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 2009 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] [-t txg]\n" 91 "\t [-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, " -t <txg> highest txg to use when " 121 "searching for uberblocks\n"); 122 (void) fprintf(stderr, "Specify an option more than once (e.g. -bb) " 123 "to make only that option verbose\n"); 124 (void) fprintf(stderr, "Default is to dump everything non-verbosely\n"); 125 exit(1); 126 } 127 128 /* 129 * Called for usage errors that are discovered after a call to spa_open(), 130 * dmu_bonus_hold(), or pool_match(). abort() is called for other errors. 131 */ 132 133 static void 134 fatal(const char *fmt, ...) 135 { 136 va_list ap; 137 138 va_start(ap, fmt); 139 (void) fprintf(stderr, "%s: ", cmdname); 140 (void) vfprintf(stderr, fmt, ap); 141 va_end(ap); 142 (void) fprintf(stderr, "\n"); 143 144 exit(1); 145 } 146 147 static void 148 dump_nvlist(nvlist_t *list, int indent) 149 { 150 nvpair_t *elem = NULL; 151 152 while ((elem = nvlist_next_nvpair(list, elem)) != NULL) { 153 switch (nvpair_type(elem)) { 154 case DATA_TYPE_STRING: 155 { 156 char *value; 157 158 VERIFY(nvpair_value_string(elem, &value) == 0); 159 (void) printf("%*s%s='%s'\n", indent, "", 160 nvpair_name(elem), value); 161 } 162 break; 163 164 case DATA_TYPE_UINT64: 165 { 166 uint64_t value; 167 168 VERIFY(nvpair_value_uint64(elem, &value) == 0); 169 (void) printf("%*s%s=%llu\n", indent, "", 170 nvpair_name(elem), (u_longlong_t)value); 171 } 172 break; 173 174 case DATA_TYPE_NVLIST: 175 { 176 nvlist_t *value; 177 178 VERIFY(nvpair_value_nvlist(elem, &value) == 0); 179 (void) printf("%*s%s\n", indent, "", 180 nvpair_name(elem)); 181 dump_nvlist(value, indent + 4); 182 } 183 break; 184 185 case DATA_TYPE_NVLIST_ARRAY: 186 { 187 nvlist_t **value; 188 uint_t c, count; 189 190 VERIFY(nvpair_value_nvlist_array(elem, &value, 191 &count) == 0); 192 193 for (c = 0; c < count; c++) { 194 (void) printf("%*s%s[%u]\n", indent, "", 195 nvpair_name(elem), c); 196 dump_nvlist(value[c], indent + 8); 197 } 198 } 199 break; 200 201 default: 202 203 (void) printf("bad config type %d for %s\n", 204 nvpair_type(elem), nvpair_name(elem)); 205 } 206 } 207 } 208 209 /* ARGSUSED */ 210 static void 211 dump_packed_nvlist(objset_t *os, uint64_t object, void *data, size_t size) 212 { 213 nvlist_t *nv; 214 size_t nvsize = *(uint64_t *)data; 215 char *packed = umem_alloc(nvsize, UMEM_NOFAIL); 216 217 VERIFY(0 == dmu_read(os, object, 0, nvsize, packed)); 218 219 VERIFY(nvlist_unpack(packed, nvsize, &nv, 0) == 0); 220 221 umem_free(packed, nvsize); 222 223 dump_nvlist(nv, 8); 224 225 nvlist_free(nv); 226 } 227 228 const char dump_zap_stars[] = "****************************************"; 229 const int dump_zap_width = sizeof (dump_zap_stars) - 1; 230 231 static void 232 dump_zap_histogram(uint64_t histo[ZAP_HISTOGRAM_SIZE]) 233 { 234 int i; 235 int minidx = ZAP_HISTOGRAM_SIZE - 1; 236 int maxidx = 0; 237 uint64_t max = 0; 238 239 for (i = 0; i < ZAP_HISTOGRAM_SIZE; i++) { 240 if (histo[i] > max) 241 max = histo[i]; 242 if (histo[i] > 0 && i > maxidx) 243 maxidx = i; 244 if (histo[i] > 0 && i < minidx) 245 minidx = i; 246 } 247 248 if (max < dump_zap_width) 249 max = dump_zap_width; 250 251 for (i = minidx; i <= maxidx; i++) 252 (void) printf("\t\t\t%u: %6llu %s\n", i, (u_longlong_t)histo[i], 253 &dump_zap_stars[(max - histo[i]) * dump_zap_width / max]); 254 } 255 256 static void 257 dump_zap_stats(objset_t *os, uint64_t object) 258 { 259 int error; 260 zap_stats_t zs; 261 262 error = zap_get_stats(os, object, &zs); 263 if (error) 264 return; 265 266 if (zs.zs_ptrtbl_len == 0) { 267 ASSERT(zs.zs_num_blocks == 1); 268 (void) printf("\tmicrozap: %llu bytes, %llu entries\n", 269 (u_longlong_t)zs.zs_blocksize, 270 (u_longlong_t)zs.zs_num_entries); 271 return; 272 } 273 274 (void) printf("\tFat ZAP stats:\n"); 275 276 (void) printf("\t\tPointer table:\n"); 277 (void) printf("\t\t\t%llu elements\n", 278 (u_longlong_t)zs.zs_ptrtbl_len); 279 (void) printf("\t\t\tzt_blk: %llu\n", 280 (u_longlong_t)zs.zs_ptrtbl_zt_blk); 281 (void) printf("\t\t\tzt_numblks: %llu\n", 282 (u_longlong_t)zs.zs_ptrtbl_zt_numblks); 283 (void) printf("\t\t\tzt_shift: %llu\n", 284 (u_longlong_t)zs.zs_ptrtbl_zt_shift); 285 (void) printf("\t\t\tzt_blks_copied: %llu\n", 286 (u_longlong_t)zs.zs_ptrtbl_blks_copied); 287 (void) printf("\t\t\tzt_nextblk: %llu\n", 288 (u_longlong_t)zs.zs_ptrtbl_nextblk); 289 290 (void) printf("\t\tZAP entries: %llu\n", 291 (u_longlong_t)zs.zs_num_entries); 292 (void) printf("\t\tLeaf blocks: %llu\n", 293 (u_longlong_t)zs.zs_num_leafs); 294 (void) printf("\t\tTotal blocks: %llu\n", 295 (u_longlong_t)zs.zs_num_blocks); 296 (void) printf("\t\tzap_block_type: 0x%llx\n", 297 (u_longlong_t)zs.zs_block_type); 298 (void) printf("\t\tzap_magic: 0x%llx\n", 299 (u_longlong_t)zs.zs_magic); 300 (void) printf("\t\tzap_salt: 0x%llx\n", 301 (u_longlong_t)zs.zs_salt); 302 303 (void) printf("\t\tLeafs with 2^n pointers:\n"); 304 dump_zap_histogram(zs.zs_leafs_with_2n_pointers); 305 306 (void) printf("\t\tBlocks with n*5 entries:\n"); 307 dump_zap_histogram(zs.zs_blocks_with_n5_entries); 308 309 (void) printf("\t\tBlocks n/10 full:\n"); 310 dump_zap_histogram(zs.zs_blocks_n_tenths_full); 311 312 (void) printf("\t\tEntries with n chunks:\n"); 313 dump_zap_histogram(zs.zs_entries_using_n_chunks); 314 315 (void) printf("\t\tBuckets with n entries:\n"); 316 dump_zap_histogram(zs.zs_buckets_with_n_entries); 317 } 318 319 /*ARGSUSED*/ 320 static void 321 dump_none(objset_t *os, uint64_t object, void *data, size_t size) 322 { 323 } 324 325 /*ARGSUSED*/ 326 void 327 dump_uint8(objset_t *os, uint64_t object, void *data, size_t size) 328 { 329 } 330 331 /*ARGSUSED*/ 332 static void 333 dump_uint64(objset_t *os, uint64_t object, void *data, size_t size) 334 { 335 } 336 337 /*ARGSUSED*/ 338 static void 339 dump_zap(objset_t *os, uint64_t object, void *data, size_t size) 340 { 341 zap_cursor_t zc; 342 zap_attribute_t attr; 343 void *prop; 344 int i; 345 346 dump_zap_stats(os, object); 347 (void) printf("\n"); 348 349 for (zap_cursor_init(&zc, os, object); 350 zap_cursor_retrieve(&zc, &attr) == 0; 351 zap_cursor_advance(&zc)) { 352 (void) printf("\t\t%s = ", attr.za_name); 353 if (attr.za_num_integers == 0) { 354 (void) printf("\n"); 355 continue; 356 } 357 prop = umem_zalloc(attr.za_num_integers * 358 attr.za_integer_length, UMEM_NOFAIL); 359 (void) zap_lookup(os, object, attr.za_name, 360 attr.za_integer_length, attr.za_num_integers, prop); 361 if (attr.za_integer_length == 1) { 362 (void) printf("%s", (char *)prop); 363 } else { 364 for (i = 0; i < attr.za_num_integers; i++) { 365 switch (attr.za_integer_length) { 366 case 2: 367 (void) printf("%u ", 368 ((uint16_t *)prop)[i]); 369 break; 370 case 4: 371 (void) printf("%u ", 372 ((uint32_t *)prop)[i]); 373 break; 374 case 8: 375 (void) printf("%lld ", 376 (u_longlong_t)((int64_t *)prop)[i]); 377 break; 378 } 379 } 380 } 381 (void) printf("\n"); 382 umem_free(prop, attr.za_num_integers * attr.za_integer_length); 383 } 384 zap_cursor_fini(&zc); 385 } 386 387 /*ARGSUSED*/ 388 static void 389 dump_zpldir(objset_t *os, uint64_t object, void *data, size_t size) 390 { 391 zap_cursor_t zc; 392 zap_attribute_t attr; 393 const char *typenames[] = { 394 /* 0 */ "not specified", 395 /* 1 */ "FIFO", 396 /* 2 */ "Character Device", 397 /* 3 */ "3 (invalid)", 398 /* 4 */ "Directory", 399 /* 5 */ "5 (invalid)", 400 /* 6 */ "Block Device", 401 /* 7 */ "7 (invalid)", 402 /* 8 */ "Regular File", 403 /* 9 */ "9 (invalid)", 404 /* 10 */ "Symbolic Link", 405 /* 11 */ "11 (invalid)", 406 /* 12 */ "Socket", 407 /* 13 */ "Door", 408 /* 14 */ "Event Port", 409 /* 15 */ "15 (invalid)", 410 }; 411 412 dump_zap_stats(os, object); 413 (void) printf("\n"); 414 415 for (zap_cursor_init(&zc, os, object); 416 zap_cursor_retrieve(&zc, &attr) == 0; 417 zap_cursor_advance(&zc)) { 418 (void) printf("\t\t%s = %lld (type: %s)\n", 419 attr.za_name, ZFS_DIRENT_OBJ(attr.za_first_integer), 420 typenames[ZFS_DIRENT_TYPE(attr.za_first_integer)]); 421 } 422 zap_cursor_fini(&zc); 423 } 424 425 static void 426 dump_spacemap(objset_t *os, space_map_obj_t *smo, space_map_t *sm) 427 { 428 uint64_t alloc, offset, entry; 429 uint8_t mapshift = sm->sm_shift; 430 uint64_t mapstart = sm->sm_start; 431 char *ddata[] = { "ALLOC", "FREE", "CONDENSE", "INVALID", 432 "INVALID", "INVALID", "INVALID", "INVALID" }; 433 434 if (smo->smo_object == 0) 435 return; 436 437 /* 438 * Print out the freelist entries in both encoded and decoded form. 439 */ 440 alloc = 0; 441 for (offset = 0; offset < smo->smo_objsize; offset += sizeof (entry)) { 442 VERIFY(0 == dmu_read(os, smo->smo_object, offset, 443 sizeof (entry), &entry)); 444 if (SM_DEBUG_DECODE(entry)) { 445 (void) printf("\t\t[%4llu] %s: txg %llu, pass %llu\n", 446 (u_longlong_t)(offset / sizeof (entry)), 447 ddata[SM_DEBUG_ACTION_DECODE(entry)], 448 (u_longlong_t)SM_DEBUG_TXG_DECODE(entry), 449 (u_longlong_t)SM_DEBUG_SYNCPASS_DECODE(entry)); 450 } else { 451 (void) printf("\t\t[%4llu] %c range:" 452 " %08llx-%08llx size: %06llx\n", 453 (u_longlong_t)(offset / sizeof (entry)), 454 SM_TYPE_DECODE(entry) == SM_ALLOC ? 