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 (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved. 23 */ 24 25 /* Portions Copyright 2010 Robert Milkowski */ 26 27 #include <mdb/mdb_ctf.h> 28 #include <sys/zfs_context.h> 29 #include <sys/mdb_modapi.h> 30 #include <sys/dbuf.h> 31 #include <sys/dmu_objset.h> 32 #include <sys/dsl_dir.h> 33 #include <sys/dsl_pool.h> 34 #include <sys/metaslab_impl.h> 35 #include <sys/space_map.h> 36 #include <sys/list.h> 37 #include <sys/spa_impl.h> 38 #include <sys/vdev_impl.h> 39 #include <sys/zap_leaf.h> 40 #include <sys/zap_impl.h> 41 #include <ctype.h> 42 #include <sys/zfs_acl.h> 43 #include <sys/sa_impl.h> 44 45 #ifdef _KERNEL 46 #define ZFS_OBJ_NAME "zfs" 47 #else 48 #define ZFS_OBJ_NAME "libzpool.so.1" 49 #endif 50 51 #ifndef _KERNEL 52 int aok; 53 #endif 54 55 static int 56 getmember(uintptr_t addr, const char *type, mdb_ctf_id_t *idp, 57 const char *member, int len, void *buf) 58 { 59 mdb_ctf_id_t id; 60 ulong_t off; 61 char name[64]; 62 63 if (idp == NULL) { 64 if (mdb_ctf_lookup_by_name(type, &id) == -1) { 65 mdb_warn("couldn't find type %s", type); 66 return (DCMD_ERR); 67 } 68 idp = &id; 69 } else { 70 type = name; 71 mdb_ctf_type_name(*idp, name, sizeof (name)); 72 } 73 74 if (mdb_ctf_offsetof(*idp, member, &off) == -1) { 75 mdb_warn("couldn't find member %s of type %s\n", member, type); 76 return (DCMD_ERR); 77 } 78 if (off % 8 != 0) { 79 mdb_warn("member %s of type %s is unsupported bitfield", 80 member, type); 81 return (DCMD_ERR); 82 } 83 off /= 8; 84 85 if (mdb_vread(buf, len, addr + off) == -1) { 86 mdb_warn("failed to read %s from %s at %p", 87 member, type, addr + off); 88 return (DCMD_ERR); 89 } 90 /* mdb_warn("read %s from %s at %p+%llx\n", member, type, addr, off); */ 91 92 return (0); 93 } 94 95 #define GETMEMB(addr, type, member, dest) \ 96 getmember(addr, #type, NULL, #member, sizeof (dest), &(dest)) 97 98 #define GETMEMBID(addr, ctfid, member, dest) \ 99 getmember(addr, NULL, ctfid, #member, sizeof (dest), &(dest)) 100 101 static int 102 getrefcount(uintptr_t addr, mdb_ctf_id_t *id, 103 const char *member, uint64_t *rc) 104 { 105 static int gotid; 106 static mdb_ctf_id_t rc_id; 107 ulong_t off; 108 109 if (!gotid) { 110 if (mdb_ctf_lookup_by_name("struct refcount", &rc_id) == -1) { 111 mdb_warn("couldn't find struct refcount"); 112 return (DCMD_ERR); 113 } 114 gotid = TRUE; 115 } 116 117 if (mdb_ctf_offsetof(*id, member, &off) == -1) { 118 char name[64]; 119 mdb_ctf_type_name(*id, name, sizeof (name)); 120 mdb_warn("couldn't find member %s of type %s\n", member, name); 121 return (DCMD_ERR); 122 } 123 off /= 8; 124 125 return (GETMEMBID(addr + off, &rc_id, rc_count, *rc)); 126 } 127 128 static boolean_t 129 strisprint(const char *cp) 130 { 131 for (; *cp; cp++) { 132 if (!isprint(*cp)) 133 return (B_FALSE); 134 } 135 return (B_TRUE); 136 } 137 138 static int verbose; 139 140 static int 141 freelist_walk_init(mdb_walk_state_t *wsp) 142 { 143 if (wsp->walk_addr == NULL) { 144 mdb_warn("must supply starting address\n"); 145 return (WALK_ERR); 146 } 147 148 wsp->walk_data = 0; /* Index into the freelist */ 149 return (WALK_NEXT); 150 } 151 152 static int 153 freelist_walk_step(mdb_walk_state_t *wsp) 154 { 155 uint64_t entry; 156 uintptr_t number = (uintptr_t)wsp->walk_data; 157 char *ddata[] = { "ALLOC", "FREE", "CONDENSE", "INVALID", 158 "INVALID", "INVALID", "INVALID", "INVALID" }; 159 int mapshift = SPA_MINBLOCKSHIFT; 160 161 if (mdb_vread(&entry, sizeof (entry), wsp->walk_addr) == -1) { 162 mdb_warn("failed to read freelist entry %p", wsp->walk_addr); 163 return (WALK_DONE); 164 } 165 wsp->walk_addr += sizeof (entry); 166 wsp->walk_data = (void *)(number + 1); 167 168 if (SM_DEBUG_DECODE(entry)) { 169 mdb_printf("DEBUG: %3u %10s: txg=%llu pass=%llu\n", 170 number, 171 ddata[SM_DEBUG_ACTION_DECODE(entry)], 172 SM_DEBUG_TXG_DECODE(entry), 173 SM_DEBUG_SYNCPASS_DECODE(entry)); 174 } else { 175 mdb_printf("Entry: %3u offsets=%08llx-%08llx type=%c " 176 "size=%06llx", number, 177 SM_OFFSET_DECODE(entry) << mapshift, 178 (SM_OFFSET_DECODE(entry) + SM_RUN_DECODE(entry)) << 179 mapshift, 180 SM_TYPE_DECODE(entry) == SM_ALLOC ? 'A' : 'F', 181 SM_RUN_DECODE(entry) << mapshift); 182 if (verbose) 183 mdb_printf(" (raw=%012llx)\n", entry); 184 mdb_printf("\n"); 185 } 186 return (WALK_NEXT); 187 } 188 189 190 static int 191 dataset_name(uintptr_t addr, char *buf) 192 { 193 static int gotid; 194 static mdb_ctf_id_t dd_id; 195 uintptr_t dd_parent; 196 char dd_myname[MAXNAMELEN]; 197 198 if (!gotid) { 199 if (mdb_ctf_lookup_by_name("struct dsl_dir", 200 &dd_id) == -1) { 201 mdb_warn("couldn't find struct dsl_dir"); 202 return (DCMD_ERR); 203 } 204 gotid = TRUE; 205 } 206 if (GETMEMBID(addr, &dd_id, dd_parent, dd_parent) || 207 GETMEMBID(addr, &dd_id, dd_myname, dd_myname)) { 208 return (DCMD_ERR); 209 } 210 211 if (dd_parent) { 212 if (dataset_name(dd_parent, buf)) 213 return (DCMD_ERR); 214 strcat(buf, "/"); 215 } 216 217 if (dd_myname[0]) 218 strcat(buf, dd_myname); 219 else 220 strcat(buf, "???"); 221 222 return (0); 223 } 224 225 static int 226 objset_name(uintptr_t addr, char *buf) 227 { 228 static int gotid; 229 static mdb_ctf_id_t os_id, ds_id; 230 uintptr_t os_dsl_dataset; 231 char ds_snapname[MAXNAMELEN]; 232 uintptr_t ds_dir; 233 234 buf[0] = '\0'; 235 236 if (!gotid) { 237 if (mdb_ctf_lookup_by_name("struct objset", 238 &os_id) == -1) { 239 mdb_warn("couldn't find struct objset"); 240 return (DCMD_ERR); 241 } 242 if (mdb_ctf_lookup_by_name("struct dsl_dataset", 243 &ds_id) == -1) { 244 mdb_warn("couldn't find struct dsl_dataset"); 245 return (DCMD_ERR); 246 } 247 248 gotid = TRUE; 249 } 250 251 if (GETMEMBID(addr, &os_id, os_dsl_dataset, os_dsl_dataset)) 252 return (DCMD_ERR); 253 254 if (os_dsl_dataset == 0) { 255 strcat(buf, "mos"); 256 return (0); 257 } 258 259 if (GETMEMBID(os_dsl_dataset, &ds_id, ds_snapname, ds_snapname) || 260 GETMEMBID(os_dsl_dataset, &ds_id, ds_dir, ds_dir)) { 261 return (DCMD_ERR); 262 } 263 264 if (ds_dir && dataset_name(ds_dir, buf)) 265 return (DCMD_ERR); 266 267 if (ds_snapname[0]) { 268 strcat(buf, "@"); 269 strcat(buf, ds_snapname); 270 } 271 return (0); 272 } 273 274 static void 275 enum_lookup(char *out, size_t size, mdb_ctf_id_t id, int val, 276 const char *prefix) 277 { 278 const char *cp; 279 size_t len = strlen(prefix); 280 281 if ((cp = mdb_ctf_enum_name(id, val)) != NULL) { 282 if (strncmp(cp, prefix, len) == 0) 283 cp += len; 284 (void) strncpy(out, cp, size); 285 } else { 286 mdb_snprintf(out, size, "? (%d)", val); 287 } 288 } 289 290 /* ARGSUSED */ 291 static int 292 zfs_params(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv) 293 { 294 /* 295 * This table can be approximately generated by running: 296 * egrep "^[a-z0-9_]+ [a-z0-9_]+( =.*)?;" *.c | cut -d ' ' -f 2 297 */ 298 static const char *params[] = { 299 "arc_reduce_dnlc_percent", 300 "zfs_arc_max", 301 "zfs_arc_min", 302 "arc_shrink_shift", 303 "zfs_mdcomp_disable", 304 "zfs_prefetch_disable", 305 "zfetch_max_streams", 306 "zfetch_min_sec_reap", 307 "zfetch_block_cap", 308 "zfetch_array_rd_sz", 309 "zfs_default_bs", 310 "zfs_default_ibs", 311 "metaslab_aliquot", 312 "reference_tracking_enable", 313 "reference_history", 314 "spa_max_replication_override", 315 "spa_mode_global", 316 "zfs_flags", 317 "zfs_txg_synctime", 318 "zfs_txg_timeout", 319 "zfs_write_limit_min", 320 "zfs_write_limit_max", 321 "zfs_write_limit_shift", 322 "zfs_write_limit_override", 323 "zfs_no_write_throttle", 324 "zfs_vdev_cache_max", 325 "zfs_vdev_cache_size", 326 "zfs_vdev_cache_bshift", 327 "vdev_mirror_shift", 328 "zfs_vdev_max_pending", 329 "zfs_vdev_min_pending", 330 "zfs_scrub_limit", 331 "zfs_no_scrub_io", 332 "zfs_no_scrub_prefetch", 333 "zfs_vdev_time_shift", 334 "zfs_vdev_ramp_rate", 335 "zfs_vdev_aggregation_limit", 336 "fzap_default_block_shift", 337 "zfs_immediate_write_sz", 338 "zfs_read_chunk_size", 339 "zfs_nocacheflush", 340 "zil_replay_disable", 341 "metaslab_gang_bang", 342 "metaslab_df_alloc_threshold", 343 "metaslab_df_free_pct", 344 "zio_injection_enabled", 345 "zvol_immediate_write_sz", 346 }; 347 348 for (int i = 0; i < sizeof (params) / sizeof (params[0]); i++) { 349 int sz; 350 uint64_t val64; 351 uint32_t *val32p = (uint32_t *)&val64; 352 353 sz = mdb_readvar(&val64, params[i]); 354 if (sz == 4) { 355 mdb_printf("%s = 0x%x\n", params[i], *val32p); 356 } else if (sz == 8) { 357 mdb_printf("%s = 0x%llx\n", params[i], val64); 358 } else { 359 mdb_warn("variable %s not found", params[i]); 360 } 361 } 362 363 return (DCMD_OK); 364 } 365 366 /* ARGSUSED */ 367 static int 368 blkptr(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv) 369 { 370 mdb_ctf_id_t type_enum, checksum_enum, compress_enum; 371 char type[80], checksum[80], compress[80]; 372 blkptr_t blk, *bp = &blk; 373 char buf[BP_SPRINTF_LEN]; 374 375 if (mdb_vread(&blk, sizeof (blkptr_t), addr) == -1) { 376 mdb_warn("failed to read blkptr_t"); 377 return (DCMD_ERR); 378 } 379 380 if (mdb_ctf_lookup_by_name("enum dmu_object_type", &type_enum) == -1 || 381 mdb_ctf_lookup_by_name("enum zio_checksum", &checksum_enum) == -1 || 382 mdb_ctf_lookup_by_name("enum zio_compress", &compress_enum) == -1) { 383 mdb_warn("Could not find blkptr enumerated types"); 384 return (DCMD_ERR); 385 } 386 387 enum_lookup(type, sizeof (type), type_enum, 388 BP_GET_TYPE(bp), "DMU_OT_"); 389 enum_lookup(checksum, sizeof (checksum), checksum_enum, 390 BP_GET_CHECKSUM(bp), "ZIO_CHECKSUM_"); 391 enum_lookup(compress, sizeof (compress), compress_enum, 392 BP_GET_COMPRESS(bp), "ZIO_COMPRESS_"); 393 394 SPRINTF_BLKPTR(mdb_snprintf, '\n', buf, bp, type, checksum, compress); 395 396 mdb_printf("%s\n", buf); 397 398 return (DCMD_OK); 399 } 400 401 /* ARGSUSED */ 402 static int 403 dbuf(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv) 404 { 405 mdb_ctf_id_t id; 406 dmu_buf_t db; 407 uintptr_t objset; 408 uint8_t level; 409 uint64_t blkid; 410 uint64_t holds; 411 char objectname[32]; 412 char blkidname[32]; 413 char path[MAXNAMELEN]; 414 415 if (DCMD_HDRSPEC(flags)) { 416 mdb_printf(" addr object lvl blkid holds os\n"); 417 } 418 419 if (mdb_ctf_lookup_by_name("struct dmu_buf_impl", &id) == -1) { 420 mdb_warn("couldn't find struct dmu_buf_impl_t"); 421 return (DCMD_ERR); 422 } 423 424 if (GETMEMBID(addr, &id, db_objset, objset) || 