1 #ifndef _BCACHE_H 2 #define _BCACHE_H 3 4 /* 5 * SOME HIGH LEVEL CODE DOCUMENTATION: 6 * 7 * Bcache mostly works with cache sets, cache devices, and backing devices. 8 * 9 * Support for multiple cache devices hasn't quite been finished off yet, but 10 * it's about 95% plumbed through. A cache set and its cache devices is sort of 11 * like a md raid array and its component devices. Most of the code doesn't care 12 * about individual cache devices, the main abstraction is the cache set. 13 * 14 * Multiple cache devices is intended to give us the ability to mirror dirty 15 * cached data and metadata, without mirroring clean cached data. 16 * 17 * Backing devices are different, in that they have a lifetime independent of a 18 * cache set. When you register a newly formatted backing device it'll come up 19 * in passthrough mode, and then you can attach and detach a backing device from 20 * a cache set at runtime - while it's mounted and in use. Detaching implicitly 21 * invalidates any cached data for that backing device. 22 * 23 * A cache set can have multiple (many) backing devices attached to it. 24 * 25 * There's also flash only volumes - this is the reason for the distinction 26 * between struct cached_dev and struct bcache_device. A flash only volume 27 * works much like a bcache device that has a backing device, except the 28 * "cached" data is always dirty. The end result is that we get thin 29 * provisioning with very little additional code. 30 * 31 * Flash only volumes work but they're not production ready because the moving 32 * garbage collector needs more work. More on that later. 33 * 34 * BUCKETS/ALLOCATION: 35 * 36 * Bcache is primarily designed for caching, which means that in normal 37 * operation all of our available space will be allocated. Thus, we need an 38 * efficient way of deleting things from the cache so we can write new things to 39 * it. 40 * 41 * To do this, we first divide the cache device up into buckets. A bucket is the 42 * unit of allocation; they're typically around 1 mb - anywhere from 128k to 2M+ 43 * works efficiently. 44 * 45 * Each bucket has a 16 bit priority, and an 8 bit generation associated with 46 * it. The gens and priorities for all the buckets are stored contiguously and 47 * packed on disk (in a linked list of buckets - aside from the superblock, all 48 * of bcache's metadata is stored in buckets). 49 * 50 * The priority is used to implement an LRU. We reset a bucket's priority when 51 * we allocate it or on cache it, and every so often we decrement the priority 52 * of each bucket. It could be used to implement something more sophisticated, 53 * if anyone ever gets around to it. 54 * 55 * The generation is used for invalidating buckets. Each pointer also has an 8 56 * bit generation embedded in it; for a pointer to be considered valid, its gen 57 * must match the gen of the bucket it points into. Thus, to reuse a bucket all 58 * we have to do is increment its gen (and write its new gen to disk; we batch 59 * this up). 60 * 61 * Bcache is entirely COW - we never write twice to a bucket, even buckets that 62 * contain metadata (including btree nodes). 63 * 64 * THE BTREE: 65 * 66 * Bcache is in large part design around the btree. 67 * 68 * At a high level, the btree is just an index of key -> ptr tuples. 69 * 70 * Keys represent extents, and thus have a size field. Keys also have a variable 71 * number of pointers attached to them (potentially zero, which is handy for 72 * invalidating the cache). 73 * 74 * The key itself is an inode:offset pair. The inode number corresponds to a 75 * backing device or a flash only volume. The offset is the ending offset of the 76 * extent within the inode - not the starting offset; this makes lookups 77 * slightly more convenient. 78 * 79 * Pointers contain the cache device id, the offset on that device, and an 8 bit 80 * generation number. More on the gen later. 