'A' : 'F', 455 (u_longlong_t)((SM_OFFSET_DECODE(entry) << 456 mapshift) + mapstart), 457 (u_longlong_t)((SM_OFFSET_DECODE(entry) << 458 mapshift) + mapstart + (SM_RUN_DECODE(entry) << 459 mapshift)), 460 (u_longlong_t)(SM_RUN_DECODE(entry) << mapshift)); 461 if (SM_TYPE_DECODE(entry) == SM_ALLOC) 462 alloc += SM_RUN_DECODE(entry) << mapshift; 463 else 464 alloc -= SM_RUN_DECODE(entry) << mapshift; 465 } 466 } 467 if (alloc != smo->smo_alloc) { 468 (void) printf("space_map_object alloc (%llu) INCONSISTENT " 469 "with space map summary (%llu)\n", 470 (u_longlong_t)smo->smo_alloc, (u_longlong_t)alloc); 471 } 472 } 473 474 static void 475 dump_metaslab(metaslab_t *msp) 476 { 477 char freebuf[5]; 478 space_map_obj_t *smo = &msp->ms_smo; 479 vdev_t *vd = msp->ms_group->mg_vd; 480 spa_t *spa = vd->vdev_spa; 481 482 nicenum(msp->ms_map.sm_size - smo->smo_alloc, freebuf); 483 484 if (dump_opt['d'] <= 5) { 485 (void) printf("\t%10llx %10llu %5s\n", 486 (u_longlong_t)msp->ms_map.sm_start, 487 (u_longlong_t)smo->smo_object, 488 freebuf); 489 return; 490 } 491 492 (void) printf( 493 "\tvdev %llu offset %08llx spacemap %4llu free %5s\n", 494 (u_longlong_t)vd->vdev_id, (u_longlong_t)msp->ms_map.sm_start, 495 (u_longlong_t)smo->smo_object, freebuf); 496 497 ASSERT(msp->ms_map.sm_size == (1ULL << vd->vdev_ms_shift)); 498 499 dump_spacemap(spa->spa_meta_objset, smo, &msp->ms_map); 500 } 501 502 static void 503 dump_metaslabs(spa_t *spa) 504 { 505 vdev_t *rvd = spa->spa_root_vdev; 506 vdev_t *vd; 507 int c, m; 508 509 (void) printf("\nMetaslabs:\n"); 510 511 for (c = 0; c < rvd->vdev_children; c++) { 512 vd = rvd->vdev_child[c]; 513 514 (void) printf("\n vdev %llu\n\n", (u_longlong_t)vd->vdev_id); 515 516 if (dump_opt['d'] <= 5) { 517 (void) printf("\t%10s %10s %5s\n", 518 "offset", "spacemap", "free"); 519 (void) printf("\t%10s %10s %5s\n", 520 "------", "--------", "----"); 521 } 522 for (m = 0; m < vd->vdev_ms_count; m++) 523 dump_metaslab(vd->vdev_ms[m]); 524 (void) printf("\n"); 525 } 526 } 527 528 static void 529 dump_dtl_seg(space_map_t *sm, uint64_t start, uint64_t size) 530 { 531 char *prefix = (void *)sm; 532 533 (void) printf("%s [%llu,%llu) length %llu\n", 534 prefix, 535 (u_longlong_t)start, 536 (u_longlong_t)(start + size), 537 (u_longlong_t)(size)); 538 } 539 540 static void 541 dump_dtl(vdev_t *vd, int indent) 542 { 543 spa_t *spa = vd->vdev_spa; 544 boolean_t required; 545 char *name[DTL_TYPES] = { "missing", "partial", "scrub", "outage" }; 546 char prefix[256]; 547 548 spa_vdev_state_enter(spa); 549 required = vdev_dtl_required(vd); 550 (void) spa_vdev_state_exit(spa, NULL, 0); 551 552 if (indent == 0) 553 (void) printf("\nDirty time logs:\n\n"); 554 555 (void) printf("\t%*s%s [%s]\n", indent, "", 556 vd->vdev_path ? vd->vdev_path : 557 vd->vdev_parent ? vd->vdev_ops->vdev_op_type : spa_name(spa), 558 required ? "DTL-required" : "DTL-expendable"); 559 560 for (int t = 0; t < DTL_TYPES; t++) { 561 space_map_t *sm = &vd->vdev_dtl[t]; 562 if (sm->sm_space == 0) 563 continue; 564 (void) snprintf(prefix, sizeof (prefix), "\t%*s%s", 565 indent + 2, "", name[t]); 566 mutex_enter(sm->sm_lock); 567 space_map_walk(sm, dump_dtl_seg, (void *)prefix); 568 mutex_exit(sm->sm_lock); 569 if (dump_opt['d'] > 5 && vd->vdev_children == 0) 570 dump_spacemap(spa->spa_meta_objset, 571 &vd->vdev_dtl_smo, sm); 572 } 573 574 for (int c = 0; c < vd->vdev_children; c++) 575 dump_dtl(vd->vdev_child[c], indent + 4); 576 } 577 578 /*ARGSUSED*/ 579 static void 580 dump_dnode(objset_t *os, uint64_t object, void *data, size_t size) 581 { 582 } 583 584 static uint64_t 585 blkid2offset(const dnode_phys_t *dnp, int level, uint64_t blkid) 586 { 587 if (level < 0) 588 return (blkid); 589 590 return ((blkid << (level * (dnp->dn_indblkshift - SPA_BLKPTRSHIFT))) * 591 dnp->dn_datablkszsec << SPA_MINBLOCKSHIFT); 592 } 593 594 static void 595 sprintf_blkptr_compact(char *blkbuf, blkptr_t *bp, int alldvas) 596 { 597 dva_t *dva = bp->blk_dva; 598 int ndvas = alldvas ? BP_GET_NDVAS(bp) : 1; 599 int i; 600 601 blkbuf[0] = '\0'; 602 603 for (i = 0; i < ndvas; i++) 604 (void) sprintf(blkbuf + strlen(blkbuf), "%llu:%llx:%llx ", 605 (u_longlong_t)DVA_GET_VDEV(&dva[i]), 606 (u_longlong_t)DVA_GET_OFFSET(&dva[i]), 607 (u_longlong_t)DVA_GET_ASIZE(&dva[i])); 608 609 (void) sprintf(blkbuf + strlen(blkbuf), "%llxL/%llxP F=%llu B=%llu", 610 (u_longlong_t)BP_GET_LSIZE(bp), 611 (u_longlong_t)BP_GET_PSIZE(bp), 612 (u_longlong_t)bp->blk_fill, 613 (u_longlong_t)bp->blk_birth); 614 } 615 616 static void 617 print_indirect(blkptr_t *bp, const zbookmark_t *zb, 618 const dnode_phys_t *dnp) 619 { 620 char blkbuf[BP_SPRINTF_LEN]; 621 int l; 622 623 ASSERT3U(BP_GET_TYPE(bp), ==, dnp->dn_type); 624 ASSERT3U(BP_GET_LEVEL(bp), ==, zb->zb_level); 625 626 (void) printf("%16llx ", 627 (u_longlong_t)blkid2offset(dnp, zb->zb_level, zb->zb_blkid)); 628 629 ASSERT(zb->zb_level >= 0); 630 631 for (l = dnp->dn_nlevels - 1; l >= -1; l--) { 632 if (l == zb->zb_level) { 633 (void) printf("L%llx", (u_longlong_t)zb->zb_level); 634 } else { 635 (void) printf(" "); 636 } 637 } 638 639 sprintf_blkptr_compact(blkbuf, bp, dump_opt['d'] > 5 ? 1 : 0); 640 (void) printf("%s\n", blkbuf); 641 } 642 643 #define SET_BOOKMARK(zb, objset, object, level, blkid) \ 644 { \ 645 (zb)->zb_objset = objset; \ 646 (zb)->zb_object = object; \ 647 (zb)->zb_level = level; \ 648 (zb)->zb_blkid = blkid; \ 649 } 650 651 static int 652 visit_indirect(spa_t *spa, const dnode_phys_t *dnp, 653 blkptr_t *bp, const zbookmark_t *zb) 654 { 655 int err; 656 657 if (bp->blk_birth == 0) 658 return (0); 659 660 print_indirect(bp, zb, dnp); 661 662 if (BP_GET_LEVEL(bp) > 0) { 663 uint32_t flags = ARC_WAIT; 664 int i; 665 blkptr_t *cbp; 666 int epb = BP_GET_LSIZE(bp) >> SPA_BLKPTRSHIFT; 667 arc_buf_t *buf; 668 uint64_t fill = 0; 669 670 err = arc_read_nolock(NULL, spa, bp, arc_getbuf_func, &buf, 671 ZIO_PRIORITY_ASYNC_READ, ZIO_FLAG_CANFAIL, &flags, zb); 672 if (err) 673 return (err); 674 675 /* recursively visit blocks below this */ 676 cbp = buf->b_data; 677 for (i = 0; i < epb; i++, cbp++) { 678 zbookmark_t czb; 679 680 SET_BOOKMARK(&czb, zb->zb_objset, zb->zb_object, 681 zb->zb_level - 1, 682 zb->zb_blkid * epb + i); 683 err = visit_indirect(spa, dnp, cbp, &czb); 684 if (err) 685 break; 686 fill += cbp->blk_fill; 687 } 688 if (!err) 689 ASSERT3U(fill, ==, bp->blk_fill); 690 (void) arc_buf_remove_ref(buf, &buf); 691 } 692 693 return (err); 694 } 695 696 /*ARGSUSED*/ 697 static void 698 dump_indirect(dnode_t *dn) 699 { 700 dnode_phys_t *dnp = dn->dn_phys; 701 int j; 702 zbookmark_t czb; 703 704 (void) printf("Indirect blocks:\n"); 705 706 SET_BOOKMARK(&czb, dmu_objset_id(&dn->dn_objset->os), 707 dn->dn_object, dnp->dn_nlevels - 1, 0); 708 for (j = 0; j < dnp->dn_nblkptr; j++) { 709 czb.zb_blkid = j; 710 (void) visit_indirect(dmu_objset_spa(&dn->dn_objset->os), dnp, 711 &dnp->dn_blkptr[j], &czb); 712 } 713 714 (void) printf("\n"); 715 } 716 717 /*ARGSUSED*/ 718 static void 719 dump_dsl_dir(objset_t *os, uint64_t object, void *data, size_t size) 720 { 721 dsl_dir_phys_t *dd = data; 722 time_t crtime; 723 char nice[6]; 724 725 if (dd == NULL) 726 return; 727 728 ASSERT3U(size, >=, sizeof (dsl_dir_phys_t)); 729 730 crtime = dd->dd_creation_time; 731 (void) printf("\t\tcreation_time = %s", ctime(&crtime)); 732 (void) printf("\t\thead_dataset_obj = %llu\n", 733 (u_longlong_t)dd->dd_head_dataset_obj); 734 (void) printf("\t\tparent_dir_obj = %llu\n", 735 (u_longlong_t)dd->dd_parent_obj); 736 (void) printf("\t\torigin_obj = %llu\n", 737 (u_longlong_t)dd->dd_origin_obj); 738 (void) printf("\t\tchild_dir_zapobj = %llu\n", 739 (u_longlong_t)dd->dd_child_dir_zapobj); 740 nicenum(dd->dd_used_bytes, nice); 741 (void) printf("\t\tused_bytes = %s\n", nice); 742 nicenum(dd->dd_compressed_bytes, nice); 743 (void) printf("\t\tcompressed_bytes = %s\n", nice); 744 nicenum(dd->dd_uncompressed_bytes, nice); 745 (void) printf("\t\tuncompressed_bytes = %s\n", nice); 746 nicenum(dd->dd_quota, nice); 747 (void) printf("\t\tquota = %s\n", nice); 748 nicenum(dd->dd_reserved, nice); 749 (void) printf("\t\treserved = %s\n", nice); 750 (void) printf("\t\tprops_zapobj = %llu\n", 751 (u_longlong_t)dd->dd_props_zapobj); 752 (void) printf("\t\tdeleg_zapobj = %llu\n", 753 (u_longlong_t)dd->dd_deleg_zapobj); 754 (void) printf("\t\tflags = %llx\n", 755 (u_longlong_t)dd->dd_flags); 756 757 #define DO(which) \ 758 nicenum(dd->dd_used_breakdown[DD_USED_ ## which], nice); \ 759 (void) printf("\t\tused_breakdown[" #which "] = %s\n", nice) 760 DO(HEAD); 761 DO(SNAP); 762 DO(CHILD); 763 DO(CHILD_RSRV); 764 DO(REFRSRV); 765 #undef DO 766 } 767 768 /*ARGSUSED*/ 769 static void 770 dump_dsl_dataset(objset_t *os, uint64_t object, void *data, size_t size) 771 { 772 dsl_dataset_phys_t *ds = data; 773 time_t crtime; 774 char used[6], compressed[6], uncompressed[6], unique[6]; 775 char blkbuf[BP_SPRINTF_LEN]; 776 777 if (ds == NULL) 778 return; 779 780 ASSERT(size == sizeof (*ds)); 781 crtime = ds->ds_creation_time; 782 nicenum(ds->ds_used_bytes, used); 783 nicenum(ds->ds_compressed_bytes, compressed); 784 nicenum(ds->ds_uncompressed_bytes, uncompressed); 785 nicenum(ds->ds_unique_bytes, unique); 786 sprintf_blkptr(blkbuf, BP_SPRINTF_LEN, &ds->ds_bp); 787 788 (void) printf("\t\tdir_obj = %llu\n", 789 (u_longlong_t)ds->ds_dir_obj); 790 (void) printf("\t\tprev_snap_obj = %llu\n", 791 (u_longlong_t)ds->ds_prev_snap_obj); 792 (void) printf("\t\tprev_snap_txg = %llu\n", 793 (u_longlong_t)ds->ds_prev_snap_txg); 794 (void) printf("\t\tnext_snap_obj = %llu\n", 795 (u_longlong_t)ds->ds_next_snap_obj); 796 (void) printf("\t\tsnapnames_zapobj = %llu\n", 797 (u_longlong_t)ds->ds_snapnames_zapobj); 798 (void) printf("\t\tnum_children = %llu\n", 799 (u_longlong_t)ds->ds_num_children); 800 (void) printf("\t\tcreation_time = %s", ctime(&crtime)); 801 (void) printf("\t\tcreation_txg = %llu\n", 802 (u_longlong_t)ds->ds_creation_txg); 803 (void) printf("\t\tdeadlist_obj = %llu\n", 804 (u_longlong_t)ds->ds_deadlist_obj); 805 (void) printf("\t\tused_bytes = %s\n", used); 806 (void) printf("\t\tcompressed_bytes = %s\n", compressed); 807 (void) printf("\t\tuncompressed_bytes = %s\n", uncompressed); 808 (void) printf("\t\tunique = %s\n", unique); 809 (void) printf("\t\tfsid_guid = %llu\n", 810 (u_longlong_t)ds->ds_fsid_guid); 811 (void) printf("\t\tguid = %llu\n", 812 (u_longlong_t)ds->ds_guid); 813 (void) printf("\t\tflags = %llx\n", 814 (u_longlong_t)ds->ds_flags); 815 (void) printf("\t\tnext_clones_obj = %llu\n", 816 (u_longlong_t)ds->ds_next_clones_obj); 817 (void) printf("\t\tprops_obj = %llu\n", 818 (u_longlong_t)ds->ds_props_obj); 819 (void) printf("\t\tbp = %s\n", blkbuf); 820 } 821 822 static void 823 dump_bplist(objset_t *mos, uint64_t object, char *name) 824 { 825 bplist_t bpl = { 0 }; 826 blkptr_t blk, *bp = &blk; 827 uint64_t itor = 0; 828 char bytes[6]; 829 char comp[6]; 830 char uncomp[6]; 831 832 if (dump_opt['d'] < 3) 833 return; 834 835 mutex_init(&bpl.bpl_lock, NULL, MUTEX_DEFAULT, NULL); 836 VERIFY(0 == bplist_open(&bpl, mos, object)); 837 if (bplist_empty(&bpl)) { 838 bplist_close(&bpl); 839 mutex_destroy(&bpl.bpl_lock); 840 return; 841 } 842 843 nicenum(bpl.bpl_phys->bpl_bytes, bytes); 844 if (bpl.bpl_dbuf->db_size == sizeof (bplist_phys_t)) { 845 nicenum(bpl.bpl_phys->bpl_comp, comp); 846 nicenum(bpl.bpl_phys->bpl_uncomp, uncomp); 847 (void) printf("\n %s: %llu entries, %s (%s/%s comp)\n", 848 name, (u_longlong_t)bpl.bpl_phys->bpl_entries, 849 bytes, comp, uncomp); 850 } else { 851 (void) printf("\n %s: %llu entries, %s\n", 852 name, (u_longlong_t)bpl.bpl_phys->bpl_entries, bytes); 853 } 854 855 if (dump_opt['d'] < 5) { 856 bplist_close(&bpl); 857 mutex_destroy(&bpl.bpl_lock); 858 return; 859 } 860 861 (void) printf("\n"); 862 863 while (bplist_iterate(&bpl, &itor, bp) == 0) { 864 char blkbuf[BP_SPRINTF_LEN]; 865 866 ASSERT(bp->blk_birth != 0); 867 sprintf_blkptr_compact(blkbuf, bp, dump_opt['d'] > 5 ? 1 : 0); 868 (void) printf("\tItem %3llu: %s\n", 869 (u_longlong_t)itor - 1, blkbuf); 870 } 871 872 bplist_close(&bpl); 873 mutex_destroy(&bpl.bpl_lock); 874 } 875 876 static avl_tree_t idx_tree; 877 static avl_tree_t domain_tree; 878 static boolean_t fuid_table_loaded; 879 880 static void 881 fuid_table_destroy() 882 { 883 if (fuid_table_loaded) { 884 zfs_fuid_table_destroy(&idx_tree, &domain_tree); 885 fuid_table_loaded = B_FALSE; 886 } 887 } 888 889 /* 890 * print uid or gid information. 891 * For normal POSIX id just the id is printed in decimal format. 892 * For CIFS files with FUID the fuid is printed in hex followed by 893 * the doman-rid string. 894 */ 895 static void 896 print_idstr(uint64_t id, const char *id_type) 897 { 898 if (FUID_INDEX(id)) { 899 char *domain; 900 901 domain = zfs_fuid_idx_domain(&idx_tree, FUID_INDEX(id)); 902 (void) printf("\t%s %llx [%s-%d]\n", id_type, 903 (u_longlong_t)id, domain, (int)FUID_RID(id)); 904 } else { 905 (void) printf("\t%s %llu\n", id_type, (u_longlong_t)id); 906 } 907 908 } 909 910 static void 911 dump_uidgid(objset_t *os, znode_phys_t *zp) 912 { 913 uint32_t uid_idx, gid_idx; 914 915 uid_idx = FUID_INDEX(zp->zp_uid); 916 gid_idx = FUID_INDEX(zp->zp_gid); 917 918 /* Load domain table, if not already loaded */ 919 if (!fuid_table_loaded && (uid_idx || gid_idx)) { 920 uint64_t fuid_obj; 921 922 /* first find the fuid object. It lives in the master node */ 923 VERIFY(zap_lookup(os, MASTER_NODE_OBJ, ZFS_FUID_TABLES, 924 8, 1, &fuid_obj) == 0); 925 zfs_fuid_avl_tree_create(&idx_tree, &domain_tree); 926 (void) zfs_fuid_table_load(os, fuid_obj, 927 &idx_tree, &domain_tree); 928 fuid_table_loaded = B_TRUE; 929 } 930 931 print_idstr(zp->zp_uid, "uid"); 932 print_idstr(zp->zp_gid, "gid"); 933 } 934 935 /*ARGSUSED*/ 936 static void 937 dump_znode(objset_t *os, uint64_t object, void *data, size_t size) 938 { 939 znode_phys_t *zp = data; 940 time_t z_crtime, z_atime, z_mtime, z_ctime; 941 char path[MAXPATHLEN * 2]; /* allow for xattr and failure prefix */ 942 int error; 943 944 ASSERT(size >= sizeof (znode_phys_t)); 945 946 error = zfs_obj_to_path(os, object, path, sizeof (path)); 947 if (error != 0) { 948 (void) snprintf(path, sizeof (path), "\?\?\?