1 /* 2 * CDDL HEADER START 3 * 4 * The contents of this file are subject to the terms of the 5 * Common Development and Distribution License (the "License"). 6 * You may not use this file except in compliance with the License. 7 * 8 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE 9 * or http://www.opensolaris.org/os/licensing. 10 * See the License for the specific language governing permissions 11 * and limitations under the License. 12 * 13 * When distributing Covered Code, include this CDDL HEADER in each 14 * file and include the License file at usr/src/OPENSOLARIS.LICENSE. 15 * If applicable, add the following below this CDDL HEADER, with the 16 * fields enclosed by brackets "[]" replaced with your own identifying 17 * information: Portions Copyright [yyyy] [name of copyright owner] 18 * 19 * CDDL HEADER END 20 */ 21 /* 22 * Copyright 2008 Sun Microsystems, Inc. All rights reserved. 23 * Use is subject to license terms. 24 */ 25 26 #include <sys/zfs_context.h> 27 #include <sys/spa.h> 28 #include <sys/dmu.h> 29 #include <sys/zap.h> 30 #include <sys/arc.h> 31 #include <sys/stat.h> 32 #include <sys/resource.h> 33 #include <sys/zil.h> 34 #include <sys/zil_impl.h> 35 #include <sys/dsl_dataset.h> 36 #include <sys/vdev.h> 37 #include <sys/dmu_tx.h> 38 39 /* 40 * The zfs intent log (ZIL) saves transaction records of system calls 41 * that change the file system in memory with enough information 42 * to be able to replay them. These are stored in memory until 43 * either the DMU transaction group (txg) commits them to the stable pool 44 * and they can be discarded, or they are flushed to the stable log 45 * (also in the pool) due to a fsync, O_DSYNC or other synchronous 46 * requirement. In the event of a panic or power fail then those log 47 * records (transactions) are replayed. 48 * 49 * There is one ZIL per file system. Its on-disk (pool) format consists 50 * of 3 parts: 51 * 52 * - ZIL header 53 * - ZIL blocks 54 * - ZIL records 55 * 56 * A log record holds a system call transaction. Log blocks can 57 * hold many log records and the blocks are chained together. 58 * Each ZIL block contains a block pointer (blkptr_t) to the next 59 * ZIL block in the chain. The ZIL header points to the first 60 * block in the chain. Note there is not a fixed place in the pool 61 * to hold blocks. They are dynamically allocated and freed as 62 * needed from the blocks available. Figure X shows the ZIL structure: 63 */ 64 65 /* 66 * This global ZIL switch affects all pools 67 */ 68 int zil_disable = 0; /* disable intent logging */ 69 70 /* 71 * Tunable parameter for debugging or performance analysis. Setting 72 * zfs_nocacheflush will cause corruption on power loss if a volatile 73 * out-of-order write cache is enabled. 74 */ 75 boolean_t zfs_nocacheflush = B_FALSE; 76 77 static kmem_cache_t *zil_lwb_cache; 78 79 static int 80 zil_dva_compare(const void *x1, const void *x2) 81 { 82 const dva_t *dva1 = x1; 83 const dva_t *dva2 = x2; 84 85 if (DVA_GET_VDEV(dva1) < DVA_GET_VDEV(dva2)) 86 return (-1); 87 if (DVA_GET_VDEV(dva1) > DVA_GET_VDEV(dva2)) 88 return (1); 89 90 if (DVA_GET_OFFSET(dva1) < DVA_GET_OFFSET(dva2)) 91 return (-1); 92 if (DVA_GET_OFFSET(dva1) > DVA_GET_OFFSET(dva2)) 93 return (1); 94 95 return (0); 96 } 97 98 static void 99 zil_dva_tree_init(avl_tree_t *t) 100 { 101 avl_create(t, zil_dva_compare, sizeof (zil_dva_node_t), 102 offsetof(zil_dva_node_t, zn_node)); 103 } 104 105 static void 106 zil_dva_tree_fini(avl_tree_t *t) 107 { 108 zil_dva_node_t *zn; 109 void *cookie = NULL; 110 111 while ((zn = avl_destroy_nodes(t, &cookie)) != NULL) 112 kmem_free(zn, sizeof (zil_dva_node_t)); 113 114 avl_destroy(t); 115 } 116 117 static int 118 zil_dva_tree_add(avl_tree_t *t, dva_t *dva) 119 { 120 zil_dva_node_t *zn; 121 avl_index_t where; 122 123 if (avl_find(t, dva, &where) != NULL) 124 return (EEXIST); 125 126 zn = kmem_alloc(sizeof (zil_dva_node_t), KM_SLEEP); 127 zn->zn_dva = *dva; 128 avl_insert(t, zn, where); 129 130 return (0); 131 } 132 133 static zil_header_t * 134 zil_header_in_syncing_context(zilog_t *zilog) 135 { 136 return ((zil_header_t *)zilog->zl_header); 137 } 138 139 static void 140 zil_init_log_chain(zilog_t *zilog, blkptr_t *bp) 141 { 142 zio_cksum_t *zc = &bp->blk_cksum; 143 144 zc->zc_word[ZIL_ZC_GUID_0] = spa_get_random(-1ULL); 145 zc->zc_word[ZIL_ZC_GUID_1] = spa_get_random(-1ULL); 146 zc->zc_word[ZIL_ZC_OBJSET] = dmu_objset_id(zilog->zl_os); 147 zc->zc_word[ZIL_ZC_SEQ] = 1ULL; 148 } 149 150 /* 151 * Read a log block, make sure it's valid, and byteswap it if necessary. 152 */ 153 static int 154 zil_read_log_block(zilog_t *zilog, const blkptr_t *bp, arc_buf_t **abufpp) 155 { 156 blkptr_t blk = *bp; 157 zbookmark_t zb; 158 uint32_t aflags = ARC_WAIT; 159 int error; 160 161 zb.zb_objset = bp->blk_cksum.zc_word[ZIL_ZC_OBJSET]; 162 zb.zb_object = 0; 163 zb.zb_level = -1; 164 zb.zb_blkid = bp->blk_cksum.zc_word[ZIL_ZC_SEQ]; 165 166 *abufpp = NULL; 167 168 /* 169 * We shouldn't be doing any scrubbing while we're doing log 170 * replay, it's OK to not lock. 171 */ 172 error = arc_read_nolock(NULL, zilog->zl_spa, &blk, 173 arc_getbuf_func, abufpp, ZIO_PRIORITY_SYNC_READ, ZIO_FLAG_CANFAIL | 174 ZIO_FLAG_SPECULATIVE | ZIO_FLAG_SCRUB, &aflags, &zb); 175 176 if (error == 0) { 177 char *data = (*abufpp)->b_data; 178 uint64_t blksz = BP_GET_LSIZE(bp); 179 zil_trailer_t *ztp = (zil_trailer_t *)(data + blksz) - 1; 180 zio_cksum_t cksum = bp->blk_cksum; 181 182 /* 183 * Validate the checksummed log block. 184 * 185 * Sequence numbers should be... sequential. The checksum 186 * verifier for the next block should be bp's checksum plus 1. 187 * 188 * Also check the log chain linkage and size used. 189 */ 190 cksum.zc_word[ZIL_ZC_SEQ]++; 191 192 if (bcmp(&cksum, &ztp->zit_next_blk.blk_cksum, 193 sizeof (cksum)) || BP_IS_HOLE(&ztp->zit_next_blk) || 194 (ztp->zit_nused > (blksz - sizeof (zil_trailer_t)))) { 195 error = ECKSUM; 196 } 197 198 if (error) { 199 VERIFY(arc_buf_remove_ref(*abufpp, abufpp) == 1); 200 *abufpp = NULL; 201 } 202 } 203 204 dprintf("error %d on %llu:%llu\n", error, zb.zb_objset, zb.zb_blkid); 205 206 return (error); 207 } 208 209 /* 210 * Parse the intent log, and call parse_func for each valid record within. 211 * Return the highest sequence number. 