1 // SPDX-License-Identifier: GPL-2.0 2 /* 3 * Primary bucket allocation code 4 * 5 * Copyright 2012 Google, Inc. 6 * 7 * Allocation in bcache is done in terms of buckets: 8 * 9 * Each bucket has associated an 8 bit gen; this gen corresponds to the gen in 10 * btree pointers - they must match for the pointer to be considered valid. 11 * 12 * Thus (assuming a bucket has no dirty data or metadata in it) we can reuse a 13 * bucket simply by incrementing its gen. 14 * 15 * The gens (along with the priorities; it's really the gens are important but 16 * the code is named as if it's the priorities) are written in an arbitrary list 17 * of buckets on disk, with a pointer to them in the journal header. 18 * 19 * When we invalidate a bucket, we have to write its new gen to disk and wait 20 * for that write to complete before we use it - otherwise after a crash we 21 * could have pointers that appeared to be good but pointed to data that had 22 * been overwritten. 23 * 24 * Since the gens and priorities are all stored contiguously on disk, we can 25 * batch this up: We fill up the free_inc list with freshly invalidated buckets, 26 * call prio_write(), and when prio_write() finishes we pull buckets off the 27 * free_inc list and optionally discard them. 28 * 29 * free_inc isn't the only freelist - if it was, we'd often to sleep while 30 * priorities and gens were being written before we could allocate. c->free is a 31 * smaller freelist, and buckets on that list are always ready to be used. 32 * 33 * If we've got discards enabled, that happens when a bucket moves from the 34 * free_inc list to the free list. 35 * 36 * There is another freelist, because sometimes we have buckets that we know 37 * have nothing pointing into them - these we can reuse without waiting for 38 * priorities to be rewritten. These come from freed btree nodes and buckets 39 * that garbage collection discovered no longer had valid keys pointing into 40 * them (because they were overwritten). That's the unused list - buckets on the 41 * unused list move to the free list, optionally being discarded in the process. 42 * 43 * It's also important to ensure that gens don't wrap around - with respect to 44 * either the oldest gen in the btree or the gen on disk. This is quite 45 * difficult to do in practice, but we explicitly guard against it anyways - if 46 * a bucket is in danger of wrapping around we simply skip invalidating it that 47 * time around, and we garbage collect or rewrite the priorities sooner than we 48 * would have otherwise. 49 * 50 * bch_bucket_alloc() allocates a single bucket from a specific cache. 51 * 52 * bch_bucket_alloc_set() allocates one bucket from different caches 53 * out of a cache set. 54 * 55 * free_some_buckets() drives all the processes described above. It's called 56 * from bch_bucket_alloc() and a few other places that need to make sure free 57 * buckets are ready. 58 * 59 * invalidate_buckets_(lru|fifo)() find buckets that are available to be 60 * invalidated, and then invalidate them and stick them on the free_inc list - 61 * in either lru or fifo order. 