81 * 82 * Index lookups are not fully abstracted - cache lookups in particular are 83 * still somewhat mixed in with the btree code, but things are headed in that 84 * direction. 85 * 86 * Updates are fairly well abstracted, though. There are two different ways of 87 * updating the btree; insert and replace. 88 * 89 * BTREE_INSERT will just take a list of keys and insert them into the btree - 90 * overwriting (possibly only partially) any extents they overlap with. This is 91 * used to update the index after a write. 92 * 93 * BTREE_REPLACE is really cmpxchg(); it inserts a key into the btree iff it is 94 * overwriting a key that matches another given key. This is used for inserting 95 * data into the cache after a cache miss, and for background writeback, and for 96 * the moving garbage collector. 97 * 98 * There is no "delete" operation; deleting things from the index is 99 * accomplished by either by invalidating pointers (by incrementing a bucket's 100 * gen) or by inserting a key with 0 pointers - which will overwrite anything 101 * previously present at that location in the index. 102 * 103 * This means that there are always stale/invalid keys in the btree. They're 104 * filtered out by the code that iterates through a btree node, and removed when 105 * a btree node is rewritten. 106 * 107 * BTREE NODES: 108 * 109 * Our unit of allocation is a bucket, and we we can't arbitrarily allocate and 110 * free smaller than a bucket - so, that's how big our btree nodes are. 111 * 112 * (If buckets are really big we'll only use part of the bucket for a btree node 113 * - no less than 1/4th - but a bucket still contains no more than a single 114 * btree node. I'd actually like to change this, but for now we rely on the 115 * bucket's gen for deleting btree nodes when we rewrite/split a node.) 116 * 117 * Anyways, btree nodes are big - big enough to be inefficient with a textbook 118 * btree implementation. 119 * 120 * The way this is solved is that btree nodes are internally log structured; we 121 * can append new keys to an existing btree node without rewriting it. This 122 * means each set of keys we write is sorted, but the node is not. 123 * 124 * We maintain this log structure in memory - keeping 1Mb of keys sorted would 125 * be expensive, and we have to distinguish between the keys we have written and 126 * the keys we haven't. So to do a lookup in a btree node, we have to search 127 * each sorted set. But we do merge written sets together lazily, so the cost of 128 * these extra searches is quite low (normally most of the keys in a btree node 129 * will be in one big set, and then there'll be one or two sets that are much 130 * smaller). 131 * 132 * This log structure makes bcache's btree more of a hybrid between a 133 * conventional btree and a compacting data structure, with some of the 134 * advantages of both. 135 * 136 * GARBAGE COLLECTION: 137 * 138 * We can't just invalidate any bucket - it might contain dirty data or 139 * metadata. If it once contained dirty data, other writes might overwrite it 140 * later, leaving no valid pointers into that bucket in the index. 141 * 142 * Thus, the primary purpose of garbage collection is to find buckets to reuse. 143 * It also counts how much valid data it each bucket currently contains, so that 144 * allocation can reuse buckets sooner when they've been mostly overwritten. 145 * 146 * It also does some things that are really internal to the btree 147 * implementation. If a btree node contains pointers that are stale by more than 148 * some threshold, it rewrites the btree node to avoid the bucket's generation 149 * wrapping around. It also merges adjacent btree nodes if they're empty enough. 150 * 151 * THE JOURNAL: 152 * 153 * Bcache's journal is not necessary for consistency; we always strictly 154 * order metadata writes so that the btree and everything else is consistent on 155 * disk in the event of an unclean shutdown, and in fact bcache had writeback 156 * caching (with recovery from unclean shutdown) before journalling was 157 * implemented. 