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, " -m metaslabs\n"); 106 (void) fprintf(stderr, " -c checksum all metadata (twice for " 107 "all data) blocks\n"); 108 (void) fprintf(stderr, " -s report stats on zdb's I/O\n"); 109 (void) fprintf(stderr, " -S <user|all>:<cksum_alg|all> -- " 110 "dump blkptr signatures\n"); 111 (void) fprintf(stderr, " -v verbose (applies to all others)\n"); 112 (void) fprintf(stderr, " -l dump label contents\n"); 113 (void) fprintf(stderr, " -L disable leak tracking (do not " 114 "load spacemaps)\n"); 115 (void) fprintf(stderr, " -U cachefile_path -- use alternate " 116 "cachefile\n"); 117 (void) fprintf(stderr, " -R read and display block from a " 118 "device\n"); 119 (void) fprintf(stderr, " -e Pool is exported/destroyed/" 120 "has altroot\n"); 121 (void) fprintf(stderr, " -p <Path to vdev dir> (use with -e)\n"); 122 (void) fprintf(stderr, " -t <txg> highest txg to use when " 123 "searching for uberblocks\n"); 124 (void) fprintf(stderr, "Specify an option more than once (e.g. -bb) " 125 "to make only that option verbose\n"); 126 (void) fprintf(stderr, "Default is to dump everything non-verbosely\n"); 127 exit(1); 128 } 129 130 /* 131 * Called for usage errors that are discovered after a call to spa_open(), 132 * dmu_bonus_hold(), or pool_match(). abort() is called for other errors. 133 */ 134 135 static void 136 fatal(const char *fmt, ...) 137 { 138 va_list ap; 139 140 va_start(ap, fmt); 141 (void) fprintf(stderr, "%s: ", cmdname); 142 (void) vfprintf(stderr, fmt, ap); 143 va_end(ap); 144 (void) fprintf(stderr, "\n"); 145 146 exit(1); 147 } 148 149 /* ARGSUSED */ 150 static void 151 dump_packed_nvlist(objset_t *os, uint64_t object, void *data, size_t size) 152 { 153 nvlist_t *nv; 154 size_t nvsize = *(uint64_t *)data; 155 char *packed = umem_alloc(nvsize, UMEM_NOFAIL); 156 157 VERIFY(0 == dmu_read(os, object, 0, nvsize, packed, DMU_READ_PREFETCH)); 158 159 VERIFY(nvlist_unpack(packed, nvsize, &nv, 0) == 0); 160 161 umem_free(packed, nvsize); 162 163 dump_nvlist(nv, 8); 164 165 nvlist_free(nv); 166 } 167 168 const char dump_zap_stars[] = "****************************************"; 169 const int dump_zap_width = sizeof (dump_zap_stars) - 1; 170 171 static void 172 dump_zap_histogram(uint64_t histo[ZAP_HISTOGRAM_SIZE]) 173 { 174 int i; 175 int minidx = ZAP_HISTOGRAM_SIZE - 1; 176 int maxidx = 0; 177 uint64_t max = 0; 178 179 for (i = 0; i < ZAP_HISTOGRAM_SIZE; i++) { 180 if (histo[i] > max) 181 max = histo[i]; 182 if (histo[i] > 0 && i > maxidx) 183 maxidx = i; 184 if (histo[i] > 0 && i < minidx) 185 minidx = i; 186 } 187 188 if (max < dump_zap_width) 189 max = dump_zap_width; 190 191 for (i = minidx; i <= maxidx; i++) 192 (void) printf("\t\t\t%u: %6llu %s\n", i, (u_longlong_t)histo[i], 193 &dump_zap_stars[(max - histo[i]) * dump_zap_width / max]); 194 } 195 196 static void 197 dump_zap_stats(objset_t *os, uint64_t object) 198 { 199 int error; 200 zap_stats_t zs; 201 202 error = zap_get_stats(os, object, &zs); 203 if (error) 204 return; 205 206 if (zs.zs_ptrtbl_len == 0) { 207 ASSERT(zs.zs_num_blocks == 1); 208 (void) printf("\tmicrozap: %llu bytes, %llu entries\n", 209 (u_longlong_t)zs.zs_blocksize, 210 (u_longlong_t)zs.zs_num_entries); 211 return; 212 } 213 214 (void) printf("\tFat ZAP stats:\n"); 215 216 (void) printf("\t\tPointer table:\n"); 217 (void) printf("\t\t\t%llu elements\n", 218 (u_longlong_t)zs.zs_ptrtbl_len); 219 (void) printf("\t\t\tzt_blk: %llu\n", 220 (u_longlong_t)zs.zs_ptrtbl_zt_blk); 221 (void) printf("\t\t\tzt_numblks: %llu\n", 222 (u_longlong_t)zs.zs_ptrtbl_zt_numblks); 223 (void) printf("\t\t\tzt_shift: %llu\n", 224 (u_longlong_t)zs.zs_ptrtbl_zt_shift); 225 (void) printf("\t\t\tzt_blks_copied: %llu\n", 226 (u_longlong_t)zs.zs_ptrtbl_blks_copied); 227 (void) printf("\t\t\tzt_nextblk: %llu\n", 228 (u_longlong_t)zs.zs_ptrtbl_nextblk); 229 230 (void) printf("\t\tZAP entries: %llu\n", 231 (u_longlong_t)zs.zs_num_entries); 232 (void) printf("\t\tLeaf blocks: %llu\n", 233 (u_longlong_t)zs.zs_num_leafs); 234 (void) printf("\t\tTotal blocks: %llu\n", 235 (u_longlong_t)zs.zs_num_blocks); 236 (void) printf("\t\tzap_block_type: 0x%llx\n", 237 (u_longlong_t)zs.zs_block_type); 238 (void) printf("\t\tzap_magic: 0x%llx\n", 239 (u_longlong_t)zs.zs_magic); 240 (void) printf("\t\tzap_salt: 0x%llx\n", 241 (u_longlong_t)zs.zs_salt); 242 243 (void) printf("\t\tLeafs with 2^n pointers:\n"); 244 dump_zap_histogram(zs.zs_leafs_with_2n_pointers); 245 246 (void) printf("\t\tBlocks with n*5 entries:\n"); 247 dump_zap_histogram(zs.zs_blocks_with_n5_entries); 248 249 (void) printf("\t\tBlocks n/10 full:\n"); 250 dump_zap_histogram(zs.zs_blocks_n_tenths_full); 251 252 (void) printf("\t\tEntries with n chunks:\n"); 253 dump_zap_histogram(zs.zs_entries_using_n_chunks); 254 255 (void) printf("\t\tBuckets with n entries:\n"); 256 dump_zap_histogram(zs.zs_buckets_with_n_entries); 257 } 258 259 /*ARGSUSED*/ 260 static void 261 dump_none(objset_t *os, uint64_t object, void *data, size_t size) 262 { 263 } 264 265 /*ARGSUSED*/ 266 void 267 dump_uint8(objset_t *os, uint64_t object, void *data, size_t size) 268 { 269 } 270 271 /*ARGSUSED*/ 272 static void 273 dump_uint64(objset_t *os, uint64_t object, void *data, size_t size) 274 { 275 } 276 277 /*ARGSUSED*/ 278 static void 279 dump_zap(objset_t *os, uint64_t object, void *data, size_t size) 280 { 281 zap_cursor_t zc; 282 zap_attribute_t attr; 283 void *prop; 284 int i; 285 286 dump_zap_stats(os, object); 287 (void) printf("\n"); 288 289 for (zap_cursor_init(&zc, os, object); 290 zap_cursor_retrieve(&zc, &attr) == 0; 291 zap_cursor_advance(&zc)) { 292 (void) printf("\t\t%s = ", attr.za_name); 293 if (attr.za_num_integers == 0) { 294 (void) printf("\n"); 295 continue; 296 } 297 prop = umem_zalloc(attr.za_num_integers * 298 attr.za_integer_length, UMEM_NOFAIL); 299 (void) zap_lookup(os, object, attr.za_name, 300 attr.za_integer_length, attr.za_num_integers, prop); 301 if (attr.za_integer_length == 1) { 302 (void) printf("%s", (char *)prop); 303 } else { 304 for (i = 0; i < attr.za_num_integers; i++) { 305 switch (attr.za_integer_length) { 306 case 2: 307 (void) printf("%u ", 308 ((uint16_t *)prop)[i]); 309 break; 310 case 4: 311 (void) printf("%u ", 312 ((uint32_t *)prop)[i]); 313 break; 314 case 8: 315 (void) printf("%lld ", 316 (u_longlong_t)((int64_t *)prop)[i]); 317 break; 318 } 319 } 320 } 321 (void) printf("\n"); 322 umem_free(prop, attr.za_num_integers * attr.za_integer_length); 323 } 324 zap_cursor_fini(&zc); 325 } 326 327 /*ARGSUSED*/ 328 static void 329 dump_zpldir(objset_t *os, uint64_t object, void *data, size_t size) 330 { 331 zap_cursor_t zc; 332 zap_attribute_t attr; 333 const char *typenames[] = { 334 /* 0 */ "not specified", 335 /* 1 */ "FIFO", 336 /* 2 */ "Character Device", 337 /* 3 */ "3 (invalid)", 338 /* 4 */ "Directory", 339 /* 5 */ "5 (invalid)", 340 /* 6 */ "Block Device", 341 /* 7 */ "7 (invalid)", 342 /* 8 */ "Regular File", 343 /* 9 */ "9 (invalid)", 344 /* 10 */ "Symbolic Link", 345 /* 11 */ "11 (invalid)", 346 /* 12 */ "Socket", 347 /* 13 */ "Door", 348 /* 14 */ "Event Port", 349 /* 15 */ "15 (invalid)", 350 }; 351 352 dump_zap_stats(os, object); 353 (void) printf("\n"); 354 355 for (zap_cursor_init(&zc, os, object); 356 zap_cursor_retrieve(&zc, &attr) == 0; 357 zap_cursor_advance(&zc)) { 358 (void) printf("\t\t%s = %lld (type: %s)\n", 359 attr.za_name, ZFS_DIRENT_OBJ(attr.za_first_integer), 360 typenames[ZFS_DIRENT_TYPE(attr.za_first_integer)]); 361 } 362 zap_cursor_fini(&zc); 363 } 364 365 static void 366 dump_spacemap(objset_t *os, space_map_obj_t *smo, space_map_t *sm) 367 { 368 uint64_t alloc, offset, entry; 369 uint8_t mapshift = sm->sm_shift; 370 uint64_t mapstart = sm->sm_start; 371 char *ddata[] = { "ALLOC", "FREE", "CONDENSE", "INVALID", 372 "INVALID", "INVALID", "INVALID", "INVALID" }; 373 374 if (smo->smo_object == 0) 375 return; 376 377 /* 378 * Print out the freelist entries in both encoded and decoded form. 379 */ 380 alloc = 0; 381 for (offset = 0; offset < smo->smo_objsize; offset += sizeof (entry)) { 382 VERIFY(0 == dmu_read(os, smo->smo_object, offset, 383 sizeof (entry), &entry, DMU_READ_PREFETCH)); 384 if (SM_DEBUG_DECODE(entry)) { 385 (void) printf("\t\t[%4llu] %s: txg %llu, pass %llu\n", 386 (u_longlong_t)(offset / sizeof (entry)), 387 ddata[SM_DEBUG_ACTION_DECODE(entry)], 388 (u_longlong_t)SM_DEBUG_TXG_DECODE(entry), 389 (u_longlong_t)SM_DEBUG_SYNCPASS_DECODE(entry)); 390 } else { 391 (void) printf("\t\t[%4llu] %c range:" 392 " %08llx-%08llx size: %06llx\n", 393 (u_longlong_t)(offset / sizeof (entry)), 394 SM_TYPE_DECODE(entry) == SM_ALLOC ? 