425 GETMEMBID(addr, &id, db, db) || 426 GETMEMBID(addr, &id, db_level, level) || 427 GETMEMBID(addr, &id, db_blkid, blkid)) { 428 return (WALK_ERR); 429 } 430 431 if (getrefcount(addr, &id, "db_holds", &holds)) { 432 return (WALK_ERR); 433 } 434 435 if (db.db_object == DMU_META_DNODE_OBJECT) 436 (void) strcpy(objectname, "mdn"); 437 else 438 (void) mdb_snprintf(objectname, sizeof (objectname), "%llx", 439 (u_longlong_t)db.db_object); 440 441 if (blkid == DMU_BONUS_BLKID) 442 (void) strcpy(blkidname, "bonus"); 443 else 444 (void) mdb_snprintf(blkidname, sizeof (blkidname), "%llx", 445 (u_longlong_t)blkid); 446 447 if (objset_name(objset, path)) { 448 return (WALK_ERR); 449 } 450 451 mdb_printf("%p %8s %1u %9s %2llu %s\n", 452 addr, objectname, level, blkidname, holds, path); 453 454 return (DCMD_OK); 455 } 456 457 /* ARGSUSED */ 458 static int 459 dbuf_stats(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv) 460 { 461 #define HISTOSZ 32 462 uintptr_t dbp; 463 dmu_buf_impl_t db; 464 dbuf_hash_table_t ht; 465 uint64_t bucket, ndbufs; 466 uint64_t histo[HISTOSZ]; 467 uint64_t histo2[HISTOSZ]; 468 int i, maxidx; 469 470 if (mdb_readvar(&ht, "dbuf_hash_table") == -1) { 471 mdb_warn("failed to read 'dbuf_hash_table'"); 472 return (DCMD_ERR); 473 } 474 475 for (i = 0; i < HISTOSZ; i++) { 476 histo[i] = 0; 477 histo2[i] = 0; 478 } 479 480 ndbufs = 0; 481 for (bucket = 0; bucket < ht.hash_table_mask+1; bucket++) { 482 int len; 483 484 if (mdb_vread(&dbp, sizeof (void *), 485 (uintptr_t)(ht.hash_table+bucket)) == -1) { 486 mdb_warn("failed to read hash bucket %u at %p", 487 bucket, ht.hash_table+bucket); 488 return (DCMD_ERR); 489 } 490 491 len = 0; 492 while (dbp != 0) { 493 if (mdb_vread(&db, sizeof (dmu_buf_impl_t), 494 dbp) == -1) { 495 mdb_warn("failed to read dbuf at %p", dbp); 496 return (DCMD_ERR); 497 } 498 dbp = (uintptr_t)db.db_hash_next; 499 for (i = MIN(len, HISTOSZ - 1); i >= 0; i--) 500 histo2[i]++; 501 len++; 502 ndbufs++; 503 } 504 505 if (len >= HISTOSZ) 506 len = HISTOSZ-1; 507 histo[len]++; 508 } 509 510 mdb_printf("hash table has %llu buckets, %llu dbufs " 511 "(avg %llu buckets/dbuf)\n", 512 ht.hash_table_mask+1, ndbufs, 513 (ht.hash_table_mask+1)/ndbufs); 514 515 mdb_printf("\n"); 516 maxidx = 0; 517 for (i = 0; i < HISTOSZ; i++) 518 if (histo[i] > 0) 519 maxidx = i; 520 mdb_printf("hash chain length number of buckets\n"); 521 for (i = 0; i <= maxidx; i++) 522 mdb_printf("%u %llu\n", i, histo[i]); 523 524 mdb_printf("\n"); 525 maxidx = 0; 526 for (i = 0; i < HISTOSZ; i++) 527 if (histo2[i] > 0) 528 maxidx = i; 529 mdb_printf("hash chain depth number of dbufs\n"); 530 for (i = 0; i <= maxidx; i++) 531 mdb_printf("%u or more %llu %llu%%\n", 532 i, histo2[i], histo2[i]*100/ndbufs); 533 534 535 return (DCMD_OK); 536 } 537 538 #define CHAIN_END 0xffff 539 /* 540 * ::zap_leaf [-v] 541 * 542 * Print a zap_leaf_phys_t, assumed to be 16k 543 */ 544 /* ARGSUSED */ 545 static int 546 zap_leaf(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv) 547 { 548 char buf[16*1024]; 549 int verbose = B_FALSE; 550 int four = B_FALSE; 551 zap_leaf_t l; 552 zap_leaf_phys_t *zlp = (void *)buf; 553 int i; 554 555 if (mdb_getopts(argc, argv, 556 'v', MDB_OPT_SETBITS, TRUE, &verbose, 557 '4', MDB_OPT_SETBITS, TRUE, &four, 558 NULL) != argc) 559 return (DCMD_USAGE); 560 561 l.l_phys = zlp; 562 l.l_bs = 14; /* assume 16k blocks */ 563 if (four) 564 l.l_bs = 12; 565 566 if (!(flags & DCMD_ADDRSPEC)) { 567 return (DCMD_USAGE); 568 } 569 570 if (mdb_vread(buf, sizeof (buf), addr) == -1) { 571 mdb_warn("failed to read zap_leaf_phys_t at %p", addr); 572 return (DCMD_ERR); 573 } 574 575 if (zlp->l_hdr.lh_block_type != ZBT_LEAF || 576 zlp->l_hdr.lh_magic != ZAP_LEAF_MAGIC) { 577 mdb_warn("This does not appear to be a zap_leaf_phys_t"); 578 return (DCMD_ERR); 579 } 580 581 mdb_printf("zap_leaf_phys_t at %p:\n", addr); 582 mdb_printf(" lh_prefix_len = %u\n", zlp->l_hdr.lh_prefix_len); 583 mdb_printf(" lh_prefix = %llx\n", zlp->l_hdr.lh_prefix); 584 mdb_printf(" lh_nentries = %u\n", zlp->l_hdr.lh_nentries); 585 mdb_printf(" lh_nfree = %u\n", zlp->l_hdr.lh_nfree, 586 zlp->l_hdr.lh_nfree * 100 / (ZAP_LEAF_NUMCHUNKS(&l))); 587 mdb_printf(" lh_freelist = %u\n", zlp->l_hdr.lh_freelist); 588 mdb_printf(" lh_flags = %x (%s)\n", zlp->l_hdr.lh_flags, 589 zlp->l_hdr.lh_flags & ZLF_ENTRIES_CDSORTED ? 590 "ENTRIES_CDSORTED" : ""); 591 592 if (verbose) { 593 mdb_printf(" hash table:\n"); 594 for (i = 0; i < ZAP_LEAF_HASH_NUMENTRIES(&l); i++) { 595 if (zlp->l_hash[i] != CHAIN_END) 596 mdb_printf(" %u: %u\n", i, zlp->l_hash[i]); 597 } 598 } 599 600 mdb_printf(" chunks:\n"); 601 for (i = 0; i < ZAP_LEAF_NUMCHUNKS(&l); i++) { 602 /* LINTED: alignment */ 603 zap_leaf_chunk_t *zlc = &ZAP_LEAF_CHUNK(&l, i); 604 switch (zlc->l_entry.le_type) { 605 case ZAP_CHUNK_FREE: 606 if (verbose) { 607 mdb_printf(" %u: free; lf_next = %u\n", 608 i, zlc->l_free.lf_next); 609 } 610 break; 611 case ZAP_CHUNK_ENTRY: 612 mdb_printf(" %u: entry\n", i); 613 if (verbose) { 614 mdb_printf(" le_next = %u\n", 615 zlc->l_entry.le_next); 616 } 617 mdb_printf(" le_name_chunk = %u\n", 618 zlc->l_entry.le_name_chunk); 619 mdb_printf(" le_name_numints = %u\n", 620 zlc->l_entry.le_name_numints); 621 mdb_printf(" le_value_chunk = %u\n", 622 zlc->l_entry.le_value_chunk); 623 mdb_printf(" le_value_intlen = %u\n", 624 zlc->l_entry.le_value_intlen); 625 mdb_printf(" le_value_numints = %u\n", 626 zlc->l_entry.le_value_numints); 627 mdb_printf(" le_cd = %u\n", 628 zlc->l_entry.le_cd); 629 mdb_printf(" le_hash = %llx\n", 630 zlc->l_entry.le_hash); 631 break; 632 case ZAP_CHUNK_ARRAY: 633 mdb_printf(" %u: array", i); 634 if (strisprint((char *)zlc->l_array.la_array)) 635 mdb_printf(" \"%s\"", zlc->l_array.la_array); 636 mdb_printf("\n"); 637 if (verbose) { 638 int j; 639 mdb_printf(" "); 640 for (j = 0; j < ZAP_LEAF_ARRAY_BYTES; j++) { 641 mdb_printf("%02x ", 642 zlc->l_array.la_array[j]); 643 } 644 mdb_printf("\n"); 645 } 646 if (zlc->l_array.la_next != CHAIN_END) { 647 mdb_printf(" lf_next = %u\n", 648 zlc->l_array.la_next); 649 } 650 break; 651 default: 652 mdb_printf(" %u: undefined type %u\n", 653 zlc->l_entry.le_type); 654 } 655 } 656 657 return (DCMD_OK); 658 } 659 660 typedef struct dbufs_data { 661 mdb_ctf_id_t id; 662 uint64_t objset; 663 uint64_t object; 664 uint64_t level; 665 uint64_t blkid; 666 char *osname; 667 } dbufs_data_t; 668 669 #define DBUFS_UNSET (0xbaddcafedeadbeefULL) 670 671 /* ARGSUSED */ 672 static int 673 dbufs_cb(uintptr_t addr, const void *unknown, void *arg) 674 { 675 dbufs_data_t *data = arg; 676 uintptr_t objset; 677 dmu_buf_t db; 678 uint8_t level; 679 uint64_t blkid; 680 char osname[MAXNAMELEN]; 681 682 if (GETMEMBID(addr, &data->id, db_objset, objset) || 683 GETMEMBID(addr, &data->id, db, db) || 684 GETMEMBID(addr, &data->id, db_level, level) || 685 GETMEMBID(addr, &data->id, db_blkid, blkid)) { 686 return (WALK_ERR); 687 } 688 689 if ((data->objset == DBUFS_UNSET || data->objset == objset) && 690 (data->osname == NULL || (objset_name(objset, osname) == 0 && 691 strcmp(data->osname, osname) == 0)) && 692 (data->object == DBUFS_UNSET || data->object == db.db_object) && 693 (data->level == DBUFS_UNSET || data->level == level) && 694 (data->blkid == DBUFS_UNSET || data->blkid == blkid)) { 695 mdb_printf("%#lr\n", addr); 696 } 697 return (WALK_NEXT); 698 } 699 700 /* ARGSUSED */ 701 static int 702 dbufs(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv) 703 { 704 dbufs_data_t data; 705 char *object = NULL; 706 char *blkid = NULL; 707 708 data.objset = data.object = data.level = data.blkid = DBUFS_UNSET; 709 data.osname = NULL; 710 711 if (mdb_getopts(argc, argv, 712 'O', MDB_OPT_UINT64, &data.objset, 713 'n', MDB_OPT_STR, &data.osname, 714 'o', MDB_OPT_STR, &object, 715 'l', MDB_OPT_UINT64, &data.level, 716 'b', MDB_OPT_STR, &blkid) != argc) { 717 return (DCMD_USAGE); 718 } 719 720 if (object) { 721 if (strcmp(object, "mdn") == 0) { 722 data.object = DMU_META_DNODE_OBJECT; 723 } else { 724 data.object = mdb_strtoull(object); 725 } 726 } 727 728 if (blkid) { 729 if (strcmp(blkid, "bonus") == 0) { 730 data.blkid = DMU_BONUS_BLKID; 731 } else { 732 data.blkid = mdb_strtoull(blkid); 733 } 734 } 735 736 if (mdb_ctf_lookup_by_name("struct dmu_buf_impl", &data.id) == -1) { 737 mdb_warn("couldn't find struct dmu_buf_impl_t"); 738 return (DCMD_ERR); 739 } 740 741 if (mdb_walk("dmu_buf_impl_t", dbufs_cb, &data) != 0) { 742 mdb_warn("can't walk dbufs"); 743 return (DCMD_ERR); 744 } 745 746 return (DCMD_OK); 747 } 748 749 typedef struct abuf_find_data { 750 dva_t dva; 751 mdb_ctf_id_t id; 752 } abuf_find_data_t; 753 754 /* ARGSUSED */ 755 static int 756 abuf_find_cb(uintptr_t addr, const void *unknown, void *arg) 757 { 758 abuf_find_data_t *data = arg; 759 dva_t dva; 760 761 if (GETMEMBID(addr, &data->id, b_dva, dva)) { 762 return (WALK_ERR); 763 } 764 765 if (dva.dva_word[0] == data->dva.dva_word[0] && 766 dva.dva_word[1] == data->dva.dva_word[1]) { 767 mdb_printf("%#lr\n", addr); 768 } 769 return (WALK_NEXT); 770 } 771 772 /* ARGSUSED */ 773 static int 774 abuf_find(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv) 775 { 776 abuf_find_data_t data; 777 GElf_Sym sym; 778 int i; 779 const char *syms[] = { 780 "ARC_mru", 781 "ARC_mru_ghost", 782 "ARC_mfu", 783 "ARC_mfu_ghost", 784 }; 785 786 if (argc != 2) 787 return (DCMD_USAGE); 788 789 for (i = 0; i < 2; i ++) { 790 switch (argv[i].a_type) { 791 case MDB_TYPE_STRING: 792 data.dva.dva_word[i] = mdb_strtoull(argv[i].a_un.a_str); 793 break; 794 case MDB_TYPE_IMMEDIATE: 795 data.dva.dva_word[i] = argv[i].a_un.a_val; 796 break; 797 default: 798 return (DCMD_USAGE); 799 } 800 } 801 802 if (mdb_ctf_lookup_by_name("struct arc_buf_hdr", &data.id) == -1) { 803 mdb_warn("couldn't find struct arc_buf_hdr"); 804 return (DCMD_ERR); 805 } 806 807 for (i = 0; i < sizeof (syms) / sizeof (syms[0]); i++) { 808 if (mdb_lookup_by_name(syms[i], &sym)) { 809 mdb_warn("can't find symbol %s", syms[i]); 810 return (DCMD_ERR); 811 } 812 813 if (mdb_pwalk("list", abuf_find_cb, &data, sym.st_value) != 0) { 814 mdb_warn("can't walk %s", syms[i]); 815 return (DCMD_ERR); 816 } 817 } 818 819 return (DCMD_OK); 820 } 821 822 /* ARGSUSED */ 823 static int 824 dbgmsg_cb(uintptr_t addr, const void *unknown, void *arg) 825 { 826 static mdb_ctf_id_t id; 827 static