1 /* 2 * CDDL HEADER START 3 * 4 * The contents of this file are subject to the terms of the 5 * Common Development and Distribution License (the "License"). 6 * You may not use this file except in compliance with the License. 7 * 8 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE 9 * or http://www.opensolaris.org/os/licensing. 10 * See the License for the specific language governing permissions 11 * and limitations under the License. 12 * 13 * When distributing Covered Code, include this CDDL HEADER in each 14 * file and include the License file at usr/src/OPENSOLARIS.LICENSE. 15 * If applicable, add the following below this CDDL HEADER, with the 16 * fields enclosed by brackets "[]" replaced with your own identifying 17 * information: Portions Copyright [yyyy] [name of copyright owner] 18 * 19 * CDDL HEADER END 20 */ 21 /* 22 * Copyright 2008 Sun Microsystems, Inc. All rights reserved. 23 * Use is subject to license terms. 24 */ 25 26 #include <stdio.h> 27 #include <stdio_ext.h> 28 #include <stdlib.h> 29 #include <ctype.h> 30 #include <sys/zfs_context.h> 31 #include <sys/spa.h> 32 #include <sys/spa_impl.h> 33 #include <sys/dmu.h> 34 #include <sys/zap.h> 35 #include <sys/fs/zfs.h> 36 #include <sys/zfs_znode.h> 37 #include <sys/vdev.h> 38 #include <sys/vdev_impl.h> 39 #include <sys/metaslab_impl.h> 40 #include <sys/dmu_objset.h> 41 #include <sys/dsl_dir.h> 42 #include <sys/dsl_dataset.h> 43 #include <sys/dsl_pool.h> 44 #include <sys/dbuf.h> 45 #include <sys/zil.h> 46 #include <sys/zil_impl.h> 47 #include <sys/stat.h> 48 #include <sys/resource.h> 49 #include <sys/dmu_traverse.h> 50 #include <sys/zio_checksum.h> 51 #include <sys/zio_compress.h> 52 #include <sys/zfs_fuid.h> 53 #include <sys/arc.h> 54 #undef ZFS_MAXNAMELEN 55 #undef verify 56 #include <libzfs.h> 57 58 const char cmdname[] = "zdb"; 59 uint8_t dump_opt[256]; 60 61 typedef void object_viewer_t(objset_t *, uint64_t, void *data, size_t size); 62 63 extern void dump_intent_log(zilog_t *); 64 uint64_t *zopt_object = NULL; 65 int zopt_objects = 0; 66 libzfs_handle_t *g_zfs; 67 boolean_t zdb_sig_user_data = B_TRUE; 68 int zdb_sig_cksumalg = ZIO_CHECKSUM_SHA256; 69 70 /* 71 * These libumem hooks provide a reasonable set of defaults for the allocator's 72 * debugging facilities. 73 */ 74 const char * 75 _umem_debug_init() 76 { 77 return ("default,verbose"); /* $UMEM_DEBUG setting */ 78 } 79 80 const char * 81 _umem_logging_init(void) 82 { 83 return ("fail,contents"); /* $UMEM_LOGGING setting */ 84 } 85 86 static void 87 usage(void) 88 { 89 (void) fprintf(stderr, 90 "Usage: %s [-udibcsvL] [-U cachefile_path] [-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 static void 129 fatal(const char *fmt, ...) 130 { 131 va_list ap; 132 133 va_start(ap, fmt); 134 (void) fprintf(stderr, "%s: ", cmdname); 135 (void) vfprintf(stderr, fmt, ap); 136 va_end(ap); 137 (void) fprintf(stderr, "\n"); 138 139 abort(); 140 } 141 142 static void 143 dump_nvlist(nvlist_t *list, int indent) 144 { 145 nvpair_t *elem = NULL; 146 147 while ((elem = nvlist_next_nvpair(list, elem)) != NULL) { 148 switch (nvpair_type(elem)) { 149 case DATA_TYPE_STRING: 150 { 151 char *value; 152 153 VERIFY(nvpair_value_string(elem, &value) == 0); 154 (void) printf("%*s%s='%s'\n", indent, "", 155 nvpair_name(elem), value); 156 } 157 break; 158 159 case DATA_TYPE_UINT64: 160 { 161 uint64_t value; 162 163 VERIFY(nvpair_value_uint64(elem, &value) == 0); 164 (void) printf("%*s%s=%llu\n", indent, "", 165 nvpair_name(elem), (u_longlong_t)value); 166 } 167 break; 168 169 case DATA_TYPE_NVLIST: 170 { 171 nvlist_t *value; 172 173 VERIFY(nvpair_value_nvlist(elem, &value) == 0); 174 (void) printf("%*s%s\n", indent, "", 175 nvpair_name(elem)); 176 dump_nvlist(value, indent + 4); 177 } 178 break; 179 180 case DATA_TYPE_NVLIST_ARRAY: 181 { 182 nvlist_t **value; 183 uint_t c, count; 184 185 VERIFY(nvpair_value_nvlist_array(elem, &value, 186 &count) == 0); 187 188 for (c = 0; c < count; c++) { 189 (void) printf("%*s%s[%u]\n", indent, "", 190 nvpair_name(elem), c); 191 dump_nvlist(value[c], indent + 8); 192 } 193 } 194 break; 195 196 default: 197 198 (void) printf("bad config type %d for %s\n", 199 nvpair_type(elem), nvpair_name(elem)); 200 } 201 } 202 } 203 204 /* ARGSUSED */ 205 static void 206 dump_packed_nvlist(objset_t *os, uint64_t object, void *data, size_t size) 207 { 208 nvlist_t *nv; 209 size_t nvsize = *(uint64_t *)data; 210 char *packed = umem_alloc(nvsize, UMEM_NOFAIL); 211 212 VERIFY(0 == dmu_read(os, object, 0, nvsize, packed)); 213 214 VERIFY(nvlist_unpack(packed, nvsize, &nv, 0) == 0); 215 216 umem_free(packed, nvsize); 217 218 dump_nvlist(nv, 8); 219 220 nvlist_free(nv); 221 } 222 223 const char dump_zap_stars[] = "****************************************"; 224 const int dump_zap_width = sizeof (dump_zap_stars) - 1; 225 226 static void 227 dump_zap_histogram(uint64_t histo[ZAP_HISTOGRAM_SIZE]) 228 { 229 int i; 230 int minidx = ZAP_HISTOGRAM_SIZE - 1; 231 int maxidx = 0; 232 uint64_t max = 0; 233 234 for (i = 0; i < ZAP_HISTOGRAM_SIZE; i++) { 235 if (histo[i] > max) 236 max = histo[i]; 237 if (histo[i] > 0 && i > maxidx) 238 maxidx = i; 239 if (histo[i] > 0 && i < minidx) 240 minidx = i; 241 } 242 243 if (max < dump_zap_width) 244 max = dump_zap_width; 245 246 for (i = minidx; i <= maxidx; i++) 247 (void) printf("\t\t\t%u: %6llu %s\n", i, (u_longlong_t)histo[i], 248 &dump_zap_stars[(max - histo[i]) * dump_zap_width / max]); 249 } 250 251 static void 252 dump_zap_stats(objset_t *os, uint64_t object) 253 { 254 int error; 255 zap_stats_t zs; 256 257 error = zap_get_stats(os, object, &zs); 258 if (error) 259 return; 260 261 if (zs.zs_ptrtbl_len == 0) { 262 ASSERT(zs.zs_num_blocks == 1); 263 (void) printf("\tmicrozap: %llu bytes, %llu entries\n", 264 (u_longlong_t)zs.zs_blocksize, 265 (u_longlong_t)zs.zs_num_entries); 266 return; 267 } 268 269 (void) printf("\tFat ZAP stats:\n"); 270 271 (void) printf("\t\tPointer table:\n"); 272 (void) printf("\t\t\t%llu elements\n", 273 (u_longlong_t)zs.zs_ptrtbl_len); 274 (void) printf("\t\t\tzt_blk: %llu\n", 275 (u_longlong_t)zs.zs_ptrtbl_zt_blk); 276 (void) printf("\t\t\tzt_numblks: %llu\n", 277 (u_longlong_t)zs.zs_ptrtbl_zt_numblks); 278 (void) printf("\t\t\tzt_shift: %llu\n", 279 (u_longlong_t)zs.zs_ptrtbl_zt_shift); 280 (void) printf("\t\t\tzt_blks_copied: %llu\n", 281 (u_longlong_t)zs.zs_ptrtbl_blks_copied); 282 (void) printf("\t\t\tzt_nextblk: %llu\n", 283 (u_longlong_t)zs.zs_ptrtbl_nextblk); 284 285 (void) printf("\t\tZAP entries: %llu\n", 286 (u_longlong_t)zs.zs_num_entries); 287 (void) printf("\t\tLeaf blocks: %llu\n", 288 (u_longlong_t)zs.zs_num_leafs); 289 (void) printf("\t\tTotal blocks: %llu\n", 290 (u_longlong_t)zs.zs_num_blocks); 291 (void) printf("\t\tzap_block_type: 0x%llx\n", 292 (u_longlong_t)zs.zs_block_type); 293 (void) printf("\t\tzap_magic: 0x%llx\n", 294 (u_longlong_t)zs.zs_magic); 295 (void) printf("\t\tzap_salt: 0x%llx\n", 296 (u_longlong_t)zs.zs_salt); 297 298 (void) printf("\t\tLeafs with 2^n pointers:\n"); 299 dump_zap_histogram(zs.zs_leafs_with_2n_pointers); 300 301 (void) printf("\t\tBlocks with n*5 entries:\n"); 302 dump_zap_histogram(zs.zs_blocks_with_n5_entries); 303 304 (void) printf("\t\tBlocks n/10 full:\n"); 305 dump_zap_histogram(zs.zs_blocks_n_tenths_full); 306 307 (void) printf("\t\tEntries with n chunks:\n"); 308 dump_zap_histogram(zs.zs_entries_using_n_chunks); 309 310 (void) printf("\t\tBuckets with n entries:\n"); 311 dump_zap_histogram(zs.zs_buckets_with_n_entries); 312 } 313 314 /*ARGSUSED*/ 315 static void 316 dump_none(objset_t *os, uint64_t object, void *data, size_t size) 317 { 318 } 319 320 /*ARGSUSED*/ 321 void 322 dump_uint8(objset_t *os, uint64_t object, void *data, size_t size) 323 { 324 } 325 326 /*ARGSUSED*/ 327 static void 328 dump_uint64(objset_t *os, uint64_t object, void *data, size_t size) 329 { 330 } 331 332 /*ARGSUSED*/ 333 static void 334 dump_zap(objset_t *os, uint64_t object, void *data, size_t size) 335 { 336 zap_cursor_t zc; 337 zap_attribute_t attr; 338 void *prop; 339 int i; 340 341 dump_zap_stats(os, object); 342 (void) printf("\n"); 343 344 for (zap_cursor_init(&zc, os, object); 345 zap_cursor_retrieve(&zc, &attr) == 0; 346 zap_cursor_advance(&zc)) { 347 (void) printf("\t\t%s = ", attr.za_name); 348 if (attr.za_num_integers == 0) { 349 (void) printf("\n"); 350 