4 * The contents of this file are subject to the terms of the
5 * Common Development and Distribution License (the "License").
6 * You may not use this file except in compliance with the License.
8 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
9 * or http://www.opensolaris.org/os/licensing.
10 * See the License for the specific language governing permissions
11 * and limitations under the License.
13 * When distributing Covered Code, include this CDDL HEADER in each
14 * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
15 * If applicable, add the following below this CDDL HEADER, with the
16 * fields enclosed by brackets "[]" replaced with your own identifying
17 * information: Portions Copyright [yyyy] [name of copyright owner]
23 * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
24 * Copyright (c) 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>
42 #include <sys/zfs_context.h>
44 #include <sys/spa_impl.h>
47 #include <sys/fs/zfs.h>
48 #include <sys/zfs_znode.h>
49 #include <sys/zfs_sa.h>
51 #include <sys/sa_impl.h>
53 #include <sys/vdev_impl.h>
54 #include <sys/metaslab_impl.h>
55 #include <sys/dmu_objset.h>
56 #include <sys/dsl_dir.h>
57 #include <sys/dsl_dataset.h>
58 #include <sys/dsl_pool.h>
59 #include <sys/dsl_bookmark.h>
62 #include <sys/zil_impl.h>
64 #include <sys/resource.h>
65 #include <sys/dmu_send.h>
66 #include <sys/dmu_traverse.h>
67 #include <sys/zio_checksum.h>
68 #include <sys/zio_compress.h>
69 #include <sys/zfs_fuid.h>
71 #include <sys/arc_impl.h>
73 #include <sys/zfeature.h>
75 #include <sys/blkptr.h>
76 #include <sys/dsl_crypt.h>
77 #include <sys/dsl_scan.h>
78 #include <sys/btree.h>
79 #include <zfs_comutil.h>
80 #include <sys/zstd/zstd.h>
82 #include <libnvpair.h>
87 #define ZDB_COMPRESS_NAME(idx) ((idx) < ZIO_COMPRESS_FUNCTIONS ? \
88 zio_compress_table[(idx)].ci_name : "UNKNOWN")
89 #define ZDB_CHECKSUM_NAME(idx) ((idx) < ZIO_CHECKSUM_FUNCTIONS ? \
90 zio_checksum_table[(idx)].ci_name : "UNKNOWN")
91 #define ZDB_OT_TYPE(idx) ((idx) < DMU_OT_NUMTYPES ? (idx) : \
92 (idx) == DMU_OTN_ZAP_DATA || (idx) == DMU_OTN_ZAP_METADATA ? \
94 (idx) == DMU_OTN_UINT64_DATA || (idx) == DMU_OTN_UINT64_METADATA ? \
95 DMU_OT_UINT64_OTHER : DMU_OT_NUMTYPES)
98 zdb_ot_name(dmu_object_type_t type
)
100 if (type
< DMU_OT_NUMTYPES
)
101 return (dmu_ot
[type
].ot_name
);
102 else if ((type
& DMU_OT_NEWTYPE
) &&
103 ((type
& DMU_OT_BYTESWAP_MASK
) < DMU_BSWAP_NUMFUNCS
))
104 return (dmu_ot_byteswap
[type
& DMU_OT_BYTESWAP_MASK
].ob_name
);
109 extern int reference_tracking_enable
;
110 extern int zfs_recover
;
111 extern unsigned long zfs_arc_meta_min
, zfs_arc_meta_limit
;
112 extern int zfs_vdev_async_read_max_active
;
113 extern boolean_t spa_load_verify_dryrun
;
114 extern int zfs_reconstruct_indirect_combinations_max
;
115 extern int zfs_btree_verify_intensity
;
117 static const char cmdname
[] = "zdb";
118 uint8_t dump_opt
[256];
120 typedef void object_viewer_t(objset_t
*, uint64_t, void *data
, size_t size
);
122 uint64_t *zopt_metaslab
= NULL
;
123 static unsigned zopt_metaslab_args
= 0;
125 typedef struct zopt_object_range
{
126 uint64_t zor_obj_start
;
127 uint64_t zor_obj_end
;
129 } zopt_object_range_t
;
130 zopt_object_range_t
*zopt_object_ranges
= NULL
;
131 static unsigned zopt_object_args
= 0;
133 static int flagbits
[256];
135 #define ZOR_FLAG_PLAIN_FILE 0x0001
136 #define ZOR_FLAG_DIRECTORY 0x0002
137 #define ZOR_FLAG_SPACE_MAP 0x0004
138 #define ZOR_FLAG_ZAP 0x0008
139 #define ZOR_FLAG_ALL_TYPES -1
140 #define ZOR_SUPPORTED_FLAGS (ZOR_FLAG_PLAIN_FILE | \
141 ZOR_FLAG_DIRECTORY | \
142 ZOR_FLAG_SPACE_MAP | \
145 #define ZDB_FLAG_CHECKSUM 0x0001
146 #define ZDB_FLAG_DECOMPRESS 0x0002
147 #define ZDB_FLAG_BSWAP 0x0004
148 #define ZDB_FLAG_GBH 0x0008
149 #define ZDB_FLAG_INDIRECT 0x0010
150 #define ZDB_FLAG_RAW 0x0020
151 #define ZDB_FLAG_PRINT_BLKPTR 0x0040
152 #define ZDB_FLAG_VERBOSE 0x0080
154 uint64_t max_inflight_bytes
= 256 * 1024 * 1024; /* 256MB */
155 static int leaked_objects
= 0;
156 static range_tree_t
*mos_refd_objs
;
158 static void snprintf_blkptr_compact(char *, size_t, const blkptr_t
*,
160 static void mos_obj_refd(uint64_t);
161 static void mos_obj_refd_multiple(uint64_t);
162 static int dump_bpobj_cb(void *arg
, const blkptr_t
*bp
, boolean_t free
,
165 typedef struct sublivelist_verify
{
166 /* FREE's that haven't yet matched to an ALLOC, in one sub-livelist */
169 /* ALLOC's without a matching FREE, accumulates across sub-livelists */
170 zfs_btree_t sv_leftover
;
171 } sublivelist_verify_t
;
174 livelist_compare(const void *larg
, const void *rarg
)
176 const blkptr_t
*l
= larg
;
177 const blkptr_t
*r
= rarg
;
179 /* Sort them according to dva[0] */
180 uint64_t l_dva0_vdev
, r_dva0_vdev
;
181 l_dva0_vdev
= DVA_GET_VDEV(&l
->blk_dva
[0]);
182 r_dva0_vdev
= DVA_GET_VDEV(&r
->blk_dva
[0]);
183 if (l_dva0_vdev
< r_dva0_vdev
)
185 else if (l_dva0_vdev
> r_dva0_vdev
)
188 /* if vdevs are equal, sort by offsets. */
189 uint64_t l_dva0_offset
;
190 uint64_t r_dva0_offset
;
191 l_dva0_offset
= DVA_GET_OFFSET(&l
->blk_dva
[0]);
192 r_dva0_offset
= DVA_GET_OFFSET(&r
->blk_dva
[0]);
193 if (l_dva0_offset
< r_dva0_offset
) {
195 } else if (l_dva0_offset
> r_dva0_offset
) {
200 * Since we're storing blkptrs without cancelling FREE/ALLOC pairs,
201 * it's possible the offsets are equal. In that case, sort by txg
203 if (l
->blk_birth
< r
->blk_birth
) {
205 } else if (l
->blk_birth
> r
->blk_birth
) {
211 typedef struct sublivelist_verify_block
{
215 * We need this to check if the block marked as allocated
216 * in the livelist was freed (and potentially reallocated)
217 * in the metaslab spacemaps at a later TXG.
219 uint64_t svb_allocated_txg
;
220 } sublivelist_verify_block_t
;
222 static void zdb_print_blkptr(const blkptr_t
*bp
, int flags
);
224 typedef struct sublivelist_verify_block_refcnt
{
225 /* block pointer entry in livelist being verified */
229 * Refcount gets incremented to 1 when we encounter the first
230 * FREE entry for the svfbr block pointer and a node for it
231 * is created in our ZDB verification/tracking metadata.
233 * As we encounter more FREE entries we increment this counter
234 * and similarly decrement it whenever we find the respective
235 * ALLOC entries for this block.
237 * When the refcount gets to 0 it means that all the FREE and
238 * ALLOC entries of this block have paired up and we no longer
239 * need to track it in our verification logic (e.g. the node
240 * containing this struct in our verification data structure
243 * [refer to sublivelist_verify_blkptr() for the actual code]
245 uint32_t svbr_refcnt
;
246 } sublivelist_verify_block_refcnt_t
;
249 sublivelist_block_refcnt_compare(const void *larg
, const void *rarg
)
251 const sublivelist_verify_block_refcnt_t
*l
= larg
;
252 const sublivelist_verify_block_refcnt_t
*r
= rarg
;
253 return (livelist_compare(&l
->svbr_blk
, &r
->svbr_blk
));
257 sublivelist_verify_blkptr(void *arg
, const blkptr_t
*bp
, boolean_t free
,
260 ASSERT3P(tx
, ==, NULL
);
261 struct sublivelist_verify
*sv
= arg
;
262 sublivelist_verify_block_refcnt_t current
= {
266 * Start with 1 in case this is the first free entry.
267 * This field is not used for our B-Tree comparisons
273 zfs_btree_index_t where
;
274 sublivelist_verify_block_refcnt_t
*pair
=
275 zfs_btree_find(&sv
->sv_pair
, ¤t
, &where
);
278 /* first free entry for this block pointer */
279 zfs_btree_add(&sv
->sv_pair
, ¤t
);
285 /* block that is currently marked as allocated */
286 for (int i
= 0; i
< SPA_DVAS_PER_BP
; i
++) {
287 if (DVA_IS_EMPTY(&bp
->blk_dva
[i
]))
289 sublivelist_verify_block_t svb
= {
290 .svb_dva
= bp
->blk_dva
[i
],
291 .svb_allocated_txg
= bp
->blk_birth
294 if (zfs_btree_find(&sv
->sv_leftover
, &svb
,
296 zfs_btree_add_idx(&sv
->sv_leftover
,
301 /* alloc matches a free entry */
303 if (pair
->svbr_refcnt
== 0) {
304 /* all allocs and frees have been matched */
305 zfs_btree_remove_idx(&sv
->sv_pair
, &where
);
314 sublivelist_verify_func(void *args
, dsl_deadlist_entry_t
*dle
)
317 struct sublivelist_verify
*sv
= args
;
319 zfs_btree_create(&sv
->sv_pair
, sublivelist_block_refcnt_compare
,
320 sizeof (sublivelist_verify_block_refcnt_t
));
322 err
= bpobj_iterate_nofree(&dle
->dle_bpobj
, sublivelist_verify_blkptr
,
325 sublivelist_verify_block_refcnt_t
*e
;
326 zfs_btree_index_t
*cookie
= NULL
;
327 while ((e
= zfs_btree_destroy_nodes(&sv
->sv_pair
, &cookie
)) != NULL
) {
328 char blkbuf
[BP_SPRINTF_LEN
];
329 snprintf_blkptr_compact(blkbuf
, sizeof (blkbuf
),
330 &e
->svbr_blk
, B_TRUE
);
331 (void) printf("\tERROR: %d unmatched FREE(s): %s\n",
332 e
->svbr_refcnt
, blkbuf
);
334 zfs_btree_destroy(&sv
->sv_pair
);
340 livelist_block_compare(const void *larg
, const void *rarg
)
342 const sublivelist_verify_block_t
*l
= larg
;
343 const sublivelist_verify_block_t
*r
= rarg
;
345 if (DVA_GET_VDEV(&l
->svb_dva
) < DVA_GET_VDEV(&r
->svb_dva
))
347 else if (DVA_GET_VDEV(&l
->svb_dva
) > DVA_GET_VDEV(&r
->svb_dva
))
350 if (DVA_GET_OFFSET(&l
->svb_dva
) < DVA_GET_OFFSET(&r
->svb_dva
))
352 else if (DVA_GET_OFFSET(&l
->svb_dva
) > DVA_GET_OFFSET(&r
->svb_dva
))
355 if (DVA_GET_ASIZE(&l
->svb_dva
) < DVA_GET_ASIZE(&r
->svb_dva
))
357 else if (DVA_GET_ASIZE(&l
->svb_dva
) > DVA_GET_ASIZE(&r
->svb_dva
))
364 * Check for errors in a livelist while tracking all unfreed ALLOCs in the
365 * sublivelist_verify_t: sv->sv_leftover
368 livelist_verify(dsl_deadlist_t
*dl
, void *arg
)
370 sublivelist_verify_t
*sv
= arg
;
371 dsl_deadlist_iterate(dl
, sublivelist_verify_func
, sv
);
375 * Check for errors in the livelist entry and discard the intermediary
380 sublivelist_verify_lightweight(void *args
, dsl_deadlist_entry_t
*dle
)
382 sublivelist_verify_t sv
;
383 zfs_btree_create(&sv
.sv_leftover
, livelist_block_compare
,
384 sizeof (sublivelist_verify_block_t
));
385 int err
= sublivelist_verify_func(&sv
, dle
);
386 zfs_btree_clear(&sv
.sv_leftover
);
387 zfs_btree_destroy(&sv
.sv_leftover
);
391 typedef struct metaslab_verify
{
393 * Tree containing all the leftover ALLOCs from the livelists
394 * that are part of this metaslab.
396 zfs_btree_t mv_livelist_allocs
;
399 * Metaslab information.
407 * What's currently allocated for this metaslab.
409 range_tree_t
*mv_allocated
;
412 typedef void ll_iter_t(dsl_deadlist_t
*ll
, void *arg
);
414 typedef int (*zdb_log_sm_cb_t
)(spa_t
*spa
, space_map_entry_t
*sme
, uint64_t txg
,
417 typedef struct unflushed_iter_cb_arg
{
421 zdb_log_sm_cb_t uic_cb
;
422 } unflushed_iter_cb_arg_t
;
425 iterate_through_spacemap_logs_cb(space_map_entry_t
*sme
, void *arg
)
427 unflushed_iter_cb_arg_t
*uic
= arg
;
428 return (uic
->uic_cb(uic
->uic_spa
, sme
, uic
->uic_txg
, uic
->uic_arg
));
432 iterate_through_spacemap_logs(spa_t
*spa
, zdb_log_sm_cb_t cb
, void *arg
)
434 if (!spa_feature_is_active(spa
, SPA_FEATURE_LOG_SPACEMAP
))
437 spa_config_enter(spa
, SCL_CONFIG
, FTAG
, RW_READER
);
438 for (spa_log_sm_t
*sls
= avl_first(&spa
->spa_sm_logs_by_txg
);
439 sls
; sls
= AVL_NEXT(&spa
->spa_sm_logs_by_txg
, sls
)) {
440 space_map_t
*sm
= NULL
;
441 VERIFY0(space_map_open(&sm
, spa_meta_objset(spa
),
442 sls
->sls_sm_obj
, 0, UINT64_MAX
, SPA_MINBLOCKSHIFT
));
444 unflushed_iter_cb_arg_t uic
= {
446 .uic_txg
= sls
->sls_txg
,
450 VERIFY0(space_map_iterate(sm
, space_map_length(sm
),
451 iterate_through_spacemap_logs_cb
, &uic
));
454 spa_config_exit(spa
, SCL_CONFIG
, FTAG
);
458 verify_livelist_allocs(metaslab_verify_t
*mv
, uint64_t txg
,
459 uint64_t offset
, uint64_t size
)
461 sublivelist_verify_block_t svb
;
462 DVA_SET_VDEV(&svb
.svb_dva
, mv
->mv_vdid
);
463 DVA_SET_OFFSET(&svb
.svb_dva
, offset
);
464 DVA_SET_ASIZE(&svb
.svb_dva
, size
);
465 zfs_btree_index_t where
;
466 uint64_t end_offset
= offset
+ size
;
469 * Look for an exact match for spacemap entry in the livelist entries.
470 * Then, look for other livelist entries that fall within the range
471 * of the spacemap entry as it may have been condensed
473 sublivelist_verify_block_t
*found
=
474 zfs_btree_find(&mv
->mv_livelist_allocs
, &svb
, &where
);
476 found
= zfs_btree_next(&mv
->mv_livelist_allocs
, &where
, &where
);
478 for (; found
!= NULL
&& DVA_GET_VDEV(&found
->svb_dva
) == mv
->mv_vdid
&&
479 DVA_GET_OFFSET(&found
->svb_dva
) < end_offset
;
480 found
= zfs_btree_next(&mv
->mv_livelist_allocs
, &where
, &where
)) {
481 if (found
->svb_allocated_txg
<= txg
) {
482 (void) printf("ERROR: Livelist ALLOC [%llx:%llx] "
483 "from TXG %llx FREED at TXG %llx\n",
484 (u_longlong_t
)DVA_GET_OFFSET(&found
->svb_dva
),
485 (u_longlong_t
)DVA_GET_ASIZE(&found
->svb_dva
),
486 (u_longlong_t
)found
->svb_allocated_txg
,
493 metaslab_spacemap_validation_cb(space_map_entry_t
*sme
, void *arg
)
495 metaslab_verify_t
*mv
= arg
;
496 uint64_t offset
= sme
->sme_offset
;
497 uint64_t size
= sme
->sme_run
;
498 uint64_t txg
= sme
->sme_txg
;
500 if (sme
->sme_type
== SM_ALLOC
) {
501 if (range_tree_contains(mv
->mv_allocated
,
503 (void) printf("ERROR: DOUBLE ALLOC: "
505 "%llu:%llu LOG_SM\n",
506 (u_longlong_t
)txg
, (u_longlong_t
)offset
,
507 (u_longlong_t
)size
, (u_longlong_t
)mv
->mv_vdid
,
508 (u_longlong_t
)mv
->mv_msid
);
510 range_tree_add(mv
->mv_allocated
,
514 if (!range_tree_contains(mv
->mv_allocated
,
516 (void) printf("ERROR: DOUBLE FREE: "
518 "%llu:%llu LOG_SM\n",
519 (u_longlong_t
)txg
, (u_longlong_t
)offset
,
520 (u_longlong_t
)size
, (u_longlong_t
)mv
->mv_vdid
,
521 (u_longlong_t
)mv
->mv_msid
);
523 range_tree_remove(mv
->mv_allocated
,
528 if (sme
->sme_type
!= SM_ALLOC
) {
530 * If something is freed in the spacemap, verify that
531 * it is not listed as allocated in the livelist.
533 verify_livelist_allocs(mv
, txg
, offset
, size
);
539 spacemap_check_sm_log_cb(spa_t
*spa
, space_map_entry_t
*sme
,
540 uint64_t txg
, void *arg
)
542 metaslab_verify_t
*mv
= arg
;
543 uint64_t offset
= sme
->sme_offset
;
544 uint64_t vdev_id
= sme
->sme_vdev
;
546 vdev_t
*vd
= vdev_lookup_top(spa
, vdev_id
);
548 /* skip indirect vdevs */
549 if (!vdev_is_concrete(vd
))
552 if (vdev_id
!= mv
->mv_vdid
)
555 metaslab_t
*ms
= vd
->vdev_ms
[offset
>> vd
->vdev_ms_shift
];
556 if (ms
->ms_id
!= mv
->mv_msid
)
559 if (txg
< metaslab_unflushed_txg(ms
))
563 ASSERT3U(txg
, ==, sme
->sme_txg
);
564 return (metaslab_spacemap_validation_cb(sme
, mv
));
568 spacemap_check_sm_log(spa_t
*spa
, metaslab_verify_t
*mv
)
570 iterate_through_spacemap_logs(spa
, spacemap_check_sm_log_cb
, mv
);
574 spacemap_check_ms_sm(space_map_t
*sm
, metaslab_verify_t
*mv
)
579 VERIFY0(space_map_iterate(sm
, space_map_length(sm
),
580 metaslab_spacemap_validation_cb
, mv
));
583 static void iterate_deleted_livelists(spa_t
*spa
, ll_iter_t func
, void *arg
);
586 * Transfer blocks from sv_leftover tree to the mv_livelist_allocs if
587 * they are part of that metaslab (mv_msid).
590 mv_populate_livelist_allocs(metaslab_verify_t
*mv
, sublivelist_verify_t
*sv
)
592 zfs_btree_index_t where
;
593 sublivelist_verify_block_t
*svb
;
594 ASSERT3U(zfs_btree_numnodes(&mv
->mv_livelist_allocs
), ==, 0);
595 for (svb
= zfs_btree_first(&sv
->sv_leftover
, &where
);
597 svb
= zfs_btree_next(&sv
->sv_leftover
, &where
, &where
)) {
598 if (DVA_GET_VDEV(&svb
->svb_dva
) != mv
->mv_vdid
)
601 if (DVA_GET_OFFSET(&svb
->svb_dva
) < mv
->mv_start
&&
602 (DVA_GET_OFFSET(&svb
->svb_dva
) +
603 DVA_GET_ASIZE(&svb
->svb_dva
)) > mv
->mv_start
) {
604 (void) printf("ERROR: Found block that crosses "
605 "metaslab boundary: <%llu:%llx:%llx>\n",
606 (u_longlong_t
)DVA_GET_VDEV(&svb
->svb_dva
),
607 (u_longlong_t
)DVA_GET_OFFSET(&svb
->svb_dva
),
608 (u_longlong_t
)DVA_GET_ASIZE(&svb
->svb_dva
));
612 if (DVA_GET_OFFSET(&svb
->svb_dva
) < mv
->mv_start
)
615 if (DVA_GET_OFFSET(&svb
->svb_dva
) >= mv
->mv_end
)
618 if ((DVA_GET_OFFSET(&svb
->svb_dva
) +
619 DVA_GET_ASIZE(&svb
->svb_dva
)) > mv
->mv_end
) {
620 (void) printf("ERROR: Found block that crosses "
621 "metaslab boundary: <%llu:%llx:%llx>\n",
622 (u_longlong_t
)DVA_GET_VDEV(&svb
->svb_dva
),
623 (u_longlong_t
)DVA_GET_OFFSET(&svb
->svb_dva
),
624 (u_longlong_t
)DVA_GET_ASIZE(&svb
->svb_dva
));
628 zfs_btree_add(&mv
->mv_livelist_allocs
, svb
);
631 for (svb
= zfs_btree_first(&mv
->mv_livelist_allocs
, &where
);
633 svb
= zfs_btree_next(&mv
->mv_livelist_allocs
, &where
, &where
)) {
634 zfs_btree_remove(&sv
->sv_leftover
, svb
);
640 * Iterate through all the sublivelists and:
641 * - report leftover frees (**)
642 * - record leftover ALLOCs together with their TXG [see Cross Check]
644 * (**) Note: Double ALLOCs are valid in datasets that have dedup
645 * enabled. Similarly double FREEs are allowed as well but
646 * only if they pair up with a corresponding ALLOC entry once
647 * we our done with our sublivelist iteration.
651 * - iterate over spacemap and then the metaslab's entries in the
652 * spacemap log, then report any double FREEs and ALLOCs (do not
656 * After finishing the Livelist Check phase and while being in the
657 * Spacemap Check phase, we find all the recorded leftover ALLOCs
658 * of the livelist check that are part of the metaslab that we are
659 * currently looking at in the Spacemap Check. We report any entries
660 * that are marked as ALLOCs in the livelists but have been actually
661 * freed (and potentially allocated again) after their TXG stamp in
662 * the spacemaps. Also report any ALLOCs from the livelists that
663 * belong to indirect vdevs (e.g. their vdev completed removal).
665 * Note that this will miss Log Spacemap entries that cancelled each other
666 * out before being flushed to the metaslab, so we are not guaranteed
667 * to match all erroneous ALLOCs.
670 livelist_metaslab_validate(spa_t
*spa
)
672 (void) printf("Verifying deleted livelist entries\n");
674 sublivelist_verify_t sv
;
675 zfs_btree_create(&sv
.sv_leftover
, livelist_block_compare
,
676 sizeof (sublivelist_verify_block_t
));
677 iterate_deleted_livelists(spa
, livelist_verify
, &sv
);
679 (void) printf("Verifying metaslab entries\n");
680 vdev_t
*rvd
= spa
->spa_root_vdev
;
681 for (uint64_t c
= 0; c
< rvd
->vdev_children
; c
++) {
682 vdev_t
*vd
= rvd
->vdev_child
[c
];
684 if (!vdev_is_concrete(vd
))
687 for (uint64_t mid
= 0; mid
< vd
->vdev_ms_count
; mid
++) {
688 metaslab_t
*m
= vd
->vdev_ms
[mid
];
690 (void) fprintf(stderr
,
691 "\rverifying concrete vdev %llu, "
692 "metaslab %llu of %llu ...",
693 (longlong_t
)vd
->vdev_id
,
695 (longlong_t
)vd
->vdev_ms_count
);
697 uint64_t shift
, start
;
698 range_seg_type_t type
=
699 metaslab_calculate_range_tree_type(vd
, m
,
701 metaslab_verify_t mv
;
702 mv
.mv_allocated
= range_tree_create(NULL
,
703 type
, NULL
, start
, shift
);
704 mv
.mv_vdid
= vd
->vdev_id
;
705 mv
.mv_msid
= m
->ms_id
;
706 mv
.mv_start
= m
->ms_start
;
707 mv
.mv_end
= m
->ms_start
+ m
->ms_size
;
708 zfs_btree_create(&mv
.mv_livelist_allocs
,
709 livelist_block_compare
,
710 sizeof (sublivelist_verify_block_t
));
712 mv_populate_livelist_allocs(&mv
, &sv
);
714 spacemap_check_ms_sm(m
->ms_sm
, &mv
);
715 spacemap_check_sm_log(spa
, &mv
);
717 range_tree_vacate(mv
.mv_allocated
, NULL
, NULL
);
718 range_tree_destroy(mv
.mv_allocated
);
719 zfs_btree_clear(&mv
.mv_livelist_allocs
);
720 zfs_btree_destroy(&mv
.mv_livelist_allocs
);
723 (void) fprintf(stderr
, "\n");
726 * If there are any segments in the leftover tree after we walked
727 * through all the metaslabs in the concrete vdevs then this means
728 * that we have segments in the livelists that belong to indirect
729 * vdevs and are marked as allocated.
731 if (zfs_btree_numnodes(&sv
.sv_leftover
) == 0) {
732 zfs_btree_destroy(&sv
.sv_leftover
);
735 (void) printf("ERROR: Found livelist blocks marked as allocated "
736 "for indirect vdevs:\n");
738 zfs_btree_index_t
*where
= NULL
;
739 sublivelist_verify_block_t
*svb
;
740 while ((svb
= zfs_btree_destroy_nodes(&sv
.sv_leftover
, &where
)) !=
742 int vdev_id
= DVA_GET_VDEV(&svb
->svb_dva
);
743 ASSERT3U(vdev_id
, <, rvd
->vdev_children
);
744 vdev_t
*vd
= rvd
->vdev_child
[vdev_id
];
745 ASSERT(!vdev_is_concrete(vd
));
746 (void) printf("<%d:%llx:%llx> TXG %llx\n",
747 vdev_id
, (u_longlong_t
)DVA_GET_OFFSET(&svb
->svb_dva
),
748 (u_longlong_t
)DVA_GET_ASIZE(&svb
->svb_dva
),
749 (u_longlong_t
)svb
->svb_allocated_txg
);
752 zfs_btree_destroy(&sv
.sv_leftover
);
756 * These libumem hooks provide a reasonable set of defaults for the allocator's
757 * debugging facilities.
760 _umem_debug_init(void)
762 return ("default,verbose"); /* $UMEM_DEBUG setting */
766 _umem_logging_init(void)
768 return ("fail,contents"); /* $UMEM_LOGGING setting */
774 (void) fprintf(stderr
,
775 "Usage:\t%s [-AbcdDFGhikLMPsvXy] [-e [-V] [-p <path> ...]] "
776 "[-I <inflight I/Os>]\n"
777 "\t\t[-o <var>=<value>]... [-t <txg>] [-U <cache>] [-x <dumpdir>]\n"
778 "\t\t[<poolname>[/<dataset | objset id>] [<object | range> ...]]\n"
779 "\t%s [-AdiPv] [-e [-V] [-p <path> ...]] [-U <cache>]\n"
780 "\t\t[<poolname>[/<dataset | objset id>] [<object | range> ...]\n"
781 "\t%s [-v] <bookmark>\n"
782 "\t%s -C [-A] [-U <cache>]\n"
783 "\t%s -l [-Aqu] <device>\n"
784 "\t%s -m [-AFLPX] [-e [-V] [-p <path> ...]] [-t <txg>] "
785 "[-U <cache>]\n\t\t<poolname> [<vdev> [<metaslab> ...]]\n"
786 "\t%s -O <dataset> <path>\n"
787 "\t%s -r <dataset> <path> <destination>\n"
788 "\t%s -R [-A] [-e [-V] [-p <path> ...]] [-U <cache>]\n"
789 "\t\t<poolname> <vdev>:<offset>:<size>[:<flags>]\n"
790 "\t%s -E [-A] word0:word1:...:word15\n"
791 "\t%s -S [-AP] [-e [-V] [-p <path> ...]] [-U <cache>] "
793 cmdname
, cmdname
, cmdname
, cmdname
, cmdname
, cmdname
, cmdname
,
794 cmdname
, cmdname
, cmdname
, cmdname
);
796 (void) fprintf(stderr
, " Dataset name must include at least one "
797 "separator character '/' or '@'\n");
798 (void) fprintf(stderr
, " If dataset name is specified, only that "
799 "dataset is dumped\n");
800 (void) fprintf(stderr
, " If object numbers or object number "
801 "ranges are specified, only those\n"
802 " objects or ranges are dumped.\n\n");
803 (void) fprintf(stderr
,
804 " Object ranges take the form <start>:<end>[:<flags>]\n"
805 " start Starting object number\n"
806 " end Ending object number, or -1 for no upper bound\n"
807 " flags Optional flags to select object types:\n"
808 " A All objects (this is the default)\n"
809 " d ZFS directories\n"
811 " m SPA space maps\n"
813 " - Negate effect of next flag\n\n");
814 (void) fprintf(stderr
, " Options to control amount of output:\n");
815 (void) fprintf(stderr
, " -b block statistics\n");
816 (void) fprintf(stderr
, " -c checksum all metadata (twice for "
817 "all data) blocks\n");
818 (void) fprintf(stderr
, " -C config (or cachefile if alone)\n");
819 (void) fprintf(stderr
, " -d dataset(s)\n");
820 (void) fprintf(stderr
, " -D dedup statistics\n");
821 (void) fprintf(stderr
, " -E decode and display block from an "
822 "embedded block pointer\n");
823 (void) fprintf(stderr
, " -h pool history\n");
824 (void) fprintf(stderr
, " -i intent logs\n");
825 (void) fprintf(stderr
, " -l read label contents\n");
826 (void) fprintf(stderr
, " -k examine the checkpointed state "
828 (void) fprintf(stderr
, " -L disable leak tracking (do not "
829 "load spacemaps)\n");
830 (void) fprintf(stderr
, " -m metaslabs\n");
831 (void) fprintf(stderr
, " -M metaslab groups\n");
832 (void) fprintf(stderr
, " -O perform object lookups by path\n");
833 (void) fprintf(stderr
, " -r copy an object by path to file\n");
834 (void) fprintf(stderr
, " -R read and display block from a "
836 (void) fprintf(stderr
, " -s report stats on zdb's I/O\n");
837 (void) fprintf(stderr
, " -S simulate dedup to measure effect\n");
838 (void) fprintf(stderr
, " -v verbose (applies to all "
840 (void) fprintf(stderr
, " -y perform livelist and metaslab "
841 "validation on any livelists being deleted\n\n");
842 (void) fprintf(stderr
, " Below options are intended for use "
843 "with other options:\n");
844 (void) fprintf(stderr
, " -A ignore assertions (-A), enable "
845 "panic recovery (-AA) or both (-AAA)\n");
846 (void) fprintf(stderr
, " -e pool is exported/destroyed/"
847 "has altroot/not in a cachefile\n");
848 (void) fprintf(stderr
, " -F attempt automatic rewind within "
849 "safe range of transaction groups\n");
850 (void) fprintf(stderr
, " -G dump zfs_dbgmsg buffer before "
852 (void) fprintf(stderr
, " -I <number of inflight I/Os> -- "
853 "specify the maximum number of\n "
854 "checksumming I/Os [default is 200]\n");
855 (void) fprintf(stderr
, " -o <variable>=<value> set global "
856 "variable to an unsigned 32-bit integer\n");
857 (void) fprintf(stderr
, " -p <path> -- use one or more with "
858 "-e to specify path to vdev dir\n");
859 (void) fprintf(stderr
, " -P print numbers in parseable form\n");
860 (void) fprintf(stderr
, " -q don't print label contents\n");
861 (void) fprintf(stderr
, " -t <txg> -- highest txg to use when "
862 "searching for uberblocks\n");
863 (void) fprintf(stderr
, " -u uberblock\n");
864 (void) fprintf(stderr
, " -U <cachefile_path> -- use alternate "
866 (void) fprintf(stderr
, " -V do verbatim import\n");
867 (void) fprintf(stderr
, " -x <dumpdir> -- "
868 "dump all read blocks into specified directory\n");
869 (void) fprintf(stderr
, " -X attempt extreme rewind (does not "
870 "work with dataset)\n");
871 (void) fprintf(stderr
, " -Y attempt all reconstruction "
872 "combinations for split blocks\n");
873 (void) fprintf(stderr
, " -Z show ZSTD headers \n");
874 (void) fprintf(stderr
, "Specify an option more than once (e.g. -bb) "
875 "to make only that option verbose\n");
876 (void) fprintf(stderr
, "Default is to dump everything non-verbosely\n");
881 dump_debug_buffer(void)
885 (void) fflush(stdout
);
886 zfs_dbgmsg_print("zdb");
891 * Called for usage errors that are discovered after a call to spa_open(),
892 * dmu_bonus_hold(), or pool_match(). abort() is called for other errors.
896 fatal(const char *fmt
, ...)
901 (void) fprintf(stderr
, "%s: ", cmdname
);
902 (void) vfprintf(stderr
, fmt
, ap
);
904 (void) fprintf(stderr
, "\n");
913 dump_packed_nvlist(objset_t
*os
, uint64_t object
, void *data
, size_t size
)
916 size_t nvsize
= *(uint64_t *)data
;
917 char *packed
= umem_alloc(nvsize
, UMEM_NOFAIL
);
919 VERIFY(0 == dmu_read(os
, object
, 0, nvsize
, packed
, DMU_READ_PREFETCH
));
921 VERIFY(nvlist_unpack(packed
, nvsize
, &nv
, 0) == 0);
923 umem_free(packed
, nvsize
);
932 dump_history_offsets(objset_t
*os
, uint64_t object
, void *data
, size_t size
)
934 spa_history_phys_t
*shp
= data
;
939 (void) printf("\t\tpool_create_len = %llu\n",
940 (u_longlong_t
)shp
->sh_pool_create_len
);
941 (void) printf("\t\tphys_max_off = %llu\n",
942 (u_longlong_t
)shp
->sh_phys_max_off
);
943 (void) printf("\t\tbof = %llu\n",
944 (u_longlong_t
)shp
->sh_bof
);
945 (void) printf("\t\teof = %llu\n",
946 (u_longlong_t
)shp
->sh_eof
);
947 (void) printf("\t\trecords_lost = %llu\n",
948 (u_longlong_t
)shp
->sh_records_lost
);
952 zdb_nicenum(uint64_t num
, char *buf
, size_t buflen
)
955 (void) snprintf(buf
, buflen
, "%llu", (longlong_t
)num
);
957 nicenum(num
, buf
, sizeof (buf
));
960 static const char histo_stars
[] = "****************************************";
961 static const uint64_t histo_width
= sizeof (histo_stars
) - 1;
964 dump_histogram(const uint64_t *histo
, int size
, int offset
)
967 int minidx
= size
- 1;
971 for (i
= 0; i
< size
; i
++) {
974 if (histo
[i
] > 0 && i
> maxidx
)
976 if (histo
[i
] > 0 && i
< minidx
)
980 if (max
< histo_width
)
983 for (i
= minidx
; i
<= maxidx
; i
++) {
984 (void) printf("\t\t\t%3u: %6llu %s\n",
985 i
+ offset
, (u_longlong_t
)histo
[i
],
986 &histo_stars
[(max
- histo
[i
]) * histo_width
/ max
]);
991 dump_zap_stats(objset_t
*os
, uint64_t object
)
996 error
= zap_get_stats(os
, object
, &zs
);
1000 if (zs
.zs_ptrtbl_len
== 0) {
1001 ASSERT(zs
.zs_num_blocks
== 1);
1002 (void) printf("\tmicrozap: %llu bytes, %llu entries\n",
1003 (u_longlong_t
)zs
.zs_blocksize
,
1004 (u_longlong_t
)zs
.zs_num_entries
);
1008 (void) printf("\tFat ZAP stats:\n");
1010 (void) printf("\t\tPointer table:\n");
1011 (void) printf("\t\t\t%llu elements\n",
1012 (u_longlong_t
)zs
.zs_ptrtbl_len
);
1013 (void) printf("\t\t\tzt_blk: %llu\n",
1014 (u_longlong_t
)zs
.zs_ptrtbl_zt_blk
);
1015 (void) printf("\t\t\tzt_numblks: %llu\n",
1016 (u_longlong_t
)zs
.zs_ptrtbl_zt_numblks
);
1017 (void) printf("\t\t\tzt_shift: %llu\n",
1018 (u_longlong_t
)zs
.zs_ptrtbl_zt_shift
);
1019 (void) printf("\t\t\tzt_blks_copied: %llu\n",
1020 (u_longlong_t
)zs
.zs_ptrtbl_blks_copied
);
1021 (void) printf("\t\t\tzt_nextblk: %llu\n",
1022 (u_longlong_t
)zs
.zs_ptrtbl_nextblk
);
1024 (void) printf("\t\tZAP entries: %llu\n",
1025 (u_longlong_t
)zs
.zs_num_entries
);
1026 (void) printf("\t\tLeaf blocks: %llu\n",
1027 (u_longlong_t
)zs
.zs_num_leafs
);
1028 (void) printf("\t\tTotal blocks: %llu\n",
1029 (u_longlong_t
)zs
.zs_num_blocks
);
1030 (void) printf("\t\tzap_block_type: 0x%llx\n",
1031 (u_longlong_t
)zs
.zs_block_type
);
1032 (void) printf("\t\tzap_magic: 0x%llx\n",
1033 (u_longlong_t
)zs
.zs_magic
);
1034 (void) printf("\t\tzap_salt: 0x%llx\n",
1035 (u_longlong_t
)zs
.zs_salt
);
1037 (void) printf("\t\tLeafs with 2^n pointers:\n");
1038 dump_histogram(zs
.zs_leafs_with_2n_pointers
, ZAP_HISTOGRAM_SIZE
, 0);
1040 (void) printf("\t\tBlocks with n*5 entries:\n");
1041 dump_histogram(zs
.zs_blocks_with_n5_entries
, ZAP_HISTOGRAM_SIZE
, 0);
1043 (void) printf("\t\tBlocks n/10 full:\n");
1044 dump_histogram(zs
.zs_blocks_n_tenths_full
, ZAP_HISTOGRAM_SIZE
, 0);
1046 (void) printf("\t\tEntries with n chunks:\n");
1047 dump_histogram(zs
.zs_entries_using_n_chunks
, ZAP_HISTOGRAM_SIZE
, 0);
1049 (void) printf("\t\tBuckets with n entries:\n");
1050 dump_histogram(zs
.zs_buckets_with_n_entries
, ZAP_HISTOGRAM_SIZE
, 0);
1055 dump_none(objset_t
*os
, uint64_t object
, void *data
, size_t size
)
1061 dump_unknown(objset_t
*os
, uint64_t object
, void *data
, size_t size
)
1063 (void) printf("\tUNKNOWN OBJECT TYPE\n");
1068 dump_uint8(objset_t
*os
, uint64_t object
, void *data
, size_t size
)
1074 dump_uint64(objset_t
*os
, uint64_t object
, void *data
, size_t size
)
1078 if (dump_opt
['d'] < 6)
1082 dmu_object_info_t doi
;
1084 VERIFY0(dmu_object_info(os
, object
, &doi
));
1085 size
= doi
.doi_max_offset
;
1087 * We cap the size at 1 mebibyte here to prevent
1088 * allocation failures and nigh-infinite printing if the
1089 * object is extremely large.
1091 oursize
= MIN(size
, 1 << 20);
1092 arr
= kmem_alloc(oursize
, KM_SLEEP
);
1094 int err
= dmu_read(os
, object
, 0, oursize
, arr
, 0);
1096 (void) printf("got error %u from dmu_read\n", err
);
1097 kmem_free(arr
, oursize
);
1102 * Even though the allocation is already done in this code path,
1103 * we still cap the size to prevent excessive printing.