boolean_t gotid; 828 static ulong_t off; 829 830 int *verbosep = arg; 831 time_t timestamp; 832 char buf[1024]; 833 834 if (!gotid) { 835 if (mdb_ctf_lookup_by_name("struct zfs_dbgmsg", &id) == -1) { 836 mdb_warn("couldn't find struct zfs_dbgmsg"); 837 return (WALK_ERR); 838 } 839 gotid = TRUE; 840 if (mdb_ctf_offsetof(id, "zdm_msg", &off) == -1) { 841 mdb_warn("couldn't find zdm_msg"); 842 return (WALK_ERR); 843 } 844 off /= 8; 845 } 846 847 848 if (GETMEMBID(addr, &id, zdm_timestamp, timestamp)) { 849 return (WALK_ERR); 850 } 851 852 if (mdb_readstr(buf, sizeof (buf), addr + off) == -1) { 853 mdb_warn("failed to read zdm_msg at %p\n", addr + off); 854 return (DCMD_ERR); 855 } 856 857 if (*verbosep) 858 mdb_printf("%Y ", timestamp); 859 860 mdb_printf("%s\n", buf); 861 862 if (*verbosep) 863 (void) mdb_call_dcmd("whatis", addr, DCMD_ADDRSPEC, 0, NULL); 864 865 return (WALK_NEXT); 866 } 867 868 /* ARGSUSED */ 869 static int 870 dbgmsg(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv) 871 { 872 GElf_Sym sym; 873 int verbose = FALSE; 874 875 if (mdb_getopts(argc, argv, 876 'v', MDB_OPT_SETBITS, TRUE, &verbose, 877 NULL) != argc) 878 return (DCMD_USAGE); 879 880 if (mdb_lookup_by_name("zfs_dbgmsgs", &sym)) { 881 mdb_warn("can't find zfs_dbgmsgs"); 882 return (DCMD_ERR); 883 } 884 885 if (mdb_pwalk("list", dbgmsg_cb, &verbose, sym.st_value) != 0) { 886 mdb_warn("can't walk zfs_dbgmsgs"); 887 return (DCMD_ERR); 888 } 889 890 return (DCMD_OK); 891 } 892 893 /*ARGSUSED*/ 894 static int 895 arc_print(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv) 896 { 897 kstat_named_t *stats; 898 GElf_Sym sym; 899 int nstats, i; 900 uint_t opt_a = FALSE; 901 uint_t opt_b = FALSE; 902 uint_t shift = 0; 903 const char *suffix; 904 905 static const char *bytestats[] = { 906 "p", "c", "c_min", "c_max", "size", NULL 907 }; 908 909 static const char *extras[] = { 910 "arc_no_grow", "arc_tempreserve", 911 "arc_meta_used", "arc_meta_limit", "arc_meta_max", 912 NULL 913 }; 914 915 if (mdb_lookup_by_name("arc_stats", &sym) == -1) { 916 mdb_warn("failed to find 'arc_stats'"); 917 return (DCMD_ERR); 918 } 919 920 stats = mdb_zalloc(sym.st_size, UM_SLEEP | UM_GC); 921 922 if (mdb_vread(stats, sym.st_size, sym.st_value) == -1) { 923 mdb_warn("couldn't read 'arc_stats' at %p", sym.st_value); 924 return (DCMD_ERR); 925 } 926 927 nstats = sym.st_size / sizeof (kstat_named_t); 928 929 /* NB: -a / opt_a are ignored for backwards compatability */ 930 if (mdb_getopts(argc, argv, 931 'a', MDB_OPT_SETBITS, TRUE, &opt_a, 932 'b', MDB_OPT_SETBITS, TRUE, &opt_b, 933 'k', MDB_OPT_SETBITS, 10, &shift, 934 'm', MDB_OPT_SETBITS, 20, &shift, 935 'g', MDB_OPT_SETBITS, 30, &shift, 936 NULL) != argc) 937 return (DCMD_USAGE); 938 939 if (!opt_b && !shift) 940 shift = 20; 941 942 switch (shift) { 943 case 0: 944 suffix = "B"; 945 break; 946 case 10: 947 suffix = "KB"; 948 break; 949 case 20: 950 suffix = "MB"; 951 break; 952 case 30: 953 suffix = "GB"; 954 break; 955 default: 956 suffix = "XX"; 957 } 958 959 for (i = 0; i < nstats; i++) { 960 int j; 961 boolean_t bytes = B_FALSE; 962 963 for (j = 0; bytestats[j]; j++) { 964 if (strcmp(stats[i].name, bytestats[j]) == 0) { 965 bytes = B_TRUE; 966 break; 967 } 968 } 969 970 if (bytes) { 971 mdb_printf("%-25s = %9llu %s\n", stats[i].name, 972 stats[i].value.ui64 >> shift, suffix); 973 } else { 974 mdb_printf("%-25s = %9llu\n", stats[i].name, 975 stats[i].value.ui64); 976 } 977 } 978 979 for (i = 0; extras[i]; i++) { 980 uint64_t buf; 981 982 if (mdb_lookup_by_name(extras[i], &sym) == -1) { 983 mdb_warn("failed to find '%s'", extras[i]); 984 return (DCMD_ERR); 985 } 986 987 if (sym.st_size != sizeof (uint64_t) && 988 sym.st_size != sizeof (uint32_t)) { 989 mdb_warn("expected scalar for variable '%s'\n", 990 extras[i]); 991 return (DCMD_ERR); 992 } 993 994 if (mdb_vread(&buf, sym.st_size, sym.st_value) == -1) { 995 mdb_warn("couldn't read '%s'", extras[i]); 996 return (DCMD_ERR); 997 } 998 999 mdb_printf("%-25s = ", extras[i]); 1000 1001 /* NB: all the 64-bit extras happen to be byte counts */ 1002 if (sym.st_size == sizeof (uint64_t)) 1003 mdb_printf("%9llu %s\n", buf >> shift, suffix); 1004 1005 if (sym.st_size == sizeof (uint32_t)) 1006 mdb_printf("%9d\n", *((uint32_t *)&buf)); 1007 } 1008 return (DCMD_OK); 1009 } 1010 1011 /* 1012 * ::spa 1013 * 1014 * -c Print configuration information as well 1015 * -v Print vdev state 1016 * -e Print vdev error stats 1017 * 1018 * Print a summarized spa_t. When given no arguments, prints out a table of all 1019 * active pools on the system. 1020 */ 1021 /* ARGSUSED */ 1022 static int 1023 spa_print(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv) 1024 { 1025 spa_t spa; 1026 const char *statetab[] = { "ACTIVE", "EXPORTED", "DESTROYED", 1027 "SPARE", "L2CACHE", "UNINIT", "UNAVAIL", "POTENTIAL" }; 1028 const char *state; 1029 int config = FALSE; 1030 int vdevs = FALSE; 1031 int errors = FALSE; 1032 1033 if (mdb_getopts(argc, argv, 1034 'c', MDB_OPT_SETBITS, TRUE, &config, 1035 'v', MDB_OPT_SETBITS, TRUE, &vdevs, 1036 'e', MDB_OPT_SETBITS, TRUE, &errors, 1037 NULL) != argc) 1038 return (DCMD_USAGE); 1039 1040 if (!(flags & DCMD_ADDRSPEC)) { 1041 if (mdb_walk_dcmd("spa", "spa", argc, argv) == -1) { 1042 mdb_warn("can't walk spa"); 1043 return (DCMD_ERR); 1044 } 1045 1046 return (DCMD_OK); 1047 } 1048 1049 if (flags & DCMD_PIPE_OUT) { 1050 mdb_printf("%#lr\n", addr); 1051 return (DCMD_OK); 1052 } 1053 1054 if (DCMD_HDRSPEC(flags)) 1055 mdb_printf("%<u>%-?s %9s %-*s%</u>\n", "ADDR", "STATE", 1056 sizeof (uintptr_t) == 4 ? 60 : 52, "NAME"); 1057 1058 if (mdb_vread(&spa, sizeof (spa), addr) == -1) { 1059 mdb_warn("failed to read spa_t at %p", addr); 1060 return (DCMD_ERR); 1061 } 1062 1063 if (spa.spa_state < 0 || spa.spa_state > POOL_STATE_UNAVAIL) 1064 state = "UNKNOWN"; 1065 else 1066 state = statetab[spa.spa_state]; 1067 1068 mdb_printf("%0?p %9s %s\n", addr, state, spa.spa_name); 1069 1070 if (config) { 1071 mdb_printf("\n"); 1072 mdb_inc_indent(4); 1073 if (mdb_call_dcmd("spa_config", addr, flags, 0, 1074 NULL) != DCMD_OK) 1075 return (DCMD_ERR); 1076 mdb_dec_indent(4); 1077 } 1078 1079 if (vdevs || errors) { 1080 mdb_arg_t v; 1081 1082 v.a_type = MDB_TYPE_STRING; 1083 v.a_un.a_str = "-e"; 1084 1085 mdb_printf("\n"); 1086 mdb_inc_indent(4); 1087 if (mdb_call_dcmd("spa_vdevs", addr, flags, errors ? 1 : 0, 1088 &v) != DCMD_OK) 1089 return (DCMD_ERR); 1090 mdb_dec_indent(4); 1091 } 1092 1093 return (DCMD_OK); 1094 } 1095 1096 /* 1097 * ::spa_config 1098 * 1099 * Given a spa_t, print the configuration information stored in spa_config. 1100 * Since it's just an nvlist, format it as an indented list of name=value pairs. 1101 * We simply read the value of spa_config and pass off to ::nvlist. 1102 */ 1103 /* ARGSUSED */ 1104 static int 1105 spa_print_config(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv) 1106 { 1107 spa_t spa; 1108 1109 if (argc != 0 || !(flags & DCMD_ADDRSPEC)) 1110 return (DCMD_USAGE); 1111 1112 if (mdb_vread(&spa, sizeof (spa), addr) == -1) { 1113 mdb_warn("failed to read spa_t at %p", addr); 1114 return (DCMD_ERR); 1115 } 1116 1117 if (spa.spa_config == NULL) { 1118 mdb_printf("(none)\n"); 1119 return (DCMD_OK); 1120 } 1121 1122 return (mdb_call_dcmd("nvlist", (uintptr_t)spa.spa_config, flags, 1123 0, NULL)); 1124 } 1125 1126 /* 1127 * ::vdev 1128 * 1129 * Print out a summarized vdev_t, in the following form: 1130 * 1131 * ADDR STATE AUX DESC 1132 * fffffffbcde23df0 HEALTHY - /dev/dsk/c0t0d0 1133 * 1134 * If '-r' is specified, recursively visit all children. 1135 * 1136 * With '-e', the statistics associated with the vdev are printed as well. 1137 */ 1138 static int 1139 do_print_vdev(uintptr_t addr, int flags, int depth, int stats, 1140 int recursive) 1141 { 1142 vdev_t vdev; 1143 char desc[MAXNAMELEN]; 1144 int c, children; 1145 uintptr_t *child; 1146 const char *state, *aux; 1147 1148 if (mdb_vread(&vdev, sizeof (vdev), (uintptr_t)addr) == -1) { 1149 mdb_warn("failed to read vdev_t at %p\n", (uintptr_t)addr); 1150 return (DCMD_ERR); 1151 } 1152 1153 if (flags & DCMD_PIPE_OUT) { 1154 mdb_printf("%#lr", addr); 1155 } else { 1156 if (vdev.vdev_path != NULL) { 1157 if (mdb_readstr(desc, sizeof (desc), 1158 (uintptr_t)vdev.vdev_path) == -1) { 1159 mdb_warn("failed to read vdev_path at %p\n", 1160 vdev.vdev_path); 1161 return (DCMD_ERR); 1162 } 1163 } else if (vdev.vdev_ops != NULL) { 1164 vdev_ops_t ops; 1165 if (mdb_vread(&ops, sizeof (ops), 1166 (uintptr_t)vdev.vdev_ops) == -1) { 1167 mdb_warn("failed to read vdev_ops at %p\n", 1168 vdev.vdev_ops); 1169 return (DCMD_ERR); 1170 } 1171 (void) strcpy(desc, ops.vdev_op_type); 1172 } else { 1173 (void) strcpy(desc, "<unknown>"); 1174 } 1175 1176 if (depth == 0 && DCMD_HDRSPEC(flags)) 1177 mdb_printf("%<u>%-?s %-9s %-12s %-*s%</u>\n", 1178 "ADDR", "STATE", "AUX", 1179 sizeof (uintptr_t) == 4 ? 