zdb.c revision 8248818d5849649ef734d62da097e90222a23763
1/*
2 * CDDL HEADER START
3 *
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.
7 *
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.
12 *
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]
18 *
19 * CDDL HEADER END
20 */
21/*
22 * Copyright 2006 Sun Microsystems, Inc.  All rights reserved.
23 * Use is subject to license terms.
24 */
25
26#pragma ident	"%Z%%M%	%I%	%E% SMI"
27
28#include <stdio.h>
29#include <stdlib.h>
30#include <sys/zfs_context.h>
31#include <sys/spa.h>
32#include <sys/spa_impl.h>
33#include <sys/dmu.h>
34#include <sys/zap.h>
35#include <sys/fs/zfs.h>
36#include <sys/zfs_znode.h>
37#include <sys/vdev.h>
38#include <sys/vdev_impl.h>
39#include <sys/metaslab_impl.h>
40#include <sys/dmu_objset.h>
41#include <sys/dsl_dir.h>
42#include <sys/dsl_dataset.h>
43#include <sys/dsl_pool.h>
44#include <sys/dbuf.h>
45#include <sys/zil.h>
46#include <sys/zil_impl.h>
47#include <sys/stat.h>
48#include <sys/resource.h>
49#include <sys/dmu_traverse.h>
50#include <sys/zio_checksum.h>
51#include <sys/zio_compress.h>
52
53const char cmdname[] = "zdb";
54uint8_t dump_opt[256];
55
56typedef void object_viewer_t(objset_t *, uint64_t, void *data, size_t size);
57
58extern void dump_intent_log(zilog_t *);
59uint64_t *zopt_object = NULL;
60int zopt_objects = 0;
61int zdb_advance = ADVANCE_PRE;
62zbookmark_t zdb_noread = { 0, 0, ZB_NO_LEVEL, 0 };
63
64/*
65 * These libumem hooks provide a reasonable set of defaults for the allocator's
66 * debugging facilities.
67 */
68const char *
69_umem_debug_init()
70{
71	return ("default,verbose"); /* $UMEM_DEBUG setting */
72}
73
74const char *
75_umem_logging_init(void)
76{
77	return ("fail,contents"); /* $UMEM_LOGGING setting */
78}
79
80static void
81usage(void)
82{
83	(void) fprintf(stderr,
84	    "Usage: %s [-udibcsvLU] [-O order] [-B os:obj:level:blkid] "
85	    "dataset [object...]\n"
86	    "       %s -C [pool]\n"
87	    "       %s -l dev\n",
88	    cmdname, cmdname, cmdname);
89
90	(void) fprintf(stderr, "	-u uberblock\n");
91	(void) fprintf(stderr, "	-d datasets\n");
92	(void) fprintf(stderr, "        -C cached pool configuration\n");
93	(void) fprintf(stderr, "	-i intent logs\n");
94	(void) fprintf(stderr, "	-b block statistics\n");
95	(void) fprintf(stderr, "	-c checksum all data blocks\n");
96	(void) fprintf(stderr, "	-s report stats on zdb's I/O\n");
97	(void) fprintf(stderr, "	-v verbose (applies to all others)\n");
98	(void) fprintf(stderr, "        -l dump label contents\n");
99	(void) fprintf(stderr, "	-L live pool (allows some errors)\n");
100	(void) fprintf(stderr, "	-O [!]<pre|post|prune|data|holes> "
101	    "visitation order\n");
102	(void) fprintf(stderr, "	-U use zpool.cache in /tmp\n");
103	(void) fprintf(stderr, "	-B objset:object:level:blkid -- "
104	    "simulate bad block\n");
105	(void) fprintf(stderr, "Specify an option more than once (e.g. -bb) "
106	    "to make only that option verbose\n");
107	(void) fprintf(stderr, "Default is to dump everything non-verbosely\n");
108	exit(1);
109}
110
111static void
112fatal(const char *fmt, ...)
113{
114	va_list ap;
115
116	va_start(ap, fmt);
117	(void) fprintf(stderr, "%s: ", cmdname);
118	(void) vfprintf(stderr, fmt, ap);
119	va_end(ap);
120	(void) fprintf(stderr, "\n");
121
122	exit(1);
123}
124
125static void
126dump_nvlist(nvlist_t *list, int indent)
127{
128	nvpair_t *elem = NULL;
129
130	while ((elem = nvlist_next_nvpair(list, elem)) != NULL) {
131		switch (nvpair_type(elem)) {
132		case DATA_TYPE_STRING:
133			{
134				char *value;
135
136				VERIFY(nvpair_value_string(elem, &value) == 0);
137				(void) printf("%*s%s='%s'\n", indent, "",
138				    nvpair_name(elem), value);
139			}
140			break;
141
142		case DATA_TYPE_UINT64:
143			{
144				uint64_t value;
145
146				VERIFY(nvpair_value_uint64(elem, &value) == 0);
147				(void) printf("%*s%s=%llu\n", indent, "",
148				    nvpair_name(elem), (u_longlong_t)value);
149			}
150			break;
151
152		case DATA_TYPE_NVLIST:
153			{
154				nvlist_t *value;
155
156				VERIFY(nvpair_value_nvlist(elem, &value) == 0);
157				(void) printf("%*s%s\n", indent, "",
158				    nvpair_name(elem));
159				dump_nvlist(value, indent + 4);
160			}
161			break;
162
163		case DATA_TYPE_NVLIST_ARRAY:
164			{
165				nvlist_t **value;
166				uint_t c, count;
167
168				VERIFY(nvpair_value_nvlist_array(elem, &value,
169				    &count) == 0);
170
171				for (c = 0; c < count; c++) {
172					(void) printf("%*s%s[%u]\n", indent, "",
173					    nvpair_name(elem), c);
174					dump_nvlist(value[c], indent + 8);
175				}
176			}
177			break;
178
179		default:
180
181			(void) printf("bad config type %d for %s\n",
182			    nvpair_type(elem), nvpair_name(elem));
183		}
184	}
185}
186
187/* ARGSUSED */
188static void
189dump_packed_nvlist(objset_t *os, uint64_t object, void *data, size_t size)
190{
191	nvlist_t *nv;
192	size_t nvsize = *(uint64_t *)data;
193	char *packed = umem_alloc(nvsize, UMEM_NOFAIL);
194
195	VERIFY(0 == dmu_read(os, object, 0, nvsize, packed));
196
197	VERIFY(nvlist_unpack(packed, nvsize, &nv, 0) == 0);
198
199	umem_free(packed, nvsize);
200
201	dump_nvlist(nv, 8);
202
203	nvlist_free(nv);
204}
205
206const char dump_zap_stars[] = "****************************************";
207const int dump_zap_width = sizeof (dump_zap_stars) - 1;
208
209static void
210dump_zap_histogram(uint64_t histo[ZAP_HISTOGRAM_SIZE])
211{
212	int i;
213	int minidx = ZAP_HISTOGRAM_SIZE - 1;
214	int maxidx = 0;
215	uint64_t max = 0;
216
217	for (i = 0; i < ZAP_HISTOGRAM_SIZE; i++) {
218		if (histo[i] > max)
219			max = histo[i];
220		if (histo[i] > 0 && i > maxidx)
221			maxidx = i;
222		if (histo[i] > 0 && i < minidx)
223			minidx = i;
224	}
225
226	if (max < dump_zap_width)
227		max = dump_zap_width;
228
229	for (i = minidx; i <= maxidx; i++)
230		(void) printf("\t\t\t%u: %6llu %s\n", i, (u_longlong_t)histo[i],
231		    &dump_zap_stars[(max - histo[i]) * dump_zap_width / max]);
232}
233
234static void
235dump_zap_stats(objset_t *os, uint64_t object)
236{
237	int error;
238	zap_stats_t zs;
239
240	error = zap_get_stats(os, object, &zs);
241	if (error)
242		return;
243
244	if (zs.zs_ptrtbl_len == 0) {
245		ASSERT(zs.zs_num_blocks == 1);
246		(void) printf("\tmicrozap: %llu bytes, %llu entries\n",
247		    (u_longlong_t)zs.zs_blocksize,
248		    (u_longlong_t)zs.zs_num_entries);
249		return;
250	}
251
252	(void) printf("\tFat ZAP stats:\n");
253
254	(void) printf("\t\tPointer table:\n");
255	(void) printf("\t\t\t%llu elements\n",
256	    (u_longlong_t)zs.zs_ptrtbl_len);
257	(void) printf("\t\t\tzt_blk: %llu\n",
258	    (u_longlong_t)zs.zs_ptrtbl_zt_blk);
259	(void) printf("\t\t\tzt_numblks: %llu\n",
260	    (u_longlong_t)zs.zs_ptrtbl_zt_numblks);
261	(void) printf("\t\t\tzt_shift: %llu\n",