212 */ 213 uint64_t 214 zil_parse(zilog_t *zilog, zil_parse_blk_func_t *parse_blk_func, 215 zil_parse_lr_func_t *parse_lr_func, void *arg, uint64_t txg) 216 { 217 const zil_header_t *zh = zilog->zl_header; 218 uint64_t claim_seq = zh->zh_claim_seq; 219 uint64_t seq = 0; 220 uint64_t max_seq = 0; 221 blkptr_t blk = zh->zh_log; 222 arc_buf_t *abuf; 223 char *lrbuf, *lrp; 224 zil_trailer_t *ztp; 225 int reclen, error; 226 227 if (BP_IS_HOLE(&blk)) 228 return (max_seq); 229 230 /* 231 * Starting at the block pointed to by zh_log we read the log chain. 232 * For each block in the chain we strongly check that block to 233 * ensure its validity. We stop when an invalid block is found. 234 * For each block pointer in the chain we call parse_blk_func(). 235 * For each record in each valid block we call parse_lr_func(). 236 * If the log has been claimed, stop if we encounter a sequence 237 * number greater than the highest claimed sequence number. 238 */ 239 zil_dva_tree_init(&zilog->zl_dva_tree); 240 for (;;) { 241 seq = blk.blk_cksum.zc_word[ZIL_ZC_SEQ]; 242 243 if (claim_seq != 0 && seq > claim_seq) 244 break; 245 246 ASSERT(max_seq < seq); 247 max_seq = seq; 248 249 error = zil_read_log_block(zilog, &blk, &abuf); 250 251 if (parse_blk_func != NULL) 252 parse_blk_func(zilog, &blk, arg, txg); 253 254 if (error) 255 break; 256 257 lrbuf = abuf->b_data; 258 ztp = (zil_trailer_t *)(lrbuf + BP_GET_LSIZE(&blk)) - 1; 259 blk = ztp->zit_next_blk; 260 261 if (parse_lr_func == NULL) { 262 VERIFY(arc_buf_remove_ref(abuf, &abuf) == 1); 263 continue; 264 } 265 266 for (lrp = lrbuf; lrp < lrbuf + ztp->zit_nused; lrp += reclen) { 267 lr_t *lr = (lr_t *)lrp; 268 reclen = lr->lrc_reclen; 269 ASSERT3U(reclen, >=, sizeof (lr_t)); 270 parse_lr_func(zilog, lr, arg, txg); 271 } 272 VERIFY(arc_buf_remove_ref(abuf, &abuf) == 1); 273 } 274 zil_dva_tree_fini(&zilog->zl_dva_tree); 275 276 return (max_seq); 277 } 278 279 /* ARGSUSED */ 280 static void 281 zil_claim_log_block(zilog_t *zilog, blkptr_t *bp, void *tx, uint64_t first_txg) 282 { 283 spa_t *spa = zilog->zl_spa; 284 int err; 285 286 /* 287 * Claim log block if not already committed and not already claimed. 288 */ 289 if (bp->blk_birth >= first_txg && 290 zil_dva_tree_add(&zilog->zl_dva_tree, BP_IDENTITY(bp)) == 0) { 291 err = zio_wait(zio_claim(NULL, spa, first_txg, bp, NULL, NULL, 292 ZIO_FLAG_MUSTSUCCEED)); 293 ASSERT(err == 0); 294 } 295 } 296 297 static void 298 zil_claim_log_record(zilog_t *zilog, lr_t *lrc, void *tx, uint64_t first_txg) 299 { 300 if (lrc->lrc_txtype == TX_WRITE) { 301 lr_write_t *lr = (lr_write_t *)lrc; 302 zil_claim_log_block(zilog, &lr->lr_blkptr, tx, first_txg); 303 } 304 } 305 306 /* ARGSUSED */ 307 static void 308 zil_free_log_block(zilog_t *zilog, blkptr_t *bp, void *tx, uint64_t claim_txg) 309 { 310 zio_free_blk(zilog->zl_spa, bp, dmu_tx_get_txg(tx)); 311 } 312 313 static void 314 zil_free_log_record(zilog_t *zilog, lr_t *lrc, void *tx, uint64_t claim_txg) 315 { 316 /* 317 * If we previously claimed it, we need to free it. 318 */ 319 if (claim_txg != 0 && lrc->lrc_txtype == TX_WRITE) { 320 lr_write_t *lr = (lr_write_t *)lrc; 321 blkptr_t *bp = &lr->lr_blkptr; 322 if (bp->blk_birth >= claim_txg && 323 !zil_dva_tree_add(&zilog->zl_dva_tree, BP_IDENTITY(bp))) { 324 (void) arc_free(NULL, zilog->zl_spa, 325 dmu_tx_get_txg(tx), bp, NULL, NULL, ARC_WAIT); 326 } 327 } 328 } 329 330 /* 331 * Create an on-disk intent log. 332 */ 333 static void 334 zil_create(zilog_t *zilog) 335 { 336 const zil_header_t *zh = zilog->zl_header; 337 lwb_t *lwb; 338 uint64_t txg = 0; 339 dmu_tx_t *tx = NULL; 340 blkptr_t blk; 341 int error = 0; 342 343 /* 344 * Wait for any previous destroy to complete. 345 */ 346 txg_wait_synced(zilog->zl_dmu_pool, zilog->zl_destroy_txg); 347 348 ASSERT(zh->zh_claim_txg == 0); 349 ASSERT(zh->zh_replay_seq == 0); 350 351 blk = zh->zh_log; 352 353 /* 354 * If we don't already have an initial log block or we have one 355 * but it's the wrong endianness then allocate one. 356 */ 357 if (BP_IS_HOLE(&blk) || BP_SHOULD_BYTESWAP(&blk)) { 358 tx = dmu_tx_create(zilog->zl_os); 359 (void) dmu_tx_assign(tx, TXG_WAIT); 360 dsl_dataset_dirty(dmu_objset_ds(zilog->zl_os), tx); 361 txg = dmu_tx_get_txg(tx); 362 363 if (!BP_IS_HOLE(&blk)) { 364 zio_free_blk(zilog->zl_spa, &blk, txg); 365 BP_ZERO(&blk); 366 } 367 368 error = zio_alloc_blk(zilog->zl_spa, ZIL_MIN_BLKSZ, &blk, 369 NULL, txg); 370 371 if (error == 0) 372 zil_init_log_chain(zilog, &blk); 373 } 374 375 /* 376 * Allocate a log write buffer (lwb) for the first log block. 377 */ 378 if (error == 0) { 379 lwb = kmem_cache_alloc(zil_lwb_cache, KM_SLEEP); 380 lwb->lwb_zilog = zilog; 381 lwb->lwb_blk = blk; 382 lwb->lwb_nused = 0; 383 lwb->lwb_sz = BP_GET_LSIZE(&lwb->lwb_blk); 384 lwb->lwb_buf = zio_buf_alloc(lwb->lwb_sz); 385 lwb->lwb_max_txg = txg; 386 lwb->lwb_zio = NULL; 387 388 mutex_enter(&zilog->zl_lock); 389 list_insert_tail(&zilog->zl_lwb_list, lwb); 390 mutex_exit(&zilog->zl_lock); 391 } 392 393 /* 394 * If we just allocated the first log block, commit our transaction 395 * and wait for zil_sync() to stuff the block poiner into zh_log. 396 * (zh is part of the MOS, so we cannot modify it in open context.) 397 */ 398 if (tx != NULL) { 399 dmu_tx_commit(tx); 400 txg_wait_synced(zilog->zl_dmu_pool, txg); 401 } 402 403 ASSERT(bcmp(&blk, &zh->zh_log, sizeof (blk)) == 0); 404 } 405 406 /* 407 * In one tx, free all log blocks and clear the log header. 408 * If keep_first is set, then we're replaying a log with no content. 409 * We want to keep the first block, however, so that the first 410 * synchronous transaction doesn't require a txg_wait_synced() 411 * in zil_create(). We don't need to txg_wait_synced() here either 412 * when keep_first is set, because both zil_create() and zil_destroy() 413 * will wait for any in-progress destroys to complete. 414 */ 415 void 416 zil_destroy(zilog_t *zilog, boolean_t keep_first) 417 { 418 const zil_header_t *zh = zilog->zl_header; 419 lwb_t *lwb; 420 dmu_tx_t *tx; 421 uint64_t txg; 422 423 /* 424 * Wait for any previous destroy to complete. 