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) 2011, 2016 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/vdev_impl.h> 40 #include <sys/zap_leaf.h> 41 #include <sys/zap_impl.h> 42 #include <ctype.h> 43 #include <sys/zfs_acl.h> 44 #include <sys/sa_impl.h> 45 #include <sys/multilist.h> 46 47 #ifdef _KERNEL 48 #define ZFS_OBJ_NAME "zfs" 49 extern int64_t mdb_gethrtime(void); 50 #else 51 #define ZFS_OBJ_NAME "libzpool.so.1" 52 #endif 53 54 #define ZFS_STRUCT "struct " ZFS_OBJ_NAME "`" 55 56 #ifndef _KERNEL 57 int aok; 58 #endif 59 60 enum spa_flags { 61 SPA_FLAG_CONFIG = 1 << 0, 62 SPA_FLAG_VDEVS = 1 << 1, 63 SPA_FLAG_ERRORS = 1 << 2, 64 SPA_FLAG_METASLAB_GROUPS = 1 << 3, 65 SPA_FLAG_METASLABS = 1 << 4, 66 SPA_FLAG_HISTOGRAMS = 1 << 5 67 }; 68 69 /* 70 * If any of these flags are set, call spa_vdevs in spa_print 71 */ 72 #define SPA_FLAG_ALL_VDEV \ 73 (SPA_FLAG_VDEVS | SPA_FLAG_ERRORS | SPA_FLAG_METASLAB_GROUPS | \ 74 SPA_FLAG_METASLABS) 75 76 static int 77 getmember(uintptr_t addr, const char *type, mdb_ctf_id_t *idp, 78 const char *member, int len, void *buf) 79 { 80 mdb_ctf_id_t id; 81 ulong_t off; 82 char name[64]; 83 84 if (idp == NULL) { 85 if (mdb_ctf_lookup_by_name(type, &id) == -1) { 86 mdb_warn("couldn't find type %s", type); 87 return (DCMD_ERR); 88 } 89 idp = &id; 90 } else { 91 type = name; 92 mdb_ctf_type_name(*idp, name, sizeof (name)); 93 } 94 95 if (mdb_ctf_offsetof(*idp, member, &off) == -1) { 96 mdb_warn("couldn't find member %s of type %s\n", member, type); 97 return (DCMD_ERR); 98 } 99 if (off % 8 != 0) { 100 mdb_warn("member %s of type %s is unsupported bitfield", 101 member, type); 102 return (DCMD_ERR); 103 } 104 off /= 8; 105 106 if (mdb_vread(buf, len, addr + off) == -1) { 107 mdb_warn("failed to read %s from %s at %p", 108 member, type, addr + off); 109 return (DCMD_ERR); 110 } 111 /* mdb_warn("read %s from %s at %p+%llx\n", member, type, addr, off); */ 112 113 return (0); 114 } 115 116 #define GETMEMB(addr, structname, member, dest) \ 117 getmember(addr, ZFS_STRUCT structname, NULL, #member, \ 118 sizeof (dest), &(dest)) 119 120 #define GETMEMBID(addr, ctfid, member, dest) \ 121 getmember(addr, NULL, ctfid, #member, sizeof (dest), &(dest)) 122 123 static boolean_t 124 strisprint(const char *cp) 125 { 126 for (; *cp; cp++) { 127 if (!isprint(*cp)) 128 return (B_FALSE); 129 } 130 return (B_TRUE); 131 } 132 133 #define NICENUM_BUFLEN 6 134 135 static int 136 snprintfrac(char *buf, int len, 137 uint64_t numerator, uint64_t denom, int frac_digits) 138 { 139 int mul = 1; 140 int whole, frac, i; 141 142 for (i = frac_digits; i; i--) 143 mul *= 10; 144 whole = numerator / denom; 145 frac = mul * numerator / denom - mul * whole; 146 return (mdb_snprintf(buf, len, "%u.%0*u", whole, frac_digits, frac)); 147 } 148 149 static void 150 mdb_nicenum(uint64_t num, char *buf) 151 { 152 uint64_t n = num; 153 int index = 0; 154 char *u; 155 156 while (n >= 1024) { 157 n = (n + (1024 / 2)) / 1024; /* Round up or down */ 158 index++; 159 } 160 161 u = &" \0K\0M\0G\0T\0P\0E\0"[index*2]; 162 163 if (index == 0) { 164 (void) mdb_snprintf(buf, NICENUM_BUFLEN, "%llu", 165 (u_longlong_t)n); 166 } else if (n < 10 && (num & (num - 1)) != 0) { 167 (void) snprintfrac(buf, NICENUM_BUFLEN, 168 num, 1ULL << 10 * index, 2); 169 strcat(buf, u); 170 } else if (n < 100 && (num & (num - 1)) != 0) { 171 (void) snprintfrac(buf, NICENUM_BUFLEN, 172 num, 1ULL << 10 * index, 1); 173 strcat(buf, u); 174 } else { 175 (void) mdb_snprintf(buf, NICENUM_BUFLEN, "%llu%s", 176 (u_longlong_t)n, u); 177 } 178 } 179 180 static int verbose; 181 182 static int 183 freelist_walk_init(mdb_walk_state_t *wsp) 184 { 185 if (wsp->walk_addr == NULL) { 186 mdb_warn("must supply starting address\n"); 187 return (WALK_ERR); 188 } 189 190 wsp->walk_data = 0; /* Index into the freelist */ 191 return (WALK_NEXT); 192 } 193 194 static int 195 freelist_walk_step(mdb_walk_state_t *wsp) 196 { 197 uint64_t entry; 198 uintptr_t number = (uintptr_t)wsp->walk_data; 199 char *ddata[] = { "ALLOC", "FREE", "CONDENSE", "INVALID", 200 "INVALID", "INVALID", "INVALID", "INVALID" }; 201 int mapshift = SPA_MINBLOCKSHIFT; 202 203 if (mdb_vread(&entry, sizeof (entry), wsp->walk_addr) == -1) { 204 mdb_warn("failed to read freelist entry %p", wsp->walk_addr); 205 return (WALK_DONE); 206 } 207 wsp->walk_addr += sizeof (entry); 208 wsp->walk_data = (void *)(number + 1); 209 210 if (SM_DEBUG_DECODE(entry)) { 211 mdb_printf("DEBUG: %3u %10s: txg=%llu pass=%llu\n", 212 number, 213 ddata[SM_DEBUG_ACTION_DECODE(entry)], 214 SM_DEBUG_TXG_DECODE(entry), 215 SM_DEBUG_SYNCPASS_DECODE(entry)); 216 } else { 217 mdb_printf("Entry: %3u offsets=%08llx-%08llx type=%c " 218 "size=%06llx", number, 219 SM_OFFSET_DECODE(entry) << mapshift, 220 (SM_OFFSET_DECODE(entry) + SM_RUN_DECODE(entry)) << 221 mapshift, 222 SM_TYPE_DECODE(entry) == SM_ALLOC ? 'A' : 'F', 223 SM_RUN_DECODE(entry) << mapshift); 224 if (verbose) 225 mdb_printf(" (raw=%012llx)\n", entry); 226 mdb_printf("\n"); 227 } 228 return (WALK_NEXT); 229 } 230 231 static int 232 mdb_dsl_dir_name(uintptr_t addr, char *buf) 233 { 234 static int gotid; 235 static mdb_ctf_id_t dd_id; 236 uintptr_t dd_parent; 237 char dd_myname[ZFS_MAX_DATASET_NAME_LEN]; 238 239 if (!gotid) { 240 if (mdb_ctf_lookup_by_name(ZFS_STRUCT "dsl_dir", 241 &dd_id) == -1) { 242 mdb_warn("couldn't find struct dsl_dir"); 243 return (DCMD_ERR); 244 } 245 gotid = TRUE; 246 } 247 if (GETMEMBID(addr, &dd_id, dd_parent, dd_parent) || 248 GETMEMBID(addr, &dd_id, dd_myname, dd_myname)) { 249 return (DCMD_ERR); 250 } 251 252 if (dd_parent) { 253 if (mdb_dsl_dir_name(dd_parent, buf)) 254 return (DCMD_ERR); 255 strcat(buf, "/"); 256 } 257 258 if (dd_myname[0]) 259 strcat(buf, dd_myname); 260 else 261 strcat(buf, "???"); 262 263 return (0); 264 } 265 266 static int 267 objset_name(uintptr_t addr, char *buf) 268 { 269 static int gotid; 270 static mdb_ctf_id_t os_id, ds_id; 271 uintptr_t os_dsl_dataset; 272 char ds_snapname[ZFS_MAX_DATASET_NAME_LEN]; 273 uintptr_t ds_dir; 274 275 buf[0] = '\0'; 276 277 if (!gotid) { 278 if (mdb_ctf_lookup_by_name(ZFS_STRUCT "objset", 279 &os_id) == -1) { 280 mdb_warn("couldn't find struct objset"); 281 return (DCMD_ERR); 282 } 283 if (mdb_ctf_lookup_by_name(ZFS_STRUCT "dsl_dataset", 284 &ds_id) == -1) { 285 mdb_warn("couldn't find struct dsl_dataset"); 286 return (DCMD_ERR); 287 } 288 289 gotid = TRUE; 290 } 291 292 if (GETMEMBID(addr, &os_id, os_dsl_dataset, os_dsl_dataset)) 293 return (DCMD_ERR); 294 295 if (os_dsl_dataset == 0) { 296 strcat(buf, "mos"); 297 return (0); 298 } 299 300 if (GETMEMBID(os_dsl_dataset, &ds_id, ds_snapname, ds_snapname) || 301 GETMEMBID(os_dsl_dataset, &ds_id, ds_dir, ds_dir)) { 302 return (DCMD_ERR); 303 } 304 305 if (ds_dir && mdb_dsl_dir_name(ds_dir, buf)) 306 return (DCMD_ERR); 307 308 if (ds_snapname[0]) { 309 strcat(buf, "@"); 310 strcat(buf, ds_snapname); 311 } 312 return (0); 313 } 314 315 static int 316 enum_lookup(char *type, int val, const char *prefix, size_t size, char *out) 317 { 318 const char *cp; 319 size_t len = strlen(prefix); 320 mdb_ctf_id_t enum_type; 321 322 if (mdb_ctf_lookup_by_name(type, &enum_type) != 0) { 323 mdb_warn("Could not find enum for %s", type); 324 return (-1); 325 } 326 327 if ((cp = mdb_ctf_enum_name(enum_type, val)) != NULL) { 328 if (strncmp(cp, prefix, len) == 0) 329 cp += len; 330 (void) strncpy(out, cp, size); 331 } else { 332 mdb_snprintf(out, size, "? (%d)", val); 333 } 334 return (0); 335 } 336 337 /* ARGSUSED */ 338 static int 339 zfs_params(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv) 340 { 341 /* 342 * This table can be approximately generated by running: 343 * egrep "^[a-z0-9_]+ [a-z0-9_]+( =.*)?;" *.c | cut -d ' ' -f 2 344 */ 345 static const char *params[] = { 346 "arc_reduce_dnlc_percent", 347 "arc_lotsfree_percent", 348 "zfs_dirty_data_max", 349 "zfs_dirty_data_sync", 350 "zfs_delay_max_ns", 351 "zfs_delay_min_dirty_percent", 352 "zfs_delay_scale", 353 "zfs_vdev_max_active", 354 "zfs_vdev_sync_read_min_active", 355 "zfs_vdev_sync_read_max_active", 356 "zfs_vdev_sync_write_min_active", 357 "zfs_vdev_sync_write_max_active", 358 "zfs_vdev_async_read_min_active", 359 "zfs_vdev_async_read_max_active", 360 "zfs_vdev_async_write_min_active", 361 "zfs_vdev_async_write_max_active", 362 "zfs_vdev_scrub_min_active", 363 "zfs_vdev_scrub_max_active", 364 "zfs_vdev_async_write_active_min_dirty_percent", 365 "zfs_vdev_async_write_active_max_dirty_percent", 366 "spa_asize_inflation", 367 "zfs_arc_max", 368 "zfs_arc_min", 369 "arc_shrink_shift", 370 "zfs_mdcomp_disable", 371 "zfs_prefetch_disable", 372 "zfetch_max_streams", 373 "zfetch_min_sec_reap", 374 "zfetch_block_cap", 375 "zfetch_array_rd_sz", 376 "zfs_default_bs", 377 "zfs_default_ibs", 378 "metaslab_aliquot", 379 "reference_tracking_enable", 380 "reference_history", 381 "spa_max_replication_override", 382 "spa_mode_global", 383 "zfs_flags", 384 "zfs_txg_timeout", 385 "zfs_vdev_cache_max", 386 "zfs_vdev_cache_size", 387 "zfs_vdev_cache_bshift", 388 "vdev_mirror_shift", 389 "zfs_scrub_limit", 390 "zfs_no_scrub_io", 391 "zfs_no_scrub_prefetch", 392 "zfs_vdev_aggregation_limit", 393 "fzap_default_block_shift", 394 "zfs_immediate_write_sz", 395 "zfs_read_chunk_size", 396 "zfs_nocacheflush", 397 "zil_replay_disable", 398 "metaslab_gang_bang", 399 "metaslab_df_alloc_threshold", 400 "metaslab_df_free_pct", 401 "zio_injection_enabled", 402 "zvol_immediate_write_sz", 403 }; 404 405 for (int i = 0; i < sizeof (params) / sizeof (params[0]); i++) { 406 int sz; 407 uint64_t val64; 408 uint32_t *val32p = (uint32_t *)&val64; 409 410 sz = mdb_readvar(&val64, params[i]); 411 if (sz == 4) { 412 mdb_printf("%s = 0x%x\n", params[i], *val32p); 413 } else if (sz == 8) { 414 mdb_printf("%s = 0x%llx\n", params[i], val64); 415 } else { 416 mdb_warn("variable %s not found", params[i]); 417 } 418 } 419 420 return (DCMD_OK); 421 } 422 423 /* ARGSUSED */ 424 static int 425 blkptr(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv) 426 { 427 char type[80], checksum[80], compress[80]; 428 blkptr_t blk, *bp = &blk; 429 char buf[BP_SPRINTF_LEN]; 430 431 if (mdb_vread(&blk, sizeof (blkptr_t), addr) == -1) { 432 mdb_warn("failed to read blkptr_t"); 433 return (DCMD_ERR); 434 } 435 436 if (enum_lookup("enum dmu_object_type", BP_GET_TYPE(bp), "DMU_OT_", 437 sizeof (type), type) == -1 || 438 enum_lookup("enum zio_checksum", BP_GET_CHECKSUM(bp), 439 "ZIO_CHECKSUM_", sizeof (checksum), checksum) == -1 || 440 enum_lookup("enum zio_compress", BP_GET_COMPRESS(bp), 441 "ZIO_COMPRESS_", sizeof (compress), compress) == -1) { 442 mdb_warn("Could not find blkptr enumerated types"); 443 return (DCMD_ERR); 444 } 445 446 SNPRINTF_BLKPTR(mdb_snprintf, '\n', buf, sizeof (buf), bp, type, 447 checksum, compress); 448 449 mdb_printf("%s\n", buf); 450 451 return (DCMD_OK); 452 } 453 454 typedef struct mdb_dmu_buf_impl { 455 struct { 456 uint64_t db_object; 457 uintptr_t db_data; 458 } db; 459 uintptr_t db_objset; 460 uint64_t db_level; 461 uint64_t db_blkid; 462 struct { 463 uint64_t rc_count; 464 } db_holds; 465 } mdb_dmu_buf_impl_t; 466 467 /* ARGSUSED */ 468 static int 469 dbuf(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv) 470 { 471 mdb_dmu_buf_impl_t db; 472 char objectname[32]; 473 char blkidname[32]; 474 char path[ZFS_MAX_DATASET_NAME_LEN]; 475 int ptr_width = (int)(sizeof (void *)) * 2; 476 477 if (DCMD_HDRSPEC(flags)) 478 mdb_printf("%*s %8s %3s %9s %5s %s\n", 479 ptr_width, "addr", "object", "lvl", "blkid", "holds", "os"); 480 481 if (mdb_ctf_vread(&db, ZFS_STRUCT "dmu_buf_impl", "mdb_dmu_buf_impl_t", 482 addr, 0) == -1) 483 return (DCMD_ERR); 484 485 if (db.db.db_object == DMU_META_DNODE_OBJECT) 486 (void) strcpy(objectname, "mdn"); 487 else 488 (void) mdb_snprintf(objectname, sizeof (objectname), "%llx", 489 (u_longlong_t)db.db.db_object); 490 491 if (db.db_blkid == DMU_BONUS_BLKID) 492 (void) strcpy(blkidname, "bonus"); 493 else 494 (void) mdb_snprintf(blkidname, sizeof (blkidname), "%llx", 495 (u_longlong_t)db.db_blkid); 496 497 if (objset_name(db.db_objset, path)) { 498 return (DCMD_ERR); 499 } 500 501 mdb_printf("%*p %8s %3u %9s %5llu %s\n", ptr_width, addr, 502 objectname, (int)db.db_level, blkidname, 503 db.db_holds.rc_count, path); 504 505 return (DCMD_OK); 506 } 507 508 /* ARGSUSED */ 509 static int 510 dbuf_stats(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv) 511 { 512 #define HISTOSZ 32 513 uintptr_t dbp; 514 dmu_buf_impl_t db; 515 dbuf_hash_table_t ht; 516 uint64_t bucket, ndbufs; 517 uint64_t histo[HISTOSZ]; 518 uint64_t histo2[HISTOSZ]; 519 int i, maxidx; 520 521 if (mdb_readvar(&ht, "dbuf_hash_table") == -1) { 522 mdb_warn("failed to read 'dbuf_hash_table'"); 523 return (DCMD_ERR); 524 } 525 526 for (i = 0; i < HISTOSZ; i++) { 527 histo[i] = 0; 528 histo2[i] = 0; 529 } 530 531 ndbufs = 0; 532 for (bucket = 0; bucket < ht.hash_table_mask+1; bucket++) { 533 int len; 534 535 if (mdb_vread(&dbp, sizeof (void *), 536 (uintptr_t)(ht.hash_table+bucket)) == -1) { 537 mdb_warn("failed to read hash bucket %u at %p", 538 bucket, ht.hash_table+bucket); 539 return (DCMD_ERR); 540 } 541 542 len = 0; 543 while (dbp != 0) { 544 if (mdb_vread(&db, sizeof (dmu_buf_impl_t), 545 dbp) == -1) { 546 mdb_warn("failed to read dbuf at %p", dbp); 547 return (DCMD_ERR); 548 } 549 dbp = (uintptr_t)db.db_hash_next; 550 for (i = MIN(len, HISTOSZ - 1); i >= 0; i--) 551 histo2[i]++; 552 len++; 553 ndbufs++; 554 } 555 556 if (len >= HISTOSZ) 557 len = HISTOSZ-1; 558 histo[len]++; 559 } 560 561 mdb_printf("hash table has %llu buckets, %llu dbufs " 562 "(avg %llu buckets/dbuf)\n", 563 ht.hash_table_mask+1, ndbufs, 564 (ht.hash_table_mask+1)/ndbufs); 565 566 mdb_printf("\n"); 567 maxidx = 0; 568 for (i = 0; i < HISTOSZ; i++) 569 if (histo[i] > 0) 570 maxidx = i; 571 mdb_printf("hash chain length number of buckets\n"); 572 for (i = 0; i <= maxidx; i++) 573 mdb_printf("%u %llu\n", i, histo[i]); 574 575 mdb_printf("\n"); 576 maxidx = 0; 577 for (i = 0; i < HISTOSZ; i++) 578 if (histo2[i] > 0) 579 maxidx = i; 580 mdb_printf("hash chain depth number of dbufs\n"); 581 for (i = 0; i <= maxidx; i++) 582 mdb_printf("%u or more %llu %llu%%\n", 583 i, histo2[i], histo2[i]*100/ndbufs); 584 585 586 return (DCMD_OK); 587 } 588 589 #define CHAIN_END 0xffff 590 /* 591 * ::zap_leaf [-v] 592 * 593 * Print a zap_leaf_phys_t, assumed to be 16k 594 */ 595 /* ARGSUSED */ 596 static int 597 zap_leaf(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv) 598 { 599 char buf[16*1024]; 600 int verbose = B_FALSE; 601 int four = B_FALSE; 602 dmu_buf_t l_dbuf; 603 zap_leaf_t l; 604 zap_leaf_phys_t *zlp = (void *)buf; 605 int i; 606 607 if (mdb_getopts(argc, argv, 608 'v', MDB_OPT_SETBITS, TRUE, &verbose, 609 '4', MDB_OPT_SETBITS, TRUE, &four, 610 NULL) != argc) 611 return (DCMD_USAGE); 612 613 l_dbuf.db_data = zlp; 614 l.l_dbuf = &l_dbuf; 615 l.l_bs = 14; /* assume 16k blocks */ 616 if (four) 617 l.l_bs = 12; 618 619 if (!