1 // SPDX-License-Identifier: GPL-2.0 2 3 #include "bcachefs.h" 4 #include "btree_key_cache.h" 5 #include "btree_update.h" 6 #include "buckets.h" 7 #include "errcode.h" 8 #include "error.h" 9 #include "journal.h" 10 #include "journal_io.h" 11 #include "journal_reclaim.h" 12 #include "replicas.h" 13 #include "sb-members.h" 14 #include "trace.h" 15 16 #include <linux/kthread.h> 17 #include <linux/sched/mm.h> 18 19 /* Free space calculations: */ 20 21 static unsigned journal_space_from(struct journal_device *ja, 22 enum journal_space_from from) 23 { 24 switch (from) { 25 case journal_space_discarded: 26 return ja->discard_idx; 27 case journal_space_clean_ondisk: 28 return ja->dirty_idx_ondisk; 29 case journal_space_clean: 30 return ja->dirty_idx; 31 default: 32 BUG(); 33 } 34 } 35 36 unsigned bch2_journal_dev_buckets_available(struct journal *j, 37 struct journal_device *ja, 38 enum journal_space_from from) 39 { 40 unsigned available = (journal_space_from(ja, from) - 41 ja->cur_idx - 1 + ja->nr) % ja->nr; 42 43 /* 44 * Don't use the last bucket unless writing the new last_seq 45 * will make another bucket available: 46 */ 47 if (available && ja->dirty_idx_ondisk == ja->dirty_idx) 48 --available; 49 50 return available; 51 } 52 53 static void journal_set_remaining(struct journal *j, unsigned u64s_remaining) 54 { 55 union journal_preres_state old, new; 56 u64 v = atomic64_read(&j->prereserved.counter); 57 58 do { 59 old.v = new.v = v; 60 new.remaining = u64s_remaining; 61 } while ((v = atomic64_cmpxchg(&j->prereserved.counter, 62 old.v, new.v)) != old.v); 63 } 64 65 static struct journal_space 66 journal_dev_space_available(struct journal *j, struct bch_dev *ca, 67 enum journal_space_from from) 68 { 69 struct journal_device *ja = &ca->journal; 70 unsigned sectors, buckets, unwritten; 71 u64 seq; 72 73 if (from == journal_space_total) 74 return (struct journal_space) { 75 .next_entry = ca->mi.bucket_size, 76 .total = ca->mi.bucket_size * ja->nr, 77 }; 78 79 buckets = bch2_journal_dev_buckets_available(j, ja, from); 80 sectors = ja->sectors_free; 81 82 /* 83 * We that we don't allocate the space for a journal entry 84 * until we write it out - thus, account for it here: 85 */ 86 for (seq = journal_last_unwritten_seq(j); 87 seq <= journal_cur_seq(j); 88 seq++) { 89 unwritten = j->buf[seq & JOURNAL_BUF_MASK].sectors; 90 91 if (!unwritten) 92 continue; 93 94 /* entry won't fit on this device, skip: */ 95 if (unwritten > ca->mi.bucket_size) 96 continue; 97 98 if (unwritten >= sectors) { 99 if (!buckets) { 100 sectors = 0; 101 break; 102 } 103 104 buckets--; 105 sectors = ca->mi.bucket_size; 106 } 107 108 sectors -= unwritten; 109 } 110 111 if (sectors < ca->mi.bucket_size && buckets) { 112 buckets--; 113 sectors = ca->mi.bucket_size; 114 } 115 116 return (struct journal_space) { 117 .next_entry = sectors, 118 .total = sectors + buckets * ca->mi.bucket_size, 119 }; 120 } 121 122 static struct journal_space __journal_space_available(struct journal *j, unsigned nr_devs_want, 123 enum journal_space_from from) 124 { 125 struct bch_fs *c = container_of(j, struct bch_fs, journal); 126 struct bch_dev *ca; 127 unsigned i, pos, nr_devs = 0; 128 struct journal_space space, dev_space[BCH_SB_MEMBERS_MAX]; 129 130 BUG_ON(nr_devs_want > ARRAY_SIZE(dev_space)); 131 132 rcu_read_lock(); 133 