262	    (u_longlong_t)zs.zs_ptrtbl_zt_shift);
263	(void) printf("\t\t\tzt_blks_copied: %llu\n",
264	    (u_longlong_t)zs.zs_ptrtbl_blks_copied);
265	(void) printf("\t\t\tzt_nextblk: %llu\n",
266	    (u_longlong_t)zs.zs_ptrtbl_nextblk);
267
268	(void) printf("\t\tZAP entries: %llu\n",
269	    (u_longlong_t)zs.zs_num_entries);
270	(void) printf("\t\tLeaf blocks: %llu\n",
271	    (u_longlong_t)zs.zs_num_leafs);
272	(void) printf("\t\tTotal blocks: %llu\n",
273	    (u_longlong_t)zs.zs_num_blocks);
274	(void) printf("\t\tzap_block_type: 0x%llx\n",
275	    (u_longlong_t)zs.zs_block_type);
276	(void) printf("\t\tzap_magic: 0x%llx\n",
277	    (u_longlong_t)zs.zs_magic);
278	(void) printf("\t\tzap_salt: 0x%llx\n",
279	    (u_longlong_t)zs.zs_salt);
280
281	(void) printf("\t\tLeafs with 2^n pointers:\n");
282	dump_zap_histogram(zs.zs_leafs_with_2n_pointers);
283
284	(void) printf("\t\tBlocks with n*5 entries:\n");
285	dump_zap_histogram(zs.zs_blocks_with_n5_entries);
286
287	(void) printf("\t\tBlocks n/10 full:\n");
288	dump_zap_histogram(zs.zs_blocks_n_tenths_full);
289
290	(void) printf("\t\tEntries with n chunks:\n");
291	dump_zap_histogram(zs.zs_entries_using_n_chunks);
292
293	(void) printf("\t\tBuckets with n entries:\n");
294	dump_zap_histogram(zs.zs_buckets_with_n_entries);
295}
296
297/*ARGSUSED*/
298static void
299dump_none(objset_t *os, uint64_t object, void *data, size_t size)
300{
301}
302
303/*ARGSUSED*/
304void
305dump_uint8(objset_t *os, uint64_t object, void *data, size_t size)
306{
307}
308
309/*ARGSUSED*/
310static void
311dump_uint64(objset_t *os, uint64_t object, void *data, size_t size)
312{
313}
314
315/*ARGSUSED*/
316static void
317dump_zap(objset_t *os, uint64_t object, void *data, size_t size)
318{
319	zap_cursor_t zc;
320	zap_attribute_t attr;
321	void *prop;
322	int i;
323
324	dump_zap_stats(os, object);
325	(void) printf("\n");
326
327	for (zap_cursor_init(&zc, os, object);
328	    zap_cursor_retrieve(&zc, &attr) == 0;
329	    zap_cursor_advance(&zc)) {
330		(void) printf("\t\t%s = ", attr.za_name);
331		if (attr.za_num_integers == 0) {
332			(void) printf("\n");
333			continue;
334		}
335		prop = umem_zalloc(attr.za_num_integers *
336		    attr.za_integer_length, UMEM_NOFAIL);
337		(void) zap_lookup(os, object, attr.za_name,
338		    attr.za_integer_length, attr.za_num_integers, prop);
339		if (attr.za_integer_length == 1) {
340			(void) printf("%s", (char *)prop);
341		} else {
342			for (i = 0; i < attr.za_num_integers; i++) {
343				switch (attr.za_integer_length) {
344				case 2:
345					(void) printf("%u ",
346					    ((uint16_t *)prop)[i]);
347					break;
348				case 4:
349					(void) printf("%u ",
350					    ((uint32_t *)prop)[i]);
351					break;
352				case 8:
353					(void) printf("%lld ",
354					    (u_longlong_t)((int64_t *)prop)[i]);
355					break;
356				}
357			}
358		}
359		(void) printf("\n");
360		umem_free(prop, attr.za_num_integers * attr.za_integer_length);
361	}
362	zap_cursor_fini(&zc);
363}
364
365static void
366dump_spacemap(objset_t *os, space_map_obj_t *smo, space_map_t *sm)
367{
368	uint64_t alloc, offset, entry;
369	int mapshift = sm->sm_shift;
370	uint64_t mapstart = sm->sm_start;
371	char *ddata[] = { "ALLOC", "FREE", "CONDENSE", "INVALID" };
372
373	if (smo->smo_object == 0)
374		return;
375
376	/*
377	 * Print out the freelist entries in both encoded and decoded form.
378	 */
379	alloc = 0;
380	for (offset = 0; offset < smo->smo_objsize; offset += sizeof (entry)) {
381		VERIFY(0 == dmu_read(os, smo->smo_object, offset,
382		    sizeof (entry), &entry));
383		if (SM_DEBUG_DECODE(entry)) {
384			(void) printf("\t\t[%4llu] %s: txg %llu, pass %llu\n",
385			    (u_longlong_t)(offset / sizeof (entry)),
386			    ddata[SM_DEBUG_ACTION_DECODE(entry)],
387			    SM_DEBUG_TXG_DECODE(entry),
388			    SM_DEBUG_SYNCPASS_DECODE(entry));
389		} else {
390			(void) printf("\t\t[%4llu]    %c  range:"
391			    " %08llx-%08llx  size: %06llx\n",
392			    (u_longlong_t)(offset / sizeof (entry)),
393			    SM_TYPE_DECODE(entry) == SM_ALLOC ? 'A' : 'F',
394			    (SM_OFFSET_DECODE(entry) << mapshift) + mapstart,
395			    (SM_OFFSET_DECODE(entry) << mapshift) + mapstart +
396			    (SM_RUN_DECODE(entry) << mapshift),
397			    (SM_RUN_DECODE(entry) << mapshift));
398			if (SM_TYPE_DECODE(entry) == SM_ALLOC)
399				alloc += SM_RUN_DECODE(entry) << mapshift;
400			else
401				alloc -= SM_RUN_DECODE(entry) << mapshift;
402		}
403	}
404	if (alloc != smo->smo_alloc) {
405		(void) printf("space_map_object alloc (%llu) INCONSISTENT "
406		    "with space map summary (%llu)\n",
407		    (u_longlong_t)smo->smo_alloc, (u_longlong_t)alloc);
408	}
409}
410
411static void
412dump_metaslab(metaslab_t *msp)
413{
414	char freebuf[5];
415	space_map_obj_t *smo = msp->ms_smo;
416	vdev_t *vd = msp->ms_group->mg_vd;
417	spa_t *spa = vd->vdev_spa;
418
419	nicenum(msp->ms_map.sm_size - smo->smo_alloc, freebuf);
420
421	if (dump_opt['d'] <= 5) {
422		(void) printf("\t%10llx   %10llu   %5s\n",
423		    (u_longlong_t)msp->ms_map.sm_start,
424		    (u_longlong_t)smo->smo_object,
425		    freebuf);
426		return;
427	}
428
429	(void) printf(
430	    "\tvdev %llu   offset %08llx   spacemap %4llu   free %5s\n",
431	    (u_longlong_t)vd->vdev_id, (u_longlong_t)msp->ms_map.sm_start,
432	    (u_longlong_t)smo->smo_object, freebuf);
433
434	ASSERT(msp->ms_map.sm_size == (1ULL << vd->vdev_ms_shift));
435
436	dump_spacemap(spa->spa_meta_objset, smo, &msp->ms_map);
437}
438
439static void
440dump_metaslabs(spa_t *spa)
441{
442	vdev_t *rvd = spa->spa_root_vdev;
443	vdev_t *vd;
444	int c, m;
445
446	(void) printf("\nMetaslabs:\n");
447
448	for (c = 0; c < rvd->vdev_children; c++) {
449		vd = rvd->vdev_child[c];
450
451		spa_config_enter(spa, RW_READER, FTAG);
452		(void) printf("\n    vdev %llu = %s\n\n",
453		    (u_longlong_t)vd->vdev_id, vdev_description(vd));
454		spa_config_exit(spa, FTAG);
455
456		if (dump_opt['d'] <= 5) {
457			(void) printf("\t%10s   %10s   %5s\n",
458			    "offset", "spacemap", "free");
459			(void) printf("\t%10s   %10s   %5s\n",
460			    "------", "--------", "----");
461		}
462		for (m = 0; m < vd->vdev_ms_count; m++)
463			dump_metaslab(vd->vdev_ms[m]);
464		(void) printf("\n");
465	}
466}
467
468static void
469dump_dtl(vdev_t *vd, int indent)
470{
471	avl_tree_t *t = &vd->vdev_dtl_map.sm_root;
472	spa_t *spa = vd->vdev_spa;
473	space_seg_t *ss;
474	vdev_t *pvd;
475	int c;
476
477	if (indent == 0)
478		(void) printf("\nDirty time logs:\n\n");
479
480	spa_config_enter(spa, RW_READER, FTAG);
481	(void) printf("\t%*s%s\n", indent, "", vdev_description(vd));
482	spa_config_exit(spa, FTAG);
483
484	for (ss = avl_first(t); ss; ss = AVL_NEXT(t, ss)) {
485		/*
486		 * Everything in this DTL must appear in all parent DTL unions.