158 * 159 * Rather, the journal is purely a performance optimization; we can't complete a 160 * write until we've updated the index on disk, otherwise the cache would be 161 * inconsistent in the event of an unclean shutdown. This means that without the 162 * journal, on random write workloads we constantly have to update all the leaf 163 * nodes in the btree, and those writes will be mostly empty (appending at most 164 * a few keys each) - highly inefficient in terms of amount of metadata writes, 165 * and it puts more strain on the various btree resorting/compacting code. 166 * 167 * The journal is just a log of keys we've inserted; on startup we just reinsert 168 * all the keys in the open journal entries. That means that when we're updating 169 * a node in the btree, we can wait until a 4k block of keys fills up before 170 * writing them out. 171 * 172 * For simplicity, we only journal updates to leaf nodes; updates to parent 173 * nodes are rare enough (since our leaf nodes are huge) that it wasn't worth 174 * the complexity to deal with journalling them (in particular, journal replay) 175 * - updates to non leaf nodes just happen synchronously (see btree_split()). 176 */ 177 178 #define pr_fmt(fmt) "bcache: %s() " fmt "\n", __func__ 179 180 #include <linux/bcache.h> 181 #include <linux/bio.h> 182 #include <linux/kobject.h> 183 #include <linux/list.h> 184 #include <linux/mutex.h> 185 #include <linux/rbtree.h> 186 #include <linux/rwsem.h> 187 #include <linux/types.h> 188 #include <linux/workqueue.h> 189 190 #include "bset.h" 191 #include "util.h" 192 #include "closure.h" 193 194 struct bucket { 195 atomic_t pin; 196 uint16_t prio; 197 uint8_t gen; 198 uint8_t last_gc; /* Most out of date gen in the btree */ 199 uint16_t gc_mark; /* Bitfield used by GC. See below for field */ 200 }; 201 202 /* 203 * I'd use bitfields for these, but I don't trust the compiler not to screw me 204 * as multiple threads touch struct bucket without locking 205 */ 206 207 BITMASK(GC_MARK, struct bucket, gc_mark, 0, 2); 208 #define GC_MARK_RECLAIMABLE 1 209 #define GC_MARK_DIRTY 2 210 #define GC_MARK_METADATA 3 211 #define GC_SECTORS_USED_SIZE 13 212 #define MAX_GC_SECTORS_USED (~(~0ULL << GC_SECTORS_USED_SIZE)) 213 BITMASK(GC_SECTORS_USED, struct bucket, gc_mark, 2, GC_SECTORS_USED_SIZE); 214 BITMASK(GC_MOVE, struct bucket, gc_mark, 15, 1); 215 216 #include "journal.h" 217 #include "stats.h" 218 struct search; 219 struct btree; 220 struct keybuf; 221 222 struct keybuf_key { 223 struct rb_node node; 224 BKEY_PADDED(key); 225 void *private; 226 }; 227 228 struct keybuf { 229 struct bkey last_scanned; 230 spinlock_t lock; 231 232 /* 233 * Beginning and end of range in rb tree - so that we can skip taking 234 * lock and checking the rb tree when we need to check for overlapping 235 * keys. 236 */ 237 struct bkey start; 238 struct bkey end; 239 240 struct rb_root keys; 241 242 #define KEYBUF_NR 500 243 DECLARE_ARRAY_ALLOCATOR(struct keybuf_key, freelist, KEYBUF_NR); 244 }; 245 246 struct bcache_device { 247 struct closure cl; 248 249 struct kobject kobj; 250 251 struct cache_set *c; 252 unsigned id; 253 #define BCACHEDEVNAME_SIZE 12 254 char name[BCACHEDEVNAME_SIZE]; 255 256 struct gendisk *disk; 257 258 unsigned long flags; 259 #define BCACHE_DEV_CLOSING 0 260 #define BCACHE_DEV_DETACHING 1 261 #define BCACHE_DEV_UNLINK_DONE 2 262 263 unsigned nr_stripes; 264 unsigned stripe_size; 265 atomic_t *stripe_sectors_dirty; 266 unsigned long *full_dirty_stripes; 267 268 unsigned long sectors_dirty_last; 269 long sectors_dirty_derivative; 270 271 struct bio_set *bio_split; 272 273 unsigned data_csum:1; 274 275 int (*cache_miss)(struct btree *, struct search *, 276 struct bio *, unsigned); 277 int (*ioctl) (struct bcache_device *, fmode_t, unsigned, unsigned long); 278 }; 279 280 struct io { 281 /* Used to track sequential IO so it can be skipped */ 282 struct hlist_node hash; 283 struct list_head lru; 284 285 unsigned long jiffies; 286 unsigned sequential; 287 sector_t last; 288 }; 289 290 struct cached_dev { 291 struct list_head list; 292 struct bcache_device disk; 293 struct block_device *bdev; 294 295 struct cache_sb sb; 296 struct bio sb_bio; 297 struct bio_vec sb_bv[1]; 298 struct closure sb_write; 299 struct semaphore sb_write_mutex; 300 301 /* Refcount on the cache set. Always nonzero when we're caching. */ 302 atomic_t count; 303 struct work_struct detach; 304 305 /* 306 * Device might not be running if it's dirty and the cache set hasn't 307 * showed up yet. 308 */ 309 atomic_t running; 310 311 /* 312 * Writes take a shared lock from start to finish; scanning for dirty 313 * data to refill the rb tree requires an exclusive lock. 314 */ 315 struct rw_semaphore writeback_lock; 316 317 /* 318 * Nonzero, and writeback has a refcount (d->count), iff there is dirty 319 * data in the cache. Protected by writeback_lock; must have an 320 * shared lock to set and exclusive lock to clear. 321 */ 322 atomic_t has_dirty; 323 324 struct bch_ratelimit writeback_rate; 325 struct delayed_work writeback_rate_update; 326 327 /* 328 * Internal to the writeback code, so read_dirty() can keep track of 329 * where it's at. 330 */ 331 sector_t last_read; 332 333 /* Limit number of writeback bios in flight */ 334 struct semaphore in_flight; 335 struct task_struct *writeback_thread; 336 struct workqueue_struct *writeback_write_wq; 337 338 struct keybuf writeback_keys; 339 340 /* For tracking sequential IO */ 341 #define RECENT_IO_BITS 7 342 #define RECENT_IO (1 << RECENT_IO_BITS) 343 struct io io[RECENT_IO]; 344 struct hlist_head io_hash[RECENT_IO + 1]; 345 struct list_head io_lru; 346 spinlock_t io_lock; 347 348 struct cache_accounting accounting; 349 350 /* The rest of this all shows up in sysfs */ 351 unsigned sequential_cutoff; 352 unsigned readahead; 353 354 unsigned verify:1; 355 unsigned bypass_torture_test:1; 356 357 unsigned partial_stripes_expensive:1; 358 unsigned writeback_metadata:1; 359 unsigned writeback_running:1; 360 unsigned char writeback_percent; 361 unsigned writeback_delay; 362 363 uint64_t writeback_rate_target; 364 int64_t writeback_rate_proportional; 365 int64_t writeback_rate_derivative; 366 int64_t writeback_rate_change; 367 368 unsigned writeback_rate_update_seconds; 369 unsigned writeback_rate_d_term; 370 unsigned writeback_rate_p_term_inverse; 371 }; 372 373 enum alloc_reserve { 374 RESERVE_BTREE, 375 RESERVE_PRIO, 376 RESERVE_MOVINGGC, 377 RESERVE_NONE, 378 RESERVE_NR, 379 }; 380 381 struct cache { 382 struct cache_set *set; 383 struct cache_sb sb; 384 struct bio sb_bio; 385 struct bio_vec sb_bv[1]; 386 387 struct kobject kobj; 388 struct block_device *bdev; 389 390 struct task_struct *alloc_thread; 391 392 struct closure prio; 393 struct prio_set *disk_buckets; 394 395 /* 396 * When allocating new buckets, prio_write() gets first dibs - since we 397 * may not be allocate at all without writing priorities and gens. 398 * prio_buckets[] contains the last buckets we wrote priorities to (so 399 * gc can mark them as metadata), prio_next[] contains the buckets 400 * allocated for the next prio write. 