62 */ 63 64 #include "bcache.h" 65 #include "btree.h" 66 67 #include <linux/blkdev.h> 68 #include <linux/kthread.h> 69 #include <linux/random.h> 70 #include <trace/events/bcache.h> 71 72 #define MAX_OPEN_BUCKETS 128 73 74 /* Bucket heap / gen */ 75 76 uint8_t bch_inc_gen(struct cache *ca, struct bucket *b) 77 { 78 uint8_t ret = ++b->gen; 79 80 ca->set->need_gc = max(ca->set->need_gc, bucket_gc_gen(b)); 81 WARN_ON_ONCE(ca->set->need_gc > BUCKET_GC_GEN_MAX); 82 83 return ret; 84 } 85 86 void bch_rescale_priorities(struct cache_set *c, int sectors) 87 { 88 struct cache *ca; 89 struct bucket *b; 90 unsigned long next = c->nbuckets * c->cache->sb.bucket_size / 1024; 91 int r; 92 93 atomic_sub(sectors, &c->rescale); 94 95 do { 96 r = atomic_read(&c->rescale); 97 98 if (r >= 0) 99 return; 100 } while (atomic_cmpxchg(&c->rescale, r, r + next) != r); 101 102 mutex_lock(&c->bucket_lock); 103 104 c->min_prio = USHRT_MAX; 105 106 ca = c->cache; 107 for_each_bucket(b, ca) 108 if (b->prio && 109 b->prio != BTREE_PRIO && 110 !atomic_read(&b->pin)) { 111 b->prio--; 112 c->min_prio = min(c->min_prio, b->prio); 113 } 114 115 mutex_unlock(&c->bucket_lock); 116 } 117 118 /* 119 * Background allocation thread: scans for buckets to be invalidated, 120 * invalidates them, rewrites prios/gens (marking them as invalidated on disk), 121 * then optionally issues discard commands to the newly free buckets, then puts 122 * them on the various freelists. 123 */ 124 125 static inline bool can_inc_bucket_gen(struct bucket *b) 126 { 127 return bucket_gc_gen(b) < BUCKET_GC_GEN_MAX; 128 } 129 130 bool bch_can_invalidate_bucket(struct cache *ca, struct bucket *b) 131 { 132 return (ca->set->gc_mark_valid || b->reclaimable_in_gc) && 133 ((!GC_MARK(b) || GC_MARK(b) == GC_MARK_RECLAIMABLE) && 134 !atomic_read(&b->pin) && can_inc_bucket_gen(b)); 135 } 136 137 void __bch_invalidate_one_bucket(struct cache *ca, struct bucket *b) 138 { 139 lockdep_assert_held(&ca->set->bucket_lock); 140 BUG_ON(GC_MARK(b) && GC_MARK(b) != GC_MARK_RECLAIMABLE); 141 142 if (GC_SECTORS_USED(b)) 143 trace_bcache_invalidate(ca, b - ca->buckets); 144 145 bch_inc_gen(ca, b); 146 b->prio = INITIAL_PRIO; 147 atomic_inc(&b->pin); 148 b->reclaimable_in_gc = 0; 149 } 150 151 static void bch_invalidate_one_bucket(struct cache *ca, struct bucket *b) 152 { 153 __bch_invalidate_one_bucket(ca, b); 154 155 fifo_push(&ca->free_inc, b - ca->buckets); 156 } 157 158 /* 159 * Determines what order we're going to reuse buckets, smallest bucket_prio() 160 * first: we also take into account the number of sectors of live data in that 161 * bucket, and in order for that multiply to make sense we have to scale bucket 162 * 163 * Thus, we scale the bucket priorities so that the bucket with the smallest 164 * prio is worth 1/8th of what INITIAL_PRIO is worth. 