487		 */
488		for (pvd = vd; pvd; pvd = pvd->vdev_parent)
489			ASSERT(vdev_dtl_contains(&pvd->vdev_dtl_map,
490			    ss->ss_start, ss->ss_end - ss->ss_start));
491		(void) printf("\t%*soutage [%llu,%llu] length %llu\n",
492		    indent, "",
493		    (u_longlong_t)ss->ss_start,
494		    (u_longlong_t)ss->ss_end - 1,
495		    (u_longlong_t)ss->ss_end - ss->ss_start);
496	}
497
498	(void) printf("\n");
499
500	if (dump_opt['d'] > 5 && vd->vdev_children == 0) {
501		dump_spacemap(vd->vdev_spa->spa_meta_objset, &vd->vdev_dtl,
502		    &vd->vdev_dtl_map);
503		(void) printf("\n");
504	}
505
506	for (c = 0; c < vd->vdev_children; c++)
507		dump_dtl(vd->vdev_child[c], indent + 4);
508}
509
510/*ARGSUSED*/
511static void
512dump_dnode(objset_t *os, uint64_t object, void *data, size_t size)
513{
514}
515
516static uint64_t
517blkid2offset(dnode_phys_t *dnp, int level, uint64_t blkid)
518{
519	if (level < 0)
520		return (blkid);
521
522	return ((blkid << (level * (dnp->dn_indblkshift - SPA_BLKPTRSHIFT))) *
523	    dnp->dn_datablkszsec << SPA_MINBLOCKSHIFT);
524}
525
526/* ARGSUSED */
527static int
528zdb_indirect_cb(traverse_blk_cache_t *bc, spa_t *spa, void *a)
529{
530	zbookmark_t *zb = &bc->bc_bookmark;
531	blkptr_t *bp = &bc->bc_blkptr;
532	dva_t *dva = &bp->blk_dva[0];
533	void *data = bc->bc_data;
534	dnode_phys_t *dnp = bc->bc_dnode;
535	char buffer[300];
536	int l;
537
538	if (bc->bc_errno) {
539		(void) sprintf(buffer,
540		    "Error %d reading <%llu, %llu, %lld, %llu>: ",
541		    bc->bc_errno,
542		    (u_longlong_t)zb->zb_objset,
543		    (u_longlong_t)zb->zb_object,
544		    (u_longlong_t)zb->zb_level,
545		    (u_longlong_t)zb->zb_blkid);
546		goto out;
547	}
548
549	if (zb->zb_level == -1) {
550		ASSERT3U(BP_GET_TYPE(bp), ==, DMU_OT_OBJSET);
551		ASSERT3U(BP_GET_LEVEL(bp), ==, 0);
552	} else {
553		ASSERT3U(BP_GET_TYPE(bp), ==, dnp->dn_type);
554		ASSERT3U(BP_GET_LEVEL(bp), ==, zb->zb_level);
555	}
556
557	if (zb->zb_level > 0) {
558		uint64_t fill = 0;
559		blkptr_t *bpx, *bpend;
560
561		for (bpx = data, bpend = bpx + BP_GET_LSIZE(bp) / sizeof (*bpx);
562		    bpx < bpend; bpx++) {
563			if (bpx->blk_birth != 0) {
564				fill += bpx->blk_fill;
565			} else {
566				ASSERT(bpx->blk_fill == 0);
567			}
568		}
569		ASSERT3U(fill, ==, bp->blk_fill);
570	}
571
572	if (zb->zb_level == 0 && dnp->dn_type == DMU_OT_DNODE) {
573		uint64_t fill = 0;
574		dnode_phys_t *dnx, *dnend;
575
576		for (dnx = data, dnend = dnx + (BP_GET_LSIZE(bp)>>DNODE_SHIFT);
577		    dnx < dnend; dnx++) {
578			if (dnx->dn_type != DMU_OT_NONE)
579				fill++;
580		}
581		ASSERT3U(fill, ==, bp->blk_fill);
582	}
583
584	(void) sprintf(buffer, "%16llx ",
585	    (u_longlong_t)blkid2offset(dnp, zb->zb_level, zb->zb_blkid));
586
587	ASSERT(zb->zb_level >= 0);
588
589	for (l = dnp->dn_nlevels - 1; l >= -1; l--) {
590		if (l == zb->zb_level) {
591			(void) sprintf(buffer + strlen(buffer), "L%llx",
592			    (u_longlong_t)zb->zb_level);
593		} else {
594			(void) sprintf(buffer + strlen(buffer), " ");
595		}
596	}
597
598out:
599	if (bp->blk_birth == 0) {
600		(void) sprintf(buffer + strlen(buffer), "<hole>");
601		(void) printf("%s\n", buffer);
602	} else {
603		// XXBP - Need to print number of active BPs here
604		(void) sprintf(buffer + strlen(buffer),
605		    "vdev=%llu off=%llx %llxL/%llxP/%llxA F=%llu B=%llu",
606		    (u_longlong_t)DVA_GET_VDEV(dva),
607		    (u_longlong_t)DVA_GET_OFFSET(dva),
608		    (u_longlong_t)BP_GET_LSIZE(bp),
609		    (u_longlong_t)BP_GET_PSIZE(bp),
610		    (u_longlong_t)DVA_GET_ASIZE(dva),
611		    (u_longlong_t)bp->blk_fill,
612		    (u_longlong_t)bp->blk_birth);
613
614		(void) printf("%s\n", buffer);
615	}
616
617	return (bc->bc_errno ? ERESTART : 0);
618}
619
620/*ARGSUSED*/
621static void
622dump_indirect(objset_t *os, uint64_t object, void *data, size_t size)
623{
624	traverse_handle_t *th;
625	uint64_t objset = dmu_objset_id(os);
626	int advance = zdb_advance;
627
628	(void) printf("Indirect blocks:\n");
629
630	if (object == 0)
631		advance |= ADVANCE_DATA;
632
633	th = traverse_init(dmu_objset_spa(os), zdb_indirect_cb, NULL, advance,
634	    ZIO_FLAG_CANFAIL);
635	th->th_noread = zdb_noread;
636
637	traverse_add_dnode(th, 0, -1ULL, objset, object);
638
639	while (traverse_more(th) == EAGAIN)
640		continue;
641
642	(void) printf("\n");
643
644	traverse_fini(th);
645}
646
647/*ARGSUSED*/
648static void
649dump_dsl_dir(objset_t *os, uint64_t object, void *data, size_t size)
650{
651	dsl_dir_phys_t *dd = data;
652	time_t crtime;
653	char used[6], compressed[6], uncompressed[6], quota[6], resv[6];
654
655	if (dd == NULL)
656		return;
657
658	ASSERT(size == sizeof (*dd));
659
660	crtime = dd->dd_creation_time;
661	nicenum(dd->dd_used_bytes, used);
662	nicenum(dd->dd_compressed_bytes, compressed);
663	nicenum(dd->dd_uncompressed_bytes, uncompressed);
664	nicenum(dd->dd_quota, quota);
665	nicenum(dd->dd_reserved, resv);
666
667	(void) printf("\t\tcreation_time = %s", ctime(&crtime));
668	(void) printf("\t\thead_dataset_obj = %llu\n",
669	    (u_longlong_t)dd->dd_head_dataset_obj);
670	(void) printf("\t\tparent_dir_obj = %llu\n",
671	    (u_longlong_t)dd->dd_parent_obj);
672	(void) printf("\t\tclone_parent_obj = %llu\n",
673	    (u_longlong_t)dd->dd_clone_parent_obj);
674	(void) printf("\t\tchild_dir_zapobj = %llu\n",
675	    (u_longlong_t)dd->dd_child_dir_zapobj);
676	(void) printf("\t\tused_bytes = %s\n", used);
677	(void) printf("\t\tcompressed_bytes = %s\n", compressed);
678	(void) printf("\t\tuncompressed_bytes = %s\n", uncompressed);
679	(void) printf("\t\tquota = %s\n", quota);
680	(void) printf("\t\treserved = %s\n", resv);
681	(void) printf("\t\tprops_zapobj = %llu\n",
682	    (u_longlong_t)dd->dd_props_zapobj);
683}
684
685/*ARGSUSED*/
686static void
687dump_dsl_dataset(objset_t *os, uint64_t object, void *data, size_t size)
688{
689	dsl_dataset_phys_t *ds = data;
690	time_t crtime;
691	char used[6], compressed[6], uncompressed[6], unique[6];
692	char blkbuf[BP_SPRINTF_LEN];
693
694	if (ds == NULL)
695		return;
696
697	ASSERT(size == sizeof (*ds));
698	crtime = ds->ds_creation_time;
699	nicenum(ds->ds_used_bytes, used);
700	nicenum(ds->ds_compressed_bytes, compressed);
701	nicenum(ds->ds_uncompressed_bytes, uncompressed);
702	nicenum(ds->ds_unique_bytes, unique);
703	sprintf_blkptr(blkbuf, BP_SPRINTF_LEN, &ds->ds_bp);
704
705	(void) printf("\t\tdataset_obj = %llu\n",
706	    (u_longlong_t)ds->ds_dir_obj);
707	(void) printf("\t\tprev_snap_obj = %llu\n",
708	    (u_longlong_t)ds->ds_prev_snap_obj);
709	(void) printf("\t\tprev_snap_txg = %llu\n",
710	    (u_longlong_t)ds->ds_prev_snap_txg);
711	(void) printf("\t\tnext_snap_obj = %llu\n",
712	    (u_longlong_t)ds->ds_next_snap_obj);
713	(void) printf("\t\tsnapnames_zapobj = %llu\n",
714	    (u_longlong_t)ds->ds_snapnames_zapobj);
715	(void) printf("\t\tnum_children = %llu\n",
716	    (u_longlong_t)ds->ds_num_children);
717	(void) printf("\t\tcreation_time = %s", ctime(&crtime));
718	(void) printf("\t\tcreation_txg = %llu\n",
719	    (u_longlong_t)ds->ds_creation_txg);
720	(void) printf("\t\tdeadlist_obj = %llu\n",
721	    (u_longlong_t)ds->ds_deadlist_obj);
722	(void) printf("\t\tused_bytes = %s\n", used);
723	(void) printf("\t\tcompressed_bytes = %s\n", compressed);
724	(void) printf("\t\tuncompressed_bytes = %s\n", uncompressed);
725	(void) printf("\t\tunique = %s\n", unique);
726	(void) printf("\t\tfsid_guid = %llu\n",
727	    (u_longlong_t)ds->ds_fsid_guid);
728	(void) printf("\t\tguid = %llu\n",
729	    (u_longlong_t)ds->ds_guid);
730	(void) printf("\t\trestoring = %llu\n",
731	    (u_longlong_t)ds->ds_restoring);
732	(void) printf("\t\tbp = %s\n", blkbuf);
733}
734
735static void
736dump_bplist(objset_t *mos, uint64_t object, char *name)
737{
738	bplist_t bpl = { 0 };
739	blkptr_t blk, *bp = &blk;
740	uint64_t itor = 0;
741	char numbuf[6];
742
743	if (dump_opt['d'] < 3)
744		return;
745
746	VERIFY(0 == bplist_open(&bpl, mos, object));
747	if (bplist_empty(&bpl)) {
748		bplist_close(&bpl);
749		return;
750	}
751
752	nicenum(bpl.bpl_phys->bpl_bytes, numbuf);
753
754	(void) printf("\n    %s: %llu entries, %s\n",
755	    name, (u_longlong_t)bpl.bpl_phys->bpl_entries, numbuf);
756
757	if (dump_opt['d'] < 5) {
758		bplist_close(&bpl);
759		return;
760	}
761
762	(void) printf("\n");
763
764	while (bplist_iterate(&bpl, &itor, bp) == 0) {
765		ASSERT(bp->blk_birth != 0);
766		// XXBP - Do we want to see all DVAs, or just one?