401 */ 402 uint64_t *prio_buckets; 403 uint64_t *prio_last_buckets; 404 405 /* 406 * free: Buckets that are ready to be used 407 * 408 * free_inc: Incoming buckets - these are buckets that currently have 409 * cached data in them, and we can't reuse them until after we write 410 * their new gen to disk. After prio_write() finishes writing the new 411 * gens/prios, they'll be moved to the free list (and possibly discarded 412 * in the process) 413 */ 414 DECLARE_FIFO(long, free)[RESERVE_NR]; 415 DECLARE_FIFO(long, free_inc); 416 417 size_t fifo_last_bucket; 418 419 /* Allocation stuff: */ 420 struct bucket *buckets; 421 422 DECLARE_HEAP(struct bucket *, heap); 423 424 /* 425 * If nonzero, we know we aren't going to find any buckets to invalidate 426 * until a gc finishes - otherwise we could pointlessly burn a ton of 427 * cpu 428 */ 429 unsigned invalidate_needs_gc; 430 431 bool discard; /* Get rid of? */ 432 433 struct journal_device journal; 434 435 /* The rest of this all shows up in sysfs */ 436 #define IO_ERROR_SHIFT 20 437 atomic_t io_errors; 438 atomic_t io_count; 439 440 atomic_long_t meta_sectors_written; 441 atomic_long_t btree_sectors_written; 442 atomic_long_t sectors_written; 443 }; 444 445 struct gc_stat { 446 size_t nodes; 447 size_t key_bytes; 448 449 size_t nkeys; 450 uint64_t data; /* sectors */ 451 unsigned in_use; /* percent */ 452 }; 453 454 /* 455 * Flag bits, for how the cache set is shutting down, and what phase it's at: 456 * 457 * CACHE_SET_UNREGISTERING means we're not just shutting down, we're detaching 458 * all the backing devices first (their cached data gets invalidated, and they 459 * won't automatically reattach). 460 * 461 * CACHE_SET_STOPPING always gets set first when we're closing down a cache set; 462 * we'll continue to run normally for awhile with CACHE_SET_STOPPING set (i.e. 463 * flushing dirty data). 464 * 465 * CACHE_SET_RUNNING means all cache devices have been registered and journal 466 * replay is complete. 467 */ 468 #define CACHE_SET_UNREGISTERING 0 469 #define CACHE_SET_STOPPING 1 470 #define CACHE_SET_RUNNING 2 471 472 struct cache_set { 473 struct closure cl; 474 475 struct list_head list; 476 struct kobject kobj; 477 struct kobject internal; 478 struct dentry *debug; 479 struct cache_accounting accounting; 480 481 unsigned long flags; 482 483 struct cache_sb sb; 484 485 struct cache *cache[MAX_CACHES_PER_SET]; 486 struct cache *cache_by_alloc[MAX_CACHES_PER_SET]; 487 int caches_loaded; 488 489 struct bcache_device **devices; 490 struct list_head cached_devs; 491 uint64_t cached_dev_sectors; 492 struct closure caching; 493 494 struct closure sb_write; 495 struct semaphore sb_write_mutex; 496 497 mempool_t *search; 498 mempool_t *bio_meta; 499 struct bio_set *bio_split; 500 501 /* For the btree cache */ 502 struct shrinker shrink; 503 504 /* For the btree cache and anything allocation related */ 505 struct mutex bucket_lock; 506 507 /* log2(bucket_size), in sectors */ 508 unsigned short bucket_bits; 509 510 /* log2(block_size), in sectors */ 511 unsigned short block_bits; 512 513 /* 514 * Default number of pages for a new btree node - may be less than a 515 * full bucket 516 */ 517 unsigned btree_pages; 518 519 /* 520 * Lists of struct btrees; lru is the list for structs that have memory 521 * allocated for actual btree node, freed is for structs that do not. 522 * 523 * We never free a struct btree, except on shutdown - we just put it on 524 * the btree_cache_freed list and reuse it later. This simplifies the 525 * code, and it doesn't cost us much memory as the memory usage is 526 * dominated by buffers that hold the actual btree node data and those 527 * can be freed - and the number of struct btrees allocated is 528 * effectively bounded. 529 * 530 * btree_cache_freeable effectively is a small cache - we use it because 531 * high order page allocations can be rather expensive, and it's quite 532 * common to delete and allocate btree nodes in quick succession. It 533 * should never grow past ~2-3 nodes in practice. 