425 */ 426 txg_wait_synced(zilog->zl_dmu_pool, zilog->zl_destroy_txg); 427 428 if (BP_IS_HOLE(&zh->zh_log)) 429 return; 430 431 tx = dmu_tx_create(zilog->zl_os); 432 (void) dmu_tx_assign(tx, TXG_WAIT); 433 dsl_dataset_dirty(dmu_objset_ds(zilog->zl_os), tx); 434 txg = dmu_tx_get_txg(tx); 435 436 mutex_enter(&zilog->zl_lock); 437 438 /* 439 * It is possible for the ZIL to get the previously mounted zilog 440 * structure of the same dataset if quickly remounted and the dbuf 441 * eviction has not completed. In this case we can see a non 442 * empty lwb list and keep_first will be set. We fix this by 443 * clearing the keep_first. This will be slower but it's very rare. 444 */ 445 if (!list_is_empty(&zilog->zl_lwb_list) && keep_first) 446 keep_first = B_FALSE; 447 448 ASSERT3U(zilog->zl_destroy_txg, <, txg); 449 zilog->zl_destroy_txg = txg; 450 zilog->zl_keep_first = keep_first; 451 452 if (!list_is_empty(&zilog->zl_lwb_list)) { 453 ASSERT(zh->zh_claim_txg == 0); 454 ASSERT(!keep_first); 455 while ((lwb = list_head(&zilog->zl_lwb_list)) != NULL) { 456 list_remove(&zilog->zl_lwb_list, lwb); 457 if (lwb->lwb_buf != NULL) 458 zio_buf_free(lwb->lwb_buf, lwb->lwb_sz); 459 zio_free_blk(zilog->zl_spa, &lwb->lwb_blk, txg); 460 kmem_cache_free(zil_lwb_cache, lwb); 461 } 462 } else { 463 if (!keep_first) { 464 (void) zil_parse(zilog, zil_free_log_block, 465 zil_free_log_record, tx, zh->zh_claim_txg); 466 } 467 } 468 mutex_exit(&zilog->zl_lock); 469 470 dmu_tx_commit(tx); 471 } 472 473 /* 474 * zil_rollback_destroy() is only called by the rollback code. 475 * We already have a syncing tx. Rollback has exclusive access to the 476 * dataset, so we don't have to worry about concurrent zil access. 477 * The actual freeing of any log blocks occurs in zil_sync() later in 478 * this txg syncing phase. 479 */ 480 void 481 zil_rollback_destroy(zilog_t *zilog, dmu_tx_t *tx) 482 { 483 const zil_header_t *zh = zilog->zl_header; 484 uint64_t txg; 485 486 if (BP_IS_HOLE(&zh->zh_log)) 487 return; 488 489 txg = dmu_tx_get_txg(tx); 490 ASSERT3U(zilog->zl_destroy_txg, <, txg); 491 zilog->zl_destroy_txg = txg; 492 zilog->zl_keep_first = B_FALSE; 493 494 /* 495 * Ensure there's no outstanding ZIL IO. No lwbs or just the 496 * unused one that allocated in advance is ok. 497 */ 498 ASSERT(zilog->zl_lwb_list.list_head.list_next == 499 zilog->zl_lwb_list.list_head.list_prev); 500 (void) zil_parse(zilog, zil_free_log_block, zil_free_log_record, 501 tx, zh->zh_claim_txg); 502 } 503 504 int 505 zil_claim(char *osname, void *txarg) 506 { 507 dmu_tx_t *tx = txarg; 508 uint64_t first_txg = dmu_tx_get_txg(tx); 509 zilog_t *zilog; 510 zil_header_t *zh; 511 objset_t *os; 512 int error; 513 514 error = dmu_objset_open(osname, DMU_OST_ANY, DS_MODE_USER, &os); 515 if (error) { 516 cmn_err(CE_WARN, "can't open objset for %s", osname); 517 return (0); 518 } 519 520 zilog = dmu_objset_zil(os); 521 zh = zil_header_in_syncing_context(zilog); 522 523 /* 524 * Claim all log blocks if we haven't already done so, and remember 525 * the highest claimed sequence number. This ensures that if we can 526 * read only part of the log now (e.g. due to a missing device), 527 * but we can read the entire log later, we will not try to replay 528 * or destroy beyond the last block we successfully claimed. 529 */ 530 ASSERT3U(zh->zh_claim_txg, <=, first_txg); 531 if (zh->zh_claim_txg == 0 && !BP_IS_HOLE(&zh->zh_log)) { 532 zh->zh_claim_txg = first_txg; 533 zh->zh_claim_seq = zil_parse(zilog, zil_claim_log_block, 534 zil_claim_log_record, tx, first_txg); 535 dsl_dataset_dirty(dmu_objset_ds(os), tx); 536 } 537 538 ASSERT3U(first_txg, ==, (spa_last_synced_txg(zilog->zl_spa) + 1)); 539 dmu_objset_close(os); 540 return (0); 541 } 542 543 /* 544 * Check the log by walking the log chain. 545 * Checksum errors are ok as they indicate the end of the chain. 546 * Any other error (no device or read failure) returns an error. 547 */ 548 /* ARGSUSED */ 549 int 550 zil_check_log_chain(char *osname, void *txarg) 551 { 552 zilog_t *zilog; 553 zil_header_t *zh; 554 blkptr_t blk; 555 arc_buf_t *abuf; 556 objset_t *os; 557 char *lrbuf; 558 zil_trailer_t *ztp; 559 int error; 560 561 error = dmu_objset_open(osname, DMU_OST_ANY, DS_MODE_USER, &os); 562 if (error) { 563 cmn_err(CE_WARN, "can't open objset for %s", osname); 564 return (0); 565 } 566 567 zilog = dmu_objset_zil(os); 568 zh = zil_header_in_syncing_context(zilog); 569 blk = zh->zh_log; 570 if (BP_IS_HOLE(&blk)) { 571 dmu_objset_close(os); 572 return (0); /* no chain */ 573 } 574 575 for (;;) { 576 error = zil_read_log_block(zilog, &blk, &abuf); 577 if (error) 578 break; 579 lrbuf = abuf->b_data; 580 ztp = (zil_trailer_t *)(lrbuf + BP_GET_LSIZE(&blk)) - 1; 581 blk = ztp->zit_next_blk; 582 VERIFY(arc_buf_remove_ref(abuf, &abuf) == 1); 583 } 584 dmu_objset_close(os); 585 if (error == ECKSUM) 586 return (0); /* normal end of chain */ 587 return (error); 588 } 589 590 /* 591 * Clear a log chain 592 */ 593 /* ARGSUSED */ 594 int 595 zil_clear_log_chain(char *osname, void *txarg) 596 { 597 zilog_t *zilog; 598 zil_header_t *zh; 599 objset_t *os; 600 dmu_tx_t *tx; 601 int error; 602 603 error = dmu_objset_open(osname, DMU_OST_ANY, DS_MODE_USER, &os); 604 if (error) { 605 cmn_err(CE_WARN, "can't open objset for %s", osname); 606 return (0); 607 } 608 609 zilog = dmu_objset_zil(os); 610 tx = dmu_tx_create(zilog->zl_os); 611 (void) dmu_tx_assign(tx, TXG_WAIT); 612 zh = zil_header_in_syncing_context(zilog); 613 BP_ZERO(&zh->zh_log); 614 dsl_dataset_dirty(dmu_objset_ds(os), tx); 615 dmu_tx_commit(tx); 616 dmu_objset_close(os); 617 return (0); 618 } 619 620 static int 621 zil_vdev_compare(const void *x1, const void *x2) 622 { 623 uint64_t v1 = ((zil_vdev_node_t *)x1)->zv_vdev; 624 uint64_t v2 = ((zil_vdev_node_t *)x2)->zv_vdev; 625 626 if (v1 < v2) 627 return (-1); 628 if (v1 > v2) 629 return (1); 630 631 return (0); 632 } 633 634 void 635 zil_add_block(zilog_t *zilog, blkptr_t *bp) 636 { 637 avl_tree_t *t = &zilog->zl_vdev_tree; 638 avl_index_t where; 639 zil_vdev_node_t *zv, zvsearch; 640 int ndvas = BP_GET_NDVAS(bp); 641 int i; 642 643 if (zfs_nocacheflush) 644 return; 645 646 ASSERT(zilog->zl_writer); 647 648 /* 649 * Even though we're zl_writer, we still need a lock because the 650 * zl_get_data() callbacks may have dmu_sync() done callbacks 651 * that will run concurrently. 