?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
Patrick Welche <prlw1@cam.ac.uk>
[netbsd-mini2440.git] / external / cddl / osnet / dist / cmd / zdb / zdb.c
blob48a84d693e4495b51703409956e0a8aa9a44f3e1
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 2008 Sun Microsystems, Inc. All rights reserved.
23 * Use is subject to license terms.
24 */
25
26 #include <stdio.h>
27 #include <stdio_ext.h>
28 #include <stdlib.h>
29 #include <ctype.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 #include <sys/zfs_fuid.h>
53 #include <sys/arc.h>
54 #undef ZFS_MAXNAMELEN
55 #undef verify
56 #include <libzfs.h>
57
58 const char cmdname[] = "zdb";
59 uint8_t dump_opt[256];
60
61 typedef void object_viewer_t(objset_t *, uint64_t, void *data, size_t size);
62
63 extern void dump_intent_log(zilog_t *);
64 uint64_t *zopt_object = NULL;
65 int zopt_objects = 0;
66 libzfs_handle_t *g_zfs;
67 boolean_t zdb_sig_user_data = B_TRUE;
68 int zdb_sig_cksumalg = ZIO_CHECKSUM_SHA256;
69
70 /*
71 * These libumem hooks provide a reasonable set of defaults for the allocator's
72 * debugging facilities.
73 */
74 const char *
75 _umem_debug_init()
76 {
77 return ("default,verbose"); /* $UMEM_DEBUG setting */
78 }
79
80 const char *
81 _umem_logging_init(void)
82 {
83 return ("fail,contents"); /* $UMEM_LOGGING setting */
84 }
85
86 static void
87 usage(void)
88 {
89 (void) fprintf(stderr,
90 "Usage: %s [-udibcsvL] [-U cachefile_path] "
91 "[-S user:cksumalg] "
92 "dataset [object...]\n"
93 " %s -C [pool]\n"
94 " %s -l dev\n"
95 " %s -R pool:vdev:offset:size:flags\n"
96 " %s [-p path_to_vdev_dir]\n"
97 " %s -e pool | GUID | devid ...\n",
98 cmdname, cmdname, cmdname, cmdname, cmdname, cmdname);
99
100 (void) fprintf(stderr, " -u uberblock\n");
101 (void) fprintf(stderr, " -d datasets\n");
102 (void) fprintf(stderr, " -C cached pool configuration\n");
103 (void) fprintf(stderr, " -i intent logs\n");
104 (void) fprintf(stderr, " -b block statistics\n");
105 (void) fprintf(stderr, " -c checksum all data blocks\n");
106 (void) fprintf(stderr, " -s report stats on zdb's I/O\n");
107 (void) fprintf(stderr, " -S <user|all>:<cksum_alg|all> -- "
108 "dump blkptr signatures\n");
109 (void) fprintf(stderr, " -v verbose (applies to all others)\n");
110 (void) fprintf(stderr, " -l dump label contents\n");
111 (void) fprintf(stderr, " -L disable leak tracking (do not "
112 "load spacemaps)\n");
113 (void) fprintf(stderr, " -U cachefile_path -- use alternate "
114 "cachefile\n");
115 (void) fprintf(stderr, " -R read and display block from a "
116 "device\n");
117 (void) fprintf(stderr, " -e Pool is exported/destroyed/"
118 "has altroot\n");
119 (void) fprintf(stderr, " -p <Path to vdev dir> (use with -e)\n");
120 (void) fprintf(stderr, "Specify an option more than once (e.g. -bb) "
121 "to make only that option verbose\n");
122 (void) fprintf(stderr, "Default is to dump everything non-verbosely\n");
123 exit(1);
124 }
125
126 static void
127 fatal(const char *fmt, ...)
128 {
129 va_list ap;
130
131 va_start(ap, fmt);
132 (void) fprintf(stderr, "%s: ", cmdname);
133 (void) vfprintf(stderr, fmt, ap);
134 va_end(ap);
135 (void) fprintf(stderr, "\n");
136
137 abort();
138 }
139
140 static void
141 dump_nvlist(nvlist_t *list, int indent)
142 {
143 nvpair_t *elem = NULL;
144
145 while ((elem = nvlist_next_nvpair(list, elem)) != NULL) {
146 switch (nvpair_type(elem)) {
147 case DATA_TYPE_STRING:
148 {
149 char *value;
150
151 VERIFY(nvpair_value_string(elem, &value) == 0);
152 (void) printf("%*s%s='%s'\n", indent, "",
153 nvpair_name(elem), value);
154 }
155 break;
156
157 case DATA_TYPE_UINT64:
158 {
159 uint64_t value;
160
161 VERIFY(nvpair_value_uint64(elem, &value) == 0);
162 (void) printf("%*s%s=%llu\n", indent, "",
163 nvpair_name(elem), (u_longlong_t)value);
164 }
165 break;
166
167 case DATA_TYPE_NVLIST:
168 {
169 nvlist_t *value;
170
171 VERIFY(nvpair_value_nvlist(elem, &value) == 0);
172 (void) printf("%*s%s\n", indent, "",
173 nvpair_name(elem));
174 dump_nvlist(value, indent + 4);
175 }
176 break;
177
178 case DATA_TYPE_NVLIST_ARRAY:
179 {
180 nvlist_t **value;
181 uint_t c, count;
182
183 VERIFY(nvpair_value_nvlist_array(elem, &value,
184 &count) == 0);
185
186 for (c = 0; c < count; c++) {
187 (void) printf("%*s%s[%u]\n", indent, "",
188 nvpair_name(elem), c);
189 dump_nvlist(value[c], indent + 8);
190 }
191 }
192 break;
193
194 default:
195
196 (void) printf("bad config type %d for %s\n",
197 nvpair_type(elem), nvpair_name(elem));
198 }
199 }
200 }
201
202 /* ARGSUSED */
203 static void
204 dump_packed_nvlist(objset_t *os, uint64_t object, void *data, size_t size)
205 {
206 nvlist_t *nv;
207 size_t nvsize = *(uint64_t *)data;
208 char *packed = umem_alloc(nvsize, UMEM_NOFAIL);
209
210 VERIFY(0 == dmu_read(os, object, 0, nvsize, packed));
211
212 VERIFY(nvlist_unpack(packed, nvsize, &nv, 0) == 0);
213
214 umem_free(packed, nvsize);
215
216 dump_nvlist(nv, 8);
217
218 nvlist_free(nv);
219 }
220
221 const char dump_zap_stars[] = "****************************************";
222 const int dump_zap_width = sizeof (dump_zap_stars) - 1;
223
224 static void
225 dump_zap_histogram(uint64_t histo[ZAP_HISTOGRAM_SIZE])
226 {
227 int i;
228 int minidx = ZAP_HISTOGRAM_SIZE - 1;
229 int maxidx = 0;
230 uint64_t max = 0;
231
232 for (i = 0; i < ZAP_HISTOGRAM_SIZE; i++) {
233 if (histo[i] > max)
234 max = histo[i];
235 if (histo[i] > 0 && i > maxidx)
236 maxidx = i;
237 if (histo[i] > 0 && i < minidx)
238 minidx = i;
239 }
240
241 if (max < dump_zap_width)
242 max = dump_zap_width;
243
244 for (i = minidx; i <= maxidx; i++)
245 (void) printf("\t\t\t%u: %6llu %s\n", i, (u_longlong_t)histo[i],
246 &dump_zap_stars[(max - histo[i]) * dump_zap_width / max]);
247 }
248
249 static void
250 dump_zap_stats(objset_t *os, uint64_t object)
251 {
252 int error;
253 zap_stats_t zs;
254
255 error = zap_get_stats(os, object, &zs);
256 if (error)
257 return;
258
259 if (zs.zs_ptrtbl_len == 0) {
260 ASSERT(zs.zs_num_blocks == 1);
261 (void) printf("\tmicrozap: %llu bytes, %llu entries\n",
262 (u_longlong_t)zs.zs_blocksize,
263 (u_longlong_t)zs.zs_num_entries);
264 return;
265 }
266
267 (void) printf("\tFat ZAP stats:\n");
268
269 (void) printf("\t\tPointer table:\n");
270 (void) printf("\t\t\t%llu elements\n",
271 (u_longlong_t)zs.zs_ptrtbl_len);
272 (void) printf("\t\t\tzt_blk: %llu\n",
273 (u_longlong_t)zs.zs_ptrtbl_zt_blk);
274 (void) printf("\t\t\tzt_numblks: %llu\n",
275 (u_longlong_t)zs.zs_ptrtbl_zt_numblks);
276 (void) printf("\t\t\tzt_shift: %llu\n",
277 (u_longlong_t)zs.zs_ptrtbl_zt_shift);
278 (void) printf("\t\t\tzt_blks_copied: %llu\n",
279 (u_longlong_t)zs.zs_ptrtbl_blks_copied);
280 (void) printf("\t\t\tzt_nextblk: %llu\n",
281 (u_longlong_t)zs.zs_ptrtbl_nextblk);
282
283 (void) printf("\t\tZAP entries: %llu\n",
284 (u_longlong_t)zs.zs_num_entries);
285 (void) printf("\t\tLeaf blocks: %llu\n",
286 (u_longlong_t)zs.zs_num_leafs);
287 (void) printf("\t\tTotal blocks: %llu\n",
288 (u_longlong_t)zs.zs_num_blocks);
289 (void) printf("\t\tzap_block_type: 0x%llx\n",
290 (u_longlong_t)zs.zs_block_type);
291 (void) printf("\t\tzap_magic: 0x%llx\n",
292 (u_longlong_t)zs.zs_magic);
293 (void) printf("\t\tzap_salt: 0x%llx\n",
294 (u_longlong_t)zs.zs_salt);
295
296 (void) printf("\t\tLeafs with 2^n pointers:\n");
297 dump_zap_histogram(zs.zs_leafs_with_2n_pointers);
298
299 (void) printf("\t\tBlocks with n*5 entries:\n");
300 dump_zap_histogram(zs.zs_blocks_with_n5_entries);
301
302 (void) printf("\t\tBlocks n/10 full:\n");
303 dump_zap_histogram(zs.zs_blocks_n_tenths_full);
304
305 (void) printf("\t\tEntries with n chunks:\n");
306 dump_zap_histogram(zs.zs_entries_using_n_chunks);
307
308 (void) printf("\t\tBuckets with n entries:\n");
309 dump_zap_histogram(zs.zs_buckets_with_n_entries);
310 }
311
312 /*ARGSUSED*/
313 static void
314 dump_none(objset_t *os, uint64_t object, void *data, size_t size)
315 {
316 }
317
318 /*ARGSUSED*/
319 void
320 dump_uint8(objset_t *os, uint64_t object, void *data, size_t size)
321 {
322 }
323
324 /*ARGSUSED*/
325 static void
326 dump_uint64(objset_t *os, uint64_t object, void *data, size_t size)
327 {
328 }
329
330 /*ARGSUSED*/
331 static void
332 dump_zap(objset_t *os, uint64_t object, void *data, size_t size)
333 {
334 zap_cursor_t zc;
335 zap_attribute_t attr;
336 void *prop;
337 int i;
338
339 dump_zap_stats(os, object);
340 (void) printf("\n");
341
342 for (zap_cursor_init(&zc, os, object);
343 zap_cursor_retrieve(&zc, &attr) == 0;
344 zap_cursor_advance(&zc)) {
345 (void) printf("\t\t%s = ", attr.za_name);
346 if (attr.za_num_integers == 0) {
347 (void) printf("\n");
348 continue;
349 }
350 prop = umem_zalloc(attr.za_num_integers *
351 attr.za_integer_length, UMEM_NOFAIL);
352 (void) zap_lookup(os, object, attr.za_name,
353 attr.za_integer_length, attr.za_num_integers, prop);
354 if (attr.za_integer_length == 1) {
355 (void) printf("%s", (char *)prop);
356 } else {
357 for (i = 0; i < attr.za_num_integers; i++) {
358 switch (attr.za_integer_length) {
359 case 2:
360 (void) printf("%u ",
361 ((uint16_t *)prop)[i]);
362 break;
363 case 4:
364 (void) printf("%u ",
365 ((uint32_t *)prop)[i]);
366 break;
367 case 8:
368 (void) printf("%lld ",
369 (u_longlong_t)((int64_t *)prop)[i]);
370 break;
371 }
372 }
373 }
374 (void) printf("\n");
375 umem_free(prop, attr.za_num_integers * attr.za_integer_length);
376 }
377 zap_cursor_fini(&zc);
378 }
379
380 /*ARGSUSED*/
381 static void
382 dump_zpldir(objset_t *os, uint64_t object, void *data, size_t size)
383 {
384 zap_cursor_t zc;
385 zap_attribute_t attr;
386 const char *typenames[] = {
387 /* 0 */ "not specified",
388 /* 1 */ "FIFO",
389 /* 2 */ "Character Device",
390 /* 3 */ "3 (invalid)",
391 /* 4 */ "Directory",
392 /* 5 */ "5 (invalid)",
393 /* 6 */ "Block Device",
394 /* 7 */ "7 (invalid)",
395 /* 8 */ "Regular File",
396 /* 9 */ "9 (invalid)",
397 /* 10 */ "Symbolic Link",
398 /* 11 */ "11 (invalid)",
399 /* 12 */ "Socket",
400 /* 13 */ "Door",
401 /* 14 */ "Event Port",
402 /* 15 */ "15 (invalid)",
403 };
404
405 dump_zap_stats(os, object);
406 (void) printf("\n");
407
408 for (zap_cursor_init(&zc, os, object);
409 zap_cursor_retrieve(&zc, &attr) == 0;
410 zap_cursor_advance(&zc)) {
411 (void) printf("\t\t%s = %lld (type: %s)\n",
412 attr.za_name, ZFS_DIRENT_OBJ(attr.za_first_integer),
413 typenames[ZFS_DIRENT_TYPE(attr.za_first_integer)]);
414 }
415 zap_cursor_fini(&zc);
416 }
417
418 static void
419 dump_spacemap(objset_t *os, space_map_obj_t *smo, space_map_t *sm)
420 {
421 uint64_t alloc, offset, entry;
422 uint8_t mapshift = sm->sm_shift;
423 uint64_t mapstart = sm->sm_start;
424 char *ddata[] = { "ALLOC", "FREE", "CONDENSE", "INVALID",
425 "INVALID", "INVALID", "INVALID", "INVALID" };
426
427 if (smo->smo_object == 0)
428 return;
429
430 /*
431 * Print out the freelist entries in both encoded and decoded form.