<object#%llu>", 949 (u_longlong_t)object); 950 } 951 952 if (dump_opt['d'] < 3) { 953 (void) printf("\t%s\n", path); 954 return; 955 } 956 957 z_crtime = (time_t)zp->zp_crtime[0]; 958 z_atime = (time_t)zp->zp_atime[0]; 959 z_mtime = (time_t)zp->zp_mtime[0]; 960 z_ctime = (time_t)zp->zp_ctime[0]; 961 962 (void) printf("\tpath %s\n", path); 963 dump_uidgid(os, zp); 964 (void) printf("\tatime %s", ctime(&z_atime)); 965 (void) printf("\tmtime %s", ctime(&z_mtime)); 966 (void) printf("\tctime %s", ctime(&z_ctime)); 967 (void) printf("\tcrtime %s", ctime(&z_crtime)); 968 (void) printf("\tgen %llu\n", (u_longlong_t)zp->zp_gen); 969 (void) printf("\tmode %llo\n", (u_longlong_t)zp->zp_mode); 970 (void) printf("\tsize %llu\n", (u_longlong_t)zp->zp_size); 971 (void) printf("\tparent %llu\n", (u_longlong_t)zp->zp_parent); 972 (void) printf("\tlinks %llu\n", (u_longlong_t)zp->zp_links); 973 (void) printf("\txattr %llu\n", (u_longlong_t)zp->zp_xattr); 974 (void) printf("\trdev 0x%016llx\n", (u_longlong_t)zp->zp_rdev); 975 } 976 977 /*ARGSUSED*/ 978 static void 979 dump_acl(objset_t *os, uint64_t object, void *data, size_t size) 980 { 981 } 982 983 /*ARGSUSED*/ 984 static void 985 dump_dmu_objset(objset_t *os, uint64_t object, void *data, size_t size) 986 { 987 } 988 989 static object_viewer_t *object_viewer[DMU_OT_NUMTYPES] = { 990 dump_none, /* unallocated */ 991 dump_zap, /* object directory */ 992 dump_uint64, /* object array */ 993 dump_none, /* packed nvlist */ 994 dump_packed_nvlist, /* packed nvlist size */ 995 dump_none, /* bplist */ 996 dump_none, /* bplist header */ 997 dump_none, /* SPA space map header */ 998 dump_none, /* SPA space map */ 999 dump_none, /* ZIL intent log */ 1000 dump_dnode, /* DMU dnode */ 1001 dump_dmu_objset, /* DMU objset */ 1002 dump_dsl_dir, /* DSL directory */ 1003 dump_zap, /* DSL directory child map */ 1004 dump_zap, /* DSL dataset snap map */ 1005 dump_zap, /* DSL props */ 1006 dump_dsl_dataset, /* DSL dataset */ 1007 dump_znode, /* ZFS znode */ 1008 dump_acl, /* ZFS V0 ACL */ 1009 dump_uint8, /* ZFS plain file */ 1010 dump_zpldir, /* ZFS directory */ 1011 dump_zap, /* ZFS master node */ 1012 dump_zap, /* ZFS delete queue */ 1013 dump_uint8, /* zvol object */ 1014 dump_zap, /* zvol prop */ 1015 dump_uint8, /* other uint8[] */ 1016 dump_uint64, /* other uint64[] */ 1017 dump_zap, /* other ZAP */ 1018 dump_zap, /* persistent error log */ 1019 dump_uint8, /* SPA history */ 1020 dump_uint64, /* SPA history offsets */ 1021 dump_zap, /* Pool properties */ 1022 dump_zap, /* DSL permissions */ 1023 dump_acl, /* ZFS ACL */ 1024 dump_uint8, /* ZFS SYSACL */ 1025 dump_none, /* FUID nvlist */ 1026 dump_packed_nvlist, /* FUID nvlist size */ 1027 dump_zap, /* DSL dataset next clones */ 1028 dump_zap, /* DSL scrub queue */ 1029 }; 1030 1031 static void 1032 dump_object(objset_t *os, uint64_t object, int verbosity, int *print_header) 1033 { 1034 dmu_buf_t *db = NULL; 1035 dmu_object_info_t doi; 1036 dnode_t *dn; 1037 void *bonus = NULL; 1038 size_t bsize = 0; 1039 char iblk[6], dblk[6], lsize[6], asize[6], bonus_size[6], segsize[6]; 1040 char aux[50]; 1041 int error; 1042 1043 if (*print_header) { 1044 (void) printf("\n Object lvl iblk dblk lsize" 1045 " asize type\n"); 1046 *print_header = 0; 1047 } 1048 1049 if (object == 0) { 1050 dn = os->os->os_meta_dnode; 1051 } else { 1052 error = dmu_bonus_hold(os, object, FTAG, &db); 1053 if (error) 1054 fatal("dmu_bonus_hold(%llu) failed, errno %u", 1055 object, error); 1056 bonus = db->db_data; 1057 bsize = db->db_size; 1058 dn = ((dmu_buf_impl_t *)db)->db_dnode; 1059 } 1060 dmu_object_info_from_dnode(dn, &doi); 1061 1062 nicenum(doi.doi_metadata_block_size, iblk); 1063 nicenum(doi.doi_data_block_size, dblk); 1064 nicenum(doi.doi_data_block_size * (doi.doi_max_block_offset + 1), 1065 lsize); 1066 nicenum(doi.doi_physical_blks << 9, asize); 1067 nicenum(doi.doi_bonus_size, bonus_size); 1068 1069 aux[0] = '\0'; 1070 1071 if (doi.doi_checksum != ZIO_CHECKSUM_INHERIT || verbosity >= 6) { 1072 (void) snprintf(aux + strlen(aux), sizeof (aux), " (K=%s)", 1073 zio_checksum_table[doi.doi_checksum].ci_name); 1074 } 1075 1076 if (doi.doi_compress != ZIO_COMPRESS_INHERIT || verbosity >= 6) { 1077 (void) snprintf(aux + strlen(aux), sizeof (aux), " (Z=%s)", 1078 zio_compress_table[doi.doi_compress].ci_name); 1079 } 1080 1081 (void) printf("%10lld %3u %5s %5s %5s %5s %s%s\n", 1082 (u_longlong_t)object, doi.doi_indirection, iblk, dblk, lsize, 1083 asize, dmu_ot[doi.doi_type].ot_name, aux); 1084 1085 if (doi.doi_bonus_type != DMU_OT_NONE && verbosity > 3) { 1086 (void) printf("%10s %3s %5s %5s %5s %5s %s\n", 1087 "", "", "", "", bonus_size, "bonus", 1088 dmu_ot[doi.doi_bonus_type].ot_name); 1089 } 1090 1091 if (verbosity >= 4) { 1092 object_viewer[doi.doi_bonus_type](os, object, bonus, bsize); 1093 object_viewer[doi.doi_type](os, object, NULL, 0); 1094 *print_header = 1; 1095 } 1096 1097 if (verbosity >= 5) 1098 dump_indirect(dn); 1099 1100 if (verbosity >= 5) { 1101 /* 1102 * Report the list of segments that comprise the object. 1103 */ 1104 uint64_t start = 0; 1105 uint64_t end; 1106 uint64_t blkfill = 1; 1107 int minlvl = 1; 1108 1109 if (dn->dn_type == DMU_OT_DNODE) { 1110 minlvl = 0; 1111 blkfill = DNODES_PER_BLOCK; 1112 } 1113 1114 for (;;) { 1115 error = dnode_next_offset(dn, 1116 0, &start, minlvl, blkfill, 0); 1117 if (error) 1118 break; 1119 end = start; 1120 error = dnode_next_offset(dn, 1121 DNODE_FIND_HOLE, &end, minlvl, blkfill, 0); 1122 nicenum(end - start, segsize); 1123 (void) printf("\t\tsegment [%016llx, %016llx)" 1124 " size %5s\n", (u_longlong_t)start, 1125 (u_longlong_t)end, segsize); 1126 if (error) 1127 break; 1128 start = end; 1129 } 1130 } 1131 1132 if (db != NULL) 1133 dmu_buf_rele(db, FTAG); 1134 } 1135 1136 static char *objset_types[DMU_OST_NUMTYPES] = { 1137 "NONE", "META", "ZPL", "ZVOL", "OTHER", "ANY" }; 1138 1139 static void 1140 dump_dir(objset_t *os) 1141 { 1142 dmu_objset_stats_t dds; 1143 uint64_t object, object_count; 1144 uint64_t refdbytes, usedobjs, scratch; 1145 char numbuf[8]; 1146 char blkbuf[BP_SPRINTF_LEN]; 1147 char osname[MAXNAMELEN]; 1148 char *type = "UNKNOWN"; 1149 int verbosity = dump_opt['d']; 1150 int print_header = 1; 1151 int i, error; 1152 1153 dmu_objset_fast_stat(os, &dds); 1154 1155 if (dds.dds_type < DMU_OST_NUMTYPES) 1156 type = objset_types[dds.dds_type]; 1157 1158 if (dds.dds_type == DMU_OST_META) { 1159 dds.dds_creation_txg = TXG_INITIAL; 1160 usedobjs = os->os->os_rootbp->blk_fill; 1161 refdbytes = os->os->os_spa->spa_dsl_pool-> 1162 dp_mos_dir->dd_phys->dd_used_bytes; 1163 } else { 1164 dmu_objset_space(os, &refdbytes, &scratch, &usedobjs, &scratch); 1165 } 1166 1167 ASSERT3U(usedobjs, ==, os->os->os_rootbp->blk_fill); 1168 1169 nicenum(refdbytes, numbuf); 1170 1171 if (verbosity >= 4) { 1172 (void) strcpy(blkbuf, ", rootbp "); 1173 sprintf_blkptr(blkbuf + strlen(blkbuf), 1174 BP_SPRINTF_LEN - strlen(blkbuf), os->os->os_rootbp); 1175 } else { 1176 blkbuf[0] = '\0'; 1177 } 1178 1179 dmu_objset_name(os, osname); 1180 1181 (void) printf("Dataset %s [%s], ID %llu, cr_txg %llu, " 1182 "%s, %llu objects%s\n", 1183 osname, type, (u_longlong_t)dmu_objset_id(os), 1184 (u_longlong_t)dds.dds_creation_txg, 1185 numbuf, (u_longlong_t)usedobjs, blkbuf); 1186 1187 dump_intent_log(dmu_objset_zil(os)); 1188 1189 if (dmu_objset_ds(os) != NULL) 1190 dump_bplist(dmu_objset_pool(os)->dp_meta_objset, 1191 dmu_objset_ds(os)->ds_phys->ds_deadlist_obj, "Deadlist"); 1192 1193 if (verbosity < 2) 1194 return; 1195 1196 if (os->os->os_rootbp->blk_birth == 0) 1197 return; 1198 1199 if (zopt_objects != 0) { 1200 for (i = 0; i < zopt_objects; i++) 1201 dump_object(os, zopt_object[i], verbosity, 1202 &print_header); 1203 (void) printf("\n"); 1204 return; 1205 } 1206 1207 dump_object(os, 0, verbosity, &print_header); 1208 object_count = 1; 1209 1210 object = 0; 1211 while ((error = dmu_object_next(os, &object, B_FALSE, 0)) == 0) { 1212 dump_object(os, object, verbosity, &print_header); 1213 object_count++; 1214 } 1215 1216 ASSERT3U(object_count, ==, usedobjs); 1217 1218 (void) printf("\n"); 1219 1220 if (error != ESRCH) { 1221 (void) fprintf(stderr, "dmu_object_next() = %d\n", error); 1222 abort(); 1223 } 1224 } 1225 1226 static void 1227 dump_uberblock(uberblock_t *ub) 1228 { 1229 time_t timestamp = ub->ub_timestamp; 1230 1231 (void) printf("Uberblock\n\n"); 1232 (void) printf("\tmagic = %016llx\n", (u_longlong_t)ub->ub_magic); 1233 (void) printf("\tversion = %llu\n", (u_longlong_t)ub->ub_version); 1234 (void) printf("\ttxg = %llu\n", (u_longlong_t)ub->ub_txg); 1235 (void) printf("\tguid_sum = %llu\n", (u_longlong_t)ub->ub_guid_sum); 1236 (void) printf("\ttimestamp = %llu UTC = %s", 1237 (u_longlong_t)ub->ub_timestamp, asctime(localtime(×tamp))); 1238 if (dump_opt['u'] >= 3) { 1239 char blkbuf[BP_SPRINTF_LEN]; 1240 sprintf_blkptr(blkbuf, BP_SPRINTF_LEN, &ub->ub_rootbp); 1241 (void) printf("\trootbp = %s\n", blkbuf); 1242 } 1243 (void) printf("\n"); 1244 } 1245 1246 static void 1247 dump_config(const char *pool) 1248 { 1249 spa_t *spa = NULL; 1250 1251 mutex_enter(&spa_namespace_lock); 1252 while ((spa = spa_next(spa)) != NULL) { 