165 */ 166 167 static inline unsigned int new_bucket_prio(struct cache *ca, struct bucket *b) 168 { 169 unsigned int min_prio = (INITIAL_PRIO - ca->set->min_prio) / 8; 170 171 return (b->prio - ca->set->min_prio + min_prio) * GC_SECTORS_USED(b); 172 } 173 174 static inline bool new_bucket_max_cmp(const void *l, const void *r, void *args) 175 { 176 struct bucket **lhs = (struct bucket **)l; 177 struct bucket **rhs = (struct bucket **)r; 178 struct cache *ca = args; 179 180 return new_bucket_prio(ca, *lhs) > new_bucket_prio(ca, *rhs); 181 } 182 183 static inline bool new_bucket_min_cmp(const void *l, const void *r, void *args) 184 { 185 struct bucket **lhs = (struct bucket **)l; 186 struct bucket **rhs = (struct bucket **)r; 187 struct cache *ca = args; 188 189 return new_bucket_prio(ca, *lhs) < new_bucket_prio(ca, *rhs); 190 } 191 192 static void invalidate_buckets_lru(struct cache *ca) 193 { 194 struct bucket *b; 195 const struct min_heap_callbacks bucket_max_cmp_callback = { 196 .less = new_bucket_max_cmp, 197 .swp = NULL, 198 }; 199 const struct min_heap_callbacks bucket_min_cmp_callback = { 200 .less = new_bucket_min_cmp, 201 .swp = NULL, 202 }; 203 204 ca->heap.nr = 0; 205 206 for_each_bucket(b, ca) { 207 if (!bch_can_invalidate_bucket(ca, b)) 208 continue; 209 210 if (!min_heap_full(&ca->heap)) 211 min_heap_push(&ca->heap, &b, &bucket_max_cmp_callback, ca); 212 else if (!new_bucket_max_cmp(&b, min_heap_peek(&ca->heap), ca)) { 213 ca->heap.data[0] = b; 214 min_heap_sift_down(&ca->heap, 0, &bucket_max_cmp_callback, ca); 215 } 216 } 217 218 min_heapify_all(&ca->heap, &bucket_min_cmp_callback, ca); 219 220 while (!fifo_full(&ca->free_inc)) { 221 if (!ca->heap.nr) { 222 /* 223 * We don't want to be calling invalidate_buckets() 224 * multiple times when it can't do anything 225 */ 226 ca->invalidate_needs_gc = 1; 227 wake_up_gc(ca->set); 228 return; 229 } 230 b = min_heap_peek(&ca->heap)[0]; 231 min_heap_pop(&ca->heap, &bucket_min_cmp_callback, ca); 232 233 bch_invalidate_one_bucket(ca, b); 234 } 235 } 236 237 static void invalidate_buckets_fifo(struct cache *ca) 238 { 239 struct bucket *b; 240 size_t checked = 0; 241 242 while (!fifo_full(&ca->free_inc)) { 243 if (ca->fifo_last_bucket < ca->sb.first_bucket || 244 ca->fifo_last_bucket >= ca->sb.nbuckets) 245 ca->fifo_last_bucket = ca->sb.first_bucket; 246 247 b = ca->buckets + ca->fifo_last_bucket++; 248 249 if (bch_can_invalidate_bucket(ca, b)) 250 bch_invalidate_one_bucket(ca, b); 251 252 if (++checked >= ca->sb.nbuckets) { 253 ca->invalidate_needs_gc = 1; 254 wake_up_gc(ca->set); 255 return; 256 } 257 } 258 } 259 260 static void invalidate_buckets_random(struct cache *ca) 261 { 262 struct bucket *b; 263 size_t checked = 0; 264 265 while (!fifo_full(&ca->free_inc)) { 266 size_t n; 267 268 get_random_bytes(&n, sizeof(n)); 269 270 n %= (size_t) (ca->sb.nbuckets - ca->sb.first_bucket); 271 n += ca->sb.first_bucket; 272 273 b = ca->buckets + n; 274 275 if (bch_can_invalidate_bucket(ca, b)) 276 bch_invalidate_one_bucket(ca, b); 277 278 if (++checked >= ca->sb.nbuckets / 2) { 279 ca->invalidate_needs_gc = 1; 280 wake_up_gc(ca->set); 281 return; 282 } 283 } 284 } 285 286 static void invalidate_buckets(struct cache *ca) 287 { 288 BUG_ON(ca->invalidate_needs_gc); 289 290 switch (CACHE_REPLACEMENT(&ca->sb)) { 291 case CACHE_REPLACEMENT_LRU: 292 invalidate_buckets_lru(ca); 293 break; 294 case