(flags & DCMD_ADDRSPEC)) { 620 return (DCMD_USAGE); 621 } 622 623 if (mdb_vread(buf, sizeof (buf), addr) == -1) { 624 mdb_warn("failed to read zap_leaf_phys_t at %p", addr); 625 return (DCMD_ERR); 626 } 627 628 if (zlp->l_hdr.lh_block_type != ZBT_LEAF || 629 zlp->l_hdr.lh_magic != ZAP_LEAF_MAGIC) { 630 mdb_warn("This does not appear to be a zap_leaf_phys_t"); 631 return (DCMD_ERR); 632 } 633 634 mdb_printf("zap_leaf_phys_t at %p:\n", addr); 635 mdb_printf(" lh_prefix_len = %u\n", zlp->l_hdr.lh_prefix_len); 636 mdb_printf(" lh_prefix = %llx\n", zlp->l_hdr.lh_prefix); 637 mdb_printf(" lh_nentries = %u\n", zlp->l_hdr.lh_nentries); 638 mdb_printf(" lh_nfree = %u\n", zlp->l_hdr.lh_nfree, 639 zlp->l_hdr.lh_nfree * 100 / (ZAP_LEAF_NUMCHUNKS(&l))); 640 mdb_printf(" lh_freelist = %u\n", zlp->l_hdr.lh_freelist); 641 mdb_printf(" lh_flags = %x (%s)\n", zlp->l_hdr.lh_flags, 642 zlp->l_hdr.lh_flags & ZLF_ENTRIES_CDSORTED ? 643 "ENTRIES_CDSORTED" : ""); 644 645 if (verbose) { 646 mdb_printf(" hash table:\n"); 647 for (i = 0; i < ZAP_LEAF_HASH_NUMENTRIES(&l); i++) { 648 if (zlp->l_hash[i] != CHAIN_END) 649 mdb_printf(" %u: %u\n", i, zlp->l_hash[i]); 650 } 651 } 652 653 mdb_printf(" chunks:\n"); 654 for (i = 0; i < ZAP_LEAF_NUMCHUNKS(&l); i++) { 655 /* LINTED: alignment */ 656 zap_leaf_chunk_t *zlc = &ZAP_LEAF_CHUNK(&l, i); 657 switch (zlc->l_entry.le_type) { 658 case ZAP_CHUNK_FREE: 659 if (verbose) { 660 mdb_printf(" %u: free; lf_next = %u\n", 661 i, zlc->l_free.lf_next); 662 } 663 break; 664 case ZAP_CHUNK_ENTRY: 665 mdb_printf(" %u: entry\n", i); 666 if (verbose) { 667 mdb_printf(" le_next = %u\n", 668 zlc->l_entry.le_next); 669 } 670 mdb_printf(" le_name_chunk = %u\n", 671 zlc->l_entry.le_name_chunk); 672 mdb_printf(" le_name_numints = %u\n", 673 zlc->l_entry.le_name_numints); 674 mdb_printf(" le_value_chunk = %u\n", 675 zlc->l_entry.le_value_chunk); 676 mdb_printf(" le_value_intlen = %u\n", 677 zlc->l_entry.le_value_intlen); 678 mdb_printf(" le_value_numints = %u\n", 679 zlc->l_entry.le_value_numints); 680 mdb_printf(" le_cd = %u\n", 681 zlc->l_entry.le_cd); 682 mdb_printf(" le_hash = %llx\n", 683 zlc->l_entry.le_hash); 684 break; 685 case ZAP_CHUNK_ARRAY: 686 mdb_printf(" %u: array", i); 687 if (strisprint((char *)zlc->l_array.la_array)) 688 mdb_printf(" \"%s\"", zlc->l_array.la_array); 689 mdb_printf("\n"); 690 if (verbose) { 691 int j; 692 mdb_printf(" "); 693 for (j = 0; j < ZAP_LEAF_ARRAY_BYTES; j++) { 694 mdb_printf("%02x ", 695 zlc->l_array.la_array[j]); 696 } 697 mdb_printf("\n"); 698 } 699 if (zlc->l_array.la_next != CHAIN_END) { 700 mdb_printf(" lf_next = %u\n", 701 zlc->l_array.la_next); 702 } 703 break; 704 default: 705 mdb_printf(" %u: undefined type %u\n", 706 zlc->l_entry.le_type); 707 } 708 } 709 710 return (DCMD_OK); 711 } 712 713 typedef struct dbufs_data { 714 mdb_ctf_id_t id; 715 uint64_t objset; 716 uint64_t object; 717 uint64_t level; 718 uint64_t blkid; 719 char *osname; 720 } dbufs_data_t; 721 722 #define DBUFS_UNSET (0xbaddcafedeadbeefULL) 723 724 /* ARGSUSED */ 725 static int 726 dbufs_cb(uintptr_t addr, const void *unknown, void *arg) 727 { 728 dbufs_data_t *data = arg; 729 uintptr_t objset; 730 dmu_buf_t db; 731 uint8_t level; 732 uint64_t blkid; 733 char osname[ZFS_MAX_DATASET_NAME_LEN]; 734 735 if (GETMEMBID(addr, &data->id, db_objset, objset) || 736 GETMEMBID(addr, &data->id, db, db) || 737 GETMEMBID(addr, &data->id, db_level, level) || 738 GETMEMBID(addr, &data->id, db_blkid, blkid)) { 739 return (WALK_ERR); 740 } 741 742 if ((data->objset == DBUFS_UNSET || data->objset == objset) && 743 (data->osname == NULL || (objset_name(objset, osname) == 0 && 744 strcmp(data->osname, osname) == 0)) && 745 (data->object == DBUFS_UNSET || data->object == db.db_object) && 746 (data->level == DBUFS_UNSET || data->level == level) && 747 (data->blkid == DBUFS_UNSET || data->blkid == blkid)) { 748 mdb_printf("%#lr\n", addr); 749 } 750 return (WALK_NEXT); 751 } 752 753 /* ARGSUSED */ 754 static int 755 dbufs(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv) 756 { 757 dbufs_data_t data; 758 char *object = NULL; 759 char *blkid = NULL; 760 761 data.objset = data.object = data.level = data.blkid = DBUFS_UNSET; 762 data.osname = NULL; 763 764 if (mdb_getopts(argc, argv, 765 'O', MDB_OPT_UINT64, &data.objset, 766 'n', MDB_OPT_STR, &data.osname, 767 'o', MDB_OPT_STR, &object, 768 'l', MDB_OPT_UINT64, &data.level, 769 'b', MDB_OPT_STR, &blkid) != argc) { 770 return (DCMD_USAGE); 771 } 772 773 if (object) { 774 if (strcmp(object, "mdn") == 0) { 775 data.object = DMU_META_DNODE_OBJECT; 776 } else { 777 data.object = mdb_strtoull(object); 778 } 779 } 780 781 if (blkid) { 782 if (strcmp(blkid, "bonus") == 0) { 783 data.blkid = DMU_BONUS_BLKID; 784 } else { 785 data.blkid = mdb_strtoull(blkid); 786 } 787 } 788 789 if (mdb_ctf_lookup_by_name(ZFS_STRUCT "dmu_buf_impl", &data.id) == -1) { 790 mdb_warn("couldn't find struct dmu_buf_impl_t"); 791 return (DCMD_ERR); 792 } 793 794 if (mdb_walk("dmu_buf_impl_t", dbufs_cb, &data) != 0) { 795 mdb_warn("can't walk dbufs"); 796 return (DCMD_ERR); 797 } 798 799 return (DCMD_OK); 800 } 801 802 typedef struct abuf_find_data { 803 dva_t dva; 804 mdb_ctf_id_t id; 805 } abuf_find_data_t; 806 807 /* ARGSUSED */ 808 static int 809 abuf_find_cb(uintptr_t addr, const void *unknown, void *arg) 810 { 811 abuf_find_data_t *data = arg; 812 dva_t dva; 813 814 if (GETMEMBID(addr, &data->id, b_dva, dva)) { 815 return (WALK_ERR); 816 } 817 818 if (dva.dva_word[0] == data->dva.dva_word[0] && 819 dva.dva_word[1] == data->dva.dva_word[1]) { 820 mdb_printf("%#lr\n", addr); 821 } 822 return (WALK_NEXT); 823 } 824 825 /* ARGSUSED */ 826 static int 827 abuf_find(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv) 828 { 829 abuf_find_data_t data; 830 GElf_Sym sym; 831 int i; 832 const char *syms[] = { 833 "ARC_mru", 834 "ARC_mru_ghost", 835 "ARC_mfu", 836 "ARC_mfu_ghost", 837 }; 838 839 if (argc != 2) 840 return (DCMD_USAGE); 841 842 for (i = 0; i < 2; i ++) { 843 switch (argv[i].a_type) { 844 case MDB_TYPE_STRING: 845 data.dva.dva_word[i] = mdb_strtoull(argv[i].a_un.a_str); 846 break; 847 case MDB_TYPE_IMMEDIATE: 848 data.dva.dva_word[i] = argv[i].a_un.a_val; 849 break; 850 default: 851 return (DCMD_USAGE); 852 } 853 } 854 855 if (mdb_ctf_lookup_by_name(ZFS_STRUCT "arc_buf_hdr", &data.id) == -1) { 856 mdb_warn("couldn't find struct arc_buf_hdr"); 857 return (DCMD_ERR); 858 } 859 860 for (i = 0; i < sizeof (syms) / sizeof (syms[0]); i++) { 861 if (mdb_lookup_by_obj(ZFS_OBJ_NAME, syms[i], &sym)) { 862 mdb_warn("can't find symbol %s", syms[i]); 863 return (DCMD_ERR); 864 } 865 866 if (mdb_pwalk("list", abuf_find_cb, &data, sym.st_value) != 0) { 867 mdb_warn("can't walk %s", syms[i]); 868 return (DCMD_ERR); 869 } 870 } 871 872 return (DCMD_OK); 873 } 874 875 876 typedef struct dbgmsg_arg { 877 boolean_t da_verbose; 878 boolean_t da_address; 879 } dbgmsg_arg_t; 880 881 /* ARGSUSED */ 882 static int 883 dbgmsg_cb(uintptr_t addr, const void *unknown, void *arg) 884 { 885 static mdb_ctf_id_t id; 886 static boolean_t gotid; 887 static ulong_t off; 888 889 dbgmsg_arg_t *da = arg; 890 time_t timestamp; 891 char buf[1024]; 892 893 if (!gotid) { 894 if (mdb_ctf_lookup_by_name(ZFS_STRUCT "zfs_dbgmsg", &id) == 895 -1) { 896 mdb_warn("couldn't find struct zfs_dbgmsg"); 897 return (WALK_ERR); 898 } 899 gotid = TRUE; 900 if (mdb_ctf_offsetof(id, "zdm_msg", &off) == -1) { 901 mdb_warn("couldn't find zdm_msg"); 902 return (WALK_ERR); 903 } 904 off /= 8; 905 } 906 907 908 if (GETMEMBID(addr, &id, zdm_timestamp, timestamp)) { 909 return (WALK_ERR); 910 } 911 912 if (mdb_readstr(buf, sizeof (buf), addr + off) == -1) { 913 mdb_warn("failed to read zdm_msg at %p\n", addr + off); 914 return (DCMD_ERR); 915 } 916 917 if (da->da_address) 918 mdb_printf("%p ", addr); 919 if (da->da_verbose) 920 mdb_printf("%Y ", timestamp); 921 922 mdb_printf("%s\n", buf); 923 924 if (da->da_verbose) 925 (void) mdb_call_dcmd("whatis", addr, DCMD_ADDRSPEC, 0, NULL); 926 927 return (WALK_NEXT); 928 } 929 930 /* ARGSUSED */ 931 static int 932 dbgmsg(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv) 933 { 934 GElf_Sym sym; 935 dbgmsg_arg_t da = { 0 }; 936 937 if (mdb_getopts(argc, argv, 938 'v', MDB_OPT_SETBITS, B_TRUE, &da.da_verbose, 939 'a', MDB_OPT_SETBITS, B_TRUE, &da.da_address, 940 NULL) != argc) 941 return (DCMD_USAGE); 942 943 if (mdb_lookup_by_obj(ZFS_OBJ_NAME, "zfs_dbgmsgs", &sym)) { 944 mdb_warn("can't find zfs_dbgmsgs"); 945 return (DCMD_ERR); 946 } 947 948 if (mdb_pwalk("list", dbgmsg_cb, &da, sym.st_value) != 0) { 949 mdb_warn("can't walk zfs_dbgmsgs"); 950 return (DCMD_ERR); 951 } 952 953 return (DCMD_OK); 954 } 955 956 /*ARGSUSED*/ 957 static int 958 arc_print(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv) 959 { 960 kstat_named_t *stats; 961 GElf_Sym sym; 962 int nstats, i; 963 uint_t opt_a = FALSE; 964 uint_t opt_b = FALSE; 965 uint_t shift = 0; 966 const char *suffix; 967 968 static const char *bytestats[] = { 969 "p", "c", "c_min", "c_max", "size", "duplicate_buffers_size", 970 "arc_meta_used", "arc_meta_limit", "arc_meta_max", 971 "arc_meta_min", "hdr_size", "data_size", "metadata_size", 972 "other_size", "anon_size", "anon_evictable_data", 973 "anon_evictable_metadata", "mru_size", "mru_evictable_data", 974 "mru_evictable_metadata", "mru_ghost_size", 975 "mru_ghost_evictable_data", "mru_ghost_evictable_metadata", 976 "mfu_size", "mfu_evictable_data", "mfu_evictable_metadata", 977 "mfu_ghost_size", "mfu_ghost_evictable_data", 978 "mfu_ghost_evictable_metadata", "evict_l2_cached", 979 "evict_l2_eligible", "evict_l2_ineligible", "l2_read_bytes", 980 "l2_write_bytes", "l2_size", "l2_asize", "l2_hdr_size", 981 "compressed_size", "uncompressed_size", "overhead_size", 982 NULL 983 }; 984 985 static const char *extras[] = { 986 "arc_no_grow", "arc_tempreserve", 987 NULL 988 }; 989 990 if (mdb_lookup_by_obj(ZFS_OBJ_NAME, "arc_stats", &sym) == -1) { 991 mdb_warn("failed to find 'arc_stats'"); 992 return (DCMD_ERR); 993 } 994 995 stats = mdb_zalloc(sym.st_size, UM_SLEEP | UM_GC); 996 997 if (mdb_vread(stats, sym.st_size, sym.st_value) == -1) { 998 mdb_warn("couldn't read 'arc_stats' at %p", sym.st_value); 999 return (DCMD_ERR); 1000 } 1001 1002 nstats = sym.st_size / sizeof (kstat_named_t); 1003 1004 /* NB: -a / opt_a are ignored for backwards compatability */ 1005 if (mdb_getopts(argc, argv, 1006 'a', MDB_OPT_SETBITS, TRUE, &opt_a, 1007 'b', MDB_OPT_SETBITS, TRUE, &opt_b, 1008 'k', MDB_OPT_SETBITS, 10, &shift, 1009 'm', MDB_OPT_SETBITS, 20, &shift, 1010 'g', MDB_OPT_SETBITS, 30, &shift, 1011 NULL) != argc) 1012 return (DCMD_USAGE); 1013 1014 if (!opt_b && !shift) 1015 shift = 20; 1016 1017 switch (shift) { 1018 case 0: 1019 suffix = "B"; 1020 break; 1021 case 10: 1022 suffix = "KB"; 1023 break; 1024 case 20: 1025 suffix = "MB"; 1026 break; 1027 case 30: 1028 suffix = "GB"; 1029 break; 1030 default: 1031 suffix = "XX"; 1032 } 1033 1034 for (i = 0; i < nstats; i++) { 1035 int j; 1036 boolean_t bytes = B_FALSE; 1037 1038 for (j = 0; bytestats[j]; j++) { 1039 if (strcmp(stats[i].name, bytestats[j]) == 0) { 1040 bytes = B_TRUE; 1041 break; 1042 } 1043 } 1044 1045 if (bytes) { 1046 mdb_printf("%-25s = %9llu %s\n", stats[i].name, 1047 stats[i].value.ui64 >> shift, suffix); 1048 } else { 1049 mdb_printf("%-25s = %9llu\n", stats[i].name, 1050 stats[i].value.ui64); 1051 } 1052 } 1053 1054 for (i = 0; extras[i]; i++) { 1055 uint64_t buf; 1056 1057 if (mdb_lookup_by_obj(ZFS_OBJ_NAME, extras[i], &sym) == -1) { 1058 mdb_warn("failed to find '%s'", extras[i]); 1059 return (DCMD_ERR); 1060 } 1061 1062 if (sym.st_size != sizeof (uint64_t) && 1063 sym.st_size != sizeof (uint32_t)) { 1064 mdb_warn("expected scalar for variable '%s'\n", 1065 extras[i]); 1066 return (DCMD_ERR); 1067 } 1068 1069 if (mdb_vread(&buf, sym.st_size, sym.st_value) == -1) { 1070 mdb_warn("couldn't read '%s'", extras[i]); 1071 return (DCMD_ERR); 1072 } 1073 1074 mdb_printf("%-25s = ", extras[i]); 1075 1076 /* NB: all the 64-bit extras happen to be byte counts */ 1077 if (sym.st_size == sizeof (uint64_t)) 1078 mdb_printf("%9llu %s\n", buf >> shift, suffix); 1079 1080 if (sym.st_size == sizeof (uint32_t)) 1081 mdb_printf("%9d\n", *((uint32_t *)&buf)); 1082 } 1083 return (DCMD_OK); 1084 } 1085 1086 typedef struct mdb_spa_print { 1087 pool_state_t spa_state; 1088 char spa_name[ZFS_MAX_DATASET_NAME_LEN]; 1089 uintptr_t spa_normal_class; 1090 } mdb_spa_print_t; 1091 1092 1093 const char histo_stars[] = "****************************************"; 1094 const int histo_width = sizeof (histo_stars) - 1; 1095 1096 static void 1097 dump_histogram(const uint64_t *histo, int size, int offset) 1098 { 1099 int i; 1100 int minidx = size - 1; 1101 int maxidx = 0; 1102 uint64_t max = 0; 1103 1104 for (i = 0; i < size; i++) { 1105 if (histo[i] > max) 1106 max = histo[i]; 1107 if (histo[i] > 0 && i > maxidx) 1108 maxidx = i; 1109 if (histo[i] > 0 && i < minidx) 1110 minidx = i; 1111 } 1112 1113 if (max < histo_width) 1114 max = histo_width; 1115 1116 for (i = minidx; i <= maxidx; i++) { 1117 mdb_printf("%3u: %6llu %s\n", 1118 i + offset, (u_longlong_t)histo[i], 1119 &histo_stars[(max - histo[i]) * histo_width / max]); 1120 } 1121 } 1122 1123 typedef struct mdb_metaslab_class { 1124 uint64_t mc_histogram[RANGE_TREE_HISTOGRAM_SIZE]; 1125 } mdb_metaslab_class_t; 1126 1127 /* 1128 * spa_class_histogram(uintptr_t class_addr) 1129 * 1130 * Prints free space histogram for a device class 1131 * 1132 * Returns DCMD_OK, or DCMD_ERR. 1133 */ 1134 static int 1135 spa_class_histogram(uintptr_t class_addr) 1136 { 1137 mdb_metaslab_class_t mc; 1138 if (mdb_ctf_vread(&mc, "metaslab_class_t", 1139 "mdb_metaslab_class_t", class_addr, 0) == -1) 1140 return (DCMD_ERR); 1141 1142 mdb_inc_indent(4); 1143 dump_histogram(mc.mc_histogram, RANGE_TREE_HISTOGRAM_SIZE, 0); 1144 mdb_dec_indent(4); 1145 return (DCMD_OK); 1146 } 1147 1148 /* 1149 * ::spa 1150 * 1151 * -c Print configuration information as well 1152 * -v Print vdev state 1153 * -e Print vdev error stats 1154 * -m Print vdev metaslab info 1155 * -M print vdev metaslab group info 1156 * -h Print histogram info (must be combined with -m or -M) 1157 * 1158 * Print a summarized spa_t. When given no arguments, prints out a table of all 1159 * active pools on the system. 