432 */
433 alloc = 0;
434 for (offset = 0; offset < smo->smo_objsize; offset += sizeof (entry)) {
435 VERIFY(0 == dmu_read(os, smo->smo_object, offset,
436 sizeof (entry), &entry));
437 if (SM_DEBUG_DECODE(entry)) {
438 (void) printf("\t\t[%4llu] %s: txg %llu, pass %llu\n",
439 (u_longlong_t)(offset / sizeof (entry)),
440 ddata[SM_DEBUG_ACTION_DECODE(entry)],
441 (u_longlong_t)SM_DEBUG_TXG_DECODE(entry),
442 (u_longlong_t)SM_DEBUG_SYNCPASS_DECODE(entry));
443 } else {
444 (void) printf("\t\t[%4llu] %c range:"
445 " %08llx-%08llx size: %06llx\n",
446 (u_longlong_t)(offset / sizeof (entry)),
447 SM_TYPE_DECODE(entry) == SM_ALLOC ? 'A' : 'F',
448 (u_longlong_t)((SM_OFFSET_DECODE(entry) <<
449 mapshift) + mapstart),
450 (u_longlong_t)((SM_OFFSET_DECODE(entry) <<
451 mapshift) + mapstart + (SM_RUN_DECODE(entry) <<
452 mapshift)),
453 (u_longlong_t)(SM_RUN_DECODE(entry) << mapshift));
454 if (SM_TYPE_DECODE(entry) == SM_ALLOC)
455 alloc += SM_RUN_DECODE(entry) << mapshift;
456 else
457 alloc -= SM_RUN_DECODE(entry) << mapshift;
458 }
459 }
460 if (alloc != smo->smo_alloc) {
461 (void) printf("space_map_object alloc (%llu) INCONSISTENT "
462 "with space map summary (%llu)\n",
463 (u_longlong_t)smo->smo_alloc, (u_longlong_t)alloc);
464 }
465 }
466
467 static void
468 dump_metaslab(metaslab_t *msp)
469 {
470 char freebuf[5];
471 space_map_obj_t *smo = &msp->ms_smo;
472 vdev_t *vd = msp->ms_group->mg_vd;
473 spa_t *spa = vd->vdev_spa;
474
475 nicenum(msp->ms_map.sm_size - smo->smo_alloc, freebuf);
476
477 if (dump_opt['d'] <= 5) {
478 (void) printf("\t%10llx %10llu %5s\n",
479 (u_longlong_t)msp->ms_map.sm_start,
480 (u_longlong_t)smo->smo_object,
481 freebuf);
482 return;
483 }
484
485 (void) printf(
486 "\tvdev %llu offset %08llx spacemap %4llu free %5s\n",
487 (u_longlong_t)vd->vdev_id, (u_longlong_t)msp->ms_map.sm_start,
488 (u_longlong_t)smo->smo_object, freebuf);
489
490 ASSERT(msp->ms_map.sm_size == (1ULL << vd->vdev_ms_shift));
491
492 dump_spacemap(spa->spa_meta_objset, smo, &msp->ms_map);
493 }
494
495 static void
496 dump_metaslabs(spa_t *spa)
497 {
498 vdev_t *rvd = spa->spa_root_vdev;
499 vdev_t *vd;
500 int c, m;
501
502 (void) printf("\nMetaslabs:\n");
503
504 for (c = 0; c < rvd->vdev_children; c++) {
505 vd = rvd->vdev_child[c];
506
507 (void) printf("\n vdev %llu\n\n", (u_longlong_t)vd->vdev_id);
508
509 if (dump_opt['d'] <= 5) {
510 (void) printf("\t%10s %10s %5s\n",
511 "offset", "spacemap", "free");
512 (void) printf("\t%10s %10s %5s\n",
513 "------", "--------", "----");
514 }
515 for (m = 0; m < vd->vdev_ms_count; m++)
516 dump_metaslab(vd->vdev_ms[m]);
517 (void) printf("\n");
518 }
519 }
520
521 static void
522 dump_dtl(vdev_t *vd, int indent)
523 {
524 avl_tree_t *t = &vd->vdev_dtl_map.sm_root;
525 space_seg_t *ss;
526 vdev_t *pvd;
527 int c;
528
529 if (indent == 0)
530 (void) printf("\nDirty time logs:\n\n");
531
532 (void) printf("\t%*s%s\n", indent, "",
533 vd->vdev_path ? vd->vdev_path :
534 vd->vdev_parent ? vd->vdev_ops->vdev_op_type :
535 spa_name(vd->vdev_spa));
536
537 for (ss = avl_first(t); ss; ss = AVL_NEXT(t, ss)) {
538 /*
539 * Everything in this DTL must appear in all parent DTL unions.
540 */
541 for (pvd = vd; pvd; pvd = pvd->vdev_parent)
542 ASSERT(vdev_dtl_contains(&pvd->vdev_dtl_map,
543 ss->ss_start, ss->ss_end - ss->ss_start));
544 (void) printf("\t%*soutage [%llu,%llu] length %llu\n",
545 indent, "",
546 (u_longlong_t)ss->ss_start,
547 (u_longlong_t)ss->ss_end - 1,
548 (u_longlong_t)(ss->ss_end - ss->ss_start));
549 }
550
551 (void) printf("\n");
552
553 if (dump_opt['d'] > 5 && vd->vdev_children == 0) {
554 dump_spacemap(vd->vdev_spa->spa_meta_objset, &vd->vdev_dtl,
555 &vd->vdev_dtl_map);
556 (void) printf("\n");
557 }
558
559 for (c = 0; c < vd->vdev_children; c++)
560 dump_dtl(vd->vdev_child[c], indent + 4);
561 }
562
563 /*ARGSUSED*/
564 static void
565 dump_dnode(objset_t *os, uint64_t object, void *data, size_t size)
566 {
567 }
568
569 static uint64_t
570 blkid2offset(const dnode_phys_t *dnp, int level, uint64_t blkid)
571 {
572 if (level < 0)
573 return (blkid);
574
575 return ((blkid << (level * (dnp->dn_indblkshift - SPA_BLKPTRSHIFT))) *
576 dnp->dn_datablkszsec << SPA_MINBLOCKSHIFT);
577 }
578
579 static void
580 sprintf_blkptr_compact(char *blkbuf, blkptr_t *bp, int alldvas)
581 {
582 dva_t *dva = bp->blk_dva;
583 int ndvas = alldvas ? BP_GET_NDVAS(bp) : 1;
584 int i;
585
586 blkbuf[0] = '\0';
587
588 for (i = 0; i < ndvas; i++)
589 (void) sprintf(blkbuf + strlen(blkbuf), "%llu:%llx:%llx ",
590 (u_longlong_t)DVA_GET_VDEV(&dva[i]),
591 (u_longlong_t)DVA_GET_OFFSET(&dva[i]),
592 (u_longlong_t)DVA_GET_ASIZE(&dva[i]));
593
594 (void) sprintf(blkbuf + strlen(blkbuf), "%llxL/%llxP F=%llu B=%llu",
595 (u_longlong_t)BP_GET_LSIZE(bp),
596 (u_longlong_t)BP_GET_PSIZE(bp),
597 (u_longlong_t)bp->blk_fill,
598 (u_longlong_t)bp->blk_birth);
599 }
600
601 static void
602 print_indirect(blkptr_t *bp, const zbookmark_t *zb,
603 const dnode_phys_t *dnp)
604 {
605 char blkbuf[BP_SPRINTF_LEN];
606 int l;
607
608 ASSERT3U(BP_GET_TYPE(bp), ==, dnp->dn_type);
609 ASSERT3U(BP_GET_LEVEL(bp), ==, zb->zb_level);
610
611 (void) printf("%16llx ",
612 (u_longlong_t)blkid2offset(dnp, zb->zb_level, zb->zb_blkid));
613
614 ASSERT(zb->zb_level >= 0);
615
616 for (l = dnp->dn_nlevels - 1; l >= -1; l--) {
617 if (l == zb->zb_level) {
618 (void) printf("L%llx", (u_longlong_t)zb->zb_level);
619 } else {
620 (void) printf(" ");
621 }
622 }
623
624 sprintf_blkptr_compact(blkbuf, bp, dump_opt['d'] > 5 ? 1 : 0);
625 (void) printf("%s\n", blkbuf);
626 }
627
628 #define SET_BOOKMARK(zb, objset, object, level, blkid) \
629 { \
630 (zb)->zb_objset = objset; \
631 (zb)->zb_object = object; \
632 (zb)->zb_level = level; \
633 (zb)->zb_blkid = blkid; \
634 }
635
636 static int
637 visit_indirect(spa_t *spa, const dnode_phys_t *dnp,
638 blkptr_t *bp, const zbookmark_t *zb)
639 {
640 int err;
641
642 if (bp->blk_birth == 0)
643 return (0);
644
645 print_indirect(bp, zb, dnp);
646
647 if (BP_GET_LEVEL(bp) > 0) {
648 uint32_t flags = ARC_WAIT;
649 int i;
650 blkptr_t *cbp;
651 int epb = BP_GET_LSIZE(bp) >> SPA_BLKPTRSHIFT;
652 arc_buf_t *buf;
653 uint64_t fill = 0;
654
655 err = arc_read_nolock(NULL, spa, bp, arc_getbuf_func, &buf,
656 ZIO_PRIORITY_ASYNC_READ, ZIO_FLAG_CANFAIL, &flags, zb);
657 if (err)
658 return (err);
659
660 /* recursively visit blocks below this */
661 cbp = buf->b_data;
662 for (i = 0; i < epb; i++, cbp++) {
663 zbookmark_t czb;
664
