1 /* SPDX-License-Identifier: GPL-2.0 */
6 * SOME HIGH LEVEL CODE DOCUMENTATION:
8 * Bcache mostly works with cache sets, cache devices, and backing devices.
10 * Support for multiple cache devices hasn't quite been finished off yet, but
11 * it's about 95% plumbed through. A cache set and its cache devices is sort of
12 * like a md raid array and its component devices. Most of the code doesn't care
13 * about individual cache devices, the main abstraction is the cache set.
15 * Multiple cache devices is intended to give us the ability to mirror dirty
16 * cached data and metadata, without mirroring clean cached data.
18 * Backing devices are different, in that they have a lifetime independent of a
19 * cache set. When you register a newly formatted backing device it'll come up
20 * in passthrough mode, and then you can attach and detach a backing device from
21 * a cache set at runtime - while it's mounted and in use. Detaching implicitly
22 * invalidates any cached data for that backing device.
24 * A cache set can have multiple (many) backing devices attached to it.
26 * There's also flash only volumes - this is the reason for the distinction
27 * between struct cached_dev and struct bcache_device. A flash only volume
28 * works much like a bcache device that has a backing device, except the
29 * "cached" data is always dirty. The end result is that we get thin
30 * provisioning with very little additional code.
32 * Flash only volumes work but they're not production ready because the moving
33 * garbage collector needs more work. More on that later.
37 * Bcache is primarily designed for caching, which means that in normal
38 * operation all of our available space will be allocated. Thus, we need an
39 * efficient way of deleting things from the cache so we can write new things to
42 * To do this, we first divide the cache device up into buckets. A bucket is the
43 * unit of allocation; they're typically around 1 mb - anywhere from 128k to 2M+
46 * Each bucket has a 16 bit priority, and an 8 bit generation associated with
47 * it. The gens and priorities for all the buckets are stored contiguously and
48 * packed on disk (in a linked list of buckets - aside from the superblock, all
49 * of bcache's metadata is stored in buckets).
51 * The priority is used to implement an LRU. We reset a bucket's priority when
52 * we allocate it or on cache it, and every so often we decrement the priority
53 * of each bucket. It could be used to implement something more sophisticated,
54 * if anyone ever gets around to it.
56 * The generation is used for invalidating buckets. Each pointer also has an 8
57 * bit generation embedded in it; for a pointer to be considered valid, its gen
58 * must match the gen of the bucket it points into. Thus, to reuse a bucket all
59 * we have to do is increment its gen (and write its new gen to disk; we batch
62 * Bcache is entirely COW - we never write twice to a bucket, even buckets that
63 * contain metadata (including btree nodes).
67 * Bcache is in large part design around the btree.
69 * At a high level, the btree is just an index of key -> ptr tuples.
71 * Keys represent extents, and thus have a size field. Keys also have a variable
72 * number of pointers attached to them (potentially zero, which is handy for
73 * invalidating the cache).
75 * The key itself is an inode:offset pair. The inode number corresponds to a
76 * backing device or a flash only volume. The offset is the ending offset of the
77 * extent within the inode - not the starting offset; this makes lookups
78 * slightly more convenient.
80 * Pointers contain the cache device id, the offset on that device, and an 8 bit
81 * generation number. More on the gen later.
83 * Index lookups are not fully abstracted - cache lookups in particular are
84 * still somewhat mixed in with the btree code, but things are headed in that
87 * Updates are fairly well abstracted, though. There are two different ways of
88 * updating the btree; insert and replace.
90 * BTREE_INSERT will just take a list of keys and insert them into the btree -
91 * overwriting (possibly only partially) any extents they overlap with. This is
92 * used to update the index after a write.
94 * BTREE_REPLACE is really cmpxchg(); it inserts a key into the btree iff it is
95 * overwriting a key that matches another given key. This is used for inserting
96 * data into the cache after a cache miss, and for background writeback, and for
97 * the moving garbage collector.
