
, NULL
, NULL
, 0);
8575 DVA_GET_OFFSET(&bp
->blk_dva
[0]);
8577 zio_checksum_compute(ck_zio
, ck
, pabd
, lsize
);
8580 "cksum=%016llx:%016llx:%016llx:%016llx\n",
8581 zio_checksum_table
[ck
].ci_name
,
8582 (u_longlong_t
)bp
->blk_cksum
.zc_word
[0],
8583 (u_longlong_t
)bp
->blk_cksum
.zc_word
[1],
8584 (u_longlong_t
)bp
->blk_cksum
.zc_word
[2],
8585 (u_longlong_t
)bp
->blk_cksum
.zc_word
[3]);
8588 printf("error %d reading block\n", error
);
8590 spa_config_exit(spa
, SCL_STATE
, FTAG
);
8595 abd_return_buf_copy(pabd
, buf
, lsize
);
8599 umem_free(lbuf
, SPA_MAXBLOCKSIZE
);
8606 zdb_embedded_block(char *thing
)
8608 blkptr_t bp
= {{{{0}}}};
8609 unsigned long long *words
= (void *)&bp
;
8613 err
= sscanf(thing
, "%llx:%llx:%llx:%llx:%llx:%llx:%llx:%llx:"
8614 "%llx:%llx:%llx:%llx:%llx:%llx:%llx:%llx",
8615 words
+ 0, words
+ 1, words
+ 2, words
+ 3,
8616 words
+ 4, words
+ 5, words
+ 6, words
+ 7,
8617 words
+ 8, words
+ 9, words
+ 10, words
+ 11,
8618 words
+ 12, words
+ 13, words
+ 14, words
+ 15);
8620 (void) fprintf(stderr
, "invalid input format\n");
8623 ASSERT3U(BPE_GET_LSIZE(&bp
), <=, SPA_MAXBLOCKSIZE
);
8624 buf
= malloc(SPA_MAXBLOCKSIZE
);
8626 (void) fprintf(stderr
, "out of memory\n");
8629 err
= decode_embedded_bp(&bp
, buf
, BPE_GET_LSIZE(&bp
));
8631 (void) fprintf(stderr
, "decode failed: %u\n", err
);
8634 zdb_dump_block_raw(buf
, BPE_GET_LSIZE(&bp
), 0);
8638 /* check for valid hex or decimal numeric string */
8640 zdb_numeric(char *str
)
8644 if (strlen(str
) == 0)
8646 if (strncmp(str
, "0x", 2) == 0 || strncmp(str
, "0X", 2) == 0)
8648 for (; i
< strlen(str
); i
++) {
8649 if (!isxdigit(str
[i
]))
8656 main(int argc
, char **argv
)
8660 objset_t
*os
= NULL
;
8664 char **searchdirs
= NULL
;
8666 char *target
, *target_pool
, dsname
[ZFS_MAX_DATASET_NAME_LEN
];
8667 nvlist_t
*policy
= NULL
;
8668 uint64_t max_txg
= UINT64_MAX
;
8669 int64_t objset_id
= -1;
8671 int flags
= ZFS_IMPORT_MISSING_LOG
;
8672 int rewind
= ZPOOL_NEVER_REWIND
;
8673 char *spa_config_path_env
, *objset_str
;
8674 boolean_t target_is_spa
= B_TRUE
, dataset_lookup
= B_FALSE
;
8675 nvlist_t
*cfg
= NULL
;
8677 dprintf_setup(&argc
, argv
);
8680 * If there is an environment variable SPA_CONFIG_PATH it overrides
8681 * default spa_config_path setting. If -U flag is specified it will
8682 * override this environment variable settings once again.