1253 if (pool == NULL) 1254 (void) printf("%s\n", spa_name(spa)); 1255 if (pool == NULL || strcmp(pool, spa_name(spa)) == 0) 1256 dump_nvlist(spa->spa_config, 4); 1257 } 1258 mutex_exit(&spa_namespace_lock); 1259 } 1260 1261 static void 1262 dump_cachefile(const char *cachefile) 1263 { 1264 int fd; 1265 struct stat64 statbuf; 1266 char *buf; 1267 nvlist_t *config; 1268 1269 if ((fd = open64(cachefile, O_RDONLY)) < 0) { 1270 (void) printf("cannot open '%s': %s\n", cachefile, 1271 strerror(errno)); 1272 exit(1); 1273 } 1274 1275 if (fstat64(fd, &statbuf) != 0) { 1276 (void) printf("failed to stat '%s': %s\n", cachefile, 1277 strerror(errno)); 1278 exit(1); 1279 } 1280 1281 if ((buf = malloc(statbuf.st_size)) == NULL) { 1282 (void) fprintf(stderr, "failed to allocate %llu bytes\n", 1283 (u_longlong_t)statbuf.st_size); 1284 exit(1); 1285 } 1286 1287 if (read(fd, buf, statbuf.st_size) != statbuf.st_size) { 1288 (void) fprintf(stderr, "failed to read %llu bytes\n", 1289 (u_longlong_t)statbuf.st_size); 1290 exit(1); 1291 } 1292 1293 (void) close(fd); 1294 1295 if (nvlist_unpack(buf, statbuf.st_size, &config, 0) != 0) { 1296 (void) fprintf(stderr, "failed to unpack nvlist\n"); 1297 exit(1); 1298 } 1299 1300 free(buf); 1301 1302 dump_nvlist(config, 0); 1303 1304 nvlist_free(config); 1305 } 1306 1307 static void 1308 dump_label(const char *dev) 1309 { 1310 int fd; 1311 vdev_label_t label; 1312 char *buf = label.vl_vdev_phys.vp_nvlist; 1313 size_t buflen = sizeof (label.vl_vdev_phys.vp_nvlist); 1314 struct stat64 statbuf; 1315 uint64_t psize; 1316 int l; 1317 1318 if ((fd = open64(dev, O_RDONLY)) < 0) { 1319 (void) printf("cannot open '%s': %s\n", dev, strerror(errno)); 1320 exit(1); 1321 } 1322 1323 if (fstat64(fd, &statbuf) != 0) { 1324 (void) printf("failed to stat '%s': %s\n", dev, 1325 strerror(errno)); 1326 exit(1); 1327 } 1328 1329 psize = statbuf.st_size; 1330 psize = P2ALIGN(psize, (uint64_t)sizeof (vdev_label_t)); 1331 1332 for (l = 0; l < VDEV_LABELS; l++) { 1333 1334 nvlist_t *config = NULL; 1335 1336 (void) printf("--------------------------------------------\n"); 1337 (void) printf("LABEL %d\n", l); 1338 (void) printf("--------------------------------------------\n"); 1339 1340 if (pread64(fd, &label, sizeof (label), 1341 vdev_label_offset(psize, l, 0)) != sizeof (label)) { 1342 (void) printf("failed to read label %d\n", l); 1343 continue; 1344 } 1345 1346 if (nvlist_unpack(buf, buflen, &config, 0) != 0) { 1347 (void) printf("failed to unpack label %d\n", l); 1348 continue; 1349 } 1350 dump_nvlist(config, 4); 1351 nvlist_free(config); 1352 } 1353 } 1354 1355 /*ARGSUSED*/ 1356 static int 1357 dump_one_dir(char *dsname, void *arg) 1358 { 1359 int error; 1360 objset_t *os; 1361 1362 error = dmu_objset_open(dsname, DMU_OST_ANY, 1363 DS_MODE_USER | DS_MODE_READONLY, &os); 1364 if (error) { 1365 (void) printf("Could not open %s\n", dsname); 1366 return (0); 1367 } 1368 dump_dir(os); 1369 dmu_objset_close(os); 1370 fuid_table_destroy(); 1371 return (0); 1372 } 1373 1374 static void 1375 zdb_leak(space_map_t *sm, uint64_t start, uint64_t size) 1376 { 1377 vdev_t *vd = sm->sm_ppd; 1378 1379 (void) printf("leaked space: vdev %llu, offset 0x%llx, size %llu\n", 1380 (u_longlong_t)vd->vdev_id, (u_longlong_t)start, (u_longlong_t)size); 1381 } 1382 1383 /* ARGSUSED */ 1384 static void 1385 zdb_space_map_load(space_map_t *sm) 1386 { 1387 } 1388 1389 static void 1390 zdb_space_map_unload(space_map_t *sm) 1391 { 1392 space_map_vacate(sm, zdb_leak, sm); 1393 } 1394 1395 /* ARGSUSED */ 1396 static void 1397 zdb_space_map_claim(space_map_t *sm, uint64_t start, uint64_t size) 1398 { 1399 } 1400 1401 static space_map_ops_t zdb_space_map_ops = { 1402 zdb_space_map_load, 1403 zdb_space_map_unload, 1404 NULL, /* alloc */ 1405 zdb_space_map_claim, 1406 NULL /* free */ 1407 }; 1408 1409 static void 1410 zdb_leak_init(spa_t *spa) 1411 { 1412 vdev_t *rvd = spa->spa_root_vdev; 1413 1414 for (int c = 0; c < rvd->vdev_children; c++) { 1415 vdev_t *vd = rvd->vdev_child[c]; 1416 for (int m = 0; m < vd->vdev_ms_count; m++) { 1417 metaslab_t *msp = vd->vdev_ms[m]; 1418 mutex_enter(&msp->ms_lock); 1419 VERIFY(space_map_load(&msp->ms_map, &zdb_space_map_ops, 1420 SM_ALLOC, &msp->ms_smo, spa->spa_meta_objset) == 0); 1421 msp->ms_map.sm_ppd = vd; 1422 mutex_exit(&msp->ms_lock); 1423 } 1424 } 1425 } 1426 1427 static void 1428 zdb_leak_fini(spa_t *spa) 1429 { 1430 vdev_t *rvd = spa->spa_root_vdev; 1431 1432 for (int c = 0; c < rvd->vdev_children; c++) { 1433 vdev_t *vd = rvd->vdev_child[c]; 1434 for (int m = 0; m < vd->vdev_ms_count; m++) { 1435 metaslab_t *msp = vd->vdev_ms[m]; 1436 mutex_enter(&msp->ms_lock); 1437 space_map_unload(&msp->ms_map); 1438 mutex_exit(&msp->ms_lock); 1439 } 1440 } 1441 } 1442 1443 /* 1444 * Verify that the sum of the sizes of all blocks in the pool adds up 1445 * to the SPA's sa_alloc total. 1446 */ 1447 typedef struct zdb_blkstats { 1448 uint64_t zb_asize; 1449 uint64_t zb_lsize; 1450 uint64_t zb_psize; 1451 uint64_t zb_count; 1452 } zdb_blkstats_t; 1453 1454 #define DMU_OT_DEFERRED DMU_OT_NONE 1455 #define DMU_OT_TOTAL DMU_OT_NUMTYPES 1456 1457 #define ZB_TOTAL DN_MAX_LEVELS 1458 1459 typedef struct zdb_cb { 1460 zdb_blkstats_t zcb_type[ZB_TOTAL + 1][DMU_OT_TOTAL + 1]; 1461 uint64_t zcb_errors[256]; 1462 int zcb_readfails; 1463 int zcb_haderrors; 1464 } zdb_cb_t; 1465 1466 static void 1467 zdb_count_block(spa_t *spa, zdb_cb_t *zcb, blkptr_t *bp, dmu_object_type_t type) 1468 { 1469 for (int i = 0; i < 4; i++) { 1470 int l = (i < 2) ? BP_GET_LEVEL(bp) : ZB_TOTAL; 1471 int t = (i & 1) ? type : DMU_OT_TOTAL; 1472 zdb_blkstats_t *zb = &zcb->zcb_type[l][t]; 1473 1474 zb->zb_asize += BP_GET_ASIZE(bp); 1475 zb->zb_lsize += BP_GET_LSIZE(bp); 1476 zb->zb_psize += BP_GET_PSIZE(bp); 1477 zb->zb_count++; 1478 } 1479 1480 if (dump_opt['S']) { 1481 boolean_t print_sig; 1482 1483 print_sig = !zdb_sig_user_data || (BP_GET_LEVEL(bp) == 0 && 1484 BP_GET_TYPE(bp) == DMU_OT_PLAIN_FILE_CONTENTS); 1485 1486 if (BP_GET_CHECKSUM(bp) < zdb_sig_cksumalg) 1487 print_sig = B_FALSE; 1488 1489 if (print_sig) { 1490 (void) printf("%llu\t%lld\t%lld\t%s\t%s\t%s\t" 1491 "%llx:%llx:%llx:%llx\n", 1492 (u_longlong_t)BP_GET_LEVEL(bp), 1493 (longlong_t)BP_GET_PSIZE(bp), 1494 (longlong_t)BP_GET_NDVAS(bp), 1495 dmu_ot[BP_GET_TYPE(bp)].ot_name, 1496 zio_checksum_table[BP_GET_CHECKSUM(bp)].ci_name, 1497 zio_compress_table[BP_GET_COMPRESS(bp)].ci_name, 1498 (u_longlong_t)bp->blk_cksum.zc_word[0], 1499 (u_longlong_t)bp->blk_cksum.zc_word[1], 1500 (u_longlong_t)bp->blk_cksum.zc_word[2], 1501 (u_longlong_t)bp->blk_cksum.zc_word[3]); 1502 } 1503 } 1504 1505 if (!dump_opt['L']) 1506 VERIFY(zio_wait(zio_claim(NULL, spa, spa_first_txg(spa), bp, 1507 NULL, NULL, ZIO_FLAG_MUSTSUCCEED)) == 0); 1508 } 1509 1510 static int 1511 zdb_blkptr_cb(spa_t *spa, blkptr_t *bp, const zbookmark_t *zb, 1512 const dnode_phys_t *dnp, void *arg) 1513 { 1514 zdb_cb_t *zcb = arg; 1515 char blkbuf[BP_SPRINTF_LEN]; 1516 1517 if (bp == NULL) 1518 return (0); 1519 1520 zdb_count_block(spa, zcb, bp, BP_GET_TYPE(bp)); 1521 1522 if (dump_opt['c'] || dump_opt['S']) { 1523 int ioerr, size; 1524 void *data; 1525 1526 size = BP_GET_LSIZE(bp); 1527 data = malloc(size); 1528 ioerr = zio_wait(zio_read(NULL, spa, bp, data, size, 1529 NULL, NULL, ZIO_PRIORITY_ASYNC_READ, 1530 ZIO_FLAG_CANFAIL | ZIO_FLAG_SCRUB, zb)); 1531 free(data); 1532 1533 /* We expect io errors on intent log */ 1534 if (ioerr && BP_GET_TYPE(bp) != DMU_OT_INTENT_LOG) { 1535 zcb->zcb_haderrors = 1; 1536 zcb->zcb_errors[ioerr]++; 1537 1538 if (dump_opt['b'] >= 2) 1539 sprintf_blkptr(blkbuf, BP_SPRINTF_LEN, bp); 1540 else 1541 blkbuf[0] = '\0'; 1542 1543 if (!dump_opt['S']) { 1544 (void) printf("zdb_blkptr_cb: " 1545 "Got error %d reading " 1546 "<%llu, %llu, %lld, %llx> %s -- skipping\n", 1547 ioerr, 1548 (u_longlong_t)zb->zb_objset, 1549 (u_longlong_t)zb->zb_object, 1550 (u_longlong_t)zb->zb_level, 1551 (u_longlong_t)zb->zb_blkid, 1552 blkbuf); 1553 } 1554 } 1555 } 1556 1557 zcb->zcb_readfails = 0; 1558 1559 if (dump_opt['b'] >= 4) { 1560 sprintf_blkptr(blkbuf, BP_SPRINTF_LEN, bp); 1561 (void) printf("objset %llu object %llu offset 0x%llx %s\n", 1562 (u_longlong_t)zb->zb_objset, 1563 (u_longlong_t)zb->zb_object, 1564 (u_longlong_t)blkid2offset(dnp, zb->zb_level, zb->zb_blkid), 1565 blkbuf); 1566 } 1567 1568 return (0); 1569 } 1570 1571 static int 1572 dump_block_stats(spa_t *spa) 1573 { 1574 zdb_cb_t zcb = { 0 }; 1575 zdb_blkstats_t *zb, *tzb; 1576 uint64_t alloc, space, logalloc; 1577 vdev_t *rvd = spa->spa_root_vdev; 1578 int leaks = 0; 1579 int c, e; 1580 1581 if (!dump_opt['S']) { 1582 (void) printf("\nTraversing all blocks %s%s%s%s...