continue; 351 } 352 prop = umem_zalloc(attr.za_num_integers * 353 attr.za_integer_length, UMEM_NOFAIL); 354 (void) zap_lookup(os, object, attr.za_name, 355 attr.za_integer_length, attr.za_num_integers, prop); 356 if (attr.za_integer_length == 1) { 357 (void) printf("%s", (char *)prop); 358 } else { 359 for (i = 0; i < attr.za_num_integers; i++) { 360 switch (attr.za_integer_length) { 361 case 2: 362 (void) printf("%u ", 363 ((uint16_t *)prop)[i]); 364 break; 365 case 4: 366 (void) printf("%u ", 367 ((uint32_t *)prop)[i]); 368 break; 369 case 8: 370 (void) printf("%lld ", 371 (u_longlong_t)((int64_t *)prop)[i]); 372 break; 373 } 374 } 375 } 376 (void) printf("\n"); 377 umem_free(prop, attr.za_num_integers * attr.za_integer_length); 378 } 379 zap_cursor_fini(&zc); 380 } 381 382 /*ARGSUSED*/ 383 static void 384 dump_zpldir(objset_t *os, uint64_t object, void *data, size_t size) 385 { 386 zap_cursor_t zc; 387 zap_attribute_t attr; 388 const char *typenames[] = { 389 /* 0 */ "not specified", 390 /* 1 */ "FIFO", 391 /* 2 */ "Character Device", 392 /* 3 */ "3 (invalid)", 393 /* 4 */ "Directory", 394 /* 5 */ "5 (invalid)", 395 /* 6 */ "Block Device", 396 /* 7 */ "7 (invalid)", 397 /* 8 */ "Regular File", 398 /* 9 */ "9 (invalid)", 399 /* 10 */ "Symbolic Link", 400 /* 11 */ "11 (invalid)", 401 /* 12 */ "Socket", 402 /* 13 */ "Door", 403 /* 14 */ "Event Port", 404 /* 15 */ "15 (invalid)", 405 }; 406 407 dump_zap_stats(os, object); 408 (void) printf("\n"); 409 410 for (zap_cursor_init(&zc, os, object); 411 zap_cursor_retrieve(&zc, &attr) == 0; 412 zap_cursor_advance(&zc)) { 413 (void) printf("\t\t%s = %lld (type: %s)\n", 414 attr.za_name, ZFS_DIRENT_OBJ(attr.za_first_integer), 415 typenames[ZFS_DIRENT_TYPE(attr.za_first_integer)]); 416 } 417 zap_cursor_fini(&zc); 418 } 419 420 static void 421 dump_spacemap(objset_t *os, space_map_obj_t *smo, space_map_t *sm) 422 { 423 uint64_t alloc, offset, entry; 424 uint8_t mapshift = sm->sm_shift; 425 uint64_t mapstart = sm->sm_start; 426 char *ddata[] = { "ALLOC", "FREE", "CONDENSE", "INVALID", 427 "INVALID", "INVALID", "INVALID", "INVALID" }; 428 429 if (smo->smo_object == 0) 430 return; 431 432 /* 433 * Print out the freelist entries in both encoded and decoded form. 434 */ 435 alloc = 0; 436 for (offset = 0; offset < smo->smo_objsize; offset += sizeof (entry)) { 437 VERIFY(0 == dmu_read(os, smo->smo_object, offset, 438 sizeof (entry), &entry)); 439 if (SM_DEBUG_DECODE(entry)) { 440 (void) printf("\t\t[%4llu] %s: txg %llu, pass %llu\n", 441 (u_longlong_t)(offset / sizeof (entry)), 442 ddata[SM_DEBUG_ACTION_DECODE(entry)], 443 (u_longlong_t)SM_DEBUG_TXG_DECODE(entry), 444 (u_longlong_t)SM_DEBUG_SYNCPASS_DECODE(entry)); 445 } else { 446 (void) printf("\t\t[%4llu] %c range:" 447 " %08llx-%08llx size: %06llx\n", 448 (u_longlong_t)(offset / sizeof (entry)), 449 SM_TYPE_DECODE(entry) == SM_ALLOC ? 'A' : 'F', 450 (u_longlong_t)((SM_OFFSET_DECODE(entry) << 451 mapshift) + mapstart), 452 (u_longlong_t)((SM_OFFSET_DECODE(entry) << 453 mapshift) + mapstart + (SM_RUN_DECODE(entry) << 454 mapshift)), 455 (u_longlong_t)(SM_RUN_DECODE(entry) << mapshift)); 456 if (SM_TYPE_DECODE(entry) == SM_ALLOC) 457 alloc += SM_RUN_DECODE(entry) << mapshift; 458 else 459 alloc -= SM_RUN_DECODE(entry) << mapshift; 460 } 461 } 462 if (alloc != smo->smo_alloc) { 463 (void) printf("space_map_object alloc (%llu) INCONSISTENT " 464 "with space map summary (%llu)\n", 465 (u_longlong_t)smo->smo_alloc, (u_longlong_t)alloc); 466 } 467 } 468 469 static void 470 dump_metaslab(metaslab_t *msp) 471 { 472 char freebuf[5]; 473 space_map_obj_t *smo = &msp->ms_smo; 474 vdev_t *vd = msp->ms_group->mg_vd; 475 spa_t *spa = vd->vdev_spa; 476 477 nicenum(msp->ms_map.sm_size - smo->smo_alloc, freebuf); 478 479 if (dump_opt['d'] <= 5) { 480 (void) printf("\t%10llx %10llu %5s\n", 481 (u_longlong_t)msp->ms_map.sm_start, 482 (u_longlong_t)smo->smo_object, 483 freebuf); 484 return; 485 } 486 487 (void) printf( 488 "\tvdev %llu offset %08llx spacemap %4llu free %5s\n", 489 (u_longlong_t)vd->vdev_id, (u_longlong_t)msp->ms_map.sm_start, 490 (u_longlong_t)smo->smo_object, freebuf); 491 492 ASSERT(msp->ms_map.sm_size == (1ULL << vd->vdev_ms_shift)); 493 494 dump_spacemap(spa->spa_meta_objset, smo, &msp->ms_map); 495 } 496 497 static void 498 dump_metaslabs(spa_t *spa) 499 { 500 vdev_t *rvd = spa->spa_root_vdev; 501 vdev_t *vd; 502 int c, m; 503 504 (void) printf("\nMetaslabs:\n"); 505 506 for (c = 0; c < rvd->vdev_children; c++) { 507 vd = rvd->vdev_child[c]; 508 509 (void) printf("\n vdev %llu\n\n", (u_longlong_t)vd->vdev_id); 510 511 if (dump_opt['d'] <= 5) { 512 (void) printf("\t%10s %10s %5s\n", 513 "offset", "spacemap", "free"); 514 (void) printf("\t%10s %10s %5s\n", 515 "------", "--------", "----"); 516 } 517 for (m = 0; m < vd->vdev_ms_count; m++) 518 dump_metaslab(vd->vdev_ms[m]); 519 (void) printf("\n"); 520 } 521 } 522 523 static void 524 dump_dtl(vdev_t *vd, int indent) 525 { 526 avl_tree_t *t = &vd->vdev_dtl_map.sm_root; 527 space_seg_t *ss; 528 vdev_t *pvd; 529 int c; 530 531 if (indent == 0) 532 (void) printf("\nDirty time logs:\n\n"); 533 534 (void) printf("\t%*s%s\n", indent, "", 535 vd->vdev_path ? vd->vdev_path : 536 vd->vdev_parent ? vd->vdev_ops->vdev_op_type : 537 spa_name(vd->vdev_spa)); 538 539 for (ss = avl_first(t); ss; ss = AVL_NEXT(t, ss)) { 540 /* 541 * Everything in this DTL must appear in all parent DTL unions. 542 */ 543 for (pvd = vd; pvd; pvd = pvd->vdev_parent) 544 ASSERT(vdev_dtl_contains(&pvd->vdev_dtl_map, 545 ss->ss_start, ss->ss_end - ss->ss_start)); 546 (void) printf("\t%*soutage [%llu,%llu] length %llu\n", 547 indent, "", 548 (u_longlong_t)ss->ss_start, 549 (u_longlong_t)ss->ss_end - 1, 550 (u_longlong_t)(ss->ss_end - ss->ss_start)); 551 } 552 553 (void) printf("\n"); 554 555 if (dump_opt['d'] > 5 && vd->vdev_children == 0) { 556 dump_spacemap(vd->vdev_spa->spa_meta_objset, &vd->vdev_dtl, 557 &vd->vdev_dtl_map); 558 (void) printf("\n"); 559 } 560 561 for (c = 0; c < vd->vdev_children; c++) 562 dump_dtl(vd->vdev_child[c], indent + 4); 563 } 564 565 /*ARGSUSED*/ 566 static void 567 dump_dnode(objset_t *os, uint64_t object, void *data, size_t size) 568 { 569 } 570 571 static uint64_t 572 blkid2offset(const dnode_phys_t *dnp, int level, uint64_t blkid) 573 { 574 if (level < 0) 575 return (blkid); 576 577 return ((blkid << (level * (dnp->dn_indblkshift - SPA_BLKPTRSHIFT))) * 578 dnp->dn_datablkszsec << SPA_MINBLOCKSHIFT); 579 } 580 581 static void 582 sprintf_blkptr_compact(char *blkbuf, blkptr_t *bp, int alldvas) 583 { 584 dva_t *dva = bp->blk_dva; 585 int ndvas = alldvas ? BP_GET_NDVAS(bp) : 1; 586 int i; 587 588 blkbuf[0] = '\0'; 589 590 for (i = 0; i < ndvas; i++) 591 (void) sprintf(blkbuf + strlen(blkbuf), "%llu:%llx:%llx ", 592 (u_longlong_t)DVA_GET_VDEV(&dva[i]), 593 (u_longlong_t)DVA_GET_OFFSET(&dva[i]), 594 (u_longlong_t)DVA_GET_ASIZE(&dva[i])); 595 596 (void) sprintf(blkbuf + strlen(blkbuf), "%llxL/%llxP F=%llu B=%llu", 597 (u_longlong_t)BP_GET_LSIZE(bp), 598 (u_longlong_t)BP_GET_PSIZE(bp), 599 (u_longlong_t)bp->blk_fill, 600 (u_longlong_t)bp->blk_birth); 601 } 602 603 static void 604 print_indirect(blkptr_t *bp, const zbookmark_t *zb, 605 const dnode_phys_t *dnp) 606 { 607 char blkbuf[BP_SPRINTF_LEN]; 608 int l; 609 610 ASSERT3U(BP_GET_TYPE(bp), ==, dnp->dn_type); 611 ASSERT3U(BP_GET_LEVEL(bp), ==, zb->zb_level); 612 613 (void) printf("%16llx ", 614 (u_longlong_t)blkid2offset(dnp, zb->zb_level, zb->zb_blkid)); 615 616 ASSERT(zb->zb_level >= 0); 617 618 for (l = dnp->dn_nlevels - 1; l >= -1; l--) { 619 if (l == zb->zb_level) { 620 (void) printf("L%llx", (u_longlong_t)zb->zb_level); 621 } else { 622 (void) printf(" "); 623 } 624 } 625 626 sprintf_blkptr_compact(blkbuf, bp, dump_opt['d'] > 5 ? 