767		(void) printf("\tItem %3llu: vdev=%llu off=%llx "
768		    "%llxL/%llxP/%llxA F=%llu B=%llu\n",
769		    (u_longlong_t)itor - 1,
770		    (u_longlong_t)DVA_GET_VDEV(&bp->blk_dva[0]),
771		    (u_longlong_t)DVA_GET_OFFSET(&bp->blk_dva[0]),
772		    (u_longlong_t)BP_GET_LSIZE(bp),
773		    (u_longlong_t)BP_GET_PSIZE(bp),
774		    (u_longlong_t)DVA_GET_ASIZE(&bp->blk_dva[0]),
775		    (u_longlong_t)bp->blk_fill,
776		    (u_longlong_t)bp->blk_birth);
777	}
778
779	bplist_close(&bpl);
780}
781
782static char *
783znode_path(objset_t *os, uint64_t object, char *pathbuf, size_t size)
784{
785	dmu_buf_t *db;
786	dmu_object_info_t doi;
787	znode_phys_t *zp;
788	uint64_t parent = 0;
789	size_t complen;
790	char component[MAXNAMELEN + 1];
791	char *path;
792	int error;
793
794	path = pathbuf + size;
795	*--path = '\0';
796
797	for (;;) {
798		error = dmu_bonus_hold(os, object, FTAG, &db);
799		if (error)
800			break;
801
802		dmu_object_info_from_db(db, &doi);
803		zp = db->db_data;
804		parent = zp->zp_parent;
805		dmu_buf_rele(db, FTAG);
806
807		if (doi.doi_bonus_type != DMU_OT_ZNODE)
808			break;
809
810		if (parent == object) {
811			if (path[0] != '/')
812				*--path = '/';
813			return (path);
814		}
815
816		if (zap_value_search(os, parent, object, component) != 0)
817			break;
818
819		complen = strlen(component);
820		path -= complen;
821		bcopy(component, path, complen);
822		*--path = '/';
823
824		object = parent;
825	}
826
827	(void) sprintf(component, "???<object#%llu>", (u_longlong_t)object);
828
829	complen = strlen(component);
830	path -= complen;
831	bcopy(component, path, complen);
832
833	return (path);
834}
835
836/*ARGSUSED*/
837static void
838dump_znode(objset_t *os, uint64_t object, void *data, size_t size)
839{
840	znode_phys_t *zp = data;
841	time_t z_crtime, z_atime, z_mtime, z_ctime;
842	char path[MAXPATHLEN * 2];	/* allow for xattr and failure prefix */
843
844	ASSERT(size >= sizeof (znode_phys_t));
845
846	if (dump_opt['d'] < 3) {
847		(void) printf("\t%s\n",
848		    znode_path(os, object, path, sizeof (path)));
849		return;
850	}
851
852	z_crtime = (time_t)zp->zp_crtime[0];
853	z_atime = (time_t)zp->zp_atime[0];
854	z_mtime = (time_t)zp->zp_mtime[0];
855	z_ctime = (time_t)zp->zp_ctime[0];
856
857	(void) printf("\tpath	%s\n",
858	    znode_path(os, object, path, sizeof (path)));
859	(void) printf("\tatime	%s", ctime(&z_atime));
860	(void) printf("\tmtime	%s", ctime(&z_mtime));
861	(void) printf("\tctime	%s", ctime(&z_ctime));
862	(void) printf("\tcrtime	%s", ctime(&z_crtime));
863	(void) printf("\tgen	%llu\n", (u_longlong_t)zp->zp_gen);
864	(void) printf("\tmode	%llo\n", (u_longlong_t)zp->zp_mode);
865	(void) printf("\tsize	%llu\n", (u_longlong_t)zp->zp_size);
866	(void) printf("\tparent	%llu\n", (u_longlong_t)zp->zp_parent);
867	(void) printf("\tlinks	%llu\n", (u_longlong_t)zp->zp_links);
868	(void) printf("\txattr	%llu\n", (u_longlong_t)zp->zp_xattr);
869	(void) printf("\trdev	0x%016llx\n", (u_longlong_t)zp->zp_rdev);
870}
871
872/*ARGSUSED*/
873static void
874dump_acl(objset_t *os, uint64_t object, void *data, size_t size)
875{
876}
877
878/*ARGSUSED*/
879static void
880dump_dmu_objset(objset_t *os, uint64_t object, void *data, size_t size)
881{
882}
883
884static object_viewer_t *object_viewer[DMU_OT_NUMTYPES] = {
885	dump_none,		/* unallocated			*/
886	dump_zap,		/* object directory		*/
887	dump_uint64,		/* object array			*/
888	dump_none,		/* packed nvlist		*/
889	dump_packed_nvlist,	/* packed nvlist size		*/
890	dump_none,		/* bplist			*/
891	dump_none,		/* bplist header		*/
892	dump_none,		/* SPA space map header		*/
893	dump_none,		/* SPA space map		*/
894	dump_none,		/* ZIL intent log		*/
895	dump_dnode,		/* DMU dnode			*/
896	dump_dmu_objset,	/* DMU objset			*/
897	dump_dsl_dir,		/* DSL directory		*/
898	dump_zap,		/* DSL directory child map	*/
899	dump_zap,		/* DSL dataset snap map		*/
900	dump_zap,		/* DSL props			*/
901	dump_dsl_dataset,	/* DSL dataset			*/
902	dump_znode,		/* ZFS znode			*/
903	dump_acl,		/* ZFS ACL			*/
904	dump_uint8,		/* ZFS plain file		*/
905	dump_zap,		/* ZFS directory		*/
906	dump_zap,		/* ZFS master node		*/
907	dump_zap,		/* ZFS delete queue		*/
908	dump_uint8,		/* zvol object			*/
909	dump_zap,		/* zvol prop			*/
910	dump_uint8,		/* other uint8[]		*/
911	dump_uint64,		/* other uint64[]		*/
912	dump_zap,		/* other ZAP			*/
913	dump_zap,		/* persistent error log		*/
914};
915
916static void
917dump_object(objset_t *os, uint64_t object, int verbosity, int *print_header)
918{
919	dmu_buf_t *db = NULL;
920	dmu_object_info_t doi;
921	dnode_t *dn;
922	void *bonus = NULL;
923	size_t bsize = 0;
924	char iblk[6], dblk[6], lsize[6], psize[6], bonus_size[6], segsize[6];
925	char aux[50];
926	int error;
927
928	if (*print_header) {
929		(void) printf("\n    Object  lvl   iblk   dblk  lsize"
930		    "  psize  type\n");
931		*print_header = 0;
932	}
933
934	if (object == 0) {
935		dn = os->os->os_meta_dnode;
936	} else {
937		error = dmu_bonus_hold(os, object, FTAG, &db);
938		if (error)
939			fatal("dmu_bonus_hold(%llu) failed, errno %u",
940			    object, error);
941		bonus = db->db_data;
942		bsize = db->db_size;
943		dn = ((dmu_buf_impl_t *)db)->db_dnode;
944	}
945	dmu_object_info_from_dnode(dn, &doi);
946
947	nicenum(doi.doi_metadata_block_size, iblk);
948	nicenum(doi.doi_data_block_size, dblk);
949	nicenum(doi.doi_data_block_size * (doi.doi_max_block_offset + 1),
950	    lsize);
951	nicenum(doi.doi_physical_blks << 9, psize);
952	nicenum(doi.doi_bonus_size, bonus_size);
953
954	aux[0] = '\0';
955
956	if (doi.doi_checksum != ZIO_CHECKSUM_INHERIT || verbosity >= 6)
957		(void) snprintf(aux + strlen(aux), sizeof (aux), " (K=%s)",
958		zio_checksum_table[doi.doi_checksum].ci_name);
959
960	if (doi.doi_compress != ZIO_COMPRESS_INHERIT || verbosity >= 6)
961		(void) snprintf(aux + strlen(aux), sizeof (aux), " (Z=%s)",
962		zio_compress_table[doi.doi_compress].ci_name);
963
964	(void) printf("%10lld  %3u  %5s  %5s  %5s  %5s  %s%s\n",
965	    (u_longlong_t)object, doi.doi_indirection, iblk, dblk, lsize,
966	    psize, dmu_ot[doi.doi_type].ot_name, aux);
967
968	if (doi.doi_bonus_type != DMU_OT_NONE && verbosity > 3) {
969		(void) printf("%10s  %3s  %5s  %5s  %5s  %5s  %s\n",
970		    "", "", "", "", bonus_size, "bonus",
971		    dmu_ot[doi.doi_bonus_type].ot_name);
972	}
973
974	if (verbosity >= 4) {
975		object_viewer[doi.doi_bonus_type](os, object, bonus, bsize);
976		object_viewer[doi.doi_type](os, object, NULL, 0);
977		*print_header = 1;
978	}
979
980	if (verbosity >= 5)
981		dump_indirect(os, object, NULL, 0);
982
983	if (verbosity >= 5) {
984		/*
985		 * Report the list of segments that comprise the object.