665 SET_BOOKMARK(&czb, zb->zb_objset, zb->zb_object,
666 zb->zb_level - 1,
667 zb->zb_blkid * epb + i);
668 err = visit_indirect(spa, dnp, cbp, &czb);
669 if (err)
670 break;
671 fill += cbp->blk_fill;
672 }
673 ASSERT3U(fill, ==, bp->blk_fill);
674 (void) arc_buf_remove_ref(buf, &buf);
675 }
676
677 return (err);
678 }
679
680 /*ARGSUSED*/
681 static void
682 dump_indirect(dnode_t *dn)
683 {
684 dnode_phys_t *dnp = dn->dn_phys;
685 int j;
686 zbookmark_t czb;
687
688 (void) printf("Indirect blocks:\n");
689
690 SET_BOOKMARK(&czb, dmu_objset_id(&dn->dn_objset->os),
691 dn->dn_object, dnp->dn_nlevels - 1, 0);
692 for (j = 0; j < dnp->dn_nblkptr; j++) {
693 czb.zb_blkid = j;
694 (void) visit_indirect(dmu_objset_spa(&dn->dn_objset->os), dnp,
695 &dnp->dn_blkptr[j], &czb);
696 }
697
698 (void) printf("\n");
699 }
700
701 /*ARGSUSED*/
702 static void
703 dump_dsl_dir(objset_t *os, uint64_t object, void *data, size_t size)
704 {
705 dsl_dir_phys_t *dd = data;
706 time_t crtime;
707 char nice[6];
708
709 if (dd == NULL)
710 return;
711
712 ASSERT3U(size, >=, sizeof (dsl_dir_phys_t));
713
714 crtime = dd->dd_creation_time;
715 (void) printf("\t\tcreation_time = %s", ctime(&crtime));
716 (void) printf("\t\thead_dataset_obj = %llu\n",
717 (u_longlong_t)dd->dd_head_dataset_obj);
718 (void) printf("\t\tparent_dir_obj = %llu\n",
719 (u_longlong_t)dd->dd_parent_obj);
720 (void) printf("\t\torigin_obj = %llu\n",
721 (u_longlong_t)dd->dd_origin_obj);
722 (void) printf("\t\tchild_dir_zapobj = %llu\n",
723 (u_longlong_t)dd->dd_child_dir_zapobj);
724 nicenum(dd->dd_used_bytes, nice);
725 (void) printf("\t\tused_bytes = %s\n", nice);
726 nicenum(dd->dd_compressed_bytes, nice);
727 (void) printf("\t\tcompressed_bytes = %s\n", nice);
728 nicenum(dd->dd_uncompressed_bytes, nice);
729 (void) printf("\t\tuncompressed_bytes = %s\n", nice);
730 nicenum(dd->dd_quota, nice);
731 (void) printf("\t\tquota = %s\n", nice);
732 nicenum(dd->dd_reserved, nice);
733 (void) printf("\t\treserved = %s\n", nice);
734 (void) printf("\t\tprops_zapobj = %llu\n",
735 (u_longlong_t)dd->dd_props_zapobj);
736 (void) printf("\t\tdeleg_zapobj = %llu\n",
737 (u_longlong_t)dd->dd_deleg_zapobj);
738 (void) printf("\t\tflags = %llx\n",
739 (u_longlong_t)dd->dd_flags);
740
741 #define DO(which) \
742 nicenum(dd->dd_used_breakdown[DD_USED_ ## which], nice); \
743 (void) printf("\t\tused_breakdown[" #which "] = %s\n", nice)
744 DO(HEAD);
745 DO(SNAP);
746 DO(CHILD);
747 DO(CHILD_RSRV);
748 DO(REFRSRV);
749 #undef DO
750 }
751
752 /*ARGSUSED*/
753 static void
754 dump_dsl_dataset(objset_t *os, uint64_t object, void *data, size_t size)
755 {
756 dsl_dataset_phys_t *ds = data;
757 time_t crtime;
758 char used[6], compressed[6], uncompressed[6], unique[6];
759 char blkbuf[BP_SPRINTF_LEN];
760
761 if (ds == NULL)
762 return;
763
764 ASSERT(size == sizeof (*ds));
765 crtime = ds->ds_creation_time;
766 nicenum(ds->ds_used_bytes, used);
767 nicenum(ds->ds_compressed_bytes, compressed);
768 nicenum(ds->ds_uncompressed_bytes, uncompressed);
769 nicenum(ds->ds_unique_bytes, unique);
770 sprintf_blkptr(blkbuf, BP_SPRINTF_LEN, &ds->ds_bp);
771
772 (void) printf("\t\tdir_obj = %llu\n",
773 (u_longlong_t)ds->ds_dir_obj);
774 (void) printf("\t\tprev_snap_obj = %llu\n",
775 (u_longlong_t)ds->ds_prev_snap_obj);
776 (void) printf("\t\tprev_snap_txg = %llu\n",
777 (u_longlong_t)ds->ds_prev_snap_txg);
778 (void) printf("\t\tnext_snap_obj = %llu\n",
779 (u_longlong_t)ds->ds_next_snap_obj);
780 (void) printf("\t\tsnapnames_zapobj = %llu\n",
781 (u_longlong_t)ds->ds_snapnames_zapobj);
782 (void) printf("\t\tnum_children = %llu\n",
783 (u_longlong_t)ds->ds_num_children);
784 (void) printf("\t\tcreation_time = %s", ctime(&crtime));
785 (void) printf("\t\tcreation_txg = %llu\n",
786 (u_longlong_t)ds->ds_creation_txg);
787 (void) printf("\t\tdeadlist_obj = %llu\n",
788 (u_longlong_t)ds->ds_deadlist_obj);
789 (void) printf("\t\tused_bytes = %s\n", used);
790 (void) printf("\t\tcompressed_bytes = %s\n", compressed);
791 (void) printf("\t\tuncompressed_bytes = %s\n", uncompressed);
792 (void) printf("\t\tunique = %s\n", unique);
793 (void) printf("\t\tfsid_guid = %llu\n",
794 (u_longlong_t)ds->ds_fsid_guid);
795 (void) printf("\t\tguid = %llu\n",
796 (u_longlong_t)ds->ds_guid);
797 (void) printf("\t\tflags = %llx\n",
798 (u_longlong_t)ds->ds_flags);
799 (void) printf("\t\tnext_clones_obj = %llu\n",
800 (u_longlong_t)ds->ds_next_clones_obj);
801 (void) printf("\t\tprops_obj = %llu\n",
802 (u_longlong_t)ds->ds_props_obj);
803 (void) printf("\t\tbp = %s\n", blkbuf);
804 }
805
806 static void
807 dump_bplist(objset_t *mos, uint64_t object, char *name)
808 {
809 bplist_t bpl = { 0 };
810 blkptr_t blk, *bp = &blk;
811 uint64_t itor = 0;
812 char bytes[6];
813 char comp[6];
814 char uncomp[6];
815
816 if (dump_opt['d'] < 3)
817 return;
818
819 mutex_init(&bpl.bpl_lock, NULL, MUTEX_DEFAULT, NULL);
820 VERIFY(0 == bplist_open(&bpl, mos, object));
821 if (bplist_empty(&bpl)) {
822 bplist_close(&bpl);
823 mutex_destroy(&bpl.bpl_lock);
824 return;
825 }
826
827 nicenum(bpl.bpl_phys->bpl_bytes, bytes);
828 if (bpl.bpl_dbuf->db_size == sizeof (bplist_phys_t)) {
829 nicenum(bpl.bpl_phys->bpl_comp, comp);
830 nicenum(bpl.bpl_phys->bpl_uncomp, uncomp);
831 (void) printf("\n %s: %llu entries, %s (%s/%s comp)\n",
832 name, (u_longlong_t)bpl.bpl_phys->bpl_entries,
833 bytes, comp, uncomp);
834 } else {
835 (void) printf("\n %s: %llu entries, %s\n",
836 name, (u_longlong_t)bpl.bpl_phys->bpl_entries, bytes);
837 }
838
839 if (dump_opt['d'] < 5) {
840 bplist_close(&bpl);
841 mutex_destroy(&bpl.bpl_lock);
842 return;
843 }
844
845 (void) printf("\n");
846
847 while (bplist_iterate(&bpl, &itor, bp) == 0) {
848 char blkbuf[BP_SPRINTF_LEN];
849
850 ASSERT(bp->blk_birth != 0);
851 sprintf_blkptr_compact(blkbuf, bp, dump_opt['d'] > 5 ? 1 : 0);
852 (void) printf("\tItem %3llu: %s\n",
853 (u_longlong_t)itor - 1, blkbuf);
854 }
855
856 bplist_close(&bpl);
857 mutex_destroy(&bpl.bpl_lock);
858 }
859
860 static avl_tree_t idx_tree;
861 static avl_tree_t domain_tree;
862 static boolean_t fuid_table_loaded;
863
864 static void
865 fuid_table_destroy()
866 {
867 if (fuid_table_loaded) {
868 zfs_fuid_table_destroy(&idx_tree, &domain_tree);
869 fuid_table_loaded = B_FALSE;
870 }
871 }
872
873 /*
874 * print uid or gid information.
875 * For normal POSIX id just the id is printed in decimal format.
876 * For CIFS files with FUID the fuid is printed in hex followed by
877 * the doman-rid string.