1 : 0); 627 (void) printf("%s\n", blkbuf); 628 } 629 630 #define SET_BOOKMARK(zb, objset, object, level, blkid) \ 631 { \ 632 (zb)->zb_objset = objset; \ 633 (zb)->zb_object = object; \ 634 (zb)->zb_level = level; \ 635 (zb)->zb_blkid = blkid; \ 636 } 637 638 static int 639 visit_indirect(spa_t *spa, const dnode_phys_t *dnp, 640 blkptr_t *bp, const zbookmark_t *zb) 641 { 642 int err; 643 644 if (bp->blk_birth == 0) 645 return (0); 646 647 print_indirect(bp, zb, dnp); 648 649 if (BP_GET_LEVEL(bp) > 0) { 650 uint32_t flags = ARC_WAIT; 651 int i; 652 blkptr_t *cbp; 653 int epb = BP_GET_LSIZE(bp) >> SPA_BLKPTRSHIFT; 654 arc_buf_t *buf; 655 uint64_t fill = 0; 656 657 err = arc_read_nolock(NULL, spa, bp, arc_getbuf_func, &buf, 658 ZIO_PRIORITY_ASYNC_READ, ZIO_FLAG_CANFAIL, &flags, zb); 659 if (err) 660 return (err); 661 662 /* recursively visit blocks below this */ 663 cbp = buf->b_data; 664 for (i = 0; i < epb; i++, cbp++) { 665 zbookmark_t czb; 666 667 SET_BOOKMARK(&czb, zb->zb_objset, zb->zb_object, 668 zb->zb_level - 1, 669 zb->zb_blkid * epb + i); 670 err = visit_indirect(spa, dnp, cbp, &czb); 671 if (err) 672 break; 673 fill += cbp->blk_fill; 674 } 675 ASSERT3U(fill, ==, bp->blk_fill); 676 (void) arc_buf_remove_ref(buf, &buf); 677 } 678 679 return (err); 680 } 681 682 /*ARGSUSED*/ 683 static void 684 dump_indirect(dnode_t *dn) 685 { 686 dnode_phys_t *dnp = dn->dn_phys; 687 int j; 688 zbookmark_t czb; 689 690 (void) printf("Indirect blocks:\n"); 691 692 SET_BOOKMARK(&czb, dmu_objset_id(&dn->dn_objset->os), 693 dn->dn_object, dnp->dn_nlevels - 1, 0); 694 for (j = 0; j < dnp->dn_nblkptr; j++) { 695 czb.zb_blkid = j; 696 (void) visit_indirect(dmu_objset_spa(&dn->dn_objset->os), dnp, 697 &dnp->dn_blkptr[j], &czb); 698 } 699 700 (void) printf("\n"); 701 } 702 703 /*ARGSUSED*/ 704 static void 705 dump_dsl_dir(objset_t *os, uint64_t object, void *data, size_t size) 706 { 707 dsl_dir_phys_t *dd = data; 708 time_t crtime; 709 char nice[6]; 710 711 if (dd == NULL) 712 return; 713 714 ASSERT3U(size, >=, sizeof (dsl_dir_phys_t)); 715 716 crtime = dd->dd_creation_time; 717 (void) printf("\t\tcreation_time = %s", ctime(&crtime)); 718 (void) printf("\t\thead_dataset_obj = %llu\n", 719 (u_longlong_t)dd->dd_head_dataset_obj); 720 (void) printf("\t\tparent_dir_obj = %llu\n", 721 (u_longlong_t)dd->dd_parent_obj); 722 (void) printf("\t\torigin_obj = %llu\n", 723 (u_longlong_t)dd->dd_origin_obj); 724 (void) printf("\t\tchild_dir_zapobj = %llu\n", 725 (u_longlong_t)dd->dd_child_dir_zapobj); 726 nicenum(dd->dd_used_bytes, nice); 727 (void) printf("\t\tused_bytes = %s\n", nice); 728 nicenum(dd->dd_compressed_bytes, nice); 729 (void) printf("\t\tcompressed_bytes = %s\n", nice); 730 nicenum(dd->dd_uncompressed_bytes, nice); 731 (void) printf("\t\tuncompressed_bytes = %s\n", nice); 732 nicenum(dd->dd_quota, nice); 733 (void) printf("\t\tquota = %s\n", nice); 734 nicenum(dd->dd_reserved, nice); 735 (void) printf("\t\treserved = %s\n", nice); 736 (void) printf("\t\tprops_zapobj = %llu\n", 737 (u_longlong_t)dd->dd_props_zapobj); 738 (void) printf("\t\tdeleg_zapobj = %llu\n", 739 (u_longlong_t)dd->dd_deleg_zapobj); 740 (void) printf("\t\tflags = %llx\n", 741 (u_longlong_t)dd->dd_flags); 742 743 #define DO(which) \ 744 nicenum(dd->dd_used_breakdown[DD_USED_ ## which], nice); \ 745 (void) printf("\t\tused_breakdown[" #which "] = %s\n", nice) 746 DO(HEAD); 747 DO(SNAP); 748 DO(CHILD); 749 DO(CHILD_RSRV); 750 DO(REFRSRV); 751 #undef DO 752 } 753 754 /*ARGSUSED*/ 755 static void 756 dump_dsl_dataset(objset_t *os, uint64_t object, void *data, size_t size) 757 { 758 dsl_dataset_phys_t *ds = data; 759 time_t crtime; 760 char used[6], compressed[6], uncompressed[6], unique[6]; 761 char blkbuf[BP_SPRINTF_LEN]; 762 763 if (ds == NULL) 764 return; 765 766 ASSERT(size == sizeof (*ds)); 767 crtime = ds->ds_creation_time; 768 nicenum(ds->ds_used_bytes, used); 769 nicenum(ds->ds_compressed_bytes, compressed); 770 nicenum(ds->ds_uncompressed_bytes, uncompressed); 771 nicenum(ds->ds_unique_bytes, unique); 772 sprintf_blkptr(blkbuf, BP_SPRINTF_LEN, &ds->ds_bp); 773 774 (void) printf("\t\tdir_obj = %llu\n", 775 (u_longlong_t)ds->ds_dir_obj); 776 (void) printf("\t\tprev_snap_obj = %llu\n", 777 (u_longlong_t)ds->ds_prev_snap_obj); 778 (void) printf("\t\tprev_snap_txg = %llu\n", 779 (u_longlong_t)ds->ds_prev_snap_txg); 780 (void) printf("\t\tnext_snap_obj = %llu\n", 781 (u_longlong_t)ds->ds_next_snap_obj); 782 (void) printf("\t\tsnapnames_zapobj = %llu\n", 783 (u_longlong_t)ds->ds_snapnames_zapobj); 784 (void) printf("\t\tnum_children = %llu\n", 785 (u_longlong_t)ds->ds_num_children); 786 (void) printf("\t\tcreation_time = %s", ctime(&crtime)); 787 (void) printf("\t\tcreation_txg = %llu\n", 788 (u_longlong_t)ds->ds_creation_txg); 789 (void) printf("\t\tdeadlist_obj = %llu\n", 790 (u_longlong_t)ds->ds_deadlist_obj); 791 (void) printf("\t\tused_bytes = %s\n", used); 792 (void) printf("\t\tcompressed_bytes = %s\n", compressed); 793 (void) printf("\t\tuncompressed_bytes = %s\n", uncompressed); 794 (void) printf("\t\tunique = %s\n", unique); 795 (void) printf("\t\tfsid_guid = %llu\n", 796 (u_longlong_t)ds->ds_fsid_guid); 797 (void) printf("\t\tguid = %llu\n", 798 (u_longlong_t)ds->ds_guid); 799 (void) printf("\t\tflags = %llx\n", 800 (u_longlong_t)ds->ds_flags); 801 (void) printf("\t\tnext_clones_obj = %llu\n", 802 (u_longlong_t)ds->ds_next_clones_obj); 803 (void) printf("\t\tprops_obj = %llu\n", 804 (u_longlong_t)ds->ds_props_obj); 805 (void) printf("\t\tbp = %s\n", blkbuf); 806 } 807 808 static void 809 dump_bplist(objset_t *mos, uint64_t object, char *name) 810 { 811 bplist_t bpl = { 0 }; 812 blkptr_t blk, *bp = &blk; 813 uint64_t itor = 0; 814 char bytes[6]; 815 char comp[6]; 816 char uncomp[6]; 817 818 if (dump_opt['d'] < 3) 819 return; 820 821 mutex_init(&bpl.bpl_lock, NULL, MUTEX_DEFAULT, NULL); 822 VERIFY(0 == bplist_open(&bpl, mos, object)); 823 if (bplist_empty(&bpl)) { 824 bplist_close(&bpl); 825 mutex_destroy(&bpl.bpl_lock); 826 return; 827 } 828 829 nicenum(bpl.bpl_phys->bpl_bytes, bytes); 830 if (bpl.bpl_dbuf->db_size == sizeof (bplist_phys_t)) { 831 nicenum(bpl.bpl_phys->bpl_comp, comp); 832 nicenum(bpl.bpl_phys->bpl_uncomp, uncomp); 833 (void) printf("\n %s: %llu entries, %s (%s/%s comp)\n", 834 name, (u_longlong_t)bpl.bpl_phys->bpl_entries, 835 bytes, comp, uncomp); 836 } else { 837 (void) printf("\n %s: %llu entries, %s\n", 838 name, (u_longlong_t)bpl.bpl_phys->bpl_entries, bytes); 839 } 840 841 if (dump_opt['d'] < 5) { 842 bplist_close(&bpl); 843 mutex_destroy(&bpl.bpl_lock); 844 return; 845 } 846 847 (void) printf("\n"); 848 849 while (bplist_iterate(&bpl, &itor, bp) == 0) { 850 char blkbuf[BP_SPRINTF_LEN]; 851 852 ASSERT(bp->blk_birth != 0); 853 sprintf_blkptr_compact(blkbuf, bp, dump_opt['d'] > 5 ? 1 : 0); 854 (void) printf("\tItem %3llu: %s\n", 855 (u_longlong_t)itor - 1, blkbuf); 856 } 857 858 bplist_close(&bpl); 859 mutex_destroy(&bpl.bpl_lock); 860 } 861 862 static avl_tree_t idx_tree; 863 static avl_tree_t domain_tree; 864 static boolean_t fuid_table_loaded; 865 866 static void 867 fuid_table_destroy() 868 { 869 if (fuid_table_loaded) { 870 zfs_fuid_table_destroy(&idx_tree, &domain_tree); 871 fuid_table_loaded = B_FALSE; 872 } 873 } 874 875 /* 876 * print uid or gid information. 877 * For normal POSIX id just the id is printed in decimal format. 878 * For CIFS files with FUID the fuid is printed in hex followed by 879 * the doman-rid string. 880 */ 881 static void 882 print_idstr(uint64_t id, const char *id_type) 883 { 884 if (FUID_INDEX(id)) { 885 char *domain; 886 887 domain = zfs_fuid_idx_domain(&idx_tree, FUID_INDEX(id)); 888 (void) printf("\t%s %llx [%s-%d]\n", id_type, 889 (u_longlong_t)id, domain, (int)FUID_RID(id)); 890 } else { 891 (void) printf("\t%s %llu\n", id_type, (u_longlong_t)id); 892 } 893 894 } 895 896 static void 897 dump_uidgid(objset_t *os, znode_phys_t *zp) 898 { 899 uint32_t uid_idx, gid_idx; 900 901 uid_idx = FUID_INDEX(zp->zp_uid); 902 gid_idx = FUID_INDEX(zp->zp_gid); 903 904 /* Load domain table, if not already loaded */ 905 if (!fuid_table_loaded && (uid_idx || gid_idx)) { 906 uint64_t fuid_obj; 907 908 /* first find the fuid object. It lives in the master node */ 909 VERIFY(zap_lookup(os, MASTER_NODE_OBJ, ZFS_FUID_TABLES, 910 8, 1, &fuid_obj) == 0); 911 (void) zfs_fuid_table_load(os, fuid_obj, 912 &idx_tree, &domain_tree); 913 fuid_table_loaded = B_TRUE; 914 } 915 916 print_idstr(zp->zp_uid, "uid"); 917 print_idstr(zp->zp_gid, "gid"); 918 } 919 920 /*ARGSUSED*/ 921 static void 922 dump_znode(objset_t *os, uint64_t object, void *data, size_t size) 923 { 924 znode_phys_t *zp = data; 925 time_t z_crtime, z_atime, z_mtime, z_ctime; 926 char path[MAXPATHLEN * 2]; /* allow for xattr and failure prefix */ 927 int error; 928 929 ASSERT(size >= sizeof (znode_phys_t)); 930 931 error = zfs_obj_to_path(os, object, path, sizeof (path)); 932 if (error != 0) { 933 (void) snprintf(path, sizeof (path), "\?