986		 */
987		uint64_t start = 0;
988		uint64_t end;
989		uint64_t blkfill = 1;
990		int minlvl = 1;
991
992		if (dn->dn_type == DMU_OT_DNODE) {
993			minlvl = 0;
994			blkfill = DNODES_PER_BLOCK;
995		}
996
997		for (;;) {
998			error = dnode_next_offset(dn, B_FALSE, &start, minlvl,
999			    blkfill);
1000			if (error)
1001				break;
1002			end = start;
1003			error = dnode_next_offset(dn, B_TRUE, &end, minlvl,
1004			    blkfill);
1005			nicenum(end - start, segsize);
1006			(void) printf("\t\tsegment [%016llx, %016llx)"
1007			    " size %5s\n", (u_longlong_t)start,
1008			    (u_longlong_t)end, segsize);
1009			if (error)
1010				break;
1011			start = end;
1012		}
1013	}
1014
1015	if (db != NULL)
1016		dmu_buf_rele(db, FTAG);
1017}
1018
1019static char *objset_types[DMU_OST_NUMTYPES] = {
1020	"NONE", "META", "ZPL", "ZVOL", "OTHER", "ANY" };
1021
1022/*ARGSUSED*/
1023static void
1024dump_dir(objset_t *os)
1025{
1026	dmu_objset_stats_t dds;
1027	uint64_t object, object_count;
1028	char numbuf[8];
1029	char blkbuf[BP_SPRINTF_LEN];
1030	char osname[MAXNAMELEN];
1031	char *type = "UNKNOWN";
1032	int verbosity = dump_opt['d'];
1033	int print_header = 1;
1034	int i, error;
1035
1036	dmu_objset_stats(os, &dds);
1037
1038	if (dds.dds_type < DMU_OST_NUMTYPES)
1039		type = objset_types[dds.dds_type];
1040
1041	if (dds.dds_type == DMU_OST_META) {
1042		dds.dds_creation_txg = TXG_INITIAL;
1043		dds.dds_last_txg = os->os->os_rootbp.blk_birth;
1044		dds.dds_objects_used = os->os->os_rootbp.blk_fill;
1045		dds.dds_space_refd =
1046		    os->os->os_spa->spa_dsl_pool->dp_mos_dir->dd_used_bytes;
1047	}
1048
1049	ASSERT3U(dds.dds_objects_used, ==, os->os->os_rootbp.blk_fill);
1050
1051	nicenum(dds.dds_space_refd, numbuf);
1052
1053	if (verbosity >= 4) {
1054		(void) strcpy(blkbuf, ", rootbp ");
1055		sprintf_blkptr(blkbuf + strlen(blkbuf),
1056		    BP_SPRINTF_LEN - strlen(blkbuf), &os->os->os_rootbp);
1057	} else {
1058		blkbuf[0] = '\0';
1059	}
1060
1061	dmu_objset_name(os, osname);
1062
1063	(void) printf("Dataset %s [%s], ID %llu, cr_txg %llu, last_txg %llu, "
1064	    "%s, %llu objects%s\n",
1065	    osname, type, (u_longlong_t)dmu_objset_id(os),
1066	    (u_longlong_t)dds.dds_creation_txg,
1067	    (u_longlong_t)dds.dds_last_txg,
1068	    numbuf,
1069	    (u_longlong_t)dds.dds_objects_used,
1070	    blkbuf);
1071
1072	dump_intent_log(dmu_objset_zil(os));
1073
1074	if (dmu_objset_ds(os) != NULL)
1075		dump_bplist(dmu_objset_pool(os)->dp_meta_objset,
1076		    dmu_objset_ds(os)->ds_phys->ds_deadlist_obj, "Deadlist");
1077
1078	if (verbosity < 2)
1079		return;
1080
1081	if (zopt_objects != 0) {
1082		for (i = 0; i < zopt_objects; i++)
1083			dump_object(os, zopt_object[i], verbosity,
1084			    &print_header);
1085		(void) printf("\n");
1086		return;
1087	}
1088
1089	dump_object(os, 0, verbosity, &print_header);
1090	object_count = 1;
1091
1092	object = 0;
1093	while ((error = dmu_object_next(os, &object, B_FALSE)) == 0) {
1094		dump_object(os, object, verbosity, &print_header);
1095		object_count++;
1096	}
1097
1098	ASSERT3U(object_count, ==, dds.dds_objects_used);
1099
1100	(void) printf("\n");
1101
1102	if (error != ESRCH)
1103		fatal("dmu_object_next() = %d", error);
1104}
1105
1106static void
1107dump_uberblock(uberblock_t *ub)
1108{
1109	time_t timestamp = ub->ub_timestamp;
1110
1111	(void) printf("Uberblock\n\n");
1112	(void) printf("\tmagic = %016llx\n", (u_longlong_t)ub->ub_magic);
1113	(void) printf("\tversion = %llu\n", (u_longlong_t)ub->ub_version);
1114	(void) printf("\ttxg = %llu\n", (u_longlong_t)ub->ub_txg);
1115	(void) printf("\tguid_sum = %llu\n", (u_longlong_t)ub->ub_guid_sum);
1116	(void) printf("\ttimestamp = %llu UTC = %s",
1117	    (u_longlong_t)ub->ub_timestamp, asctime(localtime(&timestamp)));
1118	if (dump_opt['u'] >= 3) {
1119		char blkbuf[BP_SPRINTF_LEN];
1120		sprintf_blkptr(blkbuf, BP_SPRINTF_LEN, &ub->ub_rootbp);
1121		(void) printf("\trootbp = %s\n", blkbuf);
1122	}
1123	(void) printf("\n");
1124}
1125
1126static void
1127dump_config(const char *pool)
1128{
1129	spa_t *spa = NULL;
1130
1131	mutex_enter(&spa_namespace_lock);
1132	while ((spa = spa_next(spa)) != NULL) {
1133		if (pool == NULL)
1134			(void) printf("%s\n", spa_name(spa));
1135		if (pool == NULL || strcmp(pool, spa_name(spa)) == 0)
1136			dump_nvlist(spa->spa_config, 4);
1137	}
1138	mutex_exit(&spa_namespace_lock);
1139}
1140
1141static void
1142dump_label(const char *dev)
1143{
1144	int fd;
1145	vdev_label_t label;
1146	char *buf = label.vl_vdev_phys.vp_nvlist;
1147	size_t buflen = sizeof (label.vl_vdev_phys.vp_nvlist);
1148	struct stat64 statbuf;
1149	uint64_t psize;
1150	int l;
1151
1152	if ((fd = open64(dev, O_RDONLY)) < 0) {
1153		(void) printf("cannot open '%s': %s\n", dev, strerror(errno));
1154		exit(1);
1155	}
1156
1157	if (fstat64(fd, &statbuf) != 0) {
1158		(void) printf("failed to stat '%s': %s\n", dev,
1159		    strerror(errno));
1160		exit(1);
1161	}
1162
1163	psize = statbuf.st_size;
1164	psize = P2ALIGN(psize, (uint64_t)sizeof (vdev_label_t));
1165
1166	for (l = 0; l < VDEV_LABELS; l++) {
1167
1168		nvlist_t *config = NULL;
1169
1170		(void) printf("--------------------------------------------\n");
1171		(void) printf("LABEL %d\n", l);