'A' : 'F', 395 (u_longlong_t)((SM_OFFSET_DECODE(entry) << 396 mapshift) + mapstart), 397 (u_longlong_t)((SM_OFFSET_DECODE(entry) << 398 mapshift) + mapstart + (SM_RUN_DECODE(entry) << 399 mapshift)), 400 (u_longlong_t)(SM_RUN_DECODE(entry) << mapshift)); 401 if (SM_TYPE_DECODE(entry) == SM_ALLOC) 402 alloc += SM_RUN_DECODE(entry) << mapshift; 403 else 404 alloc -= SM_RUN_DECODE(entry) << mapshift; 405 } 406 } 407 if (alloc != smo->smo_alloc) { 408 (void) printf("space_map_object alloc (%llu) INCONSISTENT " 409 "with space map summary (%llu)\n", 410 (u_longlong_t)smo->smo_alloc, (u_longlong_t)alloc); 411 } 412 } 413 414 static void 415 dump_metaslab_stats(metaslab_t *msp) 416 { 417 char maxbuf[5]; 418 space_map_t *sm = &msp->ms_map; 419 avl_tree_t *t = sm->sm_pp_root; 420 int free_pct = sm->sm_space * 100 / sm->sm_size; 421 422 nicenum(space_map_maxsize(sm), maxbuf); 423 424 (void) printf("\t %20s %10lu %7s %6s %4s %4d%%\n", 425 "segments", avl_numnodes(t), "maxsize", maxbuf, 426 "freepct", free_pct); 427 } 428 429 static void 430 dump_metaslab(metaslab_t *msp) 431 { 432 char freebuf[5]; 433 space_map_obj_t *smo = &msp->ms_smo; 434 vdev_t *vd = msp->ms_group->mg_vd; 435 spa_t *spa = vd->vdev_spa; 436 437 nicenum(msp->ms_map.sm_size - smo->smo_alloc, freebuf); 438 439 (void) printf( 440 "\tvdev %5llu offset %12llx spacemap %6llu free %5s\n", 441 (u_longlong_t)vd->vdev_id, (u_longlong_t)msp->ms_map.sm_start, 442 (u_longlong_t)smo->smo_object, freebuf); 443 444 if (dump_opt['m'] > 1) { 445 mutex_enter(&msp->ms_lock); 446 VERIFY(space_map_load(&msp->ms_map, zfs_metaslab_ops, 447 SM_FREE, &msp->ms_smo, spa->spa_meta_objset) == 0); 448 dump_metaslab_stats(msp); 449 space_map_unload(&msp->ms_map); 450 mutex_exit(&msp->ms_lock); 451 } 452 453 if (dump_opt['d'] > 5 || dump_opt['m'] > 2) { 454 ASSERT(msp->ms_map.sm_size == (1ULL << vd->vdev_ms_shift)); 455 456 mutex_enter(&msp->ms_lock); 457 dump_spacemap(spa->spa_meta_objset, smo, &msp->ms_map); 458 mutex_exit(&msp->ms_lock); 459 } 460 461 } 462 463 static void 464 dump_metaslabs(spa_t *spa) 465 { 466 vdev_t *rvd = spa->spa_root_vdev; 467 vdev_t *vd; 468 int c, m; 469 470 (void) printf("\nMetaslabs:\n"); 471 472 for (c = 0; c < rvd->vdev_children; c++) { 473 vd = rvd->vdev_child[c]; 474 475 (void) printf("\t%-10s %-19s %-15s %-10s\n", 476 "vdev", "offset", "spacemap", "free"); 477 (void) printf("\t%10s %19s %15s %10s\n", 478 "----------", "-------------------", 479 "---------------", "-------------"); 480 481 for (m = 0; m < vd->vdev_ms_count; m++) 482 dump_metaslab(vd->vdev_ms[m]); 483 (void) printf("\n"); 484 } 485 } 486 487 static void 488 dump_dtl_seg(space_map_t *sm, uint64_t start, uint64_t size) 489 { 490 char *prefix = (void *)sm; 491 492 (void) printf("%s [%llu,%llu) length %llu\n", 493 prefix, 494 (u_longlong_t)start, 495 (u_longlong_t)(start + size), 496 (u_longlong_t)(size)); 497 } 498 499 static void 500 dump_dtl(vdev_t *vd, int indent) 501 { 502 spa_t *spa = vd->vdev_spa; 503 boolean_t required; 504 char *name[DTL_TYPES] = { "missing", "partial", "scrub", "outage" }; 505 char prefix[256]; 506 507 spa_vdev_state_enter(spa); 508 required = vdev_dtl_required(vd); 509 (void) spa_vdev_state_exit(spa, NULL, 0); 510 511 if (indent == 0) 512 (void) printf("\nDirty time logs:\n\n"); 513 514 (void) printf("\t%*s%s [%s]\n", indent, "", 515 vd->vdev_path ? vd->vdev_path : 516 vd->vdev_parent ? vd->vdev_ops->vdev_op_type : spa_name(spa), 517 required ? "DTL-required" : "DTL-expendable"); 518 519 for (int t = 0; t < DTL_TYPES; t++) { 520 space_map_t *sm = &vd->vdev_dtl[t]; 521 if (sm->sm_space == 0) 522 continue; 523 (void) snprintf(prefix, sizeof (prefix), "\t%*s%s", 524 indent + 2, "", name[t]); 525 mutex_enter(sm->sm_lock); 526 space_map_walk(sm, dump_dtl_seg, (void *)prefix); 527 mutex_exit(sm->sm_lock); 528 if (dump_opt['d'] > 5 && vd->vdev_children == 0) 529 dump_spacemap(spa->spa_meta_objset, 530 &vd->vdev_dtl_smo, sm); 531 } 532 533 for (int c = 0; c < vd->vdev_children; c++) 534 dump_dtl(vd->vdev_child[c], indent + 4); 535 } 536 537 /*ARGSUSED*/ 538 static void 539 dump_dnode(objset_t *os, uint64_t object, void *data, size_t size) 540 { 541 } 542 543 static uint64_t 544 blkid2offset(const dnode_phys_t *dnp, int level, uint64_t blkid) 545 { 546 if (level < 0) 547 return (blkid); 548 549 return ((blkid << (level * (dnp->dn_indblkshift - SPA_BLKPTRSHIFT))) * 550 dnp->dn_datablkszsec << SPA_MINBLOCKSHIFT); 551 } 552 553 static void 554 sprintf_blkptr_compact(char *blkbuf, blkptr_t *bp, int alldvas) 555 { 556 dva_t *dva = bp->blk_dva; 557 int ndvas = alldvas ? BP_GET_NDVAS(bp) : 1; 558 int i; 559 560 blkbuf[0] = '\0'; 561 562 for (i = 0; i < ndvas; i++) 563 (void) sprintf(blkbuf + strlen(blkbuf), "%llu:%llx:%llx ", 564 (u_longlong_t)DVA_GET_VDEV(&dva[i]), 565 (u_longlong_t)DVA_GET_OFFSET(&dva[i]), 566 (u_longlong_t)DVA_GET_ASIZE(&dva[i])); 567 568 (void) sprintf(blkbuf + strlen(blkbuf), "%llxL/%llxP F=%llu B=%llu", 569 (u_longlong_t)BP_GET_LSIZE(bp), 570 (u_longlong_t)BP_GET_PSIZE(bp), 571 (u_longlong_t)bp->blk_fill, 572 (u_longlong_t)bp->blk_birth); 573 } 574 575 static void 576 print_indirect(blkptr_t *bp, const zbookmark_t *zb, 577 const dnode_phys_t *dnp) 578 { 579 char blkbuf[BP_SPRINTF_LEN]; 580 int l; 581 582 ASSERT3U(BP_GET_TYPE(bp), ==, dnp->dn_type); 583 ASSERT3U(BP_GET_LEVEL(bp), ==, zb->zb_level); 584 585 (void) printf("%16llx ", 586 (u_longlong_t)blkid2offset(dnp, zb->zb_level, zb->zb_blkid)); 587 588 ASSERT(zb->zb_level >= 0); 589 590 for (l = dnp->dn_nlevels - 1; l >= -1; l--) { 591 if (l == zb->zb_level) { 592 (void) printf("L%llx", (u_longlong_t)zb->zb_level); 593 } else { 594 (void) printf(" "); 595 } 596 } 597 598 sprintf_blkptr_compact(blkbuf, bp, dump_opt['d'] > 5 ? 1 : 0); 599 (void) printf("%s\n", blkbuf); 600 } 601 602 #define SET_BOOKMARK(zb, objset, object, level, blkid) \ 603 { \ 604 (zb)->zb_objset = objset; \ 605 (zb)->zb_object = object; \ 606 (zb)->zb_level = level; \ 607 (zb)->zb_blkid = blkid; \ 608 } 609 610 static int 611 visit_indirect(spa_t *spa, const dnode_phys_t *dnp, 612 blkptr_t *bp, const zbookmark_t *zb) 613 { 614 int err; 615 616 if (bp->blk_birth == 0) 617 return (0); 618 619 print_indirect(bp, zb, dnp); 620 621 if (BP_GET_LEVEL(bp) > 0) { 622 uint32_t flags = ARC_WAIT; 623 int i; 624 blkptr_t *cbp; 625 int epb = BP_GET_LSIZE(bp) >> SPA_BLKPTRSHIFT; 626 arc_buf_t *buf; 627 uint64_t fill = 0; 628 629 err = arc_read_nolock(NULL, spa, bp, arc_getbuf_func, &buf, 630 ZIO_PRIORITY_ASYNC_READ, ZIO_FLAG_CANFAIL, &flags, zb); 631 if (err) 632 return (err); 633 634 /* recursively visit blocks below this */ 635 cbp = buf->b_data; 636 for (i = 0; i < epb; i++, cbp++) { 637 zbookmark_t czb; 638 639 SET_BOOKMARK(&czb, zb->zb_objset, zb->zb_object, 640 zb->zb_level - 1, 641 zb->zb_blkid * epb + i); 642 err = visit_indirect(spa, dnp, cbp, &czb); 643 if (err) 644 break; 645 fill += cbp->blk_fill; 646 } 647 if (!err) 648 ASSERT3U(fill, ==, bp->blk_fill); 649 (void) arc_buf_remove_ref(buf, &buf); 650 } 651 652 return (err); 653 } 654 655 /*ARGSUSED*/ 656 static void 657 dump_indirect(dnode_t *dn) 658 { 659 dnode_phys_t *dnp = dn->dn_phys; 660 int j; 661 zbookmark_t czb; 662 663 (void) printf("Indirect blocks:\n"); 664 665 SET_BOOKMARK(&czb, dmu_objset_id(dn->dn_objset), 666 dn->dn_object, dnp->dn_nlevels - 1, 0); 667 for (j = 0; j < dnp->dn_nblkptr; j++) { 668 czb.zb_blkid = j; 669 (void) visit_indirect(dmu_objset_spa(dn->dn_objset), dnp, 670 &dnp->dn_blkptr[j], &czb); 671 } 672 673 (void) printf("\n"); 674 } 675 676 /*ARGSUSED*/ 677 static void 678 dump_dsl_dir(objset_t *os, uint64_t object, void *data, size_t size) 679 { 680 dsl_dir_phys_t *dd = data; 681 time_t crtime; 682 char nice[6]; 683 684 if (dd == NULL) 685 return; 686 687 ASSERT3U(size, >=, sizeof (dsl_dir_phys_t)); 688 689 crtime = dd->dd_creation_time; 690 (void) printf("\t\tcreation_time = %s", ctime(&crtime)); 691 (void) printf("\t\thead_dataset_obj = %llu\n", 692 (u_longlong_t)dd->dd_head_dataset_obj); 693 (void) printf("\t\tparent_dir_obj = %llu\n", 694 (u_longlong_t)dd->dd_parent_obj); 695 (void) printf("\t\torigin_obj = %llu\n", 696 (u_longlong_t)dd->dd_origin_obj); 697 (void) printf("\t\tchild_dir_zapobj = %llu\n", 698 (u_longlong_t)dd->dd_child_dir_zapobj); 699 nicenum(dd->dd_used_bytes, nice); 700 (void) printf("\t\tused_bytes = %s\n", nice); 701 nicenum(dd->dd_compressed_bytes, nice); 702 (void) printf("\t\tcompressed_bytes = %s\n", nice); 703 nicenum(dd->dd_uncompressed_bytes, nice); 704 (void) printf("\t\tuncompressed_bytes = %s\n", nice); 705 nicenum(dd->dd_quota, nice); 706 (void) printf("\t\tquota = %s\n", nice); 707 nicenum(dd->dd_reserved, nice); 708 (void) printf("\t\treserved = %s\n", nice); 709 (void) printf("\t\tprops_zapobj = %llu\n", 710 (u_longlong_t)dd->dd_props_zapobj); 711 (void) printf("\t\tdeleg_zapobj = %llu\n", 712 (u_longlong_t)dd->dd_deleg_zapobj); 713 (void) printf("\t\tflags = %llx\n", 714 (u_longlong_t)dd->dd_flags); 715 716 #define DO(which) \ 717 nicenum(dd->dd_used_breakdown[DD_USED_ ## which], nice); \ 718 (void) printf("\t\tused_breakdown[" #which "] = %s\n", nice) 719 DO(HEAD); 720 DO(SNAP); 721 DO(CHILD); 722 DO(CHILD_RSRV); 723 DO(REFRSRV); 724 #undef DO 725 } 726 727 /*ARGSUSED*/ 728 static void 729 dump_dsl_dataset(objset_t *os, uint64_t object, void *data, size_t size) 730 { 731 dsl_dataset_phys_t *ds = data; 732 time_t crtime; 733 char used[6], compressed[6], uncompressed[6], unique[6]; 734 char blkbuf[BP_SPRINTF_LEN]; 735 736 if (ds == NULL) 737 return; 738 739 ASSERT(size == sizeof (*ds)); 740 crtime = ds->ds_creation_time; 741 nicenum(ds->ds_used_bytes, used); 742 nicenum(ds->ds_compressed_bytes, compressed); 743 nicenum(ds->ds_uncompressed_bytes, uncompressed); 744 nicenum(ds->ds_unique_bytes, unique); 745 sprintf_blkptr(blkbuf, BP_SPRINTF_LEN, &ds->ds_bp); 746 747 (void) printf("\t\tdir_obj = %llu\n", 748 (u_longlong_t)ds->ds_dir_obj); 749 (void) printf("\t\tprev_snap_obj = %llu\n", 750 (u_longlong_t)ds->ds_prev_snap_obj); 751 (void) printf("\t\tprev_snap_txg = %llu\n", 752 (u_longlong_t)ds->ds_prev_snap_txg); 753 (void) printf("\t\tnext_snap_obj = %llu\n", 754 (u_longlong_t)ds->ds_next_snap_obj); 755 (void) printf("\t\tsnapnames_zapobj = %llu\n", 756 (u_longlong_t)ds->ds_snapnames_zapobj); 757 (void) printf("\t\tnum_children = %llu\n", 758 (u_longlong_t)ds->ds_num_children); 759 (void) printf("\t\tuserrefs_obj = %llu\n", 760 (u_longlong_t)ds->ds_userrefs_obj); 761 (void) printf("\t\tcreation_time = %s", ctime(&crtime)); 762 (void) printf("\t\tcreation_txg = %llu\n", 763 (u_longlong_t)ds->ds_creation_txg); 764 (void) printf("\t\tdeadlist_obj = %llu\n", 765 (u_longlong_t)ds->ds_deadlist_obj); 766 (void) printf("\t\tused_bytes = %s\n", used); 767 (void) printf("\t\tcompressed_bytes = %s\n", compressed); 768 (void) printf("\t\tuncompressed_bytes = %s\n", uncompressed); 769 (void) printf("\t\tunique = %s\n", unique); 770 (void) printf("\t\tfsid_guid = %llu\n", 771 (u_longlong_t)ds->ds_fsid_guid); 772 (void) printf("\t\tguid = %llu\n", 773 (u_longlong_t)ds->ds_guid); 774 (void) printf("\t\tflags = %llx\n", 775 (u_longlong_t)ds->ds_flags); 776 (void) printf("\t\tnext_clones_obj = %llu\n", 777 (u_longlong_t)ds->ds_next_clones_obj); 778 (void) printf("\t\tprops_obj = %llu\n", 779 (u_longlong_t)ds->ds_props_obj); 780 (void) printf("\t\tbp = %s\n", blkbuf); 781 } 782 783 static void 784 dump_bplist(objset_t *mos, uint64_t object, char *name) 785 { 786 bplist_t bpl = { 0 }; 787 blkptr_t blk, *bp = &blk; 788 uint64_t itor = 0; 789 char bytes[6]; 790 char comp[6]; 791 char uncomp[6]; 792 793 if (dump_opt['d'] < 3) 794 return; 795 796 mutex_init(&bpl.bpl_lock, NULL, MUTEX_DEFAULT, NULL); 797 VERIFY(0 == bplist_open(&bpl, mos, object)); 798 if (bplist_empty(&bpl)) { 799 bplist_close(&bpl); 800 mutex_destroy(&bpl.bpl_lock); 801 return; 802 } 803 804 nicenum(bpl.bpl_phys->bpl_bytes, bytes); 805 if (bpl.bpl_dbuf->db_size == sizeof (bplist_phys_t)) { 806 nicenum(bpl.bpl_phys->bpl_comp, comp); 807 nicenum(bpl.bpl_phys->bpl_uncomp, uncomp); 808 (void) printf("\n %s: %llu entries, %s (%s/%s comp)\n", 809 name, (u_longlong_t)bpl.bpl_phys->bpl_entries, 810 bytes, comp, uncomp); 811 } else { 812 (void) printf("\n %s: %llu entries, %s\n", 813 name, (u_longlong_t)bpl.bpl_phys->bpl_entries, bytes); 814 } 815 816 if (dump_opt['d'] < 5) { 817 bplist_close(&bpl); 818 mutex_destroy(&bpl.bpl_lock); 819 return; 820 } 821 822 (void) printf("\n"); 823 824 while (bplist_iterate(&bpl, &itor, bp) == 0) { 825 char blkbuf[BP_SPRINTF_LEN]; 826 827 ASSERT(bp->blk_birth != 0); 828 sprintf_blkptr_compact(blkbuf, bp, dump_opt['d'] > 5 ? 1 : 0); 829 (void) printf("\tItem %3llu: %s\n", 830 (u_longlong_t)itor - 1, blkbuf); 831 } 832 833 bplist_close(&bpl); 834 mutex_destroy(&bpl.bpl_lock); 835 } 836 837 static avl_tree_t idx_tree; 838 static avl_tree_t domain_tree; 839 static boolean_t fuid_table_loaded; 840 841 static void 842 fuid_table_destroy() 843 { 844 if (fuid_table_loaded) { 845 zfs_fuid_table_destroy(&idx_tree, &domain_tree); 846 fuid_table_loaded = B_FALSE; 847 } 848 } 849 850 /* 851 * print uid or gid information. 852 * For normal POSIX id just the id is printed in decimal format. 853 * For CIFS files with FUID the fuid is printed in hex followed by 854 * the doman-rid string. 855 */ 856 static void 857 print_idstr(uint64_t id, const char *id_type) 858 { 859 if (FUID_INDEX(id)) { 860 char *domain; 861 862 domain = zfs_fuid_idx_domain(&idx_tree, FUID_INDEX(id)); 863 (void) printf("\t%s %llx [%s-%d]\n", id_type, 864 (u_longlong_t)id, domain, (int)FUID_RID(id)); 865 } else { 866 (void) printf("\t%s %llu\n", id_type, (u_longlong_t)id); 867 } 868 869 } 870 871 static void 872 dump_uidgid(objset_t *os, znode_phys_t *zp) 873 { 874 uint32_t uid_idx, gid_idx; 875 876 uid_idx = FUID_INDEX(zp->zp_uid); 877 gid_idx = FUID_INDEX(zp->zp_gid); 878 879 /* Load domain table, if not already loaded */ 880 if (!fuid_table_loaded && (uid_idx || gid_idx)) { 881 uint64_t fuid_obj; 882 883 /* first find the fuid object. It lives in the master node */ 884 VERIFY(zap_lookup(os, MASTER_NODE_OBJ, ZFS_FUID_TABLES, 885 8, 1, &fuid_obj) == 0); 886 zfs_fuid_avl_tree_create(&idx_tree, &domain_tree); 887 (void) zfs_fuid_table_load(os, fuid_obj, 888 &idx_tree, &domain_tree); 889 fuid_table_loaded = B_TRUE; 890 } 891 892 print_idstr(zp->zp_uid, "uid"); 893 print_idstr(zp->zp_gid, "gid"); 894 } 895 896 /*ARGSUSED*/ 897 static void 898 dump_znode(objset_t *os, uint64_t object, void *data, size_t size) 899 { 900 znode_phys_t *zp = data; 901 time_t z_crtime, z_atime, z_mtime, z_ctime; 902 char path[MAXPATHLEN * 2]; /* allow for xattr and failure prefix */ 903 int error; 904 905 ASSERT(size >= sizeof (znode_phys_t)); 906 907 error = zfs_obj_to_path(os, object, path, sizeof (path)); 908 if (error != 0) { 909 (void) snprintf(path, sizeof (path), "\?\?\?<object#%llu>", 910 (u_longlong_t)object); 911 } 912 913 if (dump_opt['d'] < 3) { 914 (void) printf("\t%s\n", path); 915 return; 916 } 917 918 z_crtime = (time_t)zp->zp_crtime[0]; 919 z_atime = (time_t)zp->zp_atime[0]; 920 z_mtime = (time_t)zp->zp_mtime[0]; 921 z_ctime = (time_t)zp->zp_ctime[0]; 922 923 (void) printf("\tpath %s\n", path); 924 dump_uidgid(os, zp); 925 (void) printf("\tatime %s", ctime(&z_atime)); 926 (void) printf("\tmtime %s", ctime(&z_mtime)); 927 (void) printf("\tctime %s", ctime(&z_ctime)); 928 (void) printf("\tcrtime %s", ctime(&z_crtime)); 929 (void) printf("\tgen %llu\n", (u_longlong_t)zp->zp_gen); 930 (void) printf("\tmode %llo\n", (u_longlong_t)zp->zp_mode); 931 (void) printf("\tsize %llu\n", (u_longlong_t)zp->zp_size); 932 (void) printf("\tparent %llu\n", (u_longlong_t)zp->zp_parent); 933 (void) printf("\tlinks %llu\n", (u_longlong_t)zp->zp_links); 934 (void) printf("\txattr %llu\n", (u_longlong_t)zp->zp_xattr); 935 (void) printf("\trdev 0x%016llx\n", (u_longlong_t)zp->zp_rdev); 936 } 937 938 /*ARGSUSED*/ 939 static void 940 dump_acl(objset_t *os, uint64_t object, void *data, size_t size) 941 { 942 } 943 944 /*ARGSUSED*/ 945 static void 946 dump_dmu_objset(objset_t *os, uint64_t object, void *data, size_t size) 947 { 948 } 949 950 static object_viewer_t *object_viewer[DMU_OT_NUMTYPES] = { 951 dump_none, /* unallocated */ 952 dump_zap, /* object directory */ 