for_each_member_device_rcu(ca, c, i, 134 &c->rw_devs[BCH_DATA_journal]) { 135 if (!ca->journal.nr) 136 continue; 137 138 space = journal_dev_space_available(j, ca, from); 139 if (!space.next_entry) 140 continue; 141 142 for (pos = 0; pos < nr_devs; pos++) 143 if (space.total > dev_space[pos].total) 144 break; 145 146 array_insert_item(dev_space, nr_devs, pos, space); 147 } 148 rcu_read_unlock(); 149 150 if (nr_devs < nr_devs_want) 151 return (struct journal_space) { 0, 0 }; 152 153 /* 154 * We sorted largest to smallest, and we want the smallest out of the 155 * @nr_devs_want largest devices: 156 */ 157 return dev_space[nr_devs_want - 1]; 158 } 159 160 void bch2_journal_space_available(struct journal *j) 161 { 162 struct bch_fs *c = container_of(j, struct bch_fs, journal); 163 struct bch_dev *ca; 164 unsigned clean, clean_ondisk, total; 165 s64 u64s_remaining = 0; 166 unsigned max_entry_size = min(j->buf[0].buf_size >> 9, 167 j->buf[1].buf_size >> 9); 168 unsigned i, nr_online = 0, nr_devs_want; 169 bool can_discard = false; 170 int ret = 0; 171 172 lockdep_assert_held(&j->lock); 173 174 rcu_read_lock(); 175 for_each_member_device_rcu(ca, c, i, 176 &c->rw_devs[BCH_DATA_journal]) { 177 struct journal_device *ja = &ca->journal; 178 179 if (!ja->nr) 180 continue; 181 182 while (ja->dirty_idx != ja->cur_idx && 183 ja->bucket_seq[ja->dirty_idx] < journal_last_seq(j)) 184 ja->dirty_idx = (ja->dirty_idx + 1) % ja->nr; 185 186 while (ja->dirty_idx_ondisk != ja->dirty_idx && 187 ja->bucket_seq[ja->dirty_idx_ondisk] < j->last_seq_ondisk) 188 ja->dirty_idx_ondisk = (ja->dirty_idx_ondisk + 1) % ja->nr; 189 190 if (ja->discard_idx != ja->dirty_idx_ondisk) 191 can_discard = true; 192 193 max_entry_size = min_t(unsigned, max_entry_size, ca->mi.bucket_size); 194 nr_online++; 195 } 196 rcu_read_unlock(); 197 198 j->can_discard = can_discard; 199 200 if (nr_online < c->opts.metadata_replicas_required) { 201 ret = JOURNAL_ERR_insufficient_devices; 202 goto out; 203 } 204 205 nr_devs_want = min_t(unsigned, nr_online, c->opts.metadata_replicas); 206 207 for (i = 0; i < journal_space_nr; i++) 208 j->space[i] = __journal_space_available(j, nr_devs_want, i); 209 210 clean_ondisk = j->space[journal_space_clean_ondisk].total; 211 clean = j->space[journal_space_clean].total; 212 total = j->space[journal_space_total].total; 213 214 if (!j->space[journal_space_discarded].next_entry) 215 ret = JOURNAL_ERR_journal_full; 216 217 if ((j->space[journal_space_clean_ondisk].next_entry < 218 j->space[journal_space_clean_ondisk].total) && 219 (clean - clean_ondisk <= total / 8) && 220 (clean_ondisk * 2 > clean)) 221 set_bit(JOURNAL_MAY_SKIP_FLUSH, &j->flags); 222 else 223 clear_bit(JOURNAL_MAY_SKIP_FLUSH, &j->flags); 224 225 u64s_remaining = (u64) clean << 6; 226 u64s_remaining -= (u64) total << 3; 227 u64s_remaining = max(0LL, u64s_remaining); 228 u64s_remaining /= 4; 229 u64s_remaining = min_t(u64, u64s_remaining, U32_MAX); 230 out: 231 j->cur_entry_sectors = !ret ? j->space[journal_space_discarded].next_entry : 0; 232 j->cur_entry_error = ret; 233 journal_set_remaining(j, u64s_remaining); 234 journal_set_watermark(j); 235 236 if (!ret) 237 journal_wake(j); 238 } 239 240 /* Discards - last part of journal reclaim: */ 241 242 static bool should_discard_bucket(struct journal *j, struct journal_device *ja) 243 { 244 bool