8684 spa_config_path_env
= getenv("SPA_CONFIG_PATH");
8685 if (spa_config_path_env
!= NULL
)
8686 spa_config_path
= spa_config_path_env
;
8689 * For performance reasons, we set this tunable down. We do so before
8690 * the arg parsing section so that the user can override this value if
8693 zfs_btree_verify_intensity
= 3;
8695 struct option long_options
[] = {
8696 {"ignore-assertions", no_argument
, NULL
, 'A'},
8697 {"block-stats", no_argument
, NULL
, 'b'},
8698 {"checksum", no_argument
, NULL
, 'c'},
8699 {"config", no_argument
, NULL
, 'C'},
8700 {"datasets", no_argument
, NULL
, 'd'},
8701 {"dedup-stats", no_argument
, NULL
, 'D'},
8702 {"exported", no_argument
, NULL
, 'e'},
8703 {"embedded-block-pointer", no_argument
, NULL
, 'E'},
8704 {"automatic-rewind", no_argument
, NULL
, 'F'},
8705 {"dump-debug-msg", no_argument
, NULL
, 'G'},
8706 {"history", no_argument
, NULL
, 'h'},
8707 {"intent-logs", no_argument
, NULL
, 'i'},
8708 {"inflight", required_argument
, NULL
, 'I'},
8709 {"checkpointed-state", no_argument
, NULL
, 'k'},
8710 {"key", required_argument
, NULL
, 'K'},
8711 {"label", no_argument
, NULL
, 'l'},
8712 {"disable-leak-tracking", no_argument
, NULL
, 'L'},
8713 {"metaslabs", no_argument
, NULL
, 'm'},
8714 {"metaslab-groups", no_argument
, NULL
, 'M'},
8715 {"numeric", no_argument
, NULL
, 'N'},
8716 {"option", required_argument
, NULL
, 'o'},
8717 {"object-lookups", no_argument
, NULL
, 'O'},
8718 {"path", required_argument
, NULL
, 'p'},
8719 {"parseable", no_argument
, NULL
, 'P'},
8720 {"skip-label", no_argument
, NULL
, 'q'},
8721 {"copy-object", no_argument
, NULL
, 'r'},
8722 {"read-block", no_argument
, NULL
, 'R'},
8723 {"io-stats", no_argument
, NULL
, 's'},
8724 {"simulate-dedup", no_argument
, NULL
, 'S'},
8725 {"txg", required_argument
, NULL
, 't'},
8726 {"uberblock", no_argument
, NULL
, 'u'},
8727 {"cachefile", required_argument
, NULL
, 'U'},
8728 {"verbose", no_argument
, NULL
, 'v'},
8729 {"verbatim", no_argument
, NULL
, 'V'},
8730 {"dump-blocks", required_argument
, NULL
, 'x'},
8731 {"extreme-rewind", no_argument
, NULL
, 'X'},
8732 {"all-reconstruction", no_argument
, NULL
, 'Y'},
8733 {"livelist", no_argument
, NULL
, 'y'},
8734 {"zstd-headers", no_argument
, NULL
, 'Z'},
8738 while ((c
= getopt_long(argc
, argv
,
8739 "AbcCdDeEFGhiI:kK:lLmMNo:Op:PqrRsSt:uU:vVx:XYyZ",
8740 long_options
, NULL
)) != -1) {
8777 zfs_reconstruct_indirect_combinations_max
= INT_MAX
;
8778 zfs_deadman_enabled
= 0;
8780 /* NB: Sort single match options below. */
8782 max_inflight_bytes
= strtoull(optarg
, NULL
, 0);
8783 if (max_inflight_bytes
== 0) {
8784 (void) fprintf(stderr
, "maximum number "
8785 "of inflight bytes must be greater "
8792 key_material
= strdup(optarg
);
8793 /* redact key material in process table */
8794 while (*optarg
!= '\0') { *optarg
++ = '*'; }
8797 error
= set_global_var(optarg
);
8802 if (searchdirs
== NULL
) {
8803 searchdirs
= umem_alloc(sizeof (char *),
8806 char **tmp
= umem_alloc((nsearch
+ 1) *
8807 sizeof (char *), UMEM_NOFAIL
);
8808 memcpy(tmp
, searchdirs
, nsearch
*
8810 umem_free(searchdirs
,
8811 nsearch
* sizeof (char *));
8814 searchdirs
[nsearch
++] = optarg
;
8817 max_txg
= strtoull(optarg
, NULL
, 0);
8818 if (max_txg
< TXG_INITIAL
) {
8819 (void) fprintf(stderr
, "incorrect txg "
8820 "specified: %s\n", optarg
);
8825 spa_config_path
= optarg
;
8826 if (spa_config_path
[0] != '/') {
8827 (void) fprintf(stderr
,
8828 "cachefile must be an absolute path "
8829 "(i.e. start with a slash)\n");
8837 flags
= ZFS_IMPORT_VERBATIM
;
8840 vn_dumpdir
= optarg
;
8848 if (!dump_opt
['e'] && searchdirs
!= NULL
) {
8849 (void) fprintf(stderr
, "-p option requires use of -e\n");
8854 * ZDB does not typically re-read blocks; therefore limit the ARC
8855 * to 256 MB, which can be used entirely for metadata.