953 dump_uint64, /* object array */ 954 dump_none, /* packed nvlist */ 955 dump_packed_nvlist, /* packed nvlist size */ 956 dump_none, /* bplist */ 957 dump_none, /* bplist header */ 958 dump_none, /* SPA space map header */ 959 dump_none, /* SPA space map */ 960 dump_none, /* ZIL intent log */ 961 dump_dnode, /* DMU dnode */ 962 dump_dmu_objset, /* DMU objset */ 963 dump_dsl_dir, /* DSL directory */ 964 dump_zap, /* DSL directory child map */ 965 dump_zap, /* DSL dataset snap map */ 966 dump_zap, /* DSL props */ 967 dump_dsl_dataset, /* DSL dataset */ 968 dump_znode, /* ZFS znode */ 969 dump_acl, /* ZFS V0 ACL */ 970 dump_uint8, /* ZFS plain file */ 971 dump_zpldir, /* ZFS directory */ 972 dump_zap, /* ZFS master node */ 973 dump_zap, /* ZFS delete queue */ 974 dump_uint8, /* zvol object */ 975 dump_zap, /* zvol prop */ 976 dump_uint8, /* other uint8[] */ 977 dump_uint64, /* other uint64[] */ 978 dump_zap, /* other ZAP */ 979 dump_zap, /* persistent error log */ 980 dump_uint8, /* SPA history */ 981 dump_uint64, /* SPA history offsets */ 982 dump_zap, /* Pool properties */ 983 dump_zap, /* DSL permissions */ 984 dump_acl, /* ZFS ACL */ 985 dump_uint8, /* ZFS SYSACL */ 986 dump_none, /* FUID nvlist */ 987 dump_packed_nvlist, /* FUID nvlist size */ 988 dump_zap, /* DSL dataset next clones */ 989 dump_zap, /* DSL scrub queue */ 990 dump_zap, /* ZFS user/group used */ 991 dump_zap, /* ZFS user/group quota */ 992 dump_zap, /* snapshot refcount tags */ 993 }; 994 995 static void 996 dump_object(objset_t *os, uint64_t object, int verbosity, int *print_header) 997 { 998 dmu_buf_t *db = NULL; 999 dmu_object_info_t doi; 1000 dnode_t *dn; 1001 void *bonus = NULL; 1002 size_t bsize = 0; 1003 char iblk[6], dblk[6], lsize[6], asize[6], bonus_size[6], segsize[6]; 1004 char aux[50]; 1005 int error; 1006 1007 if (*print_header) { 1008 (void) printf("\n Object lvl iblk dblk lsize" 1009 " asize type\n"); 1010 *print_header = 0; 1011 } 1012 1013 if (object == 0) { 1014 dn = os->os_meta_dnode; 1015 } else { 1016 error = dmu_bonus_hold(os, object, FTAG, &db); 1017 if (error) 1018 fatal("dmu_bonus_hold(%llu) failed, errno %u", 1019 object, error); 1020 bonus = db->db_data; 1021 bsize = db->db_size; 1022 dn = ((dmu_buf_impl_t *)db)->db_dnode; 1023 } 1024 dmu_object_info_from_dnode(dn, &doi); 1025 1026 nicenum(doi.doi_metadata_block_size, iblk); 1027 nicenum(doi.doi_data_block_size, dblk); 1028 nicenum(doi.doi_data_block_size * (doi.doi_max_block_offset + 1), 1029 lsize); 1030 nicenum(doi.doi_physical_blks << 9, asize); 1031 nicenum(doi.doi_bonus_size, bonus_size); 1032 1033 aux[0] = '\0'; 1034 1035 if (doi.doi_checksum != ZIO_CHECKSUM_INHERIT || verbosity >= 6) { 1036 (void) snprintf(aux + strlen(aux), sizeof (aux), " (K=%s)", 1037 zio_checksum_table[doi.doi_checksum].ci_name); 1038 } 1039 1040 if (doi.doi_compress != ZIO_COMPRESS_INHERIT || verbosity >= 6) { 1041 (void) snprintf(aux + strlen(aux), sizeof (aux), " (Z=%s)", 1042 zio_compress_table[doi.doi_compress].ci_name); 1043 } 1044 1045 (void) printf("%10lld %3u %5s %5s %5s %5s %s%s\n", 1046 (u_longlong_t)object, doi.doi_indirection, iblk, dblk, lsize, 1047 asize, dmu_ot[doi.doi_type].ot_name, aux); 1048 1049 if (doi.doi_bonus_type != DMU_OT_NONE && verbosity > 3) { 1050 (void) printf("%10s %3s %5s %5s %5s %5s %s\n", 1051 "", "", "", "", bonus_size, "bonus", 1052 dmu_ot[doi.doi_bonus_type].ot_name); 1053 } 1054 1055 if (verbosity >= 4) { 1056 (void) printf("\tdnode flags: %s%s\n", 1057 (dn->dn_phys->dn_flags & DNODE_FLAG_USED_BYTES) ? 1058 "USED_BYTES " : "", 1059 (dn->dn_phys->dn_flags & DNODE_FLAG_USERUSED_ACCOUNTED) ? 1060 "USERUSED_ACCOUNTED " : ""); 1061 (void) printf("\tdnode maxblkid: %llu\n", 1062 (longlong_t)dn->dn_phys->dn_maxblkid); 1063 1064 object_viewer[doi.doi_bonus_type](os, object, bonus, bsize); 1065 object_viewer[doi.doi_type](os, object, NULL, 0); 1066 *print_header = 1; 1067 } 1068 1069 if (verbosity >= 5) 1070 dump_indirect(dn); 1071 1072 if (verbosity >= 5) { 1073 /* 1074 * Report the list of segments that comprise the object. 1075 */ 1076 uint64_t start = 0; 1077 uint64_t end; 1078 uint64_t blkfill = 1; 1079 int minlvl = 1; 1080 1081 if (dn->dn_type == DMU_OT_DNODE) { 1082 minlvl = 0; 1083 blkfill = DNODES_PER_BLOCK; 1084 } 1085 1086 for (;;) { 1087 error = dnode_next_offset(dn, 1088 0, &start, minlvl, blkfill, 0); 1089 if (error) 1090 break; 1091 end = start; 1092 error = dnode_next_offset(dn, 1093 DNODE_FIND_HOLE, &end, minlvl, blkfill, 0); 1094 nicenum(end - start, segsize); 1095 (void) printf("\t\tsegment [%016llx, %016llx)" 1096 " size %5s\n", (u_longlong_t)start, 1097 (u_longlong_t)end, segsize); 1098 if (error) 1099 break; 1100 start = end; 1101 } 1102 } 1103 1104 if (db != NULL) 1105 dmu_buf_rele(db, FTAG); 1106 } 1107 1108 static char *objset_types[DMU_OST_NUMTYPES] = { 1109 "NONE", "META", "ZPL", "ZVOL", "OTHER", "ANY" }; 1110 1111 static void 1112 dump_dir(objset_t *os) 1113 { 1114 dmu_objset_stats_t dds; 1115 uint64_t object, object_count; 1116 uint64_t refdbytes, usedobjs, scratch; 1117 char numbuf[8]; 1118 char blkbuf[BP_SPRINTF_LEN + 20]; 1119 char osname[MAXNAMELEN]; 1120 char *type = "UNKNOWN"; 1121 int verbosity = dump_opt['d']; 1122 int print_header = 1; 1123 int i, error; 1124 1125 dmu_objset_fast_stat(os, &dds); 1126 1127 if (dds.dds_type < DMU_OST_NUMTYPES) 1128 type = objset_types[dds.dds_type]; 1129 1130 if (dds.dds_type == DMU_OST_META) { 1131 dds.dds_creation_txg = TXG_INITIAL; 1132 usedobjs = os->os_rootbp->blk_fill; 1133 refdbytes = os->os_spa->spa_dsl_pool-> 1134 dp_mos_dir->dd_phys->dd_used_bytes; 1135 } else { 1136 dmu_objset_space(os, &refdbytes, &scratch, &usedobjs, &scratch); 1137 } 1138 1139 ASSERT3U(usedobjs, ==, os->os_rootbp->blk_fill); 1140 1141 nicenum(refdbytes, numbuf); 1142 1143 if (verbosity >= 4) { 1144 (void) sprintf(blkbuf + strlen(blkbuf), ", rootbp "); 1145 (void) sprintf_blkptr(blkbuf + strlen(blkbuf), 1146 BP_SPRINTF_LEN - strlen(blkbuf), os->os_rootbp); 1147 } else { 1148 blkbuf[0] = '\0'; 1149 } 1150 1151 dmu_objset_name(os, osname); 1152 1153 (void) printf("Dataset %s [%s], ID %llu, cr_txg %llu, " 1154 "%s, %llu objects%s\n", 1155 osname, type, (u_longlong_t)dmu_objset_id(os), 1156 (u_longlong_t)dds.dds_creation_txg, 1157 numbuf, (u_longlong_t)usedobjs, blkbuf); 1158 1159 dump_intent_log(dmu_objset_zil(os)); 1160 1161 if (dmu_objset_ds(os) != NULL) 1162 dump_bplist(dmu_objset_pool(os)->dp_meta_objset, 1163 dmu_objset_ds(os)->ds_phys->ds_deadlist_obj, "Deadlist"); 1164 1165 if (verbosity < 2) 1166 return; 1167 1168 if (os->os_rootbp->blk_birth == 0) 1169 return; 1170 1171 if (zopt_objects != 0) { 1172 for (i = 0; i < zopt_objects; i++) 1173 dump_object(os, zopt_object[i], verbosity, 1174 &print_header); 1175 (void) printf("\n"); 1176 return; 1177 } 1178 1179 dump_object(os, 0, verbosity, &print_header); 1180 object_count = 0; 1181 if (os->os_userused_dnode && 1182 os->os_userused_dnode->dn_type != 0) { 1183 dump_object(os, DMU_USERUSED_OBJECT, verbosity, &print_header); 1184 dump_object(os, DMU_GROUPUSED_OBJECT, verbosity, &print_header); 1185 } 1186 1187 object = 0; 1188 while ((error = dmu_object_next(os, &object, B_FALSE, 0)) == 0) { 1189 dump_object(os, object, verbosity, &print_header); 1190 object_count++; 1191 } 1192 1193 ASSERT3U(object_count, ==, usedobjs); 1194 1195 (void) printf("\n"); 1196 1197 if (error != ESRCH) { 1198 (void) fprintf(stderr, "dmu_object_next() = %d\n", error); 1199 abort(); 1200 } 1201 } 1202 1203 static void 1204 dump_uberblock(uberblock_t *ub) 1205 { 1206 time_t timestamp = ub->ub_timestamp; 1207 1208 (void) printf("Uberblock\n\n"); 1209 (void) printf("\tmagic = %016llx\n", (u_longlong_t)ub->ub_magic); 1210 (void) printf("\tversion = %llu\n", (u_longlong_t)ub->ub_version); 1211 (void) printf("\ttxg = %llu\n", (u_longlong_t)ub->ub_txg); 1212 (void) printf("\tguid_sum = %llu\n", (u_longlong_t)ub->ub_guid_sum); 1213 (void) printf("\ttimestamp = %llu UTC = %s", 1214 (u_longlong_t)ub->ub_timestamp, asctime(localtime(×tamp))); 1215 if (dump_opt['u'] >= 3) { 1216 char blkbuf[BP_SPRINTF_LEN]; 1217 sprintf_blkptr(blkbuf, BP_SPRINTF_LEN, &ub->ub_rootbp); 1218 (void) printf("\trootbp = %s\n", blkbuf); 1219 } 1220 (void) printf("\n"); 1221 } 1222 1223 static void 1224 dump_config(const char *pool) 1225 { 