534 */ 535 struct list_head btree_cache; 536 struct list_head btree_cache_freeable; 537 struct list_head btree_cache_freed; 538 539 /* Number of elements in btree_cache + btree_cache_freeable lists */ 540 unsigned btree_cache_used; 541 542 /* 543 * If we need to allocate memory for a new btree node and that 544 * allocation fails, we can cannibalize another node in the btree cache 545 * to satisfy the allocation - lock to guarantee only one thread does 546 * this at a time: 547 */ 548 wait_queue_head_t btree_cache_wait; 549 struct task_struct *btree_cache_alloc_lock; 550 551 /* 552 * When we free a btree node, we increment the gen of the bucket the 553 * node is in - but we can't rewrite the prios and gens until we 554 * finished whatever it is we were doing, otherwise after a crash the 555 * btree node would be freed but for say a split, we might not have the 556 * pointers to the new nodes inserted into the btree yet. 557 * 558 * This is a refcount that blocks prio_write() until the new keys are 559 * written. 560 */ 561 atomic_t prio_blocked; 562 wait_queue_head_t bucket_wait; 563 564 /* 565 * For any bio we don't skip we subtract the number of sectors from 566 * rescale; when it hits 0 we rescale all the bucket priorities. 567 */ 568 atomic_t rescale; 569 /* 570 * When we invalidate buckets, we use both the priority and the amount 571 * of good data to determine which buckets to reuse first - to weight 572 * those together consistently we keep track of the smallest nonzero 573 * priority of any bucket. 574 */ 575 uint16_t min_prio; 576 577 /* 578 * max(gen - last_gc) for all buckets. When it gets too big we have to gc 579 * to keep gens from wrapping around. 580 */ 581 uint8_t need_gc; 582 struct gc_stat gc_stats; 583 size_t nbuckets; 584 585 struct task_struct *gc_thread; 586 /* Where in the btree gc currently is */ 587 struct bkey gc_done; 588 589 /* 590 * The allocation code needs gc_mark in struct bucket to be correct, but 591 * it's not while a gc is in progress. Protected by bucket_lock. 592 */ 593 int gc_mark_valid; 594 595 /* Counts how many sectors bio_insert has added to the cache */ 596 atomic_t sectors_to_gc; 597 wait_queue_head_t gc_wait; 598 599 struct keybuf moving_gc_keys; 600 /* Number of moving GC bios in flight */ 601 struct semaphore moving_in_flight; 602 603 struct workqueue_struct *moving_gc_wq; 604 605 struct btree *root; 606 607 #ifdef CONFIG_BCACHE_DEBUG 608 struct btree *verify_data; 609 struct bset *verify_ondisk; 610 struct mutex verify_lock; 611 #endif 612 613 unsigned nr_uuids; 614 struct uuid_entry *uuids; 615 BKEY_PADDED(uuid_bucket); 616 struct closure uuid_write; 617 struct semaphore uuid_write_mutex; 618 619 /* 620 * A btree node on disk could have too many bsets for an iterator to fit 621 * on the stack - have to dynamically allocate them 622 */ 623 mempool_t *fill_iter; 624 625 struct bset_sort_state sort; 626 627 /* List of buckets we're currently writing data to */ 628 struct list_head data_buckets; 629 spinlock_t data_bucket_lock; 630 631 struct journal journal; 632 633 #define CONGESTED_MAX 1024 634 unsigned congested_last_us; 635 atomic_t congested; 636 637 /* The rest of this all shows up in sysfs */ 638 unsigned congested_read_threshold_us; 639 unsigned congested_write_threshold_us; 640 641 struct time_stats btree_gc_time; 642 struct time_stats btree_split_time; 643 struct time_stats btree_read_time; 644 645 atomic_long_t cache_read_races; 646 atomic_long_t writeback_keys_done; 647 atomic_long_t writeback_keys_failed; 648 649 enum { 650 ON_ERROR_UNREGISTER, 651 ON_ERROR_PANIC, 652 } on_error; 653 unsigned error_limit; 654 unsigned error_decay; 655 656 unsigned short journal_delay_ms; 657 bool expensive_debug_checks; 658 unsigned verify:1; 659 unsigned key_merging_disabled:1; 660 unsigned gc_always_rewrite:1; 661 unsigned shrinker_disabled:1; 662 unsigned copy_gc_enabled:1; 663 664 #define BUCKET_HASH_BITS 12 665 struct hlist_head bucket_hash[1 << BUCKET_HASH_BITS]; 666 }; 667 668 struct bbio { 669 unsigned submit_time_us; 670 union { 671 struct bkey key; 672 uint64_t _pad[3]; 673 /* 674 * We only need pad = 3 here because we only ever carry around a 675 * single pointer - i.e. the pointer we're doing io to/from. 676 */ 677 }; 678 struct bio bio; 679 }; 680 681 #define BTREE_PRIO