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 * Copyright 2011 Nexenta Systems, Inc. All rights reserved. 24 * Copyright (c) 2013 by Delphix. All rights reserved. 25 */ 26 27 /* Portions Copyright 2010 Robert Milkowski */ 28 29 #include <mdb/mdb_ctf.h> 30 #include <sys/zfs_context.h> 31 #include <sys/mdb_modapi.h> 32 #include <sys/dbuf.h> 33 #include <sys/dmu_objset.h> 34 #include <sys/dsl_dir.h> 35 #include <sys/dsl_pool.h> 36 #include <sys/metaslab_impl.h> 37 #include <sys/space_map.h> 38 #include <sys/list.h> 39 #include <sys/spa_impl.h> 40 #include <sys/vdev_impl.h> 41 #include <sys/zap_leaf.h> 42 #include <sys/zap_impl.h> 43 #include <ctype.h> 44 #include <sys/zfs_acl.h> 45 #include <sys/sa_impl.h> 46 47 #ifdef _KERNEL 48 #define ZFS_OBJ_NAME "zfs" 49 #else 50 #define ZFS_OBJ_NAME "libzpool.so.1" 51 #endif 52 53 #define ZFS_STRUCT "struct " ZFS_OBJ_NAME "`" 54 55 #ifndef _KERNEL 56 int aok; 57 #endif 58 59 static int 60 getmember(uintptr_t addr, const char *type, mdb_ctf_id_t *idp, 61 const char *member, int len, void *buf) 62 { 63 mdb_ctf_id_t id; 64 ulong_t off; 65 char name[64]; 66 67 if (idp == NULL) { 68 if (mdb_ctf_lookup_by_name(type, &id) == -1) { 69 mdb_warn("couldn't find type %s", type); 70 return (DCMD_ERR); 71 } 72 idp = &id; 73 } else { 74 type = name; 75 mdb_ctf_type_name(*idp, name, sizeof (name)); 76 } 77 78 if (mdb_ctf_offsetof(*idp, member, &off) == -1) { 79 mdb_warn("couldn't find member %s of type %s\n", member, type); 80 return (DCMD_ERR); 81 } 82 if (off % 8 != 0) { 83 mdb_warn("member %s of type %s is unsupported bitfield", 84 member, type); 85 return (DCMD_ERR); 86 } 87 off /= 8; 88 89 if (mdb_vread(buf, len, addr + off) == -1) { 90 mdb_warn("failed to read %s from %s at %p", 91 member, type, addr + off); 92 return (DCMD_ERR); 93 } 94 /* mdb_warn("read %s from %s at %p+%llx\n", member, type, addr, off); */ 95 96 return (0); 97 } 98 99 #define GETMEMB(addr, structname, member, dest) \ 100 getmember(addr, ZFS_STRUCT structname, NULL, #member, \ 101 sizeof (dest), &(dest)) 102 103 #define GETMEMBID(addr, ctfid, member, dest) \ 104 getmember(addr, NULL, ctfid, #member, sizeof (dest), &(dest)) 105 106 static boolean_t 107 strisprint(const char *cp) 108 { 109 for (; *cp; cp++) { 110 if (!isprint(*cp)) 111 return (B_FALSE); 112 } 113 return (B_TRUE); 114 } 115 116 static int verbose; 117 118 static int 119 freelist_walk_init(mdb_walk_state_t *wsp) 120 { 121 if (wsp->walk_addr == NULL) { 122 mdb_warn("must supply starting address\n"); 123 return (WALK_ERR); 124 } 125 126 wsp->walk_data = 0; /* Index into the freelist */ 127 return (WALK_NEXT); 128 } 129 130 static int 131 freelist_walk_step(mdb_walk_state_t *wsp) 132 { 133 uint64_t entry; 134 uintptr_t number = (uintptr_t)wsp->walk_data; 135 char *ddata[] = { "ALLOC", "FREE", "CONDENSE", "INVALID", 136 "INVALID", "INVALID", "INVALID", "INVALID" }; 137 int mapshift = SPA_MINBLOCKSHIFT; 138 139 if (mdb_vread(&entry, sizeof (entry), wsp->walk_addr) == -1) { 140 mdb_warn("failed to read freelist entry %p", wsp->walk_addr); 141 return (WALK_DONE); 142 } 143 wsp->walk_addr += sizeof (entry); 144 wsp->walk_data = (void *)(number + 1); 145 146 if (SM_DEBUG_DECODE(entry)) { 147 mdb_printf("DEBUG: %3u %10s: txg=%llu pass=%llu\n", 148 number, 149 ddata[SM_DEBUG_ACTION_DECODE(entry)], 150 SM_DEBUG_TXG_DECODE(entry), 151 SM_DEBUG_SYNCPASS_DECODE(entry)); 152 } else { 153 mdb_printf("Entry: %3u offsets=%08llx-%08llx type=%c " 154 "size=%06llx", number, 155 SM_OFFSET_DECODE(entry) << mapshift, 156 (SM_OFFSET_DECODE(entry) + SM_RUN_DECODE(entry)) << 157 mapshift, 158 SM_TYPE_DECODE(entry) == SM_ALLOC ? 'A' : 'F', 159 SM_RUN_DECODE(entry) << mapshift); 160 if (verbose) 161 mdb_printf(" (raw=%012llx)\n", entry); 162 mdb_printf("\n"); 163 } 164 return (WALK_NEXT); 165 } 166 167 static int 168 mdb_dsl_dir_name(uintptr_t addr, char *buf) 169 { 170 static int gotid; 171 static mdb_ctf_id_t dd_id; 172 uintptr_t dd_parent; 173 char dd_myname[MAXNAMELEN]; 174 175 if (!gotid) { 176 if (mdb_ctf_lookup_by_name(ZFS_STRUCT "dsl_dir", 177 &dd_id) == -1) { 178 mdb_warn("couldn't find struct dsl_dir"); 179 return (DCMD_ERR); 180 } 181 gotid = TRUE; 182 } 183 if (GETMEMBID(addr, &dd_id, dd_parent, dd_parent) || 184 GETMEMBID(addr, &dd_id, dd_myname, dd_myname)) { 185 return (DCMD_ERR); 186 } 187 188 if (dd_parent) { 189 if (mdb_dsl_dir_name(dd_parent, buf)) 190 return (DCMD_ERR); 191 strcat(buf, "/"); 192 } 193 194 if (dd_myname[0]) 195 strcat(buf, dd_myname); 196 else 197 strcat(buf, "???"); 198 199 return (0); 200 } 201 202 static int 203 objset_name(uintptr_t addr, char *buf) 204 { 205 static int gotid; 206 static mdb_ctf_id_t os_id, ds_id; 207 uintptr_t os_dsl_dataset; 208 char ds_snapname[MAXNAMELEN]; 209 uintptr_t ds_dir; 210 211 buf[0] = '\0'; 212 213 if (!gotid) { 214 if (mdb_ctf_lookup_by_name(ZFS_STRUCT "objset", 215 &os_id) == -1) { 216 mdb_warn("couldn't find struct objset"); 217 return (DCMD_ERR); 218 } 219 if (mdb_ctf_lookup_by_name(ZFS_STRUCT "dsl_dataset", 220 &ds_id) == -1) { 221 mdb_warn("couldn't find struct dsl_dataset"); 222 return (DCMD_ERR); 223 } 224 225 gotid = TRUE; 226 } 227 228 if (GETMEMBID(addr, &os_id, os_dsl_dataset, os_dsl_dataset)) 229 return (DCMD_ERR); 230 231 if (os_dsl_dataset == 0) { 232 strcat(buf, "mos"); 233 return (0); 234 } 235 236 if (GETMEMBID(os_dsl_dataset, &ds_id, ds_snapname, ds_snapname) || 237 GETMEMBID(os_dsl_dataset, &ds_id, ds_dir, ds_dir)) { 238 return (DCMD_ERR); 239 } 240 241 if (ds_dir && mdb_dsl_dir_name(ds_dir, buf)) 242 return (DCMD_ERR); 243 244 if (ds_snapname[0]) { 245 strcat(buf, "@"); 246 strcat(buf, ds_snapname); 247 } 248 return (0); 249 } 250 251 static void 252 enum_lookup(char *out, size_t size, mdb_ctf_id_t id, int val, 253 const char *prefix) 254 { 255 const char *cp; 256 size_t len = strlen(prefix); 257 258 if ((cp = mdb_ctf_enum_name(id, val)) != NULL) { 259 if (strncmp(cp, prefix, len) == 0) 260 cp += len; 261 (void) strncpy(out, cp, size); 262 } else { 263 mdb_snprintf(out, size, "? (%d)", val); 264 } 265 } 266 267 /* ARGSUSED */ 268 static int 269 zfs_params(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv) 270 { 271 /* 272 * This table can be approximately generated by running: 273 * egrep "^[a-z0-9_]+ [a-z0-9_]+( =.*)?;" *.c | cut -d ' ' -f 2 274 */ 275 static const char *params[] = { 276 "arc_reduce_dnlc_percent", 277 "zfs_arc_max", 278 "zfs_arc_min", 279 "arc_shrink_shift", 280 "zfs_mdcomp_disable", 281 "zfs_prefetch_disable", 282 "zfetch_max_streams", 283 "zfetch_min_sec_reap", 284 "zfetch_block_cap", 285 "zfetch_array_rd_sz", 286 "zfs_default_bs", 287 "zfs_default_ibs", 288 "metaslab_aliquot", 289 "reference_tracking_enable", 290 "reference_history", 291 "spa_max_replication_override", 292 "spa_mode_global", 293 "zfs_flags", 294 "zfs_txg_synctime_ms", 295 "zfs_txg_timeout", 296 "zfs_write_limit_min", 297 "zfs_write_limit_max", 298 "zfs_write_limit_shift", 299 "zfs_write_limit_override", 300 "zfs_no_write_throttle", 301 "zfs_vdev_cache_max", 302 "zfs_vdev_cache_size", 303 "zfs_vdev_cache_bshift", 304 "vdev_mirror_shift", 305 "zfs_vdev_max_pending", 306 "zfs_vdev_min_pending", 307 "zfs_scrub_limit", 308 "zfs_no_scrub_io", 309 "zfs_no_scrub_prefetch", 310 "zfs_vdev_time_shift", 311 "zfs_vdev_ramp_rate", 312 "zfs_vdev_aggregation_limit", 313 "fzap_default_block_shift", 314 "zfs_immediate_write_sz", 315 "zfs_read_chunk_size", 316 "zfs_nocacheflush", 317 "zil_replay_disable", 318 "metaslab_gang_bang", 319 "metaslab_df_alloc_threshold", 320 "metaslab_df_free_pct", 321 "zio_injection_enabled", 322 "zvol_immediate_write_sz", 323 }; 324 325 for (int i = 0; i < sizeof (params) / sizeof (params[0]); i++) { 326 int sz; 327 uint64_t val64; 328 uint32_t *val32p = (uint32_t *)&val64; 329 330 sz = mdb_readvar(&val64, params[i]); 331 if (sz == 4) { 332 mdb_printf("%s = 0x%x\n", params[i], *val32p); 333 } else if (sz == 8) { 334 mdb_printf("%s = 0x%llx\n", params[i], val64); 335 } else { 336 mdb_warn("variable %s not found", params[i]); 337 } 338 } 339 340 return (DCMD_OK); 341 } 342 343 /* ARGSUSED */ 344 static int 345 blkptr(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv) 346 { 347 mdb_ctf_id_t type_enum, checksum_enum, compress_enum; 348 char type[80], checksum[80], compress[80]; 349 blkptr_t blk, *bp = &blk; 350 char buf[BP_SPRINTF_LEN]; 351 352 if (mdb_vread(&blk, sizeof (blkptr_t), addr) == -1) { 353 mdb_warn("failed to read blkptr_t"); 354 return (DCMD_ERR); 355 } 356 357 if (mdb_ctf_lookup_by_name("enum dmu_object_type", &type_enum) == -1 || 358 mdb_ctf_lookup_by_name("enum zio_checksum", &checksum_enum) == -1 || 359 mdb_ctf_lookup_by_name("enum zio_compress", &compress_enum) == -1) { 360 mdb_warn("Could not find blkptr enumerated types"); 361 return (DCMD_ERR); 362 } 363 364 enum_lookup(type, sizeof (type), type_enum, 365 BP_GET_TYPE(bp), "DMU_OT_"); 366 enum_lookup(checksum, sizeof (checksum), checksum_enum, 367 BP_GET_CHECKSUM(bp), "ZIO_CHECKSUM_"); 368 enum_lookup(compress, sizeof (compress), compress_enum, 369 BP_GET_COMPRESS(bp), "ZIO_COMPRESS_"); 370 371 SPRINTF_BLKPTR(mdb_snprintf, '\n', buf, bp, type, checksum, compress); 372 373 mdb_printf("%s\n", buf); 374 375 return (DCMD_OK); 376 } 377 378 typedef struct mdb_dmu_buf_impl { 379 struct { 380 uint64_t db_object; 381 } db; 382 void *db_objset; 383 uint64_t db_level; 384 uint64_t db_blkid; 385 struct { 386 uint64_t rc_count; 387 } db_holds; 388 } mdb_dmu_buf_impl_t; 389 390 /* ARGSUSED */ 391 static int 392 dbuf(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv) 393 { 394 mdb_dmu_buf_impl_t db; 395 char objectname[32]; 396 char blkidname[32]; 397 char path[MAXNAMELEN]; 398 399 if (DCMD_HDRSPEC(flags)) 400 mdb_printf(" addr object lvl blkid holds os\n"); 401 402 if (mdb_ctf_vread(&db, ZFS_STRUCT "dmu_buf_impl", "mdb_dmu_buf_impl_t", 403 addr, 0) == -1) 404 return (DCMD_ERR); 405 406 if (db.db.db_object == DMU_META_DNODE_OBJECT) 407 (void) strcpy(objectname, "mdn"); 408 else 409 (void) mdb_snprintf(objectname, sizeof (objectname), "%llx", 410 (u_longlong_t)db.db.db_object); 411 412 if (db.db_blkid == DMU_BONUS_BLKID) 413 (void) strcpy(blkidname, "bonus"); 414 else 415 (void) mdb_snprintf(blkidname, sizeof (blkidname), "%llx", 416 (u_longlong_t)db.db_blkid); 417 418 if (objset_name((uintptr_t)db.db_objset, path)) { 419 return (DCMD_ERR); 420 } 421 422 mdb_printf("%p %8s %1u %9s %2llu %s\n", addr, 423 objectname, (int)db.db_level, blkidname, 424 db.db_holds.rc_count, path); 425 426 return (DCMD_OK); 427 } 428 429 /* ARGSUSED */ 430 static int 431 dbuf_stats(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv) 432 { 