1226 spa_t *spa = NULL; 1227 1228 mutex_enter(&spa_namespace_lock); 1229 while ((spa = spa_next(spa)) != NULL) { 1230 if (pool == NULL) 1231 (void) printf("%s\n", spa_name(spa)); 1232 if (pool == NULL || strcmp(pool, spa_name(spa)) == 0) 1233 dump_nvlist(spa->spa_config, 4); 1234 } 1235 mutex_exit(&spa_namespace_lock); 1236 } 1237 1238 static void 1239 dump_cachefile(const char *cachefile) 1240 { 1241 int fd; 1242 struct stat64 statbuf; 1243 char *buf; 1244 nvlist_t *config; 1245 1246 if ((fd = open64(cachefile, O_RDONLY)) < 0) { 1247 (void) printf("cannot open '%s': %s\n", cachefile, 1248 strerror(errno)); 1249 exit(1); 1250 } 1251 1252 if (fstat64(fd, &statbuf) != 0) { 1253 (void) printf("failed to stat '%s': %s\n", cachefile, 1254 strerror(errno)); 1255 exit(1); 1256 } 1257 1258 if ((buf = malloc(statbuf.st_size)) == NULL) { 1259 (void) fprintf(stderr, "failed to allocate %llu bytes\n", 1260 (u_longlong_t)statbuf.st_size); 1261 exit(1); 1262 } 1263 1264 if (read(fd, buf, statbuf.st_size) != statbuf.st_size) { 1265 (void) fprintf(stderr, "failed to read %llu bytes\n", 1266 (u_longlong_t)statbuf.st_size); 1267 exit(1); 1268 } 1269 1270 (void) close(fd); 1271 1272 if (nvlist_unpack(buf, statbuf.st_size, &config, 0) != 0) { 1273 (void) fprintf(stderr, "failed to unpack nvlist\n"); 1274 exit(1); 1275 } 1276 1277 free(buf); 1278 1279 dump_nvlist(config, 0); 1280 1281 nvlist_free(config); 1282 } 1283 1284 static void 1285 dump_label(const char *dev) 1286 { 1287 int fd; 1288 vdev_label_t label; 1289 char *buf = label.vl_vdev_phys.vp_nvlist; 1290 size_t buflen = sizeof (label.vl_vdev_phys.vp_nvlist); 1291 struct stat64 statbuf; 1292 uint64_t psize; 1293 int l; 1294 1295 if ((fd = open64(dev, O_RDONLY)) < 0) { 1296 (void) printf("cannot open '%s': %s\n", dev, strerror(errno)); 1297 exit(1); 1298 } 1299 1300 if (fstat64(fd, &statbuf) != 0) { 1301 (void) printf("failed to stat '%s': %s\n", dev, 1302 strerror(errno)); 1303 exit(1); 1304 } 1305 1306 psize = statbuf.st_size; 1307 psize = P2ALIGN(psize, (uint64_t)sizeof (vdev_label_t)); 1308 1309 for (l = 0; l < VDEV_LABELS; l++) { 1310 1311 nvlist_t *config = NULL; 1312 1313 (void) printf("--------------------------------------------\n"); 1314 (void) printf("LABEL %d\n", l); 1315 (void) printf("--------------------------------------------\n"); 1316 1317 if (pread64(fd, &label, sizeof (label), 1318 vdev_label_offset(psize, l, 0)) != sizeof (label)) { 1319 (void) printf("failed to read label %d\n", l); 1320 continue; 1321 } 1322 1323 if (nvlist_unpack(buf, buflen, &config, 0) != 0) { 1324 (void) printf("failed to unpack label %d\n", l); 1325 continue; 1326 } 1327 dump_nvlist(config, 4); 1328 nvlist_free(config); 1329 } 1330 } 1331 1332 /*ARGSUSED*/ 1333 static int 1334 dump_one_dir(char *dsname, void *arg) 1335 { 1336 int error; 1337 objset_t *os; 1338 1339 error = dmu_objset_own(dsname, DMU_OST_ANY, B_TRUE, FTAG, &os); 1340 if (error) { 1341 (void) printf("Could not open %s\n", dsname); 1342 return (0); 1343 } 1344 dump_dir(os); 1345 dmu_objset_disown(os, FTAG); 1346 fuid_table_destroy(); 1347 return (0); 1348 } 1349 1350 static void 1351 zdb_leak(space_map_t *sm, uint64_t start, uint64_t size) 1352 { 1353 vdev_t *vd = sm->sm_ppd; 1354 1355 (void) printf("leaked space: vdev %llu, offset 0x%llx, size %llu\n", 1356 (u_longlong_t)vd->vdev_id, (u_longlong_t)start, (u_longlong_t)size); 1357 } 1358 1359 /* ARGSUSED */ 1360 static void 1361 zdb_space_map_load(space_map_t *sm) 1362 { 1363 } 1364 1365 static void 1366 zdb_space_map_unload(space_map_t *sm) 1367 { 1368 space_map_vacate(sm, zdb_leak, sm); 1369 } 1370 1371 /* ARGSUSED */ 1372 static void 1373 zdb_space_map_claim(space_map_t *sm, uint64_t start, uint64_t size) 1374 { 1375 } 1376 1377 static space_map_ops_t zdb_space_map_ops = { 1378 zdb_space_map_load, 1379 zdb_space_map_unload, 1380 NULL, /* alloc */ 1381 zdb_space_map_claim, 1382 NULL, /* free */ 1383 NULL /* maxsize */ 1384 }; 1385 1386 static void 1387 zdb_leak_init(spa_t *spa) 1388 { 1389 vdev_t *rvd = spa->spa_root_vdev; 1390 1391 for (int c = 0; c < rvd->vdev_children; c++) { 1392 vdev_t *vd = rvd->vdev_child[c]; 1393 for (int m = 0; m < vd->vdev_ms_count; m++) { 1394 metaslab_t *msp = vd->vdev_ms[m]; 1395 mutex_enter(&msp->ms_lock); 1396 VERIFY(space_map_load(&msp->ms_map, &zdb_space_map_ops, 1397 SM_ALLOC, &msp->ms_smo, spa->spa_meta_objset) == 0); 1398 msp->ms_map.sm_ppd = vd; 1399 mutex_exit(&msp->ms_lock); 1400 } 1401 } 1402 } 1403 1404 static void 1405 zdb_leak_fini(spa_t *spa) 1406 { 1407 vdev_t *rvd = spa->spa_root_vdev; 1408 1409 for (int c = 0; c < rvd->vdev_children; c++) { 1410 vdev_t *vd = rvd->vdev_child[c]; 1411 for (int m = 0; m < vd->vdev_ms_count; m++) { 1412 metaslab_t *msp = vd->vdev_ms[m]; 1413 mutex_enter(&msp->ms_lock); 1414 space_map_unload(&msp->ms_map); 1415 mutex_exit(&msp->ms_lock); 1416 } 1417 } 1418 } 1419 1420 /* 1421 * Verify that the sum of the sizes of all blocks in the pool adds up 1422 * to the SPA's sa_alloc total. 1423 */ 1424 typedef struct zdb_blkstats { 1425 uint64_t zb_asize; 1426 uint64_t zb_lsize; 1427 uint64_t zb_psize; 1428 uint64_t zb_count; 1429 } zdb_blkstats_t; 1430 1431 #define DMU_OT_DEFERRED DMU_OT_NONE 1432 #define DMU_OT_TOTAL DMU_OT_NUMTYPES 1433 1434 #define ZB_TOTAL DN_MAX_LEVELS 1435 1436 typedef struct zdb_cb { 1437 zdb_blkstats_t zcb_type[ZB_TOTAL + 1][DMU_OT_TOTAL + 1]; 1438 uint64_t zcb_errors[256]; 1439 int zcb_readfails; 1440 int zcb_haderrors; 1441 } zdb_cb_t; 1442 1443 static void 1444 zdb_count_block(spa_t *spa, zdb_cb_t *zcb, blkptr_t *bp, dmu_object_type_t type) 1445 { 1446 for (int i = 0; i < 4; i++) { 1447 int l = (i < 2) ? BP_GET_LEVEL(bp) : ZB_TOTAL; 1448 int t = (i & 1) ? type : DMU_OT_TOTAL; 1449 zdb_blkstats_t *zb = &zcb->zcb_type[l][t]; 1450 1451 zb->zb_asize += BP_GET_ASIZE(bp); 1452 zb->zb_lsize += BP_GET_LSIZE(bp); 1453 zb->zb_psize += BP_GET_PSIZE(bp); 1454 zb->zb_count++; 1455 } 1456 1457 if (dump_opt['S']) { 1458 boolean_t print_sig; 1459 1460 print_sig = !zdb_sig_user_data || (BP_GET_LEVEL(bp) == 0 && 1461 BP_GET_TYPE(bp) == DMU_OT_PLAIN_FILE_CONTENTS); 1462 1463 if (BP_GET_CHECKSUM(bp) < zdb_sig_cksumalg) 1464 print_sig = B_FALSE; 1465 1466 if (print_sig) { 1467 (void) printf("%llu\t%lld\t%lld\t%s\t%s\t%s\t" 1468 "%llx:%llx:%llx:%llx\n", 1469 (u_longlong_t)BP_GET_LEVEL(bp), 1470 (longlong_t)BP_GET_PSIZE(bp), 1471 (longlong_t)BP_GET_NDVAS(bp), 1472 dmu_ot[BP_GET_TYPE(bp)].ot_name, 1473 zio_checksum_table[BP_GET_CHECKSUM(bp)].ci_name, 1474 zio_compress_table[BP_GET_COMPRESS(bp)].ci_name, 1475 (u_longlong_t)bp->blk_cksum.zc_word[0], 1476 (u_longlong_t)bp->blk_cksum.zc_word[1], 1477 (u_longlong_t)bp->blk_cksum.zc_word[2], 1478 (u_longlong_t)bp->blk_cksum.zc_word[3]); 1479 } 1480 } 1481 1482 if (!dump_opt['L']) 1483 VERIFY(zio_wait(zio_claim(NULL, spa, spa_first_txg(spa), bp, 1484 NULL, NULL, ZIO_FLAG_MUSTSUCCEED)) == 0); 1485 } 1486 1487 static int 1488 zdb_blkptr_cb(spa_t *spa, blkptr_t *bp, const zbookmark_t *zb, 1489 const dnode_phys_t *dnp, void *arg) 1490 { 1491 zdb_cb_t *zcb = arg; 1492 char blkbuf[BP_SPRINTF_LEN]; 1493 dmu_object_type_t type; 1494 boolean_t is_l0_metadata; 1495 1496 if (bp == NULL) 1497 return (0); 1498 1499 type = BP_GET_TYPE(bp); 1500 1501 zdb_count_block(spa, zcb, bp, type); 1502 1503 /* 1504 * if we do metadata-only checksumming there's no need to checksum 1505 * indirect blocks here because it is done during traverse 1506 */ 1507 is_l0_metadata = (BP_GET_LEVEL(bp) == 0 && type < DMU_OT_NUMTYPES && 1508 dmu_ot[type].ot_metadata); 1509 1510 if (dump_opt['c'] > 1 || dump_opt['S'] || 1511 (dump_opt['c'] && is_l0_metadata)) { 1512 int ioerr, size; 1513 void *data; 1514 1515 size = BP_GET_LSIZE(bp); 1516 data = malloc(size); 1517 ioerr = zio_wait(zio_read(NULL, spa, bp, data, size, 1518 NULL, NULL, ZIO_PRIORITY_ASYNC_READ, 1519 ZIO_FLAG_CANFAIL | ZIO_FLAG_SCRUB, zb)); 1520 free(data); 1521 1522 /* We expect io errors on intent log */ 1523 if (ioerr && type != DMU_OT_INTENT_LOG) { 1524 zcb->zcb_haderrors = 1; 1525 zcb->zcb_errors[ioerr]++; 1526 1527 if (dump_opt['b'] >= 2) 1528 sprintf_blkptr(blkbuf, BP_SPRINTF_LEN, bp); 1529 else 1530 blkbuf[0] = '\0'; 1531 1532 if (!dump_opt['S']) { 1533 (void) printf("zdb_blkptr_cb: " 1534 "Got error %d reading " 1535 "<%llu, %llu, %lld, %llx> %s -- skipping\n", 1536 ioerr, 1537 (u_longlong_t)zb->zb_objset, 1538 (u_longlong_t)zb->zb_object, 1539 (u_longlong_t)zb->zb_level, 1540 (u_longlong_t)zb->zb_blkid, 1541 blkbuf); 1542 } 1543 } 1544 } 1545 1546 zcb->zcb_readfails = 0; 1547 1548 if (dump_opt['b'] >= 4) { 1549 sprintf_blkptr(blkbuf, BP_SPRINTF_LEN, bp); 1550 (void) printf("objset %llu object %llu offset 0x%llx %s\n", 1551 (u_longlong_t)zb->zb_objset, 1552 (u_longlong_t)zb->zb_object, 1553 (u_longlong_t)blkid2offset(dnp, zb->zb_level, zb->zb_blkid), 1554 blkbuf); 1555 } 1556 1557 return (0); 1558 } 1559 1560 static int 1561 dump_block_stats(spa_t *spa) 1562 { 1563 zdb_cb_t zcb = { 0 }; 1564 zdb_blkstats_t *zb, *tzb; 1565 uint64_t alloc, space, logalloc; 1566 vdev_t *rvd = spa->spa_root_vdev; 1567 int leaks = 0; 1568 int c, e; 1569 1570 if (!dump_opt['S']) { 1571 (void) printf("\nTraversing all blocks %s%s%s%s%s...