\n", 1583 (dump_opt['c'] || !dump_opt['L']) ? "to verify " : "", 1584 dump_opt['c'] ? "checksums " : "", 1585 (dump_opt['c'] && !dump_opt['L']) ? "and verify " : "", 1586 !dump_opt['L'] ? "nothing leaked " : ""); 1587 } 1588 1589 /* 1590 * Load all space maps as SM_ALLOC maps, then traverse the pool 1591 * claiming each block we discover. If the pool is perfectly 1592 * consistent, the space maps will be empty when we're done. 1593 * Anything left over is a leak; any block we can't claim (because 1594 * it's not part of any space map) is a double allocation, 1595 * reference to a freed block, or an unclaimed log block. 1596 */ 1597 if (!dump_opt['L']) 1598 zdb_leak_init(spa); 1599 1600 /* 1601 * If there's a deferred-free bplist, process that first. 1602 */ 1603 if (spa->spa_sync_bplist_obj != 0) { 1604 bplist_t *bpl = &spa->spa_sync_bplist; 1605 blkptr_t blk; 1606 uint64_t itor = 0; 1607 1608 VERIFY(0 == bplist_open(bpl, spa->spa_meta_objset, 1609 spa->spa_sync_bplist_obj)); 1610 1611 while (bplist_iterate(bpl, &itor, &blk) == 0) { 1612 if (dump_opt['b'] >= 4) { 1613 char blkbuf[BP_SPRINTF_LEN]; 1614 sprintf_blkptr(blkbuf, BP_SPRINTF_LEN, &blk); 1615 (void) printf("[%s] %s\n", 1616 "deferred free", blkbuf); 1617 } 1618 zdb_count_block(spa, &zcb, &blk, DMU_OT_DEFERRED); 1619 } 1620 1621 bplist_close(bpl); 1622 } 1623 1624 zcb.zcb_haderrors |= traverse_pool(spa, zdb_blkptr_cb, &zcb); 1625 1626 if (zcb.zcb_haderrors && !dump_opt['S']) { 1627 (void) printf("\nError counts:\n\n"); 1628 (void) printf("\t%5s %s\n", "errno", "count"); 1629 for (e = 0; e < 256; e++) { 1630 if (zcb.zcb_errors[e] != 0) { 1631 (void) printf("\t%5d %llu\n", 1632 e, (u_longlong_t)zcb.zcb_errors[e]); 1633 } 1634 } 1635 } 1636 1637 /* 1638 * Report any leaked segments. 1639 */ 1640 if (!dump_opt['L']) 1641 zdb_leak_fini(spa); 1642 1643 /* 1644 * If we're interested in printing out the blkptr signatures, 1645 * return now as we don't print out anything else (including 1646 * errors and leaks). 1647 */ 1648 if (dump_opt['S']) 1649 return (zcb.zcb_haderrors ? 3 : 0); 1650 1651 alloc = spa_get_alloc(spa); 1652 space = spa_get_space(spa); 1653 1654 /* 1655 * Log blocks allocated from a separate log device don't count 1656 * as part of the normal pool space; factor them in here. 1657 */ 1658 logalloc = 0; 1659 1660 for (c = 0; c < rvd->vdev_children; c++) 1661 if (rvd->vdev_child[c]->vdev_islog) 1662 logalloc += rvd->vdev_child[c]->vdev_stat.vs_alloc; 1663 1664 tzb = &zcb.zcb_type[ZB_TOTAL][DMU_OT_TOTAL]; 1665 1666 if (tzb->zb_asize == alloc + logalloc) { 1667 if (!dump_opt['L']) 1668 (void) printf("\n\tNo leaks (block sum matches space" 1669 " maps exactly)\n"); 1670 } else { 1671 (void) printf("block traversal size %llu != alloc %llu " 1672 "(%s %lld)\n", 1673 (u_longlong_t)tzb->zb_asize, 1674 (u_longlong_t)alloc + logalloc, 1675 (dump_opt['L']) ? "unreachable" : "leaked", 1676 (longlong_t)(alloc + logalloc - tzb->zb_asize)); 1677 leaks = 1; 1678 } 1679 1680 if (tzb->zb_count == 0) 1681 return (2); 1682 1683 (void) printf("\n"); 1684 (void) printf("\tbp count: %10llu\n", 1685 (u_longlong_t)tzb->zb_count); 1686 (void) printf("\tbp logical: %10llu\t avg: %6llu\n", 1687 (u_longlong_t)tzb->zb_lsize, 1688 (u_longlong_t)(tzb->zb_lsize / tzb->zb_count)); 1689 (void) printf("\tbp physical: %10llu\t avg:" 1690 " %6llu\tcompression: %6.2f\n", 1691 (u_longlong_t)tzb->zb_psize, 1692 (u_longlong_t)(tzb->zb_psize / tzb->zb_count), 1693 (double)tzb->zb_lsize / tzb->zb_psize); 1694 (void) printf("\tbp allocated: %10llu\t avg:" 1695 " %6llu\tcompression: %6.2f\n", 1696 (u_longlong_t)tzb->zb_asize, 1697 (u_longlong_t)(tzb->zb_asize / tzb->zb_count), 1698 (double)tzb->zb_lsize / tzb->zb_asize); 1699 (void) printf("\tSPA allocated: %10llu\tused: %5.2f%%\n", 1700 (u_longlong_t)alloc, 100.0 * alloc / space); 1701 1702 if (dump_opt['b'] >= 2) { 1703 int l, t, level; 1704 (void) printf("\nBlocks\tLSIZE\tPSIZE\tASIZE" 1705 "\t avg\t comp\t%%Total\tType\n"); 1706 1707 for (t = 0; t <= DMU_OT_NUMTYPES; t++) { 1708 char csize[6], lsize[6], psize[6], asize[6], avg[6]; 1709 char *typename; 1710 1711 typename = t == DMU_OT_DEFERRED ? "deferred free" : 1712 t == DMU_OT_TOTAL ? "Total" : dmu_ot[t].ot_name; 1713 1714 if (zcb.zcb_type[ZB_TOTAL][t].zb_asize == 0) { 1715 (void) printf("%6s\t%5s\t%5s\t%5s" 1716 "\t%5s\t%5s\t%6s\t%s\n", 1717 "-", 1718 "-", 1719 "-", 1720 "-", 1721 "-", 1722 "-", 1723 "-", 1724 typename); 1725 continue; 1726 } 1727 1728 for (l = ZB_TOTAL - 1; l >= -1; l--) { 1729 level = (l == -1 ? ZB_TOTAL : l); 1730 zb = &zcb.zcb_type[level][t]; 1731 1732 if (zb->zb_asize == 0) 1733 continue; 1734 1735 if (dump_opt['b'] < 3 && level != ZB_TOTAL) 1736 continue; 1737 1738 if (level == 0 && zb->zb_asize == 1739 zcb.zcb_type[ZB_TOTAL][t].zb_asize) 1740 continue; 1741 1742 nicenum(zb->zb_count, csize); 1743 nicenum(zb->zb_lsize, lsize); 1744 nicenum(zb->zb_psize, psize); 1745 nicenum(zb->zb_asize, asize); 1746 nicenum(zb->zb_asize / zb->zb_count, avg); 1747 1748 (void) printf("%6s\t%5s\t%5s\t%5s\t%5s" 1749 "\t%5.2f\t%6.2f\t", 1750 csize, lsize, psize, asize, avg, 1751 (double)zb->zb_lsize / zb->zb_psize, 1752 100.0 * zb->zb_asize / tzb->zb_asize); 1753 1754 if (level == ZB_TOTAL) 1755 (void) printf("%s\n", typename); 1756 else 1757 (void) printf(" L%d %s\n", 1758 level, typename); 1759 } 1760 } 1761 } 1762 1763 (void) printf("\n"); 1764 1765 if (leaks) 1766 return (2); 1767 1768 if (zcb.zcb_haderrors) 1769 return (3); 1770 1771 return (0); 1772 } 1773 1774 static void 1775 dump_zpool(spa_t *spa) 1776 { 1777 dsl_pool_t *dp = spa_get_dsl(spa); 1778 int rc = 0; 1779 1780 if (dump_opt['u']) 1781 dump_uberblock(&spa->spa_uberblock); 1782 1783 if (dump_opt['d'] || dump_opt['i']) { 1784 dump_dir(dp->dp_meta_objset); 1785 if (dump_opt['d'] >= 3) { 1786 dump_bplist(dp->dp_meta_objset, 1787 spa->spa_sync_bplist_obj, "Deferred frees"); 1788 dump_dtl(spa->spa_root_vdev, 0); 1789 dump_metaslabs(spa); 1790 } 1791 (void) dmu_objset_find(spa_name(spa), dump_one_dir, NULL, 1792 DS_FIND_SNAPSHOTS | DS_FIND_CHILDREN); 1793 } 1794 1795 if (dump_opt['b'] || dump_opt['c'] || dump_opt['S']) 1796 rc = dump_block_stats(spa); 1797 1798 if (dump_opt['s']) 1799 show_pool_stats(spa); 1800 1801 if (rc != 0) 1802 exit(rc); 1803 } 1804 1805 #define ZDB_FLAG_CHECKSUM 0x0001 1806 #define ZDB_FLAG_DECOMPRESS 0x0002 1807 #define ZDB_FLAG_BSWAP 0x0004 1808 #define ZDB_FLAG_GBH 0x0008 1809 #define ZDB_FLAG_INDIRECT 0x0010 1810 #define ZDB_FLAG_PHYS 0x0020 1811 #define ZDB_FLAG_RAW 0x0040 1812 #define ZDB_FLAG_PRINT_BLKPTR 0x0080 1813 1814 int flagbits[256]; 1815 1816 static void 1817 zdb_print_blkptr(blkptr_t *bp, int flags) 1818 { 1819 dva_t *dva = bp->blk_dva; 1820 int d; 1821 1822 if (flags & ZDB_FLAG_BSWAP) 1823 byteswap_uint64_array((void *)bp, sizeof (blkptr_t)); 1824 /* 1825 * Super-ick warning: This code is also duplicated in 1826 * cmd/mdb/common/modules/zfs/zfs.c . Yeah, I hate code 1827 * replication, too. 1828 */ 1829 for (d = 0; d < BP_GET_NDVAS(bp); d++) { 1830 (void) printf("\tDVA[%d]: vdev_id %lld / %llx\n", d, 1831 (longlong_t)DVA_GET_VDEV(&dva[d]), 1832 (longlong_t)DVA_GET_OFFSET(&dva[d])); 1833 (void) printf("\tDVA[%d]: GANG: %-5s GRID: %04llx\t" 1834 "ASIZE: %llx\n", d, 1835 DVA_GET_GANG(&dva[d]) ? "TRUE" : "FALSE", 1836 (longlong_t)DVA_GET_GRID(&dva[d]), 1837 (longlong_t)DVA_GET_ASIZE(&dva[d])); 1838 (void) printf("\tDVA[%d]: :%llu:%llx:%llx:%s%s%s%s\n", d, 1839 (u_longlong_t)DVA_GET_VDEV(&dva[d]), 1840 (longlong_t)DVA_GET_OFFSET(&dva[d]), 1841 (longlong_t)BP_GET_PSIZE(bp), 1842 BP_SHOULD_BYTESWAP(bp) ? "e" : "", 1843 !DVA_GET_GANG(&dva[d]) && BP_GET_LEVEL(bp) != 0 ? 