1160 */ 1161 /* ARGSUSED */ 1162 static int 1163 spa_print(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv) 1164 { 1165 const char *statetab[] = { "ACTIVE", "EXPORTED", "DESTROYED", 1166 "SPARE", "L2CACHE", "UNINIT", "UNAVAIL", "POTENTIAL" }; 1167 const char *state; 1168 int spa_flags = 0; 1169 1170 if (mdb_getopts(argc, argv, 1171 'c', MDB_OPT_SETBITS, SPA_FLAG_CONFIG, &spa_flags, 1172 'v', MDB_OPT_SETBITS, SPA_FLAG_VDEVS, &spa_flags, 1173 'e', MDB_OPT_SETBITS, SPA_FLAG_ERRORS, &spa_flags, 1174 'M', MDB_OPT_SETBITS, SPA_FLAG_METASLAB_GROUPS, &spa_flags, 1175 'm', MDB_OPT_SETBITS, SPA_FLAG_METASLABS, &spa_flags, 1176 'h', MDB_OPT_SETBITS, SPA_FLAG_HISTOGRAMS, &spa_flags, 1177 NULL) != argc) 1178 return (DCMD_USAGE); 1179 1180 if (!(flags & DCMD_ADDRSPEC)) { 1181 if (mdb_walk_dcmd("spa", "spa", argc, argv) == -1) { 1182 mdb_warn("can't walk spa"); 1183 return (DCMD_ERR); 1184 } 1185 1186 return (DCMD_OK); 1187 } 1188 1189 if (flags & DCMD_PIPE_OUT) { 1190 mdb_printf("%#lr\n", addr); 1191 return (DCMD_OK); 1192 } 1193 1194 if (DCMD_HDRSPEC(flags)) 1195 mdb_printf("%<u>%-?s %9s %-*s%</u>\n", "ADDR", "STATE", 1196 sizeof (uintptr_t) == 4 ? 60 : 52, "NAME"); 1197 1198 mdb_spa_print_t spa; 1199 if (mdb_ctf_vread(&spa, "spa_t", "mdb_spa_print_t", addr, 0) == -1) 1200 return (DCMD_ERR); 1201 1202 if (spa.spa_state < 0 || spa.spa_state > POOL_STATE_UNAVAIL) 1203 state = "UNKNOWN"; 1204 else 1205 state = statetab[spa.spa_state]; 1206 1207 mdb_printf("%0?p %9s %s\n", addr, state, spa.spa_name); 1208 if (spa_flags & SPA_FLAG_HISTOGRAMS) 1209 spa_class_histogram(spa.spa_normal_class); 1210 1211 if (spa_flags & SPA_FLAG_CONFIG) { 1212 mdb_printf("\n"); 1213 mdb_inc_indent(4); 1214 if (mdb_call_dcmd("spa_config", addr, flags, 0, 1215 NULL) != DCMD_OK) 1216 return (DCMD_ERR); 1217 mdb_dec_indent(4); 1218 } 1219 1220 if (spa_flags & SPA_FLAG_ALL_VDEV) { 1221 mdb_arg_t v; 1222 char opts[100] = "-"; 1223 int args = 1224 (spa_flags | SPA_FLAG_VDEVS) == SPA_FLAG_VDEVS ? 0 : 1; 1225 1226 if (spa_flags & SPA_FLAG_ERRORS) 1227 strcat(opts, "e"); 1228 if (spa_flags & SPA_FLAG_METASLABS) 1229 strcat(opts, "m"); 1230 if (spa_flags & SPA_FLAG_METASLAB_GROUPS) 1231 strcat(opts, "M"); 1232 if (spa_flags & SPA_FLAG_HISTOGRAMS) 1233 strcat(opts, "h"); 1234 1235 v.a_type = MDB_TYPE_STRING; 1236 v.a_un.a_str = opts; 1237 1238 mdb_printf("\n"); 1239 mdb_inc_indent(4); 1240 if (mdb_call_dcmd("spa_vdevs", addr, flags, args, 1241 &v) != DCMD_OK) 1242 return (DCMD_ERR); 1243 mdb_dec_indent(4); 1244 } 1245 1246 return (DCMD_OK); 1247 } 1248 1249 typedef struct mdb_spa_config_spa { 1250 uintptr_t spa_config; 1251 } mdb_spa_config_spa_t; 1252 1253 /* 1254 * ::spa_config 1255 * 1256 * Given a spa_t, print the configuration information stored in spa_config. 1257 * Since it's just an nvlist, format it as an indented list of name=value pairs. 1258 * We simply read the value of spa_config and pass off to ::nvlist. 1259 */ 1260 /* ARGSUSED */ 1261 static int 1262 spa_print_config(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv) 1263 { 1264 mdb_spa_config_spa_t spa; 1265 1266 if (argc != 0 || !(flags & DCMD_ADDRSPEC)) 1267 return (DCMD_USAGE); 1268 1269 if (mdb_ctf_vread(&spa, ZFS_STRUCT "spa", "mdb_spa_config_spa_t", 1270 addr, 0) == -1) 1271 return (DCMD_ERR); 1272 1273 if (spa.spa_config == 0) { 1274 mdb_printf("(none)\n"); 1275 return (DCMD_OK); 1276 } 1277 1278 return (mdb_call_dcmd("nvlist", spa.spa_config, flags, 1279 0, NULL)); 1280 } 1281 1282 1283 1284 typedef struct mdb_range_tree { 1285 uint64_t rt_space; 1286 } mdb_range_tree_t; 1287 1288 typedef struct mdb_metaslab_group { 1289 uint64_t mg_fragmentation; 1290 uint64_t mg_histogram[RANGE_TREE_HISTOGRAM_SIZE]; 1291 uintptr_t mg_vd; 1292 } mdb_metaslab_group_t; 1293 1294 typedef struct mdb_metaslab { 1295 uint64_t ms_id; 1296 uint64_t ms_start; 1297 uint64_t ms_size; 1298 int64_t ms_deferspace; 1299 uint64_t ms_fragmentation; 1300 uint64_t ms_weight; 1301 uintptr_t ms_alloctree[TXG_SIZE]; 1302 uintptr_t ms_freetree[TXG_SIZE]; 1303 uintptr_t ms_tree; 1304 uintptr_t ms_sm; 1305 } mdb_metaslab_t; 1306 1307 typedef struct mdb_space_map_phys_t { 1308 uint64_t smp_alloc; 1309 uint64_t smp_histogram[SPACE_MAP_HISTOGRAM_SIZE]; 1310 } mdb_space_map_phys_t; 1311 1312 typedef struct mdb_space_map { 1313 uint64_t sm_size; 1314 uint8_t sm_shift; 1315 uint64_t sm_alloc; 1316 uintptr_t sm_phys; 1317 } mdb_space_map_t; 1318 1319 typedef struct mdb_vdev { 1320 uintptr_t vdev_path; 1321 uintptr_t vdev_ms; 1322 uintptr_t vdev_ops; 1323 uint64_t vdev_ms_count; 1324 uint64_t vdev_id; 1325 vdev_stat_t vdev_stat; 1326 } mdb_vdev_t; 1327 1328 typedef struct mdb_vdev_ops { 1329 char vdev_op_type[16]; 1330 } mdb_vdev_ops_t; 1331 1332 static int 1333 metaslab_stats(uintptr_t addr, int spa_flags) 1334 { 1335 mdb_vdev_t vdev; 1336 uintptr_t *vdev_ms; 1337 1338 if (mdb_ctf_vread(&vdev, "vdev_t", "mdb_vdev_t", 1339 (uintptr_t)addr, 0) == -1) { 1340 mdb_warn("failed to read vdev at %p\n", addr); 1341 return (DCMD_ERR); 1342 } 1343 1344 mdb_inc_indent(4); 1345 mdb_printf("%<u>%-?s %6s %20s %10s %9s%</u>\n", "ADDR", "ID", 1346 "OFFSET", "FREE", "FRAGMENTATION"); 1347 1348 vdev_ms = mdb_alloc(vdev.vdev_ms_count * sizeof (void *), 1349 UM_SLEEP | UM_GC); 1350 if (mdb_vread(vdev_ms, vdev.vdev_ms_count * sizeof (void *), 1351 (uintptr_t)vdev.vdev_ms) == -1) { 1352 mdb_warn("failed to read vdev_ms at %p\n", vdev.vdev_ms); 1353 return (DCMD_ERR); 1354 } 1355 1356 for (int m = 0; m < vdev.vdev_ms_count; m++) { 1357 mdb_metaslab_t ms; 1358 mdb_space_map_t sm = { 0 }; 1359 char free[NICENUM_BUFLEN]; 1360 1361 if (mdb_ctf_vread(&ms, "metaslab_t", "mdb_metaslab_t", 1362 (uintptr_t)vdev_ms[m], 0) == -1) 1363 return (DCMD_ERR); 1364 1365 if (ms.ms_sm != NULL && 1366 mdb_ctf_vread(&sm, "space_map_t", "mdb_space_map_t", 1367 ms.ms_sm, 0) == -1) 1368 return (DCMD_ERR); 1369 1370 mdb_nicenum(ms.ms_size - sm.sm_alloc, free); 1371 1372 mdb_printf("%0?p %6llu %20llx %10s ", vdev_ms[m], ms.ms_id, 1373 ms.ms_start, free); 1374 if (ms.ms_fragmentation == ZFS_FRAG_INVALID) 1375 mdb_printf("%9s\n", "-"); 1376 else 1377 mdb_printf("%9llu%%\n", ms.ms_fragmentation); 1378 1379 if ((spa_flags & SPA_FLAG_HISTOGRAMS) && ms.ms_sm != NULL) { 1380 mdb_space_map_phys_t smp; 1381 1382 if (sm.sm_phys == NULL) 1383 continue; 1384 1385 (void) mdb_ctf_vread(&smp, "space_map_phys_t", 1386 "mdb_space_map_phys_t", sm.sm_phys, 0); 1387 1388 dump_histogram(smp.smp_histogram, 1389 SPACE_MAP_HISTOGRAM_SIZE, sm.sm_shift); 1390 } 1391 } 1392 mdb_dec_indent(4); 1393 return (DCMD_OK); 1394 } 1395 1396 static int 1397 metaslab_group_stats(uintptr_t addr, int spa_flags) 1398 { 1399 mdb_metaslab_group_t mg; 1400 if (mdb_ctf_vread(&mg, "metaslab_group_t", "mdb_metaslab_group_t", 1401 (uintptr_t)addr, 0) == -1) { 1402 mdb_warn("failed to read vdev_mg at %p\n", addr); 1403 return (DCMD_ERR); 1404 } 1405 1406 mdb_inc_indent(4); 1407 mdb_printf("%<u>%-?s %15s%</u>\n", "ADDR", "FRAGMENTATION"); 1408 if (mg.mg_fragmentation == ZFS_FRAG_INVALID) 1409 mdb_printf("%0?p %15s\n", addr, "-"); 1410 else 1411 mdb_printf("%0?p %15llu%%\n", addr, mg.mg_fragmentation); 1412 1413 if (spa_flags & SPA_FLAG_HISTOGRAMS) 1414 dump_histogram(mg.mg_histogram, RANGE_TREE_HISTOGRAM_SIZE, 0); 1415 mdb_dec_indent(4); 1416 return (DCMD_OK); 1417 } 1418 1419 /* 1420 * ::vdev 1421 * 1422 * Print out a summarized vdev_t, in the following form: 1423 * 1424 * ADDR STATE AUX DESC 1425 * fffffffbcde23df0 HEALTHY - /dev/dsk/c0t0d0 1426 * 1427 * If '-r' is specified, recursively visit all children. 1428 * 1429 * With '-e', the statistics associated with the vdev are printed as well. 1430 */ 1431 static int 1432 do_print_vdev(uintptr_t addr, int flags, int depth, boolean_t recursive, 1433 int spa_flags) 1434 { 1435 vdev_t vdev; 1436 char desc[MAXNAMELEN]; 1437 int c, children; 1438 uintptr_t *child; 1439 const char *state, *aux; 1440 1441 if (mdb_vread(&vdev, sizeof (vdev), (uintptr_t)addr) == -1) { 1442 mdb_warn("failed to read vdev_t at %p\n", (uintptr_t)addr); 1443 return (DCMD_ERR); 1444 } 1445 1446 if (flags & DCMD_PIPE_OUT) { 1447 mdb_printf("%#lr\n", addr); 1448 } else { 1449 if (vdev.vdev_path != NULL) { 1450 if (mdb_readstr(desc, sizeof (desc), 1451 (uintptr_t)vdev.vdev_path) == -1) { 1452 mdb_warn("failed to read vdev_path at %p\n", 1453 vdev.vdev_path); 1454 return (DCMD_ERR); 1455 } 1456 } else if (vdev.vdev_ops != NULL) { 1457 vdev_ops_t ops; 1458 if (mdb_vread(&ops, sizeof (ops), 1459 (uintptr_t)vdev.vdev_ops) == -1) { 1460 mdb_warn("failed to read vdev_ops at %p\n", 1461 vdev.vdev_ops); 1462 return (DCMD_ERR); 1463 } 1464 (void) strcpy(desc, ops.vdev_op_type); 1465 } else { 1466 (void) strcpy(desc, "<unknown>"); 1467 } 1468 1469 if (depth == 0 && DCMD_HDRSPEC(flags)) 1470 mdb_printf("%<u>%-?s %-9s %-12s %-*s%</u>\n", 1471 "ADDR", "STATE", "AUX", 1472 sizeof (uintptr_t) == 4 ? 43 : 35, 1473 "DESCRIPTION"); 1474 1475 mdb_printf("%0?p ", addr); 1476 1477 switch (vdev.vdev_state) { 1478 case VDEV_STATE_CLOSED: 1479 state = "CLOSED"; 1480 break; 1481 case VDEV_STATE_OFFLINE: 1482 state = "OFFLINE"; 1483 break; 1484 case VDEV_STATE_CANT_OPEN: 1485 state = "CANT_OPEN"; 1486 break; 1487 case VDEV_STATE_DEGRADED: 1488 state = "DEGRADED"; 1489 break; 1490 case VDEV_STATE_HEALTHY: 1491 state = "HEALTHY"; 1492 break; 1493 case VDEV_STATE_REMOVED: 1494 state = "REMOVED"; 1495 break; 1496 case VDEV_STATE_FAULTED: 1497 state = "FAULTED"; 1498 break; 1499 default: 1500 state = "UNKNOWN"; 1501 break; 1502 } 1503 1504 switch (vdev.vdev_stat.vs_aux) { 1505 case VDEV_AUX_NONE: 1506 aux = "-"; 1507 break; 1508 case VDEV_AUX_OPEN_FAILED: 1509 aux = "OPEN_FAILED"; 1510 break; 1511 case VDEV_AUX_CORRUPT_DATA: 1512 aux = "CORRUPT_DATA"; 1513 break; 1514 case VDEV_AUX_NO_REPLICAS: 1515 aux = "NO_REPLICAS"; 1516 break; 1517 case VDEV_AUX_BAD_GUID_SUM: 1518 aux = "BAD_GUID_SUM"; 1519 break; 1520 case VDEV_AUX_TOO_SMALL: 1521 aux = "TOO_SMALL"; 1522 break; 1523 case VDEV_AUX_BAD_LABEL: 1524 aux = "BAD_LABEL"; 1525 break; 1526 case VDEV_AUX_VERSION_NEWER: 1527 aux = "VERS_NEWER"; 1528 break; 1529 case VDEV_AUX_VERSION_OLDER: 1530 aux = "VERS_OLDER"; 1531 break; 1532 case VDEV_AUX_UNSUP_FEAT: 1533 aux = "UNSUP_FEAT"; 1534 break; 1535 case VDEV_AUX_SPARED: 1536 aux = "SPARED"; 1537 break; 1538 case VDEV_AUX_ERR_EXCEEDED: 1539 aux = "ERR_EXCEEDED"; 1540 break; 1541 case VDEV_AUX_IO_FAILURE: 1542 aux = "IO_FAILURE"; 1543 break; 1544 case VDEV_AUX_BAD_LOG: 1545 aux = "BAD_LOG"; 1546 break; 1547 case VDEV_AUX_EXTERNAL: 1548 aux = "EXTERNAL"; 1549 break; 1550 case VDEV_AUX_SPLIT_POOL: 1551 aux = "SPLIT_POOL"; 1552 break; 1553 default: 1554 aux = "UNKNOWN"; 1555 break; 1556 } 1557 1558 mdb_printf("%-9s %-12s %*s%s\n", state, aux, depth, "", desc); 1559 1560 if (spa_flags & SPA_FLAG_ERRORS) { 1561 vdev_stat_t *vs = &vdev.vdev_stat; 1562 int i; 1563 1564 mdb_inc_indent(4); 1565 mdb_printf("\n"); 1566 mdb_printf("%<u> %12s %12s %12s %12s " 1567 "%12s%</u>\n", "READ", "WRITE", "FREE", "CLAIM", 1568 "IOCTL"); 1569 mdb_printf("OPS "); 1570 for (i = 1; i < ZIO_TYPES; i++) 1571 mdb_printf("%11#llx%s", vs->vs_ops[i], 1572 i == ZIO_TYPES - 1 ? "" : " "); 1573 mdb_printf("\n"); 1574 mdb_printf("BYTES "); 1575 for (i = 1; i < ZIO_TYPES; i++) 1576 mdb_printf("%11#llx%s", vs->vs_bytes[i], 1577 i == ZIO_TYPES - 1 ? "" : " "); 1578 1579 1580 mdb_printf("\n"); 1581 mdb_printf("EREAD %10#llx\n", vs->vs_read_errors); 1582 mdb_printf("EWRITE %10#llx\n", vs->vs_write_errors); 1583 mdb_printf("ECKSUM %10#llx\n", 1584 vs->vs_checksum_errors); 1585 mdb_dec_indent(4); 1586 mdb_printf("\n"); 1587 } 1588 1589 if (spa_flags & SPA_FLAG_METASLAB_GROUPS && 1590 vdev.vdev_mg != NULL) { 1591 metaslab_group_stats((uintptr_t)vdev.vdev_mg, 1592 spa_flags); 1593 } 1594 if (spa_flags & SPA_FLAG_METASLABS && vdev.vdev_ms != NULL) { 1595 metaslab_stats((uintptr_t)addr, spa_flags); 1596 } 1597 } 1598 1599 children = vdev.vdev_children; 1600 1601 if (children == 0 || !recursive) 1602 return (DCMD_OK); 1603 1604 child = mdb_alloc(children * sizeof (void *), UM_SLEEP | UM_GC); 1605 if (mdb_vread(child, children * sizeof (void *), 1606 (uintptr_t)vdev.vdev_child) == -1) { 1607 mdb_warn("failed to read vdev children at %p", vdev.vdev_child); 1608 return (DCMD_ERR); 1609 } 1610 1611 for (c = 0; c < children; c++) { 1612 if (do_print_vdev(child[c], flags, depth + 2, recursive, 1613 spa_flags)) { 1614 return (DCMD_ERR); 1615 } 1616 } 1617 1618 return (DCMD_OK); 1619 } 1620 1621 static int 1622 vdev_print(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv) 1623 { 1624 uint64_t depth = 0; 1625 boolean_t recursive = B_FALSE; 1626 int spa_flags = 0; 1627 1628 if (mdb_getopts(argc, argv, 1629 'e', MDB_OPT_SETBITS, SPA_FLAG_ERRORS, &spa_flags, 1630 'm', MDB_OPT_SETBITS, SPA_FLAG_METASLABS, &spa_flags, 1631 'M', MDB_OPT_SETBITS, SPA_FLAG_METASLAB_GROUPS, &spa_flags, 1632 'h', MDB_OPT_SETBITS, SPA_FLAG_HISTOGRAMS, &spa_flags, 1633 'r', MDB_OPT_SETBITS, TRUE, &recursive, 1634 'd', MDB_OPT_UINT64, &depth, NULL) != argc) 1635 return (DCMD_USAGE); 1636 1637 if (!