\n", 1572 (dump_opt['c'] || !dump_opt['L']) ? "to verify " : "", 1573 (dump_opt['c'] == 1) ? "metadata " : "", 1574 dump_opt['c'] ? "checksums " : "", 1575 (dump_opt['c'] && !dump_opt['L']) ? "and verify " : "", 1576 !dump_opt['L'] ? "nothing leaked " : ""); 1577 } 1578 1579 /* 1580 * Load all space maps as SM_ALLOC maps, then traverse the pool 1581 * claiming each block we discover. If the pool is perfectly 1582 * consistent, the space maps will be empty when we're done. 1583 * Anything left over is a leak; any block we can't claim (because 1584 * it's not part of any space map) is a double allocation, 1585 * reference to a freed block, or an unclaimed log block. 1586 */ 1587 if (!dump_opt['L']) 1588 zdb_leak_init(spa); 1589 1590 /* 1591 * If there's a deferred-free bplist, process that first. 1592 */ 1593 if (spa->spa_sync_bplist_obj != 0) { 1594 bplist_t *bpl = &spa->spa_sync_bplist; 1595 blkptr_t blk; 1596 uint64_t itor = 0; 1597 1598 VERIFY(0 == bplist_open(bpl, spa->spa_meta_objset, 1599 spa->spa_sync_bplist_obj)); 1600 1601 while (bplist_iterate(bpl, &itor, &blk) == 0) { 1602 if (dump_opt['b'] >= 4) { 1603 char blkbuf[BP_SPRINTF_LEN]; 1604 sprintf_blkptr(blkbuf, BP_SPRINTF_LEN, &blk); 1605 (void) printf("[%s] %s\n", 1606 "deferred free", blkbuf); 1607 } 1608 zdb_count_block(spa, &zcb, &blk, DMU_OT_DEFERRED); 1609 } 1610 1611 bplist_close(bpl); 1612 } 1613 1614 zcb.zcb_haderrors |= traverse_pool(spa, zdb_blkptr_cb, &zcb); 1615 1616 if (zcb.zcb_haderrors && !dump_opt['S']) { 1617 (void) printf("\nError counts:\n\n"); 1618 (void) printf("\t%5s %s\n", "errno", "count"); 1619 for (e = 0; e < 256; e++) { 1620 if (zcb.zcb_errors[e] != 0) { 1621 (void) printf("\t%5d %llu\n", 1622 e, (u_longlong_t)zcb.zcb_errors[e]); 1623 } 1624 } 1625 } 1626 1627 /* 1628 * Report any leaked segments. 1629 */ 1630 if (!dump_opt['L']) 1631 zdb_leak_fini(spa); 1632 1633 /* 1634 * If we're interested in printing out the blkptr signatures, 1635 * return now as we don't print out anything else (including 1636 * errors and leaks). 1637 */ 1638 if (dump_opt['S']) 1639 return (zcb.zcb_haderrors ? 3 : 0); 1640 1641 alloc = spa_get_alloc(spa); 1642 space = spa_get_space(spa); 1643 1644 /* 1645 * Log blocks allocated from a separate log device don't count 1646 * as part of the normal pool space; factor them in here. 1647 */ 1648 logalloc = 0; 1649 1650 for (c = 0; c < rvd->vdev_children; c++) 1651 if (rvd->vdev_child[c]->vdev_islog) 1652 logalloc += rvd->vdev_child[c]->vdev_stat.vs_alloc; 1653 1654 tzb = &zcb.zcb_type[ZB_TOTAL][DMU_OT_TOTAL]; 1655 1656 if (tzb->zb_asize == alloc + logalloc) { 1657 if (!dump_opt['L']) 1658 (void) printf("\n\tNo leaks (block sum matches space" 1659 " maps exactly)\n"); 1660 } else { 1661 (void) printf("block traversal size %llu != alloc %llu " 1662 "(%s %lld)\n", 1663 (u_longlong_t)tzb->zb_asize, 1664 (u_longlong_t)alloc + logalloc, 1665 (dump_opt['L']) ? "unreachable" : "leaked", 1666 (longlong_t)(alloc + logalloc - tzb->zb_asize)); 1667 leaks = 1; 1668 } 1669 1670 if (tzb->zb_count == 0) 1671 return (2); 1672 1673 (void) printf("\n"); 1674 (void) printf("\tbp count: %10llu\n", 1675 (u_longlong_t)tzb->zb_count); 1676 (void) printf("\tbp logical: %10llu\t avg: %6llu\n", 1677 (u_longlong_t)tzb->zb_lsize, 1678 (u_longlong_t)(tzb->zb_lsize / tzb->zb_count)); 1679 (void) printf("\tbp physical: %10llu\t avg:" 1680 " %6llu\tcompression: %6.2f\n", 1681 (u_longlong_t)tzb->zb_psize, 1682 (u_longlong_t)(tzb->zb_psize / tzb->zb_count), 1683 (double)tzb->zb_lsize / tzb->zb_psize); 1684 (void) printf("\tbp allocated: %10llu\t avg:" 1685 " %6llu\tcompression: %6.2f\n", 1686 (u_longlong_t)tzb->zb_asize, 1687 (u_longlong_t)(tzb->zb_asize / tzb->zb_count), 1688 (double)tzb->zb_lsize / tzb->zb_asize); 1689 (void) printf("\tSPA allocated: %10llu\tused: %5.2f%%\n", 1690 (u_longlong_t)alloc, 100.0 * alloc / space); 1691 1692 if (dump_opt['b'] >= 2) { 1693 int l, t, level; 1694 (void) printf("\nBlocks\tLSIZE\tPSIZE\tASIZE" 1695 "\t avg\t comp\t%%Total\tType\n"); 1696 1697 for (t = 0; t <= DMU_OT_NUMTYPES; t++) { 1698 char csize[6], lsize[6], psize[6], asize[6], avg[6]; 1699 char *typename; 1700 1701 typename = t == DMU_OT_DEFERRED ? "deferred free" : 1702 t == DMU_OT_TOTAL ? "Total" : dmu_ot[t].ot_name; 1703 1704 if (zcb.zcb_type[ZB_TOTAL][t].zb_asize == 0) { 1705 (void) printf("%6s\t%5s\t%5s\t%5s" 1706 "\t%5s\t%5s\t%6s\t%s\n", 1707 "-", 1708 "-", 1709 "-", 1710 "-", 1711 "-", 1712 "-", 1713 "-", 1714 typename); 1715 continue; 1716 } 1717 1718 for (l = ZB_TOTAL - 1; l >= -1; l--) { 1719 level = (l == -1 ? ZB_TOTAL : l); 1720 zb = &zcb.zcb_type[level][t]; 1721 1722 if (zb->zb_asize == 0) 1723 continue; 1724 1725 if (dump_opt['b'] < 3 && level != ZB_TOTAL) 1726 continue; 1727 1728 if (level == 0 && zb->zb_asize == 1729 zcb.zcb_type[ZB_TOTAL][t].zb_asize) 1730 continue; 1731 1732 nicenum(zb->zb_count, csize); 1733 nicenum(zb->zb_lsize, lsize); 1734 nicenum(zb->zb_psize, psize); 1735 nicenum(zb->zb_asize, asize); 1736 nicenum(zb->zb_asize / zb->zb_count, avg); 1737 1738 (void) printf("%6s\t%5s\t%5s\t%5s\t%5s" 1739 "\t%5.2f\t%6.2f\t", 1740 csize, lsize, psize, asize, avg, 1741 (double)zb->zb_lsize / zb->zb_psize, 1742 100.0 * zb->zb_asize / tzb->zb_asize); 1743 1744 if (level == ZB_TOTAL) 1745 (void) printf("%s\n", typename); 1746 else 1747 (void) printf(" L%d %s\n", 1748 level, typename); 1749 } 1750 } 1751 } 1752 1753 (void) printf("\n"); 1754 1755 if (leaks) 1756 return (2); 1757 1758 if (zcb.zcb_haderrors) 1759 return (3); 1760 1761 return (0); 1762 } 1763 1764 static void 1765 dump_zpool(spa_t *spa) 1766 { 1767 dsl_pool_t *dp = spa_get_dsl(spa); 1768 int rc = 0; 1769 1770 if (dump_opt['u']) 1771 dump_uberblock(&spa->spa_uberblock); 1772 1773 if (dump_opt['d'] || dump_opt['i'] || dump_opt['m']) { 1774 dump_dir(dp->dp_meta_objset); 1775 if (dump_opt['d'] >= 3) { 1776 dump_bplist(dp->dp_meta_objset, 1777 spa->spa_sync_bplist_obj, "Deferred frees"); 1778 dump_dtl(spa->spa_root_vdev, 0); 1779 } 1780 1781 if (dump_opt['d'] >= 3 || dump_opt['m']) 1782 dump_metaslabs(spa); 1783 1784 (void) dmu_objset_find(spa_name(spa), dump_one_dir, NULL, 1785 DS_FIND_SNAPSHOTS | DS_FIND_CHILDREN); 1786 } 1787 1788 if (dump_opt['b'] || dump_opt['c'] || dump_opt['S']) 1789 rc = dump_block_stats(spa); 1790 1791 if (dump_opt['s']) 1792 show_pool_stats(spa); 1793 1794 if (rc != 0) 1795 exit(rc); 1796 } 1797 1798 #define ZDB_FLAG_CHECKSUM 0x0001 1799 #define ZDB_FLAG_DECOMPRESS 0x0002 1800 #define ZDB_FLAG_BSWAP 0x0004 1801 #define ZDB_FLAG_GBH 0x0008 1802 #define ZDB_FLAG_INDIRECT 0x0010 1803 #define ZDB_FLAG_PHYS 0x0020 1804 #define ZDB_FLAG_RAW 0x0040 1805 #define ZDB_FLAG_PRINT_BLKPTR 0x0080 1806 1807 int flagbits[256]; 1808 1809 static void 1810 zdb_print_blkptr(blkptr_t *bp, int flags) 1811 { 1812 dva_t *dva = bp->blk_dva; 1813 int d; 1814 1815 if (flags & ZDB_FLAG_BSWAP) 1816 byteswap_uint64_array((void *)bp, sizeof (blkptr_t)); 1817 /* 1818 * Super-ick warning: This code is also duplicated in 1819 * cmd/mdb/common/modules/zfs/zfs.c . Yeah, I hate code 1820 * replication, too. 1821 */ 1822 for (d = 0; d < BP_GET_NDVAS(bp); d++) { 1823 (void) printf("\tDVA[%d]: vdev_id %lld / %llx\n", d, 1824 (longlong_t)DVA_GET_VDEV(&dva[d]), 1825 (longlong_t)DVA_GET_OFFSET(&dva[d])); 1826 (void) printf("\tDVA[%d]: GANG: %-5s GRID: %04llx\t" 1827 "ASIZE: %llx\n", d, 1828 DVA_GET_GANG(&dva[d]) ? "TRUE" : "FALSE", 1829 (longlong_t)DVA_GET_GRID(&dva[d]), 1830 (longlong_t)DVA_GET_ASIZE(&dva[d])); 1831 (void) printf("\tDVA[%d]: :%llu:%llx:%llx:%s%s%s%s\n", d, 1832 (u_longlong_t)DVA_GET_VDEV(&dva[d]), 1833 (longlong_t)DVA_GET_OFFSET(&dva[d]), 1834 (longlong_t)BP_GET_PSIZE(bp), 1835 BP_SHOULD_BYTESWAP(bp) ? "e" : "", 1836 !DVA_GET_GANG(&dva[d]) && BP_GET_LEVEL(bp) != 0 ? 