99 * There is no "delete" operation; deleting things from the index is
100 * accomplished by either by invalidating pointers (by incrementing a bucket's
101 * gen) or by inserting a key with 0 pointers - which will overwrite anything
102 * previously present at that location in the index.
104 * This means that there are always stale/invalid keys in the btree. They're
105 * filtered out by the code that iterates through a btree node, and removed when
106 * a btree node is rewritten.
110 * Our unit of allocation is a bucket, and we we can't arbitrarily allocate and
111 * free smaller than a bucket - so, that's how big our btree nodes are.
113 * (If buckets are really big we'll only use part of the bucket for a btree node
114 * - no less than 1/4th - but a bucket still contains no more than a single
115 * btree node. I'd actually like to change this, but for now we rely on the
116 * bucket's gen for deleting btree nodes when we rewrite/split a node.)
118 * Anyways, btree nodes are big - big enough to be inefficient with a textbook
119 * btree implementation.
121 * The way this is solved is that btree nodes are internally log structured; we
122 * can append new keys to an existing btree node without rewriting it. This
123 * means each set of keys we write is sorted, but the node is not.
125 * We maintain this log structure in memory - keeping 1Mb of keys sorted would
126 * be expensive, and we have to distinguish between the keys we have written and
127 * the keys we haven't. So to do a lookup in a btree node, we have to search
128 * each sorted set. But we do merge written sets together lazily, so the cost of
129 * these extra searches is quite low (normally most of the keys in a btree node
130 * will be in one big set, and then there'll be one or two sets that are much
133 * This log structure makes bcache's btree more of a hybrid between a
134 * conventional btree and a compacting data structure, with some of the
135 * advantages of both.
137 * GARBAGE COLLECTION:
139 * We can't just invalidate any bucket - it might contain dirty data or
140 * metadata. If it once contained dirty data, other writes might overwrite it
141 * later, leaving no valid pointers into that bucket in the index.
143 * Thus, the primary purpose of garbage collection is to find buckets to reuse.
144 * It also counts how much valid data it each bucket currently contains, so that
145 * allocation can reuse buckets sooner when they've been mostly overwritten.
147 * It also does some things that are really internal to the btree
148 * implementation. If a btree node contains pointers that are stale by more than
149 * some threshold, it rewrites the btree node to avoid the bucket's generation
150 * wrapping around. It also merges adjacent btree nodes if they're empty enough.
154 * Bcache's journal is not necessary for consistency; we always strictly
155 * order metadata writes so that the btree and everything else is consistent on
156 * disk in the event of an unclean shutdown, and in fact bcache had writeback
157 * caching (with recovery from unclean shutdown) before journalling was
160 * Rather, the journal is purely a performance optimization; we can't complete a
161 * write until we've updated the index on disk, otherwise the cache would be
162 * inconsistent in the event of an unclean shutdown. This means that without the
163 * journal, on random write workloads we constantly have to update all the leaf
164 * nodes in the btree, and those writes will be mostly empty (appending at most
165 * a few keys each) - highly inefficient in terms of amount of metadata writes,
166 * and it puts more strain on the various btree resorting/compacting code.
168 * The journal is just a log of keys we've inserted; on startup we just reinsert
169 * all the keys in the open journal entries. That means that when we're updating
170 * a node in the btree, we can wait until a 4k block of keys fills up before
173 * For simplicity, we only journal updates to leaf nodes; updates to parent
174 * nodes are rare enough (since our leaf nodes are huge) that it wasn't worth
175 * the complexity to deal with journalling them (in particular, journal replay)
176 * - updates to non leaf nodes just happen synchronously (see btree_split()).