CACHE_REPLACEMENT_FIFO: 295 invalidate_buckets_fifo(ca); 296 break; 297 case CACHE_REPLACEMENT_RANDOM: 298 invalidate_buckets_random(ca); 299 break; 300 } 301 } 302 303 #define allocator_wait(ca, cond) \ 304 do { \ 305 while (1) { \ 306 set_current_state(TASK_INTERRUPTIBLE); \ 307 if (cond) \ 308 break; \ 309 \ 310 mutex_unlock(&(ca)->set->bucket_lock); \ 311 if (kthread_should_stop() || \ 312 test_bit(CACHE_SET_IO_DISABLE, &ca->set->flags)) { \ 313 set_current_state(TASK_RUNNING); \ 314 goto out; \ 315 } \ 316 \ 317 schedule(); \ 318 mutex_lock(&(ca)->set->bucket_lock); \ 319 } \ 320 __set_current_state(TASK_RUNNING); \ 321 } while (0) 322 323 static int bch_allocator_push(struct cache *ca, long bucket) 324 { 325 unsigned int i; 326 327 /* Prios/gens are actually the most important reserve */ 328 if (fifo_push(&ca->free[RESERVE_PRIO], bucket)) 329 return true; 330 331 for (i = 0; i < RESERVE_NR; i++) 332 if (fifo_push(&ca->free[i], bucket)) 333 return true; 334 335 return false; 336 } 337 338 static int bch_allocator_thread(void *arg) 339 { 340 struct cache *ca = arg; 341 342 mutex_lock(&ca->set->bucket_lock); 343 344 while (1) { 345 /* 346 * First, we pull buckets off of the unused and free_inc lists, 347 * possibly issue discards to them, then we add the bucket to 348 * the free list: 349 */ 350 while (1) { 351 long bucket; 352 353 if (!fifo_pop(&ca->free_inc, bucket)) 354 break; 355 356 if (ca->discard) { 357 mutex_unlock(&ca->set->bucket_lock); 358 blkdev_issue_discard(ca->bdev, 359 bucket_to_sector(ca->set, bucket), 360 ca->sb.bucket_size, GFP_KERNEL); 361 mutex_lock(&ca->set->bucket_lock); 362 } 363 364 allocator_wait(ca, bch_allocator_push(ca, bucket)); 365 wake_up(&ca->set->btree_cache_wait); 366 wake_up(&ca->set->bucket_wait); 367 } 368 369 /* 370 * We've run out of free buckets, we need to find some buckets 371 * we can invalidate. First, invalidate them in memory and add 372 * them to the free_inc list: 373 */ 374 375 retry_invalidate: 376 allocator_wait(ca, !ca->invalidate_needs_gc); 377 invalidate_buckets(ca); 378 379 /* 380 * Now, we write their new gens to disk so we can start writing 381 * new stuff to them: 382 */ 383 allocator_wait(ca, !atomic_read(&ca->set->prio_blocked)); 384 if (CACHE_SYNC(&ca->sb)) { 385 /* 386 * This could deadlock if an allocation with a btree 387 * node locked ever blocked - having the btree node 388 * locked would block garbage collection, but here we're 389 * waiting on garbage collection before we invalidate 390 * and free anything. 391 * 392 * But this should be safe since the btree code always 393 * uses btree_check_reserve() before allocating now, and 394 * if it fails it blocks without btree nodes locked. 