43 : 35, 1180 "DESCRIPTION"); 1181 1182 mdb_printf("%0?p ", addr); 1183 1184 switch (vdev.vdev_state) { 1185 case VDEV_STATE_CLOSED: 1186 state = "CLOSED"; 1187 break; 1188 case VDEV_STATE_OFFLINE: 1189 state = "OFFLINE"; 1190 break; 1191 case VDEV_STATE_CANT_OPEN: 1192 state = "CANT_OPEN"; 1193 break; 1194 case VDEV_STATE_DEGRADED: 1195 state = "DEGRADED"; 1196 break; 1197 case VDEV_STATE_HEALTHY: 1198 state = "HEALTHY"; 1199 break; 1200 case VDEV_STATE_REMOVED: 1201 state = "REMOVED"; 1202 break; 1203 case VDEV_STATE_FAULTED: 1204 state = "FAULTED"; 1205 break; 1206 default: 1207 state = "UNKNOWN"; 1208 break; 1209 } 1210 1211 switch (vdev.vdev_stat.vs_aux) { 1212 case VDEV_AUX_NONE: 1213 aux = "-"; 1214 break; 1215 case VDEV_AUX_OPEN_FAILED: 1216 aux = "OPEN_FAILED"; 1217 break; 1218 case VDEV_AUX_CORRUPT_DATA: 1219 aux = "CORRUPT_DATA"; 1220 break; 1221 case VDEV_AUX_NO_REPLICAS: 1222 aux = "NO_REPLICAS"; 1223 break; 1224 case VDEV_AUX_BAD_GUID_SUM: 1225 aux = "BAD_GUID_SUM"; 1226 break; 1227 case VDEV_AUX_TOO_SMALL: 1228 aux = "TOO_SMALL"; 1229 break; 1230 case VDEV_AUX_BAD_LABEL: 1231 aux = "BAD_LABEL"; 1232 break; 1233 case VDEV_AUX_VERSION_NEWER: 1234 aux = "VERS_NEWER"; 1235 break; 1236 case VDEV_AUX_VERSION_OLDER: 1237 aux = "VERS_OLDER"; 1238 break; 1239 case VDEV_AUX_SPARED: 1240 aux = "SPARED"; 1241 break; 1242 case VDEV_AUX_ERR_EXCEEDED: 1243 aux = "ERR_EXCEEDED"; 1244 break; 1245 case VDEV_AUX_IO_FAILURE: 1246 aux = "IO_FAILURE"; 1247 break; 1248 case VDEV_AUX_BAD_LOG: 1249 aux = "BAD_LOG"; 1250 break; 1251 case VDEV_AUX_EXTERNAL: 1252 aux = "EXTERNAL"; 1253 break; 1254 case VDEV_AUX_SPLIT_POOL: 1255 aux = "SPLIT_POOL"; 1256 break; 1257 default: 1258 aux = "UNKNOWN"; 1259 break; 1260 } 1261 1262 mdb_printf("%-9s %-12s %*s%s\n", state, aux, depth, "", desc); 1263 1264 if (stats) { 1265 vdev_stat_t *vs = &vdev.vdev_stat; 1266 int i; 1267 1268 mdb_inc_indent(4); 1269 mdb_printf("\n"); 1270 mdb_printf("%<u> %12s %12s %12s %12s " 1271 "%12s%</u>\n", "READ", "WRITE", "FREE", "CLAIM", 1272 "IOCTL"); 1273 mdb_printf("OPS "); 1274 for (i = 1; i < ZIO_TYPES; i++) 1275 mdb_printf("%11#llx%s", vs->vs_ops[i], 1276 i == ZIO_TYPES - 1 ? "" : " "); 1277 mdb_printf("\n"); 1278 mdb_printf("BYTES "); 1279 for (i = 1; i < ZIO_TYPES; i++) 1280 mdb_printf("%11#llx%s", vs->vs_bytes[i], 1281 i == ZIO_TYPES - 1 ? "" : " "); 1282 1283 1284 mdb_printf("\n"); 1285 mdb_printf("EREAD %10#llx\n", vs->vs_read_errors); 1286 mdb_printf("EWRITE %10#llx\n", vs->vs_write_errors); 1287 mdb_printf("ECKSUM %10#llx\n", 1288 vs->vs_checksum_errors); 1289 mdb_dec_indent(4); 1290 } 1291 1292 if (stats) 1293 mdb_printf("\n"); 1294 } 1295 1296 children = vdev.vdev_children; 1297 1298 if (children == 0 || !recursive) 1299 return (DCMD_OK); 1300 1301 child = mdb_alloc(children * sizeof (void *), UM_SLEEP | UM_GC); 1302 if (mdb_vread(child, children * sizeof (void *), 1303 (uintptr_t)vdev.vdev_child) == -1) { 1304 mdb_warn("failed to read vdev children at %p", vdev.vdev_child); 1305 return (DCMD_ERR); 1306 } 1307 1308 for (c = 0; c < children; c++) { 1309 if (do_print_vdev(child[c], flags, depth + 2, stats, 1310 recursive)) 1311 return (DCMD_ERR); 1312 } 1313 1314 return (DCMD_OK); 1315 } 1316 1317 static int 1318 vdev_print(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv) 1319 { 1320 int recursive = FALSE; 1321 int stats = FALSE; 1322 uint64_t depth = 0; 1323 1324 if (mdb_getopts(argc, argv, 1325 'r', MDB_OPT_SETBITS, TRUE, &recursive, 1326 'e', MDB_OPT_SETBITS, TRUE, &stats, 1327 'd', MDB_OPT_UINT64, &depth, 1328 NULL) != argc) 1329 return (DCMD_USAGE); 1330 1331 if (!(flags & DCMD_ADDRSPEC)) { 1332 mdb_warn("no vdev_t address given\n"); 1333 return (DCMD_ERR); 1334 } 1335 1336 return (do_print_vdev(addr, flags, (int)depth, stats, recursive)); 1337 } 1338 1339 typedef struct metaslab_walk_data { 1340 uint64_t mw_numvdevs; 1341 uintptr_t *mw_vdevs; 1342 int mw_curvdev; 1343 uint64_t mw_nummss; 1344 uintptr_t *mw_mss; 1345 int mw_curms; 1346 } metaslab_walk_data_t; 1347 1348 static int 1349 metaslab_walk_step(mdb_walk_state_t *wsp) 1350 { 1351 metaslab_walk_data_t *mw = wsp->walk_data; 1352 metaslab_t ms; 1353 uintptr_t msp; 1354 1355 if (mw->mw_curvdev >= mw->mw_numvdevs) 1356 return (WALK_DONE); 1357 1358 if (mw->mw_mss == NULL) { 1359 uintptr_t mssp; 1360 uintptr_t vdevp; 1361 1362 ASSERT(mw->mw_curms == 0); 1363 ASSERT(mw->mw_nummss == 0); 1364 1365 vdevp = mw->mw_vdevs[mw->mw_curvdev]; 1366 if (GETMEMB(vdevp, struct vdev, vdev_ms, mssp) || 1367 GETMEMB(vdevp, struct vdev, vdev_ms_count, mw->mw_nummss)) { 1368 return (WALK_ERR); 1369 } 1370 1371 mw->mw_mss = mdb_alloc(mw->mw_nummss * sizeof (void*), 1372 UM_SLEEP | UM_GC); 1373 if (mdb_vread(mw->mw_mss, mw->mw_nummss * sizeof (void*), 1374 mssp) == -1) { 1375 mdb_warn("failed to read vdev_ms at %p", mssp); 1376 return (WALK_ERR); 1377 } 1378 } 1379 1380 if (mw->mw_curms >= mw->mw_nummss) { 1381 mw->mw_mss = NULL; 1382 mw->mw_curms = 0; 1383 mw->mw_nummss = 0; 1384 mw->mw_curvdev++; 1385 return (WALK_NEXT); 1386 } 1387 1388 msp = mw->mw_mss[mw->mw_curms]; 1389 if (mdb_vread(&ms, sizeof (metaslab_t), msp) == -1) { 1390 mdb_warn("failed to read metaslab_t at %p", msp); 1391 return (WALK_ERR); 1392 } 1393 1394 mw->mw_curms++; 1395 1396 return (wsp->walk_callback(msp, &ms, wsp->walk_cbdata)); 1397 } 1398 1399 /* ARGSUSED */ 1400 static int 1401 metaslab_walk_init(mdb_walk_state_t *wsp) 1402 { 1403 metaslab_walk_data_t *mw; 1404 uintptr_t root_vdevp; 1405 uintptr_t childp; 1406 1407 if (wsp->walk_addr == NULL) { 1408 mdb_warn("must supply address of spa_t\n"); 1409 return (WALK_ERR); 1410 } 1411 1412 mw = mdb_zalloc(sizeof (metaslab_walk_data_t), UM_SLEEP | UM_GC); 1413 1414 if (GETMEMB(wsp->walk_addr, struct spa, spa_root_vdev, root_vdevp) || 1415 GETMEMB(root_vdevp, struct vdev, vdev_children, mw->mw_numvdevs) || 1416 GETMEMB(root_vdevp, struct vdev, vdev_child, childp)) { 1417 return (DCMD_ERR); 1418 } 1419 1420 mw->mw_vdevs = mdb_alloc(mw->mw_numvdevs * sizeof (void *), 1421 UM_SLEEP | UM_GC); 1422 if (mdb_vread(mw->mw_vdevs, mw->mw_numvdevs * sizeof (void *), 1423 childp) == -1) { 1424 mdb_warn("failed to read root vdev children at %p", childp); 1425 return (DCMD_ERR); 1426 } 1427 1428 wsp->walk_data = mw; 1429 1430 return (WALK_NEXT); 1431 } 1432 1433 typedef struct mdb_spa { 1434 uintptr_t spa_dsl_pool; 1435 uintptr_t spa_root_vdev; 1436 } mdb_spa_t; 1437 1438 typedef struct mdb_dsl_dir { 1439 uintptr_t dd_phys; 1440 int64_t dd_space_towrite[TXG_SIZE]; 1441 } mdb_dsl_dir_t; 1442 1443 typedef struct mdb_dsl_dir_phys { 1444 uint64_t dd_used_bytes; 1445 uint64_t dd_compressed_bytes; 1446 uint64_t dd_uncompressed_bytes; 1447 } mdb_dsl_dir_phys_t; 1448 1449 typedef struct mdb_vdev { 1450 uintptr_t vdev_parent; 1451 uintptr_t vdev_ms; 1452 uint64_t vdev_ms_count; 1453 vdev_stat_t vdev_stat; 1454 } mdb_vdev_t; 1455 1456 typedef struct mdb_metaslab { 1457 space_map_t ms_allocmap[TXG_SIZE]; 1458 space_map_t ms_freemap[TXG_SIZE]; 1459 space_map_t ms_map; 1460 space_map_obj_t ms_smo; 1461 space_map_obj_t ms_smo_syncing; 1462 } mdb_metaslab_t; 1463 1464 typedef struct space_data { 1465 uint64_t ms_allocmap[TXG_SIZE]; 1466 uint64_t ms_freemap[TXG_SIZE]; 1467 uint64_t ms_map; 1468 uint64_t avail; 1469 uint64_t nowavail; 1470 } space_data_t; 1471 1472 /* ARGSUSED */ 1473 static int 1474 space_cb(uintptr_t addr, const void *unknown, void *arg) 1475 { 1476 space_data_t *sd = arg; 1477 mdb_metaslab_t ms; 1478 1479 if (GETMEMB(addr, struct metaslab, ms_allocmap, ms.ms_allocmap) || 1480 GETMEMB(addr, struct metaslab, ms_freemap, ms.ms_freemap) || 1481 GETMEMB(addr, struct metaslab, ms_map, ms.ms_map) || 1482 GETMEMB(addr, struct metaslab, ms_smo, ms.ms_smo) || 1483 GETMEMB(addr, struct metaslab, ms_smo_syncing, ms.ms_smo_syncing)) { 1484 return (WALK_ERR); 1485 } 1486 1487 sd->ms_allocmap[0] += ms.ms_allocmap[0].sm_space; 1488 sd->ms_allocmap[1] += ms.ms_allocmap[1].sm_space; 1489 sd->ms_allocmap[2] += ms.ms_allocmap[2].sm_space; 1490 sd->ms_allocmap[3] += ms.ms_allocmap[3].sm_space; 1491 sd->ms_freemap[0] += ms.ms_freemap[0].sm_space; 1492 sd->ms_freemap[1] += ms.ms_freemap[1].sm_space; 1493 sd->ms_freemap[2] += ms.ms_freemap[2].sm_space; 1494 sd->ms_freemap[3] += ms.ms_freemap[3].sm_space; 1495 sd->ms_map += ms.ms_map.sm_space; 1496 sd->avail += ms.ms_map.sm_size - ms.ms_smo.smo_alloc; 1497 sd->nowavail += ms.ms_map.sm_size - ms.ms_smo_syncing.smo_alloc; 1498 1499 return (WALK_NEXT); 1500 } 1501 1502 /* 1503 * ::spa_space [-b] 1504 * 1505 * Given a spa_t, print out it's on-disk space usage and in-core 1506 * estimates of future usage. If -b is given, print space in bytes. 1507 * Otherwise print in megabytes. 1508 */ 1509 /* ARGSUSED */ 1510 static int 1511 spa_space(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv) 1512 { 1513 mdb_spa_t spa; 1514 uintptr_t dp_root_dir; 1515 mdb_dsl_dir_t dd; 1516 mdb_dsl_dir_phys_t dsp; 1517 uint64_t children; 1518 uintptr_t childaddr; 1519 space_data_t sd; 1520 int shift = 20; 1521 char *suffix = "M"; 1522 int bits = FALSE; 1523 1524 if (mdb_getopts(argc, argv, 'b', MDB_OPT_SETBITS, TRUE, &bits, NULL) != 1525 argc) 1526 return (DCMD_USAGE); 1527 if (!(flags & DCMD_ADDRSPEC)) 1528 return (DCMD_USAGE); 1529 1530 if (bits) { 1531 shift = 0; 1532 suffix = ""; 1533 } 1534 1535 if (GETMEMB(addr, struct spa, spa_dsl_pool, spa.spa_dsl_pool) || 1536 GETMEMB(addr, struct spa, spa_root_vdev, spa.spa_root_vdev) || 1537 GETMEMB(spa.spa_root_vdev, struct vdev, vdev_children, children) || 1538 GETMEMB(spa.spa_root_vdev, struct vdev, vdev_child, childaddr) || 1539 GETMEMB(spa.spa_dsl_pool, struct dsl_pool, 1540 dp_root_dir, dp_root_dir) || 1541 GETMEMB(dp_root_dir, struct dsl_dir, dd_phys, dd.dd_phys) || 1542 GETMEMB(dp_root_dir, struct dsl_dir, 1543 dd_space_towrite, dd.dd_space_towrite) || 1544 GETMEMB(dd.dd_phys, struct dsl_dir_phys, 1545 dd_used_bytes, dsp.dd_used_bytes) || 1546 GETMEMB(dd.dd_phys, struct dsl_dir_phys, 1547 dd_compressed_bytes, dsp.dd_compressed_bytes) || 1548 GETMEMB(dd.dd_phys, struct dsl_dir_phys, 1549 dd_uncompressed_bytes, dsp.dd_uncompressed_bytes)) { 1550 return (DCMD_ERR); 1551 } 1552 1553 mdb_printf("dd_space_towrite = %llu%s %llu%s %llu%s %llu%s\n", 1554 dd.dd_space_towrite[0] >> shift, suffix, 1555 dd.dd_space_towrite[1] >> shift, suffix, 1556 dd.dd_space_towrite[2] >> shift, suffix, 1557 dd.dd_space_towrite[3] >> shift, suffix); 1558 1559 mdb_printf("dd_phys.dd_used_bytes = %llu%s\n", 1560 dsp.dd_used_bytes >> shift, suffix); 1561 mdb_printf("dd_phys.dd_compressed_bytes = %llu%s\n", 1562 dsp.dd_compressed_bytes >> shift, suffix); 1563 mdb_printf("dd_phys.dd_uncompressed_bytes = %llu%s\n", 1564 dsp.dd_uncompressed_bytes >> shift, suffix); 1565 1566 bzero(&sd, sizeof (sd)); 1567 if (mdb_pwalk("metaslab", space_cb, &sd, addr) != 0) { 1568 mdb_warn("can't walk metaslabs"); 1569 return (DCMD_ERR); 1570 } 1571 1572 mdb_printf("ms_allocmap = %llu%s %llu%s %llu%s %llu%s\n", 1573 sd.ms_allocmap[0] >> shift, suffix, 1574 sd.ms_allocmap[1] >> shift, suffix, 1575 sd.ms_allocmap[2] >> shift, suffix, 1576 sd.ms_allocmap[3] >> shift, suffix); 1577 mdb_printf("ms_freemap = %llu%s %llu%s %llu%s %llu%s\n", 1578 sd.ms_freemap[0] >> shift, suffix, 1579 sd.ms_freemap[1] >> shift, suffix, 1580 sd.ms_freemap[2] >> shift, suffix, 1581 sd.ms_freemap[3] >> shift, suffix); 1582 mdb_printf("ms_map = %llu%s\n", sd.ms_map >> shift, suffix); 1583 mdb_printf("last synced avail = %llu%s\n", sd.avail >> shift, suffix); 1584 mdb_printf("current syncing avail = %llu%s\n", 1585 sd.nowavail >> shift, suffix); 1586 1587 return (DCMD_OK); 1588 } 1589 1590 /* 1591 * ::spa_verify 1592 * 1593 * Given a spa_t, verify that that the pool is self-consistent. 1594 * Currently, it only checks to make sure that the vdev tree exists. 