ret; 245 246 spin_lock(&j->lock); 247 ret = ja->discard_idx != ja->dirty_idx_ondisk; 248 spin_unlock(&j->lock); 249 250 return ret; 251 } 252 253 /* 254 * Advance ja->discard_idx as long as it points to buckets that are no longer 255 * dirty, issuing discards if necessary: 256 */ 257 void bch2_journal_do_discards(struct journal *j) 258 { 259 struct bch_fs *c = container_of(j, struct bch_fs, journal); 260 struct bch_dev *ca; 261 unsigned iter; 262 263 mutex_lock(&j->discard_lock); 264 265 for_each_rw_member(ca, c, iter) { 266 struct journal_device *ja = &ca->journal; 267 268 while (should_discard_bucket(j, ja)) { 269 if (!c->opts.nochanges && 270 ca->mi.discard && 271 bdev_max_discard_sectors(ca->disk_sb.bdev)) 272 blkdev_issue_discard(ca->disk_sb.bdev, 273 bucket_to_sector(ca, 274 ja->buckets[ja->discard_idx]), 275 ca->mi.bucket_size, GFP_NOFS); 276 277 spin_lock(&j->lock); 278 ja->discard_idx = (ja->discard_idx + 1) % ja->nr; 279 280 bch2_journal_space_available(j); 281 spin_unlock(&j->lock); 282 } 283 } 284 285 mutex_unlock(&j->discard_lock); 286 } 287 288 /* 289 * Journal entry pinning - machinery for holding a reference on a given journal 290 * entry, holding it open to ensure it gets replayed during recovery: 291 */ 292 293 void bch2_journal_reclaim_fast(struct journal *j) 294 { 295 bool popped = false; 296 297 lockdep_assert_held(&j->lock); 298 299 /* 300 * Unpin journal entries whose reference counts reached zero, meaning 301 * all btree nodes got written out 302 */ 303 while (!fifo_empty(&j->pin) && 304 !atomic_read(&fifo_peek_front(&j->pin).count)) { 305 j->pin.front++; 306 popped = true; 307 } 308 309 if (popped) 310 bch2_journal_space_available(j); 311 } 312 313 bool __bch2_journal_pin_put(struct journal *j, u64 seq) 314 { 315 struct journal_entry_pin_list *pin_list = journal_seq_pin(j, seq); 316 317 return atomic_dec_and_test(&pin_list->count); 318 } 319 320 void bch2_journal_pin_put(struct journal *j, u64 seq) 321 { 322 if (__bch2_journal_pin_put(j, seq)) { 323 spin_lock(&j->lock); 324 bch2_journal_reclaim_fast(j); 325 spin_unlock(&j->lock); 326 } 327 } 328 329 static inline bool __journal_pin_drop(struct journal *j, 330 struct journal_entry_pin *pin) 331 { 332 struct journal_entry_pin_list *pin_list; 333 334 if (!journal_pin_active(pin)) 335 return false; 336 337 if (j->flush_in_progress == pin) 338 j->flush_in_progress_dropped = true; 339 340 pin_list = journal_seq_pin(j, pin->seq); 341 pin->seq = 0; 342 list_del_init(&pin->list); 343 344 /* 345 * Unpinning a journal entry may make journal_next_bucket() succeed, if 346 * writing a new last_seq will now make another bucket available: 347 */ 348 return atomic_dec_and_test(&pin_list->count) && 349 pin_list == &fifo_peek_front(&j->pin); 350 } 351 352 void bch2_journal_pin_drop(struct journal *j, 353 struct journal_entry_pin *pin) 354 { 355 spin_lock(&j->lock); 356 if (__journal_pin_drop(j, pin)) 357 bch2_journal_reclaim_fast(j); 358 spin_unlock(&j->lock); 359 } 360 361 static enum journal_pin_type journal_pin_type(journal_pin_flush_fn fn) 362 { 363 if (fn == bch2_btree_node_flush0 || 364 fn == bch2_btree_node_flush1) 365 return JOURNAL_PIN_btree; 366 else if (fn == bch2_btree_key_cache_journal_flush) 367 return JOURNAL_PIN_key_cache; 368 else 369 return JOURNAL_PIN_other; 370 } 371 372 void bch2_journal_pin_set(struct