(flags & DCMD_ADDRSPEC)) { 1638 mdb_warn("no vdev_t address given\n"); 1639 return (DCMD_ERR); 1640 } 1641 1642 return (do_print_vdev(addr, flags, (int)depth, recursive, spa_flags)); 1643 } 1644 1645 typedef struct mdb_metaslab_alloc_trace { 1646 uintptr_t mat_mg; 1647 uintptr_t mat_msp; 1648 uint64_t mat_size; 1649 uint64_t mat_weight; 1650 uint64_t mat_offset; 1651 uint32_t mat_dva_id; 1652 } mdb_metaslab_alloc_trace_t; 1653 1654 static void 1655 metaslab_print_weight(uint64_t weight) 1656 { 1657 char buf[100]; 1658 1659 if (WEIGHT_IS_SPACEBASED(weight)) { 1660 mdb_nicenum( 1661 weight & ~(METASLAB_ACTIVE_MASK | METASLAB_WEIGHT_TYPE), 1662 buf); 1663 } else { 1664 char size[NICENUM_BUFLEN]; 1665 mdb_nicenum(1ULL << WEIGHT_GET_INDEX(weight), size); 1666 (void) mdb_snprintf(buf, sizeof (buf), "%llu x %s", 1667 WEIGHT_GET_COUNT(weight), size); 1668 } 1669 mdb_printf("%11s ", buf); 1670 } 1671 1672 /* ARGSUSED */ 1673 static int 1674 metaslab_weight(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv) 1675 { 1676 uint64_t weight = 0; 1677 char active; 1678 1679 if (argc == 0 && (flags & DCMD_ADDRSPEC)) { 1680 if (mdb_vread(&weight, sizeof (uint64_t), addr) == -1) { 1681 mdb_warn("failed to read weight at %p\n", addr); 1682 return (DCMD_ERR); 1683 } 1684 } else if (argc == 1 && !(flags & DCMD_ADDRSPEC)) { 1685 weight = (argv[0].a_type == MDB_TYPE_IMMEDIATE) ? 1686 argv[0].a_un.a_val : mdb_strtoull(argv[0].a_un.a_str); 1687 } else { 1688 return (DCMD_USAGE); 1689 } 1690 1691 if (DCMD_HDRSPEC(flags)) { 1692 mdb_printf("%<u>%-6s %9s %9s%</u>\n", 1693 "ACTIVE", "ALGORITHM", "WEIGHT"); 1694 } 1695 1696 if (weight & METASLAB_WEIGHT_PRIMARY) 1697 active = 'P'; 1698 else if (weight & METASLAB_WEIGHT_SECONDARY) 1699 active = 'S'; 1700 else 1701 active = '-'; 1702 mdb_printf("%6c %8s ", active, 1703 WEIGHT_IS_SPACEBASED(weight) ? "SPACE" : "SEGMENT"); 1704 metaslab_print_weight(weight); 1705 mdb_printf("\n"); 1706 1707 return (DCMD_OK); 1708 } 1709 1710 /* ARGSUSED */ 1711 static int 1712 metaslab_trace(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv) 1713 { 1714 mdb_metaslab_alloc_trace_t mat; 1715 mdb_metaslab_group_t mg = { 0 }; 1716 char result_type[100]; 1717 1718 if (mdb_ctf_vread(&mat, "metaslab_alloc_trace_t", 1719 "mdb_metaslab_alloc_trace_t", addr, 0) == -1) { 1720 return (DCMD_ERR); 1721 } 1722 1723 if (!(flags & DCMD_PIPE_OUT) && DCMD_HDRSPEC(flags)) { 1724 mdb_printf("%<u>%6s %6s %8s %11s %18s %18s%</u>\n", 1725 "MSID", "DVA", "ASIZE", "WEIGHT", "RESULT", "VDEV"); 1726 } 1727 1728 if (mat.mat_msp != NULL) { 1729 mdb_metaslab_t ms; 1730 1731 if (mdb_ctf_vread(&ms, "metaslab_t", "mdb_metaslab_t", 1732 mat.mat_msp, 0) == -1) { 1733 return (DCMD_ERR); 1734 } 1735 mdb_printf("%6llu ", ms.ms_id); 1736 } else { 1737 mdb_printf("%6s ", "-"); 1738 } 1739 1740 mdb_printf("%6d %8llx ", mat.mat_dva_id, mat.mat_size); 1741 1742 metaslab_print_weight(mat.mat_weight); 1743 1744 if ((int64_t)mat.mat_offset < 0) { 1745 if (enum_lookup("enum trace_alloc_type", mat.mat_offset, 1746 "TRACE_", sizeof (result_type), result_type) == -1) { 1747 mdb_warn("Could not find enum for trace_alloc_type"); 1748 return (DCMD_ERR); 1749 } 1750 mdb_printf("%18s ", result_type); 1751 } else { 1752 mdb_printf("%<b>%18llx%</b> ", mat.mat_offset); 1753 } 1754 1755 if (mat.mat_mg != NULL && 1756 mdb_ctf_vread(&mg, "metaslab_group_t", "mdb_metaslab_group_t", 1757 mat.mat_mg, 0) == -1) { 1758 return (DCMD_ERR); 1759 } 1760 1761 if (mg.mg_vd != NULL) { 1762 mdb_vdev_t vdev; 1763 char desc[MAXNAMELEN]; 1764 1765 if (mdb_ctf_vread(&vdev, "vdev_t", "mdb_vdev_t", 1766 mg.mg_vd, 0) == -1) { 1767 return (DCMD_ERR); 1768 } 1769 1770 if (vdev.vdev_path != NULL) { 1771 char path[MAXNAMELEN]; 1772 1773 if (mdb_readstr(path, sizeof (path), 1774 vdev.vdev_path) == -1) { 1775 mdb_warn("failed to read vdev_path at %p\n", 1776 vdev.vdev_path); 1777 return (DCMD_ERR); 1778 } 1779 char *slash; 1780 if ((slash = strrchr(path, '/')) != NULL) { 1781 strcpy(desc, slash + 1); 1782 } else { 1783 strcpy(desc, path); 1784 } 1785 } else if (vdev.vdev_ops != NULL) { 1786 mdb_vdev_ops_t ops; 1787 if (mdb_ctf_vread(&ops, "vdev_ops_t", "mdb_vdev_ops_t", 1788 vdev.vdev_ops, 0) == -1) { 1789 mdb_warn("failed to read vdev_ops at %p\n", 1790 vdev.vdev_ops); 1791 return (DCMD_ERR); 1792 } 1793 (void) mdb_snprintf(desc, sizeof (desc), 1794 "%s-%llu", ops.vdev_op_type, vdev.vdev_id); 1795 } else { 1796 (void) strcpy(desc, "<unknown>"); 1797 } 1798 mdb_printf("%18s\n", desc); 1799 } 1800 1801 return (DCMD_OK); 1802 } 1803 1804 typedef struct metaslab_walk_data { 1805 uint64_t mw_numvdevs; 1806 uintptr_t *mw_vdevs; 1807 int mw_curvdev; 1808 uint64_t mw_nummss; 1809 uintptr_t *mw_mss; 1810 int mw_curms; 1811 } metaslab_walk_data_t; 1812 1813 static int 1814 metaslab_walk_step(mdb_walk_state_t *wsp) 1815 { 1816 metaslab_walk_data_t *mw = wsp->walk_data; 1817 metaslab_t ms; 1818 uintptr_t msp; 1819 1820 if (mw->mw_curvdev >= mw->mw_numvdevs) 1821 return (WALK_DONE); 1822 1823 if (mw->mw_mss == NULL) { 1824 uintptr_t mssp; 1825 uintptr_t vdevp; 1826 1827 ASSERT(mw->mw_curms == 0); 1828 ASSERT(mw->mw_nummss == 0); 1829 1830 vdevp = mw->mw_vdevs[mw->mw_curvdev]; 1831 if (GETMEMB(vdevp, "vdev", vdev_ms, mssp) || 1832 GETMEMB(vdevp, "vdev", vdev_ms_count, mw->mw_nummss)) { 1833 return (WALK_ERR); 1834 } 1835 1836 mw->mw_mss = mdb_alloc(mw->mw_nummss * sizeof (void*), 1837 UM_SLEEP | UM_GC); 1838 if (mdb_vread(mw->mw_mss, mw->mw_nummss * sizeof (void*), 1839 mssp) == -1) { 1840 mdb_warn("failed to read vdev_ms at %p", mssp); 1841 return (WALK_ERR); 1842 } 1843 } 1844 1845 if (mw->mw_curms >= mw->mw_nummss) { 1846 mw->mw_mss = NULL; 1847 mw->mw_curms = 0; 1848 mw->mw_nummss = 0; 1849 mw->mw_curvdev++; 1850 return (WALK_NEXT); 1851 } 1852 1853 msp = mw->mw_mss[mw->mw_curms]; 1854 if (mdb_vread(&ms, sizeof (metaslab_t), msp) == -1) { 1855 mdb_warn("failed to read metaslab_t at %p", msp); 1856 return (WALK_ERR); 1857 } 1858 1859 mw->mw_curms++; 1860 1861 return (wsp->walk_callback(msp, &ms, wsp->walk_cbdata)); 1862 } 1863 1864 static int 1865 metaslab_walk_init(mdb_walk_state_t *wsp) 1866 { 1867 metaslab_walk_data_t *mw; 1868 uintptr_t root_vdevp; 1869 uintptr_t childp; 1870 1871 if (wsp->walk_addr == NULL) { 1872 mdb_warn("must supply address of spa_t\n"); 1873 return (WALK_ERR); 1874 } 1875 1876 mw = mdb_zalloc(sizeof (metaslab_walk_data_t), UM_SLEEP | UM_GC); 1877 1878 if (GETMEMB(wsp->walk_addr, "spa", spa_root_vdev, root_vdevp) || 1879 GETMEMB(root_vdevp, "vdev", vdev_children, mw->mw_numvdevs) || 1880 GETMEMB(root_vdevp, "vdev", vdev_child, childp)) { 1881 return (DCMD_ERR); 1882 } 1883 1884 mw->mw_vdevs = mdb_alloc(mw->mw_numvdevs * sizeof (void *), 1885 UM_SLEEP | UM_GC); 1886 if (mdb_vread(mw->mw_vdevs, mw->mw_numvdevs * sizeof (void *), 1887 childp) == -1) { 1888 mdb_warn("failed to read root vdev children at %p", childp); 1889 return (DCMD_ERR); 1890 } 1891 1892 wsp->walk_data = mw; 1893 1894 return (WALK_NEXT); 1895 } 1896 1897 typedef struct mdb_spa { 1898 uintptr_t spa_dsl_pool; 1899 uintptr_t spa_root_vdev; 1900 } mdb_spa_t; 1901 1902 typedef struct mdb_dsl_pool { 1903 uintptr_t dp_root_dir; 1904 } mdb_dsl_pool_t; 1905 1906 typedef struct mdb_dsl_dir { 1907 uintptr_t dd_dbuf; 1908 int64_t dd_space_towrite[TXG_SIZE]; 1909 } mdb_dsl_dir_t; 1910 1911 typedef struct mdb_dsl_dir_phys { 1912 uint64_t dd_used_bytes; 1913 uint64_t dd_compressed_bytes; 1914 uint64_t dd_uncompressed_bytes; 1915 } mdb_dsl_dir_phys_t; 1916 1917 typedef struct space_data { 1918 uint64_t ms_alloctree[TXG_SIZE]; 1919 uint64_t ms_freetree[TXG_SIZE]; 1920 uint64_t ms_tree; 1921 int64_t ms_deferspace; 1922 uint64_t avail; 1923 uint64_t nowavail; 1924 } space_data_t; 1925 1926 /* ARGSUSED */ 1927 static int 1928 space_cb(uintptr_t addr, const void *unknown, void *arg) 1929 { 1930 space_data_t *sd = arg; 1931 mdb_metaslab_t ms; 1932 mdb_range_tree_t rt; 1933 mdb_space_map_t sm = { 0 }; 1934 mdb_space_map_phys_t smp = { 0 }; 1935 int i; 1936 1937 if (mdb_ctf_vread(&ms, "metaslab_t", "mdb_metaslab_t", 1938 addr, 0) == -1) 1939 return (WALK_ERR); 1940 1941 for (i = 0; i < TXG_SIZE; i++) { 1942 if (mdb_ctf_vread(&rt, "range_tree_t", 1943 "mdb_range_tree_t", ms.ms_alloctree[i], 0) == -1) 1944 return (WALK_ERR); 1945 1946 sd->ms_alloctree[i] += rt.rt_space; 1947 1948 if (mdb_ctf_vread(&rt, "range_tree_t", 1949 "mdb_range_tree_t", ms.ms_freetree[i], 0) == -1) 1950 return (WALK_ERR); 1951 1952 sd->ms_freetree[i] += rt.rt_space; 1953 } 1954 1955 if (mdb_ctf_vread(&rt, "range_tree_t", 1956 "mdb_range_tree_t", ms.ms_tree, 0) == -1) 1957 return (WALK_ERR); 1958 1959 if (ms.ms_sm != NULL && 1960 mdb_ctf_vread(&sm, "space_map_t", 1961 "mdb_space_map_t", ms.ms_sm, 0) == -1) 1962 return (WALK_ERR); 1963 1964 if (sm.sm_phys != NULL) { 1965 (void) mdb_ctf_vread(&smp, "space_map_phys_t", 1966 "mdb_space_map_phys_t", sm.sm_phys, 0); 1967 } 1968 1969 sd->ms_deferspace += ms.ms_deferspace; 1970 sd->ms_tree += rt.rt_space; 1971 sd->avail += sm.sm_size - sm.sm_alloc; 1972 sd->nowavail += sm.sm_size - smp.smp_alloc; 1973 1974 return (WALK_NEXT); 1975 } 1976 1977 /* 1978 * ::spa_space [-b] 1979 * 1980 * Given a spa_t, print out it's on-disk space usage and in-core 1981 * estimates of future usage. If -b is given, print space in bytes. 1982 * Otherwise print in megabytes. 1983 */ 1984 /* ARGSUSED */ 1985 static int 1986 spa_space(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv) 1987 { 1988 mdb_spa_t spa; 1989 mdb_dsl_pool_t dp; 1990 mdb_dsl_dir_t dd; 1991 mdb_dmu_buf_impl_t db; 1992 mdb_dsl_dir_phys_t dsp; 1993 space_data_t sd; 1994 int shift = 20; 1995 char *suffix = "M"; 1996 int bytes = B_FALSE; 1997 1998 if (mdb_getopts(argc, argv, 'b', MDB_OPT_SETBITS, TRUE, &bytes, NULL) != 1999 argc) 2000 return (DCMD_USAGE); 2001 if (!(flags & DCMD_ADDRSPEC)) 2002 return (DCMD_USAGE); 2003 2004 if (bytes) { 2005 shift = 0; 2006 suffix = ""; 2007 } 2008 2009 if (mdb_ctf_vread(&spa, ZFS_STRUCT "spa", "mdb_spa_t", 2010 addr, 0) == -1 || 2011 mdb_ctf_vread(&dp, ZFS_STRUCT "dsl_pool", "mdb_dsl_pool_t", 2012 spa.spa_dsl_pool, 0) == -1 || 2013 mdb_ctf_vread(&dd, ZFS_STRUCT "dsl_dir", "mdb_dsl_dir_t", 2014 dp.dp_root_dir, 0) == -1 || 2015 mdb_ctf_vread(&db, ZFS_STRUCT "dmu_buf_impl", "mdb_dmu_buf_impl_t", 2016 dd.dd_dbuf, 0) == -1 || 2017 mdb_ctf_vread(&dsp, ZFS_STRUCT "dsl_dir_phys", 2018 "mdb_dsl_dir_phys_t", db.db.db_data, 0) == -1) { 2019 return (DCMD_ERR); 2020 } 2021 2022 mdb_printf("dd_space_towrite = %llu%s %llu%s %llu%s %llu%s\n", 2023 dd.dd_space_towrite[0] >> shift, suffix, 2024 dd.dd_space_towrite[1] >> shift, suffix, 2025 dd.dd_space_towrite[2] >> shift, suffix, 2026 dd.dd_space_towrite[3] >> shift, suffix); 2027 2028 mdb_printf("dd_phys.dd_used_bytes = %llu%s\n", 2029 dsp.dd_used_bytes >> shift, suffix); 2030 mdb_printf("dd_phys.dd_compressed_bytes = %llu%s\n", 2031 dsp.dd_compressed_bytes >> shift, suffix); 2032 mdb_printf("dd_phys.dd_uncompressed_bytes = %llu%s\n", 2033 dsp.dd_uncompressed_bytes >> shift, suffix); 2034 2035 bzero(&sd, sizeof (sd)); 2036 if (mdb_pwalk("metaslab", space_cb, &sd, addr) != 0) { 2037 mdb_warn("can't walk metaslabs"); 2038 return (DCMD_ERR); 2039 } 2040 2041 mdb_printf("ms_allocmap = %llu%s %llu%s %llu%s %llu%s\n", 2042 sd.ms_alloctree[0] >> shift, suffix, 2043 sd.ms_alloctree[1] >> shift, suffix, 2044 sd.ms_alloctree[2] >> shift, suffix, 2045 sd.ms_alloctree[3] >> shift, suffix); 2046 mdb_printf("ms_freemap = %llu%s %llu%s %llu%s %llu%s\n", 2047 sd.ms_freetree[0] >> shift, suffix, 2048 sd.ms_freetree[1] >> shift, suffix, 2049 sd.ms_freetree[2] >> shift, suffix, 2050 sd.ms_freetree[3] >> shift, suffix); 2051 mdb_printf("ms_tree = %llu%s\n", sd.ms_tree >> shift, suffix); 2052 mdb_printf("ms_deferspace = %llu%s\n", 2053 sd.ms_deferspace >> shift, suffix); 2054 mdb_printf("last synced avail = %llu%s\n", sd.avail >> shift, suffix); 2055 mdb_printf("current syncing avail = %llu%s\n", 2056 sd.nowavail >> shift, suffix); 2057 2058 return (DCMD_OK); 2059 } 2060 2061 typedef struct mdb_spa_aux_vdev { 2062 int sav_count; 2063 uintptr_t sav_vdevs; 2064 } mdb_spa_aux_vdev_t; 2065 2066 typedef struct mdb_spa_vdevs { 2067 uintptr_t spa_root_vdev; 2068 mdb_spa_aux_vdev_t spa_l2cache; 2069 mdb_spa_aux_vdev_t spa_spares; 2070 } mdb_spa_vdevs_t; 2071 2072 static int 2073 spa_print_aux(mdb_spa_aux_vdev_t *sav, uint_t flags, mdb_arg_t *v, 2074 const char *name) 2075 { 2076 uintptr_t *aux; 2077 size_t len; 2078 int ret, i; 2079 2080 /* 2081 * Iterate over aux vdevs and print those out as well. This is a 2082 * little annoying because we don't have a root vdev to pass to ::vdev. 2083 * Instead, we print a single line and then call it for each child 2084 * vdev. 2085 */ 2086 if (sav->sav_count != 0) { 2087 v[1].a_type = MDB_TYPE_STRING; 2088 v[1].a_un.a_str = "-d"; 2089 v[2].a_type = MDB_TYPE_IMMEDIATE; 2090 v[2].a_un.a_val = 2; 2091 2092 len = sav->sav_count * sizeof (uintptr_t); 2093 aux = mdb_alloc(len, UM_SLEEP); 2094 if (mdb_vread(aux, len, sav->sav_vdevs) == -1) { 2095 mdb_free(aux, len); 2096 mdb_warn("failed to read l2cache vdevs at %p", 2097 sav->sav_vdevs); 2098 return (DCMD_ERR); 2099 } 2100 2101 mdb_printf("%-?s %-9s %-12s %s\n", "-", "-", "-", name); 2102 2103 for (i = 0; i < sav->sav_count; i++) { 2104 ret = mdb_call_dcmd("vdev", aux[i], flags, 3, v); 2105 if (ret != DCMD_OK) { 2106 mdb_free(aux, len); 2107 return (ret); 2108 } 2109 } 2110 2111 mdb_free(aux, len); 2112 } 2113 2114 return (0); 2115 } 2116 2117 /* 2118 * ::spa_vdevs 2119 * 2120 * -e Include error stats 2121 * -m Include metaslab information 2122 * -M Include metaslab group information 2123 * -h Include histogram information (requires -m or -M) 2124 * 2125 * Print out a summarized list of vdevs for the given spa_t. 2126 * This is accomplished by invoking "::vdev -re" on the root vdev, as well as 2127 * iterating over the cache devices. 2128 */ 2129 /* ARGSUSED */ 2130 static int 2131 spa_vdevs(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv) 2132 { 2133 mdb_arg_t v[3]; 2134 int ret; 2135 char opts[100] = "-r"; 2136 int spa_flags = 0; 2137 2138 if (mdb_getopts(argc, argv, 2139 'e', MDB_OPT_SETBITS, SPA_FLAG_ERRORS, &spa_flags, 2140 'm', MDB_OPT_SETBITS, SPA_FLAG_METASLABS, &spa_flags, 2141 'M', MDB_OPT_SETBITS, SPA_FLAG_METASLAB_GROUPS, &spa_flags, 2142 'h', MDB_OPT_SETBITS, SPA_FLAG_HISTOGRAMS, &spa_flags, 2143 NULL) != argc) 2144 return (DCMD_USAGE); 2145 2146 if (!