652 */ 653 mutex_enter(&zilog->zl_vdev_lock); 654 for (i = 0; i < ndvas; i++) { 655 zvsearch.zv_vdev = DVA_GET_VDEV(&bp->blk_dva[i]); 656 if (avl_find(t, &zvsearch, &where) == NULL) { 657 zv = kmem_alloc(sizeof (*zv), KM_SLEEP); 658 zv->zv_vdev = zvsearch.zv_vdev; 659 avl_insert(t, zv, where); 660 } 661 } 662 mutex_exit(&zilog->zl_vdev_lock); 663 } 664 665 void 666 zil_flush_vdevs(zilog_t *zilog) 667 { 668 spa_t *spa = zilog->zl_spa; 669 avl_tree_t *t = &zilog->zl_vdev_tree; 670 void *cookie = NULL; 671 zil_vdev_node_t *zv; 672 zio_t *zio; 673 674 ASSERT(zilog->zl_writer); 675 676 /* 677 * We don't need zl_vdev_lock here because we're the zl_writer, 678 * and all zl_get_data() callbacks are done. 679 */ 680 if (avl_numnodes(t) == 0) 681 return; 682 683 spa_config_enter(spa, SCL_STATE, FTAG, RW_READER); 684 685 zio = zio_root(spa, NULL, NULL, ZIO_FLAG_CANFAIL); 686 687 while ((zv = avl_destroy_nodes(t, &cookie)) != NULL) { 688 vdev_t *vd = vdev_lookup_top(spa, zv->zv_vdev); 689 if (vd != NULL) 690 zio_flush(zio, vd); 691 kmem_free(zv, sizeof (*zv)); 692 } 693 694 /* 695 * Wait for all the flushes to complete. Not all devices actually 696 * support the DKIOCFLUSHWRITECACHE ioctl, so it's OK if it fails. 697 */ 698 (void) zio_wait(zio); 699 700 spa_config_exit(spa, SCL_STATE, FTAG); 701 } 702 703 /* 704 * Function called when a log block write completes 705 */ 706 static void 707 zil_lwb_write_done(zio_t *zio) 708 { 709 lwb_t *lwb = zio->io_private; 710 zilog_t *zilog = lwb->lwb_zilog; 711 712 ASSERT(BP_GET_COMPRESS(zio->io_bp) == ZIO_COMPRESS_OFF); 713 ASSERT(BP_GET_CHECKSUM(zio->io_bp) == ZIO_CHECKSUM_ZILOG); 714 ASSERT(BP_GET_TYPE(zio->io_bp) == DMU_OT_INTENT_LOG); 715 ASSERT(BP_GET_LEVEL(zio->io_bp) == 0); 716 ASSERT(BP_GET_BYTEORDER(zio->io_bp) == ZFS_HOST_BYTEORDER); 717 ASSERT(!BP_IS_GANG(zio->io_bp)); 718 ASSERT(!BP_IS_HOLE(zio->io_bp)); 719 ASSERT(zio->io_bp->blk_fill == 0); 720 721 /* 722 * Now that we've written this log block, we have a stable pointer 723 * to the next block in the chain, so it's OK to let the txg in 724 * which we allocated the next block sync. 725 */ 726 txg_rele_to_sync(&lwb->lwb_txgh); 727 728 zio_buf_free(lwb->lwb_buf, lwb->lwb_sz); 729 mutex_enter(&zilog->zl_lock); 730 lwb->lwb_buf = NULL; 731 if (zio->io_error) 732 zilog->zl_log_error = B_TRUE; 733 mutex_exit(&zilog->zl_lock); 734 } 735 736 /* 737 * Initialize the io for a log block. 738 */ 739 static void 740 zil_lwb_write_init(zilog_t *zilog, lwb_t *lwb) 741 { 742 zbookmark_t zb; 743 744 zb.zb_objset = lwb->lwb_blk.blk_cksum.zc_word[ZIL_ZC_OBJSET]; 745 zb.zb_object = 0; 746 zb.zb_level = -1; 747 zb.zb_blkid = lwb->lwb_blk.blk_cksum.zc_word[ZIL_ZC_SEQ]; 748 749 if (zilog->zl_root_zio == NULL) { 750 zilog->zl_root_zio = zio_root(zilog->zl_spa, NULL, NULL, 751 ZIO_FLAG_CANFAIL); 752 } 753 if (lwb->lwb_zio == NULL) { 754 lwb->lwb_zio = zio_rewrite(zilog->zl_root_zio, zilog->zl_spa, 755 0, &lwb->lwb_blk, lwb->lwb_buf, 756 lwb->lwb_sz, zil_lwb_write_done, lwb, 757 ZIO_PRIORITY_LOG_WRITE, ZIO_FLAG_CANFAIL, &zb); 758 } 759 } 760 761 /* 762 * Start a log block write and advance to the next log block. 763 * Calls are serialized. 764 */ 765 static lwb_t * 766 zil_lwb_write_start(zilog_t *zilog, lwb_t *lwb) 767 { 768 lwb_t *nlwb; 769 zil_trailer_t *ztp = (zil_trailer_t *)(lwb->lwb_buf + lwb->lwb_sz) - 1; 770 spa_t *spa = zilog->zl_spa; 771 blkptr_t *bp = &ztp->zit_next_blk; 772 uint64_t txg; 773 uint64_t zil_blksz; 774 int error; 775 776 ASSERT(lwb->lwb_nused <= ZIL_BLK_DATA_SZ(lwb)); 777 778 /* 779 * Allocate the next block and save its address in this block 780 * before writing it in order to establish the log chain. 781 * Note that if the allocation of nlwb synced before we wrote 782 * the block that points at it (lwb), we'd leak it if we crashed. 783 * Therefore, we don't do txg_rele_to_sync() until zil_lwb_write_done(). 784 */ 785 txg = txg_hold_open(zilog->zl_dmu_pool, &lwb->lwb_txgh); 786 txg_rele_to_quiesce(&lwb->lwb_txgh); 787 788 /* 789 * Pick a ZIL blocksize. We request a size that is the 790 * maximum of the previous used size, the current used size and 791 * the amount waiting in the queue. 792 */ 793 zil_blksz = MAX(zilog->zl_prev_used, 794 zilog->zl_cur_used + sizeof (*ztp)); 795 zil_blksz = MAX(zil_blksz, zilog->zl_itx_list_sz + sizeof (*ztp)); 796 zil_blksz = P2ROUNDUP_TYPED(zil_blksz, ZIL_MIN_BLKSZ, uint64_t); 797 if (zil_blksz > ZIL_MAX_BLKSZ) 798 zil_blksz = ZIL_MAX_BLKSZ; 799 800 BP_ZERO(bp); 801 /* pass the old blkptr in order to spread log blocks across devs */ 802 error = zio_alloc_blk(spa, zil_blksz, bp, &lwb->lwb_blk, txg); 803 if (error) { 804 dmu_tx_t *tx = dmu_tx_create_assigned(zilog->zl_dmu_pool, txg); 805 806 /* 807 * We dirty the dataset to ensure that zil_sync() will 808 * be called to remove this lwb from our zl_lwb_list. 809 * Failing to do so, may leave an lwb with a NULL lwb_buf 810 * hanging around on the zl_lwb_list. 811 */ 812 dsl_dataset_dirty(dmu_objset_ds(zilog->zl_os), tx); 813 dmu_tx_commit(tx); 814 815 /* 816 * Since we've just experienced an allocation failure so we 817 * terminate the current lwb and send it on its way. 818 */ 819 ztp->zit_pad = 0; 820 ztp->zit_nused = lwb->lwb_nused; 821 ztp->zit_bt.zbt_cksum = lwb->lwb_blk.blk_cksum; 822 zio_nowait(lwb->lwb_zio); 823 824 /* 825 * By returning NULL the caller will call tx_wait_synced() 826 */ 827 return (NULL); 828 } 829 830 ASSERT3U(bp->blk_birth, ==, txg); 831 ztp->zit_pad = 0; 832 ztp->zit_nused = lwb->lwb_nused; 833 ztp->zit_bt.zbt_cksum = lwb->lwb_blk.blk_cksum; 834 bp->blk_cksum = lwb->lwb_blk.blk_cksum; 835 bp->blk_cksum.zc_word[ZIL_ZC_SEQ]++; 836 837 /* 838 * Allocate a new log write buffer (lwb). 