journal *j, u64 seq, 373 struct journal_entry_pin *pin, 374 journal_pin_flush_fn flush_fn) 375 { 376 struct journal_entry_pin_list *pin_list; 377 bool reclaim; 378 379 spin_lock(&j->lock); 380 381 if (seq < journal_last_seq(j)) { 382 /* 383 * bch2_journal_pin_copy() raced with bch2_journal_pin_drop() on 384 * the src pin - with the pin dropped, the entry to pin might no 385 * longer to exist, but that means there's no longer anything to 386 * copy and we can bail out here: 387 */ 388 spin_unlock(&j->lock); 389 return; 390 } 391 392 pin_list = journal_seq_pin(j, seq); 393 394 reclaim = __journal_pin_drop(j, pin); 395 396 atomic_inc(&pin_list->count); 397 pin->seq = seq; 398 pin->flush = flush_fn; 399 400 if (flush_fn) 401 list_add(&pin->list, &pin_list->list[journal_pin_type(flush_fn)]); 402 else 403 list_add(&pin->list, &pin_list->flushed); 404 405 if (reclaim) 406 bch2_journal_reclaim_fast(j); 407 spin_unlock(&j->lock); 408 409 /* 410 * If the journal is currently full, we might want to call flush_fn 411 * immediately: 412 */ 413 journal_wake(j); 414 } 415 416 /** 417 * bch2_journal_pin_flush: ensure journal pin callback is no longer running 418 * @j: journal object 419 * @pin: pin to flush 420 */ 421 void bch2_journal_pin_flush(struct journal *j, struct journal_entry_pin *pin) 422 { 423 BUG_ON(journal_pin_active(pin)); 424 425 wait_event(j->pin_flush_wait, j->flush_in_progress != pin); 426 } 427 428 /* 429 * Journal reclaim: flush references to open journal entries to reclaim space in 430 * the journal 431 * 432 * May be done by the journal code in the background as needed to free up space 433 * for more journal entries, or as part of doing a clean shutdown, or to migrate 434 * data off of a specific device: 435 */ 436 437 static struct journal_entry_pin * 438 journal_get_next_pin(struct journal *j, 439 u64 seq_to_flush, 440 unsigned allowed_below_seq, 441 unsigned allowed_above_seq, 442 u64 *seq) 443 { 444 struct journal_entry_pin_list *pin_list; 445 struct journal_entry_pin *ret = NULL; 446 unsigned i; 447 448 fifo_for_each_entry_ptr(pin_list, &j->pin, *seq) { 449 if (*seq > seq_to_flush && !allowed_above_seq) 450 break; 451 452 for (i = 0; i < JOURNAL_PIN_NR; i++) 453 if ((((1U << i) & allowed_below_seq) && *seq <= seq_to_flush) || 454 ((1U << i) & allowed_above_seq)) { 455 ret = list_first_entry_or_null(&pin_list->list[i], 456 struct journal_entry_pin, list); 457 if (ret) 458 return ret; 459 } 460 } 461 462 return NULL; 463 } 464 465 /* returns true if we did work */ 466 static size_t journal_flush_pins(struct journal *j, 467 u64 seq_to_flush, 468 unsigned allowed_below_seq, 469 unsigned allowed_above_seq, 470 unsigned min_any, 471 unsigned min_key_cache) 472 { 473 struct journal_entry_pin *pin; 474 size_t nr_flushed = 0; 475 journal_pin_flush_fn flush_fn; 476 u64 seq; 477 int err; 478 479 lockdep_assert_held(&j->reclaim_lock); 480 481 while (1) { 482 unsigned allowed_above = allowed_above_seq; 483 unsigned allowed_below = allowed_below_seq; 484 485 if (min_any) { 486 allowed_above |= ~0; 487 allowed_below |= ~0; 488 } 489 490 if (min_key_cache) { 491 allowed_above |= 1U << JOURNAL_PIN_key_cache; 492 allowed_below |= 1U << JOURNAL_PIN_key_cache; 493 } 494 495 cond_resched(); 496 497 j->last_flushed = jiffies; 498 499 spin_lock(&j->lock); 500 pin = journal_get_next_pin(j, seq_to_flush, allowed_below, allowed_above, &seq); 501 if (pin) { 502 BUG_ON(j->flush_in_progress); 503 j->flush_in_progress = pin; 504 j->flush_in_progress_dropped = false; 505 flush_fn = pin->flush; 506 } 507 spin_unlock(&j->lock); 508 509 if (!pin) 510 break; 511 512 if (min_key_cache && pin->flush == bch2_btree_key_cache_journal_flush) 513 min_key_cache--; 514 515 if (min_any) 516 min_any--; 517 518 err = flush_fn(j, pin, seq); 519 520 spin_lock(&j->lock); 521 /* Pin might have been dropped or rearmed: */ 522 if (likely(!err && !j->flush_in_progress_dropped)) 523 list_move(&pin->list, &journal_seq_pin(j, seq)->flushed); 524 j->flush_in_progress = NULL; 525 j->flush_in_progress_dropped = false; 526 spin_unlock(&j->lock); 527 528 wake_up(&j->pin_flush_wait); 529 530 if (err) 531 break; 532 533 nr_flushed++; 534 } 535 536 return nr_flushed; 537 } 538 539 static u64 journal_seq_to_flush(struct journal *j) 540 { 541 struct bch_fs *c = container_of(j, struct bch_fs, journal); 542 struct bch_dev *ca; 543 u64 seq_to_flush = 0; 544 unsigned iter; 545 546 spin_lock(&j->lock); 547 548 for_each_rw_member(ca, c, iter) { 549 struct journal_device *ja = &ca->journal; 550 unsigned nr_buckets, bucket_to_flush; 551 552 if (!ja->nr) 553 continue; 554 555 /* Try to keep the journal at most half full: */ 556 nr_buckets = ja->nr / 2; 557 558 /* And include pre-reservations: */ 559 nr_buckets += DIV_ROUND_UP(j->prereserved.reserved, 560 (ca->mi.bucket_size << 6) - 561 journal_entry_overhead(j)); 562 563 nr_buckets = min(nr_buckets, ja->nr); 564 565 bucket_to_flush = (ja->cur_idx + nr_buckets) % ja->nr; 566 seq_to_flush = max(seq_to_flush, 567 ja->bucket_seq[bucket_to_flush]); 568 } 569 570 /* Also flush if the pin fifo is more than half full */ 571 seq_to_flush = max_t(s64, seq_to_flush, 572 (s64) journal_cur_seq(j) - 573 (j->pin.size >> 1)); 574 spin_unlock(&j->lock); 575 576 return seq_to_flush; 577 } 578 579 /** 580 * __bch2_journal_reclaim - free up journal buckets 581 * @j: journal object 582 * @direct: direct or background reclaim? 583 * @kicked: requested to run since we last ran? 584 * Returns: 0 on success, or -EIO if the journal has been shutdown 585 * 586 * Background journal reclaim writes out btree nodes. It should be run 587 * early enough so that we never completely run out of journal buckets. 588 * 589 * High watermarks for triggering background reclaim: 590 * - FIFO has fewer than 512 entries left 591 * - fewer than 25% journal buckets free 592 * 593 * Background reclaim runs until low watermarks are reached: 594 * - FIFO has more than 1024 entries left 595 * - more than 50% journal buckets free 596 * 597 * As long as a reclaim can complete in the time it takes to fill up 598 * 512 journal entries or 25% of all journal buckets, then 599 * journal_next_bucket() should not stall. 600 */ 601 static int __bch2_journal_reclaim(struct journal *j, bool direct, bool kicked) 602 { 603 struct bch_fs *c = container_of(j, struct bch_fs, journal); 604 bool kthread = (current->flags & PF_KTHREAD) != 0; 605 u64 seq_to_flush; 606 size_t min_nr, min_key_cache, nr_flushed; 607 unsigned flags; 608 int ret = 0; 609 610 /* 611 * We can't invoke memory reclaim while holding the reclaim_lock - 612 * journal reclaim is required to make progress for memory reclaim 613 * (cleaning the caches), so