1172		(void) printf("--------------------------------------------\n");
1173
1174		if (pread64(fd, &label, sizeof (label),
1175		    vdev_label_offset(psize, l, 0)) != sizeof (label)) {
1176			(void) printf("failed to read label %d\n", l);
1177			continue;
1178		}
1179
1180		if (nvlist_unpack(buf, buflen, &config, 0) != 0) {
1181			(void) printf("failed to unpack label %d\n", l);
1182			continue;
1183		}
1184		dump_nvlist(config, 4);
1185		nvlist_free(config);
1186	}
1187}
1188
1189/*ARGSUSED*/
1190static void
1191dump_one_dir(char *dsname, void *arg)
1192{
1193	int error;
1194	objset_t *os;
1195
1196	error = dmu_objset_open(dsname, DMU_OST_ANY,
1197	    DS_MODE_STANDARD | DS_MODE_READONLY, &os);
1198	if (error) {
1199		(void) printf("Could not open %s\n", dsname);
1200		return;
1201	}
1202	dump_dir(os);
1203	dmu_objset_close(os);
1204}
1205
1206static void
1207zdb_space_map_load(spa_t *spa)
1208{
1209	vdev_t *rvd = spa->spa_root_vdev;
1210	vdev_t *vd;
1211	int c, m, error;
1212
1213	for (c = 0; c < rvd->vdev_children; c++) {
1214		vd = rvd->vdev_child[c];
1215		for (m = 0; m < vd->vdev_ms_count; m++) {
1216			metaslab_t *msp = vd->vdev_ms[m];
1217			space_map_t *sm = &msp->ms_allocmap[0];
1218			mutex_enter(&msp->ms_lock);
1219			error = space_map_load(sm, msp->ms_smo, SM_ALLOC,
1220			    spa->spa_meta_objset, msp->ms_usable_end,
1221			    sm->sm_size - msp->ms_usable_space);
1222			mutex_exit(&msp->ms_lock);
1223			if (error)
1224				fatal("%s bad space map #%d, error %d",
1225				    spa->spa_name, c, error);
1226		}
1227	}
1228}
1229
1230static int
1231zdb_space_map_claim(spa_t *spa, blkptr_t *bp, zbookmark_t *zb)
1232{
1233	dva_t *dva = &bp->blk_dva[0];
1234	uint64_t vdev = DVA_GET_VDEV(dva);
1235	uint64_t offset = DVA_GET_OFFSET(dva);
1236	uint64_t size = DVA_GET_ASIZE(dva);
1237	vdev_t *vd;
1238	metaslab_t *msp;
1239	space_map_t *allocmap, *freemap;
1240	int error;
1241
1242	if ((vd = vdev_lookup_top(spa, vdev)) == NULL)
1243		return (ENXIO);
1244
1245	if ((offset >> vd->vdev_ms_shift) >= vd->vdev_ms_count)
1246		return (ENXIO);
1247
1248	if (DVA_GET_GANG(dva)) {
1249		zio_gbh_phys_t gbh;
1250		blkptr_t blk = *bp;
1251		int g;
1252
1253		/* LINTED - compile time assert */
1254		ASSERT(sizeof (zio_gbh_phys_t) == SPA_GANGBLOCKSIZE);
1255		size = vdev_psize_to_asize(vd, SPA_GANGBLOCKSIZE);
1256		DVA_SET_GANG(&blk.blk_dva[0], 0);
1257		DVA_SET_ASIZE(&blk.blk_dva[0], size);
1258		BP_SET_CHECKSUM(&blk, ZIO_CHECKSUM_GANG_HEADER);
1259		BP_SET_PSIZE(&blk, SPA_GANGBLOCKSIZE);
1260		BP_SET_LSIZE(&blk, SPA_GANGBLOCKSIZE);
1261		BP_SET_COMPRESS(&blk, ZIO_COMPRESS_OFF);
1262		error = zio_wait(zio_read(NULL, spa, &blk,
1263		    &gbh, SPA_GANGBLOCKSIZE, NULL, NULL,
1264		    ZIO_PRIORITY_SYNC_READ,
1265		    ZIO_FLAG_CANFAIL | ZIO_FLAG_CONFIG_HELD, zb));
1266		if (error)
1267			return (error);
1268		if (BP_SHOULD_BYTESWAP(&blk))
1269			byteswap_uint64_array(&gbh, SPA_GANGBLOCKSIZE);
1270		for (g = 0; g < SPA_GBH_NBLKPTRS; g++) {
1271			if (gbh.zg_blkptr[g].blk_birth == 0)
1272				break;
1273			error = zdb_space_map_claim(spa, &gbh.zg_blkptr[g], zb);
1274			if (error)
1275				return (error);
1276		}
1277	}
1278
1279	msp = vd->vdev_ms[offset >> vd->vdev_ms_shift];
1280	allocmap = &msp->ms_allocmap[0];
1281	freemap = &msp->ms_freemap[0];
1282
1283	mutex_enter(&msp->ms_lock);
1284	if (space_map_contains(freemap, offset, size)) {
1285		mutex_exit(&msp->ms_lock);
1286		return (EAGAIN);	/* allocated more than once */
1287	}
1288
1289	if (!space_map_contains(allocmap, offset, size)) {
1290		mutex_exit(&msp->ms_lock);
1291		return (ESTALE);	/* not allocated at all */
1292	}
1293
1294	space_map_remove(allocmap, offset, size);
1295	space_map_add(freemap, offset, size);
1296
1297	mutex_exit(&msp->ms_lock);
1298
1299	return (0);
1300}
1301
1302static void
1303zdb_leak(space_map_t *sm, uint64_t start, uint64_t size)
1304{
1305	metaslab_t *msp;
1306
1307	/* LINTED */
1308	msp = (metaslab_t *)((char *)sm - offsetof(metaslab_t, ms_allocmap[0]));
1309
1310	(void) printf("leaked space: vdev %llu, offset 0x%llx, size %llu\n",
1311	    (u_longlong_t)msp->ms_group->mg_vd->vdev_id,
1312	    (u_longlong_t)start,
1313	    (u_longlong_t)size);
1314}
1315
1316static void
1317zdb_space_map_vacate(spa_t *spa)
1318{
1319	vdev_t *rvd = spa->spa_root_vdev;
1320	vdev_t *vd;
1321	int c, m;
1322
1323	for (c = 0; c < rvd->vdev_children; c++) {
1324		vd = rvd->vdev_child[c];
1325		for (m = 0; m < vd->vdev_ms_count; m++) {
1326			metaslab_t *msp = vd->vdev_ms[m];
1327			mutex_enter(&msp->ms_lock);
1328			space_map_vacate(&msp->ms_allocmap[0], zdb_leak,
1329			    &msp->ms_allocmap[0]);
1330			space_map_vacate(&msp->ms_freemap[0], NULL, NULL);
1331			mutex_exit(&msp->ms_lock);
1332		}
1333	}
1334}
1335
1336static void
1337zdb_refresh_ubsync(spa_t *spa)
1338{
1339	uberblock_t ub = { 0 };
1340	vdev_t *rvd = spa->spa_root_vdev;
1341	zio_t *zio;
1342
1343	/*
1344	 * Reload the uberblock.
1345	 */
1346	zio = zio_root(spa, NULL, NULL,
1347	    ZIO_FLAG_CANFAIL | ZIO_FLAG_SPECULATIVE);
1348	vdev_uberblock_load(zio, rvd, &ub);
1349	(void) zio_wait(zio);
1350
1351	if (ub.ub_txg != 0)
1352		spa->spa_ubsync = ub;
1353}
1354
1355/*
1356 * Verify that the sum of the sizes of all blocks in the pool adds up
1357 * to the SPA's sa_alloc total.