1595 */ 1596 /* ARGSUSED */ 1597 static int 1598 spa_verify(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv) 1599 { 1600 spa_t spa; 1601 1602 if (argc != 0 || !(flags & DCMD_ADDRSPEC)) 1603 return (DCMD_USAGE); 1604 1605 if (mdb_vread(&spa, sizeof (spa), addr) == -1) { 1606 mdb_warn("failed to read spa_t at %p", addr); 1607 return (DCMD_ERR); 1608 } 1609 1610 if (spa.spa_root_vdev == NULL) { 1611 mdb_printf("no vdev tree present\n"); 1612 return (DCMD_OK); 1613 } 1614 1615 return (DCMD_OK); 1616 } 1617 1618 static int 1619 spa_print_aux(spa_aux_vdev_t *sav, uint_t flags, mdb_arg_t *v, 1620 const char *name) 1621 { 1622 uintptr_t *aux; 1623 size_t len; 1624 int ret, i; 1625 1626 /* 1627 * Iterate over aux vdevs and print those out as well. This is a 1628 * little annoying because we don't have a root vdev to pass to ::vdev. 1629 * Instead, we print a single line and then call it for each child 1630 * vdev. 1631 */ 1632 if (sav->sav_count != 0) { 1633 v[1].a_type = MDB_TYPE_STRING; 1634 v[1].a_un.a_str = "-d"; 1635 v[2].a_type = MDB_TYPE_IMMEDIATE; 1636 v[2].a_un.a_val = 2; 1637 1638 len = sav->sav_count * sizeof (uintptr_t); 1639 aux = mdb_alloc(len, UM_SLEEP); 1640 if (mdb_vread(aux, len, 1641 (uintptr_t)sav->sav_vdevs) == -1) { 1642 mdb_free(aux, len); 1643 mdb_warn("failed to read l2cache vdevs at %p", 1644 sav->sav_vdevs); 1645 return (DCMD_ERR); 1646 } 1647 1648 mdb_printf("%-?s %-9s %-12s %s\n", "-", "-", "-", name); 1649 1650 for (i = 0; i < sav->sav_count; i++) { 1651 ret = mdb_call_dcmd("vdev", aux[i], flags, 3, v); 1652 if (ret != DCMD_OK) { 1653 mdb_free(aux, len); 1654 return (ret); 1655 } 1656 } 1657 1658 mdb_free(aux, len); 1659 } 1660 1661 return (0); 1662 } 1663 1664 /* 1665 * ::spa_vdevs 1666 * 1667 * -e Include error stats 1668 * 1669 * Print out a summarized list of vdevs for the given spa_t. 1670 * This is accomplished by invoking "::vdev -re" on the root vdev, as well as 1671 * iterating over the cache devices. 1672 */ 1673 /* ARGSUSED */ 1674 static int 1675 spa_vdevs(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv) 1676 { 1677 spa_t spa; 1678 mdb_arg_t v[3]; 1679 int errors = FALSE; 1680 int ret; 1681 1682 if (mdb_getopts(argc, argv, 1683 'e', MDB_OPT_SETBITS, TRUE, &errors, 1684 NULL) != argc) 1685 return (DCMD_USAGE); 1686 1687 if (!(flags & DCMD_ADDRSPEC)) 1688 return (DCMD_USAGE); 1689 1690 if (mdb_vread(&spa, sizeof (spa), addr) == -1) { 1691 mdb_warn("failed to read spa_t at %p", addr); 1692 return (DCMD_ERR); 1693 } 1694 1695 /* 1696 * Unitialized spa_t structures can have a NULL root vdev. 1697 */ 1698 if (spa.spa_root_vdev == NULL) { 1699 mdb_printf("no associated vdevs\n"); 1700 return (DCMD_OK); 1701 } 1702 1703 v[0].a_type = MDB_TYPE_STRING; 1704 v[0].a_un.a_str = errors ? "-re" : "-r"; 1705 1706 ret = mdb_call_dcmd("vdev", (uintptr_t)spa.spa_root_vdev, 1707 flags, 1, v); 1708 if (ret != DCMD_OK) 1709 return (ret); 1710 1711 if (spa_print_aux(&spa.spa_l2cache, flags, v, "cache") != 0 || 1712 spa_print_aux(&spa.spa_spares, flags, v, "spares") != 0) 1713 return (DCMD_ERR); 1714 1715 return (DCMD_OK); 1716 } 1717 1718 /* 1719 * ::zio 1720 * 1721 * Print a summary of zio_t and all its children. This is intended to display a 1722 * zio tree, and hence we only pick the most important pieces of information for 1723 * the main summary. More detailed information can always be found by doing a 1724 * '::print zio' on the underlying zio_t. The columns we display are: 1725 * 1726 * ADDRESS TYPE STAGE WAITER 1727 * 1728 * The 'address' column is indented by one space for each depth level as we 1729 * descend down the tree. 1730 */ 1731 1732 #define ZIO_MAXINDENT 24 1733 #define ZIO_MAXWIDTH (sizeof (uintptr_t) * 2 + ZIO_MAXINDENT) 1734 #define ZIO_WALK_SELF 0 1735 #define ZIO_WALK_CHILD 1 1736 #define ZIO_WALK_PARENT 2 1737 1738 typedef struct zio_print_args { 1739 int zpa_current_depth; 1740 int zpa_min_depth; 1741 int zpa_max_depth; 1742 int zpa_type; 1743 uint_t zpa_flags; 1744 } zio_print_args_t; 1745 1746 static int zio_child_cb(uintptr_t addr, const void *unknown, void *arg); 1747 1748 static int 1749 zio_print_cb(uintptr_t addr, const void *data, void *priv) 1750 { 1751 const zio_t *zio = data; 1752 zio_print_args_t *zpa = priv; 1753 mdb_ctf_id_t type_enum, stage_enum; 1754 int indent = zpa->zpa_current_depth; 1755 const char *type, *stage; 1756 uintptr_t laddr; 1757 1758 if (indent > ZIO_MAXINDENT) 1759 indent = ZIO_MAXINDENT; 1760 1761 if (mdb_ctf_lookup_by_name("enum zio_type", &type_enum) == -1 || 1762 mdb_ctf_lookup_by_name("enum zio_stage", &stage_enum) == -1) { 1763 mdb_warn("failed to lookup zio enums"); 1764 return (WALK_ERR); 1765 } 1766 1767 if ((type = mdb_ctf_enum_name(type_enum, zio->io_type)) != NULL) 1768 type += sizeof ("ZIO_TYPE_") - 1; 1769 else 1770 type = "?"; 1771 1772 if ((stage = mdb_ctf_enum_name(stage_enum, zio->io_stage)) != NULL) 1773 stage += sizeof ("ZIO_STAGE_") - 1; 1774 else 1775 stage = "?"; 1776 1777 if (zpa->zpa_current_depth >= zpa->zpa_min_depth) { 1778 if (zpa->zpa_flags & DCMD_PIPE_OUT) { 1779 mdb_printf("%?p\n", addr); 1780 } else { 1781 mdb_printf("%*s%-*p %-5s %-16s ", indent, "", 1782 ZIO_MAXWIDTH - indent, addr, type, stage); 1783 if (zio->io_waiter) 1784 mdb_printf("%?p\n", zio->io_waiter); 1785 else 1786 mdb_printf("-\n"); 1787 } 1788 } 1789 1790 if (zpa->zpa_current_depth >= zpa->zpa_max_depth) 1791 return (WALK_NEXT); 1792 1793 if (zpa->zpa_type == ZIO_WALK_PARENT) 1794 laddr = addr + OFFSETOF(zio_t, io_parent_list); 1795 else 1796 laddr = addr + OFFSETOF(zio_t, io_child_list); 1797 1798 zpa->zpa_current_depth++; 1799 if (mdb_pwalk("list", zio_child_cb, zpa, laddr) != 0) { 1800 mdb_warn("failed to walk zio_t children at %p\n", laddr); 1801 return (WALK_ERR); 1802 } 1803 zpa->zpa_current_depth--; 1804 1805 return (WALK_NEXT); 1806 } 1807 1808 /* ARGSUSED */ 1809 static int 1810 zio_child_cb(uintptr_t addr, const void *unknown, void *arg) 1811 { 1812 zio_link_t zl; 1813 zio_t zio; 1814 uintptr_t ziop; 1815 zio_print_args_t *zpa = arg; 1816 1817 if (mdb_vread(&zl, sizeof (zl), addr) == -1) { 1818 mdb_warn("failed to read zio_link_t at %p", addr); 1819 return (WALK_ERR); 1820 } 1821 1822 if (zpa->zpa_type == ZIO_WALK_PARENT) 1823 ziop = (uintptr_t)zl.zl_parent; 1824 else 1825 ziop = (uintptr_t)zl.zl_child; 1826 1827 if (mdb_vread(&zio, sizeof (zio_t), ziop) == -1) { 1828 mdb_warn("failed to read zio_t at %p", ziop); 1829 return (WALK_ERR); 1830 } 1831 1832 return (zio_print_cb(ziop, &zio, arg)); 1833 } 1834 1835 /* ARGSUSED */ 1836 static int 1837 zio_print(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv) 1838 { 1839 zio_t zio; 1840 zio_print_args_t zpa = { 0 }; 1841 1842 if (!(flags & DCMD_ADDRSPEC)) 1843 return (DCMD_USAGE); 1844 1845 if (mdb_getopts(argc, argv, 1846 'r', MDB_OPT_SETBITS, INT_MAX, &zpa.zpa_max_depth, 1847 'c', MDB_OPT_SETBITS, ZIO_WALK_CHILD, &zpa.zpa_type, 1848 'p', MDB_OPT_SETBITS, ZIO_WALK_PARENT, &zpa.zpa_type, 1849 NULL) != argc) 1850 return (DCMD_USAGE); 1851 1852 zpa.zpa_flags = flags; 1853 if (zpa.zpa_max_depth != 0) { 1854 if (zpa.zpa_type == ZIO_WALK_SELF) 1855 zpa.zpa_type = ZIO_WALK_CHILD; 1856 } else if (zpa.zpa_type != ZIO_WALK_SELF) { 1857 zpa.zpa_min_depth = 1; 1858 zpa.zpa_max_depth = 1; 1859 } 1860 1861 if (mdb_vread(&zio, sizeof (zio_t), addr) == -1) { 1862 mdb_warn("failed to read zio_t at %p", addr); 1863 return (DCMD_ERR); 1864 } 1865 1866 if (!(flags & DCMD_PIPE_OUT) && DCMD_HDRSPEC(flags)) 1867 mdb_printf("%<u>%-*s %-5s %-16s %-?s%</u>\n", ZIO_MAXWIDTH, 1868 "ADDRESS", "TYPE", "STAGE", "WAITER"); 1869 1870 if (zio_print_cb(addr, &zio, &zpa) != WALK_NEXT) 1871 return (DCMD_ERR); 1872 1873 return (DCMD_OK); 1874 } 1875 1876 /* 1877 * [addr]::zio_state 1878 * 1879 * Print a summary of all zio_t structures on the system, or for a particular 1880 * pool. This is equivalent to '::walk zio_root | ::zio'. 1881 */ 1882 /*ARGSUSED*/ 1883 static int 1884 zio_state(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv) 1885 { 1886 /* 1887 * MDB will remember the last address of the pipeline, so if we don't 1888 * zero this we'll end up trying to walk zio structures for a 1889 * non-existent spa_t. 1890 */ 1891 if (!(flags & DCMD_ADDRSPEC)) 1892 addr = 0; 1893 1894 return (mdb_pwalk_dcmd("zio_root", "zio", argc, argv, addr)); 1895 } 1896 1897 typedef struct txg_list_walk_data { 1898 uintptr_t lw_head[TXG_SIZE]; 1899 int lw_txgoff; 1900 int lw_maxoff; 1901 size_t lw_offset; 1902 void *lw_obj; 1903 } txg_list_walk_data_t; 1904 1905 static int 1906 txg_list_walk_init_common(mdb_walk_state_t *wsp, int txg, int maxoff) 1907 { 1908 txg_list_walk_data_t *lwd; 1909 txg_list_t list; 1910 int i; 1911 1912 lwd = mdb_alloc(sizeof (txg_list_walk_data_t), UM_SLEEP | UM_GC); 1913 if (mdb_vread(&list, sizeof (txg_list_t), wsp->walk_addr) == -1) { 1914 mdb_warn("failed to read txg_list_t at %#lx", wsp->walk_addr); 1915 return (WALK_ERR); 1916 } 1917 1918 for (i = 0; i < TXG_SIZE; i++) 1919 lwd->lw_head[i] = (uintptr_t)list.tl_head[i]; 1920 lwd->lw_offset = list.tl_offset; 1921 lwd->lw_obj = mdb_alloc(lwd->lw_offset + sizeof (txg_node_t), 1922 UM_SLEEP | UM_GC); 1923 lwd->lw_txgoff = txg; 1924 lwd->lw_maxoff = maxoff; 1925 1926 wsp->walk_addr = lwd->lw_head[lwd->lw_txgoff]; 1927 wsp->walk_data = lwd; 1928 1929 return (WALK_NEXT); 1930 } 1931 1932 static int 1933 txg_list_walk_init(mdb_walk_state_t *wsp) 1934 { 1935 return (txg_list_walk_init_common(wsp, 0, TXG_SIZE-1)); 1936 } 1937 1938 static int 1939 txg_list0_walk_init(mdb_walk_state_t *wsp) 1940 { 1941 return (txg_list_walk_init_common(wsp, 0, 0)); 1942 } 1943 1944 static int 1945 txg_list1_walk_init(mdb_walk_state_t *wsp) 1946 { 1947 return (txg_list_walk_init_common(wsp, 1, 1)); 1948 } 1949 1950 static int 1951 txg_list2_walk_init(mdb_walk_state_t *wsp) 1952 { 1953 return (txg_list_walk_init_common(wsp, 2, 2)); 1954 } 1955 1956 static int 1957 txg_list3_walk_init(mdb_walk_state_t *wsp) 1958 { 1959 return (txg_list_walk_init_common(wsp, 3, 3)); 1960 } 1961 1962 static int 1963 txg_list_walk_step(mdb_walk_state_t *wsp) 1964 { 1965 txg_list_walk_data_t *lwd = wsp->walk_data; 1966 uintptr_t addr; 1967 txg_node_t *node; 1968 int status; 1969 1970 while (wsp->walk_addr == NULL && lwd->lw_txgoff < lwd->lw_maxoff) { 1971 lwd->lw_txgoff++; 1972 wsp->walk_addr = lwd->lw_head[lwd->lw_txgoff]; 1973 } 1974 1975 if (wsp->walk_addr == NULL) 1976 return (WALK_DONE); 1977 1978 addr = wsp->walk_addr - lwd->lw_offset; 1979 1980 if (mdb_vread(lwd->lw_obj, 1981 lwd->lw_offset + sizeof (txg_node_t), addr) == -1) { 1982 mdb_warn("failed to read list element at %#lx", addr); 1983 return (WALK_ERR); 1984 } 1985 1986 status = wsp->walk_callback(addr, lwd->lw_obj, wsp->walk_cbdata); 1987 node = (txg_node_t *)((uintptr_t)lwd->lw_obj + lwd->lw_offset); 1988 wsp->walk_addr = (uintptr_t)node->tn_next[lwd->lw_txgoff]; 1989 1990 return (status); 1991 } 1992 1993 /* 1994 * ::walk spa 1995 * 1996 * Walk all named spa_t structures in the namespace. This is nothing more than 1997 * a layered avl walk. 1998 */ 1999 static int 2000 spa_walk_init(mdb_walk_state_t *wsp) 2001 { 2002 GElf_Sym sym; 2003 2004 if (wsp->walk_addr != NULL) { 2005 mdb_warn("spa walk only supports global walks\n"); 2006 return (WALK_ERR); 2007 } 2008 2009 if (mdb_lookup_by_obj(ZFS_OBJ_NAME, "spa_namespace_avl", &sym) == -1) { 2010 mdb_warn("failed to find symbol 'spa_namespace_avl'"); 2011 return (WALK_ERR); 2012 } 2013 2014 wsp->walk_addr = (uintptr_t)sym.st_value; 2015 2016 if (mdb_layered_walk("avl", wsp) == -1) { 2017 mdb_warn("failed to walk 'avl'\n"); 2018 return (WALK_ERR); 2019 } 2020 2021 return (WALK_NEXT); 2022 } 2023 2024 static int 2025 spa_walk_step(mdb_walk_state_t *wsp) 2026 { 2027 spa_t spa; 2028 2029 if (mdb_vread(&spa, sizeof (spa), wsp->walk_addr) == -1) { 2030 mdb_warn("failed to read spa_t at %p", wsp->walk_addr); 2031 return (WALK_ERR); 2032 } 2033 2034 return (wsp->walk_callback(wsp->walk_addr, &spa, wsp->walk_cbdata)); 2035 } 2036 2037 /* 2038 * [addr]::walk zio 2039 * 2040 * Walk all active zio_t structures on the system. This is simply a layered 2041 * walk on top of ::walk zio_cache, with the optional ability to limit the 2042 * structures to a particular pool. 2043 */ 2044 static int 2045 zio_walk_init(mdb_walk_state_t *wsp) 2046 { 2047 wsp->walk_data = (void *)wsp->walk_addr; 2048 2049 if (mdb_layered_walk("zio_cache", wsp) == -1) { 2050 mdb_warn("failed to walk 'zio_cache'\n"); 2051 return (WALK_ERR); 2052 } 2053 2054 return (WALK_NEXT); 2055 } 2056 2057 static int 2058 zio_walk_step(mdb_walk_state_t *wsp) 2059 { 2060 zio_t zio; 2061 2062 if (mdb_vread(&zio, sizeof (zio), wsp->walk_addr) == -1) { 2063 mdb_warn("failed to read zio_t at %p", wsp->walk_addr); 2064 return (WALK_ERR); 2065 } 2066 2067 if (wsp->walk_data != NULL && wsp->walk_data != zio.io_spa) 2068 return (WALK_NEXT); 2069 2070 return (wsp->walk_callback(wsp->walk_addr, &zio, wsp->walk_cbdata)); 2071 } 2072 2073 /* 2074 * [addr]::walk zio_root 2075 * 2076 * Walk only root zio_t structures, optionally for a particular spa_t. 2077 */ 2078 static int 2079 zio_walk_root_step(mdb_walk_state_t *wsp) 2080 { 2081 zio_t zio; 2082 2083 if (mdb_vread(&zio, sizeof (zio), wsp->walk_addr) == -1) { 2084 mdb_warn("failed to read zio_t at %p", wsp->walk_addr); 2085 return (WALK_ERR); 2086 } 2087 2088 if (wsp->walk_data != NULL && wsp->walk_data != zio.io_spa) 2089 return (WALK_NEXT); 2090 2091 /* If the parent list is not empty, ignore */ 2092 if (zio.io_parent_list.list_head.list_next != 2093 &((zio_t *)wsp->walk_addr)->io_parent_list.list_head) 2094 return (WALK_NEXT); 2095 2096 return (wsp->walk_callback(wsp->walk_addr, &zio, wsp->walk_cbdata)); 2097 } 2098 2099 #define NICENUM_BUFLEN 6 2100 2101 static int 2102 snprintfrac(char *buf, int len, 2103 uint64_t numerator, uint64_t denom, int frac_digits) 2104 { 2105 int mul = 1; 2106 int whole, frac, i; 2107 2108 for (i = frac_digits; i; i--) 2109 mul *= 10; 2110 whole = numerator / denom; 2111 frac = mul * numerator / denom - mul * whole; 2112 return (mdb_snprintf(buf, len, "%u.%0*u", whole, frac_digits, frac)); 2113 } 2114 2115 static void 2116 mdb_nicenum(uint64_t num, char *buf) 2117 { 2118 uint64_t n = num; 2119 int index = 0; 2120 char *u; 2121 2122 while (n >= 1024) { 2123 n = (n + (1024 / 2)) / 1024; /* Round up or down */ 2124 index++; 2125 } 2126 2127 u = &" \0K\0M\0G\0T\0P\0E\0"[index*2]; 2128 2129 if (index == 0) { 2130 (void) mdb_snprintf(buf, NICENUM_BUFLEN, "%llu", 2131 (u_longlong_t)n); 2132 } else if (n < 10 && (num & (num - 1)) != 0) { 2133 (void) snprintfrac(buf, NICENUM_BUFLEN, 2134 num, 1ULL << 10 * index, 2); 2135 strcat(buf, u); 2136 } else if (n < 100 && (num & (num - 1)) != 0) { 2137 (void) snprintfrac(buf, NICENUM_BUFLEN, 2138 num, 1ULL << 10 * index, 1); 2139 strcat(buf, u); 2140 } else { 2141 (void) mdb_snprintf(buf, NICENUM_BUFLEN, "%llu%s", 2142 (u_longlong_t)n, u); 2143 } 2144 } 2145 2146 /* 2147 * ::zfs_blkstats 2148 * 2149 * -v print verbose per-level information 2150 * 2151 */ 2152 static int 2153 zfs_blkstats(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv) 2154 { 2155 boolean_t verbose = B_FALSE; 2156 zfs_all_blkstats_t stats; 2157 dmu_object_type_t t; 2158 zfs_blkstat_t *tzb; 2159 uint64_t ditto; 2160 dmu_object_type_info_t dmu_ot[DMU_OT_NUMTYPES + 10]; 2161 /* +10 in case it grew */ 2162 2163 if (mdb_readvar(&dmu_ot, "dmu_ot") == -1) { 2164 mdb_warn("failed to read 'dmu_ot'"); 2165 return (DCMD_ERR); 2166 } 2167 2168 if (mdb_getopts(argc, argv, 2169 'v', MDB_OPT_SETBITS, TRUE, &verbose, 2170 NULL) != argc) 2171 return (DCMD_USAGE); 2172 2173 if (!(flags & DCMD_ADDRSPEC)) 2174 return (DCMD_USAGE); 2175 2176 if (GETMEMB(addr, struct spa, spa_dsl_pool, addr) || 2177 GETMEMB(addr, struct dsl_pool, dp_blkstats, addr) || 2178 mdb_vread(&stats, sizeof (zfs_all_blkstats_t), addr) == -1) { 2179 mdb_warn("failed to read data at %p;", addr); 2180 mdb_printf("maybe no stats? run \"zpool scrub\" first."); 2181 return (DCMD_ERR); 2182 } 2183 2184 tzb = &stats.zab_type[DN_MAX_LEVELS][DMU_OT_NUMTYPES]; 2185 if (tzb->zb_gangs != 0) { 2186 mdb_printf("Ganged blocks: %llu\n", 2187 (longlong_t)tzb->zb_gangs); 2188 } 2189 2190 ditto = tzb->zb_ditto_2_of_2_samevdev + tzb->zb_ditto_2_of_3_samevdev + 2191 tzb->zb_ditto_3_of_3_samevdev; 2192 if (ditto != 0) { 2193 mdb_printf("Dittoed blocks on same vdev: %llu\n", 2194 (longlong_t)ditto); 2195 } 2196 2197 mdb_printf("\nBlocks\tLSIZE\tPSIZE\tASIZE" 2198 "\t avg\t comp\t%%Total\tType\n"); 2199 2200 for (t = 0; t <= DMU_OT_NUMTYPES; t++) { 2201 char csize[NICENUM_BUFLEN], lsize[NICENUM_BUFLEN]; 2202 char psize[NICENUM_BUFLEN], asize[NICENUM_BUFLEN]; 2203 char avg[NICENUM_BUFLEN]; 2204 char comp[NICENUM_BUFLEN], pct[NICENUM_BUFLEN]; 2205 char typename[64]; 2206 int l; 2207 2208 2209 if (t == DMU_OT_DEFERRED) 2210 strcpy(typename, "deferred free"); 2211 else if (t == DMU_OT_TOTAL) 2212 strcpy(typename, "Total"); 2213 else if (mdb_readstr(typename, sizeof (typename), 2214 (uintptr_t)dmu_ot[t].ot_name) == -1) { 2215 mdb_warn("failed to read type name"); 2216 return (DCMD_ERR); 2217 } 2218 2219 if (stats.zab_type[DN_MAX_LEVELS][t].zb_asize == 0) 2220 continue; 2221 2222 for (l = -1; l < DN_MAX_LEVELS; l++) { 2223 int level = (l == -1 ? DN_MAX_LEVELS : l); 2224 zfs_blkstat_t *zb = &stats.zab_type[level][t]; 2225 2226 if (zb->zb_asize == 0) 2227 continue; 2228 2229 /* 2230 * Don't print each level unless requested. 2231 */ 2232 if (!verbose && level != DN_MAX_LEVELS) 2233 continue; 2234 2235 /* 2236 * If all the space is level 0, don't print the 2237 * level 0 separately. 2238 */ 2239 if (level == 0 && zb->zb_asize == 2240 stats.zab_type[DN_MAX_LEVELS][t].zb_asize) 2241 continue; 2242 2243 mdb_nicenum(zb->zb_count, csize); 2244 mdb_nicenum(zb->zb_lsize, lsize); 2245 mdb_nicenum(zb->zb_psize, psize); 2246 mdb_nicenum(zb->zb_asize, asize); 2247 mdb_nicenum(zb->zb_asize / zb->zb_count, avg); 2248 (void) snprintfrac(comp, NICENUM_BUFLEN, 2249 zb->zb_lsize, zb->zb_psize, 2); 2250 (void) snprintfrac(pct, NICENUM_BUFLEN, 2251 100 * zb->zb_asize, tzb->zb_asize, 2); 2252 2253 mdb_printf("%6s\t%5s\t%5s\t%5s\t%5s" 2254 "\t%5s\t%6s\t", 2255 csize, lsize, psize, asize, avg, comp, pct); 2256 2257 if (level == DN_MAX_LEVELS) 2258 mdb_printf("%s\n", typename); 2259 else 2260 mdb_printf(" L%d %s\n", 2261 level, typename); 2262 } 2263 } 2264 2265 return (DCMD_OK); 2266 } 2267 2268 /* ARGSUSED */ 2269 static int 2270 reference_cb(uintptr_t addr, const void *ignored, void *arg) 2271 { 2272 static int gotid; 2273 static mdb_ctf_id_t ref_id; 2274 uintptr_t ref_holder; 2275 uintptr_t ref_removed; 2276 uint64_t ref_number; 2277 boolean_t holder_is_str = B_FALSE; 2278 char holder_str[128]; 2279 boolean_t removed = (boolean_t)arg; 2280 2281 if (!gotid) { 2282 if (mdb_ctf_lookup_by_name("struct reference", &ref_id) == -1) { 2283 mdb_warn("couldn't find struct reference"); 2284 return (WALK_ERR); 2285 } 2286 gotid = TRUE; 2287 } 2288 2289 if (GETMEMBID(addr, &ref_id, ref_holder, ref_holder) || 2290 GETMEMBID(addr, &ref_id, ref_removed, ref_removed) || 2291 GETMEMBID(addr, &ref_id, ref_number, ref_number)) 2292 return (WALK_ERR); 2293 2294 if (mdb_readstr(holder_str, sizeof (holder_str), ref_holder) != -1) 2295 holder_is_str = strisprint(holder_str); 2296 2297 if (removed) 2298 mdb_printf("removed "); 2299 mdb_printf("reference "); 2300 if (ref_number != 1) 2301 mdb_printf("with count=%llu ", ref_number); 2302 mdb_printf("with tag %p", (void*)ref_holder); 2303 if (holder_is_str) 2304 mdb_printf(" \"%s\"", holder_str); 2305 mdb_printf(", held at:\n"); 2306 2307 (void) mdb_call_dcmd("whatis", addr, DCMD_ADDRSPEC, 0, NULL); 2308 2309 if (removed) { 2310 mdb_printf("removed at:\n"); 2311 (void) mdb_call_dcmd("whatis", ref_removed, 2312 DCMD_ADDRSPEC, 0, NULL); 2313 } 2314 2315 mdb_printf("\n"); 2316 2317 return (WALK_NEXT); 2318 } 2319 2320 /* ARGSUSED */ 2321 static int 2322 refcount(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv) 2323 { 2324 uint64_t rc_count, rc_removed_count; 2325 uintptr_t rc_list, rc_removed; 2326 static int gotid; 2327 static mdb_ctf_id_t rc_id; 2328 ulong_t off; 2329 2330 if (!