USHRT_MAX 682 #define INITIAL_PRIO 32768U 683 684 #define btree_bytes(c) ((c)->btree_pages * PAGE_SIZE) 685 #define btree_blocks(b) \ 686 ((unsigned) (KEY_SIZE(&b->key) >> (b)->c->block_bits)) 687 688 #define btree_default_blocks(c) \ 689 ((unsigned) ((PAGE_SECTORS * (c)->btree_pages) >> (c)->block_bits)) 690 691 #define bucket_pages(c) ((c)->sb.bucket_size / PAGE_SECTORS) 692 #define bucket_bytes(c) ((c)->sb.bucket_size << 9) 693 #define block_bytes(c) ((c)->sb.block_size << 9) 694 695 #define prios_per_bucket(c) \ 696 ((bucket_bytes(c) - sizeof(struct prio_set)) / \ 697 sizeof(struct bucket_disk)) 698 #define prio_buckets(c) \ 699 DIV_ROUND_UP((size_t) (c)->sb.nbuckets, prios_per_bucket(c)) 700 701 static inline size_t sector_to_bucket(struct cache_set *c, sector_t s) 702 { 703 return s >> c->bucket_bits; 704 } 705 706 static inline sector_t bucket_to_sector(struct cache_set *c, size_t b) 707 { 708 return ((sector_t) b) << c->bucket_bits; 709 } 710 711 static inline sector_t bucket_remainder(struct cache_set *c, sector_t s) 712 { 713 return s & (c->sb.bucket_size - 1); 714 } 715 716 static inline struct cache *PTR_CACHE(struct cache_set *c, 717 const struct bkey *k, 718 unsigned ptr) 719 { 720 return c->cache[PTR_DEV(k, ptr)]; 721 } 722 723 static inline size_t PTR_BUCKET_NR(struct cache_set *c, 724 const struct bkey *k, 725 unsigned ptr) 726 { 727 return sector_to_bucket(c, PTR_OFFSET(k, ptr)); 728 } 729 730 static inline struct bucket *PTR_BUCKET(struct cache_set *c, 731 const struct bkey *k, 732 unsigned ptr) 733 { 734 return PTR_CACHE(c, k, ptr)->buckets + PTR_BUCKET_NR(c, k, ptr); 735 } 736 737 static inline uint8_t gen_after(uint8_t a, uint8_t b) 738 { 739 uint8_t r = a - b; 740 return r > 128U ? 0 : r; 741 } 742 743 static inline uint8_t ptr_stale(struct cache_set *c, const struct bkey *k, 744 unsigned i) 745 { 746 return gen_after(PTR_BUCKET(c, k, i)->gen, PTR_GEN(k, i)); 747 } 748 749 static inline bool ptr_available(struct cache_set *c, const struct bkey *k, 750 unsigned i) 751 { 752 return (PTR_DEV(k, i) < MAX_CACHES_PER_SET) && PTR_CACHE(c, k, i); 753 } 754 755 /* Btree key macros */ 756 757 /* 758 * This is used for various on disk data structures - cache_sb, prio_set, bset, 759 * jset: The checksum is _always_ the first 8 bytes of these structs 760 */ 761 #define csum_set(i) \ 762 bch_crc64(((void *) (i)) + sizeof(uint64_t), \ 763 ((void *) bset_bkey_last(i)) - \ 764 (((void *) (i)) + sizeof(uint64_t))) 765 766 /* Error handling macros */ 767 768 #define btree_bug(b, ...) \ 769 do { \ 770 if (bch_cache_set_error((b)->c, __VA_ARGS__)) \ 771 dump_stack(); \ 772 } while (0) 773 774 #define cache_bug(c, ...) \ 775 do { \ 776 if (bch_cache_set_error(c, __VA_ARGS__)) \ 777 dump_stack(); \ 778 } while (0) 779 780 #define btree_bug_on(cond, b, ...) \ 781 do { \ 782 if (cond) \ 783 btree_bug(b, __VA_ARGS__); \ 784 } while (0) 785 786 #define cache_bug_on(cond, c, ...) \ 787 do { \ 788 if (cond) \ 789 cache_bug(c, __VA_ARGS__); \ 790 } while (0) 791 792 #define cache_set_err_on(cond, c, ...) \ 793 do { \ 794 if (cond) \ 795 bch_cache_set_error(c, __VA_ARGS__); \ 796 } while (0) 797 798 /* Looping macros */ 799 800 #define for_each_cache(ca, cs, iter) \ 801 for (iter = 0; ca = cs->cache[iter], iter < (cs)->sb.nr_in_set; iter++) 802 803 #define for_each_bucket(b, ca) \ 804 for (b = (ca)->buckets + (ca)->sb.first_bucket; \ 805 b < (ca)->buckets + (ca)->sb.nbuckets; b++) 806 807 static inline void cached_dev_put(struct cached_dev *dc) 808 { 809 if (atomic_dec_and_test(&dc->count)) 810 schedule_work(&dc->detach); 811 } 812 813 static inline bool cached_dev_get(struct cached_dev *dc) 814 { 815 if (!atomic_inc_not_zero(&dc->count)) 816 return false; 817 818 /* Paired with the mb in cached_dev_attach */ 819 smp_mb__after_atomic(); 820 return true; 821 } 822 823 /* 824 * bucket_gc_gen() returns the difference between the bucket's current gen and 825 * the oldest gen of any pointer into that bucket in the btree (last_gc). 