1105 oursize
= MIN(size
, 1 << 20);
1110 (void) printf("\t\t[]\n");
1114 (void) printf("\t\t[%0llx", (u_longlong_t
)arr
[0]);
1115 for (size_t i
= 1; i
* sizeof (uint64_t) < oursize
; i
++) {
1117 (void) printf(", %0llx", (u_longlong_t
)arr
[i
]);
1119 (void) printf(",\n\t\t%0llx", (u_longlong_t
)arr
[i
]);
1121 if (oursize
!= size
)
1122 (void) printf(", ... ");
1123 (void) printf("]\n");
1126 kmem_free(arr
, oursize
);
1131 dump_zap(objset_t
*os
, uint64_t object
, void *data
, size_t size
)
1134 zap_attribute_t attr
;
1138 dump_zap_stats(os
, object
);
1139 (void) printf("\n");
1141 for (zap_cursor_init(&zc
, os
, object
);
1142 zap_cursor_retrieve(&zc
, &attr
) == 0;
1143 zap_cursor_advance(&zc
)) {
1144 (void) printf("\t\t%s = ", attr
.za_name
);
1145 if (attr
.za_num_integers
== 0) {
1146 (void) printf("\n");
1149 prop
= umem_zalloc(attr
.za_num_integers
*
1150 attr
.za_integer_length
, UMEM_NOFAIL
);
1151 (void) zap_lookup(os
, object
, attr
.za_name
,
1152 attr
.za_integer_length
, attr
.za_num_integers
, prop
);
1153 if (attr
.za_integer_length
== 1) {
1154 if (strcmp(attr
.za_name
,
1155 DSL_CRYPTO_KEY_MASTER_KEY
) == 0 ||
1156 strcmp(attr
.za_name
,
1157 DSL_CRYPTO_KEY_HMAC_KEY
) == 0 ||
1158 strcmp(attr
.za_name
, DSL_CRYPTO_KEY_IV
) == 0 ||
1159 strcmp(attr
.za_name
, DSL_CRYPTO_KEY_MAC
) == 0 ||
1160 strcmp(attr
.za_name
, DMU_POOL_CHECKSUM_SALT
) == 0) {
1163 for (i
= 0; i
< attr
.za_num_integers
; i
++) {
1164 (void) printf("%02x", u8
[i
]);
1167 (void) printf("%s", (char *)prop
);
1170 for (i
= 0; i
< attr
.za_num_integers
; i
++) {
1171 switch (attr
.za_integer_length
) {
1173 (void) printf("%u ",
1174 ((uint16_t *)prop
)[i
]);
1177 (void) printf("%u ",
1178 ((uint32_t *)prop
)[i
]);
1181 (void) printf("%lld ",
1182 (u_longlong_t
)((int64_t *)prop
)[i
]);
1187 (void) printf("\n");
1188 umem_free(prop
, attr
.za_num_integers
* attr
.za_integer_length
);
1190 zap_cursor_fini(&zc
);
1194 dump_bpobj(objset_t
*os
, uint64_t object
, void *data
, size_t size
)
1196 bpobj_phys_t
*bpop
= data
;
1198 char bytes
[32], comp
[32], uncomp
[32];
1200 /* make sure the output won't get truncated */
1201 CTASSERT(sizeof (bytes
) >= NN_NUMBUF_SZ
);
1202 CTASSERT(sizeof (comp
) >= NN_NUMBUF_SZ
);
1203 CTASSERT(sizeof (uncomp
) >= NN_NUMBUF_SZ
);
1208 zdb_nicenum(bpop
->bpo_bytes
, bytes
, sizeof (bytes
));
1209 zdb_nicenum(bpop
->bpo_comp
, comp
, sizeof (comp
));
1210 zdb_nicenum(bpop
->bpo_uncomp
, uncomp
, sizeof (uncomp
));
1212 (void) printf("\t\tnum_blkptrs = %llu\n",
1213 (u_longlong_t
)bpop
->bpo_num_blkptrs
);
1214 (void) printf("\t\tbytes = %s\n", bytes
);
1215 if (size
>= BPOBJ_SIZE_V1
) {
1216 (void) printf("\t\tcomp = %s\n", comp
);
1217 (void) printf("\t\tuncomp = %s\n", uncomp
);
1219 if (size
>= BPOBJ_SIZE_V2
) {
1220 (void) printf("\t\tsubobjs = %llu\n",
1221 (u_longlong_t
)bpop
->bpo_subobjs
);
1222 (void) printf("\t\tnum_subobjs = %llu\n",
1223 (u_longlong_t
)bpop
->bpo_num_subobjs
);
1225 if (size
>= sizeof (*bpop
)) {
1226 (void) printf("\t\tnum_freed = %llu\n",
1227 (u_longlong_t
)bpop
->bpo_num_freed
);
1230 if (dump_opt
['d'] < 5)
1233 for (i
= 0; i
< bpop
->bpo_num_blkptrs
; i
++) {
1234 char blkbuf
[BP_SPRINTF_LEN
];
1237 int err
= dmu_read(os
, object
,
1238 i
* sizeof (bp
), sizeof (bp
), &bp
, 0);
1240 (void) printf("got error %u from dmu_read\n", err
);
1243 snprintf_blkptr_compact(blkbuf
, sizeof (blkbuf
), &bp
,
1245 (void) printf("\t%s\n", blkbuf
);
1251 dump_bpobj_subobjs(objset_t
*os
, uint64_t object
, void *data
, size_t size
)
1253 dmu_object_info_t doi
;
1256 VERIFY0(dmu_object_info(os
, object
, &doi
));
1257 uint64_t *subobjs
= kmem_alloc(doi
.doi_max_offset
, KM_SLEEP
);
1259 int err
= dmu_read(os
, object
, 0, doi
.doi_max_offset
, subobjs
, 0);
1261 (void) printf("got error %u from dmu_read\n", err
);
1262 kmem_free(subobjs
, doi
.doi_max_offset
);
1266 int64_t last_nonzero
= -1;
1267 for (i
= 0; i
< doi
.doi_max_offset
/ 8; i
++) {
1268 if (subobjs
[i
] != 0)
1272 for (i
= 0; i
<= last_nonzero
; i
++) {
1273 (void) printf("\t%llu\n", (u_longlong_t
)subobjs
[i
]);
1275 kmem_free(subobjs
, doi
.doi_max_offset
);
1280 dump_ddt_zap(objset_t
*os
, uint64_t object
, void *data
, size_t size
)
1282 dump_zap_stats(os
, object
);
1283 /* contents are printed elsewhere, properly decoded */
1288 dump_sa_attrs(objset_t
*os
, uint64_t object
, void *data
, size_t size
)
1291 zap_attribute_t attr
;
1293 dump_zap_stats(os
, object
);
1294 (void) printf("\n");
1296 for (zap_cursor_init(&zc
, os
, object
);
1297 zap_cursor_retrieve(&zc
, &attr
) == 0;
1298 zap_cursor_advance(&zc
)) {
1299 (void) printf("\t\t%s = ", attr
.za_name
);
1300 if (attr
.za_num_integers
== 0) {
1301 (void) printf("\n");
1304 (void) printf(" %llx : [%d:%d:%d]\n",
1305 (u_longlong_t
)attr
.za_first_integer
,
1306 (int)ATTR_LENGTH(attr
.za_first_integer
),
1307 (int)ATTR_BSWAP(attr
.za_first_integer
),
1308 (int)ATTR_NUM(attr
.za_first_integer
));
1310 zap_cursor_fini(&zc
);
1315 dump_sa_layouts(objset_t
*os
, uint64_t object
, void *data
, size_t size
)
1318 zap_attribute_t attr
;
1319 uint16_t *layout_attrs
;
1322 dump_zap_stats(os
, object
);
1323 (void) printf("\n");
1325 for (zap_cursor_init(&zc
, os
, object
);
1326 zap_cursor_retrieve(&zc
, &attr
) == 0;
1327 zap_cursor_advance(&zc
)) {
1328 (void) printf("\t\t%s = [", attr
.za_name
);
1329 if (attr
.za_num_integers
== 0) {
1330 (void) printf("\n");
1334 VERIFY(attr
.za_integer_length
== 2);
1335 layout_attrs
= umem_zalloc(attr
.za_num_integers
*
1336 attr
.za_integer_length
, UMEM_NOFAIL
);
1338 VERIFY(zap_lookup(os
, object
, attr
.za_name
,
1339 attr
.za_integer_length
,
1340 attr
.za_num_integers
, layout_attrs
) == 0);
1342 for (i
= 0; i
!= attr
.za_num_integers
; i
++)
1343 (void) printf(" %d ", (int)layout_attrs
[i
]);
1344 (void) printf("]\n");
1345 umem_free(layout_attrs
,
1346 attr
.za_num_integers
* attr
.za_integer_length
);
1348 zap_cursor_fini(&zc
);
1353 dump_zpldir(objset_t
*os
, uint64_t object
, void *data
, size_t size
)
1356 zap_attribute_t attr
;
1357 const char *typenames
[] = {
1358 /* 0 */ "not specified",
1360 /* 2 */ "Character Device",
1361 /* 3 */ "3 (invalid)",
1362 /* 4 */ "Directory",
1363 /* 5 */ "5 (invalid)",
1364 /* 6 */ "Block Device",
1365 /* 7 */ "7 (invalid)",
1366 /* 8 */ "Regular File",
1367 /* 9 */ "9 (invalid)",
1368 /* 10 */ "Symbolic Link",
1369 /* 11 */ "11 (invalid)",
1372 /* 14 */ "Event Port",
1373 /* 15 */ "15 (invalid)",
1376 dump_zap_stats(os
, object
);
1377 (void) printf("\n");
1379 for (zap_cursor_init(&zc
, os
, object
);
1380 zap_cursor_retrieve(&zc
, &attr
) == 0;
1381 zap_cursor_advance(&zc
)) {
1382 (void) printf("\t\t%s = %lld (type: %s)\n",
1383 attr
.za_name
, ZFS_DIRENT_OBJ(attr
.za_first_integer
),
1384 typenames
[ZFS_DIRENT_TYPE(attr
.za_first_integer
)]);
1386 zap_cursor_fini(&zc
);
1390 get_dtl_refcount(vdev_t
*vd
)
1394 if (vd
->vdev_ops
->vdev_op_leaf
) {
1395 space_map_t
*sm
= vd
->vdev_dtl_sm
;
1398 sm
->sm_dbuf
->db_size
== sizeof (space_map_phys_t
))
1403 for (unsigned c
= 0; c
< vd
->vdev_children
; c
++)
1404 refcount
+= get_dtl_refcount(vd
->vdev_child
[c
]);
1409 get_metaslab_refcount(vdev_t
*vd
)
1413 if (vd
->vdev_top
== vd
) {
1414 for (uint64_t m
= 0; m
< vd
->vdev_ms_count
; m
++) {
1415 space_map_t
*sm
= vd
->vdev_ms
[m
]->ms_sm
;
1418 sm
->sm_dbuf
->db_size
== sizeof (space_map_phys_t
))
1422 for (unsigned c
= 0; c
< vd
->vdev_children
; c
++)
1423 refcount
+= get_metaslab_refcount(vd
->vdev_child
[c
]);
1429 get_obsolete_refcount(vdev_t
*vd
)
1431 uint64_t obsolete_sm_object
;
1434 VERIFY0(vdev_obsolete_sm_object(vd
, &obsolete_sm_object
));
1435 if (vd
->vdev_top
== vd
&& obsolete_sm_object
!= 0) {
1436 dmu_object_info_t doi
;
1437 VERIFY0(dmu_object_info(vd
->vdev_spa
->spa_meta_objset
,
1438 obsolete_sm_object
, &doi
));
1439 if (doi
.doi_bonus_size
== sizeof (space_map_phys_t
)) {
1443 ASSERT3P(vd
->vdev_obsolete_sm
, ==, NULL
);
1444 ASSERT3U(obsolete_sm_object
, ==, 0);
1446 for (unsigned c
= 0; c
< vd
->vdev_children
; c
++) {
1447 refcount
+= get_obsolete_refcount(vd
->vdev_child
[c
]);
1454 get_prev_obsolete_spacemap_refcount(spa_t
*spa
)
1457 spa
->spa_condensing_indirect_phys
.scip_prev_obsolete_sm_object
;
1458 if (prev_obj
!= 0) {
1459 dmu_object_info_t doi
;
1460 VERIFY0(dmu_object_info(spa
->spa_meta_objset
, prev_obj
, &doi
));
1461 if (doi
.doi_bonus_size
== sizeof (space_map_phys_t
)) {
1469 get_checkpoint_refcount(vdev_t
*vd
)
1473 if (vd
->vdev_top
== vd
&& vd
->vdev_top_zap
!= 0 &&
1474 zap_contains(spa_meta_objset(vd
->vdev_spa
),
1475 vd
->vdev_top_zap
, VDEV_TOP_ZAP_POOL_CHECKPOINT_SM
) == 0)
1478 for (uint64_t c
= 0; c
< vd
->vdev_children
; c
++)
1479 refcount
+= get_checkpoint_refcount(vd
->vdev_child
[c
]);
1485 get_log_spacemap_refcount(spa_t
*spa
)
1487 return (avl_numnodes(&spa
->spa_sm_logs_by_txg
));
1491 verify_spacemap_refcounts(spa_t
*spa
)
1493 uint64_t expected_refcount
= 0;
1494 uint64_t actual_refcount
;
1496 (void) feature_get_refcount(spa
,
1497 &spa_feature_table
[SPA_FEATURE_SPACEMAP_HISTOGRAM
],
1498 &expected_refcount
);
1499 actual_refcount
= get_dtl_refcount(spa
->spa_root_vdev
);
1500 actual_refcount
+= get_metaslab_refcount(spa
->spa_root_vdev
);
1501 actual_refcount
+= get_obsolete_refcount(spa
->spa_root_vdev
);
1502 actual_refcount
+= get_prev_obsolete_spacemap_refcount(spa
);
1503 actual_refcount
+= get_checkpoint_refcount(spa
->spa_root_vdev
);
1504 actual_refcount
+= get_log_spacemap_refcount(spa
);
1506 if (expected_refcount
!= actual_refcount
) {
1507 (void) printf("space map refcount mismatch: expected %lld != "
1509 (longlong_t
)expected_refcount
,
1510 (longlong_t
)actual_refcount
);
1517 dump_spacemap(objset_t
*os
, space_map_t
*sm
)
1519 const char *ddata
[] = { "ALLOC", "FREE", "CONDENSE", "INVALID",
1520 "INVALID", "INVALID", "INVALID", "INVALID" };
1525 (void) printf("space map object %llu:\n",
1526 (longlong_t
)sm
->sm_object
);
1527 (void) printf(" smp_length = 0x%llx\n",
1528 (longlong_t
)sm
->sm_phys
->smp_length
);
1529 (void) printf(" smp_alloc = 0x%llx\n",
1530 (longlong_t
)sm
->sm_phys
->smp_alloc
);
1532 if (dump_opt
['d'] < 6 && dump_opt
['m'] < 4)
1536 * Print out the freelist entries in both encoded and decoded form.
1538 uint8_t mapshift
= sm
->sm_shift
;
1540 uint64_t word
, entry_id
= 0;
1541 for (uint64_t offset
= 0; offset
< space_map_length(sm
);
1542 offset
+= sizeof (word
)) {
1544 VERIFY0(dmu_read(os
, space_map_object(sm
), offset
,
1545 sizeof (word
), &word
, DMU_READ_PREFETCH
));
1547 if (sm_entry_is_debug(word
)) {
1548 uint64_t de_txg
= SM_DEBUG_TXG_DECODE(word
);
1549 uint64_t de_sync_pass
= SM_DEBUG_SYNCPASS_DECODE(word
);
1552 "\t [%6llu] PADDING\n",
1553 (u_longlong_t
)entry_id
);
1556 "\t [%6llu] %s: txg %llu pass %llu\n",
1557 (u_longlong_t
)entry_id
,
1558 ddata
[SM_DEBUG_ACTION_DECODE(word
)],
1559 (u_longlong_t
)de_txg
,
1560 (u_longlong_t
)de_sync_pass
);
1568 uint64_t entry_off
, entry_run
, entry_vdev
= SM_NO_VDEVID
;
1570 if (sm_entry_is_single_word(word
)) {
1571 entry_type
= (SM_TYPE_DECODE(word
) == SM_ALLOC
) ?
1573 entry_off
= (SM_OFFSET_DECODE(word
) << mapshift
) +
1575 entry_run
= SM_RUN_DECODE(word
) << mapshift
;
1578 /* it is a two-word entry so we read another word */
1579 ASSERT(sm_entry_is_double_word(word
));
1581 uint64_t extra_word
;
1582 offset
+= sizeof (extra_word
);
1583 VERIFY0(dmu_read(os
, space_map_object(sm
), offset
,
1584 sizeof (extra_word
), &extra_word
,
1585 DMU_READ_PREFETCH
));
1587 ASSERT3U(offset
, <=, space_map_length(sm
));
1589 entry_run
= SM2_RUN_DECODE(word
) << mapshift
;
1590 entry_vdev
= SM2_VDEV_DECODE(word
);
1591 entry_type
= (SM2_TYPE_DECODE(extra_word
) == SM_ALLOC
) ?
1593 entry_off
= (SM2_OFFSET_DECODE(extra_word
) <<
1594 mapshift
) + sm
->sm_start
;
1598 (void) printf("\t [%6llu] %c range:"
1599 " %010llx-%010llx size: %06llx vdev: %06llu words: %u\n",
1600 (u_longlong_t
)entry_id
,
1601 entry_type
, (u_longlong_t
)entry_off
,
1602 (u_longlong_t
)(entry_off
+ entry_run
),
1603 (u_longlong_t
)entry_run
,
1604 (u_longlong_t
)entry_vdev
, words
);
1606 if (entry_type
== 'A')
1612 if (alloc
!= space_map_allocated(sm
)) {
1613 (void) printf("space_map_object alloc (%lld) INCONSISTENT "
1614 "with space map summary (%lld)\n",
1615 (longlong_t
)space_map_allocated(sm
), (longlong_t
)alloc
);
1620 dump_metaslab_stats(metaslab_t
*msp
)
1623 range_tree_t
*rt
= msp
->ms_allocatable
;
1624 zfs_btree_t
*t
= &msp
->ms_allocatable_by_size
;
1625 int free_pct
= range_tree_space(rt
) * 100 / msp
->ms_size
;
1627 /* max sure nicenum has enough space */
1628 CTASSERT(sizeof (maxbuf
) >= NN_NUMBUF_SZ
);
1630 zdb_nicenum(metaslab_largest_allocatable(msp
), maxbuf
, sizeof (maxbuf
));
1632 (void) printf("\t %25s %10lu %7s %6s %4s %4d%%\n",
1633 "segments", zfs_btree_numnodes(t
), "maxsize", maxbuf
,
1634 "freepct", free_pct
);
1635 (void) printf("\tIn-memory histogram:\n");
1636 dump_histogram(rt
->rt_histogram
, RANGE_TREE_HISTOGRAM_SIZE
, 0);
1640 dump_metaslab(metaslab_t
*msp
)
1642 vdev_t
*vd
= msp
->ms_group
->mg_vd
;
1643 spa_t
*spa
= vd
->vdev_spa
;
1644 space_map_t
*sm
= msp
->ms_sm
;
1647 zdb_nicenum(msp
->ms_size
- space_map_allocated(sm
), freebuf
,
1651 "\tmetaslab %6llu offset %12llx spacemap %6llu free %5s\n",
1652 (u_longlong_t
)msp
->ms_id
, (u_longlong_t
)msp
->ms_start
,
1653 (u_longlong_t
)space_map_object(sm
), freebuf
);
1655 if (dump_opt
['m'] > 2 && !dump_opt
['L']) {
1656 mutex_enter(&msp
->ms_lock
);
1657 VERIFY0(metaslab_load(msp
));
1658 range_tree_stat_verify(msp
->ms_allocatable
);
1659 dump_metaslab_stats(msp
);
1660 metaslab_unload(msp
);
1661 mutex_exit(&msp
->ms_lock
);
1664 if (dump_opt
['m'] > 1 && sm
!= NULL
&&
1665 spa_feature_is_active(spa
, SPA_FEATURE_SPACEMAP_HISTOGRAM
)) {
1667 * The space map histogram represents free space in chunks
1668 * of sm_shift (i.e. bucket 0 refers to 2^sm_shift).
1670 (void) printf("\tOn-disk histogram:\t\tfragmentation %llu\n",
1671 (u_longlong_t
)msp
->ms_fragmentation
);
1672 dump_histogram(sm
->sm_phys
->smp_histogram
,
1673 SPACE_MAP_HISTOGRAM_SIZE
, sm
->sm_shift
);
1676 if (vd
->vdev_ops
== &vdev_draid_ops
)
1677 ASSERT3U(msp
->ms_size
, <=, 1ULL << vd
->vdev_ms_shift
);
1679 ASSERT3U(msp
->ms_size
, ==, 1ULL << vd
->vdev_ms_shift
);
1681 dump_spacemap(spa
->spa_meta_objset
, msp
->ms_sm
);
1683 if (spa_feature_is_active(spa
, SPA_FEATURE_LOG_SPACEMAP
)) {
1684 (void) printf("\tFlush data:\n\tunflushed txg=%llu\n\n",
1685 (u_longlong_t
)metaslab_unflushed_txg(msp
));
1690 print_vdev_metaslab_header(vdev_t
*vd
)
1692 vdev_alloc_bias_t alloc_bias
= vd
->vdev_alloc_bias
;
1693 const char *bias_str
= "";
1694 if (alloc_bias
== VDEV_BIAS_LOG
|| vd
->vdev_islog
) {
1695 bias_str
= VDEV_ALLOC_BIAS_LOG
;
1696 } else if (alloc_bias
== VDEV_BIAS_SPECIAL
) {
1697 bias_str
= VDEV_ALLOC_BIAS_SPECIAL
;
1698 } else if (alloc_bias
== VDEV_BIAS_DEDUP
) {
1699 bias_str
= VDEV_ALLOC_BIAS_DEDUP
;
1702 uint64_t ms_flush_data_obj
= 0;
1703 if (vd
->vdev_top_zap
!= 0) {
1704 int error
= zap_lookup(spa_meta_objset(vd
->vdev_spa
),
1705 vd
->vdev_top_zap
, VDEV_TOP_ZAP_MS_UNFLUSHED_PHYS_TXGS
,
1706 sizeof (uint64_t), 1, &ms_flush_data_obj
);
1707 if (error
!= ENOENT
) {
1712 (void) printf("\tvdev %10llu %s",
1713 (u_longlong_t
)vd
->vdev_id
, bias_str
);
1715 if (ms_flush_data_obj
!= 0) {
1716 (void) printf(" ms_unflushed_phys object %llu",
1717 (u_longlong_t
)ms_flush_data_obj
);
1720 (void) printf("\n\t%-10s%5llu %-19s %-15s %-12s\n",
1721 "metaslabs", (u_longlong_t
)vd
->vdev_ms_count
,
1722 "offset", "spacemap", "free");
1723 (void) printf("\t%15s %19s %15s %12s\n",
1724 "---------------", "-------------------",
1725 "---------------", "------------");
1729 dump_metaslab_groups(spa_t
*spa
)
1731 vdev_t
*rvd
= spa
->spa_root_vdev
;
1732 metaslab_class_t
*mc
= spa_normal_class(spa
);
1733 uint64_t fragmentation
;
1735 metaslab_class_histogram_verify(mc
);
1737 for (unsigned c
= 0; c
< rvd
->vdev_children
; c
++) {
1738 vdev_t
*tvd
= rvd
->vdev_child
[c
];
1739 metaslab_group_t
*mg
= tvd
->vdev_mg
;
1741 if (mg
== NULL
|| mg
->mg_class
!= mc
)
1744 metaslab_group_histogram_verify(mg
);
1745 mg
->mg_fragmentation
= metaslab_group_fragmentation(mg
);
1747 (void) printf("\tvdev %10llu\t\tmetaslabs%5llu\t\t"
1749 (u_longlong_t
)tvd
->vdev_id
,
1750 (u_longlong_t
)tvd
->vdev_ms_count
);
1751 if (mg
->mg_fragmentation
== ZFS_FRAG_INVALID
) {
1752 (void) printf("%3s\n", "-");
1754 (void) printf("%3llu%%\n",
1755 (u_longlong_t
)mg
->mg_fragmentation
);
1757 dump_histogram(mg
->mg_histogram
, RANGE_TREE_HISTOGRAM_SIZE
, 0);
1760 (void) printf("\tpool %s\tfragmentation", spa_name(spa
));
1761 fragmentation
= metaslab_class_fragmentation(mc
);
1762 if (fragmentation
== ZFS_FRAG_INVALID
)
1763 (void) printf("\t%3s\n", "-");
1765 (void) printf("\t%3llu%%\n", (u_longlong_t
)fragmentation
);
1766 dump_histogram(mc
->mc_histogram
, RANGE_TREE_HISTOGRAM_SIZE
, 0);
1770 print_vdev_indirect(vdev_t
*vd
)
1772 vdev_indirect_config_t
*vic
= &vd
->vdev_indirect_config
;
1773 vdev_indirect_mapping_t
*vim
= vd
->vdev_indirect_mapping
;
1774 vdev_indirect_births_t
*vib
= vd
->vdev_indirect_births
;
1777 ASSERT3P(vib
, ==, NULL
);
1781 ASSERT3U(vdev_indirect_mapping_object(vim
), ==,
1782 vic
->vic_mapping_object
);
1783 ASSERT3U(vdev_indirect_births_object(vib
), ==,
1784 vic
->vic_births_object
);
1786 (void) printf("indirect births obj %llu:\n",
1787 (longlong_t
)vic
->vic_births_object
);
1788 (void) printf(" vib_count = %llu\n",
1789 (longlong_t
)vdev_indirect_births_count(vib
));
1790 for (uint64_t i
= 0; i
< vdev_indirect_births_count(vib
); i
++) {
1791 vdev_indirect_birth_entry_phys_t
*cur_vibe
=
1792 &vib
->vib_entries
[i
];
1793 (void) printf("\toffset %llx -> txg %llu\n",
1794 (longlong_t
)cur_vibe
->vibe_offset
,
1795 (longlong_t
)cur_vibe
->vibe_phys_birth_txg
);
1797 (void) printf("\n");
1799 (void) printf("indirect mapping obj %llu:\n",
1800 (longlong_t
)vic
->vic_mapping_object
);
1801 (void) printf(" vim_max_offset = 0x%llx\n",
1802 (longlong_t
)vdev_indirect_mapping_max_offset(vim
));
1803 (void) printf(" vim_bytes_mapped = 0x%llx\n",
1804 (longlong_t
)vdev_indirect_mapping_bytes_mapped(vim
));
1805 (void) printf(" vim_count = %llu\n",
1806 (longlong_t
)vdev_indirect_mapping_num_entries(vim
));
1808 if (dump_opt
['d'] <= 5 && dump_opt
['m'] <= 3)
1811 uint32_t *counts
= vdev_indirect_mapping_load_obsolete_counts(vim
);
1813 for (uint64_t i
= 0; i
< vdev_indirect_mapping_num_entries(vim
); i
++) {
1814 vdev_indirect_mapping_entry_phys_t
*vimep
=
1815 &vim
->vim_entries
[i
];
1816 (void) printf("\t<%llx:%llx:%llx> -> "
1817 "<%llx:%llx:%llx> (%x obsolete)\n",
1818 (longlong_t
)vd
->vdev_id
,
1819 (longlong_t
)DVA_MAPPING_GET_SRC_OFFSET(vimep
),
1820 (longlong_t
)DVA_GET_ASIZE(&vimep
->vimep_dst
),
1821 (longlong_t
)DVA_GET_VDEV(&vimep
->vimep_dst
),
1822 (longlong_t
)DVA_GET_OFFSET(&vimep
->vimep_dst
),
1823 (longlong_t
)DVA_GET_ASIZE(&vimep
->vimep_dst
),
1826 (void) printf("\n");
1828 uint64_t obsolete_sm_object
;
1829 VERIFY0(vdev_obsolete_sm_object(vd
, &obsolete_sm_object
));
1830 if (obsolete_sm_object
!= 0) {
1831 objset_t
*mos
= vd
->vdev_spa
->spa_meta_objset
;
1832 (void) printf("obsolete space map object %llu:\n",
1833 (u_longlong_t
)obsolete_sm_object
);
1834 ASSERT(vd
->vdev_obsolete_sm
!= NULL
);
1835 ASSERT3U(space_map_object(vd
->vdev_obsolete_sm
), ==,
1836 obsolete_sm_object
);
1837 dump_spacemap(mos
, vd
->vdev_obsolete_sm
);
1838 (void) printf("\n");
1843 dump_metaslabs(spa_t
*spa
)
1845 vdev_t
*vd
, *rvd
= spa
->spa_root_vdev
;
1846 uint64_t m
, c
= 0, children
= rvd
->vdev_children
;
1848 (void) printf("\nMetaslabs:\n");
1850 if (!dump_opt
['d'] && zopt_metaslab_args
> 0) {
1851 c
= zopt_metaslab
[0];
1854 (void) fatal("bad vdev id: %llu", (u_longlong_t
)c
);
1856 if (zopt_metaslab_args
> 1) {
1857 vd
= rvd
->vdev_child
[c
];
1858 print_vdev_metaslab_header(vd
);
1860 for (m
= 1; m
< zopt_metaslab_args
; m
++) {
1861 if (zopt_metaslab
[m
] < vd
->vdev_ms_count
)
1863 vd
->vdev_ms
[zopt_metaslab
[m
]]);
1865 (void) fprintf(stderr
, "bad metaslab "
1867 (u_longlong_t
)zopt_metaslab
[m
]);
1869 (void) printf("\n");
1874 for (; c
< children
; c
++) {
1875 vd
= rvd
->vdev_child
[c
];
1876 print_vdev_metaslab_header(vd
);
1878 print_vdev_indirect(vd
);
1880 for (m
= 0; m
< vd
->vdev_ms_count
; m
++)
1881 dump_metaslab(vd
->vdev_ms
[m
]);
1882 (void) printf("\n");
1887 dump_log_spacemaps(spa_t
*spa
)
1889 if (!spa_feature_is_active(spa
, SPA_FEATURE_LOG_SPACEMAP
))
1892 (void) printf("\nLog Space Maps in Pool:\n");
1893 for (spa_log_sm_t
*sls
= avl_first(&spa
->spa_sm_logs_by_txg
);
1894 sls
; sls
= AVL_NEXT(&spa
->spa_sm_logs_by_txg
, sls
)) {
1895 space_map_t
*sm
= NULL
;
1896 VERIFY0(space_map_open(&sm
, spa_meta_objset(spa
),
1897 sls
->sls_sm_obj
, 0, UINT64_MAX
, SPA_MINBLOCKSHIFT
));
1899 (void) printf("Log Spacemap object %llu txg %llu\n",
1900 (u_longlong_t
)sls
->sls_sm_obj
, (u_longlong_t
)sls
->sls_txg
);
1901 dump_spacemap(spa
->spa_meta_objset
, sm
);
1902 space_map_close(sm
);
1904 (void) printf("\n");
1908 dump_dde(const ddt_t
*ddt
, const ddt_entry_t
*dde
, uint64_t index
)
1910 const ddt_phys_t
*ddp
= dde
->dde_phys
;
1911 const ddt_key_t
*ddk
= &dde
->dde_key
;
1912 const char *types
[4] = { "ditto", "single", "double", "triple" };
1913 char blkbuf
[BP_SPRINTF_LEN
];
1917 for (p
= 0; p
< DDT_PHYS_TYPES
; p
++, ddp
++) {
1918 if (ddp
->ddp_phys_birth
== 0)
1920 ddt_bp_create(ddt
->ddt_checksum
, ddk
, ddp
, &blk
);
1921 snprintf_blkptr(blkbuf
, sizeof (blkbuf
), &blk
);
1922 (void) printf("index %llx refcnt %llu %s %s\n",
1923 (u_longlong_t
)index
, (u_longlong_t
)ddp
->ddp_refcnt
,
1929 dump_dedup_ratio(const ddt_stat_t
*dds
)
1931 double rL
, rP
, rD
, D
, dedup
, compress
, copies
;
1933 if (dds
->dds_blocks
== 0)
1936 rL
= (double)dds
->dds_ref_lsize
;
1937 rP
= (double)dds
->dds_ref_psize
;
1938 rD
= (double)dds
->dds_ref_dsize
;
1939 D
= (double)dds
->dds_dsize
;
1945 (void) printf("dedup = %.2f, compress = %.2f, copies = %.2f, "
1946 "dedup * compress / copies = %.2f\n\n",
1947 dedup
, compress
, copies
, dedup
* compress
/ copies
);
1951 dump_ddt(ddt_t
*ddt
, enum ddt_type type
, enum ddt_class
class)
1953 char name
[DDT_NAMELEN
];
1956 dmu_object_info_t doi
;
1957 uint64_t count
, dspace
, mspace
;
1960 error
= ddt_object_info(ddt
, type
, class, &doi
);
1962 if (error
== ENOENT
)
1966 error
= ddt_object_count(ddt
, type
, class, &count
);
1971 dspace
= doi
.doi_physical_blocks_512
<< 9;
1972 mspace
= doi
.doi_fill_count
* doi
.doi_data_block_size
;
1974 ddt_object_name(ddt
, type
, class, name
);
1976 (void) printf("%s: %llu entries, size %llu on disk, %llu in core\n",
1978 (u_longlong_t
)count
,
1979 (u_longlong_t
)(dspace
/ count
),
1980 (u_longlong_t
)(mspace
/ count
));
1982 if (dump_opt
['D'] < 3)
1985 zpool_dump_ddt(NULL
, &ddt
->ddt_histogram
[type
][class]);
1987 if (dump_opt
['D'] < 4)
1990 if (dump_opt
['D'] < 5 && class == DDT_CLASS_UNIQUE
)
1993 (void) printf("%s contents:\n\n", name
);
1995 while ((error
= ddt_object_walk(ddt
, type
, class, &walk
, &dde
)) == 0)
1996 dump_dde(ddt
, &dde
, walk
);
1998 ASSERT3U(error
, ==, ENOENT
);
2000 (void) printf("\n");
2004 dump_all_ddts(spa_t
*spa
)
2006 ddt_histogram_t ddh_total
;
2007 ddt_stat_t dds_total
;
2009 bzero(&ddh_total
, sizeof (ddh_total
));
2010 bzero(&dds_total
, sizeof (dds_total
));
2012 for (enum zio_checksum c
= 0; c
< ZIO_CHECKSUM_FUNCTIONS
; c
++) {
2013 ddt_t
*ddt
= spa
->spa_ddt
[c
];
2014 for (enum ddt_type type
= 0; type
< DDT_TYPES
; type
++) {
2015 for (enum ddt_class
class = 0; class < DDT_CLASSES
;
2017 dump_ddt(ddt
, type
, class);
2022 ddt_get_dedup_stats(spa
, &dds_total
);
2024 if (dds_total
.dds_blocks
== 0) {
2025 (void) printf("All DDTs are empty\n");
2029 (void) printf("\n");
2031 if (dump_opt
['D'] > 1) {
2032 (void) printf("DDT histogram (aggregated over all DDTs):\n");
2033 ddt_get_dedup_histogram(spa
, &ddh_total
);
2034 zpool_dump_ddt(&dds_total
, &ddh_total
);
2037 dump_dedup_ratio(&dds_total
);
2041 dump_dtl_seg(void *arg
, uint64_t start
, uint64_t size
)
2045 (void) printf("%s [%llu,%llu) length %llu\n",
2047 (u_longlong_t
)start
,
2048 (u_longlong_t
)(start
+ size
),
2049 (u_longlong_t
)(size
));
2053 dump_dtl(vdev_t
*vd
, int indent
)
2055 spa_t
*spa
= vd
->vdev_spa
;
2057 const char *name
[DTL_TYPES
] = { "missing", "partial", "scrub",
2061 spa_vdev_state_enter(spa
, SCL_NONE
);
2062 required
= vdev_dtl_required(vd
);
2063 (void) spa_vdev_state_exit(spa
, NULL
, 0);
2066 (void) printf("\nDirty time logs:\n\n");
2068 (void) printf("\t%*s%s [%s]\n", indent
, "",
2069 vd
->vdev_path
? vd
->vdev_path
:
2070 vd
->vdev_parent
? vd
->vdev_ops
->vdev_op_type
: spa_name(spa
),
2071 required
? "DTL-required" : "DTL-expendable");
2073 for (int t
= 0; t
< DTL_TYPES
; t
++) {
2074 range_tree_t
*rt
= vd
->vdev_dtl
[t
];
2075 if (range_tree_space(rt
) == 0)
2077 (void) snprintf(prefix
, sizeof (prefix
), "\t%*s%s",
2078 indent
+ 2, "", name
[t
]);
2079 range_tree_walk(rt
, dump_dtl_seg
, prefix
);
2080 if (dump_opt
['d'] > 5 && vd
->vdev_children
== 0)
2081 dump_spacemap(spa
->spa_meta_objset
,
2085 for (unsigned c
= 0; c
< vd
->vdev_children
; c
++)
2086 dump_dtl(vd
->vdev_child
[c
], indent
+ 4);
2090 dump_history(spa_t
*spa
)
2092 nvlist_t
**events
= NULL
;
2094 uint64_t resid
, len
, off
= 0;
2099 if ((buf
= malloc(SPA_OLD_MAXBLOCKSIZE
)) == NULL
) {
2100 (void) fprintf(stderr
, "%s: unable to allocate I/O buffer\n",
2106 len
= SPA_OLD_MAXBLOCKSIZE
;
2108 if ((error
= spa_history_get(spa
, &off
, &len
, buf
)) != 0) {
2109 (void) fprintf(stderr
, "Unable to read history: "
2110 "error %d\n", error
);
2115 if (zpool_history_unpack(buf
, len
, &resid
, &events
, &num
) != 0)
2121 (void) printf("\nHistory:\n");
2122 for (unsigned i
= 0; i
< num
; i
++) {
2123 boolean_t printed
= B_FALSE
;
2125 if (nvlist_exists(events
[i
], ZPOOL_HIST_TIME
)) {
2129 tsec
= fnvlist_lookup_uint64(events
[i
],
2131 (void) localtime_r(&tsec
, &t
);
2132 (void) strftime(tbuf
, sizeof (tbuf
), "%F.%T", &t
);
2137 if (nvlist_exists(events
[i
], ZPOOL_HIST_CMD
)) {
2138 (void) printf("%s %s\n", tbuf
,
2139 fnvlist_lookup_string(events
[i
], ZPOOL_HIST_CMD
));
2140 } else if (nvlist_exists(events
[i
], ZPOOL_HIST_INT_EVENT
)) {
2143 ievent
= fnvlist_lookup_uint64(events
[i
],
2144 ZPOOL_HIST_INT_EVENT
);
2145 if (ievent
>= ZFS_NUM_LEGACY_HISTORY_EVENTS
)
2148 (void) printf(" %s [internal %s txg:%ju] %s\n",
2150 zfs_history_event_names
[ievent
],
2151 fnvlist_lookup_uint64(events
[i
],
2153 fnvlist_lookup_string(events
[i
],
2154 ZPOOL_HIST_INT_STR
));
2155 } else if (nvlist_exists(events
[i
], ZPOOL_HIST_INT_NAME
)) {
2156 (void) printf("%s [txg:%ju] %s", tbuf
,
2157 fnvlist_lookup_uint64(events
[i
],
2159 fnvlist_lookup_string(events
[i
],
2160 ZPOOL_HIST_INT_NAME
));
2162 if (nvlist_exists(events
[i
], ZPOOL_HIST_DSNAME
)) {
2163 (void) printf(" %s (%llu)",
2164 fnvlist_lookup_string(events
[i
],
2166 (u_longlong_t
)fnvlist_lookup_uint64(
2171 (void) printf(" %s\n", fnvlist_lookup_string(events
[i
],
2172 ZPOOL_HIST_INT_STR
));
2173 } else if (nvlist_exists(events
[i
], ZPOOL_HIST_IOCTL
)) {
2174 (void) printf("%s ioctl %s\n", tbuf
,
2175 fnvlist_lookup_string(events
[i
],
2178 if (nvlist_exists(events
[i
], ZPOOL_HIST_INPUT_NVL
)) {
2179 (void) printf(" input:\n");
2180 dump_nvlist(fnvlist_lookup_nvlist(events
[i
],
2181 ZPOOL_HIST_INPUT_NVL
), 8);
2183 if (nvlist_exists(events
[i
], ZPOOL_HIST_OUTPUT_NVL
)) {
2184 (void) printf(" output:\n");
2185 dump_nvlist(fnvlist_lookup_nvlist(events
[i
],
2186 ZPOOL_HIST_OUTPUT_NVL
), 8);
2188 if (nvlist_exists(events
[i
], ZPOOL_HIST_ERRNO
)) {
2189 (void) printf(" errno: %lld\n",
2190 (longlong_t
)fnvlist_lookup_int64(events
[i
],
2199 if (dump_opt
['h'] > 1) {
2201 (void) printf("unrecognized record:\n");
2202 dump_nvlist(events
[i
], 2);
2210 dump_dnode(objset_t
*os
, uint64_t object
, void *data
, size_t size
)
2215 blkid2offset(const dnode_phys_t
*dnp
, const blkptr_t
*bp
,
2216 const zbookmark_phys_t
*zb
)
2219 ASSERT(zb
->zb_level
< 0);
2220 if (zb
->zb_object
== 0)
2221 return (zb
->zb_blkid
);
2222 return (zb
->zb_blkid
* BP_GET_LSIZE(bp
));
2225 ASSERT(zb
->zb_level
>= 0);
2227 return ((zb
->zb_blkid
<<
2228 (zb
->zb_level
* (dnp
->dn_indblkshift
- SPA_BLKPTRSHIFT
))) *
2229 dnp
->dn_datablkszsec
<< SPA_MINBLOCKSHIFT
);
2233 snprintf_zstd_header(spa_t
*spa
, char *blkbuf
, size_t buflen
,
2239 zfs_zstdhdr_t zstd_hdr
;
2242 if (BP_GET_COMPRESS(bp
) != ZIO_COMPRESS_ZSTD
)
2248 if (BP_IS_EMBEDDED(bp
)) {
2249 buf
= malloc(SPA_MAXBLOCKSIZE
);
2251 (void) fprintf(stderr
, "out of memory\n");
2254 decode_embedded_bp_compressed(bp
, buf
);
2255 memcpy(&zstd_hdr
, buf
, sizeof (zstd_hdr
));
2257 zstd_hdr
.c_len
= BE_32(zstd_hdr
.c_len
);
2258 zstd_hdr
.raw_version_level
= BE_32(zstd_hdr
.raw_version_level
);
2259 (void) snprintf(blkbuf
+ strlen(blkbuf
),
2260 buflen
- strlen(blkbuf
),
2261 " ZSTD:size=%u:version=%u:level=%u:EMBEDDED",
2262 zstd_hdr
.c_len
, zfs_get_hdrversion(&zstd_hdr
),
2263 zfs_get_hdrlevel(&zstd_hdr
));
2267 pabd
= abd_alloc_for_io(SPA_MAXBLOCKSIZE
, B_FALSE
);
2268 zio
= zio_root(spa
, NULL
, NULL
, 0);
2270 /* Decrypt but don't decompress so we can read the compression header */
2271 zio_nowait(zio_read(zio
, spa
, bp
, pabd
, BP_GET_PSIZE(bp
), NULL
, NULL
,
2272 ZIO_PRIORITY_SYNC_READ
, ZIO_FLAG_CANFAIL
| ZIO_FLAG_RAW_COMPRESS
,
2274 error
= zio_wait(zio
);
2276 (void) fprintf(stderr
, "read failed: %d\n", error
);
2279 buf
= abd_borrow_buf_copy(pabd
, BP_GET_LSIZE(bp
));
2280 memcpy(&zstd_hdr
, buf
, sizeof (zstd_hdr
));
2281 zstd_hdr
.c_len
= BE_32(zstd_hdr
.c_len
);
2282 zstd_hdr
.raw_version_level
= BE_32(zstd_hdr
.raw_version_level
);
2284 (void) snprintf(blkbuf
+ strlen(blkbuf
),
2285 buflen
- strlen(blkbuf
),
2286 " ZSTD:size=%u:version=%u:level=%u:NORMAL",
2287 zstd_hdr
.c_len
, zfs_get_hdrversion(&zstd_hdr
),
2288 zfs_get_hdrlevel(&zstd_hdr
));
2290 abd_return_buf_copy(pabd
, buf
, BP_GET_LSIZE(bp
));
2294 snprintf_blkptr_compact(char *blkbuf
, size_t buflen
, const blkptr_t
*bp
,
2297 const dva_t
*dva
= bp
->blk_dva