\?\?<object#%llu>", 934 (u_longlong_t)object); 935 } 936 937 if (dump_opt['d'] < 3) { 938 (void) printf("\t%s\n", path); 939 return; 940 } 941 942 z_crtime = (time_t)zp->zp_crtime[0]; 943 z_atime = (time_t)zp->zp_atime[0]; 944 z_mtime = (time_t)zp->zp_mtime[0]; 945 z_ctime = (time_t)zp->zp_ctime[0]; 946 947 (void) printf("\tpath %s\n", path); 948 dump_uidgid(os, zp); 949 (void) printf("\tatime %s", ctime(&z_atime)); 950 (void) printf("\tmtime %s", ctime(&z_mtime)); 951 (void) printf("\tctime %s", ctime(&z_ctime)); 952 (void) printf("\tcrtime %s", ctime(&z_crtime)); 953 (void) printf("\tgen %llu\n", (u_longlong_t)zp->zp_gen); 954 (void) printf("\tmode %llo\n", (u_longlong_t)zp->zp_mode); 955 (void) printf("\tsize %llu\n", (u_longlong_t)zp->zp_size); 956 (void) printf("\tparent %llu\n", (u_longlong_t)zp->zp_parent); 957 (void) printf("\tlinks %llu\n", (u_longlong_t)zp->zp_links); 958 (void) printf("\txattr %llu\n", (u_longlong_t)zp->zp_xattr); 959 (void) printf("\trdev 0x%016llx\n", (u_longlong_t)zp->zp_rdev); 960 } 961 962 /*ARGSUSED*/ 963 static void 964 dump_acl(objset_t *os, uint64_t object, void *data, size_t size) 965 { 966 } 967 968 /*ARGSUSED*/ 969 static void 970 dump_dmu_objset(objset_t *os, uint64_t object, void *data, size_t size) 971 { 972 } 973 974 static object_viewer_t *object_viewer[DMU_OT_NUMTYPES] = { 975 dump_none, /* unallocated */ 976 dump_zap, /* object directory */ 977 dump_uint64, /* object array */ 978 dump_none, /* packed nvlist */ 979 dump_packed_nvlist, /* packed nvlist size */ 980 dump_none, /* bplist */ 981 dump_none, /* bplist header */ 982 dump_none, /* SPA space map header */ 983 dump_none, /* SPA space map */ 984 dump_none, /* ZIL intent log */ 985 dump_dnode, /* DMU dnode */ 986 dump_dmu_objset, /* DMU objset */ 987 dump_dsl_dir, /* DSL directory */ 988 dump_zap, /* DSL directory child map */ 989 dump_zap, /* DSL dataset snap map */ 990 dump_zap, /* DSL props */ 991 dump_dsl_dataset, /* DSL dataset */ 992 dump_znode, /* ZFS znode */ 993 dump_acl, /* ZFS V0 ACL */ 994 dump_uint8, /* ZFS plain file */ 995 dump_zpldir, /* ZFS directory */ 996 dump_zap, /* ZFS master node */ 997 dump_zap, /* ZFS delete queue */ 998 dump_uint8, /* zvol object */ 999 dump_zap, /* zvol prop */ 1000 dump_uint8, /* other uint8[] */ 1001 dump_uint64, /* other uint64[] */ 1002 dump_zap, /* other ZAP */ 1003 dump_zap, /* persistent error log */ 1004 dump_uint8, /* SPA history */ 1005 dump_uint64, /* SPA history offsets */ 1006 dump_zap, /* Pool properties */ 1007 dump_zap, /* DSL permissions */ 1008 dump_acl, /* ZFS ACL */ 1009 dump_uint8, /* ZFS SYSACL */ 1010 dump_none, /* FUID nvlist */ 1011 dump_packed_nvlist, /* FUID nvlist size */ 1012 dump_zap, /* DSL dataset next clones */ 1013 dump_zap, /* DSL scrub queue */ 1014 }; 1015 1016 static void 1017 dump_object(objset_t *os, uint64_t object, int verbosity, int *print_header) 1018 { 1019 dmu_buf_t *db = NULL; 1020 dmu_object_info_t doi; 1021 dnode_t *dn; 1022 void *bonus = NULL; 1023 size_t bsize = 0; 1024 char iblk[6], dblk[6], lsize[6], asize[6], bonus_size[6], segsize[6]; 1025 char aux[50]; 1026 int error; 1027 1028 if (*print_header) { 1029 (void) printf("\n Object lvl iblk dblk lsize" 1030 " asize type\n"); 1031 *print_header = 0; 1032 } 1033 1034 if (object == 0) { 1035 dn = os->os->os_meta_dnode; 1036 } else { 1037 error = dmu_bonus_hold(os, object, FTAG, &db); 1038 if (error) 1039 fatal("dmu_bonus_hold(%llu) failed, errno %u", 1040 object, error); 1041 bonus = db->db_data; 1042 bsize = db->db_size; 1043 dn = ((dmu_buf_impl_t *)db)->db_dnode; 1044 } 1045 dmu_object_info_from_dnode(dn, &doi); 1046 1047 nicenum(doi.doi_metadata_block_size, iblk); 1048 nicenum(doi.doi_data_block_size, dblk); 1049 nicenum(doi.doi_data_block_size * (doi.doi_max_block_offset + 1), 1050 lsize); 1051 nicenum(doi.doi_physical_blks << 9, asize); 1052 nicenum(doi.doi_bonus_size, bonus_size); 1053 1054 aux[0] = '\0'; 1055 1056 if (doi.doi_checksum != ZIO_CHECKSUM_INHERIT || verbosity >= 6) { 1057 (void) snprintf(aux + strlen(aux), sizeof (aux), " (K=%s)", 1058 zio_checksum_table[doi.doi_checksum].ci_name); 1059 } 1060 1061 if (doi.doi_compress != ZIO_COMPRESS_INHERIT || verbosity >= 6) { 1062 (void) snprintf(aux + strlen(aux), sizeof (aux), " (Z=%s)", 1063 zio_compress_table[doi.doi_compress].ci_name); 1064 } 1065 1066 (void) printf("%10lld %3u %5s %5s %5s %5s %s%s\n", 1067 (u_longlong_t)object, doi.doi_indirection, iblk, dblk, lsize, 1068 asize, dmu_ot[doi.doi_type].ot_name, aux); 1069 1070 if (doi.doi_bonus_type != DMU_OT_NONE && verbosity > 3) { 1071 (void) printf("%10s %3s %5s %5s %5s %5s %s\n", 1072 "", "", "", "", bonus_size, "bonus", 1073 dmu_ot[doi.doi_bonus_type].ot_name); 1074 } 1075 1076 if (verbosity >= 4) { 1077 object_viewer[doi.doi_bonus_type](os, object, bonus, bsize); 1078 object_viewer[doi.doi_type](os, object, NULL, 0); 1079 *print_header = 1; 1080 } 1081 1082 if (verbosity >= 5) 1083 dump_indirect(dn); 1084 1085 if (verbosity >= 5) { 1086 /* 1087 * Report the list of segments that comprise the object. 1088 */ 1089 uint64_t start = 0; 1090 uint64_t end; 1091 uint64_t blkfill = 1; 1092 int minlvl = 1; 1093 1094 if (dn->dn_type == DMU_OT_DNODE) { 1095 minlvl = 0; 1096 blkfill = DNODES_PER_BLOCK; 1097 } 1098 1099 for (;;) { 1100 error = dnode_next_offset(dn, 1101 0, &start, minlvl, blkfill, 0); 1102 if (error) 1103 break; 1104 end = start; 1105 error = dnode_next_offset(dn, 1106 DNODE_FIND_HOLE, &end, minlvl, blkfill, 0); 1107 nicenum(end - start, segsize); 1108 (void) printf("\t\tsegment [%016llx, %016llx)" 1109 " size %5s\n", (u_longlong_t)start, 1110 (u_longlong_t)end, segsize); 1111 if (error) 1112 break; 1113 start = end; 1114 } 1115 } 1116 1117 if (db != NULL) 1118 dmu_buf_rele(db, FTAG); 1119 } 1120 1121 static char *objset_types[DMU_OST_NUMTYPES] = { 1122 "NONE", "META", "ZPL", "ZVOL", "OTHER", "ANY" }; 1123 1124 static void 1125 dump_dir(objset_t *os) 1126 { 1127 dmu_objset_stats_t dds; 1128 uint64_t object, object_count; 1129 uint64_t refdbytes, usedobjs, scratch; 1130 char numbuf[8]; 1131 char blkbuf[BP_SPRINTF_LEN]; 1132 char osname[MAXNAMELEN]; 1133 char *type = "UNKNOWN"; 1134 int verbosity = dump_opt['d']; 1135 int print_header = 1; 1136 int i, error; 1137 1138 dmu_objset_fast_stat(os, &dds); 1139 1140 if (dds.dds_type < DMU_OST_NUMTYPES) 1141 type = objset_types[dds.dds_type]; 1142 1143 if (dds.dds_type == DMU_OST_META) { 1144 dds.dds_creation_txg = TXG_INITIAL; 1145 usedobjs = os->os->os_rootbp->blk_fill; 1146 refdbytes = os->os->os_spa->spa_dsl_pool-> 1147 dp_mos_dir->dd_phys->dd_used_bytes; 1148 } else { 1149 dmu_objset_space(os, &refdbytes, &scratch, &usedobjs, &scratch); 1150 } 1151 1152 ASSERT3U(usedobjs, ==, os->os->os_rootbp->blk_fill); 1153 1154 nicenum(refdbytes, numbuf); 1155 1156 if (verbosity >= 4) { 1157 (void) strcpy(blkbuf, ", rootbp "); 1158 sprintf_blkptr(blkbuf + strlen(blkbuf), 1159 BP_SPRINTF_LEN - strlen(blkbuf), os->os->os_rootbp); 1160 } else { 1161 blkbuf[0] = '\0'; 1162 } 1163 1164 dmu_objset_name(os, osname); 1165 1166 (void) printf("Dataset %s [%s], ID %llu, cr_txg %llu, " 1167 "%s, %llu objects%s\n", 1168 osname, type, (u_longlong_t)dmu_objset_id(os), 1169 (u_longlong_t)dds.dds_creation_txg, 1170 numbuf, (u_longlong_t)usedobjs, blkbuf); 1171 1172 dump_intent_log(dmu_objset_zil(os)); 1173 1174 if (dmu_objset_ds(os) != NULL) 1175 dump_bplist(dmu_objset_pool(os)->dp_meta_objset, 1176 dmu_objset_ds(os)->ds_phys->ds_deadlist_obj, "Deadlist"); 1177 1178 if (verbosity < 2) 1179 return; 1180 1181 if (os->os->os_rootbp->blk_birth == 0) 1182 return; 1183 1184 if (zopt_objects != 0) { 1185 for (i = 0; i < zopt_objects; i++) 1186 dump_object(os, zopt_object[i], verbosity, 1187 &print_header); 1188 (void) printf("\n"); 1189 return; 1190 } 1191 1192 dump_object(os, 0, verbosity, &print_header); 1193 object_count = 1; 1194 1195 object = 0; 1196 while ((error = dmu_object_next(os, &object, B_FALSE, 0)) == 0) { 1197 dump_object(os, object, verbosity, &print_header); 1198 object_count++; 1199 } 1200 1201 ASSERT3U(object_count, ==, usedobjs); 1202 1203 (void) printf("\n"); 1204 1205 if (error != ESRCH) 1206 fatal("dmu_object_next() = %d", error); 1207 } 1208 1209 static void 1210 dump_uberblock(uberblock_t *ub) 1211 { 1212 time_t timestamp = ub->ub_timestamp; 1213 1214 (void) printf("Uberblock\n\n"); 1215 (void) printf("\tmagic = %016llx\n", (u_longlong_t)ub->ub_magic); 1216 (void) printf("\tversion = %llu\n", (u_longlong_t)ub->ub_version); 1217 (void) printf("\ttxg = %llu\n", (u_longlong_t)ub->ub_txg); 1218 (void) printf("\tguid_sum = %llu\n", (u_longlong_t)ub->ub_guid_sum); 1219 (void) printf("\ttimestamp = %llu UTC = %s", 1220 (u_longlong_t)ub->ub_timestamp, asctime(localtime(×tamp))); 1221 if (dump_opt['u'] >= 3) { 1222 char blkbuf[BP_SPRINTF_LEN]; 1223 sprintf_blkptr(blkbuf, BP_SPRINTF_LEN, &ub->ub_rootbp); 1224 (void) printf("\trootbp = %s\n", blkbuf); 1225 } 1226 (void) printf("\n"); 1227 } 1228 1229 static void 1230 dump_config(const char *pool) 1231 { 1232 spa_t *spa = NULL; 1233 1234 mutex_enter(&spa_namespace_lock); 1235 while ((spa = spa_next(spa)) != NULL) { 1236 if (pool == NULL) 1237 (void) printf("%s\n", spa_name(spa)); 1238 if (pool == NULL || strcmp(pool, spa_name(spa)) == 0) 1239 dump_nvlist(spa->spa_config, 4); 1240 } 1241 mutex_exit(&spa_namespace_lock); 1242 } 1243 1244 static void 1245 dump_cachefile(const char *cachefile) 1246 { 1247 int fd; 1248 struct stat64 statbuf; 1249 char *buf; 1250 nvlist_t *config; 1251 1252 if ((fd = open64(cachefile, O_RDONLY)) < 0) { 1253 (void) printf("cannot open '%s': %s\n", cachefile, 1254 strerror(errno)); 1255 exit(1); 1256 } 1257 1258 if (fstat64(fd, &statbuf) != 0) { 1259 (void) printf("failed to stat '%s': %s\n", cachefile, 1260 strerror(errno)); 1261 exit(1); 1262 } 1263 1264 if ((buf = malloc(statbuf.st_size)) == NULL) { 1265 (void) fprintf(stderr, "failed to allocate %llu bytes\n", 1266 (u_longlong_t)statbuf.st_size); 1267 exit(1); 1268 } 1269 1270 if (read(fd, buf, statbuf.st_size) != statbuf.st_size) { 1271 (void) fprintf(stderr, "failed to read %llu bytes\n", 1272 (u_longlong_t)statbuf.st_size); 1273 exit(1); 1274 } 1275 1276 (void) close(fd); 1277 1278 if (nvlist_unpack(buf, statbuf.st_size, &config, 0) != 0) { 1279 (void) fprintf(stderr, "failed to unpack nvlist\n"); 1280 exit(1); 1281 } 1282 1283 free(buf); 1284 1285 dump_nvlist(config, 0); 1286 1287 nvlist_free(config); 1288 } 1289 1290 static void 1291 dump_label(const char *dev) 1292 { 1293 int fd; 1294 vdev_label_t label; 1295 char *buf = label.vl_vdev_phys.vp_nvlist; 1296 size_t buflen = sizeof (label.vl_vdev_phys.vp_nvlist); 1297 struct stat64 statbuf; 1298 uint64_t psize; 1299 int l; 1300 1301 if ((fd = open64(dev, O_RDONLY)) < 0) { 1302 (void) printf("cannot open '%s': %s\n", dev, strerror(errno)); 1303 exit(1); 1304 } 1305 1306 if (fstat64(fd, &statbuf) != 0) { 1307 (void) printf("failed to stat '%s': %s\n", dev, 1308 strerror(errno)); 1309 exit(1); 1310 } 1311 1312 psize = statbuf.st_size; 1313 psize = P2ALIGN(psize, (uint64_t)sizeof (vdev_label_t)); 1314 1315 for (l = 0; l < VDEV_LABELS; l++) { 1316 1317 nvlist_t *config = NULL; 1318 1319 (void) printf("--------------------------------------------\n"); 1320 (void) printf("LABEL %d\n", l); 1321 (void) printf("--------------------------------------------\n"); 1322 1323 if (pread64(fd, &label, sizeof (label), 1324 vdev_label_offset(psize, l, 0)) != sizeof (label)) { 1325 (void) printf("failed to read label %d\n", l); 1326 continue; 1327 } 1328 1329 if (nvlist_unpack(buf, buflen, &config, 0) != 0) { 1330 (void) printf("failed to unpack label %d\n", l); 1331 continue; 1332 } 1333 dump_nvlist(config, 4); 1334 nvlist_free(config); 1335 } 1336 } 1337 1338 /*ARGSUSED*/ 1339 static int 1340 dump_one_dir(char *dsname, void *arg) 1341 { 1342 int error; 1343 objset_t *os; 1344 1345 error = dmu_objset_open(dsname, DMU_OST_ANY, 1346 DS_MODE_USER | DS_MODE_READONLY, &os); 1347 if (error) { 1348 (void) printf("Could not open %s\n", dsname); 1349 return (0); 1350 } 1351 dump_dir(os); 1352 dmu_objset_close(os); 1353 fuid_table_destroy(); 1354 return (0); 1355 } 1356 1357 static void 1358 zdb_leak(space_map_t *sm, uint64_t start, uint64_t size) 1359 { 1360 vdev_t *vd = sm->sm_ppd; 1361 1362 (void) printf("leaked space: vdev %llu, offset 0x%llx, size %llu\n", 1363 (u_longlong_t)vd->vdev_id, (u_longlong_t)start, (u_longlong_t)size); 1364 } 1365 1366 /* ARGSUSED */ 1367 static void 1368 zdb_space_map_load(space_map_t *sm) 1369 { 1370 } 1371 1372 static void 1373 zdb_space_map_unload(space_map_t *sm) 1374 { 1375 space_map_vacate(sm, zdb_leak, sm); 1376 } 1377 1378 /* ARGSUSED */ 1379 static void 1380 zdb_space_map_claim(space_map_t *sm, uint64_t start, uint64_t size) 1381 { 1382 } 1383 1384 static space_map_ops_t zdb_space_map_ops = { 1385 zdb_space_map_load, 1386 zdb_space_map_unload, 1387 NULL, /* alloc */ 1388 zdb_space_map_claim, 1389 NULL /* free */ 1390 }; 1391 1392 static void 1393 zdb_leak_init(spa_t *spa) 1394 { 1395 vdev_t *rvd = spa->spa_root_vdev; 1396 1397 for (int c = 0; c < rvd->vdev_children; c++) { 1398 vdev_t *vd = rvd->vdev_child[c]; 1399 for (int m = 0; m < vd->vdev_ms_count; m++) { 1400 metaslab_t *msp = vd->vdev_ms[m]; 1401 mutex_enter(&msp->ms_lock); 1402 VERIFY(space_map_load(&msp->ms_map, &zdb_space_map_ops, 1403 SM_ALLOC, &msp->ms_smo, spa->spa_meta_objset) == 0); 1404 msp->ms_map.sm_ppd = vd; 1405 mutex_exit(&msp->ms_lock); 1406 } 1407 } 1408 } 1409 1410 static void 1411 zdb_leak_fini(spa_t *spa) 1412 { 1413 vdev_t *rvd = spa->spa_root_vdev; 1414 1415 for (int c = 0; c < rvd->vdev_children; c++) { 1416 vdev_t *vd = rvd->vdev_child[c]; 1417 for (int m = 0; m < vd->vdev_ms_count; m++) { 1418 metaslab_t *msp = vd->vdev_ms[m]; 1419 mutex_enter(&msp->ms_lock); 1420 space_map_unload(&msp->ms_map); 1421 mutex_exit(&msp->ms_lock); 1422 } 1423 } 1424 } 1425 1426 /* 1427 * Verify that the sum of the sizes of all blocks in the pool adds up 1428 * to the SPA's sa_alloc total. 1429 */ 1430 typedef struct zdb_blkstats { 1431 uint64_t zb_asize; 1432 uint64_t zb_lsize; 1433 uint64_t zb_psize; 1434 uint64_t zb_count; 1435 } zdb_blkstats_t; 1436 1437 #define DMU_OT_DEFERRED DMU_OT_NONE 1438 #define DMU_OT_TOTAL DMU_OT_NUMTYPES 1439 1440 #define ZB_TOTAL DN_MAX_LEVELS 1441 1442 typedef struct zdb_cb { 1443 zdb_blkstats_t zcb_type[ZB_TOTAL + 1][DMU_OT_TOTAL + 1]; 1444 uint64_t zcb_errors[256]; 1445 int zcb_readfails; 1446 int zcb_haderrors; 1447 } zdb_cb_t; 1448 1449 static void 1450 zdb_count_block(spa_t *spa, zdb_cb_t *zcb, blkptr_t *bp, dmu_object_type_t type) 1451 { 1452 for (int i = 0; i < 4; i++) { 1453 int l = (i < 2) ? BP_GET_LEVEL(bp) : ZB_TOTAL; 1454 int t = (i & 1) ? type : DMU_OT_TOTAL; 1455 zdb_blkstats_t *zb = &zcb->zcb_type[l][t]; 1456 1457 zb->zb_asize += BP_GET_ASIZE(bp); 1458 zb->zb_lsize += BP_GET_LSIZE(bp); 1459 zb->zb_psize += BP_GET_PSIZE(bp); 1460 zb->zb_count++; 1461 } 1462 1463 if (dump_opt['S']) { 1464 boolean_t print_sig; 1465 1466 print_sig = !zdb_sig_user_data || (BP_GET_LEVEL(bp) == 0 && 1467 BP_GET_TYPE(bp) == DMU_OT_PLAIN_FILE_CONTENTS); 1468 1469 if (BP_GET_CHECKSUM(bp) < zdb_sig_cksumalg) 1470 print_sig = B_FALSE; 1471 1472 if (print_sig) { 1473 (void) printf("%llu\t%lld\t%lld\t%s\t%s\t%s\t" 1474 "%llx:%llx:%llx:%llx\n", 1475 (u_longlong_t)BP_GET_LEVEL(bp), 1476 (longlong_t)BP_GET_PSIZE(bp), 1477 (longlong_t)BP_GET_NDVAS(bp), 1478 dmu_ot[BP_GET_TYPE(bp)].ot_name, 1479 zio_checksum_table[BP_GET_CHECKSUM(bp)].ci_name, 1480 zio_compress_table[BP_GET_COMPRESS(bp)].ci_name, 1481 (u_longlong_t)bp->blk_cksum.zc_word[0], 1482 (u_longlong_t)bp->blk_cksum.zc_word[1], 1483 (u_longlong_t)bp->blk_cksum.zc_word[2], 1484 (u_longlong_t)bp->blk_cksum.zc_word[3]); 1485 } 1486 } 1487 1488 if (!dump_opt['L']) 1489 VERIFY(zio_wait(zio_claim(NULL, spa, spa_first_txg(spa), bp, 1490 NULL, NULL, ZIO_FLAG_MUSTSUCCEED)) == 0); 1491 } 1492 1493 static int 1494 zdb_blkptr_cb(spa_t *spa, blkptr_t *bp, const zbookmark_t *zb, 1495 const dnode_phys_t *dnp, void *arg) 1496 { 1497 zdb_cb_t *zcb = arg; 1498 char blkbuf[BP_SPRINTF_LEN]; 1499 1500 if (bp == NULL) 1501 return (0); 1502 1503 zdb_count_block(spa, zcb, bp, BP_GET_TYPE(bp)); 1504 1505 if (dump_opt['c'] || dump_opt['S']) { 1506 int ioerr, size; 1507 void *data; 1508 1509 size = BP_GET_LSIZE(bp); 1510 data = malloc(size); 1511 ioerr = zio_wait(zio_read(NULL, spa, bp, data, size, 1512 NULL, NULL, ZIO_PRIORITY_ASYNC_READ, 1513 ZIO_FLAG_CANFAIL | ZIO_FLAG_SCRUB, zb)); 1514 free(data); 1515 1516 /* We expect io errors on intent log */ 1517 if (ioerr && BP_GET_TYPE(bp) != DMU_OT_INTENT_LOG) { 1518 zcb->zcb_haderrors = 1; 1519 zcb->zcb_errors[ioerr]++; 1520 1521 if (dump_opt['b'] >= 2) 1522 sprintf_blkptr(blkbuf, BP_SPRINTF_LEN, bp); 1523 else 1524 blkbuf[0] = '\0'; 1525 1526 if (!dump_opt['S']) { 1527 (void) printf("zdb_blkptr_cb: " 1528 "Got error %d reading " 1529 "<%llu, %llu, %lld, %llx> %s -- skipping\n", 1530 ioerr, 1531 (u_longlong_t)zb->zb_objset, 1532 (u_longlong_t)zb->zb_object, 1533 (u_longlong_t)zb->zb_level, 1534 (u_longlong_t)zb->zb_blkid, 1535 blkbuf); 1536 } 1537 } 1538 } 1539 1540 zcb->zcb_readfails = 0; 1541 1542 if (dump_opt['b'] >= 4) { 1543 sprintf_blkptr(blkbuf, BP_SPRINTF_LEN, bp); 1544 (void) printf("objset %llu object %llu offset 0x%llx %s\n", 1545 (u_longlong_t)zb->zb_objset, 1546 (u_longlong_t)zb->zb_object, 1547 (u_longlong_t)blkid2offset(dnp, zb->zb_level, zb->zb_blkid), 1548 blkbuf); 1549 } 1550 1551 return (0); 1552 } 1553 1554 static int 1555 dump_block_stats(spa_t *spa) 1556 { 1557 zdb_cb_t zcb = { 0 }; 1558 zdb_blkstats_t *zb, *tzb; 1559 uint64_t alloc, space, logalloc; 1560 vdev_t *rvd = spa->spa_root_vdev; 1561 int leaks = 0; 1562 int c, e; 1563 1564 if (!dump_opt['S']) { 1565 (void) printf("\nTraversing all blocks %s%s%s%s...