839 */ 840 nlwb = kmem_cache_alloc(zil_lwb_cache, KM_SLEEP); 841 842 nlwb->lwb_zilog = zilog; 843 nlwb->lwb_blk = *bp; 844 nlwb->lwb_nused = 0; 845 nlwb->lwb_sz = BP_GET_LSIZE(&nlwb->lwb_blk); 846 nlwb->lwb_buf = zio_buf_alloc(nlwb->lwb_sz); 847 nlwb->lwb_max_txg = txg; 848 nlwb->lwb_zio = NULL; 849 850 /* 851 * Put new lwb at the end of the log chain 852 */ 853 mutex_enter(&zilog->zl_lock); 854 list_insert_tail(&zilog->zl_lwb_list, nlwb); 855 mutex_exit(&zilog->zl_lock); 856 857 /* Record the block for later vdev flushing */ 858 zil_add_block(zilog, &lwb->lwb_blk); 859 860 /* 861 * kick off the write for the old log block 862 */ 863 dprintf_bp(&lwb->lwb_blk, "lwb %p txg %llu: ", lwb, txg); 864 ASSERT(lwb->lwb_zio); 865 zio_nowait(lwb->lwb_zio); 866 867 return (nlwb); 868 } 869 870 static lwb_t * 871 zil_lwb_commit(zilog_t *zilog, itx_t *itx, lwb_t *lwb) 872 { 873 lr_t *lrc = &itx->itx_lr; /* common log record */ 874 lr_write_t *lr = (lr_write_t *)lrc; 875 uint64_t txg = lrc->lrc_txg; 876 uint64_t reclen = lrc->lrc_reclen; 877 uint64_t dlen; 878 879 if (lwb == NULL) 880 return (NULL); 881 ASSERT(lwb->lwb_buf != NULL); 882 883 if (lrc->lrc_txtype == TX_WRITE && itx->itx_wr_state == WR_NEED_COPY) 884 dlen = P2ROUNDUP_TYPED( 885 lr->lr_length, sizeof (uint64_t), uint64_t); 886 else 887 dlen = 0; 888 889 zilog->zl_cur_used += (reclen + dlen); 890 891 zil_lwb_write_init(zilog, lwb); 892 893 /* 894 * If this record won't fit in the current log block, start a new one. 895 */ 896 if (lwb->lwb_nused + reclen + dlen > ZIL_BLK_DATA_SZ(lwb)) { 897 lwb = zil_lwb_write_start(zilog, lwb); 898 if (lwb == NULL) 899 return (NULL); 900 zil_lwb_write_init(zilog, lwb); 901 ASSERT(lwb->lwb_nused == 0); 902 if (reclen + dlen > ZIL_BLK_DATA_SZ(lwb)) { 903 txg_wait_synced(zilog->zl_dmu_pool, txg); 904 return (lwb); 905 } 906 } 907 908 /* 909 * Update the lrc_seq, to be log record sequence number. See zil.h 910 * Then copy the record to the log buffer. 911 */ 912 lrc->lrc_seq = ++zilog->zl_lr_seq; /* we are single threaded */ 913 bcopy(lrc, lwb->lwb_buf + lwb->lwb_nused, reclen); 914 915 /* 916 * If it's a write, fetch the data or get its blkptr as appropriate. 917 */ 918 if (lrc->lrc_txtype == TX_WRITE) { 919 if (txg > spa_freeze_txg(zilog->zl_spa)) 920 txg_wait_synced(zilog->zl_dmu_pool, txg); 921 if (itx->itx_wr_state != WR_COPIED) { 922 char *dbuf; 923 int error; 924 925 /* alignment is guaranteed */ 926 lr = (lr_write_t *)(lwb->lwb_buf + lwb->lwb_nused); 927 if (dlen) { 928 ASSERT(itx->itx_wr_state == WR_NEED_COPY); 929 dbuf = lwb->lwb_buf + lwb->lwb_nused + reclen; 930 lr->lr_common.lrc_reclen += dlen; 931 } else { 932 ASSERT(itx->itx_wr_state == WR_INDIRECT); 933 dbuf = NULL; 934 } 935 error = zilog->zl_get_data( 936 itx->itx_private, lr, dbuf, lwb->lwb_zio); 937 if (error) { 938 ASSERT(error == ENOENT || error == EEXIST || 939 error == EALREADY); 940 return (lwb); 941 } 942 } 943 } 944 945 lwb->lwb_nused += reclen + dlen; 946 lwb->lwb_max_txg = MAX(lwb->lwb_max_txg, txg); 947 ASSERT3U(lwb->lwb_nused, <=, ZIL_BLK_DATA_SZ(lwb)); 948 ASSERT3U(P2PHASE(lwb->lwb_nused, sizeof (uint64_t)), ==, 0); 949 950 return (lwb); 951 } 952 953 itx_t * 954 zil_itx_create(uint64_t txtype, size_t lrsize) 955 { 956 itx_t *itx; 957 958 lrsize = P2ROUNDUP_TYPED(lrsize, sizeof (uint64_t), size_t); 959 960 itx = kmem_alloc(offsetof(itx_t, itx_lr) + lrsize, KM_SLEEP); 961 itx->itx_lr.lrc_txtype = txtype; 962 itx->itx_lr.lrc_reclen = lrsize; 963 itx->itx_sod = lrsize; /* if write & WR_NEED_COPY will be increased */ 964 itx->itx_lr.lrc_seq = 0; /* defensive */ 965 966 return (itx); 967 } 968 969 uint64_t 970 zil_itx_assign(zilog_t *zilog, itx_t *itx, dmu_tx_t *tx) 971 { 972 uint64_t seq; 973 974 ASSERT(itx->itx_lr.lrc_seq == 0); 975 976 mutex_enter(&zilog->zl_lock); 977 list_insert_tail(&zilog->zl_itx_list, itx); 978 zilog->zl_itx_list_sz += itx->itx_sod; 979 itx->itx_lr.lrc_txg = dmu_tx_get_txg(tx); 980 itx->itx_lr.lrc_seq = seq = ++zilog->zl_itx_seq; 981 mutex_exit(&zilog->zl_lock); 982 983 return (seq); 984 } 985 986 /* 987 * Free up all in-memory intent log transactions that have now been synced. 988 */ 989 static void 990 zil_itx_clean(zilog_t *zilog) 991 { 992 uint64_t synced_txg = spa_last_synced_txg(zilog->zl_spa); 993 uint64_t freeze_txg = spa_freeze_txg(zilog->zl_spa); 994 list_t clean_list; 995 itx_t *itx; 996 997 list_create(&clean_list, sizeof (itx_t), offsetof(itx_t, itx_node)); 998 999 mutex_enter(&zilog->zl_lock); 1000 /* wait for a log writer to finish walking list */ 1001 while (zilog->zl_writer) { 1002 cv_wait(&zilog->zl_cv_writer, &zilog->zl_lock); 1003 } 1004 1005 /* 1006 * Move the sync'd log transactions to a separate list so we can call 1007 * kmem_free without holding the zl_lock. 1008 * 1009 * There is no need to set zl_writer as we don't drop zl_lock here 1010 */ 1011 while ((itx = list_head(&zilog->zl_itx_list)) != NULL && 1012 itx->itx_lr.lrc_txg <= MIN(synced_txg, freeze_txg)) { 1013 list_remove(&zilog->zl_itx_list, itx); 1014 zilog->zl_itx_list_sz -= itx->itx_sod; 1015 list_insert_tail(&clean_list, itx); 1016 } 1017 cv_broadcast(&zilog->zl_cv_writer); 1018 mutex_exit(&zilog->zl_lock); 1019 1020 /* destroy sync'd log transactions */ 1021 while ((itx = list_head(&clean_list)) != NULL) { 1022 list_remove(&clean_list, itx); 1023 kmem_free(itx, offsetof(itx_t, itx_lr) 1024 + itx->itx_lr.lrc_reclen); 1025 } 1026 list_destroy(&clean_list); 1027 } 1028 1029 /* 1030 * If there are any in-memory intent log transactions which have now been 1031 * synced then start up a taskq to free them. 1032 */ 1033 void 1034 zil_clean(zilog_t *zilog) 1035 { 1036 itx_t *itx; 1037 1038 mutex_enter(&zilog->zl_lock); 1039 itx = list_head(&zilog->zl_itx_list); 1040 if ((itx != NULL) && 1041 (itx->itx_lr.lrc_txg <= spa_last_synced_txg(zilog->zl_spa))) { 1042 (void) taskq_dispatch(zilog->zl_clean_taskq, 1043 (void (*)(void *))zil_itx_clean, zilog, TQ_NOSLEEP); 1044 } 1045 mutex_exit(&zilog->zl_lock); 1046 } 1047 1048 static void 1049 zil_commit_writer(zilog_t *zilog, uint64_t seq, uint64_t foid) 1050 { 1051 uint64_t txg; 1052 uint64_t commit_seq = 0; 1053 itx_t *itx, *itx_next = (itx_t *)-1; 1054 lwb_t *lwb; 1055 spa_t *spa; 1056 1057 zilog->zl_writer = B_TRUE; 1058 ASSERT(zilog->zl_root_zio == NULL); 1059 spa = zilog->zl_spa; 1060 1061 if (zilog->zl_suspend) { 1062 lwb = NULL; 1063 } else { 1064 lwb = list_tail(&zilog->zl_lwb_list); 1065 if (lwb == NULL) { 1066 /* 1067 * Return if there's nothing to flush before we 1068 * dirty the fs by calling zil_create() 1069 */ 1070 if (list_is_empty(&zilog->zl_itx_list)) { 1071 zilog->zl_writer = B_FALSE; 1072 return; 1073 } 1074 mutex_exit(&zilog->zl_lock); 1075 zil_create(zilog); 1076 mutex_enter(&zilog->zl_lock); 1077 lwb = list_tail(&zilog->zl_lwb_list); 1078 } 1079 } 1080 1081 /* Loop through in-memory log transactions filling log blocks. */ 1082 DTRACE_PROBE1(zil__cw1, zilog_t *, zilog); 1083 for (;;) { 1084 /* 1085 * Find the next itx to push: 1086 * Push all transactions related to specified foid and all 1087 * other transactions except TX_WRITE, TX_TRUNCATE, 1088 * TX_SETATTR and TX_ACL for all other files. 