(flags & DCMD_ADDRSPEC)) 2331 return (DCMD_USAGE); 2332 2333 if (!gotid) { 2334 if (mdb_ctf_lookup_by_name("struct refcount", &rc_id) == -1) { 2335 mdb_warn("couldn't find struct refcount"); 2336 return (DCMD_ERR); 2337 } 2338 gotid = TRUE; 2339 } 2340 2341 if (GETMEMBID(addr, &rc_id, rc_count, rc_count) || 2342 GETMEMBID(addr, &rc_id, rc_removed_count, rc_removed_count)) 2343 return (DCMD_ERR); 2344 2345 mdb_printf("refcount_t at %p has %llu current holds, " 2346 "%llu recently released holds\n", 2347 addr, (longlong_t)rc_count, (longlong_t)rc_removed_count); 2348 2349 if (rc_count > 0) 2350 mdb_printf("current holds:\n"); 2351 if (mdb_ctf_offsetof(rc_id, "rc_list", &off) == -1) 2352 return (DCMD_ERR); 2353 rc_list = addr + off/NBBY; 2354 mdb_pwalk("list", reference_cb, (void*)B_FALSE, rc_list); 2355 2356 if (rc_removed_count > 0) 2357 mdb_printf("released holds:\n"); 2358 if (mdb_ctf_offsetof(rc_id, "rc_removed", &off) == -1) 2359 return (DCMD_ERR); 2360 rc_removed = addr + off/NBBY; 2361 mdb_pwalk("list", reference_cb, (void*)B_TRUE, rc_removed); 2362 2363 return (DCMD_OK); 2364 } 2365 2366 /* ARGSUSED */ 2367 static int 2368 sa_attr_table(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv) 2369 { 2370 sa_attr_table_t *table; 2371 sa_os_t sa_os; 2372 char *name; 2373 int i; 2374 2375 if (mdb_vread(&sa_os, sizeof (sa_os_t), addr) == -1) { 2376 mdb_warn("failed to read sa_os at %p", addr); 2377 return (DCMD_ERR); 2378 } 2379 2380 table = mdb_alloc(sizeof (sa_attr_table_t) * sa_os.sa_num_attrs, 2381 UM_SLEEP | UM_GC); 2382 name = mdb_alloc(MAXPATHLEN, UM_SLEEP | UM_GC); 2383 2384 if (mdb_vread(table, sizeof (sa_attr_table_t) * sa_os.sa_num_attrs, 2385 (uintptr_t)sa_os.sa_attr_table) == -1) { 2386 mdb_warn("failed to read sa_os at %p", addr); 2387 return (DCMD_ERR); 2388 } 2389 2390 mdb_printf("%<u>%-10s %-10s %-10s %-10s %s%</u>\n", 2391 "ATTR ID", "REGISTERED", "LENGTH", "BSWAP", "NAME"); 2392 for (i = 0; i != sa_os.sa_num_attrs; i++) { 2393 mdb_readstr(name, MAXPATHLEN, (uintptr_t)table[i].sa_name); 2394 mdb_printf("%5x %8x %8x %8x %-s\n", 2395 (int)table[i].sa_attr, (int)table[i].sa_registered, 2396 (int)table[i].sa_length, table[i].sa_byteswap, name); 2397 } 2398 2399 return (DCMD_OK); 2400 } 2401 2402 static int 2403 sa_get_off_table(uintptr_t addr, uint32_t **off_tab, int attr_count) 2404 { 2405 uintptr_t idx_table; 2406 2407 if (GETMEMB(addr, struct sa_idx_tab, sa_idx_tab, idx_table)) { 2408 mdb_printf("can't find offset table in sa_idx_tab\n"); 2409 return (-1); 2410 } 2411 2412 *off_tab = mdb_alloc(attr_count * sizeof (uint32_t), 2413 UM_SLEEP | UM_GC); 2414 2415 if (mdb_vread(*off_tab, 2416 attr_count * sizeof (uint32_t), idx_table) == -1) { 2417 mdb_warn("failed to attribute offset table %p", idx_table); 2418 return (-1); 2419 } 2420 2421 return (DCMD_OK); 2422 } 2423 2424 /*ARGSUSED*/ 2425 static int 2426 sa_attr_print(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv) 2427 { 2428 uint32_t *offset_tab; 2429 int attr_count; 2430 uint64_t attr_id; 2431 uintptr_t attr_addr; 2432 uintptr_t bonus_tab, spill_tab; 2433 uintptr_t db_bonus, db_spill; 2434 uintptr_t os, os_sa; 2435 uintptr_t db_data; 2436 2437 if (argc != 1) 2438 return (DCMD_USAGE); 2439 2440 if (argv[0].a_type == MDB_TYPE_STRING) 2441 attr_id = mdb_strtoull(argv[0].a_un.a_str); 2442 else 2443 return (DCMD_USAGE); 2444 2445 if (GETMEMB(addr, struct sa_handle, sa_bonus_tab, bonus_tab) || 2446 GETMEMB(addr, struct sa_handle, sa_spill_tab, spill_tab) || 2447 GETMEMB(addr, struct sa_handle, sa_os, os) || 2448 GETMEMB(addr, struct sa_handle, sa_bonus, db_bonus) || 2449 GETMEMB(addr, struct sa_handle, sa_spill, db_spill)) { 2450 mdb_printf("Can't find necessary information in sa_handle " 2451 "in sa_handle\n"); 2452 return (DCMD_ERR); 2453 } 2454 2455 if (GETMEMB(os, struct objset, os_sa, os_sa)) { 2456 mdb_printf("Can't find os_sa in objset\n"); 2457 return (DCMD_ERR); 2458 } 2459 2460 if (GETMEMB(os_sa, struct sa_os, sa_num_attrs, attr_count)) { 2461 mdb_printf("Can't find sa_num_attrs\n"); 2462 return (DCMD_ERR); 2463 } 2464 2465 if (attr_id > attr_count) { 2466 mdb_printf("attribute id number is out of range\n"); 2467 return (DCMD_ERR); 2468 } 2469 2470 if (bonus_tab) { 2471 if (sa_get_off_table(bonus_tab, &offset_tab, 2472 attr_count) == -1) { 2473 return (DCMD_ERR); 2474 } 2475 2476 if (GETMEMB(db_bonus, struct dmu_buf, db_data, db_data)) { 2477 mdb_printf("can't find db_data in bonus dbuf\n"); 2478 return (DCMD_ERR); 2479 } 2480 } 2481 2482 if (bonus_tab && !TOC_ATTR_PRESENT(offset_tab[attr_id]) && 2483 spill_tab == NULL) { 2484 mdb_printf("Attribute does not exist\n"); 2485 return (DCMD_ERR); 2486 } else if (!TOC_ATTR_PRESENT(offset_tab[attr_id]) && spill_tab) { 2487 if (sa_get_off_table(spill_tab, &offset_tab, 2488 attr_count) == -1) { 2489 return (DCMD_ERR); 2490 } 2491 if (GETMEMB(db_spill, struct dmu_buf, db_data, db_data)) { 2492 mdb_printf("can't find db_data in spill dbuf\n"); 2493 return (DCMD_ERR); 2494 } 2495 if (!TOC_ATTR_PRESENT(offset_tab[attr_id])) { 2496 mdb_printf("Attribute does not exist\n"); 2497 return (DCMD_ERR); 2498 } 2499 } 2500 attr_addr = db_data + TOC_OFF(offset_tab[attr_id]); 2501 mdb_printf("%p\n", attr_addr); 2502 return (DCMD_OK); 2503 } 2504 2505 /* ARGSUSED */ 2506 static int 2507 zfs_ace_print_common(uintptr_t addr, uint_t flags, 2508 uint64_t id, uint32_t access_mask, uint16_t ace_flags, 2509 uint16_t ace_type, int verbose) 2510 { 2511 if (DCMD_HDRSPEC(flags) && !verbose) 2512 mdb_printf("%<u>%-?s %-8s %-8s %-8s %s%</u>\n", 2513 "ADDR", "FLAGS", "MASK", "TYPE", "ID"); 2514 2515 if (!verbose) { 2516 mdb_printf("%0?p %-8x %-8x %-8x %-llx\n", addr, 2517 ace_flags, access_mask, ace_type, id); 2518 return (DCMD_OK); 2519 } 2520 2521 switch (ace_flags & ACE_TYPE_FLAGS) { 2522 case ACE_OWNER: 2523 mdb_printf("owner@:"); 2524 break; 2525 case (ACE_IDENTIFIER_GROUP | ACE_GROUP): 2526 mdb_printf("group@:"); 2527 break; 2528 case ACE_EVERYONE: 2529 mdb_printf("everyone@:"); 2530 break; 2531 case ACE_IDENTIFIER_GROUP: 2532 mdb_printf("group:%llx:", (u_longlong_t)id); 2533 break; 2534 case 0: /* User entry */ 2535 mdb_printf("user:%llx:", (u_longlong_t)id); 2536 break; 2537 } 2538 2539 /* print out permission mask */ 2540 if (access_mask & ACE_READ_DATA) 2541 mdb_printf("r"); 2542 else 2543 mdb_printf("-"); 2544 if (access_mask & ACE_WRITE_DATA) 2545 mdb_printf("w"); 2546 else 2547 mdb_printf("-"); 2548 if (access_mask & ACE_EXECUTE) 2549 mdb_printf("x"); 2550 else 2551 mdb_printf("-"); 2552 if (access_mask & ACE_APPEND_DATA) 2553 mdb_printf("p"); 2554 else 2555 mdb_printf("-"); 2556 if (access_mask & ACE_DELETE) 2557 mdb_printf("d"); 2558 else 2559 mdb_printf("-"); 2560 if (access_mask & ACE_DELETE_CHILD) 2561 mdb_printf("D"); 2562 else 2563 mdb_printf("-"); 2564 if (access_mask & ACE_READ_ATTRIBUTES) 2565 mdb_printf("a"); 2566 else 2567 mdb_printf("-"); 2568 if (access_mask & ACE_WRITE_ATTRIBUTES) 2569 mdb_printf("A"); 2570 else 2571 mdb_printf("-"); 2572 if (access_mask & ACE_READ_NAMED_ATTRS) 2573 mdb_printf("R"); 2574 else 2575 mdb_printf("-"); 2576 if (access_mask & ACE_WRITE_NAMED_ATTRS) 2577 mdb_printf("W"); 2578 else 2579 mdb_printf("-"); 2580 if (access_mask & ACE_READ_ACL) 2581 mdb_printf("c"); 2582 else 2583 mdb_printf("-"); 2584 if (access_mask & ACE_WRITE_ACL) 2585 mdb_printf("C"); 2586 else 2587 mdb_printf("-"); 2588 if (access_mask & ACE_WRITE_OWNER) 2589 mdb_printf("o"); 2590 else 2591 mdb_printf("-"); 2592 if (access_mask & ACE_SYNCHRONIZE) 2593 mdb_printf("s"); 2594 else 2595 mdb_printf("-"); 2596 2597 mdb_printf(":"); 2598 2599 /* Print out inheritance flags */ 2600 if (ace_flags & ACE_FILE_INHERIT_ACE) 2601 mdb_printf("f"); 2602 else 2603 mdb_printf("-"); 2604 if (ace_flags & ACE_DIRECTORY_INHERIT_ACE) 2605 mdb_printf("d"); 2606 else 2607 mdb_printf("-"); 2608 if (ace_flags & ACE_INHERIT_ONLY_ACE) 2609 mdb_printf("i"); 2610 else 2611 mdb_printf("-"); 2612 if (ace_flags & ACE_NO_PROPAGATE_INHERIT_ACE) 2613 mdb_printf("n"); 2614 else 2615 mdb_printf("-"); 2616 if (ace_flags & ACE_SUCCESSFUL_ACCESS_ACE_FLAG) 2617 mdb_printf("S"); 2618 else 2619 mdb_printf("-"); 2620 if (ace_flags & ACE_FAILED_ACCESS_ACE_FLAG) 2621 mdb_printf("F"); 2622 else 2623 mdb_printf("-"); 2624 if (ace_flags & ACE_INHERITED_ACE) 2625 mdb_printf("I"); 2626 else 2627 mdb_printf("-"); 2628 2629 switch (ace_type) { 2630 case ACE_ACCESS_ALLOWED_ACE_TYPE: 2631 mdb_printf(":allow\n"); 2632 break; 2633 case ACE_ACCESS_DENIED_ACE_TYPE: 2634 mdb_printf(":deny\n"); 2635 break; 2636 case ACE_SYSTEM_AUDIT_ACE_TYPE: 2637 mdb_printf(":audit\n"); 2638 break; 2639 case ACE_SYSTEM_ALARM_ACE_TYPE: 2640 mdb_printf(":alarm\n"); 2641 break; 2642 default: 2643 mdb_printf(":?\n"); 2644 } 2645 return (DCMD_OK); 2646 } 2647 2648 /* ARGSUSED */ 2649 static int 2650 zfs_ace_print(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv) 2651 { 2652 zfs_ace_t zace; 2653 int verbose = FALSE; 2654 uint64_t id; 2655 2656 if (!(flags & DCMD_ADDRSPEC)) 2657 return (DCMD_USAGE); 2658 2659 if (mdb_getopts(argc, argv, 2660 'v', MDB_OPT_SETBITS, TRUE, &verbose, TRUE, NULL) != argc) 2661 return (DCMD_USAGE); 2662 2663 if (mdb_vread(&zace, sizeof (zfs_ace_t), addr) == -1) { 2664 mdb_warn("failed to read zfs_ace_t"); 2665 return (DCMD_ERR); 2666 } 2667 2668 if ((zace.z_hdr.z_flags & ACE_TYPE_FLAGS) == 0 || 2669 (zace.z_hdr.z_flags & ACE_TYPE_FLAGS) == ACE_IDENTIFIER_GROUP) 2670 id = zace.z_fuid; 2671 else 2672 id = -1; 2673 2674 return (zfs_ace_print_common(addr, flags, id, zace.z_hdr.z_access_mask, 2675 zace.z_hdr.z_flags, zace.z_hdr.z_type, verbose)); 2676 } 2677 2678 /* ARGSUSED */ 2679 static int 2680 zfs_ace0_print(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv) 2681 { 2682 ace_t ace; 2683 uint64_t id; 2684 int verbose = FALSE; 2685 2686 if (!