1358 */
1359typedef struct zdb_blkstats {
1360	uint64_t	zb_asize;
1361	uint64_t	zb_lsize;
1362	uint64_t	zb_psize;
1363	uint64_t	zb_count;
1364} zdb_blkstats_t;
1365
1366#define	DMU_OT_DEFERRED	DMU_OT_NONE
1367#define	DMU_OT_TOTAL	DMU_OT_NUMTYPES
1368
1369#define	ZB_TOTAL	ZB_MAXLEVEL
1370
1371typedef struct zdb_cb {
1372	zdb_blkstats_t	zcb_type[ZB_TOTAL + 1][DMU_OT_TOTAL + 1];
1373	uint64_t	zcb_errors[256];
1374	traverse_blk_cache_t *zcb_cache;
1375	int		zcb_readfails;
1376	int		zcb_haderrors;
1377} zdb_cb_t;
1378
1379static void
1380zdb_count_block(spa_t *spa, zdb_cb_t *zcb, blkptr_t *bp, int type)
1381{
1382	int i, error;
1383
1384	for (i = 0; i < 4; i++) {
1385		int l = (i < 2) ? BP_GET_LEVEL(bp) : ZB_TOTAL;
1386		int t = (i & 1) ? type : DMU_OT_TOTAL;
1387		zdb_blkstats_t *zb = &zcb->zcb_type[l][t];
1388
1389		zb->zb_asize += BP_GET_ASIZE(bp);
1390		zb->zb_lsize += BP_GET_LSIZE(bp);
1391		zb->zb_psize += BP_GET_PSIZE(bp);
1392		zb->zb_count++;
1393	}
1394
1395	if (dump_opt['L'])
1396		return;
1397
1398	error = zdb_space_map_claim(spa, bp, &zcb->zcb_cache->bc_bookmark);
1399
1400	if (error == 0)
1401		return;
1402
1403	if (error == EAGAIN)
1404		(void) fatal("double-allocation, bp=%p", bp);
1405
1406	if (error == ESTALE)
1407		(void) fatal("reference to freed block, bp=%p", bp);
1408
1409	(void) fatal("fatal error %d in bp %p", error, bp);
1410}
1411
1412static int
1413zdb_blkptr_cb(traverse_blk_cache_t *bc, spa_t *spa, void *arg)
1414{
1415	zbookmark_t *zb = &bc->bc_bookmark;
1416	zdb_cb_t *zcb = arg;
1417	blkptr_t *bp = &bc->bc_blkptr;
1418	dmu_object_type_t type = BP_GET_TYPE(bp);
1419	char blkbuf[BP_SPRINTF_LEN];
1420	int error = 0;
1421
1422	if (bc->bc_errno) {
1423		if (zcb->zcb_readfails++ < 10 && dump_opt['L']) {
1424			zdb_refresh_ubsync(spa);
1425			error = EAGAIN;
1426		} else {
1427			zcb->zcb_haderrors = 1;
1428			zcb->zcb_errors[bc->bc_errno]++;
1429			error = ERESTART;
1430		}
1431
1432		if (dump_opt['b'] >= 3 || (dump_opt['b'] >= 2 && bc->bc_errno))
1433			sprintf_blkptr(blkbuf, BP_SPRINTF_LEN, bp);
1434		else
1435			blkbuf[0] = '\0';
1436
1437		(void) printf("zdb_blkptr_cb: Got error %d reading "
1438		    "<%llu, %llu, %lld, %llx> %s -- %s\n",
1439		    bc->bc_errno,
1440		    (u_longlong_t)zb->zb_objset,
1441		    (u_longlong_t)zb->zb_object,
1442		    (u_longlong_t)zb->zb_level,
1443		    (u_longlong_t)zb->zb_blkid,
1444		    blkbuf,
1445		    error == EAGAIN ? "retrying" : "skipping");
1446
1447		return (error);
1448	}
1449
1450	zcb->zcb_readfails = 0;
1451
1452	ASSERT(bp->blk_birth != 0);
1453
1454	zdb_count_block(spa, zcb, bp, type);
1455
1456	if (dump_opt['b'] >= 4) {
1457		sprintf_blkptr(blkbuf, BP_SPRINTF_LEN, bp);
1458		(void) printf("objset %llu object %llu offset 0x%llx %s\n",
1459		    (u_longlong_t)zb->zb_objset,
1460		    (u_longlong_t)zb->zb_object,
1461		    (u_longlong_t)blkid2offset(bc->bc_dnode,
1462			zb->zb_level, zb->zb_blkid),
1463		    blkbuf);
1464	}
1465
1466	return (0);
1467}
1468
1469static int
1470dump_block_stats(spa_t *spa)
1471{
1472	traverse_handle_t *th;
1473	zdb_cb_t zcb = { 0 };
1474	traverse_blk_cache_t dummy_cache = { 0 };
1475	zdb_blkstats_t *zb, *tzb;
1476	uint64_t alloc, space;
1477	int leaks = 0;
1478	int advance = zdb_advance;
1479	int flags;
1480	int e;
1481
1482	zcb.zcb_cache = &dummy_cache;
1483
1484	if (dump_opt['c'])
1485		advance |= ADVANCE_DATA;
1486
1487	advance |= ADVANCE_PRUNE | ADVANCE_ZIL;
1488
1489	(void) printf("\nTraversing all blocks to %sverify"
1490	    " nothing leaked ...\n",
1491	    dump_opt['c'] ? "verify checksums and " : "");
1492
1493	/*
1494	 * Load all space maps.  As we traverse the pool, if we find a block
1495	 * that's not in its space map, that indicates a double-allocation,
1496	 * reference to a freed block, or an unclaimed block.  Otherwise we
1497	 * remove the block from the space map.  If the space maps are not
1498	 * empty when we're done, that indicates leaked blocks.
1499	 */
1500	if (!dump_opt['L'])
1501		zdb_space_map_load(spa);
1502
1503	/*
1504	 * If there's a deferred-free bplist, process that first.
1505	 */
1506	if (spa->spa_sync_bplist_obj != 0) {
1507		bplist_t *bpl = &spa->spa_sync_bplist;
1508		blkptr_t blk;
1509		uint64_t itor = 0;
1510
1511		VERIFY(0 == bplist_open(bpl, spa->spa_meta_objset,
1512		    spa->spa_sync_bplist_obj));
1513
1514		while (bplist_iterate(bpl, &itor, &blk) == 0) {
1515			zdb_count_block(spa, &zcb, &blk, DMU_OT_DEFERRED);
1516			if (dump_opt['b'] >= 4) {
1517				char blkbuf[BP_SPRINTF_LEN];
1518				sprintf_blkptr(blkbuf, BP_SPRINTF_LEN, &blk);
1519				(void) printf("[%s] %s\n",
1520				    "deferred free", blkbuf);
1521			}
1522		}
1523
1524		bplist_close(bpl);
1525	}
1526
1527	/*
1528	 * Now traverse the pool.  If we're reading all data to verify
1529	 * checksums, do a scrubbing read so that we validate all copies.
1530	 */
1531	flags = ZIO_FLAG_CANFAIL;
1532	if (advance & ADVANCE_DATA)
1533		flags |= ZIO_FLAG_SCRUB;
1534	th = traverse_init(spa, zdb_blkptr_cb, &zcb, advance, flags);
1535	th->th_noread = zdb_noread;
1536
1537	traverse_add_pool(th, 0, spa_first_txg(spa));
1538
1539	while (traverse_more(th) == EAGAIN)
1540		continue;
1541
1542	traverse_fini(th);
1543
1544	if (zcb.zcb_haderrors) {
1545		(void) printf("\nError counts:\n\n");
1546		(void) printf("\t%5s  %s\n", "errno", "count");
1547		for (e = 0; e < 256; e++) {
1548			if (zcb.zcb_errors[e] != 0) {
1549				(void) printf("\t%5d  %llu\n",
1550				    e, (u_longlong_t)zcb.zcb_errors[e]);
1551			}
1552		}
1553	}
1554
1555	/*
1556	 * Report any leaked segments.