179 #define pr_fmt(fmt) "bcache: %s() " fmt "\n", __func__
181 #include <linux/bcache.h>
182 #include <linux/bio.h>
183 #include <linux/kobject.h>
184 #include <linux/list.h>
185 #include <linux/mutex.h>
186 #include <linux/rbtree.h>
187 #include <linux/rwsem.h>
188 #include <linux/refcount.h>
189 #include <linux/types.h>
190 #include <linux/workqueue.h>
191 #include <linux/kthread.h>
201 uint8_t last_gc
; /* Most out of date gen in the btree */
202 uint16_t gc_mark
; /* Bitfield used by GC. See below for field */
206 * I'd use bitfields for these, but I don't trust the compiler not to screw me
207 * as multiple threads touch struct bucket without locking
210 BITMASK(GC_MARK
, struct bucket
, gc_mark
, 0, 2);
211 #define GC_MARK_RECLAIMABLE 1
212 #define GC_MARK_DIRTY 2
213 #define GC_MARK_METADATA 3
214 #define GC_SECTORS_USED_SIZE 13
215 #define MAX_GC_SECTORS_USED (~(~0ULL << GC_SECTORS_USED_SIZE))
216 BITMASK(GC_SECTORS_USED
, struct bucket
, gc_mark
, 2, GC_SECTORS_USED_SIZE
);
217 BITMASK(GC_MOVE
, struct bucket
, gc_mark
, 15, 1);
232 struct bkey last_scanned
;
236 * Beginning and end of range in rb tree - so that we can skip taking
237 * lock and checking the rb tree when we need to check for overlapping
245 #define KEYBUF_NR 500
246 DECLARE_ARRAY_ALLOCATOR(struct keybuf_key
, freelist
, KEYBUF_NR
);
249 struct bcache_device
{
256 #define BCACHEDEVNAME_SIZE 12
257 char name
[BCACHEDEVNAME_SIZE
];
259 struct gendisk
*disk
;
262 #define BCACHE_DEV_CLOSING 0
263 #define BCACHE_DEV_DETACHING 1
264 #define BCACHE_DEV_UNLINK_DONE 2
265 #define BCACHE_DEV_WB_RUNNING 3
266 #define BCACHE_DEV_RATE_DW_RUNNING 4
268 unsigned stripe_size
;
269 atomic_t
*stripe_sectors_dirty
;
270 unsigned long *full_dirty_stripes
;
272 struct bio_set
*bio_split
;
274 unsigned data_csum
:1;
276 int (*cache_miss
)(struct btree
*, struct search
*,
277 struct bio
*, unsigned);
278 int (*ioctl
) (struct bcache_device
*, fmode_t
, unsigned, unsigned long);
282 /* Used to track sequential IO so it can be skipped */
283 struct hlist_node hash
;
284 struct list_head lru
;
286 unsigned long jiffies
;
291 enum stop_on_failure
{
292 BCH_CACHED_DEV_STOP_AUTO
= 0,
293 BCH_CACHED_DEV_STOP_ALWAYS
,
294 BCH_CACHED_DEV_STOP_MODE_MAX
,
298 struct list_head list
;
299 struct bcache_device disk
;
300 struct block_device
*bdev
;
304 struct bio_vec sb_bv
[1];
305 struct closure sb_write
;
306 struct semaphore sb_write_mutex
;
308 /* Refcount on the cache set. Always nonzero when we're caching. */
310 struct work_struct detach
;
313 * Device might not be running if it's dirty and the cache set hasn't
319 * Writes take a shared lock from start to finish; scanning for dirty
320 * data to refill the rb tree requires an exclusive lock.
322 struct rw_semaphore writeback_lock
;
325 * Nonzero, and writeback has a refcount (d->count), iff there is dirty
326 * data in the cache. Protected by writeback_lock; must have an
327 * shared lock to set and exclusive lock to clear.