395 */ 396 if (!fifo_full(&ca->free_inc)) 397 goto retry_invalidate; 398 399 if (bch_prio_write(ca, false) < 0) { 400 ca->invalidate_needs_gc = 1; 401 wake_up_gc(ca->set); 402 } 403 } 404 } 405 out: 406 wait_for_kthread_stop(); 407 return 0; 408 } 409 410 /* Allocation */ 411 412 long bch_bucket_alloc(struct cache *ca, unsigned int reserve, bool wait) 413 { 414 DEFINE_WAIT(w); 415 struct bucket *b; 416 long r; 417 418 419 /* No allocation if CACHE_SET_IO_DISABLE bit is set */ 420 if (unlikely(test_bit(CACHE_SET_IO_DISABLE, &ca->set->flags))) 421 return -1; 422 423 /* fastpath */ 424 if (fifo_pop(&ca->free[RESERVE_NONE], r) || 425 fifo_pop(&ca->free[reserve], r)) 426 goto out; 427 428 if (!wait) { 429 trace_bcache_alloc_fail(ca, reserve); 430 return -1; 431 } 432 433 do { 434 prepare_to_wait(&ca->set->bucket_wait, &w, 435 TASK_UNINTERRUPTIBLE); 436 437 mutex_unlock(&ca->set->bucket_lock); 438 schedule(); 439 mutex_lock(&ca->set->bucket_lock); 440 } while (!fifo_pop(&ca->free[RESERVE_NONE], r) && 441 !fifo_pop(&ca->free[reserve], r)); 442 443 finish_wait(&ca->set->bucket_wait, &w); 444 out: 445 if (ca->alloc_thread) 446 wake_up_process(ca->alloc_thread); 447 448 trace_bcache_alloc(ca, reserve); 449 450 if (expensive_debug_checks(ca->set)) { 451 size_t iter; 452 long i; 453 unsigned int j; 454 455 for (iter = 0; iter < prio_buckets(ca) * 2; iter++) 456 BUG_ON(ca->prio_buckets[iter] == (uint64_t) r); 457 458 for (j = 0; j < RESERVE_NR; j++) 459 fifo_for_each(i, &ca->free[j], iter) 460 BUG_ON(i == r); 461 fifo_for_each(i, &ca->free_inc, iter) 462 BUG_ON(i == r); 463 } 464 465 b = ca->buckets + r; 466 467 BUG_ON(atomic_read(&b->pin) != 1); 468 469 SET_GC_SECTORS_USED(b, ca->sb.bucket_size); 470 471 if (reserve <= RESERVE_PRIO) { 472 SET_GC_MARK(b, GC_MARK_METADATA); 473 SET_GC_MOVE(b, 0); 474 b->prio = BTREE_PRIO; 475 } else { 476 SET_GC_MARK(b, GC_MARK_RECLAIMABLE); 477 SET_GC_MOVE(b, 0); 478 b->prio = INITIAL_PRIO; 479 } 480 481 if (ca->set->avail_nbuckets > 0) { 482 ca->set->avail_nbuckets--; 483 bch_update_bucket_in_use(ca->set, &ca->set->gc_stats); 484 } 485 486 return r; 487 } 488 489 void __bch_bucket_free(struct cache *ca, struct bucket *b) 490 { 491 SET_GC_MARK(b, 0); 492 SET_GC_SECTORS_USED(b, 0); 493 494 if (ca->set->avail_nbuckets < ca->set->nbuckets) { 495 ca->set->avail_nbuckets++; 496 bch_update_bucket_in_use(ca->set, &ca->set->gc_stats); 497 } 498 } 499 500 void bch_bucket_free(struct cache_set *c, struct bkey *k) 501 { 502 unsigned int i; 503 504 for (i = 0; i < KEY_PTRS(k); i++) 505 __bch_bucket_free(c->cache, PTR_BUCKET(c, k, i)); 506 } 507 508 int __bch_bucket_alloc_set(struct cache_set *c, unsigned int reserve, 509 struct bkey *k, bool wait) 510 { 511 struct cache *ca; 512 long b; 513 514 /* No allocation if CACHE_SET_IO_DISABLE bit is set */ 515 if (unlikely(test_bit(CACHE_SET_IO_DISABLE, &c->flags))) 516 return -1; 517 518 lockdep_assert_held(&c->bucket_lock); 519 520 bkey_init(k); 521 522 ca = c->cache; 523 b = bch_bucket_alloc(ca, reserve, wait); 524 if (b < 0) 525 return -1; 526 527 k->ptr[0] = MAKE_PTR(ca->buckets[b].gen, 528 bucket_to_sector(c, b), 529 ca->sb.nr_this_dev); 530 531 SET_KEY_PTRS(k, 1); 532 533 return 0; 534 } 535 536 int bch_bucket_alloc_set(struct cache_set *c, unsigned int reserve, 537 struct bkey *k, bool wait) 538 { 539 int ret; 540 541 mutex_lock(&c->bucket_lock); 542 ret = __bch_bucket_alloc_set(c, reserve, k, wait); 543 mutex_unlock(&c->bucket_lock); 544 return ret; 545 } 546 547 /* Sector allocator */ 548 549 struct