8857 zfs_arc_min
= 2ULL << SPA_MAXBLOCKSHIFT
;
8858 zfs_arc_max
= 256 * 1024 * 1024;
8862 * "zdb -c" uses checksum-verifying scrub i/os which are async reads.
8863 * "zdb -b" uses traversal prefetch which uses async reads.
8864 * For good performance, let several of them be active at once.
8866 zfs_vdev_async_read_max_active
= 10;
8869 * Disable reference tracking for better performance.
8871 reference_tracking_enable
= B_FALSE
;
8874 * Do not fail spa_load when spa_load_verify fails. This is needed
8875 * to load non-idle pools.
8877 spa_load_verify_dryrun
= B_TRUE
;
8880 * ZDB should have ability to read spacemaps.
8882 spa_mode_readable_spacemaps
= B_TRUE
;
8884 kernel_init(SPA_MODE_READ
);
8887 verbose
= MAX(verbose
, 1);
8889 for (c
= 0; c
< 256; c
++) {
8890 if (dump_all
&& strchr("AeEFkKlLNOPrRSXy", c
) == NULL
)
8893 dump_opt
[c
] += verbose
;
8896 libspl_set_assert_ok((dump_opt
['A'] == 1) || (dump_opt
['A'] > 2));
8897 zfs_recover
= (dump_opt
['A'] > 1);
8901 if (argc
< 2 && dump_opt
['R'])
8904 if (dump_opt
['E']) {
8907 zdb_embedded_block(argv
[0]);
8912 if (!dump_opt
['e'] && dump_opt
['C']) {
8913 dump_cachefile(spa_config_path
);
8920 return (dump_label(argv
[0]));
8922 if (dump_opt
['O']) {
8925 dump_opt
['v'] = verbose
+ 3;
8926 return (dump_path(argv
[0], argv
[1], NULL
));
8928 if (dump_opt
['r']) {
8929 target_is_spa
= B_FALSE
;
8932 dump_opt
['v'] = verbose
;
8933 error
= dump_path(argv
[0], argv
[1], &object
);
8935 fatal("internal error: %s", strerror(error
));
8938 if (dump_opt
['X'] || dump_opt
['F'])
8939 rewind
= ZPOOL_DO_REWIND
|
8940 (dump_opt
['X'] ? ZPOOL_EXTREME_REWIND
: 0);
8943 if (dump_opt
['N'] && dump_opt
['d'] == 0)
8944 dump_opt
['d'] = dump_opt
['N'];
8946 if (nvlist_alloc(&policy
, NV_UNIQUE_NAME_TYPE
, 0) != 0 ||
8947 nvlist_add_uint64(policy
, ZPOOL_LOAD_REQUEST_TXG
, max_txg
) != 0 ||
8948 nvlist_add_uint32(policy
, ZPOOL_LOAD_REWIND_POLICY
, rewind
) != 0)
8949 fatal("internal error: %s", strerror(ENOMEM
));
8954 if (strpbrk(target
, "/@") != NULL
) {
8957 target_pool
= strdup(target
);
8958 *strpbrk(target_pool
, "/@") = '\0';
8960 target_is_spa
= B_FALSE
;
8961 targetlen
= strlen(target
);
8962 if (targetlen
&& target
[targetlen
- 1] == '/')
8963 target
[targetlen
- 1] = '\0';
8966 * See if an objset ID was supplied (-d <pool>/<objset ID>).
8967 * To disambiguate tank/100, consider the 100 as objsetID
8968 * if -N was given, otherwise 100 is an objsetID iff
8969 * tank/100 as a named dataset fails on lookup.