1089 */ 1090 if (itx_next != (itx_t *)-1) 1091 itx = itx_next; 1092 else 1093 itx = list_head(&zilog->zl_itx_list); 1094 for (; itx != NULL; itx = list_next(&zilog->zl_itx_list, itx)) { 1095 if (foid == 0) /* push all foids? */ 1096 break; 1097 if (itx->itx_sync) /* push all O_[D]SYNC */ 1098 break; 1099 switch (itx->itx_lr.lrc_txtype) { 1100 case TX_SETATTR: 1101 case TX_WRITE: 1102 case TX_TRUNCATE: 1103 case TX_ACL: 1104 /* lr_foid is same offset for these records */ 1105 if (((lr_write_t *)&itx->itx_lr)->lr_foid 1106 != foid) { 1107 continue; /* skip this record */ 1108 } 1109 } 1110 break; 1111 } 1112 if (itx == NULL) 1113 break; 1114 1115 if ((itx->itx_lr.lrc_seq > seq) && 1116 ((lwb == NULL) || (lwb->lwb_nused == 0) || 1117 (lwb->lwb_nused + itx->itx_sod > ZIL_BLK_DATA_SZ(lwb)))) { 1118 break; 1119 } 1120 1121 /* 1122 * Save the next pointer. Even though we soon drop 1123 * zl_lock all threads that may change the list 1124 * (another writer or zil_itx_clean) can't do so until 1125 * they have zl_writer. 1126 */ 1127 itx_next = list_next(&zilog->zl_itx_list, itx); 1128 list_remove(&zilog->zl_itx_list, itx); 1129 zilog->zl_itx_list_sz -= itx->itx_sod; 1130 mutex_exit(&zilog->zl_lock); 1131 txg = itx->itx_lr.lrc_txg; 1132 ASSERT(txg); 1133 1134 if (txg > spa_last_synced_txg(spa) || 1135 txg > spa_freeze_txg(spa)) 1136 lwb = zil_lwb_commit(zilog, itx, lwb); 1137 kmem_free(itx, offsetof(itx_t, itx_lr) 1138 + itx->itx_lr.lrc_reclen); 1139 mutex_enter(&zilog->zl_lock); 1140 } 1141 DTRACE_PROBE1(zil__cw2, zilog_t *, zilog); 1142 /* determine commit sequence number */ 1143 itx = list_head(&zilog->zl_itx_list); 1144 if (itx) 1145 commit_seq = itx->itx_lr.lrc_seq; 1146 else 1147 commit_seq = zilog->zl_itx_seq; 1148 mutex_exit(&zilog->zl_lock); 1149 1150 /* write the last block out */ 1151 if (lwb != NULL && lwb->lwb_zio != NULL) 1152 lwb = zil_lwb_write_start(zilog, lwb); 1153 1154 zilog->zl_prev_used = zilog->zl_cur_used; 1155 zilog->zl_cur_used = 0; 1156 1157 /* 1158 * Wait if necessary for the log blocks to be on stable storage. 1159 */ 1160 if (zilog->zl_root_zio) { 1161 DTRACE_PROBE1(zil__cw3, zilog_t *, zilog); 1162 (void) zio_wait(zilog->zl_root_zio); 1163 zilog->zl_root_zio = NULL; 1164 DTRACE_PROBE1(zil__cw4, zilog_t *, zilog); 1165 zil_flush_vdevs(zilog); 1166 } 1167 1168 if (zilog->zl_log_error || lwb == NULL) { 1169 zilog->zl_log_error = 0; 1170 txg_wait_synced(zilog->zl_dmu_pool, 0); 1171 } 1172 1173 mutex_enter(&zilog->zl_lock); 1174 zilog->zl_writer = B_FALSE; 1175 1176 ASSERT3U(commit_seq, >=, zilog->zl_commit_seq); 1177 zilog->zl_commit_seq = commit_seq; 1178 } 1179 1180 /* 1181 * Push zfs transactions to stable storage up to the supplied sequence number. 1182 * If foid is 0 push out all transactions, otherwise push only those 1183 * for that file or might have been used to create that file. 1184 */ 1185 void 1186 zil_commit(zilog_t *zilog, uint64_t seq, uint64_t foid) 1187 { 1188 if (zilog == NULL || seq == 0) 1189 return; 1190 1191 mutex_enter(&zilog->zl_lock); 1192 1193 seq = MIN(seq, zilog->zl_itx_seq); /* cap seq at largest itx seq */ 1194 1195 while (zilog->zl_writer) { 1196 cv_wait(&zilog->zl_cv_writer, &zilog->zl_lock); 1197 if (seq < zilog->zl_commit_seq) { 1198 mutex_exit(&zilog->zl_lock); 1199 return; 1200 } 1201 } 1202 zil_commit_writer(zilog, seq, foid); /* drops zl_lock */ 1203 /* wake up others waiting on the commit */ 1204 cv_broadcast(&zilog->zl_cv_writer); 1205 mutex_exit(&zilog->zl_lock); 1206 } 1207 1208 /* 1209 * Called in syncing context to free committed log blocks and update log header. 1210 */ 1211 void 1212 zil_sync(zilog_t *zilog, dmu_tx_t *tx) 1213 { 1214 zil_header_t *zh = zil_header_in_syncing_context(zilog); 1215 uint64_t txg = dmu_tx_get_txg(tx); 1216 spa_t *spa = zilog->zl_spa; 1217 lwb_t *lwb; 1218 1219 mutex_enter(&zilog->zl_lock); 1220 1221 ASSERT(zilog->zl_stop_sync == 0); 1222 1223 zh->zh_replay_seq = zilog->zl_replay_seq[txg & TXG_MASK]; 1224 1225 if (zilog->zl_destroy_txg == txg) { 1226 blkptr_t blk = zh->zh_log; 1227 1228 ASSERT(list_head(&zilog->zl_lwb_list) == NULL); 1229 ASSERT(spa_sync_pass(spa) == 1); 1230 1231 bzero(zh, sizeof (zil_header_t)); 1232 bzero(zilog->zl_replay_seq, sizeof (zilog->zl_replay_seq)); 1233 1234 if (zilog->zl_keep_first) { 1235 /* 1236 * If this block was part of log chain that couldn't 1237 * be claimed because a device was missing during 1238 * zil_claim(), but that device later returns, 1239 * then this block could erroneously appear valid. 1240 * To guard against this, assign a new GUID to the new 1241 * log chain so it doesn't matter what blk points to. 1242 */ 1243 zil_init_log_chain(zilog, &blk); 1244 zh->zh_log = blk; 1245 } 1246 } 1247 1248 for (;;) { 1249 lwb = list_head(&zilog->zl_lwb_list); 1250 if (lwb == NULL) { 1251 mutex_exit(&zilog->zl_lock); 1252 return; 1253 } 1254 zh->zh_log = lwb->lwb_blk; 1255 if (lwb->lwb_buf != NULL || lwb->lwb_max_txg > txg) 1256 break; 1257 list_remove(&zilog->zl_lwb_list, lwb); 1258 zio_free_blk(spa, &lwb->lwb_blk, txg); 1259 kmem_cache_free(zil_lwb_cache, lwb); 1260 1261 /* 1262 * If we don't have anything left in the lwb list then 1263 * we've had an allocation failure and we need to zero 1264 * out the zil_header blkptr so that we don't end 1265 * up freeing the same block twice. 1266 */ 1267 if (list_head(&zilog->zl_lwb_list) == NULL) 1268 BP_ZERO(&zh->zh_log); 1269 } 1270 mutex_exit(&zilog->zl_lock); 1271 } 1272 1273 void 1274 zil_init(void) 1275 { 1276 zil_lwb_cache = kmem_cache_create("zil_lwb_cache", 1277 sizeof (struct lwb), 0, NULL, NULL, NULL, NULL, NULL, 0); 1278 } 1279 1280 void 1281 zil_fini(void) 1282 { 1283 kmem_cache_destroy(zil_lwb_cache); 1284 } 1285 1286 zilog_t * 1287 zil_alloc(objset_t *os, zil_header_t *zh_phys) 1288 { 1289 zilog_t *zilog; 1290 1291 zilog = kmem_zalloc(sizeof (zilog_t), KM_SLEEP); 1292 1293 zilog->zl_header = zh_phys; 1294 zilog->zl_os = os; 1295 zilog->zl_spa = dmu_objset_spa(os); 1296 zilog->zl_dmu_pool = dmu_objset_pool(os); 1297 zilog->zl_destroy_txg = TXG_INITIAL - 1; 1298 1299 mutex_init(&zilog->zl_lock, NULL, MUTEX_DEFAULT, NULL); 1300 1301 list_create(&zilog->zl_itx_list, sizeof (itx_t), 1302 offsetof(itx_t, itx_node)); 1303 1304 list_create(&zilog->zl_lwb_list, sizeof (lwb_t), 1305 offsetof(lwb_t, lwb_node)); 1306 1307 mutex_init(&zilog->zl_vdev_lock, NULL, MUTEX_DEFAULT, NULL); 