open_bucket { 550 struct list_head list; 551 unsigned int last_write_point; 552 unsigned int sectors_free; 553 BKEY_PADDED(key); 554 }; 555 556 /* 557 * We keep multiple buckets open for writes, and try to segregate different 558 * write streams for better cache utilization: first we try to segregate flash 559 * only volume write streams from cached devices, secondly we look for a bucket 560 * where the last write to it was sequential with the current write, and 561 * failing that we look for a bucket that was last used by the same task. 562 * 563 * The ideas is if you've got multiple tasks pulling data into the cache at the 564 * same time, you'll get better cache utilization if you try to segregate their 565 * data and preserve locality. 566 * 567 * For example, dirty sectors of flash only volume is not reclaimable, if their 568 * dirty sectors mixed with dirty sectors of cached device, such buckets will 569 * be marked as dirty and won't be reclaimed, though the dirty data of cached 570 * device have been written back to backend device. 571 * 572 * And say you've starting Firefox at the same time you're copying a 573 * bunch of files. Firefox will likely end up being fairly hot and stay in the 574 * cache awhile, but the data you copied might not be; if you wrote all that 575 * data to the same buckets it'd get invalidated at the same time. 576 * 577 * Both of those tasks will be doing fairly random IO so we can't rely on 578 * detecting sequential IO to segregate their data, but going off of the task 579 * should be a sane heuristic. 580 */ 581 static struct open_bucket *pick_data_bucket(struct cache_set *c, 582 const struct bkey *search, 583 unsigned int write_point, 584 struct bkey *alloc) 585 { 586 struct open_bucket *ret, *ret_task = NULL; 587 588 list_for_each_entry_reverse(ret, &c->data_buckets, list) 589 if (UUID_FLASH_ONLY(&c->uuids[KEY_INODE(&ret->key)]) != 590 UUID_FLASH_ONLY(&c->uuids[KEY_INODE(search)])) 591 continue; 592 else if (!bkey_cmp(&ret->key, search)) 593 goto found; 594 else if (ret->last_write_point == write_point) 595 ret_task = ret; 596 597 ret = ret_task ?: list_first_entry(&c->data_buckets, 598 struct open_bucket, list); 599 found: 600 if (!ret->sectors_free && KEY_PTRS(alloc)) { 601 ret->sectors_free = c->cache->sb.bucket_size; 602 bkey_copy(&ret->key, alloc); 603 bkey_init(alloc); 604 } 605 606 if (!ret->sectors_free) 607 ret = NULL; 608 609 return ret; 610 } 611 612 /* 613 * Allocates some space in the cache to write to, and k to point to the newly 614 * allocated space, and updates KEY_SIZE(k) and KEY_OFFSET(k) (to point to the 615 * end of the newly allocated space). 616 * 617 * May allocate fewer sectors than @sectors, KEY_SIZE(k) indicates how many 618 * sectors were actually allocated. 619 * 620 * If s->writeback is true, will not fail. 621 */ 622 bool bch_alloc_sectors(struct cache_set *c, 623 struct bkey *k, 624 unsigned int sectors, 625 unsigned int write_point, 626 unsigned int write_prio, 627 bool wait) 628 { 629 struct open_bucket *b; 630 BKEY_PADDED(key) alloc; 631 unsigned int i; 632 633 /* 634 * We might have to allocate a new bucket, which we can't do with a 635 * spinlock held. So if we have to allocate, we drop the lock, allocate 636 * and then retry. KEY_PTRS() indicates whether alloc points to 637 * allocated bucket(s). 