(flags & DCMD_ADDRSPEC)) 2147 return (DCMD_USAGE); 2148 2149 mdb_spa_vdevs_t spa; 2150 if (mdb_ctf_vread(&spa, "spa_t", "mdb_spa_vdevs_t", addr, 0) == -1) 2151 return (DCMD_ERR); 2152 2153 /* 2154 * Unitialized spa_t structures can have a NULL root vdev. 2155 */ 2156 if (spa.spa_root_vdev == NULL) { 2157 mdb_printf("no associated vdevs\n"); 2158 return (DCMD_OK); 2159 } 2160 2161 if (spa_flags & SPA_FLAG_ERRORS) 2162 strcat(opts, "e"); 2163 if (spa_flags & SPA_FLAG_METASLABS) 2164 strcat(opts, "m"); 2165 if (spa_flags & SPA_FLAG_METASLAB_GROUPS) 2166 strcat(opts, "M"); 2167 if (spa_flags & SPA_FLAG_HISTOGRAMS) 2168 strcat(opts, "h"); 2169 2170 v[0].a_type = MDB_TYPE_STRING; 2171 v[0].a_un.a_str = opts; 2172 2173 ret = mdb_call_dcmd("vdev", (uintptr_t)spa.spa_root_vdev, 2174 flags, 1, v); 2175 if (ret != DCMD_OK) 2176 return (ret); 2177 2178 if (spa_print_aux(&spa.spa_l2cache, flags, v, "cache") != 0 || 2179 spa_print_aux(&spa.spa_spares, flags, v, "spares") != 0) 2180 return (DCMD_ERR); 2181 2182 return (DCMD_OK); 2183 } 2184 2185 /* 2186 * ::zio 2187 * 2188 * Print a summary of zio_t and all its children. This is intended to display a 2189 * zio tree, and hence we only pick the most important pieces of information for 2190 * the main summary. More detailed information can always be found by doing a 2191 * '::print zio' on the underlying zio_t. The columns we display are: 2192 * 2193 * ADDRESS TYPE STAGE WAITER TIME_ELAPSED 2194 * 2195 * The 'address' column is indented by one space for each depth level as we 2196 * descend down the tree. 2197 */ 2198 2199 #define ZIO_MAXINDENT 7 2200 #define ZIO_MAXWIDTH (sizeof (uintptr_t) * 2 + ZIO_MAXINDENT) 2201 #define ZIO_WALK_SELF 0 2202 #define ZIO_WALK_CHILD 1 2203 #define ZIO_WALK_PARENT 2 2204 2205 typedef struct zio_print_args { 2206 int zpa_current_depth; 2207 int zpa_min_depth; 2208 int zpa_max_depth; 2209 int zpa_type; 2210 uint_t zpa_flags; 2211 } zio_print_args_t; 2212 2213 typedef struct mdb_zio { 2214 enum zio_type io_type; 2215 enum zio_stage io_stage; 2216 uintptr_t io_waiter; 2217 uintptr_t io_spa; 2218 struct { 2219 struct { 2220 uintptr_t list_next; 2221 } list_head; 2222 } io_parent_list; 2223 int io_error; 2224 } mdb_zio_t; 2225 2226 typedef struct mdb_zio_timestamp { 2227 hrtime_t io_timestamp; 2228 } mdb_zio_timestamp_t; 2229 2230 static int zio_child_cb(uintptr_t addr, const void *unknown, void *arg); 2231 2232 static int 2233 zio_print_cb(uintptr_t addr, zio_print_args_t *zpa) 2234 { 2235 mdb_ctf_id_t type_enum, stage_enum; 2236 int indent = zpa->zpa_current_depth; 2237 const char *type, *stage; 2238 uintptr_t laddr; 2239 mdb_zio_t zio; 2240 mdb_zio_timestamp_t zio_timestamp = { 0 }; 2241 2242 if (mdb_ctf_vread(&zio, ZFS_STRUCT "zio", "mdb_zio_t", addr, 0) == -1) 2243 return (WALK_ERR); 2244 (void) mdb_ctf_vread(&zio_timestamp, ZFS_STRUCT "zio", 2245 "mdb_zio_timestamp_t", addr, MDB_CTF_VREAD_QUIET); 2246 2247 if (indent > ZIO_MAXINDENT) 2248 indent = ZIO_MAXINDENT; 2249 2250 if (mdb_ctf_lookup_by_name("enum zio_type", &type_enum) == -1 || 2251 mdb_ctf_lookup_by_name("enum zio_stage", &stage_enum) == -1) { 2252 mdb_warn("failed to lookup zio enums"); 2253 return (WALK_ERR); 2254 } 2255 2256 if ((type = mdb_ctf_enum_name(type_enum, zio.io_type)) != NULL) 2257 type += sizeof ("ZIO_TYPE_") - 1; 2258 else 2259 type = "?"; 2260 2261 if (zio.io_error == 0) { 2262 stage = mdb_ctf_enum_name(stage_enum, zio.io_stage); 2263 if (stage != NULL) 2264 stage += sizeof ("ZIO_STAGE_") - 1; 2265 else 2266 stage = "?"; 2267 } else { 2268 stage = "FAILED"; 2269 } 2270 2271 if (zpa->zpa_current_depth >= zpa->zpa_min_depth) { 2272 if (zpa->zpa_flags & DCMD_PIPE_OUT) { 2273 mdb_printf("%?p\n", addr); 2274 } else { 2275 mdb_printf("%*s%-*p %-5s %-16s ", indent, "", 2276 ZIO_MAXWIDTH - indent, addr, type, stage); 2277 if (zio.io_waiter != 0) 2278 mdb_printf("%-16lx ", zio.io_waiter); 2279 else 2280 mdb_printf("%-16s ", "-"); 2281 #ifdef _KERNEL 2282 if (zio_timestamp.io_timestamp != 0) { 2283 mdb_printf("%llums", (mdb_gethrtime() - 2284 zio_timestamp.io_timestamp) / 2285 1000000); 2286 } else { 2287 mdb_printf("%-12s ", "-"); 2288 } 2289 #else 2290 mdb_printf("%-12s ", "-"); 2291 #endif 2292 mdb_printf("\n"); 2293 } 2294 } 2295 2296 if (zpa->zpa_current_depth >= zpa->zpa_max_depth) 2297 return (WALK_NEXT); 2298 2299 if (zpa->zpa_type == ZIO_WALK_PARENT) 2300 laddr = addr + mdb_ctf_offsetof_by_name(ZFS_STRUCT "zio", 2301 "io_parent_list"); 2302 else 2303 laddr = addr + mdb_ctf_offsetof_by_name(ZFS_STRUCT "zio", 2304 "io_child_list"); 2305 2306 zpa->zpa_current_depth++; 2307 if (mdb_pwalk("list", zio_child_cb, zpa, laddr) != 0) { 2308 mdb_warn("failed to walk zio_t children at %p\n", laddr); 2309 return (WALK_ERR); 2310 } 2311 zpa->zpa_current_depth--; 2312 2313 return (WALK_NEXT); 2314 } 2315 2316 /* ARGSUSED */ 2317 static int 2318 zio_child_cb(uintptr_t addr, const void *unknown, void *arg) 2319 { 2320 zio_link_t zl; 2321 uintptr_t ziop; 2322 zio_print_args_t *zpa = arg; 2323 2324 if (mdb_vread(&zl, sizeof (zl), addr) == -1) { 2325 mdb_warn("failed to read zio_link_t at %p", addr); 2326 return (WALK_ERR); 2327 } 2328 2329 if (zpa->zpa_type == ZIO_WALK_PARENT) 2330 ziop = (uintptr_t)zl.zl_parent; 2331 else 2332 ziop = (uintptr_t)zl.zl_child; 2333 2334 return (zio_print_cb(ziop, zpa)); 2335 } 2336 2337 /* ARGSUSED */ 2338 static int 2339 zio_print(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv) 2340 { 2341 zio_print_args_t zpa = { 0 }; 2342 2343 if (!(flags & DCMD_ADDRSPEC)) 2344 return (DCMD_USAGE); 2345 2346 if (mdb_getopts(argc, argv, 2347 'r', MDB_OPT_SETBITS, INT_MAX, &zpa.zpa_max_depth, 2348 'c', MDB_OPT_SETBITS, ZIO_WALK_CHILD, &zpa.zpa_type, 2349 'p', MDB_OPT_SETBITS, ZIO_WALK_PARENT, &zpa.zpa_type, 2350 NULL) != argc) 2351 return (DCMD_USAGE); 2352 2353 zpa.zpa_flags = flags; 2354 if (zpa.zpa_max_depth != 0) { 2355 if (zpa.zpa_type == ZIO_WALK_SELF) 2356 zpa.zpa_type = ZIO_WALK_CHILD; 2357 } else if (zpa.zpa_type != ZIO_WALK_SELF) { 2358 zpa.zpa_min_depth = 1; 2359 zpa.zpa_max_depth = 1; 2360 } 2361 2362 if (!(flags & DCMD_PIPE_OUT) && DCMD_HDRSPEC(flags)) { 2363 mdb_printf("%<u>%-*s %-5s %-16s %-16s %-12s%</u>\n", 2364 ZIO_MAXWIDTH, "ADDRESS", "TYPE", "STAGE", "WAITER", 2365 "TIME_ELAPSED"); 2366 } 2367 2368 if (zio_print_cb(addr, &zpa) != WALK_NEXT) 2369 return (DCMD_ERR); 2370 2371 return (DCMD_OK); 2372 } 2373 2374 /* 2375 * [addr]::zio_state 2376 * 2377 * Print a summary of all zio_t structures on the system, or for a particular 2378 * pool. This is equivalent to '::walk zio_root | ::zio'. 2379 */ 2380 /*ARGSUSED*/ 2381 static int 2382 zio_state(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv) 2383 { 2384 /* 2385 * MDB will remember the last address of the pipeline, so if we don't 2386 * zero this we'll end up trying to walk zio structures for a 2387 * non-existent spa_t. 2388 */ 2389 if (!(flags & DCMD_ADDRSPEC)) 2390 addr = 0; 2391 2392 return (mdb_pwalk_dcmd("zio_root", "zio", argc, argv, addr)); 2393 } 2394 2395 typedef struct mdb_multilist { 2396 uint64_t ml_num_sublists; 2397 uintptr_t ml_sublists; 2398 } mdb_multilist_t; 2399 2400 typedef struct multilist_walk_data { 2401 uint64_t mwd_idx; 2402 mdb_multilist_t mwd_ml; 2403 } multilist_walk_data_t; 2404 2405 /* ARGSUSED */ 2406 static int 2407 multilist_print_cb(uintptr_t addr, const void *unknown, void *arg) 2408 { 2409 mdb_printf("%#lr\n", addr); 2410 return (WALK_NEXT); 2411 } 2412 2413 static int 2414 multilist_walk_step(mdb_walk_state_t *wsp) 2415 { 2416 multilist_walk_data_t *mwd = wsp->walk_data; 2417 2418 if (mwd->mwd_idx >= mwd->mwd_ml.ml_num_sublists) 2419 return (WALK_DONE); 2420 2421 wsp->walk_addr = mwd->mwd_ml.ml_sublists + 2422 mdb_ctf_sizeof_by_name("multilist_sublist_t") * mwd->mwd_idx + 2423 mdb_ctf_offsetof_by_name("multilist_sublist_t", "mls_list"); 2424 2425 mdb_pwalk("list", multilist_print_cb, (void*)NULL, wsp->walk_addr); 2426 mwd->mwd_idx++; 2427 2428 return (WALK_NEXT); 2429 } 2430 2431 static int 2432 multilist_walk_init(mdb_walk_state_t *wsp) 2433 { 2434 multilist_walk_data_t *mwd; 2435 2436 if (wsp->walk_addr == NULL) { 2437 mdb_warn("must supply address of multilist_t\n"); 2438 return (WALK_ERR); 2439 } 2440 2441 mwd = mdb_zalloc(sizeof (multilist_walk_data_t), UM_SLEEP | UM_GC); 2442 if (mdb_ctf_vread(&mwd->mwd_ml, "multilist_t", "mdb_multilist_t", 2443 wsp->walk_addr, 0) == -1) { 2444 return (WALK_ERR); 2445 } 2446 2447 if (mwd->mwd_ml.ml_num_sublists == 0 || 2448 mwd->mwd_ml.ml_sublists == NULL) { 2449 mdb_warn("invalid or uninitialized multilist at %#lx\n", 2450 wsp->walk_addr); 2451 return (WALK_ERR); 2452 } 2453 2454 wsp->walk_data = mwd; 2455 return (WALK_NEXT); 2456 } 2457 2458 typedef struct txg_list_walk_data { 2459 uintptr_t lw_head[TXG_SIZE]; 2460 int lw_txgoff; 2461 int lw_maxoff; 2462 size_t lw_offset; 2463 void *lw_obj; 2464 } txg_list_walk_data_t; 2465 2466 static int 2467 txg_list_walk_init_common(mdb_walk_state_t *wsp, int txg, int maxoff) 2468 { 2469 txg_list_walk_data_t *lwd; 2470 txg_list_t list; 2471 int i; 2472 2473 lwd = mdb_alloc(sizeof (txg_list_walk_data_t), UM_SLEEP | UM_GC); 2474 if (mdb_vread(&list, sizeof (txg_list_t), wsp->walk_addr) == -1) { 2475 mdb_warn("failed to read txg_list_t at %#lx", wsp->walk_addr); 2476 return (WALK_ERR); 2477 } 2478 2479 for (i = 0; i < TXG_SIZE; i++) 2480 lwd->lw_head[i] = (uintptr_t)list.tl_head[i]; 2481 lwd->lw_offset = list.tl_offset; 2482 lwd->lw_obj = mdb_alloc(lwd->lw_offset + sizeof (txg_node_t), 2483 UM_SLEEP | UM_GC); 2484 lwd->lw_txgoff = txg; 2485 lwd->lw_maxoff = maxoff; 2486 2487 wsp->walk_addr = lwd->lw_head[lwd->lw_txgoff]; 2488 wsp->walk_data = lwd; 2489 2490 return (WALK_NEXT); 2491 } 2492 2493 static int 2494 txg_list_walk_init(mdb_walk_state_t *wsp) 2495 { 2496 return (txg_list_walk_init_common(wsp, 0, TXG_SIZE-1)); 2497 } 2498 2499 static int 2500 txg_list0_walk_init(mdb_walk_state_t *wsp) 2501 { 2502 return (txg_list_walk_init_common(wsp, 0, 0)); 2503 } 2504 2505 static int 2506 txg_list1_walk_init(mdb_walk_state_t *wsp) 2507 { 2508 return (txg_list_walk_init_common(wsp, 1, 1)); 2509 } 2510 2511 static int 2512 txg_list2_walk_init(mdb_walk_state_t *wsp) 2513 { 2514 return (txg_list_walk_init_common(wsp, 2, 2)); 2515 } 2516 2517 static int 2518 txg_list3_walk_init(mdb_walk_state_t *wsp) 2519 { 2520 return (txg_list_walk_init_common(wsp, 3, 3)); 2521 } 2522 2523 static int 2524 txg_list_walk_step(mdb_walk_state_t *wsp) 2525 { 2526 txg_list_walk_data_t *lwd = wsp->walk_data; 2527 uintptr_t addr; 2528 txg_node_t *node; 2529 int status; 2530 2531 while (wsp->walk_addr == NULL && lwd->lw_txgoff < lwd->lw_maxoff) { 2532 lwd->lw_txgoff++; 2533 wsp->walk_addr = lwd->lw_head[lwd->lw_txgoff]; 2534 } 2535 2536 if (wsp->walk_addr == NULL) 2537 return (WALK_DONE); 2538 2539 addr = wsp->walk_addr - lwd->lw_offset; 2540 2541 if (mdb_vread(lwd->lw_obj, 2542 lwd->lw_offset + sizeof (txg_node_t), addr) == -1) { 2543 mdb_warn("failed to read list element at %#lx", addr); 2544 return (WALK_ERR); 2545 } 2546 2547 status = wsp->walk_callback(addr, lwd->lw_obj, wsp->walk_cbdata); 2548 node = (txg_node_t *)((uintptr_t)lwd->lw_obj + lwd->lw_offset); 2549 wsp->walk_addr = (uintptr_t)node->tn_next[lwd->lw_txgoff]; 2550 2551 return (status); 2552 } 2553 2554 /* 2555 * ::walk spa 2556 * 2557 * Walk all named spa_t structures in the namespace. This is nothing more than 2558 * a layered avl walk. 2559 */ 2560 static int 2561 spa_walk_init(mdb_walk_state_t *wsp) 2562 { 2563 GElf_Sym sym; 2564 2565 if (wsp->walk_addr != NULL) { 2566 mdb_warn("spa walk only supports global walks\n"); 2567 return (WALK_ERR); 2568 } 2569 2570 if (mdb_lookup_by_obj(ZFS_OBJ_NAME, "spa_namespace_avl", &sym) == -1) { 2571 mdb_warn("failed to find symbol 'spa_namespace_avl'"); 2572 return (WALK_ERR); 2573 } 2574 2575 wsp->walk_addr = (uintptr_t)sym.st_value; 2576 2577 if (mdb_layered_walk("avl", wsp) == -1) { 2578 mdb_warn("failed to walk 'avl'\n"); 2579 return (WALK_ERR); 2580 } 2581 2582 return (WALK_NEXT); 2583 } 2584 2585 static int 2586 spa_walk_step(mdb_walk_state_t *wsp) 2587 { 2588 return (wsp->walk_callback(wsp->walk_addr, NULL, wsp->walk_cbdata)); 2589 } 2590 2591 /* 2592 * [addr]::walk zio 2593 * 2594 * Walk all active zio_t structures on the system. This is simply a layered 2595 * walk on top of ::walk zio_cache, with the optional ability to limit the 2596 * structures to a particular pool. 2597 */ 2598 static int 2599 zio_walk_init(mdb_walk_state_t *wsp) 2600 { 2601 wsp->walk_data = (void *)wsp->walk_addr; 2602 2603 if (mdb_layered_walk("zio_cache", wsp) == -1) { 2604 mdb_warn("failed to walk 'zio_cache'\n"); 2605 return (WALK_ERR); 2606 } 2607 2608 return (WALK_NEXT); 2609 } 2610 2611 static int 2612 zio_walk_step(mdb_walk_state_t *wsp) 2613 { 2614 mdb_zio_t zio; 2615 uintptr_t spa = (uintptr_t)wsp->walk_data; 2616 2617 if (mdb_ctf_vread(&zio, ZFS_STRUCT "zio", "mdb_zio_t", 2618 wsp->walk_addr, 0) == -1) 2619 return (WALK_ERR); 2620 2621 if (spa != 0 && spa != zio.io_spa) 2622 return (WALK_NEXT); 2623 2624 return (wsp->walk_callback(wsp->walk_addr, &zio, wsp->walk_cbdata)); 2625 } 2626 2627 /* 2628 * [addr]::walk zio_root 2629 * 2630 * Walk only root zio_t structures, optionally for a particular spa_t. 2631 */ 2632 static int 2633 zio_walk_root_step(mdb_walk_state_t *wsp) 2634 { 2635 mdb_zio_t zio; 2636 uintptr_t spa = (uintptr_t)wsp->walk_data; 2637 2638 if (mdb_ctf_vread(&zio, ZFS_STRUCT "zio", "mdb_zio_t", 2639 wsp->walk_addr, 0) == -1) 2640 return (WALK_ERR); 2641 2642 if (spa != 0 && spa != zio.io_spa) 2643 return (WALK_NEXT); 2644 2645 /* If the parent list is not empty, ignore */ 2646 if (zio.io_parent_list.list_head.list_next != 2647 wsp->walk_addr + 2648 mdb_ctf_offsetof_by_name(ZFS_STRUCT "zio", "io_parent_list") + 2649 mdb_ctf_offsetof_by_name("struct list", "list_head")) 2650 return (WALK_NEXT); 2651 2652 return (wsp->walk_callback(wsp->walk_addr, &zio, wsp->walk_cbdata)); 2653 } 2654 2655 /* 2656 * ::zfs_blkstats 2657 * 2658 * -v print verbose per-level information 2659 * 2660 */ 2661 static int 2662 zfs_blkstats(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv) 2663 { 2664 boolean_t verbose = B_FALSE; 2665 zfs_all_blkstats_t stats; 2666 dmu_object_type_t t; 2667 zfs_blkstat_t *tzb; 2668 uint64_t ditto; 2669 dmu_object_type_info_t dmu_ot[DMU_OT_NUMTYPES + 10]; 2670 /* +10 in case it grew */ 2671 2672 if (mdb_readvar(&dmu_ot, "dmu_ot") == -1) { 2673 mdb_warn("failed to read 'dmu_ot'"); 2674 return (DCMD_ERR); 2675 } 2676 2677 if (mdb_getopts(argc, argv, 2678 'v', MDB_OPT_SETBITS, TRUE, &verbose, 2679 NULL) != argc) 2680 return (DCMD_USAGE); 2681 2682 if (!