878 */
879 static void
880 print_idstr(uint64_t id, const char *id_type)
881 {
882 if (FUID_INDEX(id)) {
883 char *domain;
884
885 domain = zfs_fuid_idx_domain(&idx_tree, FUID_INDEX(id));
886 (void) printf("\t%s %llx [%s-%d]\n", id_type,
887 (u_longlong_t)id, domain, (int)FUID_RID(id));
888 } else {
889 (void) printf("\t%s %llu\n", id_type, (u_longlong_t)id);
890 }
891
892 }
893
894 static void
895 dump_uidgid(objset_t *os, znode_phys_t *zp)
896 {
897 uint32_t uid_idx, gid_idx;
898
899 uid_idx = FUID_INDEX(zp->zp_uid);
900 gid_idx = FUID_INDEX(zp->zp_gid);
901
902 /* Load domain table, if not already loaded */
903 if (!fuid_table_loaded && (uid_idx || gid_idx)) {
904 uint64_t fuid_obj;
905
906 /* first find the fuid object. It lives in the master node */
907 VERIFY(zap_lookup(os, MASTER_NODE_OBJ, ZFS_FUID_TABLES,
908 8, 1, &fuid_obj) == 0);
909 (void) zfs_fuid_table_load(os, fuid_obj,
910 &idx_tree, &domain_tree);
911 fuid_table_loaded = B_TRUE;
912 }
913
914 print_idstr(zp->zp_uid, "uid");
915 print_idstr(zp->zp_gid, "gid");
916 }
917
918 /*ARGSUSED*/
919 static void
920 dump_znode(objset_t *os, uint64_t object, void *data, size_t size)
921 {
922 znode_phys_t *zp = data;
923 time_t z_crtime, z_atime, z_mtime, z_ctime;
924 char path[MAXPATHLEN * 2]; /* allow for xattr and failure prefix */
925 int error;
926
927 ASSERT(size >= sizeof (znode_phys_t));
928
929 error = zfs_obj_to_path(os, object, path, sizeof (path));
930 if (error != 0) {
931 (void) snprintf(path, sizeof (path), "\?\?\?<object#%llu>",
932 (u_longlong_t)object);
933 }
934
935 if (dump_opt['d'] < 3) {
936 (void) printf("\t%s\n", path);
937 return;
938 }
939
940 z_crtime = (time_t)zp->zp_crtime[0];
941 z_atime = (time_t)zp->zp_atime[0];
942 z_mtime = (time_t)zp->zp_mtime[0];
943 z_ctime = (time_t)zp->zp_ctime[0];
944
945 (void) printf("\tpath %s\n", path);
946 dump_uidgid(os, zp);
947 (void) printf("\tatime %s", ctime(&z_atime));
948 (void) printf("\tmtime %s", ctime(&z_mtime));
949 (void) printf("\tctime %s", ctime(&z_ctime));
950 (void) printf("\tcrtime %s", ctime(&z_crtime));
951 (void) printf("\tgen %llu\n", (u_longlong_t)zp->zp_gen);
952 (void) printf("\tmode %llo\n", (u_longlong_t)zp->zp_mode);
953 (void) printf("\tsize %llu\n", (u_longlong_t)zp->zp_size);
954 (void) printf("\tparent %llu\n", (u_longlong_t)zp->zp_parent);
955 (void) printf("\tlinks %llu\n", (u_longlong_t)zp->zp_links);
956 (void) printf("\txattr %llu\n", (u_longlong_t)zp->zp_xattr);
957 (void) printf("\trdev 0x%016llx\n", (u_longlong_t)zp->zp_rdev);
958 }
959
960 /*ARGSUSED*/
961 static void
962 dump_acl(objset_t *os, uint64_t object, void *data, size_t size)
963 {
964 }
965
966 /*ARGSUSED*/
967 static void
968 dump_dmu_objset(objset_t *os, uint64_t object, void *data, size_t size)
969 {
970 }
971
972 static object_viewer_t *object_viewer[DMU_OT_NUMTYPES] = {
973 dump_none, /* unallocated */
974 dump_zap, /* object directory */
975 dump_uint64, /* object array */
976 dump_none, /* packed nvlist */
977 dump_packed_nvlist, /* packed nvlist size */
978 dump_none, /* bplist */
979 dump_none, /* bplist header */
980 dump_none, /* SPA space map header */
981 dump_none, /* SPA space map */
982 dump_none, /* ZIL intent log */
983 dump_dnode, /* DMU dnode */
984 dump_dmu_objset, /* DMU objset */
985 dump_dsl_dir, /* DSL directory */
986 dump_zap, /* DSL directory child map */
987 dump_zap, /* DSL dataset snap map */
988 dump_zap, /* DSL props */
989 dump_dsl_dataset, /* DSL dataset */
990 dump_znode, /* ZFS znode */
991 dump_acl, /* ZFS V0 ACL */
992 dump_uint8, /* ZFS plain file */
993 dump_zpldir, /* ZFS directory */
994 dump_zap, /* ZFS master node */
995 dump_zap, /* ZFS delete queue */
996 dump_uint8, /* zvol object */
997 dump_zap, /* zvol prop */
998 dump_uint8, /* other uint8[] */
999 dump_uint64, /* other uint64[] */
1000 dump_zap, /* other ZAP */
1001 dump_zap, /* persistent error log */
1002 dump_uint8, /* SPA history */
1003 dump_uint64, /* SPA history offsets */
1004 dump_zap, /* Pool properties */
1005 dump_zap, /* DSL permissions */
1006 dump_acl, /* ZFS ACL */
1007 dump_uint8, /* ZFS SYSACL */
1008 dump_none, /* FUID nvlist */
1009 dump_packed_nvlist, /* FUID nvlist size */
1010 dump_zap, /* DSL dataset next clones */
1011 dump_zap, /* DSL scrub queue */
1012 };
1013
1014 static void
1015 dump_object(objset_t *os, uint64_t object, int verbosity, int *print_header)
1016 {
1017 dmu_buf_t *db = NULL;
1018 dmu_object_info_t doi;
1019 dnode_t *dn;
1020 void *bonus = NULL;
1021 size_t bsize = 0;
1022 char iblk[6], dblk[6], lsize[6], asize[6], bonus_size[6], segsize[6];
1023 char aux[50];
1024 int error;
1025
1026 if (*print_header) {
1027 (void) printf("\n Object lvl iblk dblk lsize"
1028 " asize type\n");
1029 *print_header = 0;
1030 }
1031
1032 if (object == 0) {
1033 dn = os->os->os_meta_dnode;
1034 } else {
1035 error = dmu_bonus_hold(os, object, FTAG, &db);
1036 if (error)
1037 fatal("dmu_bonus_hold(%llu) failed, errno %u",
1038 object, error);
1039 bonus = db->db_data;
1040 bsize = db->db_size;
1041 dn = ((dmu_buf_impl_t *)db)->db_dnode;
1042 }
1043 dmu_object_info_from_dnode(dn, &doi);
1044
1045 nicenum(doi.doi_metadata_block_size, iblk);
1046 nicenum(doi.doi_data_block_size, dblk);
1047 nicenum(doi.doi_data_block_size * (doi.doi_max_block_offset + 1),
1048 lsize);
1049 nicenum(doi.doi_physical_blks << 9, asize);
1050 nicenum(doi.doi_bonus_size, bonus_size);
1051
1052 aux[0] = '\0';
1053
1054 if (doi.doi_checksum != ZIO_CHECKSUM_INHERIT || verbosity >= 6) {
1055 (void) snprintf(aux + strlen(aux), sizeof (aux), " (K=%s)",
1056 zio_checksum_table[doi.doi_checksum].ci_name);
1057 }
1058
1059 if (doi.doi_compress != ZIO_COMPRESS_INHERIT || verbosity >= 6) {
1060 (void) snprintf(aux + strlen(aux), sizeof (aux), " (Z=%s)",
1061 zio_compress_table[doi.doi_compress].ci_name);
1062 }
1063
1064 (void) printf("%10lld %3u %5s %5s %5s %5s %s%s\n",
1065 (u_longlong_t)object, doi.doi_indirection, iblk, dblk, lsize,
1066 asize, dmu_ot[doi.doi_type].ot_name, aux);
1067
1068 if (doi.doi_bonus_type != DMU_OT_NONE && verbosity > 3) {
1069 (void) printf("%10s %3s %5s %5s %5s %5s %s\n",
1070 "", "", "", "", bonus_size, "bonus",
1071 dmu_ot[doi.doi_bonus_type].ot_name);
1072 }
1073
1074 if (verbosity >= 4) {
1075 object_viewer[doi.doi_bonus_type](os, object, bonus, bsize);
1076 object_viewer[doi.doi_type](os, object, NULL, 0);
1077 *print_header = 1;
1078 }
1079
1080 if (verbosity >= 5)
1081 dump_indirect(dn);
1082
1083 if (verbosity >= 5) {
1084 /*
1085 * Report the list of segments that comprise the object.
1086 */
1087 uint64_t start = 0;
1088 uint64_t end;
1089 uint64_t blkfill = 1;
1090 int minlvl = 1;
1091
1092 if (dn->dn_type == DMU_OT_DNODE) {
1093 minlvl = 0;
1094 blkfill = DNODES_PER_BLOCK;
1095 }
1096
1097 for (;;) {
1098 error = dnode_next_offset(dn,
1099 0, &start, minlvl, blkfill, 0);
1100 if (error)
1101 break;
1102 end = start;
1103 error = dnode_next_offset(dn,
1104 DNODE_FIND_HOLE, &end, minlvl, blkfill, 0);
1105 nicenum(end - start, segsize);
1106 (void) printf("\t\tsegment [%016llx, %016llx)"
1107 " size %5s\n", (u_longlong_t)start,
1108 (u_longlong_t)end, segsize);
1109 if (error)
1110 break;
1111 start = end;
1112 }
1113 }
1114
1115 if (db != NULL)
1116 dmu_buf_rele(db, FTAG);
1117 }
1118
1119 static char *objset_types[DMU_OST_NUMTYPES] = {
1120 "NONE", "META", "ZPL", "ZVOL", "OTHER", "ANY" };
1121
1122 static void
1123 dump_dir(objset_t *os)
1124 {
1125 dmu_objset_stats_t dds;
1126 uint64_t object, object_count;
1127 uint64_t refdbytes, usedobjs, scratch;
1128 char numbuf[8];
1129 char blkbuf[BP_SPRINTF_LEN];
1130 char osname[MAXNAMELEN];
1131 char *type = "UNKNOWN";
1132 int verbosity = dump_opt['d'];
1133 int print_header = 1;
1134 int i, error;
1135
1136 dmu_objset_fast_stat(os, &dds);
1137
1138 if (dds.dds_type < DMU_OST_NUMTYPES)
1139 type = objset_types[dds.dds_type];
1140
1141 if (dds.dds_type == DMU_OST_META) {
1142 dds.dds_creation_txg = TXG_INITIAL;
1143 usedobjs = os->os->os_rootbp->blk_fill;
1144 refdbytes = os->os->os_spa->spa_dsl_pool->
1145 dp_mos_dir->dd_phys->dd_used_bytes;
1146 } else {
1147 dmu_objset_space(os, &refdbytes, &scratch, &usedobjs, &scratch);
1148 }
1149
1150 ASSERT3U(usedobjs, ==, os->os->os_rootbp->blk_fill);
1151
1152 nicenum(refdbytes, numbuf);
1153
1154 if (verbosity >= 4) {
1155 (void) strcpy(blkbuf, ", rootbp ");
1156 sprintf_blkptr(blkbuf + strlen(blkbuf),
1157 BP_SPRINTF_LEN - strlen(blkbuf), os->os->os_rootbp);
1158 } else {
1159 blkbuf[0] = '\0';
1160 }
1161
1162 dmu_objset_name(os, osname);
1163
1164 (void) printf("Dataset %s [%s], ID %llu, cr_txg %llu, "
1165 "%s, %llu objects%s\n",
1166 osname, type, (u_longlong_t)dmu_objset_id(os),
1167 (u_longlong_t)dds.dds_creation_txg,
1168 numbuf, (u_longlong_t)usedobjs, blkbuf);
1169
1170 dump_intent_log(dmu_objset_zil(os));
1171
1172 if (dmu_objset_ds(os) != NULL)
1173 dump_bplist(dmu_objset_pool(os)->dp_meta_objset,
1174 dmu_objset_ds(os)->ds_phys->ds_deadlist_obj, "Deadlist");
1175
1176 if (verbosity < 2)
1177 return;
1178
1179 if (os->os->os_rootbp->blk_birth == 0)
1180 return;
1181
1182 if (zopt_objects != 0) {
1183 for (i = 0; i < zopt_objects; i++)
1184 dump_object(os, zopt_object[i], verbosity,
1185 &print_header);
1186 (void) printf("\n");
1187 return;
1188 }
1189
1190 dump_object(os, 0, verbosity, &print_header);
1191 object_count = 1;
1192
1193 object = 0;
1194 while ((error = dmu_object_next(os, &object, B_FALSE, 0)) == 0) {
1195 dump_object(os, object, verbosity, &print_header);
1196 object_count++;
1197 }
1198
1199 ASSERT3U(object_count, ==, usedobjs);
1200
1201 (void) printf("\n");
1202
1203 if (error != ESRCH)
1204 fatal("dmu_object_next() = %d", error);
1205 }
1206
1207 static void
1208 dump_uberblock(uberblock_t *ub)
1209 {
1210 time_t timestamp = ub->ub_timestamp;
1211
1212 (void) printf("Uberblock\n\n");