1844 "d" : "", 1845 DVA_GET_GANG(&dva[d]) ? "g" : "", 1846 BP_GET_COMPRESS(bp) != 0 ? "d" : ""); 1847 } 1848 (void) printf("\tLSIZE: %-16llx\t\tPSIZE: %llx\n", 1849 (longlong_t)BP_GET_LSIZE(bp), (longlong_t)BP_GET_PSIZE(bp)); 1850 (void) printf("\tENDIAN: %6s\t\t\t\t\tTYPE: %s\n", 1851 BP_GET_BYTEORDER(bp) ? "LITTLE" : "BIG", 1852 dmu_ot[BP_GET_TYPE(bp)].ot_name); 1853 (void) printf("\tBIRTH: %-16llx LEVEL: %-2llu\tFILL: %llx\n", 1854 (u_longlong_t)bp->blk_birth, (u_longlong_t)BP_GET_LEVEL(bp), 1855 (u_longlong_t)bp->blk_fill); 1856 (void) printf("\tCKFUNC: %-16s\t\tCOMP: %s\n", 1857 zio_checksum_table[BP_GET_CHECKSUM(bp)].ci_name, 1858 zio_compress_table[BP_GET_COMPRESS(bp)].ci_name); 1859 (void) printf("\tCKSUM: %llx:%llx:%llx:%llx\n", 1860 (u_longlong_t)bp->blk_cksum.zc_word[0], 1861 (u_longlong_t)bp->blk_cksum.zc_word[1], 1862 (u_longlong_t)bp->blk_cksum.zc_word[2], 1863 (u_longlong_t)bp->blk_cksum.zc_word[3]); 1864 } 1865 1866 static void 1867 zdb_dump_indirect(blkptr_t *bp, int nbps, int flags) 1868 { 1869 int i; 1870 1871 for (i = 0; i < nbps; i++) 1872 zdb_print_blkptr(&bp[i], flags); 1873 } 1874 1875 static void 1876 zdb_dump_gbh(void *buf, int flags) 1877 { 1878 zdb_dump_indirect((blkptr_t *)buf, SPA_GBH_NBLKPTRS, flags); 1879 } 1880 1881 static void 1882 zdb_dump_block_raw(void *buf, uint64_t size, int flags) 1883 { 1884 if (flags & ZDB_FLAG_BSWAP) 1885 byteswap_uint64_array(buf, size); 1886 (void) write(2, buf, size); 1887 } 1888 1889 static void 1890 zdb_dump_block(char *label, void *buf, uint64_t size, int flags) 1891 { 1892 uint64_t *d = (uint64_t *)buf; 1893 int nwords = size / sizeof (uint64_t); 1894 int do_bswap = !!(flags & ZDB_FLAG_BSWAP); 1895 int i, j; 1896 char *hdr, *c; 1897 1898 1899 if (do_bswap) 1900 hdr = " 7 6 5 4 3 2 1 0 f e d c b a 9 8"; 1901 else 1902 hdr = " 0 1 2 3 4 5 6 7 8 9 a b c d e f"; 1903 1904 (void) printf("\n%s\n%6s %s 0123456789abcdef\n", label, "", hdr); 1905 1906 for (i = 0; i < nwords; i += 2) { 1907 (void) printf("%06llx: %016llx %016llx ", 1908 (u_longlong_t)(i * sizeof (uint64_t)), 1909 (u_longlong_t)(do_bswap ? BSWAP_64(d[i]) : d[i]), 1910 (u_longlong_t)(do_bswap ? BSWAP_64(d[i + 1]) : d[i + 1])); 1911 1912 c = (char *)&d[i]; 1913 for (j = 0; j < 2 * sizeof (uint64_t); j++) 1914 (void) printf("%c", isprint(c[j]) ? c[j] : '.'); 1915 (void) printf("\n"); 1916 } 1917 } 1918 1919 /* 1920 * There are two acceptable formats: 1921 * leaf_name - For example: c1t0d0 or /tmp/ztest.0a 1922 * child[.child]* - For example: 0.1.1 1923 * 1924 * The second form can be used to specify arbitrary vdevs anywhere 1925 * in the heirarchy. For example, in a pool with a mirror of 1926 * RAID-Zs, you can specify either RAID-Z vdev with 0.0 or 0.1 . 1927 */ 1928 static vdev_t * 1929 zdb_vdev_lookup(vdev_t *vdev, char *path) 1930 { 1931 char *s, *p, *q; 1932 int i; 1933 1934 if (vdev == NULL) 1935 return (NULL); 1936 1937 /* First, assume the x.x.x.x format */ 1938 i = (int)strtoul(path, &s, 10); 1939 if (s == path || (s && *s != '.' && *s != '\0')) 1940 goto name; 1941 if (i < 0 || i >= vdev->vdev_children) 1942 return (NULL); 1943 1944 vdev = vdev->vdev_child[i]; 1945 if (*s == '\0') 1946 return (vdev); 1947 return (zdb_vdev_lookup(vdev, s+1)); 1948 1949 name: 1950 for (i = 0; i < vdev->vdev_children; i++) { 1951 vdev_t *vc = vdev->vdev_child[i]; 1952 1953 if (vc->vdev_path == NULL) { 1954 vc = zdb_vdev_lookup(vc, path); 1955 if (vc == NULL) 1956 continue; 1957 else 1958 return (vc); 1959 } 1960 1961 p = strrchr(vc->vdev_path, '/'); 1962 p = p ? p + 1 : vc->vdev_path; 1963 q = &vc->vdev_path[strlen(vc->vdev_path) - 2]; 1964 1965 if (strcmp(vc->vdev_path, path) == 0) 1966 return (vc); 1967 if (strcmp(p, path) == 0) 1968 return (vc); 1969 if (strcmp(q, "s0") == 0 && strncmp(p, path, q - p) == 0) 1970 return (vc); 1971 } 1972 1973 return (NULL); 1974 } 1975 1976 /* 1977 * Read a block from a pool and print it out. The syntax of the 1978 * block descriptor is: 1979 * 1980 * pool:vdev_specifier:offset:size[:flags] 1981 * 1982 * pool - The name of the pool you wish to read from 1983 * vdev_specifier - Which vdev (see comment for zdb_vdev_lookup) 1984 * offset - offset, in hex, in bytes 1985 * size - Amount of data to read, in hex, in bytes 1986 * flags - A string of characters specifying options 1987 * b: Decode a blkptr at given offset within block 1988 * *c: Calculate and display checksums 1989 * *d: Decompress data before dumping 1990 * e: Byteswap data before dumping 1991 * *g: Display data as a gang block header 1992 * *i: Display as an indirect block 1993 * p: Do I/O to physical offset 1994 * r: Dump raw data to stdout 1995 * 1996 * * = not yet implemented 1997 */ 1998 static void 1999 zdb_read_block(char *thing, spa_t **spap) 2000 { 2001 spa_t *spa = *spap; 2002 int flags = 0; 2003 uint64_t offset = 0, size = 0, blkptr_offset = 0; 2004 zio_t *zio; 2005 vdev_t *vd; 2006 void *buf; 2007 char *s, *p, *dup, *pool, *vdev, *flagstr; 2008 int i, error, zio_flags; 2009 2010 dup = strdup(thing); 2011 s = strtok(dup, ":"); 2012 pool = s ? s : ""; 2013 s = strtok(NULL, ":"); 2014 vdev = s ? s : ""; 2015 s = strtok(NULL, ":"); 2016 offset = strtoull(s ? s : "", NULL, 16); 2017 s = strtok(NULL, ":"); 2018 size = strtoull(s ? s : "", NULL, 16); 2019 s = strtok(NULL, ":"); 2020 flagstr = s ? s : ""; 2021 2022 s = NULL; 2023 if (size == 0) 2024 s = "size must not be zero"; 2025 if (!IS_P2ALIGNED(size, DEV_BSIZE)) 2026 s = "size must be a multiple of sector size"; 2027 if (!IS_P2ALIGNED(offset, DEV_BSIZE)) 2028 s = "offset must be a multiple of sector size"; 2029 if (s) { 2030 (void) printf("Invalid block specifier: %s - %s\n", thing, s); 2031 free(dup); 2032 return; 2033 } 2034 2035 for (s = strtok(flagstr, ":"); s; s = strtok(NULL, ":")) { 2036 for (i = 0; flagstr[i]; i++) { 2037 int bit = flagbits[(uchar_t)flagstr[i]]; 2038 2039 if (bit == 0) { 2040 (void) printf("***Invalid flag: %c\n", 2041 flagstr[i]); 2042 continue; 2043 } 2044 flags |= bit; 2045 2046 /* If it's not something with an argument, keep going */ 2047 if ((bit & (ZDB_FLAG_CHECKSUM | ZDB_FLAG_DECOMPRESS | 2048 ZDB_FLAG_PRINT_BLKPTR)) == 0) 2049 continue; 2050 2051 p = &flagstr[i + 1]; 2052 if (bit == ZDB_FLAG_PRINT_BLKPTR) 2053 blkptr_offset = strtoull(p, &p, 16); 2054 if (*p != ':' && *p != '\0') { 2055 (void) printf("***Invalid flag arg: '%s'\n", s); 2056 free(dup); 2057 return; 2058 } 2059 } 2060 } 2061 2062 if (spa == NULL || strcmp(spa_name(spa), pool) != 0) { 2063 if (spa) 2064 spa_close(spa, (void *)zdb_read_block); 2065 error = spa_open(pool, spap, (void *)zdb_read_block); 2066 if (error) 2067 fatal("Failed to open pool '%s': %s", 2068 pool, strerror(error)); 2069 spa = *spap; 2070 } 2071 2072 vd = zdb_vdev_lookup(spa->spa_root_vdev, vdev); 2073 if (vd == NULL) { 2074 (void) printf("***Invalid vdev: %s\n", vdev); 2075 free(dup); 2076 return; 2077 } else { 2078 if (vd->vdev_path) 2079 (void) printf("Found vdev: %s\n", vd->vdev_path); 2080 else 2081 (void) printf("Found vdev type: %s\n", 2082 vd->vdev_ops->vdev_op_type); 2083 } 2084 2085 buf = umem_alloc(size, UMEM_NOFAIL); 2086 2087 zio_flags = ZIO_FLAG_DONT_CACHE | ZIO_FLAG_DONT_QUEUE | 2088 ZIO_FLAG_DONT_PROPAGATE | ZIO_FLAG_DONT_RETRY; 2089 2090 spa_config_enter(spa, SCL_STATE, FTAG, RW_READER); 2091 zio = zio_root(spa, NULL, NULL, 0); 2092 /* XXX todo - cons up a BP so RAID-Z will be happy */ 2093 zio_nowait(zio_vdev_child_io(zio, NULL, vd, offset, buf, size, 2094 ZIO_TYPE_READ, ZIO_PRIORITY_SYNC_READ, zio_flags, NULL, NULL)); 2095 error = zio_wait(zio); 2096 spa_config_exit(spa, SCL_STATE, FTAG); 2097 2098 if (error) { 2099 (void) printf("Read of %s failed, error: %d\n", thing, error); 2100 goto out; 2101 } 2102 2103 if (flags & ZDB_FLAG_PRINT_BLKPTR) 2104 zdb_print_blkptr((blkptr_t *)(void *) 2105 ((uintptr_t)buf + (uintptr_t)blkptr_offset), flags); 2106 else if (flags & ZDB_FLAG_RAW) 2107 zdb_dump_block_raw(buf, size, flags); 2108 else if (flags & ZDB_FLAG_INDIRECT) 2109 zdb_dump_indirect((blkptr_t *)buf, size / sizeof (blkptr_t), 2110 flags); 2111 else if (flags & ZDB_FLAG_GBH) 2112 zdb_dump_gbh(buf, flags); 