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, Version 1.0 only 6 * (the "License"). You may not use this file except in compliance 7 * with the License. 8 * 9 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE 10 * or http://www.opensolaris.org/os/licensing. 11 * See the License for the specific language governing permissions 12 * and limitations under the License. 13 * 14 * When distributing Covered Code, include this CDDL HEADER in each 15 * file and include the License file at usr/src/OPENSOLARIS.LICENSE. 16 * If applicable, add the following below this CDDL HEADER, with the 17 * fields enclosed by brackets "[]" replaced with your own identifying 18 * information: Portions Copyright [yyyy] [name of copyright owner] 19 * 20 * CDDL HEADER END 21 */ 22 /* 23 * Copyright 2005 Sun Microsystems, Inc. All rights reserved. 24 * Use is subject to license terms. 25 */ 26 27 #pragma ident "%Z%%M% %I% %E% SMI" 28 29 #include <stdio.h> 30 #include <stdlib.h> 31 #include <sys/zfs_context.h> 32 #include <sys/spa.h> 33 #include <sys/spa_impl.h> 34 #include <sys/dmu.h> 35 #include <sys/zap.h> 36 #include <sys/fs/zfs.h> 37 #include <sys/zfs_znode.h> 38 #include <sys/vdev.h> 39 #include <sys/vdev_impl.h> 40 #include <sys/metaslab_impl.h> 41 #include <sys/dmu_objset.h> 42 #include <sys/dsl_dir.h> 43 #include <sys/dsl_dataset.h> 44 #include <sys/dsl_pool.h> 45 #include <sys/dbuf.h> 46 #include <sys/zil.h> 47 #include <sys/zil_impl.h> 48 #include <sys/stat.h> 49 #include <sys/resource.h> 50 #include <sys/dmu_traverse.h> 51 #include <sys/zio_checksum.h> 52 #include <sys/zio_compress.h> 53 54 const char cmdname[] = "zdb"; 55 uint8_t dump_opt[256]; 56 57 typedef void object_viewer_t(objset_t *, uint64_t, void *data, size_t size); 58 59 extern void dump_intent_log(zilog_t *); 60 uint64_t *zopt_object = NULL; 61 int zopt_objects = 0; 62 int zdb_advance = ADVANCE_PRE; 63 zbookmark_t zdb_noread = { 0, 0, ZB_NO_LEVEL, 0 }; 64 65 /* 66 * These libumem hooks provide a reasonable set of defaults for the allocator's 67 * debugging facilities. 68 */ 69 const char * 70 _umem_debug_init() 71 { 72 return ("default,verbose"); /* $UMEM_DEBUG setting */ 73 } 74 75 const char * 76 _umem_logging_init(void) 77 { 78 return ("fail,contents"); /* $UMEM_LOGGING setting */ 79 } 80 81 static void 82 usage(void) 83 { 84 (void) fprintf(stderr, 85 "Usage: %s [-udibcsvLU] [-O order] [-B os:obj:level:blkid] " 86 "dataset [object...]\n" 87 " %s -C [pool]\n" 88 " %s -l dev\n", 89 cmdname, cmdname, cmdname); 90 91 (void) fprintf(stderr, " -u uberblock\n"); 92 (void) fprintf(stderr, " -d datasets\n"); 93 (void) fprintf(stderr, " -C cached pool configuration\n"); 94 (void) fprintf(stderr, " -i intent logs\n"); 95 (void) fprintf(stderr, " -b block statistics\n"); 96 (void) fprintf(stderr, " -c checksum all data blocks\n"); 97 (void) fprintf(stderr, " -s report stats on zdb's I/O\n"); 98 (void) fprintf(stderr, " -v verbose (applies to all others)\n"); 99 (void) fprintf(stderr, " -l dump label contents\n"); 100 (void) fprintf(stderr, " -L live pool (allows some errors)\n"); 101 (void) fprintf(stderr, " -O [!]<pre|post|prune|data|holes> " 102 "visitation order\n"); 103 (void) fprintf(stderr, " -U use zpool.cache in /tmp\n"); 104 (void) fprintf(stderr, " -B objset:object:level:blkid -- " 105 "simulate bad block\n"); 106 (void) fprintf(stderr, "Specify an option more than once (e.g. -bb) " 107 "to make only that option verbose\n"); 108 (void) fprintf(stderr, "Default is to dump everything non-verbosely\n"); 109 exit(1); 110 } 111 112 static void 113 fatal(const char *fmt, ...) 114 { 115 va_list ap; 116 117 va_start(ap, fmt); 118 (void) fprintf(stderr, "%s: ", cmdname); 119 (void) vfprintf(stderr, fmt, ap); 120 va_end(ap); 121 (void) fprintf(stderr, "\n"); 122 123 exit(1); 124 } 125 126 static void 127 dump_nvlist(nvlist_t *list, int indent) 128 { 129 nvpair_t *elem = NULL; 130 131 while ((elem = nvlist_next_nvpair(list, elem)) != NULL) { 132 switch (nvpair_type(elem)) { 133 case DATA_TYPE_STRING: 134 { 135 char *value; 136 137 VERIFY(nvpair_value_string(elem, &value) == 0); 138 (void) printf("%*s%s='%s'\n", indent, "", 139 nvpair_name(elem), value); 140 } 141 break; 142 143 case DATA_TYPE_UINT64: 144 { 145 uint64_t value; 146 147 VERIFY(nvpair_value_uint64(elem, &value) == 0); 148 (void) printf("%*s%s=%llu\n", indent, "", 149 nvpair_name(elem), (u_longlong_t)value); 150 } 151 break; 152 153 case DATA_TYPE_NVLIST: 154 { 155 nvlist_t *value; 156 157 VERIFY(nvpair_value_nvlist(elem, &value) == 0); 158 (void) printf("%*s%s\n", indent, "", 159 nvpair_name(elem)); 160 dump_nvlist(value, indent + 4); 161 } 162 break; 163 164 case DATA_TYPE_NVLIST_ARRAY: 165 { 166 nvlist_t **value; 167 uint_t c, count; 168 169 VERIFY(nvpair_value_nvlist_array(elem, &value, 170 &count) == 0); 171 172 for (c = 0; c < count; c++) { 173 (void) printf("%*s%s[%u]\n", indent, "", 174 nvpair_name(elem), c); 175 dump_nvlist(value[c], indent + 8); 176 } 177 } 178 break; 179 180 default: 181 182 (void) printf("bad config type %d for %s\n", 183 nvpair_type(elem), nvpair_name(elem)); 184 } 185 } 186 } 187 188 /* ARGSUSED */ 189 static void 190 dump_packed_nvlist(objset_t *os, uint64_t object, void *data, size_t size) 191 { 192 nvlist_t *nv; 193 size_t nvsize = *(uint64_t *)data; 194 char *packed = umem_alloc(nvsize, UMEM_NOFAIL); 195 196 dmu_read(os, object, 0, nvsize, packed); 197 198 VERIFY(nvlist_unpack(packed, nvsize, &nv, 0) == 0); 199 200 umem_free(packed, nvsize); 201 202 dump_nvlist(nv, 8); 203 204 nvlist_free(nv); 205 } 206 207 const char dump_zap_stars[] = "****************************************"; 208 const int dump_zap_width = sizeof (dump_zap_stars) - 1; 209 210 static void 211 dump_zap_histogram(uint64_t histo[ZAP_HISTOGRAM_SIZE]) 212 { 213 int i; 214 int minidx = ZAP_HISTOGRAM_SIZE - 1; 215 int maxidx = 0; 216 uint64_t max = 0; 217 218 for (i = 0; i < ZAP_HISTOGRAM_SIZE; i++) { 219 if (histo[i] > max) 220 max = histo[i]; 221 if (histo[i] > 0 && i > maxidx) 222 maxidx = i; 223 if (histo[i] > 0 && i < minidx) 224 minidx = i; 225 } 226 227 if (max < dump_zap_width) 228 max = dump_zap_width; 229 230 for (i = minidx; i <= maxidx; i++) 231 (void) printf("\t\t\t%u: %6llu %s\n", i, (u_longlong_t)histo[i], 232 &dump_zap_stars[(max - histo[i]) * dump_zap_width / max]); 233 } 234 235 static void 236 dump_zap_stats(objset_t *os, uint64_t