1213 (void) printf("\tmagic = %016llx\n", (u_longlong_t)ub->ub_magic);
1214 (void) printf("\tversion = %llu\n", (u_longlong_t)ub->ub_version);
1215 (void) printf("\ttxg = %llu\n", (u_longlong_t)ub->ub_txg);
1216 (void) printf("\tguid_sum = %llu\n", (u_longlong_t)ub->ub_guid_sum);
1217 (void) printf("\ttimestamp = %llu UTC = %s",
1218 (u_longlong_t)ub->ub_timestamp, asctime(localtime(&timestamp)));
1219 if (dump_opt['u'] >= 3) {
1220 char blkbuf[BP_SPRINTF_LEN];
1221 sprintf_blkptr(blkbuf, BP_SPRINTF_LEN, &ub->ub_rootbp);
1222 (void) printf("\trootbp = %s\n", blkbuf);
1223 }
1224 (void) printf("\n");
1225 }
1226
1227 static void
1228 dump_config(const char *pool)
1229 {
1230 spa_t *spa = NULL;
1231
1232 mutex_enter(&spa_namespace_lock);
1233 while ((spa = spa_next(spa)) != NULL) {
1234 if (pool == NULL)
1235 (void) printf("%s\n", spa_name(spa));
1236 if (pool == NULL || strcmp(pool, spa_name(spa)) == 0)
1237 dump_nvlist(spa->spa_config, 4);
1238 }
1239 mutex_exit(&spa_namespace_lock);
1240 }
1241
1242 static void
1243 dump_cachefile(const char *cachefile)
1244 {
1245 int fd;
1246 struct stat64 statbuf;
1247 char *buf;
1248 nvlist_t *config;
1249
1250 if ((fd = open64(cachefile, O_RDONLY)) < 0) {
1251 (void) printf("cannot open '%s': %s\n", cachefile,
1252 strerror(errno));
1253 exit(1);
1254 }
1255
1256 if (fstat64(fd, &statbuf) != 0) {
1257 (void) printf("failed to stat '%s': %s\n", cachefile,
1258 strerror(errno));
1259 exit(1);
1260 }
1261
1262 if ((buf = malloc(statbuf.st_size)) == NULL) {
1263 (void) fprintf(stderr, "failed to allocate %llu bytes\n",
1264 (u_longlong_t)statbuf.st_size);
1265 exit(1);
1266 }
1267
1268 if (read(fd, buf, statbuf.st_size) != statbuf.st_size) {
1269 (void) fprintf(stderr, "failed to read %llu bytes\n",
1270 (u_longlong_t)statbuf.st_size);
1271 exit(1);
1272 }
1273
1274 (void) close(fd);
1275
1276 if (nvlist_unpack(buf, statbuf.st_size, &config, 0) != 0) {
1277 (void) fprintf(stderr, "failed to unpack nvlist\n");
1278 exit(1);
1279 }
1280
1281 free(buf);
1282
1283 dump_nvlist(config, 0);
1284
1285 nvlist_free(config);
1286 }
1287
1288 static void
1289 dump_label(const char *dev)
1290 {
1291 int fd;
1292 vdev_label_t label;
1293 char *buf = label.vl_vdev_phys.vp_nvlist;
1294 size_t buflen = sizeof (label.vl_vdev_phys.vp_nvlist);
1295 struct stat64 statbuf;
1296 uint64_t psize;
1297 int l;
1298
1299 if ((fd = open64(dev, O_RDONLY)) < 0) {
1300 (void) printf("cannot open '%s': %s\n", dev, strerror(errno));
1301 exit(1);
1302 }
1303
1304 if (fstat64(fd, &statbuf) != 0) {
1305 (void) printf("failed to stat '%s': %s\n", dev,
1306 strerror(errno));
1307 exit(1);
1308 }
1309
1310 psize = statbuf.st_size;
1311 psize = P2ALIGN(psize, (uint64_t)sizeof (vdev_label_t));
1312
1313 for (l = 0; l < VDEV_LABELS; l++) {
1314
1315 nvlist_t *config = NULL;
1316
1317 (void) printf("--------------------------------------------\n");
1318 (void) printf("LABEL %d\n", l);
1319 (void) printf("--------------------------------------------\n");
1320
1321 if (pread64(fd, &label, sizeof (label),
1322 vdev_label_offset(psize, l, 0)) != sizeof (label)) {
1323 (void) printf("failed to read label %d\n", l);
1324 continue;
1325 }
1326
1327 if (nvlist_unpack(buf, buflen, &config, 0) != 0) {
1328 (void) printf("failed to unpack label %d\n", l);
1329 continue;
1330 }
1331 dump_nvlist(config, 4);
1332 nvlist_free(config);
1333 }
1334 }
1335
1336 /*ARGSUSED*/
1337 static int
1338 dump_one_dir(char *dsname, void *arg)
1339 {
1340 int error;
1341 objset_t *os;
1342
1343 error = dmu_objset_open(dsname, DMU_OST_ANY,
1344 DS_MODE_USER | DS_MODE_READONLY, &os);
1345 if (error) {
1346 (void) printf("Could not open %s\n", dsname);
1347 return (0);
1348 }
1349 dump_dir(os);
1350 dmu_objset_close(os);
1351 fuid_table_destroy();
1352 return (0);
1353 }
1354
1355 static void
1356 zdb_leak(space_map_t *sm, uint64_t start, uint64_t size)
1357 {
1358 vdev_t *vd = sm->sm_ppd;
1359
1360 (void) printf("leaked space: vdev %llu, offset 0x%llx, size %llu\n",
1361 (u_longlong_t)vd->vdev_id, (u_longlong_t)start, (u_longlong_t)size);
1362 }
1363
1364 /* ARGSUSED */
1365 static void
1366 zdb_space_map_load(space_map_t *sm)
1367 {
1368 }
1369
1370 static void
1371 zdb_space_map_unload(space_map_t *sm)
1372 {
1373 space_map_vacate(sm, zdb_leak, sm);
1374 }
1375
1376 /* ARGSUSED */
1377 static void
1378 zdb_space_map_claim(space_map_t *sm, uint64_t start, uint64_t size)
1379 {
1380 }
1381
1382 static space_map_ops_t zdb_space_map_ops = {
1383 zdb_space_map_load,
1384 zdb_space_map_unload,
1385 NULL, /* alloc */
1386 zdb_space_map_claim,
1387 NULL /* free */
1388 };
1389
1390 static void
1391 zdb_leak_init(spa_t *spa)
1392 {
1393 vdev_t *rvd = spa->spa_root_vdev;
1394
1395 for (int c = 0; c < rvd->vdev_children; c++) {
1396 vdev_t *vd = rvd->vdev_child[c];
1397 for (int m = 0; m < vd->vdev_ms_count; m++) {
1398 metaslab_t *msp = vd->vdev_ms[m];
1399 mutex_enter(&msp->ms_lock);
1400 VERIFY(space_map_load(&msp->ms_map, &zdb_space_map_ops,
1401 SM_ALLOC, &msp->ms_smo, spa->spa_meta_objset) == 0);
1402 msp->ms_map.sm_ppd = vd;
1403 mutex_exit(&msp->ms_lock);
1404 }
1405 }
1406 }
1407
1408 static void
1409 zdb_leak_fini(spa_t *spa)
1410 {
1411 vdev_t *rvd = spa->spa_root_vdev;
1412
1413 for (int c = 0; c < rvd->vdev_children; c++) {
1414 vdev_t *vd = rvd->vdev_child[c];
1415 for (int m = 0; m < vd->vdev_ms_count; m++) {
1416 metaslab_t *msp = vd->vdev_ms[m];
1417 mutex_enter(&msp->ms_lock);
1418 space_map_unload(&msp->ms_map);
1419 mutex_exit(&msp->ms_lock);
1420 }
1421 }
1422 }
1423
1424 /*
1425 * Verify that the sum of the sizes of all blocks in the pool adds up
1426 * to the SPA's sa_alloc total.
1427 */
1428 typedef struct zdb_blkstats {
1429 uint64_t zb_asize;
1430 uint64_t zb_lsize;
1431 uint64_t zb_psize;
1432 uint64_t zb_count;
1433 } zdb_blkstats_t;
1434
1435 #define DMU_OT_DEFERRED DMU_OT_NONE
1436 #define DMU_OT_TOTAL DMU_OT_NUMTYPES
1437
1438 #define ZB_TOTAL DN_MAX_LEVELS
1439
1440 typedef struct zdb_cb {
1441 zdb_blkstats_t zcb_type[ZB_TOTAL + 1][DMU_OT_TOTAL + 1];
1442 uint64_t zcb_errors[256];
1443 int zcb_readfails;
1444 int zcb_haderrors;
1445 } zdb_cb_t;
1446
1447 static void
1448 zdb_count_block(spa_t *spa, zdb_cb_t *zcb, blkptr_t *bp, dmu_object_type_t type)
1449 {
1450 for (int i = 0; i < 4; i++) {
1451 int l = (i < 2) ? BP_GET_LEVEL(bp) : ZB_TOTAL;
1452 int t = (i & 1) ? type : DMU_OT_TOTAL;
1453 zdb_blkstats_t *zb = &zcb->zcb_type[l][t];
1454
1455 zb->zb_asize += BP_GET_ASIZE(bp);
1456 zb->zb_lsize += BP_GET_LSIZE(bp);
1457 zb->zb_psize += BP_GET_PSIZE(bp);
1458 zb->zb_count++;
1459 }
1460
1461 if (dump_opt['S']) {
1462 boolean_t print_sig;
1463
1464 print_sig = !zdb_sig_user_data || (BP_GET_LEVEL(bp) == 0 &&
1465 BP_GET_TYPE(bp) == DMU_OT_PLAIN_FILE_CONTENTS);
1466
1467 if (BP_GET_CHECKSUM(bp) < zdb_sig_cksumalg)
1468 print_sig = B_FALSE;
1469
1470 if (print_sig) {
1471 (void) printf("%llu\t%lld\t%lld\t%s\t%s\t%s\t"
1472 "%llx:%llx:%llx:%llx\n",
1473 (u_longlong_t)BP_GET_LEVEL(bp),
1474 (longlong_t)BP_GET_PSIZE(bp),
1475 (longlong_t)BP_GET_NDVAS(bp),
1476 dmu_ot[BP_GET_TYPE(bp)].ot_name,
1477 zio_checksum_table[BP_GET_CHECKSUM(bp)].ci_name,
1478 zio_compress_table[BP_GET_COMPRESS(bp)].ci_name,
1479 (u_longlong_t)bp->blk_cksum.zc_word[0],
1480 (u_longlong_t)bp->blk_cksum.zc_word[1],
1481 (u_longlong_t)bp->blk_cksum.zc_word[2],
1482 (u_longlong_t)bp->blk_cksum.zc_word[3]);
1483 }
1484 }
1485
1486 if (!dump_opt['L'])
1487 VERIFY(zio_wait(zio_claim(NULL, spa, spa_first_txg(spa), bp,
1488 NULL, NULL, ZIO_FLAG_MUSTSUCCEED)) == 0);
1489 }
1490
1491 static int
1492 zdb_blkptr_cb(spa_t *spa, blkptr_t *bp, const zbookmark_t *zb,
1493 const dnode_phys_t *dnp, void *arg)
1494 {
1495 zdb_cb_t *zcb = arg;
1496 char blkbuf[BP_SPRINTF_LEN];
1497
1498 if (bp == NULL)
1499 return (0);
1500
1501 zdb_count_block(spa, zcb, bp, BP_GET_TYPE(bp));
1502
1503 if (dump_opt['c'] || dump_opt['S']) {
1504 int ioerr, size;
1505 void *data;
1506
1507 size = BP_GET_LSIZE(bp);
1508 data = malloc(size);
1509 ioerr = zio_wait(zio_read(NULL, spa, bp, data, size,
1510 NULL, NULL, ZIO_PRIORITY_ASYNC_READ,
1511 ZIO_FLAG_CANFAIL | ZIO_FLAG_SCRUB, zb));
1512 free(data);
1513
1514 /* We expect io errors on intent log */
1515 if (ioerr && BP_GET_TYPE(bp) != DMU_OT_INTENT_LOG) {
1516 zcb->zcb_haderrors = 1;
1517 zcb->zcb_errors[ioerr]++;
1518
1519 if (dump_opt['b'] >= 2)
1520 sprintf_blkptr(blkbuf, BP_SPRINTF_LEN, bp);
1521 else
1522 blkbuf[0] = '\0';
1523
1524 if (!dump_opt['S']) {
1525 (void) printf("zdb_blkptr_cb: "
1526 "Got error %d reading "
1527 "<%llu, %llu, %lld, %llx> %s -- skipping\n",
1528 ioerr,
1529 (u_longlong_t)zb->zb_objset,
1530 (u_longlong_t)zb->zb_object,
1531 (u_longlong_t)zb->zb_level,
1532 (u_longlong_t)zb->zb_blkid,
1533 blkbuf);
1534 }
1535 }
1536 }
1537
1538 zcb->zcb_readfails = 0;
1539
1540 if (dump_opt['b'] >= 4) {
1541 sprintf_blkptr(blkbuf, BP_SPRINTF_LEN, bp);
1542 (void) printf("objset %llu object %llu offset 0x%llx %s\n",
1543 (u_longlong_t)zb->zb_objset,
1544 (u_longlong_t)zb->zb_object,
1545 (u_longlong_t)blkid2offset(dnp, zb->zb_level, zb->zb_blkid),
1546 blkbuf);
1547 }
1548
1549 return (0);
1550 }
1551
1552 static int
1553 dump_block_stats(spa_t *spa)
1554 {
1555 zdb_cb_t zcb = { 0 };
1556 zdb_blkstats_t *zb, *tzb;
1557 uint64_t alloc, space, logalloc;
1558 vdev_t *rvd = spa->spa_root_vdev;
1559 int leaks = 0;
1560 int c, e;
1561
1562 if (!dump_opt['S']) {
1563 (void) printf("\nTraversing all blocks %s%s%s%s...\n",
1564 (dump_opt['c'] || !dump_opt['L']) ? "to verify " : "",
1565 dump_opt['c'] ? "checksums " : "",
1566 (dump_opt['c'] && !dump_opt['L']) ? "and verify " : "",
1567 !dump_opt['L'] ? "nothing leaked " : "");
1568 }
1569
1570 /*
1571 * Load all space maps as SM_ALLOC maps, then traverse the pool
1572 * claiming each block we discover. If the pool is perfectly
1573 * consistent, the space maps will be empty when we're done.