1837 "d" : "", 1838 DVA_GET_GANG(&dva[d]) ? "g" : "", 1839 BP_GET_COMPRESS(bp) != 0 ? "d" : ""); 1840 } 1841 (void) printf("\tLSIZE: %-16llx\t\tPSIZE: %llx\n", 1842 (longlong_t)BP_GET_LSIZE(bp), (longlong_t)BP_GET_PSIZE(bp)); 1843 (void) printf("\tENDIAN: %6s\t\t\t\t\tTYPE: %s\n", 1844 BP_GET_BYTEORDER(bp) ? "LITTLE" : "BIG", 1845 dmu_ot[BP_GET_TYPE(bp)].ot_name); 1846 (void) printf("\tBIRTH: %-16llx LEVEL: %-2llu\tFILL: %llx\n", 1847 (u_longlong_t)bp->blk_birth, (u_longlong_t)BP_GET_LEVEL(bp), 1848 (u_longlong_t)bp->blk_fill); 1849 (void) printf("\tCKFUNC: %-16s\t\tCOMP: %s\n", 1850 zio_checksum_table[BP_GET_CHECKSUM(bp)].ci_name, 1851 zio_compress_table[BP_GET_COMPRESS(bp)].ci_name); 1852 (void) printf("\tCKSUM: %llx:%llx:%llx:%llx\n", 1853 (u_longlong_t)bp->blk_cksum.zc_word[0], 1854 (u_longlong_t)bp->blk_cksum.zc_word[1], 1855 (u_longlong_t)bp->blk_cksum.zc_word[2], 1856 (u_longlong_t)bp->blk_cksum.zc_word[3]); 1857 } 1858 1859 static void 1860 zdb_dump_indirect(blkptr_t *bp, int nbps, int flags) 1861 { 1862 int i; 1863 1864 for (i = 0; i < nbps; i++) 1865 zdb_print_blkptr(&bp[i], flags); 1866 } 1867 1868 static void 1869 zdb_dump_gbh(void *buf, int flags) 1870 { 1871 zdb_dump_indirect((blkptr_t *)buf, SPA_GBH_NBLKPTRS, flags); 1872 } 1873 1874 static void 1875 zdb_dump_block_raw(void *buf, uint64_t size, int flags) 1876 { 1877 if (flags & ZDB_FLAG_BSWAP) 1878 byteswap_uint64_array(buf, size); 1879 (void) write(2, buf, size); 1880 } 1881 1882 static void 1883 zdb_dump_block(char *label, void *buf, uint64_t size, int flags) 1884 { 1885 uint64_t *d = (uint64_t *)buf; 1886 int nwords = size / sizeof (uint64_t); 1887 int do_bswap = !!(flags & ZDB_FLAG_BSWAP); 1888 int i, j; 1889 char *hdr, *c; 1890 1891 1892 if (do_bswap) 1893 hdr = " 7 6 5 4 3 2 1 0 f e d c b a 9 8"; 1894 else 1895 hdr = " 0 1 2 3 4 5 6 7 8 9 a b c d e f"; 1896 1897 (void) printf("\n%s\n%6s %s 0123456789abcdef\n", label, "", hdr); 1898 1899 for (i = 0; i < nwords; i += 2) { 1900 (void) printf("%06llx: %016llx %016llx ", 1901 (u_longlong_t)(i * sizeof (uint64_t)), 1902 (u_longlong_t)(do_bswap ? BSWAP_64(d[i]) : d[i]), 1903 (u_longlong_t)(do_bswap ? BSWAP_64(d[i + 1]) : d[i + 1])); 1904 1905 c = (char *)&d[i]; 1906 for (j = 0; j < 2 * sizeof (uint64_t); j++) 1907 (void) printf("%c", isprint(c[j]) ? c[j] : '.'); 1908 (void) printf("\n"); 1909 } 1910 } 1911 1912 /* 1913 * There are two acceptable formats: 1914 * leaf_name - For example: c1t0d0 or /tmp/ztest.0a 1915 * child[.child]* - For example: 0.1.1 1916 * 1917 * The second form can be used to specify arbitrary vdevs anywhere 1918 * in the heirarchy. For example, in a pool with a mirror of 1919 * RAID-Zs, you can specify either RAID-Z vdev with 0.0 or 0.1 . 1920 */ 1921 static vdev_t * 1922 zdb_vdev_lookup(vdev_t *vdev, char *path) 1923 { 1924 char *s, *p, *q; 1925 int i; 1926 1927 if (vdev == NULL) 1928 return (NULL); 1929 1930 /* First, assume the x.x.x.x format */ 1931 i = (int)strtoul(path, &s, 10); 1932 if (s == path || (s && *s != '.' && *s != '\0')) 1933 goto name; 1934 if (i < 0 || i >= vdev->vdev_children) 1935 return (NULL); 1936 1937 vdev = vdev->vdev_child[i]; 1938 if (*s == '\0') 1939 return (vdev); 1940 return (zdb_vdev_lookup(vdev, s+1)); 1941 1942 name: 1943 for (i = 0; i < vdev->vdev_children; i++) { 1944 vdev_t *vc = vdev->vdev_child[i]; 1945 1946 if (vc->vdev_path == NULL) { 1947 vc = zdb_vdev_lookup(vc, path); 1948 if (vc == NULL) 1949 continue; 1950 else 1951 return (vc); 1952 } 1953 1954 p = strrchr(vc->vdev_path, '/'); 1955 p = p ? p + 1 : vc->vdev_path; 1956 q = &vc->vdev_path[strlen(vc->vdev_path) - 2]; 1957 1958 if (strcmp(vc->vdev_path, path) == 0) 1959 return (vc); 1960 if (strcmp(p, path) == 0) 1961 return (vc); 1962 if (strcmp(q, "s0") == 0 && strncmp(p, path, q - p) == 0) 1963 return (vc); 1964 } 1965 1966 return (NULL); 1967 } 1968 1969 /* 1970 * Read a block from a pool and print it out. The syntax of the 1971 * block descriptor is: 1972 * 1973 * pool:vdev_specifier:offset:size[:flags] 1974 * 1975 * pool - The name of the pool you wish to read from 1976 * vdev_specifier - Which vdev (see comment for zdb_vdev_lookup) 1977 * offset - offset, in hex, in bytes 1978 * size - Amount of data to read, in hex, in bytes 1979 * flags - A string of characters specifying options 1980 * b: Decode a blkptr at given offset within block 1981 * *c: Calculate and display checksums 1982 * *d: Decompress data before dumping 1983 * e: Byteswap data before dumping 1984 * *g: Display data as a gang block header 1985 * *i: Display as an indirect block 1986 * p: Do I/O to physical offset 1987 * r: Dump raw data to stdout 1988 * 1989 * * = not yet implemented 1990 */ 1991 static void 1992 zdb_read_block(char *thing, spa_t **spap) 1993 { 1994 spa_t *spa = *spap; 1995 int flags = 0; 1996 uint64_t offset = 0, size = 0, blkptr_offset = 0; 1997 zio_t *zio; 1998 vdev_t *vd; 1999 void *buf; 2000 char *s, *p, *dup, *pool, *vdev, *flagstr; 2001 int i, error, zio_flags; 2002 2003 dup = strdup(thing); 2004 s = strtok(dup, ":"); 2005 pool = s ? s : ""; 2006 s = strtok(NULL, ":"); 2007 vdev = s ? s : ""; 2008 s = strtok(NULL, ":"); 2009 offset = strtoull(s ? s : "", NULL, 16); 2010 s = strtok(NULL, ":"); 2011 size = strtoull(s ? s : "", NULL, 16); 2012 s = strtok(NULL, ":"); 2013 flagstr = s ? s : ""; 2014 2015 s = NULL; 2016 if (size == 0) 2017 s = "size must not be zero"; 2018 if (!IS_P2ALIGNED(size, DEV_BSIZE)) 2019 s = "size must be a multiple of sector size"; 2020 if (!IS_P2ALIGNED(offset, DEV_BSIZE)) 2021 s = "offset must be a multiple of sector size"; 2022 if (s) { 2023 (void) printf("Invalid block specifier: %s - %s\n", thing, s); 2024 free(dup); 2025 return; 2026 } 2027 2028 for (s = strtok(flagstr, ":"); s; s = strtok(NULL, ":")) { 2029 for (i = 0; flagstr[i]; i++) { 2030 int bit = flagbits[(uchar_t)flagstr[i]]; 2031 2032 if (bit == 0) { 2033 (void) printf("***Invalid flag: %c\n", 2034 flagstr[i]); 2035 continue; 2036 } 2037 flags |= bit; 2038 2039 /* If it's not something with an argument, keep going */ 2040 if ((bit & (ZDB_FLAG_CHECKSUM | ZDB_FLAG_DECOMPRESS | 2041 ZDB_FLAG_PRINT_BLKPTR)) == 0) 2042 continue; 2043 2044 p = &flagstr[i + 1]; 2045 if (bit == ZDB_FLAG_PRINT_BLKPTR) 2046 blkptr_offset = strtoull(p, &p, 16); 2047 if (*p != ':' && *p != '\0') { 2048 (void) printf("***Invalid flag arg: '%s'\n", s); 2049 free(dup); 2050 return; 2051 } 2052 } 2053 } 2054 2055 if (spa == NULL || strcmp(spa_name(spa), pool) != 0) { 2056 if (spa) 2057 spa_close(spa, (void *)zdb_read_block); 2058 error = spa_open(pool, spap, (void *)zdb_read_block); 2059 if (error) 2060 fatal("Failed to open pool '%s': %s", 2061 pool, strerror(error)); 2062 spa = *spap; 2063 } 2064 2065 vd = zdb_vdev_lookup(spa->spa_root_vdev, vdev); 2066 if (vd == NULL) { 2067 (void) printf("***Invalid vdev: %s\n", vdev); 2068 free(dup); 2069 return; 2070 } else { 2071 if (vd->vdev_path) 2072 (void) printf("Found vdev: %s\n", vd->vdev_path); 2073 else 2074 (void) printf("Found vdev type: %s\n", 2075 vd->vdev_ops->vdev_op_type); 2076 } 2077 2078 buf = umem_alloc(size, UMEM_NOFAIL); 2079 2080 zio_flags = ZIO_FLAG_DONT_CACHE | ZIO_FLAG_DONT_QUEUE | 2081 ZIO_FLAG_DONT_PROPAGATE | ZIO_FLAG_DONT_RETRY; 2082 2083 spa_config_enter(spa, SCL_STATE, FTAG, RW_READER); 2084 zio = zio_root(spa, NULL, NULL, 0); 2085 /* XXX todo - cons up a BP so RAID-Z will be happy */ 2086 zio_nowait(zio_vdev_child_io(zio, NULL, vd, offset, buf, size, 2087 ZIO_TYPE_READ, ZIO_PRIORITY_SYNC_READ, zio_flags, NULL, NULL)); 2088 error = zio_wait(zio); 2089 spa_config_exit(spa, SCL_STATE, FTAG); 2090 2091 if (error) { 2092 (void) printf("Read of %s failed, error: %d\n", thing, error); 2093 goto out; 2094 } 2095 2096 if (flags & ZDB_FLAG_PRINT_BLKPTR) 2097 zdb_print_blkptr((blkptr_t *)(void *) 2098 ((uintptr_t)buf + (uintptr_t)blkptr_offset), flags); 2099 else if (flags & ZDB_FLAG_RAW) 2100 zdb_dump_block_raw(buf, size, flags); 2101 else if (flags & ZDB_FLAG_INDIRECT) 