433 #define HISTOSZ 32 434 uintptr_t dbp; 435 dmu_buf_impl_t db; 436 dbuf_hash_table_t ht; 437 uint64_t bucket, ndbufs; 438 uint64_t histo[HISTOSZ]; 439 uint64_t histo2[HISTOSZ]; 440 int i, maxidx; 441 442 if (mdb_readvar(&ht, "dbuf_hash_table") == -1) { 443 mdb_warn("failed to read 'dbuf_hash_table'"); 444 return (DCMD_ERR); 445 } 446 447 for (i = 0; i < HISTOSZ; i++) { 448 histo[i] = 0; 449 histo2[i] = 0; 450 } 451 452 ndbufs = 0; 453 for (bucket = 0; bucket < ht.hash_table_mask+1; bucket++) { 454 int len; 455 456 if (mdb_vread(&dbp, sizeof (void *), 457 (uintptr_t)(ht.hash_table+bucket)) == -1) { 458 mdb_warn("failed to read hash bucket %u at %p", 459 bucket, ht.hash_table+bucket); 460 return (DCMD_ERR); 461 } 462 463 len = 0; 464 while (dbp != 0) { 465 if (mdb_vread(&db, sizeof (dmu_buf_impl_t), 466 dbp) == -1) { 467 mdb_warn("failed to read dbuf at %p", dbp); 468 return (DCMD_ERR); 469 } 470 dbp = (uintptr_t)db.db_hash_next; 471 for (i = MIN(len, HISTOSZ - 1); i >= 0; i--) 472 histo2[i]++; 473 len++; 474 ndbufs++; 475 } 476 477 if (len >= HISTOSZ) 478 len = HISTOSZ-1; 479 histo[len]++; 480 } 481 482 mdb_printf("hash table has %llu buckets, %llu dbufs " 483 "(avg %llu buckets/dbuf)\n", 484 ht.hash_table_mask+1, ndbufs, 485 (ht.hash_table_mask+1)/ndbufs); 486 487 mdb_printf("\n"); 488 maxidx = 0; 489 for (i = 0; i < HISTOSZ; i++) 490 if (histo[i] > 0) 491 maxidx = i; 492 mdb_printf("hash chain length number of buckets\n"); 493 for (i = 0; i <= maxidx; i++) 494 mdb_printf("%u %llu\n", i, histo[i]); 495 496 mdb_printf("\n"); 497 maxidx = 0; 498 for (i = 0; i < HISTOSZ; i++) 499 if (histo2[i] > 0) 500 maxidx = i; 501 mdb_printf("hash chain depth number of dbufs\n"); 502 for (i = 0; i <= maxidx; i++) 503 mdb_printf("%u or more %llu %llu%%\n", 504 i, histo2[i], histo2[i]*100/ndbufs); 505 506 507 return (DCMD_OK); 508 } 509 510 #define CHAIN_END 0xffff 511 /* 512 * ::zap_leaf [-v] 513 * 514 * Print a zap_leaf_phys_t, assumed to be 16k 515 */ 516 /* ARGSUSED */ 517 static int 518 zap_leaf(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv) 519 { 520 char buf[16*1024]; 521 int verbose = B_FALSE; 522 int four = B_FALSE; 523 zap_leaf_t l; 524 zap_leaf_phys_t *zlp = (void *)buf; 525 int i; 526 527 if (mdb_getopts(argc, argv, 528 'v', MDB_OPT_SETBITS, TRUE, &verbose, 529 '4', MDB_OPT_SETBITS, TRUE, &four, 530 NULL) != argc) 531 return (DCMD_USAGE); 532 533 l.l_phys = zlp; 534 l.l_bs = 14; /* assume 16k blocks */ 535 if (four) 536 l.l_bs = 12; 537 538 if (!(flags & DCMD_ADDRSPEC)) { 539 return (DCMD_USAGE); 540 } 541 542 if (mdb_vread(buf, sizeof (buf), addr) == -1) { 543 mdb_warn("failed to read zap_leaf_phys_t at %p", addr); 544 return (DCMD_ERR); 545 } 546 547 if (zlp->l_hdr.lh_block_type != ZBT_LEAF || 548 zlp->l_hdr.lh_magic != ZAP_LEAF_MAGIC) { 549 mdb_warn("This does not appear to be a zap_leaf_phys_t"); 550 return (DCMD_ERR); 551 } 552 553 mdb_printf("zap_leaf_phys_t at %p:\n", addr); 554 mdb_printf(" lh_prefix_len = %u\n", zlp->l_hdr.lh_prefix_len); 555 mdb_printf(" lh_prefix = %llx\n", zlp->l_hdr.lh_prefix); 556 mdb_printf(" lh_nentries = %u\n", zlp->l_hdr.lh_nentries); 557 mdb_printf(" lh_nfree = %u\n", zlp->l_hdr.lh_nfree, 558 zlp->l_hdr.lh_nfree * 100 / (ZAP_LEAF_NUMCHUNKS(&l))); 559 mdb_printf(" lh_freelist = %u\n", zlp->l_hdr.lh_freelist); 560 mdb_printf(" lh_flags = %x (%s)\n", zlp->l_hdr.lh_flags, 561 zlp->l_hdr.lh_flags & ZLF_ENTRIES_CDSORTED ? 562 "ENTRIES_CDSORTED" : ""); 563 564 if (verbose) { 565 mdb_printf(" hash table:\n"); 566 for (i = 0; i < ZAP_LEAF_HASH_NUMENTRIES(&l); i++) { 567 if (zlp->l_hash[i] != CHAIN_END) 568 mdb_printf(" %u: %u\n", i, zlp->l_hash[i]); 569 } 570 } 571 572 mdb_printf(" chunks:\n"); 573 for (i = 0; i < ZAP_LEAF_NUMCHUNKS(&l); i++) { 574 /* LINTED: alignment */ 575 zap_leaf_chunk_t *zlc = &ZAP_LEAF_CHUNK(&l, i); 576 switch (zlc->l_entry.le_type) { 577 case ZAP_CHUNK_FREE: 578 if (verbose) { 579 mdb_printf(" %u: free; lf_next = %u\n", 580 i, zlc->l_free.lf_next); 581 } 582 break; 583 case ZAP_CHUNK_ENTRY: 584 mdb_printf(" %u: entry\n", i); 585 if (verbose) { 586 mdb_printf(" le_next = %u\n", 587 zlc->l_entry.le_next); 588 } 589 mdb_printf(" le_name_chunk = %u\n", 590 zlc->l_entry.le_name_chunk); 591 mdb_printf(" le_name_numints = %u\n", 592 zlc->l_entry.le_name_numints); 593 mdb_printf(" le_value_chunk = %u\n", 594 zlc->l_entry.le_value_chunk); 595 mdb_printf(" le_value_intlen = %u\n", 596 zlc->l_entry.le_value_intlen); 597 mdb_printf(" le_value_numints = %u\n", 598 zlc->l_entry.le_value_numints); 599 mdb_printf(" le_cd = %u\n", 600 zlc->l_entry.le_cd); 601 mdb_printf(" le_hash = %llx\n", 602 zlc->l_entry.le_hash); 603 break; 604 case ZAP_CHUNK_ARRAY: 605 mdb_printf(" %u: array", i); 606 if (strisprint((char *)zlc->l_array.la_array)) 607 mdb_printf(" \"%s\"", zlc->l_array.la_array); 608 mdb_printf("\n"); 609 if (verbose) { 610 int j; 611 mdb_printf(" "); 612 for (j = 0; j < ZAP_LEAF_ARRAY_BYTES; j++) { 613 mdb_printf("%02x ", 614 zlc->l_array.la_array[j]); 615 } 616 mdb_printf("\n"); 617 } 618 if (zlc->l_array.la_next != CHAIN_END) { 619 mdb_printf(" lf_next = %u\n", 620 zlc->l_array.la_next); 621 } 622 break; 623 default: 624 mdb_printf(" %u: undefined type %u\n", 625 zlc->l_entry.le_type); 626 } 627 } 628 629 return (DCMD_OK); 630 } 631 632 typedef struct dbufs_data { 633 mdb_ctf_id_t id; 634 uint64_t objset; 635 uint64_t object; 636 uint64_t level; 637 uint64_t blkid; 638 char *osname; 639 } dbufs_data_t; 640 641 #define DBUFS_UNSET (0xbaddcafedeadbeefULL) 642 643 /* ARGSUSED */ 644 static int 645 dbufs_cb(uintptr_t addr, const void *unknown, void *arg) 646 { 647 dbufs_data_t *data = arg; 648 uintptr_t objset; 649 dmu_buf_t db; 650 uint8_t level; 651 uint64_t blkid; 652 char osname[MAXNAMELEN]; 653 654 if (GETMEMBID(addr, &data->id, db_objset, objset) || 655 GETMEMBID(addr, &data->id, db, db) || 656 GETMEMBID(addr, &data->id, db_level, level) || 657 GETMEMBID(addr, &data->id, db_blkid, blkid)) { 658 return (WALK_ERR); 659 } 660 661 if ((data->objset == DBUFS_UNSET || data->objset == objset) && 662 (data->osname == NULL || (objset_name(objset, osname) == 0 && 663 strcmp(data->osname, osname) == 0)) && 664 (data->object == DBUFS_UNSET || data->object == db.db_object) && 665 (data->level == DBUFS_UNSET || data->level == level) && 666 (data->blkid == DBUFS_UNSET || data->blkid == blkid)) { 667 mdb_printf("%#lr\n", addr); 668 } 669 return (WALK_NEXT); 670 } 671 672 /* ARGSUSED */ 673 static int 674 dbufs(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv) 675 { 676 dbufs_data_t data; 677 char *object = NULL; 678 char *blkid = NULL; 679 680 data.objset = data.object = data.level = data.blkid = DBUFS_UNSET; 681 data.osname = NULL; 682 683 if (mdb_getopts(argc, argv, 684 'O', MDB_OPT_UINT64, &data.objset, 685 'n', MDB_OPT_STR, &data.osname, 686 'o', MDB_OPT_STR, &object, 687 'l', MDB_OPT_UINT64, &data.level, 688 'b', MDB_OPT_STR, &blkid) != argc) { 689 return (DCMD_USAGE); 690 } 691 692 if (object) { 693 if (strcmp(object, "mdn") == 0) { 694 data.object = DMU_META_DNODE_OBJECT; 695 } else { 696 data.object = mdb_strtoull(object); 697 } 698 } 699 700 if (blkid) { 701 if (strcmp(blkid, "bonus") == 0) { 702 data.blkid = DMU_BONUS_BLKID; 703 } else { 704 data.blkid = mdb_strtoull(blkid); 705 } 706 } 707 708 if (mdb_ctf_lookup_by_name(ZFS_STRUCT "dmu_buf_impl", &data.id) == -1) { 709 mdb_warn("couldn't find struct dmu_buf_impl_t"); 710 return (DCMD_ERR); 711 } 712 713 if (mdb_walk("dmu_buf_impl_t", dbufs_cb, &data) != 0) { 714 mdb_warn("can't walk dbufs"); 715 return (DCMD_ERR); 716 } 717 718 return (DCMD_OK); 719 } 720 721 typedef struct abuf_find_data { 722 dva_t dva; 723 mdb_ctf_id_t id; 724 } abuf_find_data_t; 725 726 /* ARGSUSED */ 727 static int 728 abuf_find_cb(uintptr_t addr, const void *unknown, void *arg) 729 { 730 abuf_find_data_t *data = arg; 731 dva_t dva; 732 733 if (GETMEMBID(addr, &data->id, b_dva, dva)) { 734 return (WALK_ERR); 735 } 736 737 if (dva.dva_word[0] == data->dva.dva_word[0] && 738 dva.dva_word[1] == data->dva.dva_word[1]) { 739 mdb_printf("%#lr\n", addr); 740 } 741 return (WALK_NEXT); 742 } 743 744 /* ARGSUSED */ 745 static int 746 abuf_find(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv) 747 { 748 abuf_find_data_t data; 749 GElf_Sym sym; 750 int i; 751 const char *syms[] = { 752 "ARC_mru", 753 "ARC_mru_ghost", 754 "ARC_mfu", 755 "ARC_mfu_ghost", 756 }; 757 758 if (argc != 2) 759 return (DCMD_USAGE); 760 761 for (i = 0; i < 2; i ++) { 762 switch (argv[i].a_type) { 763 case MDB_TYPE_STRING: 764 data.dva.dva_word[i] = mdb_strtoull(argv[i].a_un.a_str); 765 break; 766 case MDB_TYPE_IMMEDIATE: 767 data.dva.dva_word[i] = argv[i].a_un.a_val; 768 break; 769 default: 770 return (DCMD_USAGE); 771 } 772 } 773 774 if (mdb_ctf_lookup_by_name(ZFS_STRUCT "arc_buf_hdr", &data.id) == -1) { 775 mdb_warn("couldn't find struct arc_buf_hdr"); 776 return (DCMD_ERR); 777 } 778 779 for (i = 0; i < sizeof (syms) / sizeof (syms[0]); i++) { 780 if (mdb_lookup_by_name(syms[i], &sym)) { 781 mdb_warn("can't find symbol %s", syms[i]); 782 return (DCMD_ERR); 783 } 784 785 if (mdb_pwalk("list", abuf_find_cb, &data, sym.st_value) != 0) { 786 mdb_warn("can't walk %s", syms[i]); 787 return (DCMD_ERR); 788 } 789 } 790 791 return (DCMD_OK); 792 } 793 794 795 typedef struct dbgmsg_arg { 796 boolean_t da_verbose; 797 boolean_t da_address; 798 } dbgmsg_arg_t; 799 800 /* ARGSUSED */ 801 static int 802 dbgmsg_cb(uintptr_t addr, const void *unknown, void *arg) 803 { 804 static mdb_ctf_id_t id; 805 static boolean_t gotid; 806 static ulong_t off; 807 808 dbgmsg_arg_t *da = arg; 809 time_t timestamp; 810 char buf[1024]; 811 812 if (!gotid) { 813 if (mdb_ctf_lookup_by_name(ZFS_STRUCT "zfs_dbgmsg", &id) == 814 -1) { 815 mdb_warn("couldn't find struct zfs_dbgmsg"); 816 return (WALK_ERR); 817 } 818 gotid = TRUE; 819 if (mdb_ctf_offsetof(id, "zdm_msg", &off) == -1) { 820 mdb_warn("couldn't find zdm_msg"); 821 return (WALK_ERR); 822 } 823 off /= 8; 824 } 825 826 827 if (GETMEMBID(addr, &id, zdm_timestamp, timestamp)) { 828 return (WALK_ERR); 829 } 830 831 if (mdb_readstr(buf, sizeof (buf), addr + off) == -1) { 832 mdb_warn("failed to read zdm_msg at %p\n", addr + off); 833 return (DCMD_ERR); 834 } 835 836 if (da->da_address) 837 mdb_printf("%p ", addr); 838 if (da->da_verbose) 839 mdb_printf("%Y ", timestamp); 840 841 mdb_printf("%s\n", buf); 842 843 if (da->da_verbose) 844 (void) mdb_call_dcmd("whatis", addr, DCMD_ADDRSPEC, 0, NULL); 845 846 return (WALK_NEXT); 847 } 848 849 /* ARGSUSED */ 850 static int 851 dbgmsg(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv) 852 { 853 GElf_Sym