1574 * Anything left over is a leak; any block we can't claim (because
1575 * it's not part of any space map) is a double allocation,
1576 * reference to a freed block, or an unclaimed log block.
1577 */
1578 if (!dump_opt['L'])
1579 zdb_leak_init(spa);
1580
1581 /*
1582 * If there's a deferred-free bplist, process that first.
1583 */
1584 if (spa->spa_sync_bplist_obj != 0) {
1585 bplist_t *bpl = &spa->spa_sync_bplist;
1586 blkptr_t blk;
1587 uint64_t itor = 0;
1588
1589 VERIFY(0 == bplist_open(bpl, spa->spa_meta_objset,
1590 spa->spa_sync_bplist_obj));
1591
1592 while (bplist_iterate(bpl, &itor, &blk) == 0) {
1593 if (dump_opt['b'] >= 4) {
1594 char blkbuf[BP_SPRINTF_LEN];
1595 sprintf_blkptr(blkbuf, BP_SPRINTF_LEN, &blk);
1596 (void) printf("[%s] %s\n",
1597 "deferred free", blkbuf);
1598 }
1599 zdb_count_block(spa, &zcb, &blk, DMU_OT_DEFERRED);
1600 }
1601
1602 bplist_close(bpl);
1603 }
1604
1605 zcb.zcb_haderrors |= traverse_pool(spa, zdb_blkptr_cb, &zcb);
1606
1607 if (zcb.zcb_haderrors && !dump_opt['S']) {
1608 (void) printf("\nError counts:\n\n");
1609 (void) printf("\t%5s %s\n", "errno", "count");
1610 for (e = 0; e < 256; e++) {
1611 if (zcb.zcb_errors[e] != 0) {
1612 (void) printf("\t%5d %llu\n",
1613 e, (u_longlong_t)zcb.zcb_errors[e]);
1614 }
1615 }
1616 }
1617
1618 /*
1619 * Report any leaked segments.
1620 */
1621 if (!dump_opt['L'])
1622 zdb_leak_fini(spa);
1623
1624 /*
1625 * If we're interested in printing out the blkptr signatures,
1626 * return now as we don't print out anything else (including
1627 * errors and leaks).
1628 */
1629 if (dump_opt['S'])
1630 return (zcb.zcb_haderrors ? 3 : 0);
1631
1632 alloc = spa_get_alloc(spa);
1633 space = spa_get_space(spa);
1634
1635 /*
1636 * Log blocks allocated from a separate log device don't count
1637 * as part of the normal pool space; factor them in here.
1638 */
1639 logalloc = 0;
1640
1641 for (c = 0; c < rvd->vdev_children; c++)
1642 if (rvd->vdev_child[c]->vdev_islog)
1643 logalloc += rvd->vdev_child[c]->vdev_stat.vs_alloc;
1644
1645 tzb = &zcb.zcb_type[ZB_TOTAL][DMU_OT_TOTAL];
1646
1647 if (tzb->zb_asize == alloc + logalloc) {
1648 if (!dump_opt['L'])
1649 (void) printf("\n\tNo leaks (block sum matches space"
1650 " maps exactly)\n");
1651 } else {
1652 (void) printf("block traversal size %llu != alloc %llu "
1653 "(%s %lld)\n",
1654 (u_longlong_t)tzb->zb_asize,
1655 (u_longlong_t)alloc + logalloc,
1656 (dump_opt['L']) ? "unreachable" : "leaked",
1657 (longlong_t)(alloc + logalloc - tzb->zb_asize));
1658 leaks = 1;
1659 }
1660
1661 if (tzb->zb_count == 0)
1662 return (2);
1663
1664 (void) printf("\n");
1665 (void) printf("\tbp count: %10llu\n",
1666 (u_longlong_t)tzb->zb_count);
1667 (void) printf("\tbp logical: %10llu\t avg: %6llu\n",
1668 (u_longlong_t)tzb->zb_lsize,
1669 (u_longlong_t)(tzb->zb_lsize / tzb->zb_count));
1670 (void) printf("\tbp physical: %10llu\t avg:"
1671 " %6llu\tcompression: %6.2f\n",
1672 (u_longlong_t)tzb->zb_psize,
1673 (u_longlong_t)(tzb->zb_psize / tzb->zb_count),
1674 (double)tzb->zb_lsize / tzb->zb_psize);
1675 (void) printf("\tbp allocated: %10llu\t avg:"
1676 " %6llu\tcompression: %6.2f\n",
1677 (u_longlong_t)tzb->zb_asize,
1678 (u_longlong_t)(tzb->zb_asize / tzb->zb_count),
1679 (double)tzb->zb_lsize / tzb->zb_asize);
1680 (void) printf("\tSPA allocated: %10llu\tused: %5.2f%%\n",
1681 (u_longlong_t)alloc, 100.0 * alloc / space);
1682
1683 if (dump_opt['b'] >= 2) {
1684 int l, t, level;
1685 (void) printf("\nBlocks\tLSIZE\tPSIZE\tASIZE"
1686 "\t avg\t comp\t%%Total\tType\n");
1687
1688 for (t = 0; t <= DMU_OT_NUMTYPES; t++) {
1689 char csize[6], lsize[6], psize[6], asize[6], avg[6];
1690 char *typename;
1691
1692 typename = t == DMU_OT_DEFERRED ? "deferred free" :
1693 t == DMU_OT_TOTAL ? "Total" : dmu_ot[t].ot_name;
1694
1695 if (zcb.zcb_type[ZB_TOTAL][t].zb_asize == 0) {
1696 (void) printf("%6s\t%5s\t%5s\t%5s"
1697 "\t%5s\t%5s\t%6s\t%s\n",
1698 "-",
1699 "-",
1700 "-",
1701 "-",
1702 "-",
1703 "-",
1704 "-",
1705 typename);
1706 continue;
1707 }
1708
1709 for (l = ZB_TOTAL - 1; l >= -1; l--) {
1710 level = (l == -1 ? ZB_TOTAL : l);
1711 zb = &zcb.zcb_type[level][t];
1712
1713 if (zb->zb_asize == 0)
1714 continue;
1715
1716 if (dump_opt['b'] < 3 && level != ZB_TOTAL)
1717 continue;
1718
1719 if (level == 0 && zb->zb_asize ==
1720 zcb.zcb_type[ZB_TOTAL][t].zb_asize)
1721 continue;
1722
1723 nicenum(zb->zb_count, csize);
1724 nicenum(zb->zb_lsize, lsize);
1725 nicenum(zb->zb_psize, psize);
1726 nicenum(zb->zb_asize, asize);
1727 nicenum(zb->zb_asize / zb->zb_count, avg);
1728
1729 (void) printf("%6s\t%5s\t%5s\t%5s\t%5s"
1730 "\t%5.2f\t%6.2f\t",
1731 csize, lsize, psize, asize, avg,
1732 (double)zb->zb_lsize / zb->zb_psize,
1733 100.0 * zb->zb_asize / tzb->zb_asize);
1734
1735 if (level == ZB_TOTAL)
1736 (void) printf("%s\n", typename);
1737 else
1738 (void) printf(" L%d %s\n",
1739 level, typename);
1740 }
1741 }
1742 }
1743
1744 (void) printf("\n");
1745
1746 if (leaks)
1747 return (2);
1748
1749 if (zcb.zcb_haderrors)
1750 return (3);
1751
1752 return (0);
1753 }
1754
1755 static void
1756 dump_zpool(spa_t *spa)
1757 {
1758 dsl_pool_t *dp = spa_get_dsl(spa);
1759 int rc = 0;
1760
1761 if (dump_opt['u'])
1762 dump_uberblock(&spa->spa_uberblock);
1763
1764 if (dump_opt['d'] || dump_opt['i']) {
1765 dump_dir(dp->dp_meta_objset);
1766 if (dump_opt['d'] >= 3) {
1767 dump_bplist(dp->dp_meta_objset,
1768 spa->spa_sync_bplist_obj, "Deferred frees");
1769 dump_dtl(spa->spa_root_vdev, 0);
1770 dump_metaslabs(spa);
1771 }
1772 (void) dmu_objset_find(spa_name(spa), dump_one_dir, NULL,
1773 DS_FIND_SNAPSHOTS | DS_FIND_CHILDREN);
1774 }
1775
1776 if (dump_opt['b'] || dump_opt['c'] || dump_opt['S'])
1777 rc = dump_block_stats(spa);
1778
1779 if (dump_opt['s'])
1780 show_pool_stats(spa);
1781
1782 if (rc != 0)
1783 exit(rc);
1784 }
1785
1786 #define ZDB_FLAG_CHECKSUM 0x0001
1787 #define ZDB_FLAG_DECOMPRESS 0x0002
1788 #define ZDB_FLAG_BSWAP 0x0004
1789 #define ZDB_FLAG_GBH 0x0008
1790 #define ZDB_FLAG_INDIRECT 0x0010
1791 #define ZDB_FLAG_PHYS 0x0020
1792 #define ZDB_FLAG_RAW 0x0040
1793 #define ZDB_FLAG_PRINT_BLKPTR 0x0080
1794
1795 int flagbits[256];
1796
1797 static void
1798 zdb_print_blkptr(blkptr_t *bp, int flags)
1799 {
1800 dva_t *dva = bp->blk_dva;
1801 int d;
1802
1803 if (flags & ZDB_FLAG_BSWAP)
1804 byteswap_uint64_array((void *)bp, sizeof (blkptr_t));
1805 /*
1806 * Super-ick warning: This code is also duplicated in
1807 * cmd/mdb/common/modules/zfs/zfs.c . Yeah, I hate code
1808 * replication, too.