object) 237 { 238 int error; 239 zap_stats_t zs; 240 241 error = zap_get_stats(os, object, &zs); 242 if (error) 243 return; 244 245 if (zs.zs_ptrtbl_len == 0) { 246 ASSERT(zs.zs_num_blocks == 1); 247 (void) printf("\tmicrozap: %llu bytes, %llu entries\n", 248 (u_longlong_t)zs.zs_blocksize, 249 (u_longlong_t)zs.zs_num_entries); 250 return; 251 } 252 253 (void) printf("\tFat ZAP stats:\n"); 254 (void) printf("\t\tPointer table: %llu elements\n", 255 (u_longlong_t)zs.zs_ptrtbl_len); 256 (void) printf("\t\tZAP entries: %llu\n", 257 (u_longlong_t)zs.zs_num_entries); 258 (void) printf("\t\tLeaf blocks: %llu\n", 259 (u_longlong_t)zs.zs_num_leafs); 260 (void) printf("\t\tTotal blocks: %llu\n", 261 (u_longlong_t)zs.zs_num_blocks); 262 (void) printf("\t\tOversize blocks: %llu\n", 263 (u_longlong_t)zs.zs_num_blocks_large); 264 265 (void) printf("\t\tLeafs with 2^n pointers:\n"); 266 dump_zap_histogram(zs.zs_leafs_with_2n_pointers); 267 268 (void) printf("\t\tLeafs with n chained:\n"); 269 dump_zap_histogram(zs.zs_leafs_with_n_chained); 270 271 (void) printf("\t\tBlocks with n*5 entries:\n"); 272 dump_zap_histogram(zs.zs_blocks_with_n5_entries); 273 274 (void) printf("\t\tBlocks n/10 full:\n"); 275 dump_zap_histogram(zs.zs_blocks_n_tenths_full); 276 277 (void) printf("\t\tEntries with n chunks:\n"); 278 dump_zap_histogram(zs.zs_entries_using_n_chunks); 279 280 (void) printf("\t\tBuckets with n entries:\n"); 281 dump_zap_histogram(zs.zs_buckets_with_n_entries); 282 } 283 284 /*ARGSUSED*/ 285 static void 286 dump_none(objset_t *os, uint64_t object, void *data, size_t size) 287 { 288 } 289 290 /*ARGSUSED*/ 291 void 292 dump_uint8(objset_t *os, uint64_t object, void *data, size_t size) 293 { 294 } 295 296 /*ARGSUSED*/ 297 static void 298 dump_uint64(objset_t *os, uint64_t object, void *data, size_t size) 299 { 300 } 301 302 /*ARGSUSED*/ 303 static void 304 dump_zap(objset_t *os, uint64_t object, void *data, size_t size) 305 { 306 zap_cursor_t zc; 307 zap_attribute_t attr; 308 void *prop; 309 int i; 310 311 dump_zap_stats(os, object); 312 (void) printf("\n"); 313 314 for (zap_cursor_init(&zc, os, object); 315 zap_cursor_retrieve(&zc, &attr) == 0; 316 zap_cursor_advance(&zc)) { 317 (void) printf("\t\t%s = ", attr.za_name); 318 if (attr.za_num_integers == 0) { 319 (void) printf("\n"); 320 continue; 321 } 322 prop = umem_zalloc(attr.za_num_integers * 323 attr.za_integer_length, UMEM_NOFAIL); 324 (void) zap_lookup(os, object, attr.za_name, 325 attr.za_integer_length, attr.za_num_integers, prop); 326 if (attr.za_integer_length == 1) { 327 (void) printf("%s", (char *)prop); 328 } else { 329 for (i = 0; i < attr.za_num_integers; i++) { 330 switch (attr.za_integer_length) { 331 case 2: 332 (void) printf("%u ", 333 ((uint16_t *)prop)[i]); 334 break; 335 case 4: 336 (void) printf("%u ", 337 ((uint32_t *)prop)[i]); 338 break; 339 case 8: 340 (void) printf("%lld ", 341 (u_longlong_t)((int64_t *)prop)[i]); 342 break; 343 } 344 } 345 } 346 (void) printf("\n"); 347 umem_free(prop, attr.za_num_integers * attr.za_integer_length); 348 } 349 zap_cursor_fini(&zc); 350 } 351 352 static void 353 dump_spacemap(objset_t *os, space_map_obj_t *smo, space_map_t *sm) 354 { 355 uint64_t alloc, offset, entry; 356 int mapshift = sm->sm_shift; 357 uint64_t mapstart = sm->sm_start; 358 char *ddata[] = { "ALLOC", "FREE", "CONDENSE", "INVALID" }; 359 360 if (smo->smo_object == 0) 361 return; 362 363 /* 364 * Print out the freelist entries in both encoded and decoded form. 365 */ 366 alloc = 0; 367 for (offset = 0; offset < smo->smo_objsize; offset += sizeof (entry)) { 368 dmu_read(os, smo->smo_object, offset, sizeof (entry), &entry); 369 if (SM_DEBUG_DECODE(entry)) { 370 (void) printf("\t\t[%4llu] %s: txg %llu, pass %llu\n", 371 (u_longlong_t)(offset / sizeof (entry)), 372 ddata[SM_DEBUG_ACTION_DECODE(entry)], 373 SM_DEBUG_TXG_DECODE(entry), 374 SM_DEBUG_SYNCPASS_DECODE(entry)); 375 } else { 376 (void) printf("\t\t[%4llu] %c range:" 377 " %08llx-%08llx size: %06llx\n", 378 (u_longlong_t)(offset / sizeof (entry)), 379 SM_TYPE_DECODE(entry) == SM_ALLOC ? 'A' : 'F', 380 (SM_OFFSET_DECODE(entry) << mapshift) + mapstart, 381 (SM_OFFSET_DECODE(entry) << mapshift) + mapstart + 382 (SM_RUN_DECODE(entry) << mapshift), 383 (SM_RUN_DECODE(entry) << mapshift)); 384 if (SM_TYPE_DECODE(entry) == SM_ALLOC) 385 alloc += SM_RUN_DECODE(entry) << mapshift; 386 else 387 alloc -= SM_RUN_DECODE(entry) << mapshift; 388 } 389 } 390 if (alloc != smo->smo_alloc) { 391 (void) printf("space_map_object alloc (%llu) INCONSISTENT " 392 "with space map summary (%llu)\n", 393 (u_longlong_t)smo->smo_alloc, (u_longlong_t)alloc); 394 } 395 } 396 397 static void 398 dump_metaslab(metaslab_t *msp) 399 { 400 char freebuf[5]; 401 space_map_obj_t *smo = msp->ms_smo; 402 vdev_t *vd = msp->ms_group->mg_vd; 403 spa_t *spa = vd->vdev_spa; 404 405 nicenum(msp->ms_map.sm_size - smo->smo_alloc, freebuf); 406 407 if (dump_opt['d'] <= 5) { 408 (void) printf("\t%10llx %10llu %5s\n", 409 (u_longlong_t)msp->ms_map.sm_start, 410 (u_longlong_t)smo->smo_object, 411 freebuf); 412 return; 413 } 414 415 (void) printf( 416 "\tvdev %llu offset %08llx spacemap %4llu free %5s\n", 417 (u_longlong_t)vd->vdev_id, (u_longlong_t)msp->ms_map.sm_start, 418 (u_longlong_t)smo->smo_object, freebuf); 419 420 ASSERT(msp->ms_map.sm_size == (1ULL << vd->vdev_ms_shift)); 421 422 dump_spacemap(spa->spa_meta_objset, smo, &msp->ms_map); 423 } 424 425 static void 426 dump_metaslabs(spa_t *spa) 427 { 428 vdev_t *rvd = spa->spa_root_vdev; 429 vdev_t *vd; 430 int c, m; 431 432 (void) printf("\nMetaslabs:\n"); 433 434 for (c = 0; c < rvd->vdev_children; c++) { 435 vd = rvd->vdev_child[c]; 436 437 spa_config_enter(spa, RW_READER); 438 (void) printf("\n vdev %llu = %s\n\n", 439 (u_longlong_t)vd->vdev_id, vdev_description(vd)); 440 spa_config_exit(spa); 441 442 if (dump_opt['d'] <= 5) { 443 (void) printf("\t%10s %10s %5s\n", 444 "offset", "spacemap", "free"); 445 (void) printf("\t%10s %10s %5s\n", 446 "------", "--------", "----"); 447 } 448 for (m = 0; m < vd->vdev_ms_count; m++) 449 dump_metaslab(vd->vdev_ms[m]); 450 (void) printf("\n"); 451 } 452 } 453 454 static void 455 dump_dtl(vdev_t *vd, int indent) 456 { 457 avl_tree_t *t = &vd->vdev_dtl_map.sm_root; 458 spa_t *spa = vd->vdev_spa; 459 space_seg_t *ss; 460 vdev_t *pvd; 461 int c; 462 463 if (indent == 0) 464 (void) printf("\nDirty time logs:\n\n"); 465 466 spa_config_enter(spa, RW_READER); 467 (void) printf("\t%*s%s\n", indent, "", vdev_description(vd)); 468 spa_config_exit(spa); 469 470 for (ss = avl_first(t); ss; ss = AVL_NEXT(t, ss)) { 471 /* 472 * Everything in this DTL must appear in all parent DTL unions. 