;
2298 int ndvas
= dump_opt
['d'] > 5 ? BP_GET_NDVAS(bp
) : 1;
2301 if (dump_opt
['b'] >= 6) {
2302 snprintf_blkptr(blkbuf
, buflen
, bp
);
2304 (void) snprintf(blkbuf
+ strlen(blkbuf
),
2305 buflen
- strlen(blkbuf
), " %s", "FREE");
2310 if (BP_IS_EMBEDDED(bp
)) {
2311 (void) sprintf(blkbuf
,
2312 "EMBEDDED et=%u %llxL/%llxP B=%llu",
2313 (int)BPE_GET_ETYPE(bp
),
2314 (u_longlong_t
)BPE_GET_LSIZE(bp
),
2315 (u_longlong_t
)BPE_GET_PSIZE(bp
),
2316 (u_longlong_t
)bp
->blk_birth
);
2322 for (i
= 0; i
< ndvas
; i
++)
2323 (void) snprintf(blkbuf
+ strlen(blkbuf
),
2324 buflen
- strlen(blkbuf
), "%llu:%llx:%llx ",
2325 (u_longlong_t
)DVA_GET_VDEV(&dva
[i
]),
2326 (u_longlong_t
)DVA_GET_OFFSET(&dva
[i
]),
2327 (u_longlong_t
)DVA_GET_ASIZE(&dva
[i
]));
2329 if (BP_IS_HOLE(bp
)) {
2330 (void) snprintf(blkbuf
+ strlen(blkbuf
),
2331 buflen
- strlen(blkbuf
),
2333 (u_longlong_t
)BP_GET_LSIZE(bp
),
2334 (u_longlong_t
)bp
->blk_birth
);
2336 (void) snprintf(blkbuf
+ strlen(blkbuf
),
2337 buflen
- strlen(blkbuf
),
2338 "%llxL/%llxP F=%llu B=%llu/%llu",
2339 (u_longlong_t
)BP_GET_LSIZE(bp
),
2340 (u_longlong_t
)BP_GET_PSIZE(bp
),
2341 (u_longlong_t
)BP_GET_FILL(bp
),
2342 (u_longlong_t
)bp
->blk_birth
,
2343 (u_longlong_t
)BP_PHYSICAL_BIRTH(bp
));
2345 (void) snprintf(blkbuf
+ strlen(blkbuf
),
2346 buflen
- strlen(blkbuf
), " %s", "FREE");
2347 (void) snprintf(blkbuf
+ strlen(blkbuf
),
2348 buflen
- strlen(blkbuf
), " cksum=%llx:%llx:%llx:%llx",
2349 (u_longlong_t
)bp
->blk_cksum
.zc_word
[0],
2350 (u_longlong_t
)bp
->blk_cksum
.zc_word
[1],
2351 (u_longlong_t
)bp
->blk_cksum
.zc_word
[2],
2352 (u_longlong_t
)bp
->blk_cksum
.zc_word
[3]);
2357 print_indirect(spa_t
*spa
, blkptr_t
*bp
, const zbookmark_phys_t
*zb
,
2358 const dnode_phys_t
*dnp
)
2360 char blkbuf
[BP_SPRINTF_LEN
];
2363 if (!BP_IS_EMBEDDED(bp
)) {
2364 ASSERT3U(BP_GET_TYPE(bp
), ==, dnp
->dn_type
);
2365 ASSERT3U(BP_GET_LEVEL(bp
), ==, zb
->zb_level
);
2368 (void) printf("%16llx ", (u_longlong_t
)blkid2offset(dnp
, bp
, zb
));
2370 ASSERT(zb
->zb_level
>= 0);
2372 for (l
= dnp
->dn_nlevels
- 1; l
>= -1; l
--) {
2373 if (l
== zb
->zb_level
) {
2374 (void) printf("L%llx", (u_longlong_t
)zb
->zb_level
);
2380 snprintf_blkptr_compact(blkbuf
, sizeof (blkbuf
), bp
, B_FALSE
);
2381 if (dump_opt
['Z'] && BP_GET_COMPRESS(bp
) == ZIO_COMPRESS_ZSTD
)
2382 snprintf_zstd_header(spa
, blkbuf
, sizeof (blkbuf
), bp
);
2383 (void) printf("%s\n", blkbuf
);
2387 visit_indirect(spa_t
*spa
, const dnode_phys_t
*dnp
,
2388 blkptr_t
*bp
, const zbookmark_phys_t
*zb
)
2392 if (bp
->blk_birth
== 0)
2395 print_indirect(spa
, bp
, zb
, dnp
);
2397 if (BP_GET_LEVEL(bp
) > 0 && !BP_IS_HOLE(bp
)) {
2398 arc_flags_t flags
= ARC_FLAG_WAIT
;
2401 int epb
= BP_GET_LSIZE(bp
) >> SPA_BLKPTRSHIFT
;
2404 ASSERT(!BP_IS_REDACTED(bp
));
2406 err
= arc_read(NULL
, spa
, bp
, arc_getbuf_func
, &buf
,
2407 ZIO_PRIORITY_ASYNC_READ
, ZIO_FLAG_CANFAIL
, &flags
, zb
);
2410 ASSERT(buf
->b_data
);
2412 /* recursively visit blocks below this */
2414 for (i
= 0; i
< epb
; i
++, cbp
++) {
2415 zbookmark_phys_t czb
;
2417 SET_BOOKMARK(&czb
, zb
->zb_objset
, zb
->zb_object
,
2419 zb
->zb_blkid
* epb
+ i
);
2420 err
= visit_indirect(spa
, dnp
, cbp
, &czb
);
2423 fill
+= BP_GET_FILL(cbp
);
2426 ASSERT3U(fill
, ==, BP_GET_FILL(bp
));
2427 arc_buf_destroy(buf
, &buf
);
2435 dump_indirect(dnode_t
*dn
)
2437 dnode_phys_t
*dnp
= dn
->dn_phys
;
2439 zbookmark_phys_t czb
;
2441 (void) printf("Indirect blocks:\n");
2443 SET_BOOKMARK(&czb
, dmu_objset_id(dn
->dn_objset
),
2444 dn
->dn_object
, dnp
->dn_nlevels
- 1, 0);
2445 for (j
= 0; j
< dnp
->dn_nblkptr
; j
++) {
2447 (void) visit_indirect(dmu_objset_spa(dn
->dn_objset
), dnp
,
2448 &dnp
->dn_blkptr
[j
], &czb
);
2451 (void) printf("\n");
2456 dump_dsl_dir(objset_t
*os
, uint64_t object
, void *data
, size_t size
)
2458 dsl_dir_phys_t
*dd
= data
;
2462 /* make sure nicenum has enough space */
2463 CTASSERT(sizeof (nice
) >= NN_NUMBUF_SZ
);
2468 ASSERT3U(size
, >=, sizeof (dsl_dir_phys_t
));
2470 crtime
= dd
->dd_creation_time
;
2471 (void) printf("\t\tcreation_time = %s", ctime(&crtime
));
2472 (void) printf("\t\thead_dataset_obj = %llu\n",
2473 (u_longlong_t
)dd
->dd_head_dataset_obj
);
2474 (void) printf("\t\tparent_dir_obj = %llu\n",
2475 (u_longlong_t
)dd
->dd_parent_obj
);
2476 (void) printf("\t\torigin_obj = %llu\n",
2477 (u_longlong_t
)dd
->dd_origin_obj
);
2478 (void) printf("\t\tchild_dir_zapobj = %llu\n",
2479 (u_longlong_t
)dd
->dd_child_dir_zapobj
);
2480 zdb_nicenum(dd
->dd_used_bytes
, nice
, sizeof (nice
));
2481 (void) printf("\t\tused_bytes = %s\n", nice
);
2482 zdb_nicenum(dd
->dd_compressed_bytes
, nice
, sizeof (nice
));
2483 (void) printf("\t\tcompressed_bytes = %s\n", nice
);
2484 zdb_nicenum(dd
->dd_uncompressed_bytes
, nice
, sizeof (nice
));
2485 (void) printf("\t\tuncompressed_bytes = %s\n", nice
);
2486 zdb_nicenum(dd
->dd_quota
, nice
, sizeof (nice
));
2487 (void) printf("\t\tquota = %s\n", nice
);
2488 zdb_nicenum(dd
->dd_reserved
, nice
, sizeof (nice
));
2489 (void) printf("\t\treserved = %s\n", nice
);
2490 (void) printf("\t\tprops_zapobj = %llu\n",
2491 (u_longlong_t
)dd
->dd_props_zapobj
);
2492 (void) printf("\t\tdeleg_zapobj = %llu\n",
2493 (u_longlong_t
)dd
->dd_deleg_zapobj
);
2494 (void) printf("\t\tflags = %llx\n",
2495 (u_longlong_t
)dd
->dd_flags
);
2498 zdb_nicenum(dd->dd_used_breakdown[DD_USED_ ## which], nice, \
2500 (void) printf("\t\tused_breakdown[" #which "] = %s\n", nice)
2507 (void) printf("\t\tclones = %llu\n",
2508 (u_longlong_t
)dd
->dd_clones
);
2513 dump_dsl_dataset(objset_t
*os
, uint64_t object
, void *data
, size_t size
)
2515 dsl_dataset_phys_t
*ds
= data
;
2517 char used
[32], compressed
[32], uncompressed
[32], unique
[32];
2518 char blkbuf
[BP_SPRINTF_LEN
];
2520 /* make sure nicenum has enough space */
2521 CTASSERT(sizeof (used
) >= NN_NUMBUF_SZ
);
2522 CTASSERT(sizeof (compressed
) >= NN_NUMBUF_SZ
);
2523 CTASSERT(sizeof (uncompressed
) >= NN_NUMBUF_SZ
);
2524 CTASSERT(sizeof (unique
) >= NN_NUMBUF_SZ
);
2529 ASSERT(size
== sizeof (*ds
));
2530 crtime
= ds
->ds_creation_time
;
2531 zdb_nicenum(ds
->ds_referenced_bytes
, used
, sizeof (used
));
2532 zdb_nicenum(ds
->ds_compressed_bytes
, compressed
, sizeof (compressed
));
2533 zdb_nicenum(ds
->ds_uncompressed_bytes
, uncompressed
,
2534 sizeof (uncompressed
));
2535 zdb_nicenum(ds
->ds_unique_bytes
, unique
, sizeof (unique
));
2536 snprintf_blkptr(blkbuf
, sizeof (blkbuf
), &ds
->ds_bp
);
2538 (void) printf("\t\tdir_obj = %llu\n",
2539 (u_longlong_t
)ds
->ds_dir_obj
);
2540 (void) printf("\t\tprev_snap_obj = %llu\n",
2541 (u_longlong_t
)ds
->ds_prev_snap_obj
);
2542 (void) printf("\t\tprev_snap_txg = %llu\n",
2543 (u_longlong_t
)ds
->ds_prev_snap_txg
);
2544 (void) printf("\t\tnext_snap_obj = %llu\n",
2545 (u_longlong_t
)ds
->ds_next_snap_obj
);
2546 (void) printf("\t\tsnapnames_zapobj = %llu\n",
2547 (u_longlong_t
)ds
->ds_snapnames_zapobj
);
2548 (void) printf("\t\tnum_children = %llu\n",
2549 (u_longlong_t
)ds
->ds_num_children
);
2550 (void) printf("\t\tuserrefs_obj = %llu\n",
2551 (u_longlong_t
)ds
->ds_userrefs_obj
);
2552 (void) printf("\t\tcreation_time = %s", ctime(&crtime
));
2553 (void) printf("\t\tcreation_txg = %llu\n",
2554 (u_longlong_t
)ds
->ds_creation_txg
);
2555 (void) printf("\t\tdeadlist_obj = %llu\n",
2556 (u_longlong_t
)ds
->ds_deadlist_obj
);
2557 (void) printf("\t\tused_bytes = %s\n", used
);
2558 (void) printf("\t\tcompressed_bytes = %s\n", compressed
);
2559 (void) printf("\t\tuncompressed_bytes = %s\n", uncompressed
);
2560 (void) printf("\t\tunique = %s\n", unique
);
2561 (void) printf("\t\tfsid_guid = %llu\n",
2562 (u_longlong_t
)ds
->ds_fsid_guid
);
2563 (void) printf("\t\tguid = %llu\n",
2564 (u_longlong_t
)ds
->ds_guid
);
2565 (void) printf("\t\tflags = %llx\n",
2566 (u_longlong_t
)ds
->ds_flags
);
2567 (void) printf("\t\tnext_clones_obj = %llu\n",
2568 (u_longlong_t
)ds
->ds_next_clones_obj
);
2569 (void) printf("\t\tprops_obj = %llu\n",
2570 (u_longlong_t
)ds
->ds_props_obj
);
2571 (void) printf("\t\tbp = %s\n", blkbuf
);
2576 dump_bptree_cb(void *arg
, const blkptr_t
*bp
, dmu_tx_t
*tx
)
2578 char blkbuf
[BP_SPRINTF_LEN
];
2580 if (bp
->blk_birth
!= 0) {
2581 snprintf_blkptr(blkbuf
, sizeof (blkbuf
), bp
);
2582 (void) printf("\t%s\n", blkbuf
);
2588 dump_bptree(objset_t
*os
, uint64_t obj
, const char *name
)
2594 /* make sure nicenum has enough space */
2595 CTASSERT(sizeof (bytes
) >= NN_NUMBUF_SZ
);
2597 if (dump_opt
['d'] < 3)
2600 VERIFY3U(0, ==, dmu_bonus_hold(os
, obj
, FTAG
, &db
));
2602 zdb_nicenum(bt
->bt_bytes
, bytes
, sizeof (bytes
));
2603 (void) printf("\n %s: %llu datasets, %s\n",
2604 name
, (unsigned long long)(bt
->bt_end
- bt
->bt_begin
), bytes
);
2605 dmu_buf_rele(db
, FTAG
);
2607 if (dump_opt
['d'] < 5)
2610 (void) printf("\n");
2612 (void) bptree_iterate(os
, obj
, B_FALSE
, dump_bptree_cb
, NULL
, NULL
);
2617 dump_bpobj_cb(void *arg
, const blkptr_t
*bp
, boolean_t bp_freed
, dmu_tx_t
*tx
)
2619 char blkbuf
[BP_SPRINTF_LEN
];
2621 ASSERT(bp
->blk_birth
!= 0);
2622 snprintf_blkptr_compact(blkbuf
, sizeof (blkbuf
), bp
, bp_freed
);
2623 (void) printf("\t%s\n", blkbuf
);
2628 dump_full_bpobj(bpobj_t
*bpo
, const char *name
, int indent
)
2635 /* make sure nicenum has enough space */
2636 CTASSERT(sizeof (bytes
) >= NN_NUMBUF_SZ
);
2637 CTASSERT(sizeof (comp
) >= NN_NUMBUF_SZ
);
2638 CTASSERT(sizeof (uncomp
) >= NN_NUMBUF_SZ
);
2640 if (dump_opt
['d'] < 3)
2643 zdb_nicenum(bpo
->bpo_phys
->bpo_bytes
, bytes
, sizeof (bytes
));
2644 if (bpo
->bpo_havesubobj
&& bpo
->bpo_phys
->bpo_subobjs
!= 0) {
2645 zdb_nicenum(bpo
->bpo_phys
->bpo_comp
, comp
, sizeof (comp
));
2646 zdb_nicenum(bpo
->bpo_phys
->bpo_uncomp
, uncomp
, sizeof (uncomp
));
2647 if (bpo
->bpo_havefreed
) {
2648 (void) printf(" %*s: object %llu, %llu local "
2649 "blkptrs, %llu freed, %llu subobjs in object %llu, "
2650 "%s (%s/%s comp)\n",
2652 (u_longlong_t
)bpo
->bpo_object
,
2653 (u_longlong_t
)bpo
->bpo_phys
->bpo_num_blkptrs
,
2654 (u_longlong_t
)bpo
->bpo_phys
->bpo_num_freed
,
2655 (u_longlong_t
)bpo
->bpo_phys
->bpo_num_subobjs
,
2656 (u_longlong_t
)bpo
->bpo_phys
->bpo_subobjs
,
2657 bytes
, comp
, uncomp
);
2659 (void) printf(" %*s: object %llu, %llu local "
2660 "blkptrs, %llu subobjs in object %llu, "
2661 "%s (%s/%s comp)\n",
2663 (u_longlong_t
)bpo
->bpo_object
,
2664 (u_longlong_t
)bpo
->bpo_phys
->bpo_num_blkptrs
,
2665 (u_longlong_t
)bpo
->bpo_phys
->bpo_num_subobjs
,
2666 (u_longlong_t
)bpo
->bpo_phys
->bpo_subobjs
,
2667 bytes
, comp
, uncomp
);
2670 for (i
= 0; i
< bpo
->bpo_phys
->bpo_num_subobjs
; i
++) {
2674 VERIFY0(dmu_read(bpo
->bpo_os
,
2675 bpo
->bpo_phys
->bpo_subobjs
,
2676 i
* sizeof (subobj
), sizeof (subobj
), &subobj
, 0));
2677 error
= bpobj_open(&subbpo
, bpo
->bpo_os
, subobj
);
2679 (void) printf("ERROR %u while trying to open "
2681 error
, (u_longlong_t
)subobj
);
2684 dump_full_bpobj(&subbpo
, "subobj", indent
+ 1);
2685 bpobj_close(&subbpo
);
2688 if (bpo
->bpo_havefreed
) {
2689 (void) printf(" %*s: object %llu, %llu blkptrs, "
2692 (u_longlong_t
)bpo
->bpo_object
,
2693 (u_longlong_t
)bpo
->bpo_phys
->bpo_num_blkptrs
,
2694 (u_longlong_t
)bpo
->bpo_phys
->bpo_num_freed
,
2697 (void) printf(" %*s: object %llu, %llu blkptrs, "
2700 (u_longlong_t
)bpo
->bpo_object
,
2701 (u_longlong_t
)bpo
->bpo_phys
->bpo_num_blkptrs
,
2706 if (dump_opt
['d'] < 5)
2711 (void) bpobj_iterate_nofree(bpo
, dump_bpobj_cb
, NULL
, NULL
);
2712 (void) printf("\n");
2717 dump_bookmark(dsl_pool_t
*dp
, char *name
, boolean_t print_redact
,
2718 boolean_t print_list
)
2721 zfs_bookmark_phys_t prop
;
2722 objset_t
*mos
= dp
->dp_spa
->spa_meta_objset
;
2723 err
= dsl_bookmark_lookup(dp
, name
, NULL
, &prop
);
2729 (void) printf("\t#%s: ", strchr(name
, '#') + 1);
2730 (void) printf("{guid: %llx creation_txg: %llu creation_time: "
2731 "%llu redaction_obj: %llu}\n", (u_longlong_t
)prop
.zbm_guid
,
2732 (u_longlong_t
)prop
.zbm_creation_txg
,
2733 (u_longlong_t
)prop
.zbm_creation_time
,
2734 (u_longlong_t
)prop
.zbm_redaction_obj
);
2736 IMPLY(print_list
, print_redact
);
2737 if (!print_redact
|| prop
.zbm_redaction_obj
== 0)
2740 redaction_list_t
*rl
;
2741 VERIFY0(dsl_redaction_list_hold_obj(dp
,
2742 prop
.zbm_redaction_obj
, FTAG
, &rl
));
2744 redaction_list_phys_t
*rlp
= rl
->rl_phys
;
2745 (void) printf("\tRedacted:\n\t\tProgress: ");
2746 if (rlp
->rlp_last_object
!= UINT64_MAX
||
2747 rlp
->rlp_last_blkid
!= UINT64_MAX
) {
2748 (void) printf("%llu %llu (incomplete)\n",
2749 (u_longlong_t
)rlp
->rlp_last_object
,
2750 (u_longlong_t
)rlp
->rlp_last_blkid
);
2752 (void) printf("complete\n");
2754 (void) printf("\t\tSnapshots: [");
2755 for (unsigned int i
= 0; i
< rlp
->rlp_num_snaps
; i
++) {
2757 (void) printf(", ");
2758 (void) printf("%0llu",
2759 (u_longlong_t
)rlp
->rlp_snaps
[i
]);
2761 (void) printf("]\n\t\tLength: %llu\n",
2762 (u_longlong_t
)rlp
->rlp_num_entries
);
2765 dsl_redaction_list_rele(rl
, FTAG
);
2769 if (rlp
->rlp_num_entries
== 0) {
2770 dsl_redaction_list_rele(rl
, FTAG
);
2771 (void) printf("\t\tRedaction List: []\n\n");
2775 redact_block_phys_t
*rbp_buf
;
2777 dmu_object_info_t doi
;
2779 VERIFY0(dmu_object_info(mos
, prop
.zbm_redaction_obj
, &doi
));
2780 size
= doi
.doi_max_offset
;
2781 rbp_buf
= kmem_alloc(size
, KM_SLEEP
);
2783 err
= dmu_read(mos
, prop
.zbm_redaction_obj
, 0, size
,
2786 dsl_redaction_list_rele(rl
, FTAG
);
2787 kmem_free(rbp_buf
, size
);
2791 (void) printf("\t\tRedaction List: [{object: %llx, offset: "
2792 "%llx, blksz: %x, count: %llx}",
2793 (u_longlong_t
)rbp_buf
[0].rbp_object
,
2794 (u_longlong_t
)rbp_buf
[0].rbp_blkid
,
2795 (uint_t
)(redact_block_get_size(&rbp_buf
[0])),
2796 (u_longlong_t
)redact_block_get_count(&rbp_buf
[0]));
2798 for (size_t i
= 1; i
< rlp
->rlp_num_entries
; i
++) {
2799 (void) printf(",\n\t\t{object: %llx, offset: %llx, "
2800 "blksz: %x, count: %llx}",
2801 (u_longlong_t
)rbp_buf
[i
].rbp_object
,
2802 (u_longlong_t
)rbp_buf
[i
].rbp_blkid
,
2803 (uint_t
)(redact_block_get_size(&rbp_buf
[i
])),
2804 (u_longlong_t
)redact_block_get_count(&rbp_buf
[i
]));
2806 dsl_redaction_list_rele(rl
, FTAG
);
2807 kmem_free(rbp_buf
, size
);
2808 (void) printf("]\n\n");
2813 dump_bookmarks(objset_t
*os
, int verbosity
)
2816 zap_attribute_t attr
;
2817 dsl_dataset_t
*ds
= dmu_objset_ds(os
);
2818 dsl_pool_t
*dp
= spa_get_dsl(os
->os_spa
);
2819 objset_t
*mos
= os
->os_spa
->spa_meta_objset
;
2822 dsl_pool_config_enter(dp
, FTAG
);
2824 for (zap_cursor_init(&zc
, mos
, ds
->ds_bookmarks_obj
);
2825 zap_cursor_retrieve(&zc
, &attr
) == 0;
2826 zap_cursor_advance(&zc
)) {
2827 char osname
[ZFS_MAX_DATASET_NAME_LEN
];
2828 char buf
[ZFS_MAX_DATASET_NAME_LEN
];
2829 dmu_objset_name(os
, osname
);
2830 VERIFY3S(0, <=, snprintf(buf
, sizeof (buf
), "%s#%s", osname
,
2832 (void) dump_bookmark(dp
, buf
, verbosity
>= 5, verbosity
>= 6);
2834 zap_cursor_fini(&zc
);
2835 dsl_pool_config_exit(dp
, FTAG
);
2839 bpobj_count_refd(bpobj_t
*bpo
)
2841 mos_obj_refd(bpo
->bpo_object
);
2843 if (bpo
->bpo_havesubobj
&& bpo
->bpo_phys
->bpo_subobjs
!= 0) {
2844 mos_obj_refd(bpo
->bpo_phys
->bpo_subobjs
);
2845 for (uint64_t i
= 0; i
< bpo
->bpo_phys
->bpo_num_subobjs
; i
++) {
2849 VERIFY0(dmu_read(bpo
->bpo_os
,
2850 bpo
->bpo_phys
->bpo_subobjs
,
2851 i
* sizeof (subobj
), sizeof (subobj
), &subobj
, 0));
2852 error
= bpobj_open(&subbpo
, bpo
->bpo_os
, subobj
);
2854 (void) printf("ERROR %u while trying to open "
2856 error
, (u_longlong_t
)subobj
);
2859 bpobj_count_refd(&subbpo
);
2860 bpobj_close(&subbpo
);
2866 dsl_deadlist_entry_count_refd(void *arg
, dsl_deadlist_entry_t
*dle
)
2869 uint64_t empty_bpobj
= spa
->spa_dsl_pool
->dp_empty_bpobj
;
2870 if (dle
->dle_bpobj
.bpo_object
!= empty_bpobj
)
2871 bpobj_count_refd(&dle
->dle_bpobj
);
2876 dsl_deadlist_entry_dump(void *arg
, dsl_deadlist_entry_t
*dle
)
2878 ASSERT(arg
== NULL
);
2879 if (dump_opt
['d'] >= 5) {
2881 (void) snprintf(buf
, sizeof (buf
),
2882 "mintxg %llu -> obj %llu",
2883 (longlong_t
)dle
->dle_mintxg
,
2884 (longlong_t
)dle
->dle_bpobj
.bpo_object
);
2886 dump_full_bpobj(&dle
->dle_bpobj
, buf
, 0);
2888 (void) printf("mintxg %llu -> obj %llu\n",
2889 (longlong_t
)dle
->dle_mintxg
,
2890 (longlong_t
)dle
->dle_bpobj
.bpo_object
);
2896 dump_blkptr_list(dsl_deadlist_t
*dl
, char *name
)
2902 spa_t
*spa
= dmu_objset_spa(dl
->dl_os
);
2903 uint64_t empty_bpobj
= spa
->spa_dsl_pool
->dp_empty_bpobj
;
2905 if (dl
->dl_oldfmt
) {
2906 if (dl
->dl_bpobj
.bpo_object
!= empty_bpobj
)
2907 bpobj_count_refd(&dl
->dl_bpobj
);
2909 mos_obj_refd(dl
->dl_object
);
2910 dsl_deadlist_iterate(dl
, dsl_deadlist_entry_count_refd
, spa
);
2913 /* make sure nicenum has enough space */
2914 CTASSERT(sizeof (bytes
) >= NN_NUMBUF_SZ
);
2915 CTASSERT(sizeof (comp
) >= NN_NUMBUF_SZ
);
2916 CTASSERT(sizeof (uncomp
) >= NN_NUMBUF_SZ
);
2917 CTASSERT(sizeof (entries
) >= NN_NUMBUF_SZ
);
2919 if (dump_opt
['d'] < 3)
2922 if (dl
->dl_oldfmt
) {
2923 dump_full_bpobj(&dl
->dl_bpobj
, "old-format deadlist", 0);
2927 zdb_nicenum(dl
->dl_phys
->dl_used
, bytes
, sizeof (bytes
));
2928 zdb_nicenum(dl
->dl_phys
->dl_comp
, comp
, sizeof (comp
));
2929 zdb_nicenum(dl
->dl_phys
->dl_uncomp
, uncomp
, sizeof (uncomp
));
2930 zdb_nicenum(avl_numnodes(&dl
->dl_tree
), entries
, sizeof (entries
));
2931 (void) printf("\n %s: %s (%s/%s comp), %s entries\n",
2932 name
, bytes
, comp
, uncomp
, entries
);
2934 if (dump_opt
['d'] < 4)
2937 (void) printf("\n");
2939 dsl_deadlist_iterate(dl
, dsl_deadlist_entry_dump
, NULL
);
2943 verify_dd_livelist(objset_t
*os
)
2945 uint64_t ll_used
, used
, ll_comp
, comp
, ll_uncomp
, uncomp
;
2946 dsl_pool_t
*dp
= spa_get_dsl(os
->os_spa
);
2947 dsl_dir_t
*dd
= os
->os_dsl_dataset
->ds_dir
;
2949 ASSERT(!dmu_objset_is_snapshot(os
));
2950 if (!dsl_deadlist_is_open(&dd
->dd_livelist
))
2953 /* Iterate through the livelist to check for duplicates */
2954 dsl_deadlist_iterate(&dd
->dd_livelist
, sublivelist_verify_lightweight
,
2957 dsl_pool_config_enter(dp
, FTAG
);
2958 dsl_deadlist_space(&dd
->dd_livelist
, &ll_used
,
2959 &ll_comp
, &ll_uncomp
);
2961 dsl_dataset_t
*origin_ds
;
2962 ASSERT(dsl_pool_config_held(dp
));
2963 VERIFY0(dsl_dataset_hold_obj(dp
,
2964 dsl_dir_phys(dd
)->dd_origin_obj
, FTAG
, &origin_ds
));
2965 VERIFY0(dsl_dataset_space_written(origin_ds
, os
->os_dsl_dataset
,
2966 &used
, &comp
, &uncomp
));
2967 dsl_dataset_rele(origin_ds
, FTAG
);
2968 dsl_pool_config_exit(dp
, FTAG
);
2970 * It's possible that the dataset's uncomp space is larger than the
2971 * livelist's because livelists do not track embedded block pointers
2973 if (used
!= ll_used
|| comp
!= ll_comp
|| uncomp
< ll_uncomp
) {
2974 char nice_used
[32], nice_comp
[32], nice_uncomp
[32];
2975 (void) printf("Discrepancy in space accounting:\n");
2976 zdb_nicenum(used
, nice_used
, sizeof (nice_used
));
2977 zdb_nicenum(comp
, nice_comp
, sizeof (nice_comp
));
2978 zdb_nicenum(uncomp
, nice_uncomp
, sizeof (nice_uncomp
));
2979 (void) printf("dir: used %s, comp %s, uncomp %s\n",
2980 nice_used
, nice_comp
, nice_uncomp
);
2981 zdb_nicenum(ll_used
, nice_used
, sizeof (nice_used
));
2982 zdb_nicenum(ll_comp
, nice_comp
, sizeof (nice_comp
));
2983 zdb_nicenum(ll_uncomp
, nice_uncomp
, sizeof (nice_uncomp
));
2984 (void) printf("livelist: used %s, comp %s, uncomp %s\n",
2985 nice_used
, nice_comp
, nice_uncomp
);
2991 static avl_tree_t idx_tree
;
2992 static avl_tree_t domain_tree
;
2993 static boolean_t fuid_table_loaded
;
2994 static objset_t
*sa_os
= NULL
;
2995 static sa_attr_type_t
*sa_attr_table
= NULL
;
2998 open_objset(const char *path
, void *tag
, objset_t
**osp
)
3001 uint64_t sa_attrs
= 0;
3002 uint64_t version
= 0;
3004 VERIFY3P(sa_os
, ==, NULL
);
3006 * We can't own an objset if it's redacted. Therefore, we do this
3007 * dance: hold the objset, then acquire a long hold on its dataset, then
3008 * release the pool (which is held as part of holding the objset).
3010 err
= dmu_objset_hold(path
, tag
, osp
);
3012 (void) fprintf(stderr
, "failed to hold dataset '%s': %s\n",
3013 path
, strerror(err
));
3016 dsl_dataset_long_hold(dmu_objset_ds(*osp
), tag
);
3017 dsl_pool_rele(dmu_objset_pool(*osp
), tag
);
3019 if (dmu_objset_type(*osp
) == DMU_OST_ZFS
&& !(*osp
)->os_encrypted
) {
3020 (void) zap_lookup(*osp
, MASTER_NODE_OBJ
, ZPL_VERSION_STR
,
3022 if (version
>= ZPL_VERSION_SA
) {
3023 (void) zap_lookup(*osp
, MASTER_NODE_OBJ
, ZFS_SA_ATTRS
,
3026 err
= sa_setup(*osp
, sa_attrs
, zfs_attr_table
, ZPL_END
,
3029 (void) fprintf(stderr
, "sa_setup failed: %s\n",
3031 dsl_dataset_long_rele(dmu_objset_ds(*osp
), tag
);
3032 dsl_dataset_rele(dmu_objset_ds(*osp
), tag
);
3042 close_objset(objset_t
*os
, void *tag
)
3044 VERIFY3P(os
, ==, sa_os
);
3045 if (os
->os_sa
!= NULL
)
3047 dsl_dataset_long_rele(dmu_objset_ds(os
), tag
);
3048 dsl_dataset_rele(dmu_objset_ds(os
), tag
);
3049 sa_attr_table
= NULL
;
3054 fuid_table_destroy(void)
3056 if (fuid_table_loaded
) {
3057 zfs_fuid_table_destroy(&idx_tree
, &domain_tree
);
3058 fuid_table_loaded
= B_FALSE
;
3063 * print uid or gid information.
3064 * For normal POSIX id just the id is printed in decimal format.
3065 * For CIFS files with FUID the fuid is printed in hex followed by
3066 * the domain-rid string.
3069 print_idstr(uint64_t id
, const char *id_type
)
3071 if (FUID_INDEX(id
)) {
3074 domain
= zfs_fuid_idx_domain(&idx_tree
, FUID_INDEX(id
));
3075 (void) printf("\t%s %llx [%s-%d]\n", id_type
,
3076 (u_longlong_t
)id
, domain
, (int)FUID_RID(id
));
3078 (void) printf("\t%s %llu\n", id_type
, (u_longlong_t
)id
);
3084 dump_uidgid(objset_t
*os
, uint64_t uid
, uint64_t gid
)
3086 uint32_t uid_idx
, gid_idx
;
3088 uid_idx
= FUID_INDEX(uid
);
3089 gid_idx
= FUID_INDEX(gid
);
3091 /* Load domain table, if not already loaded */
3092 if (!fuid_table_loaded
&& (uid_idx
|| gid_idx
)) {
3095 /* first find the fuid object. It lives in the master node */
3096 VERIFY(zap_lookup(os
, MASTER_NODE_OBJ
, ZFS_FUID_TABLES
,
3097 8, 1, &fuid_obj
) == 0);
3098 zfs_fuid_avl_tree_create(&idx_tree
, &domain_tree
);
3099 (void) zfs_fuid_table_load(os
, fuid_obj
,
3100 &idx_tree
, &domain_tree
);
3101 fuid_table_loaded
= B_TRUE
;
3104 print_idstr(uid
, "uid");
3105 print_idstr(gid
, "gid");
3109 dump_znode_sa_xattr(sa_handle_t
*hdl
)
3112 nvpair_t
*elem
= NULL
;
3113 int sa_xattr_size
= 0;
3114 int sa_xattr_entries
= 0;
3116 char *sa_xattr_packed
;
3118 error
= sa_size(hdl
, sa_attr_table
[ZPL_DXATTR
], &sa_xattr_size
);
3119 if (error
|| sa_xattr_size
== 0)
3122 sa_xattr_packed
= malloc(sa_xattr_size
);
3123 if (sa_xattr_packed
== NULL
)
3126 error
= sa_lookup(hdl
, sa_attr_table
[ZPL_DXATTR
],
3127 sa_xattr_packed
, sa_xattr_size
);
3129 free(sa_xattr_packed
);
3133 error
= nvlist_unpack(sa_xattr_packed
, sa_xattr_size
, &sa_xattr
, 0);
3135 free(sa_xattr_packed
);
3139 while ((elem
= nvlist_next_nvpair(sa_xattr
, elem
)) != NULL
)
3142 (void) printf("\tSA xattrs: %d bytes, %d entries\n\n",
3143 sa_xattr_size
, sa_xattr_entries
);
3144 while ((elem
= nvlist_next_nvpair(sa_xattr
, elem
)) != NULL
) {
3148 (void) printf("\t\t%s = ", nvpair_name(elem
));
3149 nvpair_value_byte_array(elem
, &value
, &cnt
);
3150 for (idx
= 0; idx
< cnt
; ++idx
) {
3151 if (isprint(value
[idx
]))
3152 (void) putchar(value
[idx
]);
3154 (void) printf("\\%3.3o", value
[idx
]);
3156 (void) putchar('\n');
3159 nvlist_free(sa_xattr
);
3160 free(sa_xattr_packed
);
3164 dump_znode_symlink(sa_handle_t
*hdl
)
3166 int sa_symlink_size
= 0;
3167 char linktarget
[MAXPATHLEN
];
3168 linktarget
[0] = '\0';
3171 error
= sa_size(hdl
, sa_attr_table
[ZPL_SYMLINK
], &sa_symlink_size
);
3172 if (error
|| sa_symlink_size
== 0) {
3175 if (sa_lookup(hdl
, sa_attr_table
[ZPL_SYMLINK
],
3176 &linktarget
, sa_symlink_size
) == 0)
3177 (void) printf("\ttarget %s\n", linktarget
);
3182 dump_znode(objset_t
*os
, uint64_t object
, void *data
, size_t size
)
3184 char path
[MAXPATHLEN
* 2]; /* allow for xattr and failure prefix */
3186 uint64_t xattr
, rdev
, gen
;
3187 uint64_t uid
, gid
, mode
, fsize
, parent
, links
;
3189 uint64_t acctm
[2], modtm
[2], chgtm
[2], crtm
[2];
3190 time_t z_crtime
, z_atime
, z_mtime
, z_ctime
;
3191 sa_bulk_attr_t bulk
[12];
3195 VERIFY3P(os
, ==, sa_os
);
3196 if (sa_handle_get(os
, object
, NULL
, SA_HDL_PRIVATE
, &hdl
)) {
3197 (void) printf("Failed to get handle for SA znode\n");
3201 SA_ADD_BULK_ATTR(bulk
, idx
, sa_attr_table
[ZPL_UID
], NULL
, &uid
, 8);
3202 SA_ADD_BULK_ATTR(bulk
, idx
, sa_attr_table
[ZPL_GID
], NULL
, &gid
, 8);
3203 SA_ADD_BULK_ATTR(bulk
, idx
, sa_attr_table
[ZPL_LINKS
], NULL
,
3205 SA_ADD_BULK_ATTR(bulk
, idx
, sa_attr_table
[ZPL_GEN
], NULL
, &gen
, 8);
3206 SA_ADD_BULK_ATTR(bulk
, idx
, sa_attr_table
[ZPL_MODE
], NULL
,
3208 SA_ADD_BULK_ATTR(bulk
, idx
, sa_attr_table
[ZPL_PARENT
],
3210 SA_ADD_BULK_ATTR(bulk
, idx
, sa_attr_table
[ZPL_SIZE
], NULL
,
3212 SA_ADD_BULK_ATTR(bulk
, idx
, sa_attr_table
[ZPL_ATIME
], NULL
,
3214 SA_ADD_BULK_ATTR(bulk
, idx
, sa_attr_table
[ZPL_MTIME
], NULL
,
3216 SA_ADD_BULK_ATTR(bulk
, idx
, sa_attr_table
[ZPL_CRTIME
], NULL
,
3218 SA_ADD_BULK_ATTR(bulk
, idx
, sa_attr_table
[ZPL_CTIME
], NULL
,
3220 SA_ADD_BULK_ATTR(bulk
, idx
, sa_attr_table
[ZPL_FLAGS
], NULL
,
3223 if (sa_bulk_lookup(hdl
, bulk
, idx
)) {
3224 (void) sa_handle_destroy(hdl
);
3228 z_crtime
= (time_t)crtm
[0];
3229 z_atime
= (time_t)acctm
[0];
3230 z_mtime
= (time_t)modtm
[0];
3231 z_ctime
= (time_t)chgtm
[0];
3233 if (dump_opt
['d'] > 4) {
3234 error
= zfs_obj_to_path(os
, object
, path
, sizeof (path
));
3235 if (error
== ESTALE
) {
3236 (void) snprintf(path
, sizeof (path
), "on delete queue");
3237 } else if (error
!= 0) {
3239 (void) snprintf(path
, sizeof (path
),
3240 "path not found, possibly leaked");
3242 (void) printf("\tpath %s\n", path
);
3246 dump_znode_symlink(hdl
);
3247 dump_uidgid(os
, uid
, gid
);
3248 (void) printf("\tatime %s", ctime(&z_atime
));
3249 (void) printf("\tmtime %s", ctime(&z_mtime
));
3250 (void) printf("\tctime %s", ctime(&z_ctime
));
3251 (void) printf("\tcrtime %s", ctime(&z_crtime
));
3252 (void) printf("\tgen %llu\n", (u_longlong_t
)gen
);
3253 (void) printf("\tmode %llo\n", (u_longlong_t
)mode
);
3254 (void) printf("\tsize %llu\n", (u_longlong_t
)fsize
);
3255 (void) printf("\tparent %llu\n", (u_longlong_t
)parent
);
3256 (void) printf("\tlinks %llu\n", (u_longlong_t
)links
);
3257 (void) printf("\tpflags %llx\n", (u_longlong_t
)pflags
);
3258 if (dmu_objset_projectquota_enabled(os
) && (pflags
& ZFS_PROJID
)) {
3261 if (sa_lookup(hdl
, sa_attr_table
[ZPL_PROJID
], &projid
,
3262 sizeof (uint64_t)) == 0)
3263 (void) printf("\tprojid %llu\n", (u_longlong_t
)projid
);
3265 if (sa_lookup(hdl
, sa_attr_table
[ZPL_XATTR
], &xattr
,
3266 sizeof (uint64_t)) == 0)
3267 (void) printf("\txattr %llu\n", (u_longlong_t
)xattr
);
3268 if (sa_lookup(hdl
, sa_attr_table
[ZPL_RDEV
], &rdev
,
3269 sizeof (uint64_t)) == 0)
3270 (void) printf("\trdev 0x%016llx\n", (u_longlong_t
)rdev
);
3271 dump_znode_sa_xattr(hdl
);
3272 sa_handle_destroy(hdl
);
3277 dump_acl(objset_t
*os
, uint64_t object
, void *data
, size_t size
)
3283 dump_dmu_objset(objset_t
*os
, uint64_t object
, void *data
, size_t size
)
3287 static object_viewer_t
*object_viewer
[DMU_OT_NUMTYPES
+ 1] = {
3288 dump_none
, /* unallocated */
3289 dump_zap
, /* object directory */
3290 dump_uint64
, /* object array */
3291 dump_none
, /* packed nvlist */
3292 dump_packed_nvlist
, /* packed nvlist size */
3293 dump_none
, /* bpobj */
3294 dump_bpobj
, /* bpobj header */
3295 dump_none
, /* SPA space map header */
3296 dump_none
, /* SPA space map */
3297 dump_none
, /* ZIL intent log */
3298 dump_dnode
, /* DMU dnode */
3299 dump_dmu_objset
, /* DMU objset */
3300 dump_dsl_dir
, /* DSL directory */
3301 dump_zap
, /* DSL directory child map */
3302 dump_zap
, /* DSL dataset snap map */
3303 dump_zap
, /* DSL props */
3304 dump_dsl_dataset
, /* DSL dataset */
3305 dump_znode
, /* ZFS znode */
3306 dump_acl
, /* ZFS V0 ACL */
3307 dump_uint8
, /* ZFS plain file */
3308 dump_zpldir
, /* ZFS directory */
3309 dump_zap
, /* ZFS master node */
3310 dump_zap
, /* ZFS delete queue */
3311 dump_uint8
, /* zvol object */
3312 dump_zap
, /* zvol prop */
3313 dump_uint8
, /* other uint8[] */
3314 dump_uint64
, /* other uint64[] */
3315 dump_zap
, /* other ZAP */
3316 dump_zap
, /* persistent error log */
3317 dump_uint8
, /* SPA history */
3318 dump_history_offsets
, /* SPA history offsets */
3319 dump_zap
, /* Pool properties */
3320 dump_zap
, /* DSL permissions */
3321 dump_acl
, /* ZFS ACL */
3322 dump_uint8
, /* ZFS SYSACL */
3323 dump_none
, /* FUID nvlist */
3324 dump_packed_nvlist
, /* FUID nvlist size */
3325 dump_zap
, /* DSL dataset next clones */
3326 dump_zap
, /* DSL scrub queue */
3327 dump_zap
, /* ZFS user/group/project used */
3328 dump_zap
, /* ZFS user/group/project quota */
3329 dump_zap
, /* snapshot refcount tags */
3330 dump_ddt_zap
, /* DDT ZAP object */
3331 dump_zap
, /* DDT statistics */
3332 dump_znode
, /* SA object */
3333 dump_zap
, /* SA Master Node */
3334 dump_sa_attrs
, /* SA attribute registration */
3335 dump_sa_layouts
, /* SA attribute layouts */
3336 dump_zap
, /* DSL scrub translations */
3337 dump_none
, /* fake dedup BP */
3338 dump_zap
, /* deadlist */
3339 dump_none
, /* deadlist hdr */
3340 dump_zap
, /* dsl clones */
3341 dump_bpobj_subobjs
, /* bpobj subobjs */
3342 dump_unknown
, /* Unknown type, must be last */
3346 match_object_type(dmu_object_type_t obj_type
, uint64_t flags
)
3348 boolean_t match
= B_TRUE
;
3351 case DMU_OT_DIRECTORY_CONTENTS
:
3352 if (!(flags
& ZOR_FLAG_DIRECTORY
))
3355 case DMU_OT_PLAIN_FILE_CONTENTS
:
3356 if (!(flags
& ZOR_FLAG_PLAIN_FILE
))
3359 case DMU_OT_SPACE_MAP
:
3360 if (!(flags
& ZOR_FLAG_SPACE_MAP
))
3364 if (strcmp(zdb_ot_name(obj_type
), "zap") == 0) {
3365 if (!(flags
& ZOR_FLAG_ZAP
))
3371 * If all bits except some of the supported flags are
3372 * set, the user combined the all-types flag (A) with
3373 * a negated flag to exclude some types (e.g. A-f to
3374 * show all object types except plain files).