1809 */
1810 for (d = 0; d < BP_GET_NDVAS(bp); d++) {
1811 (void) printf("\tDVA[%d]: vdev_id %lld / %llx\n", d,
1812 (longlong_t)DVA_GET_VDEV(&dva[d]),
1813 (longlong_t)DVA_GET_OFFSET(&dva[d]));
1814 (void) printf("\tDVA[%d]: GANG: %-5s GRID: %04llx\t"
1815 "ASIZE: %llx\n", d,
1816 DVA_GET_GANG(&dva[d]) ? "TRUE" : "FALSE",
1817 (longlong_t)DVA_GET_GRID(&dva[d]),
1818 (longlong_t)DVA_GET_ASIZE(&dva[d]));
1819 (void) printf("\tDVA[%d]: :%llu:%llx:%llx:%s%s%s%s\n", d,
1820 (u_longlong_t)DVA_GET_VDEV(&dva[d]),
1821 (longlong_t)DVA_GET_OFFSET(&dva[d]),
1822 (longlong_t)BP_GET_PSIZE(bp),
1823 BP_SHOULD_BYTESWAP(bp) ? "e" : "",
1824 !DVA_GET_GANG(&dva[d]) && BP_GET_LEVEL(bp) != 0 ?
1825 "d" : "",
1826 DVA_GET_GANG(&dva[d]) ? "g" : "",
1827 BP_GET_COMPRESS(bp) != 0 ? "d" : "");
1828 }
1829 (void) printf("\tLSIZE: %-16llx\t\tPSIZE: %llx\n",
1830 (longlong_t)BP_GET_LSIZE(bp), (longlong_t)BP_GET_PSIZE(bp));
1831 (void) printf("\tENDIAN: %6s\t\t\t\t\tTYPE: %s\n",
1832 BP_GET_BYTEORDER(bp) ? "LITTLE" : "BIG",
1833 dmu_ot[BP_GET_TYPE(bp)].ot_name);
1834 (void) printf("\tBIRTH: %-16llx LEVEL: %-2llu\tFILL: %llx\n",
1835 (u_longlong_t)bp->blk_birth, (u_longlong_t)BP_GET_LEVEL(bp),
1836 (u_longlong_t)bp->blk_fill);
1837 (void) printf("\tCKFUNC: %-16s\t\tCOMP: %s\n",
1838 zio_checksum_table[BP_GET_CHECKSUM(bp)].ci_name,
1839 zio_compress_table[BP_GET_COMPRESS(bp)].ci_name);
1840 (void) printf("\tCKSUM: %llx:%llx:%llx:%llx\n",
1841 (u_longlong_t)bp->blk_cksum.zc_word[0],
1842 (u_longlong_t)bp->blk_cksum.zc_word[1],
1843 (u_longlong_t)bp->blk_cksum.zc_word[2],
1844 (u_longlong_t)bp->blk_cksum.zc_word[3]);
1845 }
1846
1847 static void
1848 zdb_dump_indirect(blkptr_t *bp, int nbps, int flags)
1849 {
1850 int i;
1851
1852 for (i = 0; i < nbps; i++)
1853 zdb_print_blkptr(&bp[i], flags);
1854 }
1855
1856 static void
1857 zdb_dump_gbh(void *buf, int flags)
1858 {
1859 zdb_dump_indirect((blkptr_t *)buf, SPA_GBH_NBLKPTRS, flags);
1860 }
1861
1862 static void
1863 zdb_dump_block_raw(void *buf, uint64_t size, int flags)
1864 {
1865 if (flags & ZDB_FLAG_BSWAP)
1866 byteswap_uint64_array(buf, size);
1867 (void) write(2, buf, size);
1868 }
1869
1870 static void
1871 zdb_dump_block(char *label, void *buf, uint64_t size, int flags)
1872 {
1873 uint64_t *d = (uint64_t *)buf;
1874 int nwords = size / sizeof (uint64_t);
1875 int do_bswap = !!(flags & ZDB_FLAG_BSWAP);
1876 int i, j;
1877 char *hdr, *c;
1878
1879
1880 if (do_bswap)
1881 hdr = " 7 6 5 4 3 2 1 0 f e d c b a 9 8";
1882 else
1883 hdr = " 0 1 2 3 4 5 6 7 8 9 a b c d e f";
1884
1885 (void) printf("\n%s\n%6s %s 0123456789abcdef\n", label, "", hdr);
1886
1887 for (i = 0; i < nwords; i += 2) {
1888 (void) printf("%06llx: %016llx %016llx ",
1889 (u_longlong_t)(i * sizeof (uint64_t)),
1890 (u_longlong_t)(do_bswap ? BSWAP_64(d[i]) : d[i]),
1891 (u_longlong_t)(do_bswap ? BSWAP_64(d[i + 1]) : d[i + 1]));
1892
1893 c = (char *)&d[i];
1894 for (j = 0; j < 2 * sizeof (uint64_t); j++)
1895 (void) printf("%c", isprint(c[j]) ? c[j] : '.');
1896 (void) printf("\n");
1897 }
1898 }
1899
1900 /*
1901 * There are two acceptable formats:
1902 * leaf_name - For example: c1t0d0 or /tmp/ztest.0a
1903 * child[.child]* - For example: 0.1.1
1904 *
1905 * The second form can be used to specify arbitrary vdevs anywhere
1906 * in the heirarchy. For example, in a pool with a mirror of
1907 * RAID-Zs, you can specify either RAID-Z vdev with 0.0 or 0.1 .
1908 */
1909 static vdev_t *
1910 zdb_vdev_lookup(vdev_t *vdev, char *path)
1911 {
1912 char *s, *p, *q;
1913 int i;
1914
1915 if (vdev == NULL)
1916 return (NULL);
1917
1918 /* First, assume the x.x.x.x format */
1919 i = (int)strtoul(path, &s, 10);
1920 if (s == path || (s && *s != '.' && *s != '\0'))
1921 goto name;
1922 if (i < 0 || i >= vdev->vdev_children)
1923 return (NULL);
1924
1925 vdev = vdev->vdev_child[i];
1926 if (*s == '\0')
1927 return (vdev);
1928 return (zdb_vdev_lookup(vdev, s+1));
1929
1930 name:
1931 for (i = 0; i < vdev->vdev_children; i++) {
1932 vdev_t *vc = vdev->vdev_child[i];
1933
1934 if (vc->vdev_path == NULL) {
1935 vc = zdb_vdev_lookup(vc, path);
1936 if (vc == NULL)
1937 continue;
1938 else
1939 return (vc);
1940 }
1941
1942 p = strrchr(vc->vdev_path, '/');
1943 p = p ? p + 1 : vc->vdev_path;
1944 q = &vc->vdev_path[strlen(vc->vdev_path) - 2];
1945
1946 if (strcmp(vc->vdev_path, path) == 0)
1947 return (vc);
1948 if (strcmp(p, path) == 0)
1949 return (vc);
1950 if (strcmp(q, "s0") == 0 && strncmp(p, path, q - p) == 0)
1951 return (vc);
1952 }
1953
1954 return (NULL);
1955 }
1956
1957 /*
1958 * Read a block from a pool and print it out. The syntax of the
1959 * block descriptor is:
1960 *
1961 * pool:vdev_specifier:offset:size[:flags]
1962 *
1963 * pool - The name of the pool you wish to read from
1964 * vdev_specifier - Which vdev (see comment for zdb_vdev_lookup)
1965 * offset - offset, in hex, in bytes
1966 * size - Amount of data to read, in hex, in bytes
1967 * flags - A string of characters specifying options
1968 * b: Decode a blkptr at given offset within block
1969 * *c: Calculate and display checksums
1970 * *d: Decompress data before dumping
1971 * e: Byteswap data before dumping
1972 * *g: Display data as a gang block header
1973 * *i: Display as an indirect block
1974 * p: Do I/O to physical offset
1975 * r: Dump raw data to stdout
1976 *
1977 * * = not yet implemented
1978 */
1979 static void
1980 zdb_read_block(char *thing, spa_t **spap)
1981 {
1982 spa_t *spa = *spap;
1983 int flags = 0;
1984 uint64_t offset = 0, size = 0, blkptr_offset = 0;
1985 zio_t *zio;
1986 vdev_t *vd;
1987 void *buf;
1988 char *s, *p, *dup, *pool, *vdev, *flagstr;
1989 int i, error, zio_flags;
1990
1991 dup = strdup(thing);
1992 s = strtok(dup, ":");
1993 pool = s ? s : "";
1994 s = strtok(NULL, ":");
1995 vdev = s ? s : "";
1996 s = strtok(NULL, ":");
1997 offset = strtoull(s ? s : "", NULL, 16);
1998 s = strtok(NULL, ":");
1999 size = strtoull(s ? s : "", NULL, 16);
2000 s = strtok(NULL, ":");
2001 flagstr = s ? s : "";
2002
2003 s = NULL;
2004 if (size == 0)
2005 s = "size must not be zero";
2006 if (!IS_P2ALIGNED(size, DEV_BSIZE))
2007 s = "size must be a multiple of sector size";
2008 if (!IS_P2ALIGNED(offset, DEV_BSIZE))
2009 s = "offset must be a multiple of sector size";
2010 if (s) {
2011 (void) printf("Invalid block specifier: %s - %s\n", thing, s);
2012 free(dup);
2013 return;
2014 }
2015
2016 for (s = strtok(flagstr, ":"); s; s = strtok(NULL, ":")) {
2017 for (i = 0; flagstr[i]; i++) {
2018 int bit = flagbits[(uchar_t)flagstr[i]];
2019
2020 if (bit == 0) {
2021 (void) printf("***Invalid flag: %c\n",
2022 flagstr[i]);
2023 continue;
2024 }
2025 flags |= bit;
2026
2027 /* If it's not something with an argument, keep going */
2028 if ((bit & (ZDB_FLAG_CHECKSUM | ZDB_FLAG_DECOMPRESS |
2029 ZDB_FLAG_PRINT_BLKPTR)) == 0)
2030 continue;
2031
2032 p = &flagstr[i + 1];
2033 if (bit == ZDB_FLAG_PRINT_BLKPTR)
2034 blkptr_offset = strtoull(p, &p, 16);
2035 if (*p != ':' && *p != '\0') {
2036 (void) printf("***Invalid flag arg: '%s'\n", s);
2037 free(dup);
2038 return;
2039 }
2040 }
2041 }
2042
2043 if (spa == NULL || strcmp(spa_name(spa), pool) != 0) {
2044 if (spa)
2045 spa_close(spa, (void *)zdb_read_block);
2046 error = spa_open(pool, spap, (void *)zdb_read_block);
2047 if (error)
2048 fatal("Failed to open pool '%s': %s",
2049 pool, strerror(error));
2050 spa = *spap;
2051 }
2052
2053 vd = zdb_vdev_lookup(spa->spa_root_vdev, vdev);
2054 if (vd == NULL) {
2055 (void) printf("***Invalid vdev: %s\n", vdev);
2056 free(dup);
2057 return;
2058 } else {
2059 if (vd->vdev_path)
2060 (void) printf("Found vdev: %s\n", vd->vdev_path);
2061 else
2062 (void) printf("Found vdev type: %s\n",
2063 vd->vdev_ops->vdev_op_type);
2064 }
2065
2066 buf = umem_alloc(size, UMEM_NOFAIL);
2067
2068 zio_flags = ZIO_FLAG_DONT_CACHE | ZIO_FLAG_DONT_QUEUE |
2069 ZIO_FLAG_DONT_PROPAGATE | ZIO_FLAG_DONT_RETRY;
2070
2071 spa_config_enter(spa, SCL_STATE, FTAG, RW_READER);
2072 zio = zio_root(spa, NULL, NULL, 0);
2073 /* XXX todo - cons up a BP so RAID-Z will be happy */
2074 zio_nowait(zio_vdev_child_io(zio, NULL, vd, offset, buf, size,
2075 ZIO_TYPE_READ, ZIO_PRIORITY_SYNC_READ, zio_flags, NULL, NULL));
2076 error = zio_wait(zio);
2077 spa_config_exit(spa, SCL_STATE, FTAG);
2078
2079 if (error) {
2080 (void) printf("Read of %s failed, error: %d\n", thing, error);
2081 goto out;