473 */ 474 for (pvd = vd; pvd; pvd = pvd->vdev_parent) 475 ASSERT(vdev_dtl_contains(&pvd->vdev_dtl_map, 476 ss->ss_start, ss->ss_end - ss->ss_start)); 477 (void) printf("\t%*soutage [%llu,%llu] length %llu\n", 478 indent, "", 479 (u_longlong_t)ss->ss_start, 480 (u_longlong_t)ss->ss_end - 1, 481 (u_longlong_t)ss->ss_end - ss->ss_start); 482 } 483 484 (void) printf("\n"); 485 486 if (dump_opt['d'] > 5 && vd->vdev_children == 0) { 487 dump_spacemap(vd->vdev_spa->spa_meta_objset, &vd->vdev_dtl, 488 &vd->vdev_dtl_map); 489 (void) printf("\n"); 490 } 491 492 for (c = 0; c < vd->vdev_children; c++) 493 dump_dtl(vd->vdev_child[c], indent + 4); 494 } 495 496 /*ARGSUSED*/ 497 static void 498 dump_dnode(objset_t *os, uint64_t object, void *data, size_t size) 499 { 500 } 501 502 static uint64_t 503 blkid2offset(dnode_phys_t *dnp, int level, uint64_t blkid) 504 { 505 if (level < 0) 506 return (blkid); 507 508 return ((blkid << (level * (dnp->dn_indblkshift - SPA_BLKPTRSHIFT))) * 509 dnp->dn_datablkszsec << SPA_MINBLOCKSHIFT); 510 } 511 512 /* ARGSUSED */ 513 static int 514 zdb_indirect_cb(traverse_blk_cache_t *bc, spa_t *spa, void *a) 515 { 516 zbookmark_t *zb = &bc->bc_bookmark; 517 blkptr_t *bp = &bc->bc_blkptr; 518 dva_t *dva = &bp->blk_dva[0]; 519 void *data = bc->bc_data; 520 dnode_phys_t *dnp = bc->bc_dnode; 521 char buffer[300]; 522 int l; 523 524 if (bc->bc_errno) { 525 (void) sprintf(buffer, 526 "Error %d reading <%llu, %llu, %d, %llu>: ", 527 bc->bc_errno, 528 (u_longlong_t)zb->zb_objset, 529 (u_longlong_t)zb->zb_object, 530 zb->zb_level, 531 (u_longlong_t)zb->zb_blkid); 532 goto out; 533 } 534 535 if (zb->zb_level == -1) { 536 ASSERT3U(BP_GET_TYPE(bp), ==, DMU_OT_OBJSET); 537 ASSERT3U(BP_GET_LEVEL(bp), ==, 0); 538 } else { 539 ASSERT3U(BP_GET_TYPE(bp), ==, dnp->dn_type); 540 ASSERT3U(BP_GET_LEVEL(bp), ==, zb->zb_level); 541 } 542 543 if (zb->zb_level > 0) { 544 uint64_t fill = 0; 545 blkptr_t *bpx, *bpend; 546 547 for (bpx = data, bpend = bpx + BP_GET_LSIZE(bp) / sizeof (*bpx); 548 bpx < bpend; bpx++) { 549 if (bpx->blk_birth != 0) { 550 ASSERT(bpx->blk_fill > 0); 551 fill += bpx->blk_fill; 552 } else { 553 ASSERT(bpx->blk_fill == 0); 554 } 555 } 556 ASSERT3U(fill, ==, bp->blk_fill); 557 } 558 559 if (zb->zb_level == 0 && dnp->dn_type == DMU_OT_DNODE) { 560 uint64_t fill = 0; 561 dnode_phys_t *dnx, *dnend; 562 563 for (dnx = data, dnend = dnx + (BP_GET_LSIZE(bp)>>DNODE_SHIFT); 564 dnx < dnend; dnx++) { 565 if (dnx->dn_type != DMU_OT_NONE) 566 fill++; 567 } 568 ASSERT3U(fill, ==, bp->blk_fill); 569 } 570 571 (void) sprintf(buffer, "%16llx ", 572 (u_longlong_t)blkid2offset(dnp, zb->zb_level, zb->zb_blkid)); 573 574 ASSERT(zb->zb_level >= 0); 575 576 for (l = dnp->dn_nlevels - 1; l >= -1; l--) { 577 if (l == zb->zb_level) { 578 (void) sprintf(buffer + strlen(buffer), "L%x", 579 zb->zb_level); 580 } else { 581 (void) sprintf(buffer + strlen(buffer), " "); 582 } 583 } 584 585 out: 586 if (bp->blk_birth == 0) { 587 (void) sprintf(buffer + strlen(buffer), "<hole>"); 588 (void) printf("%s\n", buffer); 589 } else { 590 // XXBP - Need to print number of active BPs here 591 (void) sprintf(buffer + strlen(buffer), 592 "vdev=%llu off=%llx %llxL/%llxP/%llxA F=%llu B=%llu", 593 (u_longlong_t)DVA_GET_VDEV(dva), 594 (u_longlong_t)DVA_GET_OFFSET(dva), 595 (u_longlong_t)BP_GET_LSIZE(bp), 596 (u_longlong_t)BP_GET_PSIZE(bp), 597 (u_longlong_t)DVA_GET_ASIZE(dva), 598 (u_longlong_t)bp->blk_fill, 599 (u_longlong_t)bp->blk_birth); 600 601 (void) printf("%s\n", buffer); 602 } 603 604 return (bc->bc_errno ? ERESTART : 0); 605 } 606 607 /*ARGSUSED*/ 608 static void 609 dump_indirect(objset_t *os, uint64_t object, void *data, size_t size) 610 { 611 traverse_handle_t *th; 612 uint64_t objset = dmu_objset_id(os); 613 int advance = zdb_advance; 614 615 (void) printf("Indirect blocks:\n"); 616 617 if (object == 0) 618 advance |= ADVANCE_DATA; 619 620 th = traverse_init(dmu_objset_spa(os), zdb_indirect_cb, NULL, advance, 621 ZIO_FLAG_CANFAIL); 622 th->th_noread = zdb_noread; 623 624 traverse_add_dnode(th, 0, -1ULL, objset, object); 625 626 while (traverse_more(th) == EAGAIN) 627 continue; 628 629 (void) printf("\n"); 630 631 traverse_fini(th); 632 } 633 634 /*ARGSUSED*/ 635 static void 636 dump_dsl_dir(objset_t *os, uint64_t object, void *data, size_t size) 637 { 638 dsl_dir_phys_t *dd = data; 639 time_t crtime; 640 char used[6], compressed[6], uncompressed[6], quota[6], resv[6]; 641 642 if (dd == NULL) 643 return; 644 645 ASSERT(size == sizeof (*dd)); 646 647 crtime = dd->dd_creation_time; 648 nicenum(dd->dd_used_bytes, used); 649 nicenum(dd->dd_compressed_bytes, compressed); 650 nicenum(dd->dd_uncompressed_bytes, uncompressed); 651 nicenum(dd->dd_quota, quota); 652 nicenum(dd->dd_reserved, resv); 653 654 (void) printf("\t\tcreation_time = %s", ctime(&crtime)); 655 (void) printf("\t\thead_dataset_obj = %llu\n", 656 (u_longlong_t)dd->dd_head_dataset_obj); 657 (void) printf("\t\tparent_dir_obj = %llu\n", 658 (u_longlong_t)dd->dd_parent_obj); 659 (void) printf("\t\tclone_parent_obj = %llu\n", 660 (u_longlong_t)dd->dd_clone_parent_obj); 661 (void) printf("\t\tchild_dir_zapobj = %llu\n", 662 (u_longlong_t)dd->dd_child_dir_zapobj); 663 (void) printf("\t\tused_bytes = %s\n", used); 664 (void) printf("\t\tcompressed_bytes = %s\n", compressed); 665 (void) printf("\t\tuncompressed_bytes = %s\n", uncompressed); 666 (void) printf("\t\tquota = %s\n", quota); 667 (void) printf("\t\treserved = %s\n", resv); 668 (void) printf("\t\tprops_zapobj = %llu\n", 669 (u_longlong_t)dd->dd_props_zapobj); 670 } 671 672 /*ARGSUSED*/ 673 static void 674 dump_dsl_dataset(objset_t *os, uint64_t object, void *data, size_t size) 675 { 676 dsl_dataset_phys_t *ds = data; 677 time_t crtime; 678 char used[6], compressed[6], uncompressed[6], unique[6], blkbuf[300]; 679 680 if (ds == NULL) 681 return; 682 683 ASSERT(size == sizeof (*ds)); 684 crtime = ds->ds_creation_time; 685 nicenum(ds->ds_used_bytes, used); 686 nicenum(ds->ds_compressed_bytes, compressed); 687 nicenum(ds->ds_uncompressed_bytes, uncompressed); 688 nicenum(ds->ds_unique_bytes, unique); 689 