3376 if ((flags
| ZOR_SUPPORTED_FLAGS
) != ZOR_FLAG_ALL_TYPES
)
3386 dump_object(objset_t
*os
, uint64_t object
, int verbosity
,
3387 boolean_t
*print_header
, uint64_t *dnode_slots_used
, uint64_t flags
)
3389 dmu_buf_t
*db
= NULL
;
3390 dmu_object_info_t doi
;
3392 boolean_t dnode_held
= B_FALSE
;
3395 char iblk
[32], dblk
[32], lsize
[32], asize
[32], fill
[32], dnsize
[32];
3396 char bonus_size
[32];
3400 /* make sure nicenum has enough space */
3401 CTASSERT(sizeof (iblk
) >= NN_NUMBUF_SZ
);
3402 CTASSERT(sizeof (dblk
) >= NN_NUMBUF_SZ
);
3403 CTASSERT(sizeof (lsize
) >= NN_NUMBUF_SZ
);
3404 CTASSERT(sizeof (asize
) >= NN_NUMBUF_SZ
);
3405 CTASSERT(sizeof (bonus_size
) >= NN_NUMBUF_SZ
);
3407 if (*print_header
) {
3408 (void) printf("\n%10s %3s %5s %5s %5s %6s %5s %6s %s\n",
3409 "Object", "lvl", "iblk", "dblk", "dsize", "dnsize",
3410 "lsize", "%full", "type");
3415 dn
= DMU_META_DNODE(os
);
3416 dmu_object_info_from_dnode(dn
, &doi
);
3419 * Encrypted datasets will have sensitive bonus buffers
3420 * encrypted. Therefore we cannot hold the bonus buffer and
3421 * must hold the dnode itself instead.
3423 error
= dmu_object_info(os
, object
, &doi
);
3425 fatal("dmu_object_info() failed, errno %u", error
);
3427 if (os
->os_encrypted
&&
3428 DMU_OT_IS_ENCRYPTED(doi
.doi_bonus_type
)) {
3429 error
= dnode_hold(os
, object
, FTAG
, &dn
);
3431 fatal("dnode_hold() failed, errno %u", error
);
3432 dnode_held
= B_TRUE
;
3434 error
= dmu_bonus_hold(os
, object
, FTAG
, &db
);
3436 fatal("dmu_bonus_hold(%llu) failed, errno %u",
3438 bonus
= db
->db_data
;
3439 bsize
= db
->db_size
;
3440 dn
= DB_DNODE((dmu_buf_impl_t
*)db
);
3445 * Default to showing all object types if no flags were specified.
3447 if (flags
!= 0 && flags
!= ZOR_FLAG_ALL_TYPES
&&
3448 !match_object_type(doi
.doi_type
, flags
))
3451 if (dnode_slots_used
)
3452 *dnode_slots_used
= doi
.doi_dnodesize
/ DNODE_MIN_SIZE
;
3454 zdb_nicenum(doi
.doi_metadata_block_size
, iblk
, sizeof (iblk
));
3455 zdb_nicenum(doi
.doi_data_block_size
, dblk
, sizeof (dblk
));
3456 zdb_nicenum(doi
.doi_max_offset
, lsize
, sizeof (lsize
));
3457 zdb_nicenum(doi
.doi_physical_blocks_512
<< 9, asize
, sizeof (asize
));
3458 zdb_nicenum(doi
.doi_bonus_size
, bonus_size
, sizeof (bonus_size
));
3459 zdb_nicenum(doi
.doi_dnodesize
, dnsize
, sizeof (dnsize
));
3460 (void) sprintf(fill
, "%6.2f", 100.0 * doi
.doi_fill_count
*
3461 doi
.doi_data_block_size
/ (object
== 0 ? DNODES_PER_BLOCK
: 1) /
3462 doi
.doi_max_offset
);
3466 if (doi
.doi_checksum
!= ZIO_CHECKSUM_INHERIT
|| verbosity
>= 6) {
3467 (void) snprintf(aux
+ strlen(aux
), sizeof (aux
) - strlen(aux
),
3468 " (K=%s)", ZDB_CHECKSUM_NAME(doi
.doi_checksum
));
3471 if (doi
.doi_compress
== ZIO_COMPRESS_INHERIT
&&
3472 ZIO_COMPRESS_HASLEVEL(os
->os_compress
) && verbosity
>= 6) {
3473 const char *compname
= NULL
;
3474 if (zfs_prop_index_to_string(ZFS_PROP_COMPRESSION
,
3475 ZIO_COMPRESS_RAW(os
->os_compress
, os
->os_complevel
),
3477 (void) snprintf(aux
+ strlen(aux
),
3478 sizeof (aux
) - strlen(aux
), " (Z=inherit=%s)",
3481 (void) snprintf(aux
+ strlen(aux
),
3482 sizeof (aux
) - strlen(aux
),
3483 " (Z=inherit=%s-unknown)",
3484 ZDB_COMPRESS_NAME(os
->os_compress
));
3486 } else if (doi
.doi_compress
== ZIO_COMPRESS_INHERIT
&& verbosity
>= 6) {
3487 (void) snprintf(aux
+ strlen(aux
), sizeof (aux
) - strlen(aux
),
3488 " (Z=inherit=%s)", ZDB_COMPRESS_NAME(os
->os_compress
));
3489 } else if (doi
.doi_compress
!= ZIO_COMPRESS_INHERIT
|| verbosity
>= 6) {
3490 (void) snprintf(aux
+ strlen(aux
), sizeof (aux
) - strlen(aux
),
3491 " (Z=%s)", ZDB_COMPRESS_NAME(doi
.doi_compress
));
3494 (void) printf("%10lld %3u %5s %5s %5s %6s %5s %6s %s%s\n",
3495 (u_longlong_t
)object
, doi
.doi_indirection
, iblk
, dblk
,
3496 asize
, dnsize
, lsize
, fill
, zdb_ot_name(doi
.doi_type
), aux
);
3498 if (doi
.doi_bonus_type
!= DMU_OT_NONE
&& verbosity
> 3) {
3499 (void) printf("%10s %3s %5s %5s %5s %5s %5s %6s %s\n",
3500 "", "", "", "", "", "", bonus_size
, "bonus",
3501 zdb_ot_name(doi
.doi_bonus_type
));
3504 if (verbosity
>= 4) {
3505 (void) printf("\tdnode flags: %s%s%s%s\n",
3506 (dn
->dn_phys
->dn_flags
& DNODE_FLAG_USED_BYTES
) ?
3508 (dn
->dn_phys
->dn_flags
& DNODE_FLAG_USERUSED_ACCOUNTED
) ?
3509 "USERUSED_ACCOUNTED " : "",
3510 (dn
->dn_phys
->dn_flags
& DNODE_FLAG_USEROBJUSED_ACCOUNTED
) ?
3511 "USEROBJUSED_ACCOUNTED " : "",
3512 (dn
->dn_phys
->dn_flags
& DNODE_FLAG_SPILL_BLKPTR
) ?
3513 "SPILL_BLKPTR" : "");
3514 (void) printf("\tdnode maxblkid: %llu\n",
3515 (longlong_t
)dn
->dn_phys
->dn_maxblkid
);
3518 object_viewer
[ZDB_OT_TYPE(doi
.doi_bonus_type
)](os
,
3519 object
, bonus
, bsize
);
3521 (void) printf("\t\t(bonus encrypted)\n");
3524 if (!os
->os_encrypted
|| !DMU_OT_IS_ENCRYPTED(doi
.doi_type
)) {
3525 object_viewer
[ZDB_OT_TYPE(doi
.doi_type
)](os
, object
,
3528 (void) printf("\t\t(object encrypted)\n");
3531 *print_header
= B_TRUE
;
3537 if (verbosity
>= 5) {
3539 * Report the list of segments that comprise the object.
3543 uint64_t blkfill
= 1;
3546 if (dn
->dn_type
== DMU_OT_DNODE
) {
3548 blkfill
= DNODES_PER_BLOCK
;
3553 /* make sure nicenum has enough space */
3554 CTASSERT(sizeof (segsize
) >= NN_NUMBUF_SZ
);
3555 error
= dnode_next_offset(dn
,
3556 0, &start
, minlvl
, blkfill
, 0);
3560 error
= dnode_next_offset(dn
,
3561 DNODE_FIND_HOLE
, &end
, minlvl
, blkfill
, 0);
3562 zdb_nicenum(end
- start
, segsize
, sizeof (segsize
));
3563 (void) printf("\t\tsegment [%016llx, %016llx)"
3564 " size %5s\n", (u_longlong_t
)start
,
3565 (u_longlong_t
)end
, segsize
);
3574 dmu_buf_rele(db
, FTAG
);
3576 dnode_rele(dn
, FTAG
);
3580 count_dir_mos_objects(dsl_dir_t
*dd
)
3582 mos_obj_refd(dd
->dd_object
);
3583 mos_obj_refd(dsl_dir_phys(dd
)->dd_child_dir_zapobj
);
3584 mos_obj_refd(dsl_dir_phys(dd
)->dd_deleg_zapobj
);
3585 mos_obj_refd(dsl_dir_phys(dd
)->dd_props_zapobj
);
3586 mos_obj_refd(dsl_dir_phys(dd
)->dd_clones
);
3589 * The dd_crypto_obj can be referenced by multiple dsl_dir's.
3590 * Ignore the references after the first one.
3592 mos_obj_refd_multiple(dd
->dd_crypto_obj
);
3596 count_ds_mos_objects(dsl_dataset_t
*ds
)
3598 mos_obj_refd(ds
->ds_object
);
3599 mos_obj_refd(dsl_dataset_phys(ds
)->ds_next_clones_obj
);
3600 mos_obj_refd(dsl_dataset_phys(ds
)->ds_props_obj
);
3601 mos_obj_refd(dsl_dataset_phys(ds
)->ds_userrefs_obj
);
3602 mos_obj_refd(dsl_dataset_phys(ds
)->ds_snapnames_zapobj
);
3603 mos_obj_refd(ds
->ds_bookmarks_obj
);
3605 if (!dsl_dataset_is_snapshot(ds
)) {
3606 count_dir_mos_objects(ds
->ds_dir
);
3610 static const char *objset_types
[DMU_OST_NUMTYPES
] = {
3611 "NONE", "META", "ZPL", "ZVOL", "OTHER", "ANY" };
3614 * Parse a string denoting a range of object IDs of the form
3615 * <start>[:<end>[:flags]], and store the results in zor.
3616 * Return 0 on success. On error, return 1 and update the msg
3617 * pointer to point to a descriptive error message.
3620 parse_object_range(char *range
, zopt_object_range_t
*zor
, char **msg
)
3623 char *p
, *s
, *dup
, *flagstr
, *tmp
= NULL
;
3628 if (strchr(range
, ':') == NULL
) {
3629 zor
->zor_obj_start
= strtoull(range
, &p
, 0);
3631 *msg
= "Invalid characters in object ID";
3634 zor
->zor_obj_end
= zor
->zor_obj_start
;
3638 if (strchr(range
, ':') == range
) {
3639 *msg
= "Invalid leading colon";
3644 len
= strlen(range
);
3645 if (range
[len
- 1] == ':') {
3646 *msg
= "Invalid trailing colon";
3651 dup
= strdup(range
);
3652 s
= strtok_r(dup
, ":", &tmp
);
3653 zor
->zor_obj_start
= strtoull(s
, &p
, 0);
3656 *msg
= "Invalid characters in start object ID";
3661 s
= strtok_r(NULL
, ":", &tmp
);
3662 zor
->zor_obj_end
= strtoull(s
, &p
, 0);
3665 *msg
= "Invalid characters in end object ID";
3670 if (zor
->zor_obj_start
> zor
->zor_obj_end
) {
3671 *msg
= "Start object ID may not exceed end object ID";
3676 s
= strtok_r(NULL
, ":", &tmp
);
3678 zor
->zor_flags
= ZOR_FLAG_ALL_TYPES
;
3680 } else if (strtok_r(NULL
, ":", &tmp
) != NULL
) {
3681 *msg
= "Invalid colon-delimited field after flags";
3687 for (i
= 0; flagstr
[i
]; i
++) {
3689 boolean_t negation
= (flagstr
[i
] == '-');
3693 if (flagstr
[i
] == '\0') {
3694 *msg
= "Invalid trailing negation operator";
3699 bit
= flagbits
[(uchar_t
)flagstr
[i
]];
3701 *msg
= "Invalid flag";
3710 zor
->zor_flags
= flags
;
3718 dump_objset(objset_t
*os
)
3720 dmu_objset_stats_t dds
= { 0 };
3721 uint64_t object
, object_count
;
3722 uint64_t refdbytes
, usedobjs
, scratch
;
3724 char blkbuf
[BP_SPRINTF_LEN
+ 20];
3725 char osname
[ZFS_MAX_DATASET_NAME_LEN
];
3726 const char *type
= "UNKNOWN";
3727 int verbosity
= dump_opt
['d'];
3728 boolean_t print_header
;
3731 uint64_t total_slots_used
= 0;
3732 uint64_t max_slot_used
= 0;
3733 uint64_t dnode_slots
;
3738 /* make sure nicenum has enough space */
3739 CTASSERT(sizeof (numbuf
) >= NN_NUMBUF_SZ
);
3741 dsl_pool_config_enter(dmu_objset_pool(os
), FTAG
);
3742 dmu_objset_fast_stat(os
, &dds
);
3743 dsl_pool_config_exit(dmu_objset_pool(os
), FTAG
);
3745 print_header
= B_TRUE
;
3747 if (dds
.dds_type
< DMU_OST_NUMTYPES
)
3748 type
= objset_types
[dds
.dds_type
];
3750 if (dds
.dds_type
== DMU_OST_META
) {
3751 dds
.dds_creation_txg
= TXG_INITIAL
;
3752 usedobjs
= BP_GET_FILL(os
->os_rootbp
);
3753 refdbytes
= dsl_dir_phys(os
->os_spa
->spa_dsl_pool
->dp_mos_dir
)->
3756 dmu_objset_space(os
, &refdbytes
, &scratch
, &usedobjs
, &scratch
);
3759 ASSERT3U(usedobjs
, ==, BP_GET_FILL(os
->os_rootbp
));
3761 zdb_nicenum(refdbytes
, numbuf
, sizeof (numbuf
));
3763 if (verbosity
>= 4) {
3764 (void) snprintf(blkbuf
, sizeof (blkbuf
), ", rootbp ");
3765 (void) snprintf_blkptr(blkbuf
+ strlen(blkbuf
),
3766 sizeof (blkbuf
) - strlen(blkbuf
), os
->os_rootbp
);
3771 dmu_objset_name(os
, osname
);
3773 (void) printf("Dataset %s [%s], ID %llu, cr_txg %llu, "
3774 "%s, %llu objects%s%s\n",
3775 osname
, type
, (u_longlong_t
)dmu_objset_id(os
),
3776 (u_longlong_t
)dds
.dds_creation_txg
,
3777 numbuf
, (u_longlong_t
)usedobjs
, blkbuf
,
3778 (dds
.dds_inconsistent
) ? " (inconsistent)" : "");
3780 for (i
= 0; i
< zopt_object_args
; i
++) {
3781 obj_start
= zopt_object_ranges
[i
].zor_obj_start
;
3782 obj_end
= zopt_object_ranges
[i
].zor_obj_end
;
3783 flags
= zopt_object_ranges
[i
].zor_flags
;
3786 if (object
== 0 || obj_start
== obj_end
)
3787 dump_object(os
, object
, verbosity
, &print_header
, NULL
,
3792 while ((dmu_object_next(os
, &object
, B_FALSE
, 0) == 0) &&
3793 object
<= obj_end
) {
3794 dump_object(os
, object
, verbosity
, &print_header
, NULL
,
3799 if (zopt_object_args
> 0) {
3800 (void) printf("\n");
3804 if (dump_opt
['i'] != 0 || verbosity
>= 2)
3805 dump_intent_log(dmu_objset_zil(os
));
3807 if (dmu_objset_ds(os
) != NULL
) {
3808 dsl_dataset_t
*ds
= dmu_objset_ds(os
);
3809 dump_blkptr_list(&ds
->ds_deadlist
, "Deadlist");
3810 if (dsl_deadlist_is_open(&ds
->ds_dir
->dd_livelist
) &&
3811 !dmu_objset_is_snapshot(os
)) {
3812 dump_blkptr_list(&ds
->ds_dir
->dd_livelist
, "Livelist");
3813 if (verify_dd_livelist(os
) != 0)
3814 fatal("livelist is incorrect");
3817 if (dsl_dataset_remap_deadlist_exists(ds
)) {
3818 (void) printf("ds_remap_deadlist:\n");
3819 dump_blkptr_list(&ds
->ds_remap_deadlist
, "Deadlist");
3821 count_ds_mos_objects(ds
);
3824 if (dmu_objset_ds(os
) != NULL
)
3825 dump_bookmarks(os
, verbosity
);
3830 if (BP_IS_HOLE(os
->os_rootbp
))
3833 dump_object(os
, 0, verbosity
, &print_header
, NULL
, 0);
3835 if (DMU_USERUSED_DNODE(os
) != NULL
&&
3836 DMU_USERUSED_DNODE(os
)->dn_type
!= 0) {
3837 dump_object(os
, DMU_USERUSED_OBJECT
, verbosity
, &print_header
,
3839 dump_object(os
, DMU_GROUPUSED_OBJECT
, verbosity
, &print_header
,
3843 if (DMU_PROJECTUSED_DNODE(os
) != NULL
&&
3844 DMU_PROJECTUSED_DNODE(os
)->dn_type
!= 0)
3845 dump_object(os
, DMU_PROJECTUSED_OBJECT
, verbosity
,
3846 &print_header
, NULL
, 0);
3849 while ((error
= dmu_object_next(os
, &object
, B_FALSE
, 0)) == 0) {
3850 dump_object(os
, object
, verbosity
, &print_header
, &dnode_slots
,
3853 total_slots_used
+= dnode_slots
;
3854 max_slot_used
= object
+ dnode_slots
- 1;
3857 (void) printf("\n");
3859 (void) printf(" Dnode slots:\n");
3860 (void) printf("\tTotal used: %10llu\n",
3861 (u_longlong_t
)total_slots_used
);
3862 (void) printf("\tMax used: %10llu\n",
3863 (u_longlong_t
)max_slot_used
);
3864 (void) printf("\tPercent empty: %10lf\n",
3865 (double)(max_slot_used
- total_slots_used
)*100 /
3866 (double)max_slot_used
);
3867 (void) printf("\n");
3869 if (error
!= ESRCH
) {
3870 (void) fprintf(stderr
, "dmu_object_next() = %d\n", error
);
3874 ASSERT3U(object_count
, ==, usedobjs
);
3876 if (leaked_objects
!= 0) {
3877 (void) printf("%d potentially leaked objects detected\n",
3884 dump_uberblock(uberblock_t
*ub
, const char *header
, const char *footer
)
3886 time_t timestamp
= ub
->ub_timestamp
;
3888 (void) printf("%s", header
? header
: "");
3889 (void) printf("\tmagic = %016llx\n", (u_longlong_t
)ub
->ub_magic
);
3890 (void) printf("\tversion = %llu\n", (u_longlong_t
)ub
->ub_version
);
3891 (void) printf("\ttxg = %llu\n", (u_longlong_t
)ub
->ub_txg
);
3892 (void) printf("\tguid_sum = %llu\n", (u_longlong_t
)ub
->ub_guid_sum
);
3893 (void) printf("\ttimestamp = %llu UTC = %s",
3894 (u_longlong_t
)ub
->ub_timestamp
, asctime(localtime(×tamp
)));
3896 (void) printf("\tmmp_magic = %016llx\n",
3897 (u_longlong_t
)ub
->ub_mmp_magic
);
3898 if (MMP_VALID(ub
)) {
3899 (void) printf("\tmmp_delay = %0llu\n",
3900 (u_longlong_t
)ub
->ub_mmp_delay
);
3901 if (MMP_SEQ_VALID(ub
))
3902 (void) printf("\tmmp_seq = %u\n",
3903 (unsigned int) MMP_SEQ(ub
));
3904 if (MMP_FAIL_INT_VALID(ub
))
3905 (void) printf("\tmmp_fail = %u\n",
3906 (unsigned int) MMP_FAIL_INT(ub
));
3907 if (MMP_INTERVAL_VALID(ub
))
3908 (void) printf("\tmmp_write = %u\n",
3909 (unsigned int) MMP_INTERVAL(ub
));
3910 /* After MMP_* to make summarize_uberblock_mmp cleaner */
3911 (void) printf("\tmmp_valid = %x\n",
3912 (unsigned int) ub
->ub_mmp_config
& 0xFF);
3915 if (dump_opt
['u'] >= 4) {
3916 char blkbuf
[BP_SPRINTF_LEN
];
3917 snprintf_blkptr(blkbuf
, sizeof (blkbuf
), &ub
->ub_rootbp
);
3918 (void) printf("\trootbp = %s\n", blkbuf
);
3920 (void) printf("\tcheckpoint_txg = %llu\n",
3921 (u_longlong_t
)ub
->ub_checkpoint_txg
);
3922 (void) printf("%s", footer
? footer
: "");
3926 dump_config(spa_t
*spa
)
3933 error
= dmu_bonus_hold(spa
->spa_meta_objset
,
3934 spa
->spa_config_object
, FTAG
, &db
);
3937 nvsize
= *(uint64_t *)db
->db_data
;
3938 dmu_buf_rele(db
, FTAG
);
3940 (void) printf("\nMOS Configuration:\n");
3941 dump_packed_nvlist(spa
->spa_meta_objset
,
3942 spa
->spa_config_object
, (void *)&nvsize
, 1);
3944 (void) fprintf(stderr
, "dmu_bonus_hold(%llu) failed, errno %d",
3945 (u_longlong_t
)spa
->spa_config_object
, error
);
3950 dump_cachefile(const char *cachefile
)
3953 struct stat64 statbuf
;
3957 if ((fd
= open64(cachefile
, O_RDONLY
)) < 0) {
3958 (void) printf("cannot open '%s': %s\n", cachefile
,
3963 if (fstat64(fd
, &statbuf
) != 0) {
3964 (void) printf("failed to stat '%s': %s\n", cachefile
,
3969 if ((buf
= malloc(statbuf
.st_size
)) == NULL
) {
3970 (void) fprintf(stderr
, "failed to allocate %llu bytes\n",
3971 (u_longlong_t
)statbuf
.st_size
);
3975 if (read(fd
, buf
, statbuf
.st_size
) != statbuf
.st_size
) {
3976 (void) fprintf(stderr
, "failed to read %llu bytes\n",
3977 (u_longlong_t
)statbuf
.st_size
);
3983 if (nvlist_unpack(buf
, statbuf
.st_size
, &config
, 0) != 0) {
3984 (void) fprintf(stderr
, "failed to unpack nvlist\n");
3990 dump_nvlist(config
, 0);
3992 nvlist_free(config
);
3996 * ZFS label nvlist stats
3998 typedef struct zdb_nvl_stats
{
4001 size_t zns_leaf_largest
;
4002 size_t zns_leaf_total
;
4003 nvlist_t
*zns_string
;
4004 nvlist_t
*zns_uint64
;
4005 nvlist_t
*zns_boolean
;
4009 collect_nvlist_stats(nvlist_t
*nvl
, zdb_nvl_stats_t
*stats
)
4011 nvlist_t
*list
, **array
;
4012 nvpair_t
*nvp
= NULL
;
4016 stats
->zns_list_count
++;
4018 while ((nvp
= nvlist_next_nvpair(nvl
, nvp
)) != NULL
) {
4019 name
= nvpair_name(nvp
);
4021 switch (nvpair_type(nvp
)) {
4022 case DATA_TYPE_STRING
:
4023 fnvlist_add_string(stats
->zns_string
, name
,
4024 fnvpair_value_string(nvp
));
4026 case DATA_TYPE_UINT64
:
4027 fnvlist_add_uint64(stats
->zns_uint64
, name
,
4028 fnvpair_value_uint64(nvp
));
4030 case DATA_TYPE_BOOLEAN
:
4031 fnvlist_add_boolean(stats
->zns_boolean
, name
);
4033 case DATA_TYPE_NVLIST
:
4034 if (nvpair_value_nvlist(nvp
, &list
) == 0)
4035 collect_nvlist_stats(list
, stats
);
4037 case DATA_TYPE_NVLIST_ARRAY
:
4038 if (nvpair_value_nvlist_array(nvp
, &array
, &items
) != 0)
4041 for (i
= 0; i
< items
; i
++) {
4042 collect_nvlist_stats(array
[i
], stats
);
4044 /* collect stats on leaf vdev */
4045 if (strcmp(name
, "children") == 0) {
4048 (void) nvlist_size(array
[i
], &size
,
4050 stats
->zns_leaf_total
+= size
;
4051 if (size
> stats
->zns_leaf_largest
)
4052 stats
->zns_leaf_largest
= size
;
4053 stats
->zns_leaf_count
++;
4058 (void) printf("skip type %d!\n", (int)nvpair_type(nvp
));
4064 dump_nvlist_stats(nvlist_t
*nvl
, size_t cap
)
4066 zdb_nvl_stats_t stats
= { 0 };
4067 size_t size
, sum
= 0, total
;
4070 /* requires nvlist with non-unique names for stat collection */
4071 VERIFY0(nvlist_alloc(&stats
.zns_string
, 0, 0));
4072 VERIFY0(nvlist_alloc(&stats
.zns_uint64
, 0, 0));
4073 VERIFY0(nvlist_alloc(&stats
.zns_boolean
, 0, 0));
4074 VERIFY0(nvlist_size(stats
.zns_boolean
, &noise
, NV_ENCODE_XDR
));
4076 (void) printf("\n\nZFS Label NVList Config Stats:\n");
4078 VERIFY0(nvlist_size(nvl
, &total
, NV_ENCODE_XDR
));
4079 (void) printf(" %d bytes used, %d bytes free (using %4.1f%%)\n\n",
4080 (int)total
, (int)(cap
- total
), 100.0 * total
/ cap
);
4082 collect_nvlist_stats(nvl
, &stats
);
4084 VERIFY0(nvlist_size(stats
.zns_uint64
, &size
, NV_ENCODE_XDR
));
4087 (void) printf("%12s %4d %6d bytes (%5.2f%%)\n", "integers:",
4088 (int)fnvlist_num_pairs(stats
.zns_uint64
),
4089 (int)size
, 100.0 * size
/ total
);
4091 VERIFY0(nvlist_size(stats
.zns_string
, &size
, NV_ENCODE_XDR
));
4094 (void) printf("%12s %4d %6d bytes (%5.2f%%)\n", "strings:",
4095 (int)fnvlist_num_pairs(stats
.zns_string
),
4096 (int)size
, 100.0 * size
/ total
);
4098 VERIFY0(nvlist_size(stats
.zns_boolean
, &size
, NV_ENCODE_XDR
));
4101 (void) printf("%12s %4d %6d bytes (%5.2f%%)\n", "booleans:",
4102 (int)fnvlist_num_pairs(stats
.zns_boolean
),
4103 (int)size
, 100.0 * size
/ total
);
4105 size
= total
- sum
; /* treat remainder as nvlist overhead */
4106 (void) printf("%12s %4d %6d bytes (%5.2f%%)\n\n", "nvlists:",
4107 stats
.zns_list_count
, (int)size
, 100.0 * size
/ total
);
4109 if (stats
.zns_leaf_count
> 0) {
4110 size_t average
= stats
.zns_leaf_total
/ stats
.zns_leaf_count
;
4112 (void) printf("%12s %4d %6d bytes average\n", "leaf vdevs:",
4113 stats
.zns_leaf_count
, (int)average
);
4114 (void) printf("%24d bytes largest\n",
4115 (int)stats
.zns_leaf_largest
);
4117 if (dump_opt
['l'] >= 3 && average
> 0)
4118 (void) printf(" space for %d additional leaf vdevs\n",
4119 (int)((cap
- total
) / average
));
4121 (void) printf("\n");
4123 nvlist_free(stats
.zns_string
);
4124 nvlist_free(stats
.zns_uint64
);
4125 nvlist_free(stats
.zns_boolean
);
4128 typedef struct cksum_record
{
4130 boolean_t labels
[VDEV_LABELS
];
4135 cksum_record_compare(const void *x1
, const void *x2
)
4137 const cksum_record_t
*l
= (cksum_record_t
*)x1
;
4138 const cksum_record_t
*r
= (cksum_record_t
*)x2
;
4139 int arraysize
= ARRAY_SIZE(l
->cksum
.zc_word
);
4142 for (int i
= 0; i
< arraysize
; i
++) {
4143 difference
= TREE_CMP(l
->cksum
.zc_word
[i
], r
->cksum
.zc_word
[i
]);
4148 return (difference
);
4151 static cksum_record_t
*
4152 cksum_record_alloc(zio_cksum_t
*cksum
, int l
)
4154 cksum_record_t
*rec
;
4156 rec
= umem_zalloc(sizeof (*rec
), UMEM_NOFAIL
);
4157 rec
->cksum
= *cksum
;
4158 rec
->labels
[l
] = B_TRUE
;
4163 static cksum_record_t
*
4164 cksum_record_lookup(avl_tree_t
*tree
, zio_cksum_t
*cksum
)
4166 cksum_record_t lookup
= { .cksum
= *cksum
};
4169 return (avl_find(tree
, &lookup
, &where
));
4172 static cksum_record_t
*
4173 cksum_record_insert(avl_tree_t
*tree
, zio_cksum_t
*cksum
, int l
)
4175 cksum_record_t
*rec
;
4177 rec
= cksum_record_lookup(tree
, cksum
);
4179 rec
->labels
[l
] = B_TRUE
;
4181 rec
= cksum_record_alloc(cksum
, l
);
4189 first_label(cksum_record_t
*rec
)
4191 for (int i
= 0; i
< VDEV_LABELS
; i
++)
4199 print_label_numbers(char *prefix
, cksum_record_t
*rec
)
4201 printf("%s", prefix
);
4202 for (int i
= 0; i
< VDEV_LABELS
; i
++)
4203 if (rec
->labels
[i
] == B_TRUE
)
4208 #define MAX_UBERBLOCK_COUNT (VDEV_UBERBLOCK_RING >> UBERBLOCK_SHIFT)
4210 typedef struct zdb_label
{
4212 nvlist_t
*config_nv
;
4213 cksum_record_t
*config
;
4214 cksum_record_t
*uberblocks
[MAX_UBERBLOCK_COUNT
];
4215 boolean_t header_printed
;
4216 boolean_t read_failed
;
4220 print_label_header(zdb_label_t
*label
, int l
)
4226 if (label
->header_printed
== B_TRUE
)
4229 (void) printf("------------------------------------\n");
4230 (void) printf("LABEL %d\n", l
);
4231 (void) printf("------------------------------------\n");
4233 label
->header_printed
= B_TRUE
;
4237 print_l2arc_header(void)
4239 (void) printf("------------------------------------\n");
4240 (void) printf("L2ARC device header\n");
4241 (void) printf("------------------------------------\n");
4245 print_l2arc_log_blocks(void)
4247 (void) printf("------------------------------------\n");
4248 (void) printf("L2ARC device log blocks\n");
4249 (void) printf("------------------------------------\n");
4253 dump_l2arc_log_entries(uint64_t log_entries
,
4254 l2arc_log_ent_phys_t
*le
, uint64_t i
)
4256 for (int j
= 0; j
< log_entries
; j
++) {
4257 dva_t dva
= le
[j
].le_dva
;
4258 (void) printf("lb[%4llu]\tle[%4d]\tDVA asize: %llu, "
4259 "vdev: %llu, offset: %llu\n",
4260 (u_longlong_t
)i
, j
+ 1,
4261 (u_longlong_t
)DVA_GET_ASIZE(&dva
),
4262 (u_longlong_t
)DVA_GET_VDEV(&dva
),
4263 (u_longlong_t
)DVA_GET_OFFSET(&dva
));
4264 (void) printf("|\t\t\t\tbirth: %llu\n",
4265 (u_longlong_t
)le
[j
].le_birth
);
4266 (void) printf("|\t\t\t\tlsize: %llu\n",
4267 (u_longlong_t
)L2BLK_GET_LSIZE((&le
[j
])->le_prop
));
4268 (void) printf("|\t\t\t\tpsize: %llu\n",
4269 (u_longlong_t
)L2BLK_GET_PSIZE((&le
[j
])->le_prop
));
4270 (void) printf("|\t\t\t\tcompr: %llu\n",
4271 (u_longlong_t
)L2BLK_GET_COMPRESS((&le
[j
])->le_prop
));
4272 (void) printf("|\t\t\t\tcomplevel: %llu\n",
4273 (u_longlong_t
)(&le
[j
])->le_complevel
);
4274 (void) printf("|\t\t\t\ttype: %llu\n",
4275 (u_longlong_t
)L2BLK_GET_TYPE((&le
[j
])->le_prop
));
4276 (void) printf("|\t\t\t\tprotected: %llu\n",
4277 (u_longlong_t
)L2BLK_GET_PROTECTED((&le
[j
])->le_prop
));
4278 (void) printf("|\t\t\t\tprefetch: %llu\n",
4279 (u_longlong_t
)L2BLK_GET_PREFETCH((&le
[j
])->le_prop
));
4280 (void) printf("|\t\t\t\taddress: %llu\n",
4281 (u_longlong_t
)le
[j
].le_daddr
);
4282 (void) printf("|\t\t\t\tARC state: %llu\n",
4283 (u_longlong_t
)L2BLK_GET_STATE((&le
[j
])->le_prop
));
4284 (void) printf("|\n");
4286 (void) printf("\n");
4290 dump_l2arc_log_blkptr(l2arc_log_blkptr_t lbps
)
4292 (void) printf("|\t\tdaddr: %llu\n", (u_longlong_t
)lbps
.lbp_daddr
);
4293 (void) printf("|\t\tpayload_asize: %llu\n",
4294 (u_longlong_t
)lbps
.lbp_payload_asize
);
4295 (void) printf("|\t\tpayload_start: %llu\n",
4296 (u_longlong_t
)lbps
.lbp_payload_start
);
4297 (void) printf("|\t\tlsize: %llu\n",
4298 (u_longlong_t
)L2BLK_GET_LSIZE((&lbps
)->lbp_prop
));
4299 (void) printf("|\t\tasize: %llu\n",
4300 (u_longlong_t
)L2BLK_GET_PSIZE((&lbps
)->lbp_prop
));
4301 (void) printf("|\t\tcompralgo: %llu\n",
4302 (u_longlong_t
)L2BLK_GET_COMPRESS((&lbps
)->lbp_prop
));
4303 (void) printf("|\t\tcksumalgo: %llu\n",
4304 (u_longlong_t
)L2BLK_GET_CHECKSUM((&lbps
)->lbp_prop
));
4305 (void) printf("|\n\n");
4309 dump_l2arc_log_blocks(int fd
, l2arc_dev_hdr_phys_t l2dhdr
,
4310 l2arc_dev_hdr_phys_t
*rebuild
)
4312 l2arc_log_blk_phys_t this_lb
;
4314 l2arc_log_blkptr_t lbps
[2];
4321 print_l2arc_log_blocks();
4322 bcopy((&l2dhdr
)->dh_start_lbps
, lbps
, sizeof (lbps
));
4324 dev
.l2ad_evict
= l2dhdr
.dh_evict
;
4325 dev
.l2ad_start
= l2dhdr
.dh_start
;
4326 dev
.l2ad_end
= l2dhdr
.dh_end
;
4328 if (l2dhdr
.dh_start_lbps
[0].lbp_daddr
== 0) {
4329 /* no log blocks to read */
4330 if (!dump_opt
['q']) {
4331 (void) printf("No log blocks to read\n");
4332 (void) printf("\n");
4336 dev
.l2ad_hand
= lbps
[0].lbp_daddr
+
4337 L2BLK_GET_PSIZE((&lbps
[0])->lbp_prop
);
4340 dev
.l2ad_first
= !!(l2dhdr
.dh_flags
& L2ARC_DEV_HDR_EVICT_FIRST
);
4343 if (!l2arc_log_blkptr_valid(&dev
, &lbps
[0]))
4346 /* L2BLK_GET_PSIZE returns aligned size for log blocks */
4347 asize
= L2BLK_GET_PSIZE((&lbps
[0])->lbp_prop
);
4348 if (pread64(fd
, &this_lb
, asize
, lbps
[0].lbp_daddr
) != asize
) {
4349 if (!dump_opt
['q']) {
4350 (void) printf("Error while reading next log "
4356 fletcher_4_native_varsize(&this_lb
, asize
, &cksum
);
4357 if (!ZIO_CHECKSUM_EQUAL(cksum
, lbps
[0].lbp_cksum
)) {
4359 if (!dump_opt
['q']) {
4360 (void) printf("Invalid cksum\n");
4361 dump_l2arc_log_blkptr(lbps
[0]);
4366 switch (L2BLK_GET_COMPRESS((&lbps
[0])->lbp_prop
)) {
4367 case ZIO_COMPRESS_OFF
:
4370 abd
= abd_alloc_for_io(asize
, B_TRUE
);
4371 abd_copy_from_buf_off(abd
, &this_lb
, 0, asize
);
4372 zio_decompress_data(L2BLK_GET_COMPRESS(
4373 (&lbps
[0])->lbp_prop
), abd
, &this_lb
,
4374 asize
, sizeof (this_lb
), NULL
);
4379 if (this_lb
.lb_magic
== BSWAP_64(L2ARC_LOG_BLK_MAGIC
))
4380 byteswap_uint64_array(&this_lb
, sizeof (this_lb
));
4381 if (this_lb
.lb_magic
!= L2ARC_LOG_BLK_MAGIC
) {
4383 (void) printf("Invalid log block magic\n\n");
4387 rebuild
->dh_lb_count
++;
4388 rebuild
->dh_lb_asize
+= asize
;
4389 if (dump_opt
['l'] > 1 && !dump_opt
['q']) {
4390 (void) printf("lb[%4llu]\tmagic: %llu\n",
4391 (u_longlong_t
)rebuild
->dh_lb_count
,
4392 (u_longlong_t
)this_lb
.lb_magic
);
4393 dump_l2arc_log_blkptr(lbps
[0]);
4396 if (dump_opt
['l'] > 2 && !dump_opt
['q'])
4397 dump_l2arc_log_entries(l2dhdr
.dh_log_entries
,
4399 rebuild
->dh_lb_count
);
4401 if (l2arc_range_check_overlap(lbps
[1].lbp_payload_start
,
4402 lbps
[0].lbp_payload_start
, dev
.l2ad_evict
) &&
4407 lbps
[1] = this_lb
.lb_prev_lbp
;
4410 if (!dump_opt
['q']) {
4411 (void) printf("log_blk_count:\t %llu with valid cksum\n",
4412 (u_longlong_t
)rebuild
->dh_lb_count
);
4413 (void) printf("\t\t %d with invalid cksum\n", failed
);
4414 (void) printf("log_blk_asize:\t %llu\n\n",
4415 (u_longlong_t
)rebuild
->dh_lb_asize
);
4420 dump_l2arc_header(int fd
)
4422 l2arc_dev_hdr_phys_t l2dhdr
, rebuild
;
4423 int error
= B_FALSE
;
4425 bzero(&l2dhdr
, sizeof (l2dhdr
));
4426 bzero(&rebuild
, sizeof (rebuild
));
4428 if (pread64(fd
, &l2dhdr
, sizeof (l2dhdr
),
4429 VDEV_LABEL_START_SIZE
) != sizeof (l2dhdr
)) {
4432 if (l2dhdr
.dh_magic
== BSWAP_64(L2ARC_DEV_HDR_MAGIC
))
4433 byteswap_uint64_array(&l2dhdr
, sizeof (l2dhdr
));
4435 if (l2dhdr
.dh_magic
!= L2ARC_DEV_HDR_MAGIC
)
4440 (void) printf("L2ARC device header not found\n\n");
4441 /* Do not return an error here for backward compatibility */
4443 } else if (!dump_opt
['q']) {
4444 print_l2arc_header();
4446 (void) printf(" magic: %llu\n",
4447 (u_longlong_t
)l2dhdr
.dh_magic
);
4448 (void) printf(" version: %llu\n",
4449 (u_longlong_t
)l2dhdr
.dh_version
);
4450 (void) printf(" pool_guid: %llu\n",
4451 (u_longlong_t
)l2dhdr
.dh_spa_guid
);
4452 (void) printf(" flags: %llu\n",
4453 (u_longlong_t
)l2dhdr
.dh_flags
);
4454 (void) printf(" start_lbps[0]: %llu\n",
4456 l2dhdr
.dh_start_lbps
[0].lbp_daddr
);
4457 (void) printf(" start_lbps[1]: %llu\n",
4459 l2dhdr
.dh_start_lbps
[1].lbp_daddr
);
4460 (void) printf(" log_blk_ent: %llu\n",
4461 (u_longlong_t
)l2dhdr
.dh_log_entries
);
4462 (void) printf(" start: %llu\n",
4463 (u_longlong_t
)l2dhdr
.dh_start
);
4464 (void) printf(" end: %llu\n",
4465 (u_longlong_t
)l2dhdr
.dh_end
);
4466 (void) printf(" evict: %llu\n",
4467 (u_longlong_t
)l2dhdr
.dh_evict
);
4468 (void) printf(" lb_asize_refcount: %llu\n",
4469 (u_longlong_t
)l2dhdr
.dh_lb_asize
);
4470 (void) printf(" lb_count_refcount: %llu\n",
4471 (u_longlong_t
)l2dhdr
.dh_lb_count
);
4472 (void) printf(" trim_action_time: %llu\n",
4473 (u_longlong_t
)l2dhdr
.dh_trim_action_time
);
4474 (void) printf(" trim_state: %llu\n\n",
4475 (u_longlong_t
)l2dhdr
.dh_trim_state
);
4478 dump_l2arc_log_blocks(fd
, l2dhdr
, &rebuild
);
4480 * The total aligned size of log blocks and the number of log blocks
4481 * reported in the header of the device may be less than what zdb
4482 * reports by dump_l2arc_log_blocks() which emulates l2arc_rebuild().