638 */ 639 640 bkey_init(&alloc.key); 641 spin_lock(&c->data_bucket_lock); 642 643 while (!(b = pick_data_bucket(c, k, write_point, &alloc.key))) { 644 unsigned int watermark = write_prio 645 ? RESERVE_MOVINGGC 646 : RESERVE_NONE; 647 648 spin_unlock(&c->data_bucket_lock); 649 650 if (bch_bucket_alloc_set(c, watermark, &alloc.key, wait)) 651 return false; 652 653 spin_lock(&c->data_bucket_lock); 654 } 655 656 /* 657 * If we had to allocate, we might race and not need to allocate the 658 * second time we call pick_data_bucket(). If we allocated a bucket but 659 * didn't use it, drop the refcount bch_bucket_alloc_set() took: 660 */ 661 if (KEY_PTRS(&alloc.key)) 662 bkey_put(c, &alloc.key); 663 664 for (i = 0; i < KEY_PTRS(&b->key); i++) 665 EBUG_ON(ptr_stale(c, &b->key, i)); 666 667 /* Set up the pointer to the space we're allocating: */ 668 669 for (i = 0; i < KEY_PTRS(&b->key); i++) 670 k->ptr[i] = b->key.ptr[i]; 671 672 sectors = min(sectors, b->sectors_free); 673 674 SET_KEY_OFFSET(k, KEY_OFFSET(k) + sectors); 675 SET_KEY_SIZE(k, sectors); 676 SET_KEY_PTRS(k, KEY_PTRS(&b->key)); 677 678 /* 679 * Move b to the end of the lru, and keep track of what this bucket was 680 * last used for: 681 */ 682 list_move_tail(&b->list, &c->data_buckets); 683 bkey_copy_key(&b->key, k); 684 b->last_write_point = write_point; 685 686 b->sectors_free -= sectors; 687 688 for (i = 0; i < KEY_PTRS(&b->key); i++) { 689 SET_PTR_OFFSET(&b->key, i, PTR_OFFSET(&b->key, i) + sectors); 690 691 atomic_long_add(sectors, 692 &c->cache->sectors_written); 693 } 694 695 if (b->sectors_free < c->cache->sb.block_size) 696 b->sectors_free = 0; 697 698 /* 699 * k takes refcounts on the buckets it points to until it's inserted 700 * into the btree, but if we're done with this bucket we just transfer 701 * get_data_bucket()'s refcount. 702 */ 703 if (b->sectors_free) 704 for (i = 0; i < KEY_PTRS(&b->key); i++) 705 atomic_inc(&PTR_BUCKET(c, &b->key, i)->pin); 706 707 spin_unlock(&c->data_bucket_lock); 708 return true; 709 } 710 711 /* Init */ 712 713 void bch_open_buckets_free(struct cache_set *c) 714 { 715 struct open_bucket *b; 716 717 while (!list_empty(&c->data_buckets)) { 718 b = list_first_entry(&c->data_buckets, 719 struct open_bucket, list); 720 list_del(&b->list); 721 kfree(b); 722 } 723 } 724 725 int bch_open_buckets_alloc(struct cache_set *c) 726 { 727 int i; 728 729 spin_lock_init(&c->data_bucket_lock); 730 731 for (i = 0; i < MAX_OPEN_BUCKETS; i++) { 732 struct open_bucket *b = kzalloc(sizeof(*b), GFP_KERNEL); 733 734 if (!b) 735 return -ENOMEM; 736 737 list_add(&b->list, &c->data_buckets); 738 } 739 740 return 0; 741 } 742 743 int bch_cache_allocator_start(struct cache *ca) 744 { 745 struct task_struct *k = kthread_run(bch_allocator_thread, 746 ca, "bcache_allocator"); 747 if (IS_ERR(k)) 748 return PTR_ERR(k); 749 750 ca->alloc_thread = k; 751 return 0; 752 } 753