331 struct bch_ratelimit writeback_rate
;
332 struct delayed_work writeback_rate_update
;
335 * Internal to the writeback code, so read_dirty() can keep track of
340 /* Limit number of writeback bios in flight */
341 struct semaphore in_flight
;
342 struct task_struct
*writeback_thread
;
343 struct workqueue_struct
*writeback_write_wq
;
345 struct keybuf writeback_keys
;
347 struct task_struct
*status_update_thread
;
348 /* For tracking sequential IO */
349 #define RECENT_IO_BITS 7
350 #define RECENT_IO (1 << RECENT_IO_BITS)
351 struct io io
[RECENT_IO
];
352 struct hlist_head io_hash
[RECENT_IO
+ 1];
353 struct list_head io_lru
;
356 struct cache_accounting accounting
;
358 /* The rest of this all shows up in sysfs */
359 unsigned sequential_cutoff
;
362 unsigned io_disable
:1;
364 unsigned bypass_torture_test
:1;
366 unsigned partial_stripes_expensive
:1;
367 unsigned writeback_metadata
:1;
368 unsigned writeback_running
:1;
369 unsigned char writeback_percent
;
370 unsigned writeback_delay
;
372 uint64_t writeback_rate_target
;
373 int64_t writeback_rate_proportional
;
374 int64_t writeback_rate_integral
;
375 int64_t writeback_rate_integral_scaled
;
376 int32_t writeback_rate_change
;
378 unsigned writeback_rate_update_seconds
;
379 unsigned writeback_rate_i_term_inverse
;
380 unsigned writeback_rate_p_term_inverse
;
381 unsigned writeback_rate_minimum
;
383 enum stop_on_failure stop_when_cache_set_failed
;
384 #define DEFAULT_CACHED_DEV_ERROR_LIMIT 64
386 unsigned error_limit
;
387 unsigned offline_seconds
;
389 char backing_dev_name
[BDEVNAME_SIZE
];
401 struct cache_set
*set
;
404 struct bio_vec sb_bv
[1];
407 struct block_device
*bdev
;
409 struct task_struct
*alloc_thread
;
412 struct prio_set
*disk_buckets
;
415 * When allocating new buckets, prio_write() gets first dibs - since we
416 * may not be allocate at all without writing priorities and gens.
417 * prio_buckets[] contains the last buckets we wrote priorities to (so
418 * gc can mark them as metadata), prio_next[] contains the buckets
419 * allocated for the next prio write.
421 uint64_t *prio_buckets
;
422 uint64_t *prio_last_buckets
;
425 * free: Buckets that are ready to be used
427 * free_inc: Incoming buckets - these are buckets that currently have
428 * cached data in them, and we can't reuse them until after we write
429 * their new gen to disk. After prio_write() finishes writing the new
430 * gens/prios, they'll be moved to the free list (and possibly discarded
433 DECLARE_FIFO(long, free
)[RESERVE_NR
];
434 DECLARE_FIFO(long, free_inc
);
436 size_t fifo_last_bucket
;
438 /* Allocation stuff: */
439 struct bucket
*buckets
;
441 DECLARE_HEAP(struct bucket
*, heap
);
444 * If nonzero, we know we aren't going to find any buckets to invalidate
445 * until a gc finishes - otherwise we could pointlessly burn a ton of
448 unsigned invalidate_needs_gc
;
450 bool discard
; /* Get rid of? */
452 struct journal_device journal
;
454 /* The rest of this all shows up in sysfs */
455 #define IO_ERROR_SHIFT 20
459 atomic_long_t meta_sectors_written
;
460 atomic_long_t btree_sectors_written
;
461 atomic_long_t sectors_written
;
463 char cache_dev_name
[BDEVNAME_SIZE
];
471 uint64_t data
; /* sectors */
472 unsigned in_use
; /* percent */
476 * Flag bits, for how the cache set is shutting down, and what phase it's at:
478 * CACHE_SET_UNREGISTERING means we're not just shutting down, we're detaching
479 * all the backing devices first (their cached data gets invalidated, and they
480 * won't automatically reattach).
482 * CACHE_SET_STOPPING always gets set first when we're closing down a cache set;
483 * we'll continue to run normally for awhile with CACHE_SET_STOPPING set (i.e.
484 * flushing dirty data).
486 * CACHE_SET_RUNNING means all cache devices have been registered and journal
487 * replay is complete.
489 * CACHE_SET_IO_DISABLE is set when bcache is stopping the whold cache set, all
490 * external and internal I/O should be denied when this flag is set.