(flags & DCMD_ADDRSPEC)) 2683 return (DCMD_USAGE); 2684 2685 if (GETMEMB(addr, "spa", spa_dsl_pool, addr) || 2686 GETMEMB(addr, "dsl_pool", dp_blkstats, addr) || 2687 mdb_vread(&stats, sizeof (zfs_all_blkstats_t), addr) == -1) { 2688 mdb_warn("failed to read data at %p;", addr); 2689 mdb_printf("maybe no stats? run \"zpool scrub\" first."); 2690 return (DCMD_ERR); 2691 } 2692 2693 tzb = &stats.zab_type[DN_MAX_LEVELS][DMU_OT_TOTAL]; 2694 if (tzb->zb_gangs != 0) { 2695 mdb_printf("Ganged blocks: %llu\n", 2696 (longlong_t)tzb->zb_gangs); 2697 } 2698 2699 ditto = tzb->zb_ditto_2_of_2_samevdev + tzb->zb_ditto_2_of_3_samevdev + 2700 tzb->zb_ditto_3_of_3_samevdev; 2701 if (ditto != 0) { 2702 mdb_printf("Dittoed blocks on same vdev: %llu\n", 2703 (longlong_t)ditto); 2704 } 2705 2706 mdb_printf("\nBlocks\tLSIZE\tPSIZE\tASIZE" 2707 "\t avg\t comp\t%%Total\tType\n"); 2708 2709 for (t = 0; t <= DMU_OT_TOTAL; t++) { 2710 char csize[NICENUM_BUFLEN], lsize[NICENUM_BUFLEN]; 2711 char psize[NICENUM_BUFLEN], asize[NICENUM_BUFLEN]; 2712 char avg[NICENUM_BUFLEN]; 2713 char comp[NICENUM_BUFLEN], pct[NICENUM_BUFLEN]; 2714 char typename[64]; 2715 int l; 2716 2717 2718 if (t == DMU_OT_DEFERRED) 2719 strcpy(typename, "deferred free"); 2720 else if (t == DMU_OT_OTHER) 2721 strcpy(typename, "other"); 2722 else if (t == DMU_OT_TOTAL) 2723 strcpy(typename, "Total"); 2724 else if (mdb_readstr(typename, sizeof (typename), 2725 (uintptr_t)dmu_ot[t].ot_name) == -1) { 2726 mdb_warn("failed to read type name"); 2727 return (DCMD_ERR); 2728 } 2729 2730 if (stats.zab_type[DN_MAX_LEVELS][t].zb_asize == 0) 2731 continue; 2732 2733 for (l = -1; l < DN_MAX_LEVELS; l++) { 2734 int level = (l == -1 ? DN_MAX_LEVELS : l); 2735 zfs_blkstat_t *zb = &stats.zab_type[level][t]; 2736 2737 if (zb->zb_asize == 0) 2738 continue; 2739 2740 /* 2741 * Don't print each level unless requested. 2742 */ 2743 if (!verbose && level != DN_MAX_LEVELS) 2744 continue; 2745 2746 /* 2747 * If all the space is level 0, don't print the 2748 * level 0 separately. 2749 */ 2750 if (level == 0 && zb->zb_asize == 2751 stats.zab_type[DN_MAX_LEVELS][t].zb_asize) 2752 continue; 2753 2754 mdb_nicenum(zb->zb_count, csize); 2755 mdb_nicenum(zb->zb_lsize, lsize); 2756 mdb_nicenum(zb->zb_psize, psize); 2757 mdb_nicenum(zb->zb_asize, asize); 2758 mdb_nicenum(zb->zb_asize / zb->zb_count, avg); 2759 (void) snprintfrac(comp, NICENUM_BUFLEN, 2760 zb->zb_lsize, zb->zb_psize, 2); 2761 (void) snprintfrac(pct, NICENUM_BUFLEN, 2762 100 * zb->zb_asize, tzb->zb_asize, 2); 2763 2764 mdb_printf("%6s\t%5s\t%5s\t%5s\t%5s" 2765 "\t%5s\t%6s\t", 2766 csize, lsize, psize, asize, avg, comp, pct); 2767 2768 if (level == DN_MAX_LEVELS) 2769 mdb_printf("%s\n", typename); 2770 else 2771 mdb_printf(" L%d %s\n", 2772 level, typename); 2773 } 2774 } 2775 2776 return (DCMD_OK); 2777 } 2778 2779 typedef struct mdb_reference { 2780 uintptr_t ref_holder; 2781 uintptr_t ref_removed; 2782 uint64_t ref_number; 2783 } mdb_reference_t; 2784 2785 /* ARGSUSED */ 2786 static int 2787 reference_cb(uintptr_t addr, const void *ignored, void *arg) 2788 { 2789 mdb_reference_t ref; 2790 boolean_t holder_is_str = B_FALSE; 2791 char holder_str[128]; 2792 boolean_t removed = (boolean_t)arg; 2793 2794 if (mdb_ctf_vread(&ref, "reference_t", "mdb_reference_t", addr, 2795 0) == -1) 2796 return (DCMD_ERR); 2797 2798 if (mdb_readstr(holder_str, sizeof (holder_str), 2799 ref.ref_holder) != -1) 2800 holder_is_str = strisprint(holder_str); 2801 2802 if (removed) 2803 mdb_printf("removed "); 2804 mdb_printf("reference "); 2805 if (ref.ref_number != 1) 2806 mdb_printf("with count=%llu ", ref.ref_number); 2807 mdb_printf("with tag %lx", ref.ref_holder); 2808 if (holder_is_str) 2809 mdb_printf(" \"%s\"", holder_str); 2810 mdb_printf(", held at:\n"); 2811 2812 (void) mdb_call_dcmd("whatis", addr, DCMD_ADDRSPEC, 0, NULL); 2813 2814 if (removed) { 2815 mdb_printf("removed at:\n"); 2816 (void) mdb_call_dcmd("whatis", ref.ref_removed, 2817 DCMD_ADDRSPEC, 0, NULL); 2818 } 2819 2820 mdb_printf("\n"); 2821 2822 return (WALK_NEXT); 2823 } 2824 2825 typedef struct mdb_refcount { 2826 uint64_t rc_count; 2827 } mdb_refcount_t; 2828 2829 typedef struct mdb_refcount_removed { 2830 uint64_t rc_removed_count; 2831 } mdb_refcount_removed_t; 2832 2833 typedef struct mdb_refcount_tracked { 2834 boolean_t rc_tracked; 2835 } mdb_refcount_tracked_t; 2836 2837 /* ARGSUSED */ 2838 static int 2839 refcount(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv) 2840 { 2841 mdb_refcount_t rc; 2842 mdb_refcount_removed_t rcr; 2843 mdb_refcount_tracked_t rct; 2844 int off; 2845 boolean_t released = B_FALSE; 2846 2847 if (!(flags & DCMD_ADDRSPEC)) 2848 return (DCMD_USAGE); 2849 2850 if (mdb_getopts(argc, argv, 2851 'r', MDB_OPT_SETBITS, B_TRUE, &released, 2852 NULL) != argc) 2853 return (DCMD_USAGE); 2854 2855 if (mdb_ctf_vread(&rc, "refcount_t", "mdb_refcount_t", addr, 2856 0) == -1) 2857 return (DCMD_ERR); 2858 2859 if (mdb_ctf_vread(&rcr, "refcount_t", "mdb_refcount_removed_t", addr, 2860 MDB_CTF_VREAD_QUIET) == -1) { 2861 mdb_printf("refcount_t at %p has %llu holds (untracked)\n", 2862 addr, (longlong_t)rc.rc_count); 2863 return (DCMD_OK); 2864 } 2865 2866 if (mdb_ctf_vread(&rct, "refcount_t", "mdb_refcount_tracked_t", addr, 2867 MDB_CTF_VREAD_QUIET) == -1) { 2868 /* If this is an old target, it might be tracked. */ 2869 rct.rc_tracked = B_TRUE; 2870 } 2871 2872 mdb_printf("refcount_t at %p has %llu current holds, " 2873 "%llu recently released holds\n", 2874 addr, (longlong_t)rc.rc_count, (longlong_t)rcr.rc_removed_count); 2875 2876 if (rct.rc_tracked && rc.rc_count > 0) 2877 mdb_printf("current holds:\n"); 2878 off = mdb_ctf_offsetof_by_name("refcount_t", "rc_list"); 2879 if (off == -1) 2880 return (DCMD_ERR); 2881 mdb_pwalk("list", reference_cb, (void*)B_FALSE, addr + off); 2882 2883 if (released && rcr.rc_removed_count > 0) { 2884 mdb_printf("released holds:\n"); 2885 2886 off = mdb_ctf_offsetof_by_name("refcount_t", "rc_removed"); 2887 if (off == -1) 2888 return (DCMD_ERR); 2889 mdb_pwalk("list", reference_cb, (void*)B_TRUE, addr + off); 2890 } 2891 2892 return (DCMD_OK); 2893 } 2894 2895 /* ARGSUSED */ 2896 static int 2897 sa_attr_table(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv) 2898 { 2899 sa_attr_table_t *table; 2900 sa_os_t sa_os; 2901 char *name; 2902 int i; 2903 2904 if (mdb_vread(&sa_os, sizeof (sa_os_t), addr) == -1) { 2905 mdb_warn("failed to read sa_os at %p", addr); 2906 return (DCMD_ERR); 2907 } 2908 2909 table = mdb_alloc(sizeof (sa_attr_table_t) * sa_os.sa_num_attrs, 2910 UM_SLEEP | UM_GC); 2911 name = mdb_alloc(MAXPATHLEN, UM_SLEEP | UM_GC); 2912 2913 if (mdb_vread(table, sizeof (sa_attr_table_t) * sa_os.sa_num_attrs, 2914 (uintptr_t)sa_os.sa_attr_table) == -1) { 2915 mdb_warn("failed to read sa_os at %p", addr); 2916 return (DCMD_ERR); 2917 } 2918 2919 mdb_printf("%<u>%-10s %-10s %-10s %-10s %s%</u>\n", 2920 "ATTR ID", "REGISTERED", "LENGTH", "BSWAP", "NAME"); 2921 for (i = 0; i != sa_os.sa_num_attrs; i++) { 2922 mdb_readstr(name, MAXPATHLEN, (uintptr_t)table[i].sa_name); 2923 mdb_printf("%5x %8x %8x %8x %-s\n", 2924 (int)table[i].sa_attr, (int)table[i].sa_registered, 2925 (int)table[i].sa_length, table[i].sa_byteswap, name); 2926 } 2927 2928 return (DCMD_OK); 2929 } 2930 2931 static int 2932 sa_get_off_table(uintptr_t addr, uint32_t **off_tab, int attr_count) 2933 { 2934 uintptr_t idx_table; 2935 2936 if (GETMEMB(addr, "sa_idx_tab", sa_idx_tab, idx_table)) { 2937 mdb_printf("can't find offset table in sa_idx_tab\n"); 2938 return (-1); 2939 } 2940 2941 *off_tab = mdb_alloc(attr_count * sizeof (uint32_t), 2942 UM_SLEEP | UM_GC); 2943 2944 if (mdb_vread(*off_tab, 2945 attr_count * sizeof (uint32_t), idx_table) == -1) { 2946 mdb_warn("failed to attribute offset table %p", idx_table); 2947 return (-1); 2948 } 2949 2950 return (DCMD_OK); 2951 } 2952 2953 /*ARGSUSED*/ 2954 static int 2955 sa_attr_print(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv) 2956 { 2957 uint32_t *offset_tab; 2958 int attr_count; 2959 uint64_t attr_id; 2960 uintptr_t attr_addr; 2961 uintptr_t bonus_tab, spill_tab; 2962 uintptr_t db_bonus, db_spill; 2963 uintptr_t os, os_sa; 2964 uintptr_t db_data; 2965 2966 if (argc != 1) 2967 return (DCMD_USAGE); 2968 2969 if (argv[0].a_type == MDB_TYPE_STRING) 2970 attr_id = mdb_strtoull(argv[0].a_un.a_str); 2971 else 2972 return (DCMD_USAGE); 2973 2974 if (GETMEMB(addr, "sa_handle", sa_bonus_tab, bonus_tab) || 2975 GETMEMB(addr, "sa_handle", sa_spill_tab, spill_tab) || 2976 GETMEMB(addr, "sa_handle", sa_os, os) || 2977 GETMEMB(addr, "sa_handle", sa_bonus, db_bonus) || 2978 GETMEMB(addr, "sa_handle", sa_spill, db_spill)) { 2979 mdb_printf("Can't find necessary information in sa_handle " 2980 "in sa_handle\n"); 2981 return (DCMD_ERR); 2982 } 2983 2984 if (GETMEMB(os, "objset", os_sa, os_sa)) { 2985 mdb_printf("Can't find os_sa in objset\n"); 2986 return (DCMD_ERR); 2987 } 2988 2989 if (GETMEMB(os_sa, "sa_os", sa_num_attrs, attr_count)) { 2990 mdb_printf("Can't find sa_num_attrs\n"); 2991 return (DCMD_ERR); 2992 } 2993 2994 if (attr_id > attr_count) { 2995 mdb_printf("attribute id number is out of range\n"); 2996 return (DCMD_ERR); 2997 } 2998 2999 if (bonus_tab) { 3000 if (sa_get_off_table(bonus_tab, &offset_tab, 3001 attr_count) == -1) { 3002 return (DCMD_ERR); 3003 } 3004 3005 if (GETMEMB(db_bonus, "dmu_buf", db_data, db_data)) { 3006 mdb_printf("can't find db_data in bonus dbuf\n"); 3007 return (DCMD_ERR); 3008 } 3009 } 3010 3011 if (bonus_tab && !TOC_ATTR_PRESENT(offset_tab[attr_id]) && 3012 spill_tab == NULL) { 3013 mdb_printf("Attribute does not exist\n"); 3014 return (DCMD_ERR); 3015 } else if (!TOC_ATTR_PRESENT(offset_tab[attr_id]) && spill_tab) { 3016 if (sa_get_off_table(spill_tab, &offset_tab, 3017 attr_count) == -1) { 3018 return (DCMD_ERR); 3019 } 3020 if (GETMEMB(db_spill, "dmu_buf", db_data, db_data)) { 3021 mdb_printf("can't find db_data in spill dbuf\n"); 3022 return (DCMD_ERR); 3023 } 3024 if (!TOC_ATTR_PRESENT(offset_tab[attr_id])) { 3025 mdb_printf("Attribute does not exist\n"); 3026 return (DCMD_ERR); 3027 } 3028 } 3029 attr_addr = db_data + TOC_OFF(offset_tab[attr_id]); 3030 mdb_printf("%p\n", attr_addr); 3031 return (DCMD_OK); 3032 } 3033 3034 /* ARGSUSED */ 3035 static int 3036 zfs_ace_print_common(uintptr_t addr, uint_t flags, 3037 uint64_t id, uint32_t access_mask, uint16_t ace_flags, 3038 uint16_t ace_type, int verbose) 3039 { 3040 if (DCMD_HDRSPEC(flags) && !verbose) 3041 mdb_printf("%<u>%-?s %-8s %-8s %-8s %s%</u>\n", 3042 "ADDR", "FLAGS", "MASK", "TYPE", "ID"); 3043 3044 if (!verbose) { 3045 mdb_printf("%0?p %-8x %-8x %-8x %-llx\n", addr, 3046 ace_flags, access_mask, ace_type, id); 3047 return (DCMD_OK); 3048 } 3049 3050 switch (ace_flags & ACE_TYPE_FLAGS) { 3051 case ACE_OWNER: 3052 mdb_printf("owner@:"); 3053 break; 3054 case (ACE_IDENTIFIER_GROUP | ACE_GROUP): 3055 mdb_printf("group@:"); 3056 break; 3057 case ACE_EVERYONE: 3058 mdb_printf("everyone@:"); 3059 break; 3060 case ACE_IDENTIFIER_GROUP: 3061 mdb_printf("group:%llx:", (u_longlong_t)id); 3062 break; 3063 case 0: /* User entry */ 3064 mdb_printf("user:%llx:", (u_longlong_t)id); 3065 break; 3066 } 3067 3068 /* print out permission mask */ 3069 if (access_mask & ACE_READ_DATA) 3070 mdb_printf("r"); 3071 else 3072 mdb_printf("-"); 3073 if (access_mask & ACE_WRITE_DATA) 3074 mdb_printf("w"); 3075 else 3076 mdb_printf("-"); 3077 if (access_mask & ACE_EXECUTE) 3078 mdb_printf("x"); 3079 else 3080 mdb_printf("-"); 3081 if (access_mask & ACE_APPEND_DATA) 3082 mdb_printf("p"); 3083 else 3084 mdb_printf("-"); 3085 if (access_mask & ACE_DELETE) 3086 mdb_printf("d"); 3087 else 3088 mdb_printf("-"); 3089 if (access_mask & ACE_DELETE_CHILD) 3090 mdb_printf("D"); 3091 else 3092 mdb_printf("-"); 3093 if (access_mask & ACE_READ_ATTRIBUTES) 3094 mdb_printf("a"); 3095 else 3096 mdb_printf("-"); 3097 if (access_mask & ACE_WRITE_ATTRIBUTES) 3098 mdb_printf("A"); 3099 else 3100 mdb_printf("-"); 3101 if (access_mask & ACE_READ_NAMED_ATTRS) 3102 mdb_printf("R"); 3103 else 3104 mdb_printf("-"); 3105 if (access_mask & ACE_WRITE_NAMED_ATTRS) 3106 mdb_printf("W"); 3107 else 3108 mdb_printf("-"); 3109 if (access_mask & ACE_READ_ACL) 3110 mdb_printf("c"); 3111 else 3112 mdb_printf("-"); 3113 if (access_mask & ACE_WRITE_ACL) 3114 mdb_printf("C"); 3115 else 3116 mdb_printf("-"); 3117 if (access_mask & ACE_WRITE_OWNER) 3118 mdb_printf("o"); 3119 else 3120 mdb_printf("-"); 3121 if (access_mask & ACE_SYNCHRONIZE) 3122 mdb_printf("s"); 3123 else 3124 mdb_printf("-"); 3125 3126 mdb_printf(":"); 3127 3128 /* Print out inheritance flags */ 3129 if (ace_flags & ACE_FILE_INHERIT_ACE) 3130 mdb_printf("f"); 3131 else 3132 mdb_printf("-"); 3133 if (ace_flags & ACE_DIRECTORY_INHERIT_ACE) 3134 mdb_printf("d"); 3135 else 3136 mdb_printf("-"); 3137 if (ace_flags & ACE_INHERIT_ONLY_ACE) 3138 mdb_printf("i"); 3139 else 3140 mdb_printf("-"); 3141 if (ace_flags & ACE_NO_PROPAGATE_INHERIT_ACE) 3142 mdb_printf("n"); 3143 else 3144 mdb_printf("-"); 3145 if (ace_flags & ACE_SUCCESSFUL_ACCESS_ACE_FLAG) 3146 mdb_printf("S"); 3147 else 3148 mdb_printf("-"); 3149 if (ace_flags & ACE_FAILED_ACCESS_ACE_FLAG) 3150 mdb_printf("F"); 3151 else 3152 mdb_printf("-"); 3153 if (ace_flags & ACE_INHERITED_ACE) 3154 mdb_printf("I"); 3155 else 3156 mdb_printf("-"); 3157 3158 switch (ace_type) { 3159 case ACE_ACCESS_ALLOWED_ACE_TYPE: 3160 mdb_printf(":allow\n"); 3161 break; 3162 case ACE_ACCESS_DENIED_ACE_TYPE: 3163 mdb_printf(":deny\n"); 3164 break; 3165 case ACE_SYSTEM_AUDIT_ACE_TYPE: 3166 mdb_printf(":audit\n"); 3167 break; 3168 case ACE_SYSTEM_ALARM_ACE_TYPE: 3169 mdb_printf(":alarm\n"); 3170 break; 3171 default: 3172 mdb_printf(":?