sym; 854 dbgmsg_arg_t da = { 0 }; 855 856 if (mdb_getopts(argc, argv, 857 'v', MDB_OPT_SETBITS, B_TRUE, &da.da_verbose, 858 'a', MDB_OPT_SETBITS, B_TRUE, &da.da_address, 859 NULL) != argc) 860 return (DCMD_USAGE); 861 862 if (mdb_lookup_by_name("zfs_dbgmsgs", &sym)) { 863 mdb_warn("can't find zfs_dbgmsgs"); 864 return (DCMD_ERR); 865 } 866 867 if (mdb_pwalk("list", dbgmsg_cb, &da, sym.st_value) != 0) { 868 mdb_warn("can't walk zfs_dbgmsgs"); 869 return (DCMD_ERR); 870 } 871 872 return (DCMD_OK); 873 } 874 875 /*ARGSUSED*/ 876 static int 877 arc_print(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv) 878 { 879 kstat_named_t *stats; 880 GElf_Sym sym; 881 int nstats, i; 882 uint_t opt_a = FALSE; 883 uint_t opt_b = FALSE; 884 uint_t shift = 0; 885 const char *suffix; 886 887 static const char *bytestats[] = { 888 "p", "c", "c_min", "c_max", "size", "duplicate_buffers_size", 889 NULL 890 }; 891 892 static const char *extras[] = { 893 "arc_no_grow", "arc_tempreserve", 894 "arc_meta_used", "arc_meta_limit", "arc_meta_max", 895 NULL 896 }; 897 898 if (mdb_lookup_by_name("arc_stats", &sym) == -1) { 899 mdb_warn("failed to find 'arc_stats'"); 900 return (DCMD_ERR); 901 } 902 903 stats = mdb_zalloc(sym.st_size, UM_SLEEP | UM_GC); 904 905 if (mdb_vread(stats, sym.st_size, sym.st_value) == -1) { 906 mdb_warn("couldn't read 'arc_stats' at %p", sym.st_value); 907 return (DCMD_ERR); 908 } 909 910 nstats = sym.st_size / sizeof (kstat_named_t); 911 912 /* NB: -a / opt_a are ignored for backwards compatability */ 913 if (mdb_getopts(argc, argv, 914 'a', MDB_OPT_SETBITS, TRUE, &opt_a, 915 'b', MDB_OPT_SETBITS, TRUE, &opt_b, 916 'k', MDB_OPT_SETBITS, 10, &shift, 917 'm', MDB_OPT_SETBITS, 20, &shift, 918 'g', MDB_OPT_SETBITS, 30, &shift, 919 NULL) != argc) 920 return (DCMD_USAGE); 921 922 if (!opt_b && !shift) 923 shift = 20; 924 925 switch (shift) { 926 case 0: 927 suffix = "B"; 928 break; 929 case 10: 930 suffix = "KB"; 931 break; 932 case 20: 933 suffix = "MB"; 934 break; 935 case 30: 936 suffix = "GB"; 937 break; 938 default: 939 suffix = "XX"; 940 } 941 942 for (i = 0; i < nstats; i++) { 943 int j; 944 boolean_t bytes = B_FALSE; 945 946 for (j = 0; bytestats[j]; j++) { 947 if (strcmp(stats[i].name, bytestats[j]) == 0) { 948 bytes = B_TRUE; 949 break; 950 } 951 } 952 953 if (bytes) { 954 mdb_printf("%-25s = %9llu %s\n", stats[i].name, 955 stats[i].value.ui64 >> shift, suffix); 956 } else { 957 mdb_printf("%-25s = %9llu\n", stats[i].name, 958 stats[i].value.ui64); 959 } 960 } 961 962 for (i = 0; extras[i]; i++) { 963 uint64_t buf; 964 965 if (mdb_lookup_by_name(extras[i], &sym) == -1) { 966 mdb_warn("failed to find '%s'", extras[i]); 967 return (DCMD_ERR); 968 } 969 970 if (sym.st_size != sizeof (uint64_t) && 971 sym.st_size != sizeof (uint32_t)) { 972 mdb_warn("expected scalar for variable '%s'\n", 973 extras[i]); 974 return (DCMD_ERR); 975 } 976 977 if (mdb_vread(&buf, sym.st_size, sym.st_value) == -1) { 978 mdb_warn("couldn't read '%s'", extras[i]); 979 return (DCMD_ERR); 980 } 981 982 mdb_printf("%-25s = ", extras[i]); 983 984 /* NB: all the 64-bit extras happen to be byte counts */ 985 if (sym.st_size == sizeof (uint64_t)) 986 mdb_printf("%9llu %s\n", buf >> shift, suffix); 987 988 if (sym.st_size == sizeof (uint32_t)) 989 mdb_printf("%9d\n", *((uint32_t *)&buf)); 990 } 991 return (DCMD_OK); 992 } 993 994 /* 995 * ::spa 996 * 997 * -c Print configuration information as well 998 * -v Print vdev state 999 * -e Print vdev error stats 1000 * 1001 * Print a summarized spa_t. When given no arguments, prints out a table of all 1002 * active pools on the system. 1003 */ 1004 /* ARGSUSED */ 1005 static int 1006 spa_print(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv) 1007 { 1008 spa_t spa; 1009 const char *statetab[] = { "ACTIVE", "EXPORTED", "DESTROYED", 1010 "SPARE", "L2CACHE", "UNINIT", "UNAVAIL", "POTENTIAL" }; 1011 const char *state; 1012 int config = FALSE; 1013 int vdevs = FALSE; 1014 int errors = FALSE; 1015 1016 if (mdb_getopts(argc, argv, 1017 'c', MDB_OPT_SETBITS, TRUE, &config, 1018 'v', MDB_OPT_SETBITS, TRUE, &vdevs, 1019 'e', MDB_OPT_SETBITS, TRUE, &errors, 1020 NULL) != argc) 1021 return (DCMD_USAGE); 1022 1023 if (!(flags & DCMD_ADDRSPEC)) { 1024 if (mdb_walk_dcmd("spa", "spa", argc, argv) == -1) { 1025 mdb_warn("can't walk spa"); 1026 return (DCMD_ERR); 1027 } 1028 1029 return (DCMD_OK); 1030 } 1031 1032 if (flags & DCMD_PIPE_OUT) { 1033 mdb_printf("%#lr\n", addr); 1034 return (DCMD_OK); 1035 } 1036 1037 if (DCMD_HDRSPEC(flags)) 1038 mdb_printf("%<u>%-?s %9s %-*s%</u>\n", "ADDR", "STATE", 1039 sizeof (uintptr_t) == 4 ? 60 : 52, "NAME"); 1040 1041 if (mdb_vread(&spa, sizeof (spa), addr) == -1) { 1042 mdb_warn("failed to read spa_t at %p", addr); 1043 return (DCMD_ERR); 1044 } 1045 1046 if (spa.spa_state < 0 || spa.spa_state > POOL_STATE_UNAVAIL) 1047 state = "UNKNOWN"; 1048 else 1049 state = statetab[spa.spa_state]; 1050 1051 mdb_printf("%0?p %9s %s\n", addr, state, spa.spa_name); 1052 1053 if (config) { 1054 mdb_printf("\n"); 1055 mdb_inc_indent(4); 1056 if (mdb_call_dcmd("spa_config", addr, flags, 0, 1057 NULL) != DCMD_OK) 1058 return (DCMD_ERR); 1059 mdb_dec_indent(4); 1060 } 1061 1062 if (vdevs || errors) { 1063 mdb_arg_t v; 1064 1065 v.a_type = MDB_TYPE_STRING; 1066 v.a_un.a_str = "-e"; 1067 1068 mdb_printf("\n"); 1069 mdb_inc_indent(4); 1070 if (mdb_call_dcmd("spa_vdevs", addr, flags, errors ? 1 : 0, 1071 &v) != DCMD_OK) 1072 return (DCMD_ERR); 1073 mdb_dec_indent(4); 1074 } 1075 1076 return (DCMD_OK); 1077 } 1078 1079 typedef struct mdb_spa_config_spa { 1080 nvlist_t *spa_config; 1081 } mdb_spa_config_spa_t; 1082 1083 /* 1084 * ::spa_config 1085 * 1086 * Given a spa_t, print the configuration information stored in spa_config. 1087 * Since it's just an nvlist, format it as an indented list of name=value pairs. 1088 * We simply read the value of spa_config and pass off to ::nvlist. 1089 */ 1090 /* ARGSUSED */ 1091 static int 1092 spa_print_config(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv) 1093 { 1094 mdb_spa_config_spa_t spa; 1095 1096 if (argc != 0 || !(flags & DCMD_ADDRSPEC)) 1097 return (DCMD_USAGE); 1098 1099 if (mdb_ctf_vread(&spa, ZFS_STRUCT "spa", "mdb_spa_config_spa_t", 1100 addr, 0) == -1) 1101 return (DCMD_ERR); 1102 1103 if (spa.spa_config == NULL) { 1104 mdb_printf("(none)\n"); 1105 return (DCMD_OK); 1106 } 1107 1108 return (mdb_call_dcmd("nvlist", (uintptr_t)spa.spa_config, flags, 1109 0, NULL)); 1110 } 1111 1112 /* 1113 * ::vdev 1114 * 1115 * Print out a summarized vdev_t, in the following form: 1116 * 1117 * ADDR STATE AUX DESC 1118 * fffffffbcde23df0 HEALTHY - /dev/dsk/c0t0d0 1119 * 1120 * If '-r' is specified, recursively visit all children. 1121 * 1122 * With '-e', the statistics associated with the vdev are printed as well. 1123 */ 1124 static int 1125 do_print_vdev(uintptr_t addr, int flags, int depth, int stats, 1126 int recursive) 1127 { 1128 vdev_t vdev; 1129 char desc[MAXNAMELEN]; 1130 int c, children; 1131 uintptr_t *child; 1132 const char *state, *aux; 1133 1134 if (mdb_vread(&vdev, sizeof (vdev), (uintptr_t)addr) == -1) { 1135 mdb_warn("failed to read vdev_t at %p\n", (uintptr_t)addr); 1136 return (DCMD_ERR); 1137 } 1138 1139 if (flags & DCMD_PIPE_OUT) { 1140 mdb_printf("%#lr", addr); 1141 } else { 1142 if (vdev.vdev_path != NULL) { 1143 if (mdb_readstr(desc, sizeof (desc), 1144 (uintptr_t)vdev.vdev_path) == -1) { 1145 mdb_warn("failed to read vdev_path at %p\n", 1146 vdev.vdev_path); 1147 return (DCMD_ERR); 1148 } 1149 } else if (vdev.vdev_ops != NULL) { 1150 vdev_ops_t ops; 1151 if (mdb_vread(&ops, sizeof (ops), 1152 (uintptr_t)vdev.vdev_ops) == -1) { 1153 mdb_warn("failed to read vdev_ops at %p\n", 1154 vdev.vdev_ops); 1155 return (DCMD_ERR); 1156 } 1157 (void) strcpy(desc, ops.vdev_op_type); 1158 } else { 1159 (void) strcpy(desc, "<unknown>"); 1160 } 1161 1162 if (depth == 0 && DCMD_HDRSPEC(flags)) 1163 mdb_printf("%<u>%-?s %-9s %-12s %-*s%</u>\n", 1164 "ADDR", "STATE", "AUX", 1165 sizeof (uintptr_t) == 4 ? 43 : 35, 1166 "DESCRIPTION"); 1167 1168 mdb_printf("%0?p ", addr); 1169 1170 switch (vdev.vdev_state) { 1171 case VDEV_STATE_CLOSED: 1172 state = "CLOSED"; 1173 break; 1174 case VDEV_STATE_OFFLINE: 1175 state = "OFFLINE"; 1176 break; 1177 case VDEV_STATE_CANT_OPEN: 1178 state = "CANT_OPEN"; 1179 break; 1180 case VDEV_STATE_DEGRADED: 1181 state = "DEGRADED"; 1182 break; 1183 case VDEV_STATE_HEALTHY: 1184 state = "HEALTHY"; 1185 break; 1186 case VDEV_STATE_REMOVED: 1187 state = "REMOVED"; 1188 break; 1189 case VDEV_STATE_FAULTED: 1190 state = "FAULTED"; 1191 break; 1192 default: 1193 state = "UNKNOWN"; 1194 break; 1195 } 1196 1197 switch (vdev.vdev_stat.vs_aux) { 1198 case VDEV_AUX_NONE: 1199 aux = "-"; 1200 break; 1201 case VDEV_AUX_OPEN_FAILED: 1202 aux = "OPEN_FAILED"; 1203 break; 1204 case VDEV_AUX_CORRUPT_DATA: 1205 aux = "CORRUPT_DATA"; 1206 break; 1207 case VDEV_AUX_NO_REPLICAS: 1208 aux = "NO_REPLICAS"; 1209 break; 1210 case VDEV_AUX_BAD_GUID_SUM: 1211 aux = "BAD_GUID_SUM"; 1212 break; 1213 case VDEV_AUX_TOO_SMALL: 1214 aux = "TOO_SMALL"; 1215 break; 1216 case VDEV_AUX_BAD_LABEL: 1217 aux = "BAD_LABEL"; 1218 break; 1219 case VDEV_AUX_VERSION_NEWER: 1220 aux = "VERS_NEWER"; 1221 break; 1222 case VDEV_AUX_VERSION_OLDER: 1223 aux = "VERS_OLDER"; 1224 break; 1225 case VDEV_AUX_UNSUP_FEAT: 1226 aux = "UNSUP_FEAT"; 1227 break; 1228 case VDEV_AUX_SPARED: 1229 aux = "SPARED"; 1230 break; 1231 case VDEV_AUX_ERR_EXCEEDED: 1232 aux = "ERR_EXCEEDED"; 1233 break; 1234 case VDEV_AUX_IO_FAILURE: 1235 aux = "IO_FAILURE"; 1236 break; 1237 case VDEV_AUX_BAD_LOG: 1238 aux = "BAD_LOG"; 1239 break; 1240 case VDEV_AUX_EXTERNAL: 1241 aux = "EXTERNAL"; 1242 break; 1243 case VDEV_AUX_SPLIT_POOL: 1244 aux = "SPLIT_POOL"; 1245 break; 1246 default: 1247 aux = "UNKNOWN"; 1248 break; 1249 } 1250 1251 mdb_printf("%-9s %-12s %*s%s\n", state, aux, depth, "", desc); 1252 1253 if (stats) { 1254 vdev_stat_t *vs = &vdev.vdev_stat; 1255 int i; 1256 1257 mdb_inc_indent(4); 1258 mdb_printf("\n"); 1259 mdb_printf("%<u> %12s %12s %12s %12s " 1260 "%12s%</u>\n", "READ", "WRITE", "FREE", "CLAIM", 1261 "IOCTL"); 1262 mdb_printf("OPS "); 1263 for (i = 1; i < ZIO_TYPES; i++) 1264 mdb_printf("%11#llx%s", vs->vs_ops[i], 1265 i == ZIO_TYPES - 1 ? "" : " "); 1266 mdb_printf("\n"); 1267 mdb_printf("BYTES "); 1268 for (i = 1; i < ZIO_TYPES; i++) 1269 mdb_printf("%11#llx%s", vs->vs_bytes[i], 1270 i == ZIO_TYPES - 1 ? "" : " "); 1271 1272 1273 mdb_printf("\n"); 1274 mdb_printf("EREAD %10#llx\n", vs->vs_read_errors); 1275 mdb_printf("EWRITE %10#llx\n", vs->vs_write_errors); 1276 mdb_printf("ECKSUM %10#llx\n", 1277 vs->vs_checksum_errors); 1278 mdb_dec_indent(4); 1279 } 1280 1281 if (stats) 1282 mdb_printf("\n"); 1283 } 1284 1285 children = vdev.vdev_children; 1286 1287 if (children == 0 || !recursive) 1288 return (DCMD_OK); 1289 1290 child = mdb_alloc(children * sizeof (void *), UM_SLEEP | UM_GC); 1291 if (mdb_vread(child, children * sizeof (void *), 1292 (uintptr_t)vdev.vdev_child) == -1) { 1293 mdb_warn("failed to read vdev children at %p", vdev.vdev_child); 1294 return (DCMD_ERR); 1295 } 1296 1297 for (c = 0; c < children; c++) { 1298 if (do_print_vdev(child[c], flags, depth + 2, stats, 1299 recursive)) 1300 return (DCMD_ERR); 1301 } 1302 1303 return (DCMD_OK); 1304 } 1305 1306 static int 1307 vdev_print(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv) 1308 { 1309 int recursive = FALSE; 1310 int stats = FALSE; 1311 uint64_t depth = 0; 1312 1313 if (mdb_getopts(argc, argv, 1314 'r', MDB_OPT_SETBITS, TRUE, &recursive, 1315 'e', MDB_OPT_SETBITS, TRUE, &stats, 1316 'd', MDB_OPT_UINT64, &depth, 1317 NULL) != argc) 1318 return (DCMD_USAGE); 1319 1320 if (!