493 #define CACHE_SET_UNREGISTERING 0
494 #define CACHE_SET_STOPPING 1
495 #define CACHE_SET_RUNNING 2
496 #define CACHE_SET_IO_DISABLE 3
501 struct list_head list
;
503 struct kobject internal
;
504 struct dentry
*debug
;
505 struct cache_accounting accounting
;
511 struct cache
*cache
[MAX_CACHES_PER_SET
];
512 struct cache
*cache_by_alloc
[MAX_CACHES_PER_SET
];
515 struct bcache_device
**devices
;
516 struct list_head cached_devs
;
517 uint64_t cached_dev_sectors
;
518 struct closure caching
;
520 struct closure sb_write
;
521 struct semaphore sb_write_mutex
;
525 struct bio_set
*bio_split
;
527 /* For the btree cache */
528 struct shrinker shrink
;
530 /* For the btree cache and anything allocation related */
531 struct mutex bucket_lock
;
533 /* log2(bucket_size), in sectors */
534 unsigned short bucket_bits
;
536 /* log2(block_size), in sectors */
537 unsigned short block_bits
;
540 * Default number of pages for a new btree node - may be less than a
543 unsigned btree_pages
;
546 * Lists of struct btrees; lru is the list for structs that have memory
547 * allocated for actual btree node, freed is for structs that do not.
549 * We never free a struct btree, except on shutdown - we just put it on
550 * the btree_cache_freed list and reuse it later. This simplifies the
551 * code, and it doesn't cost us much memory as the memory usage is
552 * dominated by buffers that hold the actual btree node data and those
553 * can be freed - and the number of struct btrees allocated is
554 * effectively bounded.
556 * btree_cache_freeable effectively is a small cache - we use it because
557 * high order page allocations can be rather expensive, and it's quite
558 * common to delete and allocate btree nodes in quick succession. It
559 * should never grow past ~2-3 nodes in practice.
561 struct list_head btree_cache
;
562 struct list_head btree_cache_freeable
;
563 struct list_head btree_cache_freed
;
565 /* Number of elements in btree_cache + btree_cache_freeable lists */
566 unsigned btree_cache_used
;
569 * If we need to allocate memory for a new btree node and that
570 * allocation fails, we can cannibalize another node in the btree cache
571 * to satisfy the allocation - lock to guarantee only one thread does
574 wait_queue_head_t btree_cache_wait
;
575 struct task_struct
*btree_cache_alloc_lock
;
578 * When we free a btree node, we increment the gen of the bucket the
579 * node is in - but we can't rewrite the prios and gens until we
580 * finished whatever it is we were doing, otherwise after a crash the
581 * btree node would be freed but for say a split, we might not have the
582 * pointers to the new nodes inserted into the btree yet.
584 * This is a refcount that blocks prio_write() until the new keys are
587 atomic_t prio_blocked
;
588 wait_queue_head_t bucket_wait
;
591 * For any bio we don't skip we subtract the number of sectors from
592 * rescale; when it hits 0 we rescale all the bucket priorities.
596 * When we invalidate buckets, we use both the priority and the amount
597 * of good data to determine which buckets to reuse first - to weight
598 * those together consistently we keep track of the smallest nonzero
599 * priority of any bucket.
604 * max(gen - last_gc) for all buckets. When it gets too big we have to gc
605 * to keep gens from wrapping around.
608 struct gc_stat gc_stats
;
610 size_t avail_nbuckets
;
612 struct task_struct
*gc_thread
;
613 /* Where in the btree gc currently is */
617 * The allocation code needs gc_mark in struct bucket to be correct, but
618 * it's not while a gc is in progress. Protected by bucket_lock.