1308 1309 avl_create(&zilog->zl_vdev_tree, zil_vdev_compare, 1310 sizeof (zil_vdev_node_t), offsetof(zil_vdev_node_t, zv_node)); 1311 1312 cv_init(&zilog->zl_cv_writer, NULL, CV_DEFAULT, NULL); 1313 cv_init(&zilog->zl_cv_suspend, NULL, CV_DEFAULT, NULL); 1314 1315 return (zilog); 1316 } 1317 1318 void 1319 zil_free(zilog_t *zilog) 1320 { 1321 lwb_t *lwb; 1322 1323 zilog->zl_stop_sync = 1; 1324 1325 while ((lwb = list_head(&zilog->zl_lwb_list)) != NULL) { 1326 list_remove(&zilog->zl_lwb_list, lwb); 1327 if (lwb->lwb_buf != NULL) 1328 zio_buf_free(lwb->lwb_buf, lwb->lwb_sz); 1329 kmem_cache_free(zil_lwb_cache, lwb); 1330 } 1331 list_destroy(&zilog->zl_lwb_list); 1332 1333 avl_destroy(&zilog->zl_vdev_tree); 1334 mutex_destroy(&zilog->zl_vdev_lock); 1335 1336 ASSERT(list_head(&zilog->zl_itx_list) == NULL); 1337 list_destroy(&zilog->zl_itx_list); 1338 mutex_destroy(&zilog->zl_lock); 1339 1340 cv_destroy(&zilog->zl_cv_writer); 1341 cv_destroy(&zilog->zl_cv_suspend); 1342 1343 kmem_free(zilog, sizeof (zilog_t)); 1344 } 1345 1346 /* 1347 * return true if the initial log block is not valid 1348 */ 1349 static boolean_t 1350 zil_empty(zilog_t *zilog) 1351 { 1352 const zil_header_t *zh = zilog->zl_header; 1353 arc_buf_t *abuf = NULL; 1354 1355 if (BP_IS_HOLE(&zh->zh_log)) 1356 return (B_TRUE); 1357 1358 if (zil_read_log_block(zilog, &zh->zh_log, &abuf) != 0) 1359 return (B_TRUE); 1360 1361 VERIFY(arc_buf_remove_ref(abuf, &abuf) == 1); 1362 return (B_FALSE); 1363 } 1364 1365 /* 1366 * Open an intent log. 1367 */ 1368 zilog_t * 1369 zil_open(objset_t *os, zil_get_data_t *get_data) 1370 { 1371 zilog_t *zilog = dmu_objset_zil(os); 1372 1373 zilog->zl_get_data = get_data; 1374 zilog->zl_clean_taskq = taskq_create("zil_clean", 1, minclsyspri, 1375 2, 2, TASKQ_PREPOPULATE); 1376 1377 return (zilog); 1378 } 1379 1380 /* 1381 * Close an intent log. 1382 */ 1383 void 1384 zil_close(zilog_t *zilog) 1385 { 1386 /* 1387 * If the log isn't already committed, mark the objset dirty 1388 * (so zil_sync() will be called) and wait for that txg to sync. 1389 */ 1390 if (!zil_is_committed(zilog)) { 1391 uint64_t txg; 1392 dmu_tx_t *tx = dmu_tx_create(zilog->zl_os); 1393 (void) dmu_tx_assign(tx, TXG_WAIT); 1394 dsl_dataset_dirty(dmu_objset_ds(zilog->zl_os), tx); 1395 txg = dmu_tx_get_txg(tx); 1396 dmu_tx_commit(tx); 1397 txg_wait_synced(zilog->zl_dmu_pool, txg); 1398 } 1399 1400 taskq_destroy(zilog->zl_clean_taskq); 1401 zilog->zl_clean_taskq = NULL; 1402 zilog->zl_get_data = NULL; 1403 1404 zil_itx_clean(zilog); 1405 ASSERT(list_head(&zilog->zl_itx_list) == NULL); 1406 } 1407 1408 /* 1409 * Suspend an intent log. While in suspended mode, we still honor 1410 * synchronous semantics, but we rely on txg_wait_synced() to do it. 1411 * We suspend the log briefly when taking a snapshot so that the snapshot 1412 * contains all the data it's supposed to, and has an empty intent log. 1413 */ 1414 int 1415 zil_suspend(zilog_t *zilog) 1416 { 1417 const zil_header_t *zh = zilog->zl_header; 1418 1419 mutex_enter(&zilog->zl_lock); 1420 if (zh->zh_claim_txg != 0) { /* unplayed log */ 1421 mutex_exit(&zilog->zl_lock); 1422 return (EBUSY); 1423 } 1424 if (zilog->zl_suspend++ != 0) { 1425 /* 1426 * Someone else already began a suspend. 1427 * Just wait for them to finish. 1428 */ 1429 while (zilog->zl_suspending) 1430 cv_wait(&zilog->zl_cv_suspend, &zilog->zl_lock); 1431 mutex_exit(&zilog->zl_lock); 1432 return (0); 1433 } 1434 zilog->zl_suspending = B_TRUE; 1435 mutex_exit(&zilog->zl_lock); 1436 1437 zil_commit(zilog, UINT64_MAX, 0); 1438 1439 /* 1440 * Wait for any in-flight log writes to complete. 1441 */ 1442 mutex_enter(&zilog->zl_lock); 1443 while (zilog->zl_writer) 1444 cv_wait(&zilog->zl_cv_writer, &zilog->zl_lock); 1445 mutex_exit(&zilog->zl_lock); 1446 1447 zil_destroy(zilog, B_FALSE); 1448 1449 mutex_enter(&zilog->zl_lock); 1450 zilog->zl_suspending = B_FALSE; 1451 cv_broadcast(&zilog->zl_cv_suspend); 1452 mutex_exit(&zilog->zl_lock); 1453 1454 return (0); 1455 } 1456 1457 void 1458 zil_resume(zilog_t *zilog) 1459 { 1460 mutex_enter(&zilog->zl_lock); 1461 ASSERT(zilog->zl_suspend != 0); 1462 zilog->zl_suspend--; 1463 mutex_exit(&zilog->zl_lock); 1464 } 1465 1466 typedef struct zil_replay_arg { 1467 objset_t *zr_os; 1468 zil_replay_func_t **zr_replay; 1469 zil_replay_cleaner_t *zr_replay_cleaner; 1470 void *zr_arg; 1471 uint64_t *zr_txgp; 1472 boolean_t zr_byteswap; 1473 char *zr_lrbuf; 1474 } zil_replay_arg_t; 1475 1476 static void 1477 zil_replay_log_record(zilog_t *zilog, lr_t *lr, void *zra, uint64_t claim_txg) 1478 { 1479 zil_replay_arg_t *zr = zra; 1480 const zil_header_t *zh = zilog->zl_header; 1481 uint64_t reclen = lr->lrc_reclen; 1482 uint64_t txtype = lr->lrc_txtype; 1483 char *name; 1484 int pass, error, sunk; 1485 1486 if (zilog->zl_stop_replay) 1487 return; 1488 1489 if (lr->lrc_txg < claim_txg) /* already committed */ 1490 return; 1491 1492 if (lr->lrc_seq <= zh->zh_replay_seq) /* already replayed */ 1493 return; 1494 1495 /* Strip case-insensitive bit, still present in log record */ 1496 txtype &= ~TX_CI; 1497 1498 /* 1499 * Make a copy of the data so we can revise and extend it. 1500 */ 1501 bcopy(lr, zr->zr_lrbuf, reclen); 1502 1503 /* 1504 * The log block containing this lr may have been byteswapped 1505 * so that we can easily examine common fields like lrc_txtype. 1506 * However, the log is a mix of different data types, and only the 1507 * replay vectors know how to byteswap their records. Therefore, if 1508 * the lr was byteswapped, undo it before invoking the replay vector. 