1557	 */
1558	if (!dump_opt['L'])
1559		zdb_space_map_vacate(spa);
1560
1561	if (dump_opt['L'])
1562		(void) printf("\n\n *** Live pool traversal; "
1563		    "block counts are only approximate ***\n\n");
1564
1565	alloc = spa_get_alloc(spa);
1566	space = spa_get_space(spa);
1567
1568	tzb = &zcb.zcb_type[ZB_TOTAL][DMU_OT_TOTAL];
1569
1570	if (tzb->zb_asize == alloc) {
1571		(void) printf("\n\tNo leaks (block sum matches space"
1572		    " maps exactly)\n");
1573	} else {
1574		(void) printf("block traversal size %llu != alloc %llu "
1575		    "(leaked %lld)\n",
1576		    (u_longlong_t)tzb->zb_asize,
1577		    (u_longlong_t)alloc,
1578		    (u_longlong_t)(alloc - tzb->zb_asize));
1579		leaks = 1;
1580	}
1581
1582	if (tzb->zb_count == 0)
1583		return (2);
1584
1585	(void) printf("\n");
1586	(void) printf("\tbp count:      %10llu\n",
1587	    (u_longlong_t)tzb->zb_count);
1588	(void) printf("\tbp logical:    %10llu\t avg: %6llu\n",
1589	    (u_longlong_t)tzb->zb_lsize,
1590	    (u_longlong_t)(tzb->zb_lsize / tzb->zb_count));
1591	(void) printf("\tbp physical:   %10llu\t avg:"
1592	    " %6llu\tcompression: %6.2f\n",
1593	    (u_longlong_t)tzb->zb_psize,
1594	    (u_longlong_t)(tzb->zb_psize / tzb->zb_count),
1595	    (double)tzb->zb_lsize / tzb->zb_psize);
1596	(void) printf("\tbp allocated:  %10llu\t avg:"
1597	    " %6llu\tcompression: %6.2f\n",
1598	    (u_longlong_t)tzb->zb_asize,
1599	    (u_longlong_t)(tzb->zb_asize / tzb->zb_count),
1600	    (double)tzb->zb_lsize / tzb->zb_asize);
1601	(void) printf("\tSPA allocated: %10llu\tused: %5.2f%%\n",
1602	    (u_longlong_t)alloc, 100.0 * alloc / space);
1603
1604	if (dump_opt['b'] >= 2) {
1605		int l, t, level;
1606		(void) printf("\nBlocks\tLSIZE\tPSIZE\tASIZE"
1607		    "\t  avg\t comp\t%%Total\tType\n");
1608
1609		for (t = 0; t <= DMU_OT_NUMTYPES; t++) {
1610			char csize[6], lsize[6], psize[6], asize[6], avg[6];
1611			char *typename;
1612
1613			typename = t == DMU_OT_DEFERRED ? "deferred free" :
1614			    t == DMU_OT_TOTAL ? "Total" : dmu_ot[t].ot_name;
1615
1616			if (zcb.zcb_type[ZB_TOTAL][t].zb_asize == 0) {
1617				(void) printf("%6s\t%5s\t%5s\t%5s"
1618				    "\t%5s\t%5s\t%6s\t%s\n",
1619				    "-",
1620				    "-",
1621				    "-",
1622				    "-",
1623				    "-",
1624				    "-",
1625				    "-",
1626				    typename);
1627				continue;
1628			}
1629
1630			for (l = ZB_TOTAL - 1; l >= -1; l--) {
1631				level = (l == -1 ? ZB_TOTAL : l);
1632				zb = &zcb.zcb_type[level][t];
1633
1634				if (zb->zb_asize == 0)
1635					continue;
1636
1637				if (dump_opt['b'] < 3 && level != ZB_TOTAL)
1638					continue;
1639
1640				if (level == 0 && zb->zb_asize ==
1641				    zcb.zcb_type[ZB_TOTAL][t].zb_asize)
1642					continue;
1643
1644				nicenum(zb->zb_count, csize);
1645				nicenum(zb->zb_lsize, lsize);
1646				nicenum(zb->zb_psize, psize);
1647				nicenum(zb->zb_asize, asize);
1648				nicenum(zb->zb_asize / zb->zb_count, avg);
1649
1650				(void) printf("%6s\t%5s\t%5s\t%5s\t%5s"
1651				    "\t%5.2f\t%6.2f\t",
1652				    csize, lsize, psize, asize, avg,
1653				    (double)zb->zb_lsize / zb->zb_psize,
1654				    100.0 * zb->zb_asize / tzb->zb_asize);
1655
1656				if (level == ZB_TOTAL)
1657					(void) printf("%s\n", typename);
1658				else
1659					(void) printf("    L%d %s\n",
1660					    level, typename);
1661			}
1662		}
1663	}
1664
1665	(void) printf("\n");
1666
1667	if (leaks)
1668		return (2);
1669
1670	if (zcb.zcb_haderrors)
1671		return (3);
1672
1673	return (0);
1674}
1675
1676static void
1677dump_zpool(spa_t *spa)
1678{
1679	dsl_pool_t *dp = spa_get_dsl(spa);
1680	int rc = 0;
1681
1682	if (dump_opt['u'])
1683		dump_uberblock(&spa->spa_uberblock);
1684
1685	if (dump_opt['d'] || dump_opt['i']) {
1686		dump_dir(dp->dp_meta_objset);
1687		if (dump_opt['d'] >= 3) {
1688			dump_bplist(dp->dp_meta_objset,
1689			    spa->spa_sync_bplist_obj, "Deferred frees");
1690			dump_dtl(spa->spa_root_vdev, 0);
1691			dump_metaslabs(spa);
1692		}
1693		dmu_objset_find(spa->spa_name, dump_one_dir, NULL,
1694		    DS_FIND_SNAPSHOTS);
1695	}
1696
1697	if (dump_opt['b'] || dump_opt['c'])
1698		rc = dump_block_stats(spa);
1699
1700	if (dump_opt['s'])
1701		show_pool_stats(spa);
1702
1703	if (rc != 0)
1704		exit(rc);
1705}
1706
1707int
1708main(int argc, char **argv)
1709{
1710	int i, c;
1711	struct rlimit rl = { 1024, 1024 };
1712	spa_t *spa;
1713	objset_t *os = NULL;
1714	char *endstr;
1715	int dump_all = 1;
1716	int verbose = 0;
1717	int error;
1718	int flag, set;
1719	vdev_knob_t *vk;
1720
1721	(void) setrlimit(RLIMIT_NOFILE, &rl);
1722
1723	dprintf_setup(&argc, argv);
1724
1725	while ((c = getopt(argc, argv, "udibcsvCLO:B:Ul")) != -1) {
1726		switch (c) {
1727		case 'u':
1728		case 'd':
1729		case 'i':
1730		case 'b':
1731		case 'c':
1732		case 's':
1733		case 'C':
1734		case 'l':
1735			dump_opt[c]++;
1736			dump_all = 0;
1737			break;
1738		case 'L':
1739			dump_opt[c]++;
1740			break;
1741		case 'O':
1742			endstr = optarg;
1743			if (endstr[0] == '!') {
1744				endstr++;
1745				set = 0;
1746			} else {
1747				set = 1;
1748			}
1749			if (strcmp(endstr, "post") == 0) {
1750				flag = ADVANCE_PRE;
1751				set = !set;
1752			} else if (strcmp(endstr, "pre") == 0) {
1753				flag = ADVANCE_PRE;
1754			} else if (strcmp(endstr, "prune") == 0) {
1755				flag = ADVANCE_PRUNE;
1756			} else if (strcmp(endstr, "data") == 0) {
1757				flag = ADVANCE_DATA;
1758			} else if (strcmp(endstr, "holes") == 0) {
1759				flag = ADVANCE_HOLES;
1760			} else {
1761				usage();
1762			}
1763			if (set)
1764				zdb_advance |= flag;
1765			else
1766				zdb_advance &= ~flag;
1767			break;
1768		case 'B':
1769			endstr = optarg - 1;
1770			zdb_noread.zb_objset = strtoull(endstr + 1, &endstr, 0);
1771			zdb_noread.zb_object = strtoull(endstr + 1, &endstr, 0);
1772			zdb_noread.zb_level = strtol(endstr + 1, &endstr, 0);
1773			zdb_noread.zb_blkid = strtoull(endstr + 1, &endstr, 16);
1774			(void) printf("simulating bad block "
1775			    "<%llu, %llu, %lld, %llx>\n",
1776			    (u_longlong_t)zdb_noread.zb_objset,
1777			    (u_longlong_t)zdb_noread.zb_object,
1778			    (u_longlong_t)zdb_noread.zb_level,
1779			    (u_longlong_t)zdb_noread.zb_blkid);
1780			break;
1781		case 'v':
1782			verbose++;
1783			break;
1784		case 'U':
1785			spa_config_dir = "/tmp";
1786			break;
1787		default:
1788			usage();
1789			break;
1790		}
1791	}
1792
1793	kernel_init(FREAD);
1794
1795	/*
1796	 * Disable vdev caching.  If we don't do this, live pool traversal
1797	 * won't make progress because it will never see disk updates.
1798	 */
1799	for (vk = vdev_knob_next(NULL); vk != NULL; vk = vdev_knob_next(vk)) {
1800		if (strcmp(vk->vk_name, "cache_size") == 0)
1801			vk->vk_default = 0;
1802	}
1803
1804	for (c = 0; c < 256; c++) {
1805		if (dump_all && c != 'L' && c != 'l')
1806			dump_opt[c] = 1;
1807		if (dump_opt[c])
1808			dump_opt[c] += verbose;
1809	}
1810
1811	argc -= optind;
1812	argv += optind;
1813
1814	if (argc < 1) {
1815		if (dump_opt['C']) {
1816			dump_config(NULL);
1817			return (0);
1818		}
1819		usage();
1820	}
1821
1822	if (dump_opt['l']) {
1823		dump_label(argv[0]);
1824		return (0);
1825	}
1826
1827	if (dump_opt['C'])
1828		dump_config(argv[0]);
1829
1830	if (strchr(argv[0], '/') != NULL) {
1831		error = dmu_objset_open(argv[0], DMU_OST_ANY,
1832		    DS_MODE_STANDARD | DS_MODE_READONLY, &os);
1833	} else {
1834		error = spa_open(argv[0], &spa, FTAG);
1835	}
1836
1837	if (error)
1838		fatal("can't open %s: error %d", argv[0], error);
1839
1840	argv++;
1841	if (--argc > 0) {
1842		zopt_objects = argc;
1843		zopt_object = calloc(zopt_objects, sizeof (uint64_t));
1844		for (i = 0; i < zopt_objects; i++) {
1845			errno = 0;
1846			zopt_object[i] = strtoull(argv[i], NULL, 0);
1847			if (zopt_object[i] == 0 && errno != 0)
1848				fatal("bad object number %s: %s",
1849				    argv[i], strerror(errno));
1850		}
1851	}
1852
1853	if (os != NULL) {
1854		dump_dir(os);
1855		dmu_objset_close(os);
1856	} else {
1857		dump_zpool(spa);
1858		spa_close(spa, FTAG);
1859	}
1860
1861	kernel_fini();
1862
1863	return (0);
1864}
1865