8971 objset_str
= strchr(target
, '/');
8972 if (objset_str
&& strlen(objset_str
) > 1 &&
8973 zdb_numeric(objset_str
+ 1)) {
8977 objset_id
= strtoull(objset_str
, &endptr
, 0);
8978 /* dataset 0 is the same as opening the pool */
8979 if (errno
== 0 && endptr
!= objset_str
&&
8982 dataset_lookup
= B_TRUE
;
8984 /* normal dataset name not an objset ID */
8985 if (endptr
== objset_str
) {
8988 } else if (objset_str
&& !zdb_numeric(objset_str
+ 1) &&
8990 printf("Supply a numeric objset ID with -N\n");
8994 target_pool
= target
;
8997 if (dump_opt
['e']) {
8998 importargs_t args
= { 0 };
9000 args
.paths
= nsearch
;
9001 args
.path
= searchdirs
;
9002 args
.can_be_active
= B_TRUE
;
9004 libpc_handle_t lpch
= {
9005 .lpc_lib_handle
= NULL
,
9006 .lpc_ops
= &libzpool_config_ops
,
9007 .lpc_printerr
= B_TRUE
9009 error
= zpool_find_config(&lpch
, target_pool
, &cfg
, &args
);
9013 if (nvlist_add_nvlist(cfg
,
9014 ZPOOL_LOAD_POLICY
, policy
) != 0) {
9015 fatal("can't open '%s': %s",
9016 target
, strerror(ENOMEM
));
9019 if (dump_opt
['C'] > 1) {
9020 (void) printf("\nConfiguration for import:\n");
9021 dump_nvlist(cfg
, 8);
9025 * Disable the activity check to allow examination of
9028 error
= spa_import(target_pool
, cfg
, NULL
,
9029 flags
| ZFS_IMPORT_SKIP_MMP
);
9033 if (searchdirs
!= NULL
) {
9034 umem_free(searchdirs
, nsearch
* sizeof (char *));
9039 * import_checkpointed_state makes the assumption that the
9040 * target pool that we pass it is already part of the spa
9041 * namespace. Because of that we need to make sure to call
9042 * it always after the -e option has been processed, which
9043 * imports the pool to the namespace if it's not in the
9046 char *checkpoint_pool
= NULL
;
9047 char *checkpoint_target
= NULL
;
9048 if (dump_opt
['k']) {
9049 checkpoint_pool
= import_checkpointed_state(target
, cfg
,
9050 &checkpoint_target
);
9052 if (checkpoint_target
!= NULL
)
9053 target
= checkpoint_target
;
9061 if (target_pool
!= target
)
9065 if (dump_opt
['k'] && (target_is_spa
|| dump_opt
['R'])) {
9066 ASSERT(checkpoint_pool
!= NULL
);
9067 ASSERT(checkpoint_target
== NULL
);
9069 error
= spa_open(checkpoint_pool
, &spa
, FTAG
);
9071 fatal("Tried to open pool \"%s\" but "
9072 "spa_open() failed with error %d\n",
9073 checkpoint_pool
, error
);
9076 } else if (target_is_spa
|| dump_opt
['R'] || objset_id
== 0) {
9077 zdb_set_skip_mmp(target
);
9078 error
= spa_open_rewind(target
, &spa
, FTAG
, policy
,
9082 * If we're missing the log device then
9083 * try opening the pool after clearing the
9086 mutex_enter(&spa_namespace_lock
);
9087 if ((spa
= spa_lookup(target
)) != NULL
&&
9088 spa
->spa_log_state
== SPA_LOG_MISSING
) {
9089 spa
->spa_log_state
= SPA_LOG_CLEAR
;
9092 mutex_exit(&spa_namespace_lock
);
9095 error
= spa_open_rewind(target
, &spa
,
9096 FTAG
, policy
, NULL
);
9099 } else if (strpbrk(target
, "#") != NULL
) {
9101 error
= dsl_pool_hold(target
, FTAG
, &dp
);
9103 fatal("can't dump '%s': %s", target
,
9106 error
= dump_bookmark(dp
, target
, B_TRUE
, verbose
> 1);
9107 dsl_pool_rele(dp
, FTAG
);
9109 fatal("can't dump '%s': %s", target
,
9114 target_pool
= strdup(target
);
9115 if (strpbrk(target
, "/@") != NULL
)
9116 *strpbrk(target_pool
, "/@") = '\0';
9118 zdb_set_skip_mmp(target
);
9120 * If -N was supplied, the user has indicated that
9121 * zdb -d <pool>/<objsetID> is in effect. Otherwise
9122 * we first assume that the dataset string is the
9123 * dataset name. If dmu_objset_hold fails with the
9124 * dataset string, and we have an objset_id, retry the
9125 * lookup with the objsetID.