(flags & DCMD_ADDRSPEC)) 2687 return (DCMD_USAGE); 2688 2689 if (mdb_getopts(argc, argv, 2690 'v', MDB_OPT_SETBITS, TRUE, &verbose, TRUE, NULL) != argc) 2691 return (DCMD_USAGE); 2692 2693 if (mdb_vread(&ace, sizeof (ace_t), addr) == -1) { 2694 mdb_warn("failed to read ace_t"); 2695 return (DCMD_ERR); 2696 } 2697 2698 if ((ace.a_flags & ACE_TYPE_FLAGS) == 0 || 2699 (ace.a_flags & ACE_TYPE_FLAGS) == ACE_IDENTIFIER_GROUP) 2700 id = ace.a_who; 2701 else 2702 id = -1; 2703 2704 return (zfs_ace_print_common(addr, flags, id, ace.a_access_mask, 2705 ace.a_flags, ace.a_type, verbose)); 2706 } 2707 2708 typedef struct acl_dump_args { 2709 int a_argc; 2710 const mdb_arg_t *a_argv; 2711 uint16_t a_version; 2712 int a_flags; 2713 } acl_dump_args_t; 2714 2715 /* ARGSUSED */ 2716 static int 2717 acl_aces_cb(uintptr_t addr, const void *unknown, void *arg) 2718 { 2719 acl_dump_args_t *acl_args = (acl_dump_args_t *)arg; 2720 2721 if (acl_args->a_version == 1) { 2722 if (mdb_call_dcmd("zfs_ace", addr, 2723 DCMD_ADDRSPEC|acl_args->a_flags, acl_args->a_argc, 2724 acl_args->a_argv) != DCMD_OK) { 2725 return (WALK_ERR); 2726 } 2727 } else { 2728 if (mdb_call_dcmd("zfs_ace0", addr, 2729 DCMD_ADDRSPEC|acl_args->a_flags, acl_args->a_argc, 2730 acl_args->a_argv) != DCMD_OK) { 2731 return (WALK_ERR); 2732 } 2733 } 2734 acl_args->a_flags = DCMD_LOOP; 2735 return (WALK_NEXT); 2736 } 2737 2738 /* ARGSUSED */ 2739 static int 2740 acl_cb(uintptr_t addr, const void *unknown, void *arg) 2741 { 2742 acl_dump_args_t *acl_args = (acl_dump_args_t *)arg; 2743 2744 if (acl_args->a_version == 1) { 2745 if (mdb_pwalk("zfs_acl_node_aces", acl_aces_cb, 2746 arg, addr) != 0) { 2747 mdb_warn("can't walk ACEs"); 2748 return (DCMD_ERR); 2749 } 2750 } else { 2751 if (mdb_pwalk("zfs_acl_node_aces0", acl_aces_cb, 2752 arg, addr) != 0) { 2753 mdb_warn("can't walk ACEs"); 2754 return (DCMD_ERR); 2755 } 2756 } 2757 return (WALK_NEXT); 2758 } 2759 2760 /* ARGSUSED */ 2761 static int 2762 zfs_acl_dump(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv) 2763 { 2764 zfs_acl_t zacl; 2765 int verbose = FALSE; 2766 acl_dump_args_t acl_args; 2767 2768 if (!(flags & DCMD_ADDRSPEC)) 2769 return (DCMD_USAGE); 2770 2771 if (mdb_getopts(argc, argv, 2772 'v', MDB_OPT_SETBITS, TRUE, &verbose, TRUE, NULL) != argc) 2773 return (DCMD_USAGE); 2774 2775 if (mdb_vread(&zacl, sizeof (zfs_acl_t), addr) == -1) { 2776 mdb_warn("failed to read zfs_acl_t"); 2777 return (DCMD_ERR); 2778 } 2779 2780 acl_args.a_argc = argc; 2781 acl_args.a_argv = argv; 2782 acl_args.a_version = zacl.z_version; 2783 acl_args.a_flags = DCMD_LOOPFIRST; 2784 2785 if (mdb_pwalk("zfs_acl_node", acl_cb, &acl_args, addr) != 0) { 2786 mdb_warn("can't walk ACL"); 2787 return (DCMD_ERR); 2788 } 2789 2790 return (DCMD_OK); 2791 } 2792 2793 /* ARGSUSED */ 2794 static int 2795 zfs_acl_node_walk_init(mdb_walk_state_t *wsp) 2796 { 2797 if (wsp->walk_addr == NULL) { 2798 mdb_warn("must supply address of zfs_acl_node_t\n"); 2799 return (WALK_ERR); 2800 } 2801 2802 wsp->walk_addr += OFFSETOF(zfs_acl_t, z_acl); 2803 2804 if (mdb_layered_walk("list", wsp) == -1) { 2805 mdb_warn("failed to walk 'list'\n"); 2806 return (WALK_ERR); 2807 } 2808 2809 return (WALK_NEXT); 2810 } 2811 2812 static int 2813 zfs_acl_node_walk_step(mdb_walk_state_t *wsp) 2814 { 2815 zfs_acl_node_t aclnode; 2816 2817 if (mdb_vread(&aclnode, sizeof (zfs_acl_node_t), 2818 wsp->walk_addr) == -1) { 2819 mdb_warn("failed to read zfs_acl_node at %p", wsp->walk_addr); 2820 return (WALK_ERR); 2821 } 2822 2823 return (wsp->walk_callback(wsp->walk_addr, &aclnode, wsp->walk_cbdata)); 2824 } 2825 2826 typedef struct ace_walk_data { 2827 int ace_count; 2828 int ace_version; 2829 } ace_walk_data_t; 2830 2831 static int 2832 zfs_aces_walk_init_common(mdb_walk_state_t *wsp, int version, 2833 int ace_count, uintptr_t ace_data) 2834 { 2835 ace_walk_data_t *ace_walk_data; 2836 2837 if (wsp->walk_addr == NULL) { 2838 mdb_warn("must supply address of zfs_acl_node_t\n"); 2839 return (WALK_ERR); 2840 } 2841 2842 ace_walk_data = mdb_alloc(sizeof (ace_walk_data_t), UM_SLEEP | UM_GC); 2843 2844 ace_walk_data->ace_count = ace_count; 2845 ace_walk_data->ace_version = version; 2846 2847 wsp->walk_addr = ace_data; 2848 wsp->walk_data = ace_walk_data; 2849 2850 return (WALK_NEXT); 2851 } 2852 2853 static int 2854 zfs_acl_node_aces_walk_init_common(mdb_walk_state_t *wsp, int version) 2855 { 2856 static int gotid; 2857 static mdb_ctf_id_t acl_id; 2858 int z_ace_count; 2859 uintptr_t z_acldata; 2860 2861 if (!gotid) { 2862 if (mdb_ctf_lookup_by_name("struct zfs_acl_node", 2863 &acl_id) == -1) { 2864 mdb_warn("couldn't find struct zfs_acl_node"); 2865 return (DCMD_ERR); 2866 } 2867 gotid = TRUE; 2868 } 2869 2870 if (GETMEMBID(wsp->walk_addr, &acl_id, z_ace_count, z_ace_count)) { 2871 return (DCMD_ERR); 2872 } 2873 if (GETMEMBID(wsp->walk_addr, &acl_id, z_acldata, z_acldata)) { 2874 return (DCMD_ERR); 2875 } 2876 2877 return (zfs_aces_walk_init_common(wsp, version, 2878 z_ace_count, z_acldata)); 2879 } 2880 2881 /* ARGSUSED */ 2882 static int 2883 zfs_acl_node_aces_walk_init(mdb_walk_state_t *wsp) 2884 { 2885 return (zfs_acl_node_aces_walk_init_common(wsp, 1)); 2886 } 2887 2888 /* ARGSUSED */ 2889 static int 2890 zfs_acl_node_aces0_walk_init(mdb_walk_state_t *wsp) 2891 { 2892 return (zfs_acl_node_aces_walk_init_common(wsp, 0)); 2893 } 2894 2895 static int 2896 zfs_aces_walk_step(mdb_walk_state_t *wsp) 2897 { 2898 ace_walk_data_t *ace_data = wsp->walk_data; 2899 zfs_ace_t zace; 2900 ace_t *acep; 2901 int status; 2902 int entry_type; 2903 int allow_type; 2904 uintptr_t ptr; 2905 2906 if (ace_data->ace_count == 0) 2907 return (WALK_DONE); 2908 2909 if (mdb_vread(&zace, sizeof (zfs_ace_t), wsp->walk_addr) == -1) { 2910 mdb_warn("failed to read zfs_ace_t at %#lx", 2911 wsp->walk_addr); 2912 return (WALK_ERR); 2913 } 2914 2915 switch (ace_data->ace_version) { 2916 case 0: 2917 acep = (ace_t *)&zace; 2918 entry_type = acep->a_flags & ACE_TYPE_FLAGS; 2919 allow_type = acep->a_type; 2920 break; 2921 case 1: 2922 entry_type = zace.z_hdr.z_flags & ACE_TYPE_FLAGS; 2923 allow_type = zace.z_hdr.z_type; 2924 break; 2925 default: 2926 return (WALK_ERR); 2927 } 2928 2929 ptr = (uintptr_t)wsp->walk_addr; 2930 switch (entry_type) { 2931 case ACE_OWNER: 2932 case ACE_EVERYONE: 2933 case (ACE_IDENTIFIER_GROUP | ACE_GROUP): 2934 ptr += ace_data->ace_version == 0 ? 2935 sizeof (ace_t) : sizeof (zfs_ace_hdr_t); 2936 break; 2937 case ACE_IDENTIFIER_GROUP: 2938 default: 2939 switch (allow_type) { 2940 case ACE_ACCESS_ALLOWED_OBJECT_ACE_TYPE: 2941 case ACE_ACCESS_DENIED_OBJECT_ACE_TYPE: 2942 case ACE_SYSTEM_AUDIT_OBJECT_ACE_TYPE: 2943 case ACE_SYSTEM_ALARM_OBJECT_ACE_TYPE: 2944 ptr += ace_data->ace_version == 0 ? 2945 sizeof (ace_t) : sizeof (zfs_object_ace_t); 2946 break; 2947 default: 2948 ptr += ace_data->ace_version == 0 ? 2949 sizeof (ace_t) : sizeof (zfs_ace_t); 2950 break; 2951 } 2952 } 2953 2954 ace_data->ace_count--; 2955 status = wsp->walk_callback(wsp->walk_addr, 2956 (void *)(uintptr_t)&zace, wsp->walk_cbdata); 2957 2958 wsp->walk_addr = ptr; 2959 return (status); 2960 } 2961 2962 /* 2963 * MDB module linkage information: 2964 * 2965 * We declare a list of structures describing our dcmds, and a function 2966 * named _mdb_init to return a pointer to our module information. 2967 */ 2968 2969 static const mdb_dcmd_t dcmds[] = { 2970 { "arc", "[-bkmg]", "print ARC variables", arc_print }, 2971 { "blkptr", ":", "print blkptr_t", blkptr }, 2972 { "dbuf", ":", "print dmu_buf_impl_t", dbuf }, 2973 { "dbuf_stats", ":", "dbuf stats", dbuf_stats }, 2974 { "dbufs", 2975 "\t[-O objset_t*] [-n objset_name | \"mos\"] " 2976 "[-o object | \"mdn\"] \n" 2977 "\t[-l level] [-b blkid | \"bonus\"]", 2978 "find dmu_buf_impl_t's that match specified criteria", dbufs }, 2979 { "abuf_find", "dva_word[0] dva_word[1]", 2980 "find arc_buf_hdr_t of a specified DVA", 2981 abuf_find }, 2982 { "spa", "?[-cv]", "spa_t summary", spa_print }, 2983 { "spa_config", ":", "print spa_t configuration", spa_print_config }, 2984 { "spa_verify", ":", "verify spa_t consistency", spa_verify }, 2985 { "spa_space", ":[-b]", "print spa_t on-disk space usage", spa_space }, 2986 { "spa_vdevs", ":", "given a spa_t, print vdev summary", spa_vdevs }, 2987 { "vdev", ":[-re]\n" 2988 "\t-r display recursively\n" 2989 "\t-e print statistics", 2990 "vdev_t summary", vdev_print }, 2991 { "zio", ":[cpr]\n" 2992 "\t-c display children\n" 2993 "\t-p display parents\n" 2994 "\t-r display recursively", 2995 "zio_t summary", zio_print }, 2996 { "zio_state", "?", "print out all zio_t structures on system or " 2997 "for a particular pool", zio_state }, 2998 { "zfs_blkstats", ":[-v]", 2999 "given a spa_t, print block type stats from last scrub", 3000 zfs_blkstats }, 3001 { "zfs_params", "", "print zfs tunable parameters", zfs_params }, 3002 { "refcount", "", "print refcount_t holders", refcount }, 3003 { "zap_leaf", "", "print zap_leaf_phys_t", zap_leaf }, 3004 { "zfs_aces", ":[-v]", "print all ACEs from a zfs_acl_t", 3005 zfs_acl_dump }, 3006 { "zfs_ace", ":[-v]", "print zfs_ace", zfs_ace_print }, 3007 { "zfs_ace0", ":[-v]", "print zfs_ace0", zfs_ace0_print }, 3008 { "sa_attr_table", ":", "print SA attribute table from sa_os_t", 3009 sa_attr_table}, 3010 { "sa_attr", ": attr_id", 3011 "print SA attribute address when given sa_handle_t", sa_attr_print}, 3012 { "zfs_dbgmsg", ":[-v]", 3013 "print zfs debug log", dbgmsg}, 3014 { NULL } 3015 }; 3016 3017 static const mdb_walker_t walkers[] = { 3018 { "zms_freelist", "walk ZFS metaslab freelist", 3019 freelist_walk_init, freelist_walk_step, NULL }, 3020 { "txg_list", "given any txg_list_t *, walk all entries in all txgs", 3021 txg_list_walk_init, txg_list_walk_step, NULL }, 3022 { "txg_list0", "given any txg_list_t *, walk all entries in txg 0", 3023 txg_list0_walk_init, txg_list_walk_step, NULL }, 3024 { "txg_list1", "given any txg_list_t *, walk all entries in txg 1", 3025 txg_list1_walk_init, txg_list_walk_step, NULL }, 3026 { "txg_list2", "given any txg_list_t *, walk all entries in txg 2", 3027 txg_list2_walk_init, txg_list_walk_step, NULL }, 3028 { "txg_list3", "given any txg_list_t *, walk all entries in txg 3", 3029 txg_list3_walk_init, txg_list_walk_step, NULL }, 3030 { "zio", "walk all zio structures, optionally for a particular spa_t", 3031 zio_walk_init, zio_walk_step, NULL }, 3032 { "zio_root", "walk all root zio_t structures, optionally for a " 3033 "particular spa_t", 3034 zio_walk_init, zio_walk_root_step, NULL }, 3035 { "spa", "walk all spa_t entries in the namespace", 3036 spa_walk_init, spa_walk_step, NULL }, 3037 { "metaslab", "given a spa_t *, walk all metaslab_t structures", 3038 metaslab_walk_init, metaslab_walk_step, NULL }, 3039 { "zfs_acl_node", "given a zfs_acl_t, walk all zfs_acl_nodes", 3040 zfs_acl_node_walk_init, zfs_acl_node_walk_step, NULL }, 3041 { "zfs_acl_node_aces", "given a zfs_acl_node_t, walk all ACEs", 3042 zfs_acl_node_aces_walk_init, zfs_aces_walk_step, NULL }, 3043 { "zfs_acl_node_aces0", 3044 "given a zfs_acl_node_t, walk all ACEs as ace_t", 3045 zfs_acl_node_aces0_walk_init, zfs_aces_walk_step, NULL }, 3046 { NULL } 3047 }; 3048 3049 static const mdb_modinfo_t modinfo = { 3050 MDB_API_VERSION, dcmds, walkers 3051 }; 3052 3053 const mdb_modinfo_t * 3054 _mdb_init(void) 3055 { 3056 return (&modinfo); 3057 } 3058