2082 }
2083
2084 if (flags & ZDB_FLAG_PRINT_BLKPTR)
2085 zdb_print_blkptr((blkptr_t *)(void *)
2086 ((uintptr_t)buf + (uintptr_t)blkptr_offset), flags);
2087 else if (flags & ZDB_FLAG_RAW)
2088 zdb_dump_block_raw(buf, size, flags);
2089 else if (flags & ZDB_FLAG_INDIRECT)
2090 zdb_dump_indirect((blkptr_t *)buf, size / sizeof (blkptr_t),
2091 flags);
2092 else if (flags & ZDB_FLAG_GBH)
2093 zdb_dump_gbh(buf, flags);
2094 else
2095 zdb_dump_block(thing, buf, size, flags);
2096
2097 out:
2098 umem_free(buf, size);
2099 free(dup);
2100 }
2101
2102 static boolean_t
2103 nvlist_string_match(nvlist_t *config, char *name, char *tgt)
2104 {
2105 char *s;
2106
2107 if (nvlist_lookup_string(config, name, &s) != 0)
2108 return (B_FALSE);
2109
2110 return (strcmp(s, tgt) == 0);
2111 }
2112
2113 static boolean_t
2114 nvlist_uint64_match(nvlist_t *config, char *name, uint64_t tgt)
2115 {
2116 uint64_t val;
2117
2118 if (nvlist_lookup_uint64(config, name, &val) != 0)
2119 return (B_FALSE);
2120
2121 return (val == tgt);
2122 }
2123
2124 static boolean_t
2125 vdev_child_guid_match(nvlist_t *vdev, uint64_t guid)
2126 {
2127 nvlist_t **child;
2128 uint_t c, children;
2129
2130 verify(nvlist_lookup_nvlist_array(vdev, ZPOOL_CONFIG_CHILDREN,
2131 &child, &children) == 0);
2132 for (c = 0; c < children; ++c)
2133 if (nvlist_uint64_match(child[c], ZPOOL_CONFIG_GUID, guid))
2134 return (B_TRUE);
2135 return (B_FALSE);
2136 }
2137
2138 static boolean_t
2139 vdev_child_string_match(nvlist_t *vdev, char *tgt)
2140 {
2141 nvlist_t **child;
2142 uint_t c, children;
2143
2144 verify(nvlist_lookup_nvlist_array(vdev, ZPOOL_CONFIG_CHILDREN,
2145 &child, &children) == 0);
2146 for (c = 0; c < children; ++c) {
2147 if (nvlist_string_match(child[c], ZPOOL_CONFIG_PATH, tgt) ||
2148 nvlist_string_match(child[c], ZPOOL_CONFIG_DEVID, tgt))
2149 return (B_TRUE);
2150 }
2151 return (B_FALSE);
2152 }
2153
2154 static boolean_t
2155 vdev_guid_match(nvlist_t *config, uint64_t guid)
2156 {
2157 nvlist_t *nvroot;
2158
2159 verify(nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE,
2160 &nvroot) == 0);
2161
2162 return (nvlist_uint64_match(nvroot, ZPOOL_CONFIG_GUID, guid) ||
2163 vdev_child_guid_match(nvroot, guid));
2164 }
2165
2166 static boolean_t
2167 vdev_string_match(nvlist_t *config, char *tgt)
2168 {
2169 nvlist_t *nvroot;
2170
2171 verify(nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE,
2172 &nvroot) == 0);
2173
2174 return (vdev_child_string_match(nvroot, tgt));
2175 }
2176
2177 static boolean_t
2178 pool_match(nvlist_t *config, char *tgt)
2179 {
2180 uint64_t guid = strtoull(tgt, NULL, 0);
2181
2182 if (guid != 0) {
2183 return (
2184 nvlist_uint64_match(config, ZPOOL_CONFIG_POOL_GUID, guid) ||
2185 vdev_guid_match(config, guid));
2186 } else {
2187 return (
2188 nvlist_string_match(config, ZPOOL_CONFIG_POOL_NAME, tgt) ||
2189 vdev_string_match(config, tgt));
2190 }
2191 }
2192
2193 static int
2194 find_exported_zpool(char *pool_id, nvlist_t **configp, char *vdev_dir)
2195 {
2196 nvlist_t *pools;
2197 int error = ENOENT;
2198 nvlist_t *match = NULL;
2199
2200 if (vdev_dir != NULL)
2201 pools = zpool_find_import_activeok(g_zfs, 1, &vdev_dir);
2202 else
2203 pools = zpool_find_import_activeok(g_zfs, 0, NULL);
2204
2205 if (pools != NULL) {
2206 nvpair_t *elem = NULL;
2207
2208 while ((elem = nvlist_next_nvpair(pools, elem)) != NULL) {
2209 verify(nvpair_value_nvlist(elem, configp) == 0);
2210 if (pool_match(*configp, pool_id)) {
2211 if (match != NULL) {
2212 (void) fatal(
2213 "More than one matching pool - "
2214 "specify guid/devid/device path.");
2215 } else {
2216 match = *configp;
2217 error = 0;
2218 }
2219 }
2220 }
2221 }
2222
2223 *configp = error ? NULL : match;
2224
2225 return (error);
2226 }
2227
2228 int
2229 main(int argc, char **argv)
2230 {
2231 int i, c;
2232 struct rlimit rl = { 1024, 1024 };
2233 spa_t *spa;
2234 objset_t *os = NULL;
2235 char *endstr;
2236 int dump_all = 1;
2237 int verbose = 0;
2238 int error;
2239 int exported = 0;
2240 char *vdev_dir = NULL;
2241
2242 (void) setrlimit(RLIMIT_NOFILE, &rl);
2243 (void) enable_extended_FILE_stdio(-1, -1);
2244
2245 dprintf_setup(&argc, argv);
2246
2247 while ((c = getopt(argc, argv, "udibcsvCLS:U:lRep:")) != -1) {
2248 switch (c) {
2249 case 'u':
2250 case 'd':
2251 case 'i':
2252 case 'b':
2253 case 'c':
2254 case 's':
2255 case 'C':
2256 case 'l':
2257 case 'R':
2258 dump_opt[c]++;
2259 dump_all = 0;
2260 break;
2261 case 'L':
2262 dump_opt[c]++;
2263 break;
2264 case 'v':
2265 verbose++;
2266 break;
2267 case 'U':
2268 spa_config_path = optarg;
2269 break;
2270 case 'e':
2271 exported = 1;
2272 break;
2273 case 'p':
2274 vdev_dir = optarg;
2275 break;
2276 case 'S':
2277 dump_opt[c]++;
2278 dump_all = 0;
2279 zdb_sig_user_data = (strncmp(optarg, "user:", 5) == 0);
2280 if (!zdb_sig_user_data && strncmp(optarg, "all:", 4))
2281 usage();
2282 endstr = strchr(optarg, ':') + 1;
2283 if (strcmp(endstr, "fletcher2") == 0)
2284 zdb_sig_cksumalg = ZIO_CHECKSUM_FLETCHER_2;
2285 else if (strcmp(endstr, "fletcher4") == 0)
2286 zdb_sig_cksumalg = ZIO_CHECKSUM_FLETCHER_4;
2287 else if (strcmp(endstr, "sha256") == 0)
2288 zdb_sig_cksumalg = ZIO_CHECKSUM_SHA256;
2289 else if (strcmp(endstr, "all") == 0)
2290 zdb_sig_cksumalg = ZIO_CHECKSUM_FLETCHER_2;
2291 else
2292 usage();
2293 break;
2294 default:
2295 usage();
2296 break;
2297 }
2298 }
2299
2300 if (vdev_dir != NULL && exported == 0) {
2301 (void) fprintf(stderr, "-p option requires use of -e\n");
2302 usage();
2303 }
2304
2305 kernel_init(FREAD);
2306 g_zfs = libzfs_init();
2307 ASSERT(g_zfs != NULL);
2308
2309 for (c = 0; c < 256; c++) {
2310 if (dump_all && c != 'l' && c != 'R')
2311 dump_opt[c] = 1;
2312 if (dump_opt[c])
2313 dump_opt[c] += verbose;
2314 }
2315
2316 argc -= optind;
2317 argv += optind;
2318
2319 if (argc < 1) {
2320 if (dump_opt['C']) {
2321 dump_cachefile(spa_config_path);
2322 return (0);
2323 }
2324 usage();
2325 }
2326
2327 if (dump_opt['l']) {
2328 dump_label(argv[0]);
2329 return (0);
2330 }
2331
2332 if (dump_opt['R']) {
2333 flagbits['b'] = ZDB_FLAG_PRINT_BLKPTR;
2334 flagbits['c'] = ZDB_FLAG_CHECKSUM;
2335 flagbits['d'] = ZDB_FLAG_DECOMPRESS;
2336 flagbits['e'] = ZDB_FLAG_BSWAP;
2337 flagbits['g'] = ZDB_FLAG_GBH;
2338 flagbits['i'] = ZDB_FLAG_INDIRECT;
2339 flagbits['p'] = ZDB_FLAG_PHYS;
2340 flagbits['r'] = ZDB_FLAG_RAW;
2341
2342 spa = NULL;
2343 while (argv[0]) {
2344 zdb_read_block(argv[0], &spa);
2345 argv++;
2346 argc--;
2347 }
2348 if (spa)
2349 spa_close(spa, (void *)zdb_read_block);
2350 return (0);
2351 }
2352
2353 if (dump_opt['C'])
2354 dump_config(argv[0]);
2355
2356 error = 0;
2357 if (exported) {
2358 /*
2359 * Check to see if the name refers to an exported zpool
2360 */
2361 char *slash;
2362 nvlist_t *exported_conf = NULL;
2363
2364 if ((slash = strchr(argv[0], '/')) != NULL)
2365 *slash = '\0';
2366
2367 error = find_exported_zpool(argv[0], &exported_conf, vdev_dir);
2368 if (error == 0) {
2369 nvlist_t *nvl = NULL;
2370
2371 if (vdev_dir != NULL) {
2372 if (nvlist_alloc(&nvl, NV_UNIQUE_NAME, 0) != 0)
2373 error = ENOMEM;
2374 else if (nvlist_add_string(nvl,
2375 zpool_prop_to_name(ZPOOL_PROP_ALTROOT),
2376 vdev_dir) != 0)
2377 error = ENOMEM;
2378 }
2379
2380 if (error == 0)
2381 error = spa_import_faulted(argv[0],
2382 exported_conf, nvl);
2383
2384 nvlist_free(nvl);
2385 }
2386
2387 if (slash != NULL)
2388 *slash = '/';
2389 }
2390
2391 if (error == 0) {
2392 if (strchr(argv[0], '/') != NULL) {
2393 error = dmu_objset_open(argv[0], DMU_OST_ANY,
2394 DS_MODE_USER | DS_MODE_READONLY, &os);
2395 } else {
2396 error = spa_open(argv[0], &spa, FTAG);
2397 }
2398 }
2399
2400 if (error)
2401 fatal("can't open %s: %s", argv[0], strerror(error));
2402
2403 argv++;
2404 if (--argc > 0) {
2405 zopt_objects = argc;
2406 zopt_object = calloc(zopt_objects, sizeof (uint64_t));
2407 for (i = 0; i < zopt_objects; i++) {
2408 errno = 0;
2409 zopt_object[i] = strtoull(argv[i], NULL, 0);
2410 if (zopt_object[i] == 0 && errno != 0)
2411 fatal("bad object number %s: %s",
2412 argv[i], strerror(errno));
2413 }
2414 }
2415
2416 if (os != NULL) {
2417 dump_dir(os);
2418 dmu_objset_close(os);
2419 } else {
2420 dump_zpool(spa);
2421 spa_close(spa, FTAG);
2422 }
2423
2424 fuid_table_destroy();
2425
2426 libzfs_fini(g_zfs);
2427 kernel_fini();
2428
2429 return (0);
2430 }
NetBSD on the FriendlyARM MINI2440