sprintf_blkptr(blkbuf, &ds->ds_bp); 690 691 (void) printf("\t\tdataset_obj = %llu\n", 692 (u_longlong_t)ds->ds_dir_obj); 693 (void) printf("\t\tprev_snap_obj = %llu\n", 694 (u_longlong_t)ds->ds_prev_snap_obj); 695 (void) printf("\t\tprev_snap_txg = %llu\n", 696 (u_longlong_t)ds->ds_prev_snap_txg); 697 (void) printf("\t\tnext_snap_obj = %llu\n", 698 (u_longlong_t)ds->ds_next_snap_obj); 699 (void) printf("\t\tsnapnames_zapobj = %llu\n", 700 (u_longlong_t)ds->ds_snapnames_zapobj); 701 (void) printf("\t\tnum_children = %llu\n", 702 (u_longlong_t)ds->ds_num_children); 703 (void) printf("\t\tcreation_time = %s", ctime(&crtime)); 704 (void) printf("\t\tcreation_txg = %llu\n", 705 (u_longlong_t)ds->ds_creation_txg); 706 (void) printf("\t\tdeadlist_obj = %llu\n", 707 (u_longlong_t)ds->ds_deadlist_obj); 708 (void) printf("\t\tused_bytes = %s\n", used); 709 (void) printf("\t\tcompressed_bytes = %s\n", compressed); 710 (void) printf("\t\tuncompressed_bytes = %s\n", uncompressed); 711 (void) printf("\t\tunique = %s\n", unique); 712 (void) printf("\t\tfsid_guid = %llu\n", 713 (u_longlong_t)ds->ds_fsid_guid); 714 (void) printf("\t\tguid = %llu\n", 715 (u_longlong_t)ds->ds_guid); 716 (void) printf("\t\trestoring = %llu\n", 717 (u_longlong_t)ds->ds_restoring); 718 (void) printf("\t\tbp = %s\n", blkbuf); 719 } 720 721 static void 722 dump_bplist(objset_t *mos, uint64_t object, char *name) 723 { 724 bplist_t bpl = { 0 }; 725 blkptr_t blk, *bp = &blk; 726 uint64_t itor = 0; 727 char numbuf[6]; 728 729 if (dump_opt['d'] < 3) 730 return; 731 732 bplist_open(&bpl, mos, object); 733 if (bplist_empty(&bpl)) { 734 bplist_close(&bpl); 735 return; 736 } 737 738 nicenum(bpl.bpl_phys->bpl_bytes, numbuf); 739 740 (void) printf("\n %s: %llu entries, %s\n", 741 name, (u_longlong_t)bpl.bpl_phys->bpl_entries, numbuf); 742 743 if (dump_opt['d'] < 5) { 744 bplist_close(&bpl); 745 return; 746 } 747 748 (void) printf("\n"); 749 750 while (bplist_iterate(&bpl, &itor, bp) == 0) { 751 ASSERT(bp->blk_birth != 0); 752 // XXBP - Do we want to see all DVAs, or just one? 753 (void) printf("\tItem %3llu: vdev=%llu off=%llx " 754 "%llxL/%llxP/%llxA F=%llu B=%llu\n", 755 (u_longlong_t)itor - 1, 756 (u_longlong_t)DVA_GET_VDEV(&bp->blk_dva[0]), 757 (u_longlong_t)DVA_GET_OFFSET(&bp->blk_dva[0]), 758 (u_longlong_t)BP_GET_LSIZE(bp), 759 (u_longlong_t)BP_GET_PSIZE(bp), 760 (u_longlong_t)DVA_GET_ASIZE(&bp->blk_dva[0]), 761 (u_longlong_t)bp->blk_fill, 762 (u_longlong_t)bp->blk_birth); 763 } 764 765 bplist_close(&bpl); 766 } 767 768 static char * 769 znode_path(objset_t *os, uint64_t object, char *pathbuf, size_t size) 770 { 771 dmu_buf_t *db; 772 dmu_object_info_t doi; 773 znode_phys_t *zp; 774 uint64_t parent = 0; 775 size_t complen; 776 char component[MAXNAMELEN + 1]; 777 char *path; 778 779 path = pathbuf + size; 780 *--path = '\0'; 781 782 for (;;) { 783 db = dmu_bonus_hold(os, object); 784 if (db == NULL) 785 break; 786 787 dmu_buf_read(db); 788 dmu_object_info_from_db(db, &doi); 789 zp = db->db_data; 790 parent = zp->zp_parent; 791 dmu_buf_rele(db); 792 793 if (doi.doi_bonus_type != DMU_OT_ZNODE) 794 break; 795 796 if (parent == object) { 797 if (path[0] != '/') 798 *--path = '/'; 799 return (path); 800 } 801 802 if (zap_value_search(os, parent, object, component) != 0) 803 break; 804 805 complen = strlen(component); 806 path -= complen; 807 bcopy(component, path, complen); 808 *--path = '/'; 809 810 object = parent; 811 } 812 813 (void) sprintf(component, "???<object#%llu>", (u_longlong_t)object); 814 815 complen = strlen(component); 816 path -= complen; 817 bcopy(component, path, complen); 818 819 return (path); 820 } 821 822 /*ARGSUSED*/ 823 static void 824 dump_znode(objset_t *os, uint64_t object, void *data, size_t size) 825 { 826 znode_phys_t *zp = data; 827 time_t z_crtime, z_atime, z_mtime, z_ctime; 828 char path[MAXPATHLEN * 2]; /* allow for xattr and failure prefix */ 829 830 ASSERT(size >= sizeof (znode_phys_t)); 831 832 if (dump_opt['d'] < 3) { 833 (void) printf("\t%s\n", 834 znode_path(os, object, path, sizeof (path))); 835 return; 836 } 837 838 z_crtime = (time_t)zp->zp_crtime[0]; 839 z_atime = (time_t)zp->zp_atime[0]; 840 z_mtime = (time_t)zp->zp_mtime[0]; 841 z_ctime = (time_t)zp->zp_ctime[0]; 842 843 (void) printf("\tpath %s\n", 844 znode_path(os, object, path, sizeof (path))); 845 (void) printf("\tatime %s", ctime(&z_atime)); 846 (void) printf("\tmtime %s", ctime(&z_mtime)); 847 (void) printf("\tctime %s", ctime(&z_ctime)); 848 (void) printf("\tcrtime %s", ctime(&z_crtime)); 849 (void) printf("\tgen %llu\n", (u_longlong_t)zp->zp_gen); 850 (void) printf("\tmode %llo\n", (u_longlong_t)zp->zp_mode); 851 (void) printf("\tsize %llu\n", (u_longlong_t)zp->zp_size); 852 (void) printf("\tparent %llu\n", (u_longlong_t)zp->zp_parent); 853 (void) printf("\tlinks %llu\n", (u_longlong_t)zp->zp_links); 854 (void) printf("\txattr %llu\n", (u_longlong_t)zp->zp_xattr); 855 (void) printf("\trdev 0x%016llx\n", (u_longlong_t)zp->zp_rdev); 856 } 857 858 /*ARGSUSED*/ 859 static void 860 dump_acl(objset_t *os, uint64_t object, void *data, size_t size) 861 { 862 } 863 864 /*ARGSUSED*/ 865 static void 866 dump_dmu_objset(objset_t *os, uint64_t object, void *data, size_t size) 867 { 868 } 869 870 static object_viewer_t *object_viewer[DMU_OT_NUMTYPES] = { 871 dump_none, /* unallocated */ 872 dump_zap, /* object directory */ 873 dump_uint64, /* object array */ 874 dump_none, /* packed nvlist */ 875 dump_packed_nvlist, /* packed nvlist size */ 876 dump_none, /* bplist */ 877 dump_none, /* bplist header */ 878 dump_none, /* SPA space map header */ 879 dump_none, /* SPA space map */ 880 dump_none, /* ZIL intent log */ 881 dump_dnode, /* DMU dnode */ 882 dump_dmu_objset, /* DMU objset */ 883 dump_dsl_dir, /* DSL directory */ 884 dump_zap, /* DSL directory child map */ 885 dump_zap, /* DSL dataset snap map */ 886 dump_zap, /* DSL props */ 887 dump_dsl_dataset, /* DSL dataset */ 888 dump_znode, /* ZFS znode */ 889 dump_acl, /* ZFS ACL */ 890 dump_uint8, /* ZFS plain file */ 891 dump_zap, /* ZFS directory */ 892 dump_zap, /* ZFS master node */ 893 dump_zap, /* ZFS delete queue */ 894 dump_uint8, /* zvol object */ 895 dump_zap, /* zvol prop */ 896 dump_uint8, /* other uint8[] */ 897 dump_uint64, /* other uint64[] */ 898 dump_zap, /* other ZAP */ 899 }; 900 901 static void 902 dump_object(objset_t *os, uint64_t object, int verbosity, int *print_header) 903 { 904 dmu_buf_t *db = NULL; 905 dmu_object_info_t doi; 906 dnode_t *dn; 907 void *bonus = NULL; 908 size_t bsize = 0; 909 char iblk[6], dblk[6], lsize[6], psize[6], bonus_size[6], segsize[6]; 910 char aux[50]; 911 int error; 912 913 if (*print_header) { 914 (void) printf("\n Object lvl iblk dblk lsize" 915 " psize type\n"); 916 *print_header = 0; 917 } 918 919 if (object == 0) { 920 dn = os->os->os_meta_dnode; 921 } else { 922 db = dmu_bonus_hold(os, object); 923 if (db == NULL) 924 fatal("dmu_bonus_hold(%llu) failed", object); 925 dmu_buf_read(db); 926 bonus = db->db_data; 927 bsize = db->db_size; 928 dn = ((dmu_buf_impl_t *)db)->db_dnode; 929 } 930 dmu_object_info_from_dnode(dn, &doi); 931 932 nicenum(doi.doi_metadata_block_size, iblk); 933 nicenum(doi.doi_data_block_size, dblk); 934 nicenum(doi.doi_data_block_size * (doi.doi_max_block_offset + 1), 935 lsize); 936 nicenum(doi.doi_physical_blks << 9, psize); 937 nicenum(doi.doi_bonus_size, bonus_size); 938 939 aux[0] = '\0'; 940 941 if (doi.doi_checksum != ZIO_CHECKSUM_INHERIT || verbosity >= 6) 942 (void) snprintf(aux + strlen(aux), sizeof (aux), " (K=%s)", 943 zio_checksum_table[doi.doi_checksum].ci_name); 944 945 if (doi.doi_compress != ZIO_COMPRESS_INHERIT || verbosity >= 6) 946 (void) snprintf(aux + strlen(aux), sizeof (aux), " (Z=%s)", 947 zio_compress_table[doi.doi_compress].ci_name); 948 949 (void) printf("%10lld %3u %5s %5s %5s %5s %s%s\n", 950 (u_longlong_t)object, doi.doi_indirection, iblk, dblk, lsize, 951 psize, dmu_ot[doi.doi_type].ot_name, aux); 952 953 if (doi.doi_bonus_type != DMU_OT_NONE && verbosity > 3) { 954 (void) printf("%10s %3s %5s %5s %5s %5s %s\n", 955 "", "", "", "", bonus_size, "bonus", 956 dmu_ot[doi.doi_bonus_type].ot_name); 957 } 958 959 if (verbosity >= 4) { 960 object_viewer[doi.doi_bonus_type](os, object, bonus, bsize); 961 object_viewer[doi.doi_type](os, object, NULL, 0); 962 *print_header = 1; 963 } 964 965 if (verbosity >= 5) 966 dump_indirect(os, object, NULL, 0); 967 968 if (verbosity >= 5) { 969 /* 970 * Report the list of segments that comprise the object. 971 */ 972 uint64_t start = 0; 973 uint64_t end; 974 uint64_t blkfill = 1; 975 int minlvl = 1; 976 977 if (dn->dn_type == DMU_OT_DNODE) { 978 minlvl = 0; 979 blkfill = DNODES_PER_BLOCK; 980 } 981 982 for (;;) { 983 error = dnode_next_offset(dn, B_FALSE, &start, minlvl, 984 blkfill); 985 if (error) 986 break; 987 end = start; 988 error = dnode_next_offset(dn, B_TRUE, &end, minlvl, 989 blkfill); 990 nicenum(end - start, segsize); 991 (void) printf("\t\tsegment [%016llx, %016llx)" 992 " size %5s\n", (u_longlong_t)start, 993 (u_longlong_t)end, segsize); 994 if (error) 995 break; 996 start = end; 997 } 998 } 999 1000 if (db != NULL) 1001 dmu_buf_rele(db); 1002 } 1003 1004 static char *objset_types[DMU_OST_NUMTYPES] = { 1005 "NONE", "META", "ZPL", "ZVOL", "OTHER", "ANY" }; 1006 1007 /*ARGSUSED*/ 1008 static void 1009 dump_dir(objset_t *os) 1010 { 1011 dmu_objset_stats_t dds; 1012 uint64_t object, object_count; 1013 char numbuf[8]; 1014 char blkbuf[300]; 1015 char osname[MAXNAMELEN]; 1016 char *type = "UNKNOWN"; 1017 int verbosity = dump_opt['d']; 1018 int print_header = 1; 1019 int i, error; 1020 1021 dmu_objset_stats(os, &dds); 1022 1023 if (dds.dds_type < DMU_OST_NUMTYPES) 1024 type = objset_types[dds.dds_type]; 1025 1026 if (dds.dds_type == DMU_OST_META) { 1027 dds.dds_creation_txg = TXG_INITIAL; 1028 dds.dds_last_txg = os->os->os_rootbp.blk_birth; 1029 dds.dds_objects_used = os->os->os_rootbp.blk_fill; 1030 dds.dds_space_refd = 1031 os->os->os_spa->spa_dsl_pool->dp_mos_dir->dd_used_bytes; 1032 } 1033 1034 ASSERT3U(dds.dds_objects_used, ==, os->os->os_rootbp.blk_fill); 1035 1036 nicenum(dds.dds_space_refd, numbuf); 1037 1038 if (verbosity >= 4) { 1039 (void) strcpy(blkbuf, ", rootbp "); 1040 sprintf_blkptr(blkbuf + strlen(blkbuf), &os->os->os_rootbp); 1041 } else { 1042 blkbuf[0] = '\0'; 1043 } 1044 1045 dmu_objset_name(os, osname); 1046 1047 (void) printf("Dataset %s [%s], ID %llu, cr_txg %llu, last_txg %llu, " 1048 "%s, %llu objects%s\n", 1049 osname, type, (u_longlong_t)dmu_objset_id(os), 1050 (u_longlong_t)dds.dds_creation_txg, 1051 (u_longlong_t)dds.dds_last_txg, 1052 numbuf, 1053 (u_longlong_t)dds.dds_objects_used, 1054 blkbuf); 1055 1056 dump_intent_log(dmu_objset_zil(os)); 1057 1058 if (dmu_objset_ds(os) != NULL) 1059 dump_bplist(dmu_objset_pool(os)->dp_meta_objset, 1060 dmu_objset_ds(os)->ds_phys->ds_deadlist_obj, "Deadlist"); 1061 1062 if (verbosity < 2) 1063 return; 1064 1065 if (zopt_objects != 0) { 1066 for (i = 0; i < zopt_objects; i++) 1067 dump_object(os, zopt_object[i], verbosity, 1068 &print_header); 1069 (void) printf("\n"); 1070 return; 1071 } 1072 1073 dump_object(os, 0, verbosity, &print_header); 1074 object_count = 1; 1075 1076 object = 0; 1077 while ((error = dmu_object_next(os, &object, B_FALSE)) == 0) { 1078 dump_object(os, object, verbosity, &print_header); 1079 object_count++; 1080 } 1081 1082 ASSERT3U(object_count, ==, dds.dds_objects_used); 1083 1084 (void) printf("\n"); 1085 1086 if (error != ESRCH) 1087 fatal("dmu_object_next() = %d", error); 1088 } 1089 1090 static void 1091 dump_uberblock(uberblock_t *ub) 1092 { 1093 time_t timestamp = ub->ub_timestamp; 1094 1095 (void) printf("Uberblock\n\n"); 1096 (void) printf("\tmagic = %016llx\n", (u_longlong_t)ub->ub_magic); 1097 (void) printf("\tversion = %llu\n", (u_longlong_t)ub->ub_version); 1098 (void) printf("\ttxg = %llu\n", (u_longlong_t)ub->ub_txg); 1099 (void) printf("\tguid_sum = %llu\n", (u_longlong_t)ub->ub_guid_sum); 1100 (void) printf("\ttimestamp = %llu UTC = %s", 1101 (u_longlong_t)ub->ub_timestamp, asctime(localtime(×tamp))); 1102 if (dump_opt['u'] >= 3) { 1103 char blkbuf[300]; 1104 sprintf_blkptr(blkbuf, &ub->ub_rootbp); 1105 (void) printf("\trootbp = %s\n", blkbuf); 1106 } 1107 (void) printf("\n"); 1108 } 1109 1110 static void 1111 dump_config(const char *pool) 1112 { 1113 spa_t *spa = NULL; 1114 1115 mutex_enter(&spa_namespace_lock); 1116 while ((spa = spa_next(spa)) != NULL) { 1117 if (pool == NULL) 1118 (void) printf("%s\n", spa_name(spa)); 1119 if (pool == NULL || strcmp(pool, spa_name(spa)) == 0) 1120 dump_nvlist(spa->spa_config, 4); 1121 } 1122 mutex_exit(&spa_namespace_lock); 1123 } 1124 1125 static void 1126 dump_label(const char *dev) 1127 { 1128 int fd; 1129 vdev_label_t label; 1130 char *buf = label.vl_vdev_phys.vp_nvlist; 1131 size_t buflen = sizeof (label.vl_vdev_phys.vp_nvlist); 1132 struct stat64 statbuf; 1133 uint64_t psize; 1134 int l; 1135 1136 if ((fd = open(dev, O_RDONLY)) < 0) { 1137 (void) printf("cannot open '%s': %s\n", dev, strerror(errno)); 1138 exit(1); 1139 } 1140 1141 if (fstat64(fd, &statbuf) != 0) { 1142 (void) printf("failed to stat '%s': %s\n", dev, 1143 strerror(errno)); 1144 exit(1); 1145 } 1146 1147 psize = statbuf.st_size; 1148 psize = P2ALIGN(psize, (uint64_t)sizeof (vdev_label_t)); 1149 1150 for (l = 0; l < VDEV_LABELS; l++) { 1151 1152 nvlist_t *config = NULL; 1153 1154 (void) printf("--------------------------------------------\n"); 1155 (void) printf("LABEL %d\n", l); 1156 (void) printf("--------------------------------------------\n"); 1157 1158 if (pread(fd, &label, sizeof (label), 1159 vdev_label_offset(psize, l, 0)) != sizeof (label)) { 1160 (void) printf("failed to read label %d\n", l); 1161 continue; 1162 } 1163 1164 if (nvlist_unpack(buf, buflen, &config, 0) != 0) { 1165 (void) printf("failed to unpack label %d\n", l); 1166 continue; 1167 } 1168 dump_nvlist(config, 4); 1169 nvlist_free(config); 1170 } 1171 } 1172 1173 /*ARGSUSED*/ 1174 static void 1175 dump_one_dir(char *dsname, void *arg) 1176 { 1177 int error; 1178 objset_t *os; 1179 1180 error = dmu_objset_open(dsname, DMU_OST_ANY, 1181 DS_MODE_STANDARD | DS_MODE_READONLY, &os); 1182 if (error) { 1183 (void) printf("Could not open %s\n", dsname); 1184 return; 1185 } 1186 dump_dir(os); 1187 dmu_objset_close(os); 1188 } 1189 1190 static void 1191 zdb_space_map_load(spa_t *spa) 1192 { 1193 vdev_t *rvd = spa->spa_root_vdev; 1194 vdev_t *vd; 1195 int c, m, error; 1196 1197 for (c = 0; c < rvd->vdev_children; c++) { 1198 vd = rvd->vdev_child[c]; 1199 for (m = 0; m < vd->vdev_ms_count; m++) { 1200 metaslab_t *msp = vd->vdev_ms[m]; 1201 space_map_t *sm = &msp->ms_allocmap[0]; 1202 mutex_enter(&msp->ms_lock); 1203 error = space_map_load(sm, msp->ms_smo, SM_ALLOC, 1204 spa->spa_meta_objset, msp->ms_usable_end, 1205 sm->sm_size - msp->ms_usable_space); 1206 mutex_exit(&msp->ms_lock); 1207 if (error) 1208 fatal("%s bad space map #%d, error %d", 1209 spa->spa_name, c, error); 1210 } 1211 } 1212 } 1213 1214 static int 1215 zdb_space_map_claim(spa_t *spa, blkptr_t *bp) 1216 { 1217 dva_t *dva = &bp->blk_dva[0]; 1218 uint64_t vdev = DVA_GET_VDEV(dva); 1219 uint64_t offset = DVA_GET_OFFSET(dva); 1220 uint64_t size = DVA_GET_ASIZE(dva); 1221 vdev_t *vd; 1222 metaslab_t *msp; 1223 space_map_t *allocmap, *freemap; 1224 int error; 1225 1226 if ((vd = vdev_lookup_top(spa, vdev)) == NULL) 1227 return (ENXIO); 1228 1229 if ((offset >> vd->vdev_ms_shift) >= vd->vdev_ms_count) 1230 return (ENXIO); 1231 1232 if (DVA_GET_GANG(dva)) { 1233 zio_gbh_phys_t gbh; 1234 blkptr_t blk = *bp; 1235 int g; 1236 1237 /* LINTED - compile time assert */ 1238 ASSERT(sizeof (zio_gbh_phys_t) == SPA_GANGBLOCKSIZE); 1239 size = vdev_psize_to_asize(vd, SPA_GANGBLOCKSIZE); 1240 DVA_SET_GANG(&blk.blk_dva[0], 0); 1241 DVA_SET_ASIZE(&blk.blk_dva[0], size); 1242 BP_SET_CHECKSUM(&blk, ZIO_CHECKSUM_GANG_HEADER); 1243 BP_SET_PSIZE(&blk, SPA_GANGBLOCKSIZE); 1244 BP_SET_LSIZE(&blk, SPA_GANGBLOCKSIZE); 1245 BP_SET_COMPRESS(&blk, ZIO_COMPRESS_OFF); 1246 error = zio_wait(zio_read(NULL, spa, &blk, 1247 &gbh, SPA_GANGBLOCKSIZE, NULL, NULL, 1248 ZIO_PRIORITY_SYNC_READ, 1249 ZIO_FLAG_CANFAIL | ZIO_FLAG_CONFIG_HELD)); 1250 if (error) 1251 return (error); 1252 if (BP_SHOULD_BYTESWAP(&blk)) 1253 byteswap_uint64_array(&gbh, SPA_GANGBLOCKSIZE); 1254 for (g = 0; g < SPA_GBH_NBLKPTRS; g++) { 1255 if (gbh.zg_blkptr[g].blk_birth == 0) 1256 break; 1257 error = zdb_space_map_claim(spa, &gbh.zg_blkptr[g]); 1258 if (error) 1259 return (error); 1260 } 1261 } 1262 1263 msp = vd->vdev_ms[offset >> vd->vdev_ms_shift]; 1264 allocmap = &msp->ms_allocmap[0]; 1265 freemap = &msp->ms_freemap[0]; 1266 1267 mutex_enter(&msp->ms_lock); 1268 if (space_map_contains(freemap, offset, size)) { 1269 mutex_exit(&msp->ms_lock); 1270 return (EAGAIN); /* allocated more than once */ 1271 } 1272 1273 if (!space_map_contains(allocmap, offset, size)) { 1274 mutex_exit(&msp->ms_lock); 1275 return (ESTALE); /* not allocated at all */ 1276 } 1277 1278 space_map_remove(allocmap, offset, size); 1279 space_map_add(freemap, offset, size); 1280 1281 mutex_exit(&msp->ms_lock); 1282 1283 return (0); 1284 } 1285 1286 static void 1287 zdb_leak(space_map_t *sm, uint64_t start, uint64_t size) 1288 { 1289 metaslab_t *msp; 1290 1291 /* LINTED */ 1292 msp = (metaslab_t *)((char *)sm - offsetof(metaslab_t, ms_allocmap[0])); 1293 1294 (void) printf("leaked space: vdev %llu, offset 0x%llx, size %llu\n", 1295 (u_longlong_t)msp->ms_group->mg_vd->vdev_id, 1296 (u_longlong_t)start, 1297 (u_longlong_t)size); 1298 } 1299 1300 static void 1301 zdb_space_map_vacate(spa_t *spa) 1302 { 1303 vdev_t *rvd = spa->spa_root_vdev; 1304 vdev_t *vd; 1305 int c, m; 1306 1307 for (c = 0; c < rvd->vdev_children; c++) { 1308 vd = rvd->vdev_child[c]; 1309 for (m = 0; m < vd->vdev_ms_count; m++) { 1310 metaslab_t *msp = vd->vdev_ms[m]; 1311 mutex_enter(&msp->ms_lock); 1312 space_map_vacate(&msp->ms_allocmap[0], zdb_leak, 1313 &msp->ms_allocmap[0]); 1314 space_map_vacate(&msp->ms_freemap[0], NULL, NULL); 1315 mutex_exit(&msp->ms_lock); 1316 } 1317 } 1318 } 1319 1320 static void 1321 zdb_refresh_ubsync(spa_t *spa) 1322 { 1323 uberblock_t ub = { 0 }; 1324 vdev_t *rvd = spa->spa_root_vdev; 1325 zio_t *zio; 1326 1327 /* 1328 * Reopen all devices to purge zdb's vdev caches. 1329 */ 1330 vdev_reopen(rvd, NULL); 1331 1332 /* 1333 * Reload the uberblock. 1334 */ 1335 zio = zio_root(spa, NULL, NULL, 1336 ZIO_FLAG_CANFAIL | ZIO_FLAG_SPECULATIVE); 1337 vdev_uberblock_load(zio, rvd, &ub); 1338 (void) zio_wait(zio); 1339 1340 if (ub.ub_txg != 0) 1341 spa->spa_ubsync = ub; 1342 } 1343 1344 /* 1345 * Verify that the sum of the sizes of all blocks in the pool adds up 1346 * to the SPA's sa_alloc total. 