\n"); 3173 } 3174 return (DCMD_OK); 3175 } 3176 3177 /* ARGSUSED */ 3178 static int 3179 zfs_ace_print(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv) 3180 { 3181 zfs_ace_t zace; 3182 int verbose = FALSE; 3183 uint64_t id; 3184 3185 if (!(flags & DCMD_ADDRSPEC)) 3186 return (DCMD_USAGE); 3187 3188 if (mdb_getopts(argc, argv, 3189 'v', MDB_OPT_SETBITS, TRUE, &verbose, TRUE, NULL) != argc) 3190 return (DCMD_USAGE); 3191 3192 if (mdb_vread(&zace, sizeof (zfs_ace_t), addr) == -1) { 3193 mdb_warn("failed to read zfs_ace_t"); 3194 return (DCMD_ERR); 3195 } 3196 3197 if ((zace.z_hdr.z_flags & ACE_TYPE_FLAGS) == 0 || 3198 (zace.z_hdr.z_flags & ACE_TYPE_FLAGS) == ACE_IDENTIFIER_GROUP) 3199 id = zace.z_fuid; 3200 else 3201 id = -1; 3202 3203 return (zfs_ace_print_common(addr, flags, id, zace.z_hdr.z_access_mask, 3204 zace.z_hdr.z_flags, zace.z_hdr.z_type, verbose)); 3205 } 3206 3207 /* ARGSUSED */ 3208 static int 3209 zfs_ace0_print(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv) 3210 { 3211 ace_t ace; 3212 uint64_t id; 3213 int verbose = FALSE; 3214 3215 if (!(flags & DCMD_ADDRSPEC)) 3216 return (DCMD_USAGE); 3217 3218 if (mdb_getopts(argc, argv, 3219 'v', MDB_OPT_SETBITS, TRUE, &verbose, TRUE, NULL) != argc) 3220 return (DCMD_USAGE); 3221 3222 if (mdb_vread(&ace, sizeof (ace_t), addr) == -1) { 3223 mdb_warn("failed to read ace_t"); 3224 return (DCMD_ERR); 3225 } 3226 3227 if ((ace.a_flags & ACE_TYPE_FLAGS) == 0 || 3228 (ace.a_flags & ACE_TYPE_FLAGS) == ACE_IDENTIFIER_GROUP) 3229 id = ace.a_who; 3230 else 3231 id = -1; 3232 3233 return (zfs_ace_print_common(addr, flags, id, ace.a_access_mask, 3234 ace.a_flags, ace.a_type, verbose)); 3235 } 3236 3237 typedef struct acl_dump_args { 3238 int a_argc; 3239 const mdb_arg_t *a_argv; 3240 uint16_t a_version; 3241 int a_flags; 3242 } acl_dump_args_t; 3243 3244 /* ARGSUSED */ 3245 static int 3246 acl_aces_cb(uintptr_t addr, const void *unknown, void *arg) 3247 { 3248 acl_dump_args_t *acl_args = (acl_dump_args_t *)arg; 3249 3250 if (acl_args->a_version == 1) { 3251 if (mdb_call_dcmd("zfs_ace", addr, 3252 DCMD_ADDRSPEC|acl_args->a_flags, acl_args->a_argc, 3253 acl_args->a_argv) != DCMD_OK) { 3254 return (WALK_ERR); 3255 } 3256 } else { 3257 if (mdb_call_dcmd("zfs_ace0", addr, 3258 DCMD_ADDRSPEC|acl_args->a_flags, acl_args->a_argc, 3259 acl_args->a_argv) != DCMD_OK) { 3260 return (WALK_ERR); 3261 } 3262 } 3263 acl_args->a_flags = DCMD_LOOP; 3264 return (WALK_NEXT); 3265 } 3266 3267 /* ARGSUSED */ 3268 static int 3269 acl_cb(uintptr_t addr, const void *unknown, void *arg) 3270 { 3271 acl_dump_args_t *acl_args = (acl_dump_args_t *)arg; 3272 3273 if (acl_args->a_version == 1) { 3274 if (mdb_pwalk("zfs_acl_node_aces", acl_aces_cb, 3275 arg, addr) != 0) { 3276 mdb_warn("can't walk ACEs"); 3277 return (DCMD_ERR); 3278 } 3279 } else { 3280 if (mdb_pwalk("zfs_acl_node_aces0", acl_aces_cb, 3281 arg, addr) != 0) { 3282 mdb_warn("can't walk ACEs"); 3283 return (DCMD_ERR); 3284 } 3285 } 3286 return (WALK_NEXT); 3287 } 3288 3289 /* ARGSUSED */ 3290 static int 3291 zfs_acl_dump(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv) 3292 { 3293 zfs_acl_t zacl; 3294 int verbose = FALSE; 3295 acl_dump_args_t acl_args; 3296 3297 if (!(flags & DCMD_ADDRSPEC)) 3298 return (DCMD_USAGE); 3299 3300 if (mdb_getopts(argc, argv, 3301 'v', MDB_OPT_SETBITS, TRUE, &verbose, TRUE, NULL) != argc) 3302 return (DCMD_USAGE); 3303 3304 if (mdb_vread(&zacl, sizeof (zfs_acl_t), addr) == -1) { 3305 mdb_warn("failed to read zfs_acl_t"); 3306 return (DCMD_ERR); 3307 } 3308 3309 acl_args.a_argc = argc; 3310 acl_args.a_argv = argv; 3311 acl_args.a_version = zacl.z_version; 3312 acl_args.a_flags = DCMD_LOOPFIRST; 3313 3314 if (mdb_pwalk("zfs_acl_node", acl_cb, &acl_args, addr) != 0) { 3315 mdb_warn("can't walk ACL"); 3316 return (DCMD_ERR); 3317 } 3318 3319 return (DCMD_OK); 3320 } 3321 3322 /* ARGSUSED */ 3323 static int 3324 zfs_acl_node_walk_init(mdb_walk_state_t *wsp) 3325 { 3326 if (wsp->walk_addr == NULL) { 3327 mdb_warn("must supply address of zfs_acl_node_t\n"); 3328 return (WALK_ERR); 3329 } 3330 3331 wsp->walk_addr += 3332 mdb_ctf_offsetof_by_name(ZFS_STRUCT "zfs_acl", "z_acl"); 3333 3334 if (mdb_layered_walk("list", wsp) == -1) { 3335 mdb_warn("failed to walk 'list'\n"); 3336 return (WALK_ERR); 3337 } 3338 3339 return (WALK_NEXT); 3340 } 3341 3342 static int 3343 zfs_acl_node_walk_step(mdb_walk_state_t *wsp) 3344 { 3345 zfs_acl_node_t aclnode; 3346 3347 if (mdb_vread(&aclnode, sizeof (zfs_acl_node_t), 3348 wsp->walk_addr) == -1) { 3349 mdb_warn("failed to read zfs_acl_node at %p", wsp->walk_addr); 3350 return (WALK_ERR); 3351 } 3352 3353 return (wsp->walk_callback(wsp->walk_addr, &aclnode, wsp->walk_cbdata)); 3354 } 3355 3356 typedef struct ace_walk_data { 3357 int ace_count; 3358 int ace_version; 3359 } ace_walk_data_t; 3360 3361 static int 3362 zfs_aces_walk_init_common(mdb_walk_state_t *wsp, int version, 3363 int ace_count, uintptr_t ace_data) 3364 { 3365 ace_walk_data_t *ace_walk_data; 3366 3367 if (wsp->walk_addr == NULL) { 3368 mdb_warn("must supply address of zfs_acl_node_t\n"); 3369 return (WALK_ERR); 3370 } 3371 3372 ace_walk_data = mdb_alloc(sizeof (ace_walk_data_t), UM_SLEEP | UM_GC); 3373 3374 ace_walk_data->ace_count = ace_count; 3375 ace_walk_data->ace_version = version; 3376 3377 wsp->walk_addr = ace_data; 3378 wsp->walk_data = ace_walk_data; 3379 3380 return (WALK_NEXT); 3381 } 3382 3383 static int 3384 zfs_acl_node_aces_walk_init_common(mdb_walk_state_t *wsp, int version) 3385 { 3386 static int gotid; 3387 static mdb_ctf_id_t acl_id; 3388 int z_ace_count; 3389 uintptr_t z_acldata; 3390 3391 if (!gotid) { 3392 if (mdb_ctf_lookup_by_name("struct zfs_acl_node", 3393 &acl_id) == -1) { 3394 mdb_warn("couldn't find struct zfs_acl_node"); 3395 return (DCMD_ERR); 3396 } 3397 gotid = TRUE; 3398 } 3399 3400 if (GETMEMBID(wsp->walk_addr, &acl_id, z_ace_count, z_ace_count)) { 3401 return (DCMD_ERR); 3402 } 3403 if (GETMEMBID(wsp->walk_addr, &acl_id, z_acldata, z_acldata)) { 3404 return (DCMD_ERR); 3405 } 3406 3407 return (zfs_aces_walk_init_common(wsp, version, 3408 z_ace_count, z_acldata)); 3409 } 3410 3411 /* ARGSUSED */ 3412 static int 3413 zfs_acl_node_aces_walk_init(mdb_walk_state_t *wsp) 3414 { 3415 return (zfs_acl_node_aces_walk_init_common(wsp, 1)); 3416 } 3417 3418 /* ARGSUSED */ 3419 static int 3420 zfs_acl_node_aces0_walk_init(mdb_walk_state_t *wsp) 3421 { 3422 return (zfs_acl_node_aces_walk_init_common(wsp, 0)); 3423 } 3424 3425 static int 3426 zfs_aces_walk_step(mdb_walk_state_t *wsp) 3427 { 3428 ace_walk_data_t *ace_data = wsp->walk_data; 3429 zfs_ace_t zace; 3430 ace_t *acep; 3431 int status; 3432 int entry_type; 3433 int allow_type; 3434 uintptr_t ptr; 3435 3436 if (ace_data->ace_count == 0) 3437 return (WALK_DONE); 3438 3439 if (mdb_vread(&zace, sizeof (zfs_ace_t), wsp->walk_addr) == -1) { 3440 mdb_warn("failed to read zfs_ace_t at %#lx", 3441 wsp->walk_addr); 3442 return (WALK_ERR); 3443 } 3444 3445 switch (ace_data->ace_version) { 3446 case 0: 3447 acep = (ace_t *)&zace; 3448 entry_type = acep->a_flags & ACE_TYPE_FLAGS; 3449 allow_type = acep->a_type; 3450 break; 3451 case 1: 3452 entry_type = zace.z_hdr.z_flags & ACE_TYPE_FLAGS; 3453 allow_type = zace.z_hdr.z_type; 3454 break; 3455 default: 3456 return (WALK_ERR); 3457 } 3458 3459 ptr = (uintptr_t)wsp->walk_addr; 3460 switch (entry_type) { 3461 case ACE_OWNER: 3462 case ACE_EVERYONE: 3463 case (ACE_IDENTIFIER_GROUP | ACE_GROUP): 3464 ptr += ace_data->ace_version == 0 ? 3465 sizeof (ace_t) : sizeof (zfs_ace_hdr_t); 3466 break; 3467 case ACE_IDENTIFIER_GROUP: 3468 default: 3469 switch (allow_type) { 3470 case ACE_ACCESS_ALLOWED_OBJECT_ACE_TYPE: 3471 case ACE_ACCESS_DENIED_OBJECT_ACE_TYPE: 3472 case ACE_SYSTEM_AUDIT_OBJECT_ACE_TYPE: 3473 case ACE_SYSTEM_ALARM_OBJECT_ACE_TYPE: 3474 ptr += ace_data->ace_version == 0 ? 3475 sizeof (ace_t) : sizeof (zfs_object_ace_t); 3476 break; 3477 default: 3478 ptr += ace_data->ace_version == 0 ? 3479 sizeof (ace_t) : sizeof (zfs_ace_t); 3480 break; 3481 } 3482 } 3483 3484 ace_data->ace_count--; 3485 status = wsp->walk_callback(wsp->walk_addr, 3486 (void *)(uintptr_t)&zace, wsp->walk_cbdata); 3487 3488 wsp->walk_addr = ptr; 3489 return (status); 3490 } 3491 3492 typedef struct mdb_zfs_rrwlock { 3493 uintptr_t rr_writer; 3494 boolean_t rr_writer_wanted; 3495 } mdb_zfs_rrwlock_t; 3496 3497 static uint_t rrw_key; 3498 3499 /* ARGSUSED */ 3500 static int 3501 rrwlock(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv) 3502 { 3503 mdb_zfs_rrwlock_t rrw; 3504 3505 if (rrw_key == 0) { 3506 if (mdb_ctf_readsym(&rrw_key, "uint_t", "rrw_tsd_key", 0) == -1) 3507 return (DCMD_ERR); 3508 } 3509 3510 if (mdb_ctf_vread(&rrw, "rrwlock_t", "mdb_zfs_rrwlock_t", addr, 3511 0) == -1) 3512 return (DCMD_ERR); 3513 3514 if (rrw.rr_writer != 0) { 3515 mdb_printf("write lock held by thread %lx\n", rrw.rr_writer); 3516 return (DCMD_OK); 3517 } 3518 3519 if (rrw.rr_writer_wanted) { 3520 mdb_printf("writer wanted\n"); 3521 } 3522 3523 mdb_printf("anonymous references:\n"); 3524 (void) mdb_call_dcmd("refcount", addr + 3525 mdb_ctf_offsetof_by_name(ZFS_STRUCT "rrwlock", "rr_anon_rcount"), 3526 DCMD_ADDRSPEC, 0, NULL); 3527 3528 mdb_printf("linked references:\n"); 3529 (void) mdb_call_dcmd("refcount", addr + 3530 mdb_ctf_offsetof_by_name(ZFS_STRUCT "rrwlock", "rr_linked_rcount"), 3531 DCMD_ADDRSPEC, 0, NULL); 3532 3533 /* 3534 * XXX This should find references from 3535 * "::walk thread | ::tsd -v <rrw_key>", but there is no support 3536 * for programmatic consumption of dcmds, so this would be 3537 * difficult, potentially requiring reimplementing ::tsd (both 3538 * user and kernel versions) in this MDB module. 3539 */ 3540 3541 return (DCMD_OK); 3542 } 3543 3544 typedef struct mdb_arc_buf_hdr_t { 3545 uint16_t b_psize; 3546 uint16_t b_lsize; 3547 struct { 3548 uint32_t b_bufcnt; 3549 uintptr_t b_state; 3550 uintptr_t b_pdata; 3551 } b_l1hdr; 3552 } mdb_arc_buf_hdr_t; 3553 3554 enum arc_cflags { 3555 ARC_CFLAG_VERBOSE = 1 << 0, 3556 ARC_CFLAG_ANON = 1 << 1, 3557 ARC_CFLAG_MRU = 1 << 2, 3558 ARC_CFLAG_MFU = 1 << 3, 3559 ARC_CFLAG_BUFS = 1 << 4, 3560 }; 3561 3562 typedef struct arc_compression_stats_data { 3563 GElf_Sym anon_sym; /* ARC_anon symbol */ 3564 GElf_Sym mru_sym; /* ARC_mru symbol */ 3565 GElf_Sym mrug_sym; /* ARC_mru_ghost symbol */ 3566 GElf_Sym mfu_sym; /* ARC_mfu symbol */ 3567 GElf_Sym mfug_sym; /* ARC_mfu_ghost symbol */ 3568 GElf_Sym l2c_sym; /* ARC_l2c_only symbol */ 3569 uint64_t *anon_c_hist; /* histogram of compressed sizes in anon */ 3570 uint64_t *anon_u_hist; /* histogram of uncompressed sizes in anon */ 3571 uint64_t *anon_bufs; /* histogram of buffer counts in anon state */ 3572 uint64_t *mru_c_hist; /* histogram of compressed sizes in mru */ 3573 uint64_t *mru_u_hist; /* histogram of uncompressed sizes in mru */ 3574 uint64_t *mru_bufs; /* histogram of buffer counts in mru */ 3575 uint64_t *mfu_c_hist; /* histogram of compressed sizes in mfu */ 3576 uint64_t *mfu_u_hist; /* histogram of uncompressed sizes in mfu */ 3577 uint64_t *mfu_bufs; /* histogram of buffer counts in mfu */ 3578 uint64_t *all_c_hist; /* histogram of compressed anon + mru + mfu */ 3579 uint64_t *all_u_hist; /* histogram of uncompressed anon + mru + mfu */ 3580 uint64_t *all_bufs; /* histogram of buffer counts in all states */ 3581 int arc_cflags; /* arc compression flags, specified by user */ 3582 int hist_nbuckets; /* number of buckets in each histogram */ 3583 } arc_compression_stats_data_t; 3584 3585 int 3586 highbit64(uint64_t i) 3587 { 3588 int h = 1; 3589 3590 if (i == 0) 3591 return (0); 3592 if (i & 0xffffffff00000000ULL) { 3593 h += 32; i >>= 32; 3594 } 3595 if (i & 0xffff0000) { 3596 h += 16; i >>= 16; 3597 } 3598 if (i & 0xff00) { 3599 h += 8; i >>= 8; 3600 } 3601 if (i & 0xf0) { 3602 h += 4; i >>= 4; 3603 } 3604 if (i & 0xc) { 3605 h += 2; i >>= 2; 3606 } 3607 if (i & 0x2) { 3608 h += 1; 3609 } 3610 return (h); 3611 } 3612 3613 /* ARGSUSED */ 3614 static int 3615 arc_compression_stats_cb(uintptr_t addr, const void *unknown, void *arg) 3616 { 3617 arc_compression_stats_data_t *data = arg; 3618 mdb_arc_buf_hdr_t hdr; 3619 int cbucket, ubucket, bufcnt; 3620 3621 if (mdb_ctf_vread(&hdr, "arc_buf_hdr_t", "mdb_arc_buf_hdr_t", 3622 addr, 0) == -1) { 3623 return (WALK_ERR); 3624 } 3625 3626 /* 3627 * Headers in the ghost states, or the l2c_only state don't have 3628 * arc buffers linked off of them. Thus, their compressed size 3629 * is meaningless, so we skip these from the stats. 3630 */ 3631 if (hdr.b_l1hdr.b_state == data->mrug_sym.st_value || 3632 hdr.b_l1hdr.b_state == data->mfug_sym.st_value || 3633 hdr.b_l1hdr.b_state == data->l2c_sym.st_value) { 3634 return (WALK_NEXT); 3635 } 3636 3637 /* 3638 * The physical size (compressed) and logical size 3639 * (uncompressed) are in units of SPA_MINBLOCKSIZE. By default, 3640 * we use the log2 of this value (rounded down to the nearest 3641 * integer) to determine the bucket to assign this header to. 3642 * Thus, the histogram is logarithmic with respect to the size 3643 * of the header. For example, the following is a mapping of the 3644 * bucket numbers and the range of header sizes they correspond to: 3645 * 3646 * 0: 0 byte headers 3647 * 1: 512 byte headers 3648 * 2: [1024 - 2048) byte headers 3649 * 3: [2048 - 4096) byte headers 3650 * 4: [4096 - 8192) byte headers 3651 * 5: [8192 - 16394) byte headers 3652 * 6: [16384 - 32768) byte headers 3653 * 7: [32768 - 65536) byte headers 3654 * 8: [65536 - 131072) byte headers 3655 * 9: 131072 byte headers 3656 * 3657 * If the ARC_CFLAG_VERBOSE flag was specified, we use the 3658 * physical and logical sizes directly. Thus, the histogram will 3659 * no longer be logarithmic; instead it will be linear with 3660 * respect to the size of the header. The following is a mapping 3661 * of the first many bucket numbers and the header size they 3662 * correspond to: 3663 * 3664 * 0: 0 byte headers 3665 * 1: 512 byte headers 3666 * 2: 1024 byte headers 3667 * 3: 1536 byte headers 3668 * 4: 2048 byte headers 3669 * 5: 2560 byte headers 3670 * 6: 3072 byte headers 3671 * 3672 * And so on. Keep in mind that a range of sizes isn't used in 3673 * the case of linear scale because the headers can only 3674 * increment or decrement in sizes of 512 bytes. So, it's not 3675 * possible for a header to be sized in between whats listed 3676 * above. 