(flags & DCMD_ADDRSPEC)) { 1321 mdb_warn("no vdev_t address given\n"); 1322 return (DCMD_ERR); 1323 } 1324 1325 return (do_print_vdev(addr, flags, (int)depth, stats, recursive)); 1326 } 1327 1328 typedef struct metaslab_walk_data { 1329 uint64_t mw_numvdevs; 1330 uintptr_t *mw_vdevs; 1331 int mw_curvdev; 1332 uint64_t mw_nummss; 1333 uintptr_t *mw_mss; 1334 int mw_curms; 1335 } metaslab_walk_data_t; 1336 1337 static int 1338 metaslab_walk_step(mdb_walk_state_t *wsp) 1339 { 1340 metaslab_walk_data_t *mw = wsp->walk_data; 1341 metaslab_t ms; 1342 uintptr_t msp; 1343 1344 if (mw->mw_curvdev >= mw->mw_numvdevs) 1345 return (WALK_DONE); 1346 1347 if (mw->mw_mss == NULL) { 1348 uintptr_t mssp; 1349 uintptr_t vdevp; 1350 1351 ASSERT(mw->mw_curms == 0); 1352 ASSERT(mw->mw_nummss == 0); 1353 1354 vdevp = mw->mw_vdevs[mw->mw_curvdev]; 1355 if (GETMEMB(vdevp, "vdev", vdev_ms, mssp) || 1356 GETMEMB(vdevp, "vdev", vdev_ms_count, mw->mw_nummss)) { 1357 return (WALK_ERR); 1358 } 1359 1360 mw->mw_mss = mdb_alloc(mw->mw_nummss * sizeof (void*), 1361 UM_SLEEP | UM_GC); 1362 if (mdb_vread(mw->mw_mss, mw->mw_nummss * sizeof (void*), 1363 mssp) == -1) { 1364 mdb_warn("failed to read vdev_ms at %p", mssp); 1365 return (WALK_ERR); 1366 } 1367 } 1368 1369 if (mw->mw_curms >= mw->mw_nummss) { 1370 mw->mw_mss = NULL; 1371 mw->mw_curms = 0; 1372 mw->mw_nummss = 0; 1373 mw->mw_curvdev++; 1374 return (WALK_NEXT); 1375 } 1376 1377 msp = mw->mw_mss[mw->mw_curms]; 1378 if (mdb_vread(&ms, sizeof (metaslab_t), msp) == -1) { 1379 mdb_warn("failed to read metaslab_t at %p", msp); 1380 return (WALK_ERR); 1381 } 1382 1383 mw->mw_curms++; 1384 1385 return (wsp->walk_callback(msp, &ms, wsp->walk_cbdata)); 1386 } 1387 1388 /* ARGSUSED */ 1389 static int 1390 metaslab_walk_init(mdb_walk_state_t *wsp) 1391 { 1392 metaslab_walk_data_t *mw; 1393 uintptr_t root_vdevp; 1394 uintptr_t childp; 1395 1396 if (wsp->walk_addr == NULL) { 1397 mdb_warn("must supply address of spa_t\n"); 1398 return (WALK_ERR); 1399 } 1400 1401 mw = mdb_zalloc(sizeof (metaslab_walk_data_t), UM_SLEEP | UM_GC); 1402 1403 if (GETMEMB(wsp->walk_addr, "spa", spa_root_vdev, root_vdevp) || 1404 GETMEMB(root_vdevp, "vdev", vdev_children, mw->mw_numvdevs) || 1405 GETMEMB(root_vdevp, "vdev", vdev_child, childp)) { 1406 return (DCMD_ERR); 1407 } 1408 1409 mw->mw_vdevs = mdb_alloc(mw->mw_numvdevs * sizeof (void *), 1410 UM_SLEEP | UM_GC); 1411 if (mdb_vread(mw->mw_vdevs, mw->mw_numvdevs * sizeof (void *), 1412 childp) == -1) { 1413 mdb_warn("failed to read root vdev children at %p", childp); 1414 return (DCMD_ERR); 1415 } 1416 1417 wsp->walk_data = mw; 1418 1419 return (WALK_NEXT); 1420 } 1421 1422 typedef struct mdb_spa { 1423 uintptr_t spa_dsl_pool; 1424 uintptr_t spa_root_vdev; 1425 } mdb_spa_t; 1426 1427 typedef struct mdb_dsl_dir { 1428 uintptr_t dd_phys; 1429 int64_t dd_space_towrite[TXG_SIZE]; 1430 } mdb_dsl_dir_t; 1431 1432 typedef struct mdb_dsl_dir_phys { 1433 uint64_t dd_used_bytes; 1434 uint64_t dd_compressed_bytes; 1435 uint64_t dd_uncompressed_bytes; 1436 } mdb_dsl_dir_phys_t; 1437 1438 typedef struct mdb_vdev { 1439 uintptr_t vdev_parent; 1440 uintptr_t vdev_ms; 1441 uint64_t vdev_ms_count; 1442 vdev_stat_t vdev_stat; 1443 } mdb_vdev_t; 1444 1445 typedef struct mdb_metaslab { 1446 space_map_t ms_allocmap[TXG_SIZE]; 1447 space_map_t ms_freemap[TXG_SIZE]; 1448 space_map_t ms_map; 1449 space_map_obj_t ms_smo; 1450 space_map_obj_t ms_smo_syncing; 1451 } mdb_metaslab_t; 1452 1453 typedef struct space_data { 1454 uint64_t ms_allocmap[TXG_SIZE]; 1455 uint64_t ms_freemap[TXG_SIZE]; 1456 uint64_t ms_map; 1457 uint64_t avail; 1458 uint64_t nowavail; 1459 } space_data_t; 1460 1461 /* ARGSUSED */ 1462 static int 1463 space_cb(uintptr_t addr, const void *unknown, void *arg) 1464 { 1465 space_data_t *sd = arg; 1466 mdb_metaslab_t ms; 1467 1468 if (GETMEMB(addr, "metaslab", ms_allocmap, ms.ms_allocmap) || 1469 GETMEMB(addr, "metaslab", ms_freemap, ms.ms_freemap) || 1470 GETMEMB(addr, "metaslab", ms_map, ms.ms_map) || 1471 GETMEMB(addr, "metaslab", ms_smo, ms.ms_smo) || 1472 GETMEMB(addr, "metaslab", ms_smo_syncing, ms.ms_smo_syncing)) { 1473 return (WALK_ERR); 1474 } 1475 1476 sd->ms_allocmap[0] += ms.ms_allocmap[0].sm_space; 1477 sd->ms_allocmap[1] += ms.ms_allocmap[1].sm_space; 1478 sd->ms_allocmap[2] += ms.ms_allocmap[2].sm_space; 1479 sd->ms_allocmap[3] += ms.ms_allocmap[3].sm_space; 1480 sd->ms_freemap[0] += ms.ms_freemap[0].sm_space; 1481 sd->ms_freemap[1] += ms.ms_freemap[1].sm_space; 1482 sd->ms_freemap[2] += ms.ms_freemap[2].sm_space; 1483 sd->ms_freemap[3] += ms.ms_freemap[3].sm_space; 1484 sd->ms_map += ms.ms_map.sm_space; 1485 sd->avail += ms.ms_map.sm_size - ms.ms_smo.smo_alloc; 1486 sd->nowavail += ms.ms_map.sm_size - ms.ms_smo_syncing.smo_alloc; 1487 1488 return (WALK_NEXT); 1489 } 1490 1491 /* 1492 * ::spa_space [-b] 1493 * 1494 * Given a spa_t, print out it's on-disk space usage and in-core 1495 * estimates of future usage. If -b is given, print space in bytes. 1496 * Otherwise print in megabytes. 1497 */ 1498 /* ARGSUSED */ 1499 static int 1500 spa_space(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv) 1501 { 1502 mdb_spa_t spa; 1503 uintptr_t dp_root_dir; 1504 mdb_dsl_dir_t dd; 1505 mdb_dsl_dir_phys_t dsp; 1506 uint64_t children; 1507 uintptr_t childaddr; 1508 space_data_t sd; 1509 int shift = 20; 1510 char *suffix = "M"; 1511 int bytes = B_FALSE; 1512 1513 if (mdb_getopts(argc, argv, 'b', MDB_OPT_SETBITS, TRUE, &bytes, NULL) != 1514 argc) 1515 return (DCMD_USAGE); 1516 if (!(flags & DCMD_ADDRSPEC)) 1517 return (DCMD_USAGE); 1518 1519 if (bytes) { 1520 shift = 0; 1521 suffix = ""; 1522 } 1523 1524 if (GETMEMB(addr, "spa", spa_dsl_pool, spa.spa_dsl_pool) || 1525 GETMEMB(addr, "spa", spa_root_vdev, spa.spa_root_vdev) || 1526 GETMEMB(spa.spa_root_vdev, "vdev", vdev_children, children) || 1527 GETMEMB(spa.spa_root_vdev, "vdev", vdev_child, childaddr) || 1528 GETMEMB(spa.spa_dsl_pool, "dsl_pool", 1529 dp_root_dir, dp_root_dir) || 1530 GETMEMB(dp_root_dir, "dsl_dir", dd_phys, dd.dd_phys) || 1531 GETMEMB(dp_root_dir, "dsl_dir", 1532 dd_space_towrite, dd.dd_space_towrite) || 1533 GETMEMB(dd.dd_phys, "dsl_dir_phys", 1534 dd_used_bytes, dsp.dd_used_bytes) || 1535 GETMEMB(dd.dd_phys, "dsl_dir_phys", 1536 dd_compressed_bytes, dsp.dd_compressed_bytes) || 1537 GETMEMB(dd.dd_phys, "dsl_dir_phys", 1538 dd_uncompressed_bytes, dsp.dd_uncompressed_bytes)) { 1539 return (DCMD_ERR); 1540 } 1541 1542 mdb_printf("dd_space_towrite = %llu%s %llu%s %llu%s %llu%s\n", 1543 dd.dd_space_towrite[0] >> shift, suffix, 1544 dd.dd_space_towrite[1] >> shift, suffix, 1545 dd.dd_space_towrite[2] >> shift, suffix, 1546 dd.dd_space_towrite[3] >> shift, suffix); 1547 1548 mdb_printf("dd_phys.dd_used_bytes = %llu%s\n", 1549 dsp.dd_used_bytes >> shift, suffix); 1550 mdb_printf("dd_phys.dd_compressed_bytes = %llu%s\n", 1551 dsp.dd_compressed_bytes >> shift, suffix); 1552 mdb_printf("dd_phys.dd_uncompressed_bytes = %llu%s\n", 1553 dsp.dd_uncompressed_bytes >> shift, suffix); 1554 1555 bzero(&sd, sizeof (sd)); 1556 if (mdb_pwalk("metaslab", space_cb, &sd, addr) != 0) { 1557 mdb_warn("can't walk metaslabs"); 1558 return (DCMD_ERR); 1559 } 1560 1561 mdb_printf("ms_allocmap = %llu%s %llu%s %llu%s %llu%s\n", 1562 sd.ms_allocmap[0] >> shift, suffix, 1563 sd.ms_allocmap[1] >> shift, suffix, 1564 sd.ms_allocmap[2] >> shift, suffix, 1565 sd.ms_allocmap[3] >> shift, suffix); 1566 mdb_printf("ms_freemap = %llu%s %llu%s %llu%s %llu%s\n", 1567 sd.ms_freemap[0] >> shift, suffix, 1568 sd.ms_freemap[1] >> shift, suffix, 1569 sd.ms_freemap[2] >> shift, suffix, 1570 sd.ms_freemap[3] >> shift, suffix); 1571 mdb_printf("ms_map = %llu%s\n", sd.ms_map >> shift, suffix); 1572 mdb_printf("last synced avail = %llu%s\n", sd.avail >> shift, suffix); 1573 mdb_printf("current syncing avail = %llu%s\n", 1574 sd.nowavail >> shift, suffix); 1575 1576 return (DCMD_OK); 1577 } 1578 1579 /* 1580 * ::spa_verify 1581 * 1582 * Given a spa_t, verify that that the pool is self-consistent. 1583 * Currently, it only checks to make sure that the vdev tree exists. 1584 */ 1585 /* ARGSUSED */ 1586 static int 1587 spa_verify(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv) 1588 { 1589 spa_t spa; 1590 1591 if (argc != 0 || !(flags & DCMD_ADDRSPEC)) 1592 return (DCMD_USAGE); 1593 1594 if (mdb_vread(&spa, sizeof (spa), addr) == -1) { 1595 mdb_warn("failed to read spa_t at %p", addr); 1596 return (DCMD_ERR); 1597 } 1598 1599 if (spa.spa_root_vdev == NULL) { 1600 mdb_printf("no vdev tree present\n"); 1601 return (DCMD_OK); 1602 } 1603 1604 return (DCMD_OK); 1605 } 1606 1607 static int 1608 spa_print_aux(spa_aux_vdev_t *sav, uint_t flags, mdb_arg_t *v, 1609 const char *name) 1610 { 1611 uintptr_t *aux; 1612 size_t len; 1613 int ret, i; 1614 1615 /* 1616 * Iterate over aux vdevs and print those out as well. This is a 1617 * little annoying because we don't have a root vdev to pass to ::vdev. 1618 * Instead, we print a single line and then call it for each child 1619 * vdev. 1620 */ 1621 if (sav->sav_count != 0) { 1622 v[1].a_type = MDB_TYPE_STRING; 1623 v[1].a_un.a_str = "-d"; 1624 v[2].a_type = MDB_TYPE_IMMEDIATE; 1625 v[2].a_un.a_val = 2; 1626 1627 len = sav->sav_count * sizeof (uintptr_t); 1628 aux = mdb_alloc(len, UM_SLEEP); 1629 if (mdb_vread(aux, len, 1630 (uintptr_t)sav->sav_vdevs) == -1) { 1631 mdb_free(aux, len); 1632 mdb_warn("failed to read l2cache vdevs at %p", 1633 sav->sav_vdevs); 1634 return (DCMD_ERR); 1635 } 1636 1637 mdb_printf("%-?s %-9s %-12s %s\n", "-", "-", "-", name); 1638 1639 for (i = 0; i < sav->sav_count; i++) { 1640 ret = mdb_call_dcmd("vdev", aux[i], flags, 3, v); 1641 if (ret != DCMD_OK) { 1642 mdb_free(aux, len); 1643 return (ret); 1644 } 1645 } 1646 1647 mdb_free(aux, len); 1648 } 1649 1650 return (0); 1651 } 1652 1653 /* 1654 * ::spa_vdevs 1655 * 1656 * -e Include error stats 1657 * 1658 * Print out a summarized list of vdevs for the given spa_t. 1659 * This is accomplished by invoking "::vdev -re" on the root vdev, as well as 1660 * iterating over the cache devices. 