622 /* Counts how many sectors bio_insert has added to the cache */
623 atomic_t sectors_to_gc
;
624 wait_queue_head_t gc_wait
;
626 struct keybuf moving_gc_keys
;
627 /* Number of moving GC bios in flight */
628 struct semaphore moving_in_flight
;
630 struct workqueue_struct
*moving_gc_wq
;
634 #ifdef CONFIG_BCACHE_DEBUG
635 struct btree
*verify_data
;
636 struct bset
*verify_ondisk
;
637 struct mutex verify_lock
;
641 struct uuid_entry
*uuids
;
642 BKEY_PADDED(uuid_bucket
);
643 struct closure uuid_write
;
644 struct semaphore uuid_write_mutex
;
647 * A btree node on disk could have too many bsets for an iterator to fit
648 * on the stack - have to dynamically allocate them
650 mempool_t
*fill_iter
;
652 struct bset_sort_state sort
;
654 /* List of buckets we're currently writing data to */
655 struct list_head data_buckets
;
656 spinlock_t data_bucket_lock
;
658 struct journal journal
;
660 #define CONGESTED_MAX 1024
661 unsigned congested_last_us
;
664 /* The rest of this all shows up in sysfs */
665 unsigned congested_read_threshold_us
;
666 unsigned congested_write_threshold_us
;
668 struct time_stats btree_gc_time
;
669 struct time_stats btree_split_time
;
670 struct time_stats btree_read_time
;
672 atomic_long_t cache_read_races
;
673 atomic_long_t writeback_keys_done
;
674 atomic_long_t writeback_keys_failed
;
676 atomic_long_t reclaim
;
677 atomic_long_t flush_write
;
678 atomic_long_t retry_flush_write
;
684 unsigned error_limit
;
685 unsigned error_decay
;
687 unsigned short journal_delay_ms
;
688 bool expensive_debug_checks
;
690 unsigned key_merging_disabled
:1;
691 unsigned gc_always_rewrite
:1;
692 unsigned shrinker_disabled
:1;
693 unsigned copy_gc_enabled
:1;
695 #define BUCKET_HASH_BITS 12
696 struct hlist_head bucket_hash
[1 << BUCKET_HASH_BITS
];
698 DECLARE_HEAP(struct btree
*, flush_btree
);
702 unsigned submit_time_us
;
707 * We only need pad = 3 here because we only ever carry around a
708 * single pointer - i.e. the pointer we're doing io to/from.
714 #define BTREE_PRIO USHRT_MAX
715 #define INITIAL_PRIO 32768U
717 #define btree_bytes(c) ((c)->btree_pages * PAGE_SIZE)
718 #define btree_blocks(b) \
719 ((unsigned) (KEY_SIZE(&b->key) >> (b)->c->block_bits))
721 #define btree_default_blocks(c) \
722 ((unsigned) ((PAGE_SECTORS * (c)->btree_pages) >> (c)->block_bits))
724 #define bucket_pages(c) ((c)->sb.bucket_size / PAGE_SECTORS)
725 #define bucket_bytes(c) ((c)->sb.bucket_size << 9)
726 #define block_bytes(c) ((c)->sb.block_size << 9)
728 #define prios_per_bucket(c) \
729 ((bucket_bytes(c) - sizeof(struct prio_set)) / \
730 sizeof(struct bucket_disk))
731 #define prio_buckets(c) \
732 DIV_ROUND_UP((size_t) (c)->sb.nbuckets, prios_per_bucket(c))
734 static inline size_t sector_to_bucket(struct cache_set
*c
, sector_t s
)
736 return s
>> c
->bucket_bits
;
739 static inline sector_t
bucket_to_sector(struct cache_set
*c
, size_t b
)
741 return ((sector_t
) b
) << c
->bucket_bits
;
744 static inline sector_t
bucket_remainder(struct cache_set
*c
, sector_t s
)
746 return s
& (c
->sb
.bucket_size
- 1);
749 static inline struct cache
*PTR_CACHE(struct cache_set
*c
,
750 const struct bkey
*k
,
753 return c
->cache
[PTR_DEV(k
, ptr
)];