3677 * 3678 * Also, the above mapping values were calculated assuming a 3679 * SPA_MINBLOCKSHIFT of 512 bytes and a SPA_MAXBLOCKSIZE of 128K. 3680 */ 3681 3682 if (data->arc_cflags & ARC_CFLAG_VERBOSE) { 3683 cbucket = hdr.b_psize; 3684 ubucket = hdr.b_lsize; 3685 } else { 3686 cbucket = highbit64(hdr.b_psize); 3687 ubucket = highbit64(hdr.b_lsize); 3688 } 3689 3690 bufcnt = hdr.b_l1hdr.b_bufcnt; 3691 if (bufcnt >= data->hist_nbuckets) 3692 bufcnt = data->hist_nbuckets - 1; 3693 3694 /* Ensure we stay within the bounds of the histogram array */ 3695 ASSERT3U(cbucket, <, data->hist_nbuckets); 3696 ASSERT3U(ubucket, <, data->hist_nbuckets); 3697 3698 if (hdr.b_l1hdr.b_state == data->anon_sym.st_value) { 3699 data->anon_c_hist[cbucket]++; 3700 data->anon_u_hist[ubucket]++; 3701 data->anon_bufs[bufcnt]++; 3702 } else if (hdr.b_l1hdr.b_state == data->mru_sym.st_value) { 3703 data->mru_c_hist[cbucket]++; 3704 data->mru_u_hist[ubucket]++; 3705 data->mru_bufs[bufcnt]++; 3706 } else if (hdr.b_l1hdr.b_state == data->mfu_sym.st_value) { 3707 data->mfu_c_hist[cbucket]++; 3708 data->mfu_u_hist[ubucket]++; 3709 data->mfu_bufs[bufcnt]++; 3710 } 3711 3712 data->all_c_hist[cbucket]++; 3713 data->all_u_hist[ubucket]++; 3714 data->all_bufs[bufcnt]++; 3715 3716 return (WALK_NEXT); 3717 } 3718 3719 /* ARGSUSED */ 3720 static int 3721 arc_compression_stats(uintptr_t addr, uint_t flags, int argc, 3722 const mdb_arg_t *argv) 3723 { 3724 arc_compression_stats_data_t data = { 0 }; 3725 unsigned int max_shifted = SPA_MAXBLOCKSIZE >> SPA_MINBLOCKSHIFT; 3726 unsigned int hist_size; 3727 char range[32]; 3728 int rc = DCMD_OK; 3729 3730 if (mdb_getopts(argc, argv, 3731 'v', MDB_OPT_SETBITS, ARC_CFLAG_VERBOSE, &data.arc_cflags, 3732 'a', MDB_OPT_SETBITS, ARC_CFLAG_ANON, &data.arc_cflags, 3733 'b', MDB_OPT_SETBITS, ARC_CFLAG_BUFS, &data.arc_cflags, 3734 'r', MDB_OPT_SETBITS, ARC_CFLAG_MRU, &data.arc_cflags, 3735 'f', MDB_OPT_SETBITS, ARC_CFLAG_MFU, &data.arc_cflags) != argc) 3736 return (DCMD_USAGE); 3737 3738 if (mdb_lookup_by_obj(ZFS_OBJ_NAME, "ARC_anon", &data.anon_sym) || 3739 mdb_lookup_by_obj(ZFS_OBJ_NAME, "ARC_mru", &data.mru_sym) || 3740 mdb_lookup_by_obj(ZFS_OBJ_NAME, "ARC_mru_ghost", &data.mrug_sym) || 3741 mdb_lookup_by_obj(ZFS_OBJ_NAME, "ARC_mfu", &data.mfu_sym) || 3742 mdb_lookup_by_obj(ZFS_OBJ_NAME, "ARC_mfu_ghost", &data.mfug_sym) || 3743 mdb_lookup_by_obj(ZFS_OBJ_NAME, "ARC_l2c_only", &data.l2c_sym)) { 3744 mdb_warn("can't find arc state symbol"); 3745 return (DCMD_ERR); 3746 } 3747 3748 /* 3749 * Determine the maximum expected size for any header, and use 3750 * this to determine the number of buckets needed for each 3751 * histogram. If ARC_CFLAG_VERBOSE is specified, this value is 3752 * used directly; otherwise the log2 of the maximum size is 3753 * used. Thus, if using a log2 scale there's a maximum of 10 3754 * possible buckets, while the linear scale (when using 3755 * ARC_CFLAG_VERBOSE) has a maximum of 257 buckets. 3756 */ 3757 if (data.arc_cflags & ARC_CFLAG_VERBOSE) 3758 data.hist_nbuckets = max_shifted + 1; 3759 else 3760 data.hist_nbuckets = highbit64(max_shifted) + 1; 3761 3762 hist_size = sizeof (uint64_t) * data.hist_nbuckets; 3763 3764 data.anon_c_hist = mdb_zalloc(hist_size, UM_SLEEP); 3765 data.anon_u_hist = mdb_zalloc(hist_size, UM_SLEEP); 3766 data.anon_bufs = mdb_zalloc(hist_size, UM_SLEEP); 3767 3768 data.mru_c_hist = mdb_zalloc(hist_size, UM_SLEEP); 3769 data.mru_u_hist = mdb_zalloc(hist_size, UM_SLEEP); 3770 data.mru_bufs = mdb_zalloc(hist_size, UM_SLEEP); 3771 3772 data.mfu_c_hist = mdb_zalloc(hist_size, UM_SLEEP); 3773 data.mfu_u_hist = mdb_zalloc(hist_size, UM_SLEEP); 3774 data.mfu_bufs = mdb_zalloc(hist_size, UM_SLEEP); 3775 3776 data.all_c_hist = mdb_zalloc(hist_size, UM_SLEEP); 3777 data.all_u_hist = mdb_zalloc(hist_size, UM_SLEEP); 3778 data.all_bufs = mdb_zalloc(hist_size, UM_SLEEP); 3779 3780 if (mdb_walk("arc_buf_hdr_t_full", arc_compression_stats_cb, 3781 &data) != 0) { 3782 mdb_warn("can't walk arc_buf_hdr's"); 3783 rc = DCMD_ERR; 3784 goto out; 3785 } 3786 3787 if (data.arc_cflags & ARC_CFLAG_VERBOSE) { 3788 rc = mdb_snprintf(range, sizeof (range), 3789 "[n*%llu, (n+1)*%llu)", SPA_MINBLOCKSIZE, 3790 SPA_MINBLOCKSIZE); 3791 } else { 3792 rc = mdb_snprintf(range, sizeof (range), 3793 "[2^(n-1)*%llu, 2^n*%llu)", SPA_MINBLOCKSIZE, 3794 SPA_MINBLOCKSIZE); 3795 } 3796 3797 if (rc < 0) { 3798 /* snprintf failed, abort the dcmd */ 3799 rc = DCMD_ERR; 3800 goto out; 3801 } else { 3802 /* snprintf succeeded above, reset return code */ 3803 rc = DCMD_OK; 3804 } 3805 3806 if (data.arc_cflags & ARC_CFLAG_ANON) { 3807 if (data.arc_cflags & ARC_CFLAG_BUFS) { 3808 mdb_printf("Histogram of the number of anon buffers " 3809 "that are associated with an arc hdr.\n"); 3810 dump_histogram(data.anon_bufs, data.hist_nbuckets, 0); 3811 mdb_printf("\n"); 3812 } 3813 mdb_printf("Histogram of compressed anon buffers.\n" 3814 "Each bucket represents buffers of size: %s.\n", range); 3815 dump_histogram(data.anon_c_hist, data.hist_nbuckets, 0); 3816 mdb_printf("\n"); 3817 3818 mdb_printf("Histogram of uncompressed anon buffers.\n" 3819 "Each bucket represents buffers of size: %s.\n", range); 3820 dump_histogram(data.anon_u_hist, data.hist_nbuckets, 0); 3821 mdb_printf("\n"); 3822 } 3823 3824 if (data.arc_cflags & ARC_CFLAG_MRU) { 3825 if (data.arc_cflags & ARC_CFLAG_BUFS) { 3826 mdb_printf("Histogram of the number of mru buffers " 3827 "that are associated with an arc hdr.\n"); 3828 dump_histogram(data.mru_bufs, data.hist_nbuckets, 0); 3829 mdb_printf("\n"); 3830 } 3831 mdb_printf("Histogram of compressed mru buffers.\n" 3832 "Each bucket represents buffers of size: %s.\n", range); 3833 dump_histogram(data.mru_c_hist, data.hist_nbuckets, 0); 3834 mdb_printf("\n"); 3835 3836 mdb_printf("Histogram of uncompressed mru buffers.\n" 3837 "Each bucket represents buffers of size: %s.\n", range); 3838 dump_histogram(data.mru_u_hist, data.hist_nbuckets, 0); 3839 mdb_printf("\n"); 3840 } 3841 3842 if (data.arc_cflags & ARC_CFLAG_MFU) { 3843 if (data.arc_cflags & ARC_CFLAG_BUFS) { 3844 mdb_printf("Histogram of the number of mfu buffers " 3845 "that are associated with an arc hdr.\n"); 3846 dump_histogram(data.mfu_bufs, data.hist_nbuckets, 0); 3847 mdb_printf("\n"); 3848 } 3849 3850 mdb_printf("Histogram of compressed mfu buffers.\n" 3851 "Each bucket represents buffers of size: %s.\n", range); 3852 dump_histogram(data.mfu_c_hist, data.hist_nbuckets, 0); 3853 mdb_printf("\n"); 3854 3855 mdb_printf("Histogram of uncompressed mfu buffers.\n" 3856 "Each bucket represents buffers of size: %s.\n", range); 3857 dump_histogram(data.mfu_u_hist, data.hist_nbuckets, 0); 3858 mdb_printf("\n"); 3859 } 3860 3861 if (data.arc_cflags & ARC_CFLAG_BUFS) { 3862 mdb_printf("Histogram of all buffers that " 3863 "are associated with an arc hdr.\n"); 3864 dump_histogram(data.all_bufs, data.hist_nbuckets, 0); 3865 mdb_printf("\n"); 3866 } 3867 3868 mdb_printf("Histogram of all compressed buffers.\n" 3869 "Each bucket represents buffers of size: %s.\n", range); 3870 dump_histogram(data.all_c_hist, data.hist_nbuckets, 0); 3871 mdb_printf("\n"); 3872 3873 mdb_printf("Histogram of all uncompressed buffers.\n" 3874 "Each bucket represents buffers of size: %s.\n", range); 3875 dump_histogram(data.all_u_hist, data.hist_nbuckets, 0); 3876 3877 out: 3878 mdb_free(data.anon_c_hist, hist_size); 3879 mdb_free(data.anon_u_hist, hist_size); 3880 mdb_free(data.anon_bufs, hist_size); 3881 3882 mdb_free(data.mru_c_hist, hist_size); 3883 mdb_free(data.mru_u_hist, hist_size); 3884 mdb_free(data.mru_bufs, hist_size); 3885 3886 mdb_free(data.mfu_c_hist, hist_size); 3887 mdb_free(data.mfu_u_hist, hist_size); 3888 mdb_free(data.mfu_bufs, hist_size); 3889 3890 mdb_free(data.all_c_hist, hist_size); 3891 mdb_free(data.all_u_hist, hist_size); 3892 mdb_free(data.all_bufs, hist_size); 3893 3894 return (rc); 3895 } 3896 3897 /* 3898 * MDB module linkage information: 3899 * 3900 * We declare a list of structures describing our dcmds, and a function 3901 * named _mdb_init to return a pointer to our module information. 3902 */ 3903 3904 static const mdb_dcmd_t dcmds[] = { 3905 { "arc", "[-bkmg]", "print ARC variables", arc_print }, 3906 { "blkptr", ":", "print blkptr_t", blkptr }, 3907 { "dbuf", ":", "print dmu_buf_impl_t", dbuf }, 3908 { "dbuf_stats", ":", "dbuf stats", dbuf_stats }, 3909 { "dbufs", 3910 "\t[-O objset_t*] [-n objset_name | \"mos\"] " 3911 "[-o object | \"mdn\"] \n" 3912 "\t[-l level] [-b blkid | \"bonus\"]", 3913 "find dmu_buf_impl_t's that match specified criteria", dbufs }, 3914 { "abuf_find", "dva_word[0] dva_word[1]", 3915 "find arc_buf_hdr_t of a specified DVA", 3916 abuf_find }, 3917 { "spa", "?[-cevmMh]\n" 3918 "\t-c display spa config\n" 3919 "\t-e display vdev statistics\n" 3920 "\t-v display vdev information\n" 3921 "\t-m display metaslab statistics\n" 3922 "\t-M display metaslab group statistics\n" 3923 "\t-h display histogram (requires -m or -M)\n", 3924 "spa_t summary", spa_print }, 3925 { "spa_config", ":", "print spa_t configuration", spa_print_config }, 3926 { "spa_space", ":[-b]", "print spa_t on-disk space usage", spa_space }, 3927 { "spa_vdevs", ":[-emMh]\n" 3928 "\t-e display vdev statistics\n" 3929 "\t-m dispaly metaslab statistics\n" 3930 "\t-M display metaslab group statistic\n" 3931 "\t-h display histogram (requires -m or -M)\n", 3932 "given a spa_t, print vdev summary", spa_vdevs }, 3933 { "vdev", ":[-remMh]\n" 3934 "\t-r display recursively\n" 3935 "\t-e display statistics\n" 3936 "\t-m display metaslab statistics (top level vdev only)\n" 3937 "\t-M display metaslab group statistics (top level vdev only)\n" 3938 "\t-h display histogram (requires -m or -M)\n", 3939 "vdev_t summary", vdev_print }, 3940 { "zio", ":[-cpr]\n" 3941 "\t-c display children\n" 3942 "\t-p display parents\n" 3943 "\t-r display recursively", 3944 "zio_t summary", zio_print }, 3945 { "zio_state", "?", "print out all zio_t structures on system or " 3946 "for a particular pool", zio_state }, 3947 { "zfs_blkstats", ":[-v]", 3948 "given a spa_t, print block type stats from last scrub", 3949 zfs_blkstats }, 3950 { "zfs_params", "", "print zfs tunable parameters", zfs_params }, 3951 { "refcount", ":[-r]\n" 3952 "\t-r display recently removed references", 3953 "print refcount_t holders", refcount }, 3954 { "zap_leaf", "", "print zap_leaf_phys_t", zap_leaf }, 3955 { "zfs_aces", ":[-v]", "print all ACEs from a zfs_acl_t", 3956 zfs_acl_dump }, 3957 { "zfs_ace", ":[-v]", "print zfs_ace", zfs_ace_print }, 3958 { "zfs_ace0", ":[-v]", "print zfs_ace0", zfs_ace0_print }, 3959 { "sa_attr_table", ":", "print SA attribute table from sa_os_t", 3960 sa_attr_table}, 3961 { "sa_attr", ": attr_id", 3962 "print SA attribute address when given sa_handle_t", sa_attr_print}, 3963 { "zfs_dbgmsg", ":[-va]", 3964 "print zfs debug log", dbgmsg}, 3965 { "rrwlock", ":", 3966 "print rrwlock_t, including readers", rrwlock}, 3967 { "metaslab_weight", "weight", 3968 "print metaslab weight", metaslab_weight}, 3969 { "metaslab_trace", ":", 3970 "print metaslab allocation trace records", metaslab_trace}, 3971 { "arc_compression_stats", ":[-vabrf]\n" 3972 "\t-v verbose, display a linearly scaled histogram\n" 3973 "\t-a display ARC_anon state statistics individually\n" 3974 "\t-r display ARC_mru state statistics individually\n" 3975 "\t-f display ARC_mfu state statistics individually\n" 3976 "\t-b display histogram of buffer counts\n", 3977 "print a histogram of compressed arc buffer sizes", 3978 arc_compression_stats}, 3979 { NULL } 3980 }; 3981 3982 static const mdb_walker_t walkers[] = { 3983 { "zms_freelist", "walk ZFS metaslab freelist", 3984 freelist_walk_init, freelist_walk_step, NULL }, 3985 { "txg_list", "given any txg_list_t *, walk all entries in all txgs", 3986 txg_list_walk_init, txg_list_walk_step, NULL }, 3987 { "txg_list0", "given any txg_list_t *, walk all entries in txg 0", 3988 txg_list0_walk_init, txg_list_walk_step, NULL }, 3989 { "txg_list1", "given any txg_list_t *, walk all entries in txg 1", 3990 txg_list1_walk_init, txg_list_walk_step, NULL }, 3991 { "txg_list2", "given any txg_list_t *, walk all entries in txg 2", 3992 txg_list2_walk_init, txg_list_walk_step, NULL }, 3993 { "txg_list3", "given any txg_list_t *, walk all entries in txg 3", 3994 txg_list3_walk_init, txg_list_walk_step, NULL }, 3995 { "zio", "walk all zio structures, optionally for a particular spa_t", 3996 zio_walk_init, zio_walk_step, NULL }, 3997 { "zio_root", 3998 "walk all root zio_t structures, optionally for a particular spa_t", 3999 zio_walk_init, zio_walk_root_step, NULL }, 4000 { "spa", "walk all spa_t entries in the namespace", 4001 spa_walk_init, spa_walk_step, NULL }, 4002 { "metaslab", "given a spa_t *, walk all metaslab_t structures", 4003 metaslab_walk_init, metaslab_walk_step, NULL }, 4004 { "multilist", "given a multilist_t *, walk all list_t structures", 4005 multilist_walk_init, multilist_walk_step, NULL }, 4006 { "zfs_acl_node", "given a zfs_acl_t, walk all zfs_acl_nodes", 4007 zfs_acl_node_walk_init, zfs_acl_node_walk_step, NULL }, 4008 { "zfs_acl_node_aces", "given a zfs_acl_node_t, walk all ACEs", 4009 zfs_acl_node_aces_walk_init, zfs_aces_walk_step, NULL }, 4010 { "zfs_acl_node_aces0", 4011 "given a zfs_acl_node_t, walk all ACEs as ace_t", 4012 zfs_acl_node_aces0_walk_init, zfs_aces_walk_step, NULL }, 4013 { NULL } 4014 }; 4015 4016 static const mdb_modinfo_t modinfo = { 4017 MDB_API_VERSION, dcmds, walkers 4018 }; 4019 4020 const mdb_modinfo_t * 4021 _mdb_init(void) 4022 { 4023 return (&modinfo); 4024 } 4025