\n", 1566 (dump_opt['c'] || !dump_opt['L']) ? "to verify " : "", 1567 dump_opt['c'] ? "checksums " : "", 1568 (dump_opt['c'] && !dump_opt['L']) ? "and verify " : "", 1569 !dump_opt['L'] ? "nothing leaked " : ""); 1570 } 1571 1572 /* 1573 * Load all space maps as SM_ALLOC maps, then traverse the pool 1574 * claiming each block we discover. If the pool is perfectly 1575 * consistent, the space maps will be empty when we're done. 1576 * Anything left over is a leak; any block we can't claim (because 1577 * it's not part of any space map) is a double allocation, 1578 * reference to a freed block, or an unclaimed log block. 1579 */ 1580 if (!dump_opt['L']) 1581 zdb_leak_init(spa); 1582 1583 /* 1584 * If there's a deferred-free bplist, process that first. 1585 */ 1586 if (spa->spa_sync_bplist_obj != 0) { 1587 bplist_t *bpl = &spa->spa_sync_bplist; 1588 blkptr_t blk; 1589 uint64_t itor = 0; 1590 1591 VERIFY(0 == bplist_open(bpl, spa->spa_meta_objset, 1592 spa->spa_sync_bplist_obj)); 1593 1594 while (bplist_iterate(bpl, &itor, &blk) == 0) { 1595 if (dump_opt['b'] >= 4) { 1596 char blkbuf[BP_SPRINTF_LEN]; 1597 sprintf_blkptr(blkbuf, BP_SPRINTF_LEN, &blk); 1598 (void) printf("[%s] %s\n", 1599 "deferred free", blkbuf); 1600 } 1601 zdb_count_block(spa, &zcb, &blk, DMU_OT_DEFERRED); 1602 } 1603 1604 bplist_close(bpl); 1605 } 1606 1607 zcb.zcb_haderrors |= traverse_pool(spa, zdb_blkptr_cb, &zcb); 1608 1609 if (zcb.zcb_haderrors && !dump_opt['S']) { 1610 (void) printf("\nError counts:\n\n"); 1611 (void) printf("\t%5s %s\n", "errno", "count"); 1612 for (e = 0; e < 256; e++) { 1613 if (zcb.zcb_errors[e] != 0) { 1614 (void) printf("\t%5d %llu\n", 1615 e, (u_longlong_t)zcb.zcb_errors[e]); 1616 } 1617 } 1618 } 1619 1620 /* 1621 * Report any leaked segments. 1622 */ 1623 if (!dump_opt['L']) 1624 zdb_leak_fini(spa); 1625 1626 /* 1627 * If we're interested in printing out the blkptr signatures, 1628 * return now as we don't print out anything else (including 1629 * errors and leaks). 1630 */ 1631 if (dump_opt['S']) 1632 return (zcb.zcb_haderrors ? 3 : 0); 1633 1634 alloc = spa_get_alloc(spa); 1635 space = spa_get_space(spa); 1636 1637 /* 1638 * Log blocks allocated from a separate log device don't count 1639 * as part of the normal pool space; factor them in here. 1640 */ 1641 logalloc = 0; 1642 1643 for (c = 0; c < rvd->vdev_children; c++) 1644 if (rvd->vdev_child[c]->vdev_islog) 1645 logalloc += rvd->vdev_child[c]->vdev_stat.vs_alloc; 1646 1647 tzb = &zcb.zcb_type[ZB_TOTAL][DMU_OT_TOTAL]; 1648 1649 if (tzb->zb_asize == alloc + logalloc) { 1650 if (!dump_opt['L']) 1651 (void) printf("\n\tNo leaks (block sum matches space" 1652 " maps exactly)\n"); 1653 } else { 1654 (void) printf("block traversal size %llu != alloc %llu " 1655 "(%s %lld)\n", 1656 (u_longlong_t)tzb->zb_asize, 1657 (u_longlong_t)alloc + logalloc, 1658 (dump_opt['L']) ? "unreachable" : "leaked", 1659 (longlong_t)(alloc + logalloc - tzb->zb_asize)); 1660 leaks = 1; 1661 } 1662 1663 if (tzb->zb_count == 0) 1664 return (2); 1665 1666 (void) printf("\n"); 1667 (void) printf("\tbp count: %10llu\n", 1668 (u_longlong_t)tzb->zb_count); 1669 (void) printf("\tbp logical: %10llu\t avg: %6llu\n", 1670 (u_longlong_t)tzb->zb_lsize, 1671 (u_longlong_t)(tzb->zb_lsize / tzb->zb_count)); 1672 (void) printf("\tbp physical: %10llu\t avg:" 1673 " %6llu\tcompression: %6.2f\n", 1674 (u_longlong_t)tzb->zb_psize, 1675 (u_longlong_t)(tzb->zb_psize / tzb->zb_count), 1676 (double)tzb->zb_lsize / tzb->zb_psize); 1677 (void) printf("\tbp allocated: %10llu\t avg:" 1678 " %6llu\tcompression: %6.2f\n", 1679 (u_longlong_t)tzb->zb_asize, 1680 (u_longlong_t)(tzb->zb_asize / tzb->zb_count), 1681 (double)tzb->zb_lsize / tzb->zb_asize); 1682 (void) printf("\tSPA allocated: %10llu\tused: %5.2f%%\n", 1683 (u_longlong_t)alloc, 100.0 * alloc / space); 1684 1685 if (dump_opt['b'] >= 2) { 1686 int l, t, level; 1687 (void) printf("\nBlocks\tLSIZE\tPSIZE\tASIZE" 1688 "\t avg\t comp\t%%Total\tType\n"); 1689 1690 for (t = 0; t <= DMU_OT_NUMTYPES; t++) { 1691 char csize[6], lsize[6], psize[6], asize[6], avg[6]; 1692 char *typename; 1693 1694 typename = t == DMU_OT_DEFERRED ? "deferred free" : 1695 t == DMU_OT_TOTAL ? "Total" : dmu_ot[t].ot_name; 1696 1697 if (zcb.zcb_type[ZB_TOTAL][t].zb_asize == 0) { 1698 (void) printf("%6s\t%5s\t%5s\t%5s" 1699 "\t%5s\t%5s\t%6s\t%s\n", 1700 "-", 1701 "-", 1702 "-", 1703 "-", 1704 "-", 1705 "-", 1706 "-", 1707 typename); 1708 continue; 1709 } 1710 1711 for (l = ZB_TOTAL - 1; l >= -1; l--) { 1712 level = (l == -1 ? ZB_TOTAL : l); 1713 zb = &zcb.zcb_type[level][t]; 1714 1715 if (zb->zb_asize == 0) 1716 continue; 1717 1718 if (dump_opt['b'] < 3 && level != ZB_TOTAL) 1719 continue; 1720 1721 if (level == 0 && zb->zb_asize == 1722 zcb.zcb_type[ZB_TOTAL][t].zb_asize) 1723 continue; 1724 1725 nicenum(zb->zb_count, csize); 1726 nicenum(zb->zb_lsize, lsize); 1727 nicenum(zb->zb_psize, psize); 1728 nicenum(zb->zb_asize, asize); 1729 nicenum(zb->zb_asize / zb->zb_count, avg); 1730 1731 (void) printf("%6s\t%5s\t%5s\t%5s\t%5s" 1732 "\t%5.2f\t%6.2f\t", 1733 csize, lsize, psize, asize, avg, 1734 (double)zb->zb_lsize / zb->zb_psize, 1735 100.0 * zb->zb_asize / tzb->zb_asize); 1736 1737 if (level == ZB_TOTAL) 1738 (void) printf("%s\n", typename); 1739 else 1740 (void) printf(" L%d %s\n", 1741 level, typename); 1742 } 1743 } 1744 } 1745 1746 (void) printf("\n"); 1747 1748 if (leaks) 1749 return (2); 1750 1751 if (zcb.zcb_haderrors) 1752 return (3); 1753 1754 return (0); 1755 } 1756 1757 static void 1758 dump_zpool(spa_t *spa) 1759 { 1760 dsl_pool_t *dp = spa_get_dsl(spa); 1761 int rc = 0; 1762 1763 if (dump_opt['u']) 1764 dump_uberblock(&spa->spa_uberblock); 1765 1766 if (dump_opt['d'] || dump_opt['i']) { 1767 dump_dir(dp->dp_meta_objset); 1768 if (dump_opt['d'] >= 3) { 1769 dump_bplist(dp->dp_meta_objset, 1770 spa->spa_sync_bplist_obj, "Deferred frees"); 1771 dump_dtl(spa->spa_root_vdev, 0); 1772 dump_metaslabs(spa); 1773 } 1774 (void) dmu_objset_find(spa_name(spa), dump_one_dir, NULL, 1775 DS_FIND_SNAPSHOTS | DS_FIND_CHILDREN); 1776 } 1777 1778 if (dump_opt['b'] || dump_opt['c'] || dump_opt['S']) 1779 rc = dump_block_stats(spa); 1780 1781 if (dump_opt['s']) 1782 show_pool_stats(spa); 1783 1784 if (rc != 0) 1785 exit(rc); 1786 } 1787 1788 #define ZDB_FLAG_CHECKSUM 0x0001 1789 #define ZDB_FLAG_DECOMPRESS 0x0002 1790 #define ZDB_FLAG_BSWAP 0x0004 1791 #define ZDB_FLAG_GBH 0x0008 1792 #define ZDB_FLAG_INDIRECT 0x0010 1793 #define ZDB_FLAG_PHYS 0x0020 1794 #define ZDB_FLAG_RAW 0x0040 1795 #define ZDB_FLAG_PRINT_BLKPTR 0x0080 1796 1797 int flagbits[256]; 1798 1799 static void 1800 zdb_print_blkptr(blkptr_t *bp, int flags) 1801 { 1802 dva_t *dva = bp->blk_dva; 1803 int d; 1804 1805 if (flags & ZDB_FLAG_BSWAP) 1806 byteswap_uint64_array((void *)bp, sizeof (blkptr_t)); 1807 /* 1808 * Super-ick warning: This code is also duplicated in 1809 * cmd/mdb/common/modules/zfs/zfs.c . Yeah, I hate code 1810 * replication, too. 1811 */ 1812 for (d = 0; d < BP_GET_NDVAS(bp); d++) { 1813 (void) printf("\tDVA[%d]: vdev_id %lld / %llx\n", d, 1814 (longlong_t)DVA_GET_VDEV(&dva[d]), 1815 (longlong_t)DVA_GET_OFFSET(&dva[d])); 1816 (void) printf("\tDVA[%d]: GANG: %-5s GRID: %04llx\t" 1817 "ASIZE: %llx\n", d, 1818 DVA_GET_GANG(&dva[d]) ? "TRUE" : "FALSE", 1819 (longlong_t)DVA_GET_GRID(&dva[d]), 1820 (longlong_t)DVA_GET_ASIZE(&dva[d])); 1821 (void) printf("\tDVA[%d]: :%llu:%llx:%llx:%s%s%s%s\n", d, 1822 (u_longlong_t)DVA_GET_VDEV(&dva[d]), 1823 (longlong_t)DVA_GET_OFFSET(&dva[d]), 1824 (longlong_t)BP_GET_PSIZE(bp), 1825 BP_SHOULD_BYTESWAP(bp) ? "e" : "", 1826 !DVA_GET_GANG(&dva[d]) && BP_GET_LEVEL(bp) != 0 ? 