4483 * This happens because dump_l2arc_log_blocks() lacks the memory
4484 * pressure valve that l2arc_rebuild() has. Thus, if we are on a system
4485 * with low memory, l2arc_rebuild will exit prematurely and dh_lb_asize
4486 * and dh_lb_count will be lower to begin with than what exists on the
4487 * device. This is normal and zdb should not exit with an error. The
4488 * opposite case should never happen though, the values reported in the
4489 * header should never be higher than what dump_l2arc_log_blocks() and
4490 * l2arc_rebuild() report. If this happens there is a leak in the
4491 * accounting of log blocks.
4493 if (l2dhdr
.dh_lb_asize
> rebuild
.dh_lb_asize
||
4494 l2dhdr
.dh_lb_count
> rebuild
.dh_lb_count
)
4501 dump_config_from_label(zdb_label_t
*label
, size_t buflen
, int l
)
4506 if ((dump_opt
['l'] < 3) && (first_label(label
->config
) != l
))
4509 print_label_header(label
, l
);
4510 dump_nvlist(label
->config_nv
, 4);
4511 print_label_numbers(" labels = ", label
->config
);
4513 if (dump_opt
['l'] >= 2)
4514 dump_nvlist_stats(label
->config_nv
, buflen
);
4517 #define ZDB_MAX_UB_HEADER_SIZE 32
4520 dump_label_uberblocks(zdb_label_t
*label
, uint64_t ashift
, int label_num
)
4524 char header
[ZDB_MAX_UB_HEADER_SIZE
];
4526 vd
.vdev_ashift
= ashift
;
4529 for (int i
= 0; i
< VDEV_UBERBLOCK_COUNT(&vd
); i
++) {
4530 uint64_t uoff
= VDEV_UBERBLOCK_OFFSET(&vd
, i
);
4531 uberblock_t
*ub
= (void *)((char *)&label
->label
+ uoff
);
4532 cksum_record_t
*rec
= label
->uberblocks
[i
];
4535 if (dump_opt
['u'] >= 2) {
4536 print_label_header(label
, label_num
);
4537 (void) printf(" Uberblock[%d] invalid\n", i
);
4542 if ((dump_opt
['u'] < 3) && (first_label(rec
) != label_num
))
4545 if ((dump_opt
['u'] < 4) &&
4546 (ub
->ub_mmp_magic
== MMP_MAGIC
) && ub
->ub_mmp_delay
&&
4547 (i
>= VDEV_UBERBLOCK_COUNT(&vd
) - MMP_BLOCKS_PER_LABEL
))
4550 print_label_header(label
, label_num
);
4551 (void) snprintf(header
, ZDB_MAX_UB_HEADER_SIZE
,
4552 " Uberblock[%d]\n", i
);
4553 dump_uberblock(ub
, header
, "");
4554 print_label_numbers(" labels = ", rec
);
4558 static char curpath
[PATH_MAX
];
4561 * Iterate through the path components, recursively passing
4562 * current one's obj and remaining path until we find the obj
4566 dump_path_impl(objset_t
*os
, uint64_t obj
, char *name
, uint64_t *retobj
)
4569 boolean_t header
= B_TRUE
;
4573 dmu_object_info_t doi
;
4575 if ((s
= strchr(name
, '/')) != NULL
)
4577 err
= zap_lookup(os
, obj
, name
, 8, 1, &child_obj
);
4579 (void) strlcat(curpath
, name
, sizeof (curpath
));
4582 (void) fprintf(stderr
, "failed to lookup %s: %s\n",
4583 curpath
, strerror(err
));
4587 child_obj
= ZFS_DIRENT_OBJ(child_obj
);
4588 err
= sa_buf_hold(os
, child_obj
, FTAG
, &db
);
4590 (void) fprintf(stderr
,
4591 "failed to get SA dbuf for obj %llu: %s\n",
4592 (u_longlong_t
)child_obj
, strerror(err
));
4595 dmu_object_info_from_db(db
, &doi
);
4596 sa_buf_rele(db
, FTAG
);
4598 if (doi
.doi_bonus_type
!= DMU_OT_SA
&&
4599 doi
.doi_bonus_type
!= DMU_OT_ZNODE
) {
4600 (void) fprintf(stderr
, "invalid bonus type %d for obj %llu\n",
4601 doi
.doi_bonus_type
, (u_longlong_t
)child_obj
);
4605 if (dump_opt
['v'] > 6) {
4606 (void) printf("obj=%llu %s type=%d bonustype=%d\n",
4607 (u_longlong_t
)child_obj
, curpath
, doi
.doi_type
,
4608 doi
.doi_bonus_type
);
4611 (void) strlcat(curpath
, "/", sizeof (curpath
));
4613 switch (doi
.doi_type
) {
4614 case DMU_OT_DIRECTORY_CONTENTS
:
4615 if (s
!= NULL
&& *(s
+ 1) != '\0')
4616 return (dump_path_impl(os
, child_obj
, s
+ 1, retobj
));
4618 case DMU_OT_PLAIN_FILE_CONTENTS
:
4619 if (retobj
!= NULL
) {
4620 *retobj
= child_obj
;
4622 dump_object(os
, child_obj
, dump_opt
['v'], &header
,
4627 (void) fprintf(stderr
, "object %llu has non-file/directory "
4628 "type %d\n", (u_longlong_t
)obj
, doi
.doi_type
);
4636 * Dump the blocks for the object specified by path inside the dataset.
4639 dump_path(char *ds
, char *path
, uint64_t *retobj
)
4645 err
= open_objset(ds
, FTAG
, &os
);
4649 err
= zap_lookup(os
, MASTER_NODE_OBJ
, ZFS_ROOT_OBJ
, 8, 1, &root_obj
);
4651 (void) fprintf(stderr
, "can't lookup root znode: %s\n",
4653 close_objset(os
, FTAG
);
4657 (void) snprintf(curpath
, sizeof (curpath
), "dataset=%s path=/", ds
);
4659 err
= dump_path_impl(os
, root_obj
, path
, retobj
);
4661 close_objset(os
, FTAG
);
4666 zdb_copy_object(objset_t
*os
, uint64_t srcobj
, char *destfile
)
4669 uint64_t size
, readsize
, oursize
, offset
;
4673 (void) printf("Copying object %" PRIu64
" to file %s\n", srcobj
,
4676 VERIFY3P(os
, ==, sa_os
);
4677 if ((err
= sa_handle_get(os
, srcobj
, NULL
, SA_HDL_PRIVATE
, &hdl
))) {
4678 (void) printf("Failed to get handle for SA znode\n");
4681 if ((err
= sa_lookup(hdl
, sa_attr_table
[ZPL_SIZE
], &size
, 8))) {
4682 (void) sa_handle_destroy(hdl
);
4685 (void) sa_handle_destroy(hdl
);
4687 (void) printf("Object %" PRIu64
" is %" PRIu64
" bytes\n", srcobj
,
4693 int fd
= open(destfile
, O_WRONLY
| O_CREAT
| O_TRUNC
, 0644);
4695 * We cap the size at 1 mebibyte here to prevent
4696 * allocation failures and nigh-infinite printing if the
4697 * object is extremely large.
4699 oursize
= MIN(size
, 1 << 20);
4701 char *buf
= kmem_alloc(oursize
, KM_NOSLEEP
);
4706 while (offset
< size
) {
4707 readsize
= MIN(size
- offset
, 1 << 20);
4708 err
= dmu_read(os
, srcobj
, offset
, readsize
, buf
, 0);
4710 (void) printf("got error %u from dmu_read\n", err
);
4711 kmem_free(buf
, oursize
);
4714 if (dump_opt
['v'] > 3) {
4715 (void) printf("Read offset=%" PRIu64
" size=%" PRIu64
4716 " error=%d\n", offset
, readsize
, err
);
4719 writesize
= write(fd
, buf
, readsize
);
4720 if (writesize
< 0) {
4723 } else if (writesize
!= readsize
) {
4724 /* Incomplete write */
4725 (void) fprintf(stderr
, "Short write, only wrote %llu of"
4726 " %" PRIu64
" bytes, exiting...\n",
4727 (u_longlong_t
)writesize
, readsize
);
4737 kmem_free(buf
, oursize
);
4743 dump_label(const char *dev
)
4745 char path
[MAXPATHLEN
];
4746 zdb_label_t labels
[VDEV_LABELS
];
4747 uint64_t psize
, ashift
, l2cache
;
4748 struct stat64 statbuf
;
4749 boolean_t config_found
= B_FALSE
;
4750 boolean_t error
= B_FALSE
;
4751 boolean_t read_l2arc_header
= B_FALSE
;
4752 avl_tree_t config_tree
;
4753 avl_tree_t uberblock_tree
;
4754 void *node
, *cookie
;
4757 bzero(labels
, sizeof (labels
));
4760 * Check if we were given absolute path and use it as is.
4761 * Otherwise if the provided vdev name doesn't point to a file,
4762 * try prepending expected disk paths and partition numbers.
4764 (void) strlcpy(path
, dev
, sizeof (path
));
4765 if (dev
[0] != '/' && stat64(path
, &statbuf
) != 0) {
4768 error
= zfs_resolve_shortname(dev
, path
, MAXPATHLEN
);
4769 if (error
== 0 && zfs_dev_is_whole_disk(path
)) {
4770 if (zfs_append_partition(path
, MAXPATHLEN
) == -1)
4774 if (error
|| (stat64(path
, &statbuf
) != 0)) {
4775 (void) printf("failed to find device %s, try "
4776 "specifying absolute path instead\n", dev
);
4781 if ((fd
= open64(path
, O_RDONLY
)) < 0) {
4782 (void) printf("cannot open '%s': %s\n", path
, strerror(errno
));
4786 if (fstat64_blk(fd
, &statbuf
) != 0) {
4787 (void) printf("failed to stat '%s': %s\n", path
,
4793 if (S_ISBLK(statbuf
.st_mode
) && zfs_dev_flush(fd
) != 0)
4794 (void) printf("failed to invalidate cache '%s' : %s\n", path
,
4797 avl_create(&config_tree
, cksum_record_compare
,
4798 sizeof (cksum_record_t
), offsetof(cksum_record_t
, link
));
4799 avl_create(&uberblock_tree
, cksum_record_compare
,
4800 sizeof (cksum_record_t
), offsetof(cksum_record_t
, link
));
4802 psize
= statbuf
.st_size
;
4803 psize
= P2ALIGN(psize
, (uint64_t)sizeof (vdev_label_t
));
4804 ashift
= SPA_MINBLOCKSHIFT
;
4807 * 1. Read the label from disk
4808 * 2. Unpack the configuration and insert in config tree.
4809 * 3. Traverse all uberblocks and insert in uberblock tree.
4811 for (int l
= 0; l
< VDEV_LABELS
; l
++) {
4812 zdb_label_t
*label
= &labels
[l
];
4813 char *buf
= label
->label
.vl_vdev_phys
.vp_nvlist
;
4814 size_t buflen
= sizeof (label
->label
.vl_vdev_phys
.vp_nvlist
);
4816 cksum_record_t
*rec
;
4820 if (pread64(fd
, &label
->label
, sizeof (label
->label
),
4821 vdev_label_offset(psize
, l
, 0)) != sizeof (label
->label
)) {
4823 (void) printf("failed to read label %d\n", l
);
4824 label
->read_failed
= B_TRUE
;
4829 label
->read_failed
= B_FALSE
;
4831 if (nvlist_unpack(buf
, buflen
, &config
, 0) == 0) {
4832 nvlist_t
*vdev_tree
= NULL
;
4835 if ((nvlist_lookup_nvlist(config
,
4836 ZPOOL_CONFIG_VDEV_TREE
, &vdev_tree
) != 0) ||
4837 (nvlist_lookup_uint64(vdev_tree
,
4838 ZPOOL_CONFIG_ASHIFT
, &ashift
) != 0))
4839 ashift
= SPA_MINBLOCKSHIFT
;
4841 if (nvlist_size(config
, &size
, NV_ENCODE_XDR
) != 0)
4844 /* If the device is a cache device clear the header. */
4845 if (!read_l2arc_header
) {
4846 if (nvlist_lookup_uint64(config
,
4847 ZPOOL_CONFIG_POOL_STATE
, &l2cache
) == 0 &&
4848 l2cache
== POOL_STATE_L2CACHE
) {
4849 read_l2arc_header
= B_TRUE
;
4853 fletcher_4_native_varsize(buf
, size
, &cksum
);
4854 rec
= cksum_record_insert(&config_tree
, &cksum
, l
);
4856 label
->config
= rec
;
4857 label
->config_nv
= config
;
4858 config_found
= B_TRUE
;
4863 vd
.vdev_ashift
= ashift
;
4866 for (int i
= 0; i
< VDEV_UBERBLOCK_COUNT(&vd
); i
++) {
4867 uint64_t uoff
= VDEV_UBERBLOCK_OFFSET(&vd
, i
);
4868 uberblock_t
*ub
= (void *)((char *)label
+ uoff
);
4870 if (uberblock_verify(ub
))
4873 fletcher_4_native_varsize(ub
, sizeof (*ub
), &cksum
);
4874 rec
= cksum_record_insert(&uberblock_tree
, &cksum
, l
);
4876 label
->uberblocks
[i
] = rec
;
4881 * Dump the label and uberblocks.
4883 for (int l
= 0; l
< VDEV_LABELS
; l
++) {
4884 zdb_label_t
*label
= &labels
[l
];
4885 size_t buflen
= sizeof (label
->label
.vl_vdev_phys
.vp_nvlist
);
4887 if (label
->read_failed
== B_TRUE
)
4890 if (label
->config_nv
) {
4891 dump_config_from_label(label
, buflen
, l
);
4894 (void) printf("failed to unpack label %d\n", l
);
4898 dump_label_uberblocks(label
, ashift
, l
);
4900 nvlist_free(label
->config_nv
);
4904 * Dump the L2ARC header, if existent.
4906 if (read_l2arc_header
)
4907 error
|= dump_l2arc_header(fd
);
4910 while ((node
= avl_destroy_nodes(&config_tree
, &cookie
)) != NULL
)
4911 umem_free(node
, sizeof (cksum_record_t
));
4914 while ((node
= avl_destroy_nodes(&uberblock_tree
, &cookie
)) != NULL
)
4915 umem_free(node
, sizeof (cksum_record_t
));
4917 avl_destroy(&config_tree
);
4918 avl_destroy(&uberblock_tree
);
4922 return (config_found
== B_FALSE
? 2 :
4923 (error
== B_TRUE
? 1 : 0));
4926 static uint64_t dataset_feature_count
[SPA_FEATURES
];
4927 static uint64_t global_feature_count
[SPA_FEATURES
];
4928 static uint64_t remap_deadlist_count
= 0;
4932 dump_one_objset(const char *dsname
, void *arg
)
4938 error
= open_objset(dsname
, FTAG
, &os
);
4942 for (f
= 0; f
< SPA_FEATURES
; f
++) {
4943 if (!dsl_dataset_feature_is_active(dmu_objset_ds(os
), f
))
4945 ASSERT(spa_feature_table
[f
].fi_flags
&
4946 ZFEATURE_FLAG_PER_DATASET
);
4947 dataset_feature_count
[f
]++;
4950 if (dsl_dataset_remap_deadlist_exists(dmu_objset_ds(os
))) {
4951 remap_deadlist_count
++;
4954 for (dsl_bookmark_node_t
*dbn
=
4955 avl_first(&dmu_objset_ds(os
)->ds_bookmarks
); dbn
!= NULL
;
4956 dbn
= AVL_NEXT(&dmu_objset_ds(os
)->ds_bookmarks
, dbn
)) {
4957 mos_obj_refd(dbn
->dbn_phys
.zbm_redaction_obj
);
4958 if (dbn
->dbn_phys
.zbm_redaction_obj
!= 0)
4959 global_feature_count
[SPA_FEATURE_REDACTION_BOOKMARKS
]++;
4960 if (dbn
->dbn_phys
.zbm_flags
& ZBM_FLAG_HAS_FBN
)
4961 global_feature_count
[SPA_FEATURE_BOOKMARK_WRITTEN
]++;
4964 if (dsl_deadlist_is_open(&dmu_objset_ds(os
)->ds_dir
->dd_livelist
) &&
4965 !dmu_objset_is_snapshot(os
)) {
4966 global_feature_count
[SPA_FEATURE_LIVELIST
]++;
4970 close_objset(os
, FTAG
);
4971 fuid_table_destroy();
4978 #define PSIZE_HISTO_SIZE (SPA_OLD_MAXBLOCKSIZE / SPA_MINBLOCKSIZE + 2)
4979 typedef struct zdb_blkstats
{
4985 uint64_t zb_ditto_samevdev
;
4986 uint64_t zb_ditto_same_ms
;
4987 uint64_t zb_psize_histogram
[PSIZE_HISTO_SIZE
];
4991 * Extended object types to report deferred frees and dedup auto-ditto blocks.
4993 #define ZDB_OT_DEFERRED (DMU_OT_NUMTYPES + 0)
4994 #define ZDB_OT_DITTO (DMU_OT_NUMTYPES + 1)
4995 #define ZDB_OT_OTHER (DMU_OT_NUMTYPES + 2)
4996 #define ZDB_OT_TOTAL (DMU_OT_NUMTYPES + 3)
4998 static const char *zdb_ot_extname
[] = {
5005 #define ZB_TOTAL DN_MAX_LEVELS
5006 #define SPA_MAX_FOR_16M (SPA_MAXBLOCKSHIFT+1)
5008 typedef struct zdb_cb
{
5009 zdb_blkstats_t zcb_type
[ZB_TOTAL
+ 1][ZDB_OT_TOTAL
+ 1];
5010 uint64_t zcb_removing_size
;
5011 uint64_t zcb_checkpoint_size
;
5012 uint64_t zcb_dedup_asize
;
5013 uint64_t zcb_dedup_blocks
;
5014 uint64_t zcb_psize_count
[SPA_MAX_FOR_16M
];
5015 uint64_t zcb_lsize_count
[SPA_MAX_FOR_16M
];
5016 uint64_t zcb_asize_count
[SPA_MAX_FOR_16M
];
5017 uint64_t zcb_psize_len
[SPA_MAX_FOR_16M
];
5018 uint64_t zcb_lsize_len
[SPA_MAX_FOR_16M
];
5019 uint64_t zcb_asize_len
[SPA_MAX_FOR_16M
];
5020 uint64_t zcb_psize_total
;
5021 uint64_t zcb_lsize_total
;
5022 uint64_t zcb_asize_total
;
5023 uint64_t zcb_embedded_blocks
[NUM_BP_EMBEDDED_TYPES
];
5024 uint64_t zcb_embedded_histogram
[NUM_BP_EMBEDDED_TYPES
]
5025 [BPE_PAYLOAD_SIZE
+ 1];
5027 hrtime_t zcb_lastprint
;
5028 uint64_t zcb_totalasize
;
5029 uint64_t zcb_errors
[256];
5033 uint32_t **zcb_vd_obsolete_counts
;
5036 /* test if two DVA offsets from same vdev are within the same metaslab */
5038 same_metaslab(spa_t
*spa
, uint64_t vdev
, uint64_t off1
, uint64_t off2
)
5040 vdev_t
*vd
= vdev_lookup_top(spa
, vdev
);
5041 uint64_t ms_shift
= vd
->vdev_ms_shift
;
5043 return ((off1
>> ms_shift
) == (off2
>> ms_shift
));
5047 * Used to simplify reporting of the histogram data.
5049 typedef struct one_histo
{
5053 uint64_t cumulative
;
5057 * The number of separate histograms processed for psize, lsize and asize.
5062 * This routine will create a fixed column size output of three different
5063 * histograms showing by blocksize of 512 - 2^ SPA_MAX_FOR_16M
5064 * the count, length and cumulative length of the psize, lsize and
5067 * All three types of blocks are listed on a single line
5069 * By default the table is printed in nicenumber format (e.g. 123K) but
5070 * if the '-P' parameter is specified then the full raw number (parseable)
5074 dump_size_histograms(zdb_cb_t
*zcb
)
5077 * A temporary buffer that allows us to convert a number into
5078 * a string using zdb_nicenumber to allow either raw or human
5079 * readable numbers to be output.
5084 * Define titles which are used in the headers of the tables
5085 * printed by this routine.
5087 const char blocksize_title1
[] = "block";
5088 const char blocksize_title2
[] = "size";
5089 const char count_title
[] = "Count";
5090 const char length_title
[] = "Size";
5091 const char cumulative_title
[] = "Cum.";
5094 * Setup the histogram arrays (psize, lsize, and asize).
5096 one_histo_t parm_histo
[NUM_HISTO
];
5098 parm_histo
[0].name
= "psize";
5099 parm_histo
[0].count
= zcb
->zcb_psize_count
;
5100 parm_histo
[0].len
= zcb
->zcb_psize_len
;
5101 parm_histo
[0].cumulative
= 0;
5103 parm_histo
[1].name
= "lsize";
5104 parm_histo
[1].count
= zcb
->zcb_lsize_count
;
5105 parm_histo
[1].len
= zcb
->zcb_lsize_len
;
5106 parm_histo
[1].cumulative
= 0;
5108 parm_histo
[2].name
= "asize";
5109 parm_histo
[2].count
= zcb
->zcb_asize_count
;
5110 parm_histo
[2].len
= zcb
->zcb_asize_len
;
5111 parm_histo
[2].cumulative
= 0;
5114 (void) printf("\nBlock Size Histogram\n");
5116 * Print the first line titles
5119 (void) printf("\n%s\t", blocksize_title1
);
5121 (void) printf("\n%7s ", blocksize_title1
);
5123 for (int j
= 0; j
< NUM_HISTO
; j
++) {
5124 if (dump_opt
['P']) {
5125 if (j
< NUM_HISTO
- 1) {
5126 (void) printf("%s\t\t\t", parm_histo
[j
].name
);
5128 /* Don't print trailing spaces */
5129 (void) printf(" %s", parm_histo
[j
].name
);
5132 if (j
< NUM_HISTO
- 1) {
5133 /* Left aligned strings in the output */
5134 (void) printf("%-7s ",
5135 parm_histo
[j
].name
);
5137 /* Don't print trailing spaces */
5138 (void) printf("%s", parm_histo
[j
].name
);
5142 (void) printf("\n");
5145 * Print the second line titles
5147 if (dump_opt
['P']) {
5148 (void) printf("%s\t", blocksize_title2
);
5150 (void) printf("%7s ", blocksize_title2
);
5153 for (int i
= 0; i
< NUM_HISTO
; i
++) {
5154 if (dump_opt
['P']) {
5155 (void) printf("%s\t%s\t%s\t",
5156 count_title
, length_title
, cumulative_title
);
5158 (void) printf("%7s%7s%7s",
5159 count_title
, length_title
, cumulative_title
);
5162 (void) printf("\n");
5167 for (int i
= SPA_MINBLOCKSHIFT
; i
< SPA_MAX_FOR_16M
; i
++) {
5170 * Print the first column showing the blocksize
5172 zdb_nicenum((1ULL << i
), numbuf
, sizeof (numbuf
));
5174 if (dump_opt
['P']) {
5175 printf("%s", numbuf
);
5177 printf("%7s:", numbuf
);
5181 * Print the remaining set of 3 columns per size:
5182 * for psize, lsize and asize
5184 for (int j
= 0; j
< NUM_HISTO
; j
++) {
5185 parm_histo
[j
].cumulative
+= parm_histo
[j
].len
[i
];
5187 zdb_nicenum(parm_histo
[j
].count
[i
],
5188 numbuf
, sizeof (numbuf
));
5190 (void) printf("\t%s", numbuf
);
5192 (void) printf("%7s", numbuf
);
5194 zdb_nicenum(parm_histo
[j
].len
[i
],
5195 numbuf
, sizeof (numbuf
));
5197 (void) printf("\t%s", numbuf
);
5199 (void) printf("%7s", numbuf
);
5201 zdb_nicenum(parm_histo
[j
].cumulative
,
5202 numbuf
, sizeof (numbuf
));
5204 (void) printf("\t%s", numbuf
);
5206 (void) printf("%7s", numbuf
);
5208 (void) printf("\n");
5213 zdb_count_block(zdb_cb_t
*zcb
, zilog_t
*zilog
, const blkptr_t
*bp
,
5214 dmu_object_type_t type
)
5216 uint64_t refcnt
= 0;
5219 ASSERT(type
< ZDB_OT_TOTAL
);
5221 if (zilog
&& zil_bp_tree_add(zilog
, bp
) != 0)
5224 spa_config_enter(zcb
->zcb_spa
, SCL_CONFIG
, FTAG
, RW_READER
);
5226 for (i
= 0; i
< 4; i
++) {
5227 int l
= (i
< 2) ? BP_GET_LEVEL(bp
) : ZB_TOTAL
;
5228 int t
= (i
& 1) ? type
: ZDB_OT_TOTAL
;
5230 zdb_blkstats_t
*zb
= &zcb
->zcb_type
[l
][t
];
5232 zb
->zb_asize
+= BP_GET_ASIZE(bp
);
5233 zb
->zb_lsize
+= BP_GET_LSIZE(bp
);
5234 zb
->zb_psize
+= BP_GET_PSIZE(bp
);
5238 * The histogram is only big enough to record blocks up to
5239 * SPA_OLD_MAXBLOCKSIZE; larger blocks go into the last,
5242 unsigned idx
= BP_GET_PSIZE(bp
) >> SPA_MINBLOCKSHIFT
;
5243 idx
= MIN(idx
, SPA_OLD_MAXBLOCKSIZE
/ SPA_MINBLOCKSIZE
+ 1);
5244 zb
->zb_psize_histogram
[idx
]++;
5246 zb
->zb_gangs
+= BP_COUNT_GANG(bp
);
5248 switch (BP_GET_NDVAS(bp
)) {
5250 if (DVA_GET_VDEV(&bp
->blk_dva
[0]) ==
5251 DVA_GET_VDEV(&bp
->blk_dva
[1])) {
5252 zb
->zb_ditto_samevdev
++;
5254 if (same_metaslab(zcb
->zcb_spa
,
5255 DVA_GET_VDEV(&bp
->blk_dva
[0]),
5256 DVA_GET_OFFSET(&bp
->blk_dva
[0]),
5257 DVA_GET_OFFSET(&bp
->blk_dva
[1])))
5258 zb
->zb_ditto_same_ms
++;
5262 equal
= (DVA_GET_VDEV(&bp
->blk_dva
[0]) ==
5263 DVA_GET_VDEV(&bp
->blk_dva
[1])) +
5264 (DVA_GET_VDEV(&bp
->blk_dva
[0]) ==
5265 DVA_GET_VDEV(&bp
->blk_dva
[2])) +
5266 (DVA_GET_VDEV(&bp
->blk_dva
[1]) ==
5267 DVA_GET_VDEV(&bp
->blk_dva
[2]));
5269 zb
->zb_ditto_samevdev
++;
5271 if (DVA_GET_VDEV(&bp
->blk_dva
[0]) ==
5272 DVA_GET_VDEV(&bp
->blk_dva
[1]) &&
5273 same_metaslab(zcb
->zcb_spa
,
5274 DVA_GET_VDEV(&bp
->blk_dva
[0]),
5275 DVA_GET_OFFSET(&bp
->blk_dva
[0]),
5276 DVA_GET_OFFSET(&bp
->blk_dva
[1])))
5277 zb
->zb_ditto_same_ms
++;
5278 else if (DVA_GET_VDEV(&bp
->blk_dva
[0]) ==
5279 DVA_GET_VDEV(&bp
->blk_dva
[2]) &&
5280 same_metaslab(zcb
->zcb_spa
,
5281 DVA_GET_VDEV(&bp
->blk_dva
[0]),
5282 DVA_GET_OFFSET(&bp
->blk_dva
[0]),
5283 DVA_GET_OFFSET(&bp
->blk_dva
[2])))
5284 zb
->zb_ditto_same_ms
++;
5285 else if (DVA_GET_VDEV(&bp
->blk_dva
[1]) ==
5286 DVA_GET_VDEV(&bp
->blk_dva
[2]) &&
5287 same_metaslab(zcb
->zcb_spa
,
5288 DVA_GET_VDEV(&bp
->blk_dva
[1]),
5289 DVA_GET_OFFSET(&bp
->blk_dva
[1]),
5290 DVA_GET_OFFSET(&bp
->blk_dva
[2])))
5291 zb
->zb_ditto_same_ms
++;
5297 spa_config_exit(zcb
->zcb_spa
, SCL_CONFIG
, FTAG
);
5299 if (BP_IS_EMBEDDED(bp
)) {
5300 zcb
->zcb_embedded_blocks
[BPE_GET_ETYPE(bp
)]++;
5301 zcb
->zcb_embedded_histogram
[BPE_GET_ETYPE(bp
)]
5302 [BPE_GET_PSIZE(bp
)]++;
5306 * The binning histogram bins by powers of two up to
5307 * SPA_MAXBLOCKSIZE rather than creating bins for
5308 * every possible blocksize found in the pool.