1347 */ 1348 typedef struct zdb_blkstats { 1349 uint64_t zb_asize; 1350 uint64_t zb_lsize; 1351 uint64_t zb_psize; 1352 uint64_t zb_count; 1353 } zdb_blkstats_t; 1354 1355 #define DMU_OT_DEFERRED DMU_OT_NONE 1356 #define DMU_OT_TOTAL DMU_OT_NUMTYPES 1357 1358 #define ZB_TOTAL ZB_MAXLEVEL 1359 1360 typedef struct zdb_cb { 1361 zdb_blkstats_t zcb_type[ZB_TOTAL + 1][DMU_OT_TOTAL + 1]; 1362 uint64_t zcb_errors[256]; 1363 traverse_blk_cache_t *zcb_cache; 1364 int zcb_readfails; 1365 int zcb_haderrors; 1366 } zdb_cb_t; 1367 1368 static blkptr_cb_t zdb_blkptr_cb; 1369 1370 static void 1371 zdb_count_block(spa_t *spa, zdb_cb_t *zcb, blkptr_t *bp, int type) 1372 { 1373 int i, error; 1374 1375 for (i = 0; i < 4; i++) { 1376 int l = (i < 2) ? BP_GET_LEVEL(bp) : ZB_TOTAL; 1377 int t = (i & 1) ? type : DMU_OT_TOTAL; 1378 zdb_blkstats_t *zb = &zcb->zcb_type[l][t]; 1379 1380 zb->zb_asize += BP_GET_ASIZE(bp); 1381 zb->zb_lsize += BP_GET_LSIZE(bp); 1382 zb->zb_psize += BP_GET_PSIZE(bp); 1383 zb->zb_count++; 1384 } 1385 1386 if (dump_opt['L']) 1387 return; 1388 1389 error = zdb_space_map_claim(spa, bp); 1390 1391 if (error == 0) 1392 return; 1393 1394 if (error == EAGAIN) 1395 (void) fatal("double-allocation, bp=%p", bp); 1396 1397 if (error == ESTALE) 1398 (void) fatal("reference to freed block, bp=%p", bp); 1399 1400 (void) fatal("fatal error %d in bp %p", error, bp); 1401 } 1402 1403 static void 1404 zdb_log_block_cb(zilog_t *zilog, blkptr_t *bp, void *arg, uint64_t first_txg) 1405 { 1406 if (bp->blk_birth < first_txg) { 1407 zdb_cb_t *zcb = arg; 1408 traverse_blk_cache_t bc = *zcb->zcb_cache; 1409 zbookmark_t *zb = &bc.bc_bookmark; 1410 1411 zb->zb_objset = bp->blk_cksum.zc_word[2]; 1412 zb->zb_blkid = bp->blk_cksum.zc_word[3]; 1413 bc.bc_blkptr = *bp; 1414 1415 (void) zdb_blkptr_cb(&bc, zilog->zl_spa, arg); 1416 } 1417 } 1418 1419 static int 1420 zdb_blkptr_cb(traverse_blk_cache_t *bc, spa_t *spa, void *arg) 1421 { 1422 zbookmark_t *zb = &bc->bc_bookmark; 1423 zdb_cb_t *zcb = arg; 1424 blkptr_t *bp = &bc->bc_blkptr; 1425 dmu_object_type_t type = BP_GET_TYPE(bp); 1426 char blkbuf[300]; 1427 int error = 0; 1428 1429 if (bc->bc_errno) { 1430 if (zcb->zcb_readfails++ < 10 && dump_opt['L']) { 1431 zdb_refresh_ubsync(spa); 1432 error = EAGAIN; 1433 } else { 1434 zcb->zcb_haderrors = 1; 1435 zcb->zcb_errors[bc->bc_errno]++; 1436 error = ERESTART; 1437 } 1438 1439 if (dump_opt['b'] >= 3 || (dump_opt['b'] >= 2 && bc->bc_errno)) 1440 sprintf_blkptr(blkbuf, bp); 1441 else 1442 blkbuf[0] = '\0'; 1443 1444 (void) printf("zdb_blkptr_cb: Got error %d reading " 1445 "<%llu, %llu, %d, %llx> %s -- %s\n", 1446 bc->bc_errno, 1447 (u_longlong_t)zb->zb_objset, 1448 (u_longlong_t)zb->zb_object, 1449 zb->zb_level, 1450 (u_longlong_t)zb->zb_blkid, 1451 blkbuf, 1452 error == EAGAIN ? "retrying" : "skipping"); 1453 1454 return (error); 1455 } 1456 1457 zcb->zcb_readfails = 0; 1458 1459 ASSERT(bp->blk_birth != 0); 1460 1461 zdb_count_block(spa, zcb, bp, type); 1462 1463 if (dump_opt['b'] >= 4) { 1464 sprintf_blkptr(blkbuf, bp); 1465 (void) printf("objset %llu object %llu offset 0x%llx %s\n", 1466 (u_longlong_t)zb->zb_objset, 1467 (u_longlong_t)zb->zb_object, 1468 (u_longlong_t)blkid2offset(bc->bc_dnode, 1469 zb->zb_level, zb->zb_blkid), 1470 blkbuf); 1471 } 1472 1473 if (type == DMU_OT_OBJSET) { 1474 objset_phys_t *osphys = bc->bc_data; 1475 zilog_t zilog = { 0 }; 1476 zilog.zl_header = &osphys->os_zil_header; 1477 zilog.zl_spa = spa; 1478 1479 zcb->zcb_cache = bc; 1480 1481 zil_parse(&zilog, zdb_log_block_cb, NULL, zcb, 1482 spa_first_txg(spa)); 1483 } 1484 1485 return (0); 1486 } 1487 1488 static int 1489 dump_block_stats(spa_t *spa) 1490 { 1491 traverse_handle_t *th; 1492 zdb_cb_t zcb = { 0 }; 1493 zdb_blkstats_t *zb, *tzb; 1494 uint64_t alloc, space; 1495 int leaks = 0; 1496 int advance = zdb_advance; 1497 int flags; 1498 int e; 1499 1500 if (dump_opt['c']) 1501 advance |= ADVANCE_DATA; 1502 1503 advance |= ADVANCE_PRUNE; 1504 1505 (void) printf("\nTraversing all blocks to %sverify" 1506 " nothing leaked ...\n", 1507 dump_opt['c'] ? "verify checksums and " : ""); 1508 1509 /* 1510 * Load all space maps. As we traverse the pool, if we find a block 1511 * that's not in its space map, that indicates a double-allocation, 1512 * reference to a freed block, or an unclaimed block. Otherwise we 1513 * remove the block from the space map. If the space maps are not 1514 * empty when we're done, that indicates leaked blocks. 1515 */ 1516 if (!dump_opt['L']) 1517 zdb_space_map_load(spa); 1518 1519 /* 1520 * If there's a deferred-free bplist, process that first. 1521 */ 1522 if (spa->spa_sync_bplist_obj != 0) { 1523 bplist_t *bpl = &spa->spa_sync_bplist; 1524 blkptr_t blk; 1525 uint64_t itor = 0; 1526 1527 bplist_open(bpl, spa->spa_meta_objset, 1528 spa->spa_sync_bplist_obj); 1529 1530 while (bplist_iterate(bpl, &itor, &blk) == 0) { 1531 zdb_count_block(spa, &zcb, &blk, DMU_OT_DEFERRED); 1532 if (dump_opt['b'] >= 4) { 1533 char blkbuf[300]; 1534 sprintf_blkptr(blkbuf, &blk); 1535 (void) printf("[%s] %s\n", 1536 "deferred free", blkbuf); 1537 } 1538 } 1539 1540 bplist_close(bpl); 1541 } 1542 1543 /* 1544 * Now traverse the pool. If we're read all data to verify checksums, 1545 * do a scrubbing read so that we validate all copies. 1546 */ 1547 flags = ZIO_FLAG_CANFAIL; 1548 if (advance & ADVANCE_DATA) 1549 flags |= ZIO_FLAG_SCRUB; 1550 th = traverse_init(spa, zdb_blkptr_cb, &zcb, advance, flags); 1551 th->th_noread = zdb_noread; 1552 1553 traverse_add_pool(th, 0, -1ULL); 1554 1555 while (traverse_more(th) == EAGAIN) 1556 continue; 1557 1558 traverse_fini(th); 1559 1560 if (zcb.zcb_haderrors) { 1561 (void) printf("\nError counts:\n\n"); 1562 (void) printf("\t%5s %s\n", "errno", "count"); 1563 for (e = 0; e < 256; e++) { 1564 if (zcb.zcb_errors[e] != 0) { 1565 (void) printf("\t%5d %llu\n", 1566 e, (u_longlong_t)zcb.zcb_errors[e]); 1567 } 1568 } 1569 } 1570 1571 /* 1572 * Report any leaked segments. 