1827 "d" : "", 1828 DVA_GET_GANG(&dva[d]) ? "g" : "", 1829 BP_GET_COMPRESS(bp) != 0 ? "d" : ""); 1830 } 1831 (void) printf("\tLSIZE: %-16llx\t\tPSIZE: %llx\n", 1832 (longlong_t)BP_GET_LSIZE(bp), (longlong_t)BP_GET_PSIZE(bp)); 1833 (void) printf("\tENDIAN: %6s\t\t\t\t\tTYPE: %s\n", 1834 BP_GET_BYTEORDER(bp) ? "LITTLE" : "BIG", 1835 dmu_ot[BP_GET_TYPE(bp)].ot_name); 1836 (void) printf("\tBIRTH: %-16llx LEVEL: %-2llu\tFILL: %llx\n", 1837 (u_longlong_t)bp->blk_birth, (u_longlong_t)BP_GET_LEVEL(bp), 1838 (u_longlong_t)bp->blk_fill); 1839 (void) printf("\tCKFUNC: %-16s\t\tCOMP: %s\n", 1840 zio_checksum_table[BP_GET_CHECKSUM(bp)].ci_name, 1841 zio_compress_table[BP_GET_COMPRESS(bp)].ci_name); 1842 (void) printf("\tCKSUM: %llx:%llx:%llx:%llx\n", 1843 (u_longlong_t)bp->blk_cksum.zc_word[0], 1844 (u_longlong_t)bp->blk_cksum.zc_word[1], 1845 (u_longlong_t)bp->blk_cksum.zc_word[2], 1846 (u_longlong_t)bp->blk_cksum.zc_word[3]); 1847 } 1848 1849 static void 1850 zdb_dump_indirect(blkptr_t *bp, int nbps, int flags) 1851 { 1852 int i; 1853 1854 for (i = 0; i < nbps; i++) 1855 zdb_print_blkptr(&bp[i], flags); 1856 } 1857 1858 static void 1859 zdb_dump_gbh(void *buf, int flags) 1860 { 1861 zdb_dump_indirect((blkptr_t *)buf, SPA_GBH_NBLKPTRS, flags); 1862 } 1863 1864 static void 1865 zdb_dump_block_raw(void *buf, uint64_t size, int flags) 1866 { 1867 if (flags & ZDB_FLAG_BSWAP) 1868 byteswap_uint64_array(buf, size); 1869 (void) write(2, buf, size); 1870 } 1871 1872 static void 1873 zdb_dump_block(char *label, void *buf, uint64_t size, int flags) 1874 { 1875 uint64_t *d = (uint64_t *)buf; 1876 int nwords = size / sizeof (uint64_t); 1877 int do_bswap = !!(flags & ZDB_FLAG_BSWAP); 1878 int i, j; 1879 char *hdr, *c; 1880 1881 1882 if (do_bswap) 1883 hdr = " 7 6 5 4 3 2 1 0 f e d c b a 9 8"; 1884 else 1885 hdr = " 0 1 2 3 4 5 6 7 8 9 a b c d e f"; 1886 1887 (void) printf("\n%s\n%6s %s 0123456789abcdef\n", label, "", hdr); 1888 1889 for (i = 0; i < nwords; i += 2) { 1890 (void) printf("%06llx: %016llx %016llx ", 1891 (u_longlong_t)(i * sizeof (uint64_t)), 1892 (u_longlong_t)(do_bswap ? BSWAP_64(d[i]) : d[i]), 1893 (u_longlong_t)(do_bswap ? BSWAP_64(d[i + 1]) : d[i + 1])); 1894 1895 c = (char *)&d[i]; 1896 for (j = 0; j < 2 * sizeof (uint64_t); j++) 1897 (void) printf("%c", isprint(c[j]) ? c[j] : '.'); 1898 (void) printf("\n"); 1899 } 1900 } 1901 1902 /* 1903 * There are two acceptable formats: 1904 * leaf_name - For example: c1t0d0 or /tmp/ztest.0a 1905 * child[.child]* - For example: 0.1.1 1906 * 1907 * The second form can be used to specify arbitrary vdevs anywhere 1908 * in the heirarchy. For example, in a pool with a mirror of 1909 * RAID-Zs, you can specify either RAID-Z vdev with 0.0 or 0.1 . 1910 */ 1911 static vdev_t * 1912 zdb_vdev_lookup(vdev_t *vdev, char *path) 1913 { 1914 char *s, *p, *q; 1915 int i; 1916 1917 if (vdev == NULL) 1918 return (NULL); 1919 1920 /* First, assume the x.x.x.x format */ 1921 i = (int)strtoul(path, &s, 10); 1922 if (s == path || (s && *s != '.' && *s != '\0')) 1923 goto name; 1924 if (i < 0 || i >= vdev->vdev_children) 1925 return (NULL); 1926 1927 vdev = vdev->vdev_child[i]; 1928 if (*s == '\0') 1929 return (vdev); 1930 return (zdb_vdev_lookup(vdev, s+1)); 1931 1932 name: 1933 for (i = 0; i < vdev->vdev_children; i++) { 1934 vdev_t *vc = vdev->vdev_child[i]; 1935 1936 if (vc->vdev_path == NULL) { 1937 vc = zdb_vdev_lookup(vc, path); 1938 if (vc == NULL) 1939 continue; 1940 else 1941 return (vc); 1942 } 1943 1944 p = strrchr(vc->vdev_path, '/'); 1945 p = p ? p + 1 : vc->vdev_path; 1946 q = &vc->vdev_path[strlen(vc->vdev_path) - 2]; 1947 1948 if (strcmp(vc->vdev_path, path) == 0) 1949 return (vc); 1950 if (strcmp(p, path) == 0) 1951 return (vc); 1952 if (strcmp(q, "s0") == 0 && strncmp(p, path, q - p) == 0) 1953 return (vc); 1954 } 1955 1956 return (NULL); 1957 } 1958 1959 /* 1960 * Read a block from a pool and print it out. The syntax of the 1961 * block descriptor is: 1962 * 1963 * pool:vdev_specifier:offset:size[:flags] 1964 * 1965 * pool - The name of the pool you wish to read from 1966 * vdev_specifier - Which vdev (see comment for zdb_vdev_lookup) 1967 * offset - offset, in hex, in bytes 1968 * size - Amount of data to read, in hex, in bytes 1969 * flags - A string of characters specifying options 1970 * b: Decode a blkptr at given offset within block 1971 * *c: Calculate and display checksums 1972 * *d: Decompress data before dumping 1973 * e: Byteswap data before dumping 1974 * *g: Display data as a gang block header 1975 * *i: Display as an indirect block 1976 * p: Do I/O to physical offset 1977 * r: Dump raw data to stdout 1978 * 1979 * * = not yet implemented 1980 */ 1981 static void 1982 zdb_read_block(char *thing, spa_t **spap) 1983 { 1984 spa_t *spa = *spap; 1985 int flags = 0; 1986 uint64_t offset = 0, size = 0, blkptr_offset = 0; 1987 zio_t *zio; 1988 vdev_t *vd; 1989 void *buf; 1990 char *s, *p, *dup, *pool, *vdev, *flagstr; 1991 int i, error, zio_flags; 1992 1993 dup = strdup(thing); 1994 s = strtok(dup, ":"); 1995 pool = s ? s : ""; 1996 s = strtok(NULL, ":"); 1997 vdev = s ? s : ""; 1998 s = strtok(NULL, ":"); 1999 offset = strtoull(s ? s : "", NULL, 16); 2000 s = strtok(NULL, ":"); 2001 size = strtoull(s ? s : "", NULL, 16); 2002 s = strtok(NULL, ":"); 2003 flagstr = s ? s : ""; 2004 2005 s = NULL; 2006 if (size == 0) 2007 s = "size must not be zero"; 2008 if (!IS_P2ALIGNED(size, DEV_BSIZE)) 2009 s = "size must be a multiple of sector size"; 2010 if (!IS_P2ALIGNED(offset, DEV_BSIZE)) 2011 s = "offset must be a multiple of sector size"; 2012 if (s) { 2013 (void) printf("Invalid block specifier: %s - %s\n", thing, s); 2014 free(dup); 2015 return; 2016 } 2017 2018 for (s = strtok(flagstr, ":"); s; s = strtok(NULL, ":")) { 2019 for (i = 0; flagstr[i]; i++) { 2020 int bit = flagbits[(uchar_t)flagstr[i]]; 2021 2022 if (bit == 0) { 2023 (void) printf("***Invalid flag: %c\n", 2024 flagstr[i]); 2025 continue; 2026 } 2027 flags |= bit; 2028 2029 /* If it's not something with an argument, keep going */ 2030 if ((bit & (ZDB_FLAG_CHECKSUM | ZDB_FLAG_DECOMPRESS | 2031 ZDB_FLAG_PRINT_BLKPTR)) == 0) 2032 continue; 2033 2034 p = &flagstr[i + 1]; 2035 if (bit == ZDB_FLAG_PRINT_BLKPTR) 2036 blkptr_offset = strtoull(p, &p, 16); 2037 if (*p != ':' && *p != '\0') { 2038 (void) printf("***Invalid flag arg: '%s'\n", s); 2039 free(dup); 2040 return; 2041 } 2042 } 2043 } 2044 2045 if (spa == NULL || strcmp(spa_name(spa), pool) != 0) { 2046 if (spa) 2047 spa_close(spa, (void *)zdb_read_block); 2048 error = spa_open(pool, spap, (void *)zdb_read_block); 2049 if (error) 2050 fatal("Failed to open pool '%s': %s", 2051 pool, strerror(error)); 2052 spa = *spap; 2053 } 2054 2055 vd = zdb_vdev_lookup(spa->spa_root_vdev, vdev); 2056 if (vd == NULL) { 2057 (void) printf("***Invalid vdev: %s\n", vdev); 2058 free(dup); 2059 return; 2060 } else { 2061 if (vd->vdev_path) 2062 (void) printf("Found vdev: %s\n", vd->vdev_path); 2063 else 2064 (void) printf("Found vdev type: %s\n", 2065 vd->vdev_ops->vdev_op_type); 2066 } 2067 2068 buf = umem_alloc(size, UMEM_NOFAIL); 2069 2070 zio_flags = ZIO_FLAG_DONT_CACHE | ZIO_FLAG_DONT_QUEUE | 2071 ZIO_FLAG_DONT_PROPAGATE | ZIO_FLAG_DONT_RETRY; 2072 2073 spa_config_enter(spa, SCL_STATE, FTAG, RW_READER); 2074 zio = zio_root(spa, NULL, NULL, 0); 2075 /* XXX todo - cons up a BP so RAID-Z will be happy */ 2076 zio_nowait(zio_vdev_child_io(zio, NULL, vd, offset, buf, size, 2077 ZIO_TYPE_READ, ZIO_PRIORITY_SYNC_READ, zio_flags, NULL, NULL)); 2078 error = zio_wait(zio); 2079 spa_config_exit(spa, SCL_STATE, FTAG); 2080 2081 if (error) { 2082 (void) printf("Read of %s failed, error: %d\n", thing, error); 2083 goto out; 2084 } 2085 2086 if (flags & ZDB_FLAG_PRINT_BLKPTR) 2087 zdb_print_blkptr((blkptr_t *)(void *) 2088 ((uintptr_t)buf + (uintptr_t)blkptr_offset), flags); 2089 else if (flags & ZDB_FLAG_RAW) 2090 zdb_dump_block_raw(buf, size, flags); 