826 */ 827 828 static inline uint8_t bucket_gc_gen(struct bucket *b) 829 { 830 return b->gen - b->last_gc; 831 } 832 833 #define BUCKET_GC_GEN_MAX 96U 834 835 #define kobj_attribute_write(n, fn) \ 836 static struct kobj_attribute ksysfs_##n = __ATTR(n, S_IWUSR, NULL, fn) 837 838 #define kobj_attribute_rw(n, show, store) \ 839 static struct kobj_attribute ksysfs_##n = \ 840 __ATTR(n, S_IWUSR|S_IRUSR, show, store) 841 842 static inline void wake_up_allocators(struct cache_set *c) 843 { 844 struct cache *ca; 845 unsigned i; 846 847 for_each_cache(ca, c, i) 848 wake_up_process(ca->alloc_thread); 849 } 850 851 /* Forward declarations */ 852 853 void bch_count_io_errors(struct cache *, blk_status_t, const char *); 854 void bch_bbio_count_io_errors(struct cache_set *, struct bio *, 855 blk_status_t, const char *); 856 void bch_bbio_endio(struct cache_set *, struct bio *, blk_status_t, 857 const char *); 858 void bch_bbio_free(struct bio *, struct cache_set *); 859 struct bio *bch_bbio_alloc(struct cache_set *); 860 861 void __bch_submit_bbio(struct bio *, struct cache_set *); 862 void bch_submit_bbio(struct bio *, struct cache_set *, struct bkey *, unsigned); 863 864 uint8_t bch_inc_gen(struct cache *, struct bucket *); 865 void bch_rescale_priorities(struct cache_set *, int); 866 867 bool bch_can_invalidate_bucket(struct cache *, struct bucket *); 868 void __bch_invalidate_one_bucket(struct cache *, struct bucket *); 869 870 void __bch_bucket_free(struct cache *, struct bucket *); 871 void bch_bucket_free(struct cache_set *, struct bkey *); 872 873 long bch_bucket_alloc(struct cache *, unsigned, bool); 874 int __bch_bucket_alloc_set(struct cache_set *, unsigned, 875 struct bkey *, int, bool); 876 int bch_bucket_alloc_set(struct cache_set *, unsigned, 877 struct bkey *, int, bool); 878 bool bch_alloc_sectors(struct cache_set *, struct bkey *, unsigned, 879 unsigned, unsigned, bool); 880 881 __printf(2, 3) 882 bool bch_cache_set_error(struct cache_set *, const char *, ...); 883 884 void bch_prio_write(struct cache *); 885 void bch_write_bdev_super(struct cached_dev *, struct closure *); 886 887 extern struct workqueue_struct *bcache_wq; 888 extern const char * const bch_cache_modes[]; 889 extern struct mutex bch_register_lock; 890 extern struct list_head bch_cache_sets; 891 892 extern struct kobj_type bch_cached_dev_ktype; 893 extern struct kobj_type bch_flash_dev_ktype; 894 extern struct kobj_type bch_cache_set_ktype; 895 extern struct kobj_type bch_cache_set_internal_ktype; 896 extern struct kobj_type bch_cache_ktype; 897 898 void bch_cached_dev_release(struct kobject *); 899 void bch_flash_dev_release(struct kobject *); 900 void bch_cache_set_release(struct kobject *); 901 void bch_cache_release(struct kobject *); 902 903 int bch_uuid_write(struct cache_set *); 904 void bcache_write_super(struct cache_set *); 905 906 int bch_flash_dev_create(struct cache_set *c, uint64_t size); 907 908 int bch_cached_dev_attach(struct cached_dev *, struct cache_set *); 909 void bch_cached_dev_detach(struct cached_dev *); 910 void bch_cached_dev_run(struct cached_dev *); 911 void bcache_device_stop(struct bcache_device *); 912 913 void bch_cache_set_unregister(struct cache_set *); 914 void bch_cache_set_stop(struct cache_set *); 915 916 struct cache_set *bch_cache_set_alloc(struct cache_sb *); 917 void bch_btree_cache_free(struct cache_set *); 918 int bch_btree_cache_alloc(struct cache_set *); 919 void bch_moving_init_cache_set(struct cache_set *); 920 int bch_open_buckets_alloc(struct cache_set *); 921 void bch_open_buckets_free(struct cache_set *); 922 923 int bch_cache_allocator_start(struct cache *ca); 924 925 void bch_debug_exit(void); 926 int bch_debug_init(struct kobject *); 927 void bch_request_exit(void); 928 int bch_request_init(void); 929 930 #endif /* _BCACHE_H */ 931