1573 */ 1574 if (!dump_opt['L']) 1575 zdb_space_map_vacate(spa); 1576 1577 if (dump_opt['L']) 1578 (void) printf("\n\n *** Live pool traversal; " 1579 "block counts are only approximate ***\n\n"); 1580 1581 alloc = spa_get_alloc(spa); 1582 space = spa_get_space(spa); 1583 1584 tzb = &zcb.zcb_type[ZB_TOTAL][DMU_OT_TOTAL]; 1585 1586 if (tzb->zb_asize == alloc) { 1587 (void) printf("\n\tNo leaks (block sum matches space" 1588 " maps exactly)\n"); 1589 } else { 1590 (void) printf("block traversal size %llu != alloc %llu " 1591 "(leaked %lld)\n", 1592 (u_longlong_t)tzb->zb_asize, 1593 (u_longlong_t)alloc, 1594 (u_longlong_t)(alloc - tzb->zb_asize)); 1595 leaks = 1; 1596 } 1597 1598 if (tzb->zb_count == 0) 1599 return (2); 1600 1601 (void) printf("\n"); 1602 (void) printf("\tbp count: %10llu\n", 1603 (u_longlong_t)tzb->zb_count); 1604 (void) printf("\tbp logical: %10llu\t avg: %6llu\n", 1605 (u_longlong_t)tzb->zb_lsize, 1606 (u_longlong_t)(tzb->zb_lsize / tzb->zb_count)); 1607 (void) printf("\tbp physical: %10llu\t avg:" 1608 " %6llu\tcompression: %6.2f\n", 1609 (u_longlong_t)tzb->zb_psize, 1610 (u_longlong_t)(tzb->zb_psize / tzb->zb_count), 1611 (double)tzb->zb_lsize / tzb->zb_psize); 1612 (void) printf("\tbp allocated: %10llu\t avg:" 1613 " %6llu\tcompression: %6.2f\n", 1614 (u_longlong_t)tzb->zb_asize, 1615 (u_longlong_t)(tzb->zb_asize / tzb->zb_count), 1616 (double)tzb->zb_lsize / tzb->zb_asize); 1617 (void) printf("\tSPA allocated: %10llu\tused: %5.2f%%\n", 1618 (u_longlong_t)alloc, 100.0 * alloc / space); 1619 1620 if (dump_opt['b'] >= 2) { 1621 int l, t, level; 1622 (void) printf("\nBlocks\tLSIZE\tPSIZE\tASIZE" 1623 "\t avg\t comp\t%%Total\tType\n"); 1624 1625 for (t = 0; t <= DMU_OT_NUMTYPES; t++) { 1626 char csize[6], lsize[6], psize[6], asize[6], avg[6]; 1627 char *typename; 1628 1629 typename = t == DMU_OT_DEFERRED ? "deferred free" : 1630 t == DMU_OT_TOTAL ? "Total" : dmu_ot[t].ot_name; 1631 1632 if (zcb.zcb_type[ZB_TOTAL][t].zb_asize == 0) { 1633 (void) printf("%6s\t%5s\t%5s\t%5s" 1634 "\t%5s\t%5s\t%6s\t%s\n", 1635 "-", 1636 "-", 1637 "-", 1638 "-", 1639 "-", 1640 "-", 1641 "-", 1642 typename); 1643 continue; 1644 } 1645 1646 for (l = ZB_TOTAL - 1; l >= -1; l--) { 1647 level = (l == -1 ? ZB_TOTAL : l); 1648 zb = &zcb.zcb_type[level][t]; 1649 1650 if (zb->zb_asize == 0) 1651 continue; 1652 1653 if (dump_opt['b'] < 3 && level != ZB_TOTAL) 1654 continue; 1655 1656 if (level == 0 && zb->zb_asize == 1657 zcb.zcb_type[ZB_TOTAL][t].zb_asize) 1658 continue; 1659 1660 nicenum(zb->zb_count, csize); 1661 nicenum(zb->zb_lsize, lsize); 1662 nicenum(zb->zb_psize, psize); 1663 nicenum(zb->zb_asize, asize); 1664 nicenum(zb->zb_asize / zb->zb_count, avg); 1665 1666 (void) printf("%6s\t%5s\t%5s\t%5s\t%5s" 1667 "\t%5.2f\t%6.2f\t", 1668 csize, lsize, psize, asize, avg, 1669 (double)zb->zb_lsize / zb->zb_psize, 1670 100.0 * zb->zb_asize / tzb->zb_asize); 1671 1672 if (level == ZB_TOTAL) 1673 (void) printf("%s\n", typename); 1674 else 1675 (void) printf(" L%d %s\n", 1676 level, typename); 1677 } 1678 } 1679 } 1680 1681 (void) printf("\n"); 1682 1683 if (leaks) 1684 return (2); 1685 1686 if (zcb.zcb_haderrors) 1687 return (3); 1688 1689 return (0); 1690 } 1691 1692 static void 1693 dump_zpool(spa_t *spa) 1694 { 1695 dsl_pool_t *dp = spa_get_dsl(spa); 1696 int rc = 0; 1697 1698 if (dump_opt['u']) 1699 dump_uberblock(&spa->spa_uberblock); 1700 1701 if (dump_opt['d'] || dump_opt['i']) { 1702 dump_dir(dp->dp_meta_objset); 1703 if (dump_opt['d'] >= 3) { 1704 dump_bplist(dp->dp_meta_objset, 1705 spa->spa_sync_bplist_obj, "Deferred frees"); 1706 dump_dtl(spa->spa_root_vdev, 0); 1707 dump_metaslabs(spa); 1708 } 1709 dmu_objset_find(spa->spa_name, dump_one_dir, NULL, 1710 DS_FIND_SNAPSHOTS); 1711 } 1712 1713 if (dump_opt['b'] || dump_opt['c']) 1714 rc = dump_block_stats(spa); 1715 1716 if (dump_opt['s']) 1717 show_pool_stats(spa); 1718 1719 if (rc != 0) 1720 exit(rc); 1721 } 1722 1723 int 1724 main(int argc, char **argv) 1725 { 1726 int i, c; 1727 struct rlimit rl = { 1024, 1024 }; 1728 spa_t *spa; 1729 objset_t *os = NULL; 1730 char *endstr; 1731 int dump_all = 1; 1732 int verbose = 0; 1733 int error; 1734 int flag, set; 1735 1736 (void) setrlimit(RLIMIT_NOFILE, &rl); 1737 1738 dprintf_setup(&argc, argv); 1739 1740 while ((c = getopt(argc, argv, "udibcsvCLO:B:Ul")) != -1) { 1741 switch (c) { 1742 case 'u': 1743 case 'd': 1744 case 'i': 1745 case 'b': 1746 case 'c': 1747 case 's': 1748 case 'C': 1749 case 'l': 1750 dump_opt[c]++; 1751 dump_all = 0; 1752 break; 1753 case 'L': 1754 dump_opt[c]++; 1755 break; 1756 case 'O': 1757 endstr = optarg; 1758 if (endstr[0] == '!') { 1759 endstr++; 1760 set = 0; 1761 } else { 1762 set = 1; 1763 } 1764 if (strcmp(endstr, "post") == 0) { 1765 flag = ADVANCE_PRE; 1766 set = !set; 1767 } else if (strcmp(endstr, "pre") == 0) { 1768 flag = ADVANCE_PRE; 1769 } else if (strcmp(endstr, "prune") == 0) { 1770 flag = ADVANCE_PRUNE; 1771 } else if (strcmp(endstr, "data") == 0) { 1772 flag = ADVANCE_DATA; 1773 } else if (strcmp(endstr, "holes") == 0) { 1774 flag = ADVANCE_HOLES; 1775 } else { 1776 usage(); 1777 } 1778 if (set) 1779 zdb_advance |= flag; 1780 else 1781 zdb_advance &= ~flag; 1782 break; 1783 case 'B': 1784 endstr = optarg - 1; 1785 zdb_noread.zb_objset = strtoull(endstr + 1, &endstr, 0); 1786 zdb_noread.zb_object = strtoull(endstr + 1, &endstr, 0); 1787 zdb_noread.zb_level = strtol(endstr + 1, &endstr, 0); 1788 zdb_noread.zb_blkid = strtoull(endstr + 1, &endstr, 16); 1789 (void) printf("simulating bad block " 1790 "<%llu, %llu, %d, %llx>\n", 1791 (u_longlong_t)zdb_noread.zb_objset, 1792 (u_longlong_t)zdb_noread.zb_object, 1793 zdb_noread.zb_level, 1794 (u_longlong_t)zdb_noread.zb_blkid); 1795 break; 1796 case 'v': 1797 verbose++; 1798 break; 1799 case 'U': 1800 spa_config_dir = "/tmp"; 1801 break; 1802 default: 1803 usage(); 1804 break; 1805 } 1806 } 1807 1808 kernel_init(FREAD); 1809 1810 for (c = 0; c < 256; c++) { 1811 if (dump_all && c != 'L' && c != 'l') 1812 dump_opt[c] = 1; 1813 if (dump_opt[c]) 1814 dump_opt[c] += verbose; 1815 } 1816 1817 argc -= optind; 1818 argv += optind; 1819 1820 if (argc < 1) { 1821 if (dump_opt['C']) { 1822 dump_config(NULL); 1823 return (0); 1824 } 1825 usage(); 1826 } 1827 1828 if (dump_opt['l']) { 1829 dump_label(argv[0]); 1830 return (0); 1831 } 1832 1833 if (dump_opt['C']) 1834 dump_config(argv[0]); 1835 1836 if (strchr(argv[0], '/') != NULL) { 1837 error = dmu_objset_open(argv[0], DMU_OST_ANY, 1838 DS_MODE_STANDARD | DS_MODE_READONLY, &os); 1839 } else { 1840 error = spa_open(argv[0], &spa, FTAG); 1841 } 1842 1843 if (error) 1844 fatal("can't open %s: error %d", argv[0], error); 1845 1846 argv++; 1847 if (--argc > 0) { 1848 zopt_objects = argc; 1849 zopt_object = calloc(zopt_objects, sizeof (uint64_t)); 1850 for (i = 0; i < zopt_objects; i++) { 1851 errno = 0; 1852 zopt_object[i] = strtoull(argv[i], NULL, 0); 1853 if (zopt_object[i] == 0 && errno != 0) 1854 fatal("bad object number %s: %s", 1855 argv[i], strerror(errno)); 1856 } 1857 } 1858 1859 if (os != NULL) { 1860 dump_dir(os); 1861 dmu_objset_close(os); 1862 } else { 1863 dump_zpool(spa); 1864 spa_close(spa, FTAG); 1865 } 1866 1867 kernel_fini(); 1868 1869 return (0); 1870 } 1871