756 static inline size_t PTR_BUCKET_NR(struct cache_set
*c
,
757 const struct bkey
*k
,
760 return sector_to_bucket(c
, PTR_OFFSET(k
, ptr
));
763 static inline struct bucket
*PTR_BUCKET(struct cache_set
*c
,
764 const struct bkey
*k
,
767 return PTR_CACHE(c
, k
, ptr
)->buckets
+ PTR_BUCKET_NR(c
, k
, ptr
);
770 static inline uint8_t gen_after(uint8_t a
, uint8_t b
)
773 return r
> 128U ? 0 : r
;
776 static inline uint8_t ptr_stale(struct cache_set
*c
, const struct bkey
*k
,
779 return gen_after(PTR_BUCKET(c
, k
, i
)->gen
, PTR_GEN(k
, i
));
782 static inline bool ptr_available(struct cache_set
*c
, const struct bkey
*k
,
785 return (PTR_DEV(k
, i
) < MAX_CACHES_PER_SET
) && PTR_CACHE(c
, k
, i
);
788 /* Btree key macros */
791 * This is used for various on disk data structures - cache_sb, prio_set, bset,
792 * jset: The checksum is _always_ the first 8 bytes of these structs
794 #define csum_set(i) \
795 bch_crc64(((void *) (i)) + sizeof(uint64_t), \
796 ((void *) bset_bkey_last(i)) - \
797 (((void *) (i)) + sizeof(uint64_t)))
799 /* Error handling macros */
801 #define btree_bug(b, ...) \
803 if (bch_cache_set_error((b)->c, __VA_ARGS__)) \
807 #define cache_bug(c, ...) \
809 if (bch_cache_set_error(c, __VA_ARGS__)) \
813 #define btree_bug_on(cond, b, ...) \
816 btree_bug(b, __VA_ARGS__); \
819 #define cache_bug_on(cond, c, ...) \
822 cache_bug(c, __VA_ARGS__); \
825 #define cache_set_err_on(cond, c, ...) \
828 bch_cache_set_error(c, __VA_ARGS__); \
833 #define for_each_cache(ca, cs, iter) \
834 for (iter = 0; ca = cs->cache[iter], iter < (cs)->sb.nr_in_set; iter++)
836 #define for_each_bucket(b, ca) \
837 for (b = (ca)->buckets + (ca)->sb.first_bucket; \
838 b < (ca)->buckets + (ca)->sb.nbuckets; b++)
840 static inline void cached_dev_put(struct cached_dev
*dc
)
842 if (refcount_dec_and_test(&dc
->count
))
843 schedule_work(&dc
->detach
);
846 static inline bool cached_dev_get(struct cached_dev
*dc
)
848 if (!refcount_inc_not_zero(&dc
->count
))
851 /* Paired with the mb in cached_dev_attach */
852 smp_mb__after_atomic();
857 * bucket_gc_gen() returns the difference between the bucket's current gen and
858 * the oldest gen of any pointer into that bucket in the btree (last_gc).
861 static inline uint8_t bucket_gc_gen(struct bucket
*b
)
863 return b
->gen
- b
->last_gc
;
866 #define BUCKET_GC_GEN_MAX 96U
868 #define kobj_attribute_write(n, fn) \
869 static struct kobj_attribute ksysfs_##n = __ATTR(n, S_IWUSR, NULL, fn)
871 #define kobj_attribute_rw(n, show, store) \
872 static struct kobj_attribute ksysfs_##n = \
873 __ATTR(n, S_IWUSR|S_IRUSR, show, store)
875 static inline void wake_up_allocators(struct cache_set
*c
)
880 for_each_cache(ca
, c
, i
)
881 wake_up_process(ca
->alloc_thread
);
884 static inline void closure_bio_submit(struct cache_set
*c
,
889 if (unlikely(test_bit(CACHE_SET_IO_DISABLE
, &c
->flags
))) {
890 bio
->bi_status
= BLK_STS_IOERR
;
894 generic_make_request(bio
);
898 * Prevent the kthread exits directly, and make sure when kthread_stop()
899 * is called to stop a kthread, it is still alive. If a kthread might be
900 * stopped by CACHE_SET_IO_DISABLE bit set, wait_for_kthread_stop() is
901 * necessary before the kthread returns.