9127 boolean_t retry
= B_TRUE
;
9129 if (dataset_lookup
== B_TRUE
) {
9131 * Use the supplied id to get the name
9134 error
= spa_open(target_pool
, &spa
, FTAG
);
9136 error
= name_from_objset_id(spa
,
9138 spa_close(spa
, FTAG
);
9144 if (objset_id
> 0 && retry
) {
9145 int err
= dmu_objset_hold(target
, FTAG
,
9148 dataset_lookup
= B_TRUE
;
9152 dmu_objset_rele(os
, FTAG
);
9155 error
= open_objset(target
, FTAG
, &os
);
9158 spa
= dmu_objset_spa(os
);
9162 nvlist_free(policy
);
9165 fatal("can't open '%s': %s", target
, strerror(error
));
9168 * Set the pool failure mode to panic in order to prevent the pool
9169 * from suspending. A suspended I/O will have no way to resume and
9170 * can prevent the zdb(8) command from terminating as expected.
9173 spa
->spa_failmode
= ZIO_FAILURE_MODE_PANIC
;
9177 if (dump_opt
['r']) {
9178 error
= zdb_copy_object(os
, object
, argv
[1]);
9179 } else if (!dump_opt
['R']) {
9180 flagbits
['d'] = ZOR_FLAG_DIRECTORY
;
9181 flagbits
['f'] = ZOR_FLAG_PLAIN_FILE
;
9182 flagbits
['m'] = ZOR_FLAG_SPACE_MAP
;
9183 flagbits
['z'] = ZOR_FLAG_ZAP
;
9184 flagbits
['A'] = ZOR_FLAG_ALL_TYPES
;
9186 if (argc
> 0 && dump_opt
['d']) {
9187 zopt_object_args
= argc
;
9188 zopt_object_ranges
= calloc(zopt_object_args
,
9189 sizeof (zopt_object_range_t
));
9190 for (unsigned i
= 0; i
< zopt_object_args
; i
++) {
9192 const char *msg
= NULL
;
9194 err
= parse_object_range(argv
[i
],
9195 &zopt_object_ranges
[i
], &msg
);
9197 fatal("Bad object or range: '%s': %s\n",
9198 argv
[i
], msg
?: "");
9200 } else if (argc
> 0 && dump_opt
['m']) {
9201 zopt_metaslab_args
= argc
;
9202 zopt_metaslab
= calloc(zopt_metaslab_args
,
9204 for (unsigned i
= 0; i
< zopt_metaslab_args
; i
++) {
9206 zopt_metaslab
[i
] = strtoull(argv
[i
], NULL
, 0);
9207 if (zopt_metaslab
[i
] == 0 && errno
!= 0)
9208 fatal("bad number %s: %s", argv
[i
],
9214 } else if (zopt_object_args
> 0 && !dump_opt
['m']) {
9215 dump_objset(spa
->spa_meta_objset
);
9220 flagbits
['b'] = ZDB_FLAG_PRINT_BLKPTR
;
9221 flagbits
['c'] = ZDB_FLAG_CHECKSUM
;
9222 flagbits
['d'] = ZDB_FLAG_DECOMPRESS
;
9223 flagbits
['e'] = ZDB_FLAG_BSWAP
;
9224 flagbits
['g'] = ZDB_FLAG_GBH
;
9225 flagbits
['i'] = ZDB_FLAG_INDIRECT
;
9226 flagbits
['r'] = ZDB_FLAG_RAW
;
9227 flagbits
['v'] = ZDB_FLAG_VERBOSE
;
9229 for (int i
= 0; i
< argc
; i
++)
9230 zdb_read_block(argv
[i
], spa
);
9233 if (dump_opt
['k']) {
9234 free(checkpoint_pool
);
9236 free(checkpoint_target
);
9240 close_objset(os
, FTAG
);
9242 spa_close(spa
, FTAG
);
9245 fuid_table_destroy();
9247 dump_debug_buffer();