5310 int bin
= highbit64(BP_GET_PSIZE(bp
)) - 1;
5312 zcb
->zcb_psize_count
[bin
]++;
5313 zcb
->zcb_psize_len
[bin
] += BP_GET_PSIZE(bp
);
5314 zcb
->zcb_psize_total
+= BP_GET_PSIZE(bp
);
5316 bin
= highbit64(BP_GET_LSIZE(bp
)) - 1;
5318 zcb
->zcb_lsize_count
[bin
]++;
5319 zcb
->zcb_lsize_len
[bin
] += BP_GET_LSIZE(bp
);
5320 zcb
->zcb_lsize_total
+= BP_GET_LSIZE(bp
);
5322 bin
= highbit64(BP_GET_ASIZE(bp
)) - 1;
5324 zcb
->zcb_asize_count
[bin
]++;
5325 zcb
->zcb_asize_len
[bin
] += BP_GET_ASIZE(bp
);
5326 zcb
->zcb_asize_total
+= BP_GET_ASIZE(bp
);
5331 if (BP_GET_DEDUP(bp
)) {
5335 ddt
= ddt_select(zcb
->zcb_spa
, bp
);
5337 dde
= ddt_lookup(ddt
, bp
, B_FALSE
);
5342 ddt_phys_t
*ddp
= ddt_phys_select(dde
, bp
);
5343 ddt_phys_decref(ddp
);
5344 refcnt
= ddp
->ddp_refcnt
;
5345 if (ddt_phys_total_refcnt(dde
) == 0)
5346 ddt_remove(ddt
, dde
);
5351 VERIFY3U(zio_wait(zio_claim(NULL
, zcb
->zcb_spa
,
5352 refcnt
? 0 : spa_min_claim_txg(zcb
->zcb_spa
),
5353 bp
, NULL
, NULL
, ZIO_FLAG_CANFAIL
)), ==, 0);
5357 zdb_blkptr_done(zio_t
*zio
)
5359 spa_t
*spa
= zio
->io_spa
;
5360 blkptr_t
*bp
= zio
->io_bp
;
5361 int ioerr
= zio
->io_error
;
5362 zdb_cb_t
*zcb
= zio
->io_private
;
5363 zbookmark_phys_t
*zb
= &zio
->io_bookmark
;
5365 mutex_enter(&spa
->spa_scrub_lock
);
5366 spa
->spa_load_verify_bytes
-= BP_GET_PSIZE(bp
);
5367 cv_broadcast(&spa
->spa_scrub_io_cv
);
5369 if (ioerr
&& !(zio
->io_flags
& ZIO_FLAG_SPECULATIVE
)) {
5370 char blkbuf
[BP_SPRINTF_LEN
];
5372 zcb
->zcb_haderrors
= 1;
5373 zcb
->zcb_errors
[ioerr
]++;
5375 if (dump_opt
['b'] >= 2)
5376 snprintf_blkptr(blkbuf
, sizeof (blkbuf
), bp
);
5380 (void) printf("zdb_blkptr_cb: "
5381 "Got error %d reading "
5382 "<%llu, %llu, %lld, %llx> %s -- skipping\n",
5384 (u_longlong_t
)zb
->zb_objset
,
5385 (u_longlong_t
)zb
->zb_object
,
5386 (u_longlong_t
)zb
->zb_level
,
5387 (u_longlong_t
)zb
->zb_blkid
,
5390 mutex_exit(&spa
->spa_scrub_lock
);
5392 abd_free(zio
->io_abd
);
5396 zdb_blkptr_cb(spa_t
*spa
, zilog_t
*zilog
, const blkptr_t
*bp
,
5397 const zbookmark_phys_t
*zb
, const dnode_phys_t
*dnp
, void *arg
)
5399 zdb_cb_t
*zcb
= arg
;
5400 dmu_object_type_t type
;
5401 boolean_t is_metadata
;
5403 if (zb
->zb_level
== ZB_DNODE_LEVEL
)
5406 if (dump_opt
['b'] >= 5 && bp
->blk_birth
> 0) {
5407 char blkbuf
[BP_SPRINTF_LEN
];
5408 snprintf_blkptr(blkbuf
, sizeof (blkbuf
), bp
);
5409 (void) printf("objset %llu object %llu "
5410 "level %lld offset 0x%llx %s\n",
5411 (u_longlong_t
)zb
->zb_objset
,
5412 (u_longlong_t
)zb
->zb_object
,
5413 (longlong_t
)zb
->zb_level
,
5414 (u_longlong_t
)blkid2offset(dnp
, bp
, zb
),
5418 if (BP_IS_HOLE(bp
) || BP_IS_REDACTED(bp
))
5421 type
= BP_GET_TYPE(bp
);
5423 zdb_count_block(zcb
, zilog
, bp
,
5424 (type
& DMU_OT_NEWTYPE
) ? ZDB_OT_OTHER
: type
);
5426 is_metadata
= (BP_GET_LEVEL(bp
) != 0 || DMU_OT_IS_METADATA(type
));
5428 if (!BP_IS_EMBEDDED(bp
) &&
5429 (dump_opt
['c'] > 1 || (dump_opt
['c'] && is_metadata
))) {
5430 size_t size
= BP_GET_PSIZE(bp
);
5431 abd_t
*abd
= abd_alloc(size
, B_FALSE
);
5432 int flags
= ZIO_FLAG_CANFAIL
| ZIO_FLAG_SCRUB
| ZIO_FLAG_RAW
;
5434 /* If it's an intent log block, failure is expected. */
5435 if (zb
->zb_level
== ZB_ZIL_LEVEL
)
5436 flags
|= ZIO_FLAG_SPECULATIVE
;
5438 mutex_enter(&spa
->spa_scrub_lock
);
5439 while (spa
->spa_load_verify_bytes
> max_inflight_bytes
)
5440 cv_wait(&spa
->spa_scrub_io_cv
, &spa
->spa_scrub_lock
);
5441 spa
->spa_load_verify_bytes
+= size
;
5442 mutex_exit(&spa
->spa_scrub_lock
);
5444 zio_nowait(zio_read(NULL
, spa
, bp
, abd
, size
,
5445 zdb_blkptr_done
, zcb
, ZIO_PRIORITY_ASYNC_READ
, flags
, zb
));
5448 zcb
->zcb_readfails
= 0;
5450 /* only call gethrtime() every 100 blocks */
5457 if (dump_opt
['b'] < 5 && gethrtime() > zcb
->zcb_lastprint
+ NANOSEC
) {
5458 uint64_t now
= gethrtime();
5460 uint64_t bytes
= zcb
->zcb_type
[ZB_TOTAL
][ZDB_OT_TOTAL
].zb_asize
;
5462 1 + bytes
/ (1 + ((now
- zcb
->zcb_start
) / 1000 / 1000));
5464 (zcb
->zcb_totalasize
- bytes
) / 1024 / kb_per_sec
;
5466 /* make sure nicenum has enough space */
5467 CTASSERT(sizeof (buf
) >= NN_NUMBUF_SZ
);
5469 zfs_nicebytes(bytes
, buf
, sizeof (buf
));
5470 (void) fprintf(stderr
,
5471 "\r%5s completed (%4dMB/s) "
5472 "estimated time remaining: %uhr %02umin %02usec ",
5473 buf
, kb_per_sec
/ 1024,
5474 sec_remaining
/ 60 / 60,
5475 sec_remaining
/ 60 % 60,
5476 sec_remaining
% 60);
5478 zcb
->zcb_lastprint
= now
;
5485 zdb_leak(void *arg
, uint64_t start
, uint64_t size
)
5489 (void) printf("leaked space: vdev %llu, offset 0x%llx, size %llu\n",
5490 (u_longlong_t
)vd
->vdev_id
, (u_longlong_t
)start
, (u_longlong_t
)size
);
5493 static metaslab_ops_t zdb_metaslab_ops
= {
5499 load_unflushed_svr_segs_cb(spa_t
*spa
, space_map_entry_t
*sme
,
5500 uint64_t txg
, void *arg
)
5502 spa_vdev_removal_t
*svr
= arg
;
5504 uint64_t offset
= sme
->sme_offset
;
5505 uint64_t size
= sme
->sme_run
;
5507 /* skip vdevs we don't care about */
5508 if (sme
->sme_vdev
!= svr
->svr_vdev_id
)
5511 vdev_t
*vd
= vdev_lookup_top(spa
, sme
->sme_vdev
);
5512 metaslab_t
*ms
= vd
->vdev_ms
[offset
>> vd
->vdev_ms_shift
];
5513 ASSERT(sme
->sme_type
== SM_ALLOC
|| sme
->sme_type
== SM_FREE
);
5515 if (txg
< metaslab_unflushed_txg(ms
))
5518 if (sme
->sme_type
== SM_ALLOC
)
5519 range_tree_add(svr
->svr_allocd_segs
, offset
, size
);
5521 range_tree_remove(svr
->svr_allocd_segs
, offset
, size
);
5528 claim_segment_impl_cb(uint64_t inner_offset
, vdev_t
*vd
, uint64_t offset
,
5529 uint64_t size
, void *arg
)
5532 * This callback was called through a remap from
5533 * a device being removed. Therefore, the vdev that
5534 * this callback is applied to is a concrete
5537 ASSERT(vdev_is_concrete(vd
));
5539 VERIFY0(metaslab_claim_impl(vd
, offset
, size
,
5540 spa_min_claim_txg(vd
->vdev_spa
)));
5544 claim_segment_cb(void *arg
, uint64_t offset
, uint64_t size
)
5548 vdev_indirect_ops
.vdev_op_remap(vd
, offset
, size
,
5549 claim_segment_impl_cb
, NULL
);
5553 * After accounting for all allocated blocks that are directly referenced,
5554 * we might have missed a reference to a block from a partially complete
5555 * (and thus unused) indirect mapping object. We perform a secondary pass
5556 * through the metaslabs we have already mapped and claim the destination
5560 zdb_claim_removing(spa_t
*spa
, zdb_cb_t
*zcb
)
5565 if (spa
->spa_vdev_removal
== NULL
)
5568 spa_config_enter(spa
, SCL_CONFIG
, FTAG
, RW_READER
);
5570 spa_vdev_removal_t
*svr
= spa
->spa_vdev_removal
;
5571 vdev_t
*vd
= vdev_lookup_top(spa
, svr
->svr_vdev_id
);
5572 vdev_indirect_mapping_t
*vim
= vd
->vdev_indirect_mapping
;
5574 ASSERT0(range_tree_space(svr
->svr_allocd_segs
));
5576 range_tree_t
*allocs
= range_tree_create(NULL
, RANGE_SEG64
, NULL
, 0, 0);
5577 for (uint64_t msi
= 0; msi
< vd
->vdev_ms_count
; msi
++) {
5578 metaslab_t
*msp
= vd
->vdev_ms
[msi
];
5580 ASSERT0(range_tree_space(allocs
));
5581 if (msp
->ms_sm
!= NULL
)
5582 VERIFY0(space_map_load(msp
->ms_sm
, allocs
, SM_ALLOC
));
5583 range_tree_vacate(allocs
, range_tree_add
, svr
->svr_allocd_segs
);
5585 range_tree_destroy(allocs
);
5587 iterate_through_spacemap_logs(spa
, load_unflushed_svr_segs_cb
, svr
);
5590 * Clear everything past what has been synced,
5591 * because we have not allocated mappings for
5594 range_tree_clear(svr
->svr_allocd_segs
,
5595 vdev_indirect_mapping_max_offset(vim
),
5596 vd
->vdev_asize
- vdev_indirect_mapping_max_offset(vim
));
5598 zcb
->zcb_removing_size
+= range_tree_space(svr
->svr_allocd_segs
);
5599 range_tree_vacate(svr
->svr_allocd_segs
, claim_segment_cb
, vd
);
5601 spa_config_exit(spa
, SCL_CONFIG
, FTAG
);
5606 increment_indirect_mapping_cb(void *arg
, const blkptr_t
*bp
, boolean_t bp_freed
,
5609 zdb_cb_t
*zcb
= arg
;
5610 spa_t
*spa
= zcb
->zcb_spa
;
5612 const dva_t
*dva
= &bp
->blk_dva
[0];
5615 ASSERT(!dump_opt
['L']);
5616 ASSERT3U(BP_GET_NDVAS(bp
), ==, 1);
5618 spa_config_enter(spa
, SCL_VDEV
, FTAG
, RW_READER
);
5619 vd
= vdev_lookup_top(zcb
->zcb_spa
, DVA_GET_VDEV(dva
));
5620 ASSERT3P(vd
, !=, NULL
);
5621 spa_config_exit(spa
, SCL_VDEV
, FTAG
);
5623 ASSERT(vd
->vdev_indirect_config
.vic_mapping_object
!= 0);
5624 ASSERT3P(zcb
->zcb_vd_obsolete_counts
[vd
->vdev_id
], !=, NULL
);
5626 vdev_indirect_mapping_increment_obsolete_count(
5627 vd
->vdev_indirect_mapping
,
5628 DVA_GET_OFFSET(dva
), DVA_GET_ASIZE(dva
),
5629 zcb
->zcb_vd_obsolete_counts
[vd
->vdev_id
]);
5635 zdb_load_obsolete_counts(vdev_t
*vd
)
5637 vdev_indirect_mapping_t
*vim
= vd
->vdev_indirect_mapping
;
5638 spa_t
*spa
= vd
->vdev_spa
;
5639 spa_condensing_indirect_phys_t
*scip
=
5640 &spa
->spa_condensing_indirect_phys
;
5641 uint64_t obsolete_sm_object
;
5644 VERIFY0(vdev_obsolete_sm_object(vd
, &obsolete_sm_object
));
5645 EQUIV(obsolete_sm_object
!= 0, vd
->vdev_obsolete_sm
!= NULL
);
5646 counts
= vdev_indirect_mapping_load_obsolete_counts(vim
);
5647 if (vd
->vdev_obsolete_sm
!= NULL
) {
5648 vdev_indirect_mapping_load_obsolete_spacemap(vim
, counts
,
5649 vd
->vdev_obsolete_sm
);
5651 if (scip
->scip_vdev
== vd
->vdev_id
&&
5652 scip
->scip_prev_obsolete_sm_object
!= 0) {
5653 space_map_t
*prev_obsolete_sm
= NULL
;
5654 VERIFY0(space_map_open(&prev_obsolete_sm
, spa
->spa_meta_objset
,
5655 scip
->scip_prev_obsolete_sm_object
, 0, vd
->vdev_asize
, 0));
5656 vdev_indirect_mapping_load_obsolete_spacemap(vim
, counts
,
5658 space_map_close(prev_obsolete_sm
);
5664 zdb_ddt_leak_init(spa_t
*spa
, zdb_cb_t
*zcb
)
5671 ASSERT(!dump_opt
['L']);
5673 bzero(&ddb
, sizeof (ddb
));
5674 while ((error
= ddt_walk(spa
, &ddb
, &dde
)) == 0) {
5676 ddt_phys_t
*ddp
= dde
.dde_phys
;
5678 if (ddb
.ddb_class
== DDT_CLASS_UNIQUE
)
5681 ASSERT(ddt_phys_total_refcnt(&dde
) > 1);
5683 for (p
= 0; p
< DDT_PHYS_TYPES
; p
++, ddp
++) {
5684 if (ddp
->ddp_phys_birth
== 0)
5686 ddt_bp_create(ddb
.ddb_checksum
,
5687 &dde
.dde_key
, ddp
, &blk
);
5688 if (p
== DDT_PHYS_DITTO
) {
5689 zdb_count_block(zcb
, NULL
, &blk
, ZDB_OT_DITTO
);
5691 zcb
->zcb_dedup_asize
+=
5692 BP_GET_ASIZE(&blk
) * (ddp
->ddp_refcnt
- 1);
5693 zcb
->zcb_dedup_blocks
++;
5696 ddt_t
*ddt
= spa
->spa_ddt
[ddb
.ddb_checksum
];
5698 VERIFY(ddt_lookup(ddt
, &blk
, B_TRUE
) != NULL
);
5702 ASSERT(error
== ENOENT
);
5705 typedef struct checkpoint_sm_exclude_entry_arg
{
5707 uint64_t cseea_checkpoint_size
;
5708 } checkpoint_sm_exclude_entry_arg_t
;
5711 checkpoint_sm_exclude_entry_cb(space_map_entry_t
*sme
, void *arg
)
5713 checkpoint_sm_exclude_entry_arg_t
*cseea
= arg
;
5714 vdev_t
*vd
= cseea
->cseea_vd
;
5715 metaslab_t
*ms
= vd
->vdev_ms
[sme
->sme_offset
>> vd
->vdev_ms_shift
];
5716 uint64_t end
= sme
->sme_offset
+ sme
->sme_run
;
5718 ASSERT(sme
->sme_type
== SM_FREE
);
5721 * Since the vdev_checkpoint_sm exists in the vdev level
5722 * and the ms_sm space maps exist in the metaslab level,
5723 * an entry in the checkpoint space map could theoretically
5724 * cross the boundaries of the metaslab that it belongs.
5726 * In reality, because of the way that we populate and
5727 * manipulate the checkpoint's space maps currently,
5728 * there shouldn't be any entries that cross metaslabs.
5729 * Hence the assertion below.
5731 * That said, there is no fundamental requirement that
5732 * the checkpoint's space map entries should not cross
5733 * metaslab boundaries. So if needed we could add code
5734 * that handles metaslab-crossing segments in the future.
5736 VERIFY3U(sme
->sme_offset
, >=, ms
->ms_start
);
5737 VERIFY3U(end
, <=, ms
->ms_start
+ ms
->ms_size
);
5740 * By removing the entry from the allocated segments we
5741 * also verify that the entry is there to begin with.
5743 mutex_enter(&ms
->ms_lock
);
5744 range_tree_remove(ms
->ms_allocatable
, sme
->sme_offset
, sme
->sme_run
);
5745 mutex_exit(&ms
->ms_lock
);
5747 cseea
->cseea_checkpoint_size
+= sme
->sme_run
;
5752 zdb_leak_init_vdev_exclude_checkpoint(vdev_t
*vd
, zdb_cb_t
*zcb
)
5754 spa_t
*spa
= vd
->vdev_spa
;
5755 space_map_t
*checkpoint_sm
= NULL
;
5756 uint64_t checkpoint_sm_obj
;
5759 * If there is no vdev_top_zap, we are in a pool whose
5760 * version predates the pool checkpoint feature.
5762 if (vd
->vdev_top_zap
== 0)
5766 * If there is no reference of the vdev_checkpoint_sm in
5767 * the vdev_top_zap, then one of the following scenarios
5770 * 1] There is no checkpoint
5771 * 2] There is a checkpoint, but no checkpointed blocks
5772 * have been freed yet
5773 * 3] The current vdev is indirect
5775 * In these cases we return immediately.
5777 if (zap_contains(spa_meta_objset(spa
), vd
->vdev_top_zap
,
5778 VDEV_TOP_ZAP_POOL_CHECKPOINT_SM
) != 0)
5781 VERIFY0(zap_lookup(spa_meta_objset(spa
), vd
->vdev_top_zap
,
5782 VDEV_TOP_ZAP_POOL_CHECKPOINT_SM
, sizeof (uint64_t), 1,
5783 &checkpoint_sm_obj
));
5785 checkpoint_sm_exclude_entry_arg_t cseea
;
5786 cseea
.cseea_vd
= vd
;
5787 cseea
.cseea_checkpoint_size
= 0;
5789 VERIFY0(space_map_open(&checkpoint_sm
, spa_meta_objset(spa
),
5790 checkpoint_sm_obj
, 0, vd
->vdev_asize
, vd
->vdev_ashift
));
5792 VERIFY0(space_map_iterate(checkpoint_sm
,
5793 space_map_length(checkpoint_sm
),
5794 checkpoint_sm_exclude_entry_cb
, &cseea
));
5795 space_map_close(checkpoint_sm
);
5797 zcb
->zcb_checkpoint_size
+= cseea
.cseea_checkpoint_size
;
5801 zdb_leak_init_exclude_checkpoint(spa_t
*spa
, zdb_cb_t
*zcb
)
5803 ASSERT(!dump_opt
['L']);
5805 vdev_t
*rvd
= spa
->spa_root_vdev
;
5806 for (uint64_t c
= 0; c
< rvd
->vdev_children
; c
++) {
5807 ASSERT3U(c
, ==, rvd
->vdev_child
[c
]->vdev_id
);
5808 zdb_leak_init_vdev_exclude_checkpoint(rvd
->vdev_child
[c
], zcb
);
5813 count_unflushed_space_cb(spa_t
*spa
, space_map_entry_t
*sme
,
5814 uint64_t txg
, void *arg
)
5816 int64_t *ualloc_space
= arg
;
5818 uint64_t offset
= sme
->sme_offset
;
5819 uint64_t vdev_id
= sme
->sme_vdev
;
5821 vdev_t
*vd
= vdev_lookup_top(spa
, vdev_id
);
5822 if (!vdev_is_concrete(vd
))
5825 metaslab_t
*ms
= vd
->vdev_ms
[offset
>> vd
->vdev_ms_shift
];
5826 ASSERT(sme
->sme_type
== SM_ALLOC
|| sme
->sme_type
== SM_FREE
);
5828 if (txg
< metaslab_unflushed_txg(ms
))
5831 if (sme
->sme_type
== SM_ALLOC
)
5832 *ualloc_space
+= sme
->sme_run
;
5834 *ualloc_space
-= sme
->sme_run
;
5840 get_unflushed_alloc_space(spa_t
*spa
)
5845 int64_t ualloc_space
= 0;
5846 iterate_through_spacemap_logs(spa
, count_unflushed_space_cb
,
5848 return (ualloc_space
);
5852 load_unflushed_cb(spa_t
*spa
, space_map_entry_t
*sme
, uint64_t txg
, void *arg
)
5854 maptype_t
*uic_maptype
= arg
;
5856 uint64_t offset
= sme
->sme_offset
;
5857 uint64_t size
= sme
->sme_run
;
5858 uint64_t vdev_id
= sme
->sme_vdev
;
5860 vdev_t
*vd
= vdev_lookup_top(spa
, vdev_id
);
5862 /* skip indirect vdevs */
5863 if (!vdev_is_concrete(vd
))
5866 metaslab_t
*ms
= vd
->vdev_ms
[offset
>> vd
->vdev_ms_shift
];
5868 ASSERT(sme
->sme_type
== SM_ALLOC
|| sme
->sme_type
== SM_FREE
);
5869 ASSERT(*uic_maptype
== SM_ALLOC
|| *uic_maptype
== SM_FREE
);
5871 if (txg
< metaslab_unflushed_txg(ms
))
5874 if (*uic_maptype
== sme
->sme_type
)
5875 range_tree_add(ms
->ms_allocatable
, offset
, size
);
5877 range_tree_remove(ms
->ms_allocatable
, offset
, size
);
5883 load_unflushed_to_ms_allocatables(spa_t
*spa
, maptype_t maptype
)
5885 iterate_through_spacemap_logs(spa
, load_unflushed_cb
, &maptype
);
5889 load_concrete_ms_allocatable_trees(spa_t
*spa
, maptype_t maptype
)
5891 vdev_t
*rvd
= spa
->spa_root_vdev
;
5892 for (uint64_t i
= 0; i
< rvd
->vdev_children
; i
++) {
5893 vdev_t
*vd
= rvd
->vdev_child
[i
];
5895 ASSERT3U(i
, ==, vd
->vdev_id
);
5897 if (vd
->vdev_ops
== &vdev_indirect_ops
)
5900 for (uint64_t m
= 0; m
< vd
->vdev_ms_count
; m
++) {
5901 metaslab_t
*msp
= vd
->vdev_ms
[m
];
5903 (void) fprintf(stderr
,
5904 "\rloading concrete vdev %llu, "
5905 "metaslab %llu of %llu ...",
5906 (longlong_t
)vd
->vdev_id
,
5907 (longlong_t
)msp
->ms_id
,
5908 (longlong_t
)vd
->vdev_ms_count
);
5910 mutex_enter(&msp
->ms_lock
);
5911 range_tree_vacate(msp
->ms_allocatable
, NULL
, NULL
);
5914 * We don't want to spend the CPU manipulating the
5915 * size-ordered tree, so clear the range_tree ops.
5917 msp
->ms_allocatable
->rt_ops
= NULL
;
5919 if (msp
->ms_sm
!= NULL
) {
5920 VERIFY0(space_map_load(msp
->ms_sm
,
5921 msp
->ms_allocatable
, maptype
));
5923 if (!msp
->ms_loaded
)
5924 msp
->ms_loaded
= B_TRUE
;
5925 mutex_exit(&msp
->ms_lock
);
5929 load_unflushed_to_ms_allocatables(spa
, maptype
);
5933 * vm_idxp is an in-out parameter which (for indirect vdevs) is the
5934 * index in vim_entries that has the first entry in this metaslab.
5935 * On return, it will be set to the first entry after this metaslab.
5938 load_indirect_ms_allocatable_tree(vdev_t
*vd
, metaslab_t
*msp
,
5941 vdev_indirect_mapping_t
*vim
= vd
->vdev_indirect_mapping
;
5943 mutex_enter(&msp
->ms_lock
);
5944 range_tree_vacate(msp
->ms_allocatable
, NULL
, NULL
);
5947 * We don't want to spend the CPU manipulating the
5948 * size-ordered tree, so clear the range_tree ops.
5950 msp
->ms_allocatable
->rt_ops
= NULL
;
5952 for (; *vim_idxp
< vdev_indirect_mapping_num_entries(vim
);
5954 vdev_indirect_mapping_entry_phys_t
*vimep
=
5955 &vim
->vim_entries
[*vim_idxp
];
5956 uint64_t ent_offset
= DVA_MAPPING_GET_SRC_OFFSET(vimep
);
5957 uint64_t ent_len
= DVA_GET_ASIZE(&vimep
->vimep_dst
);
5958 ASSERT3U(ent_offset
, >=, msp
->ms_start
);
5959 if (ent_offset
>= msp
->ms_start
+ msp
->ms_size
)
5963 * Mappings do not cross metaslab boundaries,
5964 * because we create them by walking the metaslabs.
5966 ASSERT3U(ent_offset
+ ent_len
, <=,
5967 msp
->ms_start
+ msp
->ms_size
);
5968 range_tree_add(msp
->ms_allocatable
, ent_offset
, ent_len
);
5971 if (!msp
->ms_loaded
)
5972 msp
->ms_loaded
= B_TRUE
;
5973 mutex_exit(&msp
->ms_lock
);
5977 zdb_leak_init_prepare_indirect_vdevs(spa_t
*spa
, zdb_cb_t
*zcb
)
5979 ASSERT(!dump_opt
['L']);
5981 vdev_t
*rvd
= spa
->spa_root_vdev
;
5982 for (uint64_t c
= 0; c
< rvd
->vdev_children
; c
++) {
5983 vdev_t
*vd
= rvd
->vdev_child
[c
];
5985 ASSERT3U(c
, ==, vd
->vdev_id
);
5987 if (vd
->vdev_ops
!= &vdev_indirect_ops
)
5991 * Note: we don't check for mapping leaks on
5992 * removing vdevs because their ms_allocatable's
5993 * are used to look for leaks in allocated space.
5995 zcb
->zcb_vd_obsolete_counts
[c
] = zdb_load_obsolete_counts(vd
);
5998 * Normally, indirect vdevs don't have any
5999 * metaslabs. We want to set them up for
6002 vdev_metaslab_group_create(vd
);
6003 VERIFY0(vdev_metaslab_init(vd
, 0));
6005 vdev_indirect_mapping_t
*vim __maybe_unused
=
6006 vd
->vdev_indirect_mapping
;
6007 uint64_t vim_idx
= 0;
6008 for (uint64_t m
= 0; m
< vd
->vdev_ms_count
; m
++) {
6010 (void) fprintf(stderr
,
6011 "\rloading indirect vdev %llu, "
6012 "metaslab %llu of %llu ...",
6013 (longlong_t
)vd
->vdev_id
,
6014 (longlong_t
)vd
->vdev_ms
[m
]->ms_id
,
6015 (longlong_t
)vd
->vdev_ms_count
);
6017 load_indirect_ms_allocatable_tree(vd
, vd
->vdev_ms
[m
],
6020 ASSERT3U(vim_idx
, ==, vdev_indirect_mapping_num_entries(vim
));
6025 zdb_leak_init(spa_t
*spa
, zdb_cb_t
*zcb
)
6032 dsl_pool_t
*dp
= spa
->spa_dsl_pool
;
6033 vdev_t
*rvd
= spa
->spa_root_vdev
;
6036 * We are going to be changing the meaning of the metaslab's
6037 * ms_allocatable. Ensure that the allocator doesn't try to
6040 spa
->spa_normal_class
->mc_ops
= &zdb_metaslab_ops
;
6041 spa
->spa_log_class
->mc_ops
= &zdb_metaslab_ops
;
6042 spa
->spa_embedded_log_class
->mc_ops
= &zdb_metaslab_ops
;
6044 zcb
->zcb_vd_obsolete_counts
=
6045 umem_zalloc(rvd
->vdev_children
* sizeof (uint32_t *),
6049 * For leak detection, we overload the ms_allocatable trees
6050 * to contain allocated segments instead of free segments.
6051 * As a result, we can't use the normal metaslab_load/unload
6054 zdb_leak_init_prepare_indirect_vdevs(spa
, zcb
);
6055 load_concrete_ms_allocatable_trees(spa
, SM_ALLOC
);
6058 * On load_concrete_ms_allocatable_trees() we loaded all the
6059 * allocated entries from the ms_sm to the ms_allocatable for
6060 * each metaslab. If the pool has a checkpoint or is in the
6061 * middle of discarding a checkpoint, some of these blocks
6062 * may have been freed but their ms_sm may not have been
6063 * updated because they are referenced by the checkpoint. In
6064 * order to avoid false-positives during leak-detection, we
6065 * go through the vdev's checkpoint space map and exclude all
6066 * its entries from their relevant ms_allocatable.
6068 * We also aggregate the space held by the checkpoint and add
6069 * it to zcb_checkpoint_size.
6071 * Note that at this point we are also verifying that all the
6072 * entries on the checkpoint_sm are marked as allocated in
6073 * the ms_sm of their relevant metaslab.
6074 * [see comment in checkpoint_sm_exclude_entry_cb()]
6076 zdb_leak_init_exclude_checkpoint(spa
, zcb
);
6077 ASSERT3U(zcb
->zcb_checkpoint_size
, ==, spa_get_checkpoint_space(spa
));
6079 /* for cleaner progress output */
6080 (void) fprintf(stderr
, "\n");
6082 if (bpobj_is_open(&dp
->dp_obsolete_bpobj
)) {
6083 ASSERT(spa_feature_is_enabled(spa
,
6084 SPA_FEATURE_DEVICE_REMOVAL
));
6085 (void) bpobj_iterate_nofree(&dp
->dp_obsolete_bpobj
,
6086 increment_indirect_mapping_cb
, zcb
, NULL
);
6089 spa_config_enter(spa
, SCL_CONFIG
, FTAG
, RW_READER
);
6090 zdb_ddt_leak_init(spa
, zcb
);
6091 spa_config_exit(spa
, SCL_CONFIG
, FTAG
);
6095 zdb_check_for_obsolete_leaks(vdev_t
*vd
, zdb_cb_t
*zcb
)
6097 boolean_t leaks
= B_FALSE
;
6098 vdev_indirect_mapping_t
*vim
= vd
->vdev_indirect_mapping
;
6099 uint64_t total_leaked
= 0;
6100 boolean_t are_precise
= B_FALSE
;
6102 ASSERT(vim
!= NULL
);
6104 for (uint64_t i
= 0; i
< vdev_indirect_mapping_num_entries(vim
); i
++) {
6105 vdev_indirect_mapping_entry_phys_t
*vimep
=
6106 &vim
->vim_entries
[i
];
6107 uint64_t obsolete_bytes
= 0;
6108 uint64_t offset
= DVA_MAPPING_GET_SRC_OFFSET(vimep
);
6109 metaslab_t
*msp
= vd
->vdev_ms
[offset
>> vd
->vdev_ms_shift
];
6112 * This is not very efficient but it's easy to
6113 * verify correctness.
6115 for (uint64_t inner_offset
= 0;
6116 inner_offset
< DVA_GET_ASIZE(&vimep
->vimep_dst
);
6117 inner_offset
+= 1 << vd
->vdev_ashift
) {
6118 if (range_tree_contains(msp
->ms_allocatable
,
6119 offset
+ inner_offset
, 1 << vd
->vdev_ashift
)) {
6120 obsolete_bytes
+= 1 << vd
->vdev_ashift
;
6124 int64_t bytes_leaked
= obsolete_bytes
-
6125 zcb
->zcb_vd_obsolete_counts
[vd
->vdev_id
][i
];
6126 ASSERT3U(DVA_GET_ASIZE(&vimep
->vimep_dst
), >=,
6127 zcb
->zcb_vd_obsolete_counts
[vd
->vdev_id
][i
]);
6129 VERIFY0(vdev_obsolete_counts_are_precise(vd
, &are_precise
));
6130 if (bytes_leaked
!= 0 && (are_precise
|| dump_opt
['d'] >= 5)) {
6131 (void) printf("obsolete indirect mapping count "
6132 "mismatch on %llu:%llx:%llx : %llx bytes leaked\n",
6133 (u_longlong_t
)vd
->vdev_id
,
6134 (u_longlong_t
)DVA_MAPPING_GET_SRC_OFFSET(vimep
),
6135 (u_longlong_t
)DVA_GET_ASIZE(&vimep
->vimep_dst
),
6136 (u_longlong_t
)bytes_leaked
);
6138 total_leaked
+= ABS(bytes_leaked
);
6141 VERIFY0(vdev_obsolete_counts_are_precise(vd
, &are_precise
));
6142 if (!are_precise
&& total_leaked
> 0) {
6143 int pct_leaked
= total_leaked
* 100 /
6144 vdev_indirect_mapping_bytes_mapped(vim
);
6145 (void) printf("cannot verify obsolete indirect mapping "
6146 "counts of vdev %llu because precise feature was not "
6147 "enabled when it was removed: %d%% (%llx bytes) of mapping"
6149 (u_longlong_t
)vd
->vdev_id
, pct_leaked
,
6150 (u_longlong_t
)total_leaked
);
6151 } else if (total_leaked
> 0) {
6152 (void) printf("obsolete indirect mapping count mismatch "
6153 "for vdev %llu -- %llx total bytes mismatched\n",
6154 (u_longlong_t
)vd
->vdev_id
,
6155 (u_longlong_t
)total_leaked
);
6159 vdev_indirect_mapping_free_obsolete_counts(vim
,
6160 zcb
->zcb_vd_obsolete_counts
[vd
->vdev_id
]);
6161 zcb
->zcb_vd_obsolete_counts
[vd
->vdev_id
] = NULL
;
6167 zdb_leak_fini(spa_t
*spa
, zdb_cb_t
*zcb
)
6172 boolean_t leaks
= B_FALSE
;
6173 vdev_t
*rvd
= spa
->spa_root_vdev
;
6174 for (unsigned c
= 0; c
< rvd
->vdev_children
; c
++) {
6175 vdev_t
*vd
= rvd
->vdev_child
[c
];
6177 if (zcb
->zcb_vd_obsolete_counts
[c
] != NULL
) {
6178 leaks
|= zdb_check_for_obsolete_leaks(vd
, zcb
);
6181 for (uint64_t m
= 0; m
< vd
->vdev_ms_count
; m
++) {
6182 metaslab_t
*msp
= vd
->vdev_ms
[m
];
6183 ASSERT3P(msp
->ms_group
, ==, (msp
->ms_group
->mg_class
==
6184 spa_embedded_log_class(spa
)) ?
6185 vd
->vdev_log_mg
: vd
->vdev_mg
);
6188 * ms_allocatable has been overloaded
6189 * to contain allocated segments. Now that
6190 * we finished traversing all blocks, any
6191 * block that remains in the ms_allocatable
6192 * represents an allocated block that we
6193 * did not claim during the traversal.
6194 * Claimed blocks would have been removed
6195 * from the ms_allocatable. For indirect
6196 * vdevs, space remaining in the tree
6197 * represents parts of the mapping that are
6198 * not referenced, which is not a bug.
6200 if (vd
->vdev_ops
== &vdev_indirect_ops
) {
6201 range_tree_vacate(msp
->ms_allocatable
,
6204 range_tree_vacate(msp
->ms_allocatable
,
6207 if (msp
->ms_loaded
) {
6208 msp
->ms_loaded
= B_FALSE
;
6213 umem_free(zcb
->zcb_vd_obsolete_counts
,
6214 rvd
->vdev_children
* sizeof (uint32_t *));
6215 zcb
->zcb_vd_obsolete_counts
= NULL
;
6222 count_block_cb(void *arg
, const blkptr_t
*bp
, dmu_tx_t
*tx
)
6224 zdb_cb_t
*zcb
= arg
;
6226 if (dump_opt
['b'] >= 5) {
6227 char blkbuf
[BP_SPRINTF_LEN
];
6228 snprintf_blkptr(blkbuf
, sizeof (blkbuf
), bp
);
6229 (void) printf("[%s] %s\n",
6230 "deferred free", blkbuf
);
6232 zdb_count_block(zcb
, NULL
, bp
, ZDB_OT_DEFERRED
);
6237 * Iterate over livelists which have been destroyed by the user but
6238 * are still present in the MOS, waiting to be freed
6241 iterate_deleted_livelists(spa_t
*spa
, ll_iter_t func
, void *arg
)
6243 objset_t
*mos
= spa
->spa_meta_objset
;
6245 int err
= zap_lookup(mos
, DMU_POOL_DIRECTORY_OBJECT
,
6246 DMU_POOL_DELETED_CLONES
, sizeof (uint64_t), 1, &zap_obj
);
6252 zap_attribute_t attr
;
6254 /* NULL out os prior to dsl_deadlist_open in case it's garbage */
6256 for (zap_cursor_init(&zc
, mos
, zap_obj
);
6257 zap_cursor_retrieve(&zc
, &attr
) == 0;
6258 (void) zap_cursor_advance(&zc
)) {
6259 dsl_deadlist_open(&ll
, mos
, attr
.za_first_integer
);
6261 dsl_deadlist_close(&ll
);
6263 zap_cursor_fini(&zc
);
6267 bpobj_count_block_cb(void *arg
, const blkptr_t
*bp
, boolean_t bp_freed
,
6271 return (count_block_cb(arg
, bp
, tx
));
6275 livelist_entry_count_blocks_cb(void *args
, dsl_deadlist_entry_t
*dle
)
6277 zdb_cb_t
*zbc
= args
;
6279 bplist_create(&blks
);
6280 /* determine which blocks have been alloc'd but not freed */
6281 VERIFY0(dsl_process_sub_livelist(&dle
->dle_bpobj
, &blks
, NULL
, NULL
));
6282 /* count those blocks */
6283 (void) bplist_iterate(&blks
, count_block_cb
, zbc
, NULL
);
6284 bplist_destroy(&blks
);
6289 livelist_count_blocks(dsl_deadlist_t
*ll
, void *arg
)
6291 dsl_deadlist_iterate(ll
, livelist_entry_count_blocks_cb
, arg
);
6295 * Count the blocks in the livelists that have been destroyed by the user
6296 * but haven't yet been freed.
6299 deleted_livelists_count_blocks(spa_t
*spa
, zdb_cb_t
*zbc
)
6301 iterate_deleted_livelists(spa
, livelist_count_blocks
, zbc
);
6305 dump_livelist_cb(dsl_deadlist_t
*ll
, void *arg
)
6307 ASSERT3P(arg
, ==, NULL
);
6308 global_feature_count
[SPA_FEATURE_LIVELIST
]++;
6309 dump_blkptr_list(ll
, "Deleted Livelist");
6310 dsl_deadlist_iterate(ll
, sublivelist_verify_lightweight
, NULL
);
6314 * Print out, register object references to, and increment feature counts for
6315 * livelists that have been destroyed by the user but haven't yet been freed.
6318 deleted_livelists_dump_mos(spa_t
*spa
)
6321 objset_t
*mos
= spa
->spa_meta_objset
;
6322 int err
= zap_lookup(mos
, DMU_POOL_DIRECTORY_OBJECT
,
6323 DMU_POOL_DELETED_CLONES
, sizeof (uint64_t), 1, &zap_obj
);
6326 mos_obj_refd(zap_obj
);
6327 iterate_deleted_livelists(spa
, dump_livelist_cb
, NULL
);
6331 dump_block_stats(spa_t
*spa
)
6334 zdb_blkstats_t
*zb
, *tzb
;
6335 uint64_t norm_alloc
, norm_space
, total_alloc
, total_found
;
6336 int flags
= TRAVERSE_PRE
| TRAVERSE_PREFETCH_METADATA
|
6337 TRAVERSE_NO_DECRYPT
| TRAVERSE_HARD
;
6338 boolean_t leaks
= B_FALSE
;
6340 bp_embedded_type_t i
;
6342 bzero(&zcb
, sizeof (zcb
));
6343 (void) printf("\nTraversing all blocks %s%s%s%s%s...\n\n",
6344 (dump_opt
['c'] || !dump_opt
['L']) ? "to verify " : "",
6345 (dump_opt
['c'] == 1) ? "metadata " : "",
6346 dump_opt
['c'] ? "checksums " : "",
6347 (dump_opt
['c'] && !dump_opt
['L']) ? "and verify " : "",
6348 !dump_opt
['L'] ? "nothing leaked " : "");
6351 * When leak detection is enabled we load all space maps as SM_ALLOC
6352 * maps, then traverse the pool claiming each block we discover. If
6353 * the pool is perfectly consistent, the segment trees will be empty
6354 * when we're done. Anything left over is a leak; any block we can't
6355 * claim (because it's not part of any space map) is a double
6356 * allocation, reference to a freed block, or an unclaimed log block.
6358 * When leak detection is disabled (-L option) we still traverse the
6359 * pool claiming each block we discover, but we skip opening any space
6362 bzero(&zcb
, sizeof (zdb_cb_t
));
6363 zdb_leak_init(spa
, &zcb
);
6366 * If there's a deferred-free bplist, process that first.
6368 (void) bpobj_iterate_nofree(&spa
->spa_deferred_bpobj
,
6369 bpobj_count_block_cb
, &zcb
, NULL
);
6371 if (spa_version(spa
) >= SPA_VERSION_DEADLISTS
) {
6372 (void) bpobj_iterate_nofree(&spa
->spa_dsl_pool
->dp_free_bpobj
,
6373 bpobj_count_block_cb
, &zcb
, NULL
);
6376 zdb_claim_removing(spa
, &zcb
);
6378 if (spa_feature_is_active(spa
, SPA_FEATURE_ASYNC_DESTROY
)) {
6379 VERIFY3U(0, ==, bptree_iterate(spa
->spa_meta_objset
,
6380 spa
->spa_dsl_pool
->dp_bptree_obj
, B_FALSE
, count_block_cb
,
6384 deleted_livelists_count_blocks(spa
, &zcb
);
6386 if (dump_opt
['c'] > 1)
6387 flags
|= TRAVERSE_PREFETCH_DATA
;
6389 zcb
.zcb_totalasize
= metaslab_class_get_alloc(spa_normal_class(spa
));
6390 zcb
.zcb_totalasize
+= metaslab_class_get_alloc(spa_special_class(spa
));
6391 zcb
.zcb_totalasize
+= metaslab_class_get_alloc(spa_dedup_class(spa
));
6392 zcb
.zcb_totalasize
+=
6393 metaslab_class_get_alloc(spa_embedded_log_class(spa
));
6394 zcb
.zcb_start
= zcb
.zcb_lastprint
= gethrtime();
6395 err
= traverse_pool(spa
, 0, flags
, zdb_blkptr_cb
, &zcb
);
6398 * If we've traversed the data blocks then we need to wait for those
6399 * I/Os to complete. We leverage "The Godfather" zio to wait on
6400 * all async I/Os to complete.
6402 if (dump_opt
['c']) {
6403 for (c
= 0; c
< max_ncpus
; c
++) {
6404 (void) zio_wait(spa
->spa_async_zio_root
[c
]);
6405 spa
->spa_async_zio_root
[c
] = zio_root(spa
, NULL
, NULL
,
6406 ZIO_FLAG_CANFAIL
| ZIO_FLAG_SPECULATIVE
|
6407 ZIO_FLAG_GODFATHER
);
6410 ASSERT0(spa
->spa_load_verify_bytes
);
6413 * Done after zio_wait() since zcb_haderrors is modified in
6416 zcb
.zcb_haderrors
|= err
;
6418 if (zcb
.zcb_haderrors
) {
6419 (void) printf("\nError counts:\n\n");
6420 (void) printf("\t%5s %s\n", "errno", "count");
6421 for (e
= 0; e
< 256; e
++) {
6422 if (zcb
.zcb_errors
[e
] != 0) {
6423 (void) printf("\t%5d %llu\n",
6424 e
, (u_longlong_t
)zcb
.zcb_errors
[e
]);
6430 * Report any leaked segments.