903 static inline void wait_for_kthread_stop(void)
905 while (!kthread_should_stop()) {
906 set_current_state(TASK_INTERRUPTIBLE
);
911 /* Forward declarations */
913 void bch_count_backing_io_errors(struct cached_dev
*dc
, struct bio
*bio
);
914 void bch_count_io_errors(struct cache
*, blk_status_t
, const char *);
915 void bch_bbio_count_io_errors(struct cache_set
*, struct bio
*,
916 blk_status_t
, const char *);
917 void bch_bbio_endio(struct cache_set
*, struct bio
*, blk_status_t
,
919 void bch_bbio_free(struct bio
*, struct cache_set
*);
920 struct bio
*bch_bbio_alloc(struct cache_set
*);
922 void __bch_submit_bbio(struct bio
*, struct cache_set
*);
923 void bch_submit_bbio(struct bio
*, struct cache_set
*, struct bkey
*, unsigned);
925 uint8_t bch_inc_gen(struct cache
*, struct bucket
*);
926 void bch_rescale_priorities(struct cache_set
*, int);
928 bool bch_can_invalidate_bucket(struct cache
*, struct bucket
*);
929 void __bch_invalidate_one_bucket(struct cache
*, struct bucket
*);
931 void __bch_bucket_free(struct cache
*, struct bucket
*);
932 void bch_bucket_free(struct cache_set
*, struct bkey
*);
934 long bch_bucket_alloc(struct cache
*, unsigned, bool);
935 int __bch_bucket_alloc_set(struct cache_set
*, unsigned,
936 struct bkey
*, int, bool);
937 int bch_bucket_alloc_set(struct cache_set
*, unsigned,
938 struct bkey
*, int, bool);
939 bool bch_alloc_sectors(struct cache_set
*, struct bkey
*, unsigned,
940 unsigned, unsigned, bool);
941 bool bch_cached_dev_error(struct cached_dev
*dc
);
944 bool bch_cache_set_error(struct cache_set
*, const char *, ...);
946 int bch_prio_write(struct cache
*ca
, bool wait
);
947 void bch_write_bdev_super(struct cached_dev
*, struct closure
*);
949 extern struct workqueue_struct
*bcache_wq
;
950 extern struct mutex bch_register_lock
;
951 extern struct list_head bch_cache_sets
;
953 extern struct kobj_type bch_cached_dev_ktype
;
954 extern struct kobj_type bch_flash_dev_ktype
;
955 extern struct kobj_type bch_cache_set_ktype
;
956 extern struct kobj_type bch_cache_set_internal_ktype
;
957 extern struct kobj_type bch_cache_ktype
;
959 void bch_cached_dev_release(struct kobject
*);
960 void bch_flash_dev_release(struct kobject
*);
961 void bch_cache_set_release(struct kobject
*);
962 void bch_cache_release(struct kobject
*);
964 int bch_uuid_write(struct cache_set
*);
965 void bcache_write_super(struct cache_set
*);
967 int bch_flash_dev_create(struct cache_set
*c
, uint64_t size
);
969 int bch_cached_dev_attach(struct cached_dev
*, struct cache_set
*, uint8_t *);
970 void bch_cached_dev_detach(struct cached_dev
*);
971 void bch_cached_dev_emit_change(struct cached_dev
*);
972 void bch_cached_dev_run(struct cached_dev
*);
973 void bcache_device_stop(struct bcache_device
*);
975 void bch_cache_set_unregister(struct cache_set
*);
976 void bch_cache_set_stop(struct cache_set
*);
978 struct cache_set
*bch_cache_set_alloc(struct cache_sb
*);
979 void bch_btree_cache_free(struct cache_set
*);
980 int bch_btree_cache_alloc(struct cache_set
*);
981 void bch_moving_init_cache_set(struct cache_set
*);
982 int bch_open_buckets_alloc(struct cache_set
*);
983 void bch_open_buckets_free(struct cache_set
*);
985 int bch_cache_allocator_start(struct cache
*ca
);
987 void bch_debug_exit(void);
988 int bch_debug_init(struct kobject
*);
989 void bch_request_exit(void);
990 int bch_request_init(void);
992 #endif /* _BCACHE_H */