2113 else 2114 zdb_dump_block(thing, buf, size, flags); 2115 2116 out: 2117 umem_free(buf, size); 2118 free(dup); 2119 } 2120 2121 static boolean_t 2122 nvlist_string_match(nvlist_t *config, char *name, char *tgt) 2123 { 2124 char *s; 2125 2126 if (nvlist_lookup_string(config, name, &s) != 0) 2127 return (B_FALSE); 2128 2129 return (strcmp(s, tgt) == 0); 2130 } 2131 2132 static boolean_t 2133 nvlist_uint64_match(nvlist_t *config, char *name, uint64_t tgt) 2134 { 2135 uint64_t val; 2136 2137 if (nvlist_lookup_uint64(config, name, &val) != 0) 2138 return (B_FALSE); 2139 2140 return (val == tgt); 2141 } 2142 2143 static boolean_t 2144 vdev_child_guid_match(nvlist_t *vdev, uint64_t guid) 2145 { 2146 nvlist_t **child; 2147 uint_t c, children; 2148 2149 verify(nvlist_lookup_nvlist_array(vdev, ZPOOL_CONFIG_CHILDREN, 2150 &child, &children) == 0); 2151 for (c = 0; c < children; ++c) 2152 if (nvlist_uint64_match(child[c], ZPOOL_CONFIG_GUID, guid)) 2153 return (B_TRUE); 2154 return (B_FALSE); 2155 } 2156 2157 static boolean_t 2158 vdev_child_string_match(nvlist_t *vdev, char *tgt) 2159 { 2160 nvlist_t **child; 2161 uint_t c, children; 2162 2163 verify(nvlist_lookup_nvlist_array(vdev, ZPOOL_CONFIG_CHILDREN, 2164 &child, &children) == 0); 2165 for (c = 0; c < children; ++c) { 2166 if (nvlist_string_match(child[c], ZPOOL_CONFIG_PATH, tgt) || 2167 nvlist_string_match(child[c], ZPOOL_CONFIG_DEVID, tgt)) 2168 return (B_TRUE); 2169 } 2170 return (B_FALSE); 2171 } 2172 2173 static boolean_t 2174 vdev_guid_match(nvlist_t *config, uint64_t guid) 2175 { 2176 nvlist_t *nvroot; 2177 2178 verify(nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE, 2179 &nvroot) == 0); 2180 2181 return (nvlist_uint64_match(nvroot, ZPOOL_CONFIG_GUID, guid) || 2182 vdev_child_guid_match(nvroot, guid)); 2183 } 2184 2185 static boolean_t 2186 vdev_string_match(nvlist_t *config, char *tgt) 2187 { 2188 nvlist_t *nvroot; 2189 2190 verify(nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE, 2191 &nvroot) == 0); 2192 2193 return (vdev_child_string_match(nvroot, tgt)); 2194 } 2195 2196 static boolean_t 2197 pool_match(nvlist_t *config, char *tgt) 2198 { 2199 uint64_t guid = strtoull(tgt, NULL, 0); 2200 2201 if (guid != 0) { 2202 return ( 2203 nvlist_uint64_match(config, ZPOOL_CONFIG_POOL_GUID, guid) || 2204 vdev_guid_match(config, guid)); 2205 } else { 2206 return ( 2207 nvlist_string_match(config, ZPOOL_CONFIG_POOL_NAME, tgt) || 2208 vdev_string_match(config, tgt)); 2209 } 2210 } 2211 2212 static int 2213 find_exported_zpool(char *pool_id, nvlist_t **configp, char *vdev_dir) 2214 { 2215 nvlist_t *pools; 2216 int error = ENOENT; 2217 nvlist_t *match = NULL; 2218 2219 if (vdev_dir != NULL) 2220 pools = zpool_find_import_activeok(g_zfs, 1, &vdev_dir); 2221 else 2222 pools = zpool_find_import_activeok(g_zfs, 0, NULL); 2223 2224 if (pools != NULL) { 2225 nvpair_t *elem = NULL; 2226 2227 while ((elem = nvlist_next_nvpair(pools, elem)) != NULL) { 2228 verify(nvpair_value_nvlist(elem, configp) == 0); 2229 if (pool_match(*configp, pool_id)) { 2230 if (match != NULL) { 2231 (void) fatal( 2232 "More than one matching pool - " 2233 "specify guid/devid/device path."); 2234 } else { 2235 match = *configp; 2236 error = 0; 2237 } 2238 } 2239 } 2240 } 2241 2242 *configp = error ? NULL : match; 2243 2244 return (error); 2245 } 2246 2247 int 2248 main(int argc, char **argv) 2249 { 2250 int i, c; 2251 struct rlimit rl = { 1024, 1024 }; 2252 spa_t *spa; 2253 objset_t *os = NULL; 2254 char *endstr; 2255 int dump_all = 1; 2256 int verbose = 0; 2257 int error; 2258 int exported = 0; 2259 char *vdev_dir = NULL; 2260 2261 (void) setrlimit(RLIMIT_NOFILE, &rl); 2262 (void) enable_extended_FILE_stdio(-1, -1); 2263 2264 dprintf_setup(&argc, argv); 2265 2266 while ((c = getopt(argc, argv, "udibcsvCLS:U:lRep:t:")) != -1) { 2267 switch (c) { 2268 case 'u': 2269 case 'd': 2270 case 'i': 2271 case 'b': 2272 case 'c': 2273 case 's': 2274 case 'C': 2275 case 'l': 2276 case 'R': 2277 dump_opt[c]++; 2278 dump_all = 0; 2279 break; 2280 case 'L': 2281 dump_opt[c]++; 2282 break; 2283 case 'v': 2284 verbose++; 2285 break; 2286 case 'U': 2287 spa_config_path = optarg; 2288 break; 2289 case 'e': 2290 exported = 1; 2291 break; 2292 case 'p': 2293 vdev_dir = optarg; 2294 break; 2295 case 'S': 2296 dump_opt[c]++; 2297 dump_all = 0; 2298 zdb_sig_user_data = (strncmp(optarg, "user:", 5) == 0); 2299 if (!zdb_sig_user_data && strncmp(optarg, "all:", 4)) 2300 usage(); 2301 endstr = strchr(optarg, ':') + 1; 2302 if (strcmp(endstr, "fletcher2") == 0) 2303 zdb_sig_cksumalg = ZIO_CHECKSUM_FLETCHER_2; 2304 else if (strcmp(endstr, "fletcher4") == 0) 2305 zdb_sig_cksumalg = ZIO_CHECKSUM_FLETCHER_4; 2306 else if (strcmp(endstr, "sha256") == 0) 2307 zdb_sig_cksumalg = ZIO_CHECKSUM_SHA256; 2308 else if (strcmp(endstr, "all") == 0) 2309 zdb_sig_cksumalg = ZIO_CHECKSUM_FLETCHER_2; 2310 else 2311 usage(); 2312 break; 2313 case 't': 2314 ub_max_txg = strtoull(optarg, NULL, 0); 2315 if (ub_max_txg < TXG_INITIAL) { 2316 (void) fprintf(stderr, "incorrect txg " 2317 "specified: %s\n", optarg); 2318 usage(); 2319 } 2320 break; 2321 default: 2322 usage(); 2323 break; 2324 } 2325 } 2326 2327 if (vdev_dir != NULL && exported == 0) { 2328 (void) fprintf(stderr, "-p option requires use of -e\n"); 2329 usage(); 2330 } 2331 2332 kernel_init(FREAD); 2333 g_zfs = libzfs_init(); 2334 ASSERT(g_zfs != NULL); 2335 2336 for (c = 0; c < 256; c++) { 2337 if (dump_all && c != 'l' && c != 'R') 2338 dump_opt[c] = 1; 2339 if (dump_opt[c]) 2340 dump_opt[c] += verbose; 2341 } 2342 2343 argc -= optind; 2344 argv += optind; 2345 2346 if (argc < 1) { 2347 if (dump_opt['C']) { 2348 dump_cachefile(spa_config_path); 2349 return (0); 2350 } 2351 usage(); 2352 } 2353 2354 if (dump_opt['l']) { 2355 dump_label(argv[0]); 2356 return (0); 2357 } 2358 2359 if (dump_opt['R']) { 2360 flagbits['b'] = ZDB_FLAG_PRINT_BLKPTR; 2361 flagbits['c'] = ZDB_FLAG_CHECKSUM; 2362 flagbits['d'] = ZDB_FLAG_DECOMPRESS; 2363 flagbits['e'] = ZDB_FLAG_BSWAP; 2364 flagbits['g'] = ZDB_FLAG_GBH; 2365 flagbits['i'] = ZDB_FLAG_INDIRECT; 2366 flagbits['p'] = ZDB_FLAG_PHYS; 2367 flagbits['r'] = ZDB_FLAG_RAW; 2368 2369 spa = NULL; 2370 while (argv[0]) { 2371 zdb_read_block(argv[0], &spa); 2372 argv++; 2373 argc--; 2374 } 2375 if (spa) 2376 spa_close(spa, (void *)zdb_read_block); 2377 return (0); 2378 } 2379 2380 if (dump_opt['C']) 2381 dump_config(argv[0]); 2382 2383 error = 0; 2384 if (exported) { 2385 /* 2386 * Check to see if the name refers to an exported zpool 2387 */ 2388 char *slash; 2389 nvlist_t *exported_conf = NULL; 2390 2391 if ((slash = strchr(argv[0], '/')) != NULL) 2392 *slash = '\0'; 2393 2394 error = find_exported_zpool(argv[0], &exported_conf, vdev_dir); 2395 if (error == 0) { 2396 nvlist_t *nvl = NULL; 2397 2398 if (vdev_dir != NULL) { 2399 if (nvlist_alloc(&nvl, NV_UNIQUE_NAME, 0) != 0) 2400 error = ENOMEM; 2401 else if (nvlist_add_string(nvl, 2402 zpool_prop_to_name(ZPOOL_PROP_ALTROOT), 2403 vdev_dir) != 0) 2404 error = ENOMEM; 2405 } 2406 2407 if (error == 0) 2408 error = spa_import_faulted(argv[0], 2409 exported_conf, nvl); 2410 2411 nvlist_free(nvl); 2412 } 2413 2414 if (slash != NULL) 2415 *slash = '/'; 2416 } 2417 2418 if (error == 0) { 2419 if (strchr(argv[0], '/') != NULL) { 2420 error = dmu_objset_open(argv[0], DMU_OST_ANY, 2421 DS_MODE_USER | DS_MODE_READONLY, &os); 2422 } else { 2423 error = spa_open(argv[0], &spa, FTAG); 2424 } 2425 } 2426 2427 if (error) 2428 fatal("can't open %s: %s", argv[0], strerror(error)); 2429 2430 argv++; 2431 if (--argc > 0) { 2432 zopt_objects = argc; 2433 zopt_object = calloc(zopt_objects, sizeof (uint64_t)); 2434 for (i = 0; i < zopt_objects; i++) { 2435 errno = 0; 2436 zopt_object[i] = strtoull(argv[i], NULL, 0); 2437 if (zopt_object[i] == 0 && errno != 0) 2438 fatal("bad object number %s: %s", 2439 argv[i], strerror(errno)); 2440 } 2441 } 2442 2443 if (os != NULL) { 2444 dump_dir(os); 2445 dmu_objset_close(os); 2446 } else { 2447 dump_zpool(spa); 2448 spa_close(spa, FTAG); 2449 } 2450 2451 fuid_table_destroy(); 2452 2453 libzfs_fini(g_zfs); 2454 kernel_fini(); 2455 2456 return (0); 2457 } 2458