2091 else if (flags & ZDB_FLAG_INDIRECT) 2092 zdb_dump_indirect((blkptr_t *)buf, size / sizeof (blkptr_t), 2093 flags); 2094 else if (flags & ZDB_FLAG_GBH) 2095 zdb_dump_gbh(buf, flags); 2096 else 2097 zdb_dump_block(thing, buf, size, flags); 2098 2099 out: 2100 umem_free(buf, size); 2101 free(dup); 2102 } 2103 2104 static boolean_t 2105 nvlist_string_match(nvlist_t *config, char *name, char *tgt) 2106 { 2107 char *s; 2108 2109 if (nvlist_lookup_string(config, name, &s) != 0) 2110 return (B_FALSE); 2111 2112 return (strcmp(s, tgt) == 0); 2113 } 2114 2115 static boolean_t 2116 nvlist_uint64_match(nvlist_t *config, char *name, uint64_t tgt) 2117 { 2118 uint64_t val; 2119 2120 if (nvlist_lookup_uint64(config, name, &val) != 0) 2121 return (B_FALSE); 2122 2123 return (val == tgt); 2124 } 2125 2126 static boolean_t 2127 vdev_child_guid_match(nvlist_t *vdev, uint64_t guid) 2128 { 2129 nvlist_t **child; 2130 uint_t c, children; 2131 2132 verify(nvlist_lookup_nvlist_array(vdev, ZPOOL_CONFIG_CHILDREN, 2133 &child, &children) == 0); 2134 for (c = 0; c < children; ++c) 2135 if (nvlist_uint64_match(child[c], ZPOOL_CONFIG_GUID, guid)) 2136 return (B_TRUE); 2137 return (B_FALSE); 2138 } 2139 2140 static boolean_t 2141 vdev_child_string_match(nvlist_t *vdev, char *tgt) 2142 { 2143 nvlist_t **child; 2144 uint_t c, children; 2145 2146 verify(nvlist_lookup_nvlist_array(vdev, ZPOOL_CONFIG_CHILDREN, 2147 &child, &children) == 0); 2148 for (c = 0; c < children; ++c) { 2149 if (nvlist_string_match(child[c], ZPOOL_CONFIG_PATH, tgt) || 2150 nvlist_string_match(child[c], ZPOOL_CONFIG_DEVID, tgt)) 2151 return (B_TRUE); 2152 } 2153 return (B_FALSE); 2154 } 2155 2156 static boolean_t 2157 vdev_guid_match(nvlist_t *config, uint64_t guid) 2158 { 2159 nvlist_t *nvroot; 2160 2161 verify(nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE, 2162 &nvroot) == 0); 2163 2164 return (nvlist_uint64_match(nvroot, ZPOOL_CONFIG_GUID, guid) || 2165 vdev_child_guid_match(nvroot, guid)); 2166 } 2167 2168 static boolean_t 2169 vdev_string_match(nvlist_t *config, char *tgt) 2170 { 2171 nvlist_t *nvroot; 2172 2173 verify(nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE, 2174 &nvroot) == 0); 2175 2176 return (vdev_child_string_match(nvroot, tgt)); 2177 } 2178 2179 static boolean_t 2180 pool_match(nvlist_t *config, char *tgt) 2181 { 2182 uint64_t guid = strtoull(tgt, NULL, 0); 2183 2184 if (guid != 0) { 2185 return ( 2186 nvlist_uint64_match(config, ZPOOL_CONFIG_POOL_GUID, guid) || 2187 vdev_guid_match(config, guid)); 2188 } else { 2189 return ( 2190 nvlist_string_match(config, ZPOOL_CONFIG_POOL_NAME, tgt) || 2191 vdev_string_match(config, tgt)); 2192 } 2193 } 2194 2195 static int 2196 find_exported_zpool(char *pool_id, nvlist_t **configp, char *vdev_dir) 2197 { 2198 nvlist_t *pools; 2199 int error = ENOENT; 2200 nvlist_t *match = NULL; 2201 2202 if (vdev_dir != NULL) 2203 pools = zpool_find_import_activeok(g_zfs, 1, &vdev_dir); 2204 else 2205 pools = zpool_find_import_activeok(g_zfs, 0, NULL); 2206 2207 if (pools != NULL) { 2208 nvpair_t *elem = NULL; 2209 2210 while ((elem = nvlist_next_nvpair(pools, elem)) != NULL) { 2211 verify(nvpair_value_nvlist(elem, configp) == 0); 2212 if (pool_match(*configp, pool_id)) { 2213 if (match != NULL) { 2214 (void) fatal( 2215 "More than one matching pool - " 2216 "specify guid/devid/device path."); 2217 } else { 2218 match = *configp; 2219 error = 0; 2220 } 2221 } 2222 } 2223 } 2224 2225 *configp = error ? NULL : match; 2226 2227 return (error); 2228 } 2229 2230 int 2231 main(int argc, char **argv) 2232 { 2233 int i, c; 2234 struct rlimit rl = { 1024, 1024 }; 2235 spa_t *spa; 2236 objset_t *os = NULL; 2237 char *endstr; 2238 int dump_all = 1; 2239 int verbose = 0; 2240 int error; 2241 int exported = 0; 2242 char *vdev_dir = NULL; 2243 2244 (void) setrlimit(RLIMIT_NOFILE, &rl); 2245 (void) enable_extended_FILE_stdio(-1, -1); 2246 2247 dprintf_setup(&argc, argv); 2248 2249 while ((c = getopt(argc, argv, "udibcsvCLS:U:lRep:t:")) != -1) { 2250 switch (c) { 2251 case 'u': 2252 case 'd': 2253 case 'i': 2254 case 'b': 2255 case 'c': 2256 case 's': 2257 case 'C': 2258 case 'l': 2259 case 'R': 2260 dump_opt[c]++; 2261 dump_all = 0; 2262 break; 2263 case 'L': 2264 dump_opt[c]++; 2265 break; 2266 case 'v': 2267 verbose++; 2268 break; 2269 case 'U': 2270 spa_config_path = optarg; 2271 break; 2272 case 'e': 2273 exported = 1; 2274 break; 2275 case 'p': 2276 vdev_dir = optarg; 2277 break; 2278 case 'S': 2279 dump_opt[c]++; 2280 dump_all = 0; 2281 zdb_sig_user_data = (strncmp(optarg, "user:", 5) == 0); 2282 if (!zdb_sig_user_data && strncmp(optarg, "all:", 4)) 2283 usage(); 2284 endstr = strchr(optarg, ':') + 1; 2285 if (strcmp(endstr, "fletcher2") == 0) 2286 zdb_sig_cksumalg = ZIO_CHECKSUM_FLETCHER_2; 2287 else if (strcmp(endstr, "fletcher4") == 0) 2288 zdb_sig_cksumalg = ZIO_CHECKSUM_FLETCHER_4; 2289 else if (strcmp(endstr, "sha256") == 0) 2290 zdb_sig_cksumalg = ZIO_CHECKSUM_SHA256; 2291 else if (strcmp(endstr, "all") == 0) 2292 zdb_sig_cksumalg = ZIO_CHECKSUM_FLETCHER_2; 2293 else 2294 usage(); 2295 break; 2296 case 't': 2297 ub_max_txg = strtoull(optarg, NULL, 0); 2298 if (ub_max_txg < TXG_INITIAL) { 2299 (void) fprintf(stderr, "incorrect txg " 2300 "specified: %s\n", optarg); 2301 usage(); 2302 } 2303 break; 2304 default: 2305 usage(); 2306 break; 2307 } 2308 } 2309 2310 if (vdev_dir != NULL && exported == 0) { 2311 (void) fprintf(stderr, "-p option requires use of -e\n"); 2312 usage(); 2313 } 2314 2315 kernel_init(FREAD); 2316 g_zfs = libzfs_init(); 2317 ASSERT(g_zfs != NULL); 2318 2319 for (c = 0; c < 256; c++) { 2320 if (dump_all && c != 'l' && c != 'R') 2321 dump_opt[c] = 1; 2322 if (dump_opt[c]) 2323 dump_opt[c] += verbose; 2324 } 2325 2326 argc -= optind; 2327 argv += optind; 2328 2329 if (argc < 1) { 2330 if (dump_opt['C']) { 2331 dump_cachefile(spa_config_path); 2332 return (0); 2333 } 2334 usage(); 2335 } 2336 2337 if (dump_opt['l']) { 2338 dump_label(argv[0]); 2339 return (0); 2340 } 2341 2342 if (dump_opt['R']) { 2343 flagbits['b'] = ZDB_FLAG_PRINT_BLKPTR; 2344 flagbits['c'] = ZDB_FLAG_CHECKSUM; 2345 flagbits['d'] = ZDB_FLAG_DECOMPRESS; 2346 flagbits['e'] = ZDB_FLAG_BSWAP; 2347 flagbits['g'] = ZDB_FLAG_GBH; 2348 flagbits['i'] = ZDB_FLAG_INDIRECT; 2349 flagbits['p'] = ZDB_FLAG_PHYS; 2350 flagbits['r'] = ZDB_FLAG_RAW; 2351 2352 spa = NULL; 2353 while (argv[0]) { 2354 zdb_read_block(argv[0], &spa); 2355 argv++; 2356 argc--; 2357 } 2358 if (spa) 2359 spa_close(spa, (void *)zdb_read_block); 2360 return (0); 2361 } 2362 2363 if (dump_opt['C']) 2364 dump_config(argv[0]); 2365 2366 error = 0; 2367 if (exported) { 2368 /* 2369 * Check to see if the name refers to an exported zpool 2370 */ 2371 char *slash; 2372 nvlist_t *exported_conf = NULL; 2373 2374 if ((slash = strchr(argv[0], '/')) != NULL) 2375 *slash = '\0'; 2376 2377 error = find_exported_zpool(argv[0], &exported_conf, vdev_dir); 2378 if (error == 0) { 2379 nvlist_t *nvl = NULL; 2380 2381 if (vdev_dir != NULL) { 2382 if (nvlist_alloc(&nvl, NV_UNIQUE_NAME, 0) != 0) 2383 error = ENOMEM; 2384 else if (nvlist_add_string(nvl, 2385 zpool_prop_to_name(ZPOOL_PROP_ALTROOT), 2386 vdev_dir) != 0) 2387 error = ENOMEM; 2388 } 2389 2390 if (error == 0) 2391 error = spa_import_faulted(argv[0], 2392 exported_conf, nvl); 2393 2394 nvlist_free(nvl); 2395 } 2396 2397 if (slash != NULL) 2398 *slash = '/'; 2399 } 2400 2401 if (error == 0) { 2402 if (strchr(argv[0], '/') != NULL) { 2403 error = dmu_objset_open(argv[0], DMU_OST_ANY, 2404 DS_MODE_USER | DS_MODE_READONLY, &os); 2405 } else { 2406 error = spa_open(argv[0], &spa, FTAG); 2407 } 2408 } 2409 2410 if (error) 2411 fatal("can't open %s: %s", argv[0], strerror(error)); 2412 2413 argv++; 2414 if (--argc > 0) { 2415 zopt_objects = argc; 2416 zopt_object = calloc(zopt_objects, sizeof (uint64_t)); 2417 for (i = 0; i < zopt_objects; i++) { 2418 errno = 0; 2419 zopt_object[i] = strtoull(argv[i], NULL, 0); 2420 if (zopt_object[i] == 0 && errno != 0) 2421 fatal("bad object number %s: %s", 2422 argv[i], strerror(errno)); 2423 } 2424 } 2425 2426 if (os != NULL) { 2427 dump_dir(os); 2428 dmu_objset_close(os); 2429 } else { 2430 dump_zpool(spa); 2431 spa_close(spa, FTAG); 2432 } 2433 2434 fuid_table_destroy(); 2435 2436 libzfs_fini(g_zfs); 2437 kernel_fini(); 2438 2439 return (0); 2440 } 2441