1509 */ 1510 if (zr->zr_byteswap) 1511 byteswap_uint64_array(zr->zr_lrbuf, reclen); 1512 1513 /* 1514 * If this is a TX_WRITE with a blkptr, suck in the data. 1515 */ 1516 if (txtype == TX_WRITE && reclen == sizeof (lr_write_t)) { 1517 lr_write_t *lrw = (lr_write_t *)lr; 1518 blkptr_t *wbp = &lrw->lr_blkptr; 1519 uint64_t wlen = lrw->lr_length; 1520 char *wbuf = zr->zr_lrbuf + reclen; 1521 1522 if (BP_IS_HOLE(wbp)) { /* compressed to a hole */ 1523 bzero(wbuf, wlen); 1524 } else { 1525 /* 1526 * A subsequent write may have overwritten this block, 1527 * in which case wbp may have been been freed and 1528 * reallocated, and our read of wbp may fail with a 1529 * checksum error. We can safely ignore this because 1530 * the later write will provide the correct data. 1531 */ 1532 zbookmark_t zb; 1533 1534 zb.zb_objset = dmu_objset_id(zilog->zl_os); 1535 zb.zb_object = lrw->lr_foid; 1536 zb.zb_level = -1; 1537 zb.zb_blkid = lrw->lr_offset / BP_GET_LSIZE(wbp); 1538 1539 (void) zio_wait(zio_read(NULL, zilog->zl_spa, 1540 wbp, wbuf, BP_GET_LSIZE(wbp), NULL, NULL, 1541 ZIO_PRIORITY_SYNC_READ, 1542 ZIO_FLAG_CANFAIL | ZIO_FLAG_SPECULATIVE, &zb)); 1543 (void) memmove(wbuf, wbuf + lrw->lr_blkoff, wlen); 1544 } 1545 } 1546 1547 /* 1548 * Replay of large truncates can end up needing additional txs 1549 * and a different txg. If they are nested within the replay tx 1550 * as below then a hang is possible. So we do the truncate here 1551 * and redo the truncate later (a no-op) and update the sequence 1552 * number whilst in the replay tx. Fortunately, it's safe to repeat 1553 * a truncate if we crash and the truncate commits. A create over 1554 * an existing file will also come in as a TX_TRUNCATE record. 1555 * 1556 * Note, remove of large files and renames over large files is 1557 * handled by putting the deleted object on a stable list 1558 * and if necessary force deleting the object outside of the replay 1559 * transaction using the zr_replay_cleaner. 1560 */ 1561 if (txtype == TX_TRUNCATE) { 1562 *zr->zr_txgp = TXG_NOWAIT; 1563 error = zr->zr_replay[TX_TRUNCATE](zr->zr_arg, zr->zr_lrbuf, 1564 zr->zr_byteswap); 1565 if (error) 1566 goto bad; 1567 zr->zr_byteswap = 0; /* only byteswap once */ 1568 } 1569 1570 /* 1571 * We must now do two things atomically: replay this log record, 1572 * and update the log header to reflect the fact that we did so. 1573 * We use the DMU's ability to assign into a specific txg to do this. 1574 */ 1575 for (pass = 1, sunk = B_FALSE; /* CONSTANTCONDITION */; pass++) { 1576 uint64_t replay_txg; 1577 dmu_tx_t *replay_tx; 1578 1579 replay_tx = dmu_tx_create(zr->zr_os); 1580 error = dmu_tx_assign(replay_tx, TXG_WAIT); 1581 if (error) { 1582 dmu_tx_abort(replay_tx); 1583 break; 1584 } 1585 1586 replay_txg = dmu_tx_get_txg(replay_tx); 1587 1588 if (txtype == 0 || txtype >= TX_MAX_TYPE) { 1589 error = EINVAL; 1590 } else { 1591 /* 1592 * On the first pass, arrange for the replay vector 1593 * to fail its dmu_tx_assign(). That's the only way 1594 * to ensure that those code paths remain well tested. 1595 * 1596 * Only byteswap (if needed) on the 1st pass. 1597 */ 1598 *zr->zr_txgp = replay_txg - (pass == 1); 1599 error = zr->zr_replay[txtype](zr->zr_arg, zr->zr_lrbuf, 1600 zr->zr_byteswap && pass == 1); 1601 *zr->zr_txgp = TXG_NOWAIT; 1602 } 1603 1604 if (error == 0) { 1605 dsl_dataset_dirty(dmu_objset_ds(zr->zr_os), replay_tx); 1606 zilog->zl_replay_seq[replay_txg & TXG_MASK] = 1607 lr->lrc_seq; 1608 } 1609 1610 dmu_tx_commit(replay_tx); 1611 1612 if (!error) 1613 return; 1614 1615 /* 1616 * The DMU's dnode layer doesn't see removes until the txg 1617 * commits, so a subsequent claim can spuriously fail with 1618 * EEXIST. So if we receive any error other than ERESTART 1619 * we try syncing out any removes then retrying the 1620 * transaction. 1621 */ 1622 if (error != ERESTART && !sunk) { 1623 if (zr->zr_replay_cleaner) 1624 zr->zr_replay_cleaner(zr->zr_arg); 1625 txg_wait_synced(spa_get_dsl(zilog->zl_spa), 0); 1626 sunk = B_TRUE; 1627 continue; /* retry */ 1628 } 1629 1630 if (error != ERESTART) 1631 break; 1632 1633 if (pass != 1) 1634 txg_wait_open(spa_get_dsl(zilog->zl_spa), 1635 replay_txg + 1); 1636 1637 dprintf("pass %d, retrying\n", pass); 1638 } 1639 1640 bad: 1641 ASSERT(error && error != ERESTART); 1642 name = kmem_alloc(MAXNAMELEN, KM_SLEEP); 1643 dmu_objset_name(zr->zr_os, name); 1644 cmn_err(CE_WARN, "ZFS replay transaction error %d, " 1645 "dataset %s, seq 0x%llx, txtype %llu %s\n", 1646 error, name, (u_longlong_t)lr->lrc_seq, (u_longlong_t)txtype, 1647 (lr->lrc_txtype & TX_CI) ? "CI" : ""); 1648 zilog->zl_stop_replay = 1; 1649 kmem_free(name, MAXNAMELEN); 1650 } 1651 1652 /* ARGSUSED */ 1653 static void 1654 zil_incr_blks(zilog_t *zilog, blkptr_t *bp, void *arg, uint64_t claim_txg) 1655 { 1656 zilog->zl_replay_blks++; 1657 } 1658 1659 /* 1660 * If this dataset has a non-empty intent log, replay it and destroy it. 1661 */ 1662 void 1663 zil_replay(objset_t *os, void *arg, uint64_t *txgp, 1664 zil_replay_func_t *replay_func[TX_MAX_TYPE], 1665 zil_replay_cleaner_t *replay_cleaner) 1666 { 1667 zilog_t *zilog = dmu_objset_zil(os); 1668 const zil_header_t *zh = zilog->zl_header; 1669 zil_replay_arg_t zr; 1670 1671 if (zil_empty(zilog)) { 1672 zil_destroy(zilog, B_TRUE); 1673 return; 1674 } 1675 1676 zr.zr_os = os; 1677 zr.zr_replay = replay_func; 1678 zr.zr_replay_cleaner = replay_cleaner; 1679 zr.zr_arg = arg; 1680 zr.zr_txgp = txgp; 1681 zr.zr_byteswap = BP_SHOULD_BYTESWAP(&zh->zh_log); 1682 zr.zr_lrbuf = kmem_alloc(2 * SPA_MAXBLOCKSIZE, KM_SLEEP); 1683 1684 /* 1685 * Wait for in-progress removes to sync before starting replay. 1686 */ 1687 txg_wait_synced(zilog->zl_dmu_pool, 0); 1688 1689 zilog->zl_stop_replay = 0; 1690 zilog->zl_replay_time = lbolt; 1691 ASSERT(zilog->zl_replay_blks == 0); 1692 (void) zil_parse(zilog, zil_incr_blks, zil_replay_log_record, &zr, 1693 zh->zh_claim_txg); 1694 kmem_free(zr.zr_lrbuf, 2 * SPA_MAXBLOCKSIZE); 1695 1696 zil_destroy(zilog, B_FALSE); 1697 txg_wait_synced(zilog->zl_dmu_pool, zilog->zl_destroy_txg); 1698 } 1699 1700 /* 1701 * Report whether all transactions are committed 1702 */ 1703 int 1704 zil_is_committed(zilog_t *zilog) 1705 { 1706 lwb_t *lwb; 1707 int ret; 1708 1709 mutex_enter(&zilog->zl_lock); 1710 while (zilog->zl_writer) 1711 cv_wait(&zilog->zl_cv_writer, &zilog->zl_lock); 1712 1713 /* recent unpushed intent log transactions? */ 1714 if (!list_is_empty(&zilog->zl_itx_list)) { 1715 ret = B_FALSE; 1716 goto out; 1717 } 1718 1719 /* intent log never used? */ 1720 lwb = list_head(&zilog->zl_lwb_list); 1721 if (lwb == NULL) { 1722 ret = B_TRUE; 1723 goto out; 1724 } 1725 1726 /* 1727 * more than 1 log buffer means zil_sync() hasn't yet freed 1728 * entries after a txg has committed 1729 */ 1730 if (list_next(&zilog->zl_lwb_list, lwb)) { 1731 ret = B_FALSE; 1732 goto out; 1733 } 1734 1735 ASSERT(zil_empty(zilog)); 1736 ret = B_TRUE; 1737 out: 1738 cv_broadcast(&zilog->zl_cv_writer); 1739 mutex_exit(&zilog->zl_lock); 1740 return (ret); 1741 } 1742