we can't get stuck in memory reclaim while 614 * we're holding the reclaim lock: 615 */ 616 lockdep_assert_held(&j->reclaim_lock); 617 flags = memalloc_noreclaim_save(); 618 619 do { 620 if (kthread && kthread_should_stop()) 621 break; 622 623 if (bch2_journal_error(j)) { 624 ret = -EIO; 625 break; 626 } 627 628 bch2_journal_do_discards(j); 629 630 seq_to_flush = journal_seq_to_flush(j); 631 min_nr = 0; 632 633 /* 634 * If it's been longer than j->reclaim_delay_ms since we last flushed, 635 * make sure to flush at least one journal pin: 636 */ 637 if (time_after(jiffies, j->last_flushed + 638 msecs_to_jiffies(c->opts.journal_reclaim_delay))) 639 min_nr = 1; 640 641 if (j->prereserved.reserved * 4 > j->prereserved.remaining) 642 min_nr = 1; 643 644 if (fifo_free(&j->pin) <= 32) 645 min_nr = 1; 646 647 if (atomic_read(&c->btree_cache.dirty) * 2 > c->btree_cache.used) 648 min_nr = 1; 649 650 min_key_cache = min(bch2_nr_btree_keys_need_flush(c), (size_t) 128); 651 652 trace_and_count(c, journal_reclaim_start, c, 653 direct, kicked, 654 min_nr, min_key_cache, 655 j->prereserved.reserved, 656 j->prereserved.remaining, 657 atomic_read(&c->btree_cache.dirty), 658 c->btree_cache.used, 659 atomic_long_read(&c->btree_key_cache.nr_dirty), 660 atomic_long_read(&c->btree_key_cache.nr_keys)); 661 662 nr_flushed = journal_flush_pins(j, seq_to_flush, 663 ~0, 0, 664 min_nr, min_key_cache); 665 666 if (direct) 667 j->nr_direct_reclaim += nr_flushed; 668 else 669 j->nr_background_reclaim += nr_flushed; 670 trace_and_count(c, journal_reclaim_finish, c, nr_flushed); 671 672 if (nr_flushed) 673 wake_up(&j->reclaim_wait); 674 } while ((min_nr || min_key_cache) && nr_flushed && !direct); 675 676 memalloc_noreclaim_restore(flags); 677 678 return ret; 679 } 680 681 int bch2_journal_reclaim(struct journal *j) 682 { 683 return __bch2_journal_reclaim(j, true, true); 684 } 685 686 static int bch2_journal_reclaim_thread(void *arg) 687 { 688 struct journal *j = arg; 689 struct bch_fs *c = container_of(j, struct bch_fs, journal); 690 unsigned long delay, now; 691 bool journal_empty; 692 int ret = 0; 693 694 set_freezable(); 695 696 j->last_flushed = jiffies; 697 698 while (!ret && !kthread_should_stop()) { 699 bool kicked = j->reclaim_kicked; 700 701 j->reclaim_kicked = false; 702 703 mutex_lock(&j->reclaim_lock); 704 ret = __bch2_journal_reclaim(j, false, kicked); 705 mutex_unlock(&j->reclaim_lock); 706 707 now = jiffies; 708 delay = msecs_to_jiffies(c->opts.journal_reclaim_delay); 709 j->next_reclaim = j->last_flushed + delay; 710 711 if (!time_in_range(j->next_reclaim, now, now + delay)) 712 j->next_reclaim = now + delay; 713 714 while (1) { 715 set_current_state(TASK_INTERRUPTIBLE|TASK_FREEZABLE); 716 if (kthread_should_stop()) 717 break; 718 if (j->reclaim_kicked) 719 break; 720 721 spin_lock(&j->lock); 722 journal_empty = fifo_empty(&j->pin); 723 spin_unlock(&j->lock); 724 725 if (journal_empty) 726 schedule(); 727 else if (time_after(j->next_reclaim, jiffies)) 728 schedule_timeout(j->next_reclaim - jiffies); 729 else 730 break; 731 } 732 __set_current_state(TASK_RUNNING); 733 } 734 735 return 0; 736 } 737 738 void bch2_journal_reclaim_stop(struct journal *j) 739 { 740 struct task_struct *p = j->reclaim_thread; 741 742 j->reclaim_thread = NULL; 743 744 if (p) { 745 kthread_stop(p); 746 put_task_struct(p); 