6432 leaks
|= zdb_leak_fini(spa
, &zcb
);
6434 tzb
= &zcb
.zcb_type
[ZB_TOTAL
][ZDB_OT_TOTAL
];
6436 norm_alloc
= metaslab_class_get_alloc(spa_normal_class(spa
));
6437 norm_space
= metaslab_class_get_space(spa_normal_class(spa
));
6439 total_alloc
= norm_alloc
+
6440 metaslab_class_get_alloc(spa_log_class(spa
)) +
6441 metaslab_class_get_alloc(spa_embedded_log_class(spa
)) +
6442 metaslab_class_get_alloc(spa_special_class(spa
)) +
6443 metaslab_class_get_alloc(spa_dedup_class(spa
)) +
6444 get_unflushed_alloc_space(spa
);
6445 total_found
= tzb
->zb_asize
- zcb
.zcb_dedup_asize
+
6446 zcb
.zcb_removing_size
+ zcb
.zcb_checkpoint_size
;
6448 if (total_found
== total_alloc
&& !dump_opt
['L']) {
6449 (void) printf("\n\tNo leaks (block sum matches space"
6450 " maps exactly)\n");
6451 } else if (!dump_opt
['L']) {
6452 (void) printf("block traversal size %llu != alloc %llu "
6454 (u_longlong_t
)total_found
,
6455 (u_longlong_t
)total_alloc
,
6456 (dump_opt
['L']) ? "unreachable" : "leaked",
6457 (longlong_t
)(total_alloc
- total_found
));
6461 if (tzb
->zb_count
== 0)
6464 (void) printf("\n");
6465 (void) printf("\t%-16s %14llu\n", "bp count:",
6466 (u_longlong_t
)tzb
->zb_count
);
6467 (void) printf("\t%-16s %14llu\n", "ganged count:",
6468 (longlong_t
)tzb
->zb_gangs
);
6469 (void) printf("\t%-16s %14llu avg: %6llu\n", "bp logical:",
6470 (u_longlong_t
)tzb
->zb_lsize
,
6471 (u_longlong_t
)(tzb
->zb_lsize
/ tzb
->zb_count
));
6472 (void) printf("\t%-16s %14llu avg: %6llu compression: %6.2f\n",
6473 "bp physical:", (u_longlong_t
)tzb
->zb_psize
,
6474 (u_longlong_t
)(tzb
->zb_psize
/ tzb
->zb_count
),
6475 (double)tzb
->zb_lsize
/ tzb
->zb_psize
);
6476 (void) printf("\t%-16s %14llu avg: %6llu compression: %6.2f\n",
6477 "bp allocated:", (u_longlong_t
)tzb
->zb_asize
,
6478 (u_longlong_t
)(tzb
->zb_asize
/ tzb
->zb_count
),
6479 (double)tzb
->zb_lsize
/ tzb
->zb_asize
);
6480 (void) printf("\t%-16s %14llu ref>1: %6llu deduplication: %6.2f\n",
6481 "bp deduped:", (u_longlong_t
)zcb
.zcb_dedup_asize
,
6482 (u_longlong_t
)zcb
.zcb_dedup_blocks
,
6483 (double)zcb
.zcb_dedup_asize
/ tzb
->zb_asize
+ 1.0);
6484 (void) printf("\t%-16s %14llu used: %5.2f%%\n", "Normal class:",
6485 (u_longlong_t
)norm_alloc
, 100.0 * norm_alloc
/ norm_space
);
6487 if (spa_special_class(spa
)->mc_allocator
[0].mca_rotor
!= NULL
) {
6488 uint64_t alloc
= metaslab_class_get_alloc(
6489 spa_special_class(spa
));
6490 uint64_t space
= metaslab_class_get_space(
6491 spa_special_class(spa
));
6493 (void) printf("\t%-16s %14llu used: %5.2f%%\n",
6494 "Special class", (u_longlong_t
)alloc
,
6495 100.0 * alloc
/ space
);
6498 if (spa_dedup_class(spa
)->mc_allocator
[0].mca_rotor
!= NULL
) {
6499 uint64_t alloc
= metaslab_class_get_alloc(
6500 spa_dedup_class(spa
));
6501 uint64_t space
= metaslab_class_get_space(
6502 spa_dedup_class(spa
));
6504 (void) printf("\t%-16s %14llu used: %5.2f%%\n",
6505 "Dedup class", (u_longlong_t
)alloc
,
6506 100.0 * alloc
/ space
);
6509 if (spa_embedded_log_class(spa
)->mc_allocator
[0].mca_rotor
!= NULL
) {
6510 uint64_t alloc
= metaslab_class_get_alloc(
6511 spa_embedded_log_class(spa
));
6512 uint64_t space
= metaslab_class_get_space(
6513 spa_embedded_log_class(spa
));
6515 (void) printf("\t%-16s %14llu used: %5.2f%%\n",
6516 "Embedded log class", (u_longlong_t
)alloc
,
6517 100.0 * alloc
/ space
);
6520 for (i
= 0; i
< NUM_BP_EMBEDDED_TYPES
; i
++) {
6521 if (zcb
.zcb_embedded_blocks
[i
] == 0)
6523 (void) printf("\n");
6524 (void) printf("\tadditional, non-pointer bps of type %u: "
6526 i
, (u_longlong_t
)zcb
.zcb_embedded_blocks
[i
]);
6528 if (dump_opt
['b'] >= 3) {
6529 (void) printf("\t number of (compressed) bytes: "
6531 dump_histogram(zcb
.zcb_embedded_histogram
[i
],
6532 sizeof (zcb
.zcb_embedded_histogram
[i
]) /
6533 sizeof (zcb
.zcb_embedded_histogram
[i
][0]), 0);
6537 if (tzb
->zb_ditto_samevdev
!= 0) {
6538 (void) printf("\tDittoed blocks on same vdev: %llu\n",
6539 (longlong_t
)tzb
->zb_ditto_samevdev
);
6541 if (tzb
->zb_ditto_same_ms
!= 0) {
6542 (void) printf("\tDittoed blocks in same metaslab: %llu\n",
6543 (longlong_t
)tzb
->zb_ditto_same_ms
);
6546 for (uint64_t v
= 0; v
< spa
->spa_root_vdev
->vdev_children
; v
++) {
6547 vdev_t
*vd
= spa
->spa_root_vdev
->vdev_child
[v
];
6548 vdev_indirect_mapping_t
*vim
= vd
->vdev_indirect_mapping
;
6555 zdb_nicenum(vdev_indirect_mapping_num_entries(vim
),
6556 mem
, vdev_indirect_mapping_size(vim
));
6558 (void) printf("\tindirect vdev id %llu has %llu segments "
6560 (longlong_t
)vd
->vdev_id
,
6561 (longlong_t
)vdev_indirect_mapping_num_entries(vim
), mem
);
6564 if (dump_opt
['b'] >= 2) {
6566 (void) printf("\nBlocks\tLSIZE\tPSIZE\tASIZE"
6567 "\t avg\t comp\t%%Total\tType\n");
6569 for (t
= 0; t
<= ZDB_OT_TOTAL
; t
++) {
6570 char csize
[32], lsize
[32], psize
[32], asize
[32];
6571 char avg
[32], gang
[32];
6572 const char *typename
;
6574 /* make sure nicenum has enough space */
6575 CTASSERT(sizeof (csize
) >= NN_NUMBUF_SZ
);
6576 CTASSERT(sizeof (lsize
) >= NN_NUMBUF_SZ
);
6577 CTASSERT(sizeof (psize
) >= NN_NUMBUF_SZ
);
6578 CTASSERT(sizeof (asize
) >= NN_NUMBUF_SZ
);
6579 CTASSERT(sizeof (avg
) >= NN_NUMBUF_SZ
);
6580 CTASSERT(sizeof (gang
) >= NN_NUMBUF_SZ
);
6582 if (t
< DMU_OT_NUMTYPES
)
6583 typename
= dmu_ot
[t
].ot_name
;
6585 typename
= zdb_ot_extname
[t
- DMU_OT_NUMTYPES
];
6587 if (zcb
.zcb_type
[ZB_TOTAL
][t
].zb_asize
== 0) {
6588 (void) printf("%6s\t%5s\t%5s\t%5s"
6589 "\t%5s\t%5s\t%6s\t%s\n",
6601 for (l
= ZB_TOTAL
- 1; l
>= -1; l
--) {
6602 level
= (l
== -1 ? ZB_TOTAL
: l
);
6603 zb
= &zcb
.zcb_type
[level
][t
];
6605 if (zb
->zb_asize
== 0)
6608 if (dump_opt
['b'] < 3 && level
!= ZB_TOTAL
)
6611 if (level
== 0 && zb
->zb_asize
==
6612 zcb
.zcb_type
[ZB_TOTAL
][t
].zb_asize
)
6615 zdb_nicenum(zb
->zb_count
, csize
,
6617 zdb_nicenum(zb
->zb_lsize
, lsize
,
6619 zdb_nicenum(zb
->zb_psize
, psize
,
6621 zdb_nicenum(zb
->zb_asize
, asize
,
6623 zdb_nicenum(zb
->zb_asize
/ zb
->zb_count
, avg
,
6625 zdb_nicenum(zb
->zb_gangs
, gang
, sizeof (gang
));
6627 (void) printf("%6s\t%5s\t%5s\t%5s\t%5s"
6629 csize
, lsize
, psize
, asize
, avg
,
6630 (double)zb
->zb_lsize
/ zb
->zb_psize
,
6631 100.0 * zb
->zb_asize
/ tzb
->zb_asize
);
6633 if (level
== ZB_TOTAL
)
6634 (void) printf("%s\n", typename
);
6636 (void) printf(" L%d %s\n",
6639 if (dump_opt
['b'] >= 3 && zb
->zb_gangs
> 0) {
6640 (void) printf("\t number of ganged "
6641 "blocks: %s\n", gang
);
6644 if (dump_opt
['b'] >= 4) {
6645 (void) printf("psize "
6646 "(in 512-byte sectors): "
6647 "number of blocks\n");
6648 dump_histogram(zb
->zb_psize_histogram
,
6649 PSIZE_HISTO_SIZE
, 0);
6654 /* Output a table summarizing block sizes in the pool */
6655 if (dump_opt
['b'] >= 2) {
6656 dump_size_histograms(&zcb
);
6660 (void) printf("\n");
6665 if (zcb
.zcb_haderrors
)
6671 typedef struct zdb_ddt_entry
{
6673 uint64_t zdde_ref_blocks
;
6674 uint64_t zdde_ref_lsize
;
6675 uint64_t zdde_ref_psize
;
6676 uint64_t zdde_ref_dsize
;
6677 avl_node_t zdde_node
;
6682 zdb_ddt_add_cb(spa_t
*spa
, zilog_t
*zilog
, const blkptr_t
*bp
,
6683 const zbookmark_phys_t
*zb
, const dnode_phys_t
*dnp
, void *arg
)
6685 avl_tree_t
*t
= arg
;
6687 zdb_ddt_entry_t
*zdde
, zdde_search
;
6689 if (zb
->zb_level
== ZB_DNODE_LEVEL
|| BP_IS_HOLE(bp
) ||
6693 if (dump_opt
['S'] > 1 && zb
->zb_level
== ZB_ROOT_LEVEL
) {
6694 (void) printf("traversing objset %llu, %llu objects, "
6695 "%lu blocks so far\n",
6696 (u_longlong_t
)zb
->zb_objset
,
6697 (u_longlong_t
)BP_GET_FILL(bp
),
6701 if (BP_IS_HOLE(bp
) || BP_GET_CHECKSUM(bp
) == ZIO_CHECKSUM_OFF
||
6702 BP_GET_LEVEL(bp
) > 0 || DMU_OT_IS_METADATA(BP_GET_TYPE(bp
)))
6705 ddt_key_fill(&zdde_search
.zdde_key
, bp
);
6707 zdde
= avl_find(t
, &zdde_search
, &where
);
6710 zdde
= umem_zalloc(sizeof (*zdde
), UMEM_NOFAIL
);
6711 zdde
->zdde_key
= zdde_search
.zdde_key
;
6712 avl_insert(t
, zdde
, where
);
6715 zdde
->zdde_ref_blocks
+= 1;
6716 zdde
->zdde_ref_lsize
+= BP_GET_LSIZE(bp
);
6717 zdde
->zdde_ref_psize
+= BP_GET_PSIZE(bp
);
6718 zdde
->zdde_ref_dsize
+= bp_get_dsize_sync(spa
, bp
);
6724 dump_simulated_ddt(spa_t
*spa
)
6727 void *cookie
= NULL
;
6728 zdb_ddt_entry_t
*zdde
;
6729 ddt_histogram_t ddh_total
;
6730 ddt_stat_t dds_total
;
6732 bzero(&ddh_total
, sizeof (ddh_total
));
6733 bzero(&dds_total
, sizeof (dds_total
));
6734 avl_create(&t
, ddt_entry_compare
,
6735 sizeof (zdb_ddt_entry_t
), offsetof(zdb_ddt_entry_t
, zdde_node
));
6737 spa_config_enter(spa
, SCL_CONFIG
, FTAG
, RW_READER
);
6739 (void) traverse_pool(spa
, 0, TRAVERSE_PRE
| TRAVERSE_PREFETCH_METADATA
|
6740 TRAVERSE_NO_DECRYPT
, zdb_ddt_add_cb
, &t
);
6742 spa_config_exit(spa
, SCL_CONFIG
, FTAG
);
6744 while ((zdde
= avl_destroy_nodes(&t
, &cookie
)) != NULL
) {
6746 uint64_t refcnt
= zdde
->zdde_ref_blocks
;
6747 ASSERT(refcnt
!= 0);
6749 dds
.dds_blocks
= zdde
->zdde_ref_blocks
/ refcnt
;
6750 dds
.dds_lsize
= zdde
->zdde_ref_lsize
/ refcnt
;
6751 dds
.dds_psize
= zdde
->zdde_ref_psize
/ refcnt
;
6752 dds
.dds_dsize
= zdde
->zdde_ref_dsize
/ refcnt
;
6754 dds
.dds_ref_blocks
= zdde
->zdde_ref_blocks
;
6755 dds
.dds_ref_lsize
= zdde
->zdde_ref_lsize
;
6756 dds
.dds_ref_psize
= zdde
->zdde_ref_psize
;
6757 dds
.dds_ref_dsize
= zdde
->zdde_ref_dsize
;
6759 ddt_stat_add(&ddh_total
.ddh_stat
[highbit64(refcnt
) - 1],
6762 umem_free(zdde
, sizeof (*zdde
));
6767 ddt_histogram_stat(&dds_total
, &ddh_total
);
6769 (void) printf("Simulated DDT histogram:\n");
6771 zpool_dump_ddt(&dds_total
, &ddh_total
);
6773 dump_dedup_ratio(&dds_total
);
6777 verify_device_removal_feature_counts(spa_t
*spa
)
6779 uint64_t dr_feature_refcount
= 0;
6780 uint64_t oc_feature_refcount
= 0;
6781 uint64_t indirect_vdev_count
= 0;
6782 uint64_t precise_vdev_count
= 0;
6783 uint64_t obsolete_counts_object_count
= 0;
6784 uint64_t obsolete_sm_count
= 0;
6785 uint64_t obsolete_counts_count
= 0;
6786 uint64_t scip_count
= 0;
6787 uint64_t obsolete_bpobj_count
= 0;
6790 spa_condensing_indirect_phys_t
*scip
=
6791 &spa
->spa_condensing_indirect_phys
;
6792 if (scip
->scip_next_mapping_object
!= 0) {
6793 vdev_t
*vd
= spa
->spa_root_vdev
->vdev_child
[scip
->scip_vdev
];
6794 ASSERT(scip
->scip_prev_obsolete_sm_object
!= 0);
6795 ASSERT3P(vd
->vdev_ops
, ==, &vdev_indirect_ops
);
6797 (void) printf("Condensing indirect vdev %llu: new mapping "
6798 "object %llu, prev obsolete sm %llu\n",
6799 (u_longlong_t
)scip
->scip_vdev
,
6800 (u_longlong_t
)scip
->scip_next_mapping_object
,
6801 (u_longlong_t
)scip
->scip_prev_obsolete_sm_object
);
6802 if (scip
->scip_prev_obsolete_sm_object
!= 0) {
6803 space_map_t
*prev_obsolete_sm
= NULL
;
6804 VERIFY0(space_map_open(&prev_obsolete_sm
,
6805 spa
->spa_meta_objset
,
6806 scip
->scip_prev_obsolete_sm_object
,
6807 0, vd
->vdev_asize
, 0));
6808 dump_spacemap(spa
->spa_meta_objset
, prev_obsolete_sm
);
6809 (void) printf("\n");
6810 space_map_close(prev_obsolete_sm
);
6816 for (uint64_t i
= 0; i
< spa
->spa_root_vdev
->vdev_children
; i
++) {
6817 vdev_t
*vd
= spa
->spa_root_vdev
->vdev_child
[i
];
6818 vdev_indirect_config_t
*vic
= &vd
->vdev_indirect_config
;
6820 if (vic
->vic_mapping_object
!= 0) {
6821 ASSERT(vd
->vdev_ops
== &vdev_indirect_ops
||
6823 indirect_vdev_count
++;
6825 if (vd
->vdev_indirect_mapping
->vim_havecounts
) {
6826 obsolete_counts_count
++;
6830 boolean_t are_precise
;
6831 VERIFY0(vdev_obsolete_counts_are_precise(vd
, &are_precise
));
6833 ASSERT(vic
->vic_mapping_object
!= 0);
6834 precise_vdev_count
++;
6837 uint64_t obsolete_sm_object
;
6838 VERIFY0(vdev_obsolete_sm_object(vd
, &obsolete_sm_object
));
6839 if (obsolete_sm_object
!= 0) {
6840 ASSERT(vic
->vic_mapping_object
!= 0);
6841 obsolete_sm_count
++;
6845 (void) feature_get_refcount(spa
,
6846 &spa_feature_table
[SPA_FEATURE_DEVICE_REMOVAL
],
6847 &dr_feature_refcount
);
6848 (void) feature_get_refcount(spa
,
6849 &spa_feature_table
[SPA_FEATURE_OBSOLETE_COUNTS
],
6850 &oc_feature_refcount
);
6852 if (dr_feature_refcount
!= indirect_vdev_count
) {
6854 (void) printf("Number of indirect vdevs (%llu) " \
6855 "does not match feature count (%llu)\n",
6856 (u_longlong_t
)indirect_vdev_count
,
6857 (u_longlong_t
)dr_feature_refcount
);
6859 (void) printf("Verified device_removal feature refcount " \
6860 "of %llu is correct\n",
6861 (u_longlong_t
)dr_feature_refcount
);
6864 if (zap_contains(spa_meta_objset(spa
), DMU_POOL_DIRECTORY_OBJECT
,
6865 DMU_POOL_OBSOLETE_BPOBJ
) == 0) {
6866 obsolete_bpobj_count
++;
6870 obsolete_counts_object_count
= precise_vdev_count
;
6871 obsolete_counts_object_count
+= obsolete_sm_count
;
6872 obsolete_counts_object_count
+= obsolete_counts_count
;
6873 obsolete_counts_object_count
+= scip_count
;
6874 obsolete_counts_object_count
+= obsolete_bpobj_count
;
6875 obsolete_counts_object_count
+= remap_deadlist_count
;
6877 if (oc_feature_refcount
!= obsolete_counts_object_count
) {
6879 (void) printf("Number of obsolete counts objects (%llu) " \
6880 "does not match feature count (%llu)\n",
6881 (u_longlong_t
)obsolete_counts_object_count
,
6882 (u_longlong_t
)oc_feature_refcount
);
6883 (void) printf("pv:%llu os:%llu oc:%llu sc:%llu "
6884 "ob:%llu rd:%llu\n",
6885 (u_longlong_t
)precise_vdev_count
,
6886 (u_longlong_t
)obsolete_sm_count
,
6887 (u_longlong_t
)obsolete_counts_count
,
6888 (u_longlong_t
)scip_count
,
6889 (u_longlong_t
)obsolete_bpobj_count
,
6890 (u_longlong_t
)remap_deadlist_count
);
6892 (void) printf("Verified indirect_refcount feature refcount " \
6893 "of %llu is correct\n",
6894 (u_longlong_t
)oc_feature_refcount
);
6900 zdb_set_skip_mmp(char *target
)
6905 * Disable the activity check to allow examination of
6908 mutex_enter(&spa_namespace_lock
);
6909 if ((spa
= spa_lookup(target
)) != NULL
) {
6910 spa
->spa_import_flags
|= ZFS_IMPORT_SKIP_MMP
;
6912 mutex_exit(&spa_namespace_lock
);
6915 #define BOGUS_SUFFIX "_CHECKPOINTED_UNIVERSE"
6917 * Import the checkpointed state of the pool specified by the target
6918 * parameter as readonly. The function also accepts a pool config
6919 * as an optional parameter, else it attempts to infer the config by
6920 * the name of the target pool.
6922 * Note that the checkpointed state's pool name will be the name of
6923 * the original pool with the above suffix appended to it. In addition,
6924 * if the target is not a pool name (e.g. a path to a dataset) then
6925 * the new_path parameter is populated with the updated path to
6926 * reflect the fact that we are looking into the checkpointed state.
6928 * The function returns a newly-allocated copy of the name of the
6929 * pool containing the checkpointed state. When this copy is no
6930 * longer needed it should be freed with free(3C). Same thing
6931 * applies to the new_path parameter if allocated.
6934 import_checkpointed_state(char *target
, nvlist_t
*cfg
, char **new_path
)
6937 char *poolname
, *bogus_name
= NULL
;
6938 boolean_t freecfg
= B_FALSE
;
6940 /* If the target is not a pool, the extract the pool name */
6941 char *path_start
= strchr(target
, '/');
6942 if (path_start
!= NULL
) {
6943 size_t poolname_len
= path_start
- target
;
6944 poolname
= strndup(target
, poolname_len
);
6950 zdb_set_skip_mmp(poolname
);
6951 error
= spa_get_stats(poolname
, &cfg
, NULL
, 0);
6953 fatal("Tried to read config of pool \"%s\" but "
6954 "spa_get_stats() failed with error %d\n",
6960 if (asprintf(&bogus_name
, "%s%s", poolname
, BOGUS_SUFFIX
) == -1)
6962 fnvlist_add_string(cfg
, ZPOOL_CONFIG_POOL_NAME
, bogus_name
);
6964 error
= spa_import(bogus_name
, cfg
, NULL
,
6965 ZFS_IMPORT_MISSING_LOG
| ZFS_IMPORT_CHECKPOINT
|
6966 ZFS_IMPORT_SKIP_MMP
);
6970 fatal("Tried to import pool \"%s\" but spa_import() failed "
6971 "with error %d\n", bogus_name
, error
);
6974 if (new_path
!= NULL
&& path_start
!= NULL
) {
6975 if (asprintf(new_path
, "%s%s", bogus_name
, path_start
) == -1) {
6976 if (path_start
!= NULL
)
6982 if (target
!= poolname
)
6985 return (bogus_name
);
6988 typedef struct verify_checkpoint_sm_entry_cb_arg
{
6991 /* the following fields are only used for printing progress */
6992 uint64_t vcsec_entryid
;
6993 uint64_t vcsec_num_entries
;
6994 } verify_checkpoint_sm_entry_cb_arg_t
;
6996 #define ENTRIES_PER_PROGRESS_UPDATE 10000
6999 verify_checkpoint_sm_entry_cb(space_map_entry_t
*sme
, void *arg
)
7001 verify_checkpoint_sm_entry_cb_arg_t
*vcsec
= arg
;
7002 vdev_t
*vd
= vcsec
->vcsec_vd
;
7003 metaslab_t
*ms
= vd
->vdev_ms
[sme
->sme_offset
>> vd
->vdev_ms_shift
];
7004 uint64_t end
= sme
->sme_offset
+ sme
->sme_run
;
7006 ASSERT(sme
->sme_type
== SM_FREE
);
7008 if ((vcsec
->vcsec_entryid
% ENTRIES_PER_PROGRESS_UPDATE
) == 0) {
7009 (void) fprintf(stderr
,
7010 "\rverifying vdev %llu, space map entry %llu of %llu ...",
7011 (longlong_t
)vd
->vdev_id
,
7012 (longlong_t
)vcsec
->vcsec_entryid
,
7013 (longlong_t
)vcsec
->vcsec_num_entries
);
7015 vcsec
->vcsec_entryid
++;
7018 * See comment in checkpoint_sm_exclude_entry_cb()
7020 VERIFY3U(sme
->sme_offset
, >=, ms
->ms_start
);
7021 VERIFY3U(end
, <=, ms
->ms_start
+ ms
->ms_size
);
7024 * The entries in the vdev_checkpoint_sm should be marked as
7025 * allocated in the checkpointed state of the pool, therefore
7026 * their respective ms_allocateable trees should not contain them.
7028 mutex_enter(&ms
->ms_lock
);
7029 range_tree_verify_not_present(ms
->ms_allocatable
,
7030 sme
->sme_offset
, sme
->sme_run
);
7031 mutex_exit(&ms
->ms_lock
);
7037 * Verify that all segments in the vdev_checkpoint_sm are allocated
7038 * according to the checkpoint's ms_sm (i.e. are not in the checkpoint's
7041 * Do so by comparing the checkpoint space maps (vdev_checkpoint_sm) of
7042 * each vdev in the current state of the pool to the metaslab space maps
7043 * (ms_sm) of the checkpointed state of the pool.
7045 * Note that the function changes the state of the ms_allocatable
7046 * trees of the current spa_t. The entries of these ms_allocatable
7047 * trees are cleared out and then repopulated from with the free
7048 * entries of their respective ms_sm space maps.
7051 verify_checkpoint_vdev_spacemaps(spa_t
*checkpoint
, spa_t
*current
)
7053 vdev_t
*ckpoint_rvd
= checkpoint
->spa_root_vdev
;
7054 vdev_t
*current_rvd
= current
->spa_root_vdev
;
7056 load_concrete_ms_allocatable_trees(checkpoint
, SM_FREE
);
7058 for (uint64_t c
= 0; c
< ckpoint_rvd
->vdev_children
; c
++) {
7059 vdev_t
*ckpoint_vd
= ckpoint_rvd
->vdev_child
[c
];
7060 vdev_t
*current_vd
= current_rvd
->vdev_child
[c
];
7062 space_map_t
*checkpoint_sm
= NULL
;
7063 uint64_t checkpoint_sm_obj
;
7065 if (ckpoint_vd
->vdev_ops
== &vdev_indirect_ops
) {
7067 * Since we don't allow device removal in a pool
7068 * that has a checkpoint, we expect that all removed
7069 * vdevs were removed from the pool before the
7072 ASSERT3P(current_vd
->vdev_ops
, ==, &vdev_indirect_ops
);
7077 * If the checkpoint space map doesn't exist, then nothing
7078 * here is checkpointed so there's nothing to verify.
7080 if (current_vd
->vdev_top_zap
== 0 ||
7081 zap_contains(spa_meta_objset(current
),
7082 current_vd
->vdev_top_zap
,
7083 VDEV_TOP_ZAP_POOL_CHECKPOINT_SM
) != 0)
7086 VERIFY0(zap_lookup(spa_meta_objset(current
),
7087 current_vd
->vdev_top_zap
, VDEV_TOP_ZAP_POOL_CHECKPOINT_SM
,
7088 sizeof (uint64_t), 1, &checkpoint_sm_obj
));
7090 VERIFY0(space_map_open(&checkpoint_sm
, spa_meta_objset(current
),
7091 checkpoint_sm_obj
, 0, current_vd
->vdev_asize
,
7092 current_vd
->vdev_ashift
));
7094 verify_checkpoint_sm_entry_cb_arg_t vcsec
;
7095 vcsec
.vcsec_vd
= ckpoint_vd
;
7096 vcsec
.vcsec_entryid
= 0;
7097 vcsec
.vcsec_num_entries
=
7098 space_map_length(checkpoint_sm
) / sizeof (uint64_t);
7099 VERIFY0(space_map_iterate(checkpoint_sm
,
7100 space_map_length(checkpoint_sm
),
7101 verify_checkpoint_sm_entry_cb
, &vcsec
));
7102 if (dump_opt
['m'] > 3)
7103 dump_spacemap(current
->spa_meta_objset
, checkpoint_sm
);
7104 space_map_close(checkpoint_sm
);
7108 * If we've added vdevs since we took the checkpoint, ensure
7109 * that their checkpoint space maps are empty.
7111 if (ckpoint_rvd
->vdev_children
< current_rvd
->vdev_children
) {
7112 for (uint64_t c
= ckpoint_rvd
->vdev_children
;
7113 c
< current_rvd
->vdev_children
; c
++) {
7114 vdev_t
*current_vd
= current_rvd
->vdev_child
[c
];
7115 VERIFY3P(current_vd
->vdev_checkpoint_sm
, ==, NULL
);
7119 /* for cleaner progress output */
7120 (void) fprintf(stderr
, "\n");
7124 * Verifies that all space that's allocated in the checkpoint is
7125 * still allocated in the current version, by checking that everything
7126 * in checkpoint's ms_allocatable (which is actually allocated, not
7127 * allocatable/free) is not present in current's ms_allocatable.
7129 * Note that the function changes the state of the ms_allocatable
7130 * trees of both spas when called. The entries of all ms_allocatable
7131 * trees are cleared out and then repopulated from their respective
7132 * ms_sm space maps. In the checkpointed state we load the allocated
7133 * entries, and in the current state we load the free entries.
7136 verify_checkpoint_ms_spacemaps(spa_t
*checkpoint
, spa_t
*current
)
7138 vdev_t
*ckpoint_rvd
= checkpoint
->spa_root_vdev
;
7139 vdev_t
*current_rvd
= current
->spa_root_vdev
;
7141 load_concrete_ms_allocatable_trees(checkpoint
, SM_ALLOC
);
7142 load_concrete_ms_allocatable_trees(current
, SM_FREE
);
7144 for (uint64_t i
= 0; i
< ckpoint_rvd
->vdev_children
; i
++) {
7145 vdev_t
*ckpoint_vd
= ckpoint_rvd
->vdev_child
[i
];
7146 vdev_t
*current_vd
= current_rvd
->vdev_child
[i
];
7148 if (ckpoint_vd
->vdev_ops
== &vdev_indirect_ops
) {
7150 * See comment in verify_checkpoint_vdev_spacemaps()
7152 ASSERT3P(current_vd
->vdev_ops
, ==, &vdev_indirect_ops
);
7156 for (uint64_t m
= 0; m
< ckpoint_vd
->vdev_ms_count
; m
++) {
7157 metaslab_t
*ckpoint_msp
= ckpoint_vd
->vdev_ms
[m
];
7158 metaslab_t
*current_msp
= current_vd
->vdev_ms
[m
];
7160 (void) fprintf(stderr
,
7161 "\rverifying vdev %llu of %llu, "
7162 "metaslab %llu of %llu ...",
7163 (longlong_t
)current_vd
->vdev_id
,
7164 (longlong_t
)current_rvd
->vdev_children
,
7165 (longlong_t
)current_vd
->vdev_ms
[m
]->ms_id
,
7166 (longlong_t
)current_vd
->vdev_ms_count
);
7169 * We walk through the ms_allocatable trees that
7170 * are loaded with the allocated blocks from the
7171 * ms_sm spacemaps of the checkpoint. For each
7172 * one of these ranges we ensure that none of them
7173 * exists in the ms_allocatable trees of the
7174 * current state which are loaded with the ranges
7175 * that are currently free.
7177 * This way we ensure that none of the blocks that
7178 * are part of the checkpoint were freed by mistake.
7180 range_tree_walk(ckpoint_msp
->ms_allocatable
,
7181 (range_tree_func_t
*)range_tree_verify_not_present
,
7182 current_msp
->ms_allocatable
);
7186 /* for cleaner progress output */
7187 (void) fprintf(stderr
, "\n");
7191 verify_checkpoint_blocks(spa_t
*spa
)
7193 ASSERT(!dump_opt
['L']);
7195 spa_t
*checkpoint_spa
;
7196 char *checkpoint_pool
;
7200 * We import the checkpointed state of the pool (under a different
7201 * name) so we can do verification on it against the current state
7204 checkpoint_pool
= import_checkpointed_state(spa
->spa_name
, NULL
,
7206 ASSERT(strcmp(spa
->spa_name
, checkpoint_pool
) != 0);
7208 error
= spa_open(checkpoint_pool
, &checkpoint_spa
, FTAG
);
7210 fatal("Tried to open pool \"%s\" but spa_open() failed with "
7211 "error %d\n", checkpoint_pool
, error
);
7215 * Ensure that ranges in the checkpoint space maps of each vdev
7216 * are allocated according to the checkpointed state's metaslab
7219 verify_checkpoint_vdev_spacemaps(checkpoint_spa
, spa
);
7222 * Ensure that allocated ranges in the checkpoint's metaslab
7223 * space maps remain allocated in the metaslab space maps of
7224 * the current state.
7226 verify_checkpoint_ms_spacemaps(checkpoint_spa
, spa
);
7229 * Once we are done, we get rid of the checkpointed state.
7231 spa_close(checkpoint_spa
, FTAG
);
7232 free(checkpoint_pool
);
7236 dump_leftover_checkpoint_blocks(spa_t
*spa
)
7238 vdev_t
*rvd
= spa
->spa_root_vdev
;
7240 for (uint64_t i
= 0; i
< rvd
->vdev_children
; i
++) {
7241 vdev_t
*vd
= rvd
->vdev_child
[i
];
7243 space_map_t
*checkpoint_sm
= NULL
;
7244 uint64_t checkpoint_sm_obj
;
7246 if (vd
->vdev_top_zap
== 0)
7249 if (zap_contains(spa_meta_objset(spa
), vd
->vdev_top_zap
,
7250 VDEV_TOP_ZAP_POOL_CHECKPOINT_SM
) != 0)
7253 VERIFY0(zap_lookup(spa_meta_objset(spa
), vd
->vdev_top_zap
,
7254 VDEV_TOP_ZAP_POOL_CHECKPOINT_SM
,
7255 sizeof (uint64_t), 1, &checkpoint_sm_obj
));
7257 VERIFY0(space_map_open(&checkpoint_sm
, spa_meta_objset(spa
),
7258 checkpoint_sm_obj
, 0, vd
->vdev_asize
, vd
->vdev_ashift
));
7259 dump_spacemap(spa
->spa_meta_objset
, checkpoint_sm
);
7260 space_map_close(checkpoint_sm
);
7265 verify_checkpoint(spa_t
*spa
)
7267 uberblock_t checkpoint
;
7270 if (!spa_feature_is_active(spa
, SPA_FEATURE_POOL_CHECKPOINT
))
7273 error
= zap_lookup(spa
->spa_meta_objset
, DMU_POOL_DIRECTORY_OBJECT
,
7274 DMU_POOL_ZPOOL_CHECKPOINT
, sizeof (uint64_t),
7275 sizeof (uberblock_t
) / sizeof (uint64_t), &checkpoint
);
7277 if (error
== ENOENT
&& !dump_opt
['L']) {
7279 * If the feature is active but the uberblock is missing
7280 * then we must be in the middle of discarding the
7283 (void) printf("\nPartially discarded checkpoint "
7285 if (dump_opt
['m'] > 3)
7286 dump_leftover_checkpoint_blocks(spa
);
7288 } else if (error
!= 0) {
7289 (void) printf("lookup error %d when looking for "
7290 "checkpointed uberblock in MOS\n", error
);
7293 dump_uberblock(&checkpoint
, "\nCheckpointed uberblock found:\n", "\n");
7295 if (checkpoint
.ub_checkpoint_txg
== 0) {
7296 (void) printf("\nub_checkpoint_txg not set in checkpointed "
7301 if (error
== 0 && !dump_opt
['L'])
7302 verify_checkpoint_blocks(spa
);
7309 mos_leaks_cb(void *arg
, uint64_t start
, uint64_t size
)
7311 for (uint64_t i
= start
; i
< size
; i
++) {
7312 (void) printf("MOS object %llu referenced but not allocated\n",
7318 mos_obj_refd(uint64_t obj
)
7320 if (obj
!= 0 && mos_refd_objs
!= NULL
)
7321 range_tree_add(mos_refd_objs
, obj
, 1);
7325 * Call on a MOS object that may already have been referenced.
7328 mos_obj_refd_multiple(uint64_t obj
)
7330 if (obj
!= 0 && mos_refd_objs
!= NULL
&&
7331 !range_tree_contains(mos_refd_objs
, obj
, 1))
7332 range_tree_add(mos_refd_objs
, obj
, 1);
7336 mos_leak_vdev_top_zap(vdev_t
*vd
)
7338 uint64_t ms_flush_data_obj
;
7339 int error
= zap_lookup(spa_meta_objset(vd
->vdev_spa
),
7340 vd
->vdev_top_zap
, VDEV_TOP_ZAP_MS_UNFLUSHED_PHYS_TXGS
,
7341 sizeof (ms_flush_data_obj
), 1, &ms_flush_data_obj
);
7342 if (error
== ENOENT
)
7346 mos_obj_refd(ms_flush_data_obj
);
7350 mos_leak_vdev(vdev_t
*vd
)
7352 mos_obj_refd(vd
->vdev_dtl_object
);
7353 mos_obj_refd(vd
->vdev_ms_array
);
7354 mos_obj_refd(vd
->vdev_indirect_config
.vic_births_object
);
7355 mos_obj_refd(vd
->vdev_indirect_config
.vic_mapping_object
);
7356 mos_obj_refd(vd
->vdev_leaf_zap
);
7357 if (vd
->vdev_checkpoint_sm
!= NULL
)
7358 mos_obj_refd(vd
->vdev_checkpoint_sm
->sm_object
);
7359 if (vd
->vdev_indirect_mapping
!= NULL
) {
7360 mos_obj_refd(vd
->vdev_indirect_mapping
->
7361 vim_phys
->vimp_counts_object
);
7363 if (vd
->vdev_obsolete_sm
!= NULL
)
7364 mos_obj_refd(vd
->vdev_obsolete_sm
->sm_object
);
7366 for (uint64_t m
= 0; m
< vd
->vdev_ms_count
; m
++) {
7367 metaslab_t
*ms
= vd
->vdev_ms
[m
];
7368 mos_obj_refd(space_map_object(ms
->ms_sm
));
7371 if (vd
->vdev_top_zap
!= 0) {
7372 mos_obj_refd(vd
->vdev_top_zap
);
7373 mos_leak_vdev_top_zap(vd
);
7376 for (uint64_t c
= 0; c
< vd
->vdev_children
; c
++) {
7377 mos_leak_vdev(vd
->vdev_child
[c
]);
7382 mos_leak_log_spacemaps(spa_t
*spa
)
7384 uint64_t spacemap_zap
;
7385 int error
= zap_lookup(spa_meta_objset(spa
),
7386 DMU_POOL_DIRECTORY_OBJECT
, DMU_POOL_LOG_SPACEMAP_ZAP
,
7387 sizeof (spacemap_zap
), 1, &spacemap_zap
);
7388 if (error
== ENOENT
)
7392 mos_obj_refd(spacemap_zap
);
7393 for (spa_log_sm_t
*sls
= avl_first(&spa
->spa_sm_logs_by_txg
);
7394 sls
; sls
= AVL_NEXT(&spa
->spa_sm_logs_by_txg
, sls
))
7395 mos_obj_refd(sls
->sls_sm_obj
);
7399 dump_mos_leaks(spa_t
*spa
)
7402 objset_t
*mos
= spa
->spa_meta_objset
;
7403 dsl_pool_t
*dp
= spa
->spa_dsl_pool
;
7405 /* Visit and mark all referenced objects in the MOS */
7407 mos_obj_refd(DMU_POOL_DIRECTORY_OBJECT
);
7408 mos_obj_refd(spa
->spa_pool_props_object
);
7409 mos_obj_refd(spa
->spa_config_object
);
7410 mos_obj_refd(spa
->spa_ddt_stat_object
);
7411 mos_obj_refd(spa
->spa_feat_desc_obj
);
7412 mos_obj_refd(spa
->spa_feat_enabled_txg_obj
);
7413 mos_obj_refd(spa
->spa_feat_for_read_obj
);
7414 mos_obj_refd(spa
->spa_feat_for_write_obj
);
7415 mos_obj_refd(spa
->spa_history
);
7416 mos_obj_refd(spa
->spa_errlog_last
);
7417 mos_obj_refd(spa
->spa_errlog_scrub
);
7418 mos_obj_refd(spa
->spa_all_vdev_zaps
);
7419 mos_obj_refd(spa
->spa_dsl_pool
->dp_bptree_obj
);
7420 mos_obj_refd(spa
->spa_dsl_pool
->dp_tmp_userrefs_obj
);
7421 mos_obj_refd(spa
->spa_dsl_pool
->dp_scan
->scn_phys
.scn_queue_obj
);
7422 bpobj_count_refd(&spa
->spa_deferred_bpobj
);
7423 mos_obj_refd(dp
->dp_empty_bpobj
);
7424 bpobj_count_refd(&dp
->dp_obsolete_bpobj
);
7425 bpobj_count_refd(&dp
->dp_free_bpobj
);
7426 mos_obj_refd(spa
->spa_l2cache
.sav_object
);
7427 mos_obj_refd(spa
->spa_spares
.sav_object
);
7429 if (spa
->spa_syncing_log_sm
!= NULL
)
7430 mos_obj_refd(spa
->spa_syncing_log_sm
->sm_object
);
7431 mos_leak_log_spacemaps(spa
);
7433 mos_obj_refd(spa
->spa_condensing_indirect_phys
.
7434 scip_next_mapping_object
);
7435 mos_obj_refd(spa
->spa_condensing_indirect_phys
.
7436 scip_prev_obsolete_sm_object
);
7437 if (spa
->spa_condensing_indirect_phys
.scip_next_mapping_object
!= 0) {
7438 vdev_indirect_mapping_t
*vim
=
7439 vdev_indirect_mapping_open(mos
,
7440 spa
->spa_condensing_indirect_phys
.scip_next_mapping_object
);
7441 mos_obj_refd(vim
->vim_phys
->vimp_counts_object
);
7442 vdev_indirect_mapping_close(vim
);
7444 deleted_livelists_dump_mos(spa
);
7446 if (dp
->dp_origin_snap
!= NULL
) {
7449 dsl_pool_config_enter(dp
, FTAG
);
7450 VERIFY0(dsl_dataset_hold_obj(dp
,
7451 dsl_dataset_phys(dp
->dp_origin_snap
)->ds_next_snap_obj
,
7453 count_ds_mos_objects(ds
);
7454 dump_blkptr_list(&ds
->ds_deadlist
, "Deadlist");
7455 dsl_dataset_rele(ds
, FTAG
);
7456 dsl_pool_config_exit(dp
, FTAG
);
7458 count_ds_mos_objects(dp
->dp_origin_snap
);
7459 dump_blkptr_list(&dp
->dp_origin_snap
->ds_deadlist
, "Deadlist");
7461 count_dir_mos_objects(dp
->dp_mos_dir
);
7462 if (dp
->dp_free_dir
!= NULL
)
7463 count_dir_mos_objects(dp
->dp_free_dir
);
7464 if (dp
->dp_leak_dir
!= NULL
)
7465 count_dir_mos_objects(dp
->dp_leak_dir
);
7467 mos_leak_vdev(spa
->spa_root_vdev
);
7469 for (uint64_t class = 0; class < DDT_CLASSES
; class++) {
7470 for (uint64_t type
= 0; type
< DDT_TYPES
; type
++) {
7471 for (uint64_t cksum
= 0;
7472 cksum
< ZIO_CHECKSUM_FUNCTIONS
; cksum
++) {
7473 ddt_t
*ddt
= spa
->spa_ddt
[cksum
];
7474 mos_obj_refd(ddt
->ddt_object
[type
][class]);
7480 * Visit all allocated objects and make sure they are referenced.