1661 */ 1662 /* ARGSUSED */ 1663 static int 1664 spa_vdevs(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv) 1665 { 1666 spa_t spa; 1667 mdb_arg_t v[3]; 1668 int errors = FALSE; 1669 int ret; 1670 1671 if (mdb_getopts(argc, argv, 1672 'e', MDB_OPT_SETBITS, TRUE, &errors, 1673 NULL) != argc) 1674 return (DCMD_USAGE); 1675 1676 if (!(flags & DCMD_ADDRSPEC)) 1677 return (DCMD_USAGE); 1678 1679 if (mdb_vread(&spa, sizeof (spa), addr) == -1) { 1680 mdb_warn("failed to read spa_t at %p", addr); 1681 return (DCMD_ERR); 1682 } 1683 1684 /* 1685 * Unitialized spa_t structures can have a NULL root vdev. 1686 */ 1687 if (spa.spa_root_vdev == NULL) { 1688 mdb_printf("no associated vdevs\n"); 1689 return (DCMD_OK); 1690 } 1691 1692 v[0].a_type = MDB_TYPE_STRING; 1693 v[0].a_un.a_str = errors ? "-re" : "-r"; 1694 1695 ret = mdb_call_dcmd("vdev", (uintptr_t)spa.spa_root_vdev, 1696 flags, 1, v); 1697 if (ret != DCMD_OK) 1698 return (ret); 1699 1700 if (spa_print_aux(&spa.spa_l2cache, flags, v, "cache") != 0 || 1701 spa_print_aux(&spa.spa_spares, flags, v, "spares") != 0) 1702 return (DCMD_ERR); 1703 1704 return (DCMD_OK); 1705 } 1706 1707 /* 1708 * ::zio 1709 * 1710 * Print a summary of zio_t and all its children. This is intended to display a 1711 * zio tree, and hence we only pick the most important pieces of information for 1712 * the main summary. More detailed information can always be found by doing a 1713 * '::print zio' on the underlying zio_t. The columns we display are: 1714 * 1715 * ADDRESS TYPE STAGE WAITER 1716 * 1717 * The 'address' column is indented by one space for each depth level as we 1718 * descend down the tree. 1719 */ 1720 1721 #define ZIO_MAXINDENT 24 1722 #define ZIO_MAXWIDTH (sizeof (uintptr_t) * 2 + ZIO_MAXINDENT) 1723 #define ZIO_WALK_SELF 0 1724 #define ZIO_WALK_CHILD 1 1725 #define ZIO_WALK_PARENT 2 1726 1727 typedef struct zio_print_args { 1728 int zpa_current_depth; 1729 int zpa_min_depth; 1730 int zpa_max_depth; 1731 int zpa_type; 1732 uint_t zpa_flags; 1733 } zio_print_args_t; 1734 1735 typedef struct mdb_zio { 1736 enum zio_type io_type; 1737 enum zio_stage io_stage; 1738 void *io_waiter; 1739 uint64_t io_timestamp; 1740 void *io_spa; 1741 struct { 1742 struct { 1743 void *list_next; 1744 } list_head; 1745 } io_parent_list; 1746 int io_error; 1747 } mdb_zio_t; 1748 1749 static int zio_child_cb(uintptr_t addr, const void *unknown, void *arg); 1750 1751 static int 1752 zio_print_cb(uintptr_t addr, const mdb_zio_t *zio, zio_print_args_t *zpa) 1753 { 1754 mdb_ctf_id_t type_enum, stage_enum; 1755 int indent = zpa->zpa_current_depth; 1756 const char *type, *stage; 1757 uintptr_t laddr; 1758 1759 if (indent > ZIO_MAXINDENT) 1760 indent = ZIO_MAXINDENT; 1761 1762 if (mdb_ctf_lookup_by_name("enum zio_type", &type_enum) == -1 || 1763 mdb_ctf_lookup_by_name("enum zio_stage", &stage_enum) == -1) { 1764 mdb_warn("failed to lookup zio enums"); 1765 return (WALK_ERR); 1766 } 1767 1768 if ((type = mdb_ctf_enum_name(type_enum, zio->io_type)) != NULL) 1769 type += sizeof ("ZIO_TYPE_") - 1; 1770 else 1771 type = "?"; 1772 1773 if (zio->io_error == 0) { 1774 stage = mdb_ctf_enum_name(stage_enum, zio->io_stage); 1775 if (stage != NULL) 1776 stage += sizeof ("ZIO_STAGE_") - 1; 1777 else 1778 stage = "?"; 1779 } else { 1780 stage = "FAILED"; 1781 } 1782 1783 if (zpa->zpa_current_depth >= zpa->zpa_min_depth) { 1784 if (zpa->zpa_flags & DCMD_PIPE_OUT) { 1785 mdb_printf("%?p\n", addr); 1786 } else { 1787 mdb_printf("%*s%-*p %-5s %-16s ", indent, "", 1788 ZIO_MAXWIDTH - indent, addr, type, stage); 1789 if (zio->io_waiter) 1790 mdb_printf("%?p\n", zio->io_waiter); 1791 else 1792 mdb_printf("-\n"); 1793 } 1794 } 1795 1796 if (zpa->zpa_current_depth >= zpa->zpa_max_depth) 1797 return (WALK_NEXT); 1798 1799 if (zpa->zpa_type == ZIO_WALK_PARENT) 1800 laddr = addr + mdb_ctf_offsetof_by_name(ZFS_STRUCT "zio", 1801 "io_parent_list"); 1802 else 1803 laddr = addr + mdb_ctf_offsetof_by_name(ZFS_STRUCT "zio", 1804 "io_child_list"); 1805 1806 zpa->zpa_current_depth++; 1807 if (mdb_pwalk("list", zio_child_cb, zpa, laddr) != 0) { 1808 mdb_warn("failed to walk zio_t children at %p\n", laddr); 1809 return (WALK_ERR); 1810 } 1811 zpa->zpa_current_depth--; 1812 1813 return (WALK_NEXT); 1814 } 1815 1816 /* ARGSUSED */ 1817 static int 1818 zio_child_cb(uintptr_t addr, const void *unknown, void *arg) 1819 { 1820 zio_link_t zl; 1821 mdb_zio_t zio; 1822 uintptr_t ziop; 1823 zio_print_args_t *zpa = arg; 1824 1825 if (mdb_vread(&zl, sizeof (zl), addr) == -1) { 1826 mdb_warn("failed to read zio_link_t at %p", addr); 1827 return (WALK_ERR); 1828 } 1829 1830 if (zpa->zpa_type == ZIO_WALK_PARENT) 1831 ziop = (uintptr_t)zl.zl_parent; 1832 else 1833 ziop = (uintptr_t)zl.zl_child; 1834 1835 if (mdb_ctf_vread(&zio, ZFS_STRUCT "zio", "mdb_zio_t", ziop, 0) == -1) 1836 return (WALK_ERR); 1837 1838 return (zio_print_cb(ziop, &zio, arg)); 1839 } 1840 1841 /* ARGSUSED */ 1842 static int 1843 zio_print(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv) 1844 { 1845 mdb_zio_t zio; 1846 zio_print_args_t zpa = { 0 }; 1847 1848 if (!(flags & DCMD_ADDRSPEC)) 1849 return (DCMD_USAGE); 1850 1851 if (mdb_getopts(argc, argv, 1852 'r', MDB_OPT_SETBITS, INT_MAX, &zpa.zpa_max_depth, 1853 'c', MDB_OPT_SETBITS, ZIO_WALK_CHILD, &zpa.zpa_type, 1854 'p', MDB_OPT_SETBITS, ZIO_WALK_PARENT, &zpa.zpa_type, 1855 NULL) != argc) 1856 return (DCMD_USAGE); 1857 1858 zpa.zpa_flags = flags; 1859 if (zpa.zpa_max_depth != 0) { 1860 if (zpa.zpa_type == ZIO_WALK_SELF) 1861 zpa.zpa_type = ZIO_WALK_CHILD; 1862 } else if (zpa.zpa_type != ZIO_WALK_SELF) { 1863 zpa.zpa_min_depth = 1; 1864 zpa.zpa_max_depth = 1; 1865 } 1866 1867 if (mdb_ctf_vread(&zio, ZFS_STRUCT "zio", "mdb_zio_t", addr, 0) == -1) 1868 return (DCMD_ERR); 1869 1870 if (!(flags & DCMD_PIPE_OUT) && DCMD_HDRSPEC(flags)) 1871 mdb_printf("%<u>%-*s %-5s %-16s %-?s%</u>\n", ZIO_MAXWIDTH, 1872 "ADDRESS", "TYPE", "STAGE", "WAITER"); 1873 1874 if (zio_print_cb(addr, &zio, &zpa) != WALK_NEXT) 1875 return (DCMD_ERR); 1876 1877 return (DCMD_OK); 1878 } 1879 1880 /* 1881 * [addr]::zio_state 1882 * 1883 * Print a summary of all zio_t structures on the system, or for a particular 1884 * pool. This is equivalent to '::walk zio_root | ::zio'. 1885 */ 1886 /*ARGSUSED*/ 1887 static int 1888 zio_state(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv) 1889 { 1890 /* 1891 * MDB will remember the last address of the pipeline, so if we don't 1892 * zero this we'll end up trying to walk zio structures for a 1893 * non-existent spa_t. 1894 */ 1895 if (!(flags & DCMD_ADDRSPEC)) 1896 addr = 0; 1897 1898 return (mdb_pwalk_dcmd("zio_root", "zio", argc, argv, addr)); 1899 } 1900 1901 typedef struct txg_list_walk_data { 1902 uintptr_t lw_head[TXG_SIZE]; 1903 int lw_txgoff; 1904 int lw_maxoff; 1905 size_t lw_offset; 1906 void *lw_obj; 1907 } txg_list_walk_data_t; 1908 1909 static int 1910 txg_list_walk_init_common(mdb_walk_state_t *wsp, int txg, int maxoff) 1911 { 1912 txg_list_walk_data_t *lwd; 1913 txg_list_t list; 1914 int i; 1915 1916 lwd = mdb_alloc(sizeof (txg_list_walk_data_t), UM_SLEEP | UM_GC); 1917 if (mdb_vread(&list, sizeof (txg_list_t), wsp->walk_addr) == -1) { 1918 mdb_warn("failed to read txg_list_t at %#lx", wsp->walk_addr); 1919 return (WALK_ERR); 1920 } 1921 1922 for (i = 0; i < TXG_SIZE; i++) 1923 lwd->lw_head[i] = (uintptr_t)list.tl_head[i]; 1924 lwd->lw_offset = list.tl_offset; 1925 lwd->lw_obj = mdb_alloc(lwd->lw_offset + sizeof (txg_node_t), 1926 UM_SLEEP | UM_GC); 1927 lwd->lw_txgoff = txg; 1928 lwd->lw_maxoff = maxoff; 1929 1930 wsp->walk_addr = lwd->lw_head[lwd->lw_txgoff]; 1931 wsp->walk_data = lwd; 1932 1933 return (WALK_NEXT); 1934 } 1935 1936 static int 1937 txg_list_walk_init(mdb_walk_state_t *wsp) 1938 { 1939 return (txg_list_walk_init_common(wsp, 0, TXG_SIZE-1)); 1940 } 1941 1942 static int 1943 txg_list0_walk_init(mdb_walk_state_t *wsp) 1944 { 1945 return (txg_list_walk_init_common(wsp, 0, 0)); 1946 } 1947 1948 static int 1949 txg_list1_walk_init(mdb_walk_state_t *wsp) 1950 { 1951 return (txg_list_walk_init_common(wsp, 1, 1)); 1952 } 1953 1954 static int 1955 txg_list2_walk_init(mdb_walk_state_t *wsp) 1956 { 1957 return (txg_list_walk_init_common(wsp, 2, 2)); 1958 } 1959 1960 static int 1961 txg_list3_walk_init(mdb_walk_state_t *wsp) 1962 { 1963 return (txg_list_walk_init_common(wsp, 3, 3)); 1964 } 1965 1966 static int 1967 txg_list_walk_step(mdb_walk_state_t *wsp) 1968 { 1969 txg_list_walk_data_t *lwd = wsp->walk_data; 1970 uintptr_t addr; 1971 txg_node_t *node; 1972 int status; 1973 1974 while (wsp->walk_addr == NULL && lwd->lw_txgoff < lwd->lw_maxoff) { 1975 lwd->lw_txgoff++; 1976 wsp->walk_addr = lwd->lw_head[lwd->lw_txgoff]; 1977 } 1978 1979 if (wsp->walk_addr == NULL) 1980 return (WALK_DONE); 1981 1982 addr = wsp->walk_addr - lwd->lw_offset; 1983 1984 if (mdb_vread(lwd->lw_obj, 1985 lwd->lw_offset + sizeof (txg_node_t), addr) == -1) { 1986 mdb_warn("failed to read list element at %#lx", addr); 1987 return (WALK_ERR); 1988 } 1989 1990 status = wsp->walk_callback(addr, lwd->lw_obj, wsp->walk_cbdata); 1991 node = (txg_node_t *)((uintptr_t)lwd->lw_obj + lwd->lw_offset); 1992 wsp->walk_addr = (uintptr_t)node->tn_next[lwd->lw_txgoff]; 1993 1994 return (status); 1995 } 1996 1997 /* 1998 * ::walk spa 1999 * 2000 * Walk all named spa_t structures in the namespace. This is nothing more than 2001 * a layered avl walk. 2002 */ 2003 static int 2004 spa_walk_init(mdb_walk_state_t *wsp) 2005 { 2006 GElf_Sym sym; 2007 2008 if (wsp->walk_addr != NULL) { 2009 mdb_warn("spa walk only supports global walks\n"); 2010 return (WALK_ERR); 2011 } 2012 2013 if (mdb_lookup_by_obj(ZFS_OBJ_NAME, "spa_namespace_avl", &sym) == -1) { 2014 mdb_warn("failed to find symbol 'spa_namespace_avl'"); 2015 return (WALK_ERR); 2016 } 2017 2018 wsp->walk_addr = (uintptr_t)sym.st_value; 2019 2020 if (mdb_layered_walk("avl", wsp) == -1) { 2021 mdb_warn("failed to walk 'avl'\n"); 2022 return (WALK_ERR); 2023 } 2024 2025 return (WALK_NEXT); 2026 } 2027 2028 static int 2029 spa_walk_step(mdb_walk_state_t *wsp) 2030 { 2031 spa_t spa; 2032 2033 if (mdb_vread(&spa, sizeof (spa), wsp->walk_addr) == -1) { 2034 mdb_warn("failed to read spa_t at %p", wsp->walk_addr); 2035 return (WALK_ERR); 2036 } 2037 2038 return (wsp->walk_callback(wsp->walk_addr, &spa, wsp->walk_cbdata)); 2039 } 2040 2041 /* 2042 * [addr]::walk zio 2043 * 2044 * Walk all active zio_t structures on the system. This is simply a layered 2045 * walk on top of ::walk zio_cache, with the optional ability to limit the 2046 * structures to a particular pool. 2047 */ 2048 static int 2049 zio_walk_init(mdb_walk_state_t *wsp) 2050 { 2051 wsp->walk_data = (void *)wsp->walk_addr; 2052 2053 if (mdb_layered_walk("zio_cache", wsp) == -1) { 2054 mdb_warn("failed to walk 'zio_cache'\n"); 2055 return (WALK_ERR); 2056 } 2057 2058 return (WALK_NEXT); 2059 } 2060 2061 static int 2062 zio_walk_step(mdb_walk_state_t *wsp) 2063 { 2064 mdb_zio_t zio; 2065 2066 if (mdb_ctf_vread(&zio, ZFS_STRUCT "zio", "mdb_zio_t", 2067 wsp->walk_addr, 0) == -1) 2068 return (WALK_ERR); 2069 2070 if (wsp->walk_data != NULL && wsp->walk_data != zio.io_spa) 2071 return (WALK_NEXT); 2072 2073 return (wsp->walk_callback(wsp->walk_addr, &zio, wsp->walk_cbdata)); 2074 } 2075 2076 /* 2077 * [addr]::walk zio_root 2078 * 2079 * Walk only root zio_t structures, optionally for a particular spa_t. 2080 */ 2081 static int 2082 zio_walk_root_step(mdb_walk_state_t *wsp) 2083 { 2084 mdb_zio_t zio; 2085 2086 if (mdb_ctf_vread(&zio, ZFS_STRUCT "zio", "mdb_zio_t", 2087 wsp->walk_addr, 0) == -1) 2088 return (WALK_ERR); 2089 2090 if (wsp->walk_data != NULL && wsp->walk_data != zio.io_spa) 2091 return (WALK_NEXT); 2092 2093 /* If the parent list is not empty, ignore */ 2094 if ((uintptr_t)zio.io_parent_list.list_head.list_next != 2095 wsp->walk_addr + 2096 mdb_ctf_offsetof_by_name(ZFS_STRUCT "zio", "io_parent_list") + 2097 mdb_ctf_offsetof_by_name("struct list", "list_head")) 2098 return (WALK_NEXT); 2099 2100 return (wsp->walk_callback(wsp->walk_addr, &zio, wsp->walk_cbdata)); 2101 } 2102 2103 #define NICENUM_BUFLEN 6 2104 2105 static int 2106 snprintfrac(char *buf, int len, 2107 uint64_t numerator, uint64_t denom, int frac_digits) 2108 { 2109 int mul = 1; 2110 int whole, frac, i; 2111 2112 for (i = frac_digits; i; i--) 2113 mul *= 10; 2114 whole = numerator / denom; 2115 frac = mul * numerator / denom - mul * whole; 2116 return (mdb_snprintf(buf, len, "%u.%0*u", whole, frac_digits, frac)); 2117 } 2118 2119 static void 2120 mdb_nicenum(uint64_t num, char *buf) 2121 { 2122 uint64_t n = num; 2123 int index = 0; 2124 char *u; 2125 2126 while (n >= 1024) { 2127 n = (n + (1024 / 2)) / 1024; /* Round up or down */ 2128 index++; 2129 } 2130 2131 u = &" \0K\0M\0G\0T\0P\0E\0"[index*2]; 2132 2133 if (index == 0) { 2134 (void) mdb_snprintf(buf, NICENUM_BUFLEN, "%llu", 2135 (u_longlong_t)n); 2136 } else if (n < 10 && (num & (num - 1)) != 0) { 2137 (void) snprintfrac(buf, NICENUM_BUFLEN, 2138 num, 1ULL << 10 * index, 2); 2139 strcat(buf, u); 2140 } else if (n < 100 && (num & (num - 1)) != 0) { 2141 (void) snprintfrac(buf, NICENUM_BUFLEN, 2142 num, 1ULL << 10 * index, 1); 2143 strcat(buf, u); 2144 } else { 2145 (void) mdb_snprintf(buf, NICENUM_BUFLEN, "%llu%s", 2146 (u_longlong_t)n, u); 2147 } 2148 } 2149 2150 /* 2151 * ::zfs_blkstats 2152 * 2153 * -v print verbose per-level information 2154 * 2155 */ 2156 static int 2157 zfs_blkstats(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv) 2158 { 2159 boolean_t verbose = B_FALSE; 2160 zfs_all_blkstats_t stats; 2161 dmu_object_type_t t; 2162 zfs_blkstat_t *tzb; 2163 uint64_t ditto; 2164 dmu_object_type_info_t dmu_ot[DMU_OT_NUMTYPES + 10]; 2165 /* +10 in case it grew */ 2166 2167 if (mdb_readvar(&dmu_ot, "dmu_ot") == -1) { 2168 mdb_warn("failed to read 'dmu_ot'"); 2169 return (DCMD_ERR); 2170 } 2171 2172 if (mdb_getopts(argc, argv, 2173 'v', MDB_OPT_SETBITS, TRUE, &verbose, 2174 NULL) != argc) 2175 return (DCMD_USAGE); 2176 2177 if (!(flags & DCMD_ADDRSPEC)) 2178 return (DCMD_USAGE); 2179 2180 if (GETMEMB(addr, "spa", spa_dsl_pool, addr) || 2181 GETMEMB(addr, "dsl_pool", dp_blkstats, addr) || 2182 mdb_vread(&stats, sizeof (zfs_all_blkstats_t), addr) == -1) { 2183 mdb_warn("failed to read data at %p;", addr); 2184 mdb_printf("maybe no stats? run \"zpool scrub\" first."); 2185 return (DCMD_ERR); 2186 } 2187 2188 tzb = &stats.zab_type[DN_MAX_LEVELS][DMU_OT_TOTAL]; 2189 if (tzb->zb_gangs != 0) { 2190 mdb_printf("Ganged blocks: %llu\n", 2191 (longlong_t)tzb->zb_gangs); 2192 } 2193 2194 ditto = tzb->zb_ditto_2_of_2_samevdev + tzb->zb_ditto_2_of_3_samevdev + 2195 tzb->zb_ditto_3_of_3_samevdev; 2196 if (ditto != 0) { 2197 mdb_printf("Dittoed blocks on same vdev: %llu\n", 2198 (longlong_t)ditto); 2199 } 2200 2201 mdb_printf("\nBlocks\tLSIZE\tPSIZE\tASIZE" 2202 "\t avg\t comp\t%%Total\tType\n"); 2203 2204 for (t = 0; t <= DMU_OT_TOTAL; t++) { 2205 char csize[NICENUM_BUFLEN], lsize[NICENUM_BUFLEN]; 2206 char psize[NICENUM_BUFLEN], asize[NICENUM_BUFLEN]; 2207 char avg[NICENUM_BUFLEN]; 2208 char comp[NICENUM_BUFLEN], pct[NICENUM_BUFLEN]; 2209 char typename[64]; 2210 int l; 2211 2212 2213 if (t == DMU_OT_DEFERRED) 2214 strcpy(typename, "deferred free"); 2215 else if (t == DMU_OT_OTHER) 2216 strcpy(typename, "other"); 2217 else if (t == DMU_OT_TOTAL) 2218 strcpy(typename, "Total"); 2219 else if (mdb_readstr(typename, sizeof (typename), 2220 (uintptr_t)dmu_ot[t].ot_name) == -1) { 2221 mdb_warn("failed to read type name"); 2222 return (DCMD_ERR); 2223 } 2224 2225 if (stats.zab_type[DN_MAX_LEVELS][t].zb_asize == 0) 2226 continue; 2227 2228 for (l = -1; l < DN_MAX_LEVELS; l++) { 2229 int level = (l == -1 ? DN_MAX_LEVELS : l); 2230 zfs_blkstat_t *zb = &stats.zab_type[level][t]; 2231 2232 if (zb->zb_asize == 0) 2233 continue; 2234 2235 /* 2236 * Don't print each level unless requested. 2237 */ 2238 if (!verbose && level != DN_MAX_LEVELS) 2239 continue; 2240 2241 /* 2242 * If all the space is level 0, don't print the 2243 * level 0 separately. 2244 */ 2245 if (level == 0 && zb->zb_asize == 2246 stats.zab_type[DN_MAX_LEVELS][t].zb_asize) 2247 continue; 2248 2249 mdb_nicenum(zb->zb_count, csize); 2250 mdb_nicenum(zb->zb_lsize, lsize); 2251 mdb_nicenum(zb->zb_psize, psize); 2252 mdb_nicenum(zb->zb_asize, asize); 2253 mdb_nicenum(zb->zb_asize / zb->zb_count, avg); 2254 (void) snprintfrac(comp, NICENUM_BUFLEN, 2255 zb->zb_lsize, zb->zb_psize, 2); 2256 (void) snprintfrac(pct, NICENUM_BUFLEN, 2257 100 * zb->zb_asize, tzb->zb_asize, 2); 2258 2259 mdb_printf("%6s\t%5s\t%5s\t%5s\t%5s" 2260 "\t%5s\t%6s\t", 2261 csize, lsize, psize, asize, avg, comp, pct); 2262 2263 if (level == DN_MAX_LEVELS) 2264 mdb_printf("%s\n", typename); 2265 else 2266 mdb_printf(" L%d %s\n", 2267 level, typename); 2268 } 2269 } 2270 2271 return (DCMD_OK); 2272 } 2273 2274 /* ARGSUSED */ 2275 static int 2276 reference_cb(uintptr_t addr, const void *ignored, void *arg) 2277 { 2278 static int gotid; 2279 static mdb_ctf_id_t ref_id; 2280 uintptr_t ref_holder; 2281 uintptr_t ref_removed; 2282 uint64_t ref_number; 2283 boolean_t holder_is_str = B_FALSE; 2284 char holder_str[128]; 2285 boolean_t removed = (boolean_t)arg; 2286 2287 if (!gotid) { 2288 if (mdb_ctf_lookup_by_name("struct reference", &ref_id) == -1) { 2289 mdb_warn("couldn't find struct reference"); 2290 return (WALK_ERR); 2291 } 2292 gotid = TRUE; 2293 } 2294 2295 if (GETMEMBID(addr, &ref_id, ref_holder, ref_holder) || 2296 GETMEMBID(addr, &ref_id, ref_removed, ref_removed) || 2297 GETMEMBID(addr, &ref_id, ref_number, ref_number)) 2298 return (WALK_ERR); 2299 2300 if (mdb_readstr(holder_str, sizeof (holder_str), ref_holder) != -1) 2301 holder_is_str = strisprint(holder_str); 2302 2303 if (removed) 2304 mdb_printf("removed "); 2305 mdb_printf("reference "); 2306 if (ref_number != 1) 2307 mdb_printf("with count=%llu ", ref_number); 2308 mdb_printf("with tag %p", (void*)ref_holder); 2309 if (holder_is_str) 2310 mdb_printf(" \"%s\"", holder_str); 2311 mdb_printf(", held at:\n"); 2312 2313 (void) mdb_call_dcmd("whatis", addr, DCMD_ADDRSPEC, 0, NULL); 2314 2315 if (removed) { 2316 mdb_printf("removed at:\n"); 2317 (void) mdb_call_dcmd("whatis", ref_removed, 2318 DCMD_ADDRSPEC, 0, NULL); 2319 } 2320 2321 mdb_printf("\n"); 2322 2323 return (WALK_NEXT); 2324 } 2325 2326 /* ARGSUSED */ 2327 static int 2328 refcount(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv) 2329 { 2330 uint64_t rc_count, rc_removed_count; 2331 uintptr_t rc_list, rc_removed; 2332 static int gotid; 2333 static mdb_ctf_id_t rc_id; 2334 ulong_t off; 2335 boolean_t released = B_FALSE; 2336 2337 if (!(flags & DCMD_ADDRSPEC)) 2338 return (DCMD_USAGE); 2339 2340 if (mdb_getopts(argc, argv, 2341 'r', MDB_OPT_SETBITS, B_TRUE, &released, 2342 NULL) != argc) 2343 return (DCMD_USAGE); 2344 2345 if (!gotid) { 2346 /* 2347 * The refcount structure is different when compiled debug 2348 * vs nondebug. Therefore, we want to make sure we get the 2349 * refcount definition from the ZFS module, in case it has 2350 * been compiled debug but genunix is nondebug. 2351 */ 2352 if (mdb_ctf_lookup_by_name(ZFS_STRUCT "refcount", 2353 &rc_id) == -1) { 2354 mdb_warn("couldn't find struct refcount"); 2355 return (DCMD_ERR); 2356 } 2357 gotid = TRUE; 2358 } 2359 2360 if (GETMEMBID(addr, &rc_id, rc_count, rc_count)) 2361 return (DCMD_ERR); 2362 2363 if (GETMEMBID(addr, &rc_id, rc_removed_count, rc_removed_count)) { 2364 mdb_printf("refcount_t at %p has %llu current holds\n", 2365 addr, (longlong_t)rc_count); 2366 return (DCMD_OK); 2367 } 2368 2369 mdb_printf("refcount_t at %p has %llu current holds, " 2370 "%llu recently released holds\n", 2371 addr, (longlong_t)rc_count, (longlong_t)rc_removed_count); 2372 2373 if (rc_count > 0) 2374 mdb_printf("current holds:\n"); 2375 if (mdb_ctf_offsetof(rc_id, "rc_list", &off) == -1) 2376 return (DCMD_ERR); 2377 rc_list = addr + off/NBBY; 2378 mdb_pwalk("list", reference_cb, (void*)B_FALSE, rc_list); 2379 2380 if (released) { 2381 if (rc_removed_count > 0) 2382 mdb_printf("released holds:\n"); 2383 if (mdb_ctf_offsetof(rc_id, "rc_removed", &off) == -1) 2384 return (DCMD_ERR); 2385 rc_removed = addr + off/NBBY; 2386 mdb_pwalk("list", reference_cb, (void*)B_TRUE, rc_removed); 2387 } 2388 2389 return (DCMD_OK); 2390 } 2391 2392 /* ARGSUSED */ 2393 static int 2394 sa_attr_table(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv) 2395 { 2396 sa_attr_table_t *table; 2397 sa_os_t sa_os; 2398 char *name; 2399 int i; 2400 2401 if (mdb_vread(&sa_os, sizeof (sa_os_t), addr) == -1) { 2402 mdb_warn("failed to read sa_os at %p", addr); 2403 return (DCMD_ERR); 2404 } 2405 2406 table = mdb_alloc(sizeof (sa_attr_table_t) * sa_os.sa_num_attrs, 2407 UM_SLEEP | UM_GC); 