747 } 748 } 749 750 int bch2_journal_reclaim_start(struct journal *j) 751 { 752 struct bch_fs *c = container_of(j, struct bch_fs, journal); 753 struct task_struct *p; 754 int ret; 755 756 if (j->reclaim_thread) 757 return 0; 758 759 p = kthread_create(bch2_journal_reclaim_thread, j, 760 "bch-reclaim/%s", c->name); 761 ret = PTR_ERR_OR_ZERO(p); 762 if (ret) { 763 bch_err_msg(c, ret, "creating journal reclaim thread"); 764 return ret; 765 } 766 767 get_task_struct(p); 768 j->reclaim_thread = p; 769 wake_up_process(p); 770 return 0; 771 } 772 773 static int journal_flush_done(struct journal *j, u64 seq_to_flush, 774 bool *did_work) 775 { 776 int ret; 777 778 ret = bch2_journal_error(j); 779 if (ret) 780 return ret; 781 782 mutex_lock(&j->reclaim_lock); 783 784 if (journal_flush_pins(j, seq_to_flush, 785 (1U << JOURNAL_PIN_key_cache)| 786 (1U << JOURNAL_PIN_other), 0, 0, 0) || 787 journal_flush_pins(j, seq_to_flush, 788 (1U << JOURNAL_PIN_btree), 0, 0, 0)) 789 *did_work = true; 790 791 spin_lock(&j->lock); 792 /* 793 * If journal replay hasn't completed, the unreplayed journal entries 794 * hold refs on their corresponding sequence numbers 795 */ 796 ret = !test_bit(JOURNAL_REPLAY_DONE, &j->flags) || 797 journal_last_seq(j) > seq_to_flush || 798 !fifo_used(&j->pin); 799 800 spin_unlock(&j->lock); 801 mutex_unlock(&j->reclaim_lock); 802 803 return ret; 804 } 805 806 bool bch2_journal_flush_pins(struct journal *j, u64 seq_to_flush) 807 { 808 bool did_work = false; 809 810 if (!test_bit(JOURNAL_STARTED, &j->flags)) 811 return false; 812 813 closure_wait_event(&j->async_wait, 814 journal_flush_done(j, seq_to_flush, &did_work)); 815 816 return did_work; 817 } 818 819 int bch2_journal_flush_device_pins(struct journal *j, int dev_idx) 820 { 821 struct bch_fs *c = container_of(j, struct bch_fs, journal); 822 struct journal_entry_pin_list *p; 823 u64 iter, seq = 0; 824 int ret = 0; 825 826 spin_lock(&j->lock); 827 fifo_for_each_entry_ptr(p, &j->pin, iter) 828 if (dev_idx >= 0 829 ? bch2_dev_list_has_dev(p->devs, dev_idx) 830 : p->devs.nr < c->opts.metadata_replicas) 831 seq = iter; 832 spin_unlock(&j->lock); 833 834 bch2_journal_flush_pins(j, seq); 835 836 ret = bch2_journal_error(j); 837 if (ret) 838 return ret; 839 840 mutex_lock(&c->replicas_gc_lock); 841 bch2_replicas_gc_start(c, 1 << BCH_DATA_journal); 842 843 /* 844 * Now that we've populated replicas_gc, write to the journal to mark 845 * active journal devices. This handles the case where the journal might 846 * be empty. Otherwise we could clear all journal replicas and 847 * temporarily put the fs into an unrecoverable state. Journal recovery 848 * expects to find devices marked for journal data on unclean mount. 849 */ 850 ret = bch2_journal_meta(&c->journal); 851 if (ret) 852 goto err; 853 854 seq = 0; 855 spin_lock(&j->lock); 856 while (!ret) { 857 struct bch_replicas_padded replicas; 858 859 seq = max(seq, journal_last_seq(j)); 860 if (seq >= j->pin.back) 861 break; 862 bch2_devlist_to_replicas(&replicas.e, BCH_DATA_journal, 863 journal_seq_pin(j, seq)->devs); 864 seq++; 865 866 spin_unlock(&j->lock); 867 ret = bch2_mark_replicas(c, &replicas.e); 868 spin_lock(&j->lock); 869 } 870 spin_unlock(&j->lock); 871 err: 872 ret = bch2_replicas_gc_end(c, ret); 873 mutex_unlock(&c->replicas_gc_lock); 874 875 return ret; 876 } 877