2102 zdb_dump_indirect((blkptr_t *)buf, size / sizeof (blkptr_t), 2103 flags); 2104 else if (flags & ZDB_FLAG_GBH) 2105 zdb_dump_gbh(buf, flags); 2106 else 2107 zdb_dump_block(thing, buf, size, flags); 2108 2109 out: 2110 umem_free(buf, size); 2111 free(dup); 2112 } 2113 2114 static boolean_t 2115 nvlist_string_match(nvlist_t *config, char *name, char *tgt) 2116 { 2117 char *s; 2118 2119 if (nvlist_lookup_string(config, name, &s) != 0) 2120 return (B_FALSE); 2121 2122 return (strcmp(s, tgt) == 0); 2123 } 2124 2125 static boolean_t 2126 nvlist_uint64_match(nvlist_t *config, char *name, uint64_t tgt) 2127 { 2128 uint64_t val; 2129 2130 if (nvlist_lookup_uint64(config, name, &val) != 0) 2131 return (B_FALSE); 2132 2133 return (val == tgt); 2134 } 2135 2136 static boolean_t 2137 vdev_child_guid_match(nvlist_t *vdev, uint64_t guid) 2138 { 2139 nvlist_t **child; 2140 uint_t c, children; 2141 2142 verify(nvlist_lookup_nvlist_array(vdev, ZPOOL_CONFIG_CHILDREN, 2143 &child, &children) == 0); 2144 for (c = 0; c < children; ++c) 2145 if (nvlist_uint64_match(child[c], ZPOOL_CONFIG_GUID, guid)) 2146 return (B_TRUE); 2147 return (B_FALSE); 2148 } 2149 2150 static boolean_t 2151 vdev_child_string_match(nvlist_t *vdev, char *tgt) 2152 { 2153 nvlist_t **child; 2154 uint_t c, children; 2155 2156 verify(nvlist_lookup_nvlist_array(vdev, ZPOOL_CONFIG_CHILDREN, 2157 &child, &children) == 0); 2158 for (c = 0; c < children; ++c) { 2159 if (nvlist_string_match(child[c], ZPOOL_CONFIG_PATH, tgt) || 2160 nvlist_string_match(child[c], ZPOOL_CONFIG_DEVID, tgt)) 2161 return (B_TRUE); 2162 } 2163 return (B_FALSE); 2164 } 2165 2166 static boolean_t 2167 vdev_guid_match(nvlist_t *config, uint64_t guid) 2168 { 2169 nvlist_t *nvroot; 2170 2171 verify(nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE, 2172 &nvroot) == 0); 2173 2174 return (nvlist_uint64_match(nvroot, ZPOOL_CONFIG_GUID, guid) || 2175 vdev_child_guid_match(nvroot, guid)); 2176 } 2177 2178 static boolean_t 2179 vdev_string_match(nvlist_t *config, char *tgt) 2180 { 2181 nvlist_t *nvroot; 2182 2183 verify(nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE, 2184 &nvroot) == 0); 2185 2186 return (vdev_child_string_match(nvroot, tgt)); 2187 } 2188 2189 static boolean_t 2190 pool_match(nvlist_t *config, char *tgt) 2191 { 2192 uint64_t guid = strtoull(tgt, NULL, 0); 2193 2194 if (guid != 0) { 2195 return ( 2196 nvlist_uint64_match(config, ZPOOL_CONFIG_POOL_GUID, guid) || 2197 vdev_guid_match(config, guid)); 2198 } else { 2199 return ( 2200 nvlist_string_match(config, ZPOOL_CONFIG_POOL_NAME, tgt) || 2201 vdev_string_match(config, tgt)); 2202 } 2203 } 2204 2205 static int 2206 find_exported_zpool(char *pool_id, nvlist_t **configp, char *vdev_dir) 2207 { 2208 nvlist_t *pools; 2209 int error = ENOENT; 2210 nvlist_t *match = NULL; 2211 2212 if (vdev_dir != NULL) 2213 pools = zpool_find_import_activeok(g_zfs, 1, &vdev_dir); 2214 else 2215 pools = zpool_find_import_activeok(g_zfs, 0, NULL); 2216 2217 if (pools != NULL) { 2218 nvpair_t *elem = NULL; 2219 2220 while ((elem = nvlist_next_nvpair(pools, elem)) != NULL) { 2221 verify(nvpair_value_nvlist(elem, configp) == 0); 2222 if (pool_match(*configp, pool_id)) { 2223 if (match != NULL) { 2224 (void) fatal( 2225 "More than one matching pool - " 2226 "specify guid/devid/device path."); 2227 } else { 2228 match = *configp; 2229 error = 0; 2230 } 2231 } 2232 } 2233 } 2234 2235 *configp = error ? NULL : match; 2236 2237 return (error); 2238 } 2239 2240 int 2241 main(int argc, char **argv) 2242 { 2243 int i, c; 2244 struct rlimit rl = { 1024, 1024 }; 2245 spa_t *spa; 2246 objset_t *os = NULL; 2247 char *endstr; 2248 int dump_all = 1; 2249 int verbose = 0; 2250 int error; 2251 int exported = 0; 2252 char *vdev_dir = NULL; 2253 2254 (void) setrlimit(RLIMIT_NOFILE, &rl); 2255 (void) enable_extended_FILE_stdio(-1, -1); 2256 2257 dprintf_setup(&argc, argv); 2258 2259 while ((c = getopt(argc, argv, "udibcmsvCLS:U:lRep:t:")) != -1) { 2260 switch (c) { 2261 case 'u': 2262 case 'd': 2263 case 'i': 2264 case 'b': 2265 case 'c': 2266 case 'm': 2267 case 's': 2268 case 'C': 2269 case 'l': 2270 case 'R': 2271 dump_opt[c]++; 2272 dump_all = 0; 2273 break; 2274 case 'L': 2275 dump_opt[c]++; 2276 break; 2277 case 'v': 2278 verbose++; 2279 break; 2280 case 'U': 2281 spa_config_path = optarg; 2282 break; 2283 case 'e': 2284 exported = 1; 2285 break; 2286 case 'p': 2287 vdev_dir = optarg; 2288 break; 2289 case 'S': 2290 dump_opt[c]++; 2291 dump_all = 0; 2292 zdb_sig_user_data = (strncmp(optarg, "user:", 5) == 0); 2293 if (!zdb_sig_user_data && strncmp(optarg, "all:", 4)) 2294 usage(); 2295 endstr = strchr(optarg, ':') + 1; 2296 if (strcmp(endstr, "fletcher2") == 0) 2297 zdb_sig_cksumalg = ZIO_CHECKSUM_FLETCHER_2; 2298 else if (strcmp(endstr, "fletcher4") == 0) 2299 zdb_sig_cksumalg = ZIO_CHECKSUM_FLETCHER_4; 2300 else if (strcmp(endstr, "sha256") == 0) 2301 zdb_sig_cksumalg = ZIO_CHECKSUM_SHA256; 2302 else if (strcmp(endstr, "all") == 0) 2303 zdb_sig_cksumalg = ZIO_CHECKSUM_FLETCHER_2; 2304 else 2305 usage(); 2306 break; 2307 case 't': 2308 ub_max_txg = strtoull(optarg, NULL, 0); 2309 if (ub_max_txg < TXG_INITIAL) { 2310 (void) fprintf(stderr, "incorrect txg " 2311 "specified: %s\n", optarg); 2312 usage(); 2313 } 2314 break; 2315 default: 2316 usage(); 2317 break; 2318 } 2319 } 2320 2321 if (vdev_dir != NULL && exported == 0) { 2322 (void) fprintf(stderr, "-p option requires use of -e\n"); 2323 usage(); 2324 } 2325 2326 kernel_init(FREAD); 2327 g_zfs = libzfs_init(); 2328 ASSERT(g_zfs != NULL); 2329 2330 for (c = 0; c < 256; c++) { 2331 if (dump_all && c != 'l' && c != 'R') 2332 dump_opt[c] = 1; 2333 if (dump_opt[c]) 2334 dump_opt[c] += verbose; 2335 } 2336 2337 argc -= optind; 2338 argv += optind; 2339 2340 if (argc < 1) { 2341 if (dump_opt['C']) { 2342 dump_cachefile(spa_config_path); 2343 return (0); 2344 } 2345 usage(); 2346 } 2347 2348 if (dump_opt['l']) { 2349 dump_label(argv[0]); 2350 return (0); 2351 } 2352 2353 if (dump_opt['R']) { 2354 flagbits['b'] = ZDB_FLAG_PRINT_BLKPTR; 2355 flagbits['c'] = ZDB_FLAG_CHECKSUM; 2356 flagbits['d'] = ZDB_FLAG_DECOMPRESS; 2357 flagbits['e'] = ZDB_FLAG_BSWAP; 2358 flagbits['g'] = ZDB_FLAG_GBH; 2359 flagbits['i'] = ZDB_FLAG_INDIRECT; 2360 flagbits['p'] = ZDB_FLAG_PHYS; 2361 flagbits['r'] = ZDB_FLAG_RAW; 2362 2363 spa = NULL; 2364 while (argv[0]) { 2365 zdb_read_block(argv[0], &spa); 2366 argv++; 2367 argc--; 2368 } 2369 if (spa) 2370 spa_close(spa, (void *)zdb_read_block); 2371 return (0); 2372 } 2373 2374 if (dump_opt['C']) 2375 dump_config(argv[0]); 2376 2377 error = 0; 2378 if (exported) { 2379 /* 2380 * Check to see if the name refers to an exported zpool 2381 */ 2382 char *slash; 2383 nvlist_t *exported_conf = NULL; 2384 2385 if ((slash = strchr(argv[0], '/')) != NULL) 2386 *slash = '\0'; 2387 2388 error = find_exported_zpool(argv[0], &exported_conf, vdev_dir); 2389 if (error == 0) { 2390 nvlist_t *nvl = NULL; 2391 2392 if (vdev_dir != NULL) { 2393 if (nvlist_alloc(&nvl, NV_UNIQUE_NAME, 0) != 0) 2394 error = ENOMEM; 2395 else if (nvlist_add_string(nvl, 2396 zpool_prop_to_name(ZPOOL_PROP_ALTROOT), 2397 vdev_dir) != 0) 2398 error = ENOMEM; 2399 } 2400 2401 if (error == 0) 2402 error = spa_import_verbatim(argv[0], 2403 exported_conf, nvl); 2404 2405 nvlist_free(nvl); 2406 } 2407 2408 if (slash != NULL) 2409 *slash = '/'; 2410 } 2411 2412 if (error == 0) { 2413 if (strchr(argv[0], '/') != NULL) { 2414 error = dmu_objset_own(argv[0], DMU_OST_ANY, 2415 B_TRUE, FTAG, &os); 2416 } else { 2417 error = spa_open(argv[0], &spa, FTAG); 2418 } 2419 } 2420 2421 if (error) 2422 fatal("can't open %s: %s", argv[0], strerror(error)); 2423 2424 argv++; 2425 if (--argc > 0) { 2426 zopt_objects = argc; 2427 zopt_object = calloc(zopt_objects, sizeof (uint64_t)); 2428 for (i = 0; i < zopt_objects; i++) { 2429 errno = 0; 2430 zopt_object[i] = strtoull(argv[i], NULL, 0); 2431 if (zopt_object[i] == 0 && errno != 0) 2432 fatal("bad object number %s: %s", 2433 argv[i], strerror(errno)); 2434 } 2435 } 2436 2437 if (os != NULL) { 2438 dump_dir(os); 2439 dmu_objset_disown(os, FTAG); 2440 } else { 2441 dump_zpool(spa); 2442 spa_close(spa, FTAG); 2443 } 2444 2445 fuid_table_destroy(); 2446 2447 libzfs_fini(g_zfs); 2448 kernel_fini(); 2449 2450 return (0); 2451 } 2452