2408 name = mdb_alloc(MAXPATHLEN, UM_SLEEP | UM_GC); 2409 2410 if (mdb_vread(table, sizeof (sa_attr_table_t) * sa_os.sa_num_attrs, 2411 (uintptr_t)sa_os.sa_attr_table) == -1) { 2412 mdb_warn("failed to read sa_os at %p", addr); 2413 return (DCMD_ERR); 2414 } 2415 2416 mdb_printf("%<u>%-10s %-10s %-10s %-10s %s%</u>\n", 2417 "ATTR ID", "REGISTERED", "LENGTH", "BSWAP", "NAME"); 2418 for (i = 0; i != sa_os.sa_num_attrs; i++) { 2419 mdb_readstr(name, MAXPATHLEN, (uintptr_t)table[i].sa_name); 2420 mdb_printf("%5x %8x %8x %8x %-s\n", 2421 (int)table[i].sa_attr, (int)table[i].sa_registered, 2422 (int)table[i].sa_length, table[i].sa_byteswap, name); 2423 } 2424 2425 return (DCMD_OK); 2426 } 2427 2428 static int 2429 sa_get_off_table(uintptr_t addr, uint32_t **off_tab, int attr_count) 2430 { 2431 uintptr_t idx_table; 2432 2433 if (GETMEMB(addr, "sa_idx_tab", sa_idx_tab, idx_table)) { 2434 mdb_printf("can't find offset table in sa_idx_tab\n"); 2435 return (-1); 2436 } 2437 2438 *off_tab = mdb_alloc(attr_count * sizeof (uint32_t), 2439 UM_SLEEP | UM_GC); 2440 2441 if (mdb_vread(*off_tab, 2442 attr_count * sizeof (uint32_t), idx_table) == -1) { 2443 mdb_warn("failed to attribute offset table %p", idx_table); 2444 return (-1); 2445 } 2446 2447 return (DCMD_OK); 2448 } 2449 2450 /*ARGSUSED*/ 2451 static int 2452 sa_attr_print(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv) 2453 { 2454 uint32_t *offset_tab; 2455 int attr_count; 2456 uint64_t attr_id; 2457 uintptr_t attr_addr; 2458 uintptr_t bonus_tab, spill_tab; 2459 uintptr_t db_bonus, db_spill; 2460 uintptr_t os, os_sa; 2461 uintptr_t db_data; 2462 2463 if (argc != 1) 2464 return (DCMD_USAGE); 2465 2466 if (argv[0].a_type == MDB_TYPE_STRING) 2467 attr_id = mdb_strtoull(argv[0].a_un.a_str); 2468 else 2469 return (DCMD_USAGE); 2470 2471 if (GETMEMB(addr, "sa_handle", sa_bonus_tab, bonus_tab) || 2472 GETMEMB(addr, "sa_handle", sa_spill_tab, spill_tab) || 2473 GETMEMB(addr, "sa_handle", sa_os, os) || 2474 GETMEMB(addr, "sa_handle", sa_bonus, db_bonus) || 2475 GETMEMB(addr, "sa_handle", sa_spill, db_spill)) { 2476 mdb_printf("Can't find necessary information in sa_handle " 2477 "in sa_handle\n"); 2478 return (DCMD_ERR); 2479 } 2480 2481 if (GETMEMB(os, "objset", os_sa, os_sa)) { 2482 mdb_printf("Can't find os_sa in objset\n"); 2483 return (DCMD_ERR); 2484 } 2485 2486 if (GETMEMB(os_sa, "sa_os", sa_num_attrs, attr_count)) { 2487 mdb_printf("Can't find sa_num_attrs\n"); 2488 return (DCMD_ERR); 2489 } 2490 2491 if (attr_id > attr_count) { 2492 mdb_printf("attribute id number is out of range\n"); 2493 return (DCMD_ERR); 2494 } 2495 2496 if (bonus_tab) { 2497 if (sa_get_off_table(bonus_tab, &offset_tab, 2498 attr_count) == -1) { 2499 return (DCMD_ERR); 2500 } 2501 2502 if (GETMEMB(db_bonus, "dmu_buf", db_data, db_data)) { 2503 mdb_printf("can't find db_data in bonus dbuf\n"); 2504 return (DCMD_ERR); 2505 } 2506 } 2507 2508 if (bonus_tab && !TOC_ATTR_PRESENT(offset_tab[attr_id]) && 2509 spill_tab == NULL) { 2510 mdb_printf("Attribute does not exist\n"); 2511 return (DCMD_ERR); 2512 } else if (!TOC_ATTR_PRESENT(offset_tab[attr_id]) && spill_tab) { 2513 if (sa_get_off_table(spill_tab, &offset_tab, 2514 attr_count) == -1) { 2515 return (DCMD_ERR); 2516 } 2517 if (GETMEMB(db_spill, "dmu_buf", db_data, db_data)) { 2518 mdb_printf("can't find db_data in spill dbuf\n"); 2519 return (DCMD_ERR); 2520 } 2521 if (!TOC_ATTR_PRESENT(offset_tab[attr_id])) { 2522 mdb_printf("Attribute does not exist\n"); 2523 return (DCMD_ERR); 2524 } 2525 } 2526 attr_addr = db_data + TOC_OFF(offset_tab[attr_id]); 2527 mdb_printf("%p\n", attr_addr); 2528 return (DCMD_OK); 2529 } 2530 2531 /* ARGSUSED */ 2532 static int 2533 zfs_ace_print_common(uintptr_t addr, uint_t flags, 2534 uint64_t id, uint32_t access_mask, uint16_t ace_flags, 2535 uint16_t ace_type, int verbose) 2536 { 2537 if (DCMD_HDRSPEC(flags) && !verbose) 2538 mdb_printf("%<u>%-?s %-8s %-8s %-8s %s%</u>\n", 2539 "ADDR", "FLAGS", "MASK", "TYPE", "ID"); 2540 2541 if (!verbose) { 2542 mdb_printf("%0?p %-8x %-8x %-8x %-llx\n", addr, 2543 ace_flags, access_mask, ace_type, id); 2544 return (DCMD_OK); 2545 } 2546 2547 switch (ace_flags & ACE_TYPE_FLAGS) { 2548 case ACE_OWNER: 2549 mdb_printf("owner@:"); 2550 break; 2551 case (ACE_IDENTIFIER_GROUP | ACE_GROUP): 2552 mdb_printf("group@:"); 2553 break; 2554 case ACE_EVERYONE: 2555 mdb_printf("everyone@:"); 2556 break; 2557 case ACE_IDENTIFIER_GROUP: 2558 mdb_printf("group:%llx:", (u_longlong_t)id); 2559 break; 2560 case 0: /* User entry */ 2561 mdb_printf("user:%llx:", (u_longlong_t)id); 2562 break; 2563 } 2564 2565 /* print out permission mask */ 2566 if (access_mask & ACE_READ_DATA) 2567 mdb_printf("r"); 2568 else 2569 mdb_printf("-"); 2570 if (access_mask & ACE_WRITE_DATA) 2571 mdb_printf("w"); 2572 else 2573 mdb_printf("-"); 2574 if (access_mask & ACE_EXECUTE) 2575 mdb_printf("x"); 2576 else 2577 mdb_printf("-"); 2578 if (access_mask & ACE_APPEND_DATA) 2579 mdb_printf("p"); 2580 else 2581 mdb_printf("-"); 2582 if (access_mask & ACE_DELETE) 2583 mdb_printf("d"); 2584 else 2585 mdb_printf("-"); 2586 if (access_mask & ACE_DELETE_CHILD) 2587 mdb_printf("D"); 2588 else 2589 mdb_printf("-"); 2590 if (access_mask & ACE_READ_ATTRIBUTES) 2591 mdb_printf("a"); 2592 else 2593 mdb_printf("-"); 2594 if (access_mask & ACE_WRITE_ATTRIBUTES) 2595 mdb_printf("A"); 2596 else 2597 mdb_printf("-"); 2598 if (access_mask & ACE_READ_NAMED_ATTRS) 2599 mdb_printf("R"); 2600 else 2601 mdb_printf("-"); 2602 if (access_mask & ACE_WRITE_NAMED_ATTRS) 2603 mdb_printf("W"); 2604 else 2605 mdb_printf("-"); 2606 if (access_mask & ACE_READ_ACL) 2607 mdb_printf("c"); 2608 else 2609 mdb_printf("-"); 2610 if (access_mask & ACE_WRITE_ACL) 2611 mdb_printf("C"); 2612 else 2613 mdb_printf("-"); 2614 if (access_mask & ACE_WRITE_OWNER) 2615 mdb_printf("o"); 2616 else 2617 mdb_printf("-"); 2618 if (access_mask & ACE_SYNCHRONIZE) 2619 mdb_printf("s"); 2620 else 2621 mdb_printf("-"); 2622 2623 mdb_printf(":"); 2624 2625 /* Print out inheritance flags */ 2626 if (ace_flags & ACE_FILE_INHERIT_ACE) 2627 mdb_printf("f"); 2628 else 2629 mdb_printf("-"); 2630 if (ace_flags & ACE_DIRECTORY_INHERIT_ACE) 2631 mdb_printf("d"); 2632 else 2633 mdb_printf("-"); 2634 if (ace_flags & ACE_INHERIT_ONLY_ACE) 2635 mdb_printf("i"); 2636 else 2637 mdb_printf("-"); 2638 if (ace_flags & ACE_NO_PROPAGATE_INHERIT_ACE) 2639 mdb_printf("n"); 2640 else 2641 mdb_printf("-"); 2642 if (ace_flags & ACE_SUCCESSFUL_ACCESS_ACE_FLAG) 2643 mdb_printf("S"); 2644 else 2645 mdb_printf("-"); 2646 if (ace_flags & ACE_FAILED_ACCESS_ACE_FLAG) 2647 mdb_printf("F"); 2648 else 2649 mdb_printf("-"); 2650 if (ace_flags & ACE_INHERITED_ACE) 2651 mdb_printf("I"); 2652 else 2653 mdb_printf("-"); 2654 2655 switch (ace_type) { 2656 case ACE_ACCESS_ALLOWED_ACE_TYPE: 2657 mdb_printf(":allow\n"); 2658 break; 2659 case ACE_ACCESS_DENIED_ACE_TYPE: 2660 mdb_printf(":deny\n"); 2661 break; 2662 case ACE_SYSTEM_AUDIT_ACE_TYPE: 2663 mdb_printf(":audit\n"); 2664 break; 2665 case ACE_SYSTEM_ALARM_ACE_TYPE: 2666 mdb_printf(":alarm\n"); 2667 break; 2668 default: 2669 mdb_printf(":?\n"); 2670 } 2671 return (DCMD_OK); 2672 } 2673 2674 /* ARGSUSED */ 2675 static int 2676 zfs_ace_print(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv) 2677 { 2678 zfs_ace_t zace; 2679 int verbose = FALSE; 2680 uint64_t id; 2681 2682 if (!(flags & DCMD_ADDRSPEC)) 2683 return (DCMD_USAGE); 2684 2685 if (mdb_getopts(argc, argv, 2686 'v', MDB_OPT_SETBITS, TRUE, &verbose, TRUE, NULL) != argc) 2687 return (DCMD_USAGE); 2688 2689 if (mdb_vread(&zace, sizeof (zfs_ace_t), addr) == -1) { 2690 mdb_warn("failed to read zfs_ace_t"); 2691 return (DCMD_ERR); 2692 } 2693 2694 if ((zace.z_hdr.z_flags & ACE_TYPE_FLAGS) == 0 || 2695 (zace.z_hdr.z_flags & ACE_TYPE_FLAGS) == ACE_IDENTIFIER_GROUP) 2696 id = zace.z_fuid; 2697 else 2698 id = -1; 2699 2700 return (zfs_ace_print_common(addr, flags, id, zace.z_hdr.z_access_mask, 2701 zace.z_hdr.z_flags, zace.z_hdr.z_type, verbose)); 2702 } 2703 2704 /* ARGSUSED */ 2705 static int 2706 zfs_ace0_print(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv) 2707 { 2708 ace_t ace; 2709 uint64_t id; 2710 int verbose = FALSE; 2711 2712 if (!(flags & DCMD_ADDRSPEC)) 2713 return (DCMD_USAGE); 2714 2715 if (mdb_getopts(argc, argv, 2716 'v', MDB_OPT_SETBITS, TRUE, &verbose, TRUE, NULL) != argc) 2717 return (DCMD_USAGE); 2718 2719 if (mdb_vread(&ace, sizeof (ace_t), addr) == -1) { 2720 mdb_warn("failed to read ace_t"); 2721 return (DCMD_ERR); 2722 } 2723 2724 if ((ace.a_flags & ACE_TYPE_FLAGS) == 0 || 2725 (ace.a_flags & ACE_TYPE_FLAGS) == ACE_IDENTIFIER_GROUP) 2726 id = ace.a_who; 2727 else 2728 id = -1; 2729 2730 return (zfs_ace_print_common(addr, flags, id, ace.a_access_mask, 2731 ace.a_flags, ace.a_type, verbose)); 2732 } 2733 2734 typedef struct acl_dump_args { 2735 int a_argc; 2736 const mdb_arg_t *a_argv; 2737 uint16_t a_version; 2738 int a_flags; 2739 } acl_dump_args_t; 2740 2741 /* ARGSUSED */ 2742 static int 2743 acl_aces_cb(uintptr_t addr, const void *unknown, void *arg) 2744 { 2745 acl_dump_args_t *acl_args = (acl_dump_args_t *)arg; 2746 2747 if (acl_args->a_version == 1) { 2748 if (mdb_call_dcmd("zfs_ace", addr, 2749 DCMD_ADDRSPEC|acl_args->a_flags, acl_args->a_argc, 2750 acl_args->a_argv) != DCMD_OK) { 2751 return (WALK_ERR); 2752 } 2753 } else { 2754 if (mdb_call_dcmd("zfs_ace0", addr, 2755 DCMD_ADDRSPEC|acl_args->a_flags, acl_args->a_argc, 2756 acl_args->a_argv) != DCMD_OK) { 2757 return (WALK_ERR); 2758 } 2759 } 2760 acl_args->a_flags = DCMD_LOOP; 2761 return (WALK_NEXT); 2762 } 2763 2764 /* ARGSUSED */ 2765 static int 2766 acl_cb(uintptr_t addr, const void *unknown, void *arg) 2767 { 2768 acl_dump_args_t *acl_args = (acl_dump_args_t *)arg; 2769 2770 if (acl_args->a_version == 1) { 2771 if (mdb_pwalk("zfs_acl_node_aces", acl_aces_cb, 2772 arg, addr) != 0) { 2773 mdb_warn("can't walk ACEs"); 2774 return (DCMD_ERR); 2775 } 2776 } else { 2777 if (mdb_pwalk("zfs_acl_node_aces0", acl_aces_cb, 2778 arg, addr) != 0) { 2779 mdb_warn("can't walk ACEs"); 2780 return (DCMD_ERR); 2781 } 2782 } 2783 return (WALK_NEXT); 2784 } 2785 2786 /* ARGSUSED */ 2787 static int 2788 zfs_acl_dump(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv) 2789 { 2790 zfs_acl_t zacl; 2791 int verbose = FALSE; 2792 acl_dump_args_t acl_args; 2793 2794 if (!