7482 uint64_t object
= 0;
7483 while (dmu_object_next(mos
, &object
, B_FALSE
, 0) == 0) {
7484 if (range_tree_contains(mos_refd_objs
, object
, 1)) {
7485 range_tree_remove(mos_refd_objs
, object
, 1);
7487 dmu_object_info_t doi
;
7489 dmu_object_info(mos
, object
, &doi
);
7490 if (doi
.doi_type
& DMU_OT_NEWTYPE
) {
7491 dmu_object_byteswap_t bswap
=
7492 DMU_OT_BYTESWAP(doi
.doi_type
);
7493 name
= dmu_ot_byteswap
[bswap
].ob_name
;
7495 name
= dmu_ot
[doi
.doi_type
].ot_name
;
7498 (void) printf("MOS object %llu (%s) leaked\n",
7499 (u_longlong_t
)object
, name
);
7503 (void) range_tree_walk(mos_refd_objs
, mos_leaks_cb
, NULL
);
7504 if (!range_tree_is_empty(mos_refd_objs
))
7506 range_tree_vacate(mos_refd_objs
, NULL
, NULL
);
7507 range_tree_destroy(mos_refd_objs
);
7511 typedef struct log_sm_obsolete_stats_arg
{
7512 uint64_t lsos_current_txg
;
7514 uint64_t lsos_total_entries
;
7515 uint64_t lsos_valid_entries
;
7517 uint64_t lsos_sm_entries
;
7518 uint64_t lsos_valid_sm_entries
;
7519 } log_sm_obsolete_stats_arg_t
;
7522 log_spacemap_obsolete_stats_cb(spa_t
*spa
, space_map_entry_t
*sme
,
7523 uint64_t txg
, void *arg
)
7525 log_sm_obsolete_stats_arg_t
*lsos
= arg
;
7527 uint64_t offset
= sme
->sme_offset
;
7528 uint64_t vdev_id
= sme
->sme_vdev
;
7530 if (lsos
->lsos_current_txg
== 0) {
7531 /* this is the first log */
7532 lsos
->lsos_current_txg
= txg
;
7533 } else if (lsos
->lsos_current_txg
< txg
) {
7534 /* we just changed log - print stats and reset */
7535 (void) printf("%-8llu valid entries out of %-8llu - txg %llu\n",
7536 (u_longlong_t
)lsos
->lsos_valid_sm_entries
,
7537 (u_longlong_t
)lsos
->lsos_sm_entries
,
7538 (u_longlong_t
)lsos
->lsos_current_txg
);
7539 lsos
->lsos_valid_sm_entries
= 0;
7540 lsos
->lsos_sm_entries
= 0;
7541 lsos
->lsos_current_txg
= txg
;
7543 ASSERT3U(lsos
->lsos_current_txg
, ==, txg
);
7545 lsos
->lsos_sm_entries
++;
7546 lsos
->lsos_total_entries
++;
7548 vdev_t
*vd
= vdev_lookup_top(spa
, vdev_id
);
7549 if (!vdev_is_concrete(vd
))
7552 metaslab_t
*ms
= vd
->vdev_ms
[offset
>> vd
->vdev_ms_shift
];
7553 ASSERT(sme
->sme_type
== SM_ALLOC
|| sme
->sme_type
== SM_FREE
);
7555 if (txg
< metaslab_unflushed_txg(ms
))
7557 lsos
->lsos_valid_sm_entries
++;
7558 lsos
->lsos_valid_entries
++;
7563 dump_log_spacemap_obsolete_stats(spa_t
*spa
)
7565 if (!spa_feature_is_active(spa
, SPA_FEATURE_LOG_SPACEMAP
))
7568 log_sm_obsolete_stats_arg_t lsos
;
7569 bzero(&lsos
, sizeof (lsos
));
7571 (void) printf("Log Space Map Obsolete Entry Statistics:\n");
7573 iterate_through_spacemap_logs(spa
,
7574 log_spacemap_obsolete_stats_cb
, &lsos
);
7576 /* print stats for latest log */
7577 (void) printf("%-8llu valid entries out of %-8llu - txg %llu\n",
7578 (u_longlong_t
)lsos
.lsos_valid_sm_entries
,
7579 (u_longlong_t
)lsos
.lsos_sm_entries
,
7580 (u_longlong_t
)lsos
.lsos_current_txg
);
7582 (void) printf("%-8llu valid entries out of %-8llu - total\n\n",
7583 (u_longlong_t
)lsos
.lsos_valid_entries
,
7584 (u_longlong_t
)lsos
.lsos_total_entries
);
7588 dump_zpool(spa_t
*spa
)
7590 dsl_pool_t
*dp
= spa_get_dsl(spa
);
7593 if (dump_opt
['y']) {
7594 livelist_metaslab_validate(spa
);
7597 if (dump_opt
['S']) {
7598 dump_simulated_ddt(spa
);
7602 if (!dump_opt
['e'] && dump_opt
['C'] > 1) {
7603 (void) printf("\nCached configuration:\n");
7604 dump_nvlist(spa
->spa_config
, 8);
7611 dump_uberblock(&spa
->spa_uberblock
, "\nUberblock:\n", "\n");
7616 if (dump_opt
['d'] > 2 || dump_opt
['m'])
7617 dump_metaslabs(spa
);
7619 dump_metaslab_groups(spa
);
7620 if (dump_opt
['d'] > 2 || dump_opt
['m']) {
7621 dump_log_spacemaps(spa
);
7622 dump_log_spacemap_obsolete_stats(spa
);
7625 if (dump_opt
['d'] || dump_opt
['i']) {
7627 mos_refd_objs
= range_tree_create(NULL
, RANGE_SEG64
, NULL
, 0,
7629 dump_objset(dp
->dp_meta_objset
);
7631 if (dump_opt
['d'] >= 3) {
7632 dsl_pool_t
*dp
= spa
->spa_dsl_pool
;
7633 dump_full_bpobj(&spa
->spa_deferred_bpobj
,
7634 "Deferred frees", 0);
7635 if (spa_version(spa
) >= SPA_VERSION_DEADLISTS
) {
7636 dump_full_bpobj(&dp
->dp_free_bpobj
,
7637 "Pool snapshot frees", 0);
7639 if (bpobj_is_open(&dp
->dp_obsolete_bpobj
)) {
7640 ASSERT(spa_feature_is_enabled(spa
,
7641 SPA_FEATURE_DEVICE_REMOVAL
));
7642 dump_full_bpobj(&dp
->dp_obsolete_bpobj
,
7643 "Pool obsolete blocks", 0);
7646 if (spa_feature_is_active(spa
,
7647 SPA_FEATURE_ASYNC_DESTROY
)) {
7648 dump_bptree(spa
->spa_meta_objset
,
7650 "Pool dataset frees");
7652 dump_dtl(spa
->spa_root_vdev
, 0);
7655 for (spa_feature_t f
= 0; f
< SPA_FEATURES
; f
++)
7656 global_feature_count
[f
] = UINT64_MAX
;
7657 global_feature_count
[SPA_FEATURE_REDACTION_BOOKMARKS
] = 0;
7658 global_feature_count
[SPA_FEATURE_BOOKMARK_WRITTEN
] = 0;
7659 global_feature_count
[SPA_FEATURE_LIVELIST
] = 0;
7661 (void) dmu_objset_find(spa_name(spa
), dump_one_objset
,
7662 NULL
, DS_FIND_SNAPSHOTS
| DS_FIND_CHILDREN
);
7664 if (rc
== 0 && !dump_opt
['L'])
7665 rc
= dump_mos_leaks(spa
);
7667 for (f
= 0; f
< SPA_FEATURES
; f
++) {
7671 if (!(spa_feature_table
[f
].fi_flags
&
7672 ZFEATURE_FLAG_PER_DATASET
)) {
7673 if (global_feature_count
[f
] == UINT64_MAX
)
7675 if (!spa_feature_is_enabled(spa
, f
)) {
7676 ASSERT0(global_feature_count
[f
]);
7679 arr
= global_feature_count
;
7681 if (!spa_feature_is_enabled(spa
, f
)) {
7682 ASSERT0(dataset_feature_count
[f
]);
7685 arr
= dataset_feature_count
;
7687 if (feature_get_refcount(spa
, &spa_feature_table
[f
],
7688 &refcount
) == ENOTSUP
)
7690 if (arr
[f
] != refcount
) {
7691 (void) printf("%s feature refcount mismatch: "
7692 "%lld consumers != %lld refcount\n",
7693 spa_feature_table
[f
].fi_uname
,
7694 (longlong_t
)arr
[f
], (longlong_t
)refcount
);
7697 (void) printf("Verified %s feature refcount "
7698 "of %llu is correct\n",
7699 spa_feature_table
[f
].fi_uname
,
7700 (longlong_t
)refcount
);
7705 rc
= verify_device_removal_feature_counts(spa
);
7708 if (rc
== 0 && (dump_opt
['b'] || dump_opt
['c']))
7709 rc
= dump_block_stats(spa
);
7712 rc
= verify_spacemap_refcounts(spa
);
7715 show_pool_stats(spa
);
7721 rc
= verify_checkpoint(spa
);
7724 dump_debug_buffer();
7729 #define ZDB_FLAG_CHECKSUM 0x0001
7730 #define ZDB_FLAG_DECOMPRESS 0x0002
7731 #define ZDB_FLAG_BSWAP 0x0004
7732 #define ZDB_FLAG_GBH 0x0008
7733 #define ZDB_FLAG_INDIRECT 0x0010
7734 #define ZDB_FLAG_RAW 0x0020
7735 #define ZDB_FLAG_PRINT_BLKPTR 0x0040
7736 #define ZDB_FLAG_VERBOSE 0x0080
7738 static int flagbits
[256];
7739 static char flagbitstr
[16];
7742 zdb_print_blkptr(const blkptr_t
*bp
, int flags
)
7744 char blkbuf
[BP_SPRINTF_LEN
];
7746 if (flags
& ZDB_FLAG_BSWAP
)
7747 byteswap_uint64_array((void *)bp
, sizeof (blkptr_t
));
7749 snprintf_blkptr(blkbuf
, sizeof (blkbuf
), bp
);
7750 (void) printf("%s\n", blkbuf
);
7754 zdb_dump_indirect(blkptr_t
*bp
, int nbps
, int flags
)
7758 for (i
= 0; i
< nbps
; i
++)
7759 zdb_print_blkptr(&bp
[i
], flags
);
7763 zdb_dump_gbh(void *buf
, int flags
)
7765 zdb_dump_indirect((blkptr_t
*)buf
, SPA_GBH_NBLKPTRS
, flags
);
7769 zdb_dump_block_raw(void *buf
, uint64_t size
, int flags
)
7771 if (flags
& ZDB_FLAG_BSWAP
)
7772 byteswap_uint64_array(buf
, size
);
7773 VERIFY(write(fileno(stdout
), buf
, size
) == size
);
7777 zdb_dump_block(char *label
, void *buf
, uint64_t size
, int flags
)
7779 uint64_t *d
= (uint64_t *)buf
;
7780 unsigned nwords
= size
/ sizeof (uint64_t);
7781 int do_bswap
= !!(flags
& ZDB_FLAG_BSWAP
);
7788 hdr
= " 7 6 5 4 3 2 1 0 f e d c b a 9 8";
7790 hdr
= " 0 1 2 3 4 5 6 7 8 9 a b c d e f";
7792 (void) printf("\n%s\n%6s %s 0123456789abcdef\n", label
, "", hdr
);
7794 #ifdef _LITTLE_ENDIAN
7795 /* correct the endianness */
7796 do_bswap
= !do_bswap
;
7798 for (i
= 0; i
< nwords
; i
+= 2) {
7799 (void) printf("%06llx: %016llx %016llx ",
7800 (u_longlong_t
)(i
* sizeof (uint64_t)),
7801 (u_longlong_t
)(do_bswap
? BSWAP_64(d
[i
]) : d
[i
]),
7802 (u_longlong_t
)(do_bswap
? BSWAP_64(d
[i
+ 1]) : d
[i
+ 1]));
7805 for (j
= 0; j
< 2 * sizeof (uint64_t); j
++)
7806 (void) printf("%c", isprint(c
[j
]) ? c
[j
] : '.');
7807 (void) printf("\n");
7812 * There are two acceptable formats:
7813 * leaf_name - For example: c1t0d0 or /tmp/ztest.0a
7814 * child[.child]* - For example: 0.1.1
7816 * The second form can be used to specify arbitrary vdevs anywhere
7817 * in the hierarchy. For example, in a pool with a mirror of
7818 * RAID-Zs, you can specify either RAID-Z vdev with 0.0 or 0.1 .
7821 zdb_vdev_lookup(vdev_t
*vdev
, const char *path
)
7829 /* First, assume the x.x.x.x format */
7830 i
= strtoul(path
, &s
, 10);
7831 if (s
== path
|| (s
&& *s
!= '.' && *s
!= '\0'))
7833 if (i
>= vdev
->vdev_children
)
7836 vdev
= vdev
->vdev_child
[i
];
7837 if (s
&& *s
== '\0')
7839 return (zdb_vdev_lookup(vdev
, s
+1));
7842 for (i
= 0; i
< vdev
->vdev_children
; i
++) {
7843 vdev_t
*vc
= vdev
->vdev_child
[i
];
7845 if (vc
->vdev_path
== NULL
) {
7846 vc
= zdb_vdev_lookup(vc
, path
);
7853 p
= strrchr(vc
->vdev_path
, '/');
7854 p
= p
? p
+ 1 : vc
->vdev_path
;
7855 q
= &vc
->vdev_path
[strlen(vc
->vdev_path
) - 2];
7857 if (strcmp(vc
->vdev_path
, path
) == 0)
7859 if (strcmp(p
, path
) == 0)
7861 if (strcmp(q
, "s0") == 0 && strncmp(p
, path
, q
- p
) == 0)
7869 name_from_objset_id(spa_t
*spa
, uint64_t objset_id
, char *outstr
)
7873 dsl_pool_config_enter(spa
->spa_dsl_pool
, FTAG
);
7874 int error
= dsl_dataset_hold_obj(spa
->spa_dsl_pool
, objset_id
,
7877 (void) fprintf(stderr
, "failed to hold objset %llu: %s\n",
7878 (u_longlong_t
)objset_id
, strerror(error
));
7879 dsl_pool_config_exit(spa
->spa_dsl_pool
, FTAG
);
7882 dsl_dataset_name(ds
, outstr
);
7883 dsl_dataset_rele(ds
, NULL
);
7884 dsl_pool_config_exit(spa
->spa_dsl_pool
, FTAG
);
7889 zdb_parse_block_sizes(char *sizes
, uint64_t *lsize
, uint64_t *psize
)
7891 char *s0
, *s1
, *tmp
= NULL
;
7896 s0
= strtok_r(sizes
, "/", &tmp
);
7899 s1
= strtok_r(NULL
, "/", &tmp
);
7900 *lsize
= strtoull(s0
, NULL
, 16);
7901 *psize
= s1
? strtoull(s1
, NULL
, 16) : *lsize
;
7902 return (*lsize
>= *psize
&& *psize
> 0);
7905 #define ZIO_COMPRESS_MASK(alg) (1ULL << (ZIO_COMPRESS_##alg))
7908 zdb_decompress_block(abd_t
*pabd
, void *buf
, void *lbuf
, uint64_t lsize
,
7909 uint64_t psize
, int flags
)
7911 boolean_t exceeded
= B_FALSE
;
7913 * We don't know how the data was compressed, so just try
7914 * every decompress function at every inflated blocksize.
7916 void *lbuf2
= umem_alloc(SPA_MAXBLOCKSIZE
, UMEM_NOFAIL
);
7917 int cfuncs
[ZIO_COMPRESS_FUNCTIONS
] = { 0 };
7918 int *cfuncp
= cfuncs
;
7919 uint64_t maxlsize
= SPA_MAXBLOCKSIZE
;
7920 uint64_t mask
= ZIO_COMPRESS_MASK(ON
) | ZIO_COMPRESS_MASK(OFF
) |
7921 ZIO_COMPRESS_MASK(INHERIT
) | ZIO_COMPRESS_MASK(EMPTY
) |
7922 (getenv("ZDB_NO_ZLE") ? ZIO_COMPRESS_MASK(ZLE
) : 0);
7923 *cfuncp
++ = ZIO_COMPRESS_LZ4
;
7924 *cfuncp
++ = ZIO_COMPRESS_LZJB
;
7925 mask
|= ZIO_COMPRESS_MASK(LZ4
) | ZIO_COMPRESS_MASK(LZJB
);
7926 for (int c
= 0; c
< ZIO_COMPRESS_FUNCTIONS
; c
++)
7927 if (((1ULL << c
) & mask
) == 0)
7931 * On the one hand, with SPA_MAXBLOCKSIZE at 16MB, this
7932 * could take a while and we should let the user know
7933 * we are not stuck. On the other hand, printing progress
7934 * info gets old after a while. User can specify 'v' flag
7935 * to see the progression.
7938 lsize
+= SPA_MINBLOCKSIZE
;
7941 for (; lsize
<= maxlsize
; lsize
+= SPA_MINBLOCKSIZE
) {
7942 for (cfuncp
= cfuncs
; *cfuncp
; cfuncp
++) {
7943 if (flags
& ZDB_FLAG_VERBOSE
) {
7944 (void) fprintf(stderr
,
7945 "Trying %05llx -> %05llx (%s)\n",
7946 (u_longlong_t
)psize
,
7947 (u_longlong_t
)lsize
,
7948 zio_compress_table
[*cfuncp
].\
7953 * We randomize lbuf2, and decompress to both
7954 * lbuf and lbuf2. This way, we will know if
7955 * decompression fill exactly to lsize.
7957 VERIFY0(random_get_pseudo_bytes(lbuf2
, lsize
));
7959 if (zio_decompress_data(*cfuncp
, pabd
,
7960 lbuf
, psize
, lsize
, NULL
) == 0 &&
7961 zio_decompress_data(*cfuncp
, pabd
,
7962 lbuf2
, psize
, lsize
, NULL
) == 0 &&
7963 bcmp(lbuf
, lbuf2
, lsize
) == 0)
7969 umem_free(lbuf2
, SPA_MAXBLOCKSIZE
);
7971 if (lsize
> maxlsize
) {
7974 if (*cfuncp
== ZIO_COMPRESS_ZLE
) {
7975 printf("\nZLE decompression was selected. If you "
7976 "suspect the results are wrong,\ntry avoiding ZLE "
7977 "by setting and exporting ZDB_NO_ZLE=\"true\"\n");
7984 * Read a block from a pool and print it out. The syntax of the
7985 * block descriptor is:
7987 * pool:vdev_specifier:offset:[lsize/]psize[:flags]
7989 * pool - The name of the pool you wish to read from
7990 * vdev_specifier - Which vdev (see comment for zdb_vdev_lookup)
7991 * offset - offset, in hex, in bytes
7992 * size - Amount of data to read, in hex, in bytes
7993 * flags - A string of characters specifying options
7994 * b: Decode a blkptr at given offset within block
7995 * c: Calculate and display checksums
7996 * d: Decompress data before dumping
7997 * e: Byteswap data before dumping
7998 * g: Display data as a gang block header
7999 * i: Display as an indirect block
8000 * r: Dump raw data to stdout
8005 zdb_read_block(char *thing
, spa_t
*spa
)
8007 blkptr_t blk
, *bp
= &blk
;
8008 dva_t
*dva
= bp
->blk_dva
;
8010 uint64_t offset
= 0, psize
= 0, lsize
= 0, blkptr_offset
= 0;
8015 char *s
, *p
, *dup
, *vdev
, *flagstr
, *sizes
, *tmp
= NULL
;
8017 boolean_t borrowed
= B_FALSE
, found
= B_FALSE
;
8019 dup
= strdup(thing
);
8020 s
= strtok_r(dup
, ":", &tmp
);
8022 s
= strtok_r(NULL
, ":", &tmp
);
8023 offset
= strtoull(s
? s
: "", NULL
, 16);
8024 sizes
= strtok_r(NULL
, ":", &tmp
);
8025 s
= strtok_r(NULL
, ":", &tmp
);
8026 flagstr
= strdup(s
? s
: "");
8030 if (!zdb_parse_block_sizes(sizes
, &lsize
, &psize
))
8031 s
= "invalid size(s)";
8032 if (!IS_P2ALIGNED(psize
, DEV_BSIZE
) || !IS_P2ALIGNED(lsize
, DEV_BSIZE
))
8033 s
= "size must be a multiple of sector size";
8034 if (!IS_P2ALIGNED(offset
, DEV_BSIZE
))
8035 s
= "offset must be a multiple of sector size";
8037 (void) printf("Invalid block specifier: %s - %s\n", thing
, s
);
8041 for (s
= strtok_r(flagstr
, ":", &tmp
);
8043 s
= strtok_r(NULL
, ":", &tmp
)) {
8044 for (i
= 0; i
< strlen(flagstr
); i
++) {
8045 int bit
= flagbits
[(uchar_t
)flagstr
[i
]];
8048 (void) printf("***Ignoring flag: %c\n",
8049 (uchar_t
)flagstr
[i
]);
8055 p
= &flagstr
[i
+ 1];
8056 if (*p
!= ':' && *p
!= '\0') {
8057 int j
= 0, nextbit
= flagbits
[(uchar_t
)*p
];
8058 char *end
, offstr
[8] = { 0 };
8059 if ((bit
== ZDB_FLAG_PRINT_BLKPTR
) &&
8061 /* look ahead to isolate the offset */
8062 while (nextbit
== 0 &&
8063 strchr(flagbitstr
, *p
) == NULL
) {
8066 if (i
+ j
> strlen(flagstr
))
8069 nextbit
= flagbits
[(uchar_t
)*p
];
8071 blkptr_offset
= strtoull(offstr
, &end
,
8074 } else if (nextbit
== 0) {
8075 (void) printf("***Ignoring flag arg:"
8076 " '%c'\n", (uchar_t
)*p
);
8081 if (blkptr_offset
% sizeof (blkptr_t
)) {
8082 printf("Block pointer offset 0x%llx "
8083 "must be divisible by 0x%x\n",
8084 (longlong_t
)blkptr_offset
, (int)sizeof (blkptr_t
));
8087 if (found
== B_FALSE
&& strlen(flagstr
) > 0) {
8088 printf("Invalid flag arg: '%s'\n", flagstr
);
8092 vd
= zdb_vdev_lookup(spa
->spa_root_vdev
, vdev
);
8094 (void) printf("***Invalid vdev: %s\n", vdev
);
8099 (void) fprintf(stderr
, "Found vdev: %s\n",
8102 (void) fprintf(stderr
, "Found vdev type: %s\n",
8103 vd
->vdev_ops
->vdev_op_type
);
8106 pabd
= abd_alloc_for_io(SPA_MAXBLOCKSIZE
, B_FALSE
);
8107 lbuf
= umem_alloc(SPA_MAXBLOCKSIZE
, UMEM_NOFAIL
);
8111 DVA_SET_VDEV(&dva
[0], vd
->vdev_id
);
8112 DVA_SET_OFFSET(&dva
[0], offset
);
8113 DVA_SET_GANG(&dva
[0], !!(flags
& ZDB_FLAG_GBH
));
8114 DVA_SET_ASIZE(&dva
[0], vdev_psize_to_asize(vd
, psize
));
8116 BP_SET_BIRTH(bp
, TXG_INITIAL
, TXG_INITIAL
);
8118 BP_SET_LSIZE(bp
, lsize
);
8119 BP_SET_PSIZE(bp
, psize
);
8120 BP_SET_COMPRESS(bp
, ZIO_COMPRESS_OFF
);
8121 BP_SET_CHECKSUM(bp
, ZIO_CHECKSUM_OFF
);
8122 BP_SET_TYPE(bp
, DMU_OT_NONE
);
8123 BP_SET_LEVEL(bp
, 0);
8124 BP_SET_DEDUP(bp
, 0);
8125 BP_SET_BYTEORDER(bp
, ZFS_HOST_BYTEORDER
);
8127 spa_config_enter(spa
, SCL_STATE
, FTAG
, RW_READER
);
8128 zio
= zio_root(spa
, NULL
, NULL
, 0);
8130 if (vd
== vd
->vdev_top
) {
8132 * Treat this as a normal block read.
8134 zio_nowait(zio_read(zio
, spa
, bp
, pabd
, psize
, NULL
, NULL
,
8135 ZIO_PRIORITY_SYNC_READ
,
8136 ZIO_FLAG_CANFAIL
| ZIO_FLAG_RAW
, NULL
));
8139 * Treat this as a vdev child I/O.
8141 zio_nowait(zio_vdev_child_io(zio
, bp
, vd
, offset
, pabd
,
8142 psize
, ZIO_TYPE_READ
, ZIO_PRIORITY_SYNC_READ
,
8143 ZIO_FLAG_DONT_CACHE
| ZIO_FLAG_DONT_PROPAGATE
|
8144 ZIO_FLAG_DONT_RETRY
| ZIO_FLAG_CANFAIL
| ZIO_FLAG_RAW
|
8145 ZIO_FLAG_OPTIONAL
, NULL
, NULL
));
8148 error
= zio_wait(zio
);
8149 spa_config_exit(spa
, SCL_STATE
, FTAG
);
8152 (void) printf("Read of %s failed, error: %d\n", thing
, error
);
8156 uint64_t orig_lsize
= lsize
;
8158 if (flags
& ZDB_FLAG_DECOMPRESS
) {
8159 boolean_t failed
= zdb_decompress_block(pabd
, buf
, lbuf
,
8160 lsize
, psize
, flags
);
8162 (void) printf("Decompress of %s failed\n", thing
);
8166 buf
= abd_borrow_buf_copy(pabd
, lsize
);
8170 * Try to detect invalid block pointer. If invalid, try
8173 if ((flags
& ZDB_FLAG_PRINT_BLKPTR
|| flags
& ZDB_FLAG_INDIRECT
) &&
8174 !(flags
& ZDB_FLAG_DECOMPRESS
)) {
8175 const blkptr_t
*b
= (const blkptr_t
*)(void *)
8176 ((uintptr_t)buf
+ (uintptr_t)blkptr_offset
);
8177 if (zfs_blkptr_verify(spa
, b
, B_FALSE
, BLK_VERIFY_ONLY
) ==
8179 abd_return_buf_copy(pabd
, buf
, lsize
);
8182 boolean_t failed
= zdb_decompress_block(pabd
, buf
,
8183 lbuf
, lsize
, psize
, flags
);
8184 b
= (const blkptr_t
*)(void *)
8185 ((uintptr_t)buf
+ (uintptr_t)blkptr_offset
);
8186 if (failed
|| zfs_blkptr_verify(spa
, b
, B_FALSE
,
8187 BLK_VERIFY_LOG
) == B_FALSE
) {
8188 printf("invalid block pointer at this DVA\n");
8194 if (flags
& ZDB_FLAG_PRINT_BLKPTR
)
8195 zdb_print_blkptr((blkptr_t
*)(void *)
8196 ((uintptr_t)buf
+ (uintptr_t)blkptr_offset
), flags
);
8197 else if (flags
& ZDB_FLAG_RAW
)
8198 zdb_dump_block_raw(buf
, lsize
, flags
);
8199 else if (flags
& ZDB_FLAG_INDIRECT
)
8200 zdb_dump_indirect((blkptr_t
*)buf
,
8201 orig_lsize
/ sizeof (blkptr_t
), flags
);
8202 else if (flags
& ZDB_FLAG_GBH
)
8203 zdb_dump_gbh(buf
, flags
);
8205 zdb_dump_block(thing
, buf
, lsize
, flags
);
8208 * If :c was specified, iterate through the checksum table to
8209 * calculate and display each checksum for our specified
8212 if ((flags
& ZDB_FLAG_CHECKSUM
) && !(flags
& ZDB_FLAG_RAW
) &&
8213 !(flags
& ZDB_FLAG_GBH
)) {
8215 (void) printf("\n");
8216 for (enum zio_checksum ck
= ZIO_CHECKSUM_LABEL
;
8217 ck
< ZIO_CHECKSUM_FUNCTIONS
; ck
++) {
8219 if ((zio_checksum_table
[ck
].ci_flags
&
8220 ZCHECKSUM_FLAG_EMBEDDED
) ||
8221 ck
== ZIO_CHECKSUM_NOPARITY
) {
8224 BP_SET_CHECKSUM(bp
, ck
);
8225 spa_config_enter(spa
, SCL_STATE
, FTAG
, RW_READER
);
8226 czio
= zio_root(spa
, NULL
, NULL
, ZIO_FLAG_CANFAIL
);
8229 if (vd
== vd
->vdev_top
) {
8230 zio_nowait(zio_read(czio
, spa
, bp
, pabd
, psize
,
8232 ZIO_PRIORITY_SYNC_READ
,
8233 ZIO_FLAG_CANFAIL
| ZIO_FLAG_RAW
|
8234 ZIO_FLAG_DONT_RETRY
, NULL
));
8236 zio_nowait(zio_vdev_child_io(czio
, bp
, vd
,
8237 offset
, pabd
, psize
, ZIO_TYPE_READ
,
8238 ZIO_PRIORITY_SYNC_READ
,
8239 ZIO_FLAG_DONT_CACHE
|
8240 ZIO_FLAG_DONT_PROPAGATE
|
8241 ZIO_FLAG_DONT_RETRY
|
8242 ZIO_FLAG_CANFAIL
| ZIO_FLAG_RAW
|
8243 ZIO_FLAG_SPECULATIVE
|
8244 ZIO_FLAG_OPTIONAL
, NULL
, NULL
));
8246 error
= zio_wait(czio
);
8247 if (error
== 0 || error
== ECKSUM
) {
8248 zio_t
*ck_zio
= zio_root(spa
, NULL
, NULL
, 0);
8250 DVA_GET_OFFSET(&bp
->blk_dva
[0]);
8252 zio_checksum_compute(ck_zio
, ck
, pabd
, lsize
);
8253 printf("%12s\tcksum=%llx:%llx:%llx:%llx\n",
8254 zio_checksum_table
[ck
].ci_name
,
8255 (u_longlong_t
)bp
->blk_cksum
.zc_word
[0],
8256 (u_longlong_t
)bp
->blk_cksum
.zc_word
[1],
8257 (u_longlong_t
)bp
->blk_cksum
.zc_word
[2],
8258 (u_longlong_t
)bp
->blk_cksum
.zc_word
[3]);
8261 printf("error %d reading block\n", error
);
8263 spa_config_exit(spa
, SCL_STATE
, FTAG
);
8268 abd_return_buf_copy(pabd
, buf
, lsize
);
8272 umem_free(lbuf
, SPA_MAXBLOCKSIZE
);
8279 zdb_embedded_block(char *thing
)
8282 unsigned long long *words
= (void *)&bp
;
8286 bzero(&bp
, sizeof (bp
));
8287 err
= sscanf(thing
, "%llx:%llx:%llx:%llx:%llx:%llx:%llx:%llx:"
8288 "%llx:%llx:%llx:%llx:%llx:%llx:%llx:%llx",
8289 words
+ 0, words
+ 1, words
+ 2, words
+ 3,
8290 words
+ 4, words
+ 5, words
+ 6, words
+ 7,
8291 words
+ 8, words
+ 9, words
+ 10, words
+ 11,
8292 words
+ 12, words
+ 13, words
+ 14, words
+ 15);
8294 (void) fprintf(stderr
, "invalid input format\n");
8297 ASSERT3U(BPE_GET_LSIZE(&bp
), <=, SPA_MAXBLOCKSIZE
);
8298 buf
= malloc(SPA_MAXBLOCKSIZE
);
8300 (void) fprintf(stderr
, "out of memory\n");
8303 err
= decode_embedded_bp(&bp
, buf
, BPE_GET_LSIZE(&bp
));
8305 (void) fprintf(stderr
, "decode failed: %u\n", err
);
8308 zdb_dump_block_raw(buf
, BPE_GET_LSIZE(&bp
), 0);
8313 main(int argc
, char **argv
)
8316 struct rlimit rl
= { 1024, 1024 };
8318 objset_t
*os
= NULL
;
8322 char **searchdirs
= NULL
;
8324 char *target
, *target_pool
, dsname
[ZFS_MAX_DATASET_NAME_LEN
];
8325 nvlist_t
*policy
= NULL
;
8326 uint64_t max_txg
= UINT64_MAX
;
8327 int64_t objset_id
= -1;
8329 int flags
= ZFS_IMPORT_MISSING_LOG
;
8330 int rewind
= ZPOOL_NEVER_REWIND
;
8331 char *spa_config_path_env
, *objset_str
;
8332 boolean_t target_is_spa
= B_TRUE
, dataset_lookup
= B_FALSE
;
8333 nvlist_t
*cfg
= NULL
;
8335 (void) setrlimit(RLIMIT_NOFILE
, &rl
);
8336 (void) enable_extended_FILE_stdio(-1, -1);
8338 dprintf_setup(&argc
, argv
);
8341 * If there is an environment variable SPA_CONFIG_PATH it overrides
8342 * default spa_config_path setting. If -U flag is specified it will
8343 * override this environment variable settings once again.
8345 spa_config_path_env
= getenv("SPA_CONFIG_PATH");
8346 if (spa_config_path_env
!= NULL
)
8347 spa_config_path
= spa_config_path_env
;
8350 * For performance reasons, we set this tunable down. We do so before
8351 * the arg parsing section so that the user can override this value if
8354 zfs_btree_verify_intensity
= 3;
8356 while ((c
= getopt(argc
, argv
,
8357 "AbcCdDeEFGhiI:klLmMo:Op:PqrRsSt:uU:vVx:XYyZ")) != -1) {
8393 zfs_reconstruct_indirect_combinations_max
= INT_MAX
;
8394 zfs_deadman_enabled
= 0;
8396 /* NB: Sort single match options below. */
8398 max_inflight_bytes
= strtoull(optarg
, NULL
, 0);
8399 if (max_inflight_bytes
== 0) {
8400 (void) fprintf(stderr
, "maximum number "
8401 "of inflight bytes must be greater "
8407 error
= set_global_var(optarg
);
8412 if (searchdirs
== NULL
) {
8413 searchdirs
= umem_alloc(sizeof (char *),
8416 char **tmp
= umem_alloc((nsearch
+ 1) *
8417 sizeof (char *), UMEM_NOFAIL
);
8418 bcopy(searchdirs
, tmp
, nsearch
*
8420 umem_free(searchdirs
,
8421 nsearch
* sizeof (char *));
8424 searchdirs
[nsearch
++] = optarg
;
8427 max_txg
= strtoull(optarg
, NULL
, 0);
8428 if (max_txg
< TXG_INITIAL
) {
8429 (void) fprintf(stderr
, "incorrect txg "
8430 "specified: %s\n", optarg
);
8435 spa_config_path
= optarg
;
8436 if (spa_config_path
[0] != '/') {
8437 (void) fprintf(stderr
,
8438 "cachefile must be an absolute path "
8439 "(i.e. start with a slash)\n");
8447 flags
= ZFS_IMPORT_VERBATIM
;
8450 vn_dumpdir
= optarg
;
8458 if (!dump_opt
['e'] && searchdirs
!= NULL
) {
8459 (void) fprintf(stderr
, "-p option requires use of -e\n");
8462 if (dump_opt
['d'] || dump_opt
['r']) {
8463 /* <pool>[/<dataset | objset id> is accepted */
8464 if (argv
[2] && (objset_str
= strchr(argv
[2], '/')) != NULL
&&
8465 objset_str
++ != NULL
) {
8468 objset_id
= strtoull(objset_str
, &endptr
, 0);
8469 /* dataset 0 is the same as opening the pool */
8470 if (errno
== 0 && endptr
!= objset_str
&&
8472 target_is_spa
= B_FALSE
;
8473 dataset_lookup
= B_TRUE
;
8474 } else if (objset_id
!= 0) {
8475 printf("failed to open objset %s "
8476 "%llu %s", objset_str
,
8477 (u_longlong_t
)objset_id
,
8481 /* normal dataset name not an objset ID */
8482 if (endptr
== objset_str
) {
8490 * ZDB does not typically re-read blocks; therefore limit the ARC
8491 * to 256 MB, which can be used entirely for metadata.
8493 zfs_arc_min
= zfs_arc_meta_min
= 2ULL << SPA_MAXBLOCKSHIFT
;
8494 zfs_arc_max
= zfs_arc_meta_limit
= 256 * 1024 * 1024;
8498 * "zdb -c" uses checksum-verifying scrub i/os which are async reads.
8499 * "zdb -b" uses traversal prefetch which uses async reads.
8500 * For good performance, let several of them be active at once.
8502 zfs_vdev_async_read_max_active
= 10;
8505 * Disable reference tracking for better performance.
8507 reference_tracking_enable
= B_FALSE
;
8510 * Do not fail spa_load when spa_load_verify fails. This is needed
8511 * to load non-idle pools.
8513 spa_load_verify_dryrun
= B_TRUE
;
8515 kernel_init(SPA_MODE_READ
);
8518 verbose
= MAX(verbose
, 1);
8520 for (c
= 0; c
< 256; c
++) {
8521 if (dump_all
&& strchr("AeEFklLOPrRSXy", c
) == NULL
)
8524 dump_opt
[c
] += verbose
;
8527 libspl_assert_ok
= (dump_opt
['A'] == 1) || (dump_opt
['A'] > 2);
8528 zfs_recover
= (dump_opt
['A'] > 1);
8532 if (argc
< 2 && dump_opt
['R'])
8535 if (dump_opt
['E']) {
8538 zdb_embedded_block(argv
[0]);
8543 if (!dump_opt
['e'] && dump_opt
['C']) {
8544 dump_cachefile(spa_config_path
);
8551 return (dump_label(argv
[0]));
8553 if (dump_opt
['O']) {
8556 dump_opt
['v'] = verbose
+ 3;
8557 return (dump_path(argv
[0], argv
[1], NULL
));
8559 if (dump_opt
['r']) {
8562 dump_opt
['v'] = verbose
;
8563 error
= dump_path(argv
[0], argv
[1], &object
);
8566 if (dump_opt
['X'] || dump_opt
['F'])
8567 rewind
= ZPOOL_DO_REWIND
|
8568 (dump_opt
['X'] ? ZPOOL_EXTREME_REWIND
: 0);
8570 if (nvlist_alloc(&policy
, NV_UNIQUE_NAME_TYPE
, 0) != 0 ||
8571 nvlist_add_uint64(policy
, ZPOOL_LOAD_REQUEST_TXG
, max_txg
) != 0 ||
8572 nvlist_add_uint32(policy
, ZPOOL_LOAD_REWIND_POLICY
, rewind
) != 0)
8573 fatal("internal error: %s", strerror(ENOMEM
));
8578 if (strpbrk(target
, "/@") != NULL
) {
8581 target_pool
= strdup(target
);
8582 *strpbrk(target_pool
, "/@") = '\0';
8584 target_is_spa
= B_FALSE
;
8585 targetlen
= strlen(target
);
8586 if (targetlen
&& target
[targetlen
- 1] == '/')
8587 target
[targetlen
- 1] = '\0';
8589 target_pool
= target
;
8592 if (dump_opt
['e']) {
8593 importargs_t args
= { 0 };
8595 args
.paths
= nsearch
;
8596 args
.path
= searchdirs
;
8597 args
.can_be_active
= B_TRUE
;
8599 error
= zpool_find_config(NULL
, target_pool
, &cfg
, &args
,
8600 &libzpool_config_ops
);
8604 if (nvlist_add_nvlist(cfg
,
8605 ZPOOL_LOAD_POLICY
, policy
) != 0) {
8606 fatal("can't open '%s': %s",
8607 target
, strerror(ENOMEM
));
8610 if (dump_opt
['C'] > 1) {
8611 (void) printf("\nConfiguration for import:\n");
8612 dump_nvlist(cfg
, 8);
8616 * Disable the activity check to allow examination of
8619 error
= spa_import(target_pool
, cfg
, NULL
,
8620 flags
| ZFS_IMPORT_SKIP_MMP
);
8624 if (searchdirs
!= NULL
) {
8625 umem_free(searchdirs
, nsearch
* sizeof (char *));
8630 * import_checkpointed_state makes the assumption that the
8631 * target pool that we pass it is already part of the spa
8632 * namespace. Because of that we need to make sure to call
8633 * it always after the -e option has been processed, which
8634 * imports the pool to the namespace if it's not in the
8637 char *checkpoint_pool
= NULL
;
8638 char *checkpoint_target
= NULL
;
8639 if (dump_opt
['k']) {
8640 checkpoint_pool
= import_checkpointed_state(target
, cfg
,
8641 &checkpoint_target
);
8643 if (checkpoint_target
!= NULL
)
8644 target
= checkpoint_target
;
8652 if (target_pool
!= target
)
8656 if (dump_opt
['k'] && (target_is_spa
|| dump_opt
['R'])) {
8657 ASSERT(checkpoint_pool
!= NULL
);
8658 ASSERT(checkpoint_target
== NULL
);
8660 error
= spa_open(checkpoint_pool
, &spa
, FTAG
);
8662 fatal("Tried to open pool \"%s\" but "
8663 "spa_open() failed with error %d\n",
8664 checkpoint_pool
, error
);
8667 } else if (target_is_spa
|| dump_opt
['R'] || objset_id
== 0) {
8668 zdb_set_skip_mmp(target
);
8669 error
= spa_open_rewind(target
, &spa
, FTAG
, policy
,
8673 * If we're missing the log device then
8674 * try opening the pool after clearing the
8677 mutex_enter(&spa_namespace_lock
);
8678 if ((spa
= spa_lookup(target
)) != NULL
&&
8679 spa
->spa_log_state
== SPA_LOG_MISSING
) {
8680 spa
->spa_log_state
= SPA_LOG_CLEAR
;
8683 mutex_exit(&spa_namespace_lock
);
8686 error
= spa_open_rewind(target
, &spa
,
8687 FTAG
, policy
, NULL
);
8690 } else if (strpbrk(target
, "#") != NULL
) {
8692 error
= dsl_pool_hold(target
, FTAG
, &dp
);
8694 fatal("can't dump '%s': %s", target
,
8697 error
= dump_bookmark(dp
, target
, B_TRUE
, verbose
> 1);
8698 dsl_pool_rele(dp
, FTAG
);
8700 fatal("can't dump '%s': %s", target
,
8705 zdb_set_skip_mmp(target
);
8706 if (dataset_lookup
== B_TRUE
) {
8708 * Use the supplied id to get the name
8711 error
= spa_open(target
, &spa
, FTAG
);
8713 error
= name_from_objset_id(spa
,
8715 spa_close(spa
, FTAG
);
8721 error
= open_objset(target
, FTAG
, &os
);
8723 spa
= dmu_objset_spa(os
);
8726 nvlist_free(policy
);
8729 fatal("can't open '%s': %s", target
, strerror(error
));
8732 * Set the pool failure mode to panic in order to prevent the pool
8733 * from suspending. A suspended I/O will have no way to resume and
8734 * can prevent the zdb(8) command from terminating as expected.
8737 spa
->spa_failmode
= ZIO_FAILURE_MODE_PANIC
;
8741 if (dump_opt
['r']) {
8742 error
= zdb_copy_object(os
, object
, argv
[1]);
8743 } else if (!dump_opt
['R']) {
8744 flagbits
['d'] = ZOR_FLAG_DIRECTORY
;
8745 flagbits
['f'] = ZOR_FLAG_PLAIN_FILE
;
8746 flagbits
['m'] = ZOR_FLAG_SPACE_MAP
;
8747 flagbits
['z'] = ZOR_FLAG_ZAP
;
8748 flagbits
['A'] = ZOR_FLAG_ALL_TYPES
;
8750 if (argc
> 0 && dump_opt
['d']) {
8751 zopt_object_args
= argc
;
8752 zopt_object_ranges
= calloc(zopt_object_args
,
8753 sizeof (zopt_object_range_t
));
8754 for (unsigned i
= 0; i
< zopt_object_args
; i
++) {
8758 err
= parse_object_range(argv
[i
],
8759 &zopt_object_ranges
[i
], &msg
);
8761 fatal("Bad object or range: '%s': %s\n",
8762 argv
[i
], msg
? msg
: "");
8764 } else if (argc
> 0 && dump_opt
['m']) {
8765 zopt_metaslab_args
= argc
;
8766 zopt_metaslab
= calloc(zopt_metaslab_args
,
8768 for (unsigned i
= 0; i
< zopt_metaslab_args
; i
++) {
8770 zopt_metaslab
[i
] = strtoull(argv
[i
], NULL
, 0);
8771 if (zopt_metaslab
[i
] == 0 && errno
!= 0)
8772 fatal("bad number %s: %s", argv
[i
],
8778 } else if (zopt_object_args
> 0 && !dump_opt
['m']) {
8779 dump_objset(spa
->spa_meta_objset
);
8784 flagbits
['b'] = ZDB_FLAG_PRINT_BLKPTR
;
8785 flagbits
['c'] = ZDB_FLAG_CHECKSUM
;
8786 flagbits
['d'] = ZDB_FLAG_DECOMPRESS
;
8787 flagbits
['e'] = ZDB_FLAG_BSWAP
;
8788 flagbits
['g'] = ZDB_FLAG_GBH
;
8789 flagbits
['i'] = ZDB_FLAG_INDIRECT
;
8790 flagbits
['r'] = ZDB_FLAG_RAW
;
8791 flagbits
['v'] = ZDB_FLAG_VERBOSE
;
8793 for (int i
= 0; i
< argc
; i
++)
8794 zdb_read_block(argv
[i
], spa
);
8797 if (dump_opt
['k']) {
8798 free(checkpoint_pool
);
8800 free(checkpoint_target
);
8804 close_objset(os
, FTAG
);
8806 spa_close(spa
, FTAG
);
8809 fuid_table_destroy();
8811 dump_debug_buffer();