(flags & DCMD_ADDRSPEC)) 2795 return (DCMD_USAGE); 2796 2797 if (mdb_getopts(argc, argv, 2798 'v', MDB_OPT_SETBITS, TRUE, &verbose, TRUE, NULL) != argc) 2799 return (DCMD_USAGE); 2800 2801 if (mdb_vread(&zacl, sizeof (zfs_acl_t), addr) == -1) { 2802 mdb_warn("failed to read zfs_acl_t"); 2803 return (DCMD_ERR); 2804 } 2805 2806 acl_args.a_argc = argc; 2807 acl_args.a_argv = argv; 2808 acl_args.a_version = zacl.z_version; 2809 acl_args.a_flags = DCMD_LOOPFIRST; 2810 2811 if (mdb_pwalk("zfs_acl_node", acl_cb, &acl_args, addr) != 0) { 2812 mdb_warn("can't walk ACL"); 2813 return (DCMD_ERR); 2814 } 2815 2816 return (DCMD_OK); 2817 } 2818 2819 /* ARGSUSED */ 2820 static int 2821 zfs_acl_node_walk_init(mdb_walk_state_t *wsp) 2822 { 2823 if (wsp->walk_addr == NULL) { 2824 mdb_warn("must supply address of zfs_acl_node_t\n"); 2825 return (WALK_ERR); 2826 } 2827 2828 wsp->walk_addr += 2829 mdb_ctf_offsetof_by_name(ZFS_STRUCT "zfs_acl", "z_acl"); 2830 2831 if (mdb_layered_walk("list", wsp) == -1) { 2832 mdb_warn("failed to walk 'list'\n"); 2833 return (WALK_ERR); 2834 } 2835 2836 return (WALK_NEXT); 2837 } 2838 2839 static int 2840 zfs_acl_node_walk_step(mdb_walk_state_t *wsp) 2841 { 2842 zfs_acl_node_t aclnode; 2843 2844 if (mdb_vread(&aclnode, sizeof (zfs_acl_node_t), 2845 wsp->walk_addr) == -1) { 2846 mdb_warn("failed to read zfs_acl_node at %p", wsp->walk_addr); 2847 return (WALK_ERR); 2848 } 2849 2850 return (wsp->walk_callback(wsp->walk_addr, &aclnode, wsp->walk_cbdata)); 2851 } 2852 2853 typedef struct ace_walk_data { 2854 int ace_count; 2855 int ace_version; 2856 } ace_walk_data_t; 2857 2858 static int 2859 zfs_aces_walk_init_common(mdb_walk_state_t *wsp, int version, 2860 int ace_count, uintptr_t ace_data) 2861 { 2862 ace_walk_data_t *ace_walk_data; 2863 2864 if (wsp->walk_addr == NULL) { 2865 mdb_warn("must supply address of zfs_acl_node_t\n"); 2866 return (WALK_ERR); 2867 } 2868 2869 ace_walk_data = mdb_alloc(sizeof (ace_walk_data_t), UM_SLEEP | UM_GC); 2870 2871 ace_walk_data->ace_count = ace_count; 2872 ace_walk_data->ace_version = version; 2873 2874 wsp->walk_addr = ace_data; 2875 wsp->walk_data = ace_walk_data; 2876 2877 return (WALK_NEXT); 2878 } 2879 2880 static int 2881 zfs_acl_node_aces_walk_init_common(mdb_walk_state_t *wsp, int version) 2882 { 2883 static int gotid; 2884 static mdb_ctf_id_t acl_id; 2885 int z_ace_count; 2886 uintptr_t z_acldata; 2887 2888 if (!gotid) { 2889 if (mdb_ctf_lookup_by_name("struct zfs_acl_node", 2890 &acl_id) == -1) { 2891 mdb_warn("couldn't find struct zfs_acl_node"); 2892 return (DCMD_ERR); 2893 } 2894 gotid = TRUE; 2895 } 2896 2897 if (GETMEMBID(wsp->walk_addr, &acl_id, z_ace_count, z_ace_count)) { 2898 return (DCMD_ERR); 2899 } 2900 if (GETMEMBID(wsp->walk_addr, &acl_id, z_acldata, z_acldata)) { 2901 return (DCMD_ERR); 2902 } 2903 2904 return (zfs_aces_walk_init_common(wsp, version, 2905 z_ace_count, z_acldata)); 2906 } 2907 2908 /* ARGSUSED */ 2909 static int 2910 zfs_acl_node_aces_walk_init(mdb_walk_state_t *wsp) 2911 { 2912 return (zfs_acl_node_aces_walk_init_common(wsp, 1)); 2913 } 2914 2915 /* ARGSUSED */ 2916 static int 2917 zfs_acl_node_aces0_walk_init(mdb_walk_state_t *wsp) 2918 { 2919 return (zfs_acl_node_aces_walk_init_common(wsp, 0)); 2920 } 2921 2922 static int 2923 zfs_aces_walk_step(mdb_walk_state_t *wsp) 2924 { 2925 ace_walk_data_t *ace_data = wsp->walk_data; 2926 zfs_ace_t zace; 2927 ace_t *acep; 2928 int status; 2929 int entry_type; 2930 int allow_type; 2931 uintptr_t ptr; 2932 2933 if (ace_data->ace_count == 0) 2934 return (WALK_DONE); 2935 2936 if (mdb_vread(&zace, sizeof (zfs_ace_t), wsp->walk_addr) == -1) { 2937 mdb_warn("failed to read zfs_ace_t at %#lx", 2938 wsp->walk_addr); 2939 return (WALK_ERR); 2940 } 2941 2942 switch (ace_data->ace_version) { 2943 case 0: 2944 acep = (ace_t *)&zace; 2945 entry_type = acep->a_flags & ACE_TYPE_FLAGS; 2946 allow_type = acep->a_type; 2947 break; 2948 case 1: 2949 entry_type = zace.z_hdr.z_flags & ACE_TYPE_FLAGS; 2950 allow_type = zace.z_hdr.z_type; 2951 break; 2952 default: 2953 return (WALK_ERR); 2954 } 2955 2956 ptr = (uintptr_t)wsp->walk_addr; 2957 switch (entry_type) { 2958 case ACE_OWNER: 2959 case ACE_EVERYONE: 2960 case (ACE_IDENTIFIER_GROUP | ACE_GROUP): 2961 ptr += ace_data->ace_version == 0 ? 2962 sizeof (ace_t) : sizeof (zfs_ace_hdr_t); 2963 break; 2964 case ACE_IDENTIFIER_GROUP: 2965 default: 2966 switch (allow_type) { 2967 case ACE_ACCESS_ALLOWED_OBJECT_ACE_TYPE: 2968 case ACE_ACCESS_DENIED_OBJECT_ACE_TYPE: 2969 case ACE_SYSTEM_AUDIT_OBJECT_ACE_TYPE: 2970 case ACE_SYSTEM_ALARM_OBJECT_ACE_TYPE: 2971 ptr += ace_data->ace_version == 0 ? 2972 sizeof (ace_t) : sizeof (zfs_object_ace_t); 2973 break; 2974 default: 2975 ptr += ace_data->ace_version == 0 ? 2976 sizeof (ace_t) : sizeof (zfs_ace_t); 2977 break; 2978 } 2979 } 2980 2981 ace_data->ace_count--; 2982 status = wsp->walk_callback(wsp->walk_addr, 2983 (void *)(uintptr_t)&zace, wsp->walk_cbdata); 2984 2985 wsp->walk_addr = ptr; 2986 return (status); 2987 } 2988 2989 typedef struct mdb_zfs_rrwlock { 2990 kthread_t *rr_writer; 2991 boolean_t rr_writer_wanted; 2992 } mdb_zfs_rrwlock_t; 2993 2994 static uint_t rrw_key; 2995 2996 /* ARGSUSED */ 2997 static int 2998 rrwlock(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv) 2999 { 3000 mdb_zfs_rrwlock_t rrw; 3001 3002 if (rrw_key == 0) { 3003 if (mdb_ctf_readsym(&rrw_key, "uint_t", "rrw_tsd_key", 0) == -1) 3004 return (DCMD_ERR); 3005 } 3006 3007 if (mdb_ctf_vread(&rrw, "rrwlock_t", "mdb_zfs_rrwlock_t", addr, 3008 0) == -1) 3009 return (DCMD_ERR); 3010 3011 if (rrw.rr_writer != NULL) { 3012 mdb_printf("write lock held by thread %p\n", rrw.rr_writer); 3013 return (DCMD_OK); 3014 } 3015 3016 if (rrw.rr_writer_wanted) { 3017 mdb_printf("writer wanted\n"); 3018 } 3019 3020 mdb_printf("anonymous references:\n"); 3021 (void) mdb_call_dcmd("refcount", addr + 3022 mdb_ctf_offsetof_by_name(ZFS_STRUCT "rrwlock", "rr_anon_rcount"), 3023 DCMD_ADDRSPEC, 0, NULL); 3024 3025 mdb_printf("linked references:\n"); 3026 (void) mdb_call_dcmd("refcount", addr + 3027 mdb_ctf_offsetof_by_name(ZFS_STRUCT "rrwlock", "rr_linked_rcount"), 3028 DCMD_ADDRSPEC, 0, NULL); 3029 3030 /* 3031 * XXX This should find references from 3032 * "::walk thread | ::tsd -v <rrw_key>", but there is no support 3033 * for programmatic consumption of dcmds, so this would be 3034 * difficult, potentially requiring reimplementing ::tsd (both 3035 * user and kernel versions) in this MDB module. 3036 */ 3037 3038 return (DCMD_OK); 3039 } 3040 3041 /* 3042 * MDB module linkage information: 3043 * 3044 * We declare a list of structures describing our dcmds, and a function 3045 * named _mdb_init to return a pointer to our module information. 3046 */ 3047 3048 static const mdb_dcmd_t dcmds[] = { 3049 { "arc", "[-bkmg]", "print ARC variables", arc_print }, 3050 { "blkptr", ":", "print blkptr_t", blkptr }, 3051 { "dbuf", ":", "print dmu_buf_impl_t", dbuf }, 3052 { "dbuf_stats", ":", "dbuf stats", dbuf_stats }, 3053 { "dbufs", 3054 "\t[-O objset_t*] [-n objset_name | \"mos\"] " 3055 "[-o object | \"mdn\"] \n" 3056 "\t[-l level] [-b blkid | \"bonus\"]", 3057 "find dmu_buf_impl_t's that match specified criteria", dbufs }, 3058 { "abuf_find", "dva_word[0] dva_word[1]", 3059 "find arc_buf_hdr_t of a specified DVA", 3060 abuf_find }, 3061 { "spa", "?[-cv]", "spa_t summary", spa_print }, 3062 { "spa_config", ":", "print spa_t configuration", spa_print_config }, 3063 { "spa_verify", ":", "verify spa_t consistency", spa_verify }, 3064 { "spa_space", ":[-b]", "print spa_t on-disk space usage", spa_space }, 3065 { "spa_vdevs", ":", "given a spa_t, print vdev summary", spa_vdevs }, 3066 { "vdev", ":[-re]\n" 3067 "\t-r display recursively\n" 3068 "\t-e print statistics", 3069 "vdev_t summary", vdev_print }, 3070 { "zio", ":[cpr]\n" 3071 "\t-c display children\n" 3072 "\t-p display parents\n" 3073 "\t-r display recursively", 3074 "zio_t summary", zio_print }, 3075 { "zio_state", "?", "print out all zio_t structures on system or " 3076 "for a particular pool", zio_state }, 3077 { "zfs_blkstats", ":[-v]", 3078 "given a spa_t, print block type stats from last scrub", 3079 zfs_blkstats }, 3080 { "zfs_params", "", "print zfs tunable parameters", zfs_params }, 3081 { "refcount", ":[-r]\n" 3082 "\t-r display recently removed references", 3083 "print refcount_t holders", refcount }, 3084 { "zap_leaf", "", "print zap_leaf_phys_t", zap_leaf }, 3085 { "zfs_aces", ":[-v]", "print all ACEs from a zfs_acl_t", 3086 zfs_acl_dump }, 3087 { "zfs_ace", ":[-v]", "print zfs_ace", zfs_ace_print }, 3088 { "zfs_ace0", ":[-v]", "print zfs_ace0", zfs_ace0_print }, 3089 { "sa_attr_table", ":", "print SA attribute table from sa_os_t", 3090 sa_attr_table}, 3091 { "sa_attr", ": attr_id", 3092 "print SA attribute address when given sa_handle_t", sa_attr_print}, 3093 { "zfs_dbgmsg", ":[-va]", 3094 "print zfs debug log", dbgmsg}, 3095 { "rrwlock", ":", 3096 "print rrwlock_t, including readers", rrwlock}, 3097 { NULL } 3098 }; 3099 3100 static const mdb_walker_t walkers[] = { 3101 { "zms_freelist", "walk ZFS metaslab freelist", 3102 freelist_walk_init, freelist_walk_step, NULL }, 3103 { "txg_list", "given any txg_list_t *, walk all entries in all txgs", 3104 txg_list_walk_init, txg_list_walk_step, NULL }, 3105 { "txg_list0", "given any txg_list_t *, walk all entries in txg 0", 3106 txg_list0_walk_init, txg_list_walk_step, NULL }, 3107 { "txg_list1", "given any txg_list_t *, walk all entries in txg 1", 3108 txg_list1_walk_init, txg_list_walk_step, NULL }, 3109 { "txg_list2", "given any txg_list_t *, walk all entries in txg 2", 3110 txg_list2_walk_init, txg_list_walk_step, NULL }, 3111 { "txg_list3", "given any txg_list_t *, walk all entries in txg 3", 3112 txg_list3_walk_init, txg_list_walk_step, NULL }, 3113 { "zio", "walk all zio structures, optionally for a particular spa_t", 3114 zio_walk_init, zio_walk_step, NULL }, 3115 { "zio_root", 3116 "walk all root zio_t structures, optionally for a particular spa_t", 3117 zio_walk_init, zio_walk_root_step, NULL }, 3118 { "spa", "walk all spa_t entries in the namespace", 3119 spa_walk_init, spa_walk_step, NULL }, 3120 { "metaslab", "given a spa_t *, walk all metaslab_t structures", 3121 metaslab_walk_init, metaslab_walk_step, NULL }, 3122 { "zfs_acl_node", "given a zfs_acl_t, walk all zfs_acl_nodes", 3123 zfs_acl_node_walk_init, zfs_acl_node_walk_step, NULL }, 3124 { "zfs_acl_node_aces", "given a zfs_acl_node_t, walk all ACEs", 3125 zfs_acl_node_aces_walk_init, zfs_aces_walk_step, NULL }, 3126 { "zfs_acl_node_aces0", 3127 "given a zfs_acl_node_t, walk all ACEs as ace_t", 3128 zfs_acl_node_aces0_walk_init, zfs_aces_walk_step, NULL }, 3129 { NULL } 3130 }; 3131 3132 static const mdb_modinfo_t modinfo = { 3133 MDB_API_VERSION, dcmds, walkers 3134 }; 3135 3136 const mdb_modinfo_t * 3137 _mdb_init(void) 3138 { 3139 return (&modinfo); 3140 } 3141